J. Math. Sci. Univ. Tokyo 2 (1995), 441–498. A construction of the fundamental solution for Schrödinger equations By Naoto Kumano-go Abstract. In this paper, applying the skip method in Fujiwara [3] to the theory of multi-products of Fourier integral operators in Kitada and H.Kumano-go [6], we give a construction of the fundamental solution for the Cauchy problem of a pseudo-differential equation of Schrödinger’s type. We regard this construction as a multi-product of Fourier integral operators, and investigate the pointwise convergence of the phase function and that of the symbol. Here we use neither the solution of the Hamilton-Jacobi equation in [6] nor the action of the classical orbit in [2],[4]. 0. Introduction Let Lh be a Schrödinger operator defined by Lh ≡ i∂t + Kh (t, X, Dx ) on [0, T ] with a parameter 0 < h < 1. For sufficiently small T0 (0 < T0 ≤ T ), let Uh (t, s) (0 ≤ s ≤ t ≤ T0 ) be the fundamental solution for the operator Lh such that (0.1) Lh Uh (t, s) = 0 on [s, T0 ] Uh (s, s) = I (0 ≤ s ≤ T0 ) . Noting that (0.2) Kh (t, X, Dx )u(x) ≡ eix·ξ Kh (t, x, ξ)û(ξ)d−ξ (u ∈ S, d−ξ ≡ (2π)−n dξ) , 1991 Mathematics Subject Classification. 58G15. 441 Primary 35S05; Secondary 58D30, 47G30, 442 Naoto Kumano-go one may ask whether Uh (t, s)u(x) = eix·ξ ei t s Kh (τ,x,ξ)dτ û(ξ)d−ξ ? Of course, the answer is “ No ”. However, this simple idea is our starting point. In what follows, we set t i st Kh (τ )dτ (X, Dx )u(x) ≡ eix·ξ ei s Kh (τ,x,ξ)dτ û(ξ)d−ξ (u ∈ S) . (0.3) e By the result of Kitada and H.Kumano-go [6], we have the following: R 2n with a parameter Let Hh (t, x, ξ) be a real-valued function on [0, T ]×R R 2n 0 < h < 1, which has a continuous derivative ∂ξα Dxβ Hh (t, x, ξ) on [0, T ]×R for any α, β, satisfying (0.4) |∂ξα Dxβ {hρ−δ Hh (t, hδ x, h−ρ ξ)}| Cα,β x; ξ 2−|α+β| (|α + β| ≤ 1) ≤ Cα,β (|α + β| ≥ 2) , on [0, T ] × R 2n . Here Cα,β is a constant independent of a parameter 0 < h < 1. Let Wh (t, x, ξ) be a complex-valued function on [0, T ] × R 2n with a parameter 0 < h < 1, which has a continuous derivative ∂ξα Dxβ Wh (t, x, ξ) on [0, T ] × R 2n for any α, β, satisfying (0.5) |∂ξα Dxβ {Wh (t, hδ x, h−ρ ξ)}| ≤ Cα,β on [0, T ] × R 2n . Here Cα,β is a constant independent of a parameter 0 < h < 1. Set Kh (t, x, ξ) ≡ Hh (t, x, ξ) + Wh (t, x, ξ) (0.6) Lh ≡ i∂t + Kh (t, X, Dx ) . Then, using the solution φh (t, s)(x, ξ) of the Hamilton-Jacobi equation −∂t φh (t, s)(x, ξ) + Hh (t, x, ∇x φh (t, s)(x, ξ)) = 0 on [s, T0 ] × R 2n (0.7) φh (s, s)(x, ξ) = x · ξ , Fundamental solution for Schrödinger equations 443 we can construct Uh (t, s) from the Fourier integral operator I(φh (t, s)) with the phase function φh (t, s)(x, ξ) and the symbol 1, by the method of the successive approximation and by H.Kumano-go-Taniguchi theorem. Especially, in the case Wh (t, x, ξ) ≡ 0, we have Uh (t, s) = (0.8) lim |∆t,s |→0 I(φh (t, t1 ))I(φh (t1 , t2 )) · · · I(φh (tν , s)) , where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν ≥ tν+1 ≡ s (≥ 0) is an arbitrary division of interval [s,t] into subintervals, and |∆t,s | ≡ max |tj−1 − tj |. 1≤j≤ν+1 Now, in this case Wh (t, x, ξ) ≡ 0, by applying Fujiwara’s skip method in [3] to Kitada and H.Kumano-go’s theory, we can construct Uh (t, s) directly with (0.8). Here the following facts are important: (A) φh (t, θ)#φh (θ, s) = φh (t, s) (T0 ≥ t ≥ θ ≥ s ≥ 0) (B) There exist a constant Cl and rh (t, θ, s)(x, ξ) ∈ B0ρ,δ (h) (T0 ≥ t ≥ θ ≥ s ≥ 0) such that I(φh (t, θ))I(φh (θ, s)) = I(φh (t, s)) + rh (t, θ, s)(φh (t, s); X, Dx ) and (0) |rh (t, θ, s)|l ≤ Cl |t − θ||θ − s| . The arguments about the solution of the Hamilton-Jacobi equation in Kitada and H.Kumano-go [6], however, seem to be too technical. Then, can we construct Uh (t, s) in a simple style without the solution of the HamiltonJacobi equation ? On the other hand, we can rewrite (0.8) with Uh (t, s) = lim |∆t,s |→0 ei(φh (t,t1 )(x,ξ)−x·ξ) (X, Dx )ei(φh (t1 ,t2 )(x,ξ)−x·ξ) (X, Dx ) · · · ei(φh (tν ,s)(x,ξ)−x·ξ) (X, Dx ) . Furthermore, φh (tj−1 , tj )(x, ξ) − x · ξ is approximated to i tj−1 tj−1 tj Hh (τ, x, ξ)dτ . H (τ )dτ h (X, Dx ) ? Our answer Then, can we construct Uh (t, s) from e tj is to get the following expression for general cases including Wh (t, x, ξ) ≡ 0: (0.9) Uh (t, s) = lim i e t t1 Kh (τ )dτ |∆t,s |→0 i stν Kh (τ )dτ ···e i (X, Dx )e (X, Dx ) . t 1 t2 Kh (τ )dτ (X, Dx ) 444 Naoto Kumano-go Here we consider (A) and (B). If we replace φh (t, s)(x, ξ) by x · ξ + t s Hh (τ, x, ξ)dτ , then (A) does not hold any longer and (B) requires some improvement. Therefore, in order to get (0.9), we have to make the best of the theory of multi-products of phase functions in Kitada and H.Kumano-go [6]. Main Theorems The main theorems for a Schrödinger equation are Theorem 5.1, 5.3 and 5.4. For the details, see Section 5. Roughly speaking, Theorem 5.1, 5.3 and 5.4 are as follows: R 2n with a parameter Let Hh (t, x, ξ) be a real-valued function on [0, T ]×R R 2n 0 < h < 1, which has a continuous derivative ∂ξα Dxβ Hh (t, x, ξ) on [0, T ]×R for any α, β, satisfying (0.10) |∂ξα Dxβ {hρ−δ Hh (t, hδ x, h−ρ ξ)}| Cα,β x; ξ 2−|α+β| (|α + β| ≤ 1) ≤ Cα,β (|α + β| ≥ 2) , on [0, T ] × R 2n . Here Cα,β is a constant independent of a parameter 0 < h < 1. Let Wh (t, x, ξ) be a complex-valued function on [0, T ] × R 2n with a parameter 0 < h < 1, which has a continuous derivative ∂ξα Dxβ Wh (t, x, ξ) on [0, T ] × R 2n for any α, β, satisfying C0,0 x; ξ (α = β = 0) |∂ α Dxβ {Wh (t, hδ x, h−ρ ξ)}| ≤ ξ (|α + β| ≥ 1) Cα,β (0.11) δ −ρ Im{Wh (t, h x, h ξ)} ≥ −C0,0 , is a constant independent of a parameter 0 < on [0, T ] × R 2n . Here Cα,β h < 1. Set Kh (t, x, ξ) ≡ Hh (t, x, ξ) + Wh (t, x, ξ) (0.12) Lh ≡ i∂t + Kh (t, X, Dx ) . Furthermore, let T0 (0 < T0 ≤ T ) be sufficiently small. Fundamental solution for Schrödinger equations 445 Theorem 5.1. For any ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0), there exists a Fourier integral operator ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) with a phase function φh (∆t0 ,tν+1 )(x, ξ) and a symbol ph (∆t0 ,tν+1 )(x, ξ) such that (0.13) ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) i =e t 0 t1 i ···e Kh (τ )dτ tν tν+1 i (X, Dx )e Kh (τ )dτ t 1 t2 Kh (τ )dτ (X, Dx ) (X, Dx ) . Furthermore, the set {φh (∆t0 ,tν+1 )}∆t0 ,tν+1 is bounded in Bδ−ρ,2 ρ,δ (h), and the 0 set {ph (∆t0 ,tν+1 )}∆t0 ,tν+1 is bounded in Bρ,δ (h). Theorem 5.3. For any (T0 ≥) t ≥ s (≥ 0), there exist a phase function φh (t, s)(x, ξ) and a symbol ph (t, s)(x, ξ) such that φh (∆t,s ) uniformly converges to φh (t, s) in Bδ−ρ,2 ρ,δ (h) as |∆t,s | tends to 0, and ph (∆t,s ) uniformly converges to ph (t, s) 0,1 in Bρ,δ (h) as |∆t,s | tends to 0. Furthermore, it follows that (0.14) eiφh (t,s)(x,ξ)−ix ·ξ ph (t, s)(x, ξ) ν+1 ··· exp i(xj−1 − xj ) · ξ j = lim Os — |∆t,s |→0 tj−1 +i Kh (τ, xj−1 , ξ j )dτ j=1 dx1 d−ξ 1 · · · dxν d−ξ ν , tj where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0), x0 ≡ x, xν+1 ≡ x and ξ ν+1 ≡ ξ. Theorem 5.4. The Fourier integral operator Uh (t, s) ≡ ph (t, s)(φh (t, s); X, Dx ) with the phase function φh (t, s)(x, ξ) and the symbol ph (t, s)(x, ξ) in Theorem 5.3, has the following form: (0.15) Uh (t, s) = lim i e t t1 Kh (τ )dτ |∆t,s |→0 i stν Kh (τ )dτ ···e i (X, Dx )e (X, Dx ) , t 1 t2 Kh (τ )dτ (X, Dx ) 446 Naoto Kumano-go where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν ≥ tν+1 ≡ s (≥ 0). Furthermore, Uh (t, s) is the fundamental solution for the operator Lh . An example of a Schrödinger operator Lh : n h∂ h∂ 1 xj xk aj,k (t) Lh ≡ i∂t + h i i j,k=1 h∂ xk + cj,k (t)xj xk + bj,k (t)xj i n h∂ xj + bj (t)xj + c(t, x) aj (t) + i Remark. (0.17) j=1 + d(t, x). Here aj,k (t), bj,k (t), cj,k (t), aj (t) and bj (t) are real-valued continuous functions on [0, T ]. c(t, x) is a real-valued function on [0, T ] × R 2n which has a continuous and bounded derivative Dxβ c(t, x) on [0, T ] × R 2n for any β. d(t, x) is a complex-valued function on [0, T ] × R 2n , which has a continuous and bounded derivative Dxβ d(t, x) on [0, T ] × R 2n for any β = 0, and whose imaginary part Im{d(t, x)} is lower bounded on [0, T ] × R 2n . The Plan of This Paper The plan of this paper is as follows. m,λ Section 1. We introduce the classes of symbols Bm ρ,δ (h), Bρ,δ (h) and m,λ the families of pseudo-differential operators B m ρ,δ (h), B ρ,δ (h) in Kitada and H.Kumano-go [6]. Section 2. We define the class of phase functions Pρ,δ (τ, h, {κl }∞ l=0 ) and investigate the properties of the multi-products of phase functions in more detail. It is important that the multi-product of phase functions (φ1,h #φ2,h # · · · #φν+1,h )(x, ξ) is ‘ close ’ to x · ξ + ν+1 j=1 Jj,h (x, ξ) where Jj,h (x, ξ) ≡ φj,h (x, ξ) − x · ξ. This section plays an essential role to prove the pointwise convergence of the phase function in Section 5. Section 3. We introduce the families of Fourier integral operators m,λ Bm ρ,δ (φh ), B ρ,δ (φh ) in Kitada and H.Kumano-go [6]. Fundamental solution for Schrödinger equations 447 Section 4. We apply Fujiwara’s skip method in [3] to Kitada and H.Kumano-go’s theory of multi-products of Fourier integral operators, and investigate some properties. This section plays an essential role to prove the convergence of the symbol in Section 5. Section 5. We give a construction of the fundamental solution for the Cauchy problem of a pseudo-differential equation of Schrödinger’s type. We regard this construction as a multi-product of Fourier integral operators, and investigate the pointwise convergence of the phase function and that of the symbol. Acknowledgements. I would like to express my sincere gratitude to Professor K. Kataoka, Professor H. Komatsu, Professor D. Fujiwara, Professor K. Taniguchi, and Professor K. Yajima for helpful discussions and suggestions. It was very happy that I could learn Fujiwara’s skip method from Professor D. Fujiwara, and H.Kumano-go-Taniguchi theorem from Professor K. Taniguchi. Moreover, Lemma 4.7 is a revised version of that in Fujiwara [4]. Notation For x = (x1 , . . . , xn ) ∈ R nx , ξ = (ξ1 , . . . , ξn ) ∈ R nξ and multi-indices of non-negative integers α = (α1 , . . . , αn ), β = (β1 , . . . , βn ), we employ the usual notation: |α| = α1 + · · · + αn , |β| = β1 + · · · + βn , x · ξ = x1 ξ1 + · · · + xn ξn , x = (1 + |x|2 )1/2 , x; ξ = (1 + |x|2 + |ξ|2 )1/2 , ∂ξj = ∂ ∂ , Dxj = −i , ∂ξα = ∂ξα11 · · · ∂ξαnn , Dxβ = Dxβ11 · · · Dxβnn . ∂ξj ∂xj S denotes the Schwartz space of rapidly decreasing C ∞ -functions on R n . R 2n ) denotes the set of C ∞ -functions on R 2n whose derivatives are all B(R R 2n ) denotes the set of C ∞ -functions g(x, ξ) on R 2n whose bounded. B 1,∞ (R R 2n ) derivatives ∂ξα Dxβ g(x, ξ) are bounded for |α+β| ≥ 1. For u ∈ S, f ∈ B(R R 2n ), we define semi-norms |u|l,S , |f |l,B and |g|l,B1,∞ (l = and g ∈ B1,∞ (R 0, 1, 2, . . . ), respectively by |u|l,S ≡ max sup |x k ∂xα u(x)| (l = 0, 1, 2, . . . ) , k+|α|≤l x 448 Naoto Kumano-go |f |l,B ≡ max sup |∂ξα Dxβ f (x, ξ)| (l = 0, 1, 2, . . . ) , |α+β|≤l x,ξ and |g|l,B1,∞ ≡ sup |g(x, ξ)|/x; ξ (l = 0) x,ξ sup |g(x, ξ)|/x; ξ + x,ξ max sup |∂ξα Dxβ g(x, ξ)| 1≤|α+β|≤l x,ξ (l = 1, 2, . . . ) . R 2n ) are Fréchet spaces with these semi-norms. R 2n ) and B 1,∞ (R Then, S, B(R The Fourier transform û(ξ) ≡ F[u](ξ) and the inverse Fourier transform F[v](x) are defined respectively by û(ξ) ≡ F[u](ξ) ≡ e−ix·ξ u(x)dx (u ∈ S) F[v](x) ≡ eix·ξ v(ξ)d−ξ (v ∈ S), where d−ξ ≡ (2π)−n dξ. Os — See [7]. 1. ·dyd−η denotes the usual oscillatory integrals. Pseudo-Differential Operators In this section, we introduce the basic properties of the pseudodifferential operators defined by Kitada and H.Kumano-go in [6]. For the details, see [6]. In what follows, for simplicity, let n0 ≡ 2[n/2 + 1]. Definition 1.1 ( A Family of Pseudo-Differential Operators B m ρ,δ (h) ). (1) We say that a family {ph }0<h<1 of C ∞ -functions ph ≡ ph (x, ξ, x , ξ , x ) belongs to the class {Bm ρ,δ (h)}0<h<1 (m ∈ R , 0 ≤ δ ≤ ρ ≤ 1) of symbols, if ph satisfies (1.1) sup sup 0<h<1 x,ξ,x ,ξ ,x |∂ξα Dxβ ∂ξα Dxβ Dxβ ph (x, ξ, x , ξ , x )| hm+ρ|α+α |−δ|β+β +β | < ∞, for any α, β, α , β , β . We write {ph }0<h<1 ∈ {Bm ρ,δ (h)}0<h<1 , or (h). simply ph ∈ Bm ρ,δ Fundamental solution for Schrödinger equations 449 (2) For ph (x, ξ, x , ξ , x ) ∈ Bm ρ,δ (h), we define a family {Ph }0<h<1 of pseudo-differential operators Ph ≡ ph (X, Dx , X , Dx , X ) by (1.2) ph (X, Dx , X , Dx , X )u(x) 1 1 2 2 e−i(y ·η +y ·η ) ph (x, η 1 , x + y 1 , η 2 , x + y 1 + y 2 ) ≡ Os — × u(x + y 1 + y 2 )dy 2 d−η 2 dy 1 d−η 1 (u ∈ S) . m Bm We write {Ph }0<h<1 ∈ {B ρ,δ (h)}0<h<1 , or simply Ph ∈ B ρ,δ (h). Remark. (m) (l = 0, 1, 2, . . . ) by 1◦ . For ph ∈ Bm ρ,δ (h), we define semi-norms |ph |l (1.3) (m) |ph |l ≡ max sup x,ξ,x ,ξ ,x 2◦ . (1.4) sup |α+β+α +β +β |≤l 0<h<1 |∂ξα Dxβ ∂ξα Dxβ Dxβ ph (x, ξ, x , ξ , x )| hm+ρ|α+α |−δ|β+β +β | . Then Bm ρ,δ (h) is a Fréchet space with these semi-norms. For ph (x, ξ, x , ξ , x ) ∈ Bm ph by ρ,δ (h), we define ph (x, ξ, x , ξ , x ) ≡ h−m ph (hδ x, h−ρ ξ, hδ x , h−ρ ξ , hδ x ) . Then we have ph ∈ B00,0 (h), (1.5) and (1.6) (0) (m) | ph |l ( in B00,0 (h)) = |ph |l ( in Bm ρ,δ (h)) . 3◦ . Symbols ph (x, ξ), ph (ξ, x ) ∈ Bm ρ,δ (h) are often called single symbols. ◦ 4 . For symbols ph (x, ξ), ph (ξ, x ), ph (x, ξ, x ) ∈ Bm ρ,δ (h), we have the representation formulae: ph (X, Dx )u(x) = eix·ξ ph (x, ξ)û(ξ)d−ξ (u ∈ S) , (1.7) F[ph (Dx , X )u](ξ) = e−ix ·ξ ph (ξ, x )u(x )dx (u ∈ S) , (1.8) (1.9) ph (X, Dx , X )u(x) ei(x−x )·ξ ph (x, ξ, x )u(x )dx d−ξ (u ∈ S) . = Os — 450 Naoto Kumano-go Definition 1.2 ( A Family of Pseudo-Differential Operators B m,λ ρ,δ (h) ). (1) We say that a family {ph }0<h<1 of C ∞ -functions ph ≡ ph (x, ξ) belongs to the class {Bm,λ ρ,δ (h)}0<h<1 (m, λ ∈ R , 0 ≤ δ ≤ ρ ≤ 1) of −δ symbols, if ph (x, ξ)h x; hρ ξ −λ belongs to Bm ρ,δ (h). m,λ We write {ph }0<h<1 ∈ {Bm,λ ρ,δ (h)}0<h<1 , or simply ph ∈ Bρ,δ (h). (2) For ph (x, ξ) ∈ Bm,λ ρ,δ (h), we define a family {Ph }0<h<1 of pseudodifferential operators Ph ≡ ph (X, Dx ) by ph (X, Dx )u(x) ≡ (1.10) eix·ξ ph (x, ξ)û(ξ)d−ξ (u ∈ S) . m,λ B m,λ We write {Ph }0<h<1 ∈ {B ρ,δ (h)}0<h<1 , or simply Ph ∈ B ρ,δ (h). Remark. 1◦ . For ph ∈ Bm,λ ρ,δ (h), we define semi-norms |ph |l (m,λ) (1.11) (m,λ) |ph |l ≡ max sup sup (l = 0, 1, 2, . . . ) by |∂ξα Dxβ {ph (x, ξ)h−δ x; hρ ξ −λ }| hm+ρ|α|−δ|β| |α+β|≤l 0<h<1 x,ξ Then Bm,λ ρ,δ (h) is a Fréchet space with these semi-norms. 2◦ . For ph (x, ξ) ∈ Bm,λ ph by ρ,δ (h), we define ph (x, ξ) ≡ h−m ph (hδ x, h−ρ ξ) . (1.12) Then we have ph ∈ B0,λ 0,0 (h) , (1.13) and (1.14) (0,λ) | ph |l (m,λ) ( in B0,λ 0,0 (h)) = |ph |l ( in Bm,λ ρ,δ (h)) . . Fundamental solution for Schrödinger equations 451 Proposition 1.3 ( Continuity on S ). (1) There exist a constant Mm,h,l (depending only on m, h) and integers l , l such that (1.15) |ph (X, Dx , X , Dx , X )u|l,S ≤ Mm,h,l |ph |l |u|l ,S (m) (ph (x, ξ, x , ξ , x ) ∈ Bm ρ,δ (h), u ∈ S) . (2) There exist a constant Mm,λ,h,l (depending only on m, λ, h) and integers lλ , lλ (depending only on λ) such that (m,λ) |ph (X, Dx )u|l,S ≤ Mm,λ,h,l |ph |l (1.16) λ |u|lλ ,S (ph (x, ξ) ∈ Bm,λ ρ,δ (h), u ∈ S) . Proof. See [6]. Proposition 1.4 ( Multi-products ). There exist constants A and Cl , satisfying the following: m For pj,h (x, ξ, x ) ∈ Bρ,δj (h) (ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1), define qν+1,h (x, ξ, x ) by (1.17) qν+1,h (x, ξ, x ) ≡ Os — ··· exp(−i ν yj · ηj ) j=1 × ν pj,h (x + ȳ j−1 , ξ + η j , x + ȳ j ) j=1 · pν+1,h (x + ȳ ν , ξ, x )dy ν d−η ν · · · dy 1 d−η 1 (ȳ 0 ≡ 0, ȳ j ≡ y 1 + · · · + y j , j = 1, 2, . . . , ν) . Then it follows that (1.18) (1.19) qν+1,h (x, ξ, x ) ∈ Bρ,δν+1 (h) , m̄ qν+1,h (X, Dx , X ) = p1,h (X, Dx , X )p2,h (X, Dx , X ) · · · pν+1,h (X, Dx , X ) , 452 Naoto Kumano-go and (1.20) (m̄ ) |qν+1,h |l ν+1 ≤ Cl A ν+1 ν+1 max l1 +l2 +···+lν+1 ≤l (m ) j |pj,h |lj +3n , 0 j=1 where m̄ν+1 ≡ m1 + m2 + · · · + mν+1 . Proof. See [6]. Proposition 1.5 ( Simplified Symbols ). There exists a constant Cl satisfying the following: For ph (x, ξ, x , ξ , x ) ∈ Bm ρ,δ (h), define ph,L (x, ξ, x ) and ph,R (x, ξ, x ), respectively by −iy·η ph,L (x, ξ, x ) ≡ Os — e ph (x, ξ + η, x + y, ξ, x )dyd−η −iy·η (1.21) e ph (x, ξ, x + y, ξ − η, x )dyd−η . ph,R (x, ξ, x ) ≡ Os — Then it follows that (1.22) (1.23) ph,L (x, ξ, x ), ph,R (x, ξ, x ) ∈ Bm ρ,δ (h) , ph (X, Dx , X , Dx , X ) = ph,L (X, Dx , X ) = ph,R (X, Dx , X ) , and (1.24) (m) |ph,L |l (m) , |ph,R |l (m) ≤ Cl |ph |l+2n0 . Proof. See [6]. Proposition 1.6 ( The Inverse of (I − ph (X, Dx , X )) ). Let the constant A be the same as that in Proposition 1.4. There exists a constant Cl satisfying the following: (0) If ph (x, ξ, x ) ∈ B0ρ,δ (h) satisfies |ph |3n0 ≤ 1/(2A), there exists qh (x, ξ, x ) ∈ B0ρ,δ (h) such that (1.25) I = qh (X, Dx , X )(I − ph (X, Dx , X )) = (I − ph (X, Dx , X ))qh (X, Dx , X ) , Fundamental solution for Schrödinger equations 453 and (0) |qh |l (1.26) (0) ≤ Cl {max(1, |ph |l+3n0 )}l . Proof. See [6]. Proposition 1.7 ( L 2 -Boundedness ). There exists a constant C such that (1.27) ph (X, Dx , X )uL 2 ≤ Chm |ph |3n0 uL 2 (m) (ph (x, ξ, x ) ∈ Bm ρ,δ (h), u ∈ S) . Proof. See [6] and [7]. Definition 1.8 ( Hilbert Spaces, H 0,h , H 2,h ). (1) We define the Hilbert space H 0,h as the completion of S with respect to the norm u0,h ≡ hρ−δ uL 2 . (1.28) (2) We define the Hilbert space H 2,h as the completion of S with respect to the norm (1.29) u2,h ≡ (h −δ 1/2 α ρ β x) (h Dx ) u(x)L2 2 . |α+β|≤2 2. Phase Functions In this section, we define phase functions and investigate the properties of multi-products of phase functions in more detail. This section plays an essential role to prove the pointwise convergence of the phase function in Section 5. The definition of phase functions below is a little different from that in Kitada and H.Kumano-go [6]. This definition is essential in this paper. 454 Naoto Kumano-go Definition 2.1 ( A Family of Phase Functions Pρ,δ (τ, h, {κl }∞ l=0 ) ). For τ ≥ 0 and a sequence of positive constants {κl }∞ (κ = 1, κl ≤ l=0 0 ∞ κl+1 ), we say that a family {φh }0<h<1 of real-valued C -functions φh (x, ξ) belongs to the class {Pρ,δ (τ, h, {κl }∞ l=0 )}0<h<1 of phase functions, if Jh (x, ξ) ≡ φh (x, ξ) − x · ξ satisfies (2.1) sup 0<h<1 sup |α+β|≤1 + |∂ξα Dxβ {hρ−δ Jh (hδ x, h−ρ ξ)}| x; ξ 2−|α+β| x,ξ 2≤|α+β|≤2+l sup |∂ξα Dxβ {hρ−δ Jh (hδ x, h−ρ ξ)}| x,ξ ≤ κl τ , for any l = 0, 1, 2, . . . . We write {φh }0<h<1 ∈ {Pρ,δ (τ, h, {κl }∞ l=0 )}0<h<1 , or simply φh ∈ ∞ Pρ,δ (τ, h, {κl }l=0 ). Furthermore, for simplicity, we sometimes omit {κl }∞ l=0 and write simply Pρ,δ (τ, h). Remark. For φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), we define φh and Jh respectively by (2.2) φh (x, ξ) ≡ hρ−δ φh (hδ x, h−ρ ξ) h (x, ξ) ≡ hρ−δ Jh (hδ x, h−ρ ξ) . J Then we have (2.3) h ∈ P0,0 (τ, h, {κ}∞ ) . φ l=0 Proposition 2.2. ∞ τj < 1/2. Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and j=1 Then, for any ν = 1, 2, 3, . . . , 0 < h < 1 and (x, ξ) ∈ R 2n , the solution j j {Xν,h , Ξν,h }νj=1 (x, ξ) (∈ R 2νn ) of the equation (2.4) j j−1 j , Ξν,h ) X = ∇ξ φj,h (Xν,h ν,h j j j+1 = ∇x φj+1,h (Xν,h , Ξν,h ) Ξν,h 0 ≡ x, Ξ ν+1 ≡ ξ) , (j = 1, . . . , ν, Xν,h ν,h Fundamental solution for Schrödinger equations j ,Ξ j }ν (x, ξ) (∈ R 2νn ) of the equation and the solution {X ν,h ν,h j=1 j j,h (X j−1 , Ξ j ) X = ∇ξ φ ν,h ν,h ν,h j j j+1 ) Ξν,h = ∇x φj+1,h (Xν,h , Ξ ν,h 0 ν+1 ≡ ξ) , (j = 1, . . . , ν, Xν,h ≡ x, Ξ ν,h (2.5) exist uniquely. Furthermore, it follows that j j (x, ξ) ∈ C ∞ (R R 2n ) (j = 0, 1, . . . , ν) Xν,h (x, ξ), X ν,h (2.6) j j (x, ξ) ∈ C ∞ (R R 2n ) (j = 1, 2, . . . , ν + 1) , (x, ξ), Ξ Ξν,h ν,h and (2.7) j (x, ξ) = h−δ X j (hδ x, h−ρ ξ) (j = 0, 1, . . . , ν) X ν,h ν,h j (x, ξ) = hρ Ξ j (hδ x, h−ρ ξ) (j = 1, 2, . . . , ν + 1) . Ξ ν,h ν,h Proof. See [6]. Definition 2.3 ( Multi-products of Phase Functions ). ∞ τj < 1/2 . Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and j=1 For ν = 1, 2, . . . , we define the #-(ν + 1) products (2.8) Φν+1,h ≡ φ1,h #φ2,h # · · · #φν+1,h 1,h #φ 2,h # · · · #φ ν+1,h , ν+1,h ≡ φ Φ respectively by (2.9) ν+1 ν j−1 j j j Φν+1,h (x, ξ) ≡ j=1 φj,h (Xν,h , Ξν,h ) − j=1 Xν,h · Ξν,h ν+1 ν j,h (X ν+1,h (x, ξ) ≡ φ j ·Ξ j−1 , Ξ j ) − X j . Φ ν,h ν,h ν,h ν,h j=1 j=1 Remark. Set Jj,h (x, ξ) ≡ φj,h (x, ξ) − x · ξ (j = 1, 2, . . . , ν + 1) (2.10) j,h (x, ξ) − x · ξ (j = 1, 2, . . . , ν + 1) , j,h (x, ξ) ≡ φ J 455 456 Naoto Kumano-go and (2.11) J ν+1,h (x, ξ) ≡ Φν+1,h (x, ξ) − x · ξ ν+1,h (x, ξ) − x · ξ . J ν+1,h (x, ξ) ≡ Φ Then we get (2.12) ν+1 ν+1,h (x, ξ) = J j,h (X j−1 , Ξ j ) J ν,h ν,h j=1 ν j −X j−1 ) · (Ξ j − ξ) (X − ν,h ν,h ν,h j=1 ν+1 j,h )(X j−1 , Ξ j ). ∇(x,ξ)J ν+1,h (x, ξ) = (∇(x,ξ) J ν,h j=1 ν,h Furthermore, using (2.7), we have (2.13) ν+1,h (x, ξ) = hρ−δ Φν+1,h (hδ x, h−ρ ξ) . Φ Proposition 2.4. (1) Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and ∞ τj ≤ 1/4. j=1 Then it follows that (2.14) j j−1 |, |Ξ j−1 − Ξ j | ≤ 2τj x; ξ |Xν,h − X ν,h ν,h ν,h j−1 −1 2 x; ξ ≤ x + θ(X j ν,h − x); ξ + θ(Ξν,h − ξ) ≤ 2x; ξ j − ξ))| ≤ |f |0,B j−1 − x), ξ + θ(Ξ |f (x + θ(X ν,h ν,h j−1 − x), ξ + θ(Ξ j − ξ))| ≤ 2|g| 1,∞ x; ξ |g(x + θ(X 0,B ν,h ν,h ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1, |θ| ≤ 1, 0 < h < 1, . ν+1 ≡ X ν , Ξ 0 ≡ Ξ 1 R 2n ), X R 2n ), g(x, ξ) ∈ B1,∞ (R f (x, ξ) ∈ B(R ν,h ν,h ν,h ν,h (2) There exists a constant c0 satisfying the following: ∞ τj ≤ 1/4. Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and j=1 Fundamental solution for Schrödinger equations 457 Then it follows that → → j −X j−1 )|, |∇x (Ξ j−1 − Ξ j )| ≤ c τj ∇ ( X | x 0 ν,h ν,h ν,h ν,h → → j j−1 )|, |∇ξ (Ξ j−1 − Ξ j )| ≤ c τj |∇ξ (Xν,h − X 0 ν,h ν,h ν,h (ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1, 0 < h < 1, ν+1 ≡ X ν , Ξ 0 ≡ Ξ 1 ). X ν,h ν,h ν,h ν,h (2.15) (3) For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants cl (depending only on κ1 , κ2 , . . . , κl−1 , κl ) and cl−1 , c l−1 (depending only on κ1 , κ2 , . . . , κl−1 ), satisfying the following: ∞ ) (j = 1, 2, . . . ) and τj ≤ 1/4. Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 j=1 Then it follows that (2.16) → → j − X j−1 )|, |∂ α Dxβ ∇x (Ξ j−1 − Ξ j )| ≤ c |∂ α Dxβ ∇x (X ξ ξ ν,h ν,h ν,h ν,h |α+β| τj → → j − X j−1 )|, |∂ α Dxβ ∇ξ (Ξ j−1 − Ξ j )| ≤ c |∂ξα Dxβ ∇ξ (X ξ ν,h ν,h ν,h ν,h |α+β| τj |∂ α Dβ f (x + θ(X j−1 − x), ξ + θ(Ξ j − ξ))| ≤ c |f | x ξ ν,h ν,h |α+β|−1 |α+β|,B β j−1 − x), ξ + θ(Ξ j − ξ))| ≤ c |∂ξα Dx g(x + θ(X ν,h ν,h |α+β|−1 |g||α+β|,B1,∞ ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1, |θ| ≤ 1, |α + β| ≥ 1, 0 < h < 1, . f (x, ξ) ∈ B(R ν+1 ≡ X ν , Ξ 0 ≡ Ξ 1 R 2n ), X R 2n ), g(x, ξ) ∈ B1,∞ (R ν,h ν,h ν,h ν,h Proof. By induction. See [6]. Proposition 2.5. For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist a constant c0 > 1(inde∞ pendent of {κl }l=0 ) and a sequence of positive constants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl ) (depending only on κ1 , κ2 , . . . , κl ), satisfying the following: ∞ (1) Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ )(j = 1, 2, . . . ) and τj ≤ 1/4. l=0 j=1 Then, for any ν = 1, 2, . . . , it follows that (2.17) Φν+1,h ≡ (φ1,h #φ2,h # · · · #φν+1,h ) ∈ Pρ,δ (c0 τ̄ν+1 , h, {κl }∞ l=0 ) 1,h #φ 2,h # · · · #φ ν+1,h ) ∈ P0,0 (c0 τ̄ν+1 , h, {κ }∞ ) , ν+1,h ≡ (φ Φ l l=0 458 Naoto Kumano-go where τ̄ν+1 ≡ τ1 + τ2 + · · · + τν+1 . (2) Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 )(j = 1, 2, . . . ) and ∞ τj ≤ 1/(4c0 ). j=1 Then, for any ν = 1, 2, . . . , it follows that Φν+2,h ≡ (φ1,h #φ2,h # · · · #φν+1,h #φν+2,h ) (2.18) = (φ1,h #φ2,h # · · · #φν+1,h )#φν+2,h = φ1,h #(φ2,h # · · · #φν+1,h #φν+2,h ) . Furthermore, if {Xh , Ξh }(x, ξ) is the solution of the equations (2.19) Xh = ∇ξ Φν+1,h (x, Ξh ) Ξh = ∇x φν+2,h (Xh , ξ) , then it follows that j (x, ξ) Xν+1,h (2.20) = j Xν,h (x, Ξh ) (j = 1, 2, . . . , ν) Xh (x, ξ) (j = ν + 1) , and (2.21) j (x, ξ) Ξν+1,h = j Ξν,h (x, Ξh ) (j = 1, 2, . . . , ν) Ξh (x, ξ) (j = ν + 1) . Proof. See [6]. Lemma 2.6. (1) Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and ∞ τj ≤ 1/4. j=1 Then it follows that (2.22) j − ξ) + θ η ±1 j−1 − x) + θ y; ξ + θ(Ξ x + θ(X ν,h ν,h ≤ 4y; η x; ξ ±1 (ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1, |θ|, |θ |, |θ | ≤ 1, 0 < h < 1) . Fundamental solution for Schrödinger equations 459 (2) For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), λ ∈ R , α and β , there exists a constant Cλ,α ,β ,l−1 (depending only on λ, α , β , κ1 , κ2 , . . . , κl−1 ) satisfying the following: ∞ Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ τj ≤ 1/4. l=0 ) (j = 1, 2, . . . ) and j=1 Then it follows that j−1 − x) + θ y; ξ + θ(Ξ j − ξ) + θ η λ | (2.23) |∂ξα Dxβ ∂ηα Dyβ x + θ(X ν,h ν,h ≤ Cλ,α ,β ,|α+β|−1 y; η |λ| x; ξ λ (ν = 1, 2, . . . , j = 1, 2, . . . , ν + 1, |θ|, |θ |, |θ | ≤ 1, 0 < h < 1) . Proof. By induction. Note Proposition 2.4. 3 j=1 For φj,h ∈ Pρ,δ (τj , h)(j = 1, 3) and φ2,ϑ,h ∈ Pρ,δ (τ2 , h)(0 ≤ ϑ ≤ 1) with j ,Ξ j }2 (x, ξ) be the solution of the equations τj < 1/2, let {X 2,ϑ,h 2,ϑ,h j=1 (2.24) 1 1,h (X 0 ,Ξ 1 ), X2,ϑ,h = ∇ξ φ 2,ϑ,h 2,ϑ,h 1 1 2,ϑ,h (X 2 = ∇x φ Ξ 2,ϑ,h , Ξ2,ϑ,h ) 2,ϑ,h 1 2 2 X 2,ϑ,h = ∇ξ φ2,ϑ,h (X2,ϑ,h , Ξ2,ϑ,h ), 2 3 2 Ξ 2,ϑ,h = ∇x φ3,h (X2,ϑ,h , Ξ2,ϑ,h ) 0 3 (X 2,ϑ,h ≡ x, Ξ2,ϑ,h ≡ ξ, 0 ≤ ϑ ≤ 1) . 2,ϑ,h (x, ξ) be continuously differentiable on 0 ≤ ϑ ≤ 1. Furthermore, let φ Set (2.25) 2,ϑ,h (x, ξ) − x · ξ J2,ϑ,h (x, ξ) ≡ φ 1,h #φ 2,ϑ,h #φ 3,h )(x, ξ) − x · ξ . J 3,ϑ,h (x, ξ) ≡ (φ Then we get the following lemma. 460 Naoto Kumano-go Lemma 2.7. ∞ For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ) and {χl }l=0 (0 ≤ χl ≤ χl+1 ), there exist a constant Cl and a sequence of positive constants {χl }∞ l=0 (0 ≤ χl ≤ χl+1 , χl ≤ χl ) (depending only on κ1 , κ2 , . . . , κl and χ0 , χ1 , . . . , χl ), satisfying the following: ∞ Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 )(j = 1, 3), φ2,ϑ,h ∈ Pρ,δ (τ2 , h, {κl }l=0 )(0 ≤ 3 2,ϑ,h (x, ξ) satisfy ϑ ≤ 1) and τj ≤ 1/4, and let ∂ϑ J j=1 (2.26) 2 2 |∂ϑ J2,ϑ,h (x, ξ)| ≤ χ0 (τ2 ) x; ξ 2,ϑ,h (x, ξ)| ≤ χ|α+β| (τ2 )2 x; ξ |∂ξα Dxβ ∇(x,ξ) ∂ϑ J (0 < h < 1, 0 ≤ ϑ ≤ 1) . Then it follows that (2.27) j |, |∂ α Dxβ ∂ϑ Ξ j | ≤ C|α+β| (τ2 )2 x; ξ |∂ξα Dxβ ∂ϑ X ξ 2,ϑ,h 2,ϑ,h (j = 1, 2, 0 < h < 1, 0 ≤ ϑ ≤ 1) , and (2.28) (x, ξ)| ≤ χ0 (τ2 )2 x; ξ 2 |∂ J ϑ 3,ϑ,h |∂ξα Dxβ ∇(x,ξ) ∂ϑJ 3,ϑ,h (x, ξ)| ≤ χ|α+β| (τ2 )2 x; ξ (0 < h < 1, 0 ≤ ϑ ≤ 1) . Proof. By induction. Note Proposition 2.4 and Lemma 2.6. Let φj,h ∈ Pρ,δ (τj , h)(j = 1, 2, . . . ) and ∞ τj < 1/2. j=1 For ν = 1, 2, . . . and 0 ≤ ϑ ≤ 1, set ν+1 (x, ξ) ≡ ϑΦ (x, ξ) + (1 − ϑ) x · ξ + J (x, ξ) Φ ν+1,ϑ,h ν+1,h j,h j=1 (2.29) ν+1 j,h (x, ξ) , ν+1,h (x, ξ) + (1 − ϑ) x · ξ + J ν+1,ϑ,h (x, ξ) ≡ ϑΦ Φ j=1 and (2.30) J ν+1,ϑ,h (x, ξ) ≡ Φν+1,ϑ,h (x, ξ) − x · ξ ν+1,ϑ,h (x, ξ) − x · ξ . J ν+1,ϑ,h (x, ξ) ≡ Φ Fundamental solution for Schrödinger equations 461 Then we have (2.31) Φν+1,1,h (x, ξ) = Φν+1,h (x, ξ), ν+1 (x, ξ) = x · ξ + Jj,h (x, ξ) Φ ν+1,0,h j=1 ν+1,h (x, ξ), ν+1,1,h (x, ξ) = Φ Φ ν+1 j,h (x, ξ) , ν+1,0,h (x, ξ) = x · ξ + J Φ j=1 and (2.32) ν+1,ϑ,h (x, ξ) = hρ−δ Φν+1,ϑ,h (hδ x, h−ρ ξ) . Φ Furthermore, we get the following lemma. Lemma 2.8. For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist a constant c0 > 1(inde ∞ pendent of {κl }∞ l=0 ) and sequences of positive constants {κl }l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl ), {χl }∞ l=0 (0 ≤ χl ≤ χl+1 )(depending only on κ1 , κ2 , . . . , κl ), satisfying the following: ∞ )(j = 1, 2, . . . ) and τj ≤ 1/4. Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 j=1 Then it follows that (2.33) Φν+1,ϑ,h ∈ Pρ,δ (c0 τ̄ν+1 , h, {κl }∞ l=0 ) ν+1,ϑ,h ∈ P0,0 (c0 τ̄ν+1 , h, {κ }∞ ) , Φ l l=0 and (2.34) 2 2 |∂ϑJ ν+1,ϑ,h (x, ξ)| ≤ χ0 (τ̄ν+1 ) x; ξ |∂ξα Dxβ ∇(x,ξ) ∂ϑJ ν+1,ϑ,h (x, ξ)| ≤ χ|α+β| (τ̄ν+1 )2 x; ξ (ν = 1, 2, . . . , 0 ≤ ϑ ≤ 1, 0 < h < 1) , where τ̄ν+1 ≡ τ1 + τ2 + · · · + τν+1 . 462 Naoto Kumano-go Proof. From (2.12), we have (2.35) ∂ϑJ ν+1,ϑ,h (x, ξ) = − + ν+1 1 → 0 j=1 ν j −X j−1 ) · (Ξ j − ξ) (X ν,h ν,h ν,h j=1 j,h )(x + θ(X j−1 − x), ξ + θ(Ξ j − ξ))dθ (∇(x,ξ) J ν,h ν,h j−1 Xν,h − x · j − ξ , Ξ ν,h and (2.36) ∇(x,ξ) ∂ϑJ ν+1,ϑ,h (x, ξ) ν+1 1 → j,h )(x + θ(X j−1 − x), ξ + θ(Ξ j − ξ))dθ = (∇(x,ξ) ∇(x,ξ) J ν,h ν,h j=1 0 j−1 Xν,h − x · j − ξ . Ξ ν,h Using Proposition 2.4, we get (2.34). 3. Fourier Integral Operators In this section, we introduce the basic properties of the Fourier integral operators defined by Kitada and H.Kumano-go in [6]. For the details, see [6]. In what follows, for simplicity, let n0 ≡ 2[n/2 + 1]. m ∗ Definition 3.1 ( Fourier Integral Operators B m ρ,δ (φh ), B ρ,δ (φh ) ). (1) For 0 ≤ τ < 1, φh ∈ Pρ,δ (τ, h) and ph (x, ξ) ∈ Bm ρ,δ (h), we define a family of Fourier integral operators ph (φh ; X, Dx ) by (3.1) ph (φh ; X, Dx )u(x) = eiφh (x,ξ) ph (x, ξ)û(ξ)d−ξ (u ∈ S). Bm We write {ph (φh ; X, Dx )}0<h<1 ∈ {B ρ,δ (φh )}0<h<1 , or simply m ph (φh ; X, Dx ) ∈ B ρ,δ (φh ). We sometimes use this definition for 0 ≤ τ < ∞. ( See Theorem 4.1.) Fundamental solution for Schrödinger equations 463 (2) For 0 ≤ τ < 1, φh ∈ Pρ,δ (τ, h) and qh (ξ, x ) ∈ Bm ρ,δ (h), we define a family of conjugate Fourier integral operators qh (φ∗h ; Dx , X ) by ∗ (3.2) F[qh (φ ; Dx , X )u](ξ) = e−iφh (x ,ξ) qh (ξ, x )u(x )dx (u ∈ S). ∗ Bm We write {qh (φ∗h ; Dx , X )}0<h<1 ∈ {B ρ,δ (φh )}0<h<1 , or simply m ∗ ∗ qh (φh ; Dx , X ) ∈ B ρ,δ (φh ). Definition 3.2 ( Fourier Integral Operators B m,λ ρ,δ (φh ) ). For 0 ≤ τ < 1, φh ∈ Pρ,δ (τ, h) and ph (x, ξ) ∈ Bm,λ ρ,δ (h), we define a family of Fourier integral operators ph (φh ; X, Dx ) by (3.3) ph (φh ; X, Dx )u(x) = eiφh (x,ξ) ph (x, ξ)û(ξ)d−ξ (u ∈ S). We write {ph (φh ; X, Dx )}0<h<1 B m,λ ρ,δ (φh ). ph (φh ; X, Dx ) ∈ ( See Theorem 4.2.) B m,λ {B ρ,δ (φh )}0<h<1 , ∈ or simply We sometimes use this definition for 0 ≤ τ < ∞. In the following proposition, it is a revised point that Mm,h,l , Mm,λ,h,l , l and l are independent of the choice of phase functions as well as the choice of symbols. Proposition 3.3 ( Continuity on S ). (1) For any 0 ≤ τ0 < 1 and {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist a constant Mm,h,l (depending only on m, h, τ0 , {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 )) and integers l , l such that (m) |ph (φh ; X, Dx )u|l,S ≤ Mm,h,l |ph |l |u|l ,S (m) |qh (φ∗h ; Dx , X )u|l,S ≤ Mm,h,l |qh |l |u|l ,S (3.4) (0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), ph (x, ξ), qh (ξ, x ) ∈ Bm ρ,δ (h), u ∈ S) . (2) For any 0 ≤ τ0 < 1 and {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist a constant Mm,λ,h,l (depending only on m, λ, h, τ0 , {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 )) and integers lλ , lλ (depending only on λ) such that (3.5) (m,λ) |ph (φh ; X, Dx )u|l,S ≤ Mm,λ,h,l |ph |l λ |u|lλ ,S m,λ (0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), ph (x, ξ) ∈ Bρ,δ (h), u ∈ S) . 464 Naoto Kumano-go Proof. See [6]. m ∗ ∗ B m (φ ) ⊂ B m+m (h)). Bm Bm Proposition 3.4 (B ρ,δ (φh ), B ρ,δ (φh )◦B ρ,δ (φh )◦B ρ,δ h ρ,δ For any 0 ≤ τ0 < 1 and {κl }∞ (κ = 1, κ ≤ κ ), there exists a l l+1 l=0 0 constant Cl+2n0 (depending only on τ0 , κ1 , κ2 , . . . , κl+2n0 ) satisfying the following: m Let 0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), ph (x, ξ) ∈ B ρ,δ (h) and m+m (h) such qh (ξ, x ) ∈ B m ρ,δ (h). Then there exist rh (x, ξ), rh (ξ, x ) ∈ B ρ,δ that ph (φh ; X, Dx )qh (φ∗h ; Dx , X ) = rh (X, Dx ) (3.6) qh (φ∗h ; Dx , X )ph (φh ; X, Dx ) = rh (Dx , X ) , and (3.7) (m+m ) |rh |l (m+m ) , |rh |l (m) (m ) ≤ Cl+2n0 |ph |l+2n0 |qh |l+2n0 . Proof. See [6]. m m+m Bm Bm Proposition 3.5 ( B m (φh ) ). ρ,δ (φh ), B ρ,δ (φh )◦B ρ,δ (h)◦B ρ,δ (h) ⊂ B ρ,δ ∞ For any 0 ≤ τ0 < 1 and {κl }l=0 (κ0 = 1, κl ≤ κl+1 ), there exists a constant Cl+2n0 −1 (depending only on τ0 , κ1 , κ2 , . . . , κl+2n0 −1 ) satisfying the following: m Let 0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), ph (x, ξ) ∈ B ρ,δ (h) and qh (x, ξ) ∈ m+m (h) such that Bm ρ,δ (h). Then there exist rh (x, ξ), rh (x, ξ) ∈ B ρ,δ (3.8) ph (X, Dx )qh (φh ; X, Dx ) = rh (φh ; X, Dx ) qh (φh ; X, Dx )ph (X, Dx ) = rh (φh ; X, Dx ) , and (3.9) (m+m ) |rh |l (m+m ) , |rh |l Proof. See [6]. (m) (m ) ≤ Cl+2n0 −1 |ph |l+2n0 |qh |l+2n0 . Fundamental solution for Schrödinger equations 465 m ∗ ∗ B m (h) ⊂ B m+m (φ∗ ) ). Bm Proposition 3.6 ( B m ρ,δ (φh ), B ρ,δ (φh )◦B ρ,δ (h)◦B ρ,δ h ρ,δ (κ = 1, κ ≤ κ ), there exists a For any 0 ≤ τ0 < 1 and {κl }∞ 0 l l+1 l=0 constant Cl+2n0 −1 (depending only on τ0 , κ1 , κ2 , . . . , κl+2n0 −1 ) satisfying the following: m Let 0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), ph (ξ, x ) ∈ B ρ,δ (h) and m+m (h) such qh (ξ, x ) ∈ B m ρ,δ (h). Then there exist rh (ξ, x ), rh (ξ, x ) ∈ B ρ,δ that (3.10) ph (Dx , X )qh (φ∗h ; Dx , X ) = rh (φ∗h ; Dx , X ) qh (φ∗h ; Dx , X )ph (Dx , X ) = rh (φ∗h ; Dx , X ) , and (3.11) (m+m ) |rh |l (m+m ) , |rh |l (m) (m ) ≤ Cl+2n0 −1 |ph |l+2n0 |qh |l+2n0 . Proof. See [6]. Proposition 3.7 ( L 2 -Boundedness ). For any 0 ≤ τ0 < 1 and {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exists a constant C5n0 (depending only on τ0 , κ1 , κ2 , . . . , κ5n0 ) such that (3.12) (m) ph (φh ; X, Dx )uL2 ≤ C5n0 hm |ph |5n0 uL2 (m) qh (φ∗h ; Dx , X )uL 2 ≤ C5n0 hm |qh |5n0 uL 2 (0 ≤ τ ≤ τ0 , φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ), m ph (x, ξ), qh (ξ, x ) ∈ Bρ,δ (h), u ∈ S) . Proof. See [6]. Proposition 3.8 ( Inverse of I(φh ), I(φ∗h ) ). For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants 0 < τ0 < 1(depending only on κ1 , κ2 , . . . , κ5n0 ) and Cl+7n0 (depending only on κ1 , κ2 , . . . , κl+7n0 ), satisfying the following: Let φh ∈ Pρ,δ (τ, h, {κl }∞ l=0 ) and 0 ≤ τ ≤ τ0 . 466 Naoto Kumano-go Then there exist rh (ξ, x ), rh (x, ξ) ∈ B0ρ,δ (h) such that (3.13) I(φh )rh (φ∗h ; Dx , X ) = rh (φ∗h ; Dx , X )I(φh ) = I I(φ∗h )rh (φh ; X, Dx ) = rh (φh ; X, Dx )I(φ∗h ) = I , and |rh |l , |rh |l (0) (3.14) (0) ≤ Cl+7n0 . Proof. See [6]. 4. Multi-Products of Fourier Integral Operators In this section, we apply Fujiwara’s skip method in [3] to Kitada and H.Kumano-go’s theory of multi-products of Fourier integral operators. This section plays an essential role to prove the pointwise convergence of the symbol in Section 5. m2 m1 +m2 1 Theorem 4.1 ( B m (φ1,h #φ2,h ) ). ρ,δ (φ1,h ) ◦ B ρ,δ (φ2,h ) ⊂ B ρ,δ ∞ For any {κl }l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants C3l+2n+1 (de pending only on κ1 , κ2 , . . . , κ3l+2n+1 ) and C3l+2n+3 (depending only on κ1 , κ2 , . . . , κ3l+2n+3 ), satisfying the following: mj Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ), pj,h (x, ξ) ∈ Bρ,δ (h)(j = 1, 2), τ1 + τ2 ≤ 1/4, and {Xh , Ξh }(x, ξ) be the solution of the equations (4.1) Xh = ∇ξ φ1,h (x, Ξh ) Ξh = ∇x φ2,h (Xh , ξ) . 1 +m2 Then there exist qh (x, ξ), rh (x, ξ) ∈ Bm (h) such that ρ,δ (4.2) qh (x, ξ) = p1,h (x, Ξh (x, ξ))p2,h (Xh (x, ξ), ξ) + rh (x, ξ) , (4.3) p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) = qh (φ1,h #φ2,h ; X, Dx ) , (4.4) (m1 +m2 ) |qh |l (m ) (m ) 1 2 ≤ C3l+2n+1 |p1,h |3l+2n+1 |p2,h |3l+2n+1 , Fundamental solution for Schrödinger equations 467 and (4.5) (m1 +m2 ) |rh |l 1 1 ≤ C3l+2n+3 (τ1 |p1,h |3l+2n+3 + max |∂ξj p1,h |3l+2n+3 ) (m ) (m +ρ) 1≤j≤n × (m2 ) (τ2 |p2,h |3l+2n+3 (m −δ) 2 + max |Dxj p2,h |3l+2n+3 ). 1≤j≤n Proof. (I) For u ∈ S, we can write (4.6) p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx )u(x) = ei(φ1,h #φ2,h )(x,ξ) qh (x, ξ)û(ξ)d−ξ , where (4.7) ψh (x, ξ , x , ξ) ≡ φ1,h (x, ξ ) − x · ξ + φ2,h (x , ξ) − (φ1,h #φ2,h )(x, ξ) , and (4.8) Set (4.9) qh (x, ξ) ≡ Os — eiψh (x,ξ ,x ,ξ) p1,h (x, ξ )p2,h (x , ξ)dx d−ξ . q (x, ξ) ≡ h−(m1 +m2 ) qh (hδ x, h−ρ ξ) h pj,h (x, ξ) ≡ h−mj pj,h (hδ x, h−ρ ξ) ∈ B00,0 (h) (j = 1, 2) ψh (x, ξ , x , ξ) ≡ hρ−δ ψh (hδ x, h−ρ ξ , hδ x , h−ρ ξ) . By a change of variables: (x, ξ, x , ξ ) → (hδ x, h−ρ ξ, hδ x , h−ρ ξ ), we have −2σ −2nσ (4.10) Os — eih ψh (x,ξ ,x ,ξ) qh (x, ξ) = h · p1,h (x, ξ ) p2,h (x , ξ)dx d−ξ , where σ ≡ (ρ − δ)/2. Furthermore, set −δ δ −ρ Xh (x, ξ) ≡ h Xh (h x, h ξ) h (x, ξ) ≡ hρ Ξh (hδ x, h−ρ ξ) Ξ (4.11) Jj,h (x, ξ) ≡ hρ−δ φj,h (hδ x, h−ρ ξ) − x · ξ (j = 1, 2) . 468 Naoto Kumano-go h (x, ξ) + hσ y, ξ = Ξ h (x, ξ) + hσ η, we have By a change of variables: x = X qh (x, ξ) = Os — (4.12) h (x, ξ) + hσ η) eiϕh (y,η;x,ξ) p1,h (x, Ξ h (x, ξ) + hσ y, ξ)dyd−η , · p2,h (X where h (x, Ξ h (x, ξ) + hσ η, X h (x, ξ) + hσ y, ξ) . ϕ h (y, η; x, ξ) ≡ h−2σ ψ (4.13) Setting (4.14) h (η; x, ξ) ≡ A h (y; x, ξ) ≡ B 1 0 1 → 1,h )(x, Ξ h (x, ξ) + θhσ η)dθη · η (1 − θ)(∇ξ ∇ξ J → σ 0 (1 − θ)(∇x ∇x J2,h )(Xh (x, ξ) + θh y, ξ)dθy · y , we can write (4.15) h (η; x, ξ) + B h (y; x, ξ) . ϕ h (y, η; x, ξ) = −y · η + A (II) For 0 ≤ ϑ ≤ 1, set h (ϑη; x, ξ) + B h (y; x, ξ) . ϕ ϑ,h (y, η; x, ξ) ≡ −y · η + A (4.16) Then we have (4.17) h (y, η; x, ξ) ϕ 1,h (y, η; x, ξ) = ϕ h (y; x, ξ) . ϕ 0,h (y, η; x, ξ) = −y · η + B Furthermore, we get (4.18) h (η; x, ξ) = ∇η A h (y; x, ξ) = ∇y B 1 → σ 0 (∇ξ ∇ξ J1,h )(x, Ξh (x, ξ) + θh η)dθη 1 0 → 2,h )(X h (x, ξ) + θhσ y, ξ)dθy , (∇x ∇x J and (4.19) h )(ϑη; x, ξ) ϑ,h (y, η; x, ξ) = −y + ϑ(∇η A ∇η ϕ h )(y; x, ξ) . ϑ,h (y, η; x, ξ) = −η + (∇y B ∇y ϕ Fundamental solution for Schrödinger equations 469 By Proposition 2.4, there exists a constant Cl (depending only on κ1 , κ2 , . . . , κl ) such that (4.20) h )(ϑη; x, ξ)| ≤ C |∂ξα Dxβ ∂ηα (∇η A |α+β+α | τ1 η (0 ≤ ϑ ≤ 1) h )(y; x, ξ)| ≤ C |∂ξα Dxβ Dyβ (∇y B |α+β+β | τ2 y , and (4.21) −1 ϑ,h (y, η; x, ξ) ≤ 3y; η 3 y; η ≤ ∇(y,η) ϕ |∂ α Dxβ ∂ηα Dyβ ∇(y,η) ϕ ϑ,h (y, η; x, ξ)| ≤ C|α+β+α +β | y; η ξ 2 ϑ,h (y, η; x, ξ)| ≤ C|α+β+α |∂ξα Dxβ ∂ηα Dyβ ϕ +β | y; η (0 ≤ ϑ ≤ 1) . h + hσ (ϑη)) h + (III) Setting f (ϑ; y, η; x, ξ) ≡ eiϕϑ,h (y,η;x,ξ) p1,h (x, Ξ p2,h (X 1 hσ y, ξ) and using the formula: f (1) = f (0) + 0 f (ϑ)dϑ, we have (4.22) h) p1,h (x, Ξ qh (x, ξ) = h + hσ y, ξ)dyd−η · Os — e−iy·η eiBh (y;x,ξ) p2,h (X 1 f (ϑ; y, η; x, ξ)dϑdyd−η + Os — 0 h ) h , ξ) + p2,h (X rh (x, ξ) , = p1,h (x, Ξ where (4.23) rh (x, ξ) ≡ 1 rϑ,h (x, ξ)dϑ , 0 and (4.24) (e−iy·η η) rϑ,h (x, ξ) ≡ Os — h )(ϑη) h + hσ (ϑη)) · i(∇η A p1,h (x, Ξ h + hσ (ϑη)) p1,h )(x, Ξ + hσ (∇ξ h + hσ y, ξ)dyd−η . p2,h (X × eiAh (ϑη) eiBh (y) 470 Naoto Kumano-go Integrating by parts, we get rϑ,h (x, ξ) = Os — (4.25) eiϕϑ,h (y,η;x,ξ) aϑ,h (y, η; x, ξ)dyd−η , where h )(ϑη) h + hσ (ϑη)) p1,h (x, Ξ aϑ,h (y, η; x, ξ) ≡ i(∇η A h + hσ (ϑη)) p1,h )(x, Ξ + hσ (∇ξ h )(y) h + hσ y, ξ) p2,h (X · (∇y B h + hσ y, ξ) . p2,h )(X + hσ (−i∇x (4.26) (IV) Set aϑ,h,α,β (y, η; x, ξ) eiϕϑ,h (y,η;x,ξ) α β iϕ ϑ,h (y,η;x,ξ) aϑ,h (y, η; x, ξ) . ≡ ∂ξ Dx e (4.27) By (4.20), (4.21) and Proposition 2.4, there exists a constant Cl (depending only on κ1 , κ2 , . . . , κl ) such that 2|α+β|+2 (4.28) |∂ξγ Dxµ ∂ηα Dyβ aϑ,h,α,β (y, η; x, ξ)| ≤ C|α+β+α +β +γ+µ| y; η (0) (0) p1,h ||α+β+α +γ+µ| + max |∂ξj p1,h ||α+β+α +γ+µ| ) × (τ1 | 1≤j≤n (0) (0) p2,h ||α+β+β +γ+µ| + max |Dxj p2,h ||α+β+β +γ+µ| ) × (τ2 | 1≤j≤n (0 ≤ ϑ ≤ 1) . Hence, for l = 2|α + β| + 2n + 3, we can write (4.29) ∂ξα Dxβ rϑ,h (x, ξ) = = Os — eiϕϑ,h (y,η;x,ξ) aϑ,h,α,β (y, η; x, ξ)dyd−η eiϕϑ,h (1 + i∇(y,η) · ∇(y,η) ϕ ϑ,h )l · ∇(y,η) ϕ ϑ,h −2l aϑ,h,α,β (y, η; x, ξ)dyd−η , Fundamental solution for Schrödinger equations 471 and there exists a constant C3l+2n+3 (depending only on κ1 , κ2 , . . . , κ3l+2n+3 ) such that (0) (4.30) | rϑ,h |l ≤ C3l+2n+3 (τ1 | p1,h |3l+2n+3 + max |∂ξj p1,h |3l+2n+3 ) (0) (0) 1≤j≤n (0) (0) p2,h |3l+2n+3 + max |Dxj p2,h |3l+2n+3 ) (0 ≤ ϑ ≤ 1) . × (τ2 | 1≤j≤n From (4.23), we have (4.31) (0) | rh |l ≤ C3l+2n+3 (τ1 | p1,h |3l+2n+3 + max |∂ξj p1,h |3l+2n+3 ) (0) (0) 1≤j≤n (0) (0) p2,h |3l+2n+3 + max |Dxj p2,h |3l+2n+3 ) . × (τ2 | 1≤j≤n Furthermore, we can get the estimate for qh (x, ξ) in a way similar to this. 1 ,λ1 2 ,λ2 1 +m2 ,λ1 +λ2 Bm Bm (φ1,h )◦B (φ2,h ) ⊂ B m (φ1,h #φ2,h )). Theorem 4.2 (B ρ,δ ρ,δ ρ,δ ∞ For any λ1 , λ2 ∈ R and {κl }l=0 (κ0 = 1, κl ≤ κl+1 ), there exists a constant Cλ1 ,λ2 ,lλ1 ,λ2 (depending only on λ1 , λ2 , κ1 , κ2 , . . . , κlλ1 ,λ2 ) satisfying the following: mj ,λj Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ (h)(j = 1, 2) and τ1 + l=0 ), pj,h (x, ξ) ∈ Bρ,δ 1 +m2 ,λ1 +λ2 τ2 ≤ 1/4. Then there exists qh (x, ξ) ∈ Bm (h) such that ρ,δ (4.32) p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) = qh (φ1,h #φ2,h ; X, Dx ) , and (4.33) (m1 +m2 ,λ1 +λ2 ) |qh |l (m1 ,λ1 ) ≤ Cλ1 ,λ2 ,lλ1 ,λ2 |p1,h |lλ 1 ,λ2 (m2 ,λ2 ) |p2,h |lλ 1 ,λ2 , where lλ1 ,λ2 ≡ 3l + 2n + 1 + [|λ1 | + |λ2 |]. Proof. (I) is the same as that in the proof of Theorem 4.1 except for the following modification: (4.34) pj,h (x, ξ) ≡ h−mj pj,h (hδ x, h−ρ ξ) ∈ B0,0 j (h) (j = 1, 2) . 0,λ 472 Naoto Kumano-go (II) By Proposition 2.4, there exists a constant Cl (depending only on κ1 , κ2 , . . . , κl ) such that −1 3 y; η ≤ ∇(y,η) ϕ h (y, η; x, ξ) ≤ 3y; η β β |∂ξα Dx ∂ηα Dy ∇(y,η) ϕ h (y, η; x, ξ)| ≤ C|α+β+α (4.35) +β | y; η α β α β 2 h (y, η; x, ξ)| ≤ C|α+β+α |∂ξ Dx ∂η Dy ϕ +β | y; η . (III) Set pj,h (x, ξ) ≡ pj,h (x, ξ)x; ξ −λj ∈ B00,0 (h) (j = 1, 2) . (4.36) Then we have qh (x, ξ) = Os — (4.37) eiϕh (y,η;x,ξ) ah (y, η; x, ξ)dyd−η , where h + hσ η) h + hσ y, ξ) ah (y, η; x, ξ) ≡ p1,h (x, Ξ p2,h (X (4.38) h + hσ η λ1 X h + hσ y; ξ λ2 . × x; Ξ (IV) Set (4.39) eiϕh (y,η;x,ξ) ah,α,β (y, η; x, ξ) ≡ ∂ξα Dxβ eiϕh (y,η;x,ξ) ah (y, η; x, ξ) . By (4.35), Proposition 2.4 and Lemma 2.6, there exists a constant Cλ1 ,λ2 ,l (depending only on λ1 , λ2 , κ1 , κ2 , . . . , κl ) such that (4.40) ah,α,β (y, η; x, ξ)| |∂ξγ Dxµ ∂ηα Dyβ ≤ Cλ1 ,λ2 ,|α+β+α +β +γ+µ| y; η |λ1 |+|λ2 |+2|α+β| x; ξ λ1 +λ2 × | p1,h ||α+β+α +γ+µ| | p2,h ||α+β+β +γ+µ| . (0) (0) Hence, for l = 2|α + β| + 2n + 1 + [|λ1 | + |λ2 |], we can write α β (4.41) ∂ξ Dx qh (x, ξ) = Os — eiϕh (y,η;x,ξ) ah,α,β (y, η; x, ξ)dyd−η = eiϕh (1 + i∇(y,η) · ∇(y,η) ϕ h )l · ∇(y,η) ϕ h −2l ah,α,β (y, η; x, ξ)dyd−η Fundamental solution for Schrödinger equations 473 and there exists a constant Cλ1 ,λ2 ,lλ1 ,λ2 (depending only on λ1 , λ2 , κ1 , κ2 , . . . , κlλ1 ,λ2 ) such that (0,λ1 +λ2 ) | qh |l (4.42) ≤ Cλ1 ,λ2 ,lλ1 ,λ2 | p1,h |lλ (0) | p2,h |lλ (0) 1 ,λ2 1 ,λ2 . Therefore, we can get (4.33). Theorem 4.3 ( H.Kumano-go-Taniguchi theorem ). For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants 0 < τ0 < 1/(4c0 ) (depending only on κ1 , κ2 , . . . , κ5n0 ) and C3l+23n0 (depending only on κ1 , κ2 , . . . , κ3l+23n0 ) satisfying the following: mj Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ), pj,h (x, ξ) ∈ Bρ,δ (h)(j = 1, 2, . . . ) and ∞ τj ≤ τ0 . j=1 m̄ Then, for any ν = 1, 2, . . . , there exists rν+1,h (x, ξ) ∈ Bρ,δν+1 (h) such that (4.43) rν+1,h (φ1,h #φ2,h # · · · #φν+1,h ; X, Dx ) = p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) · · · pν+1,h (φν+1,h ; X, Dx ) , and (4.44) (m̄ν+1 ) |rν+1,h |l ≤ (C3l+23n0 )ν+1 ν+1 (m ) j |pj,h |l+11n , 0 j=1 where m̄ν+1 ≡ m1 + m2 + · · · + mν+1 . Proof. (I) By Proposition 2.5(1), there exist a constant c0 > 1(independent ∞ of {κl }∞ l=0 ) and a sequence of positive constants {κl }l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl )(depending only on κ1 , κ2 , . . . , κl ), satisfying the following: ∞ For any φj,h ∈ Pρ,δ (τj , h, {κl }∞ )(j = 1, 2, . . . ) with τj ≤ 1/4, it l=0 j=1 follows that (4.45) Φj,h = (φ1,h #φ2,h # · · · #φj,h ) ∈ Pρ,δ (c0 τ¯j , h, {κl }) . 474 Naoto Kumano-go (II) By Proposition 3.8, there exists a constant 0 < τ0 < 1/(4c0 )(depending only on κ1 , κ2 , . . . , κ5n0 ) satisfying the following: ∞ For any φj,h ∈ Pρ,δ (τj , h, {κl }∞ τj ≤ τ0 /c0 (≡ l=0 )(j = 1, 2, . . . ) with τ0 ), there exist sj,h (ξ, x ), tj,h (ξ, x ) ∈ B0ρ,δ (h) such that (4.46) j=1 I(φj,h )sj,h (φ∗j,h ; Dx , X ) = I tj,h (Φ∗j,h ; Dx , X )I(Φj,h ) = I . (III) For any j = 1, 2, . . . , set I(Φj−1,h )I(φj,h ) (4.47) qj,h (X, Dx ) ≡ ∗ ∗ ◦ sj,h (φj,h ; Dx , X )pj,h (φj,h ; X, Dx ) tj,h (Φj,h ; Dx , X ) , where Φ0,h ≡ x · ξ. Then we have (4.48) rν+1,h (φ1,h #φ2,h # · · · #φν+1,h ; X, Dx ) = p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) · · · pν+1,h (φν+1,h ; X, Dx ) = q1,h (X, Dx )q2,h (X, Dx ) · · · qν+1,h (X, Dx )I(Φν+1,h ) . By Proposition 3.5, 1.4, 3.4, 3.6 and Theorem 4.1, we get (4.44). Definition 4.4. For τ ≥ 0 and a sequence of positive constants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), we say that a family {ωh }0<h<1 of complex-valued C ∞ -functions ωh (x, ξ) belongs to the class {Ωρ,δ (τ, h, {κl }∞ h (x, ξ) ≡ l=0 )}0<h<1 , if ω δ −ρ ωh (h x, h ξ) satisfies ωh (x, ξ)} ≥ −κ0 τ inf inf Im{ 0<h<1 x,ξ (4.49) h (x,ξ)| sup sup |ωx;ξ + 0<h<1 x,ξ sup |∂ξα Dxβ ω h (x, ξ)| 1≤|α+β|≤1+l x,ξ ≤ κl τ , for any l = 0, 1, 2, . . . . We write {ωh }0<h<1 ∈ {Ωρ,δ (τ, h, {κl }∞ l=0 )}0<h<1 , or simply ωh ∈ Ωρ,δ (τ, h, {κl }∞ ). l=0 Fundamental solution for Schrödinger equations 475 Proposition 4.5. For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist a constant c0 > 1(inde∞ pendent of {κl }l=0 ) and a sequence of positive constants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl ) (depending only on κ1 , κ2 , . . . , κl ), satisfying the following: ∞ Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ) and ωj,h ∈ Ωρ,δ (τj , h, {κl }l=0 ) (j = 1, ∞ τj ≤ 1/4. 2, . . . ) with j=1 Then, for any ν = 1, 2, . . . , it follows that ∞ Φν+1,h ≡ (φ1,h #φ2,h # · · · #φν+1,h ) ∈ Pρ,δ (c0 τ̄ν+1 , h, {κl }l=0 ) ν+1 (4.50) j−1 j ωj,h (Xν,h , Ξν,h ) ∈ Ωρ,δ (c0 τ̄ν+1 , h, {κl }∞ l=0 ) , j=1 where τ̄ν+1 ≡ τ1 + τ2 + · · · + τν+1 . Proof. By Proposition 2.4 and 2.5. Theorem 4.6 ( Fujiwara’s skip method ). For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants 0 < τ0 < 1/(4c0 ) (depending only on κ1 , κ2 , . . . , κ5n0 ) and C3l+36n0 , C3l+36n0 (depending only on κ1 , κ2 , . . . , κ3l+36n0 ) satisfying the following: ∞ For φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ) and ωj,h ∈ Ωρ,δ (τj , h, {κl }l=0 )(j = 1, 2, . . . ) ∞ τj ≤ τ0 , set with j=1 (4.51) φ1,2,...,ν+1,h (x, ξ) ≡ (φ1,h #φ2,h # · · · #φν+1,h )(x, ξ) ν+1 j−1 j ω1,2,...,ν+1,h (x, ξ) ≡ ωj,h (Xν,h , Ξν,h ), j=1 and (4.52) pj,h (x, ξ) ≡ exp iωj,h (x, ξ) ∈ B0ρ,δ (h) p# (x, ξ) ≡ exp iω (x, ξ) ∈ B0ρ,δ (h) . 1,2,...,ν+1,h 1,2,...,ν+1,h Then, for any ν = 1, 2, . . . , there exists Υ1,2,...,ν+1,h (x, ξ) ∈ B0ρ,δ (h) such that (4.53) p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) · · · pν+1,h (φν+1,h ; X, Dx ) = p# 1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) + Υ1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) , 476 (4.54) Naoto Kumano-go (0) |Υ1,2,...,ν+1,h |l ≤ C3l+36n (τ̄ν+1 )2 , 0 and (4.55) (0) |(p# 1,2,...,ν+1,h + Υ1,2,...,ν+1,h )|l ≤ C3l+36n . 0 Proof. (I) By Proposition 2.5 and Proposition 4.5, there exist a constant c0 > 1 and a sequence of positive constants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl ) (depending only on κ1 , κ2 , . . . , κl ) such that (4.56) φ1,2,...,ν,h ∈ Pρ,δ (c0 τ̄ν , h, {κl }∞ l=0 ), ∞ φ ν+1,h ∈ Pρ,δ (c0 τν+1 , h, {κl }l=0 ) ω1,2,...,ν,h ∈ Ωρ,δ (c0 τ̄ν , h, {κl }∞ l=0 ), ων+1,h ∈ Ωρ,δ (c0 τν+1 , h, {κl }∞ l=0 ) . Hence, there exist a constant Cl > 1(depending only on κ1 , κ2 , . . . , κl ) such that # (0) (0) |p1,2,...,ν,h |l ≤ Cl , |pν+1,h |l ≤ Cl (ρ) (−δ) |∂ξk p# ≤ Cl τν+1 1,2,...,ν,h |l ≤ Cl τ̄ν , |Dxk pν+1,h |l (4.57) ν (ν = 1, 2, . . . , k = 1, 2, . . . , n, p# ≡ pj,h , τ̄ν ≡ τj ) . j,h j=1 By Theorem 4.1 and Proposition 2.5(2), there exist a constant C3l+2n+3 0 (depending only on κ1 , κ2 , . . . , κ3l+2n+3 ) and r1,2,...,ν+1,h (x, ξ) ∈ Bρ,δ (h) such that (4.58) p# 1,2,...,ν,h (φ1,2,...,ν,h ; X, Dx )pν+1,h (φν+1,h ; X, Dx ) = p# 1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) + r1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) , and (4.59) (0) |r1,2,...,ν+1,h |l ≤ C3l+2n+3 (τ̄ν )τν+1 . Fundamental solution for Schrödinger equations 477 For simplicity, set # P1,2,...,ν+1,h ≡ p# 1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) (4.60) R1,2,...,ν+1,h ≡ r1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) . Furthermore, replacing {φ1,h , φ2,h , . . . , φν+1,h } and {ω1,h , ω2,h , . . . , ων+1,h } by {φj+1,h , φj+2,h , . . . , φj+ν+1,h } and {ωj+1,h , ωj+2,h , . . . , ωj+ν+1,h }, set # Pj+1,j+2,...,j+ν+1,h and Rj+1,j+2,...,j+ν+1,h . (II) We use Fujiwara’s skip method: Using (4.58) inductively, we can get (4.61) p1,h (φ1,h ; X, Dx )p2,h (φ2,h ; X, Dx ) · · · pν+1,h (φν+1,h ; X, Dx ) = p# 1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) + Υ1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) , where (4.62) Υ1,2,...,ν+1,h (φ1,2,...,ν+1,h ; X, Dx ) Rj0 +1,j0 +2,...,j1 ,h Rj1 +1,j1 +2,...,j2 ,h ≡ · · · Rjs−1 +1,js−1 +2,...,js ,h Pj#s +1,js +2,...,ν+1,h , stands for the summation with respect to the sequences of integers (j1 , j2 , . . . , js−1 , js ) with the property (4.63) 0 ≡ j0 < j1 − 1 < j1 < j2 − 1 < · · · < js−1 < js − 1 < js ≤ ν + 1 , and, in the special case of js = ν + 1, we set Pj#s +1,js +2,...,ν+1,h ≡ I. Now, let C3l+23n0 is the same constant as that in Theorem 4.3. By (4.57), (4.59), (4.62), (4.63) and Theorem 4.3, we have (4.64) (0) |Υ1,2,...,ν+1,h |l (C3l+23n0 )s+1 Cl+11n ≤ 0 · s k=1 ≤ j k −1 C3(l+11n0 )+2n+3 ( τk )τjk k =jk−1 +1 C3l+23n0 Cl+11n 0 ν+1 j=1 ≤ C3l+36n (τ̄ν+1 )2 . 0 1+ (τ̄ν+1 )τj C3l+23n0 C3l+36n 0 −1 478 Naoto Kumano-go Here C3l+36n is a constant depending only on κ1 , κ2 , . . . , κ3l+36n0 . 0 The following lemma is a revised version of that in Fujiwara [4]. Lemma 4.7. ∞ For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ) and {χl }l=0 (0 ≤ χl ≤ χl+1 ), there exists a constant C3l+4n+5 (depending only on κ1 , κ2 , . . . , κ3l+4n+5 and χ0 , χ1 , . . . , χ3l+4n+5 ) satisfying the following: 0 Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ), pj,h (x, ξ) ∈ Bρ,δ (h) (j = 1, 3), φ2,ϑ,h ∈ 3 0 τj ≤ 1/(4c0 ). Pρ,δ (τ2 , h, {κl }∞ l=0 ), p2,ϑ,h (x, ξ) ∈ Bρ,δ (h) (0 ≤ ϑ ≤ 1) and j=1 2,ϑ,h (0 ≤ ϑ ≤ 1) and let φ Furthermore, let ∂ϑ p2,ϑ,h (x, ξ) ∈ satisfy (2.26). Then, there exists qϑ,h (x, ξ) ∈ B0ρ,δ (h) such that 0,1 Bρ,δ (h) (4.65) qϑ,h (φ1,h #φ2,ϑ,h #φ3,h ; X, Dx ) = p1,h (φ1,h ; X, Dx )p2,ϑ,h (φ2,ϑ,h ; X, Dx )p3,h (φ3,h ; X, Dx ) , ∂ϑ qϑ,h (x, ξ) ∈ B0,1 ρ,δ (h) , (4.66) and (4.67) (0,1) |∂ϑ qϑ,h |l (0) (0) ≤ C3l+4n+5 |p1,h |3l+4n+5 |p3,h |3l+4n+5 (0) (0,1) × (τ2 )2 |p2,h |3l+4n+5 + |∂ϑ p2,ϑ,h |3l+4n+5 (0 ≤ ϑ ≤ 1) . Proof. (I) For u ∈ S, we can write (4.68) qϑ,h (x, ξ) = h −4nσ Os — −2σ ψ ϑ,h (x,ξ 1 ,x1 ,ξ 2 ,x2 ,ξ) eih p2,ϑ,h (x1 , ξ 2 ) p3,h (x2 , ξ)dx1 d−ξ 1 dx2 d−ξ 2 , × p1,h (x, ξ 1 ) Fundamental solution for Schrödinger equations where (4.69) ϑ,h (x, ξ 1 , x1 , ξ 2 , x2 , ξ) ≡ φ 1,h (x, ξ 1 ) − x1 · ξ 1 + φ 2,ϑ,h (x1 , ξ 2 ) ψ 3,h (x2 , ξ) − x2 · ξ 2 + φ 2,ϑ,h #φ 3,h )(x, ξ) , 1,h #φ − (φ and σ ≡ (ρ − δ)/2. j }2 (x, ξ) be the solution of the equations (2.24). j ,Ξ Let {X 2,ϑ,h 2,ϑ,h j=1 By a change of variables: (4.70) 1 (x, ξ) + hσ y 1 , ξ 1 = Ξ 1 (x, ξ) + hσ η 1 x1 = X 2,ϑ,h 2,ϑ,h 2 (x, ξ) + hσ y 2 , ξ 2 = Ξ 2 (x, ξ) + hσ η 2 , x2 = X 2,ϑ,h 2,ϑ,h we can write qϑ,h (x, ξ) = Os — (4.71) eiϕϑ,h (y 1 ,η 1 ,y 2 ,η 2 ;x,ξ) · aϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ)dy 1 d−η 1 dy 2 d−η 2 , where σ 1 1 p1,h (x, Ξ aϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) ≡ 2,ϑ,h + h η ) (4.72) σ 1 2 σ 2 1 × p2,ϑ,h (X 2,ϑ,h + h y , Ξ2,ϑ,h + h η ) σ 2 2 × p3,h (X 2,ϑ,h + h y , ξ) , and (4.73) ϕ ϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) ≡ h−2σ ψϑ,h (x, ξ 1 , x1 , ξ 2 , x2 , ξ) = −y 1 · η 1 − y 2 · η 2 + y 1 · η 2 1 → σ 1 1 1 1 + (1 − θ)(∇ξ ∇ξ J1,h )(x, Ξ 2,ϑ,h + θh η )dθη · η 0 1 + 0 → σ 1 2 σ 2 1 1 1 (1 − θ)(∇x ∇x J2,ϑ,h )(X 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθy · y +2 0 1 → σ 1 2 σ 2 1 2 1 (1 − θ)(∇x ∇ξ J2,ϑ,h )(X 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθy · η 479 480 Naoto Kumano-go 1 + 0 1 + 0 → σ 1 2 σ 2 2 2 1 (1 − θ)(∇ξ ∇ξ J2,ϑ,h )(X 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθη · η → σ 2 2 2 2 (1 − θ)(∇x ∇x J3,h )(X 2,ϑ,h + θh y , ξ)dθy · y . Furthermore, we have (4.74) → 1 + 1 (∇ ∇ J σ 1 1 1 ∇ = −y 1ϕ ϑ,h ξ ξ 1,h )(x, Ξ2,ϑ,h + θh η )dθη η 0 ∇y 1 ϕ ϑ,h = −η 1 + η 2 1 → σ 1 2 σ 2 1 1 + 0 (∇x ∇x J2,ϑ,h )(X2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθy → σ 1 2 σ 2 2 2,ϑ,h )(X 1 + 1 (∇ξ ∇x J 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθη 0 ∇η 2 ϕ ϑ,h = y 1 − y 2 1 → σ 1 2 σ 2 1 2,ϑ,h )(X 1 + 0 (∇x ∇ξ J 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθy 1 → σ 1 2 σ 2 2 2,ϑ,h )(X 1 + 0 (∇ξ ∇ξ J 2,ϑ,h + θh y , Ξ2,ϑ,h + θh η )dθη 1 → σ 1 2 3,h )(X 2 ∇y 2 ϕ ϑ,h = −η 2 + 0 (∇x ∇x J 2,ϑ,h + θh y , ξ)dθy . (II) By (4.72), (4.73), (4.74) and Proposition 2.4, there exists a constant C (independent of {κl }∞ l=0 ) and Cl (depending only on κ1 , κ2 , . . . , κl ) such that −1 1 1 2 2 C y ; η ; y ; η ≤ ∇(y1 ,η1 ,y2 ,η2 ) ϕ ϑ,h ≤ Cy 1 ; η 1 ; y 2 ; η 2 α β α1 β 1 α 2 β 2 ϑ,h | |∂ Dx ∂η1 Dy1 ∂η2 Dy2 ∇(y1 ,η1 ,y2 ,η2 ) ϕ ξ 1 1 2 2 ≤ C|α+β+α (4.75) 1 +β 1 +α2 +β 2 | y ; η ; y ; η 1 2 1 2 |∂ξα Dxβ ∂ηα1 Dyβ1 ∂ηα2 Dyβ2 ϕ ϑ,h | 1 1 2 2 2 ≤ C|α+β+α 1 +β 1 +α2 +β 2 | y ; η ; y ; η , and (4.76) 1 1 2 2 |∂ξα Dxβ ∂ηα1 Dyβ1 ∂ηα2 Dyβ2 aϑ,h | ≤ C|α+β+α p1,h ||α+β+α1 | 1 +β 1 +α2 +β 2 |−1 | (0) (0) (0) × | p2,ϑ,h ||α+β+β 1 +α2 | | p3,h ||α+β+β 2 | (0 ≤ ϑ ≤ 1) , Fundamental solution for Schrödinger equations 481 where y; η; y ; η ≡ (1 + |y|2 + |η|2 + |y |2 + |η |2 )1/2 . Furthermore, by (4.72), (4.73), Lemma 2.6, 2.7, and Proposition 2.4, there exists a constant Cl (depending only on κ1 , κ2 , . . . , κl and χ0 , χ1 , . . . , χl ) such that (4.77) 1 1 2 2 ϑ,h | |∂ξα Dxβ ∂ηα1 Dyβ1 ∂ηα2 Dyβ2 ∂ϑ ϕ 2 1 1 2 2 3 ≤ C|α+β+α 1 +β 1 +α2 +β 2 |+1 (τ2 ) y ; η ; y ; η x; ξ , and (4.78) 1 1 2 2 |∂ξα Dxβ ∂ηα1 Dyβ1 ∂ηα2 Dyβ2 ∂ϑ aϑ,h | 1 2 ≤ C|α+β+α 1 +β 1 +α2 +β 2 | y ; η x; ξ (0) (0) · | p1,h ||α+β+α1 |+1 | p3,h ||α+β+β 2 |+1 (0) (0,1) × (τ2 )2 | p2,ϑ,h ||α+β+β 1 +α2 |+1 + |∂ϑ p2,ϑ,h ||α+β+β 1 +α2 | (0 ≤ ϑ ≤ 1) , Set (4.79) bϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) ≡ ∂ϑ aϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) + i∂ϑ ϕ ϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) aϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) . Then we have (4.80) ∂ϑ qϑ,h (x, ξ) = Os — eiϕϑ,h (y 1 ,η 1 ,y 2 ,η 2 ;x,ξ) · bϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ)dy 1 d−η 1 dy 2 d−η 2 . (III) Set (4.81) eiϕϑ,h (y bϑ,h,α,β (y 1 , η 1 , y 2 , η 2 ; x, ξ) 1 ,η 1 ,y 2 ,η 2 ;x,ξ) 1 1 2 2 ≡ ∂ξα Dxβ eiϕϑ,h (y ,η ,y ,η ;x,ξ)bϑ,h (y 1 , η 1 , y 2 , η 2 ; x, ξ) . 482 Naoto Kumano-go By (4.75)-(4.79), there exists a constant Cl (depending only on κ1 , κ2 , . . . , κl and χ0 , χ1 , . . . , χl ) such that 1 2 1 2 (4.82) |∂ξγ Dxµ ∂ηα1 Dyβ1 ∂ηα2 Dyβ2 bϑ,h,α,β (y 1 , η 1 , y 2 , η 2 ; x, ξ)| 1 1 2 2 2|α+β|+3 x; ξ ≤ C|α+β+α 1 +β 1 +α2 +β 2 +γ+µ|+1 y ; η ; y ; η (0) (0) × | p1,h ||α+β+α1 +γ+µ|+1 | p3,h ||α+β+β 2 +γ+µ|+1 (0) (0,1) × (τ2 )2 | p2,ϑ,h ||α+β+β 1 +α2 +γ+µ|+1 + |∂ϑ p2,ϑ,h ||α+β+β 1 +α2 +γ+µ| (0 ≤ ϑ ≤ 1) . Hence, for l = 2|α + β| + 4n + 4, we can write (4.83) ∂ξα Dxβ ∂ϑ qϑ,h (x, ξ) 1 1 2 2 eiϕϑ,h (y ,η ,y ,η ;x,ξ) = Os — · bϑ,h,α,β (y 1 , η 1 , y 2 , η 2 ; x, ξ)dy 1 d−η 1 dy 2 d−η 2 = eiϕϑ,h (1 + i∇(y1 ,η1 ,y2 ,η2 ) · ∇(y1 ,η1 ,y2 ,η2 ) ϕ ϑ,h )l × ∇(y1 ,η1 ,y2 ,η2 ) ϕ ϑ,h −2l bϑ,h,α,β (y 1 , η 1 , y 2 , η 2 ; x, ξ)dy 1 d−η 1 dy 2 d−η 2 and there exists a constant C3l+4n+5 (depending only on κ1 , κ2 , . . . , κ3l+4n+5 ) such that (0,1) (4.84) |∂ϑ qϑ,h |l (0) (0) ≤ C3l+4n+5 | p1,h |3l+4n+5 | p3,h |3l+4n+5 (0) (0,1) p2,ϑ,h |3l+4n+5 + |∂ϑ p2,ϑ,h |3l+4n+5 ) × ((τ2 )2 | (0 ≤ ϑ ≤ 1) . Therefore, we get (4.67). In Theorem 4.6, for ν = 1, 2, . . . and 0 ≤ ϑ ≤ 1, set (4.85) p1,2,...,ν+1,ϑ,h (x, ξ) ν+1 j−1 j ϑωj,h (Xν,h , Ξν,h ) + (1 − ϑ)ωj,h (x, ξ) ≡ exp i j=1 + ϑΥ1,2,...,ν+1,h (x, ξ) . Fundamental solution for Schrödinger equations 483 Then we have # p1,2,...,ν+1,1,h (x, ξ) = (p1,2,...,ν+1,h + Υ1,2,...,ν+1,h )(x, ξ) ν+1 p1,2,...,ν+1,0,h (x, ξ) = exp i ωj,h (x, ξ) . (4.86) j=1 Furthermore, we get the following lemma. Lemma 4.8. For any {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), there exist constants 0 < τ0 < 1/(4c0 ) (depending only on κ1 , κ2 , . . . , κ5n0 ) and C3l+36n0 , C3l+36n0 (depending only on κ1 , κ2 , . . . , κ3l+36n0 ) satisfying the following: ∞ Let φj,h ∈ Pρ,δ (τj , h, {κl }∞ l=0 ) and ωj,h ∈ Ωρ,δ (τj , h, {κl }l=0 )(j = 1, 2, . . . ) ∞ τj ≤ τ0 . with j=1 Then, it follows that (4.87) p1,2,...,ν+1,ϑ,h (x, ξ) ∈ B0ρ,δ (h) 0,1 (h) (0 ≤ ϑ ≤ 1) , ∂ϑ p1,2,...,ν+1,ϑ,h (x, ξ) ∈ Bρ,δ and (4.88) ≤ C3l+36n 0 (0) |p1,2,...,ν+1,ϑ,h |l (0,1) |∂ϑ p1,2,...,ν+1,ϑ,h |l ≤ C3l+36n (τ̄ν+1 )2 (0 ≤ ϑ ≤ 1) . 0 Proof. Note that (4.89) ν+1 j−1 , Ξ j ) − ω (x, ξ) ω j,h (X j,h ν,h ν,h j=1 ν+1 1 = → j−1 − x), ξ + θ(Ξ j − ξ))dθ (∇(x,ξ) ω j,h )(x + θ(X ν,h ν,h j=1 0 j−1 Xν,h − x . · j − ξ Ξ ν,h By Proposition 2.4, 4.5 and Theorem 4.6, we get (4.88). 484 5. Naoto Kumano-go A Construction of the Fundamental Solution In this section, using the theory developed in Section 1, 2, 3 and 4, we give a construction of the fundamental solution for the Cauchy problem of a pseudo-differential equation of Schrödinger’s type. We regard this construction as a multi-product of Fourier integral operators, and investigate the pointwise convergence of the phase function and that of the symbol. R 2n with a parameter Let Hh (t, x, ξ) be a real-valued function on [0, T ]×R β α R 2n 0 < h < 1, which has a continuous derivative ∂ξ Dx Hh (t, x, ξ) on [0, T ]×R for any α, β, satisfying (5.1) |∂ξα Dxβ {hρ−δ Hh (t, hδ x, h−ρ ξ)}| Cα,β x; ξ 2−|α+β| (|α + β| ≤ 1) ≤ Cα,β (|α + β| ≥ 2) , on [0, T ] × R 2n . Here Cα,β is a constant independent of a parameter 0 < h < 1. Let Wh (t, x, ξ) be a complex-valued function on [0, T ] × R 2n with a parameter 0 < h < 1, which has a continuous derivative ∂ξα Dxβ Wh (t, x, ξ) on [0, T ] × R 2n for any α, β, satisfying C0,0 x; ξ (α = β = 0) |∂ α Dxβ {Wh (t, hδ x, h−ρ ξ)}| ≤ ξ (|α + β| ≥ 1) Cα,β (5.2) δ −ρ Im{Wh (t, h x, h ξ)} ≥ −C0,0 , on [0, T ] × R 2n . Here Cα,β is a constant independent of a parameter 0 < h < 1. Set Kh (t, x, ξ) ≡ Hh (t, x, ξ) + Wh (t, x, ξ) (5.3) Lh ≡ i∂t + Kh (t, X, Dx ) . For sufficiently small T0 (0 < T0 ≤ T ), we consider the fundamental solution Uh (t, s) (0 ≤ s ≤ t ≤ T0 ) for the operator Lh such that (5.4) Lh Uh (t, s) = 0 on [s, T0 ] Uh (s, s) = I (0 ≤ s ≤ T0 ) . Fundamental solution for Schrödinger equations 485 Our purpose is to construct Uh (t, s) by (5.5) i Uh (t, s) = lim e t t1 Kh (τ )dτ |∆t,s |→0 · · · ei tν s i (X, Dx )e Kh (τ )dτ t 1 t2 Kh (τ )dτ (X, Dx ) (X, Dx ) , where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν ≥ tν+1 ≡ s (≥ 0) is an arbitrary division of interval [s,t] into subintervals, |∆t,s | ≡ max |tj−1 − tj |, and 1≤j≤ν+1 i e (5.6) tj−1 tj Kh (τ )dτ ≡ (X, Dx )u(x) i eix·ξ e tj−1 tj Kh (τ,x,ξ)dτ û(ξ)d−ξ (u ∈ S) . Now, set (5.7) t φh (t, s)(x, ξ) ≡ x · ξ + s Hh (τ, x, ξ)dτ t ωh (t, s)(x, ξ) ≡ s Wh (τ, x, ξ)dτ (0 ≤ s ≤ t ≤ T ) . and (5.8) ph (t, s)(x, ξ) ≡ exp iωh (t, s)(x, ξ) (0 ≤ s ≤ t ≤ T ) . Then we can write (5.9) ei t s Kh (τ )dτ (X, Dx )u(x) ≡ ph (t, s)(φh (t, s); X, Dx )u(x) . t Therefore, we regard the operator ei s Kh (τ )dτ (X, Dx ) as the Fourier integral operator ph (t, s)(φh (t, s); X, Dx ) with the phase function φh (t, s)(x, ξ) and the symbol ph (t, s)(x, ξ), and investigate the properties of i st Kh (τ )dτ e (X, Dx ). Theorem 5.1. Under the assumptions (5.1), (5.2) and (5.3), there exist sufficiently small T0 (0 < T0 ≤ T ), constants C1,l , c and a sequences of positive con stants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ), satisfying the following: 486 Naoto Kumano-go For any ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0), there exist a phase function φh (∆t0 ,tν+1 )(x, ξ) ∈ Pρ,δ (c (t0 − tν+1 ), h, {κl }∞ l=0 ) and a sym0 bol ph (∆t0 ,tν+1 )(x, ξ) ∈ Bρ,δ (h) such that (5.10) i ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) = e i ·e t 1 t2 Kh (τ )dτ i (X, Dx ) · · · e tν tν+1 t 0 t1 Kh (τ )dτ Kh (τ )dτ (X, Dx ) (X, Dx ) , and (0) |ph (∆t0 ,tν+1 )|l (5.11) ≤ C1,l . Furthermore, the following equation holds: (5.12) eiφh (∆t0 ,tν+1 )(x,ξ)−ix ·ξ ph (∆t0 ,tν+1 )(x, ξ) ν+1 = Os — ··· exp i(xj−1 − xj ) · ξ j j=1 tj−1 +i j−1 Kh (τ, x j , ξ )dτ dx1 d−ξ 1 · · · dxν d−ξ ν , tj where x0 ≡ x, xν+1 ≡ x and ξ ν+1 ≡ ξ. Proof. (I) By (5.1), (5.2) and (5.7), there exist a constant c > 0 and a sequence of positive constants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 ) such that (5.13) φh (t, s)(x, ξ) ∈ Pρ,δ (c(t − s), h, {κl }∞ l=0 ) ωh (t, s)(x, ξ) ∈ Ωρ,δ (c(t − s), h, {κl }∞ l=0 ) . Let c0 > 1 be the same constant as that in Proposition 4.5, T ≡ min{T, 1/(4c0 c)} and c ≡ c0 c. For any ∆t0 ,tν+1 : (T ≥) t0 ≥ t1 ≥ · · · ≥ j j , Ξν,h }νj=1 (x, ξ) be the solution of the equations tν+1 (≥ 0), let {Xν,h (5.14) j j−1 j X = ∇ξ φh (tj−1 , tj )(Xν,h , Ξν,h ) ν,h j j j+1 = ∇x φh (tj , tj+1 )(Xν,h , Ξν,h ) Ξν,h 0 ≡ x, Ξ ν+1 ≡ ξ) , (j = 1, . . . , ν, Xν,h ν,h Fundamental solution for Schrödinger equations 487 and set φh (∆t0 ,tν+1 )(x, ξ) ≡ (φh (t0 , t1 )#φh (t1 , t2 )# · · · #φh (tν , tν+1 ))(x, ξ) ν+1 j−1 j ωh (tj−1 , tj )(Xν,h , Ξν,h ). ωh (∆t0 ,tν+1 )(x, ξ) ≡ (5.15) j=1 Then, by Proposition 2.5 and 4.5, there exists a sequence of positive con stants {κl }∞ l=0 (κ0 = 1, κl ≤ κl+1 , κl ≤ κl ) such that ∞ φh (∆t0 ,tν+1 )(x, ξ) ∈ Pρ,δ (c (t − s), h, {κl }l=0 ) ωh (∆t0 ,tν+1 )(x, ξ) ∈ Ωρ,δ (c (t − s), h, {κl }∞ l=0 ) (ν = 0, 1, 2, . . . , T ≥ t0 ≥ t1 ≥ · · · ≥ tν+1 ≥ 0) . (5.16) Furthermore, set (5.17) (∆ )(x, ξ) ≡ exp iω (∆ )(x, ξ) ∈ B0ρ,δ (h) . p# t ,t t ,t h 0 ν+1 0 ν+1 h such that Then, by (5.16), there exists a constant C1,l (0) |p# h (∆t0 ,tν+1 )|l (5.18) ≤ C1,l (ν = 0, 1, 2, . . . , T ≥ t0 ≥ t1 ≥ · · · ≥ tν+1 ≥ 0) . (II) We choose sufficiently small T0 (0 < T0 < T ) such that Theorem 4.6 satisfying the following: holds. By Theorem 4.6, there exists a constant C1,l For any ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0), there exists Υh (∆t0 ,tν+1 )(x, ξ) ∈ B0ρ,δ (h) such that (5.19) ph (t0 , t1 )(φh (t0 , t1 ); X, Dx )ph (t1 , t2 )(φh (t1 , t2 ); X, Dx ) = · · · ph (tν , tν+1 )(φh (tν , tν+1 ); X, Dx ) # ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) + Υh (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) , and (5.20) (0) |Υh (∆t0 ,tν+1 )|l ≤ C1,l (t0 − tν+1 )2 . 488 Naoto Kumano-go (III) Set ph (∆t0 ,tν+1 ) = p# h (∆t0 ,tν+1 ) + Υh (∆t0 ,tν+1 ) . (5.21) Then we get (5.11). Lemma 5.2. (1) There exists a sequence of positive constants {χl }∞ l=0 (0 ≤ χl ≤ χl+1 ) satisfying the following: For any ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0), there exists φϑ,h (∆t0 ,tν+1 ) ∈ P0,0 (c (t0 − tν+1 ), h, {κl }∞ l=0 ) (0 ≤ ϑ ≤ 1) such that h (∆t ,t )(x, ξ) φ1,h (∆t0 ,tν+1 )(x, ξ) = φ 0 ν+1 φ0,h (∆t ,t )(x, ξ) = φh (t0 , tν+1 )(x, ξ) , (5.22) 0 ν+1 and (∆ )(x, ξ)| ≤ χ0 (t0 − tν+1 )2 x; ξ 2 |∂ J ϑ ϑ,h t0 ,tν+1 ϑ,h (∆t ,t )(x, ξ)| ≤ χ|α+β| (t0 − tν+1 )2 x; ξ (5.23) |∂ξα Dxβ ∇(x,ξ) ∂ϑ J 0 ν+1 (0 < h < 1, 0 ≤ ϑ ≤ 1) , ϑ,h (∆t ,t )(x, ξ) − x · ξ. ϑ,h (∆t ,t )(x, ξ) ≡ φ where J 0 ν+1 0 ν+1 (2) There exist constants C1,l , C2,l , satisfying the following: For any ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0), there exists pϑ,h (∆t0 ,tν+1 )(x, ξ) ∈ B0ρ,δ (h) (0 ≤ ϑ ≤ 1) such that (5.24) p1,h (∆t0 ,tν+1 )(x, ξ) = ph (∆t0 ,tν+1 )(x, ξ) p0,h (∆t0 ,tν+1 )(x, ξ) = ph (t0 , tν+1 )(x, ξ) , 0,1 ∂ϑ pϑ,h (∆t0 ,tν+1 )(x, ξ) ∈ Bρ,δ (h) (0 ≤ ϑ ≤ 1) , (5.25) and (5.26) (0) ≤ C1,l |pϑ,h (∆t0 ,tν+1 )|l (0,1) |∂ϑ pϑ,h (∆t0 ,tν+1 )|l ≤ C2,l (t0 − tν+1 )2 (0 ≤ ϑ ≤ 1) . Fundamental solution for Schrödinger equations 489 Proof. Set h (∆t ,t )(x, ξ) + (1 − ϑ)φ h (t0 , tν+1 )(x, ξ) , ϑ,h (∆t ,t )(x, ξ) ≡ ϑφ (5.27) φ 0 ν+1 0 ν+1 and (5.28) pϑ,h (∆t0 ,tν+1 )(x, ξ) ≡ exp i ϑωh (∆t0 ,tν+1 )(x, ξ) + (1 − ϑ)ωh (t0 , tν+1 )(x, ξ) + ϑΥh (∆t0 ,tν+1 )(x, ξ) . By Lemma 2.8 and Lemma 4.8, we get (1) and (2). Theorem 5.3. (1) There exists a constant C3,l satisfying the following: For any (T0 ≥) t ≥ s (≥ 0), let ∆t,s be an arbitrary division of interval [s, t] into subintervals, and ∆t,s be an arbitrary refinement of ∆t,s . Then it follows that (5.29) |φh (∆t,s ) − φh (∆t,s )|l (δ−ρ,2) ≤ C3,l |∆t,s |(t − s) . Furthermore, there exists a phase function φh (t, s) ∈ Pρ,δ (c (t − s), h, {κl }∞ l=0 ) such that (5.30) (δ−ρ,2) |φh (∆t,s ) − φh (t, s)|l ≤ C3,l |∆t,s |(t − s) . (2) There exists a constant C4,l satisfying the following: For any (T0 ≥) t ≥ s (≥ 0), let ∆t,s be an arbitrary division of interval [s, t] into subintervals, and ∆t,s be an arbitrary refinement of ∆t,s . Then it follows that (5.31) |ph (∆t,s ) − ph (∆t,s )|l (0,1) ≤ C4,l |∆t,s |(t − s) . 490 Naoto Kumano-go Furthermore, there exists a symbol ph (t, s) ∈ B0ρ,δ (h) such that (0,1) |ph (∆t,s ) − ph (t, s)|l (5.32) ≤ C4,l |∆t,s |(t − s) . and (0) |ph (t, s)|l (5.33) ≤ C1,l . (3) Furthermore, the following equation holds: (5.34) eiφh (t,s)(x,ξ)−ix ·ξ ph (t, s)(x, ξ) ν+1 ··· exp i(xj−1 − xj ) · ξ j = lim Os — |∆t,s |→0 tj−1 +i Kh (τ, xj−1 , ξ j )dτ j=1 dx1 d−ξ 1 · · · dxν d−ξ ν , tj where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0), x0 ≡ x, xν+1 ≡ x and ξ ν+1 ≡ ξ. Proof. (I) Set (5.35) ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) , and (5.36) ∆t,s t0 ≡ t0,0 ≥ t0,1 ≥ · · · ≥ t0,k0 ≡ t1 t1 ≡ t1,0 ≥ t1,1 ≥ · · · ≥ t1,k1 ≡ t2 ................................. : tν ≡ tν,0 ≥ tν,1 ≥ · · · ≥ tν,k ≡ tν+1 ν (kj > 0, j = 0, 1, 2, . . . , ν) . Fundamental solution for Schrödinger equations 491 Then we have h (∆ ) − φ h (∆t,s ) (5.37) φ t,s ν+1 h (∆ = φ t0 ,tj−1 )#φh (∆tj−1 ,tj )#φh (∆tj ,tν+1 ) j=1 h (∆ h (tj−1 , tj )#φ h (∆t ,t ) −φ )# φ t0 ,tj−1 j ν+1 = ν+1 j=1 1 0 ϑ,h (∆ h (∆t ,t ) dϑ . h (∆ ∂ϑ φ )# φ )# φ t0 ,tj−1 tj−1 ,tj j ν+1 By Lemma 2.7, Lemma 5.2 and Theorem 5.1, there exists a constant χl such that h (∆ ) − φ h (∆t,s )}| |∂ξα Dxβ {φ t,s ν+1 1 ≤ χ|α+β|−1 (tj−1 − tj )2 x; ξ 2 dϑ (5.38) j=1 0 ≤ χ|α+β|−1 |∆t,s |(t − s)x; ξ 2 . Hence we get (1). (II) Define qj,ϑ,h (∆t,s , ∆t,s )(x, ξ) ∈ B0ρ,δ (h) by (5.39) qj,ϑ,h (∆t,s , ∆t,s )(φh (∆t0 ,tj−1 )#φϑ,h (∆tj−1 ,tj )#φh (∆tj ,tν+1 ); X, Dx ) ≡ ph (∆t0 ,tj−1 )(φh (∆t0 ,tj−1 ); X, Dx ) ◦ pϑ,h (∆tj−1 ,tj )(φϑ,h (∆tj−1 ,tj ); X, Dx ) ◦ ph (∆tj ,tν+1 )(φh (∆tj ,tν+1 ); X, Dx ) . Then we have (5.40) qj,1,h (∆t,s , ∆t,s )(x, ξ) = ph (∆t0 ,tj , ∆tj+1 ,tν+1 )(x, ξ) qj,0,h (∆t,s , ∆t,s )(x, ξ) = ph (∆t0 ,tj−1 , ∆tj ,tν+1 )(x, ξ) . 492 Naoto Kumano-go Hence, it follows that (5.41) ph (∆t,s ) − ph (∆t,s ) ν+1 = ph (∆t0 ,tj , ∆tj+1 ,tν+1 ) − ph (∆t0 ,tj−1 , ∆tj ,tν+1 ) j=1 = ν+1 1 j=1 0 ∂ϑ qj,ϑ,h (∆t,s , ∆t,s ) dϑ . By Lemma 4.7, Lemma 5.2 and Theorem 5.1, we get (2). (III) From (1) and (2), (3) is clear. Theorem 5.4. The Fourier integral operator Uh (t, s) ≡ ph (t, s)(φh (t, s); X, Dx ) with the phase function φh (t, s)(x, ξ) and the symbol ph (t, s)(x, ξ) in Theorem 5.3, satisfies the following: (1) The operators Uh (t, s) : L 2 → L 2 , H 2,h → H 2,h are uniformly bounded for 0 ≤ s ≤ t ≤ T0 and 0 < h < 1. (2) The operators Kh (t, X, Dx )Uh (t, s), ∂t Uh (t, s) : H 2,h → H 0,h are uniformly bounded for 0 ≤ s ≤ t ≤ T0 and 0 < h < 1. (3) The following relation holds: (5.42) Uh (t, s) = i lim e t t1 Kh (τ )dτ |∆t,s |→0 · · · ei tν s i (X, Dx )e Kh (τ )dτ t 1 t2 Kh (τ )dτ (X, Dx ) (X, Dx ) , where ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν ≥ tν+1 ≡ s (≥ 0) is an arbitrary division of interval [s,t] into subintervals, and |∆t,s | ≡ max |tj−1 − tj |. 1≤j≤ν+1 (4) Uh (t, s) is the fundamental solution for the operator Lh such that (5.43) Lh Uh (t, s) = 0 on [s, T0 ] Uh (s, s) = I (0 ≤ s ≤ T0 ) . Fundamental solution for Schrödinger equations 493 Proof. (0) By Proposition 3.7, Theorem 4.2 and Theorem 5.3, (1) and (2) are clear. We have only to check (3) and (4) for u ∈ S. (I) By Proposition 3.3 and Theorem 5.1, there exist a constant M1,h,l and an integer l1 such that |ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx )u|l,S ≤ M1,h,l |u|l1 ,S ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0) . (5.44) Furthermore, we can write (5.45) ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) − ph (t0 , tν+1 )(φh (t0 , tν+1 ); X, Dx ) u(x) 1 iφϑ,h (∆t0 ,tν+1 )(x,ξ) − ∂ϑ pϑ,h (∆t0 ,tν+1 )(x, ξ)û(ξ)dξ dϑ e = 0 1 = qϑ,h (∆t0 ,tν+1 )(φϑ,h (∆t0 ,tν+1 ); X, Dx )u(x)dϑ , 0 where (5.46) qϑ,h (∆t0 ,tν+1 )(x, ξ) ≡∂ϑ pϑ,h (∆t0 ,tν+1 )(x, ξ) + i∂ϑ φϑ,h (∆t0 ,tν+1 )(x, ξ)pϑ,h (∆t0 ,tν+1 )(x, ξ) . Hence, by Proposition 3.3, Theorem 5.1 and Lemma 5.2, there exist a constant M2,h,l and an integer l2 such that (5.47) ph (∆t0 ,tν+1 )(φh (∆t0 ,tν+1 ); X, Dx ) − ph (t0 , tν+1 )(φh (t0 , tν+1 ); X, Dx ) u ≤ M2,h,l (t0 − tν+1 ) |u|l2 ,S l,S 2 ∆t0 ,tν+1 : (T0 ≥) t0 ≥ t1 ≥ · · · ≥ tν+1 (≥ 0) . 494 Naoto Kumano-go Similarly, there exist a constant M3,h,l and an integer l3 such that (5.48) i∂t + K(t, X, Dx ) ph (t, t1 )(φh (t, t1 ); X, Dx )u l,S ≤ M3,h,l (t − t1 )|u|l3 ,S (T0 ≥ t ≥ t1 ≥ 0) . (II) Let ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) and let ∆t,s be any refinement of ∆t,s . Then we can write (5.49) ph (∆t,s )(φh (∆t,s ); X, Dx ) − ph (∆t,s )(φh (∆t,s ); X, Dx ) u = ν+1 ph (∆t0 ,tj−1 )(φh (∆t0 ,tj−1 ); X, Dx ) j=1 ◦ ph (∆tj−1 ,tj )(φh (∆tj−1 ,tj ); X, Dx ) − ph (tj−1 , tj )(φh (tj−1 , tj ); X, Dx ) ◦ ph (∆tj ,tν+1 )(φh (∆tj ,tν+1 ); X, Dx )u . Hence, using (5.44) and (5.47), there exist a constant M4,h,l and an integer l4 such that (5.50) ph (∆t,s )(φh (∆t,s ); X, Dx ) − ph (∆t,s )(φh (∆t,s ); X, Dx ) u l,S ≤ M4,h,l |∆t,s |(t − s)|u|l4 ,S ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) . Therefore, there exists Uh (t, s)u ∈ S such that Uh (t, s)u = (5.51) lim |∆t,s |→0 ph (∆t,s )(φh (∆t,s ); X, Dx )u ( uniformly on [s, T0 ]) . Similarly, using (5.44), (5.47) and (5.48), there exist a constant M5,h,l and an integer l5 such that (5.52) ∂t ph (∆t,s )(φh (∆t,s ); X, Dx ) − ∂t ph (∆t,s )(φh (∆t,s ); X, Dx ) u l,S ≤ M5,h,l |∆t,s ||u|l5 ,S ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) . Fundamental solution for Schrödinger equations 495 Therefore, there exists Vh (t, s)u ∈ S such that (5.53) Vh (t, s)u = lim |∆t,s |→0 ∂t ph (∆t,s )(φh (∆t,s ); X, Dx )u ( uniformly on [s, T0 ]) . Hence, Uh (t, s)u is differentiable and ∂t Uh (t, s)u = Vh (t, s)u. Furthermore, using (5.44) and (5.48), there exist a constant M6,h,l and an integer l6 such that (5.54) i∂t + Kh (t, X, Dx ) ph (∆t,s )(φh (∆t,s ); X, Dx )u l,S ≤ M6,h,l (t − t1 )|u|l6 ,S ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) . Therefore, we have (5.55) lim |∆t,s |→0 i∂t + Kh (t, X, Dx ) ph (∆t,s )(φh (∆t,s ); X, Dx )u = 0 As a consequence, we get (5.56) Lh Uh (t, s)u = 0 on [s, T0 ] Uh (s, s)u = u (0 ≤ s ≤ T0 ) . (III) By Theorem 5.1 and Theorem 5.3, there exist a constant M7,h,l and an integer l7 such that (5.57) ph (∆t,s )(φh (∆t,s ); X, Dx ) − ph (t, s)(φh (t, s); X, Dx ) u l,S ≤ M7,h,l |∆t,s |(t − s)|u|l7 ,S ∆t,s : (T0 ≥) t ≡ t0 ≥ t1 ≥ · · · ≥ tν+1 ≡ s (≥ 0) . Hence, we have (5.58) Uh (t, s) = ph (t, s)(φh (t, s); X, Dx ) . (VI) Finally, we prove the uniqueness of Uh (t, s): 496 Naoto Kumano-go Let uh (t, s) ∈ C 1 ([s, T0 ]; S) be a solution of the Cauchy problem such that (5.59) Lh uh (t, s) = 0 on [s, T0 ] uh (s, s) = 0 . Set (5.60) Kh (t, ξ, x ) ≡ Kh (t, x , ξ) Lh ≡ i∂t + Kh (t, Dx , X ) . Then there exists Wh (t, x, ξ) such that Lh = i∂t + Hh (t, X, Dx ) + Wh (t, X, Dx ) (5.61) and (5.62) C0,0 x; ξ |∂ α Dxβ {W (t, hδ x, h−ρ ξ)}| ≤ ξ h Cα,β , Im{Wh (t, hδ x, h−ρ ξ)} ≤ C0,0 (α = β = 0) (|α + β| ≥ 1) is a constant independent of 0 < h < 1. on [0, T0 ] × R 2n . Here Cα,β Hence, for any 0 ≤ θ ≤ T0 , there exists a fundamental solution U (t, θ) (0 ≤ t ≤ θ) for the operator Lh such that Lh Uh (t, θ) = 0 (5.63) on [0, θ] Uh (θ, θ) = I . For any v(t) ∈ C 0 ([s, T0 ]; S), we set (5.64) zh (t) ≡ −i t T0 Uh (t, θ)v(θ)dθ (s ≤ t ≤ T0 ) . Then we have (5.65) Lh zh (t) = v(t) on [s, T0 ] zh (T0 ) = 0 . Fundamental solution for Schrödinger equations 497 Hence, using (5.59) and (5.65), we have (5.66) T0 (uh (t, s), v(t))dt s = s T0 (uh (t, s), Lh zh (t))dt t=T0 + = − i(uh (t, s), zh (t)) t=s T0 (Lh uh (t, s), zh (t))dt s = 0 (v(t) ∈ C 0 ([s, T0 ]; S)) . Therefore, we get uh (t, s) = 0 on [s, T0 ]. References [1] [2] [3] [4] [5] [6] [7] [8] [9] Feynman, R. P., Space-time approach to non relativistic quantum mechanics, Rev. of Modern Phys. 20 (1948), 367–387. Fujiwara, D., Remarks on convergence of some Feynman path integrals, Duke Math J. 47 (1980), 559–600. Fujiwara, D., The stationary phase method with an estimate of the remainder term on a space of large dimension, Nagoya Math. J. 124 (1991), 61–97. Fujiwara, D., Some Feynman Path Integrals As Oscillatory Integrals Over A Sobolev Manifold, preprint (1993). Kitada, H., Fourier integral operators with weighted symbols and micro-local resolvent estimates, J. Math. Soc. Japan 39 (1987), 455–476. Kitada, H. and H. Kumano-go, A family of Fourier integral operators and the fundamental solution for a Schrödinger equation, Osaka J. Math. 18 (1981), 291–360. Kumano-go, H., PSEUDO-DIFFERENTIAL OPERATORS, MIT press, Cambridge, Massachusetts and London, England, 1983. Kumano-go, H. and K. Taniguchi, Fourier integral operators of multi-phase and the fundamental solution for a hyperbolic system, Funkcial. Ekvac. 22 (1979), 161–196. Taniguchi, K., Multi-products of Fourier integral operators and the fundamental solution for a hyperbolic system with involutive characteristics, Osaka J. Math. 21 (1984), 169–224. (Received December 22, 1994) 498 Naoto Kumano-go Graduate School of Mathematical Sciences University of Tokyo 3-8-1, Komaba Meguro-ku, Tokyo 153 Japan