Floquet Perturbation Theory Formalism and Applicat

ACCEPTED MANUSCRIPT • OPEN ACCESS
Floquet perturbation theory: formalism and application to low-frequency

limit
To cite this article before publication: Martin Rodriguez-Vega et al 2018 New J. Phys. in press https://doi.org/10.1088/1367-2630/aade37
Manuscript version: Accepted Manuscript

Accepted Manuscript is “the version of the article accepted for publication including all changes made as a result of the peer review process,
and which may also include the addition to the article by IOP Publishing of a header, an article ID, a cover sheet and/or an ‘Accepted
Manuscript’ watermark, but excluding any other editing, typesetting or other changes made by IOP Publishing and/or its licensors”
This Accepted Manuscript is © 2018 The Author(s). Published by IOP Publishing Ltd on behalf of Deutsche Physikalische
Gesellschaft.
As the Version of Record of this article is going to be / has been published on a gold open access basis under a CC BY 3.0 licence, this Accepted
Manuscript is available for reuse under a CC BY 3.0 licence immediately.
Everyone is permitted to use all or part of the original content in this article, provided that they adhere to all the terms of the licence
https://creativecommons.org/licences/by/3.0
Although reasonable endeavours have been taken to obtain all necessary permissions from third parties to include their copyrighted content
within this article, their full citation and copyright line may not be present in this Accepted Manuscript version. Before using any content from this
article, please refer to the Version of Record on IOPscience once published for full citation and copyright details, as permissions may be required.
All third party content is fully copyright protected and is not published on a gold open access basis under a CC BY licence, unless that is
specifically stated in the figure caption in the Version of Record.
View the article online for updates and enhancements.
This content was downloaded from IP address 191.101.212.12 on 01/09/2018 at 01:58

Page 1 of 28 AUTHOR SUBMITTED MANUSCRIPT - NJP-108725.R2
1
2
3
4
pt
5
6
7
8
9
Floquet Perturbation Theory: Formalism and
cri
10
11 Application to Low-Frequency Limit
12
13
14 M Rodriguez-Vega1,3 , M Lentz2 , and B Seradjeh1,3
15 1
Department of Physics, Indiana University, Bloomington, IN 47405 USA
16
us
2
17 Department of Physics, Syracuse University, Syracuse, NY 13244 USA
3
18 Max Planck Institute for the Physics of Complex Systems, Nöthnitzer Straße 38,
19 Dresden 01187 Germany
20 E-mail: babaks@indiana.edu
21
22
23
24
25
26
27
28
an
Abstract. We develop a low-frequency perturbation theory in the extended Floquet
Hilbert space of a periodically driven quantum systems, which puts the high-
and low-frequency approximations to the Floquet theory on the same footing. It
captures adiabatic perturbation theories recently discussed in the literature as well as
diabatic deviation due to Floquet resonances. For illustration, we apply our Floquet
perturbation theory to a driven two-level system as in the Schwinger-Rabi and the
dM
29 Landau-Zener-Stückelberg-Majorana models. We reproduce some known expressions
30 for transition probabilities in a simple and systematic way and clarify and extend
31 their regime of applicability. We then apply the theory to a periodically-driven system
32 of fermions on the lattice and obtain the spectral properties and the low-frequency
33
dynamics of the system.
34
35
36
37
38
pte
39
40
41
42
43
44
45
46
ce
47
48
49
50
51
52
Ac
53
54
55
56
57
58
59
60
AUTHOR SUBMITTED MANUSCRIPT - NJP-108725.R2 Page 2 of 28
1
2
3 Floquet Perturbation Theory: Formalism and Application to Low-Frequency Limit 2
4
pt
5 1. Introduction
6
7 Driving a system’s parameters periodically in time leads to qualitatively new phenomena
8 that are absent in equilibrium. Well-known examples of such phenomena in classical
9
cri
10 systems include parametric resonance and stability [1]. In quantum systems, a
11 well-known consequence of periodic driving is the Rabi oscillation in a two-level
12 system [2]. More recently, the repertoire of such phenomena has been expanded
13
14 to many-body quantum systems [3–5], including the appearance of non-equilibrium
15 topological phases [6–19] and, in the presence of interactions and/or disorder, many-
16
us
body localized phases [20–22] that exhibit subharmonic oscillations, thus realizing a
17
18 time crystal [23–26]. Also, recent experimental advances have allowed the realization of
19 driven optical lattices [27–31].
20 Analytically, the appearance of these novel features is usually understood within a
21
22 high-frequency approximation, e.g. the rotating-wave approximation, Floquet-Magnus
23
24
25
26
27
28
an
expansion, and Brillouin-Wigner theory [32–36]. These approximations often break
down as frequency is lowered below the typical energy scale of the static system, such
as the bandwidth or an equilibrium insulating gap. Though in certain cases other
perturbative schemes, such as the Schrieffer-Wolff theory [37], provide valuable insight
dM
away from the high-frequency regime, understanding the low-frequency behavior of these
29
30 novel phases remains challenging. In the opposite limit of vanishingly small frequency,
31 one may expect the dynamics be governed by adiabatic evolution. Perturbative
32 methods to account for diabatic correction to this adiabatic evolution have been
33
34 developed [38–42]. However, the connection between these methods and the Floquet
35 theory used for higher frequencies is not clear.
36 In this paper, we develop a systematic perturbation theory based on the Floquet
37
38 theorem within the extended Floquet Hilbert space furnishing the steady states of
pte
39 a periodically driven quantum system [43, 44]. Our approach is general and works
40 whenever an operator in the Floquet Hamiltonian describing the dynamics of the
41
42 system in the extended Floquet Hilbert space can be taken to be small. Indeed, we
43 show how this Floquet perturbation theory leads to perturbative expansion both in the
44 high- and the low-frequency limits. In both cases, we reproduce previous results in a
45
46 compact and efficient way and show how higher-order terms is worked out systematically.
ce
47 Moreover, using this formalism we expand the applicability of these results and show
48 when deviations are expected. In the low-frequency limit, we clarify the deviations
49
50 from adiabatic evolution near quasienergy resonances [47] that lead to Rabi oscillations.
51 Finally, using our Floquet perturbation theory, we study a system of non-interacting
52 fermions moving on a driven one-dimensional lattice at low frequency [18, 48]. We
Ac
53
54
derive the Floquet spectrum and show when the low-frequency limit does and does not
55 approach the adiabatic evolution.
56 We note that in the low-frequency limit the periodicity assumed in the Floquet
57
58
theory is not a real restriction for reproducing the results of the adiabatic perturbation
59 theory for a general drive. Basically, in this limit one can think of any drive as one
60
1
2
4
pt
5 big cycle of a periodic drive and find the desired evolution at any time mid cycle. The
6 additional periodic structure in the Floquet theory is important only if when one wishes
7
to study the Floquet spectra of an actual periodic drive. One may call this low-frequency
8
9 Floquet perturbation theory the “Floquet adiabatic perturbation theory;” however, this
cri
10 term is already used in the literature [49–52] to describe the evolution of a driven system
11 when a parameter of the drive is slowly varied. To avoid confusion, we do not use this
12
13 terminology.
14 The paper is organized as follows. In Section 2, the Floquet perturbation theory
15 is developed within the extended Floquet Hilbert space and used to derive high- and
16
us
17 low-frequency series expansions of the Floquet spectrum. Formal aspects of the theory
18 are presented in Appendix A. In Section 3, we illustrate the formalism by applying
19 it to transition probabilities in a driven two-level system, described separately by the
20
21 Rabi-Schwinger and the Landau-Zener-Stückelberg-Majorana models. In Section 4, we
22 develop the degenerate low-frequency Floquet perturbation theory and demonstrate its
23
24
25
26
27
28
an
application near quasienergy degeneracies in the low-frequency regime of the Landau-
Zener model as well as the driven Su-Schrieffer-Heeger model of non-interacting fermions
moving on a one-dimensional lattice. We conclude with a summary and outlook in
Section 5. Some technical details of our calculations are given in Appendix B.
dM
29
30 2. Floquet Perturbation Theory
31
32
2.1. Floquet Theory and the Extended Floquet Hilbert Space
33
34
Floquet theorem is the statement that the solution to a differential equation with
35
36 periodic coefficients can be written as a phase factor multiplied by a periodic function.
37 A direct consequence of this statement in the condensed matter setting is the Bloch
38
theorem for the solution to the Schrödinger equation in the presence of a spatially
pte
39
40 periodic potential due to a lattice. In our discussion, we reserve the Floquet theorem
41 for a system with parameters that are periodic in time, t. The details of the Floquet
42
theory formalism are presented in Appendix A; here, we provide a summary.
43
44 For a Hamiltonian Ĥ(t) = Ĥ(t + T ) with period T = 2π/Ω, Floquet theorem states
45 that the time-dependent Shrödinger equation i dt d
|ψ(t)i = Ĥ(t)|ψ(t)i takes steady-state
46 solutions of the form
ce
47
48 |ψα (t)i = e−iα t |φα (t)i, (1)
49
50 where the quasienergy α is a conserved quantity and the periodic Floquet mode
51 |φα (t)i = |φα (t + T )i satisfies the Floquet Schrödinger equation
52
Ac
53

d
54 Ĥ(t) − i |φα (t)i = α |φα (t)i. (2)
dt
55
56 The |φα (t)i form a time-dependent orthonormal basis for the Hilbert space H and can
57 d
be viewed as the eigenstates of the time-dependent Floquet Hamiltonian Ĥ(t) − i dt with
58
59 time-independent eigenvalues belonging to the Floquet zone, α ∈ [−Ω/2, Ω/2]. Using
60
1
2
4
pt
5 Floquet theorem, the evolution operator
6 Z t
7 Û (t, t0 ) = Texp −i Ĥ(s)ds , (3)
8 t0
9 with t0 < t and Texp the time-ordered exponential, can be decomposed as
cri
10
11 Û (t, t0 ) = Φ̂(t)e−i(t−t0 )ĤF Φ̂(t0 )† , (4)
12
13 where
14 X
15 e−itĤF = e−iα t |φα (0)i hφα (0)| , (5a)
16 α
us
17
X
Φ(t) = |φα (t)i hφα (0)| , (5b)
18
α
19
20 define, respectively, the Floquet Hamiltonian ĤF and the micromotion operator Φ̂(t).
21 ˆ We could choose a different boundary condition by a change
Here, we set Φ̂(0) = I.
22
23
24
25
26
27
28
an
of basis to |γα i = Γ̂† |φα (0)i, where Γ̂ is a unitary operator. In this basis, the Floquet
Hamiltonian is Γ̂† ĤF Γ̂ and the micromotion operator Φ̂Γ (t) = α |φα (t)i hγα | = Φ̂(t)Γ̂,
P
with Φ̂Γ (0) = Γ̂. This freedom can lead to different truncated Floquet perturbative
expansions, if Γ̂ depends on the perturbation parameter itself [33, 34]. We shall see an
example of this in Sec. 2.3. The evolution operator is independent of this choice.
dM
29
The structure we have described above can be formalized in terms of an extended
30
31 Floquet Hilbert space F = H ⊗I , where the auxiliary space I is the space of bounded
32 periodic function over [0, T ) [44]. We denote the states in H , I , and F respectively
33 ˆ.
“ and Ô
by |·i , |·), and |·ii and the operators acting on each respective space as Ô, O,
34
35 The space I is spanned by a continuous orthonormal basis {|t)}, 0 ≤ t < T ,
36 Z T
dt
37 0 0
(t |t) = T δ(t − t ), |t)(t| = I,“ (6)
38 0 T
pte
39
40
where I“ is the identity operator in I . The auxiliary space I is also spanned by the
41 orthonormal Fourier basis
42 Z T
dt
43 |n) = e−inΩt |t) , n ∈ Z, (7)
44 0 T
45 satisfying
46 X
“
ce
47 (n|m) = δnm , |n)(n| = I. (8)

48 n∈Z
49
50 We note
51
X
|t) = einΩt |n). (9)
52
Ac
n∈Z
53
54 A loop in H is a one-parameter family of states |φ(t)i that is cyclic, i.e. |φ(T )i =
55 |φ(0)i. It can be lifted to a loop in F given by |φt ii := |φ(t)i |t). Associated with any
56
57 loop |φt ii ∈ F is the center of the loop,
58 Z T
dt
59 |φii ≡ |φt ii . (10)
60 0 T
1
2
4
pt
5 We define the Fourier-integral and the time-derivative operators,
6 Z T
dt
7 µ̂ ˆ
ˆn = I ⊗ |t)e+inΩt (t| , (11a)
8 0 T
9 Ẑˆt = Iˆ ⊗
X
|n)nΩ(n|, (11b)
cri
10
n∈Z
11
12 such that for a loop |φt ii ∈ F ,
13 Z T
dt
14 ˆ
µ̂n |φii = e+inΩt |φt ii ≡ |φn ii , (12)
15 0 T
16 is the n-th Fourier integral and
us
17 Z T
18 ˆ d |φ(t)i dt
Ẑt |φii = i |t) ≡ i |dφ/dtii , (13)
19 0 dt T
20
21 is the center of the time-derivative of the loop. Then, the Floquet Schrödinger Eq. (2)
22 can be written in F as
23
24
25
26
27
28
where Ĥ
ˆ − Ẑˆ ) |φ ii = |φ ii ,
(Ĥ t α
ˆ = T Ĥ(t) ⊗ |t)(t| dt .
R
0 T
α α
an (14)
Note, however, that the set {|φα ii} of solutions to Eq. (14) is not large enough to
furnish a complete basis for F . Noting that [Ĥ ˆ , µ̂
ˆn ] = 0 and
dM
29
30 ˆ , Ẑˆ ] = nΩµ̂
[µ̂ n t ˆ ,
n (15)
31
32
we can write instead
33 ˆ − Ẑˆ ) |φ ii = |φ ii ,
(Ĥ (16)
t αn αn αn
34
35 where αn ≡ α +nΩ and |φαn ii ≡ µ̂ ˆn |φα ii. Indeed, µ̂ ˆ − Ẑˆ ,
ˆn is the ladder operator for Ĥ
36 mapping the solution |φα ii with quasienergy α to |φαn ii with quasienergy α +nΩ. Now,
37
38 the solutions |φαn ii to Eq. (16) provide a full basis for F .
pte
39
40 2.2. Floquet Perturbation Theory
41
42 Let us recap the Floquet perturbation theory [43] in the above language. The Floquet
43
44 Schrödinger equation (16) can be inverted in F to give the Floquet Green’s function
45 X |φαn ii hhφαn |
ˆ () ≡ ( − Ĥ
Ĝ ˆ + Ẑˆ )−1 = . (17)
46 t
− α −nΩ
ce
47 αn
48 The Floquet Green’s function can be employed to calculate a variety of responses of the
49
50 driven system. In this work, we focus on its application to perturbation theory.
51 For the periodic Hamiltonian Ĥ(t) = Ĥ0 (t) + V̂ (t), where Ĥ0 (t) is the unperturbed
52 Hamiltonian and V̂ (t) is the perturbing potential (both having a common period T ),
Ac
53 ˆ = Ĥˆ + V̂ˆ in F . Paralleling the conventional time-independent

54
we lift Ĥ(t) to Ĥ 0
55 perturbation expansion in F , we then expand the solutions to the Floquet Schrödinger
56 equation as
57
58 α = α(0) + α(1) + α(2) + · · · , (18a)
59
60
|φα ii = |φα(0) ii + |φα(1) ii + |φα(2) ii + · · · , (18b)
1
2
4
ˆ − Ẑˆ ) |φ ii =
pt
5 with (Ĥ0 α(0) α(0) |φα(0) ii, to find for i ≥ 1,
6
7 α(i) = hhφα(0) | V̂ˆ |φα(i−1) ii , (19a)
8
" i−1
#
ˆ ˆ ˆ ˆ
X
9 |φ ii = P̂ Ĝ P̂ V̂ |φ
α(i) α 0α α ii −
α(i−1) α(i−j) |φα(j) ii , (19b)
cri
10 j=1
11
12 ˆ ˆ ˆ −1 is the Floquet Green’s function for Ĥ
ˆ , P̂ˆ =
where Ĝ 0α = (α(0) − Ĥ0 + Ẑt ) 0 α
13 ˆ
I − |φα(0) ii hhφα(0) | projects to the subspace of F that is orthogonal to |φα(0) ii, and
14
15 we have assumed the standard normalization hhφα(0) |φα ii = 1.
16 Explicitly,
us
17 Z T
18 dt
α(1) = hφα(0) (t)| V̂ (t) |φα(0) (t)i , (20a)
19 0 T
20
X hhφβn(0) | V̂ˆ |φα(0) ii
21 |φα(1) ii = |φβn(0) ii , (20b)
22 α(0) − β(0) −nΩ
23
24
25
26
27
28
and
α(2) =
(β,n)6=(α,0)
(β,n)6=(α,0)
an
| hhφα(0) | V̂ˆ |φβn(0) ii |2
α(0) − β(0) −nΩ
dM
R 2
29 T
hφ (t)| e +inΩt
V̂ (t) |φ (t)i dt
α(0) β(0)
30 0 T
X
= , (21)
31 α(0) − β(0) −nΩ
32 n 6= 0
β 6= α
33
34 etc.
35
36
37 2.3. High-Frequency Expansion
38
As an example, we derive a high-frequency expansion using the Floquet perturbation
pte
39
40 theory (see Ref. [34] for a detailed discussion). Assuming the frequency is larger
41 than the typical quasienergy, we shall take the unperturbed Hamiltonian Ĥ ˆ = 0 and
0
42 ˆ ˆ
43
V̂ = Ĥ . Therefore, α(0) = 0, and |φα(0) ii can be obtained from lifting an arbitrary time-
44 independent set |φα i in H to F . Since the unperturbed quasienergies are degenerate
45 for the same n, we employ degenerate perturbation theory, noting the matrix elements:
46
ˆ |φ ii = hφ | Ĥ (0) |φ i ,
ce
47 hhφαn | Ĥ βn α β (22)
48 T
where the Fourier components Ĥ (n) = 0 einΩt Ĥ(t) dt
R
49 T
. Thus, at the lowest order, |φα i
(0)
50 are chosen as the eigenstates of Ĥ . After some algebra, using Eqns. (20a), (20b), and
51 (21), we find
52
Ac
53 α(1) = hφα | Ĥ (0) |φα i , (23a)

54 X [Ĥ (−n) , Ĥ (n) ]
55 α(2) = hφα | |φα i , (23b)
56 2nΩ
n6=0
57
58 X Ĥ (−n) einΩt
59 |φα(1) (t)i = |φα i . (23c)
nΩ
60 n6=0
1
2
4
pt
5 Thus, in the basis {|φα i}, the quasienergies are obtained by diagonalizing
6 X [Ĥ (−n) , Ĥ (n) ]
7 ĤF ≡ Ĥ (0) + + O(1/Ω2 ), (24)
8 n6=0
2nΩ
9
cri
10 and the micromotion takes the form
11 X einΩt
ˆ
X
12 Φ̂(t) = |φα (t)i hφα | ≈ I + Ĥ (−n)
nΩ
13 α n6=0
14
" #
X Ĥ (−n) einΩt − H (n) e−inΩt
15 ≈ exp i . (25)
16 2inΩ
us
n6=0
17
18 This is indeed the same expression obtained using other high-frequency expansions, such
19 as van-Vleck perturbation theory [33–35]. We note that in this basis, the boundary
20 ˆ Instead, T log[Φ̂(t)] dt = 0, again in
condition Φ̂(0) = Iˆ + n6=0 Ĥ (−n) /(nΩ) 6= I.
P R
21 0 T
22 agreement with the van-Vleck theory.
23
24
25
26
27
28
ĤF 7→ Φ̂(0)ĤF Φ̂† (0)
an
We can restore the boundary condition to identity by the unitary transformation
Φ̂(0) |φα i = |φα (0)i to the basis of perturbed Floquet modes. In this basis, we obtain
X [Ĥ (−n) , Ĥ (n) ] + [Ĥ (0) , Ĥ (n) ] + [Ĥ (−n) , Ĥ (0) ]

dM
29
≈ Ĥ (0) + , (26)
2nΩ
30 n6=0
31 and
32 X einΩt − 1
|φα (t)i hφα (0)| ≈ Iˆ +
X
33 Φ̂(t) = Ĥ (−n)
34 α n6=0
nΩ
35 " #
36 X einΩt/2 Ĥ (−n) + e−inΩt/2 Ĥ (n) nΩt
37 ≈ exp i sin . (27)
nΩ 2
38 n6=0
pte
39 In the last step, we have written the micromotion in a form that is manifestly unitary.
40
Note that now Φ̂(0) = α |φα (0)i hφα (0)| = Iˆ by orthonormality of the Floquet modes.
P
41
42 This boundary condition and Eqns. (26) and (27) agree with those obtained using the
43 Floquet-Magnus expansion [33–35].
44
45
46 2.4. Low-Frequency Expansion
ce
47
48 A perturbative expansion at low frequencies can be obtained by rescaling time to τ = Ωt
49 and noting that the periodicity of the Hamiltonian Ĥ(τ ) is maintained when translating
50
51 τ → τ + 2π. The Floquet Schrödinger equation in rescaled units read
52 ˆ − ΩẐˆ ) |φ ii = |φ ii ,
(Ĥ (28)
Ac
τ αn αn αn
53
54 where the dimensionless Ẑˆτ = Iˆ ⊗ n |n)n(n|, and Ĥ ˆ is defined below Eq. (14).
P
55
56
One may now attempt a perturbative expansion at low frequencies taking −ΩẐˆτ as
57 the perturbation operator. However, there is a subtlety that must be addressed: the
58 Floquet perturbation theory we developed in the previous section takes Ẑˆτ as part of
59
60
the unperturbed Hamiltonian. This is necessary to ensure that the eigenvalues of the
1
2
4
pt
5 unperturbed operator have the same modular structure as the final quasienergies; that
6 is, if is a quasienergy obtained from the the perturbative solution of Eq. (28), then
7 ˆ
+ nΩ for any n ∈ Z should also be a quasienergy solution of Eq. (28). By, taking Ĥ
8
9 as the unperturbed operator without including Ẑˆτ , the eigenvalues of the unperturbed
cri
10 Hamiltonian will no longer be modular. Indeed, the eigenstates of Ĥ ˆ are nothing but
11 the eigenstates of the instantaneous Hamiltonian Ĥ(τ ) lifted to F :
12
13 ˆ |ψ ii = E |ψ ii ,
Ĥ (29)
ατ ατ ατ
14
15 where |ψατ ii = |ψα (τ )i |τ ), Ĥ(τ ) |ψα (τ )i = Eα (τ ) |ψα (τ )i, and Eατ = Eα (τ ). The
16 eigenvalues Eατ of Ĥˆ are, therefore, not modular, unlike the eigenvalues of Ĥ ˆ − ΩẐˆ .
us
αn
17
18
To avoid confusion, let us note that here τ is simply a label indexing the eigenvalues and
ˆ , even though the operator itself does not depend on a specific choice
eigenstates of Ĥ
19
20 of this label.
21 ˆ − ΩẐˆ
22
Therefore, in order to use perturbation theory to build the spectrum of Ĥ τ
23
24
25
26
27
28
ˆ
an
as a power series over the spectrum of Ĥ , we need to amend our Floquet perturbation
theory to ensure we obtain a modular spectrum. This can be done by using the general
relationship, employed in writing Eq. (16), between the modular Floquet spectrum and
the Fourier integrals of the loop in F obtained by lifting the loop of Floquet modes in
H . Starting with the zeroth order solutions |ψατ ii in Eq. (29), we first use perturbation
dM
29
theory to find |ψατ (i) ii to the desired order i. The modular spectrum is then found by
30
31 taking the Fourier transform of this loop in F ,
32 Z 2π
33 ˆn |ψα(i) ii =
|φαn(i) ii = µ̂ e+inτ |ψατ (i) ii d̄τ, (30)
34 0
35 ˆ − ΩẐˆ with a modular
where d̄τ ≡ dτ /(2π). This defines the proper eigenstate of Ĥ τ
36
37 eigenvalue αn(i) = α(i) +nΩ, and the quasienergy
38 ˆ |φ ii − Ω hhφ | Ẑˆ |φ
α(i) = hhφα(0) | Ĥ α(i−1) ii . (31)
pte
39 α(i) α(0)
40 This equation follows from the Floquet Schrödinger equation and noting that
41 R 2π
42 hhφα(0) |φα ii = 0 hhψατ (0) |ψατ ii d̄τ = 1. Note that for i ≥ 1 the first term vanishes.
43 Explicitly, for the first few terms we find,
44 Z 2π
45 α(0) = Eα (τ ) d̄τ ; (32a)
46 0
ce
47 Z 2π
1 ∂
48 α(1) = Ω hψα (τ )| |ψα (τ )i d̄τ ; (32b)
49 0 i ∂τ
50 Z 2π X 1 ∂
2
hψ β | |ψα i
51 α(2) = Ω2 i ∂τ
d̄τ ; (32c)
52 0 E α (τ ) − Eβ (τ )
β6=α
Ac
53
54
and
55
Z 2π X hψβ | 1 ∂ |ψα i
56 |φαn(1) ii = Ω i ∂τ
e+inτ |ψβτ ii d̄τ. (33)
0 E α (τ ) − E β (τ )
57 β6=α
58 There is one final loose end we now address: there is a gauge freedom in the choice of
59
60
the instantaneous basis |ψα (τ )i that we need to fix. Explicitly, |κα (τ )i = e−iΛα (τ ) |ψα (τ )i
1
2
4
pt
5 with Λα (2π) = Λα (0) is another basis, satisfying
6 ∂ ∂Λα
7 [Ĥ(τ ) − iΩ ] |κα (τ )i = [Eα (τ ) − Ω ] |κα (τ )i
8 ∂τ ∂τ
9 ∂
− iΩe−iΛα (τ ) |ψα (τ )i . (34)
cri
10 ∂τ
11 Thus, fixing the gauge by setting Ω∂Λα /∂τ = Eα (τ ) − α(0) , that is,
12
13 1 τ
Z
14 Λα (τ ) = [Eα (s) − α(0) ]ds, (35)
Ω 0
15
16 up to a constant, we obtain
us
17 ∂
18 [Ĥ(τ ) − iΩ ] |κα (τ )i = α(0) |κα (τ )i + O(Ω). (36)
19 ∂τ
20 This is the zeroth-order Floquet Schrödinger equation.‡ Therefore, we must indeed
21 choose |φα(0) ii = |κα ii. This gauge-fixing was previously used by Martiskainen and
22
23
24
25
26
27
28
an
Moiseyev [45]; however, they only justified its use numerically by showing it improves
the accuracy of the perturbative expansion. Here, we see that this gauge must be
fixed for consistency of the adiabatic solution as the zeroth order term in the general
low-frequency Floquet perturbation theory.
Our low-frequency expansion is obtained for periodically driven systems using
dM
29 Floquet perturbation theory. However, this same approach can be used for general
30
31
non-periodic and slowly driven systems by treating the whole evolution as one long
32 single cycle of a periodic drive. In this way, we can connect our results to other low-
33 frequency approximations, such as the adiabatic perturbation theory [38–40] and the
34
35
adiabatic-impulse theory [41, 42]. Our expressions obtained above for the evolution
36 of the states are closely related to those obtained using the adiabatic perturbation
37 theory [39, 40]. Our method is, however, simpler in its structure and casts the entire
38
procedure in the language of time-independent perturbation theory in the extended
pte
39
40 Floquet Hilbert space. The adiabatic-impulse theory [41,42], on the other hand, requires
41 the identification of special points during the drive where Landau-Zener transitions are
42
likely to occur. These points are then treated separately from the rest of the drive,
43
44 which is taken to be adiabatic. The accuracy of this approach depends strongly on the
45 specific shape of the drive and lacks a natural low-frequency perturbation parameter.
46
In contrast, our approach, similar to the adiabatic perturbation theory, can be used for
ce
47
48 any drive protocol. Finally, the low-frequency Floquet perturbation theory formulated
49 in this work naturally connects to other approximate methods for higher frequencies
50
that also use the structure of the extended Floquet Hilbert space.
51
52 ‡ For completeness, we note that theR gauge fixing can be done entirely in F by defining the gauge
Ac
53 ˆ ) = 2π e−iΛα (τ ) |ψ ii hhψ | d̄τ . Then, |κ ii = exp(−iΛ̂

transformation operator exp(−iΛ̂ ˆ ) |ψ ii, and
0 ατ ατ ατ ατ
54 ˆ ), Ẑˆ ] = 2π dΛα (τ ) |ψ ii hhψ | d̄τ .
R
Eq. (36) follows from the commutation relation [exp(−iΛ̂
55 τ 0 dτ ατ ατ
56
57
58
59
60
1
2
4
pt
5 3. Applications
6
7 3.1. Schwinger-Rabi Model at Low Frequency
8
9 As an example, take the matrix Hamiltonian [53]
cri
10
11 H(t) = B(t) · σ, (37)
12
13 of a spin- 21 particle in a magnetic field B(t) rotating at frequency Ω and a fixed angle θ
14 with the z-direction. (Here, σ is the vector of Pauli matrices and we set the magnetic
15 moment to unity.) Taking n(t) = (sin θ cos Ωt, sin θ sin Ωt, cos θ) to be the unit vector
16
us
17 in the direction of the magnetic field, and rescaling time as before to τ = Ωt, we have
18 H(τ ) = Bn(τ ) · σ.
19 The exact solution to the Schrödinger equation is found by going to the rotating
20
21 frame given by the periodic unitary transformation S(τ ) = eiτ (1+σz )/2 , where the
22 Hamiltonian is
23
24
25
26
27
28
Ω
an

= − + B sin θσx + B cos θ −
2
d
Hrot = S(τ )H(τ )S † (t) − iΩS(τ ) S † (τ )
dτ
Ω
2

σz . (38)
dM
29 Since this is now time-independent, the solutions are pfound as the eigenstates |χ± i of Hrot
30 with eigenvalues ± = Be± , with e± = −(Ω/2B)± (cos θ − Ω/2B)2 + B 2 sin θ2 . In the
31 original frame, we find the Floquet steady states |φ± (τ )i = S † (τ ) |χ± i with eigenvalues
32
33 ± as quasienergies. To compare these exact solutions with the low-frequency Floquet
34 perturbation theory, we impose the normalization hχ± (Ω)|χ± (Ω = 0)i = 1; then, the
35 Fourier transform of the the loop |φ± (t)i lifted to F reads
36 " #
37 (cos θ ± e± )δn,+1 |n)
38 |φ±n ii = , (39)
sin θ δn,0 (1 ± e± )f∓ (θ/2)
pte
39
40
with f+ (x) = sin x and f− (x) = cos x. Expanding in powers of Ω, we find
41
42 ±(0) = ±B; (40a)
43
44 ±(1) = −Ωf∓2 (θ/2); (40b)
45 Ω2
46 ±(2) =± sin2 θ, etc; (40c)
8B
ce
47 " #
48 ±f∓ (θ/2)δn,+1
49 |φ±n(0) ii = |n); (40d)
f± (θ/2)δn,0
50
51
" #
Ω sin θ −f± (θ/2)δn,+1
52 |φ±n(1) ii = |n), etc. (40e)
Ac
53 4B ±f∓ (θ/2)δn,0
54
55 The low-frequency Floquet perturbation series is based on the instantaneous
56 spectrum of H(τ ) of spin Pauli matrices along n(τ ), given by the eigenstates
57 " #
58 ±f∓ (θ/2)e−iτ
|ψ± (τ )i =
59 f± (θ/2)
60
1
2
4
pt
5 with eigenvalues E± (τ ) = ±B. Since the instantaneous eigenvalues are time-
R 2π
6 independent, the gauge Λ± = 0. Thus, at the lowest order, ±(0) = 0 E± (τ ) d̄τ = ±B,
7 R 2π
and |ψ±n(0) ii = 0 e+inτ |ψ± (τ )i |τ ) d̄τ , which reproduce Eqs. (40a) and (40d). Noting
8
9 1 ∂
cri
10 hψ± (τ )| |ψ± (τ )i = −f∓2 (θ/2), (41a)
i ∂τ
11
1 ∂ 1
12 hψ+ (τ )| |ψ− (τ )i = sin θ, (41b)
13 i ∂τ 2
14 and using Eqs. (32b), (33), and (32c), we find precisely Eqs. (40b), (40e), and (40c).
15
Thus, our low-frequency Floquet perturbation theory yields just the same leading order
16
us
17 terms as those obtained from the exact solution.
18 We shall now consider a more general periodic Hamiltonian H(τ ) = d(τ ) · σ, where
19
d(τ ) = d(τ )n(τ ) is a vector, whose direction, n, as well as its magnitude, d, change
20
¯ Then, the Schrödinger
R 2π
21 periodically in time. We denote the average 0 d(τ ) d̄τ ≡ d.
22 equation is not in general exactly solvable. Transforming to the rotating frame, for
23
24
25
26
27
28
an
example, will not produce a time-independent Hamiltonian any more. A high-frequency
expansion can be developed in the rotating frame; however, these expansion fail at low
enough frequency. Instead, we shall use the Floquet perturbation
instantaneous spectrum given by eigenstates |ψ± (τ )i = e∓iΛ(τ )

" theory based on the
±f∓ (θ(τ )/2)e −iϕ(τ )
#
dM
29 f± (θ(τ )/2)
30 and eigenvalues E± (τ ) = ±d(τ ), where θ and ϕ are the polar angles of n and the gauge
31 Rτ
¯
Λ(τ ) = Ω1 0 [d(s) − d]ds. Now,
32
33 1 ∂ 1 ± cos θ dϕ
34 hψ± | |ψ± i = , (42a)
i ∂τ 2 dτ
35
36
1 ∂ sin θ dϕ i dθ
hψ− | |ψ+ i = − e−2iΛ . (42b)
37 i ∂τ 2 dτ 2 dτ
38
So,
pte
39
2π
40
Z
±(0) = ± ¯
d(τ ) d̄τ = ±d; (43a)
41
0
42 2π
1 ± cos θ(τ ) dϕ
Z
43
±(1) = Ω d̄τ ; (43b)
44 0 2 dτ
45
Ω2 2π ( dϕ )2 sin2 θ(τ ) + ( dτ
dθ 2
)
Z
46 ±(2) =± dτ
d̄τ . (43c)
ce
47 8 0 d(τ )
48
49
The first-order correction to the Floquet steady state reads
50 Ω dϕ dθ
sin θ ∓ i dτ
51 |ψ±(1) (τ )i = ± dτ
e∓2iΛ(τ ) |ψ∓ (τ )i . (44)
52 2 d(τ )
Ac
53
54 3.2. Driven Landau-Zener Model
55
56 Consider the Hamiltonian
57
58 H(τ ) = af (τ )σx + bσy , (45)
59
60
1
2
4
pt
5 where f (τ ) is a periodic function satisfying f (0) = −f (π) = 1 and a, b > 0. Half way
6 during the cycle, the first term switches sign, thus realizing the usual situation in the
7
Landau-Zener p model for large frequencies. In our general notation, H(τ ) = d(τ )n(τ ) · σ,
8
9 where d(τ ) = a2 f 2 (τ ) + b2 , θ = π/2, and cos φ(τ ) = bf (τ )/d(τ ). The minimum gap
cri
10 in the instantaneous spectrum is b obtained when f vanishes. We take a b.
11 For b = 0, the exact solution for the evolution operator is U (τ ) =
12 Rτ
13 exp[−i(a/Ω)F (τ )σx ], where F (τ ) = 0 f (s)ds. Thus, the Floquet spectrum is given
14 by the quasienergies ±(0) = ±aF (2π)/(2π) = ±af (0) and Floquet steady states
15 |φ±τ (0) ii = √12 [1 ± 1]T e∓i(a/Ω)Λ(τ ) |τ ), with the micromotion phase Λ(τ ) = F (τ ) − f (0) τ .
16
us
17 For b/Ω 1 or b/a 1, we expect the operator bσy to be “small’ compared to
18 Ĥ0 − ΩẐˆ with H0 (τ ) = af (τ )σx ; thus, we may use the Floquet perturbation theory in
ˆ
19
either the high-frequency limit Ω b or the the high-amplitude limit a b to find
20
21 corrections to the Floquet spectrum:
22
23
24
25
26
27
28
±(1)
±(2)
|φ±(1) ii =
= 0,
=±
2
b X
Ω n an
|gn+ (2a/Ω)|2
(2a/Ω)f (0) + n
b X [gn± (2a/Ω)]∗ gn−m ±
,
(a/Ω) 1
√
"
1
#
|m),
(46a)
(46b)
(46c)
dM
29 Ω n,m (2a/Ω)f (0) ± n i 2 ∓1
30 R 2π ∓
31 where gn± (a/Ω) = 0 e−inτ e±i(a/Ω)Λ(τ ) d̄τ . Note that [gn± (a/Ω)]∗ = g−n (a/Ω). Thus,
32 starting from an initial state |φ±(0) (0)i, the probability of transitioning to state |φ∓(0) (τ )i
33
34 at time τ is,
35 2
P± (τ ) = hφ∓(0) (τ )| U (τ ) |φ±(0) (0)i
36 2
37

X i

38 = hφ∓(0) (τ )|e− Ω α τ φα (τ )i hφα (0)|φ±(0) (0)i

pte
39 α
40 i
≈ hφ∓(0) (τ )|e− Ω ± τ φ±(1) (τ )i hφ±(0) (0)|φ±(0) (0)i

41
42 i
2
43 + hφ∓(0) (τ )|e− Ω ∓ τ φ∓(0) (τ )i hφ∓(1) (0)|φ±(0) (0)i ,

44
45 (47)
46
where the approximation is to the second-order in b. This yields, after some algebra,
ce
47
48
Z τ 2
b ∓i(2a/Ω)F (s)

49 P± (τ ) =
e ds . (48)
50 Ω 0
51 This expression can also be obtained [54,55] directly from the Schrödinger equation
52
written in the adiabatic basis, |ψ(τ )i = α=± Aα (τ )e−i±(0) τ /Ω |φα(0) (τ )i, which reads
P
Ac
53
54 dA± b
55 = ± e±i(2a/Ω)F (τ ) A∓ (τ ). (49)
56
dτ Ω
57 Starting with initial state |φ±(0) (0)i and assuming A± (τ ) = 1 + O(b/Ω) during the
58
Rτ
evolution, we find A∓ (τ ) ≈ ∓(b/Ω) 0 e∓i(2a/Ω)F (s) ds + O(b2 /Ω2 ). Our derivation using
59
60
the more systematic Floquet perturbation theory, apart from being an application of
1
2
b/a = 0.1, f=cos
4 b/Ω = 0.5, b/a = 0.02
(c) ��
pt
5 (a) FPT
��
6 exact
P− (π)
��
7
P−
LZ
��
8 FPT
��
��
9 exact
��
cri
��
10 ��
b/Ω = 1.8, b/a = 0.0399113 � �� b/a ��
= 0.1, ��
f=cos ��
11 (b)�� (d) ��
FPT
12 ��
�� exact
13
P− (2π)
��
P−
14 ��
15 �� FPT
��
16 exact
us
��
17 0 π
2 π 3π
2
2π � � ��
18 τ a/Ω
19
1
20 Figure 1. The transition probability in the driven Landau-Zener model for the
21 drive function f (τ ) = cos τ , calculated via Floquet perturbation theory (FPT) and
22 numerically exactly, for (a) b/Ω = 0.5, a/b = 25 and (b) b/Ω = 1.8, a/b = 45.1. In
23
24
25
26
27
28
(a)��
b/Ω = 0.9, b/a = 0.0375
FPT
(c) ��
��
an
(c) and (d) a/b = 20. The Landau-Zener probability for the linear ramp at half cycle,
Eq. (51), is shown by the horizontal grid line in (a) and (b) and by the dashed curve
in (c).
b/a = 0.1, f=linear
FPT
dM
�� exact
29 �� exact
P− (π)
�� LZ
P−
30 ��
31 ��
32 ��
��
33 b/Ω = 1.8, b/a = 0.0352941 � ��b/a =��
0.1, f=linear
��
34 (b)�� (d) ��
FPT FPT
35 ��
exact
exact
P− (2π)
36 ��
P−
��
37 ��
38 ��
��
pte
39 ��
40 0 π
2 π 3π
2
2π � � ��
41
τ a/Ω
1
42
43 Figure 2. The transition probability in the driven Landau-Zener model for the
44 linear drive function, Eq. (50), calculated via Floquet perturbation theory (FPT) and
45 numerically exactly, for (a) b/Ω = 0.9, a/b = 24 and (b) b/Ω = 1.8, a/b = 51. In (c)
46 and (d) a/b = 20. The Landau-Zener probability at half cycle, Eq. (51), is shown by
ce
47 the horizontal grid line in (a) and (b) and by the dashed curve in (c).
48
49
50 the formalism, shows that this result is valid not only when b/Ω 1, but also when
51
b a, even near for resonant b ≈ Ω/2 a. In the latter case, the usual rotating-wave
52
Ac
53 approximation for large frequencies fails; however, the highly oscillating phase factor
54 suppresses higher order corrections.
55
In Figs. 1 and 2, we compare P− (τ ) obtained by the exact numerical solution with
56
57
58
59
60
1
2
4
pt
5 the Floquet perturbation theory for f (τ ) = cos(τ ) as well as the sawtooth-linear function
6
(
2
τ − π2

7 π
0 ≤ τ ≤ π,
f (τ ) = 2 3π
(50)
8 π 2
− τ π ≤ τ ≤ 2π.
9
cri
10 Panels (a) and (b) in each figure show two choices of parameters at or near resonance
11 b/Ω = 0.5. We note that while agreement is good in the first half of the cycle, it
12 becomes less reliable in the second half. In fact, the end-of-cycle behavior of the Floquet
13
14 perturbation theory is sensitive to the choice of parameters: for small final values of
15 P− (τ ) around τ = 2π the agreement is reasonably good, but for certain larger final
16
us
values, as in Fig. 2(b), the Floquet perturbation result becomes less reliable.
17
18 One way to understand these variations is to compare Floquet perturbation theory
19 and exact results at half- and full-cycle. In panels (c) and (d) of each figure we
20 compare our results for τ = π and τ = 2π, respectively, at a fixed ratio b/a 1
21
22 as a/Ω is varied. For the cosine ramp, the Floquet perturbation theory gives,
23
24
25
26
27
28
J0 and H0 are, respectively,
p Bessel
an
P− (π) = π 2 b2 ([H0 (2a/Ω)]2 + [J0 (2a/Ω)]2 ) , and P− (2π) = 4π 2 (b/Ω)2 [J0 (2a/Ω)]2 , where
p and 2Struve2
functions. For the linear ramp, we
Rx
find P− (π) = [F− ( a/Ω) + F+ ( a/Ω)]π b /(aΩ) where F± (x) = 0 f± (πy 2 /2)dy are
2 2
Fresnel integrals. We note that for large a/Ω 1, the prefactor approaches 1/2 and
dM
this reproduces the classic Landau-Zener-Stückelberg-Majorana formula [56–59],
29
30 PLZ = 1 − exp[−π 2 b2 /(2aΩ)] ≈ π 2 b2 /(2aΩ). (51)
31
32 However, the perturbative result starts to fail for larger a/Ω and fixed b/Ω since the
33 magnitude of the exponent in PLZ becomes large. We show PLZ in panel (c) of each
34
35
figure. We note that while the classic result is in very good agreement with the exact
36 result, it misses the oscillations as a function of a/Ω. By contrast, perturbative result
37 captures these oscillations very well.
38
We also see the sporadic nature of the agreement between Floquet perturbation
pte
39
40 theory and the exact result in the second half-cycle. As seen in panels (d) of each
41 figure, the final value P− (2π) shows large oscillations as a function of a/Ω and vanishes
42
periodically. The Floquet perturbation theory result captures these oscillations and, in
43
44 particular, the zeros of P− (2π) remarkably well. This feature is similar to the coherent
45 destruction of transitions discussed in a periodically driven double-well potential [60–63].
46
For larger values of a/Ω going beyond its applicability, the Floquet perturbation theory
ce
47
48 overshoots the amplitude of the oscillations, eventually giving unphysical values larger
49 than unity. This is due to the non-unitary nature of perturbative expansion of the
50
micromotion operator. We expect that a more controlled expansion that respects the
51
52 unitarity of the micromotion operator, similar to the Floquet-Magnus expansion, should
Ac
53 resolve this problem.

54
55
56 4. Degenerate Low-Frequency Floquet Perturbation Theory
57
58 In the preceding discussion we have assumed the quasienergy spectrum is non-
59
60
degenerate. We shall now address the case where this is not true; as we will see this has
1
2
4
pt
5 interesting consequences for the dynamics. Degenerate Floquet perturbation theory has
6 been employeed before to investigate multiphoton excitations and heating processes in
7
driven optical lattices. [64, 65]
8
9
cri
10 4.1. Formalism
11
12 First, let us say a few words about the different conventions for the range of
13 quasienergies, which we formally restricted to the first Floquet zone [−Ω/2, Ω/2]. This
14
15 choice is entirely arbitrary, of course, and it may be more suitable in some problems to
16 make other choices. Shifting a quasienergy αn 7→ αn +mΩ maps |φαn ii 7→ µ̂ ˆm |φαn ii =
us
17 |φαn+m ii. Doing so for all or just a subset of quasienergies has no physical effect. For
18
19 example, in Eq. (17) all such shifts can be absorbed into a shift of the summation
20 variable n. Now, in Floquet perturbation theory, even if the unperturbed quasienergies
21 α(0) are in the first Floquet zone, the corrections, Eq. (19a), may not be. Thus, in this
22
23
24
25
26
27
28
section, we shall relax this condition.
an
Degeneracies arise when two or more quasienergies rα , labeled by r, coincide when
shifted by integer multiples of frequency, nr . For accidental or symmetry related
degeneracies, nr = 0, and one might as well restrict the unperturbed quasienergies to
the first Floquet zone from the outset. However, for low-frequency Floquet perturbation
dM
29 theory, where even the unperturbed quasienergies need to be calculated according to
30
31 Eq. (32a), it is more convenient to allow unperturbed quasienergies take values outside
32 the first Floquet zone. In this case, we shall assume the shifts nr are unique. (Of course,
33 one may always set one of them to zero.) For initially non-degenerate instantaneous
34
35 energy eigenvalues Eα 6= Eβ , this happens only when the associated quasienergies
36 become resonant at the Floquet zone center or edges; thus, α = β + nΩ = mΩ/2
37 for some integer m.
38
This is an unusual situation in textbook perturbation theory, since the zeroth-order
pte
39
40 quantities are resonant up to first order in the small quantity Ω. Nevertheless, we may
41 still proceed by trying to find the proper superpositions of degenerate states that resolve
42
43 the degeneracy. Define
44
X
|χrα ii = crs ˆ s
α µ̂nr |φα(0) ii , (52)
45 s
46
with coefficients crs
to be determined by solving
ce
47 α
48 ˆ − ΩẐˆ ) |χr ii = (mΩ/2 + r ) |χr ii + O(Ω2 ).
(Ĥ (53)
α α(1) α
49
50 This yields,
51 X
52
Wαrq cqs s rs
α = α(1) cα , (54)
Ac
53 q
54 where Wα is a matrix with diagonal elements Wαrr = rα +nr Ω − mΩ/2, and the off-
55 diagonal elements,
56
57 ˆ†nr Ẑˆ µ̂
Wαrs ≡ −Ω hhφrα | µ̂ ˆns |φsα ii , r 6= s
58 Z 2π
1 ∂ s
59 =Ω ei(ns −nr )τ hφrα (τ )| |φ (τ )i d̄τ, (55)
60 0 i ∂τ α
1
2
4
pt
5 where we have used hφrα (τ )|φsα (τ )i = 0 for r 6= s. The eigenvalue set of equations in (54)
6 constitute the first-order degenerate low-frequency Floquet perturbation theory.
7
8
9 4.2. Floquet Resonances in Landau-Zener Model
cri
10
11 Let us revisit the driven Landau-Zener model, Eq. (45), in the low-frequency Floquet
12 perturbation theory. Here, we shall take f (τ ) = cos τ . The instantaneous eigenvalues
13 √
are E± (τ ) = ± a2 cos2 τ + b2 ; so, the lowest-order quasienergies are
14
15 2b
±(0) = ± E −a2 /b2 ,

16 (56a)
π
us
17 b
E(−a2 /b2 ) − K(−a2 /b2 ) (Ω/b)2 ,

18 ±(2) = ± (56b)
19 4π
20 where E(x) and K(x) are complete elliptic integrals. (The first-order correction to
21 quasienergy vanishes.) For Ω/b 1, one can always find quasienergy degeneracies by
22
23
24
25
26
27
28
frequencies. an
tuning a/b and sufficiently large shifts n. However, we note quasienergy degeneracies
also occur for Ω/b & 1, so the associated dynamics is not restricted to (very) low
To the lowest order in Ω, the adiabatic solutions are

Z ∓iΛ(τ ) " −iϕ(τ ) #
dM
29
e e
|φ±(0) ii = √ |τ ) d̄τ, (57)
30 2 ±1
31 Rτ
32 where cot ϕ(τ ) = (a/b) cos τ and the gauge Λ(τ ) = Ω1 0 [E+ (s) − +(0) ]ds can also be
33 expressed in terms of an incomplete elliptical integral. Near a quasienergy degeneracy,
34 ˆn |φ−(0) ii},
+ = nΩ/2 + ∆ and − = −nΩ/2 − ∆; thus, in the adiabatic basis {|φ+(0) ii , µ̂
35
36 m = n, and,
37 " #
38 ∆ Ωzn∗
W = , (58)
pte
39 Ωzn −∆
40 R 2π
41 where zn = − 12 0 (dϕ/dτ )einτ e2iΛ(τ ) d̄τ is nonzero only for odd n. We derive a closed-
42
form expression for zn in Appendix B. Writing zn = −iζn |zn | with ζn = −sgn(Im zn ),
43
44 we find the solutions
45 θn θn
46 |χ+ ii = cos |φ+(0) ii + iζn sin µ̂ ˆn |φ−(0) ii , (59a)
2 2
ce
47
θn θn
48 |χ− ii = sin |φ+(0) ii − iζn cos µ̂ ˆn |φ−(0) ii , (59b)
49 2 2
50 with 0 < θn < π and tan θn = Ω|zn |/∆. The quasienergy degeneracy is lifted to
51 p
52 ± = +nΩ/2 ± ∆2 + |zn |2 Ω2 . (60)
Ac
53
54 ˆn |φ+(0) ii ±
Exactly at the degeneracy, ± = +nΩ/2 ± Ω|zn |, θn = π/2, and |χ± ii = √12 (µ̂
55 iζn |φ−(0) ii).
56
57 In Fig. 3, we show the transition probability and the expectation value of energy at
58 two such quasienergy degeneracies, one at a lower frequency Ω/b = 0.75 and the other
59 at a much higher frequency Ω/b = 2.45 with, respectively, n = 3 and n = 1. In both
60
1
2
4
pt
��
5 (a) FPT �� (c) FPT
��
6 adiab �� adiab
7 �� exact
hHi
exact
P−
��
8 ��
-��
9 ��
cri
-��
10 ��
11 � � � � ��
12 ��
(b) FPT �� (d) FPT
13 �� adiab �� adiab
14 ��
hHi
exact exact
P−
��
15 ��
-��
16 ��
us
17 -��
��
18 � � � � � � � � � � � �
19 cycle cycle
20
1
21 Figure 3. The transition probability and energy vs. cycle time in the driven Landau-
22 Zener model for the drive protocol f (τ ) = cos τ , starting from the ground state of
23
24
25
26
27
28
an
initial Hamiltonian, calculated with degenerate low-frequency Floquet perturbation
theory (FPT), the adiabatic approximation, and numerically exactly. The parameters
a/b = 0.691508, Ω/b = 0.75 (a, c) and a/b = 0.631747, Ω/b = 2.45 (b, d), correspond
to quasienergy degeneracies with, respectively, n = 3 and n = 1.
dM
29 cases, we see that the low-frequency Floquet perturbation theory works well. This is
30
31 remarkable for these frequencies are not particularly low compared to the minimum gap
32 of 2b. In fact, in the second case, the frequency Ω/b = 2.45 is larger than the gap, which is
33 why it was chosen to obtain the lowest value of n = 1. In this case, numerically calculated
34
35 quasienergies are far from Floquet zone edges and simply eyeballing their evolution with
36 a/b does not hint at an avoid quasienergy degeneracy in the adiabatic approximation.
37 Nevertheless, the lowest-order degenerate low-frequency Floquet perturbation theory
38
produces a reasonably accurate result.
pte
39
40 A few remarks are in order. The low frequency oscillations observed in Fig. 3 are
41 manifestations of Rabi-like oscillations between resonant states |φ+(0) ii and µ̂ ˆn |φ−(0) ii;
42
43 at degeneracy, the frequency of these oscillations is |zn |Ω. These oscillations indicate
44 a breakdown of adiabaticity [46, 47]: the system starting in the adiabatic state |φ−(0) ii
45 would transition out fully over 1/(4|zn |) cycles.
46
However, one must be quite careful about statements of adiabatic breakdown at low
ce
47
48 frequencies. For smooth drive protocols, the splitting |zn | is typically exponentially small
49 at low frequencies, meaning that a non-adiabatic transition would take exponentially
50
51 long times. For example, in Appendix B we show that for Ω/b 1, a quasienergy
52 degeneracy of the lowest shift order is obtained for a ∼ Ω with n ∼ bΩ and the
Ac
53 quasienergy splitting vanishes as (Ω/b)2b/Ω . Thus, the transition time out of the
54
55 adiabatic evolution diverges as a factorially large number ∼ e−2(b/Ω) log(b/Ω) . This trend
56 is seen in Fig. 3. The period of Rabi oscillations is multiplied by a factor of about six
57 when going from a principal resonance at n = 1 and Ω/b = 2.45 to the next resonance
58
59 at n = 3 and Ω = 0.7. Using low-frequency Floquet perturbation theory, we found that
60
1
2
4
pt
5 for the first resonance in the driven Landau-Zener model with n = 7 and Ω/b = 0.29,
6 this period rises up to more than 2 × 105 cycles.
7
We have shown here that the celebrated adiabatic approximation is the lowest order
8
9 of the more general low-frequency Floquet perturbation theory, which is built on the
cri
10 Floquet Green’s function. The latter correctly accounts for corrections to the adiabatic
11 approximation and, in particular, its potential breakdown due to Rabi oscillations at
12
13 quasienergy degeneracies. We must note, too, that not all quasienergy resonances at
14 low frequency exhibit Rabi oscillations. The low-frequency Floquet perturbation theory
15 shows when adiabatic evolution is preserved due to protected quasienergy crossings. For
16
us
17 example, in our driven Landau-Zener model, zn = 0 for even n. At such crossings, Rabi
18 oscillations become infinitely long, regardless of frequency, restoring adiabatic evolution
19 in the degenerate subspace.
20
21
22 4.3. Driven Su-Schrieffer-Heeger Model at Low Frequency
23
24
25
26
27
28
an
As a second application, we apply the (degenerate) low-frequency Floquet perturbation
theory to a driven lattice model of non-interacting fermions, namely the Su-Schrieffer-
Heeger (SSH) model. Both the static [66, 67] and driven [18, 48] versions of this model
show distinct topological phases that are distinguished by the appearance of protected
dM
29 bound states at the edges of the lattice with open boundary conditions or a topological
30
31 winding number in the Brillouin zone for a system with periodic boundary conditions.
32 The Hamiltonian for the SSH model is written as
33 N −1
34 wr ξˆr+1 ξˆr + Ĥb + h.c.,
X †
Ĥ = − (61)
35
r=1
36
37 where ξˆr† is the creation operator of a (spinless) fermion at site r, wr = w + (−1)r δ is the
38
modulated hopping amplitude, and Ĥb is a boundary Hamiltonian that depends on the
pte
39
40 choice of boundary conditions. In the following we take w > 0 without loss of generality.
41 For periodic boundary conditions, Ĥb = wN ξˆN † ˆ
ξ1 , and even N we label the two-point
42
unit cells with s = b 2 c, arrange the lattice operators into the spinor Ξ̂†s = (ξˆr† , ξˆr+1
r+1 †
),
43 P iks
44 and write the mode expansion Ξ̂s = s e Ξ̂k with lattice momentum k ∈ [−π, π] to
45 find Ĥ = k Ξ̂†k hk Ξ̂k , with matrix Hamiltonian hk = dk · σ,
P
46
k k
ce
47 dkx + idky ≡ dk = 2e ik/2

w cos + iδ sin , (62)
48 2 2
49 p
50 and energies Ek± = ±2 w2 cos2 (k/2) + δ 2 sin2 (k/2). We also note that the Hamiltonian
51
k k
can be mapped unitarily to ei 4 σz hk e−i 4 σz = (w cos k2 )σx + (δ sin k2 )σy .
52
Now, we consider the driven SSH model, where δ(τ ) is a periodic function of
Ac
53
54 time with frequency Ω. We shall assume below that |δ| ≤ w and denote the average
R 2π
55 δ̄ = 0 δ(τ ) d̄τ . In the low-frequency limit, to the lowest order in Ω, the adiabatic
56
57
solutions are
" #
58 e∓iΛk (τ )−ikσz /4 eiϕk (τ )
59 |φk±(0) (τ )i = √ , (63)
2 ±1
60
1
2
4 ×10−2
pt
0.5
5 (a)
6 (c)
6
1 − Fk (τ )
7
k /Ω
0.0 4
8
9 2 FPT
cri
adiab
10
11 −0.5 0
0.5
12 (b)
6 (d)
13
1 − Fk (τ )
14
k /Ω
0.0 4
15
16 2 FPT
us
adiab ×0.1
17
18 −0.5 0 π 3π
−1 −0.5 0 0.5 1 0 π 2π
19 2 2
k/π τ
20
21
22 Figure 4. Spectral properties of the driven Su-Schrieffer-Heeger model as a function of
23
24
25
26
27
28
an
momentum k for drive protocol δ(τ ) = δ̄+δ0 cos τ . The quasienergy spectrum shown in
(a) and (b) is found using exact numerical calculation (black) and the (degenerate) low-
frequency Floquet perturbation theory (orange) for two different frequencies in (a) and
(b). The insets show a closeup around quasienergy degeneracies. At these quasienergy
degeneracies we show in (c) and (d) the infidelity in a single cycle, 1 − F, where F is
dM
the overlap between the numerically exact solution and the adiabatic solution (purple)
29 or the lowest-order solution of the degenerate Floquet perturbation theory (orange).
30 The parameters are δ̄/w = 0.2, δ0 /w = 0.1, and Ω/w = 0.8 (a, c) Ω/w = 2.1 (b, d),
31 and k/π = 0.767 (c), k/π = 0.675 (d).
32
33
34 Rτ
with Λk (τ ) = Ω1 0 Ek+ (s) − k+(0) ds and tan ϕk (τ ) = −[δ(τ )/w] tan k2 . The

35
36 quasienergies in the leading-order non-degenerate low-frequency Floquet perturbation
37 theory are k± (Ω) = k±(0) + k±(1) + k±(2) + · · ·, where
38 Z 2π
pte
39
40
k±(0) = Ek± (τ ) d̄τ, (64a)
0
41
42
k±(1) = 0, (64b)
2 2π 2
43 Ω |∂ϕk /∂τ |
Z
44 k±(2) = d̄τ. (64c)
8 0 Ek± (τ )
45
46 For a given Ω < 4w there are a set of points ±k ℘ (Ω) in the Brillouin zone where
ce
47 quasienergies become degenerate: k℘ + = k℘ − +nk℘ Ω, with d4|δ̄|/Ωe ≤ nk℘ ≤ b4w/Ωc.
48
49
Near these points we must employ degenerate low-frequency Floquet perturbation
℘ ℘± ± ± ˆ
50 theory. Expand #k± = ±(n # + ∆k ) and |χ
" k Ω/2 " k ii = ck+ |φ#k+(0) ii + ck− µ̂nk℘ |φk−(0) ii, to
℘
51
"
c± c± ∆℘k Ωzk℘∗
52 find Wk℘ k+
= ℘± k+
with Wk℘ = . Here, zk℘ ≡ zk,nk℘ , where
c± c± Ωzk℘ −∆℘k
Ac
k(1)
53 k− k−
54 2π
1 ∂ϕ
Z
55 zk,n =− einτ e2iΛk (τ ) d̄τ, (65)
56 2 0 ∂τ
57 q
58 determines the gap opening at the degeneracy point as ℘±k(1) = ± ∆℘k 2 + Ω2 |zk℘ |2 and
59 ℘±
60
the solutions |χk ii in a fashion similar to Eqs. (4.2).
1
2
4
pt
5 (a) (b) n=3
0.06 n=1 n=1
6 10−3
10 −2
7 10−7
n=5
|zk,n|
0.04 10−3
8 n=2 n=2
10−11 10−4
9 1 2 3 4 5 6 7 1 10
cri
0.02 n n
10 n=3 n=4
11
0.00
12 0 0.25 0.5 0.75 10 0.25 0.5 0.75 1
13 k/π k/π
14
15 Figure 5. The off-diagonal element |zk,n | for a Floquet resonance of order n in
16
us
the driven Su-Schrieffer-Heeger model with the drive protocols δ(τ ) = δ̄ + δ0 cos τ (a),
17
and the smoothed two-step protocol δ(τ ) = δ̄ + δ0 arctan (B sin τ ) (b). The values of
18
parameters are: δ̄ = 0.2, δ0 = 0.1, Ω/w = 0.8 and B = 20. The insets show |zk,n | as a
19
20 function of n for the momenta indicated by the symbols.
21
22
23
24
25
26
27
28
In Fig. 4 we compare the spectral measures obtained from the adiabatic
an
approximation and the low-frequency Floquet perturbation theory with the numerically
exact solution for a smooth drive protocol δ(τ ) = δ̄ + δ0 cos τ . The quasienergies found
using degenerate low-frequency Floquet perturbation theory match the exact solution
remarkably well. The infidelity of an approximate solution, |φ(τ )i, is defined as 1 − F,
dM
29 where F(τ ) = | hφe (τ )|φ(τ )i |2 and |φe (τ )i is the exact solution. Near a quasienergy
30
31 degeneracy point, the infidelity of the adiabatic approximation increases dramatically
32 as the frequency increases. The infidelity of the solution obtained using the degenerate
33 low-frequency Floquet perturbation theory, on the other hand, is not only small relative
34
35 to the adiabatic approximation, but it also remains small on the absolute scale even
36 for larger frequencies. This demonstrates the consistency and accuracy of the Floquet
37 perturbation theory.
38
As the frequency is lowered, the order nk℘ increases as ∼ 1/Ω. For a smooth
pte
39
40 drive protocol, the adiabatic limit is obtained as the off-diagonal element zk℘ vanishes.
41 For example, for the sinusoidal drive protocol zk℘ vanishes exponentially similar to the
42
43 Floquet resonances in the Landau-Zener model. However, for a drive protocol with
44 sharp features, such as a step-wise protocol, the approach to the adiabatic limit may be
45 slower or even violated. In Fig. 5 we show the dependence of zk,n on k and the order
46
n of the quasienergy degeneracy. Indeed, as shown in Fig. 5(a) and the inset, for the
ce
47
48 sinusoidal drive protocol, zk,n vanishes with n in an exponential manner. However, for
49 step-wise protocol, shown in Fig. 5(b), zk,n approaches a limiting value as n increases
50
51 that depends on the parity of n.
52
Ac
53
5. Summary and Outlook
54
55
56 Floquet perturbation theory recasts time-dependent perturbation theory of a
57 periodically driven quantum system in terms of a time-independent perturbation theory
58
in the extended Floquet Hilbert space of the periodic operators. This formalism is
59
60
1
2
4
pt
5 transparent and strips away cumbersome book-keeping that is required to track the
6 time-evolution in the original Hilbert space. Using this formalism, we developed a
7
low-frequency perturbation theory, which connects naturally with the high-frequency
8
9 expansions of the Floquet dynamic.
cri
10 While we reproduced some results already reported in the literature, for example
11 in two-level systems, our approach allowed us to clarify and extend the range of
12
13 applicability of these results. Additionally, using the formalism in this paper one can
14 readily obtain the full-cycle dynamics not usually accessible in traditional approaches.
15 For example, the usual treatment of the Landau-Zener model assumes an infinite
16
us
17 duration for the transition. However, the actual transition takes a finite time [55, 68].
18 In experimentally relevant situations the drive may take a comparable time as the time
19 needed for the transition. In these cases, the Floquet perturbation theory is useful and
20
21 provides a more detailed description of the dynamics within the drive cycle.
22 In the low-frequency limit, we obtained a systematic and compact derivation
23
24
25
26
27
28
an
of the adiabatic perturbation theory. Moreover, in this formalism, the occurrence
of quasienergy degeneracies that cause diabatic deviations via Rabi oscillations is
easily and accurately captured using a degenerate low-frequency Floquet perturbation
theory. We saw that for typical, smooth drive protocols, such as sinusoidal, the
dM
29 approach to the adiabatic limit is exponential since the matrix element leading to Rabi
30 oscillations vanishes (super-)exponentially in 1/Ω. However, for drive protocols with
31 sharp features, such as step-wise, this matrix element may approach an asymptotic
32
33 value, thus invalidating the adiabatic approximation. In such scenarios, the Floquet
34 perturbation theory is essential for describing the correct dynamics of the system at low
35 frequencies.
36
37 We briefly mention some interesting problems for future application of our work.
38 On the technical side, it would be useful to improve the Floquet perturbation theory
pte
39 for the expansion of the Floquet modes by ensuring the unitarity of the micromotion
40
41 operator, perhaps along the lines of the Floquet-Magnus expansion [33,34]. This would,
42 for example, improve the accuracy of the calculated transition probabilities in the driven
43 Landau-Zener model and avoid divergence for slow drive. Other interesting problems
44
45 for which the Floquet perturbation theory might be useful are the fate and structure
46 of Floquet topological [18] and Floquet many-body localized [21, 23] phases at low
ce
47 frequencies.
48
49
50 Acknowledgments
51
52 This work was supported in part by the National Science Foundation CAREER award
Ac
53
54 DMR-1350663, the US-Israel Binational Science Foundation under grant No. 2014245,
55 and the College of Arts and Sciences at Indiana University (B.S. and M.R.V.), as well
56 as the Indiana University REU program through NSF grant PHY-1460882 (M.L.). B.S.
57
58 thanks the hospitality of Aspen Center for Physics, supported by NSF grant PHY-
59 1607611, where parts of this work were performed.
60
1
2
4
pt
5 Appendix A. Floquet Theory
6
7 In this Appendix, we provide more detail for the Floquet theory in the extended Floquet
8 Hilbert space.
9
cri
10
11 Appendix A.1. Periodic Hamiltonians
12
13 For a periodich Hamiltoniani Ĥ(t) = Ĥ(t + T ), the full-period evolution operator,
14 RT
15
Û (T ) = Texp −i 0 Ĥ(s)ds , is unitary and can be written as exp(−iT ĤF ) for some
16 Hermitian operator ĤF . The eigenvalues of Û (T ) are phases e−iα T with eigenstates
us
17
18
|φα i,
19 Û (T )|φα i = e−iα T |φα i. (A.1)
20
21 Starting with a state |ψα (0)i = |φα i, we have |ψα (T )i = Û (T ) |φα i = e−iα T |φα i. Thus,
22
23
24
25
26
27
28
an
defining |φα (t)i ≡ eiα t |ψα (t)i, we find |φα (T )i = |φα i ≡ |φα (0)i, i.e. |φα (t)i are periodic
and |ψα (t)i = e−iα t |φα (t)i. This is the Floquet theorem.
Since Û (t) is unitary, it follows immediately that α ∈ R and {|φα (0)i} is an
orthonormal basis for H . Therefore, the |φα (t)i = eiα t Û (t)|φα (0)i also form an
orthonormal basis for H . Since the quasienergies are modular, defined only through
dM
29 the eigenvalues e−iα T of Û (T ), we may restrict them to be in the first Floquet zone,
30 α ∈ [−Ω/2, Ω/2]. Using this structure, the evolution operator is completely defined by
31
32 its action Û (t)|φα (0)i = e−iα t |φα (t)i; thus,
33
X
Û (t) = e−iα t |φα (t)ihφα (0)| =: Φ̂(t)e−itĤF , (A.2)
34
α
35
36 where the sum over α is understood as an integral whenever α is continuous, and
37 X
38 Φ̂(t) = |φα (t)ihφα (0)|, (A.3)
pte
39 α
40
X
−itĤF
e = e−itα |φα (0)ihφα (0)|. (A.4)
41 α
42
43 Here Φ̂(t) = Φ̂(t + T ) is a unitary periodic operator, with the boundary condition
44 Φ̂(0) = I,ˆ called the micromotion operator. It produces the periodic evolution of the
45
46 Floquet modes, |φα (t)i = Φ̂(t) |φα (0)i. The time-independent, Hermitian operator ĤF
ce
47 is called the Floquet Hamiltonian.

48
49
50 Appendix A.2. Micromotion
51
52 In the above decomposition, we chose to set the initial time t0 = 0. This choice is
Ac
53 arbitrary; we could choose any other time within a cycle 0 ≤ t0 < T . In general, for
54 t > t0 we have:
55 Z t
56
Û (t, t0 ) = Texp −i Ĥ(s)ds (A.5)
57 t0
58 X
59 = e−iα (t−t0 ) |φα (t)i hφα (t0 )| (A.6)
60 α
1
2
4
pt
5 = Φ̂(t, t0 )e−i(t−t0 )ĤF (t0 ) (A.7)
6 =e −i(t−t0 )ĤF (t)
Φ̂(t, t0 ), (A.8)
7
8 where
9
cri
X
10 Φ̂(t, t0 ) = |φα (t)i hφα (t0 )| = Φ̂(t)Φ̂(t0 )† , (A.9)
11 α
12 −isĤF (t0 )
X
13 e = e−isα |φα (t0 )i hφα (t0 )|
14 α
15 = Φ̂(t0 )e−isĤF Φ̂(t0 )† . (A.10)
16
us
17 The two-time micromotion operator produces the periodic evolution of Floquet modes,
18 Φ̂(t, t0 ) |φα (t0 )i = |φα (t)i.
19 d
20 We note that, since [Ĥ(t) − i dt ] |φα (t)i = α |φα (t)i, the time-dependent Floquet
21 Hamiltonian can be resolved as
22
23
24
25
26
27
28
d
Ĥ(t) − i =
dt
X
α
an
α |φα (t)i hφα (t)| ≡ ĤF (t). (A.11)
Moreover, ĤF and ĤF (t0 ), are all unitarily equivalent. Especially, the eigenvalues of
ĤF (t0 ) do not depend on the initial time. However, when certain approximate methods
dM
are used to find ĤF (t0 ) it turns out the eigenvalues acquire a spurious dependence on
29
30 t0 . One way to avoid this problem is by using the decomposition Φ̂(t, t0 ) = Φ̂(t)Φ̂(t0 )†
31 to write
32
33 Û (t, t0 ) = Φ̂(t)e−i(t−t0 )ĤF Φ̂(t0 )† . (A.12)
34
35 The dependence on t0 is then entirely accounted for in the micromotion operator.
36
37
Appendix A.3. Floquet Hilbert Space
38
pte
39 The structure we have described in the previous section can be formalized in terms of an
40
41 extended Floquet Hilbert space F = H ⊗ I , where the auxiliary space I is the space
42 of bounded periodic function over [0, T ). It is spanned by a continuous orthonormal
43 basis {|t)}, 0 ≤ t < T ,
44
45 (t0 |t) = T δ(t − t0 ), (A.13)
46 Z T
dt
ce
47 |t)(t| = I.“ (A.14)

48 0 T
49
50 Here, I“ is the identity operator in I . Equivalently, it is also spanned by the orthonormal
51 Fourier basis
52 Z T
dt
Ac
53 |n) = e−inΩt |t) , n ∈ Z, (A.15)

54 0 T
55 which satisfy
56
57 (n|m) = δnm , (A.16)
58 X
“
|n)(n| = I. (A.17)
59
60 n∈Z
1
2
4
pt
Using the Poisson summation formula, n∈Z einΩt = T p∈Z δ(t − pT ), these relations
P P
5
6 can be inverted to give
7 X
8 |t) = einΩt |n). (A.18)
9 n∈Z
cri
10 Note that if we extended the range of t to R periodically by defining |t + T ) = |t), then
11
0
X X
12 “ − t0 ) := T
(t0 |t) = T δ(t δ(t − t0 − pT ) = e−inΩ(t−t ) . (A.19)
13 p∈Z p∈Z
14
15 Now, a loop in H given by the one-parameter family of states |φ(t)i, with
16 |φ(T )i = |φ(0)i, can be lifted to a loop in F given by |φt ii := |φ(t)i |t). Associated with
us
17
any loop in F is the center
18 Z T
19 dt
20 |φii := |φt ii . (A.20)
0 T
21
22 From the center of a lifted loop in F one can in turn obtain the corresponding loop in
23
24
25
26
27
28
H by the projection
|φ(t)i = (t |φii . an
We also note that |φii = n∈Z |φ(n) i |n), where the Fourier components
Z T
P
dt
(A.21)
dM
(n)
29 |φ i = e+inΩt |φ(t)i . (A.22)
30 0 T
31 Similarly, a two-parameter periodic family of operators Â(t, t0 ) : H → H with
32
33 Â(t + T, t0 ) = Â(t, t0 + T ) = Â(t, t0 ) is lifted to an operator Âˆ : F → F defined as
34
Z T
ˆ dt dt0 X 0
35 Â = |t)Â(t, t0 )(t0 | = |n)Â(n,n ) (n0 |, (A.23)
36 0 T T 0
n,n ∈Z
37
38
with
T
dt dt0
Z
pte
39 (n,n0 ) 0 0
40
Â = einΩt e−in Ωt Â(t, t0 ) . (A.24)
0 T T
41
42 Two special cases arise when Âˆ is diagonal in either I or H . As an example of
43 “ − t0 ) and Â(n,n0 ) = Â(n−n0 ) = T ei(n−n0 )Ωt Â(t)dt/T ;
the former case, Â(t, t0 ) = Â(t) T δ(t
R
44 0
45 so,
46
Z T
ˆ dt X
|n)Â(n−m) (m|.
ce
47 Â = |t)Â(t)(t| = (A.25)
48 0 T n,m∈Z
49
50 As a simple case of the latter case, we take Âˆ to be the identity in H , so that Â(t, t0 ) is
51 just a complex-valued function A(t, t0 ) of its arguments. As an important example, we
52 “ − t0 ) =: µ
choose A(t, t0 ) = e+inΩt δ(t “n (t, t0 ) to find
Ac
53
54
Z T
dt
55 ˆ
µ̂n = |t)e+inΩt (t| . (A.26)
56 0 T
57 Then,
58 Z T
dt
59 ˆ
|φn ii := µ̂n |φii = e+inΩt |φt ii . (A.27)
60 0 T
1
2
4
pt
5 ˆn acting on the center of a loop in F yields the Fourier component n of the
Thus, µ̂
6 loop.
7 ∂ “
Another important example is obtained by the choice A(t, t0 ) = i ∂t δ(t − t0 ), for
8
9 which we have the operator
cri
10 Z T
∂“ dt dt0 X
11 ˆ
Ẑt = |t)i δ(t − t0 )(t0 | = nΩ|n)(n|, (A.28)
12 0 ∂t T T n∈Z
13
Then,
14
T 0
15 ∂ 0 dt
Z
16 Ẑˆt |φt ii = [i δ(t 0
− t)] |φ(t)i |t )
∂t0 T
us
0
17
18 d
= [i|φ(t)i]|t) ≡ i |(dφ/dt)t ii , (A.29)
19 dt
20 where |(dφ/dt)t ii is defined as the loop dtd
|φ(t)i lifted to F . In particular,
21
22 ∂
hhφ0t0 | Ẑˆt |φt ii = hφ0 (t)| i |φ(t)i T δ(t − t0 ).
23
24
25
26
27
28
We also note that,
Ẑˆ |φii = i |dφ/dtii .
t
∂t
an (A.30)
(A.31)
A loop of periodic family of bases |φα (t)i for H can be lifted to a basis |φαt ii for
dM
29 F , since it may be easily seen
30
31 hhφαt |φβt0 ii = δαβ T δ(t − t0 ), (A.32)
32 Z TX
dt
33 |φαt ii hhφαt | ˆ
= I. (A.33)
34 0 α
T
35
A different basis is obtained by the Fourier transform of the loop in F , i.e. |φαn ii =
36
37 ˆn |φα ii. To see that this is a complete basis for F , first note that
µ̂
38 Z T
0 dt dt0
e−inΩt e+imΩt hhφαt |φβt0 ii
pte
39 hhφαn |φβm ii =
40 0 T T
41
Z T
dt
42 = δαβ ei(m−n)Ωt = δαβ δnm ; (A.34)
43 0 T
44 and, second,
45 Z TX
X
−inΩ(t−t0 ) dt dt0
46 |φαn ii hhφαn | = e |φαt ii hhφαt |
0
T T
ce
47 αn 0 αn
48 Z TX
49
dt ˆ
= |φαt ii hhφαt | = I. (A.35)
50 0 α
T
51
52 For a time-independent basis |φα i in H , we obtain |φαn ii = |φα i |n), which is obviously
Ac
53 a basis for F .
54 Now, we can write the Floquet Schrödinger equation in F as
55
56 ˆ − Ẑˆ ) |φ ii = |φ ii .
(Ĥ (A.36)
t α α α
57
58 Note that
59
ˆn , Ẑˆt ] = nΩµ̂
[µ̂ ˆn . (A.37)
60
1
2
4
ˆ , µ̂ ˆ − Ẑˆ , mapping the solution
pt
5 Since [Ĥ ˆn ] = 0, we see that µ̂
ˆn is a ladder operator for Ĥ t
6 |φα ii with quasienergy α to |φαn ii with quasienergy α +nΩ. Thus, the solutions to the
7
Floquet Schrödinger equation
8
9 ˆ − Ẑˆ ) |φ ii = |φ ii ,
(Ĥ (A.38)
cri
t αn αn αn
10
11 where αn := α +nΩ provides a full basis for F .
12
13
14 Appendix B. Low-frequency Floquet perturbation theory of driven
15 Landau-Zener model
16
us
17
In this Appendix, we derive analytic expressions for the driven Landau-Zener model in
18
19 the low-frequency regime.
20 The gauge Λ in the adiabatic solutions, Eq. (57), is
21
1 τ
Z
22
23
24
25
26
27
28
Λ(τ ) =
Rρp
=
Ω 0
b
Ω
[E+ (s) − +(0) ]ds
an
E fτ π/2| − a2 /b2 − fτ E(−a2 /b2 ) ,

(B.1)
with E(ρ|x) = 0 1 − x sin2 αdα an incomplete elliptic integral, and fτ = 2{τ /π} − 1
dM
a periodic saw-tooth function ({x} is the fractional part of x). For asymptotic values of
29
30 a/b, we may expand the elliptic functions to find
31 a2
32 Λ(a b) = sin(2τ ) + O(a/b)3 , (B.2)
33 8bΩ
a
34 Λ(a b) ≈ A sin(2τ ) + O((b/a) log(b/a)), (B.3)
35 Ω
36 For a b, we have substituted sgn(fτ )[1 − |fτ | − cos(fτ π/2)]) ≈ A sin(2τ ), with
37 √
A = ( 2 − 1)/2.
38
For these asymptotic expressions of the gauge, Λ = Λ0 sin(2τ ), we expand
pte
39
40 X
41
e2iΛ0 sin(2τ ) = Jk (2Λ0 )e2ikτ , (B.4)
42 k∈Z
43 in Bessel functions Jk . Then, using the integral
44 Z 2π
45 x sin τ eimτ p
d̄τ = i m
( 1 + 1/x2 − 1/x)|m| , (B.5)
46 0 1 + x 2 cos2 τ
ce
47 for odd m and the fact that the integral vanishes for even m, we find zn = 0 for even n
48
49 and, for odd n = 2q + 1,
50 iX
51 z2q+1 = (−1)q+k J2k (Λ0 )y |2(q+k)+1| (B.6)
2 k∈Z
52
Ac
53 ∞
X
54 = (−1)q i Jq,p (Λ0 )y 2p+1 (B.7)
55 p=0
56 p
57 where y = 1 + (b/a)2 − (b/a) ∈ [0, 1], and Jq,p (x) = 12 [Jq−p (2x) + Jq+p+1 (2x)].
58 For small Ω/b 1, the smallest shift is obtained when a/b 1 to be q ∼ b/Ω.
59
60
In this limit, y ≈ a/2b. Assuming a/Ω . 1, we find Λ0 = a2 /(8bΩ) 1. The biggest
1
2
4
pt
5 contribution to z2q+1 would not come from the smallest power of y, since this is multiplied
6 by Jq (2Λ0 ), which is suppressed by the factorial q!. Instead, it comes from the term with
7
J0 , that is
8
9 z2q+1 ≈ y 2q+1 ∼ (Ω/b)2b/Ω , (B.8)
cri
10
11 as discussed in Sec. 4.2.
12
13
14 References
15
16 [1] V. I. Arnold, Geometrical Methods in the Theory of Ordinary Differential Equations, 2nd ed.
us
17 (Springer, New York, 1988).
18 [2] C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Atom-Photon Interactions (Wiley-VCH
19 Verlag GmbH, 1998).
20 [3] P. Bordia et al., Nature Phys. 13,460 (2017).
21 [4] F. Görg et al., Nature 553, 481 (2018).
22
23
24
25
26
27
28
an
[5] R. Desbuquois et al., Phys. Rev. A 96, 053602 (2017).
[6] T. Oka and H. Aoki, Phys. Rev. B 79, 081406 (2009).
[7] L. Jiang et al., Phys. Rev. Lett. 106, 220402 (2011).
[8] N. H. Lindner, G. Refael, and V. Galitski, Nature Phys. 7, 490 (2011).
[9] M. C. Rechtsman et al., Nature 496, 196 (2013).
[10] Y. H. Wang, H. Steinberg, P. Jarillo-Herrero, and N. Gedik, Science 342, 453 (2013).
dM
29 [11] J. Cayssol, B. Dóra, F. Simon, and R. Moessner, Physica status solidi (RRL) 7, 101 (2013).
30 [12] A. Kundu, H. Fertig, and B. Seradjeh, Phys. Rev. Lett. 113, 236803 (2014).
31 [13] D. Carpentier, P. Delplace, M. Fruchart, and K. Gawȩdzki, Phys. Rev. Lett. 114, 106806 (2015).
32 [14] J.-Y. Zou and B.-G. Liu, Phys. Rev. B 93, 205435 (2016).
33 [15] J. Klinovaja, P. Stano, and D. Loss, Phys. Rev. Lett. 116 (2016).
34 [16] M. Thakurathi, D. Loss, and J. Klinovaja, Phys. Rev. B 95, 155407 (2017).
35 [17] A. Eckardt, Rev. Mod. Phys. 89, 011004 (2017).
36
[18] M. Rodriguez-Vega and B. Seradjeh, preprint arXiv:1706.05303.
37
[19] M. Rodriguez-Vega, H. A. Fertig, and B. Seradjeh, preprint arXiv:1803.10646.
38
[20] A. Lazarides, A. Das, and R. Moessner, Phys. Rev. Lett. 112, 150401 (2014).
pte
39
40 [21] D. A. Abanin, W. De Roeck, and F. Huveneers, Ann. Phys. 372, 1 (2016).
41 [22] R. Moessner and S. L. Sondhi, Nature Phys. 13, 424 (2017).
42 [23] V. Khemani, A. Lazarides, R. Moessner, and S. L. Sondhi, Phys. Rev. Lett. 116, 250401 (2016).
43 [24] D. V. Else, B. Bauer, and C. Nayak, Phys. Rev. Lett. 117, 090402 (2016).
44 [25] J. Zhang et al., Nature 543, 217 (2017).
45 [26] S. Choi et al., Nature 543, 221 (2017).
46 [27] M. Reitter et al., Phys. Rev. Lett. 119, 200402 (2017).
ce
47 [28] M. Lohse et al., Nature Phys. 12, 350 (2016).

48 [29] Yu-Ao Chen et al., Phys. Rev. Lett. 107, 210405 (2011).
49
[30] J. Struck, J. Simonet, and K. Sengstock, Phys. Rev. A 90, 031601(R) (2014).
50
[31] N. Fläschner et al., Nature Phys. 14, 265 (2018).
51
[32] F. Casas, J. A. Oteo, and J. Ros, J. Phys. A: Math. Gen. 34, 3379 (2001).
52
Ac
53 [33] E. S. Mananga and T. Charpentier, J. Chem. Phys. 135, 044109 (2011).

54 [34] A. Eckardt and E. Anisimovas, New J. Phys. 17, 093039 (2015).
55 [35] T. Mikami et al., Phys. Rev. B 93, 144307 (2016).
56 [36] P. Mohan, R. Saxena, A. Kundu, and S. Rao, Phys. Rev. B 94, 235419 (2016).
57 [37] M. Bukov, M. Kolodrubetz, and A. Polkovnikov, Phys. Rev. Lett. 116, 125301 (2016).
58 [38] S. Teufel, Adiabatic Perturbation Theory in Quantum Dynamics (Springer, Berlin, 2003).
59 [39] G. Rigolin, G. Ortiz, and V. H. Ponce, Phys. Rev. A 78, 052508 (2008).
60
1
2
4
pt
5 [40] G. Rigolin and G. Ortiz, Phys. Rev. Lett. 104, 170406 (2010).
6 [41] S. N. Shevchenko, S. Ashhab, and F. Nori, Phys. Rep. 492, 1 (2010).
7 [42] B. Mukherjee, P. Mohan, D. Sen, and K. Sengupta, preprint arXiv:1709.06554.
8 [43] J. H. Shirley, Phys. Rev. 138, B979 (1965).
9 [44] H. Sambe, Phys. Rev. A 7, 2203 (1973).
cri
10 [45] H. Martiskainen and N. Moiseyev, Phys. Rev. A 91, 023416 (2015).
11 [46] M. H. S. Amin, Phys. Rev. Lett. 102, 220401 (2009).
12 [47] A. Russomanno and G. E. Santoro, J. Stat. Mech. 2017, 103104 (2017).
13 [48] Q. Cheng et al., preprint arXiv:1804.05134.
14
[49] H. Wang, L. Zhou, and J. Gong, Phys. Rev. B 91, 085420 (2015).
15
[50] P. Weinberg et al., Phys. Rep. 688, 1 (2017).
16
us
17 [51] K. Drese and M. Holthaus, Eur. Phys. J. D 5, 119 (1999).
18 [52] V. Novičenko, E. Anisimovas, and G. Juzeliūnas, Phys. Rev. A 95, 023615 (2017).
19 [53] J. Schwinger, Phys. Rev. 51, 648 (1937).
20 [54] J. P. Davis and P. Pechukas, J. Chem. Phys. 64, 3129 (1976).
21 [55] M. V. Berry, Proc. R. Soc. A 429, 61 (1990).
22
23
24
25
26
27
28
[56]
[57]
[58]
[59]
[60]
[61]
L. Landau, Phys. Zeit. Sowj. 2, 46 (1932).
E. Majorana, Nuo. Cim. 9, 43 (1932). an

C. Zener, Proc. R. Soc. Lond. A 137, 696 (1932).
E. C. G. Stückelberg, Helv. Phys. Act. 5, 369 (1932).
F. Grossmann, T. Dittrich, P. Jung, and P. Hänggi, Phys. Rev. Lett. 67, 516 (1991).
F. Großmann, P. Jung, T. Dittrich, and P. Hänggi, Zeit. Phys. B 84, 315 (1991).
dM
[62] J. M. Gomez Llorente and J. Plata, Phys. Rev. A 45, R6958(R) (1992).
29
30 [63] J. M. Gomez Llorente and J. Plata, Phys. Rev. E 49, 3547 (1994).
31 [64] M. Weinberg, et al., Phys. Rev. A 92, 043621 (2015).
32 [65] C. Sträter, and A. Eckardt, Z. Naturforsch., A: Phys. Sci. 71, 909 (2016).
33 [66] W. P. Su, J. R. Schrieffer, and A. J. Heeger, Phys. Rev. Lett. 42, 1698 (1979).
34 [67] E. J. Meier, F. A. An, and B. Gadway, Nature Comm. 7, 13986 (2016).
35 [68] Z. Sun, L. Zhou, G. Xiao, D. Poletti, and J. Gong, Phys. Rev. A 93, 012121 (2016).
36
37
38
pte
39
40
41
42
43
44
45
46
ce
47
48
49
50
51
52
Ac
53
54
55
56
57
58
59
60

Floquet Perturbation Theory Formalism and Applicat

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Floquet Perturbation Theory Formalism and Applicat

Uploaded by

Copyright:

Available Formats

ACCEPTED MANUSCRIPT • OPEN ACCESS