Skip to main content
PMC home page
Nature Portfolio logoLink to Nature Portfolio
. 2015 Mar 30;6:6708. doi: 10.1038/ncomms7708

Ultrafast and reversible control of the exchange interaction in Mott insulators

J H Mentink 1,a,*, K Balzer 1, M Eckstein 1
PMCID: PMC4389227  PMID: 25819547

Abstract

The strongest interaction between microscopic spins in magnetic materials is the exchange interaction Jex. Therefore, ultrafast control of Jex holds the promise to control spins on ultimately fast timescales. We demonstrate that time-periodic modulation of the electronic structure by electric fields can be used to reversibly control Jex on ultrafast timescales in extended antiferromagnetic Mott insulators. In the regime of weak driving strength, we find that Jex can be enhanced and reduced for frequencies below and above the Mott gap, respectively. Moreover, for strong driving strength, even the sign of Jex can be reversed and we show that this causes time reversal of the associated quantum spin dynamics. These results suggest wide applications, not only to control magnetism in condensed matter systems, for example, via the excitation of spin resonances, but also to assess fundamental questions concerning the reversibility of the quantum many-body dynamics in cold atom systems.

Inline graphicElectronic interactions underlie the exchange interaction responsible for the magnetic ordering and dynamics of magnetic materials. Here, Mentink et al. theoretically demonstrate the ultrafast and reversible tuning of the exchange interaction in Mott insulators driven by a time-periodic electric field.

Controlling magnetically ordered systems on sub-picosecond timescales is currently a widely studied research area owing to the joint fundamental interest and technological demand for faster and more energy-efficient magnetic storage1. The fastest pathways to reverse magnetic order utilize the exchange interaction Jex between microscopic magnetic moments2,3,4,5,6, which can exceed external magnetic fields by orders of magnitude. Because Jex relies on the electrostatic Coulomb repulsion and the Pauli principle rather than on magnetic dipole forces, it may be modified directly by the action of a laser pulse on the electronic state. This implies an appealing and largely unexplored scenario to control magnetism on the fastest possible timescale. Recently, several experimental studies on magnetic materials have discussed an ultrafast modification of Jex or change of the type of magnetic order by creating a nonequilibrium electron distribution (by photodoping or laser heating)7,8,9,10,11,12,13,14,15. In these cases the spin dynamics after the excitation strongly depends on the relaxation of the electrons, thereby hindering a direct and reversible control of the spin degrees of freedom alone. On the other hand, reversible electrical control of Jex was recently demonstrated in a multiferroic solid state system, where the bond alignment can be changed by a static electric field16. Clearly, a natural goal is to achieve control of Jex, which is both reversible and ultrafast, that is, it is active while a laser pulse is on, but leaves the electronic state unexcited after the pulse is switched off.

A versatile approach to reversibly control the dynamics of quantum systems is given by the rectification of time-periodic perturbations. The use of periodic driving to control the dynamics of a quantum system is known in many areas of physics, for example, through effective conservative forces resulting from the AC Stark effect or through the coherent destruction of tunnelling17,18. For extended solid state systems, it is well known that particles in a tight-binding band subject to periodic driving evolve under an effective Hamiltonian that has a different band structure17,18,19,20,21,22. On the other hand, the control of the exchange interaction requires to understand how the driving influences both the band structure and the electronic correlations, which determine Jex in equilibrium. This is a highly challenging problem in general, since it implies the solution of a strongly time-dependent many-body problem of an extended system.

In recent years, important insights into the control of Jex have been obtained by studying the effect of periodic driving for one and two spin systems, leading to light-induced Kondo effects23, the design of an effective low-energy spin Hamiltonian in bosonic double-well systems24, as well as an optically induced RKKY interaction between localized spins in semiconductor quantum dots by virtual excitation of delocalized excitons25,26. The latter has been shown to be effective in extended systems as well, in particular for the dilute ferromagnetic (FM) semiconductors27,28. However, since the spin dynamics in ferromagnets requires a change of the total angular momentum, it is difficult to induce fast dynamics by modifying Jex. To the contrary, extended antiferromagnetic (AFM) systems do not suffer from the angular momentum bottleneck, and hence can provide novel opportunities for the ultrafast control of the spin dynamics by modifying Jex.

In this paper, we demonstrate that it is possible to reversibly control Jex in extended AFM Mott insulators by periodically modulating the electronic structure with a frequency ω higher than Jex/ħ, but not resonant to electronic excitations. We investigate a simple driving scheme using time-periodic electric fields, which can be realized both in solid state systems and for cold atoms, and hence suggests wide applications: besides the possibility of manipulating magnetism in solids, for example, via the excitation of spin resonances, we find that in the extreme limit of strong driving one may even achieve a sign reversal of Jex, which is equivalent of letting the system evolve backwards in time and may allow for addressing fundamental questions concerning the reversibility of quantum many-body dynamics29,30,31 in cold atom experiments. Furthermore, we show that considerable insight can be obtained from analytical Floquet theory for a few-site cluster under continuous driving, which predicts reversible enhancement, reduction and even complete sign-change of the exchange interaction. The relevance of these results for extended many-body systems may not be clear a priori, since in this case a true quasi-steady driven state may always become infinitely excited32. For the relatively short-term dynamics of interest here, the predictions of Floquet theory are nevertheless correct, as we demonstrate using numerical calculations for both the classical spin dynamics in high-dimensional Mott insulators and the quantum spin dynamics in low-dimensional Mott insulators.

Results

Floquet theory for a two-site cluster

In this work, we study the repulsive Hubbard model as a model for strongly interacting electrons on a lattice. The Hamiltonian is given by

graphic file with name ncomms7708-m1.jpg

where Inline graphic creates an electron at site i with spin σ=↑, ↓, t0 is the hopping between nearest-neighbour sites, and U the repulsive on-site interaction. Arbitrary time-dependent electric fields E(t) are incorporated by adding a Peierls phase to the hopping matrix elements (see Methods). Below we set ħ=1 and measure energy and time in units of the hopping t0 and the inverse hopping, respectively. Electric fields are measured in units of t0/ea, where a is the lattice spacing and Inline graphic the electron charge.

For half-filling and Inline graphic, the Hubbard model describes a Mott insulator with one electron per site, in which the remaining spin degrees of freedom are coupled by an AFM exchange interaction Inline graphic. The simplest analytical understanding for this result is obtained already for two electrons on two Hubbard sites: for total Sz=0 we have four states. In the atomic limit, two of them (|↑, ↓› and |↓, ↑›) are singly occupied sites at E1=0, while the other two states involve a doubly occupied and empty site at energy E2=U (|↑↓, 0› and |0, ↑↓›). In the presence of hopping, the degeneracy is lifted and the lowest states become singlet and triplet states at energies Inline graphic and ET=0, respectively. The low-energy spectrum is thus described by a spin Hamiltonian 2JexS1S2 with Inline graphic. This analytical understanding from the cluster is useful since lattice effects beyond the lowest-order perturbative result only appear in the order Inline graphic. In the same spirit, to gain theoretical insight into the modification of Jex by periodic driving, we first consider the same two-site Hubbard cluster and employ Floquet’s theorem33,34, the analogue of Bloch’s theorem in time. When the Hamiltonian is periodic in time with a period T=2π/ω, solutions of the time-dependent Schrödinger equation are given in the form Inline graphic, where Inline graphic is time periodic and Inline graphic is a quasi-energy defined up to multiples of ω. The Floquet picture describes a system that undergoes virtual absorption and emission of an arbitrary number of photons, as depicted in Fig. 1c for the Mott–Hubbard systems. The unperturbed Floquet sectors are described by the time-averaged Hamiltonian shifted by nω, and mixing between these Floquet sectors results in a renormalization of quasi-energy levels. A natural procedure is then to adopt an ‘adiabatic’ principle in which the driving amplitude varies slowly as compared with the driving frequency, Inline graphic, and identify the amplitude-dependent singlet-triplet splitting Inline graphic in the quasi-energy spectrum with the (time dependent) exchange interaction that describes the spin dynamics in the laser driven system on timescales much slower than the driving period T.

Figure 1. Floquet spectrum of a two-site Hubbard model.

Figure 1

Open in a new tab

(a) Floquet spectrum for U=10 and ω=6 as a function of Floquet amplitude Inline graphic. Thin dashed lines indicate the spectrum without driving. The exchange interaction (Jex) is extracted from the difference between the lowest singlet (ES) and triplet (ET) levels. The result is shown in b for the same U=10 and two different frequencies above (ω=16, blue solid line) and below (ω=6, red dashed line) the Mott gap. For large driving strength Inline graphic, a reversal of Jex is possible. In addition, thin blue dotted (ω=16) and thin red dash dotted (ω=6) lines indicate the modification of Jex as obtained within second-order perturbation theory. (c) Illustration of the modification of Jex in the Floquet picture. Red wavy lines indicate the virtual absorption and emission of an arbitrary number n of photons with frequency ω. This induces a coupling to excited states in different Floquet sectors displaced by an energy U±nω.

The numerical solution of the Floquet spectrum and various analytically tractable limiting cases for the two-site Hubbard model are detailed in the methods section. For a tight-binding model driven by an electric field E(t)=E0 cos(ωt), the coupling between Floquet sectors is controlled by the dimensionless Floquet amplitude

graphic file with name ncomms7708-m14.jpg

and the time averaging of H corresponds to a coherent reduction of the tunnelling amplitude17,18 by a factor Inline graphic, where J0 is the Bessel function (see Methods). In Fig. 1, we display the Floquet spectrum and the exchange splitting Inline graphic for a half-filled two-site Hubbard model. In contrast to the limit Inline graphic, t0, where the only effect would be a renormalization of the hopping by Inline graphic and a corresponding reduction of the exchange splitting at large U by a factor Inline graphic, one can see that Jex can be both increased and decreased for finite ω, depending on the driving. This is clear already in the perturbative limit for Inline graphic and Inline graphic, which is given by Inline graphic with (see Methods)

graphic file with name ncomms7708-m23.jpg

and indicated with thin (dash-)dotted lines in the right panel of Fig. 1. The last term of equation (3) is the reduction of the exchange due to coherent reduction of the tunnelling, while the first two terms derive from the coupling to the m=±1 Floquet sector with effectively shifted charge-transfer energies U±ω. The net effect is an enhancement (reduction) of Jex for driving frequencies below (above) the Coulomb energy U. For sufficiently strong driving one can even reverse the sign of Jex, thus leading to the remarkable finding of a FM exchange coupling in the half-filled Hubbard model. This happens when Inline graphic is of order one, such that coupling to higher Floquet sectors with effective Coulomb energy U-mω becomes strongly enhanced, while the direct exchange path is reduced by coherent destruction of tunnelling ~Inline graphic. For larger Floquet amplitudes, the direct exchange path again increases due to the oscillating behaviour of the Bessel function. In the remainder of this paper, we verify that these predictions from the two-site Floquet picture can remain valid for extended condensed matter systems at off-resonant driving with finite pulse duration, in spite of the possible importance of higher-order processes such as multi-photon absorption in the strongly driven regime, and the limited number of cycles in the pulse.

Mean-field spin dynamics in a high-dimensional lattice

A direct prediction of the Floquet theory (equation (3)) is the enhancement (reduction) of the exchange interaction for driving below (above) the Mott gap with weak amplitudes (Inline graphic). A large class of materials for which this might be relevant are three-dimensional bulk systems, such as transition metal oxides, whose low-energy spin dynamics contain uniform spin resonances that can be conveniently described in mean-field theory. For large dimensions, also a numerical solution of the nonequilibrium electron dynamics in the Hubbard model is possible using the dynamical mean-field theory (DMFT, see Methods).

Within DMFT, the equilibrium solution of the Hubbard model at half-filling and low temperature is the Néel state. To asses the exchange interaction in this state, we study the excitation of resonances in the AFM phase in a transverse magnetic field Bx, a setup that was pioneered in ref. 15 for resonant photoexcitation. In equilibrium, the balance of Bx and Jex gives rise to a canting of the magnetic sublattices out of the y–z plane. If Jex is modified under periodic driving, the sub-lattice magnetizations are no longer aligned with the effective field Beff given by external and exchange fields, as illustrated in the inset of Fig. 2. This implies a rotation of the spins in the plane perpendicular to Bx (leaving the total angular momentum Sx conserved), from which the time-dependent modification of the exchange interaction is calculated (see Methods). Hence, in our calculations, Jex is defined as the parameter that describes best the observed transverse spin dynamics as is obtained from the solution of the full electron problem. We stress that this allows us to quantify changes of the exchange interaction independent of the exchange energy stored in the system. This is particularly important for the regime in which absorption is not negligible, where the laser excites mobile carriers that transfer their energy to the spin background on an ultrafast timescale and thus reduce the ordered moment.

Figure 2. Laser-induced modification of the exchange interaction.

Figure 2

Open in a new tab

(a,b) Time-dependent change of the exchange interaction (ΔJex, thick lines) during the action of a laser pulse, for driving frequencies ω=3 (a) and ω=12 (b) below and above the Mott gap, respectively. Different colours correspond to results obtained with different amplitude E0 of the electric field, increasing from light to dark. Thin lines show the time dependence of the electric field (E-field). Numerical results were obtained using dynamical mean-field theory (DMFT) for the hyper-cubic lattice at U=10 and initial temperature T=0.025. (c) The driving susceptibility Inline graphic for Inline graphic for frequencies above (blue, right vertical axis) and below gap (red, left vertical axis), obtained from DMFT for the hyper-cubic lattice (disks), from the numerical Floquet spectrum of a two-site Hubbard cluster (solid lines), and from the perturbative result equation (3) (dashed lines). The inset illustrates the canted geometry of the two sub-lattice magnetizations S1,2 (black arrows) induced by a static transverse magnetic field Bx (grey arrow). In equilibrium, the effective magnetic field Inline graphic (green arrow) is collinear with S1. A modification of the exchange interaction (ΔJex) would rotate the effective field (light blue arrow) with respect to S1 causing the excitation of a spin resonance. In the DMFT calculations, ΔJex is computed from the observed spin precession in this canted geometry.

We have implemented the DMFT solution of the Hubbard model in a time-dependent external electric field for the infinite-dimensional hyper-cubic lattice with density of states Inline graphic (refs 35, 36). The electric field is pointing along the body diagonal of the lattice and represents a laser pulse with frequency ω and a Gaussian envelope that contains 15 cycles per pulse, that is, Inline graphic with tc=15π/(2.1ω). Figure 2a,b show time traces of the induced change of the exchange interaction Inline graphic for one driving frequency below (a) and above (b) gap, as extracted from the time evolution of the spin degrees of freedom during the pulse. In accordance with the prediction of the Floquet theory, we observe an enhancement (reduction) of the exchange interaction during the application of the field with a frequency below (above) gap. The frequency ω=3 in Fig. 2a is far from the resonance ω≈U, and we observe that Inline graphic after the pulse, demonstrating the reversibility of the effect. Conversely, the driving frequency ω=12 in Fig. 2b is chosen close to the edge of the upper Hubbard band where we observe significant absorption and transient behaviour after the pulse. Hence, although the exchange interaction is modified in this case as well, the effect is not reversible. Note also that the time reached in the present simulations is too short for the photoexcited carriers to relax, hence a description in terms of a quasi-stationary photo-doped state discussed earlier15 is not yet valid.

A quantitative comparison with the Floquet theory is shown in the bottom panel of Fig. 2, where the ‘driving susceptibility’ Inline graphic for Inline graphic is plotted as a function of the driving frequency. Solid discs show the DMFT results as obtained by a linear fit through the dependence of the ratio ΔJex(t)/Jex on Inline graphic at its maximum. Dashed and solid lines show the results based on the perturbative Floquet formula (equation (3)) and the full Floquet spectrum (non-perturbative in t0/U), evaluated from the derivative Inline graphic at Inline graphic. As expected, in the vicinity of the band edge (|ω−U|~2), strong deviation is found since band absorption is not captured in a cluster picture. Away from the band edge, however, the frequency dependence matches very well, being even in quantitative agreement for the lowest frequencies below gap. This demonstrates the usefulness of the Floquet theory for understanding how off-resonant periodic driving modifies the exchange interaction in extended condensed matter systems by photo-assisted hopping.

One-dimensional quantum spin dynamics

An intriguing prediction of the Floquet analysis is the existence of amplitude and frequency ranges in which the exchange coupling becomes FM. Such a sign-change of Jex cannot cause a transition to a FM state since the Hubbard model equation (1) conserves the total spin. However, even if the system remains AFM, a change of sign of Jex by periodic driving allows for a very non-trivial and unique way to control the spin dynamics, namely, to reverse the time evolution of the undriven system. Such time reversal can be anticipated by considering a pure Heisenberg spin Hamiltonian Inline graphic, which gives an accurate description of the low-energy spin dynamics in the half-filled Hubbard model at Inline graphic if the system is not electronically excited. In the absence of driving, the propagation over a time interval t is given by the evolution operator Inline graphic. Such evolution can exactly be reversed by the propagation with an exchange interaction Inline graphic of opposite sign over a time interval Inline graphic, since for the FM time evolution operator we have Inline graphic, that is, the two time evolution operators are exactly inverse to each other.

To demonstrate that periodic driving of the Hubbard model at large U indeed yields the anticipated time reversal of the spin degrees of freedom, we consider a chain of L=10 sites and compute the dynamics using exact diagonalization techniques (see Methods). The system is initially prepared in a classical Néel state Inline graphic and is evolved under the unperturbed Hamiltonian (1). In a quantum Heisenberg model, the classical Néel state is a highly excited state the energy of which exceeds the thermal energy at the Néel temperature, such that no remanent long-range order is expected at long times, apart from finite size effects. In one dimension, not even the ground state displays long-range order. As a consequence of the spin-flip terms Inline graphic in the effective AFM Heisenberg model, we thus observe a rapid decay of the total staggered magnetization Inline graphic (Fig. 3b). After this initial free evolution to a state where long-range order is suppressed, we ramp on a time-periodic electric field (Fig. 3a), with Floquet amplitude Inline graphic and frequency ω/U=0.6 such that the Floquet theory for a two-site model predicts a reversal of the exchange coupling. Under the periodic driving, one indeed observes a near perfect reversal of the dynamics of M(t) in Fig. 3b, which almost completely recovers to the initial value M(t=0) around t≈100. Subsequently, M(t) is reduced again by further evolution in the reverse direction, as a consequence of the spin-flip terms in the FM model. This continues until the field is ramped off, after which one observes that the free evolution brings the system again back to the initial state, from which the same rapid decay of M(t) is observed as for the initial free evolution. Hence, we conclude that the periodic driving allows for a reversible control of the spin dynamics for the timescale considered in our simulations. This is further confirmed by the time evolution of the total double occupation Inline graphic, which has the same mean value before and after the driving, demonstrating that electronic excitations due to the driving are negligible. The weak oscillations in d(t) are caused by switching on the hopping at t=0, while the increased mean value of d(t) during driving is due to photo-assisted hopping processes.

Figure 3. Time reversal of the spin dynamics by periodic driving.

Figure 3

Open in a new tab

(a) Field envelope with cosine-shaped ramps of length Δt=15 around tf=45 and t=112.5. The bar below the field envelope indicates forward (grey) and backward (white) time evolution when the field is off and on, respectively. (b) Time evolution of staggered magnetization M (red) and total double occupation d (blue), scaled by a factor of 50, of a 10-site Hubbard chain, showing free evolution for times t<37.5 and t>120 and evolution under an additional periodic driving at frequency ω/U=0.6 and Floquet amplitude Inline graphic in between. (c) Build up and diminishing of the spin–spin correlation function Inline graphic. Numerical results were obtained by exact diagonalization for U=50 and open boundary conditions, starting from a classical Néel state. The colour bar indicates the value of Inline graphic, ranging from −0.25 (dark red) through 0 (white) to 0.25 (dark blue).

Time reversal can be demonstrated not only on the level of local observables. Figure 3c displays the evolution of the spin–spin correlation function Inline graphic as a function of distance l and time (Nl is the number of site pairs with distance l). Starting from the initial uncorrelated product state, correlations build up under the evolution of H. Even though the system is small, the spreading of correlations resembles the light-cone effect that has been observed in quantum many-body systems after a quench37, that is, correlations stay zero outside the light cone |l|≤2εt, where ε is a maximal mode velocity38, while short-range AFM correlations emerge inside the light cone. Further, under the action of the periodic driving, the spin–spin correlations diminish with the same speed, restoring the initially uncorrelated state.

To determine quantitatively how well the time evolution is reversed in our simulations, we computed the difference ΔM=1−M(t*) between the initial magnetization in the Néel state and the magnetization M(t*) at the revival time t* for different values of U (Fig. 4). In all simulations, the system is evolved forward in time for a given time tf=10, after which the field is ramped on for a period Δt=10. As before, we choose Inline graphic and ω=0.6U, which gives the same relative change of Jex for all sufficiently large values Inline graphic. The observed scaling ΔM~1/U2 indicates that in the current setup the deviation from perfect reversal originates from small electronic excitations above the gap, which arises from switching on the hopping in the beginning of the simulation and the ramping on of the field. Since the dynamics of the electronically excited states is not captured by the spin Hamiltonian, it is not time-reversed. While the electronic excitation can be further reduced by slow ramping, the ultimate limit is given by non-Heisenbergian terms in the effective spin Hamiltonian, for which perfect reversal under periodic driving is not expected. For the half-filled Hubbard model in equilibrium, the leading-order correction to the Heisenberg model appears in the order Inline graphic in the strong-coupling expansion. At least for small times t, the contribution of a perturbation δH proportional to Inline graphic to the time-reversed Hamiltonian would lead to a scaling ΔM~1/U3, which is smaller than the electronic excitation in the present case. Note, however, that the times reached in the current simulations are nevertheless long enough to observe near perfect reversal even from a state without magnetic order, in which spin correlations have spread throughout the full chain (Fig. 3).

Figure 4. Quality of the time reversal.

Figure 4

Open in a new tab

(a) Time evolution of the staggered magnetization M in a 10-site Hubbard chain, where after fixed forward propagation time tf=10 the field is ramped up within a time interval Δt=10. Different colours correspond to different values of U, as indicated by the corresponding coloured labels. (b) Difference ΔM between the initial staggered magnetization and the staggered magnetization at the revival time t* as a function of U. The error bars represent the magnitude of the short-time fluctuations of M close to t*. In all calculations, the driving frequency ω/U=0.6 and Floquet amplitude Inline graphic were used.

Discussion

Our results demonstrate ultrafast and reversible electrical control of the exchange interaction in extended fermionic many-body systems by modulation with time-periodic electric fields. These numerical results have recently been corroborated analytically in the high-frequency limit, by showing that the effective spin model for the two-site cluster remains valid for the driven lattice39. We emphasize that Floquet amplitudes Inline graphic are well accessible for condensed matter systems, which would lead to relative changes of Jex up to 1%. For example, for a frequency ħω=1 eV and a lattice spacing of 2 Å, a laser fluence of 1 mJ cm−2 in a 100-fs pulse corresponds to a Floquet parameter Inline graphic. While realistic condensed matter systems usually involve several correlated bands, we think that our current results are already quite robust for the AFM oxides governed by superexchange interactions. Similar as in the single-band model studied here, superexchange interactions are governed by virtual charge excitations, which will be reversibly modified through the mechanism of photo-assisted hopping between different Floquet sectors. This is further supported by recently presented experiments on canted AFM oxides40 using terahertz emission spectroscopy41, which show the first experimental evidence of reversibly controlling exchange interactions by off-resonant pumping below the charge-transfer gap. Furthermore, equation (3) implies a strong enhancement of off-resonant effects at low frequencies. As an extreme limit of this one can anticipate control of spin dynamics by few-cycle terahertz pulses, using both the coupling between the spins to the magnetic field of the light42, and the modifications of Jex predicted by our work. At the same time, extensions to multi-band models are very interesting and relevant to perform as they usually involve multiple possibly competing exchange paths, which generally yield quantitative differences as well as different dependencies on the driving frequency due to the presence of additional resonances, potentially enabling to achieve even stronger effects. In this connection, we also mention that extensions to multi-band models may be relevant too for the description of metallic ferromagnets. Here static electronic structure calculations already indicate that exchange integrals involving electrons excited to higher bands considerably differ from those in the ground state43. Furthermore, a more accurate quantitative description of photo-excited states would also include the effect of dynamic screening. However, we anticipate that this will be important only if mobile carriers are injected, that is, for photo-doping excitations, while it will be a secondary effect for the off-resonant driving investigated here, which leaves the electron distribution unchanged.

While the single-band model is clearly a minimal model for application to condensed matter systems, fermionic cold atom systems resemble the single-band model very accurately. Moreover, such systems may realize the large amplitudes Inline graphic needed to achieve the FM exchange at frequencies sufficiently far from resonances nω=U, which are difficult to realize for most condensed matter systems. It will, therefore, be of fundamental interest to investigate the reversal of the exchange interaction and the associated time reversal of the quantum spin dynamics in cold atom experiments. In these systems recently, great progress has been made to prepare and measure systems with single-site spatial resolution44,45. Furthermore, various single-particle Floquet Hamiltonians could be realized in the limit of strong periodic driving without substantial heating by inter-band absorption (see, for example, ref. 46). Cold atoms in optical lattices have been successfully used as a quantum simulator for the dynamics of a quantum quenches in the Bose–Hubbard model, starting from an artificially prepared charge-ordered phase47, which suggest similar techniques to probe the behaviour of spin systems under time reversal. In the methods section (cf. equation (7)), we show that analogous time reversal can be achieved by modulating the amplitude of the hopping instead of its phase, which is easier to control in cold atoms. An intriguing problem to study both theoretically and experimentally is the fundamental question how well the time evolution can be reversed after (much) longer forward evolution time and investigate systematically the influence of small deviations from perfect time reversal. Furthermore, a study of the Loschmidt echo and dynamical phase transitions48 in cold atoms might be possible by including additional perturbations to the back propagation.

Methods

Floquet theory

The Floquet quasi-energy spectrum can be obtained from the ansatz Inline graphic by expanding Inline graphic in a Fourier series Inline graphic, where Inline graphic is referred to as the component of the wave function in the mth Floquet sector. The Schrödinger equation then achieves a block-matrix structure

graphic file with name ncomms7708-m59.jpg

where Inline graphic are the Fourier components of the Hamiltonian. Different from the usual discussions of Floquet theory for single-particle Hamiltonians, here we focus on the effect of periodic driving on an electronic spectrum that is controlled by electronic correlations.

Time-dependent electric fields are incorporated into the Hubbard Hamiltonian (1) by adding a time-dependent Peierls phase to the hopping matrix elements, tij(t)=t0 exp[ieaAij(t)], where Aij is the projection of the vector potential along the direction from site i to j (choosing a gauge with zero scalar potential and E(t)=−∂tA(t)). For the one-dimensional chain with electric field E0 sin(ωt) along the chain, this implies Inline graphic. The Fourier components of the Hamiltonian are thus given by

graphic file with name ncomms7708-m62.jpg

plus the additional (time-independent) interaction part in the m=0 component, where Jm(x) is the mth Bessel function, and the dimensionless parameter Inline graphic measures the strength of the perturbation (cf. equation (2)). For the numerical determination of the Floquet energies, one truncates the number of Floquet sectors in equation (4) to |n|≤N, and increases N to reach convergence. The determination of a many-body Floquet spectrum thus requires the diagonalization of a matrix of dimension N × D, where D is the dimension of the Hilbert space. The results presented in Fig. 1 are converged with N=8.

In the limit of large frequency, Inline graphic, t0, Floquet sectors in equation (4) are separated in energy, and one can restrict oneself to the lowest sector m=0. This is equivalent to replacing the Hamiltonian with its time average, which leads to the renormalization of the hopping by Inline graphic, and a corresponding reduction of the exchange by a factor Inline graphic. In the perturbative limit where both Inline graphic and Inline graphic, we expand the Bessel functions Jn(x)~xn for x→0. To lowest order only states of the m=0 and m=±1 Floquet sectors have to be taken into account, and the result given by equation (3) follows from standard second-order perturbation theory. Furthermore, an interesting limit for the Mott regime is given by Inline graphic, but allowing for fields of arbitrary amplitude. Because all terms Hm for m≠0 are proportional to t0, the perturbative shift of the spin states in the m=0 sector is given by a sum over all second-order processes containing precisely one virtual hopping to a higher Floquet sector and back, yielding

graphic file with name ncomms7708-m70.jpg

The unperturbed exchange is modified by a factor dependent only on ω/U and Inline graphic. For the parameters Inline graphic and ω/U=0.6 chosen for the time reversal, for example, the factor is given by −0.95 and indicates a near perfect sign reversal.

Finally, we note that a similar analysis is possible for the case of periodic modulation of the hopping amplitude, taking t0(t)=t0(1+A cos(ωt)). As above, we obtain for Inline graphic

graphic file with name ncomms7708-m74.jpg

yielding, for example, a perfect sign reversal for driving above gap ω/U=1.2 at A≈0.67. Other than for the field driven case, the perturbation is purely harmonic, and only Floquet sectors m=±1 enter this expression.

Dynamical mean-field theory

To solve the electron dynamics in the Hubbard model, we use nonequilibriumDMFT35,49. The electric field of the laser is incorporated by the Peierls substitution (see Methods, Floquet theory), so that the light matter interaction within the single-band model is treated to all orders. Within DMFT50, which becomes exact in the limit of infinite dimensions51, local correlation functions are obtained from an effective impurity model in which one site of the lattice is coupled to a non-interacting, self-consistently determined bath. The impurity model is solved within the perturbative hybridization expansion (non-crossing approximation52). The accuracy of this approach has been tested in equilibrium and for the short-time dynamics by comparison with higher-order hybridization expansions, as well as with numerically exact Quantum Monte Carlo52,53,54, which revealed good agreement both in the AFM and paramagnetic Mott insulator regime Inline graphic. A detailed description of the formalism and of our numerical implementation is given in ref. 15, which studies the same setup (that is, a hyper-cubic lattice with electric field along the body diagonal), yet for different electric fields.

Determination of J ex in DMFT

To asses the exchange interaction from mean-field spin dynamics, we investigate the AFM phase of the Hubbard model supplemented with a term Inline graphic, which couples the spin Inline graphic to a homogeneous magnetic field Bx along the x axis (Inline graphic denotes the Pauli matrices; α=x, y, z). DMFT allows us to compute the time-dependent expectation value of the electron spin ‹S1,2› on the two magnetic sublattices. Assuming a rigid macrospin model, the time-dependent exchange interaction can be inferred from these results by inverting the Landau–Lifshitz equation for the dynamics of spins on the mean field Inline graphic. It was shown15 that this approach compares well with the definition of exchange interactions from a time-dependent response formalism55, as well as to the analytical perturbative result Inline graphic in equilibrium at large U. In the transverse field, the equilibrium exchange interaction can be determined from the canting induced by Inline graphic, yielding Inline graphic. Out of equilibrium, we obtain Inline graphic,

graphic file with name ncomms7708-m84.jpg

where the time averaging is done to extract only the low-frequency component, similar as in the Floquet theory. Note that by calculating the exchange interaction in this way, we have automatically projected out dynamical changes in the (time averaged) local magnetization |‹S1,2›|, and slight changes of the local moments (double occupation) as a result of the virtual absorption and emission of photons.

Exact diagonalization

To compute the time evolution of the (driven) one-dimensional Hubbard model from the Schrödinger equation with a time-dependent Hamiltonian H(t) and a given initial state Inline graphic, we use the Krylov technique56 in combination with a commutator-free exponential time-propagation scheme57. While the Krylov method provides efficient approximations to the time propagator, which are important to treat large Hilbert spaces, the commutator-free exponential time-propagation scheme is related to the Magnus expansion, and, preserving unitarity, allows for a high-order accurate integration of the Schrödinger equation in time.

Author contributions

The Floquet theory was developed by M.E. and J.H.M. DMFT simulations were run by J.H.M., while K.B. obtained the exact diagonalization results. All authors contributed to the discussion and interpretation of the results and the writing of the manuscript.

Additional information

How to cite this article: Mentink, J. H. et al. Ultrafast and reversible control of the exchange interaction in Mott insulators. Nat. Commun. 6:6708 doi: 10.1038/ncomms7708 (2015).

Acknowledgments

We thank U. Bovensiepen, M.I. Katsnelson, A.V. Kimel, A. Lichtenstein, R.V. Mikhaylovskiy, T. Oka, A. Secchi, N. Tsuji and Ph. Werner for fruitful discussions. The exact diagonalization calculations were run in part on the supercomputer HLRN-II of the North-German Supercomputing Alliance. J.H.M. acknowledges funding from the Nederlandse Organisatie voor Wetenschappelijk onderzoek (NWO) by a Rubicon Grant.

References

  1. Kirilyuk A., Kimel A. V. & Rasing Th. Ultrafast optical manipulation of magnetic order. Rev. Mod. Phys. 82, 2731–2784 (2010) . [Google Scholar]
  2. Radu I. et al. Transient ferromagnetic-like state mediating ultrafast reversal of antiferromagnetically coupledspins. Nature 472, 205–208 (2011) . [DOI] [PubMed] [Google Scholar]
  3. Mentink J. H. et al. Ultrafast spin dynamics in multisublattice magnets. Phys. Rev. Lett. 108, 057202 (2012) . [DOI] [PubMed] [Google Scholar]
  4. Wienholdt S., Hinzke D., Carva K., Oppeneer P. M. & Nowak U. Orbital-resolved spin model for thermal mag-netization switching in rare-earth-based ferrimagnets. Phys. Rev. B 88, 020406 (2013) . [Google Scholar]
  5. Baryakhtar V. G., Butrim V. I. & Ivanov B. A. Exchange relaxation as the mechanism of ultrafast spin re-orientation in two-sublattice ferrimagnets. JETP Lett. 98, 289–293 (2013) . [Google Scholar]
  6. Evans R. F. L., Ostler T. A., Chantrell R. W., Radu I. & Rasing Th. Ultrafast thermally induced magnetic switching in synthetic ferrimagnets. Appl. Phys. Lett. 104, 082410 (2014) . [Google Scholar]
  7. Rhie H.-S., Dürr H.A. & Eberhardt W. Femtosecond electron and spin dynamics in ni/w(110) films. Phys. Rev. Lett. 90, 247201 (2003) . [DOI] [PubMed] [Google Scholar]
  8. Ju G. et al. Ultrafast generation of ferromagnetic or-der via a laser-induced phase transformation in ferh thin films. Phys. Rev. Lett. 93, 197403 (2004) . [DOI] [PubMed] [Google Scholar]
  9. Thiele J., Buess M. & Back C. H. Spin dynamics of the antiferromagnetic-to-ferromagnetic phase transition in ferh on a sub-picosecond time scale. Appl. Phys. Lett. 85, 2857 (2004) . [Google Scholar]
  10. Wall S., Prabhakaran D., Boothroyd A. T. & Cavalleri A. Ultrafast coupling between light, coherent lattice vibrations, and the magnetic structure of semicovalent LaMnO3. Phys. Rev. Lett. 103, 097402 (2009) . [DOI] [PubMed] [Google Scholar]
  11. Först M. et al. Driving magnetic order in a manganite by ultrafast lattice excitation. Phys. Rev. B 84, 241104 (2011) . [Google Scholar]
  12. Carley R. et al. Femtosecond laser excitation drives ferromagnetic gadolinium out of magnetic equilibrium. Phys. Rev. Lett. 109, 057401 (2012) . [DOI] [PubMed] [Google Scholar]
  13. Li T. et al. Femtosecond switching of magnetism via strongly correlated spin-charge quantum excitations. Nature 496, 69–73 (2013) . [DOI] [PubMed] [Google Scholar]
  14. Matsubara M. et al. Tuning the ultrafast spin dynamics in carrier-density-controlled ferromagnets. Preprint at http://arxiv.org/abs/1304.2509 (2013) .
  15. Mentink J. H. & Eckstein M. Ultrafast quenching of the exchange interaction in a Mott insulator. Phys. Rev. Lett. 113, 057201 (2014) . [DOI] [PubMed] [Google Scholar]
  16. Ryan P. J. et al. Reversible control of magnetic inter-actions by electric field in a single-phase material. Nat. Commun. 4, 1334 (2013) . [DOI] [PubMed] [Google Scholar]
  17. Dunlap D. H. & Kenkre V. M. Dynamical localization of a particle moving under inuence of an electric field. Phys. Rev. B 34, 3625–3633 (1986) . [DOI] [PubMed] [Google Scholar]
  18. Grossmann F., Dittrich T., Jung P. & Hänggi P. Coherent destruction of tunneling. Phys. Rev. Lett. 67, 516–519 (1991) . [DOI] [PubMed] [Google Scholar]
  19. Lignier H. et al. Dynamical control of matter-wave tunneling in periodic potentials. Phys. Rev. Lett. 99, 220403 (2007) . [DOI] [PubMed] [Google Scholar]
  20. Oka T. & Aoki H. Photovoltaic Hall effect in graphene. Phys. Rev. B 79, 081406 (2009) . [Google Scholar]
  21. Tsuji N., Oka T., Werner P. & Aoki H. Dynamical band ipping in fermionic lattice systems: an ac-field-driven change of the interaction from repulsive to attractive. Phys. Rev. Lett. 106, 236401 (2011) . [DOI] [PubMed] [Google Scholar]
  22. Wang Y. H., Steinberg H., Jarillo-Herrero P. & Gedik N. Observation of Floquet-Bloch states on the surface of a topological insulator. Science 342, 453–457 (2013) . [DOI] [PubMed] [Google Scholar]
  23. Shahbazyan T., Perakis I. & Raikh M. Spin correlations in nonlinear optical response: Light-induced kondo effect. Phys. Rev. Lett. 84, 5896–5899 (2000) . [DOI] [PubMed] [Google Scholar]
  24. Chen Y.-A., Nascimbeńe S., Aidelsburger M., Atala M., Trotzky S. & Bloch I. Controlling correlated tunnel-ing and superexchange interactions with ac-driven optical lattices. Phys. Rev. Lett. 107, 210405 (2011) . [DOI] [PubMed] [Google Scholar]
  25. Piermarocchi C., Pochung Chen L., Sham, & Steel D. Optical RKKY interaction between charged semicon-ductor quantum dots. Phys. Rev. Lett. 89, 167402 (2002) . [DOI] [PubMed] [Google Scholar]
  26. Piermarocchi C. & Quinteiro G. F. Coherent optical control of spin-spin interaction in doped semiconductors. Phys. Rev. B 70, 235210 (2004) . [Google Scholar]
  27. Fernández-Rossier J., Piermarocchi C., Chen P., Mac-Donald A. & Sham L. Coherently photoinduced ferromagnetism in diluted magnetic semiconductors. Phys. Rev. Lett. 93, 127201 (2004) . [DOI] [PubMed] [Google Scholar]
  28. Wang J., Cotoros I., Dani K., Liu X., Furdyna J. & Chemla D. Ultrafast enhancement of ferromagnetism via photoexcited holes in GaMnAs. Phys. Rev. Lett. 98, 217401 (2007) . [DOI] [PubMed] [Google Scholar]
  29. Lebowitz J. L. Boltzmann’s entropy and time’s arrow. Phys. Today 46, 32–38 (1993) . [Google Scholar]
  30. Polkovnikov A. Microscopic diagonal entropy and its connection to basic thermodynamic relations. Ann. Phys. 326, 486–499 (2011) . [Google Scholar]
  31. Caneva T., Silva A., Fazio R., Lloyd S., Calarco T. & Montangero S. Complexity of controlling quantum many-body dynamics. Phys. Rev. A 89, 042322 (2014) . [Google Scholar]
  32. D’Alessio L. & Rigol M. Long-time behavior of isolated periodically driven interacting lattice systems. Phys. Rev. X 4, 041048 (2014) . [Google Scholar]
  33. Floquet G. Sur les équations différentielles linéaires à coefficients périodiques. Ann. Sci. Ec. Normale Super 12, 47–88 (1883) . [Google Scholar]
  34. Grifoni M. & Hänggi P. Driven quantum tunnelling. Phys. Rep. 304, 229–354 (1998) . [Google Scholar]
  35. Aoki H., Tsuji N., Eckstein M., Kollar M., Oka T. & Werner P. Nonequilibrium dynamical mean-field theory and its applications. Rev. Mod. Phys. 86, 779–837 (2014) . [Google Scholar]
  36. Turkowski V. & Freericks J. K. Nonlinear response of bloch electrons in infinite dimensions. Phys. Rev. B 71, 085104 (2005) . [Google Scholar]
  37. Cheneau M. et al. Light-cone-like spreading of correlations in a quantum many-body system. Nature 481, 484–487 (2012) . [DOI] [PubMed] [Google Scholar]
  38. Calabrese P. & Cardy J. Time dependence of correlation functions following a quantum quench. Phys. Rev. Lett. 96, 136801 (2006) . [DOI] [PubMed] [Google Scholar]
  39. Itin A. P. & Katsnelson M. I. Effective Hamiltonians for fastly driven many-body lattice systems. Preprint at http://arxiv.org/abs/1401.0402v2 (2015) . [DOI] [PubMed]
  40. Mikhaylovskiy R. V. et al. Inverse magneto-refraction as a mechanism for laser modification of spin-spin exchange parameters and subsequent Terahertz emission from iron oxides. Preprint at http://arxiv.org/abs/1412.7094 (2014) .
  41. Mikhaylovskiy R. V., Hendry E., Kruglyak V. V., Pisarev R. V. & Kimel Th. Rasing, and A. V. Terahertz emission spectroscopy of laser-induced spin dynamics in TmFeO3 and ErFeO3 orthoferrites. Phys. Rev. B 90, 184405 (2014) . [Google Scholar]
  42. Kampfrath T. et al. Coherent Terahertz control of antiferromagnetic spin waves. Nat. Photon 5, 31–34 (2011) . [Google Scholar]
  43. Zhang G. P., Gu M. & Wu X. S. Ultrafast reduction in exchange interaction by a laser pulse: alternative path to femtomagnetism. J. Phys. Condens. Matter 26, 376001 (2014) . [DOI] [PubMed] [Google Scholar]
  44. Bakr W. S. et al. Probing the superuid-to-Mott insulator transition at the single-atom level. Science 329, 547–550 (2010) . [DOI] [PubMed] [Google Scholar]
  45. Endres M. et al. Single-site- and single-atom-resolved measurement of correlation functions. Appl. Phys. B 113, 27–39 (2013) . [Google Scholar]
  46. Struck J. et al. Tunable gauge potential for neutral and spinless particles in driven optical lattices. Phys. Rev. Lett. 108, 225304 (2012) . [DOI] [PubMed] [Google Scholar]
  47. Trotzky S. et al. Probing the relaxation towards equilibrium in an isolated strongly correlated one-dimensional bose gas. Nat. Phys. 8, 325–330 (2012) . [Google Scholar]
  48. Heyl M., Polkovnikov A. & Kehrein S. Dynamical quantum phase transitions in the transverse-field Ising model. Phys. Rev. Lett. 110, 135704 (2013) . [DOI] [PubMed] [Google Scholar]
  49. Freericks J. K., Turkowski V. M. & Zlatić V. Nonequilibrium dynamical mean-field theory. Phys. Rev. Lett. 97, 266408 (2006) . [DOI] [PubMed] [Google Scholar]
  50. Georges A., Kotliar G., Krauth W. & Rozenberg M. J. Dynamical mean-field theory of strongly correlated fermion systems and the limit of infinite dimensions. Rev. Mod. Phys. 68, 13–125 (1996) . [Google Scholar]
  51. Metzner W. & Vollhardt D. Correlated lattice fermions in d=∞ dimensions. Phys. Rev. Lett. 62, 324–327 (1989) . [DOI] [PubMed] [Google Scholar]
  52. Eckstein M. & Werner P. Nonequilibrium dynamical mean-field calculations based on the noncrossing approximation and its generalizations. Phys. Rev. B 82, 115115 (2010) . [Google Scholar]
  53. Eckstein M. & Werner P. Photoinduced states in a Mott insulator. Phys. Rev. Lett. 110, 126401 (2013) . [DOI] [PubMed] [Google Scholar]
  54. Werner P., Tsuji N. & Eckstein M. Nonthermal symmetry-broken states in the strongly interacting hubbard model. Phys. Rev. B 86, 205101 (2012) . [Google Scholar]
  55. Secchi A., Brener S., Lichtenstein A. I. & Katsnelson M. I. Non-equilibrium magnetic interactions in strongly correlated systems. Ann. Phys. 333, 221–271 (2013) . [Google Scholar]
  56. Hochbruck M. & Lubich C. On Krylov subspace ap-proximations to the matrix exponential operator. SIAM J. Numer. Anal. 34, 1911–1925 (1997) . [Google Scholar]
  57. Alvermann A. & Fehske H. High-order commutator-free exponential time-propagation of driven quantum sys-tems. J. Comp. Phys. 230, 5930–5956 (2011) . [Google Scholar]

Articles from Nature Communications are provided here courtesy of Nature Publishing Group

RESOURCES