betaFeedback

WorksPhysics and Astronomy

Coherent control of the translational and point group symmetries of crystals with light

Abstract

We use theory and first-principles calculations to explore mechanisms for control of the translational and point group symmetries of crystals in ultrafast optical experiments. We focus in particular on mechanisms that exploit anharmonic (biquadratic) lattice couplings between a driven infrared-active phonon mode and other modes at arbitrary wave vector, which are always allowed by symmetry in any space group. We use Floquet theory to develop a general phase diagram depicting the various dynamical regimes accessible to materials, with simulated dynamics to illustrate how the biquadratic coupling changes materials structure depending on both extrinsic factors (light pulse characteristics) and intrinsic materials parameters (phonon frequencies, phonon coupling strengths). We use our phase diagram, in conjunction with density functional theory calculations, both to suggest experiments to reveal hidden structural order in perovskite KTaO3_3, and to provide additional insights into recently reported experiments on SrTiO3_3 and LiNbO3_3.

Introduction

Phase transitions in crystals are often characterized in terms of symmetry changes. For example, structural phase transitions, which involve some change to the symmetry of the lattice, are ubiquitous in some classes of materials, complex oxides in particular, and have been studied extensively for many decades [1, 2, 3, 4]. Phase transitions involving broken time-reversal symmetry give rise to magnetic materials and have been similarly well-studied [2, 5]. In recent years, attention has turned to more exotic phase transitions. Broken rotational symmetries of the electronic states (but not seen in the lattice) give rise to strong correlations in electronic nematic systems [6, 7]. The relative twist between layers of stacked two-dimensional materials, such as graphene, controls the overall point group and translational symmetry of the system and gives rise to significant changes in the density of states and corresponding electronic properties, depending on the twist angle [8]. In each of these cases, detailed study of the relevant phase transitions has revealed new physical insights and advanced our fundamental understanding of condensed matter. In addition to their fundamental importance, materials that undergo phase transitions are of great practical interest because they provide opportunities for the control of functional properties with external fields. Control of the polarization with electric fields in ferroelectric materials is exploited in certain types of random access memory [9]. The key component of many sensors and actuators is a piezoelectric material [10], which exhibits large changes in its electrical polarization in response to external stress [11] (and vice versa [12]). The challenge for condensed matter and materials scientists is that the property of interest may not always couple directly with an external field.

For example, the magnetic and electronic properties of many ABO3_3 perovskites oxides (where A and B are cations and O is oxygen) are associated with lattice (phonon) modes that do not couple directly to external fields. It is possible to identify mechanisms that couple these modes by symmetry to others that are, say, electric field-controllable thereby giving indirect control of properties with an external field [13, 14, 15, 16, 17]. However, not all materials exhibit the required crystallographic symmetries for such mechanisms.

The development of bright mid-infrared and THz laser sources, capable of resonantly exciting phonons in crystals, has created new opportunities for the control of functional properties with external fields, namely light. One such mechanism involves resonantly exciting an infrared (IR)-active phonon mode of the crystal to large amplitude (QIRQ_{IR}), which induces quasi-static displacements of some Raman-active modes (QRQ_R) via anharmonic coupling of the form QIR2QRQ_{IR}^{2}Q_R. In these so-called nonlinear phononics experiments, the light pulse can induce a transition to a metastable phase with properties that are either difficult or impossible to access in the equilibrium structure at a given temperature, for example, metal-insulator phase transitions [18], superconductivity [19, 20], and changes in orbital ordering [21]. In most of the experiments that have been reported so far, the optically excited IR mode couples to Raman modes that are also at the Brillouin zone center, such that the induced metastable phase has the same translational symmetry as the ground-state structure. Is it possible to identify other anharmonic lattice couplings, involving phonon modes at non-zero wave vector, which would allow us to dynamically stabilize phases with a translation symmetry different to that of the parent ground-state structure?

In this work, we use theory and first-principles calculations to explore and elucidate anharmonic lattice coupling pathways involving a zone-center IR-active mode and modes at arbitrary wave vector, Qq⃗Q_{\vec{q}}. We show that biquadratic coupling between these modes, QIR2Qq⃗2Q_{IR}^{2}Q_{\vec{q}}^{2}, which is always allowed by symmetry in any space group, can be exploited in ultrafast optical experiments such that resonant excitation of an IR-active mode at the zone center can dynamically induce phases with translational symmetry different to that of the ground state through coupling to and condensation of modes at arbitrary (non-zero) wave vector. This biquadratic coupling potentially offers a new pathway through which the functional properties of materials can be dynamically controlled. We use classical Floquet theory to map out a general phase diagram showing the various dynamical regimes that may be accessed with this biquadratic coupling in modern ultrafast optical experiments; this phase diagram is our key result and is shown in Figure 1. Our approach highlights how the nonlinear phononics mechanism can be viewed as a specific example of a more general parametric amplification process and reveals dynamical regimes that may be missed in the time-averaged theories commonly used to study these processes. We show how different parts of the phase diagram can be accessed by tuning the light polarization, frequency and peak electric field (extrinsic experimental parameters), and the relative frequencies and strength of the coupling between the pumped IR and coupled modes (determined by intrinsic materials parameters, such as the crystal structure and chemical composition). The phase diagram thus functions as a tool that can be used to interpret existing experiments (such as recent work demonstrating transient – that is, not long-lasting or deterministic – polarization switching in LiNbO3_3 [22]) and to design future experiments. We use our phase diagram, in combination with first-principles DFT calculations, to both interpret the results of a recently reported experiment on SrTiO3_3 showing light control of translational symmetry [23], and to suggest future experiments that can access hidden structural order in perovskite KTaO3_3. Our work goes further than previously published studies by exploring a much greater range of dynamical regimes within a single coherent framework.

Dynamical regimes accessible through biquadratic coupling between an optically pumped IR-active phonon and a mode at arbitrary wave vector

Figure 1. Dynamical regimes accessible through biquadratic coupling between an optically pumped IR-active phonon, QIRQ_{\mathrm{IR}}, and a mode at arbitrary wave vector, Qq~Q_{\tilde{q}}. The Floquet phase diagram (center) is shown as a function of the frequency ratio of the biquadratically coupled mode (ωq~\omega_{\tilde{q}}) to the excited IR mode (ωIR\omega_{\mathrm{IR}}) versus the driving strength ϵ∝DIR,q~QIR2\epsilon\propto D_{\mathrm{IR},\tilde{q}}Q_{\mathrm{IR}}^2 (see full definition below Eqn. (6)). The simulated dynamics, structural changes, and schematic mechanisms are shown in the color-coordinated panels. When Qq~Q_{\tilde{q}} has a positive force constant (real frequency) at equilibrium (top-half, ωq~/ωIR>0\omega_{\tilde{q}}/\omega_{\mathrm{IR}} > 0), trivially damped motion is expected in much of the phase diagram (I – white). For negative biquadratic coupling (DIR,q~<0D_{\mathrm{IR},\tilde{q}} < 0), Qq~Q_{\tilde{q}} can be frozen in by optical excitation of an IR phonon, leading to novel transient structural phases (II – gray). Near ωq~/ωIR=1\omega_{\tilde{q}}/\omega_{\mathrm{IR}} = 1, parametric oscillation of Qq~Q_{\tilde{q}} is possible (III – blue). In the purple region of the phase diagram (Region IV) we imagine a high-symmetry reference structure for the material of interest, which may be virtual, located at the saddle point (see inset). Qq~Q_{\tilde{q}} has a negative force constant in this phase and drives a non-transient structural transition to a lower-symmetry phase. In the low-symmetry phase Qq~Q_{\tilde{q}} can be displaced quasi-statically by optical excitation of QIRQ_{\mathrm{IR}}; this is the conventional nonlinear phononics effect. For positive coupling between Qq~Q_{\tilde{q}} and the IR mode (DIR,q~>0D_{\mathrm{IR},\tilde{q}} > 0), the symmetry broken by Qq~Q_{\tilde{q}} can be (re)introduced into the structure (V – green), beyond which Qq~Q_{\tilde{q}} again parametrically oscillates.

Theoretical Model

Simple model of biquadratic coupling

We start by writing down a general equation for describing the response of a centrosymmetric crystal to resonant excitation of an IR-active phonon mode by a short, intense mid-IR pulse. We consider the case where the dominant dynamics induced are those of the excited IR mode, QIRQ_{IR}, and another mode coupled to it at arbitrary wave vector q⃗\vec{q}, Qq⃗Q_{\vec{q}}. The lattice potential energy is assumed to be:

U=12KIRQIR2+12Kq⃗Qq⃗2+DIR,q⃗QIR2Qq⃗2+14Dq⃗Qq⃗4−ΔP⃗⋅E⃗,(1)U = \frac{1}{2}K_{IR}Q_{IR}^{2} + \frac{1}{2}K_{\vec{q}}Q_{\vec{q}}^{2} + D_{IR,\vec{q}}Q_{IR}^{2}Q_{\vec{q}}^{2} + \frac{1}{4}D_{\vec{q}}Q_{\vec{q}}^{4} - \vec{\Delta P}\cdot\vec{E}, \tag*{(1)}

where KIRK_{IR} and Kq⃗K_{\vec{q}} are the force constants of the relevant phonon modes at harmonic order, Dq⃗D_{\vec{q}} is the fourth-order force constant for Qq⃗Q_{\vec{q}}, and DIR,q⃗D_{IR,\vec{q}} is the biquadratic coupling coefficient coupling QIRQ_{IR} and Qq⃗Q_{\vec{q}}. The polarization change in the crystal due to the electric field E⃗\vec{E} of the light pulse is given to linear order by ΔP⃗=Z~∗QIR\Delta\vec{P} = \tilde{Z}^{*}Q_{IR}, where Z~∗\tilde{Z}^{*} is the mode-effective charge of the excited IR mode (as defined in [24]). In previous work, we showed that higher-order terms in the polarization are important for understanding changes in the optical properties of materials due to optical excitation [25]; we ignore those terms in this work to simplify the analysis, and because here we are focused on understanding structural changes due to optical excitation. Also note that we ignore explicit electron-phonon interactions – the electrons adiabatically follow the excited phonon coordinates and the equations of motion are classical. Recent work [26] on the quantum theory of lattice dynamics in the presence of external driving fields has validated this classical approach; the framework developed here could also be adapted to incorporate the quantum theory of Ref. 26.

As mentioned above, much of the previous work on nonlinear phononics has focused on systems where the excited IR mode is coupled to Raman-active modes at the zone center through a term of the form QIR2QRQ_{IR}^{2}Q_{R}. Whether or not such a term is an allowed invariant in the lattice potential is determined by crystallographic symmetry and hence only some materials are candidates for nonlinear phononics experiments that exploit this type of cubic anharmonic coupling. In addition, since the term QIR2QRQ_{IR}^{2}Q_{R} involves modes at the zone center only, there is no change in the translational symmetry of the crystal upon optical excitation. In contrast, the biquadratic term QIR2Qq⃗2Q_{IR}^{2}Q_{\vec{q}}^{2} shown in Eqn. (1) is always allowed by symmetry between all modes and in all space groups (biquadratic coupling between QIRQ_{IR} and a phonon at q⃗=0⃗\vec{q} = \vec{0} has been explored in previous work [27]). How can this biquadratic coupling be exploited for light control of properties? By rearranging terms in Eqn. (1), we see that the effective force constant of Qq⃗Q_{\vec{q}} (K~q⃗\tilde{K}_{\vec{q}}) is renormalized by the motion of the IR mode:

K~q⃗=Kq⃗+2DIR,q⃗QIR2.(2)\widetilde{K}_{\vec{q}} = K_{\vec{q}} + 2D_{IR,\vec{q}}Q_{IR}^{2}. \tag*{(2)}

Displacing the IR mode with DIR,q⃗>0D_{IR,\vec{q}}>0 increases the force constant K~q⃗\widetilde{K}_{\vec{q}}, thereby stiffening this mode. In contrast, when DIR,q⃗<0D_{IR,\vec{q}}<0 the effective force constant K~q⃗\widetilde{K}_{\vec{q}} decreases. If the IR mode displacement and the magnitude of the biquadratic coupling are sufficiently large compared to the equilibrium force constant Kq⃗K_{\vec{q}} (that is, when Kq⃗≤2DIR,q⃗QIR2K_{\vec{q}}\leq2D_{IR,\vec{q}}Q_{IR}^{2}), then Qq⃗Q_{\vec{q}} will ‘freeze in’, resulting in a new structural phase with symmetry broken by QIRQ_{IR} and Qq⃗Q_{\vec{q}}. Exciting IR modes to large amplitude is now possible with current bright mid-IR and THz sources, but the IR motion is oscillatory. Can oscillatory motion of the IR mode induce new transient structural phases via biquadratic coupling to Qq⃗Q_{\vec{q}}? In what follows, we use Floquet theory, first-principles calculations, and dynamical simulations to uncover new dynamical regimes accessible through the biquadratic coupling in Eqn. (1).

Floquet Theory

Consider the equations of motion that are derived from Eqn. (1) by taking derivatives with respect to the mode coordinates, QIRQ_{IR} and Qq⃗Q_{\vec{q}}. We find,

Q¨IR+2γIRQ˙IR+1MIR(KIR+2DIR,q⃗Qq⃗2)QIR=Z~∗MIRE,Q¨q⃗+2γq⃗Q˙q⃗+1Mq⃗(K~q⃗+2DIR,q⃗QIR2)Qq⃗+Dq⃗Mq⃗Qq⃗3=0.(3)\begin{aligned} \ddot{Q}_{IR}+2\gamma_{IR}\dot{Q}_{IR}+\frac{1}{M_{IR}}\left(K_{IR}+2D_{IR,\vec{q}}Q_{\vec{q}}^{2}\right)Q_{IR}&=\frac{\widetilde{Z}^{*}}{M_{IR}}E,\\ \ddot{Q}_{\vec{q}}+2\gamma_{\vec{q}}\dot{Q}_{\vec{q}}+\frac{1}{M_{\vec{q}}}\left(\widetilde{K}_{\vec{q}}+2D_{IR,\vec{q}}Q_{IR}^{2}\right)Q_{\vec{q}}+\frac{D_{\vec{q}}}{M_{\vec{q}}}Q_{\vec{q}}^{3}&=0. \tag*{(3)} \end{aligned}

To simplify the analysis we ignore all terms that are higher than harmonic order except for the biquadratic coupling and reintroduce the Dq⃗Qq⃗4D_{\vec{q}}Q_{\vec{q}}^{4} term only when necessary to guarantee a finite minimum in UU with respect to Qq⃗Q_{\vec{q}}. Additionally, we assume that the pumped IR mode is undamped (γIR=0\gamma_{IR}=0), thereby allowing for the exploration of the long timescale dynamics. We justify this assumption a posteriori by numerical exploration of the dynamics, finding that our simulated short-timescale dynamics of Qq⃗Q_{\vec{q}} are comparable to the Floquet results (which assume continuous periodic driving).

After resonant excitation of the IR mode with a Gaussian pulse we find sinusoidal periodic motion for QIRQ_{IR} of the form

QIR(t)=2Z~∗E~KIRcos⁡(ωIRt+ϕ).(4)Q_{IR}(t)=2\frac{\widetilde{Z}^{*}\widetilde{E}}{K_{IR}}\cos\left(\omega_{IR}t+\phi\right). \tag*{(4)}

Here E~\widetilde{E} is defined as ηE0τfIR\eta E_{0}\tau f_{IR}, where E0E_{0} is the peak electric field, τ\tau is the full-width at half-maximum of the electric field, fIRf_{IR} is the (linear) IR frequency, and η\eta characterizes the shape of the pulse (for a Gaussian pulse η=(π/2)32ln⁡(2)≈1.67\eta=\sqrt{\frac{(\pi/2)^{3}}{2\ln(2)}}\approx1.67). Then ητfIR\eta\tau f_{IR} measures the number of cycles the IR mode is driven by the electric field of the light pulse and the peak IR displacement is QIR,0=2Z~∗E~KIRQ_{IR,0}=2\frac{\widetilde{Z}^{*}\widetilde{E}}{K_{IR}}. We neglect ϕ\phi in what follows. QIRQ_{IR} is therefore a periodic driver for the general lattice motion. This can be contrasted with other recent Floquet approaches, where off-resonant periodic excitation of the electronic states is motivating the Floquet approach and enabling access to novel nonequilibrium phases [28, 29, 30].

Eqn. (4) assumes that the biquadratic coupling between QIRQ_{IR} and Qq⃗Q_{\vec{q}} is zero (DIR,q⃗=0D_{IR,\vec{q}}=0). This assumption can be justified from a perturbation theory perspective, since Qq⃗Q_{\vec{q}} is initially characterized by small fluctuations about zero amplitude. However, it obviously will not hold when Qq⃗Q_{\vec{q}} is displaced to large amplitudes. In this scenario, the force constant of the IR mode can also be renormalized by Qq⃗Q_{\vec{q}}, ΔKIR≈2DIR,q⃗Qq⃗2\Delta K_{IR}\approx2D_{IR,\vec{q}}Q_{\vec{q}}^{2}, such that the IR mode also freezes in. Using Eqn. (4) for QIRQ_{IR} in the second line of Eqn. (3), rescaling time to θ=ωIRt\theta=\omega_{IR}t, and collecting terms, we find the following form,

ddθx(θ)=A(θ)x(θ),(5)\frac{d}{d\theta}x(\theta)=A(\theta)x(\theta), \tag*{(5)}

where we have defined the vector x(θ)=(Qq⃗,ddθQq⃗)x(\theta)=\left(Q_{\vec{q}},\frac{d}{d\theta}Q_{\vec{q}}\right) and the matrix

A(θ)=(01−(δ+ϵ+ϵcos⁡(2θ))−ν).(6)A(\theta)=\begin{pmatrix}0&1\\-\left(\delta+\epsilon+\epsilon\cos(2\theta)\right)&-\nu\end{pmatrix}. \tag*{(6)}

The dimensionless parameters δ=1ωIR2Kq⃗Mq⃗=ωq⃗2ωIR2\delta=\frac{1}{\omega_{IR}^{2}}\frac{K_{\vec{q}}}{M_{\vec{q}}}=\frac{\omega_{\vec{q}}^{2}}{\omega_{IR}^{2}}, ϵ=DIR,q⃗QIR,02Mq⃗ωIR2\epsilon=\frac{D_{IR,\vec{q}}Q_{IR,0}^{2}}{M_{\vec{q}}\omega_{IR}^{2}} (see Appendix A for unit conversions), and ν=2γq⃗ωIR\nu=\frac{2\gamma_{\vec{q}}}{\omega_{IR}} measure the square-frequency ratio, the driving strength, and effective damping for the driven mode Qq⃗Q_{\vec{q}} due to the motion of QIRQ_{IR}, respectively. The matrix A(θ)A(\theta) is periodic, with period π\pi (that is, A(θ+π)=A(θ)A(\theta+\pi)=A(\theta)) and therefore represents a Hill’s equation. The solutions to this equation (which may not be analytic), as well as how they depend on the parameters δ\delta, ϵ\epsilon and ν\nu, can be obtained using the standard techniques of Floquet theory [31, 32].

Our primary interest is to find solutions to Eqn. (5) that predict exponential growth in Qq⃗Q_{\vec{q}}. These represent dynamical regimes in which optical pumping of an IR phonon induces large-amplitude changes in Qq⃗Q_{\vec{q}}, thereby driving a transient structural phase transition that could be resolved experimentally. We use the main results of our Floquet analysis to create a phase diagram, shown in the central panel of Figure 1, from Eqn. (5) depicting these dynamical regimes (see Appendix B for more details).

First-Principles Calculations and Dynamical Simulation

To connect the Floquet theory to real materials and experiments we find parameters for Eqns. (3) & (6) using density functional theory (DFT). Calculations were performed using VASP 6.2.0 [33, 34, 35], using the projector augmented-wave (PAW) method in the local-density approximation (LDA) [36]. The following states were included in the valence of the relevant PAW potentials: 3s23p64s13s^{2}3p^{6}4s^{1} for K, 5p65d46s15p^{6}5d^{4}6s^{1} for Ta, 4s24p65s24s^{2}4p^{6}5s^{2} for Sr, 3s23p63d34s13s^{2}3p^{6}3d^{3}4s^{1} for Ti, and 2s22p42s^{2}2p^{4} for O. A force convergence tolerance of 10−310^{-3} eV/Å was used for all calculations with a 4×4×44 \times4 \times4 Monkhorst-Pack k-point grid (for a 2×2×22 \times2 \times2 formula unit supercell of the primitive cubic perovskite unit cell) and plane-wave energy cutoffs of 600 eV (KTaO3_3) and 700 eV (SrTiO3_3). These values were chosen to converge phonon frequencies, calculated with density functional perturbation theory [37] (DFPT), to within 5 cm−1^{-1} when compared to incrementing the k-point grid to 8×8×88 \times8 \times8, and the energy cutoff to 800 eV for both KTaO3_3 and SrTiO3_3. Our converged cubic lattice constants for KTaO3_3 (3.959 Å) and SrTiO3_3 (3.859 Å) are underestimated compared to the experimental lattice constants [38, 39, 40] of 3.988 Å and 3.905 Å, respectively, as expected in LDA.

We focus on pairs of IR modes and modes characterized by arbitrary wave vector that exhibit strong coupling at biquadratic order. To find these strongly coupled mode pairs, we calculated a series of phonon dispersion curves for structures in which an IR-active mode has been ‘frozen in’ at varying amplitude. Modes that exhibit large changes in their frequencies as a function of IR mode amplitude were selected for further study. Quadratic, biquadratic, and quartic coupling coefficients were calculated via a series of symmetry-constrained frozen-phonon calculations. Mode-effective charges were calculated as defined in Ref. [24], which were found to be consistent with values obtained from modern theory of polarization calculations [41, 42, 43] conducted on +/-10 pm meshes of the IR-active phonons. Non-equilibrium phonon dispersion curves were calculated with Quantum Espresso (plane-wave energy cutoff of 70 Ry, 8×8×88 \times8 \times8 Monkhorst-Pack k-point grid in the 5-atom cubic unit cell of KTaO3_3), using Garrity-Bennett-Rabe-Vanderbilt pseudopotentials [44, 45], which confirmed the qualitative features of the induced dynamical instabilities, i.e. which modes develop imaginary frequencies when the IR mode amplitude is increased along a fixed polarization direction beyond a critical amplitude. These dispersion curves (Figure 2) were calculated on a regular 10×10×1010 \times10 \times10 q-point grid, with the inclusion of a simple acoustic sum rule. Symmetry assignment of structural phases and irreducible representations of phonon modes were generated with the ISOTROPY Software Suite [46, 47].

Phonon dispersion curves for KTaO3

Figure 2. Phonon dispersion curves from our DFT calculations for KTaO3_3 in the cubic perovskite structure at equilibrium (left) and with the 16.7 THz IR phonon displaced along [111] with an amplitude of 25 pm in the 5 atom cell (right). The IR phonon displacement splits and softens a branch at the R-point, leading to a driven instability of this mode. Modes with imaginary frequencies are depicted as having negative frequencies.

To explore the transient dynamics of the regions found in the Floquet analysis with numerical simulation of Eqns. (3) & (6), we use a Runge-Kutta 5(4) method for numerical integration, as implemented in NumPy [48]. Simulations were performed with Gaussian electric field pulses with duration τ=500\tau= 500 fs, varied peak electric fields (E0E_0), IR phonon frequency, and peak field time set at t=0t = 0. First-principles calculations of the damping parameters γIR/q⃗\gamma_{IR/\vec{q}} are computationally expensive. As a result, we have explored the dynamics as a function of damping parameter, ranging from 0 THz to 1 THz, showing only selected dynamics to highlight the response in different regions of the phase diagram. For initial conditions, QIRQ_{IR} is taken to be at rest long before the Gaussian pulse is present, and Qq⃗Q_{\vec{q}} is given an amplitude of 0.1 pm to simulate a weak fluctuation about the average equilibrium value. Note that all parameters are included when simulating Eqn. (3), in contrast to the numerical solution of Eqn. (6) where approximations have been made (see discussion in Sec. IIB).

Results and Discussion

Floquet phase diagram and dynamics

The phase diagram (center-panel of Figure 1) shows different dynamical regimes for Qq⃗Q_{\vec{q}}. The vertical axis is the ratio of frequencies δ=ωq⃗/ωIR\sqrt{\delta} = \omega_{\vec{q}}/\omega_{IR}, while the horizontal axis is ϵ\epsilon, the driving strength parameter (DIR,q⃗QIR,02/Mq⃗ωIR2)(D_{IR,\vec{q}}Q_{IR,0}^{2}/M_{\vec{q}}\omega_{IR}^{2}). The sign of the anharmonic coupling DIR,q⃗D_{IR,\vec{q}} is an intrinsic material property that dictates which side of the phase diagram is accessible – positive coupling increases K~q⃗\tilde{K}_{\vec{q}} (right) and negative coupling decreases K~q⃗\tilde{K}_{\vec{q}} (left).

While the Floquet analysis informs us of regions of exponential growth or decay of Qq⃗Q_{\vec{q}}, it does not provide compact analytic results of the phase boundaries or dynamics in a given region that might inform future ultrafast experiments. We anticipate this need and include approximate analytic results, expressed in terms of microscopic materials parameters, in what follows and in Appendix § 1.1.3.

In the subsections below we discuss the basic physical mechanism in each region, followed by a materials example, and finally show the simulated dynamics of Qq⃗Q_{\vec{q}}. This follows the layout of the color-coded panels extending from the phase diagram in Figure 1. Where appropriate we discuss previous theoretical and experimental work. We start with the top half of the phase diagram where the high-symmetry equilibrium structure is stable (δ>0\sqrt{\delta} > 0).

Region I – Trivially damped motion

For most of the upper-half of the phase diagram, trivial exponential decay is expected, suggesting that optically excited IR modes will not induce large amplitude dynamical behavior in Qq⃗Q_{\vec{q}}. Due to the exponentially decaying response of QIRQ_{IR} and Qq⃗Q_{\vec{q}}, a long-lived change in crystal symmetry is not expected. Recall that the biquadratic coupling pathway is allowed for all modes at all wave vectors. For weak driving of an IR phonon, our expectation is that most, or all of the coupled modes will be in this region. As can be seen from Figure 1, a combination of external and intrinsic materials parameters (ϵ\epsilon) are needed to move one, or several, modes into another region of the phase diagram. We now focus on negative biquadratic coupling (ϵ<0\epsilon< 0) and small δ\sqrt{\delta} where, as mentioned in § 1.1 A, a strong enough drive can decrease K~q⃗\widetilde{K}_{\vec{q}} such that Qq⃗Q_{\vec{q}} freezes into the structure with a nonzero amplitude (Region II).

Region II – Access to new phases: Novel exponential growth and symmetry control

Region II describes a region in which we expect Qq⃗Q_{\vec{q}} to grow exponentially, thereby inducing a transient structural phase transition. Recent experimental work [23] has shown that optical pumping of an IR phonon in the cubic phase of the perovskite SrTiO3_3 at 135 K can induce a phonon mode that involves rotations of the TiO6_6 octahedra, doubling the size of the unit cell and changing the translational symmetry. SrTiO3_3 undergoes this structural phase transition with temperature at about 110 K – optical excitation of an IR phonon effectively increases the transition temperature. As far as we are aware, this is the first experimental realization of this dynamical region. We show that in addition to changing the transition temperatures for structural phase transitions, it is also possible to induce structural phases that are not present at all in the equilibrium phase diagram with temperature.

Taking Qq⃗(θ)∝eμθQ_{\vec{q}}(\theta) \propto e^{\mu\theta}, with μ\mu as a dimensionless exponential growth parameter (μωIR\mu\omega_{IR} is the dimensional growth parameter), inserting this functional form in Eqn. (5), and neglecting the cos⁡(2θ)\cos(2\theta) term (which is equivalent to time averaging), we find the following relation

μ2+νμ+(δ+ϵ)=0,(7)\mu^2 + \nu\mu+ (\delta+ \epsilon) = 0, \tag*{(7)}

which has the exponential growth solution μII,+=12(−ν+ν2−4(δ−∣ϵ∣))\mu_{II,+} = \frac{1}{2}\left(-\nu+ \sqrt{\nu^2 - 4(\delta- |\epsilon|)}\right) in Region II.

This shows that exponential growth of Qq⃗Q_{\vec{q}} in Region II is affected by damping, as well as the pulse characteristics through ϵ\epsilon. The phase boundary between Region I and Region II is defined by μ=0\mu= 0, which is solved when ∣ϵ∣=δ|\epsilon| = \delta. This result is surprisingly independent of the damping ν\nu, suggesting that although damping influences the growth rate of Qq⃗Q_{\vec{q}} once in Region II it does not influence the location of the phase boundary. Comparing this analytic result with that from the Floquet analysis, we find that ∣ϵ∣=δ|\epsilon| = \delta is a good approximation for the phase boundary. The analytic result works well for small δ\delta, i.e. when ωq⃗≪ωIR\omega_{\vec{q}} \ll\omega_{IR}, but may slightly underestimate the value of ϵ\epsilon needed to reach this regime (Supplementary Figure S1). This is attributed to neglecting the oscillatory component of the IR motion in the development of Eqn. 7.

Floquet phase diagram with approximate analytic results for the phase boundaries

Figure S1. Floquet phase diagram with approximate analytic results for the phase boundaries (red, dashed). The analytic expressions can be found in the main text below Eqn. (7) (∣ϵ∣=δ|\epsilon|=\delta), in § 1.2 A 5 (ϵ=∣δ∣\epsilon=|\delta|), and in Appendix § 1.1.3.

To find the growth rate of Qq⃗Q_{\vec{q}} in terms of microscopic parameters we need to rescale μII,+\mu_{II,+} by ωIR\omega_{IR} to account for the definition of dimensionless time θ=ωIRt\theta= \omega_{IR}t. The growth rate then becomes,

μII,+ωIR=−γq⃗+γq⃗2−(ωq⃗2−∣DIR,q⃗∣Mq⃗QIR,02).(8)\mu_{II,+}\omega_{IR} = -\gamma_{\vec{q}} + \sqrt{\gamma_{\vec{q}}^2 - \left(\omega_{\vec{q}}^2 - \frac{|D_{IR,\vec{q}}|}{M_{\vec{q}}}Q_{IR,0}^2\right)}. \tag*{(8)}

Notice that on the phase boundary the terms in the parentheses of Eqn. (8) cancel, as a result μII,+=0\mu_{II,+} = 0. We can express the phase boundary in terms of microscopic parameters and pulse characteristics to derive a threshold field, which takes the form

E0τ≥12KIRηfIRZ∗Kq⃗∣DIR,q⃗∣.(9)E_0\tau\ge\frac{1}{2}\frac{K_{IR}}{\eta f_{IR} Z^*}\sqrt{\frac{K_{\vec{q}}}{|D_{IR,\vec{q}}|}}. \tag*{(9)}

In this analysis, if E0τE_0\tau satisfies this inequality, Qq⃗Q_{\vec{q}} grows exponentially, inducing a transient structural phase that changes the point group and translational symmetry of the crystal.

Beyond the threshold electric field, we can approximate the induced amplitude of Qq⃗Q_{\vec{q}} and its renormalized frequency by time-averaging Eqn. (1). From this point of view, the curvature of the average potential energy landscape is decreasing and becoming negative with respect to Qq⃗Q_{\vec{q}}. That is, a minimum in the potential energy landscape develops beyond the threshold electric field in which Qq⃗≠0Q_{\vec{q}} \ne0. This new minimum survives the oscillatory 2ωIR2\omega_{IR} component of the IR motion (Figure 1, Region II). The motion of the IR phonon renormalizes both the amplitude of Qq⃗Q_{\vec{q}},

Qq⃗,±∗(QIR)=±∣Kq⃗+2DIR,q⃗⟨QIR2⟩∣/Dq⃗,(10)Q_{\vec{q},\pm}^{*}(Q_{IR}) = \pm\sqrt{\left|K_{\vec{q}} + 2D_{IR,\vec{q}}\langle Q_{IR}^{2}\rangle\right|/D_{\vec{q}}}, \tag*{(10)}

and its frequency,

ω∗(QIR)=2∣Kq⃗+2DIR,q⃗⟨QIR2⟩∣/Mq⃗,(11)\omega^{*}(Q_{IR}) = \sqrt{2\left|K_{\vec{q}} + 2D_{IR,\vec{q}}\langle Q_{IR}^{2}\rangle\right|/M_{\vec{q}}}, \tag*{(11)}

where we have explicitly shown the dependence of the new minimum and frequency on the IR phonon amplitude.

The condensation of Qq⃗Q_{\vec{q}} to nonzero amplitude modulates the existing crystal structure with a lengthscale derived from the wave vector q⃗\vec{q}. If q⃗=ζ1b⃗1+ζ2b⃗2+ζ3b⃗3\vec{q} = \zeta_1\vec{b}_1+\zeta_2\vec{b}_2+\zeta_3\vec{b}_3 where b⃗i\vec{b}_i define the reciprocal lattice vectors of a lattice defined by a⃗i\vec{a}_i so that a⃗i⋅b⃗j=2πδij\vec{a}_i\cdot\vec{b}_j = 2\pi\delta_{ij}, the condensation of Qq⃗Q_{\vec{q}} transiently induces a new periodicity to the crystal with modulation vector λ⃗=ζ1a⃗1+ζ2a⃗2+ζ3a⃗3\vec{\lambda} = \zeta_1\vec{a}_1 + \zeta_2\vec{a}_2 + \zeta_3\vec{a}_3, where each ζi\zeta_i may lead to (in)commensurability with its equilibrium lattice vector a⃗i\vec{a}_i.

We now turn to the example of KTaO3_3, one of the few perovskites that remains cubic (Pm3ˉmPm\bar{3}m space group) at all temperatures; the 0 K phonon dispersion curve in Figure 2 shows all modes with real frequencies. KTaO3_3 features three sets of triply degenerate IR-active optical phonons and no Raman-active phonons. As a result, new structural phases cannot be dynamically induced by relying on the conventional QIR2QRQ_{\mathrm{IR}}^2Q_{\mathrm{R}} coupling. In contrast, every phonon at every wave vector is accessible through the biquadratic coupling shown in Eqn. (1). The challenge now is to find which modes couple most strongly to an IR active phonon with a frequency that is accessible in a modern ultrafast optical experiment.

We explore this by calculating nonequilibrium phonon dispersion curves as a function of all IR phonons polarized along the [100], [110], and [111] crystallographic directions. The nonequilibrium phonon dispersion curves for the highest frequency (16.7 THz) IR-active phonon displaced along the [111] direction, which we will use as a representative example, are shown in Figure 2. As the amplitude of the IR phonon increases, we see a branch split at the R-point, with one mode softening and becoming unstable, that is, its frequency becomes imaginary (or its force constant becomes negative). Inspection of the displacement pattern and symmetry of this mode shows that it is an out-of-phase octahedral tilt (transforming as the irreducible representation R4+R_4^+) about the [111] axis (Region II of Figure 1). Octahedral tilts are common structural distortions in perovskites [1, 49] but are not present in KTaO3_3 at any temperature. Here we show that a transient structural phase of KTaO3_3 involving octahedral tilts can be induced by biquadratic coupling to an optically excited IR-active phonon.

Since the mode that freezes in is at the Brillouin zone edge and has wave vector QR=2πa(12,12,12)Q_{\mathrm{R}} = \frac{2\pi}{a}\left(\frac{1}{2},\frac{1}{2},\frac{1}{2}\right) (where aa is the cubic lattice parameter) the unit cell of the cubic phase is doubled to accommodate the octahedral tilt pattern. The transient unit cell lattice parameters (a⃗i′\vec{a}_i') are given by a⃗1′=a⃗2+a⃗3\vec{a}_1' = \vec{a}_2 + \vec{a}_3, a⃗2′=a⃗1+a⃗3\vec{a}_2' = \vec{a}_1 + \vec{a}_3, and a⃗3′=a⃗1+a⃗2\vec{a}_3' = \vec{a}_1 + \vec{a}_2, where the unprimed quantities denote the equilibrium lattice parameters and a⃗1=a⃗2=a⃗3\vec{a}_1 = \vec{a}_2 = \vec{a}_3 since KTaO3_3 is cubic at equilibrium. This shows that light-induced structural phases can change the translational symmetry of crystals on ultrafast timescales. In Sec. IIIB we show that the changes to point group and translational symmetry are polarization dependent, allowing for unprecedented control of crystalline symmetry.

Using the DFT-derived parameters in Table I we can estimate the critical electric field (Eqn. (9)) needed to enter Region II of the phase diagram and induce a new structural phase in cubic KTaO3_3. For a 500 fs Gaussian electric field pulse with 16.7 THz carrier frequency, we find a threshold field of 4.8 MV/cm. We note that this value ignores Fresnel reflection, higher-order anharmonicities, and damping.

We explore the dynamics for the excitation of the 16.7 THz IR active phonon polarized along the [111] direction as a function of peak electric field in the Region II panel inset of Figure 1. The frequency ratio (δ=0.24\sqrt{\delta} = 0.24) of the two phonons involved places us on the gray arrow originating on the vertical axis in the phase diagram and pointing into the gray region. A peak electric field of 3.7 MV/cm (ϵ=−0.033\epsilon= -0.033) is not strong enough to overcome the intrinsic restoring force such that any transient motion imparted on QqˉQ_{\bar{q}} is damped away, hence the amplitude of QqˉQ_{\bar{q}} is zero. That is, the threshold from Region I to Region II is not crossed. By increasing the peak electric field to 5.3 MV/cm (ϵ=−0.066\epsilon= -0.066), just beyond the critical field, a new energy minimum develops for QqˉQ_{\bar{q}} at Qqˉ,+∗Q_{\bar{q},+}^{*} (Eq. (10)). Because the critical condition has just been reached, the effective potential is shallow and the new resonant frequency ω∗\omega^{*} (Eq. (11)) is small. Increasing the peak electric field further to 6.4 MV/cm (ϵ=−0.099\epsilon= -0.099) further establishes the Region II behavior. The renormalized frequency ω∗\omega^{*} has increased as expected, the location of the new energy minimum (Qqˉ,+∗Q_{\bar{q},+}^{*}) is at a larger amplitude, and the enhanced growth rate (μII,+\mu_{II,+}) towards the new phase is visually apparent. This is a consequence of the field-dependent growth rate in Eqn. (8), giving access to the new structural phase ≈0.5\approx0.5 ps earlier compared to the 5.3 MV/cm peak electric field. We expect the threshold values of E0τE_{0}\tau (Eqn. (9)) to be accessible with current mid-IR laser sources at 16.7 THz [50, 51].

QuantityQIRQ_{IR}QqˉQ_{\bar{q}} Reg IIQqˉQ_{\bar{q}} Reg III
Frequencyff[THz]16.733.9815.19
Force constantKK[eV/Å2^2]18.361.0415.16
Reduced massMM[u]16.0316.0016.06
Mode-effective chargeZ~∗\tilde{Z}^{*}[e−^{-}]13.03--
Biquadratic couplingDIR,qˉD_{IR,\bar{q}}[eV/Å4^4]--1.1210.00

Table 1. Characteristics of the high-frequency IR mode (QIRQ_{IR}) and biquadratically coupled RR-point modes (QqˉQ_{\bar{q}}) in the cubic phase of KTaO3_3 from our DFT calculations needed to realize dynamical regimes in Region II and Region III. The IR mode is polarized along [111] in Region II and along [100] in Region III. In Region III, the excited IR mode and QqˉQ_{\bar{q}} are very strongly coupled and DIR,qˉD_{IR,\bar{q}} varies from 10–20 eV/Å4^4, depending on the precise amplitudes of QIRQ_{IR} and QqˉQ_{\bar{q}} used to calculate it. We used DIR,qˉ=10D_{IR,\bar{q}} = 10 eV/Å4^4 for our dynamical simulations. The units of reduced mass are in atomic mass units, u.

Region III – Parametric Oscillation

In the previous section, driving an IR-active phonon condensed a single phonon, QqˉQ_{\bar{q}}, transiently altering the crystal symmetry. In complex crystals there are many modes of arbitrary wave vectors, all of which are accessible to the biquadratic coupling. As a result, we expect the dynamical response of the crystal to resonant IR-phonon drive to be rich with complex dynamical motion on many length scales involving many phonons. In this section, we show that parametrically driven oscillatory motion of other QqˉQ_{\bar{q}}’s is also expected, adding a complexity of unexplored time scales and microscopic motion to the dynamical response of the crystal.

We highlight the parametric oscillation of QqˉQ_{\bar{q}} through the optical excitation of an IR-active phonon in the upper right of the phase diagram (positive ϵ\epsilon, blue region). Near ϵ=0\epsilon= 0, this region grows from δ=1\delta= 1, the oscillatory component of the driver is twice the fundamental frequency of QqˉQ_{\bar{q}} (cosine term in Eqn. (6)). This suggests that Region III describes parametric oscillation of QqˉQ_{\bar{q}}. We therefore expect rapidly growing oscillatory motion of QqˉQ_{\bar{q}} in this region. Derivation of an approximate phase boundary, exponential growth rate, and the effect of intrinsic damping are included in Appendix C, following standard approaches to solutions in dynamical systems [52].

QuantityQIRQ_{IR}QqˉQ_{\bar{q}} Reg IVQqˉQ_{\bar{q}} Reg V
Frequencyff[THz]5.152.68ii2.68ii
Force constantKK[eV/Å2^2]6.25-0.47-0.47
Reduced massMM[u]57.6016.0016.00
Mode-effective chargeZ~∗\tilde{Z}^{*}[e−^{-}]11.90--
Biquadratic couplingDIR,qˉD_{IR,\bar{q}}[eV/Å4^4]--0.290.85

Table 2. Characteristics of the IR mode (QIRQ_{IR}) and biquadratically coupled RR-point mode (QqˉQ_{\bar{q}}) in the cubic phase of SrTiO3_3 from our DFT calculations needed to realize dynamical regimes in Region VI and Region V. The IR mode is polarized along [110] in Region VI and along [1] in Region V. The units of reduced mass are in atomic mass units, u.

To give a quantitative example of the parametric oscillation process, we return to KTaO3_3, and look for positive biquadratic coupling. We identify strong positive biquadratic coupling between the 16.7 THz IR phonon, now polarized along the [100] direction, and a phonon at the MM-point of the Brillouin zone that transforms like the irreducible representation M3M_{3} with frequency 15.2 THz (Figure 1, Region III Structural Changes); the parameters we used for our dynamical simulations, calculated from first principles, are shown in Table I. The frequency ratio (δ=0.91\sqrt{\delta} = 0.91) of the two phonons places us on the blue arrow originating from the vertical axis of the phase diagram, near the phase boundary. At this frequency ratio, parametric oscillation of the M3M_{3} phonon is expected when 0.11<ϵ<0.340.11 < \epsilon< 0.34 (Eqn. (14)) corresponding to peak electric fields between 2.3 MV/cm <E0<4.0< E_{0} < 4.0 MV/cm. This is consistent with the simulated dynamics in Figure 1, Region III, which we will now discuss.

For E0=2.2E_{0} = 2.2 MV/cm (ϵ=0.10\epsilon= 0.10) the amplitude of QqˉQ_{\bar{q}} appears identical to the horizontal axis in the structural dynamics panel of Figure 1, Region III. That is, the dynamics are those of Region I.

For E0=3.3E_0 = 3.3 MV/cm (ϵ=0.23\epsilon= 0.23) we see large amplification of the M3−M_3^{-} phonon which grows over ≈1\approx1 ps. The large amplification of QqˉQ_{\bar{q}} is due to the oscillation of its effective potential through the QIRQ_{IR}. This parametric oscillation mechanism is shown schematically in Figure 1, Region III. The effective potential oscillates at a frequency of 2ωIR2\omega_{IR} as a result of the IR phonon drive, stiffening the potential through the positive DIR,qˉD_{IR,\bar{q}} (negative DIR,qˉD_{IR,\bar{q}} would soften the potential). A half-cycle of QqˉQ_{\bar{q}} motion is shown in four steps with QqˉQ_{\bar{q}} starting at the apex of a cycle. During the first step (1), QqˉQ_{\bar{q}} falls toward the static equilibrium point as the potential softens to its time-averaged value (Eqn. (2)). In the second step (2), QqˉQ_{\bar{q}}’s kinetic energy allows it to move beyond the equilibrium point as the potential continues to soften to its undriven value. In the third step (3), QqˉQ_{\bar{q}} begins losing kinetic energy while the potential stiffens back to the time-averaged value. In the final step (4), QqˉQ_{\bar{q}} reaches the height of its half-cycle motion as the potential stiffens back to its apex. QIRQ_{IR}’s effect on the potential increases the amplitude of QqˉQ_{\bar{q}} with each half-cycle of motion, as expected in a parametric oscillation process. The growth in QqˉQ_{\bar{q}} then decays back to zero amplitude due to the “back action” on QIRQ_{IR}. That is, QIRQ_{IR} is transiently driven with a finite amount of energy, as QqˉQ_{\bar{q}} grows in amplitude it must gain energy from QIRQ_{IR}, thereby decreasing the overall amplitude of the IR mode. Additionally, as the amplitude of QqˉQ_{\bar{q}} grows, the frequency of QIRQ_{IR} will be modified via the biquadratic coupling, tuning the frequency ratio δ\sqrt{\delta} away from Region III (see the 2DIR,qˉQqˉ22D_{IR,\bar{q}}Q_{\bar{q}}^2 term in Eqn. (3)). This suggests that other changes in the IR phonon frequency or amplitude, e.g. through other anharmonic couplings, have a similar detrimental effect on the parametric oscillation process.

For E0=4.4E_0 = 4.4 MV/cm (ϵ=0.41\epsilon= 0.41) we are again in a trivial region of the phase diagram where excitation of QIRQ_{IR} has negligible effect on the amplitude of QqˉQ_{\bar{q}}. This is because the effective frequency of QqˉQ_{\bar{q}} (due to QIRQ_{IR}) is driven out of sync with the oscillation in its potential energy landscape. This conversely explains why amplification was not seen for the E0=2.2E_0 = 2.2 MV/cm case, the effective frequency of QqˉQ_{\bar{q}} was not driven high enough to sync up with QIRQ_{IR}.

Since the parametric oscillation process is general, we expect many modes to parametrically oscillate for a large enough drive, the effects of which we expect will be seen in structure factor analysis of diffuse scattering following IR excitation. That being said, parametric oscillation processes and the phase boundary between region I and region III are sensitive to damping (Appendix § 1.1.3). As a result, the experimental observation of parametric oscillation between the modes used in this illustrative example, even with the large coupling between them, may be hindered by damping.

We are unaware of any experimental work showing parametric oscillation through the anharmonic lattice potential in the nonlinear phononics literature. A recent work proposed parametric oscillation of an IR-active phonon through nonlinear (in QIRQ_{IR}) contributions to the polarizability as an explanation for IR-resonant enhancement of the reflectivity in the reststrahlen band in SiC [53]; we have ignored the nonlinear polarizability contribution to the dynamics in this work, although this would certainly be an interesting path to pursue for future work.

What happens in crystals where a structural mode is already present in the equilibrium phase? We now consider the lower half of the phase diagram (imaginary δ\sqrt{\delta}), and look at Region IV to show that the conventional nonlinear phononics response may be activated by the condensation of a mode responsible for the structural phase transition.

Region IV – Conventional Nonlinear Phononics

We now briefly consider Region IV in the lower half of the phase diagram in Figure 1. This region corresponds to the conventional nonlinear phononics effect, which has been the focus of many previous studies [18, 20, 54, 55, 56, 57].

For this region, and in the lower part of the phase diagram generally, it is helpful to consider some (possibly virtual) high-symmetry reference phase of the material of interest. In this phase, QqˉQ_{\bar{q}} has a negative force constant (Kqˉ<0K_{\bar{q}} < 0) and δ\sqrt{\delta} is imaginary. The Floquet analysis predicts exponential growth of QqˉQ_{\bar{q}} even in the absence of an excited IR phonon (ϵ=0\epsilon= 0) because QqˉQ_{\bar{q}} is independently ‘unstable’ and induces a non-transient structural phase transition. The symmetry of the reference phase is lowered by QqˉQ_{\bar{q}} such that QqˉQ_{\bar{q}} becomes fully symmetric (transforms like the identical representation) in the new structural phase. It is therefore Raman-active about its new minimum and described by a new mode QRQ_R (that is Qqˉ→Qqˉ,±∗(0)+QRQ_{\bar{q}} \rightarrow Q_{\bar{q},\pm}^{*}(0) + Q_R). The biquadratic coupling term between an IR-active mode and QqˉQ_{\bar{q}} in the high-symmetry phase transforms into two lower-order terms describing a renormalization of the IR force constant of the form KIR→KIR+(2DIR,qˉQqˉ,±∗(0)2)K_{IR} \rightarrow K_{IR} + (2D_{IR,\bar{q}}Q_{\bar{q},\pm}^{*}(0)^2), and a new linear-quadratic coupling is created of the form AQIR2QRAQ_{IR}^2Q_R (where A=2DIR,qˉQqˉ,±∗(0)A = 2D_{IR,\bar{q}}Q_{\bar{q},\pm}^{*}(0)).

When an IR phonon is excited in the low-symmetry phase it will exert a unidirectional force on other modes coupled to it through the AQIR2QRAQ_{IR}^2Q_R term and the QRQ_R modes will be unidirectionally displaced from their equilibrium amplitudes. The sign of the biquadratic coupling parameter DIR,qˉD_{IR,\bar{q}} dictates the direction in which QRQ_R will be displaced: for DIR,qˉ<0D_{IR,\bar{q}} < 0 (Figure 1, left side of Region IV), QRQ_R is pushed away from the saddle point (the amplitude of QRQ_R increases compared to its equilibrium value), whereas for DIR,qˉ>0D_{IR,\bar{q}} > 0, QRQ_R is pushed towards the saddle point (the amplitude of QRQ_R decreases relative to its equilibrium value). Since QRQ_R is fully symmetric as mentioned above, unidirectional displacements in this region do not change the crystal point group or translational symmetry [58]. We will return to this point in Region V, where symmetry can be restored by driving QIRQ_{IR}.

We focus our theoretical development of this region on SrTiO3_3, which has been studied in several recent nonlinear phononics experiments [59, 60]. Our development here is intended to demonstrate conceptual features of the Floquet phase diagram, with a detailed description saved for a future publication.

To explore the dynamics and connect with the theoretical development of § 1.1 A and II B, we first calculate parameters for Eqn. (1) by considering an IR-active phonon at 5.15 THz in the cubic phase of SrTiO3_3, which we take as our high-symmetry reference phase (see Table II). We again consider coupling to a mode with R4+R_4^+ symmetry, which corresponds to an out-of-phase tilt of the TiO6_6 octahedra [61]. As mentioned above, this mode drives a structural phase transition in SrTiO3_3 at about 110 K to a phase with I4/mcmI4/mcm symmetry. Hence, in the cubic phase at 0 K this octahedral tilt mode has an imaginary frequency, 2.68ii THz. For this particular mode pairing, δ=0.52i\sqrt{\delta}=0.52i, which places us in a region represented by the purple arrow originating from the vertical axis in Figure 1, Region IV.

We perform our simulations in a structural phase in which the octahedral tilt mode has frozen in to the cubic structure to produce the low-symmetry (low temperature) tetragonal I4/mcmI4/mcm phase. In this phase, the octahedral tilt mode has an equilibrium amplitude of -72 pm from our DFT calculations and we denote it as QRQ_R (the negative amplitude signifies the left minimum of the double-well potential in Figure 1 Region IV). The triply degenerate IR phonon at 5.15 THz in the cubic phase is split in the tetragonal phase into a mode polarized along the axis about which the octahedra rotate (5.54 THz from our calculations), and a doubly degenerate mode polarized along the in-plane direction, perpendicular to the axis about which the octahedra rotate (5.04 THz). In this section we focus on this latter set of modes.

Excitation of an IR phonon polarized along the in-plane direction quasi-statically displaces QRQ_R away from the saddle point shown in the Region IV inset in Figure 1; that is, the amplitude of QRQ_R transiently increases. As the peak electric field increases from 5 MV/cm to 7 MV/cm and 9 MV/cm, the amplitude increases to -85 pm, -97 pm and -112 pm (corresponding to an ϵ\epsilon of -0.11, -0.22, -0.37).

In this section, we have shown that the conventional nonlinear phononics effect can be incorporated into a more general parametric amplification framework that is enabled by biquadratic coupling between modes in a proximal higher symmetry parent phase that is either real or virtual. In the next section, we go one step further and show that higher symmetry phases may be stabilized through IR-active phonon drive.

Region V – (Re)Introducing Symmetry

In this final section, we show that crystal phases of proximal high-symmetry parent phases can be stabilized in the transient response to IR-phonon drive, (re)introducing symmetry elements into the transient crystal structure.

For this region it is again helpful to consider some high-symmetry reference phase of the material of interest, as was done in the discussion of Region IV. In this reference phase, QqˉQ_{\bar{q}} has a negative force constant (Kqˉ<0K_{\bar{q}}<0) and δ\sqrt{\delta} is imaginary – the high-symmetry reference structure is a saddle-point of the energy. With DIR,qˉ>0D_{IR,\bar{q}}>0 and a sufficient drive ϵ\epsilon, Floquet theory predicts exponential decay of QqˉQ_{\bar{q}}. That is, Floquet theory predicts the high-symmetry reference phase is stable for large enough IR-active phonon drive, essentially stabilizing a saddle-point of the energy landscape. The negative Kˉqˉ\bar{K}_{\bar{q}} is overcome and made positive by the driven oscillating potential through the large amplitude IR-active phonon motion and biquadratic coupling. This is the phonon counterpart to the classic rigid pendulum problem where driving its pivot periodically can stabilize the inverted solution [62, 63].

Mathematically, the results are identical to those found in § 1.2 A 2, but both δ\delta and ϵ\epsilon have changed sign. Rather than making Kˉqˉ\bar{K}_{\bar{q}} negative by driving the IR mode, Kˉqˉ\bar{K}_{\bar{q}} starts negative and is driven to a positive value, collapsing the time-averaged double-well back to a stable single-well. The phase boundary is therefore defined by, in analogy to the discussion below Eqn. (7), ϵ=∣δ∣\epsilon=|\delta|. In this way the minimum associated with Qqˉ,±(0)Q_{\bar{q},\pm}(0) have moved to zero, restoring a higher-symmetry crystal configuration by reintroducing the symmetry elements of QqˉQ_{\bar{q}}.

To illustrate this point, we focus on the out-of-plane polarized IR mode in SrTiO3_3 at 5.54 THz mentioned in § 1.2 A 4. In contrast to the in-plane excitation, the sign of the biquadratic coupling is positive and therefore ϵ>0\epsilon>0 (Table II). Exciting this phonon pushes the octahedral tilts (QRQ_R) from their starting amplitude of -72 pm towards zero amplitude. For ϵ=0.21\epsilon=0.21 (E0=4.0E_0=4.0 MV/cm), the response is still of the conventional nonlinear phononics type. That is, the amplitude of QRQ_R changes but the point group and translational symmetry of the crystal is preserved. For ϵ=0.65\epsilon=0.65 (E0=7.0E_0=7.0 MV/cm), we crossover the phase boundary from Region IV to Region V, and the point group and translational symmetry elements are reintroduced so that the average transient structure appears cubic (Pm3ˉmPm\bar{3}m, space group #221).

In the transient response, as QIRQ_{IR} dissipates energy it eventually becomes unable to sustain QRQ_R about the cubic structure. When this happens QRQ_R will fall back into either of the double-well minima (Figure 1, Region V). We expect that in experiments this process will depend sensitively on details of the pulse characteristics, damping, initial conditions, and the boundary of the illuminated region of the crystal [22, 64, 65]. This suggests that with these approximations deterministic switching from one double-well minimum to another is not possible by this pathway. This was pointed out in a recent study where a multi-pulse sequence was needed to switch the polarization of KNbO3_3 [66], and may partly explain the lack of switching and transient recovery of polarization seen in an IR phonon pumped experiment in LiNbO3_3 [22] (as pointed out in [64, 65]).

For large enough drive, another phase boundary is crossed to the parametric oscillation regime (Region III). In the transient response, in order to cross to the parametric oscillation regime, the trajectory must pass through Region V. That is, the transient response will first be stabilized in the high-symmetry structure before the mode parametrically oscillates. This is shown in the simulated dynamics of Figure 1, Region V for ϵ=0.93\epsilon= 0.93 and E0=8.4E_0 = 8.4 MV/cm. Increasing ϵ\epsilon further will traverse an alternating series of phase boundaries of reintroducing symmetry and parametric oscillation regimes, though we expect this to be largely inaccessible in experiment due to the presence of other anharmonicity or destruction/melting of the crystal.

Mode- and polarization-selective response

In the development of the work presented so far, it is implicit that the response is frequency dependent. That is, the choice of IR phonon to excite, along with its polarization, may strongly affect the resulting transient structural phase transition. In this section, we illustrate the mode- and polarization-selective features of the lattice response by focusing on Region II of the phase diagram for KTaO3_3 and SrTiO3_3. For KTaO3_3 we can estimate the critical fields needed to induce new structural phases since the cubic phase is stable in DFT. For SrTiO3_3, since the cubic phase is unstable in DFT (there are phonon modes with imaginary frequencies), we report only the new structural phases induced, which we expect to be relevant to experiments just above the structural phase transition temperature of 110 K. We note that a general exploration of the polarization dependence for all regions is possible with ab initio techniques, but quite computationally expensive due to the enormous number of anharmonic pathways allowed. This is particularly true for Region III, where phonons at all frequencies and all wave vectors may be involved.

In both KTaO3_3 and SrTiO3_3, and in perovskite materials in general, the most common structural instabilities are associated with the R4+R_4^+ and M3+M_3^+ zone-edge phonons, that is, these modes often appear with imaginary frequencies in the cubic phase (readily calculated using DFT) and often drive structural phase transitions. The displacement patterns for these phonons represent out-of-phase (R4+R_4^+), and in-phase (M3+M_3^+) octahedral tilts of the TaO6_6 or TiO6_6 octahedra about the cubic crystallographic axes.

It is convenient to introduce a notation favored in the complex oxide community, after Glazer [67, 68], which describes in-phase (+) and out-of-phase (-) octahedral tilts as a list about the cubic-crystallographic axes. In the example described in Figure 1 for Region II, the equilibrium state of KTaO3_3 is described by the label a0a0a0a^0a^0a^0 indicating that the aa, bb, and cc lattice constants are all equal and that there are no octahedral tilts about any crystallographic axis. The octahedral tilt pattern induced in Region II (associated with the R4+R_4^+ [111] phonon of the cubic phase) following excitation of an IR phonon polarized along the [111]-direction is labeled a−a−a−a^-a^-a^-, signifying that the octahedral tilts about each crystallographic axes are all out-of-phase with respect to each other but are of the same amplitude. The octahedral tilt patterns in KTaO3_3 associated with transient structural phase transitions following excitation of various IR-active phonons polarized along the [100], [110], and [111] directions are shown in Table III (Table IV for SrTiO3_3). The entries shown represent the first modes that develop negative force constants (imaginary frequencies) with respect to an increase in amplitude of a given IR mode (see Figure 2). We expect that polarization directions between the principal crystallographic axes will give octahedral tilt patterns between those listed in Table III, e.g. for a 16.7 THz pulse polarized between [111] and [110] we expect an octahedral tilt pattern of a−a−b−a^-a^-b^-, which corresponds to the C2/cC2/c (C2hC_{2h}) space group. Note that for the [100] direction, the two directions of Qq⃗Q_{\vec{q}} shown in Tables III & IV have the same energy decrease in DFT, so either direction, or both directions may be seen in experiment.

Depending on the polarization direction, although the octahedral tilts dominate the induced Qq⃗Q_{\vec{q}} structural change, other structural distortions may be present, including strain (Supplementary § 3.2). We find that other structural distortions associated with A-site and B-site motion and deformation of the oxygen octahedra tend to be small compared to the octahedral tilt components of the motion (Supplementary § 3.3).

We speculate that the strong coupling between IR phonons and octahedral tilt modes is a consequence of the geometric network of bonds in both KTaO3_3 and SrTiO3_3, and is likely general to perovskites. To illustrate this, consider the 16.7 THz IR phonon polarized along [100] direction in KTaO3_3, where at the critical field, the shortest time-averaged Ta-O bond is along the polarization direction and its length has decreased by ≈±10%\approx\pm10\% (≈±20\approx\pm20 pm). This shortest Ta-O bond is energetically unfavorable. Rotating the octahedra by displacing the A5A_5 symmetry Qq⃗Q_{\vec{q}} mode (Table III) accommodates this unfavorable condition by increasing the length of this Ta-O bond towards its equilibrium value (Supplementary Fig. S4).

Summary and Conclusions

We have used a combination of first-principles DFT calculations and Floquet theory to develop a phase diagram depicting the various dynamical regimes accessible to materials given ultrafast optical excitation of an IR-active phonon quadratically coupled to another mode at arbitrary wave vector. We have shown that crystal point group and translational symmetries may be introduced or removed via various mechanisms, depending on which dynamical regime is accessed in a given experiment. Our phase diagram is intended to frame theoretical

Frequency [THz]Polarization DirectionSpace group induced by QIRQ_{IR}Irrep of Qˉq\bar{Q}_{q} in space group induced by QIRQ_{IR}Space group induced by QIR+QˉqQ_{IR}+\bar{Q}_{q}Octahedral tilt patternSpace group induced by octahedral tiltCritical field [MV/cm]
16.7[100]P4mm (C4vC_{4v})A5(a,0)A_{5}(a,0)Ima2 (C2vC_{2v})a0b−b−a^{0}b^{-}b^{-}Imma (D2hD_{2h})4.5
5.5[100]P4mm (C4vC_{4v})orororor18.0
3.3[100]P4mm (C4vC_{4v})A5(a,a)A_{5}(a,a)Fmm2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})1.8
16.7[110]Amm2 (C2vC_{2v})T4T_{4}Ima2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})4.6
5.5[110]Amm2 (C2vC_{2v})Y4Y_{4}Pmc2₁ (C2vC_{2v})a0a0b+a^{0}a^{0}b^{+}P4/mbm (D4hD_{4h})11.6
3.3[110]Amm2 (C2vC_{2v})T4T_{4}Ima2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})2.3
16.7[111]R3m (C3vC_{3v})4.8
5.5[111]R3m (C3vC_{3v})T2T_{2}R3m (C3vC_{3v})a−a−a−a^{-}a^{-}a^{-}R3ˉ\bar{3}c (D3dD_{3d})15.1
3.3[111]R3m (C3vC_{3v})2.8

Table 3. Induced instabilities (Region II) in KTaO₃ for resonant excitation of IR-active phonons polarized along different directions from our DFT calculations. Within a half cycle of the IR-active phonon, the space group symmetry is lowered (third column). As the IR-active phonon rings, Qˉq\bar{Q}_{q} condenses into the crystal, further lowering the space group symmetry (fifth column), altering the point group and translational symmetry of the crystal. Qˉq\bar{Q}_{q}, labeled by the irreducible representation (Irrep) in the space group induced by the IR-active phonon excitation (fourth column), is primarily associated with TaO₆ octahedral tilt patterns (sixth column), but may also include other subtle distortions (Supplementary Sec. S3). The space group induced by the octahedral tilts alone is given in the seventh column. The critical electric field needed to condense Qˉq\bar{Q}_{q} for a 500 fs duration Gaussian electric field pulse is given in the last column. For the 5.5 THz IR phonon polarized along the [110] direction, our calculations suggest that higher-order lattice anharmonicity may be needed to describe the condensation of Qˉq\bar{Q}_{q}.

Frequency [THz]Polarization DirectionSpace group induced by QIRQ_{IR}Irrep of Qˉq\bar{Q}_{q} in space group induced by QIRQ_{IR}Space group induced by QIR+QˉqQ_{IR}+\bar{Q}_{q}Octahedral tilt patternSpace group induced by octahedral tilt
16.8[100]P4mm (C4vC_{4v})A5(a,0)A_{5}(a,0)Ima2 (C2vC_{2v})a0b−b−a^{0}b^{-}b^{-}Imma (D2hD_{2h})
5.2[100]P4mm (C4vC_{4v})orororor
1.9[100]P4mm (C4vC_{4v})A5(a,a)A_{5}(a,a)Fmm2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})
16.8[110]Amm2 (C2vC_{2v})T4T_{4}Ima2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})
5.2[110]Amm2 (C2vC_{2v})T4T_{4}Ima2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})
1.9[110]Amm2 (C2vC_{2v})T4T_{4}Ima2 (C2vC_{2v})a0a0b−a^{0}a^{0}b^{-}I4/mcm (D4hD_{4h})
16.8[111]R3m (C3vC_{3v})T2T_{2}R3c (C3vC_{3v})a−a−a−a^{-}a^{-}a^{-}R3ˉ\bar{3}c (D3dD_{3d})
5.2[111]R3m (C3vC_{3v})T2T_{2}R3c (C3vC_{3v})a−a−a−a^{-}a^{-}a^{-}R3ˉ\bar{3}c (D3dD_{3d})
1.9[111]R3m (C3vC_{3v})T2T_{2}R3c (C3vC_{3v})a−a−a−a^{-}a^{-}a^{-}R3ˉ\bar{3}c (D3dD_{3d})

Table 4. Transiently induced structural phases (Region II) in SrTiO₃ for resonant excitation of IR-active phonons polarized along different directions. Since the cubic phase of SrTiO₃ is dynamically unstable at 0 K in DFT, we only report the predicted structural phases, assuming that the critical field can be controlled by proximity to the 110 K phase transition temperature. Within a half cycle of the IR-active phonon, the space group symmetry is lowered (third column). As the excited IR-active phonon rings, Qˉq\bar{Q}_{q} freezes into the crystal, further lowering the space group symmetry (fifth column), altering the point group and translational symmetry of the crystal. Qˉq\bar{Q}_{q}, labeled by the irreducible representation (Irrep) in the space group induced by the IR-active phonon excitation (fourth column), is primarily associated with TiO₆ octahedral tilt patterns (sixth column), but may also include other subtle distortions (Supplementary Sec. S3). The space group induced by the octahedral tilts alone is given in the seventh column.

and experimental work in the nonlinear phononics field where the transient response of the crystal may approach the Floquet regime, as justified by our dynamical simulations with parameters derived from first-principles calculations. Although we have ignored damping of the IR phonon in this work, we note that the inclusion of damping will quantitatively alter some of our results. That is, for a given IR-active phonon, larger peak electric fields and/or pulse durations may be needed to observe the desired effect.

As an example of point group and translational symmetry control, we have shown that in the perovskite KTaO3_3, which is cubic at all temperatures at equilibrium, zone-edge octahedral tilts can be induced by coupling to an optically excited IR phonon, revealing a hidden structural phase. The explored polarization dependence of this phenomenon suggests that octahedral tilt modes (and potentially other kinds of structural distortion patterns) can be precisely controlled with light. We expect that the susceptibility of a given material to this kind of control is tied to the frequency of Qq~Q_{\tilde q}, with lower frequency ωq~\omega_{\tilde q} modes more likely to induce transient structural phase transitions. Accompanying this lowering in symmetry, we expect other phonons to be amplified to large oscillatory motion through parametric oscillation. Resolving this motion experimentally will require measuring the structural response of the crystal on multiple timescales using, for example, ultrafast IR pump/diffuse X-ray scattering probes. Furthermore, our results suggest that continued development of tunable lasers spanning the 1 THz – 20 THz spectral range will enable further exploration of the structural and functional response of crystals to light.

Finally, we note that in the development of the Floquet phase diagram in Figure 1 we have focused on coupling between the driven IR-active phonon and other phonons, however the form of Eqn. (1) is general. That is, replacing Qq~Q_{\tilde q} with an arbitrary order parameter describing, for example, magnetism or orbital order, preserves the symmetries of the model. The approach here is therefore general and can be applied to understand any IR-phonon-driven phase change.

Acknowledgments

JZK and NAB were supported by the Department of Energy – Office of Basic Energy Sciences under award DE-SC0019414. GK was supported by the Cornell Center for Materials Research with funding from the NSF MRSEC program (Grant No. DMR-1719875). Computational resources were provided by the Cornell Center for Advanced Computing.

A Unit Conversion

The proper unit conversions are required in order for ϵ\epsilon to be a dimensionless quantity. ϵ\epsilon, when simplified and written in terms of pulse characteristics and microscopic parameters

ϵ=(ηπ)2DIR,q~Z~∗2τ2E02Mq~KIR2(12)\epsilon= \left(\frac{\eta}{\pi}\right)^2 \frac{D_{IR,\tilde q}\tilde Z^{*2}\tau^2 E_0^2}{M_{\tilde q}K_{IR}^2} \tag*{(12)}

where (ηπ)2\left(\frac{\eta}{\pi}\right)^2 is a dimensionless coefficient and η\eta is defined by the pulse shape (see below Eqn. (4)). Given the following choice of units for the other terms: DIR,q~D_{IR,\tilde q} [eV/Å4^4], Z~∗\tilde Z^* [e−^-], τ\tau [ps], E0E_0 [MV/cm], Mq~M_{\tilde q} [u (atomic mass units)], and KIRK_{IR} [eV/Å2^2]; a scale factor of 0.9648533 is needed.

B Constructing the Floquet Phase Diagram

In order to construct the phase diagram for the dynamical response described by Eqn. (5) we recall the main results of Floquet theory [31, 32].

The fundamental solutions of Eqn. (5) are of the form, xi(θ)=λiθ/πpi(θ)x_i(\theta) = \lambda_i^{\theta/\pi}p_i(\theta) where pi(θ)=pi(θ+π)p_i(\theta) = p_i(\theta+\pi) shares the periodicity of A(θ)A(\theta) and λi\lambda_i are the so-called characteristic multipliers. This shows that solutions are generally periodic in time and may grow or decay exponentially. The λi\lambda_i are found numerically by integrating Eqn. (5) over one period given a set of linearly independent initial conditions X(θ0)=(x1(θ0),x2(θ0),...,xN(θ0))X(\theta_0) = (x_1(\theta_0), x_2(\theta_0), ..., x_N(\theta_0)). The transformation matrix B=X(θ0)−1X(θ0+π)B = X(\theta_0)^{-1}X(\theta_0+\pi) is diagonalized to find its eigenvalues—the characteristic multipliers λi\lambda_i. Three scenarios are possible: ∣λi∣>1|\lambda_i| > 1 corresponding to exponential growth, ∣λi∣<1|\lambda_i| < 1 corresponding to exponential decay, and ∣λi∣=1|\lambda_i| = 1 corresponding to stability of the ithi^{th} solution. Here, ∣λi∣=1|\lambda_i| = 1 define phase boundaries between regions of exponential growth and decay.

We numerically integrate Eqn. (6) from θ∈[0,π]\theta\in[0,\pi] for a mesh of δ\delta, ϵ\epsilon, and γ\gamma with initial conditions x1(0)=(1,0)x_1(0) = (1,0) and x2(0)=(0,1)x_2(0) = (0,1). x1(0)x_1(0) corresponds to a physical scenario where at θ=0\theta= 0, Qq~Q_{\tilde q} is displaced but its velocity Q˙q~\dot{Q}_{\tilde q} is zero. Conversely, x2(0)x_2(0) corresponds to a scenario where at θ=0\theta= 0, Qq~=0Q_{\tilde q} = 0 and the velocity is nonzero. The eigenvalues of the transformation matrix BB define the characteristic multipliers, which are analyzed to construct the phase boundaries (∣λ∣=1|\lambda| = 1) and regions of exponential growth (∣λ∣>1|\lambda| > 1) and decay (∣λ∣<1|\lambda| < 1) in Qq~Q_{\tilde q}.

C Parametric Oscillation Derivation

In this Appendix, we derive expressions for the approximate phase boundary, an exponential growth rate, and peak growth rate including the effects of damping for parametric oscillation in Region III. We anticipate that these expressions will be useful in future experimental work exploring this phenomenon.

To accommodate the exponential growth predicted by the Floquet analysis and the expected periodic motion, we assume a solution of the form Qq~=A(τ)cos⁡(τ)+B(τ)sin⁡(τ)Q_{\tilde q} = A(\tau)\cos(\tau) + B(\tau)\sin(\tau). Inserting this ansatz into Eqn. (5), we find for the exponential growth parameters AA and BB,

ddτ(AB)=12[0−1+(δ+ϵ)−ϵ/21−(δ+ϵ)−ϵ/20](AB)(13)\frac{d}{d\tau}\begin{pmatrix} A \\ B \end{pmatrix} = \frac{1}{2} \begin{bmatrix} 0 & -1+(\delta+\epsilon)-\epsilon/2 \\ 1-(\delta+\epsilon)-\epsilon/2 & 0 \end{bmatrix} \begin{pmatrix} A \\ B \end{pmatrix} \tag*{(13)}

where A¨\ddot{A} and B¨\ddot{B} have been neglected in the spirit of the slowly varying envelope approximation [69]. We have also neglected high-harmonic terms and damping to find this form. Eqn. (13) has exponential solutions with their growth rate found by solving for the eigenvalues of the matrix on the right-hand side. We find μV,±=±12(ϵ/2)2−(1−(δ+ϵ))2\mu_{V,\pm}=\pm\frac{1}{2}\sqrt{(\epsilon/2)^2-(1-(\delta+\epsilon))^2}, where μV,+\mu_{V,+} gives rise to exponential growth. The phase boundary is again identified by setting μ=0\mu=0. We find the following conditions for the phase boundary:

(ϵ2)2≥(1−(δ+ϵ))2\left(\frac{\epsilon}{2}\right)^2 \ge\left(1-(\delta+\epsilon)\right)^2
23(1−δ)≤ϵ≤2(1−δ)(14)\frac{2}{3}(1-\delta)\le\epsilon\le2(1-\delta) \tag*{(14)}

Maximum exponential growth is found at ϵ∗=43(1−δ)\epsilon^*=\frac{4}{3}(1-\delta) with a growth rate of μ(ϵ∗)=∣1−δ∣23\mu(\epsilon^*)=\frac{|1-\delta|}{2\sqrt{3}}.

To account for the effect of damping we assume the standard result from damped oscillators Qq~∝eϵ~τ/2Q_{\tilde{q}}\propto e^{\tilde{\epsilon}\tau/2} so that exponential growth is only expected when μV,+\mu_{V,+} is greater than ν/2\nu/2. This alters Eqn. (14) so that the phase boundary is defined by

ϵ≥43((1−δν)−12(1−δν)2−3ν2)\epsilon\ge\frac{4}{3}\left((1-\delta_\nu)-\frac{1}{2}\sqrt{(1-\delta_\nu)^2-3\nu^2}\right)
ϵ≤43((1−δν)+12(1−δν)2−3ν2)(15)\epsilon\le\frac{4}{3}\left((1-\delta_\nu)+\frac{1}{2}\sqrt{(1-\delta_\nu)^2-3\nu^2}\right) \tag*{(15)}

Here δν=δ−ν2/4\delta_\nu=\delta-\nu^2/4 accounts for the frequency shift imparted by the damping. This relation requires ∣1−δν∣≥3ν|1-\delta_\nu|\ge\sqrt{3}\nu and ∣ϵ∣≥2ν|\epsilon|\ge2\nu for exponential growth. That is, there is a range of ωq~\omega_{\tilde{q}} near ωIR\omega_{IR} that will not exhibit parametric oscillation and a larger drive ϵ\epsilon is needed to overcome the damping for ωq~\omega_{\tilde{q}} outside this range. The decrease in size of Region III is taken up by Region I and Region V (Figure S2). The phase boundaries between Region V and Regions I & IV represents the only phase boundaries in Figure 1 sensitive to damping of Qq~Q_{\tilde{q}}.

Floquet phase diagram with analytic phase boundaries including damping

Figure S2. Floquet phase diagram with analytic phase boundaries including damping (ν=0.2,0.4\nu= 0.2, 0.4). The red lines in the figure show the threshold values of δ=ω~q⃗/ωIR\sqrt{\delta} = \tilde{\omega}_{\vec{q}}/\omega_{IR} and ϵ=DIR,q⃗/Mq⃗ωIR2\epsilon= D_{IR,\vec{q}}/M_{\vec{q}}\omega_{IR}^{2} needed to enter the parametric oscillation regime near δ=1\sqrt{\delta} = 1. The analytic expressions can be found in Appendix § 1.1.3 of the main text.

References

Supplementary Information: Coherent control of the translational and point group symmetries of crystals with light

Guru Khalsa; Jeffrey Z. Kaaret; Nicole A. Benedek

Approximate solutions to the Floquet phase boundaries

Approximate analytic results for the phase boundaries are shown in Figure S1. This suggests that the analytic results given in the main text are good approximations for the transition from trivial motion (white regions – Region I) to novel dynamical response discussed in the main text (shaded regions), compared to those found with Floquet theory.

Figure S2 shows the affect of damping (ν=2γq~/ωIR\nu=2\gamma_{\tilde q}/\omega_{\mathrm{IR}}) on the phase boundary between trivial damped motion of Qq~Q_{\tilde q} (white region – Region I) and exponentially growing parametric oscillation of Qq~Q_{\tilde q} (blue region – Region III). The red lines in the figure show the threshold values the driving strength (∣ϵ∣≥2ν|\epsilon|\geq2\nu) and the frequency ratio δν\sqrt{\delta_\nu} (∣1−δν∣≥3ν|1-\delta_\nu|\geq\sqrt{3}\nu, where δν=δ−ν2/4\delta_\nu=\delta-\nu^2/4) that must be overcome to see the parametric oscillatory behavior (Appendix § 1.1.3 in the main text for further discussion).

The effect of strain and linear-quadratic coupling

Large strain states are expected to be observed on timescales comparable to the phonon-phonon scattering component of the lifetime [S1]. That is, when the energy transfers from QIRQ_{\mathrm{IR}} and Qq~Q_{\tilde q} to the acoustic branch. We note that a nearly instantaneous strain response (on the scale of the electronic scattering rate ∼1\sim1 ps) is also expected as the electrons accommodate the increased energy density from the optical field, but we expect this to be a small effect. Here, we discuss how this strain may impact the dynamics of the phase boundary, focusing on Region II. Including strain components in Eqn. (1) gives

U=12KIRQIR2+12Kq⃗Qq⃗2+12Cϵϵ2+αIRϵQIR2+αq⃗ϵQq⃗2+DIR,q⃗QIR2Qq⃗2−ΔP⃗⋅E⃗.(16)\begin{aligned} U &= \frac{1}{2}K_{IR}Q_{IR}^{2} + \frac{1}{2}K_{\vec{q}}Q_{\vec{q}}^{2} + \frac{1}{2}C_{\epsilon}\epsilon^{2} \\ &\quad+ \alpha_{IR}\epsilon Q_{IR}^{2} + \alpha_{\vec{q}}\epsilon Q_{\vec{q}}^{2} + D_{IR,\vec{q}}Q_{IR}^{2}Q_{\vec{q}}^{2} - \Delta\vec{P}\cdot\vec{E}. \tag*{(16)} \end{aligned}

where ϵ\epsilon is a strain coordinate, CϵC_{\epsilon} is an elastic stiffness, and αIR/q⃗\alpha_{IR/\vec{q}} is the coupling between strain and QIR/q⃗Q_{IR/\vec{q}}. In this equation, ϵ\epsilon represents the strain induced by the excitation and will generally be a linear combination of the equilibrium strain tensor components. CϵC_{\epsilon}, and αIR/q⃗\alpha_{IR/\vec{q}} are therefore effective elastic parameters associated with the induced strain response. Note that in the calculation of the strain coefficients, only the lattice constants are allowed to change. That is, the only atomic motion included is due to the IR-active phonon.

Taking the same steps as done in Sec. II A we define an effective force constant for Qq⃗Q_{\vec{q}}:

K~q⃗(QIR,ϵ)=Kq⃗+2αq⃗ϵ+2DIR,q⃗QIR2.(17)\widetilde{K}_{\vec{q}}(Q_{IR},\epsilon) = K_{\vec{q}} + 2\alpha_{\vec{q}}\epsilon+ 2D_{IR,\vec{q}}Q_{IR}^{2}. \tag*{(17)}

We find the peak strain amplitude ϵ∗\epsilon^{*} by solving −∂U∂ϵ=0-\frac{\partial U}{\partial\epsilon}=0, while keeping Qq⃗=0Q_{\vec{q}}=0. Resulting in: ϵ∗=αIRQIR2/Cϵ\epsilon^{*}=\alpha_{IR}Q_{IR}^{2}/C_{\epsilon}, which is substituted into Eqn. (17) to find,

K~q⃗(QIR)=Kq⃗+2(αIRαq⃗Cϵ+DIR,q⃗)QIR2.(18)\widetilde{K}_{\vec{q}}(Q_{IR}) = K_{\vec{q}} + 2\left(\frac{\alpha_{IR}\alpha_{\vec{q}}}{C_{\epsilon}} + D_{IR,\vec{q}}\right)Q_{IR}^{2}. \tag*{(18)}

This highlights the qualitative finding that if αIRαq⃗Cϵ\frac{\alpha_{IR}\alpha_{\vec{q}}}{C_{\epsilon}} has the same sign as DIR,q⃗D_{IR,\vec{q}}, then strain and the biquadratic coupling work in concert to decrease the critical field. If they differ in sign, strain and the biquadratic coupling work in opposition to increase the threshold field. We are unaware of any a priori strategy for anticipating the relative sign of the coupling and therefore expect theoretical modeling will generally be necessary for this level of detail in the definition of the phase boundary. We note that this framework for including the effect of other modes in the phase boundary applies to any linear-quadratic third-order coupling term. Therefore, the conventional nonlinear phononics effect, with anharmonic coupling ∝QIR2QR\propto Q_{\mathrm{IR}}^{2}Q_{\mathrm{R}} can also be treated on equal footing with strain, having only the quantitative effect on the phase boundary.

As an example of this, we focus on the Region II phase boundary for KTaO3_3 for excitation of the 16.7 THz IR-active phonon polarized along the [100] direction. The phase boundary in the absence of strain is found to occur at 4.6 MV/cm for a 500 fs duration Gaussian pulse. We find that αIRαq~/Cϵ\alpha_{\mathrm{IR}}\alpha_{\tilde{q}}/C_{\epsilon} is negative. That is, strain and the biquadratic coupling work in concert to condense Qq~Q_{\tilde{q}}. We find the phase boundary including strain to be 3.8 MV/cm.

Condensed value of $Q_{\tilde{q}}$ as a function of peak electric field without and with strain

Figure S3. Condensed value of Qq~Q_{\tilde{q}} as a function of peak electric field for without (black) and with (red) strain for the 16.7 THz IR active phonon in KTaO3_3 polarized along the [100] direction.

Eigenmodes of the induced instability Qq~Q_{\tilde{q}}

The induced structural instabilities due to IR-active phonon pumping are primarily octahedral tilts/rotations, as discussed in the main text, but involve smaller distortions that are also allowed by symmetry. An eigenmode decomposition of the Qq~Q_{\tilde{q}}’s found in Table III of the main text, referenced to the parent Pm3ˉ\bar{3}m structure, is given in Tables S1, S2, and S3. The symmetry labels and directions were generated by the ISOTROPY Software Suite [46, 47].

Table S1. Real-space eigendisplacements of the induced instabilities for IR phonon excitation along the crystallographic [100] direction in KTaO3_3, found in Table III of the main text. The IR phonon lowers the symmetry of the equilibrium Pm3ˉ\bar{3}m (OhO_h) structure to P4mm (C4vC_{4v}). The irreducible representation (Irrep) of Q~q⃗\widetilde{Q}_{\vec{q}} in P4mm is A5_5 (second column) which has two high symmetry directions shown in the top and bottom of the table. The decomposition of Q~q⃗\widetilde{Q}_{\vec{q}} in the Pm3ˉ\bar{3}m reference is given in the last columns. The sum of the squares of the eigenmode components is unity.

Frequency [THz]Irrep of Q~q⃗\widetilde{Q}_{\vec{q}} in space group induced by QIRQ_{IR}R5+R_5^+[101]:KR4−R_4^-[101]:TaR4+R_4^+[101]:OR5+R_5^+[101]:O
16.70.0607260.9787660.195523
5.5A5_5(a,0)-0.318172-0.0592780.9459800.019344
3.3-0.0199620.9837220.178586
16.7--0.0607260.9787660.195523
5.5A5_5(a,a)-0.3181720.0592780.9459800.019344
3.3--0.0199620.9837220.178586

Table S2. Real-space eigendisplacements of the induced instabilities for IR phonon excitation along the crystallographic [110] direction in KTaO3_3, found in Table III of the main text. The IR phonon lowers the symmetry of the equilibrium Pm3ˉ\bar{3}m (OhO_h) structure to Amm2 (C2vC_{2v}). The irreducible representation (Irrep) of Q~q⃗\widetilde{Q}_{\vec{q}} in Amm2 is T4_4 for the 16.7 THz and 3.3 THz phonons, and Y4_4 for the 5.5 THz IR phonon (second column) shown in the top and bottom of the table, respectively. The decomposition of Q~q⃗\widetilde{Q}_{\vec{q}} in the Pm3ˉ\bar{3}m reference is given in the last columns. The sum of the squares of the eigenmode components is unity.

Frequency [THz]Irrep of Q~q⃗\widetilde{Q}_{\vec{q}} in space group induced by QIRQ_{IR}R4−R_4^-[11]:TaR3+R_3^+[10]:OR4+R_4^+[100]:O
16.7T4_4-0.061179-0.026906-0.997764
3.3-0.058514-0.041206-0.997436
5.5Y4_40.5431050.1267060.026610-0.8164690.147150

Table S3. Real-space eigendisplacements of the induced instabilities for IR phonon excitation along the crystallographic [111] direction in KTaO3_3, found in Table III of the main text. The IR phonon lowers the symmetry of the equilibrium Pm3ˉ\bar{3}m (OhO_h) structure to R3m (C3vC_{3v}). The irreducible representation (Irrep) of Q~q⃗\widetilde{Q}_{\vec{q}} in R3m is T2_2. The decomposition of Q~q⃗\widetilde{Q}_{\vec{q}} in the Pm3ˉ\bar{3}m reference is given in the last column where only the R4+R_4^+[111]:O, representing a pure out-of-phase octahedral tilt, is seen. The sum of the squares of the eigenmode components is unity.

Frequency [THz]Irrep of Q~q⃗\widetilde{Q}_{\vec{q}} in space group induced by QIRQ_{IR}Decomposition of Q~q⃗\widetilde{Q}_{\vec{q}} in Pm3ˉ\bar{3}m
16.7R4+R_4^+[111]:O; 1.000000
5.5T2_21.000000
3.31.000000

Zone-edge phonon response to IR-active phonon excitation: bond-length changes

The strong coupling between the zone-edge phonon Qq⃗Q_{\vec{q}} and the IR-active phonon QIRQ_{IR} leading to new translational and point group symmetries in KTaO3_3 found in § 1.2 B, can be understood in terms of the geometric network of bonds (see last two paragraphs of § 1.2 B). The large changes in bond lengths due to the IR phonon excitation are energetically unfavorable for the crystal. Octahedral tilts associated with zone-edge phonons provide a pathway for the crystal to mitigate this situation by increasing the shortest bond length. This is shown in Figure S4 where the shortest time-averaged Ta-O bond length is shown as a function of IR-active phonon displacement (blue), with the inclusion of the R4+R_{4}^{+} octahedral tilt (red) and the full A5A_{5} phonon (green). Once the threshold value of QIRQ_{IR} is reached, both octahedral tilts and the other subtle structural distortions contributing to the A5A_{5} mode, increase the shortest time-averaged bond length, lowering the configuration energy of the crystal lattice.

Change in the shortest time-averaged Ta-O bond length due to the IR phonon, the IR phonon with the $R_{4}^{+}$ octahedral tilts, and the IR phonon with the total $Q_{\vec{q}}$

Figure S4. Change is the shortest time-averaged Ta-O bond length due to the IR phonon (blue), the IR phonon with the R4+R_{4}^{+} octahedral tilts (red), and the IR phonon with the total Qq⃗Q_{\vec{q}} (green). The solid black lines show the threshold value of QIRQ_{IR} and the Ta-O bond length. The structural changes were found by adding the Qq⃗Q_{\vec{q}} amplitudes predicted by Eqn. (10) with parameters from Table I into the equilibrium KTaO3_3 crystal structure to calculate the bond lengths.

Supplementary references

[S1] J. R. Hortensius, D. Afanasiev, A. Sasani, E. Bousquet, and A. D. Caviglia, “Ultrafast strain engineering and coherent structural dynamics from resonantly driven optical phonons in LaAlO3_3,” npj Quantum Materials, vol. 5, p. 95, Dec. 2020.

[46] H. T. Stokes, D. M. Hatch, and B. J. Campbell, “ISOTROPY Software Suite.” iso.byu.edu.

[47] B. J. Campbell, H. T. Stokes, D. E. Tanner, and D. M. Hatch, “ISODISPLACE: a web-based tool for exploring structural distortions,” Journal of Applied Crystallography, vol. 39, pp. 607–614, Aug 2006.

References

  1. [1]V. M. Goldschmidt, “Die Gesetze der Kristallochemie,” Naturwissenschaften, vol. 14, pp. 477–485, May 1926.
  2. [2]“29 - on the theory of phase transitions,” in Collected Papers of L.D. Landau (D. TER HAAR, ed.), pp. 193–216, Pergamon, 1965.
  3. [3]G. Shirane, “Neutron scattering studies of structural phase transitions at Brookhaven,” Rev. Mod. Phys., vol. 46, pp. 437–449, July 1974.DOI
  4. [4]W. Li, X. Qian, and J. Li, “Phase transitions in 2D materials,” Nature Reviews Materials, vol. 6, pp. 829–846, Sept. 2021.DOI
  5. [5]J. Tolédano and P. Tolédano, The Landau Theory of Phase Transitions. World Scientific Publishing Company, 1987.
  6. [6]R. A. Borzi, S. A. Grigera, J. Farrell, R. S. Perry, S. J. S. Lister, S. L. Lee, D. A. Tennant, Y. Maeno, and A. P. Mackenzie, “Formation of a Nematic Fluid at High Fields in Sr3Ru2O7,” Science, vol. 315, pp. 214–217, Jan. 2007.
  7. [7]N. K. Gupta, C. McMahon, R. Sutarno, T. Shi, R. Gong, H. I. Wei, K. M. Shen, F. He, Q. Ma, M. Dragomir, B. D. Gaulin, and D. G. Hawthorn, “Vanishing nematic order beyond the pseudogap phase in overdoped cuprate superconductors,” Proceedings of the National Academy of Sciences, vol. 118, p. e2106881118, Aug. 2021.
  8. [8]Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, E. Kaxiras, and P. Jarillo-Herrero, “Unconventional superconductivity in magic-angle graphene superlattices,” Nature, vol. 556, pp. 43–50, Mar. 2018.DOI
  9. [9]X. Zhang, M. Takahashi, K. Takeuchi, and S. Sakai, “64 kbit Ferroelectric-Gate-Transistor-Integrated NAND Flash Memory with 7.5 V Program and Long Data Retention,” Japanese Journal of Applied Physics, vol. 51, p. 04DD01, Apr. 2012.DOI
  10. [10]P. Jiao, K.-J. I. Egbe, Y. Xie, A. M. Nazar, and A. H. Alavi, “Piezoelectric Sensing Techniques in Structural Health Monitoring: A State-of-the-Art Review,” Sensors, vol. 20, July 2020.
  11. [11]P. Curie and J. Curie, “Développement par compression de l’électricité polaire dans les cristaux hémièdres à faces inclinées,” Bulletin de Minéralogie, vol. 3, no. 4, pp. 90–93, 1880.
  12. [12]G. Lippmann, “Principe de la conservation de l'électricité, ou second principe de la théorie des phénomènes électriques,” Journal de Physique Théorique et Appliquée, vol. 10, no. 1, pp. 381–394, 1881.
  13. [13]F. Bousquet, M. Dawber, N. Stucki, C. Lichtensteiger, P. Hermet, S. Gariglio, J.-M. Triscone, and P. Ghosez, “Improper ferroelectricity in perovskite oxide artificial superlattices,” Nature, vol. 452, pp. 732–736, Apr. 2008.DOI
  14. [14]T. Fukushima, A. Stroppa, S. Picozzi, and J. M. Perez-Mato, “Large ferroelectric polarization in the new double perovskite NaLaMnWO6 induced by non-polar instabilities,” Phys. Chem. Chem. Phys., vol. 13, pp. 12186–12190, June 2011.
  15. [15]N. A. Benedek and C. J. Fennie, “Hybrid improper ferroelectricity: A mechanism for controllable polarization-magnetization coupling,” Phys. Rev. Lett., vol. 106, p. 107204, Mar. 2011.
  16. [16]N. A. Benedek, A. T. Mulder, and C. J. Fennie, “Polar octahedral rotations: A path to new multifunctional materials,” Journal of Solid State Chemistry, vol. 195, pp. 11–20, Nov. 2012.
  17. [17]N. A. Benedek and M. A. Hayward, “Hybrid Improper Ferroelectricity: A Theoretical, Computational, and Synthetic Perspective,” Annual Review of Materials Research, vol. 52, pp. 331–355, July 2022.DOI
  18. [18]M. Rini, R. T. Tobey, N. Dean, S. Kaiser, A. Dienst, M. C. Hoffmann, S. Pyon, T. Takayama, H. Takagi, and A. Cavalleri, “Control of the electronic phase of a manganite by mode-selective vibrational excitation,” Nature, vol. 449, pp. 72–74, Sept. 2007.pubmed.ncbi.nlm.nih.gov/17805291
  19. [19]D. Fausti, R. I. Tobey, N. Dean, S. Kaiser, A. Dienst, M. C. Hoffmann, S. Pyon, T. Takayama, H. Takagi, and A. Cavalleri, “Light-Induced Superconductivity in a Stripe-Ordered Cuprate,” Science, vol. 331, pp. 189–191, Jan. 2011.DOI
  20. [20]R. Mankowsky, A. Subedi, M. Först, S. O. Mariager, M. Chollet, H. T. Lemke, J. S. Robinson, J. M. Glownia, M. P. Minitti, A. Frano, M. Fechner, N. A. Spaldin, T. Loew, B. Keimer, A. Georges, and A. Cavalleri, “Nonlinear lattice dynamics as a basis for enhanced superconductivity in YBa₂Cu₃O₆.₅,” Nature, vol. 516, pp. 71–73, Dec. 2014.DOI
  21. [21]T. A. Miller, R. W. Chhajlany, L. Tagliacozzo, B. Green, S. Kovalev, D. Prabhakaran, M. Lewenstein, M. Gensch, and S. Wall, “Terahertz field control of in-plane orbital order in La₀.₅Sr₁.₅MnO₄,” Nature Communications, vol. 6, p. 8175, Sept. 2015.DOI
  22. [22]R. Mankowsky, A. von Hoegen, M. Först, and A. Cavalleri, “Ultrafast Reversal of the Ferroelectric Polarization,” Phys. Rev. Lett., vol. 118, p. 197601, May 2017.DOI
  23. [23]M. Fechner, M. Först, G. Ornstein, V. Krapivin, A. S. Disa, M. Buzzi, A. von Hoegen, G. de la Pena, Q. L. Nguyen, R. Mankowsky, M. Sander, H. Lemke, Y. Deng, M. Trigo, and A. Cavalleri, “Quenched lattice fluctuations in optically driven SrTiO₃,” 2023.arxiv.org/abs/2301.08703
  24. [24]X. Gonze and C. Lee, “Dynamical matrices, Born effective charges, dielectric permittivity tensors, and interatomic force constants from density-functional perturbation theory,” Phys. Rev. B, vol. 55, pp. 10355–10368, Apr. 1997.DOI
  25. [25]G. Khalsa, N. A. Benedek, and J. Moses, “Ultrafast Control of Material Optical Properties via the Infrared Resonant Raman Effect,” Phys. Rev. X, vol. 11, p. 021067, June 2021.DOI
  26. [26]F. Caruso and M. Zacharias, “Quantum theory of light-driven coherent lattice dynamics,” Phys. Rev. B, vol. 107, p. 054102, Feb. 2023.DOI
  27. [27]A. Subedi, A. Cavalleri, and A. Georges, “Theory of nonlinear phononics for coherent light control of solids,” Phys. Rev. B, vol. 89, p. 220301, June 2014.DOI
  28. [28]M. Claassen, H.-C. Jiang, B. Moritz, and T. P. Devereaux, “Dynamical time-reversal symmetry breaking and photo-induced chiral spin liquids in frustrated Mott insulators,” Nature Communications, vol. 8, p. 1192, Oct. 2017.DOI
  29. [29]T. Oka and S. Kitamura, “Floquet Engineering of Quantum Materials,” Annual Review of Condensed Matter Physics, vol. 10, no. 1, pp. 387–408, 2019.DOI
  30. [30]A. de la Torre, D. M. Kennes, M. Claassen, S. Gerber, J. W. McIver, and M. A. Sentef, “Colloquium: Nonthermal pathways to ultrafast control in quantum materials,” Rev. Mod. Phys., vol. 93, p. 041002, Oct. 2021.DOI
  31. [31]W. Magnus and S. Winkler, Hill’s Equation. Dover Books on Mathematics Series, Dover Publications, 2004.
  32. [32]I. Kovacic, R. Rand, and S. Mohamed Sah, “Mathieu’s Equation and Its Generalizations: Overview of Stability Charts and Their Features,” Applied Mechanics Reviews, vol. 70, Feb. 2018. 020802.DOI
  33. [33]G. Kresse and J. Hafner, “Ab initio molecular dynamics for liquid metals,” Phys. Rev. B, vol. 47, pp. 558–561, Jan. 1993.pubmed.ncbi.nlm.nih.gov/10004490
  34. [34]G. Kresse and J. Furthmüller, “Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set,” Computational Materials Science, vol. 6, pp. 15–50, July 1996.DOI
  35. [35]G. Kresse and J. Furthmüller, “Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set,” Phys. Rev. B, vol. 54, pp. 11169–11186, Oct. 1996.pubmed.ncbi.nlm.nih.gov/9984901
  36. [36]G. Kresse and D. Joubert, “From ultrasoft pseudopotentials to the projector augmented-wave method,” Phys. Rev. B, vol. 59, pp. 1758–1775, Jan. 1999.DOI
  37. [37]S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, “Phonons and related crystal properties from density-functional perturbation theory,” Rev. Mod. Phys., vol. 73, pp. 515–562, July 2001.DOI
  38. [38]P. Voudsen, “A Study of the Unit-Cell Dimensions and Symmetry of certain Ferroelectric Compounds of Niobium and Tantalum at Room Temperature,” Acta Crystallographica, vol. 4, pp. 373–376, July 1951.DOI
  39. [39]T. Ishihara, N. S. Baik, N. Ono, H. Nishiguchi, and Y. Takita, “Effects of crystal structure on photolysis of H₂O on K–Ta mixed oxide,” Journal of Photochemistry and Photobiology A: Chemistry, vol. 167, pp. 149–157, Oct. 2004.DOI
  40. [40]M. Schmidbauer, A. Kwasniewski, and J. Schwarzkopf, “High-precision absolute lattice parameter determination of SrTiO₃, DyScO₃ and NdGaO₃ single crystals,” Acta Crystallographica Section B, vol. 68, pp. 8–14, Feb. 2012.DOI
  41. [41]R. D. King-Smith and D. Vanderbilt, “Theory of polarization of crystalline solids,” Phys. Rev. B, vol. 47, pp. 1651–1654, Jan. 1993.pubmed.ncbi.nlm.nih.gov/10006189
  42. [42]R. Resta, “Macroscopic Electric Polarization as a Geometric Quantum Phase,” Europhysics Letters, vol. 22, p. 133, Apr. 1993.DOI
  43. [43]D. Vanderbilt and R. D. King-Smith, “Electric polarization as a bulk quantity and its relation to surface charge,” Phys. Rev. B, vol. 48, pp. 4442–4455, Aug. 1993.pubmed.ncbi.nlm.nih.gov/10008920
  44. [44]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. D. Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, M. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, “QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials,” Journal of Physics: Condensed Matter, vol. 21, p. 395502, Sept. 2009.DOI
  45. [45]K. F. Garrity, J. W. Bennett, K. M. Rabe, and D. Vanderbilt, “Pseudopotentials for high-throughput DFT calculations,” Computational Materials Science, vol. 81, pp. 446–452, Jan. 2014.DOI
  46. [46]H. T. Stokes, D. M. Hatch, and B. J. Campbell, “ISOTROPY Software Suite.” iso.byu.edu.
  47. [47]B. J. Campbell, H. T. Stokes, D. E. Tanner, and D. M. Hatch, “ISODISPLACE: a web-based tool for exploring structural distortions,” Journal of Applied Crystallography, vol. 39, pp. 607–614, Aug. 2006.
  48. [48]C. R. Harris, K. J. Millman, S. J. van der Walt, R. Gommers, P. Virtanen, D. Cournapeau, E. Wieser, J. Taylor, S. Berg, N. J. Smith, R. Kern, M. Picus, S. Hoyer, M. H. van Kerkwijk, M. Brett, A. Haldane, J. F. del Río, M. Wiebe, P. Peterson, P. Gérard-Marchant, K. Sheppard, T. Reddy, W. Veësser, H. Abbasi, C. Gohlke, and T. E. Oliphant, “Array programming with NumPy,” Nature, vol. 585, pp. 357–362, Sept. 2020.DOI
  49. [49]M. W. Lufaso and P. M. Woodward, “Prediction of the crystal structures of perovskites using the software program SPuDS,” Acta Crystallographica Section B, vol. 57, pp. 725–738, Dec 2001.pubmed.ncbi.nlm.nih.gov/11717471
  50. [50]A. Sell, A. Leitenstorfer, and R. Huber, “Phase-locked generation and field-resolved detection of widely tunable terahertz pulses with amplitudes exceeding 100 MV/cm,” Opt. Lett., vol. 33, pp. 2767–2769, Dec. 2008.pubmed.ncbi.nlm.nih.gov/19037420
  51. [51]M. Knorr, J. Raab, M. Tauer, P. Merkl, D. Peller, E. Wittmann, E. Riedle, C. Lange, and R. Huber, “Phase-locked multi-terahertz electric fields exceeding 13 MV/cm at a 190 kHz repetition rate,” Opt. Lett., vol. 42, pp. 4367–4370, Nov. 2017.DOI
  52. [52]J. V. José and E. J. Saletan, Classical Dynamics: A Contemporary Approach, p. 382–491. Cambridge University Press, 1998.
  53. [53]A. Cartella, T. F. Nova, M. Fechner, R. Merlin, and A. Cavalleri, “Parametric amplification of optical phonons,” Proceedings of the National Academy of Sciences, vol. 115, pp. 12148–12151, Nov. 2018.DOI
  54. [54]M. Först, C. Manzoni, S. Kaiser, Y. Tomioka, Y. Tokura, R. Merlin, and A. Cavalleri, “Nonlinear phononics as an ultrafast route to lattice control,” Nature Physics, vol. 7, pp. 854–856, Nov. 2011.DOI
  55. [55]M. Först, R. Mankowsky, and A. Cavalleri, “Mode-Selective Control of the Crystal Lattice,” Accounts of Chemical Research, vol. 48, pp. 380–387, Feb. 2015.DOI
  56. [56]G. Khalsa and N. A. Benedek, “Ultrafast optically induced ferromagnetic/anti-ferromagnetic phase transition in GdTiO₃ from first principles,” npj Quantum Materials, vol. 3, p. 15, Mar. 2018.arxiv.org/abs/1803.01323
  57. [57]A. S. Disa, T. F. Nova, and A. Cavalleri, “Engineering crystal structures with light,” Nature Physics, vol. 17, pp. 1087–1092, Oct. 2021.DOI
  58. [58]We note that the condensation of Q̄q can create other Raman-active phonons via additional anharmonic lattice energy terms which may be accessed in some crystals through the conventional nonlinear phononics effect [70]. For simplicity, we ignore these pathways in our analysis since, though their inclusion is straightforward, it can quickly become unnecessarily cumbersome for our purposes.
  59. [59]T. F. Nova, A. S. Disa, M. Fechner, and A. Cavalleri, “Metastable ferroelectricity in optically strained SrTiO₃,” Science, vol. 364, pp. 1075–1079, June 2019.DOI
  60. [60]X. Li, T. Qi, J. Zhang, E. Baldini, J. Lu, A. M. Rappe, and K. A. Nelson, “Terahertz field-induced ferroelectricity in quantum paraelectric SrTiO₃,” Science, vol. 364, pp. 1079–1082, June 2019.DOI
  61. [61]P. A. Fleury, J. F. Scott, and J. M. Worlock, “Soft Phonon Modes and the 110°K Phase Transition in SrTiO₃,” Phys. Rev. Lett., vol. 21, pp. 16–19, July 1968.DOI
  62. [62]A. Stephenson, “XX. On Induced Stability,” The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, vol. 15, pp. 233–236, Apr. 1908.DOI
  63. [63]P. L. Kapitsa, “Pendulum with vibrating suspension,” Usp. Phys. Sciences, vol. 44, pp. 7–20, May 1951.
  64. [64]T. Mertelj and V. V. Kabanov, “Comment on “Ultrafast Reversal of the Ferroelectric Polarization”,” Phys. Rev. Lett., vol. 123, p. 129701, Sept. 2019.DOI
  65. [65]V. A. Abalmassov, “Ultrafast reversal of the ferroelectric polarization by a midinfrared pulse,” Phys. Rev. B, vol. 101, p. 014102, Jan. 2020.DOI
  66. [66]P. Chen, C. Paillard, H. J. Zhao, J. Íñiguez, and L. Bellaiche, “Deterministic control of ferroelectric polarization by ultrafast laser pulses,” Nature Communications, vol. 13, p. 2566, May 2022.DOI
  67. [67]A. M. Glazer, “The Classification of Tilted Octahedra in Perovskites,” Acta Crystallographica Section B, vol. 28, pp. 3384–3392, Nov. 1972.DOI
  68. [68]A. M. Glazer, “Simple Ways of Determining Perovskite Structures,” Acta Crystallographica Section A, vol. 31, pp. 756–762, Nov. 1975.DOI
  69. [69]F. Arecchi and R. Bonifacio, “Theory of Optical Maser Amplifiers,” IEEE Journal of Quantum Electronics, vol. 1, no. 4, pp. 169–178, 1965.

Paper details

vlnc:W_q9y62unp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents