Nonequilibrium steady states and transient dynamics of conventional superconductors under phonon driving

Yuta Murakami,1 Naoto Tsuji,2 Martin Eckstein,3 and Philipp Werner1 1Department of Physics, University of Fribourg, 1700 Fribourg, Switzerland 2RIKEN Center for Emergent Matter Science (CEMS), Wako 351-0198, Japan 3Max Planck Research Department for Structural Dynamics, University of Hamburg-CFEL, 22761 Hamburg, Germany (Dated: July 20, 2017) We perform a systematic analysis of the influence of phonon driving on the superconducting Hol- stein model coupled to heat baths by studying both the transient dynamics and the nonequilibrium steady state (NESS) in the weak and strong electron-phonon coupling regimes. Our study is based on the nonequilibrium dynamical mean-field theory, and for the NESS we present a Floquet formu- lation adapted to electron-phonon systems. The analysis of the phonon propagator suggests that the effective attractive interaction can be strongly enhanced in a parametric resonant regime because of the Floquet side bands of phonons. While this may be expected to enhance the superconductivity (SC), our fully self-consistent calculations, which include the effects of heating and nonthermal dis- tributions, show that the parametric phonon driving generically results in a suppression or complete melting of the SC order. In the strong coupling regime, the NESS always shows a suppression of the SC gap, the SC order parameter and the superfluid density as a result of the driving, and this ten- dency is most prominent at the parametric resonance. Using the real-time nonequilibrium DMFT formalism, we also study the dynamics towards the NESS, which shows that the heating effect dom- inates the transient dynamics, and SC is weakened by the external modulations, in particular at the parametric resonance. In the weak coupling regime, we find that the SC fluctuations above the transition temperature are generally weakened under the driving. The strongest suppression occurs again around the parametric resonances because of the efficient energy absorption.

PACS numbers: 71.10.Fd

I. INTRODUCTION fects of the dynamical part of the Hamiltonian, it has been argued that the parametric driving of the Raman modes through nonlinear couplings with the mid-infrared The prospect of nonequilibrium exploration and con- phonon modes11,12 or the excited mid-infrared phonon trol of material properties has caught the interest and through a nonlinear electron-phonon coupling13,16 ef- attention of broad segments of the condensed matter 1 fectively pushes the system into the stronger-coupling community. In particular, the resonant excitation of regime, leading to an enhancement of the SC. Still, many mid-infrared phonon modes has opened a new path- important questions remain to be answered. In particu- way for manipulating properties such as metallicity,2 3 4–9 lar, when the enhancement of the SC is attributed to a magnetism and superconductivity (SC), and for mod- change of the time-averaged Hamiltonian or the effective ifying lattice structures in ways which may not be achiev- interaction, it is often assumed that the system remains able in equilibrium. Especially, the light-induced SC-like in thermal equilibrium. However, in general, a period- behavior in the optical conductivity for high Tc cuprates ically driven and interacting many-body system is ex- and doped fullerides4–8 has stimulated several theoretical 10–19 pected to heat up (either by resonant excitations, or on investigations. So far, two possible scenarios for the a longer timescale by multi-photon absorption processes) enhancement of SC have been discussed in the literature and eventually reach an infinite temperature state.20,21 based on the enhanced pairing interaction out of equilib- This would imply a melting of long-range order after a rium: (i) Under time-periodic driving, the time-averaged long time, even if the pairing interactions are enhanced by Hamiltonian can be modified in such a way that super- 10,14,18 the driving. In a real system, this detrimental heating ef- conductivity is favored, and (ii) the interaction can fect may potentially be circumvented in two ways: Either be effectively enhanced in the driven state by dynamical arXiv:1702.02942v3 [cond-mat.supr-con] 19 Jul 2017 11–13,16 the heating processes are slow enough so that an enhance- effects. Both effects can be induced by the reso- ment of a long-range order can be observed at least tran- nant excitation of mid-infrared phonon modes, and pre- siently, or the heating is eventually balanced by energy vious theoretical investigations have tried to clarify their dissipation to the environment, thus leading to the so- influence on SC. called nonequilibrium steady state (NESS). The transient Concerning the effect of time-periodic driving on the dynamics has been studied by looking at the SC instabil- time-averaged Hamiltonian, it has been pointed out that ities of the normal phase, in the form of negative eigen- the resonantly excited mid-infrared phonon can lead to values of the transient pairing susceptibility.11,17 While favorable parameter changes,10,14,18 such as an orbital such negative susceptibilities can occur for short times, imbalance of the Coulomb interaction. As for the ef- it is an open question on which time scale the long-range 2 order could evolve in such a situation, and whether a po- a) tential increase can be fast enough to overcome the heat- H ing. Even less is known about the nature of the NESSs, Ω mix where excitation and dissipation are balanced. For an understanding of the light-enhanced SC, it is therefore W (t) System Hsys essential to go beyond the study of effective interactions and Hamiltonians, and to consider a formalism that can I(t) take account of the transient or steady-state modifica- Bath (in equilibrium) Hbath tions of the electron and phonon distribution functions, and of the SC order or fluctuations, in the nonequilibrium state. b) Keldysh contour In this paper, we focus on the dynamical aspect of 1 a phonon-driven system and discuss the influence of C the parametric phonon driving and the resulting heat- C2 ing and nonthermal distribution on the SC. Specifi- FIG. 1: (a) Schematic picture of the setup we consider in the cally, we study the time-periodic nonequilibrium steady paper. Ω is the frequency of the external field, W (t) is the state (NESS) and the nonequilibrium transient dynam- energy injected from the field, while I(t) is the energy flow ics of the Holstein model, a prototypical model of from the system to the bath. (b) The Keldysh contour. electron-phonon systems, using the nonequilibrium dy- namical mean-field theory (DMFT) combined with the Migdal approximation.22–25 To investigate the time- phonon driving problem can be found in Sec. III E. The periodic NESS of the electron-phonon (el-ph) system results of our systematic analysis are shown in Sec. IV. with externally driven phonons, we extend the Floquet In Sec. IV A, we focus on the modification of the phonon DMFT formalism, which has been previously only ap- propagator in the driven state in the weak coupling limit, plied to purely electronic systems,26–31 to the Holstein and show that an enhancement of the attractive interac- case. tion is realized in some parameter regimes, which naively The Floquet DMFT is based on the Floquet Green’s suggests an enhancement of SC. Sections IV B and IV C function formalism, which can naturally take into ac- are dedicated to the strong coupling regime. In Sec. IV B, count the effect of many-body interactions, periodic driv- we study the NESS under the parametric phonon driving ing by external fields and a dissipative coupling to the using the Floquet DMFT and in Sec. IV C we discuss the environment on equal footing. The Floquet DMFT for transient dynamics toward the NESS. In Sec. IV D, we el-ph systems developed here is therefore not limited to focus on the weak coupling regime, and study how the SC the SC order, and should be useful to study a wide range fluctuations above Tc react to the phonon modulation. A of nonequilibrium situations that involve phonon excita- summary and conclusions are provided in Sec. V tions triggered by “nonlinear phononics”.32 Our systematic analysis, which covers the weak and strong coupling regimes, reveals that the parametric II. MODEL phonon driving generally suppresses the SC (see Fig. 8 and Fig. 15), even though in some parameter regimes In this paper, we consider the case where the system the retarded attractive interaction is increased and coupled to heat baths is periodically driven by an exter- an enhancement of the SC is naively expected (see nal field. Such a situation is expressed by the Hamilto- Fig. 4(a)(d)). In particular in the parametrically reso- nian nant regime, where the system absorbs energy most ef-

ficiently independent of the coupling strength, the SC Htot(t) = Hsys(t) + Hmix + Hbath, (1) order and fluctuations are strongly suppressed. The paper is organized as follows. Section II describes see Fig. 1(a). Here Hsys(t) is the Hamiltonian for the the setup for the Holstein model under external phonon system, which includes terms representing periodic exci- driving and attached to thermal baths. The framework tations. Hbath is the Hamiltonian for heat baths, while used to study the problem is explained in Sec. III and Hmix represents the coupling between the system and the the nonequilibrium Green’s functions are introduced in baths. When the system is continuously excited by pe- Sec. III A. In Sec. III B, we discuss the heat bath di- riodic fields, it is expected that it finally reaches a time- rectly coupled to the phonon degrees of freedom, and in periodic nonequilibrium steady state (NESS) in which Sec. III C we introduce the Floquet Green’s function and the energy injection from the field and the energy flow derive its expression for the case of parametric phonon into the baths are balanced. This situation is different driving. In Sec. III D, we explain the Floquet DMFT for from isolated systems under periodic driving.20,21,33,34 the Holstein model and provide expressions for relevant The system Hamiltonian Hsys(t) consists of the un- physical quantities. Important remarks about the usage perturbed part Hsys,0 and the external periodic field, and limitations of the Floquet DMFT for the parametric Hext(t). Here, we focus on a simple but relevant exam- 3 ple of an electron-phonon coupled system, the Holstein and model, with the Hamiltonian  2 2  X Pi Xi X † X Type 2 : Hph(t) = ω0(t) + (5) H , = v (c cj,σ + h.c.) µ ni 4 4 sys 0 − i,σ − i hi,ji,σ i with ω0(t) = ω0 + ∆ω0 cos(Ωt). X † X † + ω a ai + g (a + ai)(ni 1), (2) 0 i i − i i The first type, Eq. (4), is included in excitations through some of the nonlinear couplings between the mid-infrared † mode and the Raman mode. Its potential effect on su- where ci is the creation operator for an electron with spin σ at site i, v is the electron hopping, µ is the perconductivity has been discussed in Ref. 11. Potential † † effects of the second type of excitations, Eq. (5), have electron chemical potential, ni = ci,↑ci,↑ + ci,↓ci,↓, ω0 † been discussed in Ref. 12. As we will show below, it is the bare phonon frequency, ai is the creation oper- turns out that the type 1 and type 2 excitations give ator for the Einstein phonon, and g is the el-ph cou- qualitatively the same results concerning the enhance- pling. The retarded attractive interaction between elec- ment of the attractive interaction and the suppression of trons is mediated by the phonons, which leads to an s- SC in the NESS and the transient dynamics. Besides wave superconducting state at low enough temperatures. 1 P these two types of excitations, one may also want to con- The SC order parameter is defined as φ = c c T N i i↓ i↑ sider the case where 1 R ω (t)dt = ω as in Ref. 11. which is assumed to be real without loss of generality.h i T 0 0X 0 1 R T 6 The strength of the phonon-mediated attractive interac- If T 0 ω0X (t)dt < ω0, the system goes to the stronger- tion is characterized by the dimensionless el-ph coupling, coupling regime and the SC should be enhanced. How- 2 R λ = ρ0(0)g D (ω = 0), where ρ0(ω) is the density of ever, since this effect is rather trivial we do not discuss states− for the free electrons, and D is the retarded phonon it in this paper. propagator.35 As for the bath, we use free electron and free phonon baths, which are introduced in detail in the following section.

Recent developments in THz or mid-infrared laser techniques make it possible to selectively and strongly ex- III. FORMALISM cite a certain phonon mode. The infrared active phonons can couple to the electron sector through further exci- A. Nonequilibrium Green’s functions tation of another phonon mode (Raman mode) that is linearly coupled to the electron sector11,36 or through a nonlinear-coupling mechanism.8,13,37 Our present inter- The transient time evolution toward the NESS can be est is in the former case. Considering the mid-infrared described in the nonequilibrium Green’s function method 22 phonon oscillations as an external driving force to the formulated on the Kadanoff-Baym (KB) contour KB, C Raman mode through the nonlinear coupling, we study where we define the contour-ordered Green’s functions, the following type of excitations for the phonon sector: 0 † 0 G(t, t ) i CKB c(t)c (t ) , (6a) 2 2 0 ≡ − hT 0i X Pi X Xi D(t, t ) i CKB X(t)X(t ) , (6b) Hph(t) = ω0P (t) + ω0X (t) . (3) ≡ − hT i 4 4 i i for the electrons and phonons, respectively. Here c† is

Here Hph(t) consists of the time-dependent part (Hext(t)) the creation operator for an electron in the system, X P † represents the phonon distortion and C is the contour- and the time-independent part (ω a ai in H , ). KB 0 i i sys 0 ordering operator on the KB contour.T All the operators ω0P (t) and ω0X (t) are periodic in time with a period 2π † † are in the Heisenberg picture, and indicates the T , Xi = ai + a and Pi = (ai a )/i. This Ω i i average with respect to the initial ensemble.h· · · i In this for- type≡ of modulation of the phonon degrees− of freedom has malism, the system is in equilibrium at the initial time been discussed as a potential origin of the enhancement t and is then driven by the periodic external field. In of superconductivity.11,12,17 0 Secs. IV C,IV D we use this formalism to study the tran- sient dynamics. The DMFT formalism for the real-time dynamics, both in electron and electron-phonon-coupled To be specific, we focus on two types of periodic systems, has been discussed in previous works22–24,38 and phonon excitations, will not be presented in detail here. The KB formalism can be inefficient for studying the 2 2 X Pi X Xi NESS, which the system reaches in the long-time limit. Type 1 : H (t) = ω + ω X (t) (4) ph 0 4 0 4 Since the initial correlations are wiped out through the i i coupling to the thermal bath, they should be irrelevant with ω0X (t) = ω0 + ∆ω0 cos(Ωt), to the NESS.22 Hence, we can simplify the problem by 4 assuming t0 = and neglecting the left-mixing, right- the self-energy correction Σbath, which is characterized mixing and Matsubara−∞ components in the KB formalism by (which reduces to the Keldysh formalism), and focus di- R rectly on the NESS.61 In the Keldysh formalism, we focus Γ(ω) ImΣ (ω). (10) ≡ − bath on the contour K = 1 2 [see Fig. 1(b)], and hence only need to considerC C ∪ C All of the physical components of Σbath can be ob- tained through the fluctuation-dissipation theorem for  11 0 12 0  Σ (ω).22 The total self-energy is the sum of the con- ˇ 0 G (t, t ) G (t, t ) bath G(t, t ) 21 0 22 0 . (7) ≡ G (t, t ) G (t, t ) tributions from the bath and the interactions in Hsys,

ij 0 0 Σ = Σ + Σ . (11) Here G (t, t ) indicates t i and t j. tot bath int The physical representation∈ C for the∈ Green’s C function is defined as In the following, we use a heat bath with a finite band width,  R K    G G ˇ ˇ ˇ† ˇ 1 1 1 G A = Lσˇ3GL with L = − , (8) 0 G 1 1 s 2 ≡ √2  ω  Γ(ω) = γel 1 , (12) whereσ ˇ3 is the third component of the Pauli matrices. − Webath This is the so-called Larkin-Ovchinnikov form. We can to avoid the divergence of the bath self-energy in the time apply the same transformation to the phonon Green’s 0 function D. representation at t = t . We note that this bath tends If there is no interaction or coupling to external baths, to suppress the SC order, since it reduces the lifetime of quasiparticles. one can evaluate the bare Green’s function G0 for elec- trons (D0 for phonons), even in the presence of an exter- As for the heat bath attached to the phonon degrees of nal field. The correlations coming from the interaction freedom, we consider a free-boson bath as in the Caldeira- and baths can be taken into account through the self- Leggett model,39 which is expressed as energy Σ for electrons (Π for phonons). Since the initial X † † correlation can be neglected, the Dyson equation in the H , = p(a + ai)(b + bi,p), (13a) mix ph V i i,p Keldysh formalism becomes simple: i,p X † Z ∞ Hbath,ph = ωpb bi,p. (13b) −1 0 0 i,p dt1[G0 (t, t1) Σ(t, t1)] G(t1, t ) = Iδ(t t ), i,p −∞ − · − (9a) Here b† is the creation operator for a bath phonon with Z ∞ i,p −1 0 0 an index p, which is attached to the system phonon at dt1[D0 (t, t1) Π(t, t1)] D(t1, t ) = Iδ(t t ). −∞ − · − site i. We can trace out the bath phonons in the path- (9b) integral formalism, which yields an additional self-energy correction to the phonon Green’s function in the system, Here indicates the matrix product, I is the 2 2 iden- · R −1 0 × 0 0 X 2 0 tity matrix, dt1G0 (t, t1)G0(t1, t ) = Iδ(t t ) (the Πi,bath(t, t ) = p D0,p(t, t ), (14) −1 − V same holds for D0 ). Precisely speaking, Eq. (9b) is true p as long as the average phonon distortion X(t) remains 0 D (t, t0) = i[θ (t, t0) + f (ω )]e−i(t−t )ωp zero, which is the case considered in thish paper.i When 0,p c b p − 0 0 −i(t −t)ωp X(t) is finite, the phonon propagator has an additional i[θc(t , t) + fb(ωp)]e , (15) contributionh i i X(t) X(t0) . − − h ih i where D0,p is the bare bath phonon propagator in equi- librium, fb is the boson distribution function with the 0 B. Thermal baths bath temperature T , and θc(t, t ) is the Heaviside step function along the Keldysh contour. We note that the total phonon self-energy includes the contribution from In this paper, we consider two types of thermal baths. the bath and that from the interaction in H , One is attached to the electron degrees of freedom and sys the other is attached to the phonon degrees of freedom. Πtot = Πbath + Πint. (16) For the bath attached to the electrons we employ the so-called B¨uttiker model, which consists of multiple sets Since Π (t, t0) is time-translation invariant, we can of free electrons in equilibrium connected to each site of bath consider its Fourier components. The retarded part is the system. The model has been used in the previous 22,29,31 Floquet DMFT studies for electron systems. Here X 2ωp ΠR (ω) = 2 , (17) we briefly review the model. Due to the noninteracting bath Vp (ω + i0+)2 ω2 nature of the bath, it can be traced out exactly to yield p − p 5 and the bath information is contained in its spectrum this function is defined as22

R B (ω) = ImΠ (ω) Fmn(ω) (22) bath − bath ≡ X T ∞ 2 Z Z m+n = π p [δ(ω ωp) δ(ω + ωp)]. (18) 1 i(ω+ Ω)tr i(m−n)Ωtav dtav dtre 2 e F (tr; tav). p V − − T 0 −∞ This representation has some important properties. At When Bbath(ω) is given, the self-energy from the bath can be expressed as each ω, Fmn(ω) can be regarded as a matrix whose in- dices are m and n (the Floquet matrix). In this rep- Z 0 R 1 Bbath(ω ) 0 resentation, the convolution of two functions satisfying Πbath(ω) = 0 + dω , (19a) Eq. (21) in the two-time representation becomes the π ω ω + i0 22 A R −∗ product of the Floquet matrices at each ω. This sim- Πbath(ω) = Πbath(ω) , (19b) plifies the Dyson equation, Eq. (9), to K βω  R A Πbath(ω) = coth [Πbath(ω) Πbath(ω)] −1 2 − [G0 (ω) Σ(ω)] G(ω) = I, (23a) − · βω  −1 = 2i coth B (ω). (19c) [D0 (ω) Π(ω)] D(ω) = I. (23b) − 2 bath − · Here G(ω) indicates the Larkin-Ovchinnikov form with In the following, we consider a bath spectrum of the each physical component expressed in the Floquet rep- form resentation, and is the matrix product with respect to ·   the Floquet representation and the Larkin-Ovchinnikov 2 γ γ form. Hence the Dyson equation, which has been origi- Bbath(ω) = 2 2 2 2 . V (ω ωD) + γ − (ω + ωD) + γ nally a Volterra equation in the two-time representation, − (20) becomes a matrix inversion problem in the Floquet rep- resentation. In addition, from the definition it follows Here we assume γ = ωD, which yields an almost linear that Fm,n(ω) = Fm+l,n+l(ω lΩ). We deal with this structure in B (ω) around ω [ ωD, ωD]. This can be − bath ∈ − redundancy by using the reduced Brillouin-zone scheme regarded as an Ohmic bath with a soft cut-off. Because Ω Ω restricting the frequency to ω [ 2 , 2 ). of the non-vanishing real part of the bath self-energy, the When an external periodic∈ field− is applied to the bath coupling softens the phonon frequency, leading to a electrons, we need to incorporate this effect into the 2 stronger attractive interaction. We put γ = /ωD in ph V bare electron propagator, which acquires a time-periodic the following. structure as in Eq. (21). The corresponding expressions have been discussed in the previous works for the single- band28 and multi-band31 systems, so we do not repeat this here. Instead, we focus on the direct phonon modu- lation originating from a THz or mid-infrared laser pump, C. Floquet Green’s functions Eq. (3). In this case, we need to first derive the expres- sion for the bare phonon propagator in the presence of such a modulation, in order to understand the effect of We apply the Keldysh formalism introduced in direct phonon excitations and to construct the perturba- Sec. III A to a time-periodic NESS realized under an tion theory on top of it. external periodic field, where we assume that relevant The free phonon system under the periodic driving is physical observables become periodic in time. We can described by the Hamiltonian take advantage of this to simplify the Keldysh formalism. For example, the Green’s function becomes periodic with P 2 X2 H (t) = ω P (t) + ω X (t) . (24) respect to the average time tav, ph 0 4 0 4 R,A,K R,A,K G (tr; tav) =G (tr; tav + T ), The operator X in the Heisenberg representation satisfies (21) R,A,K R,A,K the equation D (tr; tav) =D (tr; tav + T ),   0 t+t0 2 ∂tω0P (t) where t = t t and t = . The self-energies ∂ + ∂t ω0P (t)ω0X (t) X(t) = 0. (25) r av 2 − t ω (t) − have the same− property in the NESS. Hence, the full 0P R,A,K R,A,K tav-dependence of G (tr; tav) and D (tr; tav) con- From this, we find that tains redundant information. We can eliminate this re- dundancy by considering the Floquet representation of 2 ∂tω0P (t) ∂t + ∂t ω0P (t)ω0X (t) −1 0 − ω0P (t) − 0 these functions in the following manner. Let us assume D0 (t, t ) δ(t t ) ≡ 2ω0P (t) − that a function F (tr; tav) has the periodicity of Eq. (21). The Floquet representation (the Floquet matrix form) of (26) 6 satisfies the time domain as

Z 0 0 −1 R 0 0 R 0 0 i(S(t)−S(t )) −i(S(t)−S(t )) dtD¯ (t, t¯)D (t,¯ t ) = δ(t t ), (27a) D0 (t, t ) = θ(t, t )i[e e ], (33) 0 0 − − Z ∆ω0 ¯ −1 ¯ K ¯ 0 where S(t) tω0 + Ω sin(Ωt). dtD0 (t, t)D0 (t, t ) = 0, (27b) ≡ Z −1 A 0 0 dtD¯ (t, t¯)D (t,¯ t ) = δ(t t ). (27c) D. Dynamical mean-field theory 0 0 −

Hence, in the Dyson equation for the full phonon Green’s In order to evaluate the full Green’s functions, we em- −1 0 −1 0 22,40–42 function, Eq. (9b), we can use D0 (t, t ) = ID0 (t, t ). ploy the dynamical mean-field theory (DMFT), −1 0 In the Floquet representation, D0 (t, t ) becomes which becomes exact in the limit of infinite spatial dimensions.42 The idea of DMFT is to map the lattice inv (ω + nΩ)(ω + mΩ)[ω ]mn [ω X ]mn [D ]−1 (ω) = 0P − 0 , problem to an effective impurity problem. In the present 0 mn 2 case of the Holstein model with thermal baths, the form (28) of the effective impurity model in the path-integral for- malism is where Z 0 † − 0 0 0 T ˆ 1 ˆ Z Simp = i dtdt Ψ (t)( 0 (t, t ) Σbath(t, t ))Ψ(t ) 1 i(m−n)Ωtav [ω0X ]mn = dtave ω0X (tav), (29a) CK G − T 0 Z −1 0 0 0 D0 (t, t ) Πbath(t, t ) 0 Z T i(m−n)Ωtav + i dtdt X(t) − X(t ) inv 1 e 2 [ω0P ]mn = dtav . (29b) CK T 0 ω0P (tav) Z † ig dtX(t)Ψ (t)ˆσ3Ψ(t), (34) We note that the inverse of the Floquet matrix of the re- − CK R −1 + −1 tarded part, [D0 (ω)] , is equal to [D0(ω + i0 )] . In the numerical implementation, we solve the Dyson equa- where tion for the phonon Green’s function using Eq. (28) and −1 0 0 0 ˆ (t, t ) = [i∂tIˆ+ µσˆ ]δC (t, t ) ∆(ˆ t, t ), (35) Eq. (23b). G0 3 K − R denotes an integral on the Keldysh contour, Ψ†(t) Now we focus on more specific situations described CK † ≡ by the driving protocols of type 1 [Eq. (4)] and type 2 [c↑(t), c↓(t)] a Nambu spinor, and a hat symbol a 2 2 [Eq. (5)]. For the excitation of type 1 [Eq. (4)], the in- matrix in the Nambu form. In this expression, the ther-× verse of the bare phonon Green’s function is obtained by mal baths and P (phonon momentum) in the impu- substituting rity problem have been integrated out.24 The hybridiza- tion function ∆ˆ is determined such that the impurity ∆ω 0 † 0 0 Green’s function, Gˆimp(t, t ) = i TC Ψ(t)Ψ (t ) , and [ω0X ]mn = ω0δm,n + (δm,n+1 + δm,n−1), (30a) − h K i 2 the impurity self-energy Σˆ imp,int are identical to the lo- inv 1 ˆ 0 [ω0P ]mn = δm,n, (30b) cal lattice Green’s function for the electrons, Gloc(t, t ) = ω0 † 0 i TC Ψi(t)Ψ (t ) , and the local lattice self-energy, re- − h K i i R 0 spectively. into Eq. (28). Since the equation of D0 (t, t ) turns out to be the Mathieu’s differential equation (see also Since we are considering a time-periodic NESS, the effective impurity model also has the same time period- Sec. III E), we can express it as a linear combination of ˆ its two Floquet solutions. icity, where the hybridization function, ∆, is assumed to show the periodic behavior of Eq. (21). Hence the Dyson For the excitation of type 2 [Eq. (5)], the inverse of the equation involved in the solution of the effective impu- bare phonon Green’s function is obtained from rity model can be dealt with as an inverse problem of the Floquet matrix. Precisely speaking, the matrix to ∆ω0 invert involves three indices, i.e., those of the Floquet, [ω X ]mn = ω δm,n + (δm,n + δm,n− ), (31) 0 0 2 +1 1 Larkin-Ovchinnikov, and Nambu forms. ∞  2l+r inv 1 X (2l + r)! ∆ω0 [ω0P ]mn = − (32) ω0 (l + r)!l! 2ω0 l=0 1. Migdal approximation 1  ∆ω r 1 + r 2 + r ∆ω2  = − 0 F , ; 1 + r; 0 , 2 1 2 In order to solve the effective impurity ω0 2ω0 2 2 ω0 model, we employ the self-consistent Migdal 23–25,35,43–47 where we put r = m n and 2F1(α, β; γ; z) is the hyper- approximation, which is justified when − R geometric function. One can also easily express D0 in the phonon frequency ω0 is small enough compared 7 to the electron bandwidth.35,43–45 Here, the electron (1) The work done by the field is self-energy (Σ)ˆ and phonon self-energy (Π) are given by ( ∆ω0Ω sin(Ωt) P 2 Xi (t) (Type 1), 0 2 0 0 W (t) = − 4 ih i Σˆ int(t, t ) = ig Dimp(t, t )ˆσ3Gˆimp(t, t )ˆσ3, ∆ω0Ω sin(Ωt) P 2 2 (36) 4 i( Pi (t) + Xi (t)) (Type 2), 0 2 0 0 − h i h i Πint(t, t ) = ig Tr[ˆσ3Gˆimp(t, t )ˆσ3Gˆimp(t , t)], (40) − where we do not explicitly write the Hartree term. Re- where X2(t) can be obtained from the lesser part of the ˆ 0 h i i member that in DMFT Σint(t, t ) is identified with the phonon Green’s function. lattice self-energy of the electrons at each momentum. (2) The dissipated energy is The Hartree term is proportional to X(t) , and, in the h i half-filled case, which we focus on in this paper, it is I(t) = Iph(t) + Iel(t) exactly zero. Away from half-filling, X(t) can be finite h i = i [Hph mix(t),Hsys(t)] i [Hel mix(t),Hsys(t)] , and time dependent. Still, since we are considering homo- − h i − h (41)i geneous excitations, this term can be regarded as a time- dependent chemical potential and does not affect the dy- where namics of relevant physical quantities without couplings X < 0 to the B¨uttiker-type bath. The self-consistent Migdal I (t) = i ∂t[Di Πi, ] (t, t ) t0 t (42) ph ∗ ph bath | = approximation is a conserving approximation, which al- i lows one to trace the energy flow from the system to the thermal baths. and X ˆ ˆ ˆ < Iel(t) = Tr[Σel bath ~ ∆i ~ Gii] (t, t) + H.c. i X < µ Tr[ˆσ Σˆ Gˆii] (t, t) + H.c − 3 · el bath ~ 2. Observables i X < Tr[Σˆ ,i Gˆii Σˆ ] (t, t) + H.c. (43) − int ~ ~ el bath i We summarize the physical observables that are used to discuss the results in the paper. Let Here the symbol indicates the convolution along the ∗ us denote the Heisenberg representation as (t) = Keldysh contour K and the symbol ~ represents a multi- O C tot(t0, t) tot(t, t0), where tot(t, t0) is the unitary evo- plication of the 2 2 Nambu matrices and the convolution 0 R 0 U OU U ˆ ˆ 0 × P ¯ ¯ 0 ¯ lution operator from time t0 to t. i.e., [A ~ B]α,α (t, t ) = γ C dtAα,γ (t, t)Bγ,α (t, t ). Tr refers to the trace of Nambu matrices. The change of the system energy can be expressed as

dEsys(t) The system observables (that Hsys involves) can be = ∂tH (t) + i [H (t),H (t)] . (37) dt h sys i h mix sys i expressed in the following manner, in analogy to the Kadanoff-Baym formalism.23 The first term represents the work done by the external 1. Kinetic energy field, which we express as W (t). The second term repre- sents the energy dissipation from the system to the bath, X † E (t) = vi,j c (t)cj,σ(t) kin − h i,σ i which we express as I(t). If the system is in a time- i,j,σ periodic NESS, the average− energy injected from the field = iTr[∆ˆ Gˆ ]<(t, t). (44) and dissipated to the baths should be the same, − ~ loc W = I,¯ (38) 2. Interaction energy where the overline indicates the time average over one X ˆ † ˆ ˆ ˆ < EnX(t) = g Xi(t)Ψi σˆ3Ψi(t) = iTr[Σint ~ Gloc] (t, t). 1 R T h i − i period, T 0 dt. We also note that the energy dissipated from the system should be equal to the energy injected (45) to the bath. This leads to 3. Phonon energy I¯ = i [H (t),H (t)] , (39) h mix bath i ω0X (t) X ω0P (t) X E (t) = (iD<(t, t)) + (iD< (t, t)). see also Appendix A for more details. ph 4 i 4 PP,i i i In the following, we give the explicit expression for (46) each term in Eq. (37) using the self-energies and Green’s 0 0 functions. Here DPP(t, t ) = i TcP (t)P (t ) . This can be ob- tained by taking derivatives− h of D(it, t0).23 The Larkin- 8

Ovchinnikov form of this function becomes ω0 =0.4,Ω =0.7,γph =0.02,ωD =0.6

∆ω0 =0.13:Volterra 0 8 0 1 0 2δ(t t ) ∆ω0 =0.14:Volterra D (t, t ) = ∂ ∂ 0 D(t, t ) I. PP 0 t t − ∆ω0 =0.13:Floquet ω0P (t)ω0P (t ) − ω0P (t) ∆ω0 =0.14:Floquet (47) 4 ) r t We note that the derivative in the two-time represen- ( 0 R tation becomes a multiplication of factors depending on D ω, m and n in the Floquet representation , which can be 4 used for the numerical implementation. Another important quantity is the optical conductiv- 8 0 0 ity, i.e. σαα(t, t ) δ jα(t) /δEα(t ), where jα(t) is ≡ h i 0 h i 0 50 100 150 200 250 300 350 the current for the direction α, Eα(t ) is the probe field, tr and δ indicates the functional derivative.22 This can 0 be expressed in terms of the susceptibility χαα(t, t ) 0 ≡ FIG. 2: Comparison between solutions of the Dyson equation δ jα(t) /δAα(t ) as h i obtained from the Volterra equation, Eq. (9b), and the Flo- quet representation, Eq. (23b). D¯ R(t ) is DR(t ; t ) averaged Z t r r av 0 over tav. Only the self-energy correction from the phonon σαα(t, t ) = χαα(t, t¯)dt.¯ (48) − t0 bath is considered.

0 dia 0 χαα(t, t ) consists of the diamagnetic term χαα(t, t ) and −i(ν+mΩ)t −iνt para 0 22 e components in addition to e . Previous the paramagnetic term χαα (t, t ). In principle, the studies29,31 have mainly focused on the diagonal compo- paramagnetic term includes the vertex correction, but nent (m = 0), while in the present work, we will also in the present case, since the phonon driving does not discuss the off-diagonal components. break the parity symmetry (k k), its contribution vanishes. In the end, we obtain↔ the − following expression in the Nambu form, E. Breakdown of the Floquet Green’s function method Z dΦ() χdia(t, t0) = iδ(t t0) d Tr[ˆσ Gˆ<(t, t)], (49a) − d 3  Z n It is important to remark that there are instabilities para 0 ˆR 0 ˆ< 0 χ (t, t ) = i dΦ() Tr[G (t, t )G (t , t) in the system under the excitation Eq. (4). Without the heat bath and the el-ph coupling, the equation for the o ˆ< 0 ˆA 0 + Tr[G (t, t )G (t , t)] , phonon Green’s function [Eq. (27) combined with Eq. (4)] becomes the so-called Mathieu’s differential equation. It (49b) is known that the solution of this equation is unstable n where  denotes the energy of noninteracting elec- around 2 Ω = ω0, with n a natural number. In this un-  2 stable regime, the amplitude of oscillations in the solution 1 P ∂k dΦ() trons, Φ() = δ( k), = Ωvol k,σ ∂kα d of the equation monotonically increases, and there is no 2 − 1 P ∂ k well-defined Floquet form of the Green’s function since k ∂k2 δ( k), and Ωvol is the volume of the Ωvol α − we cannot define the Fourier components. system. In particular, for the Bethe lattice, Φ() = The heat bath and the el-ph coupling introduce a (N/3d)[(W/2)2 2]ρ () is used, where N is the system 0 damping effect through the self-energy corrections. Even size and 2d the− coordination number.48 though the damping effect can reduce the unstable 0 0 In the NESS, σ(t, t ) and χ(t, t ) become time periodic, regime,49,50 the Green’s function still exhibits an instabil- and their Floquet expressions are connected through ity when the strength of the external field is large enough. In such cases the Floquet Green’s function cannot be de- χmn(ω) σmn(ω) = . (50) fined and the Floquet formalism becomes inapplicable. i(ω + nΩ) Physically, this means that the NESS, if any, is unstable against infinitesimally small perturbations to the phonon To obtain an intuitive picture, let us imagine that part. E (t) = δE e−iνt is applied as a probe field. The in- α 0 In Fig. 2, we demonstrate that even with the phonon duced current is bath the behavior of the retarded part of the phonon X −i(ν+mΩ)t Green’s function changes qualitatively as a function of δjα(t) = δE0 e σm0(ν). (51) m driving amplitude. The correct results are obtained by solving the Dyson equation in the time domain We note that in this expression we unfold the restricted (differential-integral Volterra equation), Eq. (9b), which ij zone scheme for σm0(ν) by using σm0(ν) = σm+l,l(ν produces either a stable or unstable solution depending lΩ). Equation (51) means that the response includes− on the excitation parameter. For comparison, we solve 9

for the SC, is modulated under the phonon driving in the weak coupling limit. Our discussion is based on the Migdal-Eliashberg theory, where the retarded attractive T interaction mediated by the phonons is represented by the phonon propagator, g2DR(ω), and g2DR(0) is con- ventionally taken as a measure for the strength of the attractive interaction. In the weak coupling limit, the renormalization of the phonon (the self-energy correc- SC 2 R tion) is small and g D0 (ω) serves as a measure of the ef- λ =1 fective attractive interaction. With this idea in mind, the λ ¯ R 1 R T R ∞ R time-averaged D0 (0) = T 0 dtav −∞ dtrD0 (tr; tav) FIG. 3: Schematic picture of the parameter regimes discussed has been analyzed for the type 2 excitation in Ref. 12 in this paper. The dashed line represents λ = 1. The effect of the type 1 excitation, on the other hand, has been discussed in Ref. 11 using the polaron pic- the Dyson equation in the Floquet form, Eq. (23b), and ture and the Lang-Firsov transformation. The polaron approximation works well when the phonon timescale carry out an inverse transformation to the two-time rep- 51 resentation. These two ways of solving the Dyson equa- is comparable or faster than the electronic timescale. tion give the same solution, when the solution is sta- Within this framework, the hopping of polarons is renor- malized from the bare electron hopping parameter by the ble. However, we have found an unphysical solution for 2 2 the unstable regime in the Floquet form that is different Frank-Condon factor, exp( g /ω0), and the effective at- − 2g2 from the transient solution of the Volterra equation in tractive interaction is λ0 = . However, in many ma- − ω0 the time domain. In particular, the unphysical solution terials the phonon frequencies are much smaller than the violates the initial condition (the sum rule in the ω space) bandwidth, in which case the Migdal-Eliashberg theory R 0 ∂tD0 (t, t ) t0=t = 2ω0, see also Sec. IV A. is applicable. We note that these two different theories Therefore,| when− using the Floquet form of the Green’s can lead to different predictions concerning the behavior function, one has to be careful about the unstable regime, of the SC order. The isotope effect serves as an example. where the system cannot reach a physically stable NESS. By the substitution of isotopes, usually ω0 decreases as Ω 1 The excitation Eq. (4) at the parametric resonance = √ (M is the mass of the atom), while λ0 does not 2 ∝ M ω0 is potentially in such a regime. However, there is change. In the polaron picture, the system moves to the no such problem in the calculation of the transient time stronger coupling regime, since the Frank-Condon fac- evolution on the Kadanoff-Baym contour. tor decreases. The convex shape of the superconducting phase boundary in the weak-coupling regime (see Fig. 3) thus implies an increase of the transition temperature. IV. RESULTS On the other hand, the BCS theory, to which the Migdal- Eliashberg theory reduces in the weak coupling regime, In this paper, we investigate the effect of parametric predicts the opposite behavior. Here, the transition tem- 1 perature is proportional to ω0 exp( ), which de- phonon driving on superconductivity (SC) both in the − λ0ρ0(0) weak coupling (λ 1, the red region in Fig. 3) and creases if ω0 is reduced by the substitution of isotopes. strong coupling (λ 1, the blue region in Fig. 3) regimes. Hence for realistic phonon frequencies we should discuss In this paper, we use∼ the semi-elliptic density of states, the effects of parametric phonon excitations within the 1 p 2 2 Migdal-Eliashberg theory. ρ0(ω) = πv2 4v∗  , i.e. the Bethe lattice. We set 2 ∗ − v∗ = 1, i.e. the electron bandwidth W is 4, and we focus In Fig. 4, we show the numerical results for ¯ R on the half-filled case. We choose the bare phonon fre- the time-averaged phonon propagator, D0 (ω) = T ∞ quency ω0 = 0.4, which is much smaller than the electron 1 R R R iωtr T 0 dtav −∞ dtrD0 (tr; tav)e , which corresponds to bandwidth. If we take the bandwidth as 0.4 eV, the bare the diagonal components in the Floquet representation, phonon frequency is 40 meV, corresponding to a period for both types of excitations. In the evaluation, we for- of about 100 fs. The gap size of the SC spectrum around R−1 mally invert the Floquet matrix of D0 . Therefore, the λ 1 is approximately 10 meV, i.e. this period is about ∼ results also include unphysical ones as we explained in 400 fs. Our unit temperature corresponds to 0.1 eV or Sec. III E, but we can detect the unstable regime from 1160 K. the Mathieu’s equation. In Fig. 4(a)(d), we plot the change in the effective attractive interaction expressed ¯ R as (ReD0 (0) ReD0,eq(0))/ReD0,eq(0) in the plane of A. Effective attractive interaction in the weak − ∆ω0 and Ω. It turns out that both excitations result in coupling limit : Bare phonon propagator a similar behavior. There is a strong enhancement of the ¯ R attractive interaction D0 (0) at Ω . ω0. On the other In this section, we discuss how the effective attractive hand, for Ω & ω0, the attractive interaction is strongly interaction mediated by phonons, which is responsible suppressed and it can even become repulsive as is ana- 10

1 1 (a) (b) (c) 0.8 0.5 0.6

Ω 0 0.4

Type 1 -0.5 0.2

0 -1 0 0.1 0.2 ∆ω0 1 1 (d) (e) (f) 0.8 0.5 0.6

Ω 0 0.4 Type 2 -0.5 0.2

0 -1 0 0.1 0.2 ∆ω0

¯ R FIG. 4: Properties of D0 (ω) for the type 1 excitation (a)(b)(c) and the type 2 excitation (d)(e)(f). Here we choose ω0 = 0.4. ¯ R (a)(d) show the change of the effective attractive interaction expressed as (ReD0 (0) − ReD0,eq(0))/ReD0,eq(0). The yellow and green dotted lines in (a) indicate the boundaries of the unstable regimes of the Mathieu equation around Ω = 2ω0 and Ω = ω0. ¯ R ¯ R (b)(c)(e)(f) show the excitation-frequency dependence of D0 (ω) along the black dashed line in (a)(d). (b)(e) show ReD0 (ω) ¯ 1 ¯ R and (c)(f) show B0(ω) ≡ − π ImD0 (ω). The dotted color lines (Ω = 0.75) in (b)(c) correspond to an unphyscial solution. The 0 spectra are broadened by adding a damping term −γ∂tδ(t − t ) with γ = 0.005 in Eq. (26). lytically shown in Ref. 12. dotted lines in Fig. 4(a). Therefore, when m = 2n, This behavior is related to the position of the Flo- the unstable regime overlaps with the Ω-regime in which ¯ R quet sidebands of the phonons in the phonon spectrum, D0 (0) exhibits drastic changes. On the other hand, when ¯ 1 ¯ R B0(ω) π ImD0 (ω). At Ω . ω0, the first replica peak m = 2n 1 the existence of the unstable regime does not ≡ − − ¯ R ( ω0 Ω) is just above ω = 0. At Ω & ω0, the first side leave any trace in D0 (0). It turns out, instead, that the peak∼ moves− to ω < 0 and there emerges a negative peak effect of the instability emerges in the finite frequency ¯ R in the spectrum at ω > 0, which is the first sideband of part of D0 (ω). As can be seen in the result for Ω = 0.7 the phonon peak at ω0, in the time-averaged spectrum and Ω = 0.75 in Fig. 4 (b)(c), there is a large change − B¯ (ω0) R ¯ ¯ R R 0 0 in D¯ (ω) around ω ω0. The results for Ω = 0.75 are B0(ω), see Fig. 4 (c)(f). Since D0 (ω) = dω ω−ω0+i0+ , 0 unphysical as discussed∼ in Sec. III E, and the spectrum in the former case (Ω . ω0) the sideband acts as an additional phonon coupled to electrons with a positive shows dominant negative weight at ω > 0, which leads R ∞ ¯ frequency, which leads to an additional attractive in- to the violation of the sum rule, 2ω0 = −∞ ωB(ω)dω. ¯ R teraction. In the latter case, the sideband resembles a Based on the behavior of D0 (0), we thus do not expect negative frequency phonon attached to electrons, which any enhancement of SC around the parametric resonant yields an effective repulsive interaction. There are anal- regime at m = 2n 1 for the type 1 excitation, and the ¯ R − ogous singularities around Ω = ω/n related to the zero analysis of D0 (0) results in a picture which differs from crossing of the n-th Floquet sideband of the phonons. the enhancement of SC at the parametric resonance dis- Since the spectral weight of such high-order sidebands is cussed in Ref. 11. For the type 2 excitation, there is small, a substantial effect only appears when the exci- no unstable solution, and we do not observe any large ¯ R tation strength ∆ω0 is large enough. Even though the change in D0 (ω) near ω = ω0, see Fig. 4 (e)(f). strong enhancement of the attractive interaction is re- Even though the increased attractive interaction 2 ¯ R stricted to some energy window smaller than ω0, see g D0 (0) seems a plausible scenario for the enhancement Fig. 4(b)(e), in the weak coupling limit, the effect of the of SC, this naive expectation ignores the following issues: energy window is smaller than that of the strength of the (i) The system gains energy from the external field, which attractive interaction as can be seen from the BCS ex- results in heating and nonthermal distribution functions. pression for the transition temperature. The former is a (ii) When the electron-phonon coupling is not small, the prefactor, while the latter is in the exponential. Hence phonon spectrum is renormalized and broadened, and the ¯ R it is natural to expect that this may lead to an enhance- behavior of D might be different from D0. (iii) D0 (0) is R ment of SC assuming that the electron distribution is the an average of D0 (tr; tav), and hence includes informa- same as in equilibrium. tion of infinite past and infinite future times. Therefore, ¯ R Here we note that the unstable regime in the case of it is not clear to what extent D0 (0) is relevant to the 2ω0 the type 1 excitation is located around Ω = m . The transient dynamics of the superconducting order. boundary of the unstable region is depicted by colored In the following sections, using the nonequilibrium 11

0 (a) through the electron-phonon coupling. In equilibrium, Eq:β =120 ∆ω0 =0.04 because of the strong el-ph coupling, the sign change Eq:β =50 ) ∆ω0 =0.05 ¯ R ω 10 ∆ω0 =0.03 ∆ω =0.08 in ReD (ω) around the renormalized phonon frequency, ( 0 ωr, is smeared out, see β = 50 (above Tc) in Fig. 5(a). R R

¯ ¯ D Below Tc [β = 120 in Fig. 5(a)], even though D (0)

e 20 | |

R remains almost unchanged, a strong renormalization oc- curs for ω = 0. This is caused by the opening of the SC 6 30 gap, within which the scattering between electrons and 0.2 0.1 0.0ω 0.1 0.2 phonons is suppressed. The peak in ReD¯ R(ω) around | | 0 ω = 0.08 corresponds to this energy scale. This effect is (b) Eq:β =120 ∆ω0 =0.03 known as the phonon anomaly in strongly coupled el- ∆ω =0.01 24,52–54 ) 0 ∆ω0 =0.04 ph superconductors, see also Fig. 6(c). In the ω 10 ∆ω0 =0.02 ∆ω =0.05 ( 0 NESS, as in the case of the bare phonon propagator, R

¯ the sign of the change in the effective attractive interac- D 2 R e 20 tion g D¯ (0) depends on the excitation frequency Ω, de- R spite the strong renormalization and the absence of sharp structures. Namely, for Ω ω , the strength g2D¯ R(0) 30 . r | | 0.2 0.1 0.0ω 0.1 0.2 increases [see Fig. 5(a)], while for Ω & ωr, it decreases [see Fig. 5(b)]. However, this enhancement of D turns out not to lead to an enhancement of the superconduc- FIG. 5: Real part of the retarded and time-averaged full tivity. phonon propagator, ReD¯ R(ω), for g = 0.41, ω = 0.4, β = 0 Figure 6 illustrates how the (time-averaged) nonequi- 120 and the bath parameters γph = 0.02, ωD = 0.6, γel = librium spectrum and the distribution functions for the 0.005,Webath = 2.0. The equilibrium phonon propagator in the normal phase at β = 50 is also shown in (a) for compari- electrons and phonons change as we increase the am- son. (a) is for Ω = 0.125 and (b) is for Ω = 0.4. plitude of the excitation. Here we choose Ω = 0.125 for which an enhancement of the effective attractive in- teraction has been observed [Fig. 5(a)]. The nonequi- DMFT and the Holstein model, we provide a compre- librium spectrum is defined as A¯(ω) 1 ImG¯R(ω) hensive study of the effects of the parametric phonon ≡ − π and the distribution function is f¯el(ω) N¯(ω)/A¯(ω) driving on SC which takes the above issues into account. ¯ 1 ¯< ≡¯R(<) with N(ω) = 2π ImG (ω). Here G (ω) = We also note that it turns out that the type 1 and type ∞ 1 R T R R(<) iωtr 2 excitations give the same results on a qualitative level. T 0 dtav −∞ dtrG (tr; tav)e . As we increase the Therefore, in the main text, we only show the results for excitation amplitude ∆ω0, the size of the SC gap de- the type 1 excitation, and present the results for the type creases and eventually closes completely, see Fig. 6(a). At 2 excitation in Appendix B. the same time, the slope of the nonthermal distribution function for the electrons decreases around ω = 0 and high energy states become more occupied, which roughly resembles the results for equilibrium states with higher B. Nonequilibrium steady states under phonon temperatures, see Fig. 6(b). driving One interesting feature in the nonequilibrium SC be- fore melting is that there emerges a dip in A¯(ω) and a In this section, we study the effects of the phonon driv- hump in f¯el(ω) at ω = Ω/2. These features are only ing on SC in the strong electron-phonon coupling regime pronounced in the NESS with SC and disappear in the by studying the nonequilibrium steady states. In the fol- normal state. They can be explained as follows as a con- lowing we use, as a representative set of parameters for sequence of the two-band structure in the SC phase. The the strong coupling regime, g = 0.41, ω0 = 0.4, β = 120 hump in f¯el(ω) results from excitations from the lower and the bath parameters γph = 0.02, ωD = 0.6, γel = band of the Bogoliubov quasiparticles ( Ek) to the up- 0.005,W = 2.0. This gives λ = 1.08, the renormal- − Ω ebath per band (Ek), which become efficient at Ek = 2 , where ized phonon frequency ωr 0.2, and the inverse tran- E is the energy of the quasiparticles. Although Bogoli- ∼ k sition temperature βc = 55. γel is chosen to be much ubov excitations can recombine or relax within the bands smaller than the SC gap in order not to break the SC or- due to the scattering with phonons and the coupling to der, while γph is chosen so that the damping of phonons the environment, the continuous creation of particles at (the phonon peak width in the spectrum) mainly origi- Ω the particular energy Ek = 2 can lead to an enhanced nates from the electron-phonon interaction. We have also distribution in the steady state at the corresponding en- considered stronger g, different temperatures and both ergy. The dip in the spectral function can be explained as stronger and weaker γph and γel, and confirmed that the a hybridization effect between the first Floquet sideband behavior discussed below does not change qualitatively. of the upper (lower) Bogoliubov band with the lower (up- ¯ R Ω First, in Fig. 5, we show ReD (ω), which includes per) Bogoliubov band, which occurs at Ek = 2 . The the effects of Ω and ∆ω0 as well as the renormalization hybridization opens a gap at the crossing point, which 12

1.0 1.0 Eq ∆ω0 =0.05 ∆ω0 =0.03 ∆ω =0.08 0.8 0 0.8 ∆ω0 =0.04 ) ) 0.6 0.6 ω ω ( ( l e ¯ A

0.4 ¯ f 0.4 Ω = 0.2 0.125 0.4 0.2 0.15 0.8 (a) 0.2 0.0 0.1 0.0 0.1 0.0 ω 0.2 0.1 0.0ω 0.1 0.2 1.0

0.8 FIG. 7: Time-averaged distribution functions (solid lines) for g = 0.41, ω0 = 0.4, β = 120 and the bath parameters γph = ) 0.6 ω 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0 and ∆ω0 = 0.04. ( l e Dashed lines show the corresponding equilibrium distribution ¯ f 0.4 functions with the effective temperatures estimated from the 0.2 slope ∂ωf¯el(ω)|ω=0. (b) 0.0 0.1 0.0 0.1 ω librium distribution for ω > 0. Therefore, it is expected

(c) 1.0 that superconductivity will be weaker than what is ex- 12 pected from this effective temperature, which is indeed ) ω

( the case, as shown below. h 0.5 p ) ¯ 8 f

ω In Fig. 8, we summarize our representative Floquet ( 0.0 ¯ B 0.0 0.1 0.2 DMFT results in the strong el-ph coupling regime for ω 4 quantities time-averaged over one period. The SC or- der parameter is generally reduced from the equilibrium 0 value (in fact, despite an extensive search, we could not 0.0 0.2 ω 0.4 find any parameter set where the SC order is enhanced in the strong-coupling regime, λ 1). In particular, around Ω = 0.4, which is almost twice∼ the renormalized phonon FIG. 6: Time-averaged spectrum and distribution functions frequency, the superconducting order vanishes even for for g = 0.41, ω0 = 0.4, β = 120 and the bath parameters phonon modulations with small amplitude ∆ω . This γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0 and the ex- 0 citation frequency Ω = 0.125. (a) Time-averaged electron is consistent with the amount of energy injected by the spectrum. (b) Time-averaged electron distribution function. excitation, and the effective temperature estimated from (c) Time-averaged phonon spectrum. The inset shows the the nonthermal distribution function ∂ωf¯el(ω) ω=0, see time-averaged phonon distribution function. Fig. 8(b). We note that the peak position in Fig.| 8(b) slightly shifts to higher energy as we increase the ex- citation strength. This is because the injected energy is not fully developed in our case because of the finite changes the renormalization of the phonons, which af- correlations, resulting in a dip in the spectrum. fects the value of the parametric resonance ( 2ωr). We The phonon spectra, which are defined in the same also note that the SC order parameter is always∼ lower manner as the electron spectra using D, show a hard- than the equilibrium value at Teff. This is consistent with ening of the renormalized mode as we increase the field the fact that the equilibrium distribution function at Teff strength [Fig. 6(c)], again consistent with the behavior underestimates the occupancy of the states at ω > 0. 23 seen in equilibrium with increased temperature. At the Let us briefly comment on the effect of stronger cou- same time, the phonon distribution function defined as plings to heat baths, since one expects in this case f¯ (ω) 1 ImD¯ <(ω)/B¯(ω) shows an enhancement of ph ≡ − 2π more efficient energy absorption. We have checked that the phonon occupation [see the inset in Fig. 6(c)]. When stronger couplings to the phonon bath lead to a broader we further increase the excitation strength, the Floquet phonon peak in the spectrum, which results in a stronger sidebands of the phonons become more prominent [see suppression of SC and higher effective temperatures at ∆ω = 0.08 data in Fig. 6(c)], in which case the spectrum Ω . ωr. Stronger couplings to the electron bath gener- does not resemble that at elevated temperatures. ally lead to a lower effective temperature and a less dra- Next, let us discuss how the nonthermal distribution matic suppression of SC. However, the equilibrium value is different from a thermal one. In Fig. 7, we show the of the SC order becomes smaller and in any case we only comparison between the nonthermal distribution and the find a suppression of SC as far as we have investigated. thermal distribution with an effective temperature de- So far we have focused on the time-averaged val- rived from βeff = 4∂ωf¯el(ω) ω=0. We can see that this ues, whereas in the NESS all the quantities are oscil- thermal distribution− always underestimates| the nonequi- lating around the averaged value during one period. In 13

more 0.74 0.70 less attractive attractive

) 0.76 Eq ) t t

( FDMF:Ave ( n n i i 0.75 k ∆ω0 =0.03 k E 0.78 ∆ω0 =0.04 E ∆ω0 =0.05 (a) Ω =0.125 (b) Ω =0.2 0.80 0.00 0.04 0.08 0.00 0.05 0.10 Eint(t) Eint(t) 0.65 0.70

) 0.70 ) t t ( ( n n i i 0.75 k k

E 0.75 E

(c) Ω =0.4 (d) Ω =0.8 0.80 0.80 0.00 0.05 0.10 0.05 0.10 Eint(t) Eint(t)

FIG. 9: Trajectories of Ekin(t) and Eint(t) = EnX(t) + Eph(t) in the NESS for g = 0.41, ω0 = 0.4, β = 120 and the bath parameters γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0. Solid loops indicate that the NESS is in the SC state, while dashed loops are for the normal state. In addition, we show the time-averaged values E¯kin and E¯int of the NESS for dif- ferent excitation strength (violet curves), and the equilib- rium Ekin and Eint for different temperatures (black lines). π 3π Black open circles indicate the values at t = 2Ω , 2Ω , where the Hamiltonian temporarily becomes the same as the non- perturbed one for ∆ω = 0.04.

FIG. 8: Time-averaged results for g = 0.41, ω0 = 0.4, β = In Fig. 10, we show the optical conductivity in the 120 and the bath parameters γph = 0.02, ωD = 0.6, γel = NESS. In equilibrium, for T < Tc, the real part shows a 0.005,Webath = 2.0. (a) Dependence of the SC order param- reduction of the spectral weight below the SC gap and eter φsc,av on Ω for some ∆ω0. The open symbols show the the imaginary part exhibits a 1/ω dependence, whose results estimated from the effective temperature. In the pink coefficient corresponds to the superfluid density. In the ¯ R (blue) background region, |D (0)| increases (decreases) when NESS, the optical conductivity σ(t, t0), like any correla- ∆ω0 changes from 0 to 0.02. (b) The dependence of Etot,av tion function, can be written in the Floquet matrix form on Ω for some ∆ω0. The inset is the effective temperature Eq.(22), see the discussion around Eq. (51). In the diag- estimated from the slope ∂ωf¯el(ω)|ω=0. The vertical lines in- dicate the positions of the parametric resonance in the weakly onal component in the NESSs, the weight of the real part inside the SC gap increases with increasing field strength, driven limit (Ω = ωr, 2ωr). and becomes close to the result at elevated temperatures [see Fig. 10(a)]. Meanwhile, the 1/ω component in the Fig. 9, we plot the trajectories of Ekin(t) and Eint(t) = imaginary part is suppressed, which indicates the sup- EnX(t) + Eph(t) for the NESS over one period, and com- pression of the superfluid density and is consistent with ¯ ¯ pare them to the averaged Ekin and Eint of the NESS the closure of the gap and the decrease of the SC order (violet lines) and the equilibrium Ekin and Eint for dif- parameter [see Fig. 10(b)]. In the off-diagonal compo- ferent temperatures (black lines). First, we can see that nent, one can also observe a 1/ω around ω = 0 behavior the results for the time averaged values of Ekin and Eint both in the real and imaginary parts in the SC phase [see ∗ are in general not on top of the equilibrium curve, and Fig. 10(c)(d)]. (We note that σm,n(ω) = σ−m,−n( ω).) that the deviation strongly depends on the excitation fre- This is caused from the imperfect cancellation between− quency. This indicates that it is difficult to fully describe χpara and χdia, and corresponds to the following asymp- the properties of NESSs through an analogy to equilib- totic form for t t0 of the conductivity, rium states with elevated temperatures. We also notice | − | → ∞ 0 0 X −imΩt that the oscillations around the averaged values are large, σS(t, t ) θ(t t ) nS,me . (52) and that they become largest at the parametric reso- ≡ − m nance, Ω = 0.4 2ωr. Even at the times where the ' ∗ Hamiltonian becomes temporarily identical to the unper- We note that nS,m can be a complex number and nS,m = π π 3 nS,−m. This indicates that if a delta-function electric- turbed one (t = 2Ω , 2Ω ) the energies can strongly deviate from the thermal line. This indicates that the dynamics field pulse is applied to the system, the induced current which occurs during one period cannot be captured using continues to flow forever (supercurrent) oscillating with an adiabatic picture either in these cases. a period of 2π/Ω, and the phase of nS,m determines the 14

(a) (b) (a) Ω =0.8 (b) Ω =0.4 ) ) . ∆ω =0.03 Eq:β =120 . 0.04 ∆ω = 0.04

U 0.02 0.02 0 U 0 . . 0.04 −0.04 A ω . ∆ =0 04 Eq:β =50 A 0 ) ) ( ( t t

−

( 0.08 0.08 ( ) ∆ω =0.05 ) 0 0 C − C ω ω S S ( ( φ φ 0

0 0.02 0.02 0 0 σ σ e m R I 0.00 0.00 0 0 20 40 60 80 0 20 40 60 80 0.1 0.0ω 0.1 0.1 0.0ω 0.1 t t (c) (d) (c) Ω =0.125 (d) Ω =0.04 ) ) . . Eq:Reσ00(ω) 0.04 0.04 U Eq:Reσ (ω) U .

00 . A A ) ) ( ( t t

( ( ) ) 0 C C ω ω S S ( ( φ φ 0

0 0.02 0.02 1 1 σ σ e m R 0 I 0.00 0.00 0 20 40 60 80 0 20 40 60 80 0.1 0.0ω 0.1 0.1 0.0ω 0.1 t t

FIG. 10: Optical conductivity (σm0(ω)) for g = 0.41, ω0 = FIG. 11: Time evolution of the SC order parameter under the 0.4, β = 120 and the bath parameters γph = 0.02, ωD = periodic driving of type 1 for various frequencies and ampli- 0.6, γel = 0.005,Webath = 2.0 and the excitation Ω = 0.125. tudes for g = 0.41, ω0 = 0.4, β = 120 and the bath parameters ∆ω0 = 0.03, 0.04 and the equilibrium state at β = 120 are in γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0. Horizontal SC, and the rest are in the normal phase. (a)(b) The diagonal black dotted lines indicate the equilibrium value, while the component of the optical conductivity, σ00(ω). (c)(d) The off- blue and red vertical lines show t = π/Ω and t = 2π/Ω, re- diagonal component of the optical conductivity, σ10(ω). The spectively. The arrow in panel (c) indicates the value at the dotted line and the dashed and dot-dashed lines show the NESS. equilibrium σ00(ω) with indicated temperatures for compar- ison. The red allows in (c) point to the 1/ω divergence for should increase. However, in the initial dynamics for ∆ω0 = 0.03. At ω = 0 in the SC states, the delta function is located at each m, which is not shown here. non-adiabatic excitations, the solid (dashed) lines show a positive (negative) hump, which is opposite to the above expectation. The hump becomes smaller with decreasing phase of the oscillations in the suppercurrent. The 1/ω Ω, and the dynamics becomes more consistent with the component in σm0(ω) corresponds to nS,m. For example, expectations from the adiabatic picture. We also note without the phonon driving nS, 1.18, while for ∆ω = 0 ≈ 0 that, during the first period, one can see a significant 0.03 nS, 0.62 and nS, 0.05 i0.01. 0 ≈ 1 ≈ − transient enhancement of the order parameter in some cases, see Fig. 11. There the time-averaged φSC(t) for one period can become larger than the equilibrium value C. Transient dynamics towards the NESS if we carefully choose the phase of the excitation, i.e. the sign of ∆ω0.

In order to study the transient dynamics towards the As time evolves further, φSC(t) is however strongly NESS, we consider a situation where the system at time suppressed from the initial value. One can see that the t = 0 is in equilibrium (attached to the bath) and at fastest suppression occurs around Ω = 0.4 2ωr, which t > 0 is exposed to a phonon driving of type 1 with is at the parametric resonance. This is consistent∼ with the analysis of the NESS with the Floquet DMFT, where ω0X(t) = ω0 + ∆ω0 sin(Ωt). (53) we have observed the strongest energy injection and the strongest increase of the effective temperature in this In order to follow the transient dynamics, we use the driving regime. The results in Fig. 11 show that even nonequilibrium DMFT formulated on the Kadanoff- 22–24 in the transient dynamics, the parametric resonance is Baym contour. For this, we first need to evaluate associated with a rapid destruction of the SC order. the bare phonon propagator subject to the periodic driv- ing by solving Eq. (27). Apart from that, the calculation In Fig. 12, we show how the energies approach the is identical to the scheme explained in Refs. 23,24. NESS described by the Floquet DMFT. These results In Fig. 11, we show the transient dynamics of the demonstrate that it takes several cycles for the system to reach the NESS. SC order parameter, φSC(t), for various excitation con- ditions. First we note that when ω0X becomes small To gain more insights into the dynamics of the order with g and T (< Tc) fixed, φSC becomes larger because parameter, we compare it with the linear-response result. the system moves to the stronger coupling regime. This In the linear-response regime, the variation of the order leads to the expectation that in the adiabatic regime, parameter can be expressed as φSC becomes larger with decreasing ω0X. In other words, Z for ∆ω0 > 0 (solid lines in Fig. 11) φSC should ini- ¯ 2 ¯ ¯ ∆B0(t) = dtχB0,X (t t)F (t), (54) tially decrease, while for ∆ω0 < 0 (dashed lines) it − 15

0.5 Ω =0.8 (a) (b) (a) + +:FDMFT 0.04 0.04 0.6 0.2 ) ) ) :FDMFT ) t t t t ( ( ( − − ( n C C X i S S k n ∆ω0 = 0.7 0.3 φ 0.02 φ 0.02 E E 0.04 0.04 0.8 0.4 − (b) Ω =0.4 0.00 0.00 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 t t t t 0.5 Ω =0.125 Ω =0.04 (c) (d) (c) (d) 0.5 0.04 0.04 0.6 ) ) ) ) t t t t ( ( ( ( 0.4 t C C h o S S p t

0.7 φ 0.02 φ 0.02 E E 0.3 0.8 0.2 0.00 0.00 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 t t t t

FIG. 12: Time evolution of the energies for the type 1 excita- FIG. 14: Comparison between the full dynamics of the SC or- tion with Ω = 0.4 and |∆ω0| = 0.04 for g = 0.41, ω0 = 0.4, β = der parameter (solid lines) and the prediction from the linear- 120 and the bath parameters γph = 0.02, ωD = 0.6, γel = response theory (dashed lines) for g = 0.41, ω0 = 0.4, β = 120 0.005,Webath = 2.0. The sign of ∆ω0 is indicated in the fig- with the bath γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0. ures. Horizontal black dotted lines indicate the equilibrium The sign of ∆ω0 is indicated in the figures. Horizontal black value, while the blue and red vertical lines show t = π/Ω and dotted lines indicate the equilibrium values, while the blue t = 2π/Ω, respectively. and red vertical lines show t = π/Ω and t = 2π/Ω, re- spectively. Dash-dotted curves show the results of the adi- abatic approximation for the time-dependent Hamiltonian at β = 120. 0.02

) † † t 0.00 2 P

( 2 ∂t χB0,X (t) t=0 = 8igω0 i ( c c + ci↓ci↑)Xi . If we 2 i↑ i↓

X | h − i ,

0 0.02 decrease Ω with ∆ω0 fixed (as we approach the adiabatic B

χ regime), the contribution from the initial positive hump 0.04 2 in χB0,X (t) decreases, and the hump in φSC becomes 0.06 smaller. At longer times, φSC(t) starts to deviate from 0 20 40 60 80 100 the linear-response result, which generally overestimates t φSC(t). This indicates that the general effect of the non-linear component is to reduce the superconducting order. It is FIG. 13: χ 2 (t) for g = 0.41, ω0 = 0.4, β = 120 and the B0,X also illustrative to compare the results to the expecta- bath parameters γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0. tion in the adiabatic limit, in which the temperature is fixed to the bath temperature and the phonon potential 2 is changed (see the dash-dotted curves in Fig. 14). (Note with χ 2 (t) iθ(t) [B (t),X (0)] , B = B0,X 0 0 that in an isolated system, the adiabatic limit would be P † † ≡ − h i i ci↑ci↓ + H.c., and F (t) = ∆ω0 sin(Ωt). The linear- defined by constant entropy instead of constant temper- response function can be numerically evaluated by con- ature.) The comparison confirms that the dynamics ap- sidering, instead of the periodic driving, Hex(t) = proaches the adiabatic expectation as we decrease the P 2 df δ(t) i Xi and taking df small enough. The result- excitation frequency. ing response function is shown in Fig. 13. The oscillation 62 frequency in χ 2 (t) is ω 0.4 2ω . . B0,X ∼ ∼ r In Fig. 14, we compare the full dynamics (solid lines) D. Superconducting fluctuations above Tc under and the linear-response component (dashed lines), which phonon driving is obtained by taking the convolution, Eq. (54). The dif- ference corresponds to the contributions from the non- So far we have focused on the superconducting phase in linear components. The linear-response theory captures the strong-coupling regime and showed that the phonon the initial behavior of the order parameter including the driving generally weakens the superconductivity and that temporal enhancement or suppression. The initial hump this tendency is particularly pronounced in the paramet- whose direction is opposite to the adiabatic expectation ric resonance regime. Now we move on to the weak- is also captured. This hump originates from the posi- coupling regime (λ 1), where the BCS theory has usu-  2 tive value of χB0,X (t) just after t = 0. Since the first ally been applied. In this regime the transition tempera- 2 derivative of χB0,X (t) at t = 0 is zero, as can be ana- ture becomes very small and a direct investigation of the lytically shown, this corresponds to the positive value of superconducting state is numerically difficult. Hence we 16

-1 -1 1.0 10 0.26 0.34 10 (a)∆ω = 0.15 (b)∆ω = 0.06 Ω = 0 − 0 − 0.28 0.36 -2 0.4 0.8 0.3 0.38 10 0.28 0.34 0.55 0.7 0.3 0.36 0.6 0.75 ) 0.32 0.4 ) 0.3 t t 0.6 0 0

-3 o o , , 10 t 0.32 t 0.65 t t E E ( ( 0.2

r r Ω = i i 0.4 a a -4 ∆ ∆ p p 10 0.5 0.65 0.1 χ χ 0.7 -2 0.55 0.2 10 10-5 0.6 0.75 0.0 (a) ∆ω0 = 0.06 (b) ∆ω0 = 0.06 0.0 − 10-6 − 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 t t t t 10-1 10-1 10-2 10-2 FIG. 16: Time evolution of the total energy of the system

) ) Ω = under the periodic driving of type 1. The parameters are g =

0 -3 0 -3 , 10 Ω = , 10 0.5 0.65 t t

( ( 0.7 0.3, ω0 = 0.4, β = 320 and no heat bath is attached. Different

r r 0.55 i 0.26 0.34 i a -4 a -4 0.6 0.75 p 10 0.28 0.36 p 10 colors indicate different Ω. Panel (a) shows the results around χ 0.3 0.38 χ 10-5 0.32 0.4 10-5 Ω = ωr, while panel (b) is for the driving frequencies around (c)∆ω0 = 0.15 (d)∆ω0 = 0.15 Ω = 2ωr. The data at the resonance are shown by bold lines. 10-6 − 10-6 − 0 20 40 60 80 0 20 40 60 80 t t 10-1 10-1 temperature (down to β = 640), g and the couplings ∆ω0 = ∆ω0 = 0.02 0.06 0.02 0.1 to the heat baths, we have confirmed that the behavior 0.02 0−.1 0.02 0.1

) 0−.06 0.1 ) 0−.06 0−.14 described below is the generic one in the weak-coupling 0 0 , − , 0.06 0.14 t t

( ( − − r r regime. In Fig. 15(a)(c), we show the results around i i a a p p

χ χ Ω = ωr. We remind the reader that from the analy- R sis of D0 in Sec. IV A, an strong enhancement of D (0) (e) Ω =0.16 (f) Ω =0.16 10-2 10-2 is expected for Ω ωr. However, what we observe is 0 20 40 60 80 0 20 40 60 80 . t t that generally in the presence of periodic phonon mod- ulations the lifetime of the fluctuations becomes shorter FIG. 15: Transient dynamics of the SC fluctuations above than without phonon modulations and that at Ω = ωr the Tc under the periodic driving of type 1, χpair(t, 0). The pa- lifetime is particularly strongly reduced. This tendency rameters are g = 0.3, ω0 = 0.4, β = 320 and no heat bath becomes clearer with increasing ∆ω0. In Fig. 16(a), we is attached. Panels (a)(c) are for driving frequencies around show the evolution of the energy absorption around the Ω = ωr, while (b)(d) are for frequencies around Ω = 2ωr. The parametric resonance, Ω = ωr. It turns out that at the data at the resonance are shown by bold lines. (e)(f) show the parametric resonance the energy absorption is also very results for Ω = 0.16 and various ∆ω0. Dotted lines are the efficient. These results suggest that even though there linear component (in ∆ω0) estimated from the ∆ω0 = ±0.02 might be some enhancement of the attractive interac- data. The dash-dotted black line in each panel is the equilib- tion, the energy absorption and resulting disturbance of rium result. long-range correlations is the dominant effect in a wide parameter range. In particular, at the parametric res- study the time evolution of the superconducting fluctu- onance, the maximized energy absorption rate leads to ations in the normal state, in systems with and without a particularly rapid heating of the system, and efficient phonon driving. Specifically, we evaluate the dynami- destruction of the SC fluctuations. We also note that cal pair susceptibility under the excitation described by the decay of the fluctuations accelerates as t increases, Eq. (53). For this, as in the equilibrium case,24 we add as is evidenced by the concave form of χR (t, 0) in the 0 pair Hex(t) = Fex(t)B0 on top of the phonon modulation, log scale. This can also be interpreted as a heating ef- P † † where B0 = i(ci↑ci↓ + ci↓ci↑) and Fex(t) = df δ(t). We fect. The more energy the system gains as t increases, take df small enough and compute the time evolution of the faster the decay of the fluctuation becomes. B0 to obtain the dynamical pair susceptibility, This dominant heating effect and quick destruction of the SC fluctuation can also be observed in another res- R 0 0 1 0 χpair(t, t ) = iθ(t t ) [B0(t),B0(t )] . (55) onant regime Ω 2ωr, where in the polaron picture an − − N h i enhancement of the∼ superconductivity is expected,11 see Fig. 15(b)(d). Here, Ω = 0.65 2ω shows the faster In the normal state in equilibrium, this susceptibility de- ∼ r cays as t t0 increases. The decay time increases as we decay than the neighboring driving frequencies. The effi- approach− the boundary to the superconducting state and cient absorption of energy resulting from this parametric diverges at the boundary. The interesting question thus phonon driving is shown in Fig. 16(b). is how this decay depends on the phonon driving in the Now, we point out that there is a regime where the weak-coupling regime. phonon driving slightly enhances the SC fluctuations. In In Fig. 15, we show the results for g = 0.3, ω0 = Fig. 15(e)(f), we show the SC fluctuations for a smaller 0.4, β = 320 without heat baths. This condition cor- excitation frequency than those above, where less heat- responds to λ 0.23 and ωr = 0.32, and the inverse ing is expected, and for various excitation strengths. In ∼ R transition temperature is βc 1000. By changing the all cases, χ (t, 0) stays around the equilibrium curve, ∼ pair 17 but exhibits a shift above or below the equilibrium value presence of the renormalization from the electron-phonon depending on the sign of ∆ω0. For small amplitudes, coupling, the effective attractive interaction character- R ¯ R the change of χpair(t, 0) with respect to the equilibrium ized by D (ω = 0) shows an enhancement in some pa- R rameter regime. However, taking into account heating value χpair,eq(t) is linear in ∆ω0. In order to iden- tify the component linear in ∆ω0, we rescale the differ- and the nonthermal nature of the driven state, the ulti- ence of χR (t, 0) χR (t) at the smallest amplitude mate effect of the driving turns out to be a suppression pair − pair,eq ∆ω0 = 0.02 to larger amplitudes, and depict the results of the SC gap, the SC order parameter or the superfluid with dashed lines in Fig. 15(e). The linear components density both in the time-periodic steady states and the consistently explain the general behavior. The enhance- transient dynamics. The strongest suppression of SC is ment of the SC fluctuations for ∆ω < 0 can be ex- observed in the parametric resonant regime, Ω 2ωr. 0 ∼ plained as follows. At t [0, π/Ω] [0, 20], the phonon frequency is temporarily∈ reduced and∼ the system is tem- In the weak-coupling regime, we have studied the SC porarily in a stronger coupling regime compared to equi- fluctuations above Tc. We have found that, generally, librium (this is not the effect of the Floquet sidebands). they are suppressed by the phonon driving, in the sense Hence the decay of the SC fluctuations is slower in this that the pairing susceptibility decays more rapidly in the time interval. During t [π/Ω, 2π/Ω] the situation is op- driven state. The decay becomes particularly fast in the posite and we observe a∈ faster decay. However, because parametric resonant regimes, which can be attributed to R of the enhancement during t [0, π/Ω], χpair(t, 0) stays a particularly efficient energy absorption. above the equilibrium curve.∈ The opposite is the case when ∆ω0 is positive. From these analyses, we conclude that the previously The non-linear component can result in a slight but predicted enhancement of the superconductivity in the systematic increase of the SC fluctuations, in particular parametric resonant regime is not a general result. In at ∆ω0 = 0.06, see the difference between the solid the simple Holstein model with a phonon energy scale ± and dotted lines in Fig. 15(e). We note that both the smaller than that of electrons, such effects, if any, are heating and the creation of the Floquet sidebands of the overwhelmed by the absorption of energy in a wide pa- phonons, which leads to the enhancement of the attrac- rameter range, and particularly strongly so near the para- tive interaction, start from the second order in the field metric resonances. Even though our analysis focuses on a strength. Therefore, this observation may indicate that specific model, it demonstrates the importance of a care- the enhancement of the attractive interaction slightly ful treatment of heating effects and/or nonthermal distri- dominates the heating effect. However this enhancement butions when discussing the potential effects of phonon is very subtle and as we further increase the field strength driving. We note that in addition to the coupling of light this enhancement becomes weak and eventually the SC to phonons, one could also take into account the direct fluctuations decay faster than in the equilibrium case, see coupling of light to electrons, by means of a Peierls sub- Fig. 15(f). stitution. This can be expected to lead to further heating Finally we note that even though we have also studied effects. a parameter set for ω0 > W in the weak coupling regime with various excitation conditions, we could not observe As for the relation to the K3C60 experiment mentioned 63 any enhancement of the SC fluctuation. in the introduction,8 from our analysis it seems likely that the apparent enhancement of the SC originates from the static change of the Hamiltonian14,18 or some other V. CONCLUSIONS dynamical effects13,16,55, which are not captured in the present simple model. Since alkali doped fullerides are We have systematically investigated the effects of multi-orbital systems and an effective negative Hund’s the parametric phonon driving on the superconductivity coupling from the Jahn-Teller screening plays an impor- (SC) using the simple Holstein model and the nonequi- tant role,56–59 we may need to go beyond the Holstein librium dynamical mean-field theory (DMFT). In order model to fully understand the mechanism. to study the time-periodic nonequilibrium steady state, we extended the Floquet DMFT formalism, which has Finally, we remark that the diagrammatic method for previously been applied to purely electronic systems, to steady states and the transient dynamics used here can this electron-phonon model. We have also studied the be easily extended to other scenarios for the enhance- transient dynamics under periodic phonon driving with ment of SC, such as non-linear couplings to electrons,13,16 the nonequilibrium DMFT formulated on the Kadanoff- which is an interesting direction for future studies. The Baym contour. This simulation method, which does not investigation of nonequilibrium states with an unconven- involve a gradient approximation, can describe dynamics tional pairing such as the d-wave superconductivity in on arbitrarily fast timescales. cuprates is also an interesting future work. In particular In the strong electron-phonon coupling regime, we have it is important to understand whether or not a particu- studied both the nonequilibrium steady states and the lar mechanism for the enhancement of SC survives even transient dynamics towards these states. Even in the if heating effects are taken into account. 18

Acknowledgments tendency becomes strong at Ω 2ωr and/or at Ω ωr as in the case of the type 1 excitation.' ' The authors wish to thank D. Goleˇz,H. Strand and J. Okamoto for fruitful discussions. YM is supported by the Swiss National Science Foundation through NCCR Mar- 1. Nonequilibrium steady states vel. PW acknowledges support from FP7 ERC Starting Grant No. 278023. NT is supported by JSPS KAKENHI Here we show the results for the nonequilibrium steady (Grant No. 16K17729). ME acknowledges support by states for the type 2 excitation in the strong electron- the Deutsche Forschungsgemeinschaft within the Sonder- phonon coupling regime as in Sec. IV B. Even in this forschungsbereich 925 (projects B4). regime, the type 2 excitation shows an increase of the ef- fective attractive interaction characterized by g2D¯ R(0) | | at Ω . ωr, and a decrease at Ω & ωr. In Fig. 17, we show Appendix A: The expression for the energy the time-averaged spectra and distribution functions for dissipation in the time-periodic NESS the electrons and phonons for Ω = 0.125, where we ob- served an increase of the attractive interaction. The gen- The explicit expression for Eq. (39) is eral features are the same as in the case of the type 1 ex- citation (Fig. 6). The SC gap is suppressed with increas- I = I¯el + I¯ph ing driving amplitude and is completely wiped out at ∆ω0 0.05, see Fig. 17(a). The slope of the electron dis- = i [Hph mix(t),Hph bath(t)] + i [Hel mix(t),Hel bath(t)] . ≥ h i h (A1) i tribution functions [Fig. 17(b)] at ω = 0 decreases com- pared to the equilibrium case, which resembles the effect Using the bath’s fluctuation-dissipation relation, we ob- of heating. Precisely speaking, the equilibrium distribu- tain tion functions at the effective temperature underestimate the number of the occupied states above the Fermi level. Z Ω/2 The characteristic features in the NESS with SC are the ¯ X dω X Iel = i (ω + nΩ)Γ(ω + nΩ) kink in A¯(ω) and the hump in f¯(ω) at ω = Ω/2. The − −Ω/2 2π i,σ n renormalized phonon frequency increases as we increase  K R A G (ω) Ff (ω + nΩ)(G (ω) G (ω)) , the strength of the excitation, which is naturally expected i,σ,nn − i,σ,nn − i,σ,nn (A2) from the equilibrium state with elevated temperatures, see Fig. 17(c). When we further increase the excitation βω with Ff (ω) = tanh( ) and strength, the Floquet sidebands of the phonons become 2 more prominent [see ∆ω = 0.08 data in Fig. 17(c)], in Z Ω/2 which case the spectrum does not resemble that of the ¯ i X dω X Iph = (ω + nΩ)Bbath(ω + nΩ) equilibrium state with elevated temperatures. 2 2π i −Ω/2 n  K R A Di,nn(ω) Fb(ω + nΩ)(Di,nn(ω) Di,nn(ω)) , 1.0 1.0 Eq ∆ω0 =0.05 (b) − − ∆ω =0.03 (A3) 0.8 0 ∆ω0 =0.08 0.8 ∆ω0 =0.04 ) ) 0.6 0.6 ω ω ( ( βω l e ¯ A with F (ω) = coth( ). f b 2 0.4 ¯ 0.4 These expressions imply that the energy dissipation 0.2 0.2 ¯ (a) I is related to the violation of the system’s fluctuation- 0.0 0.0 dissipation relation in the time-periodic NESS. This is 0.1 0.0ω 0.1 0.1 0.0ω 0.1 the Floquet generalization of the so-called Harada-Sasa 1.0 (c) (d) relation.60 We have confirmed that this expression is in- 12

deed satisfied within the Floquet DMFT implemented ) )

8 ω ω ( ( with the self-consistent Migdal approximation. h 0.5 p ¯ B f ¯ 4

0 0.0 Appendix B: Results for the type 2 excitation 0.0 0.2 ω 0.4 0.0 0.1ω 0.2

In this section, we show the results for the the type 2 FIG. 17: Time-averaged spectra and distribution functions for excitation [Eq. (5)], whose effect has been discussed in the type 2 excitation for g = 0.41, ω0 = 0.4, β = 120 and the Ref. 12. It turns out that the general behavior is very bath parameters γph = 0.02, ωD = 0.6, γel = 0.005,Webath = similar to the results for the type 1 excitation [Eq. (4)] 2.0 and the excitation Ω = 0.125. (a) Time-averaged electron shown in the main text. The ultimate effect of the type spectrum. (b) Time-averaged electron distribution function. 2 excitation is to generally suppress the superconductiv- (c) Time-averaged phonon spectrum. (d) Time-averaged ity both in the strong- and weak-coupling regimes. This phonon distribution function. 19

more (b) Ω =0.4 less attractive ∆ω0 = 0.02 0.02 attractive 0.04 0.04 −0.04 0.04 0.08 −0.08 ) )

t − t ( ( C C S S

φ 0.02 φ 0.02

(a) Ω =0.8 0.00 0.00 0 20 40 60 80 0 20 40 60 80 t t (c) Ω =0.125 (d) Ω =0.04 0.04 0.04 ) ) t t ( ( C C S S

φ 0.02 φ 0.02

0.00 0.00 0 20 40 60 80 0 20 40 60 80 t t

FIG. 19: Time evolution of the SC order parameter un- der the periodic driving of type 2 for various conditions for g = 0.41, ω0 = 0.4, β = 120 with the bath γph = 0.02, ωD = 0.6, γel = 0.005,Webath = 2.0. Horizontal black dotted lines indicate the equilibrium value, while the blue and red vertical lines show t = π/Ω and t = 2π/Ω.

2. Transient dynamics towards the NESS

In Fig. 19, we show the transient dynamics towards the NESS under the type 2 parametric phonon driving. It corresponds to Fig. 11 in the main part. One again finds that (i) the decease of the order parameter is fastest FIG. 18: Time-averaged results for the type 2 excitation around Ω = 0.4 2ωr, (ii) the initial hump goes in the for g = 0.41, ω0 = 0.4, β = 120 and the bath parameters opposite direction' from the adiabatic excitation, and be- γ = 0.02, ω = 0.6, γ = 0.005,W = 2.0. (a) Depen- ph D el ebath comes less prominent as we approach the adiabatic limit, dence of φsc,av on Ω for some ∆ω0. The open symbols show the results estimated from the effective temperature. In the and (iii) depending on the phase of the excitation, there red (blue) background region, |D¯ R(0)| increases (decreases) can be an enhancement of the SC order parameter during when ∆ω0 changes from 0 to 0.02. (b) Dependence of Etot,av the first cycle. Quantitatively speaking, the speed of the on Ω for some ∆ω0. The inset is the effective temperature melting of SC is slightly slower than in the type 1 case at estimated from the slope ∂ωf¯el(ω)|ω=0. The vertical lines in- the same set of the parameters. This is consistent with dicate the position of the parametric resonance in the weakly the results for the NESS, where the order parameter is driven limit (Ω = ωr, 2ωr). more robust compared to the type 1 and the effective temperature is lower than the type 1 case. The summary of the results for the type 2 excitation is shown in Fig. 18. Generally, in the presence of the phonon driving the SC order is weakened, and this ten- 3. Superconducting fluctuations above Tc dency becomes strong around Ω = 2ωr as in the type 1 case, see Fig. 18 (a). Again the order parameter evalu- Here we show the SC fluctuations above Tc in the weak ated from the effective temperature overestimates its size coupling regime and under the type 2 driving. Figure 20 in the NESS, but its general behavior can be qualitatively shows the results corresponding to Fig. 15 in the main reproduced. The (time-averaged) total energy and the ef- text. First we can again see that generally the decay fective temperature shows a peak around Ω = 2ωr, see of the fluctuations in the driven system is faster than Fig. 18(b). The peak position is shifted to higher fre- in the absence of phonon driving, which indicates that quency as the excitation strength is increased, which is the driving does not enhance the superconductivity, see attributed to the hardening of the renormalized phonon Fig. 20 (a-d). As in the type 1 case, around Ω = ωr, frequency as in the type 1 excitation. Quantitatively, un- the decay becomes particularly fastest at Ω = ωr, even der the type 2 excitation the SC order is more robust than though at ∆Ω = 0.06 the tendency is not so clear. Near − under the type 1 excitation, which is consistent with the Ω = 2ωr, there is again a point where the decay becomes lower energy and the lower effective temperatures (com- particularly fastest. We also note the concave form of pare Figs. 8 and 18). the χpair(t, 0) curve in the log scale, which is explained by the continuous heating of the system. 20

10-1 10-1 In Fig. 20 (e)(f), we show the results for an excita- Ω = 0.26 0.34 0.28 0.36 -2 0.3 0.38 10 tion frequency small compared to ωr. With the weak to

) 0.32 0.4 )

0 0 -3 moderate excitation strength, depending on the phase of , , 10 t t ( (

r r Ω = i i ∆ω0, χpair(t, 0) is shifted above or below the equilibrium

a a -4

p p 10 0.5 0.65

χ χ 0.7 -2 0.55 curve, which is consistently explained by the linear com- 10 -5 0.6 0.75 10 ponent in ∆ω0 (the dashed lines). The non-linear compo- (a) ∆ω0 = 0.06 (b) ∆ω0 = 0.06 − 10-6 − nent slightly but systematically shifts the SC fluctuations 0 20 40 60 80 0 20 40 60 80 t t above the linear component, which indicates that there 10-1 10-1 is more than the heating effect. However, this slight en- 10-2 10-2 hancement is washed out when we further increase the

) ) excitation strength. These features are the same as in

0 -3 0 -3 , 10 Ω = , 10 t t ( (

r r Ω = the type 1 case.

i 0.26 0.34 i

a -4 a -4

p 10 0.28 0.36 p 10 0.5 0.65 χ 0.3 0.38 χ 0.55 0.7 10-5 0.32 0.4 10-5 0.6 0.75 (c)∆ω0 = 0.15 (d)∆ω0 = 0.15 10-6 − 10-6 − 0 20 40 60 80 0 20 40 60 80 t t 10-1 10-1 ∆ω0 = ∆ω0 = 0.02 0.06 0.02 0.1 0.02 0−.1 0.02 0.1

) 0−.06 0.1 ) 0−.06 0−.14 0 0 , − , 0.06 0.14 t t

( ( − − r r i i a a p p χ χ

(e) Ω =0.16 (f) Ω =0.16 10-2 10-2 0 20 40 60 80 0 20 40 60 80 t t

FIG. 20: χpair(t, 0) above Tc under the periodic driving of type 2. The parameters are g = 0.3, ω0 = 0.4, β = 320 without heat baths. Lines with different colors indicate different Ω. (a) (c) are for the cases around Ω = ωr, while (b)(d) are for the cases around Ω = 2ωr. (e)(f) show the results for Ω = 0.16 and various ∆ω0. Dotted lines are the linear component (in ∆ω0) estimated from the ∆ω0 = ±0.02 data. The dash- dotted black line in each panel is the equilibrium result.

1 C. Giannetti et al., Advances in Physics 65, 58 (2016). 19 M. Sch¨utt,P. P. Orth, A. Levchenko, and R. M. Fernandes, 2 M. Rini et al., Nature 449, 72 (2007). arXiv:1702.04793 (2017). 3 M. F¨orst et al., Phys. Rev. B 84, 241104 (2011). 20 L. D’Alessio and M. Rigol, Phys. Rev. X 4, 041048 (2014). 4 D. Fausti et al., Science 331, 189 (2011). 21 P. Ponte, A. Chandran, Z. Papi, and D. A. Abanin, Annals 5 S. Kaiser et al., Phys. Rev. B 89, 184516 (2014). of Physics 353, 196 (2015). 6 W. Hu, S. Kaiser, D. Nicoletti, and C. Hunt, Nature ma- 22 H. Aoki et al., Rev. Mod. Phys. 86, 779 (2014). terials 13, 705 (2014). 23 Y. Murakami, P. Werner, N. Tsuji, and H. Aoki, Phys. 7 R. Mankowsky et al., Nature 516, 71 (2014). Rev. B 91, 045128 (2015). 8 M. Mitrano et al., Nature 530, 461 (2016), letter. 24 Y. Murakami, P. Werner, N. Tsuji, and H. Aoki, Phys. 9 A. Cantaluppi et al., arXiv:1705.05939 (2017). Rev. B 93, 094509 (2016). 10 M. A. Sentef, A. F. Kemper, A. Georges, and C. Kollath, 25 Y. Murakami, P. Werner, N. Tsuji, and H. Aoki, Phys. Phys. Rev. B 93, 144506 (2016). Rev. B 94, 115126 (2016). 11 M. Knap et al., Phys. Rev. B 94, 214504 (2016). 26 P. Schmidt and H. Monien, arXiv:0202046 (2002). 12 A. Komnik and M. Thorwart, The European Physical 27 A. V. Joura, J. K. Freericks, and T. Pruschke, Phys. Rev. Journal B 89, 244 (2016). Lett. 101, 196401 (2008). 13 D. M. Kennes, E. Y. Wilner, D. R. Reichman, and A. J. 28 N. Tsuji, T. Oka, and H. Aoki, Phys. Rev. B 78, 235124 Millis, Nat Phys 13, 479 (2017), article. (2008). 14 M. Kim et al., Phys. Rev. B 94, 155152 (2016). 29 N. Tsuji, T. Oka, and H. Aoki, Phys. Rev. Lett. 103, 15 J.-i. Okamoto, A. Cavalleri, and L. Mathey, Phys. Rev. 047403 (2009). Lett. 117, 227001 (2016). 30 W.-R. Lee and K. Park, Phys. Rev. B 89, 205126 (2014). 16 M. A. Sentef, arXiv:1702.00952 (2017). 31 T. Mikami et al., Phys. Rev. B 93, 144307 (2016). 17 M. Babadi et al., arXiv:1702.02531 (2017). 32 M. F¨orst et al., Nat. Phys. 7, 854 (2011). 18 G. Mazza and A. Georges, arXiv:1702.04675 (2017). 33 T. Kuwahara, T. Mori, and K. Saito, Annals of Physics 21

367, 96 (2016). (1994). 34 T. Mori, T. Kuwahara, and K. Saito, Phys. Rev. Lett. 116, 50 C. Zerbe and P. H¨anggi,Phys. Rev. E 52, 1533 (1995). 120401 (2016). 51 Y. Murakami, P. Werner, N. Tsuji, and H. Aoki, Phys. 35 J. Bauer, J. E. Han, and O. Gunnarsson, Phys. Rev. B 84, Rev. B 88, 125126 (2013). 184531 (2011). 52 J. D. Axe and G. Shirane, Phys. Rev. Lett. 30, 214 (1973). 36 A. Subedi, A. Cavalleri, and A. Georges, Phys. Rev. B 89, 53 P. B. Allen, V. N. Kostur, N. Takesue, and G. Shirane, 220301 (2014). Phys. Rev. B 56, 5552 (1997). 37 R. Singla et al., Phys. Rev. Lett. 115, 187401 (2015). 54 F. Weber et al., Phys. Rev. Lett. 101, 237002 (2008). 38 P. Werner and M. Eckstein, Phys. Rev. B 88, 165108 55 A. Nava et al., arXiv:1704.05613 (2017). (2013). 56 M. Capone, M. Fabrizio, C. Castellani, and E. Tosatti, 39 A. O. Caldeira and A. J. Leggett, Phys. Rev. Lett. 46, 211 Rev. Mod. Phys. 81, 943 (2009). (1981). 57 Y. Nomura, S. Sakai, M. Capone, and R. Arita, Science 40 W. Metzner and D. Vollhardt, Phys. Rev. Lett. 62, 324 Advances 1, e1500568 (2015). (1989). 58 K. Steiner, S. Hoshino, Y. Nomura, and P. Werner, Phys. 41 A. Georges and G. Kotliar, Phys. Rev. B 45, 6479 (1992). Rev. B 94, 075107 (2016). 42 A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg, 59 S. Hoshino and P. Werner, arXiv:1609.00136 (2016). Rev. Mod. Phys. 68, 13 (1996). 60 T. Harada and S.-i. Sasa, Phys. Rev. Lett. 95, 130602 43 J. K. Freericks, Phys. Rev. B 50, 403 (1994). (2005). 44 61 M. Capone and S. Ciuchi, Phys. Rev. Lett. 91, 186405 We assume that at t0 = −∞ the system is free and de- (2003). coupled from the baths, and that the interactions and the 45 J. Hague and N. dAmbrumenil, J. Low Temp. Phys. 151, coupling to the baths are adiabatically turned on. 1149 (2008). 62 With the phonon bath it turns out that the phonon renor- 46 N. S¨akkinen,Y. Peng, H. Appel, and R. van Leeuwen, The malization and the effective phonon-phonon interaction Journal of Chemical Physics 143, (2015). discussed in Ref. 24 become weaker than without. Hence 47 2 M. Sch¨uler,J. Berakdar, and Y. Pavlyukh, Phys. Rev. B the oscillation frequency in χB0,X remains close to 2ωr in 93, 054303 (2016). the present case. 48 L.-F. Arsenault and A.-M. S. Tremblay, Phys. Rev. B 88, 63 As far as the system is in the weak-coupling regime the 205109 (2013). Migdal approximation may still work. 49 C. Zerbe, P. Jung, and P. H¨anggi,Phys. Rev. E 49, 3626