Mathematical Physics, Analysis and Geometry 4: 1–36, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
1
On the Law of Multiplication of Random Matrices VLADIMIR VASILCHUK Université Paris 7 Denis Diderot, Mathématiques, case 7012, Paris, France, Institute for Low Temperature Physics, 47 Lenin ave., 310164 Kharkov, Ukraine (Received: 22 February 2001; in revised form: 2 April 2001) Abstract. We recover Voiculescu’s results on multiplicative free convolutions of probability measures by different techniques which were already developed by Pastur and Vasilchuk for the law of addition of random matrices. Namely, we study the normalized eigenvalue counting measure of the product of two n × n unitary matrices and the measure of the product of three n × n Hermitian (or real symmetric) positive matrices rotated independently by random unitary (or orthogonal) Haar distributed matrices. We establish the convergence in probability as n → ∞ to a limiting nonrandom measure and obtain functional equations for the Herglotz and Stieltjes transforms of that limiting measure. Mathematics Subject Classifications (2000): 15A52, 60B99, 60F05. Key words: random matrices, counting measure, limit laws.
1. Introduction This paper deals with the eigenvalue distribution of products of n×n random matrices. We consider two models: (i) the product of two unitary (resp. orthogonal) matrices and (ii) the product of three Hermitian (or real symmetric) positive matrices. We study the eigenvalue distribution of these two ensembles in the limit n → ∞. Namely, we express the limiting normalized counting measure of eigenvalues of the product via the limits of the same counting measures of the corresponding factors. We assume that these exist and that factors are randomly rotated one with respect to another by a unitary (or orthogonal) random matrix uniformly distributed over the group U(n) (resp. O(n)). In this paper, under weaker assumptions, we obtain the analog of Voiculescu’s results concerning the free multiplicative convolutions of probability measures on the real axis and unit circle. They were studied within the context of free (noncommutative) probability theory introduced by Voiculescu at the beginning of the 90’s (see [2, 8, 10] for results and references). This theory deals with free random variables (operators in von Neumann algebras) that can be modeled by unitary (orthogonal) invariant random matrices [5, 6]. The notion of S-transform introduced in this theory allows one to generalize the functional equations for transforms of limiting counting measures of certain multiplicative unitary invariant models first proposed and studied by Marchenko and Pastur [4].
2
VLADIMIR VASILCHUK
Motivation for studying multiplicative random matrix ensembles is given by the fact that they appear in some physics studies (see, e.g., [3]). We use a simple method of deriving the functional equations for limiting eigenvalue distributions. It is a natural extension of the method proposed in [5] to study an additive analog of our ensembles. The basic idea is the same as in [4]: to study not the counting measure itself but rather some integral transforms that are generating functions of the moments of that measure. We derive functional relations for these transforms using the resolvent identity and differential identities for expectations of smooth functions with respect to the Haar measure of U(n) (or O(n)). The paper is organized as follows. In Section 2, we state and discuss the main results (Theorem 2.2 for unitary matrices and Theorem 2.1 for Hermitian positive matrices). In Section 3, we prove auxiliary Theorems 3.1 and 3.2 concerning Hermitian positive definite matrices under the conditions of uniform in n boundedness of the fourth moment of the normalized counting measure of the factors. In Section 4, we use these results to prove Theorem 2.1. Here the main condition is the uniform boundedness of the second moment of the normalized counting measure of the factors. In Section 5 we prove Theorems 5.1 and 5.2 and then prove Theorem 2.2 giving the solution of the problem for unitary matrices. In Section 6 we prove the auxiliary facts that we need. We also describe generalizations of our results in the cases of orthogonal and real symmetric matrices. 2. Models and Main Results We study two ensembles of n × n random matrices Vn and Hn of the form: Vn = V1,nV2,n ,
(2.1)
where V1,n = Wn∗ Sn Wn ,
V2,n = Un∗ Tn Un
and 1/2
1/2
Hn = H1,n H2,n H1,n ,
(2.2)
where H1,n = Wn∗ An Wn ,
H2,n = Un∗ Bn Un .
We assume that Sn , Tn , An , Bn , Un and Wn are mutually independent. Sn and Tn are random unitary matrices, Un and Wn are unitary (resp. orthogonal) random matrices uniformly distributed over the unitary (orthogonal) group U(n) (resp. O(n)) with respect to the Haar measure. An and Bn are Hermitian positive definite random matrices.
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
3
We will restrict ourselves to the case of Hermitian matrices (resp. to the group U(n)). The results for symmetric real matrices (i.e. for the group O(n)) are similar, although their proofs are more difficult (see Section 6). We are interested in the asymptotic behavior, as n → ∞, of the normalized eigenvalue counting measure (NCM) νn of the ensemble (2.1), whose value on any Borel set ⊂ [0, 2π ] is given by νn () =
#{µ(n) i ∈ } , n
(2.3)
where µ(n) are the eigenvalues of Vn . We are also interested in the asymptotic i behavior of the NCM Nn of the ensemble (2.2), whose value on any Borel set ⊂ R is given by Nn () =
#{λ(n) i ∈ } , n
(2.4)
where λ(n) i are the eigenvalues of Hn . The problem was studied recently by Voiculescu [2, 8, 10] within the context of free (noncommutative) probability. Combining Voiculescu’s results on free multiplicative convolution of measures having nonzero first moment [2, 10] with results of asymptotic freeness of n × n Haar distributed unitary matrices and nonrandom diagonal matrices [6, 9, 10], one can easily obtain the following result: PROPOSITION. If the matrices Sn , Tn , An and Bn are nonrandom, if the norms of An and Bn are uniformly bounded in n, i.e. their NCMs N1,n and N2,n have compact support uniformly in n, and if these measures have weak limits as n → ∞ ν1,n → ν1 , N1,n → N1 ,
ν2,n → ν2 , N2,n → N2 ,
(2.5) (2.6)
where ν1,n and ν2,n are the NCM’s of Sn and Tn , then the NCM’s (2.3) and (2.4) of (2.1) and (2.2) converge weakly with probability 1 to nonrandom measures ν and N. Here and below the convergence with probability 1 is understood as that in the natural probability spaces ˜ = ˜ n, n , (2.7) = n
n
where n is the probability space of matrices (2.1), that is the product of respective ˜n spaces of Sn and Tn and two copies of the group U(n) for Un and Wn , and where is the probability space of matrices (2.2), that is the product of respective spaces of An and Bn and two copies of the group U(n) for Un and Wn . Besides, according to [10], one can define the S-transforms S1 , S2 and S of the measures ν1 , ν2 and ν, respectively, and the S-transforms S˜1 , S˜2 and S˜ of the
4
VLADIMIR VASILCHUK
measures N1 , N2 and N (see our remark after Theorem 2.2) and one can find the following simple expressions S and S˜ via S1,2 and S˜1,2 S = S1 S2 ,
S˜ = S˜1 S˜2 .
The proof of the asymptotic freeness of n × n Haar distributed unitary matrices and nonrandom diagonal matrices having, uniformly in n, compactly supported NCMs in [6, 8], is based on the asymptotic analysis of the expectations of normalized traces of mixed products of matrices Sn , Tn , Un and Wn and An , Bn , Un and Wn , respectively. It requires a considerable amount of combinatorial analysis, the existence of all moments measures Nr , r = 1, 2, and their rather regular behavior as n → ∞ to obtain the convergence of expectations. In this paper we obtain analogous results under more weak assumptions by a method that does not involve combinatorics. This is because we work with the Stieltjes transforms of the measures (2.4) and (2.6) and with the Herglotz transforms of the measures (2.3) and (2.5). We directly derive functional equations for their limits by using simple identities for expectations of matrix-valued functions with respect to the Haar measure (Proposition 3.2 below) and elementary facts concerning resolvents of Hermitian and unitary matrices. This method was already used in [5] for the additive random ensembles. We list below the properties of the Stieltjes and Herglotz transforms that we will need below (see, e.g., [1]). PROPOSITION 2.1. Let m(dλ) , Im z = 0 s(z) = R λ−z
(2.8)
be the Stieltjes transform of a probability measure m on R, then (i) s(z) is analytic in C \ R and |s(z)| |Im z|−1 . (ii) (iii)
Im s(z)Im z > 0, lim y |s(iy)| = 1.
y→∞
(2.9) Im z = 0.
(2.10) (2.11)
(iv) For any continuous function ϕ with compact support we have the Frobenius– Perron inversion formula 1 φ(λ)m(dλ) = lim φ(λ)Im s(λ + iε). (2.12) ε→0 π R R (v) Conversely, any function satisfying (2.9)–(2.11) is the Stieltjes transform of a probability measure and this one-to-one correspondence between measures and their Stieltjes transforms is continuous for the topology of weak convergence for measures and for the topology of convergence on compact subsets of C \ R for the Stieltjes transforms.
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
PROPOSITION 2.2. Let 2π iθ e +z µ(dθ), t (z) = eiθ − z 0
|z| < 1
5
(2.13)
be the Herglotz transform of a probability measure µ on [0, 2π ], then (i) t (z) is analytic for |z| < 1 and |t (z) − 1| 2|z|(1 − |z|)−1 , (ii)
t (0) = 1,
Re t (z) > 0,
|z| < 1. |z| < 1.
(2.14) (2.15)
(iii) For any continuous on [0, 2π ] function ϕ we have the inversion formula 2π 2π 1 φ(θ)µ(dθ) = lim φ(θ)Re t (re−iθ ) dθ. (2.16) r→1 2π − 0 0 (iv) Conversely, any function satisfying (2.14)–(2.15) is the Herglotz transform of a probability measure on the unit circle. This one-to-one correspondence between measures and their Herglotz transforms is continuous for the topology of weak convergence for measures and for the topology of convergence on compact subsets of {z ∈ C | |z| < 1} for the Herglotz transforms. Now we state our main results. Since the set of eigenvalues of unitary and Hermitian matrices are unitary invariant, we can replace matrices (2.1) and (2.2) by Vn = Sn Un∗ Tn Un
(2.17)
Hn = An1/2 Un∗ Bn Un An1/2 ,
(2.18)
and
where Sn , Tn , An , Bn and Un are as in (2.1) and (2.2). However, it is useful to keep in mind that the problem is symmetric in Sn and Tn and in An and Bn (as we will see below). THEOREM 2.1. Let Hn be a positive definite random n × n matrix of the form (2.2). Assume that the normalized counting measures N1,n , N2,n of An , Bn converge weakly in probability as n → ∞ to nonrandom probability measures N1 , N2 . We also assume +∞ λ2 E{Nr,n(dλ)} m(2) < ∞, r = 1, 2, (2.19) sup n
0
and
+∞
mr =
λNr (dλ) > 0, 0
r = 1, 2,
(2.20)
6
VLADIMIR VASILCHUK
i.e. the measures N1 , N2 are not concentrated at zero. Then the normalized counting measure Nn of Hn converge in probability to a nonrandom probability measure whose Stieltjes transform +∞ N(dλ) , Im z = 0 (2.21) f (z) = λ−z 0 is the unique solution of the system f (z)(1 + zf (z)) = 1 (z)2 (z), z2 (z) , 1 (z) = f2 1 + zf (z) z1 (z) 2 (z) = f1 1 + zf (z)
(2.22)
in the class of functions f (z), 1,2 (z) which are analytic for Im z = 0 and which satisfy (2.9)–(2.11) and zr (z) = −mr + O(|Im z|−1 ), |Re z| |Im z|, z → ∞.
r = 1, 2, (2.23)
f1 (z), f2 (z) are the Stieltjes transforms of N1 , N2 and E{ · } denotes the expectation with respect to the probability measure generated by An , Bn , Un and Wn . THEOREM 2.2. Let Vn be a random n × n matrix of the form (2.1). Assume that the normalized counting measures ν1,n , ν2,n of Sn , Tn converge weakly in probability as n → ∞ to nonrandom probability measures on the unit circle ν1 , ν2 . Then the normalized counting measure νn of Vn converge in probability to a nonrandom probability measure ν whose Herglotz transform 2π iµ e +z ν(dµ), |z| < 1 (2.24) h(z) = eiµ − z 0 is the unique solution of the system h2 (z) = 1 + 4z1 (z)2 (z), 2z2 (z) , h(z) = h2 1 + h(z) 2z1 (z) h(z) = h1 1 + h(z)
(2.25)
in the class of functions h(z), 1,2 (z) which are analytic for |z| < 1 and which satisfy (2.14)–(2.15) and |1,2 (z)| (1 − |z|)−1 ,
|z| < 1.
h1,2 (z) are the Herglotz transforms of the measures ν1,2 .
(2.26)
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
7
Both theorems will be proved in Sections 3 and 5. Here we interpret them in terms of S-transform introduced by Voiculescu in the context of C ∗ -algebras. 2.1. VOICULESCU ’ S FORMULATION Consider a probability measure µ on the unit circle and assume that its first moment is nonzero 2π eiθ µ(dθ) = 0. µ1 = 0
Consider the function ϕµ (z) = −
1 + t (z−1 ) 2
where t (z) is the Herglotz transform of µ. Since ϕµ (z) = µ1 + o(1), z → ∞, then, according to the local inversion theorem, there exists a unique inverse function χµ (ϕ) of ϕµ (z), χµ (ϕµ (z)) = z defined and analytic in a neighborhood of −1 and assuming its values in a neighborhood of infinity. On the other hand, for any probability measure m on the real nonnegative semi-axis having nonzero first moment ∞ λm(dλ) > 0, m1 = 0
we can consider the function ϕm (z) = −(1 + z−1 s(z−1 )), where s(z) is the Stieltjes transform of the measure m. Since ϕm (0) = m1 , then, according to the local inversion theorem, ϕm (z) also has a unique inverse function χm (ϕ) defined and analytic in a neighborhood of zero and assuming its values in a neighborhood of zero. Denote Sµ (ϕ) = χµ (ϕ)ϕ −1 (1 + ϕ), Sm (ϕ) = χm (ϕ)ϕ −1 (1 + ϕ) and, following Voiculescu [10], call Sµ (ϕ) the S-transform of the probability measure µ on the unit circle and Sm (ϕ) the S-transform of the probability measure m on the real nonnegative semi-axis. By using the S-transforms S1 , S2 of ν1 , ν2 , we can rewrite (2.25) in the form 1 + ψ(z) 1 + ψ(z) , z1 (z) z2 (z) 1 + ψ(z) , r = 1, 2, Sr (ψ(z)) = − zr (z)
S(ψ(z)) =
(2.27) (2.28)
8
VLADIMIR VASILCHUK
where ψ(z) = ϕ(z−1) = −(1 + h(z))/2 and S(ψ) denotes the S-transform of the limiting normalized counting measure ν of (2.1), whose Herglotz transform is h(z). Then we derive from (2.27) Voiculescu’s very simple expression of S via S1 and S2 S(ψ) = S1 (ψ)S2 (ψ).
(2.29)
It is easy to check that (2.22) leads to the relations (2.27) for ψ(z) = −(1 + zf (z)) and S(ψ) denotes the S-transform of the limiting normalized counting measure N of (2.2), whose Stieltjes transform is f (z). Thus, (2.22) leads to the same expression (2.29) where S1,2 will be the S-transforms of the measures N1,2 . The relation (2.29) was obtained by Voiculescu in the context of C ∗ -algebra studies (see [9, 10] for results and references). 3. Convergence with Probability 1 for Nonrandom An , Bn In this section, we start the proof of Theorem 2.1. As a first step we prove the following theorem: THEOREM 3.1. Let Hn be a positive definite nonrandom n × n matrix of the form (2.2) in which An and Bn are nonrandom Hermitian positive matrices, Un and Wn are random independent unitary matrices distributed according to the Haar measure on U(n). Assume that the normalized counting measures N1,n , N2,n of An and Bn converge weakly as n → ∞ to nonrandom probability measures N1 , N2 , +∞ +∞ λNr,n (dλ) = mr = λNr (dλ) > 0, r = 1, 2, (3.1) lim n→∞
0
+∞
sup n
0
λ4 Nr (dλ) m4 < ∞.
(3.2)
0
Then the normalized counting measure Nn of Hn converges with probability 1 to a nonrandom probability measure N whose Stieltjes transform f (z) (2.21) is the unique solution of (2.22) in the class of functions f (z), 1,2 (z) which are analytic for z ∈ C \ R+ and which satisfy (2.9)–(2.11) and (2.23). We use the technique introduced in [5]. Let us recall its basic means. First we collect elementary facts of linear algebra. PROPOSITION 3.1. Let Mn be the algebra of linear endomorphisms of Cn equipped with the norm induced by the standard Euclidean norm of Cn . Then (i) If {Mj k }nj,k=1 is the matrix of M ∈ Mn in any orthonormal basis of Cn then |Mj k | ||M||.
(3.3)
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
9
(ii) If Tr M denotes the trace of M ∈ Mn , then |Tr M| n||M||, |Tr M1 M2 |2 Tr M1 M1∗ Tr M2 M2∗ , (3.4) where M ∗ is the adjoint of M. Furthermore, if P ∈ Mn is positive definite, then |Tr MP | ||M||Tr P .
(3.5)
G(z) = (M − z)−1
(3.6)
(iii) Let be the resolvent of M ∈ Mn . It is defined for all nonreal z, Im z = 0 if M is Hermitian. It is defined for all z, |z| = 1 if M is unitary. (iv) If G(z1 ) and G(z2 ) are defined, G(z1 ) − G(z2 ) = (z1 − z2 )G(z1 )G(z2 ).
(3.7)
(v) If M ∈ Mn is Hermitian and Im z = 0, then ||G(z)|| |Im z|−1 .
(3.8)
(vi) If M ∈ Mn is invertible and if {Gj k (z)}nj,k=1 is the matrix of its resolvent G(z) in any orthonormal basis, then |Gj k (z)|
||M −1 || . 1 − |z| ||M −1 ||
(3.9)
(vii) If M1 , M2 ∈ Mn their resolvents G1 (z), G2 (z) satisfy the “resolvent identity” G2 (z) = G1 (z) − G1 (z)(M2 − M1 )G2 (z).
(3.10)
(viii) The differential G (z) of the resolvent G(z) = (M − z)−1 viewed as function of M satisfies G (z) · X = −G(z)XG(z)
(3.11)
for any X ∈ Mn . In particular, ||G (z)|| ||G(z)||2 |Im z|−2 .
(3.12)
Now we present the main technical tool. PROPOSITION 3.2 ([5]). Let .: Mn → C be of class C 1 . Then, for any M ∈ Mn and any Hermitian X ∈ Mn : . (U ∗ MU ) · [X, U ∗ MU ] dU = 0, (3.13) U(n)
where [M1 , M2 ] denotes the commutator M1 M2 − M1 M2 and the integral denotes integration over U(n) with respect to the Haar measure.
10
VLADIMIR VASILCHUK
Proof. cf. [5], Proposition 3.2. The integral .(eiεX U ∗ MU e−iεX )[X, U ∗ MU ] dU = 0 U(n)
does not depend on ε. The derivative at ε = 0 gives (3.13).
✷
PROPOSITION 3.3. The system (2.22) has a unique solution in the class of functions f (z), 1,2 (z) which are analytic for Im z = 0 and which satisfy (2.9)–(2.11) and (2.23). Proof. Assume that there exist two solutions (f , 1,2 ) and (f , 1,2 ) of (2.22). Denote δf = f −f , δ1,2 = 1,2 −1,2 . Then, by using (2.22) and the following relations +∞ λNr (dλ) , r = 1, 2 fr (z) = −z−1 + z−1 λ−z 0 we obtain a linear system for δφ = zδf , δ1 , δ2 : (1 + a1 (z))δφ − b1 (z)δ1 − c1 (z)δ2 = 0, (1 + a2 (z))δφ − b2 (z)δ2 = 0, (1 + a3 (z))δφ − b3 (z)δ1 = 0,
(3.14)
where b1 = z2 , c1 = z1 , a1 = zf + zf , z2 J2 (s2 , s2 ) J2 (s2 , s2 ) , b = z , a2 = 2 (1 + zf )(1 + zf ) 1 + zf +∞ λN2 (dλ) J2 (z , z ) = (λ − z )(λ − z ) 0
(3.15) s2, =
1
z, 2 , + zf , (3.16)
and a3 , b3 can be obtained from a2 and b2 by replacing N2 and 2 by N1 and 1 in the above formulas. For any y0 > 0, consider the domain E(y0 ) = {z ∈ C | |Im z| y0 , |Re z| |Im z|}.
(3.17)
Due to condition (2.23) and the first equation of the system (2.22), we have for z ∈ E(y0 ) 1 + zf , (z) = −z−1 (m1 m2 + O(|Im z|−1 )), Besides, if
t (z , z ) = 0
+∞
λm(dλ) (λ − z )(λ − z )
z → ∞.
(3.18)
11
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
and m is a probability measure having finite second moment, then we have for z , z ∈ E(y0 ), +∞ λm(dλ)| |z z t (z , z ) − 0 +∞ 2 λ (z + z − λ)m(dλ) = (λ − z )(λ − z ) 0 +∞ λ2 m(dλ), 6y0−1 0
i.e.
+∞
z z t (z , z ) =
λm(dλ) + O(|Im z|−1 ),
0
z , z → ∞, z , z ∈ E(y0 ). Thus, from the relation above, (3.15), (3.16), (3.18) and condition (2.23), we obtain that for z → ∞, z ∈ E(y0 ) sr (z)sr (z)Jr (sr (z), sr (z)) = mr + o(1),
r = 1, 2.
Hence a1 (z) = −2 + o(1), c1 (z) = m1 + o(1),
a2,3 (z) = −1 + o(1), b2 (z) = m1 + o(1),
b1 (z) = m2 + o(1), b3 (z) = m2 + o(1).
Thus the determinant −(1 + a1 )b2 b3 + b1 b2 (1 + a3 ) + c1 b3 (1 + a2 ) of the system (3.15) is asymptotically equal to m1 m2 > 0. We conclude that if y0 is sufficiently large in (3.17), then (3.15) has only the trivial solution, i.e. (2.22) is uniquely soluble. ✷ Proof of Theorem 3.1. Due to the unitary invariance of eigenvalues of Hermitian matrices, we can assume without loss of generality that W = I in (2.2), i.e. we can work with the random matrix (2.18). We will omit below the subindex n in all cases when it will not lead to confusion. Consider the matrices 1/2 1/2 Hˆ n = H2,n H1,n H2,n
H˜ n = H1,nH2,n ,
(3.19)
and their resolvents 1/2 −1/2 ˜ G(z) = (H˜ n − z)−1 = H1,n G(z)H1,n ,
(3.20)
1/2 ˜ −1/2 ˆ G(z) = (Hˆ n − z)−1 = H2,n G(z)H 2,n 1/2
1/2
−1/2
−1/2
= H2,n H1,n G(z)H1,n H2,n ,
(3.21)
12
VLADIMIR VASILCHUK
where G(z) = (Hn − z)−1 is the resolvent of (2.2). Because of the trace property we derive from (3.20) ˜ ˆ n−1 Tr G(z) = n−1 Tr G(z) = n−1 Tr G(z) +∞ Nn (dλ) = gn (z), = λ−z 0
(3.22)
where gn (z) is Stieltjes transform of the NCM of (2.2). Consider the resolvent identity (3.10) for the pair (H˜ , 0): ˜ ˜ G(z)H 1 H2 = zG(z) + I.
(3.23)
Using Proposition 3.2 with .(M) = (H1 M − z)−1 ab , in view of (3.11), we obtain that ˜ 1 [X, H2 ]G)ab = 0. (GH
(3.24)
Choosing the matrix X with only Xab = 0,7 we obtain ˜ ˜ ˜ ˜ (3.25) (G(z)H 1 )aa (H2 G(z))bb = (G(z)H1 H2 )aa Gbb (z). Applying to this quantity n−1 na,b=1 and taking into account (3.23), we obtain the relation δ1,n (z)δ2,n(z) = gn (z) + zgn (z)gn (z),
(3.26)
where −1 ˜ δ1,n (z) = n−1 Tr G(z)H 1 = n Tr G(z)H1 , −1 ˜ ˆ δ2,n (z) = n−1 Tr G(z)H 2 = n Tr G(z)H2 .
(3.27) (3.28)
Introduce now the centered quantities gn◦ (z) = gn (z) − fn (z),
◦ δ2,n (z) = δ2,n (z) − 2,n (z),
(3.29)
where fn (z) = gn (z),
2,n (z) = δ2,n (z).
(3.30)
With these notations, (3.26) becomes (1 + zfn (z))fn(z) = 1,n (z)2,n(z) + r1,n (z),
(3.31)
7 In fact, in this relation (and in similar relations below) we can substitute only Hermitian matrices, e.g. (X(r) )pq = (δa,p δb,q + δa,q δb,p )/2 or (X(i) )pq = −i(δa,p δb,q − δa,q δb,p )/2 (here δ is the Kronecker symbol). Nevertheless, adding the relations obtained, we get relation (3.24) with (X)pq = (X(r) + iX(i) )pq = δa,p δb,q .
13
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
where ◦ (z)δ2,n(z) − gn◦ (z)zgn(z). r1,n (z) = δ1,n
(3.32)
In the next Theorem 3.2, we show that there exist y0 > 0, and C(y0 ) > 0, both independent of n, such that for z ∈ E(y0 ) (3.17) the variances v1 (z) = (gn◦ (z))2,
◦ v1+r (z) = (δr,n (z))2 ,
◦ v3 (z) = (δ2,n (z))2
satisfy v1 (z)
C(y0 ) , n2
v2 (z)
C(y0 ) , n2
v3 (z)
C(y0 ) . n2
(3.33)
In addition, according to Proposition 3.1, (3.22), (3.27) and (3.23), we have the estimates gn (z) |Im z|−1 ,
δr,n (z) m4 |Im z|−1 , Im z = 0, 1/4
r = 1, 2
and |zgn (z)| 1 + m4 |Im zt|−1 , Im z = 0. 1/2
These bounds and the Cauchy–Schwarz inequality for the average · imply that uniformly in n for z ∈ E(y0 ) |r1,n (z)| v2 v3 + (1 + m4 y0−1 )v1 √ C(y0 ) C(y0 ) 1/2 (1 + m4 y0−1 ) = O(n−1 ). + 2 n n 1/2 1/2
1/2
1/2
(3.34)
Now consider the matrix ∗ ˜ B. Y1 = B −1 U G(z)AU
(3.35)
It is clear that δ1,n (z) = n−1 Tr Y1 . On the other hand, using the resolvent identity ˜ ˜ (3.7) for the pair (G(z), G(0)), we obtain Y1 = B −1 + zZ1 ,
(3.36)
where ∗ ˜ . Z1 = B −1 U G(z)U
(3.37)
By using for the following function of the unitary matrix U ∗ ˜ )ac (B −1 U G(z)U
the analog of (3.13) derived from the left shift invariance of the Haar measure, we obtain ∗ ∗ ∗ ˜ ˜ ˜ ])ac − (H2−1 U G(z)H (B −1 [X, U G(z)U 1 U [H2 , X]U G(z)U )ac = 0.
14
VLADIMIR VASILCHUK
Choosing the matrix X with only Xbc = 0, we get ∗ ∗ ˜ ˜ )cc − (Y1 )ab (U G(z)U )cc (B −1 )ab (U G(z)U −1 ∗ ˜ = (Z1 )ab Icc − (Y1 B )ab (BU G(z)U )cc . Applying the operation n−1 nc=1 to (3.38), we obtain the matrix relation
Z1 = B −1 fn (z) − Y1 gn (z) + Y1 T ∗ δ2,n (z).
(3.38)
(3.39)
Regrouping the terms in (3.39) according to (3.29) and using (3.36), we obtain Z1 =
2 (z) Y1 B −1 + R, 1 + zfn (z)
(3.40)
where ◦ R = (1 + zfn(z))−1 (Y1 B −1 δ2,n (z) − Y1 gn◦ (z)).
(3.41)
Substituting (3.40) in (3.36), we get Y1 (B − z2 (z)(1 + zfn(z))−1 I ) = I + zRB.
(3.42)
Besides, using (3.20), (3.22) and the resolvent identity (3.23), we obtain z2,n (z) = −n−1 Tr H2 + n−1 Tr G(z)H1 H22 H1 , 1/2 2 1/2 ˆ z1,n (z) = −n−1 Tr H1 + n−1 Tr G(z)H 2 H1 H2 1/2
1/2
(3.43) (3.44)
and 1 + zfn (z) 1/2 1/2 = z−1 −n−1 Tr H1 H2 + n−1 Tr G(z)(H1 H2 H1 )2 .
(3.45)
In addition, using Proposition 6.1 with M1 = A and M2 = B, we obtain n−1 Tr H1 H2 = n−1 Tr H1 n−1 Tr H2
(3.46)
2 n−1 Tr (H1 H2 )2 2 n−1 Tr H12 n−1 Tr H2 + 2 + n−1 Tr H1 n−1 Tr H22 − 2 2 + n−1 Tr H1 n−1 Tr H2 + +n−3 Tr H12 n−1 Tr H22 .
(3.47)
and
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
15
Besides, using (3.5), (3.2), the trace property and (3.46), (3.47), we obtain −1 2 −1 n Tr G(z) H 1/2H2 H 1/2 2 n Tr (H1 H2 ) 1 1 |Im z| 2 8m4 8µ2 , |Im z| |Im z| 1/2 2 1/2 −1 −1
n Tr G(z)H 1/2H 2 H 1/2 n Tr H1 H2 H1 2 1 1 |Im z| −1 n Tr H1 n−1 Tr H22 = |Im z| 3/2
3/4
m µ2 4 |Im z| |Im z| and, similarly, 3/2 3/4 −1 m4 µ2 1/2 2 1/2 n Tr G(z)H ˆ , H H 2 1 1 |Im z| |Im z|
where
µ2 = max sup r=1,2
n
1/2
λ Nr,n (dλ) m4 . 2
From (3.43), (3.44), (3.45) and from the above estimates, we derive for z ∈ E(y0 ) and y0 sufficiently large, uniformly in n, zfn (z) = −1 + O(|Im z|−1 ).
(3.48)
In addition, according to condition (3.1), we have for some n sufficiently large and for all n n |1 + zfn (z)| |z|−1 (m1 m2 )/2, −1 2|z| (z) , r = 1, 2 r,n mr
(3.49)
and mr zr,n (z) Im z , 1 + zfn (z) 2m1 m2 m1 m2 zfn (z) Im z , r = 1, 2. Im r,n (z) 2mr
Im
(3.50)
Thus, the matrix P = B − z2,n (z)(1 + zfn (z))−1 I
(3.51)
16
VLADIMIR VASILCHUK
is uniformly in n invertible for z ∈ E(y0 ) and for y0 sufficiently large, we have ||P −1 ||
2m1 . |Im z|
(3.52)
Thus, we obtain from (3.42) z2 (z) + zRT P −1 , Y1 = G2 1 + zfn(z) where G2 (z) = (B − z)−1 . Applying the operation n−1 Tr to this relation and using (3.41), we obtain z2 (z) + r2,n (z), (3.53) 1 (z) = f2,n 1 + zfn (z) where −1
+∞
f2,n (z) = n Tr G2 (z) = 0
and r2,n (z) =
N2,n (dλ) , λ−z
Im z = 0
−1 z ∗ −1 ◦ ˜ n Tr U G(z)AU P δ2,n (z) − 1 + zf (z) ∗ −1 ◦ ˜ P gn (z) . −n−1 Tr G(z)BU GAU
(3.54)
(3.55)
Using the arguments above for the matrices ∗ ˜ ˜ U (z) = U G(z)U = (U AU ∗ B − z)−1 , G Z2 = U ∗ GU U A Y2 = AU ∗ BGU U A−1 ,
in which the roles A and B are interchanged, we obtain the analogous to (3.53) relation for 2,n (z). Thus, we obtain the system of relations fn (z) + zfn2 (z) = 1,n (z)2,n(z) + r1,n (z), z2,n (z) + r2,n (z), 1,n (z) = f2,n 1 + zfn(z) z1,n (z) + r3,n (z), 2,n (z) = f1,n 1 + zfn(z) where fr,n (z) = n−1 Tr Gr (z) =
+∞ 0
Nr,n (dλ) , λ−z
(3.56)
r = 1, 2.
(3.57)
In addition, according to (3.33), (3.49), (3.52) and the Cauchy–Schwarz inequality, we have for z ∈ E(y0 ) |rs,n (z)|
16 1/4 1/2 1/2 1/2 m4 v5−s + m4 v1 = O(n−1 ), m4−s
s = 2, 3.
(3.58)
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
17
Besides, for z ∈ C \ R+,ξ where ξ > 0 and R+,ξ is a ξ -neighborhood of the real positive semi-axis R+ , we have the following estimates: |fn (z)| ξ −1 ,
−1 |(n) . 1,2 (z)| m4 ξ 1/4
These estimates imply that {fn (z)} and {(n) 1,2 (z)} are sequences of analytic functions, bounded uniformly in n for z ∈ C \ R+,ξ . Thus these sequences are relatively compact with respect to uniform convergence on any compact subset of C\R+,ξ . In addition, according to the hypothesis of the theorem, the normalized counting measures N1,n , N2,n of H1,n , H2,n converge weakly to limiting measures N1 , N2 . Thus their Stieltjes transforms (3.57) converge uniformly on every compact of E(y0 ), y0 > ξ to the Stieltjes transforms f1,2 (z) of N1,2 . Hence, if y0 is large enough, there exist three functions f , 1 , 2 which are analytic in E(y0 ) and which satisfy the limiting system (2.22) for z ∈ E(y0 ). Its unique solubility in (3.17) where y0 is large enough is proved in Proposition 3.3. Besides, the three functions fn , 1 , 2 are a-priori analytic for z ∈ C \ R+ . Thus, their limits f , 1 , 2 are also analytic for z ∈ C \ R+ . In view of the weak compactness of probability measures and the continuity of the one-to-one correspondence between nonnegative measures and their Stieltjes transforms (see Proposition 2.1(v)), there exists a unique nonnegative measure N such that f admits the representation (2.21). This measure N is a probability measure in view of (3.48). We conclude that the whole sequence {fn(z)} converges uniformly on every compact subset of C \ R+,ξ , ξ > 0 to the limiting function f (z) verifying (2.22). This result, Theorem 3.2 and the Borel–Cantelli lemma imply that the sequence {gn (z)} where gn (z) is defined in (3.22) converge with probability 1 to f (z) for any fixed z ∈ E(y0 ). Since the convergence of a sequence of analytical functions on any countable set having an accumulation point in their common domain of definition implies the uniform convergence of the sequence on any compact of the domain, we obtain the convergence gn (z) to f (z) with probability 1 on any compact of C \ R+,ξ . Due to the continuity of the one-to-one correspondence between probability measures and their Stieltjes transforms, the normalized counting measure (NCM) of the eigenvalues of random matrix (2.2) converge weakly with probability 1 to the non-random measure N whose Stieltjes transform (2.21) satisfies (2.22). Theorem 3.1 is proved. ✷ THEOREM 3.2. Let Hn be a random matrix of the form (2.2) satisfying the conditions of Theorem 3.1. Denote gn (z) = n−1 Tr (Hn − z)−1 = n−1 Tr (H˜ n − z)−1 , δr,n (z) = n−1 Tr Hr,n (H˜ n − z)−1 , r = 1, 2, where H˜ n is defined in (3.19).
(3.59)
18
VLADIMIR VASILCHUK
Then there exist y0 and C(y0 ), both positive and independent of n, such that the variances of the random variables (3.59) satisfy for any z ∈ E(y0 ) C(y0 ) , (3.60) v1 = |gn (z) − gn (z)|2 n2 C(y0 ) , r = 1, 2. (3.61) v1+r = |δr,n (z) − δr,n (z)|2 n2 Proof. We will derive and analyze the system of inequalities vi
3
1/2 1/2
αij (y0 )vi vj
j =1,j =i
+
βi (y0 ) , n2
i = 1, 2, 3.
(3.62)
Below we will use the notations g(z) and δr (z) for gn (z) and δr,n (z), r = 1, 2 and the notations 1 and 2 for two values z1 and z2 of the complex spectral parameter z. We assume that |Im z1,2 | y0 > 0. Consider the matrix ∗ ˜ Q1 = g ◦ (1)z2−1 BU G(2)AU , (3.63) where g ◦ (1) = g(1) − g(1). It follows from (3.23) that n−1 Tr Q1 for z1 = z and z2 = z is the variance (3.60), that we denote v1 (z): v1 (z) = n−1 Tr Q1 |z1 =z,z2 =z = |g ◦ (z)|2.
(3.64)
By using for the following function of the unitary matrix U ∗ ◦ −1 ∗ ˜ ˜ z2 BU G(2)AU U G(1)U cd
aa
that is the analog of (3.13) derived from the left shift invariance of the Haar measure, we obtain ∗ ∗ ˜ ˜ ])aa (BU G(2)AU )cd − z−1 (([X, U G(1)U 2
∗ ∗ ∗ ˜ ˜ ˜ [B, X]U G(1)U )aa (BU G(2)AU )cd + −(U G(1)AU ∗ ◦ ∗ ˜ ˜ )aa (B[X, U G(2)AU ])cd − +(U G(1)U ∗ ◦ ∗ ∗ ˜ ˜ ˜ )aa (BU G(2)AU [B, X]U G(2)U )cd ) = 0. −(U G(1)U
Choosing in the above relation a matrix X with only Xbd = 0, applying to the result the operation n−2 na,d=1 and taking into account that ◦
g (z) = n
−1
n
G◦aa (z),
a=1
we obtain the matrix relation ∗ ˜ ˜ 2 (1)AU ∗ − G ˜ 2 (1)AU ∗ B)+ BU G z2−1 n−2 BU G(2)A(U ∗ ˜ + +Bg ◦ (1)δ1 (2) − g ◦ (1)BU G(2)AU ◦ ∗ ˜ +g (1)BU G(2)AU (1 + z2 g(2))− ∗ ˜ Bδ1 (2) = 0. −g ◦ (1)BU G(2)AU
(3.65)
19
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
Regrouping the terms in (3.65) according to (3.29) and taking into account that ∗ ∗ ˜ ˜ B − B) = U A−1 G(2)AU B, z2−1 (BU G(2)AU
we obtain Q1 (z2 f (2) − 1 (2)B) ∗ ˜ = −g ◦ (1)g ◦ (2)BU G(2)AU + ◦ ◦ −1 ˜ + g (1)δ1 (2)U A G(2)AU ∗ B + R˜ 1 ,
(3.66)
where ∗ ˜ ˜ 2 (1)AU ∗ − G ˜ 2 (1)AU ∗ B). BU G R˜ 1 = n−2 z2−1 BU G(2)A(U
(3.67)
Besides, according to (3.49) and (3.50), the matrix P2 = (z2 f (2) − 1 (2)B) = 1 (2)(z2 f (2)−1 1 (2) − B) is uniformly in n invertible for z ∈ E(y0 ) and ||P2−1 ||
8 . m2
Multiplying (3.66) by P2−1 from the right and applying the operation n−1 Tr to the result, we obtain in the left-hand side the variance v1 (z) of (3.60). In view of (3.9), (3.49), (3.20) and the trace property, the terms in right-hand side can be estimated for z2 ∈ E(y0 ) as follows 1/2 8m4 −1 ◦ ◦ −1 1/2 1/2 ∗ v1 ; (i) |g (1)g (2)n Tr P2 BU A G(2)A U | m2 y0 (ii) |g ◦ (1)δ1◦ (2)n−1 Tr P2−1 U A−1/2 G(2)A1/2 U ∗ B| (iii)
1/2
8m4 1/2 1/2 v v ; m2 y0 1 2
|z2 |−1 |n−3 Tr P2−1 BU A1/2 G(2)A1/2 (U ∗ BU A1/2 G2 (1)A1/2 U ∗ − −3 −A1/2 G2 (1)A1/2 U ∗ B)| n−2 16m4 m−1 2 y0 .
−1 1/2 to the first inequality of (3.62), in These bounds lead for 8m4 m−1 2 y0 which 1/2
−1 −1 −3 α12 (y0 ) = 16m4 m−1 2 y0 , α13 (y0 ) = 0, β1 = 32m4 m2 y0 . 1/2
(3.68)
To obtain the second inequality of the system, consider the matrix ˜ Q2 = δ1◦ (1)G(2)A.
(3.69)
Applying to Q2 operation n−1 Tr and setting z1 = z, z2 = z, we obtain the variance v2 of (3.61). On the other hand, using Proposition 3.2 for the function ◦ ˜ ˜ .(M) = (G(1)A) aa (G(2)A)cd ,
20
VLADIMIR VASILCHUK
˜ where G(z) = (AU ∗ MU − z)−1 , we obtain relation ˜ ˜ ˜ G(1)A G(2)A+ n−2 (−U ∗ BU G(1)A ˜ ˜ H˜ G(2)A)− ˜ +G(1)A G(1) ◦ ∗ ˜ −δ1 (1)δ1 (2)U BU G(2)A+
˜ ˜ = 0. + δ1◦ (1)G(2)A +δ1◦ (1)g(2)z2 G(2)A
(3.70)
We do this by identifying M and B and performing almost the same procedure as that used in the derivation of (3.65), in particular, choosing for the matrix X the matrix with only the (c, b)th entry nonzero. Regrouping terms in (3.70) according ˜ ˜ = zA−1 G(2) + A−1 , we obtain to (3.29) and taking into account that U ∗ BU G(z) ((1 + z2 f (2))I − z2 1 (2)A−1 )Q2 ˜ ˜ = δ1◦ (1)δ1◦ (2)U ∗ BU G(2)A − δ1◦ (1)g ◦ (2)z2 G(2)A + R˜ 2 ,
(3.71)
where ˜ ˜ ˜ G(1)A G(2)A − R˜ 2 = n−2 (U ∗ BU G(1)A ˜ ˜ ˜ ˜ −G(1)AG(1)H G(2)A).
(3.72)
Multiplying (3.71) by A from the left, we get ˜ − P˜ Q2 = ((1 + z2 f (2))−1 (δ1◦ (1)δ1◦ (2)H˜ G(2)A ◦ ◦ ˜ ˜ −δ1 (1)g (2)z2 AG(2)A + AR2 ),
(3.73)
where P˜ = A−z2 1 (2)(1+z2 f (2))−1 I . It follows from (3.50) that for z2 ∈ E(y0 ) where y0 is sufficiently large, the matrix P˜ is invertible uniformly in n and 2m2 . ||P˜ −1 || |Im z2 | Multiplying (3.73) by P˜ −1 from the left and applying the operation n−1 Tr to the result, we obtain in the left-hand side the variance v2 (z) of (3.61). In view of (3.9), (3.49), (3.20) and the trace property, the terms in right-hand side can be estimated for z ∈ E(y0 ) as follows 1/4 8m4 −1 ◦ ◦ −1 −1 ˜ v2 ; (i) |1 + z2 f (2)| |δ1 (1)δ1 (2)n Tr P H G(2)A| m1 y0 (ii) |1 + z2 f (2)|−1 |δ1◦ (1)g ◦ (2)n−1 Tr P˜ −1 z2 AG(2)A| (iii)
|1 + z2 f (2)|−1 |n−3 Tr P˜ −1 A(H G(1)AG(1)AG(2)− 5/4 −2 −G(1)AG(1)H G(2)A)| n−2 16m4 m−1 1 y0 .
1/2
16m4 1/2 1/2 v v ; m1 2 1
21
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
−1 These bounds lead for 8m4 m−1 1 y0 1/2 to the second inequality of (3.62), in which 1/4
α21 (y0 ) = 32m4 m−1 1 , 1/2
−2 β2 = 32m4 m−1 1 y0 . 5/4
α23 (y0 ) = 0,
(3.74)
∗ ˜ Using the arguments above for the matrix Q3 = δ2◦ (1)BU G(2)U , we obtain the third inequality of (3.62) where
α31 = α21 = 32m4 m−1 1 , α32 = α23 = 0, 5/4 −2 β3 = β2 = 32m4 m−1 1 y0 . 1/2
(3.75)
Introducing new variables 1/2 1/2
u1 = y0 v1 ,
1/2
u2 = v2 ,
1/2
u3 = v3 ,
we obtain from (3.68), (3.74) and (3.75) the system u2i
3
j =1,j =i
aij ui uj +
γi , n2
i = 1, 2, 3,
(3.76) −1/2
where the coefficients {aij , i = j } have the form aij = y0 bij with bij and γi bounded in y0 and n as y0 → ∞ and n → ∞. By choosing y0 sufficiently large (and then fixing it), we can guarantee that 0 aij 1/4, i = j . Summing the three relations (3.76), we can write the result in the form (cu, u) γ /n2 where γ = γ1 + γ2 + γ3 and the minimum eigenvalue of the matrix c is 1/2. Thus we obtain the bounds (3.60) and (3.61). Theorem 3.2 is proved. ✷
4. Convergence in Probability for Random An and Bn According to Theorem 3.2, the randomness of Un in (2.2) (or (2.18)) already allows us to prove that the variance of the Stieltjes transform of the NCM (2.4) vanishes as n → ∞. Therefore, we have only to prove that the additional randomness due to the matrices An and Bn in (2.2) does not destroy this property. We will prove this fact first for An and Bn whose norms are uniformly bounded in n (see Lemma 4.1 below), and then for the general case of Theorem 2.1 by using some truncation procedure. PROPOSITION 4.1 ([5]). Let {mn } be a sequence of random measures defined on some probability space and {sn } be the sequence of their Stieltjes transforms. Then: (i) The sequence mn converges in probability to a nonrandom probability measure m if and only if the sequence {sn } converges in probability for any fixed z
22
VLADIMIR VASILCHUK
belonging to some compact set K ⊂ {z ∈ C | Im z > 0} to the Stieltjes transform f of the measure m. (ii) We can replace the requirement of their convergence for any z belonging to a certain compact of C± by the convergence for any z belonging to any interval of the imaginary axis, i.e. for z = iy, y ∈ [y1 , y2 ], y1 > 0. (iii) If {mn } is a sequence of random nonnegative measures converging weakly in probability to a nonrandom nonnegative measure m, then the Stieltjes transforms sn of mn and the Stieltjes transform s of m are related by (4.1) lim E sup |sn (z) − s(z)| = 0 n→∞
z∈K
for any compact of C± . Proof. cf. [5], Proposition 4.1.
✷
LEMMA 4.1. Let Hn be the random n×n matrix of the form (2.2) in which An and Bn are random positive definite Hermitian matrices, Un and Vn are random unitary matrices distributed each according to the Haar measure on U(n) and An , Bn , Un and Vn are mutually independent. Assume that the normalized counting measures Nr,n , r = 1, 2 of matrices An and Bn converge weakly in probability as n → ∞ to the nonrandom nonnegative and probability measures Nr , r = 1, 2 respectively and that sup ||An || T < ∞,
sup ||Bn || T < ∞,
n
(4.2)
n
and condition (2.20) of Theorem 2.2 is fulfilled. Then the normalized counting measure Nn of Hn converges weakly in probability to a nonrandom probability measure whose Stieltjes transform (2.21) is a unique solution of the system (2.22) in the class of functions f (z), 1,2 (z) analytic for z ∈ C \ R+ and satisfying conditions (2.9)–(2.11) and (2.23). Proof. In view of Proposition 4.1(ii) it suffices to show that lim E{|gn (z) − f (z)|} = 0
n→∞
for any z belonging to some interval of the imaginary axis, i.e. z = iy,
y ∈ [y1 , y2 ],
0 < y1 < y2 < ∞.
Condition (4.2) implies +∞ λNr,n (dλ) − mr ε → 0, P
r = 1, 2, as n → ∞.
(4.3)
(4.4)
0
Thus, for any sufficiently small η > 0 there exists n(η) such that for all n n(η) +∞ 12 10 mr λNr,n (dλ) mr 1 − η, r = 1, 2. (4.5) P 11 11 0
23
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
Denote by ∗ the event whose probability is written in the left-hand side of (4.5) and by E∗ { · } the average over product of ∗ and two copies of the group U (n) for Un and Wn . As a result of (4.5), for z ∈ [y1 , y2 ], we have lim E{|gn (z) − f (z)|} 2y1−1 η + lim E∗ {|gn (z) − f (z)|}
n→∞
n→∞
and for all n n(η) and any realization of An and Bn on ∗ , we have +∞ 12 10 mr λNr,n (dλ) mr . 11 11 0
(4.6)
Thus, to complete the proof we have to show that for any z = iy, y ∈ [y1 , y2 ] lim E∗ {|gn (z) − f (z)|} = 0.
(4.7)
n→∞
Since the relation (4.4) and condition (4.2) of the lemma imply evidently the conditions (3.1) and (3.2) of Theorems 3.1 and 3.2, all the results obtained in these theorems are valid in our case for any fixed realization of random matrices An and Bn on ∗ . In addition, all n-independent estimating quantities entering various bounds in the proofs of these theorems and depending on the moments m1 , m2 , m4 in (3.1), (3.2) and on n , ξ and y0 will now depend on m1 , m2 , T and on n(η), ξ and y1 , y2 in (4.3), but not on any particular realization of random matrices An and Bn . Using (3.33), we can write that E∗ {|gn (z) − gn (z)|} E∗ {|v1 (z)|} 1/2
C , n
where the symbol · denotes, as above, the expectation with respect the Haar measure on U (n). Thus, it suffices to show lim E∗ {|gn (z) − f (z)|} = 0,
n→∞
z = iy, y ∈ [y1 , y2 ],
(4.8)
where y1 is big enough. Introduce the quantities γn (y) = iy(gn (iy) − f (iy)),
γr,n (y) = δr,n (iy) − r (iy),
r = 1, 2.
By using the first equation of system (2.22) and first relation of (3.56), we can write the identity γn (y)(1 + iy(gn (iy) + f (iy))) = iy2 (iy)γ1,n (y) + iyδ1,n (iy)γ2,n (y) + iyr1,n (iy).
(4.9)
On other hand, using the integral representation (2.8), from the first two equations of (2.22), we obtain 1 + iyf (iy) = ρ2 (t2 (y)),
(4.10)
24
VLADIMIR VASILCHUK
where
+∞
ρ2 (z) = 1 + zf2 (z) = 0
λN2 (dλ) , λ−z
t2 (y) =
iy2 (iy) . 1 + iyf (iy)
(4.11)
Using (4.10), we obtain γn (y) = [ρ2 (t2,n (y)) − ρ2 (t2 (y))] + ε1,n (y),
(4.12)
where ε1,n (y) = [1 + iygn (iy) − ρ2 (t2,n (y))], iyδ2,n (iy) . t2,n (y) = 1 + iygn (iy)
(4.13)
We have E∗ {|ε1,n (y)|} E∗ {|1 + iygn (iy) − ρ2,n(t2,n (y))|} + +E∗ {|ρ2,n(t2,n (y)) − ρ2 (t2,n (y))|}, where
ρ2,n (z) = 1 + zg2,n (z) = 0
+∞
λN2,n (dλ) λ−z
(4.14)
(4.15)
and where g2,n (z) is the Stieltjes transform of the random NCM N2,n of H2,n (cf. (3.57)) +∞ N2,n (dλ) −1 . g2,n (z) = n Tr G2 (z) = λ−z 0 Besides, by using two first relations of (3.56), we obtain 1 + iygn (iy) = ρ2,n (t2,n (y)) + rˆ2,n (iy)t2,n (y) − r1,n (iy)(1 + gn (iy))−1 , where r2,n (z) =
−1 z ◦ n Tr Y1 P˜ −1 δ2,n (z)− 1 + zgn (z) −n−1 Tr U G(z)AU BU ∗ P˜ −1 gn◦ (z) ,
P˜ −1 = G2 (t2,n (z)), gn◦ (z) = gn (z) − gn (z), ◦ (z) = δr,n (z) − δr,n (z), δr,n
r = 1, 2
are the respective random variables centered by the partial expectations with respect to the Haar measure. Using (4.6) we obtain the analogs of (3.49) and (3.50).
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
25
In our case, we have for all n n(η) and y ∈ [y1 , y2 ], y1 sufficiently large |1 + iyfn (iy)| y −1 (m1 m2 )/3, 3y , |t2,n (y)| m1 y . Im t2,n (y) 3m1
(4.16)
In addition, we have the analoge of (3.58) C . n This leads to the following bound for the first term in the right-hand side of (4.14): | r2,n (z)|
E∗ {|1 + iygn (iy) − ρ2,n(t2,n (y))|} 1 ∗ 3y2 ∗ E {|r1,n (iy)|} E {|ˆr2,n (iy)|} + m1 m2 C . n As for the second term, using (4.16) we obtain for it the following bound E∗ {|ρ2,n(t2,n (y)) − ρ2 (t2,n (y))|} 3y2 sup E∗ {|g2,n (z) − f2 (z)|} m1 z∈K 3y2 sup E{|g2,n (z) − f2 (z)|}, m1 z∈K
(4.17)
where K is a compact subset of C+ y1 3y2 K = z ∈ C : Im z , |z| . 3m1 m1 The right-hand side of (4.17) tends to zero as n → ∞ in view of the hypothesis of Theorem 2.1 and Proposition 3.3. Thus, there exist 0 < y1 < y2 < ∞ such that for for all y ∈ [y1 , y2 ], lim E∗ {|ε1,n (y)|} = 0.
n→∞
(4.18)
Analogous arguments show that limn→∞ E∗ {|ε2,n(y)|} = 0, ε2,n (y), where ε2,n (y) is defined from (4.13) and (4.11) by interchanging the indexes 1 and 2. Consider now the first term in the right-hand side of (4.12). In view of (4.11) and (4.15), we can write this term in the form [ρ2 (t2,n (y)) − ρ2 (t2 (y))] iy t2,n (y) J2 γn + J2 γ2,n =− 1 + iyf (iy) 1 + iygn (iy) = −a1 γn + b2 γ2,n ,
26
VLADIMIR VASILCHUK
where J2 , a2 and b2 are defined by formulas (3.15) and (3.16), in which we have to replace 2 , 2 , f and f by δ2,n , 2 , gn and f , respectively. Denote by . = {.ij }3i,j =1 the matrix defined by the left-hand side of system (3.15) and by H = {Hi }3i=1 the vector with components H1 = γn , H2 = γ1,n, H3 = γ2,n . It is easy to check using (4.6) that for all n n(η), y ∈ [y1 , y2 ] and y1 sufficiently large, we have m1 m2 det . , 2 i.e. matrix . is uniformly invertible in n and y. Then we have from (4.9) E{|(.H)1 |} E{|yr1,n |}.
(4.19)
Besides, relations (4.12) and (4.18) lead to E{|(.H)2 |} E{|ε1,n|}.
(4.20)
Interchanging indices 1 and 2, in the above arguments we also obtain that E{|(.H)3 |} E{|ε2,n|}.
(4.21)
Denote by || · ||1 the l 1 -norm of C3 and by || · || the induced matrix norm. Then we have E{||H||1 } E{||.−1 .H||} E1/2 {||.−1 ||2 }E1/2 {||.H||21 }.
(4.22)
It follows from our arguments above that all entries of the matrices . and .−1 and all components of the vector H are bounded uniformly in n and in realizations of random matrices An , Bn , Un and Wn in (2.2). Thus, we have ||.−1 ||
3
|(.−1 )ij | C,
||.H||1
i,j =1
3
|.ij ||H|j C.
i,j =1
These bounds and (4.19)–(4.22) imply that E{||H||1 } C 3/2 (E{|r1,n |} + E{|ε1,n|} + E{|ε2,n|})1/2 . In view of (4.18), this inequality imply (4.8), i.e. the assertion of the lemma.
✷
Proof of Theorem 2.1. For any T > 0 introduce the matrices ATn and BnT reT the NCMs of placing eigenvalues An and Bn lying in ]T , ∞[ by T . Denote by Nr,n T T T An and Bn . It is clear that Nr,n converges weakly in probability to NrT as n → ∞ where NrT is defined via Nr by the formula ⊂ ]−∞, T [, Nr (), T = {T }, (4.23) Nr () = Nr ([T , ∞[), 0, ⊂ ]T , ∞[,
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
27
for any Borel set ∈ R. Now denote T = (H1 )1/2H2T (H1 )1/2 , H T = (H2T )1/2 H1 (H2T )1/2, H H T = (H2T )1/2 H1T (H2T )1/2 , where H1T = V ∗ AT V ,
H2T = U ∗ B T U.
nT and H nT − HnT do not exceed It is clear that the dimensions of ranges of Hn − H T T the ranges of H1,n − H1,n and of H2,n − H2,n, respectively, i.e. they do not exceed the number of those [A]i and [B]i which are larger than T . Therefore, the NCMs nT (λ), N nT (λ) and NnT (λ) of matrices Hn , H T , H T and HnT satisfy the Nn (λ), N inequalities +∞ #{|[B]i | T } T = N2,n (dµ), (4.24) |Nn (λ) − Nn (λ)| n T +∞ nT (λ) − NnT (λ)| #{|[A]i | T } = N1,n (dµ). (4.25) |N n T nT , N nT and NnT , respecgnT (z) and gnT (z) be Stieltjes transforms of N Let gnT (z), tively. Because of the trace property, we have nT − z)−1 = n−1 Tr (H nT − z)−1 = gnT (z). gnT (z) = n−1 Tr (H Besides, it follows by definition and from (4.24) and (4.25), that if gn (z) is the Stieltjes transform of Nn , then we have uniformly on z ∈ E(y0 ) gnT (z) − gn (z)| + | gnT (z) − gnT (z)| |gnT (z) − gn (z)| | +∞ +∞ 1 N1,n (dµ) + N2,n (dµ) . y0 T T Hence E{|gnT (z) − gn (z)|} +∞ +∞ 1 E{N1,n (dµ)} + E{N2,n(dµ)} . y0 T T
(4.26)
Since the norms of matrices H1T and H2T are bounded, the results of the Lemma 4.1 is applicable to the function gnT (z), so that, in particular, for n → ∞ it converges in probability to a function f T (z) obeying the system f T (z)(1 + zf T (z)) = T1 (z)T2 (z), zT2 (z) T T , 1 (z) = f2 1 + zf T (z) zT1 (z) . T2 (z) = f1T 1 + zf T (z)
28
VLADIMIR VASILCHUK
T In addition, since E{gnT (z)} and E{δr,n (z)}, r = 1, 2 are bounded uniformly in n and T for z ∈ C \ R+,ξ
1 |E{gnT (z)}| , ξ 1 +∞ T T λE{Nr,n (dλ)} |E{δr,n (z)}| ξ 0 (m(2))1/2 1 +∞ , λE{Nr,n (dλ)} ξ 0 ξ we have |f T (z)|
1 , ξ
|T1,2 (z)|
(m(2) )1/2 . ξ
(4.27)
Besides, as a result of (3.43), (3.44) and (3.45), we have 1 + zf T (z) = −z−1 mT1 mT2 + O(|Im z|−1 ) , zT1,2 (z) = −mT1,2 + O(|Im z|−1 ), where
mTr =
0
+∞
T λE{Nr,n (dλ)},
(4.28)
r = 1, 2.
Besides, for any sequence Tk → ∞, the sequences of analytic functions {f Tk (z)} k (z)} are relatively compact with respect to uniform convergence on any and {T1,2 Tk compact subset of C \ R+,ξ . In addition, according to (4.23), the measures N1,2 converge weakly to the limiting measures N1,2 and mT1,2 → m1,2 ,
T → ∞.
(4.29)
Hence, there exist three analytic functions f (z), 1 (z), 2 (z) verifying (2.22). (4.28) and (4.29) imply that f (z), 1 (z), 2 (z) satisfy the conditions of Proposition 3.3 and, hence, they are uniquely defined. Furthermore, for z ∈ E(y0 ), we have E{|gn (z) − f (z)|} T
T
E{|gn (z) − gn k (z)|} + E{|gn k (z) − f Tk (z)|} + |f Tk (z) − f (z)|, where f Tk (z) denotes a convergent subsequence of f Tk (z). Hence, in view of (4.26), the arguments above, and Lemma 6.1, we conclude that for each z ∈ E(y0 ) lim E{|gn (z) − f (z)|} = 0.
n→∞
In view of Proposition 4.1, we conclude that the NCM of random matrices (2.2) converge weakly in probability to the nonrandom measure N whose Stieltjes transform f (z) satisfies (2.22). Theorem 2.1 is proved. ✷
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
29
5. Convergence with Probability 1 for Nonrandom Sn , T n and in Probability for Random Sn , T n The proof of Theorem 2.2 follows the scheme of the proof of Theorem 2.1. In this section, we briefly describe the steps that corespond to those used in the previous sections. The first step of the proof of Theorem 2.2 consists in the following statements (cf. Theorem 3.1 and Proposition 3.3). The following Theorem 5.1 generalizes Voiculescu’s result on the multiplicative free convolution of probability measures on the unit circle [8] proved under the condition that the NCMs ν1,2 have nonzero first moments, i.e. 2π eiθ νr (dθ) = 0, r = 1, 2. 0
THEOREM 5.1 ([7]). Let Vn be a random n × n matrix of the form (2.1) in which Sn and Tn are nonrandom unitary matrices, Un and Wn are random independent unitary matrices distributed each according to the Haar measure on U(n). Assume that the normalized counting measures νr,n , r = 1, 2 of Sn and Tn converge weakly as n → ∞ to the nonnegative and probability measures on the unit circle νr , r = 1, 2, respectively. Then the normalized counting measure of νn converges in probability to a nonrandom nonnegative and probability measure ν whose Herglotz transform (2.24) is the unique solution of the system (2.25) in the class of functions h(z), 1,2 (z) analytic for |z| < 1 and satisfying conditions (2.14)–(2.15) and (2.26). PROPOSITION 5.1. System (2.25) has a unique solution in the class of functions h(z), 1,2 (z) analytic for |z| < 1 and satisfying conditions (2.14)–(2.15) and (2.26). Proof. Assume that there exist two solutions (h , 1,2 ) and (h , 1,2 ) of the system. Denote δh = h − h , δ1,2 = 1,2 − 1,2 and δf = f − f , where f , = (h, − 1)/2z. Then, by using (2.24) and the integral representation (2.13) for h1,2, we obtain the linear system for δf , δ1,2 (1 + a1 (z))δf − b1 (z)δ1 − c1 (z)δ2 = 0, a2 (z)δf + δ1 − b2 (z)δ2 = 0, a3 (z)δf − b3 (z)δ1 + δ2 = 0,
(5.1)
where b1 = 2 , c1 = 1 , a1 = z(f + f ), s I2 (s2 , s2 ) I2 (s2 , s2 ) z, , 2 , b = z , s = , a2 = z 2 2 2 1 + zf 1 + zf 1 + zf , 2π ν2 (dθ) I2 (z , z ) = iθ − z )(eiθ − z ) (e 0
(5.2) (5.3)
30
VLADIMIR VASILCHUK
and a3 , b3 can be obtained from a2 and b2 by replacing subindexes 2 by 1 in above formulas. For any 0 < d0 1/4 consider the domain D(d0 ) = {z ∈ C : |z| d0 }.
(5.4)
By using (2.14) and (2.26), we obtain for z ∈ D(d0 ) that , I1,2 (s , s ) 4 s 1/2, 1,2 1,2 1,2 and, hence, ar = o(1),
r = 1, 2, 3,
b2,3 = o(1),
z → 0.
Thus, the determinant 1+ a1 − b2 b3 + b1 a2 + c1 a3 − a1 b2 b3 + b1 b3 a3 + c1 a2 b3 of the system (5.1) is equal asymptotically to 1. We conclude that if d0 in (5.4) is small enough, then system (5.1) has only trivial solution, i.e. system (2.25) is uniquely soluble. ✷ The proof of Theorem 5.1 coincides with the proof of Theorem 3.1 modulo to the substitution of matrices Sn and Tn instead of matrices An and Bn (and V1 and V2 instead of H1 and H2 correspondently). The proof for the unitary case is much simpler, e.g. the bounds analogous to (3.49) and (3.50) will be follows (and also uniform in n) 1 , |fn (z)| 1 − |z| z2,n (z) 1 + zf (z) 1/2, n
|2,n (z)|
1 , 1 − |z|
|z| < 1/4.
Thus, for |z| < 1/4 the matrix T − z2,n(z)(1 + zfn (z))−1 I will be invertible uniformly in n and we need not require the first moments of measures νr,n , r = 1, 2 to be nonzero. The proof is also based on the following properties of variances. THEOREM 5.2 ([7]). Let Vn be the random matrix of the form (2.1) satisfying the condition of Theorem 5.1. Denote gn (z) = n−1 Tr (V − z)−1 ,
δr,n (z) = n−1 Tr Vr (V − z)−1 ,
r = 1, 2. (5.5)
Then there exist d0 and C(d0 ), both independent of n and such that the variances of random variables (5.5) admit the bounds for |z| d0 < 1 C(d0 ) , n2 C(d0 ) = |δr,n (z) − δr,n (z)|2 , n2
v1 = |(gn (z) − gn (z))|2 v1+r
(5.6) r = 1, 2.
(5.7)
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
31
Proof. The proof follows the scheme of the proof of Theorem 3.2, i.e. we derive and analyze the system of inequalities (3.62) (see [7] for more details). ✷ In addition, using the following relation between Stieltjes and Herglotz transforms h(z) − 1 , f (z) = 2z one can easily establish that (2.25) and (2.22) are equivalent. Now we are going to prove Theorem 2.2. According to Theorem 5.2, the randomness of Un in (2.1) (or (2.17)) an already provides vanishing variance of the gn (z) and, hence, it also provides vanishing of the variance of the Herglotz transform of the NCM (2.3) (see (5.6)). Thus we have only to prove that the additional randomness due to the matrices Sn and Tn in (2.1) does not destroy this property. We use the following analog of Proposition 4.1. PROPOSITION 5.2. Let {µn } be a sequence of random measures on the unit circle defined on a certain probability space and {hn } be the sequence of their Herglotz transforms, then (i) the sequence {µn } converges in probability to a nonrandom probability measure µ on the unit circle if and only if the sequence {hn } converges in probability for any fixed z belonging to a certain compact K ⊂ {z ∈ C | |z| < 1} to the Herglotz transform h of the measure µ; (ii) if {µn } is a sequence of random nonnegative measures on the unit circle converging weakly in probability to a nonrandom nonnegative measure µ and if hn (z) are the Herglotz transforms of µn and h(z) is the Herglotz transform of µ then the functions h(z) − 1 hn (z) − 1 , p(z) = pn (z) = 2z 2z are related as follows: lim E sup |pn (z) − p(z)| = 0 n→∞
z∈K
for any compact of {z ∈ C | |z| < 1}. This proposition can be easily proved by repeating the proof of Proposition 4.1 from [5]. Proof of Theorem 2.2. In view of Proposition 5.2, it suffices to show that lim E{|qn (z) − h(z)|} = 2|z| lim E{|gn (z) − f (z)|} = 0
n→∞
n→∞
for any z belonging to a certain compact D(d0 ) of {z ∈ C : |z| < 1}, where f (z) = (h(z) − 1)/2z and 2π iµ e +z νn (dµ), |z| < 1. qn (z) = eiµ − z 0
32
VLADIMIR VASILCHUK
The results obtained in Theorems 5.1 and 5.2 are valid in our case for any fixed realization of random matrices Sn and Tn . In addition, all n-independent estimating quantities entering various bounds in the proofs of these theorems and depending on d0 will depend now also on d0 but not on any particular realization of random matrices Sn and Tn . We will denote below all these quantities simply by the unique symbol C that may have different value in different formulas. In particular, denoting as above by · the expectation with respect to the Haar measure and using (5.6), we can write that 1/2
E{|gn (z) − gn (z)|} E{|v1 (z)|}
C . n2
Thus, it suffices to show lim E{|gn (z) − f (z)|} = 0,
n→∞
z ∈ D(d0 ),
(5.8)
where d0 is small enough. Introduce the quantities γn (z) = (gn (z) − f (z)),
γr,n (z) = δr,n (z) − r (z),
r = 1, 2. (5.9)
The first equation of system (2.25) and (3.31) leads to the identity γn (z)(1 + z(gn (z) + f (z))) = 2 (z)γ1,n(z) + δ1,n (z)γ2,n(z) + r1,n (z).
(5.10)
By using the second equation of system (2.25), we can write the identity γ1,n (z) = f2 (t2,n (z)) − f2 (t2 (z)) + ε1,n (z),
(5.11)
where 2π h2 (z) − 1 ν2 (dθ) = , iθ − z 2z e 0 ε1,n (z) = δ1,n (z) − f2 (t2,n (z)), zδ2,n (z) z2 (z) , t2 (y) = . t2,n (z) = 1 + zgn (z) 1 + zf (z)
f2 (z) =
(5.12) (5.13) (5.14)
We have E{|ε1,n (y)|} E{|δ1,n (z) − g2,n (t2,n (z))|} + +E{|g2,n (t2,n (z)) − f2 (t2,n (z))|}.
(5.15)
The analogs of (3.53), and (3.54) in our case are r2,n (z), δ1,n (z) = g2,n (zδ2,n (z)(1 + zgn (z))−1 ) +
(5.16)
33
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
where g2,n (z) is random function defined as follows (cf. (3.54)) g2,n (z) = n−1 Tr G2 (z) =
2π
0
ν2,n(dθ) , eiθ − z
−1 z ◦ n Tr Y1 P −1 δ2,n (z) − r2,n (z) = 1 + zgn (z) −n−1 Tr U G(z)SU T U ∗ P −1 gn◦ (z) , P −1 = G2 (t2,n (z)), and gn◦ (z) = gn (z) − gn (z),
◦ δr,n (z) = δr,n (z) − δr,n (z),
r = 1, 2 (5.17)
are the respective random variables centralized by the partial expectations with respect to the Haar measure. In addition, we have the analog of (3.58) | r2,n (z)|
C . n
This leads to the following bound for the first term in the right-hand side of (5.15): r2,n (iy)|} E{|δ1,n (z) − g2,n (t2,n (z))|} E{|
C . n
The universal bounds |gn (z)|
1 , 1 − |z|
|δr,n |
1 , 1 − |z|
r = 1, 2
are valid on all realizations of random matrices Sn and Tn which imply that for z ∈ D(d0 ), d0 1/4 |tr,n (z)| 1/2,
r = 1, 2.
(5.18)
Thus E{|g2,n (t2,n (z)) − f2 (t2,n (z))|} sup E{|g2,n(ζ ) − f2 (ζ )|}.
(5.19)
|ζ |1/2
The right-hand side of this inequality tends to zero as n → ∞ in view of hypothesis of the Theorem 2.2 and Proposition 5.2. Thus we have limn→∞ E{|ε1,n(z)|} = 0 for all z ∈ D(d0 ). Analogous arguments show that limn→∞ E{|ε2,n (z)|} = 0, where ε2,n (z) is defined from (5.13) and from (5.14) by interchanging the indices 1 and 2. Thus we have lim E{|εr,n (z)|} = 0,
n→∞
r = 1, 2.
(5.20)
34
VLADIMIR VASILCHUK
Consider now the first term in the right-hand side of (5.11). In view of (5.12) we can write this term in the form f2 (t1,n (z)) − f2 (t1 (z)) z zt1,n (z) I2 γ n + I2 γ2,n =− 1 + zf (z) 1 + zf (z) = −a2 γn + b2 γ2,n ,
(5.21)
where I2 , a2 and b2 are defined by formulas (5.3) and (5.3), in which we have to replace 2 , 2 , f and f by δ2,n , 2 , gn and f , respectively. Denote by . = {.ij }3i,j =1 the matrix defined by the left-hand side of system (5.1) and by H = {Hi }3i=1 the vector with components H1 = γn , H2 = γ1,n, H3 = γ2,n . Thus, the rest of the proof of the theorem corresponds line by line to the proof of Lemma 4.1 after (4.19). ✷
6. Appendix PROPOSITION 6.1. Let Mr , r = 1, 2 be nonrandom n × n matrices and Un unitary random matrix uniformly distributed over the unitary group U(n) with respect to the Haar measure. Then (i) n−1 Tr M1 Un∗ M2 Un = n−1 Tr M1 n−1 Tr M2 ; (ii) n−1 Tr (M1 Un∗ M2 Un )2 (1 − n−2 ) 2 2 = n−1 Tr M1 n−1 Tr M22 + n−1 Tr M12 n−1 Tr M2 − 2 2 − n−1 Tr M1 n−1 Tr M2 − − n−3 Tr M12 n−1 Tr M22 ,
(6.1)
(6.2)
where · denotes the average over the unitary group U(n). Proof. (i) Using Proposition 3.2 with .(M2 ) = (M1 U ∗ M2 U )ab we obtain (M1 [X, U ∗ M2 U ])ab = 0. Choosing the matrix X with only (a, b)th nonzero entry and applying the operation n−2 na,b=1 to this relation, we get (6.1). (ii) On the other hand, using Proposition 3.2 with .(M2 ) = (M1 U ∗ M2 U M1 U ∗ M2 U )ab , after the same procedure, we obtain n−1 Tr (M1 Un∗ M2 Un )2 = n−1 Tr M1 n−1 Tr (M1 U ∗ M22 U )+ + n−1 Tr (M12 U ∗ M2 U )n−1 Tr M2 −
− (n−1 Tr (M1 U ∗ M2 U ))2 .
(6.3)
35
ON THE LAW OF MULTIPLICATION OF RANDOM MATRICES
Besides, using Proposition 3.2 with .(M2 ) = (M1 U ∗ M2 U )cc (M1 U ∗ M2 U )ab , we obtain (M1 [X, U ∗ M2 U ])cc (M1 U ∗ M2 U )ab + +(M1 U ∗ M2 U )cc (M1 [X, U ∗ M2 U ])ab = 0. Choosing the matrix X with only (a, b)th nonzero entry and applying the operation n−3 na,b,c=1 to this relation, we get (n−1 Tr (M1 U ∗ M2 U ))2 = (n−1 Tr M1 )2 (n−1 Tr M2 )2 + +n−2 n−1 Tr M12 Un∗ M22 Un −
−n−2 n−1 Tr (M1 Un∗ M2 Un )2 . Substituting the relation above in (6.3) and using (6.1) we obtain (6.2).
✷
REMARK 6.1. In this paper we deal with unitary and Hermitian matrices, i.e. we assume that the matrices Sn , Tn , Un and Wn in (2.1) are unitary and An and Bn in (2.2) are Hermitian. It is natural also to consider the case of orthogonal Sn , Tn , Un and Wn and real symmetric An and Bn . This case can be handled by using the analogue of formula (3.13) of the orthogonal group O(n). Indeed, it is easy to see that this analog has the form . (O MO) · [X, O MO] dO = 0, O(n)
where O is the transpose of O and X is a real antisymmetric matrix. By using this formula we obtain instead of (3.25) ˜ 1 )ab (2 G) ˜ bb − (GH ˜ ab ˜ 1 )aa (H2 G) (GH ˜ 1 H2 )ab G ˜ bb − (GH ˜ ab . ˜ 1 H2 )aa G = (GH The second terms in both sides of this formula give two additional terms in (3.32) ˜ 1 ) H2 G ˜ 1 H2 ) G. ˜ − n−2 Tr (GH ˜ n−2 Tr (GH These terms, however, produce the asymptotically vanishing contribution to the remainder (3.32), because, in view of (3.5), (3.8) and (3.9) we have in the case of real symmetric An and Bn −2 n Tr (GH ˜ 1 ) H2 G ˜ 1 H2 ) G ˜ − n−2 Tr (GH ˜ 2 m2/4 ny02 4 and in the case of orthogonal Sn and Tn −2 n Tr (GV1 ) V2 G − n−2 Tr (GV1 V2 ) G
2 . n(1 − d0 )2
36
VLADIMIR VASILCHUK
Similar, and also negligible as n → ∞ terms appear in formulas (3.55), (3.41), (3.67) and (3.72) of the proof of Theorems 3.1 and 3.2. As a result, in this case we obtain the same system (2.25), defining the Herglotz transform of the limiting eigenvalue counting measure of the analogue of (2.1) with the orthogonal Sn and Tn and orthogonal Haar-distributed Un and Wn and the same system (2.22), defining the Stieltjes transform of the limiting eigenvalue counting measure of the analogue of (2.2) with the real symmetric An and Bn and orthogonal Haar-distributed Un and Wn .
Acknowledgements I am thankful to Prof. Anne Boutet de Monvel, Prof. A. Khorunzhy and Prof. L. Pastur for numerous helpful discussions. I also thanks the Ministère des Affaires Etrangères de France for financial support. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Akhiezer, N. I. and Glazman, I. M.: Theory of Linear Operators in Hilbert Space, Frederick Ungar Publishing Co., New York, 1963. Bercovici, H. and Voiculescu, D. V.: Free convolution of measures with unbounded support, Indiana Univ. Math. J. 42 (1993), 733–773. Janik, R. A., Nowak, M. A., Papp, G., Wambach, J., and Zahed, I.: Nonhermitian random matrix models: a free random variable approach, Phys. Rev. E 55 (1997), 4100–4106. Marchenko, V. A. and Pastur, L. A.: Distribution of eigenvalues for some sets of random matrices, Mat. Sb. (N.S.) 72 (114) (1967), 507–536 (Russian). Pastur, L. and Vasilchuk, V.: On the law of addition of random matrices, Comm. Math. Phys. 214 (2000), 249–286. Speicher, R.: Free convolution and the random sum of matrices, Publ. Res. Inst. Math. Sci. 29 (1993), 731–744. Vasilchuk, V.: On the law of multiplication of unitary random matrices, Mat. Fiz. Anal. Geom. 7 (2000), 266–283 (Russian). Voiculescu, D. V.: Limit laws for random matrices and free products, Invent. Math. 104 (1991), 201–220. Voiculescu, D.: A strengthened asymptotic freeness result for random matrices with applications to free entropy, Internat. Math. Res. Notices 1998, No. 1, 41–62. Voiculescu, D. V., Dykema, K. J., and Nica, A.: Free Random Variables, CRM Monograph Series, Amer. Math. Soc., Providence, RI, 1992.
Mathematical Physics, Analysis and Geometry 4: 37–49, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
37
Smoothing Properties of the Heat Semigroups Associated to Hamiltonians Describing Point Interactions in One and Two Dimensions A. BEN AMOR and PH. BLANCHARD Universität Bielefeld, Fakultät für Physik, D-33615, Bielefeld, Germany (Received: 30 June 2000; in final form: 12 April 2001) Abstract. Smoothing properties of the heat semigroups associated to Hamiltonians describing point interactions in one and two dimensions are investigated. A construction of Hamiltonians describing point interaction on Lp -spaces is then derived and a full description of their spectra is given. Particularly, we prove the p-independence of their spectra and the exponential growth of the p-norms of such semigroups for large time. Mathematics Subject Classifications (2000): primary 79-XX, secondary 81Q10. Key words: heat semigroup, point interaction, smoothing property, spectrum.
1. Introduction By the heat semigroup we mean the semigroup associated to the equation ∂ψ = −H ψ, ∂t
(1)
where H is a selfadjoint operator and we will denote it by exp(−tH ). The most important case in applications is when H is a generalized Schrödinger operator on Rd , i.e. a perturbation of minus the Laplacian by a suitable measure (for example, in the generalized Kato class) and especially by a potential. Constructions and investigations of the heat semigroups exp(−tH ), where H is a Schrödinger operator, is subject of extensive literature, rich in its methods (analytic, probabilistic, or related to operator theory) and contents. In a survey paper, B. Simon [19] proved the Lp smoothing property of exp(−tH ) with H = −+V for a large class of potentials: If the negative part of the potential is in the Kato class and its positive part is in the local Kato class, then for every t > 0, exp(−tH ) maps continuously Lp into Lq , 1 p q ∞. In the same paper, Lp -growth of these semigroups is also studied. For instance, it is proved that for the class of potentials described above, exp(−tH )p,p has an exponential growth for large t, namely exp(−tH )p,p C exp(αt)
(2)
38
A. BEN AMOR AND PH. BLANCHARD
for large t, where C is a positive constant and α ∈ R. Let us emphasize that for d = 1, 2 and V < 0, and due to the existence of negative eigenvalues of H , one has α > 0. Among other interesting features it is proved in [19] that the spectral bound, inf σ (H ) = − lim t −1 ln exp(−tH )p,p t →+∞
is p-independent. This result is now a well-known fact. For instance, R. Hempel and J. Voigt [15, 16] proved that for certain potentials (including those in the Kato class) the spectrum of H is p-independent as well. A more detailed answer to the question asked by B. Simon about p-independence of spectra of Schrödinger operators can be found in [16]. Later on, Ph. Blanchard and Z. M. Ma [4], proved also the Lp -smoothing property for the heat semigroups associated to H = −+µ on Rd (d 3), where µ is a signed measure whose positive part is smooth and negative part is in the generalized Kato class. In an abstract setting, P. Stollmann and J. Voigt [20] investigated the properties of exp(−tHµ ), Hµ = − + µ on the space L2 (X, m), where m is a measure whose support is X. Recently (cf. [13] and references therein), a particular interest was showed for the study of smoothing properties of exp(−tH ), where H = − 12 +V on the scale of Bessel potential spaces on Rd . In [14] (Theorem 1.1), the authors proved that for V ∈ Kd,loc with negative part V − ∈ Kd , the boundedness of exp(−tH ) from p,s+2 p,s Lp into Lloc is equivalent to V ∈ Lloc , while A. Gulisashvili [13] gave a sharp p,s+2 estimate of the Lp − Lloc norm of exp(−tH ) for V ∈ Lp,s ∩ Kd , where Kd is the Kato class of potentials and Kd,loc the local Kato class of potentials. Hamiltonians describing point interactions, or Schrödinger operators with point interactions, are operators corresponding to perturbations of minus the Laplacian by the linear combination of Dirac measures. The mathematical setting and the description of such Hamiltonians are now quite well-known [3] and a new approach to handling them is given in [5]. It is known [3] that for d = 2, 3, there is only one kind of point interaction that can be denoted by Hα = − + αδ. In contrast, in one dimension there are more possibilities of point interactions. However, we will concentrate here on a one-parameter family corresponding to the δ -interaction [3], which we shall also denote by Hα , −∞ < α +∞. This family is related to selfadjoint extensions of the operator − with its domain the Sobolev space H02 (R \ {0}) and determined by the boundary condition: ϕ (0+ ) = ϕ (0− ),
ϕ(0+ ) − ϕ(0− ) = αϕ (0),
(3)
for every ϕ in the domain of Hα . Hence, up to now operators describing point interactions in Rd do not fit into a standard situation and, in any case, the theory developed in [20] does not include such operators. However, we are saved by the explicit knowledge of the kernels associated to Hα [1]. Our aim in this paper is to use these formulas to establish p, q smoothing properties of Hα on Lp (Rd )
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
39
for d = 1, 2. As a consequence, we get a construction of the operators Hα on Lp spaces for 1 p < +∞ which we denote by Hα,p (for p = 2 we omit the subscript p). Then combining the techniques used by J. Voigt and P. Stollmann [15, 16] and the new functional calculus developed by E. B. Davies [8], we prove the pindependence of their spectra and derive the exponential growth of the p-norm of exp(−tHα,p ). Such a construction was done by S. Albeverio et al. [2] using the ‘family of pseudo-resolvent’, thereby exploiting the expression of the resolvent kernel of Hα . They conclude the construction for d = 1, p ∈ [1, +∞) or d = 2, p ∈ ]1, +∞[ or d = 3, p ∈ ] 32 , 3[ and the same thing for the C0 -semigroup exp(−tHα ). We will use here the reversed strategy: Using the explicit formula of the heat kernel associated to Hα (cf. [1]) which we denote, as in [1], by P α (t; x, y), we prove that for d = 1, 2, this kernel also defines a bounded linear operator from Lp into Lq for 1 p q +∞. This is the well-known ‘smoothing property’. Moreover, this kernel even defines a strongly continuous semigroup on Lp for 1 p < +∞. Then we denote by −Hα,p the generator of exp(−tHα ) on Lp (1 p < +∞). Using the integral representation of the resolvent function [9], p. 55 (which is the Laplace transform of the semigroup), we get that for k 2 such that Im(k) > 0 and Re(k 2 ) < min(s(Hα ), s(Hα,p )), (Hα,p − k 2 )−1 is a kernel operator whose kernel is Gk , where Gk is the kernel of (Hα − k 2 )−1 . Hence, for d = 1, 2, the construction we propose in this paper includes the one made in [2]. A natural question arises: why is it so? This is related to the properties of the heat kernel. For instance, the heat kernel has better properties than the resolvent kernel. A good example is a comparison between the resolvent and the heat kernel of −. Unfortunately, this method does not work in three dimensions for the reason that we will explain at the end of the paper. For the notations we will adopt those used in [3]. So Hα is the Hamiltoniandescribing interaction placed at the origin in the space L2 , which corresponds in one dimension to the δ -interaction. We shall denote the free Hamiltonian by H0 . The space Lp (Rd ) is denoted simply by Lp for every 1 p +∞. For every linear closed operator T , the spectrum of T is denoted by σ (T ), whereas its spectral bound is denoted by s(T ) and is defined by [9] s(T ) = inf{Re(λ), λ ∈ σ (T )}.
(4)
2. Smoothing Property in Two Dimensions Following the notation of [1], we denote the heat kernel of exp(−tHα ) by P α (t; x, y) for every t > 0. It is given by [1] +∞ 1 e(−αu) α × t u−1 P (t; x, y) = P (t; x, y) + 2π 0 $(u) +∞ |x|2 +|y|2 |x||y| z dz du, (5) (z − 1)u−1 z−u e(−z 4t ) K0 × 2t 1
40
A. BEN AMOR AND PH. BLANCHARD
where
1 |x − y|2 P (t; x, y) = exp − 4π t 4t
is the kernel of the free Hamiltonian and K0 is the MacDonald function. We will also denote by P˜α (t; x, y) the difference P α (t; x, y) − P (t; x, y). A way to prove the smoothing property of exp(−tHα ) has been developed in [19]: first prove the boundedness from L∞ into L∞ and the boundedness from L1 into L∞ and then use the Riesz–Thorin convexity theorem [18] to conclude. The first step is given by this lemma. LEMMA 2.1. For every t > 0, we have P α (t; x, y) dy < +∞. sup x∈R2
(6)
R2
Proof. Since P (t; x, y) dy = 1, sup x∈R2
R2
we just have to prove that P˜α (t; x, y) dy < +∞. sup x∈R2
(7)
R2
Set
I (x) =
and
P˜α (t; x, y) dy R2
J (x) =
2 +|y|2 4t
(−z |x|
e R2
Then 1 I (x) = 2π
+∞
t 0
u−1
)
K0
e−αu $(u)
|x||y| z dy. 2t
+∞
(z − 1)u−1 z−u J (x) dz du.
1
Let us make a suitable estimate for J (x). Using polar coordinates, we get +∞ |x|2 r2 r|x| r dr. e(−z 4t ) K0 z J (x) = 2π e(−z 4t ) 2t 0 With the change of variable s = z(r|x|/2t) one get t 2 2t 2 +∞ (− 2z|x| 2s ) e sK0 (s) ds J (x) = 2π z|x| 0
(8)
41
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
which is equal to (cf. [12], p. 717, 3) 1 2 1 z|x|2 z|x|2 2 ( z|x|2 ) (−z |x| ) 4t 8t , e W− 1 ,0 2π e 2 2 t 4t
(9)
where Wχ,µ is the Whittaker function. On the other hand, one has (cf. [17], p. 305) 1 z W− 1 ,0 (z) = z 2 e( 2 ) which gives 2
2π t J (x) = √ . 2z Thus
+∞ −αu +∞ 1 ue t (z − 1)u−1 z−u−1 dz du sup I (x) = √ $(u) 2 2 0 1 x∈R 1 = √ ν(te−α ), 2
(10)
where the function ν is defined by (cf. [10]): +∞ xs ds. ν(x) = $(s + 1) 0 From this follows PROPOSITION 2.1. For every t > 0 and every p such that 1 p +∞ the operator exp(−tHα ): Lp → Lp
(11)
is bounded and defines a strongly continuous semigroup for 1 p < +∞. Proof. First let us denote by C(t) = 1 + √12 ν(te−α ). For 1 < p +∞, let q be the conjugate of p: p −1 + q −1 = 1. Then, by Hölder inequality, we have for every f ∈ Lp p1 p p α p q |exp(−tHα )f (x)| (C(t)) P (t; x, y)|f (y)| dy , (12) R2
thereby
p exp(−tHα f (x) dx R2 p p q (C(t)) |f (y)| R2
α
P (t; x, y) dx dy.
Now using the symmetry property of the heat kernel, we get p p |exp(−tHα )f (x)| dx (C(t)) |f |p dy. R2
(13)
R2
R2
(14)
42
A. BEN AMOR AND PH. BLANCHARD
To prove that it defines a strongly continuous semigroup, let us denote by S(t) the operators whose kernel is P˜ α , then we have S(t)p,p √12 ν(te−α ). By the properties of the function ν [10], p. 219, we have limt →0 ν(t) = 0, we then get lim exp(−tHα ) = I.
(15)
t →0
Now defining Tα (t) by Tα (t) = exp(−tHα ) for t > 0 and Tα (0) = I , we get a strongly continuous semigroup which we shall denote by exp(−tHα ). For p = 1, the proof is straightforward. ✷ We are now in a position to prove boundedness from L1 into L∞ . LEMMA 2.2. For every t > 0 we have sup P α (t; x, y) < +∞.
(16)
x,y∈R2
Proof. To achieve this goal, we are going to use the construction of the Hamiltonian Hα via Dirichlet forms as done in [3] which we recall here. We omit the case α = ∞, which corresponds to the free Hamiltonian. Let ϕα be the following function ϕα (x) = H0(1)(2ie−2πα+0(1) |x|),
x ∈ R2 \ {0},
(17)
where H0(1) is the Hankel function. Denote by Hϕα the operator associated to the local positive Dirichlet form ¯ α2 dx. ∇f ∇gϕ (18) Eϕα : D(Eϕα ) ⊂ L2 (ϕα2 ), Eϕα (f, g) = R2
Then the Hamiltonian Hα is related to this Dirichlet form by [3] Hα = ϕα [Hϕα − βI ]ϕα −1 ,
(19)
where β = 4e2(−2πα+0(1)) . Clearly exp(−tHα ) = eβt ϕα exp(−tHϕα )ϕα −1 . Now if we denote by q α (t; x, y) the heat kernel of the operator exp(−tHϕα ), then P α (t; x, y) = eβt ϕα (x)ϕα −1 (y)q α (t; x, y). Now since the Dirichlet form Eϕα is local and positive, then the kernel q α (t; x, y) is Markovian [11], hence 0 < q α (t; x, y) 1. Now we have on the diagonal the following estimate: P α (t; x, x) eβt .
(20)
Using the Chapman–Kolmogorov equation α P α (s; x, z)P α (t; z, y) dz, P (t + s; x, y) =
(21)
R2
we get P α (t; x, y) eβt which completes the proof. We now formulate the following theorem:
✷
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
43
THEOREM 2.1. For every t > 0 and every 1 p q ∞, the operator exp(−tHα ): Lp → Lq
(22)
is bounded. Theorem 2.1 gives a variety of properties of the heat semigroup exp(−tHα ) known for a large class of Schrödinger operators, at least for operators H = − + V where V is in the Kato class. For every 1 p < +∞, let us denote by −Hα,p the generator of exp(−tHα ) on Lp which will be denoted exp(−tHα,p ). For p = 2, we omit the index p. Interpretation of point interaction on Lp spaces as an extension of − on D0 = {f ∈ C0∞ (Rd ), f (0) = 0} was done in [6]. There the authors show that point interaction can be defined on Lp for d = 1, 2 and 1 < p < +∞ or d = 3 and 32 < p < 3 as negative generators of analytic semigroups. Their construction is essentially based on estimations of the resolvent of Hα with α = 1. We here emphasize that for d = 1, 2 we have fewer restrictions on p than in [2] or in [6]. Now the question of the p-independence of the spectra of Hα arises. Let us note that for p = +∞ one cannot hope to get the inclusion σ (Hα ) ⊂ σ (Hα,p ). For the eigenfuction of Hα which is equal to i (1) H (2ieβ |x|), x = 0, 4 0 where β = −2π α + 0(1), is unbounded.
(23)
PROPOSITION 2.2. For every t > 0 and p such that 1 p < +∞ we have (i) The spectrum of Hα,p is p-independent. (ii) Every isolated eigenvalue of Hα of algebraic multiplicity m is an isolated eigenvalue of Hα,p with the same multiplicity and conversely. (iii) The spectral bound of Hα,p satisfies − lim t −1 ln exp(−tHα,p )p,p = −4 exp(2(2α + 0(1))). t →∞
(24)
Proof. The proof of assertion (ii) is as in [15], (iii) follows from (i) and the characterization of the spectral bound of generators associated to strongly continuous semigroups [9], p. 299. So the important point is to prove (i). Following R. Hempel and J. Voigt [15], we are going to prove first that σ (Hα ) ⊂ σ (Hα,p ).
(25)
A crucial argument to prove (25) is that, for every 1 p q < +∞ and every t > 0, the operator exp(−tHα,p ) maps continuously Lp into Lq , for exp(−tHα,p ) has the same kernel as exp(−tHα ). Once this is observed, one can continue the proof of (25) as in [20]. Let us prove now the reversed inclusion. We may suppose that 2 p +∞, and then conclude by duality. Given ξ ∈ ρ(Hα ) and f ∈ Lp ∩ L2 , then in fact
44
A. BEN AMOR AND PH. BLANCHARD
exp(−tHα )f = exp(−tHα,p )f , now by the integral representation of the resolvent function [9], we have for every ξ such that Re(ξ ) < min(s(Hα ), s(Hα,p )), Rξ (Hα )f = Rξ (Hα,p )f which implies that Hα f = Hα,p f . Since ρ(Hα ) is connected, we get, for every ξ ∈ ρ(Hα ) and every f ∈ Lp ∩ L2 , (Hα − ξ I )−1 (Hα,p − ξ I )f = f.
(26)
It is now sufficient to prove that T (ξ ) = (Hα − ξ I )−1 is bounded as an operator on Lp . Indeed, from the kernel formula of (Hα − ξ I )−1 [3], we get that T (ξ ) is a closed operator in L∞ whose domain is the whole space L∞ , hence by the Banach theorem we conclude that T (ξ ) is bounded on L∞ . On the other hand, it is bounded on L2 , thus by the Riesz–Thorin convexity theorem, we get the boundedness of ✷ T (ξ ) on Lp for every 2 p +∞ which implies the result. Remark 2.1. From Proposition (2.2), we get exp(−tHα,p )p,p ∼ exp(4t exp(2α + 0(1))),
(27)
for large t which expresses the exponential growth of the p-norm of the operator exp(−tHα,p ), while for small t we have exp(−tHα,p )p,p 1 +
C . |log(t)|
(28)
In [4] it is proved that for 1 p < +∞ and every f ∈ Lp the function exp(−tH ) tends to zero at infinity. The same phenomenon occurs in our situation. PROPOSITION 2.3. For 1 p < +∞ and f ∈ Lp we have lim |exp(−tHα,p )f (x)| = 0.
(29)
|x|→∞
Proof. We give the proof for p = 2, for p = 2, the proof is more or less the same. The proof is based on the explicit formula of the heat kernel [1], Equation (3.16), namely,
A
P (t; x, y) = P (t; x, y) +
e− t
α
+∞
×
r
(4π t|x||y|) u−1 1
0
(r + 1)u+ 2
where (|x| + |y|)2 A= 4
A 0 e− t r K
and
0 (z) = K
1 2
0
+∞ u −αu
t e × $(u)
|x||y| (r + 1) dr du, 2t
2z exp(z)K0 (z). π
(30)
45
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
Let us recall that [1] 0 (r)| = M < +∞. sup |K
(31)
r0
Taking into account that lim exp(−tH0 )f (x) = 0,
|x|→+∞
we should just prove that +∞ u −αu +∞ t e r u−1 1 × 1 1 $(u) 0 (4π t|x| 2 ) 0 (r + 1)u+ 2 A e− t A 0 |x||y| (r + 1) f (y) dy dr du e− t r K × 2t R2 (4π t|y|)
(32)
tends to zero for |x| → ∞. We denote by A(x) the last term in Equation (32), then 1 − (|x|+|y|)2 4t e |f (y)| dy A(x) M 1 R2 |y| 2 for which, by Hölder inequality, we obtain 12 +∞ 1 s2 − e 2t ds . A(x) 2π 2 Mf L2
(33)
0
This yields |exp(−tHα,p )f (x)| |exp(−tH0 )f (x)| + C
ν(te−α ) 1
|x| 2
f L2 ,
where C > 0, and this completes the proof.
(34) ✷
3. Smoothing Property in One Dimension In one dimension, the Hamiltonian Hα corresponds for α = 0 to the free Hamiltonian, so in this section we will omit the case α = 0. For α = 0, the heat kernel of Hα is given by [1] (Formula 3.4), sgn(xy) − (|x|+|y|)2 4t e + P α (t; x, y) = P (t; x, y) + √ 4π t sgn(xy) +∞ − 2 u − (|x|+|y|+u)2 4t e α e du, +2 √ α 4π t 0 where P (t; x, y) = √
1 4π t
e−
|x−y|2 4t
(35)
46
A. BEN AMOR AND PH. BLANCHARD
is the heat kernel of the free one-dimensional Hamiltonian. The arguments used to prove the p, q-smoothing property of the heat semigroup in one dimension are quite similar to those used in the previous section. Indeed, we have the following easy lemma: LEMMA 3.1. For every t > 0, we have sup |P α (t; x, y)| < +∞. x∈R
Proof. A direct computation shows that, for every α > 0, we have √ sup |P α (t; x, y)| C1 (t) = 2 + 2 4π t. x∈R
(36)
R
(37)
R
While for α < 0, the following estimate holds true: 2 √ 4t α π t exp 2 . sup |P (t; x, y)| 2 + |α| α x∈R R
(38) ✷
LEMMA 3.2. For every t > 0, we have 1 C1 + C2 t − 2 ; α > 0, α sup |P (t; x, y)| 4t − 12 C1 + C2 t + exp( α 2 ); α < 0, x,y∈R with positive constants Cj and Cj . Now using Lemmas 3.1, 3.2, and the Riesz–Thorin theorem, one can easily establish the following theorem: THEOREM 3.1. For every t > 0 and every 1 p q ∞, the operator exp(−tHα ): Lp → Lq
(39)
is bounded and, for p = q < +∞, it defines a strongly continuous semigroup. Now, as in the first section, we denote by −Hα,p the generator of the operator exp(−tHα ) on the space Lp and for p = 2 we will omit the subscript p. The spectral properties of the operators Hα,p can be easily investigated using the smoothing properties of their semigroups. PROPOSITION 3.1. For every 1 p < +∞ we have (i) σ (Hα,p ) is p-independent. (ii) Every isolated eigenvalue of Hα of algebraic multiplicity m is an isolated eigenvalue of Hα,p with the same multiplicity 4 and, conversely. ; α < 0, (iii) lim t −1 ln exp(−tHα,p )p,p = α 2 t →+∞ 0; α > 0.
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
47
We shall prove only assertion (i). To this aim we use a new method which relies on the new functional calculus introduced by E. B. Davies [8]. The application of this calculus requires an estimate for the resolvent function and the the spectra of Hα,p must be real. Thanks to the explicit formula of the kernel of the operator (Hα − k 2 )−1 , one can prove a suitable estimate for the resolvent functions. Let us first recall that for k 2 ∈ ρ(Hα ), Im k > 0 the kernel of (Hα − k 2 )−1 is given by [3], p. 92, i α sgn(xy) ik(|x|+|y|) e . (40) Gk (x, y) = eik|x−y| + 2k 2(−ikα + 2) It is obvious that Gk (x, y) is also the kernel of (Hα,p −k 2 )−1 for every k 2 ∈ ρ(Hα )∩ ρ(Hα,p ) with Im(k) > 0. LEMMA 3.3. For every k 2 ∈ ρ(Hα ) such that Im k > 0, we have 4 . sup |Gk (x, y)| dy |Im(k 2 )| x∈R R
(41)
Proof. We have
−Im(k)(|x|+|y|) 1 −Im(k)|x−y| α e e + , |Gk (x, y)| |2k| 2(−ikα + 2) which yields |α| 1 1 + . |Gk (x, y)| dy |k|Im(k) |(−ikα + 2)| Im(k) R Observing that 1 |k|Im(k) |Re(k)|Im(k) = |Im(k 2 )| 2 and that 2 − ik |Re(k)|, α we get the desired estimate.
(42)
(43)
✷
Using estimate (41) we conclude that for every k 2 ∈ ρ(Hα ) with Im k > 0 the operator whose kernel is Gk defines a bounded operator on Lp for 1 p +∞. This operator is nothing else but (Hα,p − k 2 )−1 , thereby giving that C \ R ⊂ ρ(Hα,p ).
(44)
Thus, σ (Hα,p ) ⊂ R. Now the operators Hα,p satisfies all hypotheses (especially H 1) required by the functional calculus [8]. Proof of Proposition 3.1. Using Lemma 4 in [7] and the fact that for every ξ ∈ ρ(Hα ) ∩ ρ(Hα,p ) and every f ∈ Lp ∩ Lq , we have Rξ (Hα )f = Rξ (Hα,p )f , we get the spectral p-independence. ✷
48
A. BEN AMOR AND PH. BLANCHARD
Remark 3.1. From Proposition 3.1(iii), we observe that for α < 0 the Lp norm of exp(−tHα,p ) has an exponential growth for large t, this is, however, similar to the case of perturbations of the Laplacian by negative potentials. While for α > 0, exp(−tHα,p )p,p behaves like exp(−tH )p,p for large t. So that the operator exp(−tHα,p ) in one dimension behaves somewhat different as in two dimensions where the behavior does not depend on the sign of α. We close this section with the following proposition: PROPOSITION 3.2. For 1 p < +∞ and f ∈ Lp we have lim | exp(−tHα,p )f (x)| = 0.
(45)
|x|→∞
Remark 3.2. We here studied only the case of a one-parameter family of point interaction, corresponding to δ -interaction. However, the same method applies for the four-paramater family of self-adjoint extensions. The method we give here does not work for d = 3 for the simple reason that for each α = +∞ we have P α (t; x, y) dy = +∞. sup x∈R3
R3
In fact, for α 0, we have [1] P α (t; x, y)
2t P (t; |x| + |y|), |x||y|
(46)
where P (t; |x| + |y|) =
1 (4π t)
3 2
e−
(|x|+|y|)2 4t
.
Hence,
|x|2 4π 2 t |x| α ˜ D−2 √ e− 2t , P (t; x, y) dy |x| t R3
(47)
where Dν is the cylindrical hypergeometric function [17]. This implies that P˜ α (t; x, y) dy = +∞. lim x→0
R3
Similarly, one can prove the same result for positive α. References 1.
Albeverio, S., Brze´zniak, Z. and D¸abrowski, L.: Fundamental solution of the heat and Schrödinger equations with point interaction, J. Funct. Anal. 130(1) (1995), 220–254.
SMOOTHING PROPERTIES OF HEAT SEMIGROUPS
2. 3. 4.
5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
49
Albeverio, S., Brze´zniak, Z. and D¸abrowski, L.: The heat equation with point interaction in Lp spaces, Integral Equations Operator Theory 21(2) (1995), 127–138. Albeverio, S., Gesztezy, F., Høegh-Krohn, R. and Holden, H.: Solvable Models in Quantum Mechanics, Texts and Monogr. Phys., Springer-Verlag, New York, 1988. Blanchard, Ph. and Ma, Z. M.: Semigroup of Schrödinger operators with potentials given by Radon measures, In: Stochastic Processes, Physics and Geometry (Ascona and Locarno, 1988), World Scientific, Teaneck, NJ, 1990, pp. 160–195. Caspers, W. and Clément, P.: A different approach to singular solutions, Differential Integral Equations 7(5–6) (1994), 1227–1240. Caspers, W. and Clément, Ph.: Point interactions in Lp , Semigroup Forum 46(2) (1993), 253– 265. Davies, E. B.: Lp spectral independence and L1 analyticity, J. London Math. Soc. (2) 52(1) (1995), 177–184. Davies, E. B.: The functional calculus, J. London Math. Soc. (2) 52(1) (1995), 166–176. Engel, K. J. and Nagel, R.: One-Parameter Semigroups of Linear Evolution Equations, Springer, New York, 2000. Erdélyi, A., Magnus, W., Oberhettinger, F. and Tricomi, F. G.: Tables of Integral Transforms, Vol. 3, McGraw-Hill, New York, 1954. Fukushima, M.: Dirichlet Forms and Markov Processes, North-Holland, Amsterdam, 1980. Gradshteyn, I. S. and Ryzhik, I. M.: Table of Integrals, Series, and Products, Academic Press, New York, 1965. Gulisashvili, A.: Sharp estimates in smoothing theorems for Schrödinger semigroups, J. Funct. Anal. 170(1) (2000), 161–187. Gulisashvili, A. and Kon, M. A.: Exact smoothing properties of Schrödinger semigroups, Amer. J. Math. 118(6) (1996), 1215–1248. Hempel, R. and Voigt, J.: The spectrum of a Schrödinger operator in Lp (Rν ) is p-independent, Comm. Math. Phys. 104(2) (1986), 243–250. Hempel, R. and Voigt, J.: On the Lp -spectrum of Schrödinger operators, J. Math. Anal. Appl. 121(1) (1987), 138–159. Magnus, W., Oberhettinger, F. and Soni, R. Pal.: Formulas and Theorems for the Special Functions of Mathematical Physics, Springer, New York, 1966. Reed, M. and Simon, B.: Methods of Modern Mathematical Physics, Vol. 2, Fourier Analysis and Self-adjointness, Academic Press, New York, 19xx. Simon, B.: Schrödinger semigroups, Bull. Amer. Math. Soc. (N.S.) 7(3) (1982), 447–526. Stollmann, P. and Voigt, J.: Perturbation of Dirichlet forms by measures, Potential Anal. 5(2) (1996), 109–138.
Mathematical Physics, Analysis and Geometry 4: 51–63, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
51
On the Tidal Motion Around the Earth Complicated by the Circular Geometry of the Ocean’s Shape Without Coriolis Forces Dedicated to Professor L. V. Ovsyannikov on the occasion of his 80th birthday RANIS N. IBRAGIMOV Department of Applied Mathematics, University of Waterloo, Waterloo, ON, N2L 3G1, Canada. e-mail:
[email protected] (Received: 10 February 2000; in revised form: 29 March 2001) Abstract. The Cauchy–Poisson free boundary problem on the stationary motion of a perfect incompressible fluid circulating around the Earth is considered in this paper. Rotation plays a significant role in the early stages of the formation of solitary waves. However, these effects are less important on the solitary waves once they are formed. Therefore, for simplicity, rotation is not included for these simulations. The main concern is to find the inverse conformal mapping of the unknown free boundary in the hodograph plane onto some fixed mapping in the physical domain. The approximate solution to the problem is derived as the application of such a method. The behaviour of tidal waves around the Earth is discussed. It is shown that one of the features of the positively curved bottom is that the problem admits two different higher-order systems of shallow water equations, while the classical problem for the flat bottom admits only one system. Mathematics Subject Classification (2000): 30C20. Key words: free boundary, inverse conformal mapping.
1. Introduction The Cauchy–Poisson problem on the stationary motion of a perfect fluid which has a free boundary and has a solid bottom represented by a circle with a sufficiently large radius, is considered in this paper. For simplicity, the fluid is not considered in the real, rotating reference frame of the Earth. However, we note that the presence of Coriolis effects do not change the qualitive analysis of the presented method. We have shown in [4] that such a problem can be associated with a two-dimensional model to an oceanic motion around the Earth, since we consider strictly longitudinal flow. Since the problem is a free boundary problem, the analysis is rather difficult. Permanent water waves have been considered in a large number of papers. However, most researchers are concerned with fluid motion which is infinitely deep
52
RANIS N. IBRAGIMOV
and extends infinitely both rightward and leftward (see Crapper [1], Stoker [9] or Stokes [10] for the history). Such problems are usually called Stokes’s problem if the surface tension is neglected and Wilton’s problem if the surface tension is taken into account (see [6] for more details). We consider water waves for which the ratio of the depth of fluid above the circular bottom to the radius of the circle is small (shallow water). Our primary concern is to find the conformal mapping (for the Stokes problem) of the unknown free boundary onto fixed mapping. The resulting Dirichlet problem can be solved numerically using Okamoto’s method [3]. A more detailed structure of the bifurcation of solutions for the related problem was numerically computed by Fujita et al. [3]. The existence of nontrivial solutions for the analogously reduced Dirichlet problem can be found in [4] and [7] as well as in the classical literature (see, e.g., [8] or [9]). Higher-order shallow water equations in the nonstationary case are derived in this paper. It is shown that the present problem admits two different systems of shallow water equations, while the classical problem for the flat bottom admits only one system (see [2]). We note that papers [3, 6, 7] are concerned with fluid whose surface tension is taken into account. In fact, the surface tension plays the role of a ‘regulator’ of the problem which substantially simplifies the analysis. Furthermore, the nature of the problem requires that the surface tension should be neglected. Thus, this paper represents a more systematic approach to the problem. The paper aims to investigate the problem by using a conformal mapping which distinguishes it from [3, 4, 6, 7]. 2. Basic Equations The analysis of this problem is performed in the following notation: R is the radius of the circle, r is a distance from the origin, θ is a polar angle, h0 is the undisturbed level of the liquid above the circle, and h = h (θ) is the level of the disturbance of the free boundary above the circle. For the sake of simplicity, we assume that the pressure is constant on the free boundary. The stream function ψ = ψ(r, θ) defines the velocity vector, i.e., 1 v r = − ψθ , r
v θ = ψr .
Hence, irrotational motion of an ideal incompressible fluid of the constant pressure in the homogeneous gravity field g = const is described by the stream function ψ in the domain h = {(r, θ) : 0 θ 2π, R r R + h0 + h(θ)}, which is bounded by the bottom R = {(r, θ) : r = R, θ ∈ [0, 2π ]} and the free boundary with equation h = {(r, θ) : r = R + h0 + h(θ), θ ∈ [0, 2π ]}. Note that
53
TIDAL MOTION AROUND THE EARTH
ψ is an harmonic function in h , since we assumed that the flow is irrotational. More specifically, we assume that the fluid is incompressible and inviscid and that the flow is stationary. Then the problem is to find the function h (θ) and the stationary, irrotational flow beneath the free boundary r = R + h0 + h (θ) given by the stream function ψ which satisfy the following differential equations:
ψ = 0(in h ),
ψ =0
|∇ψ|2 + 2gh = constant 1 2
2π
ψ =a
(on R ),
(on h ),
(1) (2)
(on h ),
(R + h0 + h(θ))2 dθ = π(R + h0 )2 ,
(3)
0
where a = const denotes the flow rate. Equations (1) to (3) represent the free boundary Cauchy–Poisson problem in which the boundary h is unknown as well as the stream function. 3. The Inverse Transforms Principle 3.1. CONSTANT FLOW The exact solution h≡0
and
ψ = ψ0 = a log r
(4)
of Equations (1)–(3) corresponds to the constant flow with an undisturbed free boundary. The trivial solution (4) represents a flow whose streamlines are concentric circles with the common center at the origin. The following nondimensional quantities are introduced: √ h0 gh0 h0 F = , h = h0 h , ψ = aψ , ε= , r = R + h0 r , R a where F is a Froude number and R is used as a vertical scale. We consider ε as the small parameter of the problem. After dropping the prime, Equations (1)–(3) are written by ψ , h , and (r , θ) as follows: (ε) ψ = 0
(5)
(in h ),
ψ = 0 (on R ),
(6)
ψ = 1 (on h ),
(7)
∇(ε)ψ 2 + 2F −2 h = constant 1 2
0
2π
(8)
(on h ),
(1 + ε + εh(θ))2 dθ = π(1 + ε)2
(on h ).
(9)
54
RANIS N. IBRAGIMOV
Here the Laplace and gradient operators are given by ε∂θ 2 2 , ∂r , (ε) = (ε∂θ ) + [(1 + εr)∂r ] , ∇(ε) = (1 + εr) where the subscripts imply differentiation. We further consider the complex potential ω(ζ ) = ϕ + iψ, where ζ = (1 + εr)eiθ is the independent complex variable and ϕ(ζ ) is the velocity potential which is characterized by the analyticity of ϕ + iψ, i.e., ϕr =
εψθ , (1 + εr)
εϕθ = −ψr . (1 + εr)
We note that the complex velocity dω/dζ is a single-valued analytic function of ζ , although ω is not single-valued. In fact, when we turn around the bottom r = 1 2π once, ϕ increases by − 0 ψr (1, θ)dθ which has a positive sign by the maximum principle (Hopf’s lemma). Hence, if we remove the width of annulus region θ = 0, r ∈ [1, 1 + ε], then at every point (r, θ), ω(ζ ) is single-valued analytic function which maps the rectangular (in the ω(α)-hodograph plane) domain with 2π and ψ ∈ [0, 1] ϕ ∈ 0, − log(1 + ε) as coordinates onto the annulus h0 = {(r, θ) : 1 < r < 1 + ε, θ ∈ [0, 2π ]}. We represent the constant flow (4) by ω(ζ ) = ϕ + iψ =
i log(1 + εr) − θ , log(1 + ε)
(10)
where r = ξ0 (ψ) and θ = η0 (ϕ) transform the rectangular domain in the hodograph plane into h0 . Consequently, each conformal mapping by the function ω(ζ )
Figure 1.
TIDAL MOTION AROUND THE EARTH
55
between hodograph and physical planes represents an irrotational flow in the physical ζ plane. Furthermore, Equations (10) implies that (11) η0 (ϕ) = −ϕ log(1 + ε) and ξ0 (ψ) = ε −1 [1 + ε]ψ − 1 .
3.2. REDUCTION ONTO THE BOUNDARY Now a two-dimensional infinitesimal disturbance ξ and η is superimposed on ξ0 and η0 . Then the resulting transform components are ξ = ξ0 + ξ ,
η = η0 + η .
The perturbed quantities ξ and η are assumed to be small quantities so that the nontrivial solution is close to the trivial one. Then, with Equation (11), the inverse transform can be combined to form εξ + iη , log ζ = ψ log(1 + ε) − iϕ log(1 + ε) + log 1 + 1 + εξ0 since log(1 + εξ0 ) = ψ log(1 + ε). Consequently, the polar angle θ and radius r are given by the equations εξ −1 ψ r=ε (1 + ε) 1 + −1 . θ = −ϕ log(1 + ε) + η , 1 + εξ0 Since the motion is irrotational, we can decrease the dimension of the problem by one. In other words, we introduce the boundary value for the function ξ and reduce the basic equations to the quantities which arise from the condition on the free boundary ψ = 1. To this end, we introduce the regular function f (ω) = α +iβ such that 1 (λϕ , −ηϕ ). (α, β) = 2 (λ + η2 )ϕ Then the nonlinear boundary condition (8) can be reduced to the differential equation of the conservation form for f (ω) by virtue of the following lemma: LEMMA 1. Let the function λ (ϕ, ψ) defined by 1+
εξ = eλ(ϕ,ψ) 1 + εξ0
(12)
is differentiable at least once. We introduce the derivative operator ∂n along the normal to ψ = 1 by ∂n µ = λψ |ψ=1 , where µ(ϕ) = λ(ϕ, 1). Then, on the free boundary, the following three relations hold:
1 + ε 2 2µ e , (13) βϕ = −∂n α = −∂ϕ ∂n τ + 2 where τ = (b − µF −2 )(1 + ε)2 , and b = const is the Bernoulli constant.
56
RANIS N. IBRAGIMOV
Proof. Since the velocity potential ω defined by Equation (10) is an analytic function, ξ and η are single-valued and they satisfy the Cauchy–Riemann equations ε ξϕ = ηψ , 1 + εξ
ε ε 2 (1 + εξ )−1 ξψ − ξ0ψ = −ηϕ 1 + εξ (1 + εξ0 )
which can be simplified as λϕ = ηψ ,
λψ = −ηϕ .
(14)
From Equations (11), (14) and presentation ζ = (1 + εξ )eiη , it follows that 2 dζ = ε 2 (ε −1 + 1)2 e2λ λ2 + (1 + εξ )2 η2 , (15) ϕ ϕ dϕ since ξ0 = 1 on the free boundary. By virtue of Equations (14), (15) and the presentation ηϕ = − log(1 + ε) − ∂n µ, the Bernoulli equation (8) takes the form 1 2λ 2 2 2 (1 + εξ ) (log(1 + ε) + ∂n µ) − τ e λϕ + (1 + ε)2
1 2
= 0,
which can be transformed in to the conservation law, (1 + ε 2 ) 2µ ∂ τ+ e . α(ϕ, 1) = ∂ϕ 2
(16)
Thus, the first equation of Equations (13) holds due to the analyticity of the function f (ω) and the second equation of Equations (13) follows from the definition of the normal derivative operator ∂n . Finally, the last equation is the consequence of the ✷ changing of the order of differentiation ∂ϕ ∂n µ = ∂n µϕ . Note that the function µ(ϕ) is found through the analyticity of f (ω) and thus transformation ξ(ϕ, ψ) is determined by definition (12) as 1 + ξ0 (ψ) (eλ(ϕ,ψ) − 1). ξ = ε
3.3. SOLUTION TO THE DIRICHLET PROBLEM IN A FIXED DOMAIN In view of Lemma 1, it follows from the definition for the function α(ϕ, 1) and Equation (16) that integrating Equations (13) over ϕ along ψ = 1 leads to the following equation on the free boundary: 4τ e2µ [log(1 + ε) + ∂n µ] + ∂n ([2τ + 1 + ε 2 ]e2µ ) − εδ0 = 0,
57
TIDAL MOTION AROUND THE EARTH
where δ0 is the constant of integrating which represent the horizontal impulse flow. Finally, simplifying the last equation, we arrive at the Dirihlet problem in the fixed domain λϕϕ + λψψ = 0 λ(ϕ, 0) = 0,
(0 < ψ < 1),
(17)
λ(ϕ, 1) = µ,
(18)
ε 1 −2 ∂n µ + − µ∂n µF − b + 1 − F 4 2 δe−2µ log(1 + ε) (µF −2 − b) + = 0, + 2 (1 + ε)2 −2
(19)
where we denote δ = δ0 /8. Now the problem (Equations (17) to (19)) is reduced to finding one function µ(ϕ) since if the function µ is known, then function λ(ϕ, ψ) is defined as the solution of the mixed problem for the Laplace equation (17) and the boundary conditions (18). In particular, λψ |ψ=1 can be considered as the result of the action of the operator ∂n on the function µ. Namely, we represent λ (ϕ, ψ) by the Fourier series (see, for example, [8] or [9]). Then the dependence between the nth Fourier coefficients of functions λ, µ and ∂n µ is given by [λ(ϕ, ψ)]n =
sinh nψ µn , sinh n
[µ(ϕ)]n = µn ,
[∂n µ(ϕ)]n = µn cot n, (20)
in which µ(ϕ) = µn einϕ (summation is assumed). Thus, problem (17)–(19) is written in terms of [µ(ϕ)]n only. Since [∂n µ]n are given by Equation (20), we can represent the disturbance µ(ϕ) by the expansion in series with respect to parameter ε (see also [8]). Consequently, we apply the stretching transformation and expansion ∞ ε i ε µ¯ i , ε(∂¯n µ) ¯ i (µ, ∂n µ) =
(i = 0, ∞)
(21)
i=0
which is characteristic of shallow water. Substitution of representation (21) into Equation (19) and elimination mod ε 3 (neglecting of the terms with ε m , m 4) yields the approximate solution of the form [µi ]n = [µi (b, F )]n as follows:
b F −2 1 b −2 µ1 − [1 − F + 4b](∂n µ)1 − 2δµ1 + δ − + ε −2δ + + 2 4 2 4
F −2 1 b F −2 µ1 + µ2 + (∂n µ)1 − + ε 3δ − − 6 4 4 2 2
58
RANIS N. IBRAGIMOV
1 −2 2 − [1 − F + 4b](∂n µ)2 + 4δµ1 + 2δµ1 − 2δµ2 + µ1 (∂n µ)1 + 4
F −2 F −2 1 b F −2 µ1 − µ2 + µ3 + (∂n µ)2 − + ε −4δ + + 8 6 4 2 2 3
−
1 [1 − F −2 + 4b](∂n µ)3 − 4
4 − 6δµ1 − 4δµ21 − δµ31 + 4δµ2 − 2δµ3 + 4δµ1 µ2 + µ1 (∂n µ)2 + 3 (22) + (∂n µ)1 µ2 + o(ε 4 ). Thus, in view of Equations (20), Equation (22) represents the recurrent system of algebraic equations for determination of all [µi ]n , where the horizontal impulse flow has asymptotic δ = b/2. The shape of the free boundary h (θ) can be determined numerically using Okmaoto’s method [3]. The existence of exact solution (ψ, h) can be established analytically by the Fixed Point Theorem (see, for example, [7] or [4]). 4. Behavior of Tides Waves 4.1. EXISTENCE OF STATIONARY WAVES The main concern of this section is the evolution of tides around the Earth in time t. In order to bring out the essential parameters of the problem, the dimensional fundamental equations, Equations (1) to (2), are written, in the nonstationary case, as follows: (ψ) = 0 (in h ), ψθ = 0 (on R ), rht + ψθ + ψr hθ = 0 (on h ), r ψθ2 2 + ψr + rghθ = 0 (on h ), −hθ ψt θ + r ψt r + 2 r2 θ 2
where = (∂θθ + r 2 ∂rr + r∂r ). The perturbed quantities h and ψ are interrelated as h = h ,
ψ =−
γ log r + ψ , 2π
TIDAL MOTION AROUND THE EARTH
59
where γ is the intensity of the vortex. For the small disturbances, we obtain the linear problem in the domain D0 = {(r, θ) : R r R + h0 , 0 θ 2π } as follows:
(ψ ) = 0 (in D0 ),
(23)
ψθ = 0 (on R ),
(24)
hθ +
γ hθ ψθ − = 0 (on h0 ), r 2π r 2
r 2 ψtr −
γ ψ + rghθ = 0 (on h0 ). 2π rθ
(25) (26)
Since Equations (23)–(26) are linear, the method of superposition is applicable (see also Friedrichs [2]). Hence it is sufficient to look for periodic solutions of the form (h , ψ ) = (H, )(r)) exp{i(kθ − wt)}
(27)
in which the wave number k is a given real quantity and eigenvalues w give the different modes of the tide’s wave propagation. Substitution of representation (27) into Equations (23)–(26) leads to the expression )(r) = c(r k − R 2k r −k ) and to the equations kc kγ − (r k − R 2k r −k ) = 0, H w+ 2 2π r r kγ kg H − w+ (kr k−1 + kR 2k r −k−1 ) = 0 cr 2π r 2 in which c is a constant of integration. Consequently, the determinantal equation for the longitudinal tide wave is as follows:
kg (R + h0 )k−1 − R 2k (R + h0 )−k−1 kγ (R+h0 ) − . (28) w=± (R + h0 )k−1 + R 2k (R + h0 )−k−1 2π(R + h0 )2 Thus surface tide waves (on the constant flow) are dispersive with two different modes of propagation. Simplification of relation (28) shows that the tide wave is propagated with a speed γ w tanh [k ln(1 + ε)] = ± gh0 − . a0 = 2 k kRh0(1 + ε) 2π R (1 + ε)2
60
RANIS N. IBRAGIMOV
Hence, the condition of the existence of stationary tide waves (a0 = 0) is 1 |γ | 2π Rε − 2 (1 + ε)2 gh0 .
4.2. SPLITTING PHENOMENA FOR SHALLOW WATER EQUATIONS We suppose that the parameter ε is infinitesimally small. So we consider R as the natural physical scale. Note that kinematic condition can be written as the mass balance equation. Namely, R+h −1 v θ dr = 0, (29) ht + (R + h) ∂θ R
since the radial velocity component is given by R+h vθθ dr. v r = −r −1 R
Hence, the mass balance equation (29) takes the form rht + ∂θ (uh) = 0 (on h ), where the average velocity u (θ, t) is defined by the relation R+h −1 v θ (r, θ, t)dr. u(θ, t) = h R
To go further, it is better to introduce an nondimensionalization here. We put t=
R t, U
ψ = h0 U ψ ,
u = U u ,
where U is a unit of velocity. Hereafter, the prime will be omitted. Then the impulse equation is written as 1 ε 2 ψθ2 ε 2 hθ ψt θ 2 ∂θ + ψt r + + ψr + − (1 + εh)2 2(1 + εh) (1 + εh)2 hθ = 0. (30) + (1 + εh) We represent function ψ by the Lagrangian expansion (see also [8] thei stream (i) ε ψ . Then the Laplace equation takes the form (mod ε 2 ) or [2]) ψ = ∞ i=0 (1) (0) (0) + ε ψrr + 2rψrr + ψr(0) + ψrr (0) (2) (1) (0) + ψrr + 2rψrr + r 2 ψrr + ψr(1) + rψr(0) = 0. (31) + ε 2 ψθθ Equation (31) represents the recurrent system of differential equations for the determination of ψ (i) as the solution of the Cauchy problem with boundary conditions
TIDAL MOTION AROUND THE EARTH
61
ψ(0, θ, t) = 0, ψ(1 + h, θ, t) = uh for ψ (0) and zero boundary conditions for ψ (1) and ψ (2) . Hence, the function ψ (mod ε 2 ) is as follows: 2 r r + ψ = ur + ε u − uh 2 2 r2 r3 2 2r 2r + ε −uθθ + uh + uθθ h − uh . (32) 6 4 6 4 We use the Tailor expansion (1 + εh)−1 = 1 − εh + (εh)2 + · · ·
(33)
to write Equation (30) as ψt r + 12 (ε 2 ψθ2 + ψr2 )θ − ε 2 hθ ψt θ + 1 + (εh − 1) εh(ψr2 )θ + εhh θ + hθ = 0. 2
(34)
Let us multiply Equation (30) by (1 + εh)2 and then use expansion (33). Then Equation (30) becomes ψt r + 12 (ε 2 ψθ2 + ψr2 )θ − ε 2 hθ ψt θ + 1 + εh ψt r (2 + εh) + (ψr2 )θ + hθ + hθ = 0. 2
(35)
If we substitute ψ defined by Equation (32) into (34), we obtain the following equation of the shallow water theory: h u 2 ut + uuθ + hθ + ε ut − ht + u hθ − hhθ + 2 2 3 h h2 h + ε 2 hhθ uθt − uθθt + ut + uθθ ht + hhθ u2θ + 3 4 3 2 3 1 1 2 2 (36) + h uθ uθθ + 4 uuθ + hu hθ − 3 uθ uθθ − 3 uuθθθ + h hθ = 0. Consequently, the substitution of ψ (32) into (35) yields h u h u2 3 ut + uuθ + hθ + ε ut − ht + uut + hθ + huuθ + 2 2 2 2 2 2 h 9 h + ε 2 − hhθ uθt − uθθt + h2 ut + ht uθθ − uhht + 3 4 3 3 1 1 + hhθ u2θ + h2 uθ uθθ + h2 uuθ + u2 hhθ − uθθθ + 4 4 3 1 uh 3 hθ = 0. + h2 uθ + uhθ + 4 2 2
(37)
62
RANIS N. IBRAGIMOV
Equations (36) and (37) supplied with the kinematic condition ∂t εh2 + 2h + 2∂θ (uh) = 0
(38)
represent two systems of the shallow water equations. To verify that these two systems are different, we compute their first integrals in the stationary case as follows: 1 2 1 3 2 2 2 2 h + ε 2c h − h + ε 3c h − h = J1 , 2 3 1 2 17 2 2 1 2 2 1 2 c h − c h − ch = J2 , h + εc h + ε 2 2 4 2 where c, J1 , J2 are constants of integrating. Obviously, J1 = J2 . At first sight, it seems that the problem (23)–(26) does not have a unique solution because of that fact. However, it can be shown that the solution of the problem is invariant with respect to the decomposition of the function which represents the free boundary. Since it is not difficult, the proof is omitted. Acknowledgements I wish to express my gratitude to Prof. N. Makarenko of Novosibirsk State University, Russia, for his helpful comments and advice throughout this study. In addition, I would like to thank Prof. L.V. Ovsyannikov for valuable discussions on this subject at seminars at Lavryentiev’s Institute of Hydrodynamics, Russia. I am grateful to Prof. H. Okamoto of the Research Institute of Mathematical Sciences (Kyoto University) for arranging my visit to Japan and for encouraging discussions. Also, I am grateful to Professor K. Lamb of the University of Waterloo, Canada, for valuable discussions and support. References 1. 2. 3. 4. 5. 6. 7.
Crapper, G. D.: Introduction to Water Waves, Ellis Horwood, London, 1984. Friedrichs, K. O. and Hyers, D. H.: The existence of solitary waves, Comm. Pure. Appl. Math. 7 (1954). Fujita, H., Okamoto, H. and Shoji, M.: A numerical approach to a free boundary problem of a circulating perfect fluid, Japan J. Appl. Math. 2 (1985), 197–210. Ibragimov, R. N.: Stationary surface waves on circular liquid layer, Quaest. Math. 23(1) (2000). Levi-Civita, T.: Determination rigoureuse des ondes permanents d’ampleur finie, Math. Ann. 93 (1925), 264–314. Okamoto, H.: Nonstationary free boundary problem for perfect fluid with surface tension, J. Math. Soc. Japan 38(3) (1986). Okamoto, H. and Shoji, M.: On the existence of progressive waves in the flow of perfect fluid around a circle, In: T. Nishida, M. Mimura and H. Fujii (eds), Patterns and Waves, 1986, pp. 631–644.
TIDAL MOTION AROUND THE EARTH
8. 9. 10. 11.
63
Ovsjannikov, L. V. and Makarenko, N. I.: Nonlinear Problems of Surface and Internal Waves Theory, Nauka, Novosibirsk, 1985. Stoker, J. J.: Water Waves, Interscience Publishers, New York, 1957. Stokes, G. G.: On the theory of oscillatory waves, Trans. Cambridge Philos. Soc. 8 (1847), 441–455. Lamb, K. G.: Are solitary waves solitons? Stud. Appl. Math. 101 (1998), 298–308.
Mathematical Physics, Analysis and Geometry 4: 65–96, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
65
From the Solution of the Tsarev System to the Solution of the Whitham Equations TAMARA GRAVA Department of Mathematics, University of Maryland, College Park 20742-4015, U.S.A. and Department of Mathematics, Imperial College, London SW7 2BZ, U.K. e-mail:
[email protected]. (Received: 24 October 2000; in final form: 30 May 2001) Abstract. We study the Cauchy problem for the Whitham modulation equations for increasing smooth initial data. The Whitham equations are a collection of one-dimensional quasi-linear hyperbolic systems. This collection of systems is enumerated by the genus g = 0, 1, 2, . . . of the corresponding hyperelliptic Riemann surface. Each of these systems can be integrated by the socalled hodograph transformation introduced by Tsarev. A key step in the integration process is the solution of the Tsarev linear overdetermined system. For each g > 0, we construct the unique solution of the Tsarev system, which matches the genus g + 1 and g − 1 solutions on the transition boundaries. Mathematics Subject Classifications (2000): 35Q53, 58F07. Key words: Whitham equations, hyperelliptic Riemann surfaces, linear overdetermined systems of Euler–Poisson Darboux type.
1. Introduction The Whitham equations are a collection of one-dimensional quasi-linear hyperbolic systems of the form [1–3] ∂ui ∂ui − λi (u1 , u2 , . . . , u2g+1 ) = 0, ∂t ∂x x, t, ui ∈ R, i = 1, . . . , 2g + 1, g = 0, 1, 2, . . . ,
(1.1)
with the ordering u1 > u2 > · · · > u2g+1 . For a given g, the system (1.1) is called g-phase Whitham equations. For g > 0, the speeds λi (u1 , u2 , . . . , u2g+1 ), i = 1, 2, . . . , 2g + 1, depend through u1 , . . . , u2g+1 on complete hyperelliptic integrals of genus g. For this reason, the g-phase system is also called a genus g system. The zero-phase Whitham equation has the form ∂u ∂u − 6u = 0, (1.2) ∂t ∂x where we use the notation u1 = u. Equations (1.1) were found by Whitham [1] in the single-phase case g = 1 and more generally by Flaschka, Forest and McLaughlin [2] in the multi-phase case.
66
TAMARA GRAVA
The Whitham equations were also found in [3] when studying the zero dispersion limit of the Korteweg–de Vries equation. The hyperbolic nature of the equations was found by Levermore [4]. In this paper we study the initial-value problem of the Whitham equations for increasing smooth (C ∞ ) initial data u(x, t = 0) = u0 (x). The initial-value problem consists of the following. We consider the evolution on the x − u plane of the initial curve u(x, t = 0) = u0 (x) according to the zerophase equation (1.2). The solution u(x, t) of (1.2), with the initial data u0 (x), is given by the characteristic equation x = −6tu + f (u),
(1.3)
where f (u)|t =0 is the inverse function of u0 (x). The solution u(x, t) in (1.3) is globally well defined only for 0 t < t0 , where t0 = 16 minu∈R [f (u)] is the time of gradient catastrophe of (1.3). Near the point of gradient catastrophe and for a short time t > t0 , the evolving curve is given by a multivalued function with three branches u1 (x, t) > u2 (x, t) > u3 (x, t), which evolve according to the one-phase Whitham equations (see Figure 1). Outside the multivalued region, the solution is given by the zero-phase solution u(x, t) defined in (1.3). On the phase transition boundary, the zero-phase solution and the one-phase solution are C 1 -smoothly attached (see Figure 1). Since the Whitham equations are hyperbolic, other points of gradient catastrophe can appear in the branches u1 (x, t) > u2 (x, t) > u3 (x, t) themselves or in u(x, t). In general, for t > t0 , the evolving curve is given by a multivalued function with an odd number of branches u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t), g 0.
(a)
(b)
Figure 1. In picture (a), the dashed line represents the formal solution of the zero-phase equation and the continuous line represents the solution of the one-phase equations. The solution (u1 (x, t) , u2 (x, t), u3 (x, t)) of the one-phase equations and the position of the boundaries x − (t) and x + (t) are to be determined from the conditions u(x − (t), t) = u1 (x − (t), t), ux (x − (t), t) = u1x (x − (t), t), u(x + (t), t) = u3 (x + (t), t), ux (x + (t), t) = u3x (x + (t), t), where u(x, t) is the solution of the zero-phase equation.
67
THE TSAREV AND WHITHAM EQUATIONS
These branches evolve according to the g-phase Whitham equations. The g-phase solutions for different g must be glued together in order to produce a C 1 -smooth curve in the (x, u) plane evolving smoothly with t (see Figure 1b). The initialvalue problem of the Whitham equations is to determine, for almost all t > 0 and x, the phase g(x, t) 0 and the corresponding branches u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t) from the initial data x = f (u)|t =0 . For generic initial data, it is not known whether the solution of the Whitham equations has a finite genus. Some results in this direction have been obtained in [5, 6]. Using the geometricHamiltonian structure [7] of the Whitham equations, Tsarev [8] showed that these equations can be locally integrated by a generalization of the method of characteristic. Namely, he proved that if the functions wi = wi (u1 , u2 , . . . , u2g+1 ), i = 1, . . . , 2g + 1, solve the linear over-determined system 1 ∂λi ∂wi = [wi − wj ], ∂uj λi − λj ∂uj
i, j = 1, 2, . . . , 2g + 1, i = j,
(1.4)
where λi = λi (u1 , u2 , . . . , u2g+1 ), i = 1, . . . , 2g + 1, are the speeds in (1.1), then the solution u(x, t) = (u1 (x, t), u2 (x, t), . . . , u2g+1 (x, t)) of the so-called hodograph transformation x = −λi (u)t + wi (u),
i = 1, . . . , 2g + 1,
(1.5)
satisfies system (1.1). Conversely, any solution (u1 (x, t), u2 (x, t), . . . , u2g+1 (x, t)) of (1.1) can be obtained in this way. Furthermore, the solution wi (u), i = 1, . . . , 2g + 1, g 0, of (1.4) must satisfy some natural matching conditions which guarantee that the g-phase solutions of (1.1) for different g are glued together in order to produce a C 1 -smooth curve in the (x, u) plane. Tsarev theorem relies on two factors: (a) the existence of a solution of the linear over-determined system (1.4); (b) the existence of a real solution u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t) of the hodograph transformation (1.5). In this paper, we investigate problem (a). We construct a new expression for the solution of the Tsarev system. This construction enables us to extend the solution of the Tsarev system, for g > 1, from analytic initial data with polynomial or exponential growth at infinity [15] to any smooth initial data. In addition, this new expression has the advantage that: (i) it is possible to evaluate explicitly the Jacobian of the hodograph transformation (see (5.45)). It turns out that the determinant of the Jacobian is propor2g+1 tional to the product j =1 g (ui ; u), where the function g (r; u) can be explicitly obtained from the initial data; (ii) it is simpler to study the hodograph transformation near the phase transition boundary.
68
TAMARA GRAVA
The investigation of the initial value problem of the Whitham equations was initiated by Gurevich and Pitaevskii [9]. In the case g 1, they solved Equations (1.1) for step-like initial data and studied numerically the case of cubic initial data. Krichever [10] introduced an algebro-geometric procedure to integrate (1.4). Based on this procedure, Potemin [11] obtained the explicit solution of the system (1.4) for cubic initial data. Kudashev and Sharapov in [12] and Gurevich, Krylov and El in [13] connected the solution of the Tsarev system (1.4) to the solution of some linear overdetermined systems of the Euler–Poisson–Darboux type introduced in [14]. The structure of the solution of the systems of the Euler–Poisson–Darboux type was in depth investigated by Tian in [5], where he obtained for g 1 the solution of the Tsarev system from the solution of the systems of the Euler–Poisson–Darboux type for any smooth monotone initial data. Furthermore, he partly solved problem (b) proving the solvability of the hodograph transformation for g 1. The structure of the multiphase solutions has been investigated by Tian [15] and later by El [16]. In [15], Tian built the solution of the Tsarev system for g > 1 for generic polynomial initial data. Again, such a solution is constructed in terms of solutions of the linear overdetermined systems of the Euler–Poisson–Darboux type. Tian’s formula is not a-priori generalizable to analytic initial data u0 (x) which are bounded at infinity or to any smooth initial data. In this paper, we obtain a new expression for the solution wi (u), i = 1, . . . , 2g + 1, of the Tsarev system for polynomial initial data. Such an expression enables us to generalize Tian’s result to any increasing smooth initial data. Furthermore, we prove that the solution obtained is unique. This paper is organized as follows. In Section 2 we give some background to Abelian differentials on hyperelliptic Riemann surfaces. We describe the Whitham equations in Section 3 where we show that the solution of the Tsarev system with some given natural matching conditions is unique. In Section 4, we construct a new formula for the solution of the Tsarev system (1.4) for polynomial initial data. Then we show that the formula obtained can be extended to any smooth increasing initial data. In Section 5, we prove that such formula guarantees that the g-phase solutions of (1.1) for different g are glued together in order to produce a C 1 -smooth multivalued curve in the x − u plane evolving smoothly with time. Our conclusions are drawn in Section 6. 2. Riemann Surfaces and Abelian Differentials: Notations and Definitions Let
Sg := P = (r, µ), µ2 =
2g+1
(r − uj ) ,
(2.1)
j =1
be the hyperelliptic Riemann surface of genus g 0 with real branch points g u1 > u2 > · · · > u2g+1 . We choose the basis {αj , βj }j =1 of the homology group H1 (Sg ) so that αj lies fully on the upper sheet and encircles clockwise the
69
THE TSAREV AND WHITHAM EQUATIONS
interval [u2j , u2j −1 ], j = 1, . . . , g, while βj emerges on the upper sheet on the cut [u2j , u2j −1 ], passes anti-clockwise to the lower sheet through the cut (−∞, u2g+1 ] and return to the initial point through the lower sheet. The one-forms that are analytic on the closed Riemann surface Sg except for a finite number of points are called Abelian differentials. We define on Sg the following Abelian differentials [17]: (1) The canonical basis of holomorphic one-forms or Abelian differentials of the first kind φ1 , φ2 . . . φg : φk (r) =
r g−1 γ1k + r g−2 γ2k + · · · + γgk µ(r)
k = 1, . . . , g.
dr,
The constants γik are uniquely determined by the normalization conditions φk = δj k , j, k = 1, . . . , g.
(2.2)
(2.3)
αj
We remark that the holomorphic differential having all its α-periods equal to zero is identically zero [17]. g
(2) The set σk , k 0, g 0, of the Abelian differentials of the second kind with a pole of order 2k + 2 at infinity, with asymptotic behavior 1 3 g (2.4) σk (r) = r k− 2 + O(r − 2 ) dr for large r and normalized by the condition g σk = 0, j = 1, . . . , g.
(2.5)
αj
We use the notation g
σ0 (r) = dp g (r),
g
12σ1 (r) = dq g (r), g
g 0.
(2.6)
g
In the literature, the differentials dp (r) and dq (r) are called quasi-momentum g and quasi-energy, respectively [7]. The explicit formula for the differentials σk , k 0, is given by the expression g
Pk (r) dr, µ(r) g k , Pk (r) = r g+k + c1k r g+k−1 + c2k r g+k−2 + · · · + cg+k g
σk (r) =
(2.7)
where the coefficients cik = cik (u), u = (u1 , u2 , . . . , u2g+1 ), i = 1, . . . , g + k, are uniquely determined by (2.4) and (2.5). (3) The Abelian differential of the third kind ωqq0 (r) with first-order poles at the points Q = (q, µ(q)) and Q0 = (q0 , µ(q0 )) with residues ±1, respectively. Its periods are normalized by the relation ωqq0 (r) = 0, j = 1, . . . , g. (2.8) αj
70
TAMARA GRAVA
2.1. RIEMANN BILINEAR RELATIONS Let ω1 and ω2 be two Abelian differentials on the Riemann surface Sg . We suppose that ω1 has some poles with nonzero residue so that the integral d−1 ω1 has logarithm singularities on Sg . Let s be the path connecting the singular points of d−1 ω1 . We have the following relation:
g ω1 ω2 − ω2 ω1 + $(d−1 ω1 )ω2 j =1
αj
= 2π i
βj
αj
βj
s
Res[(d−1 ω1 )ω2 ],
(2.9)
Sg −s
difference of the values of d−1 ω1 on the two sides of the where $(d−1 ω1 ) is the cut s and the quantity Sg −s Res[(d−1 ω1 )ω2 ] is the sum of the residues of the differential (d−1 ω1 )ω2 on the cut surface Sg − s. This formula is known as the Riemann bilinear period relation [18]. Assuming ω1 = ωqq0 and ω2 = ωpp0 in (2.9) we obtain q p ωqq0 = ωpp0 . (2.10) p0
q0
Differentiating the above expression, with respect to p and q we obtain the identity dq [ωqq0 (p)] = dp [ωpp0 (q)],
(2.11)
where dq and dp denote differentiation with respect to q and p, respectively. In the following we mainly use the normalized Abelian differential of the third g kind ωz (r) which has simple poles at the points Q± (z) = (z, ±µ(z)) with residue ±1 respectively. g The differential ωz (r) is explicitly given by the expression dr µ(z) − = φk (r) µ(r) r − z k=1 g
ωzg (r)
αk
dt µ(z) , µ(t) t − z
(2.12)
where φk (r), k = 1, . . . , g, is the normalized basis of holomorphic differentials. g Using the explicit expression of the φk (r)’s in (2.2), we write ωz (r) in the form dr µ(z) r g−j − dr, = Nj (z, u) µ(r) r − z j =1 µ(r) g
ωzg (r)
(2.13)
where Nj (z, u) =
g k=1
γjk αk
dt µ(z) µ(t) t − z
(2.14)
71
THE TSAREV AND WHITHAM EQUATIONS
and the coefficients γjk have been defined in (2.2). In order to provide a more useful expression for the Nj ’s, we apply the Riemann bilinear relation (2.9) to g g the differentials σm (r) and ωz (r), getting Q+ (z) σmg (ξ ) = − Res [ωzg (r) d−1 σmg (r)], m = 0, . . . , g, r=∞ Q− (z)
g 4 = − −µ(z)*mg + Nj (z)+m+1−j . (2.15) 2m + 1 j =1 In the above formula, *mg = 1 for m = g and zero otherwise, and the +l ’s are the coefficients of the expansion for ξ → ∞ of +2 +1 +l 1 −g− 12 =ξ + 2 + ··· + l + ··· . (2.16) +0 + µ(ξ ) ξ ξ ξ We define +k = 0 for k < 0. Solving (2.15) for Nj (z, u), we obtain +˜ N1 (z, u) 0 N2 (z, u) +˜ 1 ··· = ··· N (z, u) +˜ g−1 g −µ(z) +˜ g
0 +˜ 0 ··· +˜ g−2 +˜ g−1
0 0 ··· ··· ···
Q+ (z) g − 14 Q− (z) σ0 (ξ ) 0 + (z) g 0 − 3 QQ− (z) σ1 (ξ ) 4 ··· , · Q·+· (z) g 2g−1 0 − Q− (z) σg−1 (ξ ) 4 Q+ (z) g +˜ 0 − 2g+1 Q− (z) σg (ξ ) 4 (2.17)
··· ··· ··· +˜ 0 +˜ 1
where the +˜ k ’s are the coefficients of the expansion for ξ → ∞ of +˜ 1 +˜ 2 +˜ l g+ 21 ˜ + 2 + ··· + l + ··· . +0 + µ(ξ ) = ξ ξ ξ ξ
(2.18)
From the relation (2.17), we obtain the identity which will be useful later Q+ (z) g+1 1 g ˜ (2k − 1)+g+1−k σk−1 (ξ ). µ(z) = 4 k=1 Q− (z)
(2.19)
The next proposition is also important for our subsequent considerations. g
PROPOSITION 2.1. The Abelian differentials of the second kind σk (r), k 0, defined in (2.4) satisfy the relations 1 1 1 1 g dr Res ωrg (z) zk+ 2 , (2.20) σk (r) = Res ωzg (r) zk− 2 dz = − 2 z=∞ 2k + 1 z=∞ g
g
where ωz (r) has been defined in (2.12), ωr (z) is the normalized Abelian differential of the third kind with simple poles at the points Q± (r) = (r, ±µ(r)) with residue ±1, respectively, and dr denotes differentiation with respect to r.
72
TAMARA GRAVA g
1
g
Proof. The differential Resz=∞ [ωz (r) zk− 2 dz] is normalized because ωz (r) is a normalized differential. From (2.13) and (2.17) it can be easily shown that 1 1 3 Res ωzg (r) zk− 2 dz = r k− 2 dr + O(r − 2 ) dr
z=∞
g
for r → ∞.
1
Therefore, Resz=∞ [ωz (r) zk− 2 dz] coincides with the normalized Abelian differeng tial of the second kind σk (r). For proving the second equality in (2.20) we consider the integral in the z variable 1 g g k+ 12 k+ 21 g dz (ωz (r) z ) = z (dz ωz (r)) + (k + 12 )(ωz (r) zk+ 2 ), 0= C∞
C∞
C∞
(2.21) where C∞ is a close contour around the point at infinity. From (2.11), we obtain g g the identity dz ωz (r) = dr ωr (z). Substituting the above identity in the right-hand side of (2.21), we obtain the second relation in (2.20). ✷
3. Preliminaries on the Theory of the Whitham Equations The speeds λi (u1 , u2 , . . . , u2g+1 ) of the g-phase Whitham equations (1.1) are given by the ratio [1, 2]: dq g (r) , i = 1, 2, . . . , 2g + 1, (3.1) λi (u) = dp g (r) r=ui where dp g (r) and dq g (r) have been defined in (2.6). In the case g = 0 dp 0 (r) = √
dr , r −u
12r − 6u dq 0 (r) = √ dr, r −u
(3.2)
so that one obtains the zero-phase Whitham equation (1.2). For monotonically increasing smooth initial data x = f (u)|t =0 , the solution of the zero-phase equation (1.2) is obtained by the method of characteristic [1] and is given by the expression x = −6tu + f (u).
(3.3)
The zero-phase solution is globally well-defined only for 0 t < t0 , where t0 = 16 minu∈R [f (u)] is the time of gradient catastrophe of the solution (3.3). The breaking is caused by an inflection point in the initial data. For t t0 , we expect to have single, double and higher phase solutions. For higher genus the Whitham equations can be locally integrated using a generalization of the characteristic equation (3.3). We have the following theorem of Tsarev [8]
73
THE TSAREV AND WHITHAM EQUATIONS
THEOREM 3.1. If wi (u) solves the linear over-determined system ∂wi = aij [wi − wj ], ∂uj 1 ∂λi , aij = λi − λj ∂uj
i, j = 1, 2, . . . , 2g + 1, i = j, (3.4)
then the solution (u1 (x, t), u2 (x, t), . . . , u2g+1 (x, t)) of the hodograph transformation x = −λi (u) t + wi (u),
i = 1, . . . , 2g + 1,
(3.5)
satisfies system (1.1). Conversely, any solution (u1 , u2 , . . . , u2g+1 ) of (1.1) can be obtained in this way in a neighborhood (x0 , t0 ) where the uix ’s are not vanishing. To guarantee that the g-phase solutions for different g are attached continuously, the following natural matching conditions involving f (u) must be imposed on wi (u1 , u2 , . . . , u2g+1 ), i = 1, . . . , 2g + 1, g > 0. When ul = ul+1 , 1 l 2g, g
wl (u1 , . . . , ul−1 , ul , ul , ul+2 , . . . , u2g+1 ) g = wl+1 (u1 , . . . , ul−1 , ul , ul , ul+2 , . . . , u2g+1 )
(3.6)
and, for 1 i 2g + 1, i = l, l + 1, g
wi (u1 , . . . , ul−1 , ul , ul , ul+2 , . . . , u2g+1 ) g−1
= wi
(u1 , . . . , uˆ l , uˆ l , . . . , u2g+1 ).
(3.7)
The superscript g and g − 1 in the wi ’s specify the corresponding genus and the hat denotes the variable that have been dropped. When g = 1 we have that w11 (u1 , u1 , u3 ) = w21 (u1 , u1 , u3 ),
w21 (u1 , u3 , u3 ) = w31 (u1 , u3 , u3 ) (3.8)
and w31 (u1 , u1 , u3 ) = f (u3 ),
w11 (u1 , u3 , u3 ) = f (u1 ),
(3.9)
where f (u) is the initial data. In the following, we will sometimes omit the superscript g when we are referring to genus g quantities. We remark that the λi (u)’s in (3.1) satisfy the matching conditions (3.6)–(3.8) and, for g = 1, we have λ1 (u1 , u3 , u3 ) = 6u1 ,
λ3 (u1 , u1 , u3 ) = 6u3 .
The matching conditions (3.6)–(3.9) guarantee that the solution of the Tsarev system (3.4) is unique.
74
TAMARA GRAVA g
THEOREM 3.2. If the initial data f (u) ≡ 0, then the solution wi (u) of (3.4) with matching conditions (3.6)–(3.9) is identically zero for 1 i 2g + 1, for any g 0 and for all u1 > u2 > · · · > u2g+1 . Proof. The proof is obtained by induction on g. The statement is satisfied for g = 0. For g = 1, we repeat the arguments of [5]. We fix u2 and we consider Equation (3.4) with matching conditions (3.8)–(3.9), namely ∂w11 = a13 [w11 − w31 ], ∂u3
∂w31 = a31 [w31 − w11 ], ∂u1
w11 (u1 , u2 , u2 ) = f (u1 ) ≡ 0,
w31 (u2 , u2 , u3 ) = f (u3 ) ≡ 0.
We can regard each of the above equations as a first-order linear ordinary differential equation with a nonhomogeneous term. Integrating them, we obtain a couple integral equation. By the standard contraction mapping method, it can be shown that when f (u) ≡ 0, this system has only the zero solution, i.e. w11 ≡ w31 ≡ 0 for (u1 , u3 ) satisfying u1 > u2 > u3 . Because of the arbitrariness of u2 , w11 and w31 vanish as a function of (u1 , u2 , u3 ) and, therefore, by (3.4), so does w21 (u). Now we suppose the theorem true for genus g − 1 and we prove it for genus g. We fix g g u2 > u3 > · · · > u2g and we consider Equation (3.4) for w1 and w2g+1 with the matching conditions (3.6)–(3.7), namely g ∂ g g w1 = a1(2g+1) w1 − w2g+1 , ∂u2g+1 g ∂ g g w2g+1 = a(2g+1)1 w2g+1 − w1 , (3.10) ∂u1 g g−1 w1 (u1 , u2 , . . . , u2g , u2g ) = w1 (u1 , u2 , . . . , u2g−1 , uˆ 2g , uˆ 2g ) ≡ 0, g
g−1
w2g+1 (u2 , u2 , . . . , u2g , u2g+1 ) = w2g+1(uˆ 2 , uˆ 2 , u3 , . . . , u2g , u2g+1 ) ≡ 0. Repeating the arguments developed for genus g = 1, we may conclude that g g w1 (u) ≡ w2g+1(u) ≡ 0, for arbitrary u1 > u2 > · · · > u2g+1 . We then repeat the above argument fixing u1 > u3 > · · · > u2g−1 > u2g+1 and considering g g Equation (3.4) for w2 (u) and w2g (u) with the matching conditions (3.6)–(3.7), namely g ∂ g g w2 = a2(2g) w2 − w2g , ∂u2g g ∂ g g w2g = a(2g)2 w2g − w2 , ∂ug 2 (3.11) w2 (u1 , u2 , . . . , u2g−1 , u2g+1 , u2g+1 ) g−1
(u1 , u2 , . . . , u2g−1 , uˆ 2g+1 , uˆ 2g+1 ) ≡ 0, g g−1 w2g (u1 , u1 , u3 , . . . , u2g , u2g+1 ) = w2g (uˆ 1 , uˆ 1 , u3 , . . . , u2g , u2g+1 ) = w2
g
g
≡ 0.
It can be easily shown that also w2 (u) ≡ w2g (u) ≡ 0 for arbitrary u1 > u2 > · · · > u2g+1 . Repeating these arguments some other g − 2 times, we conclude that
THE TSAREV AND WHITHAM EQUATIONS
75
g
wi (u) ≡ 0 for 1 i g, g + 2 i 2g + 1 and for arbitrary u1 > u2 > · · · > u2g+1 . Applying (3.4) and the matching conditions (3.6)–(3.7), we can prove that g ✷ also wg+1 (u) is identically zero. The theorem is then proved. The solution of the Tsarev system (3.4) with the matching conditions (3.6)– (3.9) has been obtained in [15, 10] for monotonically increasing polynomial initial data of the form x = fa (u) = c0 + c1 u + · · · + ck uk + · · · ,
(3.12)
where we assume that only a finite number of ck is different from zero. For such initial data, the wi (u)’s which satisfy (3.4) and the matching conditions (3.6)–(3.9) are given by the expression [15] ds g (r) , i = 1, . . . , 2g + 1. (3.13) wi (u) = dp g (r) r=ui The differential ds g (r) in (3.13) is given by ds (r) = g
∞ k=0
2k k! g ck σk (r), (2k − 1)!!
(3.14)
g
and the differentials σk (r), k 0 have been defined in (2.7). For the initial data (3.12), Tian [15] has reduced the expression of the wi (u)’s in (3.13) to the form
wi (u) =
g ∂ j qk γk ∂ui k=1 j
∂γ1 ∂ui
,
i = 1, 2 . . . , 2g + 1, 1 j g,
(3.15)
j
where the γk ’s are the normalization constants of the holomorphic differential φj (r) in (2.2) and the functions qk = qk (u), k = 1, . . . , g, solve the linear overdetermined system of the Euler–Darboux–Poisson type [15] ∂ 2 qk (u) ∂qk (u) ∂qk (u) = − , ∂ui ∂uj ∂ui ∂uj i, j = 1, . . . , 2g + 1, k = 1, . . . , g, g−k 2g−1 1 d 1 u−k+ 2 ug− 2 fa(k−1) (u) , qk (u, u, . . . , u) = (2g − 1)!! du 2(ui − uj )
(3.16)
2g+1
where fa(k−1)(u) is the (k − 1)th derivative of the polynomial initial data fa (u) defined in (3.12). Thus, the solution of the Tsarev system is reduced to the solution of the above systems of equations. The systems (3.16) can be integrated for any smooth initial data [5].
76
TAMARA GRAVA
THEOREM 3.3. Let f (u) be a smooth function with domain (a, b), −∞ a < b +∞. The initial value problem ∂ 2 qk (u) ∂qk (u) ∂qk (u) = − , ∂ui ∂uj ∂ui ∂uj i = j, i, j = 1, . . . , 2g + 1, g > 0, qk (u, u, . . . , u) = Fk (u), 2(ui − uj )
(3.17) (3.18)
2g+1
Fk (u) =
g−k 2(g−1) 1 d 1 u−k+ 2 g−k ug− 2 f (k−1)(u) , (2g − 1)!! du
(3.19)
with the ordering b > u1 > u2 > · · · > u2g+1 > a, has one and only one solution. The solution is symmetric and is given by 1 1 1 1 qk (u) = ··· dξ1 dξ2 . . . dξ2g (1 + ξ2g )g−1 × C −1 −1 −1 × (1 + ξ2g−1 )g− 2 . . . (1 + ξ3 ) 2 (1 + ξ1 )− 2 × 3
×
Fk (
1
1+ξ2g 2 1+ξ1 (. . . ( 1+ξ ( 2 u1 2 2
+
1
1−ξ1 u2 ) 2
+
1−ξ2 u3 ) 2
+ · · ·) +
(1 − ξ1 )(1 − ξ2 ) . . . (1 − ξ2g )
k = 1, . . . , g, 2g where C = m=1 Cm , 1 m (1 + µ) 2 −1 dµ, Cm = √ 1−µ −1
1−ξ2g u2g+1 ) 2
,
(3.20)
m > 0.
(3.21)
Proof. To prove the theorem we follow the procedure in [5]. We start with the following lemma. LEMMA 3.4 [5]. The system 2(z − y)hzy = hz − ρhy , h(z, z) = s(z),
ρ > 0,
(3.22)
has, for any smooth initial data s(z), one and only one solution. Moreover, the solution can be written explicitly 1 1+µ s( 2 z + 1−µ y) ρ 1 2 (1 + µ) 2 −1 dµ, (3.23) h(z, y) = √ Cρ −1 1−µ where Cρ has been defined in (3.21). Using the above lemma, the linear over-determined systems (3.17) can be integrated for any smooth initial data in the following way. Suppose that qk (u1 , u2 , . . . , u2g+1 ) is the solution of (3.17)–(3.19).
77
THE TSAREV AND WHITHAM EQUATIONS
Clearly Ak (u1 , u2g+1 ) = qk (u1 , u1 , . . . , u1 , u2g+1 ) 2g
satisfies ∂ 2 Ak ∂Ak ∂Ak = − 2g , ∂u1 ∂u2g+1 ∂u1 ∂u2g+1 Ak (u, u) = Fk (u), 2(u1 − u2g+1 )
(3.24)
which, by Lemma 3.4, implies that 1 1+ξ 1−ξ Fk ( 2 2g u1 + 2 2g u2g+1 ) 1 (1 + ξ2g )g−1 dξ2g . (3.25) Ak (u1 , u2g+1 ) = C2g −1 1 − ξ2g For each fixed u2g+1 , the function Bk (u1 , u2g , u2g+1 ) = qk (u1 , . . . , u1 , u2g , u2g+1 ) 2g−1
satisfies ∂ 2 Bk ∂Bk ∂Bk = − (2g − 1) , ∂u1 ∂u2g ∂u1 ∂u2g Bk (u, u, u2g+1 ) = Ak (u, u2g+1 ).
2(u1 − u2g )
(3.26)
Again using Lemma 3.4, we obtain 1 1 1 3 dξ2g dξ2g−1 (1 + ξ2g )g−1 (1 + ξ2g−1)g− 2 × Bk (u1 , u2g , u2g+1 ) = C2g C2g−1 −1 −1 ×
Fk (
1+ξ2g 1+ξ2g−1 ( 2 u1 2
1+ξ
+ 22g−1 u2g ) 1 − ξ2g 1 − ξ2g−1
1−ξ2g u2g+1 ) 2
.
(3.27)
Continuing the process of integration, we obtain the solution (3.20). The uniqueness follows from Lemma 3.4 and the above arguments. The boundary conditions (3.18)–(3.19) are clearly satisfied. The symmetry property is obtained considering the solution hk (u) = qk (u) − qk (P (u)), where P (u) is any reordering of the ui ’s. Such a solution satisfies (3.17) with Fk (u) = 0 and therefore, by construction, equals zero. ✷ The formula for the wi ’s obtained in (3.13) is valid only for a Taylor series with an infinite radius of convergence. This corresponds to an increasing initial data which is the sum of exponentials, sine, cosine, and polynomials. Therefore, it is not obvious that formula (3.15) can be extended to analytic initial data which have a different behavior at infinity or to any smooth initial data.
78
TAMARA GRAVA
In the next section we provide a new expression for the wi (u)’s equivalent to (3.15) which enables us to make such an extension. Namely, we show that these wi (u)’s are the unique solution of the Tsarev system (3.4) with matching conditions (3.6)–(3.9) for any monotonically increasing smooth initial data. Furthermore, this new expression enables us to evaluate the Jacobian of the hodograph transformation (3.5) very easily. Namely, the Jacobian turns out to be propor2g+1 tional to the product j =1 g (ui ; u), where the function g (r; u) satisfies a linear overdetermined system of the Euler–Poisson–Darboux type. 4. Solution of the Tsarev System In this section, we build the solution of the Tsarev system (3.4) with matching conditions (3.6)–(3.9) for monotonically increasing smooth initial data. We consider initial data of the form x = f (u)|t =0 , where f (u) is a monotonically increasing function. The domain of f is the interval (a, b), where −∞ a < b +∞, and the range of f is the real line (−∞, +∞). In order to obtain such a solution, we need the following technical lemma: LEMMA 4.1. The differential ds g (r) defined in (3.14) can be written in the form
2g+1 R g (r) dr, (4.1) ∂uk 5 g (r; u) dr + ds g (r) = 2µ(r) ∂r 5 g (r; u) + µ(r) k=1 where
! F (z) dz , 5 (r; u) = − Res z=∞ 2µ(z)(z − r) k = 1, . . . g, z fa (ξ ) dξ, F (z) = √ z−ξ 0 g
2g+1
R (r) = 2 g
(4.3)
2g+1
∂uk qg (u)
g
(r − un ) +
n=1,n=k
k=1
+
! zg−k F (z) dz qk (u) = − Res , z=∞ 2µ(z) (4.2)
k g qk (u) (2n − 1)+˜ k−n Pn−1 (r),
k=1
(4.4)
n=1
g the polynomials Pn (r), n 0, have been defined in (2.7), the +˜ k ’s have been defined in (2.18) and fa (ξ ) is the analytic initial data (3.12). Proof. Using the second identity in (2.20), we rewrite the differential ds g (r) defined in (3.14) in the form # " (4.5) ds g (r) = −dr Res[ωrg (z)F (z)] , z=∞
79
THE TSAREV AND WHITHAM EQUATIONS g
where ωr (z) has been defined in (2.12) and F (z) is the Abel transform defined in (4.3) of the analytic initial data (3.12). The identity (4.5) can be checked in a g straightforward manner. Using the explicit expression of ωr (z) in (2.13), we obtain ds (r) = 2dr (µ(r)5 (r; u)) + g
g
g
qk (u)
k=1
k
g (2n − 1)+˜ k−n σn−1 (r),
(4.6)
n=1
where 5 g (r; u) and qk (u) have been defined in (4.2). From (4.2) we get the relations 5 g (r; u) 5 g (ui ; u) − = 2∂ui 5 g (r; u), r − ui r − ui and for g = 0 we define u1 = u and
2∂ui qg (u) = 5 g (ui ; u)
(4.7)
2∂u q0 (u) := 5 0 (u; u) = fa (u). Using (4.7) we transform the expression for ds g (r) in (4.6) to the form (4.1).
✷
The relation (4.1) enables us to write the quantities ds g (r) , i = 1, . . . , 2g + 1, wi (u) = dp g (r) r=ui
in (3.13) in the form
2g+1 1 2∂ui qg (u) (ui − un ) + wi (u) = g P0 (ui ) n=1,n=i +
g k=1
k g qk (u) (2n − 1)+˜ k−n Pn−1 (ui ) .
(4.8)
n=1
We observe that in the formula (4.8) all the information on the initial data is contained in the functions qk (u). The functions qk = qk (u), k = 1, . . . , g, solve the linear over-determined system (3.16). THEOREM 4.2 (Main Theorem). Let f (u) be a smooth monotonically increasing function with domain (a, b), −∞ a < b +∞ and range (−∞, +∞). If qk = qk (u1 , u2 , . . . , u2g+1 ), 1 k g, is the symmetric solution of the linear over-determined system (3.17)–(3.19), with the ordering b > u1 > u2 > · · · > u2g+1 > a, then wi (u), i = 1, . . . , 2g + 1, defined by 2g+1 1 2∂ui qg (u) (ui − un ) + wi (u) = g P0 (ui ) n=1,n=i
g k g qk (u) (2n − 1)+˜ k−n Pn−1 (ui ) , (4.9) + k=1
n=1
solves the Tsarev system (3.4) with matching conditions (3.6)–(3.9).
80
TAMARA GRAVA
Proof. We consider the nontrivial case where qk (u) ≡ 0, k = 1, . . . , g, and ∂uj qg (u) ≡ 0, j = 1, . . . , 2g + 1. The proof of the theorem consists of two parts. (a) The wi (u)’s defined in (4.9) satisfy (3.4). Using the definition of wi (u) in (4.9) we have the following relation: 2g+1 2g+1 n=1, n=i (ui − un ) n=1, n=i (ui − un ) ∂uj ∂ui qg (u) − 2 ∂ui qg (u) × ∂uj wi (u) = 2 g g P0 (ui ) P0 (ui ) g ∂uj P0 (ui ) 1 × + + g ui − uj P0 (ui ) g
g g Pn−1 (ui ) (2n − 1) g qk (u)+˜ k−n , + ∂ uj P (u ) i 0 k=n n=1 i = j, i, j = 1, . . . , 2g + 1.
(4.10)
From (3.4), we obtain g g g 1 ∂λi Pk (ui ) Pk (uj ) ∂ Pk (ui ) = − , ∂uj P0g (ui ) λi − λj ∂uj P0g (ui ) P0g (uj ) i = j, i, j = 1, . . . , 2g + 1, k 1
(4.11)
and, from [15], we have g
∂uj P0 (ui ) 1 1 ∂λi 1 − =− , g λi − λj ∂uj 2 ui − uj P0 (ui ) i = j, i, j = 1, . . . , 2g + 1,
(4.12)
g
where λi (u) has been defined in (3.1) and Pk (r) has been defined in (2.4). Using the definition of +˜ k in (2.18), we get k 1 ∂ +˜ k =− . +˜ k−m um−1 j ∂uj 2 m=1
(4.13)
It is easy to verify that the functions Fk (u) and the solutions qk (u), k = 1, . . . , g, of (3.17)–(3.19) satisfy the relations 2g + 1 Fk+1 (u) + u∂u Fk+1 (u), k = 1, . . . , g − 1, 2 1 ∂ui qk (u) = qk+1 (u) + ui ∂ui qk+1 (u), 2 i = 1, . . . , 2g + 1, k = 1, . . . , g − 1, g > 0.
∂u Fk (u) =
g > 0, (4.14)
Repeatedly applying the relations (4.14), we obtain the following expression for ∂uj qk (u): 1 g−k qm+k (u)um−1 + uj ∂uj qg (u), j 2 m=1 g−k
∂uj qk (u) =
k = 1, . . . , g − 1.
(4.15)
81
THE TSAREV AND WHITHAM EQUATIONS
From (3.17) and (4.11)–(4.15) we can write ∂uj wi (u), i = j , in (4.10) in the form 2g+1 n=1,n=i (ui − un ) ∂ui qg (u) − ∂uj qg (u) − ∂uj wi (u) = g ui − uj P0 (ui ) 2g+1 g ∂uj P0 (ui ) 1 n=1,n=i (ui − un ) ∂ui qg (u) + −2 + g g ui − uj P0 (ui ) P0 (ui ) g−1 g−k g g Pn−1 (ui ) 1 ˜ +k−n (2n − 1) g qm+k (u)um−1 + + j 2 P (u ) i 0 k=n n=1 m=1
g g−k +˜ k−n uj ∂uj qg (u) + + k=n
g g g k−n Pn−1 (ui ) 1˜ m−1 (2n − 1) g qk (u) − +k−n−m uj + + 2 P (u ) i 0 n=1 k=n m=1 g (2n − 1) + n=1
1 ∂λi λi − λj ∂uj
g g g Pn−1 (ui ) Pn−1 (uj ) − g qk (u)+˜ k−n . g P0 (ui ) P0 (uj ) k=n (4.16)
Simplifying, we obtain 2g+1 ∂uj wi (u) = −
− un ) ∂u qg (u) g uj )P0 (ui ) j
n=1,n=i (ui
(ui − 2g+1
n=1,n=i (ui − g P0 (ui )
+
un )
1 ∂λi ∂ui qg (u) +2 λi − λj ∂uj g g Pn−1 (ui ) ∂λi 1 − (2n − 1) + g λi − λj ∂uj n=1 P0 (ui )
+
g g Pn−1 (uj ) qk (u)+˜ k−n + − g P0 (uj ) k=l g
P (ui ) g−k (2n − 1) n−1 +˜ k−n uj ∂uj qg (u), + g P0 (ui ) k=n n=1 g
g
i = j, i, j = 1, . . . , 2g + 1. Adding and subtracting the quantity ∂λi 1 wj λi − λj ∂uj
(4.17)
82
TAMARA GRAVA
to (4.17), we can reduce it to the form ∂λi 1 [wi − wj ] λi − λj ∂uj 2g+1 2 ∂λi n=1,n=j (uj − un ) − = ∂uj qg (u) g λi − λj ∂uj P0 (uj )
2g+1 g g g Pn−1 (ui ) n=1,n=i (ui − un ) g−k . (4.18) + (2n − 1) g +˜ k−n uj − g (ui − uj )P0 (ui ) P (u ) i 0 n=1 k=n
∂ uj w i −
The term in parentheses in the right-hand side of (4.18) does not depend on the initial data f (u). It is identically zero for the initial data (3.12) because, in such a case, the wi ’s satisfy (3.4) [15]. Therefore we conclude that ∂ uj w i −
1 ∂λi [wi − wj ] = 0 λi − λj ∂uj
(4.19)
for any smooth monotonically increasing initial data x = f (u)|t =0 . (b) The wi (u)’s satisfy the matching conditions (3.6)–(3.9). In the following, we use the superscript g to denote the corresponding genus of the quantities we are referring to. We have the following relations. When ul = ul+1 = v for 1 l 2g, the +˜ k ’s defined in (2.18) satisfy g +˜ k (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g−1 = +˜ k (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 )−
− v +˜ k−1 (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ), g−1
k 1, g > 1,
(4.20)
and the qk (u)’s defined in (3.20) satisfy g
qk (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 )− g
− vqk+1 (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g−1
= qk (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ), k = 1, . . . , g − 1, g > 1.
(4.21)
For ui = ul = ul+1 = v, we have g−1
∂ui qg−1 (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 )− 1 − qgg (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) 2 = (ui − v)∂ui qgg (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ),
(4.22)
which follows from (4.14). g g Next we study the behavior of the polynomials Pk (r) = Pk (r; u), k 0, when √ two branch points become coincident. For this purpose, let us consider ul = v+ *,
83
THE TSAREV AND WHITHAM EQUATIONS
√
g
ul+1 = v − *, where 0 < * 1. When l is odd, the polynomial Pk (r; u) defined in (2.7) satisfies the following expansion [20]: √ √ g Pk (r; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) g−1
= (r − v)Pk
(r)+
g−1 * g−1 g−1 ˜ − (r − v) r g−1−k ∂v Nk (v) + O(* 2 ). (4.23) + σk (v) ∂v µ(v) 2 k=1
In the above formula, the normalized Abelian differential of the second kind g−1 Pk (r) g−1 dr, σk (r) = µ(r) ˜ with pole at infinity of order 2k + 2 and with asymptotic behavior (2.4) is defined on the Riemann surface µ˜ 2 = (r − u1 )(r − u2 ) . . . (r − ul−1 )(r − ul+2 ) . . . (r − u2g+1 ),
(4.24)
the quantity g−1
g−1
σk
(v) =
Pk (v) µ(v) ˜
g−1
and the Nk (v)’s have been defined in (2.17). When l is even, we have [20] √ √ g Pk (r; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) Q+ (v) (r − v) g−1 g−1 g−1 µ(r)ω ˜ σk (ξ ), (r − v)Pk (r) − v (r) log * Q− (v)
(4.25)
where ωvg−1(r) is the normalized Abelian differential of the third kind defined on the Riemann surface (4.24) and with simple poles at the points Q± (v) = (v, ±µ(v)) ˜ with residue ±1, respectively. From (4.23) and (4.25), we have that for ui = ul = ul+1 = v g
Pk (ui ; u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g−1
= (ui − v)Pk
(ui ),
k 0.
(4.26)
Using the relations (4.20)–(4.22) and (4.26), we obtain for i = l, l + 1, i = 1, . . . , 2g + 1, g
wi (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) 2g+1 k=1,k=i, l, l+1 (ui − uk ) g−1 ∂ui qg−1 + =2 g−1 P0 (ui )
84
TAMARA GRAVA g−1
+
(2m − 1)
m=1 g
+
Pm−1 (ui ) g−1
g−1
g−1
(ui ) k=m
P0
g−1
(2m − 1)
m=1
Pm−1 (ui ) g−1 P0 (ui )
g−1
qk
g−1 +˜ g−m
g−1 +˜ k−m +
2g+1 −
k=1,k=i, l, l+1 (ui g−1 P0 (ui )
− uk )
qgg (u)|ul =ul+1 , (4.27)
with g−1 g−1 +˜ k = +˜ k (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ),
k0
and g−1
g−1
= qk
qk
(u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ),
k = 1, . . . , g − 1.
The above reduces to the form g
wi (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g−1
= wi
+
(u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 )+
2g+1 g g−1 Pm−1 (ui ) g−1 k=1,k=i, l, l+1 (ui − uk ) +˜ g−m − qgg (u)|ul =ul+1 . (2m − 1) g−1 g−1 P (u ) P (u ) i i 0 0 m=1 (4.28)
Using (2.19), the term in parentheses in the right-hand side of (4.28) turns out to be identically zero. Therefore, the boundary conditions (3.7) are satisfied for any smooth monotonically increasing initial data. For studying the boundary conditions √ (3.6), we need to√distinguish between l odd or even. Let us define ul = v + *, ul+1 = v − *, where 0 < * 1. g From (4.23) when l is odd√ the polynomial Pk (r; u) defined in (2.7) has the following expansion at r = v ± *: √ √ √ g Pk (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) √ g−1 g−1 g−1 ˜ + 2∂v Pk (v)) + O(* 2 ), = ± * Pk (v) + *(σk (v)∂v µ(v) k 0, (4.29) so that √ √ √ g Pk (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) lim g √ √ √ *→0 P (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) 0 g−1
=
Pk
(v)
g−1 P0 (v)
.
(4.30)
85
THE TSAREV AND WHITHAM EQUATIONS g−1
g−1
We remark that here and below Pk (v) = Pk (v; u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ). When l is even, we have from (4.25) √ √ √ g Pk (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) Q+ (v) √ g−1 1 g−1 µ(v) ˜ σk (ξ ), (4.31) ± * Pk (v) − − log * Q (v) so that
√ √ √ g Pk (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) lim g √ √ √ *→0 P (v ± *; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 ) 0 Q+ (v) g−1 (ξ ) Q− (v) σk = Q+ (v) g−1 . (ξ ) Q− (v) σ0
(4.32)
Using (4.20), (4.21) and (4.30) we deduce that for l odd and ul = ul+1 = v g
wl (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g−1 g−1 k 2µ˜ 2 (v) ∂ g g−1 g−1 Pn−1 (v) ˜ = g−1 qg (u) + qk (2n − 1)+k−n g−1 + P0 (v) ∂ul P0 (v) ul =ul+1 k=1 n=1 +
qgg (u)|ul =ul+1
g g−1 g−1 Pn−1 (v) ˜ (2n − 1)+g−n g−1 , P0 (v) n=1
(4.33)
where g−1
qk
g−1
= qk
(u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 )
and g−1 g−1 +˜ k = +˜ k (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ).
Since the functions qk (u) in (3.20) are symmetric with respect to u1 , u2 , . . . , u2g+1 , we immediately obtain that ∂ ∂ g g qg (u) = qg (u) . (4.34) ∂ul ∂ul+1 ul =ul+1 ul =ul+1 Therefore, combining (4.33) and (4.34), we deduce that g
wl (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g = wl+1 (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ),
l odd.
When l is even, using (4.20), (4.21) and (4.32), we obtain g
wl (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) g = wl+1 (u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 )
(4.35)
86
TAMARA GRAVA
=
g−1 k=1
+
g−1 qk
k
(2n −
n=1
qgg (u)|{ul =ul+1 =v}
Q+ (v) g−1 Q− (v) σn−1 (ξ ) g−1 1)+˜ k−n Q+ (v) g−1 + (ξ ) Q− (v) σ0 g
(2k −
k=1
Q+ (v) g−1 Q− (v) σk−1 (ξ ) g−1 1)+˜ g−k Q+ (v) g−1 . (ξ ) Q− (v) σ0
(4.36)
We conclude from (4.35) and (4.36) that the boundary condition (3.6) is satisfied. When g = 1 we deduce from (4.35) w11 (u1 , u1 , u3 ) = w21 (u1 , u1 , u3 ) and from (4.9) and (4.26) w31 (u1 , u1 , u3 ) = 2(u3 − u1 )∂u3 q1 (u1 , u1 , u3 ) + q1 (u1 , u1 , u3 ). From (3.17) and (3.20) we get the relation q1 (u1 , u1 , u3 ) = f (u3 ) + 2(u1 − u3 )∂u3 q1 (u1 , u1 , u3 ) so that w31 (u1 , u1 , u3 ) = f (u3 ). An analogous result can be obtain when u2 = u3 , so that the boundary conditions (3.8)–(3.9) are satisfied. The theorem is then proved. ✷ Remark 4.3. Expression (3.15) obtained by Tian [15] for the polynomial initial data (3.12) is equivalent to (4.9) for any smooth monotonically increasing initial data. Indeed, (3.15) can be reduced to (4.9) using (4.14) and the identity γ j 0 0 ··· 0 +0 1 γ2j +1 + 0 · · · 0 0 ∂ ··· ··· ··· ··· ··· ··· j ∂ui + · · · +0 0 γg−1 g−2 +g−3 j +g−1 +g−2 · · · +1 +0 γg σ0 (r)φj (r) Resr=ui [ dr ] σ (r)φ (r) 3 Resr=ui [ 1 dr j ] 1 , (4.37) = ··· σg−2 (r)φj (r) 4 ] (2g − 3) Resr=ui [ dr σ (r)φ (r) (2g − 1) Resr=ui [ g−1 dr j ] where the σk (r)’s are the Abelian differentials of the second kind defined in (2.4), φj (r) is the j th holomorphic differential defined in (2.2) and Resr=ui is the residue evaluated at r = ui . The +k ’s are the coefficients of the expansion defined in (2.16). The above identity can be obtained from Proposition 5.4 (see below). Formula (3.15) is not effective for proving Theorem 4.2.
87
THE TSAREV AND WHITHAM EQUATIONS
5. C 1 -Smoothness of the Solution of the Hodograph Transformation In the following, we show that the g-phase solutions of the Whitham equations u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t) for different g 0 describe a C 1 -smooth multivalued curve in the x–u plane evolving smoothly in time. The matching conditions (3.6)–(3.9) guarantee that the C 1 -smoothness is preserved on the phase transition boundary where two Riemann invariants coalesce. THEOREM 5.1 ([8]). Let us suppose that the g-phase solution defined by (3.5) exists for some x and t > 0. Then on the solution of (3.5) (∂x ui (x, t))−1 = ∂ui (−λi (u)t + wi (u)),
i = 1, . . . , 2g + 1.
Proof. Deriving with respect to x Equation (3.5), we obtain ∂uj (−λi (u)t + wi (u))∂x uj . 1 = ∂ui (−λi (u)t + wi (u))∂x ui +
(5.38)
(5.39)
j =i
By (1.4) we have ∂uj (−λi (u)t + wi (u)) = −∂uj λi (u)t + ∂uj wi (u), i = j, i, j = 1, . . . , 2g + 1 ∂λi 1 (−λi (u)t + wi (u)) − (−λj (u)t + wj (u)) = λi − λj ∂uj = 0.
(5.40)
Thus (5.39) becomes 1 = ∂ui (−λi (u)t + wi (u))∂x ui ,
i = 1, . . . , 2g + 1.
(5.41) ✷
In the next theorem, we evaluate explicitly ∂ui (−λi (u)t + wi (u)), i = 1, . . . , 2g + 1. THEOREM 5.2. On the solution of the hodograph transformation (3.5), the following relation is satisfied: 2g+1 m=1 (ui − um ) ∂ m=i g (ui ; u), (−λi (u)t + wi (u)) = g ∂ui P0 (ui ) i = 1, . . . , 2g + 1, g > 0, (5.42) where
2g+1
g (r; u) = ∂r 5 g (r; u) +
k=1
∂uk 5 g (r; u)
(5.43)
88
TAMARA GRAVA
and 5 g (r; u) satisfies the linear overdetermined system ∂ g ∂ ∂2 5 (r; u) − 5 g (r; u) = 2(ui − uj ) 5 g (r; u), ∂ui ∂uj ∂ui ∂uj i, j = 1, . . . , 2g + 1, ∂ ∂2 ∂ g 5 (r; u) − 2 5 g (r; u) = 2(r − uj ) 5 g (r; u), ∂r ∂uj ∂r∂uj
i = j,
(5.44)
j = 1, . . . , 2g + 1,
2g f (g)(r). 5 g (r; r, . . . , r ) = (2g + 1)!! 2g+1
Remark 5.3. For polynomial initial data, the solution of the above initial value problem coincides with the function 5 g (r; u) defined in (4.2). The integration procedure of (5.44) is analogous to the one illustrated in the proof of Theorem 3.17. The function 5 g (r; u) that solves (5.44) satisifes the relations (4.7). We observe that the determinant of the Jacobian of the hodograph transformation can be obtained easily from (5.42), namely 2g+1 2g+1 m=1 (ui − um ) m=i det(Jacobian) = . (5.45) g (ui ; u) g P (u ) i 0 i=1 Therefore, the above determinant is nondegenerate if g (ui ; u) = 0, i = 1, . . . , 2g + 1. Proof of Theorem 5.2. From a generalization of a result in [4], we obtain g g 1 ∂ Pk (r) ∂ Pk (ui ) = , (5.46) ∂ui P0g (ui ) 2 ∂r P0g (r) r=ui so that ∂ (−λi (u)t + wi (u)) ∂ui 2g+1 2g+1 g k=1 m=1 (ui − uj ) ∂ P1 (r) k=i m=k,i ∂ui qg (u)+ + 2 = −6t g ∂r P0g (r) r=ui P0 (ui ) 2g+1 2g+1 m=1 (ui − um ) m=1 (ui − um ) m=i m=i g 2 (∂ui ) qg (u) − 2 ∂ui qg (u)∂ui P0 (ui )+ +2 g g 2 P0 (ui ) (P0 (ui )) g g g ∂ Pn−1 (r) 1 (2n − 1) qm (u)+˜ m−n + + 2 ∂r P g (r) n=1
(2n − g
+
n=1
r=ui m=n 0 g g P (ui ) 1) n−1 ∂u (qm (u)+˜ m−n ). g P0 (ui ) m=n i
(5.47)
89
THE TSAREV AND WHITHAM EQUATIONS
In the above relation we need to compute the following derivatives: g−m ∂ui (qm (u)+˜ m−n ) = ui +˜ m−n ∂ui qg (u),
(5.48)
which are obtained from (2.18) and (4.14) and g
g
g
∂ui P0 (ui ) = ∂r P0 (r)|r=ui + ∂ui P0 (r)|r=ui , g
where P0 (r) = r g + α10 r g−1 + · · · + αg0 . In order to obtain the derivative of the normalization constants α10 , α20 , . . . , αg0 , we need the following proposition. PROPOSITION 5.4 ([19]).√Let be ω1 (r) and ω2 (r) two normalized Abelian differentials on Sg . Let ξ = 1/ r be the local coordinate at infinity and ak1 ξ k dξ, ω2 = ak2 ξ k dξ. ω1 = k
k
Define the bilinear product Vω1 ω2 =
1 a−k−2 ak2 k0
k+1
,
then ω1 (r)ω2 (r) ∂ , Vω1 ω2 = Res r=ui ∂ui dr
i = 1, . . . , 2g + 1,
(5.49)
where Resr=ui (ω1 (r)ω2 (r))/dr is the residue of the differential (ω1 (r)ω2 (r))/dr evaluated at r = ui . Applying the above proposition to σ0 and σk , k = nontrivial simplifications we obtain 1 0 α0 0 0 1 0 ui 1 0 0 α2 1 ∂ 1 ··· ··· ··· = − ... − ··· ∂ui α 0 2 ug−1 2 ug−3 ug−4 · · · i i i g−1 g−1 g−2 g−3 αg0 ui ui ··· ui +˜ 0 0 +˜ 0 +˜ 1 g P (ui ) 1 ··· ··· + 2g+1 0 2 k=1,k=i (ui − uk ) +˜ ˜ + g−2 g−3 +˜ g−1 +˜ g−2 1 g 3P1 (ui ) ... × , g (2g − 3)P (u ) i g−2 g (2g − 1)Pg−1 (ui )
0, . . . , g − 1 and after · · · 0 α10 · · · 0 α20 ··· ··· ··· + 0 0 0 αg−1 1 0 αg0 0 ··· 0 0 ··· 0 ··· ··· ··· × · · · +˜ 0 0 · · · +˜ 1 +˜ 0
(5.50)
90
TAMARA GRAVA
where the +˜ k ’s have been defined in (2.18). From the above formula, we obtain g g g ∂ui P0 (ui ) = ∂r P0 (r)|r=ui + ∂ui P0 (r)|r=ui 1 g = ∂r P0 (r)|r=ui + 2 g g g P (ui ) 1 g g−m (2n − 1)Pn−1 (ui ) ui +˜ m−n . + 2g+1 0 2 k=1,k=i (ui − uk ) n=1 m=n (5.51) Using relations (5.48) and (5.51), we simplify (5.47) to the form ∂ (−λi (u)t + wi (u)) ∂ui 2g+1 ∂ui (2 m=1 (ui − um )∂ui qg (u)) m=i + = g P0 (ui ) g g g g P (r) P (r) 1 + (2n − 1) n−1 qm +˜ m−n + + ∂r −12t 1g g 2 P0 (r) n=1 P0 (r) m=n 2g+1
m=1 (ui − um ) m=i ∂ui qg (u) . +2 g P0 (r) r=ui
(5.52)
The above expression can be written in the form ∂ (−λi (u)t + wi (u)) ∂ui g g ∂ −xP0 (r) − 12tP1 (r) + R g (r) + = g ∂r 2P0 (r) r=ui 2g+1
2g+1 m=1 (ui − um ) ∂u qg (u) − ∂u qg (u) m=i i k 2(∂ui )2 qg + , + g u − u P0 (ui ) i k k=1
(5.53)
k=i
where the polynomial R g (r) has been defined in (4.4). We remark that such a polynomial has been defined for the analytic initial data (3.12), but it can be naturally extended to any smooth initial data. Applying the relations (3.17) and (4.7) to the last term of the above expression we obtain ∂u qg (u) − ∂u qg (u) i k u − u i k k=1
2g+1
2(∂ui ) qg + 2
k=i
2g+1
= 2(∂ui ) qg + 2 2
k=1 k=i
∂ui ∂uk qg
91
THE TSAREV AND WHITHAM EQUATIONS 2g+1 ∂uk 5 g (ui ; u) = (∂r + ∂ui )5 (r; u) r=ui + g
k=1 k=i
= g (ui ; u), where the function g (r; u) has been defined in (5.43). Substituting the above relation in (5.53), we obtain ∂ (−λi (u)t + wi (u)) ∂ui g g ∂ −xP0 (r) − 12tP1 (r) + R g (r) + = g ∂r 2P0 (r) r=ui 2g+1 m=1 (ui − um ) m=i + g (ui ; u). g P0 (ui )
(5.54)
We observe that g
g
−xP0 (r) − 12tP1 (r) + R g (r) ≡ 0,
g > 0,
on the solution of the Whitham equations. Indeed the hodograph transformation (3.5) is equivalent to imposing 2g + 1 zeros on the above polynomial which has degree 2g. Consequently, such a polynomial is identically zero. Therefore, we can simplify (5.54) to the form (5.42) when u1 > u2 > · · · > u2g+1 satisfy the g-phase Whitham equations, g > 0. ✷ From (5.42) it is clear that the ui ’s evolve in such a way that the graph of their curve is certainly C 1 -smooth in the x − u plane for u1 > u2 > · · · > u2g+1 . Indeed, g (r; u) is a smooth function of r and u when the initial data (5.44) is smooth. The next lemma shows that this property is preserved when two Riemann invariants coalesce. LEMMA 5.5. When ul = ul+1 the following relations are satisfied: ∂ui (−λi (u)t + wi (u))|{ul =ul+1 =v} g−1
= ∂ui (−λi
g−1
t + wi
),
i = l, l + 1,
(5.55)
where g−1
λi
g−1 wi
g−1
= λi =
(u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ) g−1 wi (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ).
Regarding ∂x ul and ∂x ul+1 , we have lim
ul →ul+1
and
! 1 1 − = 0. (5.56) ∂ul (−λl (u)t + wl (u)) ∂ul+1 (−λl+1 (u)t + wl+1 (u))
92
TAMARA GRAVA
Proof. We analyze the behavior of the derivatives ∂ui (−λi (u)t + wi (u)), i = 1, . . . , 2g + 1 when ul = ul+1 = v. We first prove (5.55). From (4.23) and (4.25) we have that, for ui = ul = ul+1 = v , g
g−1
Pk (ui ; u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 ) = (ui − v)Pk g ∂r Pk (r; u1 , . . . , ul−1 , v, v, ul+2 , . . . , u2g+1 )|r=ui g−1 g−1 = P (ui ) + (ui − v)∂r P (r) , k
k
(ui ),
k 0, (5.57)
r=ui
where g−1
Pk
g−1
(r) = Pk
(r; u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ),
k 0.
Applying (4.20), (4.21) and (5.57) to expression (5.52), we obtain g g ∂u (−λ (u)t + w (u)) {ul =v i
i
i
ul+1 =v}
! g−1 g−1 (r)|r=ui 1 g−1 ∂r P0 + + = − −λi t + wi g−1 2(ui − v) 2P0 (ui ) 2 k=i (ui − uk ) ∂ q g (u)|{ul =ul+1 =v} + + 2(ui − v) g−1 2 g ∂u P0 (ui ) i
4 k=i (ui − uk ) k=i m=k,i (ui − um ) + 2(ui − v) × + g−1 g−1 P0 (ui ) P0 (ui ) × ∂ui qgg (u){ul =u =v} + l+1 g g qg (u)|{ul =ul+1 =v} g−1 g−1 (2n − 1)+˜ g−n ∂r Pn−1 (r)|r=ui + + g−1 2P0 (ui ) n=1
g 1 g−1 g−1 (2n − 1)Pn−1 (ui )+˜ g−n + + (ui − v) n=1
g−1 g−1 1 g−1 g−1 g−1 ˜ g−1 ∂r (2n − 1)Pn−1 (r) qm +m−n − 12tP1 (r) + + g−1 2P0 (ui ) r=u i m=n n=1 g−1
g−1 1 g−1 g−1 g−1 (2n − 1)Pn−1 (ui ) qmg−1 +˜ m−n − 12tP1 (ui ) , + g−1 2(ui − v)P0 (ui ) n=1 m=n (5.58) where
2g+1
=
k=1 k=l,l+1
and
2g+1
=
m=1 m=l,l+1
.
93
THE TSAREV AND WHITHAM EQUATIONS
Using the relation (2.19), we obtain g
g−1 g−1 (2n − 1)+˜ g−n ∂r Pn−1 (r)r=ui = 2
k=i (ui
n=1
− uk )
(5.59)
(ui − uj )
j =i
and g
g−1 g−1 (2n − 1)+˜ g−n Pn−1 (ui ) =
n=1
k=i
(ui − uk ).
(5.60)
Substituting (5.59) and (5.60) in (5.58) and applying relation (4.22), we can easily obtain (5.55). For proving (5.56), we consider only l even. Analogous considerations can be done for l odd. Using expansion (4.25) and relation (2.13), we have √ √ g ∂r Pk (r; u1 , . . . , ul−1 , v + *, v − *, ul+2 , . . . , u2g+1 )) g−1
Pk + so that
g−1
(r) + (r − v)∂r Pk (r)+ g−1 g−1 ∂r ((r − v) j =1 Nj (v)r g−1−j ) log *
Q+ (v) Q− (v)
g−1
σk
(ξ ),
k 0, (5.61)
√ √ *, v − *, ul+2 , . . . , u2g+1 )|r=v±√* √ g P0 (v ± *) Q+ (v) g−1 g−1 g−1−j g−1 (ξ ) v σk (v) Q− (v) σk g−1 − Q+ (v) g−1 Nj (v)± − log * Q+ (v) g−1 µ(v) ˜ σ (ξ ) σ (ξ ) j =1 − − g
∂r Pk (r; u1 , . . . , ul−1 , v +
Q (v)
0
σ0 (v)σk (v)
Q (v)
0
√
* log2 *. ± Q+ (v) g−1 2 ( Q− (v) σ0 (ξ ))
(5.62)
Substituting (4.32), (4.36) and (5.62) in (5.56) and using the explicit expression (5.38) of the derivative, we obtain g g ∂u (−λ (u)t + w (u)) {ul =v+√* l
l
l
√ ul+1 =v− *}
g−1 √ σ0 (v) 1 2 × log * − * log * Q+ (v) g−1 2 (ξ ) Q− (v) σ0 Q+ (v) g−1 ! g−1 (ξ ) Z g−1 (v) σ0 (v) Q− (v) Z − g−1 × Q+ (v) g−1 Q+ (v) g−1 σ0 (v) σ (ξ ) σ (ξ ) − − 0 0 Q (v) Q (v)
and
g g ∂ul+1 (−λl+1 (u)t + wl+1 (u))
√ {ul =v+ * √ ul+1 =v− *}
(5.63)
94
TAMARA GRAVA
g−1 √ 1 σ0 (v) × log * + * log2 * Q+ (v) g−1 2 σ (ξ ) − Q (v) 0 Q+ (v) g−1 ! g−1 (ξ ) Z g−1 (v) σ0 (v) Q− (v) Z − g−1 , × Q+ (v) g−1 Q+ (v) g−1 σ0 (v) (ξ ) Q− (v) σ0 (ξ ) Q− (v) σ0
(5.64)
where g−1
Z g−1 (r) = −12tσ1
(r) +
g−1 g−1 g−1 g−1 (2n − 1)σn−1 (r) qmg−1 +˜ m−n + m=n
n=1
2g+1
+ 2µ(r) ˜ dr
∂uk qgg (u){ul =u
l+1 =v}
.
k=1
In the above formulas, Z g−1 (v) =
Z g−1 (r) , dr r=v
qmg−1 = qmg−1 (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 )
and +˜ mg−1 = +˜ mg−1 (u1 , . . . , ul−1 , ul+2 , . . . , u2g+1 ). Substituting (5.63) and (5.64) into (5.56), it is easy to verify that the limit (5.56) is satisfied. ✷ Combining Theorem 5.2 and Lemma 5.5, we conclude that the g-phase solutions u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t) for different g 0 are glued together in order to produce a C 1 -smooth multivalued curve in the x−u plane evolving smoothly with time. 6. Conclusion In this work, we have obtained a new formula for the solution wi (u), i = 1, . . . , 2g + 1, of the Tsarev system with matching conditions (3.6)–(3.9) for any increasing smooth initial data. As in [15], this new formula is expressed through some auxiliary functions which solve the linear overdetermined systems of the Euler–Poisson–Darboux type. We have shown that the matching conditions (3.6)– (3.9) for the wi (u)’s guarantee that the solution of the Tsarev system is unique and that the g-phase solutions of the Whitham equations for different g, are glued together in order to produce a C 1 -smooth multivalued curve in the (x, u) plane evolving smoothly with time. The new formula (4.9) enabled us to obtain a simple expression for the Jacobian of the hodograph transformation. We believe that this formula will be useful in the future to investigate the solvability of the hodograph transformation for g > 1. Indeed, it is still an open problem to prove the existence
THE TSAREV AND WHITHAM EQUATIONS
95
of a real solution u1 (x, t) > u2 (x, t) > · · · > u2g+1 (x, t), g > 1, of the hodograph transformation (1.5) for all x and t > 0 and for generic smooth initial data.
Acknowledgements I am indebted to Professor Boris Dubrovin who posed me the problem of this work and gave me many hints on how to reach the solution. I am grateful to Professor Sergei Novikov for his suggestions during the preparation of the manuscript. This work was partially supported by a CNR grant 203.01.70, by a grant of S. Novikov, and by a Marie Curie Fellowship. I wish to thank the anonymous referee for the suggested improvements to the manuscript. I am also grateful to T. Liverpool for carefully reading the manuscript.
References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
13. 14. 15. 16. 17.
Whitham, G. B.: Linear and Nonlinear Waves, Wiley, New York, 1974. Flaschka, H., Forest, M. and McLaughlin, D. H.: Multiphase averaging and the inverse spectral solution of the Korteweg–de Vries equations, Comm. Pure Appl. Math. 33 (1980), 739–784. Lax, P. D. and Levermore, C. D.: The small dispersion limit of the Korteweg–de Vries equation, I, II, III, Comm. Pure Appl. Math. 36 (1983), 253–290, 571–593, 809–830. Levermore, C. D.: The hyperbolic nature of the zero dispersion KdV limit, Comm. Partial Differential Equations 13 (1988), 495–514. Fei Ran Tian: Oscillations of the zero dispersion limit of the Korteweg–de Vries equations, Comm. Pure Appl. Math. 46 (1993), 1093–1129. Grava, T.: Existence of a global solution of the Whitham equations, Theoret. Math. Phys. 122(1) (2000), 46–58. Dubrovin, B. and Novikov, S. P.: Hydrodynamic of weakly deformed soliton lattices. Differential geometry and Hamiltonian theory, Russian Math. Surveys 44(6) (1989), 35–124. Tsarev, S. P.: Poisson brackets and one-dimensional Hamiltonian systems of hydrodynamic type, Soviet Math. Dokl. 31 (1985), 488–491. Gurevich, A. G. and Pitaevskii, L. P.: Nonstationary structure of collisionless shock waves, JEPT Lett. 17 (1973), 193–195. Krichever, I. M.: The method of averaging for two dimensional integrable equations, Funct. Anal. Appl. 22 (1988), 200–213. Potemin, G. V.: Algebraic-geometric construction of self-similar solutions of the Whitham equations, Uspekhi Mat. Nauk 43(5) (1988), 211–212. Kudashev, V. R. and Sharapov, S. E.: Inheritance of KdV symmetries under Whitham averaging and hydrodynamic symmetry of the Whitham equations, Theoret. Math. Phys. 87(1) (1991), 40–47. Gurevich, A. V., Krylov, A. L. and El, G. A.: Evolution of a Riemann wave in dispersive hydrodynamics, Soviet Phys. JEPT 74(6) (1992), 957–962. Eisenhart, L. P.: Ann. of Math. 120 (1918), 262. Fei Ran Tian: The Whitham type equations and linear over-determined systems of Euler– Poisson–Darboux type, Duke Math. J. 74 (1994), 203–221. El, G. A.: Generating function of the Whitham–KdV hierarchy and effective solution of the Cauchy problem, Phys. Lett. A 222 (1996), 393–399. Springer, G.: Introduction to Riemann Surfaces, Addison-Wesley, Reading, Mass., 1957.
96
TAMARA GRAVA
18.
Rodin, Yu. L.: The Riemann Boundary Value Problem on Riemann Surfaces, Math. Appl. Soviet Ser., D. Reidel, Dordrecht, 1987. Dubrovin, B.: Lectures on 2-D Topological Field Theory, Lecture Notes in Math. 1620, Springer-Verlag, Berlin, 1996. Fay, J.: Theta Functions on Riemann Surface, Lecture Notes in Math. 352, Springer-Verlag, Heidelberg, 1973.
19. 20.
Mathematical Physics, Analysis and Geometry 4: 97–130, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
97
A General Framework for Localization of Classical Waves: I. Inhomogeneous Media and Defect Eigenmodes ABEL KLEIN1 and ANDREW KOINES2, 1 University of California, Irvine, Department of Mathematics, Irvine, CA 92697-3875, U.S.A.
e-mail:
[email protected] 2 Texas A&M University, Department of Mathematics, College Station, TX 77843-3368, U.S.A. (Received: 7 February 2001) Abstract. We introduce a general framework for studying the localization of classical waves in inhomogeneous media, which encompasses acoustic waves with position dependent compressibility and mass density, elastic waves with position dependent Lamé moduli and mass density, and electromagnetic waves with position dependent magnetic permeability and dielectric constant. We also allow for anisotropy. We develop mathematical methods to study wave localization in inhomogeneous media. We show localization for local perturbations (defects) of media with a spectral gap, and study midgap eigenmodes. Mathematics Subject Classifications (2000): 35Q60, 35Q99, 78A99, 78A48, 74J99, 35P99, 47F05. Key words: wave localization, inhomogeneous media, defects, midgap eigenmodes.
1. Introduction We provide a general framework for studying localization of acoustic waves, elastic waves, and electromagnetic waves in inhomogeneous media, i.e., the existence of acoustic, elastic, and electromagnetic waves such that almost all of the wave’s energy remains in a fixed bounded region uniformly over time. Our general framework encompasses acoustic waves with position dependent compressibility and mass density, elastic waves with position dependent Lamé moduli and mass density, and electromagnetic waves with position dependent magnetic permeability and dielectric constant. We also allow for anisotropy. In this first article we develop mathematical methods to study wave localization in inhomogeneous media. As an application we show localization for local perturbations (defects) of media with a gap in the spectrum, and study midgap eigenmodes. In the second article [17], we use the methods developed in this article This work was partially supported by NSF Grant DMS-9800883. Current address: Orange Coast College, Department of Mathematics, Costa Mesa, CA 92626,
U.S.A.
98
ABEL KLEIN AND ANDREW KOINES
to study wave localization for random perturbations of periodic media with a gap in the spectrum. Our results extend the work of Figotin and Klein [7–11, 16] in several ways: (1) We study a general class of classical waves which includes acoustic, electromagnetic and elastic waves as special cases. (2) We allow for more than one inhomogeneous coefficient (e.g., electromagnetic waves in media where both the magnetic permeability and the dielectric constant are position dependent). (3) We allow for anisotropy in our wave equations. (4) In [17] we prove strong dynamical localization in random media, using the recent results of Germinet and Klein [14] on strong dynamical localization and of Klein, Koines and Seifert [18] on a generalized eigenfunction expansion for classical wave operators. Previous results on localization of classical waves in inhomogeneous media [7– 11, 16, 3] considered only the case of one inhomogeneous coefficient. Acoustic and electromagnetic waves were treated separately. Elastic waves were not discussed. Our approach to the mathematical study of localization of classical waves, as in the work of Figotin and Klein, is operator theoretic and reminiscent of quantum mechanics. It is based on the fact that many wave propagation phenomena in classical physics are governed by equations that can be recast in abstract Schrödinger form [21, 8, 16]. The corresponding self-adjoint operator, which governs the dynamics, is a first order partial differential operator, but its spectral theory may be studied through an auxiliary self-adjoint, second order partial differential operator. These second-order classical wave operators are analogous to Schrödinger operators in quantum mechanics. The method is particularly suitable for the study of phenomena historically associated with quantum mechanical electron waves, especially Anderson localization in random media [7, 8, 11, 16] and midgap defect eigenmodes [9, 10]. Physically interesting inhomogeneous media give rise to nonsmooth coefficients in the classical wave equations, and hence in their classical wave operators (e.g., a medium composed of two different homogeneous materials will be represented by piecewise constant coefficients). Thus we make no assumptions about the smoothness of the coefficients of classical wave operators. Since we allow two inhomogeneous coefficients, we have to deal with domain questions for the quadratic forms associated with classical wave operators. We must also take into account that many classical wave equations come with auxiliary conditions, and the corresponding classical wave operators are not elliptic (e.g., the Maxwell operator – see [21, 20, 8, 11, 16]). This paper is organized as follows: In Section 2 we introduce our framework for studying classical waves. We discuss classical wave equations in inhomogeneous media and wave localization. We define first and second order classical wave operators, and use them to rewrite the wave equations in abstract Schrödinger form. We state our results on wave localization created by defects. In Section 3 we study classical wave operators, obtaining the technical tools that are needed for proving localization in inhomogeneous and random media. We study finite volume classi-
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
99
cal wave operators, discuss interior estimates, give an improved resolvent decay estimate in a gap, prove a Simon–Lieb-type inequality and an eigenfunction decay inequality. In Section 4 we study periodic classical wave operators, and prove a theorem that gives the spectrum of a periodic classical wave operator in terms of the spectra of its restriction to finite cubes with periodic boundary condition. In Section 5 we study the effect of defects on classical wave operators, and give the proofs and details of the results on defects and wave localization stated in Section 2.
2. The Mathematical Framework 2.1. CLASSICAL WAVE EQUATIONS Many classical wave equations in a linear, lossless, inhomogeneous medium can be written as first order equations of the form: ∂ ψt (x) = D∗ φt (x), ∂t ∂ R(x)−1 φt (x) = −Dψt (x), ∂t K(x)−1
(2.1)
where x ∈ Rd (space), t ∈ R (time), ψt (x) ∈ Cn and φt (x) ∈ Cm are physical quantities that describe the state of the medium at position x and time t, D is an m×n matrix whose entries are first-order partial differential operators with constant coefficients (see Definition 2.1), D∗ is the formal adjoint of D, and, K(x) and R(x) are n × n and m × m positive, invertible matrices, uniformly bounded from above and away from 0, that describe the medium at position x (see Definition 2.3). In addition, D satisfies a partial ellipticity property (see Definition 2.2), and there may be auxiliary conditions to be satisfied by the quantities ψt (x) and φt (x). The physical quantities ψt (x) and φt (x) then satisfy second-order wave equations, with the same auxiliary conditions: ∂2 ψt (x) = −K(x)D∗ R(x)Dψt (x), ∂t 2 ∂2 φt (x) = −R(x)DK(x)D∗ φt (x). ∂t 2
(2.2) (2.3)
Conversely, given (2.2) (or (2.3)), we may write this equation in the form (2.1) by introducing an appropriate quantity φt (x) (or ψt (x)), which will then satisfy Equation (2.3) (or (2.2)). The medium is called homogeneous if the coefficient matrices K(x) and R(x) are constant, i.e., they do not depend on the position x. Otherwise the medium is said to be inhomogeneous.
100
ABEL KLEIN AND ANDREW KOINES
EXAMPLES. Electromagnetic waves: Maxwell equations are given by (2.1) with d = n = m = 3, ψt (x) the magnetic field, φt (x) the electric field, D = D∗ the curl, Dφ = ∇ × φ, K(x) =
1 I3 µ(x)
and
R(x) =
1 I3 , ε(x)
with µ(x) the magnetic permeability and ε(x) the dielectric constant. (By Ik we denote the k × k identity matrix.) The auxiliary conditions are ∇ · µψt = 0 and ∇ · εφt = 0. Acoustic waves: The acoustic equations in d dimensions may be written as (2.1), with n = 1, m = d, ψt (x) the pressure, φt (x) the velocity, D the gradient, Dφ = ∇φ, D∗ ψ = −∇ · ψ,
K(x) =
1 I1 κ(x)
and
R(x) =
1 Id , (x)
with κ(x) the compressibility and (x) the mass density. The auxiliary condition is ∇ × φt = 0. The usual second-order acoustic equation for the pressure is then given by (2.2). Elastic waves: The equations of motion for linear elasticity, in an isotropic medium, can be written as the second-order wave equation ρ(x)
∂2 ψt (x) = − ∇[λ(x) + 2µ(x)]∇ ∗ + ∇ × µ(x)∇× ψt (x), 2 ∂t
(2.4)
where x ∈ R3 , ψt (x) is the medium displacement, ρ(x) is the mass density, and, λ(x) and µ(x) are the Lamé moduli. It is of the form of Equation (2.2), with n = 3, D the differential operator given by Dψ = (∇ ∗ ψ) ⊕ (∇ × ψ)
(a 4 × 3 matrix),
K(x) =
1 I3 , ρ(x)
and R(x) = (λ(x) + 2µ(x))I1 ⊕ µ(x)I3
(a 4 × 4 matrix).
2.2. WAVE EQUATIONS IN ABSTRACT SCHRÖDINGER FORM The wave equation (2.1) may be rewritten in abstract Schrödinger form [21, 8, 16]: d t = Wt , dt ψ where t = φ t and −i
W=
(2.5)
t
0 iR(x)D
−iK(x)D∗ 0
.
(2.6)
101
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
The (first-order) classical wave operator W is formally (and can be defined as) a self-adjoint operator on the Hilbert space (2.7) H = L2 Rd , K(x)−1 dx; Cn ⊕ L2 Rd , R(x)−1 dx; Cm , where, for a k × k positive invertible matrix-valued measurable function S(x), we set L2 Rd , S(x)−1 dx; Ck
= f : Rd → Ck ; f, S(x)−1 f L2 (Rd ,dx;Ck ) < ∞ . The auxiliary conditions to the wave equation are imposed by requiring the solutions to Equation (2.5) to also satisfy ⊥ t , t = PW
(2.8)
⊥ denotes the orthogonal projection onto the orthogonal complement of where PW the kernel of W. The solutions to Equations (2.5) and (2.8) are of the form ⊥ 0 , t = eit W PW
0 ∈ H .
(2.9)
The energy density at time t of a solution ≡ t (x) = (ψt (x), φt (x)) of the wave equation (2.1) is given by E (t, x) = 12 ψ(x), K(x)−1 ψt (x)Cn + φt (x), R(x)−1 φt (x)Cm . (2.10) The wave energy, a conserved quantity, is thus given by E = 12 t 2H
for any t.
(2.11)
Note that (2.9) gives the finite energy solutions to the wave equation (2.1). 2.3. WAVE LOCALIZATION Let = t (x) be a finite energy solution of the wave equation (2.1). There are many criteria for wave localization, e.g.: Simple localization: Almost all of the wave’s energy remains in a fixed bounded region at all times, more precisely: 1 E (t, x) dx = 1. (2.12) lim inf R→∞ t E |x|R Moment localization: For some (or we may require for all) q > 0, we have q |x|q E (t, x) dx = 12 sup |x| 2 t 2H < ∞. (2.13) sup t
Rd
t
102
ABEL KLEIN AND ANDREW KOINES
Exponential localization (in the L2 -sense): For some C < ∞ and m > 0, we have sup χx E (t, ·) = √1 sup χx t H Ce−m|x| (2.14) t
2
2
t
for all x ∈ Rd , where χx denotes the characteristic function of a cube of side 1 centered at x. It is easy to see that exponential localization implies moment localization for all q > 0, and moment localization for some q > 0 implies simple localization. The fact that finite energy solutions are given by (2.9) suggests a method to obtain localized waves: if 0 ∈ H is an eigenfunction for the classical wave operator W with nonzero eigenvalue ω, i.e., W0 = ω0 with ω q= 0, then the wave t = eit ω 0 exhibits simple localization. If in addition |x| 2 0 2H < ∞, we have moment localization. If 0 is exponentially decaying (in the L2 -sense), we have exponential localization. In a homogenous medium, a classical wave operator cannot have nonzero eigenvalues. (This can be shown using the Fourier transform.) Thus an appropriate inhomogenous medium is required to produce nonzero eigenvalues, and hence localized waves. In Subsection 2.5 we will see that we can produce eigenvalues in spectral gaps of classical wave operators by introducing defects, i.e., by making local changes in the medium. Moreover, the corresponding waves will exhibit exponential localization. In the sequel [17] we show that random changes in the media can produce Anderson localization in spectral gaps of periodic classical wave operators. 2.4. CLASSICAL WAVE OPERATORS We now introduce the mathematical machinery needed to make the preceding discussion mathematically rigorous. It is convenient to work on L2 (Rd , dx; Ck ) instead of the weighted space 2 L (Rd , S(x)−1 dx; Ck ). To do so, note that the operator VS , given by multiplication by the matrix S(x)−1/2 , is a unitary map from the Hilbert space L2 (Rd , S(x)−1 dx;
= (VK ⊕ VR )W(V ∗ ⊕ V ∗ ), we have Ck ) to L2 (Rd , dx; Ck ), and if we set W K R √ √ ∗ 0√ −i K(x)D R(x)
= √ , (2.15) W 0 i R(x)D K(x) a formally self-adjoint operator on L2 (Rd , dx; Cn ) ⊕ L2 (Rd , dx; Cm ). In addition, if S− I S(x) S+ I with 0 < S− S+ < ∞, as it will be the case in this article, it turns out that if ϕ = VS ϕ, then the functions ϕ(x) and
ϕ (x) share the same decay and growth proporties (e.g., exponential or polynomial decay). Thus it will suffice for us to work on L2 (Rd , dx; Ck ), and we will do so in the remainder of this article. We set H (k) = L2 (Rd , dx; Ck ).
(2.16)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
103
Given a closed densely defined operator T on a Hilbert space H , we will denote its kernel by ker T and its range by ran T ; note ker T ∗ T = ker T . If T is selfadjoint, it leaves invariant the orthogonal complement of its kernel; the restriction of T to (ker T )⊥ will be denoted by T⊥ . Note that T⊥ is a self-adjoint operator on the Hilbert space (ker T )⊥ = PT⊥ H , where PT⊥ denotes the orthogonal projection onto (ker T )⊥ . DEFINITION 2.1. A constant coefficient, first-order, partial differential operator D from H (n) to H (m) (CPDO(1) n,m ) is of the form D = D(−i∇), where, for a d-component vector k, D(k) is the m × n matrix D(k)r,s = ar,s · k, ar,s ∈ Cd . (2.17) D(k) = D(k)r,s r=1,...,m ; s=1,...,n
We set D+ = sup{D(k); k ∈ Cd , |k| = 1},
(2.18)
so D(k) D+ |k| for all k ∈ Cd . Note that D+ is bounded by the norm of the matrix [|ar,s |] r=1,...,m . s=1,...,n Defined on D(D) = {ψ ∈ H (n) : Dψ ∈ H (m) in distributional sense},
(2.19)
∞ d n a CPDO(1) n,m D is a closed, densely defined operator, and C0 (R ; C ) (the space of infinitely differentiable functions with compact support) is an operator core for D. We will denote by D∗ the CPDO(1) m,n given by the formal adjoint of the matrix in (2.17).
DEFINITION 2.2. A CPDO(1) n,m D is said to be partially elliptic if there exists a (1) ⊥ CPDOn,q D (for some q), satisfying the following two properties: D⊥ D∗ = 0, D∗ D + (D⊥ )∗ D⊥ ,[(−-) ⊗ In ],
(2.20) (2.21)
with , > 0 being a constant. (- = ∇ · ∇ is the Laplacian on L2 (Rd , dx); In denotes the n × n identity matrix.) If D is partially elliptic, we have H (n) = ker D⊥ ⊕ ker D,
(2.22)
D∗ D + (D⊥ )∗ D⊥ = (D∗ D)⊥ ⊕ ((D⊥ )∗ D⊥ )⊥ .
(2.23)
and
Note that D is elliptic if and only it is partially elliptic with D⊥ = 0. Note also ∗ that a CPDO(1) n,m D may be partially elliptic with D not being partially elliptic [18, Remark 1.1].
104
ABEL KLEIN AND ANDREW KOINES
DEFINITION 2.3. A coefficient operator S on H (n) (COn ) is a bounded, invertible operator given by multiplication by a coefficient matrix: an n×n matrix -valued measurable function S(x) on Rd , satisfying S− In S(x) S+ In ,
with 0 < S− S+ < ∞.
(2.24)
DEFINITION 2.4. A multiplicative coefficient, first-order, partial differential operator from H (n) to H (m) (MPDO(1) n,m ) is of the form √ √ 1 (2.25) A = RD K on D(A) = K − 2 D(D), where D is a CPDO(1) n,m , K is a COn , and R is a COm . (We will write AK,R for A whenever it is necessary to make explicit the dependence on the medium, i.e., on the coefficient operators. D does not depend on the medium, so it will be omitted in the notation.) √ √ ∗ KD∗ R An MPDO(1) n,m A is a closed, densely defined operator with A = − 12 ∞ C0 (Rd ; Cn ) is an operator core for A. an MPDO(1) m,n . Note that K The following quantity will appear often in estimates: (2.26) /A ≡ D+ R+ K+ . DEFINITION 2.5. A first-order classical operator (CWO(1) n,m ) is an operator of the form 0 −iA∗ on H (n+m) ∼ (2.27) WA = = H (n) ⊕ H (m) , iA 0 ∗ where A is an MPDO(1) n,m . If either D or D is partially elliptic, WA will also be called partially elliptic. √ √ (1) SWD S, where A CWO(1) n,m is a self-adjoint MPDOn+m,n+m : WA = S = K ⊕ R is a COn+m and WD is a self-adjoint CPDO(1) n+m,n+m . (Note that our definition of a first-order classical wave operator is more restrictive than the one used in [18]. The definition of partial ellipticity is also different; [18] requires both D and D∗ to be partially eliptic.) The Schrödinger-like Equation (2.5) for classical waves with the auxiliary condition (2.8) may be written in the form:
−i
∂ t = (WA )⊥ t , ∂t
t ∈ (ker WA )⊥ = (ker A)⊥ ⊕ (ker A∗ )⊥ ,
(2.28)
with WA a CWO(1) n+m as in (2.27). Its solutions are of the form t = eit (WA )⊥ 0 ,
0 ∈ (ker WA )⊥ ,
which is just another way of writing (2.9).
(2.29)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
Since
(WA )2 =
0 A∗ A 0 AA∗
105
(2.30)
,
if t = (ψt , φt ) ∈ H (n) ⊕ H (m) is a solution of (2.28), then its components satisfy the second-order wave equations (2.2) and (2.3), plus the auxiliary conditions, which may be all written in the form ∂2 ψt = −(A∗ A)⊥ ψt , ∂t 2 ∂2 φt = −(AA∗ )⊥ φt , ∂t 2
with ψt ∈ (ker A)⊥ ,
(2.31)
with φt ∈ (ker A∗ )⊥ .
(2.32)
The solutions to (2.31) may be written as 1 1 ψt = cos t(A∗ A)⊥2 ψ0 + sin t(A∗ A)⊥2 η0 ,
ψ0 , η0 ∈ (ker A)⊥ ,
(2.33)
with a similar expression for the solutions of (2.32). The operators (A∗ A)⊥ and (AA∗ )⊥ are unitarily equivalent (see Lemma A.1): the operator U defined by −1
U ψ = A(A∗ A)⊥ 2 ψ
1
for ψ ∈ ran(A∗ A)⊥2 ,
(2.34)
extends to a unitary operator from (ker A)⊥ to (ker A∗ )⊥ , and (AA∗ )⊥ = U (A∗ A)⊥ U ∗ . In addition, if 1 U= √ 2
IA iU
IA −iU
(2.35)
,
with IA the identity on (ker A)⊥ ,
(2.36)
U is a unitary operator from (ker A)⊥ ⊕ (ker A)⊥ to (ker A)⊥ ⊕ (ker A∗ )⊥ , and we have the unitary equivalence: 1 1 (2.37) U∗ (WA )⊥ U = (A∗ A)⊥2 ⊕ −(A∗ A)⊥2 . Thus the operator (A∗ A)⊥ contains full information about the spectral theory of the operator (WA )⊥ (e.g., [8, 18]). In particular 1 1 (2.38) σ ((WA )⊥ ) = σ (A∗ A)⊥2 ∪ −σ (A∗ A)⊥2 , and to find all eigenvalues and eigenfunctions for (WA )⊥ , it is necessary and sufficient to find all eigenvalues and eigefunctions for (A∗ A)⊥ . For if (A∗ A)⊥ ψω2 = ω2 ψω2 , with ω = 0, ψω2 = 0, we have i i (2.39) (WA )⊥ ψω2 , ± Aψω2 = ±ω ψω2 , ± Aψω2 . ω ω
106
ABEL KLEIN AND ANDREW KOINES
Conversely, if (WA )⊥ (ψ±ω , φ±ω ) = ±ω(ψ±ω , φ±ω ), with ω = 0, it follows that (see [18, Proposition 5.2]) (A∗ A)⊥ ψ±ω = ω2 ψ±ω
and
i φ±ω = ± Aψ±ω . ω
(2.40)
DEFINITION 2.6. A second-order classical wave operator on H (n) (CWO(2) n ) is for some m. (We write W an operator W = A∗ A, with A an MPDO(1) K,R = n,m (2) ∗ AK,R AK,R .) If D in (2.25) is partially elliptic, the CWOn will also be called partially elliptic. Note that a first-order classical wave operator WA is partially elliptic if and only if one of the two second-order classical wave operators A∗ A and AA∗ is partially elliptic. DEFINITION 2.7. A classical wave operator (CWO) is either a CWO(1) n or a . If the operator W is a CWO, we call W a proper CWO. CWO(2) ⊥ n Remark 2.8. A proper classical wave operator W has a trivial kernel by construction, so 0 is not an eigenvalue. However, using a dilation argument, one can show that 0 is in the spectrum of W⊥ [18, Theorem A.1], so W⊥ and W have the same spectrum and essential spectrum.
2.5. DEFECTS AND WAVE LOCALIZATION We now describe our results on defects and wave localization. We can produce eigenvalues in spectral gaps of classical wave operators by introducing defects. Moreover, the corresponding waves exhibit exponential localization. The proofs and details are given in Section 5. A defect is a modification of a given medium in a bounded domain. Two media, described by coefficient matrices K0 (x), R0 (x) and K(x), R(x), are said to differ by a defect, if they are the same outside some bounded set 5, i.e., K0 (x) = K(x) / 5. The defect is said to be supported by the bounded and R0 (x) = R(x) if x ∈ set 5. We recall that the essential spectrum σess(H ) of an operator H consists of all the points of its spectrum, σ (H ), which are not isolated eigenvalues with finite multiplicity. Figotin and Klein [9] showed that the essential spectrum of Acoustic and Maxwell operators are not changed by defects. We extend this result to the class of classical wave operators: the essential spectrum of a partially elliptic classical wave operator (first or second order) is not changed by defects. THEOREM 2.9. Let W0 and W be partially elliptic classical wave operators for two media which differ by a defect. Then σess (W ) = σess (W0 ).
(2.41)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
107
If (a, b) is a gap in the spectrum of W0 , the spectrum of W in (a, b) consists of at most isolated eigenvalues with finite multiplicity, the corresponding eigenmodes decaying exponentially fast away from the defect, with a rate depending on the distance from the eigenvalue to the edges of the gap. In view of the unitary equivalence (2.37), Theorem 2.9 is an immediate corollary to Theorem 5.1 and Corollary 5.3. For second-order classical wave operators, the exponential decay of an eigenmode is given in (5.11). For first-order classical wave operators, the exponential decay of an eigenmode follows from (2.40), (5.11), and (3.19). We now turn to the existence of midgap eigenmodes and exponentially localized waves. The next theorem shows that one can design simple defects which generate eigenvalues in a specified subinterval of a spectral gap of W0 , extending [9, Theorem 2] to the class of classical wave operators. We insert a defect that changes the value of K0 (x) and R0 (x) inside a bounded set of ‘size’ 8 to given positive constants K and R. If (a, b) is a gap in the spectrum of W0 , we will show that we can deposit an eigenvalue of W √ inside any specified closed subinterval of (a, b), by inserting such a defect with 8/ KR large enough. We provide estimates on how large is ‘large enough’. Note that the corresponding eigenmode is exponentially decaying by Theorem 2.9, so we construct an exponentially localized wave. THEOREM 2.10 (Existence of exponentially localized waves). Let (a, b) be a gap in the spectrum of a partially elliptic classical wave operator W0 = WK0 ,R0 , select µ ∈ (a, b), and pick δ > 0 such that the interval [µ − δ, µ + δ] is contained in the gap, i.e., [µ − δ, µ + δ] ⊂ (a, b). Given an open bounded set 5, x0 ∈ 5, 0 < K, R, 8 < ∞, we introduce a defect that produces coefficient matrices K(x) and R(x) that are constant in the set 58 = x0 + 8(5 − x0 ), with K(x) = KIn
and
R(x) = RIm
for x ∈ 58 .
(2.42)
Then there is a finite constant C, satisfying an explicit lower bound depending only on the order ( first or second ) of the classical wave operator, and on D+ , µ, δ, and the geometry of 5, such that if √
8 KR
> C,
(2.43)
then the operator W = WK,R has at least one eigenvalue in the interval [µ − δ, µ + δ]. Theorem 2.10 follows from Theorem 5.4 and (2.37). For second-order operators the explicit lower bound is given in (5.16), for first-order operators it can be calculated from (5.16) and (2.37).
108
ABEL KLEIN AND ANDREW KOINES
3. Properties of Classical Wave Operators In this section we discuss several important properties of classical wave operators, which provide the necessary technical tools for proving localization in inhomogeneous and random (see [17]) media. 3.1. A TRACE ESTIMATE Partially elliptic second-order classical wave operators satisfy a trace estimate that provides a crucial ingredient for many results. THEOREM 3.1 ([18, Theorem 1.1]). Let W be a partially elliptic second-order classical wave operator on H (n) , and let PW⊥ denote the orthogonal projection onto (ker W )⊥ . Then tr(V ∗ PW⊥ (W + I )−2r V ) Cd,n,K± ,R± ,D+ ,D+⊥ ,, V 2∞,2 < ∞,
(3.1)
for r ν, where ν is the smallest integer satisfying ν > d/4. V is the bounded operator on H (n) given by multiplication by an n × n matrix-valued measurable function V (x), with χy,1 (x)V ∗ (x)V (x)∞ < ∞. (3.2) V 2∞,2 = y∈Zd
(χy,L denotes the characteristic function of a cube of side length L centered at y.) The constant Cd,n,K± ,R± ,D+ ,D+⊥ ,, depends only on the fixed parameters d, n, K± , R± , D+ , D+⊥ , ,. 3.2. FINITE VOLUME CLASSICAL WAVE OPERATORS Throughout this paper we use two norms in Rd and Cd : |x| =
d
1/2 |xi |2
,
(3.3)
i=1
x = max{|xi |, i = 1, . . . , d}.
(3.4)
We set Br (x) to be the open ball in Rd , centered at x with radius r > 0: Br (x) = {y ∈ Rd ; |y − x| < r}.
(3.5)
By =L (x) we denote the open cube in Rd , centered at x with side L > 0 =L (x) = {y ∈ Rd ; y − x < L/2},
(3.6)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
109
and by =L (x) the closed cube. By = we will always denote some open cube =L (x). We will identify a closed cube =L (x) with a torus in the usual way, and use the following distance in the torus: √ d L for y, y ∈ =L (x). (3.7) dL (y, y ) = min |y − y + m| d 2 m∈LZ We set H=(n) = L2 (=, dx; Cn ). A CPDO(1) n,m D defines a closed densely defined operator D= from H=(n) to H=(m) with periodic boundary condition; an operator core ∞ is given by Cper (=, Cn ), the infinitely differentiable, periodic Cn -valued functions on =. The restriction of a COn S to = gives the bounded, invertible operator S= on H=(n) . Given an MPDO(1) n,m A as in (2.25), we define its restriction A= to the cube = with periodic boundary condition by A= =
R= D= K=
−1
on D(A= ) = K= 2 D(D= ),
(3.8)
a closed, densely defined operator on H=(n) . The restriction W= of the second-order classical wave operator W = A∗ A to = with periodic boundary condition is now defined as W= = A∗= A= . If the CPDO(1) n,m D is partially elliptic, then the restriction D= is also partially elliptic, in the sense that Equations (2.20) and (2.21) hold for D= , (D⊥ )= , and -= . (-= is the Laplacian on L2 (=, dx) with periodic boundary condition.) This can be easily seen by using the Fourier transform; here the use of periodic boundary condition plays a crucial role. We also have (2.22) and (2.23) with H=(n) . (n) , Wx,L , and so on. If = = =L (x), we write Hx,L Given a second-order classical wave operator W on H (n) , we define its finite volume resolvent on a cube = by R= (z) = (W= − z)−1
for z ∈ / σ (W= ).
(3.9)
If W is partially elliptic, it turns out that (W= )⊥ has compact resolvent, i.e., / σ (W= ). Note that it suffices to prove the R= (z)PW⊥= is a compact operator for z ∈ statement for z = −1. We will prove a stronger statement. In what follows, we write F G if the positive self-adjoint operators F and G are unitarily equivalent, and we write F ! G if F J for some positive selfadjoint operator J G. Note that, if 0 F ! G, then tr f (G) tr f (F ), for any positive, decreasing function f on [0, ∞). PROPOSITION 3.2. Let W be a partially elliptic second-order classical wave operator. Then for any finite cube = and p > d/2 we have (3.10) tr (W= + 1)−p PW⊥= n tr (K− R− ,(−-= ) + 1)−p < ∞.
110
ABEL KLEIN AND ANDREW KOINES
Proof. Using Lemma A.1 and (2.24), we get (W= )⊥ = K= D∗= R= D= K= ⊥ R− K= D∗= D= K= ⊥ R− (D= K= D∗= )⊥ K− R− (D= D∗= )⊥ K− R− (D∗= D= )⊥ .
(3.11)
It follows from (3.11), (2.23), and (2.21) that tr (W= + 1)−p PW⊥= −p = tr (W= )⊥ + 1 tr (K− R− (D∗= D= )⊥ + 1)−p −p tr K− R− (D∗= D= )⊥ ⊕ (D⊥ )∗= (D⊥ )= ⊥ + 1 tr (K− R− ,[(−-= ) ⊗ In ] + 1)−p = n tr (K− R− ,(−-= ) + 1)−p < ∞,
(3.12) ✷
if p > d/2. Since (W= )⊥ 0 has compact resolvent, we may define NW= (E) = tr χ(−∞,E) ((W= )⊥ ),
(3.13)
the number of eigenvalues of (W= )⊥ that are less than E. If E 0, we have NW= (E) = 0, and if E > 0, NW= (E) is the number of eigenvalues of W= (or (W= )⊥ ) in the interval (0, E). Notice that NW= (E) is the distribution function of the measure nW= (dE) given by (3.14) h(E)nW= (dE) = tr(h((W= )⊥ )), for positive continuous functions h of a real variable. We have the following ‘a priori’ estimate: LEMMA 3.3. Let W be a partially elliptic second-order classical wave operator. Then for any finite cube = and E > 0 we have d2 E E nCd |=|, (3.15) NW= (E) nN−-= K− R− , K− R− , where Cd is some finite constant depending only on the dimension d. Proof. We have NW= (E) NK− R− (D∗= D= ) (E) NK− R− (D∗= D= +(D⊥ )∗= (D⊥ )= ) (E) NK− R− ,[(−-= )⊗In ] (E) = nN−-=
E , K− R− ,
(3.16) (3.17) (3.18)
111
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
where (3.16) follows from (3.11) and the Min-max Principle, (3.17) follows from (2.23), (3.18) follows from (2.21), plus a simple computation for the equality. The second inequality in (3.15) is given by a standard estimate. ✷
3.3. AN INTERIOR ESTIMATE The following interior estimate is an adaptation of [18, Theorem 4.1] to both finite or infinite volume. LEMMA 3.4. Let W = A∗ A be a second-order classical wave operator, and let = denote either an open cube or Rd . Let ρ ∈ C01 (=) and τ ∈ L∞ loc (=, dx), with 0 ρ(x) τ (x) and |∇ρ(x)| cτ (x) a.e., where c is a finite constant. Then, for any ψ ∈ D(W= ) we have 1 2 2 2 2 ρA= ψ aτ W= ψ + + 4c /A τ ψ2 (3.19) a for all a > 0, where /A is given in (2.26). Proof. This is proved as [18, Theorem 4.1], keeping track of the constants.
✷
3.4. IMPROVED RESOLVENT DECAY ESTIMATES IN A GAP We adapt an argument of Barbaroux, Combes and Hislop [1] to second-order classical wave operators, obtaining an improvement on the rate of decay given by the usual Combes–Thomas argument (e.g., [7, Lemma 12], [8, Lemma 15]). Our proof, while based on [1, Lemma 3.1], is otherwise different from the proof for Schrödinger operators, as we use an argument based on quadratic forms avoiding the analytic continuation of the operators. This way we can accomodate the nonsmoothness of the coefficients of our classical wave operators. We will prove the decay estimate for both infinite and finite volumes (with periodic boundary condition). We start with infinite volume. Recall that Br (x) denotes the open ball of radius r centered at x. THEOREM 3.5. Let W = A∗ A be a second-order classical wave operator with a spectral gap (a, b). Then for any E ∈ (a, b) and 8, 8 > 0 we have
χB8 (x) R(E)χB8 (y) CE emE (8+8 ) e−mE |x−y| for all x, y ∈ Rd , with 1 1 (E − a)(b − E) mE = 4/A (a + b + 2)(b + 1) 4/A
(3.20)
(3.21)
112
ABEL KLEIN AND ANDREW KOINES
and
CE = max
a + b + 2 4(b + 1) , . E−a b−E
(3.22)
In addition, χB8 (x) AR(E)χB8 (y) 1 CE 2E + 16/2A 2 emE (8+8 +1) e−mE |x−y|
(3.23)
for all x, y ∈ Rd with |x − y| 8 + 8 + 1. Proof. We start by defining the operators formally given by Wα = eα·x W e−α·x ,
α ∈ Rd .
To do so, let us consider the bounded operator √ √ Gα = RD(α) K, Gα |α|/A .
(3.24)
(3.25)
Then Aα = eα·x Ae−α·x = A + iGα on D(A), (A∗ )α = eα·x A∗ e−α·x = A∗ + iG∗α on D(A∗ ),
(3.26) (3.27)
are closed, densely defined operators. (Note (A∗ )α = (Aα )∗ .) We define Wα = (A∗ )α Aα as a quadratric form. More precisely, for each α ∈ Rd ,we define a quadratic form with domain D(A) by
(3.28) Wα [ψ] = (A∗ )∗α ψ, Aα ψ . Note that if α = 0, W = W0 is the closed, nonegative quadratic form associated to the classical wave operator W . It follows from (3.26) and (3.27) that Wα [ψ] − W [ψ] = 2i ReAψ, Gα ψ − Gα ψ, Gα ψ,
(3.29)
so 1 Wα [ψ] − W [ψ] Gα ψ 4Aψ2 + Gα ψ2 2 1 1 + s |α|2 /2A ψ2 2sW [ψ] + 2 s
(3.30)
for any s > 0. It follows [15, Theorem VI.1.33] that Wα is a closed sectorial form on the form domain of W . We define Wα as the unique m-sectorial operator associated with it [15, Theorem VI.2.1]. For each α ∈ Rd , we set W (α) = W − G∗α Gα , θ (α) = 1 + |α|2 /2A ,
(3.31) (3.32)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
113
note W (α) + θ (α) 1.
(3.33)
It follows from (3.29) that
− 1 − 1 W (α) + θ (α) 2 (Wa − E) W (α) + θ (α) 2 − 1 − 1 = W (α) + θ (α) 2 W (α) − E W (α) + θ (α) 2 + iYα ,
(3.34)
as everywhere defined quadratic forms, where (α) 1 1 W + θ (α) − 2 W (α) − E W (α) + θ (α) − 2 1 + θ (α) + E < ∞, (3.35) and − 1 − 1 Yα = W (α) + θ (α) 2 A∗ Gα + G∗α A W (α) + θ (α) 2
(3.36)
extends to a bounded self-adjoint operator with Yα 2|α|/A ,
(3.37)
in view of (3.25), (3.31), (3.32), (3.33), and (α) 1 A W + θ (α) − 2 ψ 2 (3.38) 1 1
(α) − − = W + θ (α) 2 ψ, A∗ A W (α) + θ (α) 2 ψ − 1 − 1
= ψ2 + W (α) + θ (α) 2 ψ, G∗α Gα − θ (α) W (α) + θ (α) 2 ψ − 1 2 (3.39) ψ2 − W (α) + θ (α) 2 ψ ψ2 . Since (a, b) is a gap in the spectrum of W and E ∈ (a, b), we have that the interval b − E − |α|2 /2A a−E , a + θ (α) b + θ (α) − |α|2 /2A b − E − |α|2 /2A a−E (3.40) , = b+1 a + 1 + |α|2 /2A is a gap in the spectrum of the operator
W (α) + θ (α)
− 12
− 1 W (α) − E W (α) + θ (α) 2 ,
containing 0, as long as √ b−E . |α| < /A
(3.41)
(3.42)
114
ABEL KLEIN AND ANDREW KOINES
We now use [1, Lemma 3.1] to conclude, if in addition to (3.42) we also require 1 (E − a)(b − E − |α|2 /2A ) |α| < , (3.43) 4/A (a + 1 + |α|2 /2A )(b + 1) that 0 is not in the spectrum of the operator in (3.34), and (α) 1 1 W + θ (α) − 2 (Wa − E) W (α) + θ (α) − 2 −1 b+1 a + 1 + |α|2 /2A , . 5α ≡ 2 max E−a b − E − |α|2 /2A
(3.44)
Since 1 D(Wα ) ⊂ D(Wα ) = D(W ) = D W (α) + θ (α) 2 ,
(3.45)
we may use (3.33) and (3.44) to obtain, for all φ ∈ D(Wα ), (Wa − E)φ − 1 − 1 1 W (α) + θ (α) 2 (Wa − E) W (α) + θ (α) 2 W (α) + θ (α) 2 φ (α) 1 W + θ (α) 2 φ 5−1 φ. 5−1 α α
(3.46)
Since (3.46) holds for all α ∈ Rd , and Wα∗ = W−α , W (α) = W (−α) , θ (α) = θ (−α) , and 5(α) = 5(−α) , we see that we also have (3.46) for Wα∗ . We can conclude that E∈ / σ (Wα ) and (Wα − E)−1 5α , (3.47) for all α ∈ Rd satisfying (3.42) and (3.43). We now take |α| mE , where mE is given √ in (3.21). Then both (3.42) and (3.43) are satisfied, and we also have |α|/A (b − E)/2, so 5α CE , with CE as in (3.22), and (3.47) gives Rα (E) CE ,
with Rα (E) = (Wα − E)−1 .
We may now prove (3.20). Let x0 , y0 ∈ Rd , 8 > 0, and take α = We have
(3.48) mE (x −y0 ). |x0 −y0 | 0
χB8 (x0 ) R(E)χB8 (y0 ) = χB8 (x0 ) e−α·x Rα (E)eα·x χB8 (y0 ) = e−m|x0 −y0 | χB8 (x0 ) e−α·(x−x0) Rα (E)eα·(x−y0) χB8 (y0 ) ,
(3.49)
so χB8 (x0 ) R(E)χB8 (y0 ) CE χB8 (x0 ) e−α·(x−x0 ) ∞ χB8 (y0 ) ea·(x−y0) ∞ e−m|x0 −y0 | .
(3.50)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
115
Since χB8 (x0 ) e±α·(x−x0) ∞ e|α|8 = emE 8 ,
(3.51)
(3.20) follows from (3.50) and (3.51). To prove (3.23), we use Lemma 3.4. We let |x0 − y0 | 8 + 8 + 1, and pick ρ ∈ C01 (Rd ), with χB8 (x0 ) ρ χB8+1 (x0 ) and |∇ρ(x)| 2. We have χB8 (x0 ) AR(E)χB8 (y0 ) ρAR(E)χB8 (y0 ) 1 2E + 16/2A 2 χB8+1 (x0 ) R(E)χB8 (y0 ) ,
(3.52) ✷
so (3.23) now follows from (3.20)
We now turn to the torus, i.e., we prove a version of Theorem 3.5 for the restriction of a second-order classical wave operator to a cube = with periodic boundary condition. We use the distance (3.7) in the torus. THEOREM 3.6. W be a second-order classical wave operator whose restriction with periodic boundary condition to a cube =L (x0 ) has a spectral gap (a, b). Then for any E ∈ (a, b) and 8 > 0, with L > 28 + 8, we have χB8 (x) Rx0 ,L (E)χB8 (y)x0 ,L CE e2mE,L,8 8 e−mE,L,8 dL (x,y) for all x, y ∈ =L (x0 ), where mE,L,8
mE = , cL,8
with cL,8 =
√ 2 d 1−
2(8+3) L−2
+1 ,
(3.53)
(3.54)
and mE and CE are as in Theorem 3.5. Proof. Let us fix x1 , y1 ∈ =L (x0 ), by redefining the coefficient operators we may assume x0 = 0 = 12 (x1 + y1 ) and x1 , y1 ∈ = L (0). In particular, dL (x1 , y1 ) = 2 |x1 − y1 |. Let L > 28 + 8, we pick a real valued function ξ ∈ C01 (R) with 0 ξ(t) 1 for all t ∈ R, such that ξ(t) = 1 for |t|
L 8 + , 4 2
ξ(t) = 0 for |t|
L − 1, 2
and
−1 L 8 − −2 for all t ∈ R. |ξ (t)| 4 2 We set (x) = di=1 ξ(xi ) for x ∈ Rd . Notice supp (x) ⊂ =L (0).
116
ABEL KLEIN AND ANDREW KOINES
We now proceed as in the proof of Theorem 3.5 with =L (0) substituted for Rd and definition (3.24) replaced by (W0,L )α = e(x)α·x W0,L e−(x)α·x ,
α ∈ Rd ,
and instead of (3.25), consider the bounded operators (G0,L )α = R0,L D ∇ (x)α · x 0,L K0,L . Since ∇ (x)α · x
√ ( L2 − 1) d L 4
−
8 2
−2
+ 1 |α| = cL,8 |α|
(3.55)
(3.56)
(3.57)
for all x ∈ =L (0), with cL,8 as in (3.54), we have (G0,L )α cL,8 |α|/A .
(3.58)
We now proceed as in the proof of Theorem 3.5, except that we must now substitute cL,8 |α| for |α| in the estimates. Thus, if |α| mE /cL,8 , we conclude that E∈ / σ ((W0,L)α ) and (R0,L)α (E) CE , with (R0,L )α (E) = (W0,L )α − E −1 . (3.59) To prove (3.53), we take α=
mE (x1 − y1 ), cL,8 |x1 − y1 |
and complete the proof of as before (with x1 , y1 substituted for x, y in (3.53)), as χB8 (x1 ) e±(x)α·x−α·x1 ∞ = χB8 (x1 ) e±α·(x−x1) ∞ emE,L,8 8 .
(3.60) ✷
3.5. A SIMON – LIEB - TYPE INEQUALITY The norm in H=(r) and also the corresponding operator norm will both be denoted by = , or x,L in case = = =L (x). (We omit r from the notation.) If =1 ⊂ =2 are open cubes (possibly the whole space), let J==12 : H=(r)1 → H=(r)2 be the canonical 2 injection. If =i = =Li (xi ), i = 1, 2, we write xx21 ,L ,L1 for the operator norm from =L (x2 )
,L2 = J=L 2(x1 ) . H=(r)L (x1 ) to H=(r)L (x2 ) , and Jxx12,L 1 1 2 1 Given a function φ ∈ L∞ (Rd ), with supp φ ⊂ =, we do not distinguish in the notation between φ as a multiplication operator on H=(r) and on H (r) . If D is a 1 d CPDO(1) n,m , and φ ∈ C0 (R ), real-valued, with supp φ ⊂ =1 ⊂ =2 , we can verify that, as operators,
D=2 J==12 φ = J==12 D=1 φ
on D(D=1 ).
(3.61)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
117
It follows for the MPDO(1) n,m A that A=2 J==12 φ = J==12 A=1 φ
on D(A=1 ).
(3.62)
We set A[φ] =
√
√ RD[φ] K,
(3.63)
where D[φ], given by multiplication by the matrix valued function D(−i∇φ(x)), is a bounded operator from H (n) to H (m) , with norm bounded by D+ ∇φ∞ . Thus A[φ] is a bounded operator given by multiplication by a matrix-valued, measurable function, with (see (2.26)) A[φ] /A ∇φ∞ .
(3.64)
We denote by A= [φ] its restriction to the cube =; it also satisfies the bound (3.64). We will use the fact that A= R= (z) is a bounded operator with (3.65) A= R= (z)2= R= (z)= |z|R= (z)= + 1 . The basic tool to relate the finite volume resolvents in different scales is the smooth resolvent identity (SRI) (see [2, 7, 8, 18]). LEMMA 3.7 (SRI). Let W = A∗ A be a second-order classical wave operator, and let =1 ⊂ =2 be either open cubes or Rd , and let φ ∈ C01 (Rd ) with supp φ ⊂ =1 . Then, for any z ∈ / σ (W=1 ) ∪ σ (W=2 ) we have R=2 (z)φJ==12 = J==12 φR=1 (z) + R=2 (z)A∗=2 [φ]J==12 A=1 R=1 (z) − − R=2 (z)A∗=2 J==12 A=1 [φ]R=1 (z) as bounded operators from H=(nL to H=(n)L . 1 2 Proof. This lemma can be proved as [18, Lemma 7.2].
(3.66)
✷
We will now state and prove a Simon–Lieb-type inequality (SLI) for secondorder classical wave operators. This estimate is a crucial ingredient in the multiscale analysis proofs of localization for random operators, where it is used to obtain decay in a larger scale from decay in a given scale [13, 12, 5, 2, 7, 8, 14]. Let us fix q ∈ N. (In [17] we will work with a periodic background medium, and we will take q to be the period.) We will take cubes =L (x) centered at sites x ∈ qZd with side L ∈ 2qN (so in a periodic background medium with period q the background medium will be the same in all cubes in a given scale L). For such
118
ABEL KLEIN AND ANDREW KOINES
cubes (with L 4q), we set L d ϒL (x) = y ∈ qZ ; y − x = − q , 2
L (x) = =L−q (x)\=L−3q (x) = =q (y), ϒ
(3.67) (3.68)
y∈ϒL (x)
L (x) = = 3q (x)\= 5q (x), ϒ L− 2 L− 2 χy,q a.e., Mx,L = χϒ L (x) =
(3.69) (3.70)
y∈ϒL (x)
Mx,L = χϒL (x).
(3.71)
Note that |ϒL (x)| d(L − 2q + 1)d−1 .
(3.72)
In addition each cube =L (x) will be equipped with a function φx,L constructed in the following way: we fix an even function ξ ∈ C01 (R) with 0 ξ(t) 1 for all t ∈ R, such that ξ(t) = 1 for |t| q/4 , ξ(t) = 0 for |t| 3q/4, and |ξ (t)| 3/q for all t ∈ R. (Such a function always exists.) We define 1, if |t| L2 − 5q , ξL (t) = (3.73) L 43q L 3q ξ |t| − 2 − 2 , if |t| 2 − 2 , and set φx,L (y) = φL (y − x)
for y ∈ Rd , with φL (y) =
d
ξL (yi ).
(3.74)
i=1
We have φx,L ∈ C01 (Rd ), with supp φx,L ⊂ =L (x) and 0 φx,L 1. By construction, we have χx, L − 5q φx,L = χx, L − 5q , 2
4
2
4
Mx,L (∇φx,L ) = ∇φx,L ,
χx, L − 3q φx,L = x,L , 2 4 √ 3 d . |∇φx,L | q
(3.75) (3.76)
Similarly, we also construct a function ρx,L ∈ C01 (Rd ), 0 ρx,L 1, such that Mx,L , Mx,L ρx,L = √ 5 d . |∇ρx,L | q
Mx,L ρx,L = ρx,L,
(3.77) (3.78)
LEMMA 3.8 (SLI). Let W be a second-order classical wave operator. Let x, y ∈ qZd , L, 8 ∈ 2qN, and 5 be a set, with 5 ⊂ =8−3q (y) ⊂ =L−3q (x). Then, if z∈ / σ (Wx,L) ∪ σ (Wy,8 ), we have Mx,L Rx,L(z)χ5 x,L γz My,8 Ry,8(z)χ5 y,8 Mx,L Rx,L (z)My,8x,L ,
(3.79)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
with
√ 1 6 d 100d 2 2 /A 2|z| + 2 /A , γz = q q
119
(3.80)
where /A is given in (2.26). Proof. We proceed as in [7, Lemma 26]. Using (3.75), (3.66), and Mx,L φy,8 = 0, we obtain x,L χ5 Mx,L Rx,L (z)Jy,8 x,L x,L = Mx,L Rx,L (z)Jy,8 φy,8 χ5 = Mx,L Rx,L(z)A∗x,L [φy,8 ]Jy,8 Ay,8 Ry,8 (z)χ5 − x,L Ay,8 [φy,8 ]Ry,8(z)χ5 . −Mx,L Rx,L (z)A∗x,LJy,8
(3.81)
We now use (3.76) and (3.64) to get x,L Ay,8 Ry,8(z)χ5 x,L Mx,L Rx,L(z)A∗x,L [φy,8 ]Jy,8 y,8 x,L = Mx,L Rx,L(z)My,8 A∗x,L [φy,8 ]Jy,8 My,8 Ay,8 Ry,8 (z)χ5 x,L y,8 √ 3 d /A Mx,L Rx,L (z)My,8x,L My,8 Ay,8 Ry,8 (z)χ5 y,8 , q
(3.82)
and x,L Ay,8 [φy,8 ]Ry,8 (z)χ5 x,L Mx,L Rx,L(z)A∗x,L Jy,8 x,L = Mx,L Rx,L(z)A∗x,L Ay,8 [φy,8 ]My,8 Ry,8 (z)χ5 x,L My,8 Jy,8 √ 3 d /A Mx,L Rx,L(z)A∗x,L My,8 x,L My,8 Ry,8(z)χ5 y,8 q √ 3 d /A = My,8 Ax,L Rx,L (¯z)Mx,L x,L My,8 Ry,8 (z)χ5 y,8 . q
(3.83)
(n) and a > 0 We now appeal to Lemma 3.4 using (3.77) and (3.78). For ψ ∈ Hy,8 we get
My,8 Ay,8 Ry,8 (z)χ5 ψ2y,8 ρy,8 Ay,8 Ry,8 (z)χ5 ψ2y,8 aMy,8 Wy,8 Ry,8 (z)χ5 ψ2y,8 + 1 100d 2 + 2 /A My,8 Ry,8(z)χ5 ψ2y,8 + a q 100d 2 1 2 a|z| + + 2 /A My,8 Ry,8(z)χ5 ψ2y,8 . a q
(3.84)
Choosing a = |z|−1 , we get My,8 Ay,8 Ry,8 (z)χ5 y,8 1 100d 2 2 2|z| + 2 /A My,8 Ry,8 (z)χ5 y,8 . q
(3.85)
120
ABEL KLEIN AND ANDREW KOINES
Similarly, we get My,8 Ax,L Rx,L (¯z)Mx,L x,L 1 100d 2 2 2|z| + 2 /A My,8 Rx,L (¯z)Mx,L x,L q 1 100d 2 2 = 2|z| + 2 /A Mx,L Rx,L (z)My,8x,L . q
(3.86)
Since x,L χ5 x,L Mx,L Rx,L(z)χ5 x,L = Mx,L Rx,L (z)Jy,8 y,8 ,
the lemma follows from (3.81)–(3.86).
(3.87) ✷
3.6. THE EIGENFUNCTION DECAY INEQUALITY The eigenfunction decay inequality (EDI) estimates decay for generalized eigenfunctions from decay of finite volume resolvents. We start by introducing generalized eigenfunctions for classical wave operators. (We refer to [18] for the details.) Given ν > d/4, we define the weighted spaces (we will omit ν from the notation) H±(r) as follows: H±(r) = L2 (Rd , (1 + |x|2 )±2ν dx; Cr ). H−(r) is the space of polynomially L2 -bounded functions. The sesquilinear form (3.88) φ1 , φ2 H (r) ,H (r) = φ1 (x) · φ2 (x) dx, +
−
where φ1 ∈ H+(r) and φ2 ∈ H−(r) , makes H+(r) and H−(r) conjugate duals to each other. By O † we will denote the adjoint of an operator O with respect to this duality. By construction, H+(r) ⊂ H (r) ⊂ H−(r) , the natural injections ı+ : H+(r) → H (r) and ı− : H (r) → H−(r) being continuous with dense range, with ı+† = ı− . Given a second-order classical wave operator W = A∗ A, where A is a MPDO(1) n,m , (n) we define operators W± on H as follows: A+ is the restriction of the operator A to H+(n) , i.e., A+ is the operator from H+(n) to H+(m) with domain D(A+ ) = {φ ∈ D(A) ∩ H+(n) ; Aφ ∈ H+(m) }, defined by A+ φ = Aφ for φ ∈ D(A+ ). A+ is a closed densely defined operator, and we set A− = (A∗+ )† , a closed densely defined operator from H−(n) to H−(m) . We define W+ = A∗+ A+ = A†− A+ , which is a closed densely defined operator on H+(n) with domain D(W+ ) = {φ ∈ D(W ) ∩ H+(n) ; W φ ∈ H+(n) }, and W+ φ = W φ for φ ∈ D(W+ ) [18, Theorem 4.2]. We define W− = W+† , a closed densely defined operator on H−(n) . Note that W is the restriction of W− to H (n) .
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
121
A measurable function ψ: Rd → Cn is said to be a generalized eigenfunction of W with generalized eigenvalue λ, if ψ ∈ H−(n) (for some ν > d/4) and is an eigenfunction for W− with eigenvalue λ, i.e., ψ ∈ D(W− ) and W− ψ = λψ. In other words, ψ ∈ H−(n) and W+ φ, ψH (n) ,H (n) = λφ, ψH (n) ,H (n) +
−
+
−
for all φ ∈ D(W+ ).
(3.89)
Eigenfunctions of W are always generalized eigenfunctions. Conversely, if a generalized eigenfunction is in H (n) , then it is a bona fide eigenfunction. LEMMA 3.9 (EDI). Let W be a second-order classical wave operator, and let ψ be a generalized eigenfunction of W with generalized eigenvalue E. Let x ∈ qZd and L ∈ 2qN be such that E ∈ / σ (Wx,L ). Then for any set 5 ⊂ =L−3q (x) we have χ5 ψ γE Mx,L Rx,L(E)χ5 x,L Mx,L ψ,
(3.90)
with γE as in (3.80). Proof. We fix ν > d/4 such that ψ ∈ D(W− ) and W− ψ = Eψ. We write Rd . Jx,L = Jx,L (n) , Using [18, Lemma 4.1], we can show that, weakly in Hx,L ∗ ∗ ∗ χ5 ψ = χ5 Jx,L φx,L ψ = χ5 Rx,L (E)(Wx,L − E)Jx,L φx,L ψ Jx,L ∗ ∗ = χ5 Rx,L (E)Ax,L Jx,L A[φx,L ]ψ + ∗ A∗ [φx,L ]A− ψ. + χ5 Rx,L (E)Jx,L
Proceeding as in the proof of Lemma 3.8, we have χ5 Rx,L (E)A∗ J ∗ A[φx,L ]ψ x,L x,L x,L ∗ ∗ = χ5 Rx,L (E)Ax,L Mx,L Jx,L A[φx,L ]Mx,L ψ x,L √ 3 d ∗ /A χ5 Rx,L(E)A∗x,L Mx,L ψx,L Mx,L x,L Jx,L q √ 3 d ∗ /A = Mx,L Ax,L Rx,L(E)χ5 x,L Jx,L Mx,L ψx,L q √ 1 3 d 100d 2 2 /A 2|E| + 2 /A × q q ×Mx,L Rx,L(E)χ5 x,L Mx,L ψ. Similarly, χ5 Rx,L (E)J ∗ A∗ [φx,L ]A− ψ x,L x,L ∗ ∗ = χ5 Rx,L (E)Mx,L Jx,L A [φx,L ] Mx,L A− ψ x,L √ 3 d /A Mx,L Rx,L (E)χ5 x,L Mx,L A− ψ, q
(3.91)
(3.92)
(3.93)
122
ABEL KLEIN AND ANDREW KOINES
and, using Lemma 3.4, which is also valid for the operator A− (see [18, Theorem 4.1], we have Mx,L A− ψ2 ρx,L A− ψ2
1 100d 2 aMx,L W− ψ2 + + 2 /A Mx,L ψ2 a q 1 100d = a|E|2 + + 2 /2A Mx,L ψ2 , a q
(3.94)
for any a > 0. Choosing a = |E|−1 in (3.94), (3.90) follows from (3.91)–(3.94).
✷
4. Periodic Classical Wave Operators In this section we study classical wave operators in periodic media. The main theorem gives the spectrum of a periodic classical wave operator in terms of the spectra of its restriction to finite cubes with periodic boundary condition. DEFINITION 4.1. A coefficient operator S on H (n) is periodic with period q > 0 if S(x) = S(x + qj ) for all x ∈ Rd and j ∈ Zd . DEFINITION 4.2. A medium is called periodic if the coefficient operators K and R that describe the medium are periodic with the same period q. (We will always take the period q ∈ N without loss of generality.) The corresponding classical wave operators will be said to be periodic with period q (q-periodic). If k, n ∈ N, we say that k ! n if n ∈ kN and that k ≺ n if k ! n and k = n. THEOREM 4.3. Let W be a q-periodic second-order classical wave operator. Let {8n ; n = 0, 1, 2, . . .} be a sequence in N such that 80 = q and 8n ≺ 8n+1 for each n = 0, 1, 2, . . .. Then σ (W0,8n ) ⊂ σ (W0,8n+1 ) ⊂ σ (W )
for all n = 0, 1, 2, . . . ,
(4.1)
and σ (W ) =
∞
σ (W0,8n ).
(4.2)
n=0
Proof. The analogous result for periodic Schrödinger operators is well known [6]. Periodic acoustic and Maxwell operators are treated in [7, Theorem 14] and [8, Theorem 25], respectively. We will sketch a proof, using Floquet theory. We refer to [19, Section XIII.6] for the definitions and notations of direct integrals of Hilbert spaces.
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
123
˘ 2π (0) the dual basic cell, We let Q = =q (0) be the basic period cell, Q˜ = = q L L d ˘ L (x) = y ∈ R ; xi − yi < xi + , i = 1, . . . , d . = 2 2 ˘ q (0), but we will not since it will make no difference (We should also take Q = = in what follows.) For any r ∈ N we define the Floquet transform ⊕ (r) ˜ dk; HQ(r) HQ(r) dk ≡ L2 Q, (4.3) F :H → Q˜
by (F ψ)(k, x) =
q 2π
d2
eik·(x−m) ψ(x − m),
˜ x ∈ Q, k ∈ Q,
(4.4)
m∈qZd
if ψ has compact support; it extends by continuity to a unitary operator. The q-periodic operator W is decomposable in this direct integral representation, more precisely, ⊕ ∗ WQ (k) dk, (4.5) F WF = Q˜
where for each k ∈ Rd we set DQ (k) to be the restriction to Q with periodic boundary condition of the operator given by the matrix D(−i∇ + k) (see (2.17) ), a closed, densely defined operator, and let WQ (k) = A∗Q (k)AQ (k) with AQ (k) = RQ DQ (k) KQ . (If for p ∈ 2π /qZd , Up denotes the unitary operator on HQ(r) given by multiplication by the function e−ip·x , then for all k ∈ Rd we have WQ (k + p) = Up∗ WQ (k)Up .) Since AQ (k + h) − AQ (k) |h|/A ,
(4.6)
follows from the resolvent identity that the map k ∈ Rd (→ (WQ (k) + I )−1 ∈ B HQ(n)
(4.7)
is operator norm continuous, so we conclude from (4.5) that σ (WQ (k)). σ (W ) =
(4.8)
k∈Q˜
If 8 ∈ qZd , similar considerations apply to the operator W0,8 , which is q-periodic on the torus =8 (0). The Floquet transform (r) → HQ(r) (4.9) F8 : H0,8 d ˜ k∈ 2π 8 Z ∩Q
124
ABEL KLEIN AND ANDREW KOINES
is a unitary operator now defined by d2 q (F8 ψ)(k, x) = 8 d
eik·(x−m) ψ(x − m),
(4.10)
˘ 8 (0) m∈qZ ∩=
where x ∈ Q,
k∈
2π d ˜ Z ∩ Q, 8
(r) ψ ∈ H0,8 ,
ψ(x − m) being properly interpreted in the torus =8 (0). We also have WQ (k), F8 W0,8 F8∗ =
(4.11)
d ˜ k∈ 2π 8 Z ∩Q
and σ (W0,8 ) =
σ (WQ (k)).
(4.12)
d ˜ k∈ 2π 8 Z ∩Q
Theorem 4.3 follows from (4.8) and (4.12).
✷
5. Defects and Midgap Eigenmodes We now prove the results in Subsection 2.5. THEOREM 5.1 (Stability of essential spectrum). Let W0 and W be second-order partially elliptic classical wave operators for two media which differ by a defect. Then σess (W ) = σess (W0 ).
(5.1)
Proof. We will first prove the theorem when the defect only changes R, i.e., we will show σess (WK0 ,R ) = σess (WK0 ,R0 ). The general case will follow, using Remark 2.8 and Lemma A.1, as then σess (WK0 ,R0 ) = σess (WK0 ,R ) = σess (WK0 ,R )⊥ = σess (WR,K0 )⊥ = σess (WR,K0 ) = σess (WR,K ) = σess (WR,K )⊥ = σess (WK,R )⊥ = σess(WK,R ).
(5.2)
(5.3)
To prove (5.2), we proceed as in [9, Theorem 1]. Let T (x) = R(x) − R0 (x), by our hypotheses it is a bounded, measurable, self-adjoint matrix-valued function with compact support. We write T (x) = T+ (x) − T− (x),
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
125
with T± (x) the positive/negative part of the self-adjoint matrix T (x). We let T± denote the bounded operators given by the matrices T± (x), they would be coefficients operators except for the fact that the functions T± (x) have compact support, so they are not bounded away from zero. We may still define operators define nonnegative self-adjoint operators WK0 ,T± . We have WK0 ,R = (WK0 ,R0 + WK0 ,T+ ) − WK0 ,T− ,
(5.4)
as quadratic forms. (Note that Q(WK0 ,R ) = Q(WK0 ,R0 ) ⊂ Q(WK0 ,T± ), where Q(W ) denotes the form domain of the operator W .) Thus (5.2) follows from [19, Corollary 4 to Theorem XIII.14] and the following lemma. LEMMA 5.2. Let WK,R be a second-order partially elliptic classical wave operator, and let T be like a coefficient operator, except for the fact that the function T (x) has compact support, so it is not bounded away from zero (i.e., T− = 0). Then (5.5) tr (WK,R + I )−r WK,T (WK,R + I )−r < ∞ if r ν + 1, where ν is the smallest integer satisfying ν > d/4. Proof. Let 5 denote the support of T (x), we pick a function ρ ∈ C01 (Rd ) with
= supp ρ is a compact set. We have χ5 ρ(x) χ5
, where 5 T T+ ρ 2 T+ R−−1 ρ 2 R.
(5.6)
Thus, using HS to denote the Hilbert–Schmidt norm, and setting c = ∇ρ∞ , we have (5.7) tr (WK,R + I )−r WK,T (WK,R + I )−r T+ R−−1 tr (WK,R + I )−r A∗K,R ρ 2 AK,R (WK,R + I )−r 2 (5.8) = T+ R−−1 ρAK,R (WK,R + I )−r HS −1 −r 2 + (5.9) T+ R− χ5
WK,R (WK,R + I ) HS −r 2 + 1 + 4c2 /2AK,R χ5
(WK,R + I ) HS 2 −r+1 −r χ5 + (W + I ) + T+ R−−1 χ5
(WK,R + I )
K,R HS HS 2 2 2 −r (5.10) + 1 + 4c /AK,R χ5
(WK,R + I ) HS < ∞, where the final bound in (5.10) follows from Theorem 3.1 if r − 1 ν. To go from (5.8) to (5.9) we used Lemma 3.4. ✷ This finishes the proof of Theorem 5.1.
✷
COROLLARY 5.3 (Behavior of midgap eigenmodes). Let W0 and W be secondorder partially elliptic classical wave operators for two media which differ by
126
ABEL KLEIN AND ANDREW KOINES
a defect. If (a, b) is a gap in the spectrum of W0 , the spectrum of W in (a, b) consists of at most isolated eigenvalues with finite multiplicity, the corresponding eigenmodes decaying exponentially fast from the defect, with a rate depending on the distance from the eigenvalue to the edges of the gap. If the defect is supported by some ball Br (x0 ), and E ∈ (a, b) is an eigenvalue for W with a corresponding eigenmode ψ, ψ = 1, then χx ψ2
√ 1 1 d 2CE /A0 E 2 + emE 2E + 16 /2A0 2 emE ( 2 +r+2) e−|x−x0 |
(5.11)
for all x ∈ Rd such that √ d + r + 3, |x − x0 | 2 where mE and CE are as in Theorem 3.5. Proof. By Theorem 5.1 W has no essential spectrum in (a, b). Thus, if E ∈ σ (W ) ∩ (a, b), it must be an isolated eigenvalue with finite multiplicity; let ψ be a corresponding eigenvector. To estimate the decay of ψ we have to deal with the fact that the form domains of W and W0 may be different, and ψ may not be in the form domain of W0 . Thus we pick ρ ∈ C 1 (Rd ) such that 1 − χBr+2 (x0 ) (x) ρ(x) 1 − χBr+1 (x0 ) (x),
|∇ρ(x)| 2.
(5.12)
Since W and W0 differ by a defect supported by Br (x0 ), it follows from (2.25) that Dρ ≡ ρD(A) = ρD(A0 ), and Aϕ = A0 ϕ for ϕ ∈ Dρ . Thus, if φ ∈ D(A0 ), we have A0 φ, A0 ρψ = = = = = = =
A0 φ, Aρψ = A0 φ, ρAψ + A0 φ, A[ρ]ψ ρA0 φ, Aψ + A0 φ, A[ρ]ψ ρA0 φ, Aψ + A0 φ, A[ρ]ψ A0 ρφ, Aψ − A0 [ρ]φ, Aψ + A0 φ, A[ρ]ψ Aρφ, Aψ − A0 [ρ]φ, Aψ + A0 φ, A[ρ]ψ ρφ, W ψ − A0 [ρ]φ, Aψ + A0 φ, A0 [ρ]ψ Eρφ, ψ − A0 [ρ]φ, Aψ + A0 φ, A0 [ρ]ψ.
(5.13)
−1
Taking φ = (W0 − E) χx ψ, we get
χx ψ2 = − A0 [ρ](W0 − E)−1 χx ψ, Aψ +
+ A0 (W0 − E)−1 χx ψ, A0 [ρ]ψ
= − A0 [ρ]χBr+2 (x0 ) (W0 − E)−1 χx ψ, Aψ +
+ χBr+2 (x0 ) A0 (W0 − E)−1 χx ψ, A0 [ρ]ψ √ 2/A0 EχBr+2 (x0 ) (W0 − E)−1 χx + + χBr+2 (x0 ) A0 (W0 − E)−1 χx ψ2 , where we used (3.64) and Aψ = ψ, W ψ = Eψ . 2
2
(5.14)
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
127
The estimate (5.11) now follows from (5.14), using (3.20) and (3.23) in Theorem 3.5. ✷ The next theorem shows that one can design simple defects which generate eigenvalues in a specified subinterval of a spectral gap of W0 , extending [9, Theorem 2] to the class of classical wave operators. Let 5 be a an open bounded subset of Rd , x0 ∈ 5. Typically, we take 5 to be the cube =1 (x0 ), or the ball B1 (x0 ). We set 58 = x0 + 8(5 − x0 ) for 8 > 0. We insert a defect that changes the value of K0 (x) and R0 (x) inside 58 to given positive constants K and R. If (a, b) is a gap in the spectrum of W0 , we will show that we can deposit an eigenvalue of W inside √ any specified closed subinterval of (a, b), by inserting such a defect with 8/ KR large enough, how large depending only on D+ , the geometry of 5, and the specified closed subinterval. THEOREM 5.4 (Creation of midgap eigenvalues). Let (a, b) be a gap in the spectrum of a second-order partially elliptic classical wave operator W0 = WK0 ,R0 , select µ ∈ (a, b), and pick δ > 0 such that the interval [µ − δ, µ + δ] is contained in the gap, i.e., [µ − δ, µ + δ] ⊂ (a, b). Given an open bounded set 5, x0 ∈ 5, 0 < K, R, 8 < ∞, we introduce a defect that produces coefficient matrices K(x) and R(x) that are constant in the set 58 = x0 + 8(5 − x0 ), with K(x) = KIn
and
R(x) = RIm
for x ∈ 58 .
(5.15)
If 12 µ δ 2 2 > D+ inf ∇η2 + ∇η2 + -η2 , √ δ µ KR 8
√
(5.16)
where the infimum is taken over all real valued C 2 -functions η on Rd with support in 5 and η2 = 1, the operator W = WK,R has at least one eigenvalue in the interval [µ − δ, µ + δ]. Proof. We proceed as in [9, Theorem 2]. In view of Corollary 5.3, it suffices to show that σ (W ) ∩ [µ − δ, µ + δ] = ∅
(5.17)
if (5.16) is satisfied. To prove (5.17), it suffices to find ϕ ∈ D(W ) such that (W − µ)ϕ δϕ. (5.18) To do so, we will construct a function ϕ ∈ D(W ), with ϕ = 1 and support in 58 , such that (5.18) holds. In this case the inequality (5.18) takes the following simple form: (KRD∗ D − µ)ϕ δ, (5.19)
128
ABEL KLEIN AND ANDREW KOINES
which is the same as ∗ (D D − µ )ϕ δ ,
(5.20)
with µ = µ/KR and δ = δ/KR. We start by constructing generalized eigenfunctions for the nonnegative operator D∗ D corresponding to µ . In order to do this, we consider κ ∈ Sd , pick an eigenvalue λ = λκ > 0 and a corresponding eigenvector ξ = ξκ,λ ∈ Cn , |ξ | = 1, of the n × n matrix D(κ)∗ D(κ) (see (2.17)). We set µ λ κ·x
f (x) = fκ,λ,ξ (x) = ei
ξ ∈ C ∞ (Rn ; Cn ).
(5.21)
Note that, pointwise, we have |f (x)| = 1, and (D∗ Df )(x) = µ f (x).
(5.22)
To produce the desired ϕ satisfying (5.18), we will restrict f to 58 in suitable manner, and prove (5.20). To do so, let η8 be a real valued C 2 function on Rd with support in 58 and η8 2 = 1. We set ϕ(x) = η8 (x)f (x),
note
ϕ = η8 2 = 1.
(5.23)
We have ϕ ∈ D(D∗ D) with support in 58 , and (D∗ D − µ )ϕ ! ∗ µ ∗ D (−i∇η8 )D(κ)f + = D (−i∇)D(−i∇η8 ) f + λ ! µ ∗ D (κ)D(−i∇η8 )f. + λ Thus
! ∗ (D D − µ )ϕ D 2 -η8 2 + 2 µ D 2 ∇η8 2 . + λ +
(5.24)
(5.25)
We now use a scaling argument (i.e., write η8 (x) = η(8−1 (x − x0 ) + x0 )) to conclude that to obtain (5.20), it suffices to find η ∈ C 2 (Rd , R) with support in 5, η2 = 1, and a unit vector κ ∈ Rd , such that ! −2 2 −1 µ D 2 ∇η2 δ , (5.26) 8 D+ -η2 + 28 λ + which will be satisfied if
√ √ 8−2 KRD+2 -η2 + 28−1 KRD+ µ∇η2 δ,
where we used the fact that λ D+2 . Thus (5.20) holds if (5.16) is satisfied.
(5.27) ✷
A GENERAL FRAMEWORK FOR LOCALIZATION OF CLASSICAL WAVES: I
129
Appendix: A Useful Lemma The following well known lemma (e.g., [4, Lemma 2]) is used throughout this paper. We recall that, given a closed densely defined operator T on a Hilbert space H, we denote its kernel by ker T and its range by ran T . If T is self-adjoint, it leaves invariant the orthogonal complement of its kernel; the restriction of T to (ker T )⊥ is denoted by T⊥ , a self-adjoint operator on the Hilbert space (ker T )⊥ . LEMMA A.1. Let B be a closed, densely defined operator from the Hilbert space H1 to the Hilbert space H2 . Then the operators (B ∗ B)⊥ and (BB ∗ )⊥ are unitarily equivalent. More precisely, the operator U defined by −1
U ψ = B(B ∗ B)⊥ 2 ψ
1
for ψ ∈ ran(B ∗ B)⊥2 ,
(A.1)
extends to a unitary operator from (ker B)⊥ to (ker B ∗ )⊥ , and (BB ∗ )⊥ = U (B ∗ B)⊥ U ∗ .
(A.2)
Acknowledgements The authors thanks Maximilian Seifert for many discussions and suggestions. A. Klein also thanks Alex Figotin, François Germinet, and Svetlana Jitomirskaya for enjoyable discussions. References 1. 2. 3.
4. 5. 6. 7. 8. 9. 10. 11.
Barbaroux, J. M., Combes, J. M. and Hislop, P. D.: Localization near band edges for random Schrödinger operators, Helv. Phys. Acta 70 (1997), 16–43 . Combes, J. M. and Hislop, P. D.: Localization for some continuous, random Hamiltonian in d-dimension, J. Funct. Anal. 124 (1994), 149–180. Combes, J. M., Hislop, P. D. and Tip, A.: Band edge localization and the density of states for acoustic and electromagnetic waves in random media, Ann. Inst. H. Poincaré Phys. Théor. 70 (1999), 381–428. Deift, P. A.: Applications of a commutation formula, Duke Math. J. 45 (1978), 267–310. von Dreifus, H. and Klein, A.: A new proof of localization in the Anderson tight binding model, Comm. Math. Phys. 124 (1989), 285–299. Eastham, M.: The Spectral Theory of Periodic Differential Equations, Scottish Academic Press, 1973. Figotin, A. and Klein, A.: Localization of classical waves I: Acoustic waves, Comm. Math. Phys. 180 (1996), 439–482. Figotin, A. and Klein, A.: Localization of classical waves II: Electromagnetic waves, Comm. Math. Phys. 184 (1997), 411–441. Figotin, A. and Klein, A.: Localized classical waves created by defects, J. Statist. Phys. 86 (1997), 165–177. Figotin, A. and Klein, A.: Midgap defect modes in dielectric and acoustic media, SIAM J. Appl. Math. 58 (1998), 1748–1773. Figotin, A. and Klein, A.: Localization of light in lossless inhomogeneous dielectrics, J. Opt. Soc. Amer. A 15 (1998), 1423–1435.
130 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.
ABEL KLEIN AND ANDREW KOINES
Fröhlich, J., Martinelli, F., Scoppola, E. and Spencer, T.: Constructive proof of localization in the Anderson tight binding model, Comm. Math. Phys. 101 (1985), 21–46. Fröhlich, J. and Spencer, T.: Absence of diffusion with Anderson tight binding model for large disorder or low energy, Comm. Math. Phys. 88 (1983), 151–184. Germinet, F. and Klein, A.: Bootstrap multiscale analysis and localization in random media, Comm. Math. Phys., to appear. Kato, T.: Perturbation Theory for Linear Operators, Springer-Verlag, New York, 1976. Klein, A.: Localization of light in randomized periodic media, In: J.-P. Fouque (ed.), Diffuse Waves in Complex Media, Kluwer, Dordrecht, 1999, pp. 73–92. Klein, A. and Koines, A.: A general framework for localization of classical waves: II. Random media, in preparation. Klein, A., Koines, A. and Seifert, M.: Generalized eigenfunctions for waves in inhomogeneous media, J. Funct. Anal., to appear. Reed, M. and Simon, B.: Methods of Modern Mathematical Physics, Vol. IV, Analysis of Operators, Academic Press, New York, 1978. Schulenberger, J. and Wilcox, C.: Coerciveness inequalities for nonelliptic systems of partial differential equations, Arch. Rational Mech. Anal. 88 (1971), 229–305. Wilcox, C.: Wave operators and asymptotic solutions of wave propagation problems of classical physics, Arch. Rational Mech. Anal. 22 (1966), 37–78.
Mathematical Physics, Analysis and Geometry 4: 131–146, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
131
Toda Equations, bi-Hamiltonian Systems, and Compatible Lie Algebroids ATTILIO MEUCCI Bain & Co, via Crocefisso 10, I-20122 Milan, Italy. e-mail:
[email protected] (Received: 23 March 2001) Abstract. We present the bi-Hamiltonian structure of Toda3 , a dynamical system studied by Kupershmidt as a restriction of the discrete KP hierarchy. We derive this structure by a suitable reduction of the set of maps from Zd to GL(3, R), in the framework of Lie algebroids. Mathematics Subject Classifications (2000): 37K10, 70H06. Key words: Toda lattice, Lie algebroids, bi-Hamiltonian manifolds, Marsden–Ratiu reduction.
1. Introduction It is well known (see [12] and references therein) that the periodic Toda lattice is a bi-Hamiltonian system and that its integrability properties can be easily derived by its bi-Hamiltonian structure. In [11], this structure is investigated by means of a new approach: it stems from a reduction process of a special kind of Lie algebroids [6]. This approach parallels the work of [2], where the ‘continuous counterpart’ of the Toda lattice is studied, namely, the KdV equation. Indeed, the KdV equation is a bi-Hamiltonian system obtained by reducing the space Map(S 1 , gl(2, R)) of C ∞ maps from S 1 to gl(2, R). If instead of the space Map(S 1 , gl(2, R)) one considers the space Map(S 1 , gl(3, R)), one obtains the Boussineq hierarchy, which also displays a bi-Hamiltonian structure. The discrete version of the KdV equation, the Toda lattice, is obtained in [11] by replacing the circle S 1 with the cyclic group Zd and the algebra gl(2, R) with the group GL(2, R): therefore, the space to reduce is Map(Zd , GL(2, R)). In this paper we analyze the discrete version of the Boussineq equation. We consider therefore the reduction of the space Map(Zd , GL(3, R)). The equations we obtain also display a bi-Hamiltonian structure, which, to the best of our knowledge, is not known in the literature. We call Toda3 the integrable dynamical system that arises naturally from this structure. This dynamical system is studied under a different perspective by Kupershmidt in [5]. The plan of the paper is the following. In Section 2 we introduce the geometrical structure of the phase space to reduce: the set of maps Map(Zd , GL(3, R)). Following the analysis in [11], this space can be endowed with the structure of
132
ATTILIO MEUCCI
a Poisson bi-anchored manifold. In Section 3 we present the reduction of the Poisson bi-anchored manifold Map(Zd , GL(3, R)). This reduction is an adaptation of the Marsden–Ratiu reduction scheme for Poisson manifolds [9] and gives rise to a bi-Hamiltonian manifold. In Section 4 we apply the theory of Gelfand and Zakharevich [4] to study the integrability properties of the bi-Hamiltonian flows obtained previously. The last section contains an example. 2. Poisson Bi-anchored Manifolds In this section we introduce the geometric objects we need for our approach to the Toda lattice. The discussion closely follows [11]. We will endow the manifold M of the maps from the cyclic group Zd to GL(3, R) with several structures, namely a Poisson tensor and two compatible Lie algebroids (see [6]) suitably soldered together: this will make M into a Poisson bi-anchored manifold. First we introduce the manifold M. A point q of M is simply a d-tuple of invertible 3 × 3 matrices (1) q = q1, . . . , qd , where
q1k q2k q3k q k = q4k q5k q6k . q7k q8k q9k
(2)
We will always be dealing with d-tuples of matrices and the following condition is supposed to hold throughout the discussion: (·)k+d = (·)k .
(3)
For convenience, we will say that a matrix as in (2) represents a d-tuple as in (1). Vector fields on M will be represented by d-tuples of 3 × 3 matrices q˙ k whose entries are functions of the point q ∈ M. The same way, one-forms on M are represented as d-tuples of 3 × 3 matrices α k where each entry is a function of the point. The value of the one-form α on the vector field q˙ is given by the scalar function: α, q ˙ =
d
Tr α k q˙ k .
(4)
k=1
Next, we endow M with a Poisson manifold structure. A quick computation shows that the map P : T ∗ M → T M defined as q˙ k = P (α)k = q k α k b − bα k q k ,
(5)
133
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
where b is any fixed matrix, is indeed a Poisson tensor, i.e., it defines a Poisson bracket {f, g} = df, P dg . In order to recover the Toda lattice we choose 1 0 0 b = 0 0 0 . (6) 0 0 0 To obtain the Hamiltonian vector field XH associated with a (Hamiltonian) function H : M → R, we simply have to plug its differential α = dH into Equation (5). At this point we endow M with an additional structure: a pencil of Lie algebroids. We recall (see [6]) that (M, E, A, {·, ·}) is a Lie algebroid if (i) E is a vector bundle on M (ii) {·, ·} is a bilinear composition law on the sections of E that makes them into a Lie algebra (iii) the map A: E → T M, called an anchor, is a Lie algebra morphism: A({s, t}) = [A(s), A(t)],
(7)
where [·, ·] is the usual commutator of vector fields.
If two different Lie algebroid structures (M, E, A, {·, ·}) and (M, E, A , {·, ·} ) coexist on the same manifold M and vector bundle E we can consider the pencil of brackets {s, t}λ = {s, t} + λ{s, t}
(8)
and the pencil of maps
Aλ (s) = A(s) + λA (s),
(9)
where λ is a complex parameter. The two Lie algebroid structures are said to be compatible if (M, E, Aλ , {·, ·}λ ) is a Lie algebroid for every value of λ. In our case, we consider the trivial vector bundle E = M × [Mat(3, R)]d , where d (·) denotes the d-times Cartesian product (·)×· · ·×(·). The sections of this bundle are represented by d-tuples of 3 × 3 matrices s k whose entries are functions of the point q ∈ M. Then we define the pencil of anchors Aλ : E → T M as (10) q˙ k = Aλ (s)k = s k+1 q k + λb − q k + λb s k , and a composition law {·, ·}λ as
{s, t}kλ = ∂A(s)t k − ∂A(t )s k + t k , s k + λ ∂A (s) t k − ∂A (t )s k ,
(11)
˙ It is where by ∂q˙ t we mean the derivative of the section t along the vector field q. easy to prove that the manifold (M, E, Aλ , {·, ·}λ ) is a pencil of Lie algebroids. The pencil of Lie algebroids structure defines a useful relation among oneforms. We define a one-form α to be related with a one-form β (and we denote this by α ∼ β) if A∗ α = A ∗ β. In order to explicitly calculate this relation we
134
ATTILIO MEUCCI
need to derive the expression of the dual pencil of anchors A∗λ = A∗ + λA ∗ . An element ξ of the dual vector bundle E ∗ can be naturally identified with a point of E by means of the pairing ξ, s =
d
Tr ξ k s k .
k=1
Therefore the dual pencil A∗λ : T ∗ M → E ∗ can be viewed as a map that with a oneform α associates a d-tuple ξ k of matrices whose entries are functions of the point q ∈ M. A quick calculation shows that A∗λ reads: (12) ξ k = A∗λ (α)k = q k−1 + λb α k−1 − α k q k + λb . Thus far we have defined a Poisson structure and a pencil of Lie algebroids on the manifold M. We arrived at the last step: soldering these structures by means of two maps J, J : T ∗ M → E that verify the following conditions: if α ∼ β, then P (α) = A (J α + J β), P (β) = A(J α + J ).
(13) (14)
It is easy to see that the intertwining maps defined as J : sk = αk q k ,
J : s k = −α k b,
satisfy the above relation. This ends the definition of the geometrical structures of the manifold M = Map(Zd , GL(3, R)): it is a Poisson manifold endowed with two compatible Lie algebroid structures that define a relation on one-forms and two intertwining maps that solder everything. We call such a structure a Poisson bi-anchored manifold. 3. The Reduction In this section we perform the reduction of the manifold M = Map(Zd , GL(3, R)) introduced in the previous section. Combining a restriction and a projection we obtain a new manifold N of lower dimension endowed with the same geometrical structure as the original manifold M, but with the important additional property of being bi-Hamiltonian. As a first step, we consider the distribution Im P + Im A . It is easy to verify by formulas ( 5) and (10) that this distribution is integrable and therefore it foliates Map(Zd , GL(3, R)) in maximal integral leaves, which are the 5d-dimensional hyperplanes of the form k k k q1 q2 q3 k (15) q = q4k ν5k ν6k , q7k ν8k ν9k
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
135
where the νik are constants. The restriction process we mentioned above consists in selecting one of these leaves. To obtain the Toda3 system, we pick the leaf L defined by points of the form k k k q1 q2 q3 (16) q k = q4k 0 0 . q7k 1 0 The pencil of anchors allows us to define another distribution, namely D = A(ker A ), which is integrable. As opposed to the GL(2, R) case, this distribution is not tangent to L. We are interested only in the restriction of this distribution to the leaf L, which we denote by D|L . The distribution E = D|L ∩ T L of L is also integrable and an explicit computation shows that E is spanned by the vector fields of the form q˙1k = 0, q˙5k = 0,
q˙2k = µk q2k − q3k s8k , q˙3k = µk−1 q3k , q˙4k = −µk+1 q4k , q˙6k = 0, q˙7k = τ k q4k − µk q7k , q˙8k = 0, q˙9k = 0
(17)
for arbitrary µk , τ k . From this expression we see that along the vector fields in E the following equations are satisfied: k+1 k k k+2 • • q2 q4 + q3k+1 q7k = 0, q4 q3 = 0. q˙1k = 0, This means that the distribution E admits the three invariants a1k = q1k , a2k = q2k+1 q4k + q3k+1 q7k ,
(18) (19)
a3k = q4k q3k+2 .
(20)
At this point we can operate the projection we mentioned above: we define the reduced manifold N to be the quotient of the leaf L with respect to the foliation induced by the distribution E. By (18), (19) and (20) we argue that N is a 3d-dimensional manifold that can be regarded as R3d , endowed with the set of coordinates (a1k , a2k , a3k )k=1,...,d . The above formulas also yield the expression of the canonical projection π : L → N . After obtaining the reduced manifold N we endow it with a Poisson structure. This can be done observing that (M, P , D, L) is Poisson reducible, in the terminology of [9]. In this context, to find the expression of the reduction of P we have to extend a generic one-form ϕ on N to a one-form α on M (possibly defined only at the points of the leaf L) which annihilates the distribution D. This means: α, D = 0,
˙ α, q ˙ = ϕ, π∗ q .
Let us denote by α = ext(ϕ) any such extension. Then the expression p (ϕ) = π∗ ◦ P (ext(ϕ))
(21)
This is true in general, provided that for any two sections s, t such that A (s) = 0 and A (t) = 0 we have {s, t} = 0. It is evident from (11) that this condition holds in our case.
136
ATTILIO MEUCCI
does not depend on the of ext(ϕ) and determines a Poisson structure on N .
dchoice k If we denote by ϕ = k=1 ϕ1 da1k + ϕ2k da2k + ϕ3k da3k the generic one-form on N , an easy calculation shows that an extension α = ext(ϕ) has the form ϕ3k q3k+2 + q2k+1 ϕ2k ϕ2k q3k+1 ϕ1k (22) αk = ϕ2k−1 q4k−1 α5k q3k+1 ϕ3k−1 q4k−1 , k−2 k−2 k−1 k−1 k ϕ3 q4 + q7 ϕ2 α8 α9k where α9k = q7k−1 ϕ3k−1 q3k+1 − q2k ϕ3k−2 q4k−2 + α5k−1 . As we said, this matrix is not completely determined: the components α5k , α8k are free, since the extension gives an equivalence class of one-forms. Now we can apply formula (21) to obtain the expression of the reduced Poisson tensor p : a˙ 1k = ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k , a˙ 2k = ϕ1k − ϕ1k+1 a2k − ϕ2k+1 a3k + ϕ2k−1 a3k−1 , a˙ 3k = ϕ1k − ϕ1k+2 a3k .
(23)
Recalling the periodic Toda case [11], to find another Poisson structure we have to determine the reduced relation on the one-forms of N . Therefore we need the expression of the reduced dual pencil of anchors a∗λ . To have this, in turn, we have to define a proper reduced vector bundle U based on N on which the reduced anchors act: aλ : U → T N . It is convenient to define first the dual vector bundle U∗ and the dual pencil a∗λ . The definition of the vector bundle U and the pencil aλ will then follow by duality. A lenghty computation [10] allows us to find all these characters. We are only interested in the reduced relation on one-forms, which turns out to be the following: ϕ ∼ ψ (i.e., a∗ (ϕ) = a ∗ (ψ)) if and only if they satisfy ψ1k − ψ1k+1 = ϕ1k a1k − ϕ1k+1 a1k+1 + ϕ2k−1 a2k−1 − ϕ2k+1 a2k+1 + + ϕ3k−2 a3k−2 − ϕ3k+1 a3k+1 ,
(24)
−ψ2k−2 a3k−2 + ψ2k a3k−1 = ϕ1k−2 a3k−2 − ϕ1k+1 a3k−1 − ϕ2k−2 a3k−2 a1k−1 + + ϕ2k a3k−1 a1k − ϕ3k−2 a3k−2 a2k−1 + ϕ3k−1 a3k−1 a2k−1 , ψ2k a2k − ψ2k−1 a2k−1 + ψ3k a3k − ψ3k−2 a3k−2 =
ϕ1k−1 a2k−1 − ϕ1k+1 a2k + + ϕ2k−2 a3k−2 − ϕ2k+1 a3k
(25) (26)
+ a1k ϕ2k a2k − ϕ2k−1 a2k−1 + a1k ϕ3k a3k − ϕ3k−2 a3k−2 .
Now we focus on the special feature of the reduced Poisson bi-anchored manifold N . By Equations (24), (25) and (26) for a fixed one-form ϕ there is a whole class
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
137
of one-forms [ψ] = ψ + ker a ∗ that is related with it. In the reduced structure, though, ker a ∗ ⊂ ker p . Therefore the following tensor, p(ϕ) := p (ψ),
(27)
is well defined. A lengthy calculation shows that the bracket induced by this tensor verifies the Jacobi identity. Furthermore, p is compatible with p , i.e., the pencil pλ = p + λp is a Poisson tensor for all values of the complex parameter λ. Explicitly, a˙ = p(ϕ) reads a˙ 1k = a1k ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k + + a2k ϕ1k+1 − a2k−1 ϕ1k−1 + a3k ϕ2k+1 − a3k−2 ϕ2k−2 , a˙ 2k = a1k ϕ1k a2k + ϕ2k−1 a3k−1 − (28) k+1 k k+1 k+1 k k+2 k k−1 k−1 − a1 ϕ1 a2 + ϕ2 a3 + ϕ1 a3 − ϕ1 a3 + + a2k ϕ2k−1 a2k−1 − ϕ2k+1 a2k+1 + ϕ3k−2 a3k−2 + ϕ3k−1 a3k−1 − − ϕ3k a3k − ϕ3k+1 a3k+1 , a˙ 3k = a3k ϕ1k a1k − ϕ1k+2 a1k+2 + ϕ2k−1 a2k−1 + ϕ2k a2k − ϕ2k+1 a2k+1 − ϕ2k+2 a2k+2 + + ϕ3k−2 a3k−2 + ϕ3k−1 a3k−1 − ϕ3k+1 a3k+1 − ϕ3k+2 a3k+2 . Our goal has been achieved: we arrived at a bi-Hamiltonian manifold by means of a systematic procedure of reduction of the original Poisson bi-anchored manifold. Now we have to investigate the information provided by the bi-Hamiltonian structure. 4. The Toda3 System In this section we show how the bi-Hamiltonian structure obtained in the previous section defines specific vector fields and, moreover, accounts for their integrability. To obtain an integrable system we will focus on the Casimirs of the Poisson pencil, i.e., the functions C such that their differentials are in the kernel of Pλ . Indeed, if we expand a Casimir in powers of λ, H i λi , (29) C= i
it is immediate to check that the coefficients Hi satisfy the Lenard relations P (dHi ) = −P (dHi+1 ).
(30)
It is easily shown (see, e.g., [7]) that this in turn implies that the Hi ’s are in involution with respect to both Poisson bracket. If these coefficients are enough, the We recall that a manifold endowed with two compatible Poisson tensors is said to be
bi-Hamiltonian.
138
ATTILIO MEUCCI
system is integrable in the classical sense of Liouville and Arnold [1]. Nevertheless, in general finding the Casimirs of a Poisson pencil is not easy. In the present case we can make use of the following PROPOSITION 1. If hk solves hk hk+1 hk+2 = a1k+2 + λ hk hk+1 + a2k+1 hk + a3k ,
(31)
then C(λ) = h1 · · · hd is a Casimir of the Poisson pencil (23)–(28). The solutions hk and thus the Casimirs C can be calculated explicitly as Laurent series in the parameter λ. Proof. See the appendix. ✷ We call (31) the characteristic equation. Proposition 1 allows us in principle to calculate the Casimirs of the Poisson pencil pλ , but the computation is lengthy. Fortunately there is a shortcut: if in (31) we set hk =
ψk+1 µ, ψk
(32)
the characteristic equation becomes the linear system 0 = ψk+3 µ3 − a1k+2 + λ ψk+2 µ2 − a2k+1 ψk+1 µ − a3k ψk .
(33)
We can express (33) in matrix form as Lψ = 0, where L is the matrix −µ3 µ2 a11 + λ µa21 µ2 a12 + λ L= a31 µa22 .. . 0 −µ3 0
(34)
0
a3d−1 .. . .. .
µa2d
−µ3 a3d .. . 0 .. d−1 2 3 . µ a1 + λ −µ a3d−2 µa2d−1 µ2 a1d + λ
and ψ is the vector of the ‘homogeneous coordinates’ ψ1 ψ = ... . ψd For (34) to admit nontrivial solutions we must have det L = 0. It can be proved that the cyclicity of the matrix L implies that its determinant is a polynomial of degree 3 in µd . Therefore we must have 0 = det L = ±µ3d + K1 µ2d + K2 µd + K3 ,
(35)
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
139
where K1 , K2 , K3 are polynomials in λ (in particular, C3 does not depend on λ). By Proposition 1 and Equation (32), for all µ that satisfy (35) we have that µd = h1 . . . hd is a Casimir, thus K1 , K2 and K3 are Casimirs as well. Their coefficients provide all information about the geometry of the system at hand. The family of dynamical systems associated with the Casimirs K1 , K2 and K3 is what we call Toda3 . It can be proved that they coincide with Kupershmidt’s reduction of the discrete KP hierarchy [5]. We will show in an example how the integrability of these systems stems from their bi-Hamiltonian structure. 5. An Example of the Toda3 System To illustrate how the scheme described above works we consider the specific case where d = 4. This example is easy to handle, but at the same time general enough. In order to make the equations easier to read we will change notation: we set a1k = bk , a2k = ak , a3k = ck , ϕ1k = βk , ϕ2k = αk , ϕ3k = γk . Thus our bi-Hamiltonian manifold N becomes R12
with coordinates (a1 , . . . , c4 ). The Poisson pencil pλ associates with a one-form 4k=1 (αk dak + βk dbk + γk dck ) the vector field b˙k = (bk + λ)(αk−1 ak−1 − αk ak + γk+2 ck+2 − γk ck ) + + ak βk+1 − ak−1 βk−1 + ck αk+1 − ck+2 αk+2 , a˙ k = (ck−1 αk−1 + ak βk )(bk + λ) − − (ck αk+1 + ak βk+1 )(bk+1 + λ) + ck βk+2 − βk−1 ck−1 + + ak (αk−1 ak−1 − αk+1 ak+1 + γk−1 ck−1 − γk ck − − γk+1 ck+1 + γk+2 ck+2 ), c˙k = ck (βk (bk + λ) − βk+2 (bk+2 + λ)) + + ck (−αk+2 ak+2 + αk ak − αk+1 ak+1 + αk−1 ak−1 + + γk−1 ck−1 − γk+1 ck+1 ),
(36)
where k = 1, 2, 3, 4. Equation (35) provides the Casimirs of the Poisson pencil (36). Indeed, from that equation one expects to find only three Casimirs of the whole pencil. The interesting feature of this approach is that in fact we obtain four Casimirs which prove to be enough to guarantee integrability. The three coefficients are K1 (λ) = λ4 + λ3 C1 + λ2 H1 + λH2 + H3 , K2 (λ) = λ2 C2 + λC3 + H4 , K3 (λ) = C4 , where C1 = b1 + b2 + b3 + b4 , H1 = a1 + a2 + a3 + a4 + b1 b2 + b2 b3 + b3 b4 + b4 b1 + b1 b3 + b2 b4 ,
140
ATTILIO MEUCCI
H2 = c1 + c2 + c3 + c4 + b1 b2 b3 + b2 b3 b4 + b3 b4 b1 + b4 b1 b2 + + b1 (a2 + a3 ) + b2 (a3 + a4 ) + b3 (a4 + a1 ) + b4 (a1 + a2 ), H3 = b1 b2 b3 b4 + b1 c2 + b2 c3 + b3 c4 + b4 c1 + a1 a3 + a2 a4 + + b1 b2 a3 + b2 b3 a4 + b3 b4 a1 + b4 b1 a2 , C2 = −c2 c4 − c1 c3 , C3 = −b1 c2 c4 − b2 c3 c1 − b3 c4 c2 − b4 c1 c3 + + c1 a3 a4 + c2 a4 a1 + c3 a1 a2 + c4 a2 a3 , H4 = −a1 a2 a3 a4 + b1 a2 a3 c4 + b2 a3 a4 c1 + b3 a4 a1 c2 + b4 a1 a2 c3 + + a1 c2 c3 + a2 c3 c4 + a3 c4 c1 + a4 c1 c2 − b2 b4 c1 c3 − b1 b3 c2 c4 , C4 = c1 c2 c3 c4 . Nonetheless, the second coefficient K2 (λ) is composed itself of two independent Casimirs of the pencil: λ2 C2 and λC3 + H4 . Therefore we obtain the four Casimirs: K1 (λ) = λ4 + λ3 C1 + λ2 H1 + λH2 + H3 , K2
(λ) = λC3 + H4 , K3 (λ) = C4 .
K2 (λ) = C2 ,
In the theory of Gelfand–Zakharevich the important objects are these eight functions C1 , . . . , C4 , H1 , . . . , H4 . Since the K’s are Casimir functions of the Poisson pencil, we must have: p (dH1 ) = −p(dC1 ), p (dH3 ) = −p(dH2 ),
p (dH2 ) = −p(dH1 ), p (dH4 ) = −p(dC3 )
(37)
and 0 = p (dC1 ) = p (dC2 ) = p (dC3 ) = p (dC4 ) = p(dC2 ) = p(dH3 ) = p(dC4 ) = p(dH4 ).
(38)
The 4-particle periodic Toda3 system are the vector fields X1 , . . . , X4 defined as follows: X1 = p (dH1 ),
X2 = p (dH2 ),
X3 = p (dH3 ),
X4 = p (dH4 ).
The explicit expression of these four vector fields is X1 :
X2 :
X3 :
b˙1 = a4 − a1 , a˙ 1 = a1 (b2 − b1 ) − c1 + c4 , c˙1 = c1 (b3 − b1 ), b˙1 = c3 − c1 − (b3 + b4 )a1 + (b2 + b3 )a4 , a˙ 1 = a1 (−b4 b1 − b1 b3 + b2 b3 + b4 b2 − a4 + a2 )− −c1 (b4 + b1 ) + c4 (b2 + b3 ), c˙1 = c1 (−b4 b1 − b1 b2 + b2 b3 + b3 b4 − a4 − a1 + a2 + a3 ), b˙1 = −b4 c1 + b2 c3 − (a3 + b3 b4 )a1 + (a2 + b2 b3 )a4 , a˙ 1 = a1 (−b3 b4 b1 + b2 b3 b4 − b3 a4 − b1 a3 + b2 a3 + b4 a2 −
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
X4 :
141
−c3 + c2 ) − c1 (a4 + b4 b1 ) + c4 (a2 + b2 b3 ), c˙1 = c1 (−b4 b1 b2 + b2 b3 b4 − b4 a1 − b2 a4 + b2 a3 + b4 a2 − c4 + c2 ), b˙1 = −c1 c4 a3 + c3 c4 a2 , a˙ 1 = −c4 b1 c1 a3 − c4 c1 c3 + c1 b2 c4 a3 + c1 c4 c2 , c˙1 = c1 (−a4 a1 c2 + b1 c4 c2 + a2 a3 c4 − b3 c2 c4 ).
Of course, in the above formulas the cyclic condition holds and yields the other components. Due to (37) these vector fields are bi-Hamiltonian. In order to show that they are integrable we choose a symplectic leaf of the Poisson tensor p . It can be shown that this leaf is given by C1 = constant, C3 = constant,
C2 = constant, C4 = constant.
Therefore, the leaf is an eight-dimensional symplectic submanifold of the twelvedimensional original phase space. As we said in Section 4, the functions H1 , . . . , H4 commute with respect to both Poisson brackets. Therefore their restrictions to the symplectic leaf also commute, and, since they can be checked to be functionally independent, constitute an integrable system. 6. Conclusions In this paper we showed that reducing a special Poisson bi-anchored manifold, namely the set of maps from Zd to GL(3, R), we obtain a new bi-Hamiltonian structure. This bi-Hamiltonian structure gives rise to an integrable system, which we called Toda3 and already appeared in the work of Kupershmidt [5]. This system represents a generalization of the periodic Toda lattice (which corresponds to GL(2, R)). There are several further developments of this approach. First of all, it is easy to endow the set of maps from Zd to GL(n, R) for a generic n ∈ N with the structure of bi-anchored Poisson manifold. It is possible to show that the reduction of these manifolds gives rise to other Toda systems, which are the discrete analog of the Gelfand–Dickey hierarchies [3]. Secondly (see [8, 10]) the study of the conservation laws of the periodic Toda lattice allows to define the discrete analog [5] of the KP equations on the Sato Grassmannian [13]. These represent flows on an infinite-dimensional phase space that admit invariant submanifolds. These submanifolds are the different phase spaces of the Toda system, and the restriction of the KP equation to these phase spaces are the Toda equations. This way it is possible to extend to the discrete case the description given for the continuous case in [2], where the KdV hierarchy and the (usual) KP equations are considered.
142
ATTILIO MEUCCI
Appendix. Proof of Proposition 1 We will split the proof in three parts. (1) We are looking for Casimir functions, i.e. exact one-forms in the kernel of the Poisson pencil pλ , which we rewrite like this a˙ 1k = a1k + λ ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k − − ϕ1k−1 a2k−1 + ϕ1k+1 a2k − ϕ2k−2 a3k−2 + ϕ2k+1 a3k , a˙ 2k = ϕ2k−1 a3k−1 a1k + λ − ϕ2k+1 a3k a1k+1 + λ −
(39)
− ϕ1k−1 a3k−1 + ϕ1k+2 a3k + ϕ3k−1 a3k−1 a2k − ϕ3k a3k a2k + + a2k ϕ1k a1k + λ − ϕ1k+1 a1k+1 + λ + ϕ2k−1 a2k−1 − − ϕ2k+1 a2k+1 + ϕ3k−2 a3k−2 − ϕ3k+1 a3k+1 , a˙ 3k = a3k ϕ1k a1k + λ − ϕ1k+1 a1k+1 + λ + ϕ2k−1 a2k−1 − − ϕ2k+1 a2k+1 + ϕ3k−2 a3k−2 − ϕ3k+1 a3k+1 + + a3k ϕ1k+1 a1k+1 + λ − ϕ1k+2 a1k+2 + λ + ϕ2k a2k − − ϕ2k+2 a2k+2 + ϕ3k−1 a3k−1 − ϕ3k+2 a3k+2 . From this expression, it is straightforward to see that in order for the one-form ϕ to be in the kernel of the Poisson pencil it is enough that it satisfies the following equations: 0 = a1k + λ ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k − − ϕ1k−1 a2k−1 + ϕ1k+1 a2k − ϕ2k−2 a3k−2 + ϕ2k+1 a3k , 0 = ϕ2k−1 a3k−1 a1k + λ − ϕ2k+1 a3k a1k+1 + λ − 0 =
− ϕ1k−1 a3k−1 + ϕ1k+2 a3k + ϕ3k−1 a3k−1 a2k − ϕ3k a3k a2k , ϕ1k a1k + λ − ϕ1k+1 a1k+1 + λ + ϕ2k−1 a2k−1 − − ϕ2k+1 a2k+1 + ϕ3k−2 a3k−2 − ϕ3k+1 a3k+1 .
(40)
(41)
A change of variables reduces this system to a single equation of Riccati type. Indeed, let hk be any solution of what we call the characteristic equation, which is of Riccati type: (42) hk hk+1 hk+2 = a1k+2 + λ hk hk+1 + a2k+1 hk + a3k . Let us set ϕ1k+2 = hk hk+1 ρk ,
ϕ2k+1 = hk ρk ,
ϕ3k = ρk .
Let us choose the multiplier ρ is such a way that the following equation holds: ϕ1k a1k + λ + ϕ2k−1 a2k−1 + ϕ2k a2k + ϕ3k−2 a3k−2 + ϕ3k−1 a3k−1 + ϕ3k a3k = L, where L is any constant different from zero.
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
143
(2) Such a one-form ϕ solves (40) and therefore it is an element of the kernel of the Poisson pencil pλ . Indeed, let us consider a one-form ϕ that satisfies the system: hk hk+1 hk+2 = a1k+2 + λ hk hk+1 + a2k+1 hk + a3k , ϕ1k = hk−2 hk−1 ϕ3k−2 ,
ϕ2k = hk−1 ϕ3k−1 ,
(43)
L = ϕ1k (a1k + λ) + ϕ2k−1 a2k−1 + ϕ2k a2k + + ϕ3k−2 a3k−2 + ϕ3k−1 a3k−1 + ϕ3k a3k . We will show that ϕ satisfies the following equations: 0 = a1k + λ ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k − − ϕ1k−1 a2k−1 + ϕ1k+1 a2k − ϕ2k−2 a3k−2 + ϕ2k+1 a3k , 0 = ϕ2k−1 a3k−1 a1k + λ − ϕ2k+1 a3k a1k+1 + λ −
(44)
− ϕ1k−1 a3k−1 + ϕ1k+2 a3k + ϕ3k−1 a3k−1 a2k − ϕ3k a3k a2k , 0 = ϕ1k a1k + λ − ϕ1k+1 a1k+1 + λ + ϕ2k−1 a2k−1 − ϕ2k+1 a2k+1 + + ϕ3k−2 a3k−2 − ϕ3k+1 a3k+1 . We immediately see that ϕ fulfills the third equation of (44): just compare with the last equation of (43). Before we verify that the two other equations are satisfied as well, we need a formula. Namely, we have hk−3 hk−2 hk−1 ϕ3k−3 − hk−2 hk−1 a1k + λ ϕ3k−2 = hk−1 ϕ3k−1 a2k + ϕ3k a3k . (45) We will show now that ϕ satisfies the first equation of (44): 0 = a1k + λ ϕ2k−1 a2k−1 − ϕ2k a2k + ϕ3k−2 a3k−2 − ϕ3k a3k − − ϕ1k−1 a2k−1
+
ϕ1k+1 a2k
−
ϕ2k−2 a3k−2
+
(46)
ϕ2k+1 a3k .
From the last equation of (43) we obtain
L = ϕ3k−2 hk−2 hk−1 (a1k + λ) + hk−2 ϕ3k−2 a2k−1 + hk−1 ϕ3k−1 a2k + ϕ3k−2 a3k−2 + + ϕ3k−1 a3k−1 + ϕ3k a3k = ϕ3k−2 (hk−2 hk−1 (a1k + λ) + hk−2 a2k−1 + a3k−2 ) + hk−1 ϕ3k−1 a2k + ϕ3k−1 a3k−1 + ϕ3k a3k = ϕ3k−2 (hk−2 hk−1 hk ) + ϕ3k−1 (hk−1 a2k + a3k−1 ) + ϕ3k a3k . This implies ϕ3k−2 (hk−2 hk−1 hk ) = ϕ3k−1 (hk−1 hk hk+1 − hk−1 a2k − a3k−1 ) + ϕ3k (hk a2k+1 ) + ϕ3k+1 a3k+1 and the result follows using the characteristic equation hk−1 hk hk+1 − a2k hk−1 − a3k−1 = (a1k+1 + λ)hk−1 hk .
144
ATTILIO MEUCCI
We replace in this expression the values of ϕ1k , ϕ2k obtained from (43), and we multiply both sides by hk−2 hk−1 . We obtain that (46) holds if and only if 0 = hk−2 hk−1 a1k + λ hk−2 ϕ3k−2 a2k−1 − hk−1 ϕ3k−1 a2k + ϕ3k−2 a3k−2 − ϕ3k a3k − − hk−2 hk−1 hk−3 hk−2 ϕ3k−3 a2k−1 + hk−2 hk−1 hk−1 hk ϕ3k−1 a2k − − hk−2 hk−1 hk−3 ϕ3k−3 a3k−2 + hk−2 hk−1 hk ϕ3k a3k . Rearranging the terms in the last expression, we end up having to prove that
k−1 k k k ϕ a + ϕ a hk−3 h k−2 hk−1 ϕ3k−3 − hk−2 hk−1 a1k + λ ϕ3k−2 hk−1 2 3 3 3 0 = det hk−2 a2k−1 a3k−2 hk−2 hk−1 hk − hk−2 hk−1 a1k + λ but this is true, since the characteristic equation and (45) hold. We will show now that ϕ satisfies the second equation of (44): 0 = ϕ2k−1 a3k−1 a1k + λ − ϕ2k+1 a3k a1k+1 + λ − − ϕ1k−1 a3k−1 + ϕ1k+2 a3k + ϕ3k−1 a3k−1 a2k − ϕ3k a3k a2k .
(47)
We replace in this expression the values of ϕ1k , ϕ2k obtained from (43), and we multiply both sides by hk−1 . We obtain that (47) holds if and only if: 0 = hk−1 hk−2 ϕ3k−2 a3k−1 a1k + λ − hk−1 hk ϕ3k a3k a1k+1 + λ − − hk−1 hk−3 hk−2 ϕ3k−3 a3k−1 + hk−1 hk hk+1 ϕ3k a3k + + hk−1 ϕ3k−1 a3k−1 a2k − hk−1 ϕ3k a3k a2k . Rearranging the terms in the last expression, we end up having to prove that
k−1 hk−1 hk hk+1 − hk−1 hk a1k+1 + λ −a2k a3 0 = det ϕ3k a3k hk−1 hk−3 hk−2 ϕ3k−3 − hk−1 hk−2 ϕ3k−2 a1k + λ − ϕ3k−1 a2k but this is true, since the characteristic equation and (45) hold. (3) Furthermore, ϕ is exact. The system hk hk+1 hk+2 = a1k+2 + λ hk hk+1 + a2k+1 hk + a3k , ϕ2k = hk−1 ϕ3k−1 , ϕ1k = hk−2 hk−1 ϕ3k−2 , L = ϕ1k a1k + λ + ϕ2k−1 a2k−1 + ϕ2k a2k + ϕ3k−2 a3k−2 + ϕ3k−1 a3k−1 + ϕ3k a3k is equivalent to
hk hk+1 hk+2 = a1k+2 + λ hk hk+1 + a2k+1 hk + a3k ,
ϕ2k = hk−1 ϕ3k−1 , ϕ1k = hk−1 hk−2 ϕ3k−2 , 0 = L − ϕ3k a3k + ϕ2k a1k+1 + λ hk − ϕ1k hk − ϕ2k hk hk+1 . Combining the first and the fourth equations we obtain ϕ3k a1k+2 + λ hk+1 + ϕ3k a2k+1 L = − + ϕ3k hk+1 hk+2 − ϕ2k (a1k+1 + λ) + ϕ1k + ϕ2k hk+1 . hk
145
TODA EQUATIONS, BI-HAMILTONIAN SYSTEMS, AND COMPATIBLE LIE ALGEBROIDS
We will use this in the next calculation. We move on to evaluating d k+2 k+2 ϕ1 a˙ 1 + ϕ2k+1 a˙ 2k+1 + ϕ3k a˙ 3k k=1
=
d
hk hk+1 ϕ3k a˙ 1k+2 + hk ϕ3k a˙ 2k+1 + ϕ3k a˙ 3k
k=1
=
d
ϕ3k hk hk+1 a˙ 1k+2 + hk a˙ 2k+1 + a˙ 3k
k=1
=
d
ϕ3k
k=1
= =
d
˙ k hk+1 hk+2 + hk h˙ k+1 hk+2 + hk hk+1 h˙ k+2 − hk+2 − a1 + λ h˙ k hk+1 − a1k+2 + λ hk h˙ k+1 − a2k+1 h˙ k
k=1
ϕ3k h˙ k hk+1 hk+2 + ϕ3k hk h˙ k+1 hk+2 + ϕ3k hk hk+1 h˙ k+2 − k+2 k k+2 k −ϕ3 a1 + λ h˙ k hk+1 − ϕ3 a1 + λ hk h˙ k+1 − ϕ3k a2k+1 h˙ k
k=1
k˙ hk+1 hk+2 + ϕ2k+1 h˙ k+1 hk+2 + ϕ1k+2 h˙ k+2 − ϕk3 hkk+2 k+2 k+1 k+1 k − ϕ2 a1 + λ h˙ k+1 −h˙ k ϕ3 (a1 + λ)hk+1 + ϕ3 a2
d
k+1 ˙ k+2 ˙ k˙ h h + ϕ h + ϕ − h h h ϕ k k+1 k+2 k+1 k+2 k+2 3 2 1 d −h˙ k ϕ k + ϕ k hk+1 + ϕ k hk+1 hk+2 − L − ϕ k a k+1 + λ − = 2 1 1 2 3 hk k=1 k+1 k+2 k+2 ˙ ˙ −ϕ2 a1 + λ hk+1 ϕ1 hk+2 . k L d k k+1 k k ˙ −hk ϕ1 + ϕ2 hk+1 + ϕ3 hk+1 hk+2 − hk − ϕ2 a1 + λ − −ϕ3k hk+1 hk+2 + ϕ2k+1 hk+2 − ϕ2k+1 a1k+2 + λ h˙ k+1 k=1 =
d
• L = L log(h1 · · · hd ) . h˙ k hk k=1
The last equation follows from L being a constant (independent of both site k and variables aik ). Therefore d k k ϕ1 da1 + ϕ2k da2k + ϕ3k da3k = d(L log(h1 · · · hd )). k=1
Thus, ϕ is exact, and C = h1 . . . hd is a Casimir of the Poisson pencil. This ends the proof. Acknowledgements I wish to thank Marco Pedroni for his suggestions and continuous support, as well as my former advisor, Franco Magri, for useful discussions on the topic.
146
ATTILIO MEUCCI
References 1. 2. 3. 4.
5. 6. 7. 8.
9. 10. 11. 12. 13.
Arnold, V. I.: Mathematical Methods of Classical Mechanics, Grad. Texts in Math., Springer, New York, 1989. Falqui, G., Magri, F. and Pedroni, M.: Bihamiltonian geometry, Darboux coverings, and linearization of the KP hierarchy, Comm. Math. Phys. 197 (1998), 303–324. Gelfand, I. M. and Dickey, L. A.: Fractional powers of operators and Hamiltonian systems, Funct. Anal. Appl. 10 (1976), 259–273. Gelfand, I. M. and Zakharevich, I.: On the local geometry of a bi-Hamiltonian structure, In: L. Corvin et al. (eds), The Gelfand Mathematical Seminars 1990–1992, Birkhäuser, Boston, 1993, pp. 51–112. Kupershmidt, B. A.: Discrete Lax equations and differential-difference calculus, Asterisque 123 (1985), 212–245. Mackenzie, K.: Lie Groupoids and Lie Algebroids in Differential Geometry 124, Cambridge Univ. Press, Cambridge, 1987. Magri, F.: Eight lectures on integrable systems, In: Integrability of Nonlinear Systems, Lecture Notes in Phys. 495, Springer, New York, 1997, pp. 256–296. Magri, F.: The bihamiltonian route to Sato Grassmannian, Proc. CRM Conf. on Bispectral Problems, Montreal, CRM Proc. Lecture Notes 14, Amer. Math. Soc., Providence, 1998, pp. 203–209. Marsden, J. E. and Ratiu, T.: Reduction of Poisson manifolds, Lett. Math. Phys. 11 (1986), 161–169. Meucci, A.: The bi-Hamiltonian route to the discrete Sato Grassmannian, PhD thesis, Universitá Statale di Milano, 1999. Meucci, A.: Compatible lie algebroids and the periodic Toda lattice, To appear in J. Geom. Phys. Morosi, C. and Pizzocchero, L.: R-matrix theory, formal Casimirs and the periodic Toda lattice, J. Math. Phys. 37 (1996), 4484–4513. Sato, M. and Sato, Y.: Soliton equations as dynamical systems on infinite-dimensional Grassmannian manifolds, In: P. Lax and H. Fujita (eds), Nonlinear PDE’s in Applied Sciences (US/Japan Seminar, Tokio), North-Holland, Amsterdam, 1982, pp. 259–271.
Mathematical Physics, Analysis and Geometry 4: 147–195, 2001. © 2001 Kluwer Academic Publishers. Printed in the Netherlands.
147
Ergodicity for the Randomly Forced 2D Navier–Stokes Equations SERGEI KUKSIN1 and ARMEN SHIRIKYAN2
1 Department of Mathematics, Heriot-Watt University, Edinburgh EH14 4AS, Scotland, U.K. e-mail:
[email protected] and Steklov Institute of Mathematics, 8 Gubkina St., 117966 Moscow, Russia. 2 Department of Mathematics, Heriot-Watt University, Edinburgh EH14 4AS, Scotland, U.K. e-mail:
[email protected] and Institute of Mechanics of MSU, 1 Michurinskii Av., 119899 Moscow, Russia.
(Received: 18 April 2001) Abstract. We study space-periodic 2D Navier–Stokes equations perturbed by an unbounded random kick-force. It is assumed that Fourier coefficients of the kicks are independent random variables all of whose moments are bounded and that the distributions of the first N0 coefficients (where N0 is a sufficiently large integer) have positive densities against the Lebesgue measure. We treat the equation as a random dynamical system in the space of square integrable divergence-free vector fields. We prove that this dynamical system has a unique stationary measure and study its ergodic properties. Mathematics Subject Classifications (2000): 37H99, 35Q30. Key words: Navier–Stokes equations, kick-force, stationary measure, random dynamical system, Ruelle–Perron–Frobenius theorem.
0. Introduction We continue our study of the randomly forced 2D space-periodic Navier–Stokes system (NS), started in [KS1, KS2]. That is, we consider the equations u˙ − νu + (u, ∇)u + ∇p = ηω (t, x),
div u = 0,
(0.1)
where x ∈ T = R /Z , 0 < ν 1 is the viscosity, u = u(t, x) is the velocity field, and p = p(t, x) is the pressure. Equations (0.1) are supplemented by the conditions 2
2
u ≡ η ≡ 0,
2
div η = 0.
The brackets · signify the space averaging. The right-hand side ηω is a random process with range in the functional space H = {u ∈ L2 (T2 , R2 ) : div u = 0, u = 0}, and Equations (0.1) defines a random dynamical system in H . We provide H with the usual orthonormal basis {e1 , e2 , . . .} formed by the trigonometric vector fields
148
SERGEI KUKSIN AND ARMEN SHIRIKYAN
Figure 1. Evolution defined by (0.1), (0.2).
2 2 Cs −s sin(s · x) and Cs −s cos(s · x), s ∈ Z2 \ {0}. The ej ’s are eigenvectors of s1 s1 the Laplacian, −ej = αj ej . We assume that the eigenvalues αj are indexed in non-decreasing order. In [KS1], we consider the NS equations forced by a bounded random kick-force ηω =
δ(t − kT )ηk (x),
ηk =
∞
bj ξj k ej (x),
(0.2)
j =1
k∈Z
where bj 0 are some constants such that b2 := b12 + b22 + · · · < ∞, and {ξj k } are independent random variables. It is assumed in [KS1] that the distribution D(ξj k ) of the random variable ξj k is k-independent and has the form D(ξj k ) = pj (r) dr
for j 1, k ∈ Z,
(0.3)
where the pj ’s are Lipschitz continuous functions such that pj (0) > 0, supp pj ⊂ [−1, 1]. Let {St , t 0} be flow-maps of the free NS equation (0.1) with η ≡ 0. If u(t, x) is a solution for (0.1) with a kick-force (0.2) normalised to be a continuous from the right curve in H , then for any integer k and for t ∈ [T k, T (k + 1)] we have (see Figure 1) t < T (k + 1), St −T k (u(T k)), (0.4) u(t) = t = T (k + 1). ST (u(T k)) + ηk , Accordingly, long-time behaviour of solutions for (0.1), (0.2) is described by longtime behaviour of solutions for the following random dynamical system with discrete time: uk = S(uk−1 ) + ηk ,
(0.5)
where S = ST and uk = u(T k, ·) ∈ H . In [KS1], we show that if relations (0.3) hold with densities pj as above and bj = 0
for 1 j N0
(0.6)
ERGODICITY FOR THE RANDOMLY FORCED 2D NAVIER–STOKES EQUATIONS
149
for some finite N0 = N0 (ν, b) 1, then the random dynamical system (0.5) has in H a unique stationary measure λ. Moreover, if (uk , k 0) satisfies (0.5) for k > 0 and u0 = u, then λ as k → ∞
D(uk )
(0.7)
for any choice of the initial vector u ∈ H .! We note that if bj = 0 for all j 1 and bj2 < ∞, then these results apply to Equation (0.1) with any ν > 0, any T > 0 and with arbitrarily large kick-force η as above. See the introduction to [KS1] for discussion of this result and see [G, KS1] for its relations with statistical hydrodynamics. Next, E, Mattingly, and Sinai [EMS] and Bricmont, Kupiainen, and Lefevere [BKL] considered the 2D NS equations perturbed by a white noise force η=
∞
bj w˙ j (t)ej (x),
j =1
where w1 , w2 , . . . are independent standard Brownian motions. Under the assumption that bj = 0 for 1 j N0 (ν) and bj = 0 for j > N with some ∞ > N N0 (ν), they obtained results similar to those reviewed above. We do not discuss these results here, but we mention that, as it is shown in [BKL], for almost all initial functions u(0, x) the distribution of a solution converges to the stationary measure exponentially fast. In this work we study the NS equation with unbounded random kick-forces. That is, we consider Equations (0.1), (0.2), where the independent random variables ξj k have k-independent distributions as in (0.3), the densities pj are absolutely continuous and everywhere positive, ∞ ∂pj (r) ∂r dr < ∞ for all j 1; −∞ (0.8) pj (r) > 0 for all j 1, r ∈ R, and decay at infinity faster than any negative degree of r. We consider in fact the following two extreme cases which are allowed by our techniques: (A) (finite moments) the densities pj satisfy (0.8) and ∞ |r|m pj (r) dr Cm for all m 1, j 1,
(0.9)
−∞
with some fixed constants Cm , m 1. ! In [KS1], we study in fact the system (0.3) restricted to the domain of attainability from zero A,
which is a compact subset of H , invariant for (0.3), and prove that the restricted system has a unique stationary measure and satisfies (0.7) for u ∈ A. In the short paper [KS2] we show that this measure is a unique stationary measure for the system in the whole space H and prove that (0.7) holds for any u ∈ H.
150
SERGEI KUKSIN AND ARMEN SHIRIKYAN
(B) (finite second exponential moments) the densities pj satisfy (0.8) and ∞ 2 e%0 r pj (r) dr C0 for any j 1, −∞
with some fixed positive constants %0 and C0 ;
We stress that in (A) and (B) it is not assumed that rpj (r) dr = 0. For any s > 0, we denote H s = H ∩ H s (T2 ; R2 ), where H s (T2 ; R2 ) is the Sobolev space of order s with the corresponding norm · s . MAIN THEOREM. Let us assume that condition (A) is satisfied and ∞
bj2 αjs < ∞
for some s > 0.
j =1
Then there is an integer N0 < ∞ with the following property: if (0.6) holds, then the random dynamical system (0.5) has a unique stationary measure λ such that |u|m λ(du) < ∞ for all m 1. H
Moreover, the following assertions hold: (a) λ(H s ) = 1; (b) if (uk , k 0) is a solution of Equation (0.5) with a deterministic initial function u0 = u, then, for λ-almost all u, convergence (0.7) holds. Moreover, Ef (uk ) → (λ, f )
as k → ∞,
(0.10) p
where f is any continuous function on H s such that |f (u)| C1 + C2 us for some finite constants C1 , C2 , and p. (c) if bj = 0 for all j , then supp λ = H , and convergence (0.10) holds uniformly in u ∈ H s , us R, for any R > 0. 2 Finally, if condition (B) is also satisfied, then H eβ|u| λ(du) < ∞ for some β > 0, and convergence (0.10) holds for λ-almost all u ∈ H and any function f ∈ C(H s ) such that |f (u)| C exp(σ uκs ), where the positive constants σ and κ are sufficiently small. If s > 1, then the delta-function is a continuous functional on the space H s . Accordingly, if s > 1 and u(t, x) is a solution for (0.1) such that u(0, x) = u0 ∈ H , then, for λ-almost all u0 ∈ H , the correlation tensor of the solution j Eui (k, x)u i(k, y)j converges as k → ∞ to the correlation tensor of the measure λ, equal to u (x)u (y)λ(du). If u0 is an arbitrary vector in H , then in this statement the convergence should be replaced by the Cesàro convergence. The proof of the Main Theorem remains true if condition (A) is replaced by the following weaker assumption with M 20:
ERGODICITY FOR THE RANDOMLY FORCED 2D NAVIER–STOKES EQUATIONS
151
(AM ) the densities pj satisfy (0.8), and (0.9) holds for m M and all j 1. In this case, the stationary measure λ has M M finite moments, where M goes to infinity with M, and (0.10) holds for any continuous functional f : H → R satisfying the inequality |f (u)| C1 + C2 uM s . The proof of the Main Theorem, which occupies Sections 1–5, follows the scheme developed in [KS1] to work with bounded kick-forces. It is based on a Foia¸s–Prodi type reduction of (0.5) to a finite-dimensional abstract Gibbs system which has a unique stationary solution due to a version of the Ruelle–Perron– Frobenius theorem. In fact, the Main Theorem can be strengthened as follows: AMPLIFICATION. Under the assumption of the above theorem, convergence (0.10) holds for any u ∈ H , uniformly on bounded subsets of H . This result can be derived from the Main Theorem (and some intermediate assertions), using the methods of [KS2]. Since the corresponding arguments differ from those used in this work, we shall present them in another publication. NOTATION
We denote by Z be the set of all integers and by Z0 be the set of non-positive integers. Let X be a topological space. We shall use the following notation. [B]X is the closure in the space X of its subset B. BX (x, r) is a closed ball in X of radius r centred at x ∈ X. B(X) is the σ -algebra of Borel subsets of X. P (X) is the set of probability measures on (X, B(X)). C(X) is the space of real-valued continuous functions on X. Cb (X) is the space of bounded functions f ∈ C(X). It is endowed with the supremum-norm f ∞ . L1 (X, µ) is the space of Borel functions on X with finite norm f µ := f (x) dµ(x). X
The integral of a function f (x) over the space X with respect to a measure µ will sometimes be denoted by (µ, f ): f (x) dµ(x) = f dµ. (µ, f ) = X
X
D(ξ ) is the distribution of a random variable ξ . a ∨ b (a ∧ b) is the maximum (minimum) of real numbers a and b. We denote by Ci , i = 1, 2, . . . , unessential positive constants.
152
SERGEI KUKSIN AND ARMEN SHIRIKYAN
1. Preliminaries: Equations, Estimates and the Markov Chain 1.1. DESCRIPTION OF THE CLASS