HAL Id: hal-01005809
https://hal.archives-ouvertes.fr/hal-01005809
Submitted on 13 Jun 2014
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Time-decay of semigroups generated by dissipative
Schrödinger operators
Xue Ping Wang
To cite this version:
Xue Ping Wang. Time-decay of semigroups generated by dissipative Schrödinger operators. Journal of Differential Equations, Elsevier, 2012, 253 (12), pp.3523-3542. �hal-01005809�
SCHR ¨ODINGER OPERATORS
XUE PING WANG
Abstract. We establish a representation formula for semigroup of contraction in terms of global limiting absorption principle from the upper-half complex plane. As applications, we prove time-decay estimates of the semigroup of contractions generated by −iH where H is a dissipative Schr¨odinger operator.
1. Introduction
Let H = −∆ + V (x) be the Schr¨odinger operator with a complex-valued potential V satisfying V = V1 − iV2, where V1 and V2 are real functions satisfying V2(x) ≥ 0 and
V2(x) > 0 on some non-trivial open set. Suppose that
|Vj(x)| ≤ Chxi−ρ0, x ∈ Rn, (1.1) ass1
for some ρ0 > 1. Here hxi = (1 + |x|2)1/2. Mild local singularities can be included
with little additional effort. Denote H0 = −∆ and H1 = −∆ + V1. H defined
on D(−∆) is maximally dissipative and the numerical range of H is contained in {z; ℜz ≥ −R, −R ≤ ℑz ≤ 0} for some R > 0.
2. Some abstract results
Let H1 and V2 be selfadjoint operators on some Hilbert space H, with H1
semi-bounded from below, V2 ≥ 0 and relatively compact with respect to H1. H = H1− iV2
is maximally dissipative on H. Let S(t) := e−itH, t ≥ 0, be the strongly continuous
semigroup generated by −iH. In this Section, we give two results on S(t) based on the existence of a limiting absorption principle for H on the whole real axis. There results are to be applied in the next Section to a class of dissipative Schr¨odinger operators on Rn, n ≥ 2.
Since V2 is H1 compact, one sees that ∀δ0 > 0, ∃R1 > 0 such that the numerical range
of H is contained in the sector
{z ∈ C; ℜz > −R1, −δ0 ≤ arg(z + R1) ≤ 0}.
2000 Mathematics Subject Classification. 35J10, 35P15, 47A55.
Key words and phrases. Time-decay, semigroup of contractions, non-selfadjoint Schr¨odinger operators.
Research supported in part by the French National Research Project NONAa, No. ANR-08-BLAN-0228-01, on Spectral and microlocal analysis of non-selfadjoint operators.
2 XUE PING WANG
Let δ0 < π be fixed. For ǫ0 > 0 small enough and for all ǫ ∈]0, ǫ0], the set {λeiǫ; λ ∈
R, |λ| > R1} is contained in the resolvent set of H. Assume that there exists a dense subset D ⊂ H such that {f ∈ D ∩ D(H); Hf ∈ D} is dense in H and
• For any λ ∈ R, the limit
hR(λ + i0)f, gi = lim
ǫ→0+
hR(λ + iǫ)f, gi (2.1) ass1
exists for any f, g ∈ D and is continuous in λ ∈ R.
• There exist R > 1, k ∈ N∗, ρ, σ > 0 with ρ + kσ > 1 such that for any f, g in D,
λ → hR(λ + i0)f, gi is Ck for |λ| > R and
| d
j
dλjhR(λ + i0)f, gi| ≤ Cf,ghλi
−ρ−jσ, (2.2)
ass2
for j = 0, 1, . . . , k and |λ| > R.
Theorem 2.1. Under the conditions (2.1) and (ass1 2.2), one hasass2 he−itHf, gi = 1
2πi Z
R
e−itλhR(λ + i0)f, gidλ, t > 0, (2.3)
for f, g ∈ D.
Proof. Denote Hǫ = e−iǫH, ǫ > 0. Then the numerical range of Hǫ is contained in
Nǫ := {e−iǫz; ℜz > −R1, −δ0 ≤ arg(z + R1) ≤ 0}
for some R1 > 1. Therefore for each ǫ > 0 small enough, Hǫ − iR1ǫ is maximally
dissipative and strictly m-sectorial and −iHǫ generates a semigroup e−itHǫ, t ≥ 0, which
can be represented in a usual way (cf. [12], pp 489-491). For Rk 0 > ǫR1, let Γǫ,R be a
contour in ρ(Hǫ) composed of the segment {ℑz = R0, ℜz ∈ [−R1− 1, R1+ 1]} and the
two rays (−R1− 1 + iR0) + ei(π+ǫ/2)R+ and (R1+ 1 + iR0) + e−iǫ/2R+ (running from the
infinity with arg z = −π + ǫ2 to the infinity with arg z = −2ǫ). Then one has e−itHǫ = 1 2πi Z Γǫ,R0 e−itz(H ǫ− z)−1dz := Fǫ(t), (2.4) Uepsilon
for t > 0. In addition, one has the estimate
ke−itHǫk ≤ etR1ǫ, t ≥ 0. (2.5) norm1
Under the condition (2.2), by an argument of perturbation, one can deduce that forass2 j = 0, 1, · · · , k, | d j dλjhR(λe iη)f, gi| ≤ C f,ghλi−ρ−jσ, (2.6)
uniformly in λ ≥ 1 and η > 0 small enough. Making use of techniques of oscillatory integrals, we deduce from (ass12.1) and (ass22.2) that the integral
hF (t)f, gi := 1 2πi
Z
R
e−itλhR(λ + i0)f, gidλ, (2.7)
converges for f, g ∈ D. By the same method, one can show that the integral defining Kǫ(t) converges uniformly in ǫ > 0 and one has
hF (t)f, gi = lim
R,ǫ→0+
for any f, g ∈ D. The bound on kFǫ(t)k shows that |hF (t)f, gi| ≤ kf kkgk. Therefore
F (t) can be extended to a contraction on H, still denoted by F (t) and limǫ→0+Fǫ(t) =
F (t) weakly for t > 0.
For f ∈ D ∩ D(H) with Hf ∈ D and g ∈ D, one has Z
Γǫ,R0
d dthe
−itz(H
ǫ− z)−1f, gidz = −ie−iǫ
Z
Γǫ,R0
he−itz(H
ǫ− z)−1Hf, gidz
converges uniformly in ǫ, R0 > 0 small. It follows that hF (t)f, gi is differentiable in
t > 0 and
hdF (t)
dt f, gi = h−iF (t)Hf, gi = h−iHF (t)f, gi.
By an argument of density, we deduce that for any f ∈ D(H), t → F (t)f is weakly differentiable and
dF (t)
dt f = −iF (t)Hf = −iHF (t)f. (2.8)
To show that F (t) = e−itH, we prove that F (t) → 1 weakly as t → 0
+. For each
t ∈]0, 1], take R0 = t−1 ≥ 1. Making a change of variables, we obtain that
Fǫ(t) = 1 2πi Z Γǫ,1 e−iζζ−1(Hǫ− ζ t) −1dζ t , t ∈]0, 1]. Noticing that 1 2πi Z Γǫ,1 e−iζζ−1dζ = −1 for every ǫ > 0, one deduces
h(Fǫ(t) − 1)f, gi = 1 2πi Z Γǫ,1 e−iζζ−1h(H −ζe iǫ t ) −1Hf, gidζ (2.9)
for f ∈ D(H). For ζ ∈ Γǫ,1, one has |ζ| ≥ 12 and |e−iζ| ≤ C uniformly in ǫ > 0 small.
By the condition (2.2),ass2
|h(H −ζe iǫ t ) −1Hf, gi| ≤ C f,g( t |ζ|) ρ, t ∈]0, 1],
for any f, g ∈ vD with Hf ∈ D, uniformly in ǫ > 0. It follows that
|h(Fǫ(t) − 1)f, gi| ≤ C′tρ, t ∈]0, 1], (2.10)
uniformly in ǫ. Taking the limit ǫ → 0 in the above inequality, we obtain
|h(F (t) − 1)f, gi| ≤ C′tρ. (2.11)
This proves that limt→0+h(F (t) − 1)f, gi = 0 for any g ∈ D and f ∈ D ∩ D(H) with
Hf ∈ D. Since kF (t) − 1k ≤ 2, an argument of density shows that F (t)f → f weakly for any f ∈ H as t → 0+.
Now for any f ∈ D(H), let f (t) = F (t)f and u(t) = e−itHf . Then f (t) ∈ D(H) for
any t > 0 and one has d
dshF (s)e
4 XUE PING WANG
for any g ∈ H. Integrating the above equation gives that hf (t), gi − hu(t), gi = c for some constant c. Since both f (t) and u(t) converge weakly to f as t → 0+, one has
c = 0, hence hf (t), gi = hu(t), gi for any g ∈ H. This proves that F (t)f = e−itHf for
any f ∈ D(H). Since D(H) is dense in H, F (t) coincides with the semigroup generated
by −iH. ¤
Another consequence of global limiting absorption principle is the Kato’s smoothness estimate for semigroup of contractions which is useful for dissipative quantum scattering. We give below a simple proof, using the theory of selfadjoint dilation. See also roy[16] in some special case.
Let H be maximal dissipative on a Hilbert space H. −iH is generator of a semigroup of contractions T (s) = e−isH, t ≥ 0. According to the theory of Foia¸s-Sz. Nagy
(Masson,1967, Ch.III, § 9), ∃ a Hilbert space G ⊃ H and a unitary group U (t) = e−itG
on G such that
Π0U (s)|H = T (s), s ≥ 0, (2.12)
where Π0 : G → H is the projection. G is called a selfadjoint dilation of H.
th2.2 Theorem 2.2. Assume that there exits A : H → H continuous such that sup λ∈R,δ∈]0,1] kA(H − (λ + iδ))−1A∗k ≤ γ. (2.13) Then Z ∞ 0 (kAT (s)f k2+ kAT (s)∗f k2)ds ≤ 2γkf k2, f ∈ H. (2.14)
Proof. Let G be a selfadjoint dilation of H. Then
Π0(G − z)−1|H = (H − z)−1, Π0(G − z)−1|H = (H∗− z)−1,
for ℑz > 0. Therefore
k(AΠ0)(G − z)−1(AΠ0)∗k ≤ γ, 0 < |ℑz| ≤ 1.
By Kato’s smoothness estimate for selfadjoint operators,
Z ∞ −∞ kAΠ0U (s)gk2 ds ≤ Ckgk2, g ∈ G, with C = sup 0<ℑz≤1 k(AΠ0)[(G − z)−1− (G − z)−1)(AΠ0)∗k ≤ 2γ.
For g = f ∈ H, one has
Z ∞
0
(kAT (s)f k2+ kAT (s)∗f k2) ds ≤ 2γkf k2, f ∈ H.
3. Dissipative Schr¨odinger operators
Let R(z) = (H − z)−1, z 6∈ σ(H). Let R
j(z) = (Hj − z)−1. Denote L2,s =
L2(Rn; hxisdx) and kf k
s = kf kL2,s. It is well know that if n ≥ 3, the limit
F0 = lim
z→0,z6∈R+
R0(z) : L2,s→ L2,−s
exists if s > 1.
prop1 Proposition 3.1. Assume that n ≥ 3 and ρ0 > 2. Then one has
(a). 1 + F0V is invertible on L2,−s for any s ∈]1, ρ0/2[ and there exists c0 > 0 such
that the limit
R(λ + i0) = lim
ǫ→0+
R(λ + iǫ) : L2,s→ L2,−s exists for s > 1 and λ ∈ [−c0, c0].
(b). Zero is not an accumulating point of the eigenvalues of H and there exists δ0 > 0
such that
σ(H) ⊂ {z; −π + δ0 ≤ arg z ≤ 0}. (3.1)
This result is proved by X.P. Wang 2009.
th1 Theorem 3.2. Assume that n ≥ 3 and ρ0 > 2. Then
(a). For any s > 1
sup
ǫ∈]0,1],λ∈R
hλi12khxi−sR(λ + iǫ)hxi−sk < ∞ (3.2) global-resolv
The limit
R(λ + i0) = lim
ǫ→0+
R(λ + iǫ) : L2,s→ L2,−s
exists for s > 1 and is continuous on R.
(b) Let k ∈ N. Assume that |(x · ∇)jV (x)| ≤ Chxi−ρ0, j = 0, 1, · · · , k, ρ
0 > 2. Then for j = 0, 1, · · · , k + 1 and s > j + 1 2, one has khxi−s dj dλjR(λ + i0)hxi −sk ≤ C shλi− j+1 2 , (3.3) for λ > 1.
Ideas. For λ > 0, one uses the equation R(z) = (1 + R0(z)V )−1R0(z) by the argument
of S. Agmon (1975) (ρ0 > 1). For more general situations (an abstract Mourre’s theory
for dissipative operators), see J. Royer (Commun. in PDE, to appear). For λ near 0, we apply Proposition . n ≥ 3 and ρ0 > 2 are needed.
prop1
4. Optimal time-decay rate
As an application of the representation of the semigroup, we can show the following dispersive estimate for dissipative Schr¨odinger operators.
dispersive Theorem 4.1. Assume n = 3 and |V (x)| + |x · ∇V (x)| ≤ Chxi−ρ0, ρ
0 > 2. Then one has ke−itHf k L∞ ≤ Ct− 3 2kf kL1, ∀ f ∈ L1(R3), t > 0. (4.1) dispative-disp
6 XUE PING WANG
To prove the Theorem, it suffices to prove |hU (t)u, vi| ≤ C|t|−3/2kuk
L1kvkL1, u, v ∈ C0∞.
Distinguish two regimes: λ ∈ [−R, R] and |λ| > R. Consider only the case λ ≥ 0. By the change of variable λ → λ2 and an integration by parts, we are led to prove
| Z
e−itλ2ρ(λ)hG′(λ)u, vi dλ| ≤ C|t|−1/2kuk1kvk1. (4.2) est1
Here ρ is an appropriate cut-off and G(λ) = r(λ2) which can be written as G(λ) =
G0(λ)(1 + V G0(λ))−1.
One can calculate
G′(λ) = (1 − G0(λ)V )G′0(λ)(1 + V G0(λ))−1 (4.3) G
and G′
0(λ) is the operator with integral kernel: ie
iλ|x−y|
4π .
Take ρ with support in ]λ0− δ, λ0+ δ[, δ > 0 to be adjusted. If one replaces λ by λ0
in (1 − G0(λ)V ) and (1 + V G0(λ))−1 in (
G
4.3), one sees that
| Z R3 x Z R3 y Z R
e−itλ2+iλ|x−y|ρ(λ)˜u(x)˜v(y)dλdxdy| ≤ C|t|−1/2kuk1kvk1.
Note that V G0(λ) : L1 → L1 is H¨older-continuous in λ. For λ ∈ supp ρ, one expands
(1 + V G0(λ))−1= ∞
X
k=0
(−1)k(S0D(λ))kS0
with S0 = 1 + V G0(λ0), D(λ) = V (G0(λ) − G0(λ0)). The integral kernel of D(λ) is
V(x)e
iλ|x−y|− eiλ0|x−y| 4π|x − y| .
One can prove that Z R kFλ→τρ(λ)(S0D(λ))kS0uk1dτ ≤ (CVδǫ)kkuk1. It follows that | Z R
hG′0(λ)e−itλ2ρ(λ)(S0D(λ))kS0u, A(λ)vidλ| ≤ |t|−1/2(CVδǫ)kkuk1kvk1.
Here A(λ) = 1 − V G0(λ0) or D(λ).
Taking δ s.t. CVδǫ < 1 and summing up in k, we obtain the desired estimate (
est1
4.2) for each fixed energy.
Little modification is needed when λ0 = 0.
The high energy estimate in the limiting absorption principle implies k(V G0(λ))kuk1 ≤ Ckλ−(k−2)kuk1
Lemma 4.2. Let fk : R → R be a sequence of continuous functions such that |fk(λ)| ≤ Ckhλi−k+2, Z R hτ iǫ0| ˆf k(τ )|dτ ≤ Ck.
Then for χ a cut-off for the interval [R, ∞[, one has Z
|dχfk(τ )|dτ ≤ C1khRi
−(k−2)ǫ0
2(1+ǫ0).
One uses the series for λ large
(1 + V G0(λ))−1 = ∞
X
k=0
(−1)k(V G0(λ))k
By the decay condition on V , one has Z
hτ iǫ0kF
λ→τρ(λ)(V G0(λ))kuk1dτ ≤ Ckkuk1,
uniformly in R. Here ρ is a cut-off for [R, ∞[. The Lemma gives Z
R
kFλ→τρ(V G0(λ))kuk1dτ ≤ CkR−
(k−2)ǫ0 2(1+ǫ0)kuk1.
For R > 1 large enough,
CkR−(k−2)ǫ02(1+ǫ0) < ǫk, ǫ < 1.
One can deduce that
| Z
e−itλ2ρ(λ)hG′0(λ)(V G0(λ))ku, A(λ)vidλ| ≤ |t|−1/2Cǫkkuk1kvk1.
Taking the summation in k, we obtain the desired high energy estimate. ¤
This result is to compare with the dispersive estimate for selfadjoint operator H1
(V2 = 0). Assume that n = 3 and |V1(x)| ≤ Chxi−ρ0, ρ0 > 2 and that 0 is not an
eigenvalue nor a resonance of H1. Then one has
ke−itH1P
acf kL∞ ≤ Ct− 3
2kf kL1, ∀ f ∈ L1(R3), t 6= 0, (4.4)
where Pac is the projection onto the absolutely continuous spectral subspace of H1.
As a consequence of Theorem 4.1, one hasdispersive
Corollary 4.3. Under the condition of Theorem 4.1, one has for any s ∈ [0, 3/2] anddispersive s′ > s,
khxi−s′e−itHhxi−s′kL(L2) ≤ Chti−s, t > 0. (4.5)
Proof. For s = 3/2 and s′ > 3/2, it follows from Theorem dispersive4.1 that the above estimate
8 XUE PING WANG
References
ag [1] S. Agmon, Spectral properties of Schr¨odinger operators and scattering theory. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) (1975), no. 2, 151-218.
dav1 [2] E. B. Davies, Two-channel Hamiltonians and the optical model for nuclear scattering, Ann. Inst. H. Poincar´e, Vol. XXIX(4) (1978), 395-413.
dav2 [3] E. B. Davies, Non-unitary scattering and capture. I. Hilbert space theory, Ann. Inst. H. Poincar´e, Commun. Math. Phys., 71(1980), 277-288.
dav3 [4] E. B. Davies, Non-unitary scattering and capture. II. Quantum dynamical semigroup theory, Ann. Inst. H. Poincar´e, Vol. XXXII(4) (1978), 361-375.
dav4 [5] E. B. Davies, Non-self-adjoint differential operators, Bull. London Math. Soc., 34(2002), 513-532.
god [6] M. Goldberg, Dispersive bounds for the three-dimensional Schr¨odinger equation with almost crit-ical potentials, GAFA, 16(2006), 517-536.
fes [7] H. Feshbach, Unified theory of nuclear reactions. I., Ann. Phys., 5(1958), 357-390.
hor [8] L. H¨ormander, The Analysis of Linear Partial Differential Operators. II, Springer, 1983.
ij [9] A. Ionescu and D. Jerison, On the absence of positive eigenvalues of Schr¨odinger operators with rough potentials, Geom. and Funct. Anal., 13(2003), 1029-181.
jk [10] A. Jensen and T. Kato, Spectral properties of Schr¨odinger operators and time decay of wave functions, Duke Math. J., 46(1979), 583-611.
ky [11] T. Kako, K. Yajima, Spectral and scattering theory for a class of non-selfadjoint operators. Sci. Papers College Gen. Ed. Univ. Tokyo 26 (1976), no. 2, 73-89
k [12] T. Kato, Wave operators and similarity for some non-selfadjoint operators, Math. Annalen, 162(1966), 258-279.
kat2 [13] T. Kato, Perturbation Theory of Linear Operators, Springer-Verlag, Berlin, 1980.
mar [14] Ph. A. martin, Scattering theory with dissipative intercations and time delay, Nuovo Cim., (30)B (1975), 217-238.
pav1 [15] B. S. Pavlov, Selfadjoint dilation of the dissipative Schr¨odinger operator and its resolution in terms of eigenfunctions, Math. URSS Sbornik, 31(4)(1977), 457-478.
roy [16] J. Royer, Limiting absorption principle for the dissipative Helmholtz equation, Commun. in PDE, to appear.
sa [17] Y. Saito, The principle of limiting absorption for the nonselfadjoint Schr¨odinger operator in RN(N 6= 2). Publ. Res. Inst. Math. Sci. 9 (1973/74), 397-428.
sa2 [18] Y. Saito, The principle of limiting absorption for the non-selfadjoint Schr¨odinger operator in R2.
Osaka J. Math. 11 (1974), 295-306.
sim [19] B. Simon, Phase space analysis of simple scattering systems: extensions of some work of Enss, Duke Math. J., 40(1)(1980), 119-168.
sch [20] J. T. Schwartz, Some non-selfadjoint operators, Comm. Pure and Appl. Math., 13(1960), 609-639.
szf [21] B. Sz. Nagy, C. Foia¸s, Harmonic analysis of operators on Hilber space, North Holland, Amsterdam, 1970.
w2 [22] X.P. Wang: Asymptotic expansion in time of the Schr¨odinger group on conical manifolds, Ann, Inst. Fourier (Grenoble), 56(2006), 1903-1945
w3 [23] X. P. Wang, Number of eigenvalues for dissipative Schr¨odinger operators under weak perturbation, preprint 2009.
D´epartement de Math´ematiques, UMR 6629 CNRS, Universit´e de Nantes, 44322 Nantes Cedex 3 France, E-mail: xue-ping.wang@univ-nantes.fr