LIST OF TABLES
5.3 Microscopic Model and Dynamical Simulation Microscopic ModelMicroscopic Model
due to a stronger electron-phonon coupling. First principle calculations suggested that all the π΄π phonons in Ca2RuO4involve rotation or tilting of octahedra (M.-C.
Lee, C. H. Kim, Kwak, J. Kim, et al., 2018), which inevitably changes the Ru-O bond length and generatesπΈπ-symmetry modulation of each octahedron. Therefore, the clear dichotomy shows thatππ andππ eigenmodes of 3.7 THz phonon exhibit distinct behaviors as the QO is manipulated by light.
5.3 Microscopic Model and Dynamical Simulation
a
x z y
b
PES QΞ΅
QΞΈ
QΞ΅
QΞΈ
10 8
P Amp.6
-2 -1 0 1 2 F>Fc
F<Fc
2 1 0
2.0 1.5 1.0 0.5 0.0
-0.5 Time (ps) F>Fc
F<Fc
Q ΞΈQ Ξ΅
10 8 6 4 2
QOCP Amp. 0
1200 800 400
0 F
Q Q b
c
d ( )
( )
( ) ( )
Figure 5.7: Microscopic model simulation results. (a) Schematic of PES of the quadrupolar ordered phase calculated from the microscopic model (Methods). The corresponding real-space pseudospin distribution and lattice distortion for each valley are shown by side. (b)(c) Time evolution of ππ and πππ π ππ ππ with pump fluence lower and higher than the critical fluence. (d) Pump fluence dependence of amplitude of QOCP alongππ (red) andππ (blue) coordinates calculated based on the dynamical simulation. The dashed line denotes the first switching critical fluence.
and amplitude proportional to the fluence πΉ with a scaling factor π΄. By further including a damping term with constantπΎwhich can be experimentally determined, we can write out the equation of motion ofππandππ:
π2ππ/π(π‘) π π‘2 +2πΎ
πππ/π(π‘) π π‘
+ ππ(ππ(π‘), ππ(π‘))
πππ/π(π‘) = π΄πΉexp
4 ln(2)π‘2 π2
. (5.1) Simulation results
The time evolution of the structural order parametersππ andππ pumped with two characteristic fluences (πΉ > πΉπ and πΉ < πΉπ) are shown in Figures 5.7(b) and (c).
As πΉ > πΉπ, a clear change of sign in offset of both πs can be unambiguously observed, which underlines a switch of QO from the initial valley to another hidden valley. By performing FFT to time traces after the system is stabilized as we vary the only free parameter πΉ, we can track the fluence dependence of the amplitude of ππ and ππ modes. As expected, when πΉ < πΉπ, the amplitude of both ππ and ππ coordinates exhibits the same linear dependence akin to a conventional phonon;
as the system transiently evolves into the two excited states accompanied by the QO switching (πΉ > πΉπ), phonons along both coordinates display an deviation from perfect linearity and the nonlinearity ofππ is more drastic thanππ [Figure 5.7(d)].
Not only does this result suggest that the nonlinear behavior of the coupled phonon signifies the ultrafast switch of QO, but it reproduces the probe-energy-dependent measurements in Figure 5.6, whereππ-sensitive probes indeed show a more dramatic deviation from linearity than theππ-sensitive probes.
Note that there is an aperiodic oscillation of QOCP amplitude over fluence as πΉ > πΉπ. The nonmonotonic behavior arises from the back-and-forth switching between different valleys as πΉ increases. This indicates πΉπ corresponds to the fluence where the first reversal occurs πΉπ,π=
1. When πΉ is slightly higher than πΉπ
1, the system relaxes to and stays in the π₯/π¦-compressed valley. When πΉ > πΉπ
2, the system will relax back to the original valley. With even higher πΉ, the system can switch between different valleys back and forth transiently before it settles into one valley. Now we provide an explanation on why the switch between valleys will generate the aperiodic oscillation. Let us consider two fluences: πΉ
1 = πΉπ
1βπΏ and πΉ2 = πΉπ
1+πΏ, where πΏ is a small positive value. Since πΉ
2 > πΉ
1, the system will roll over the barrier and move to the new valley, which obviously travels a longer distance and thus the damping will decrease the amplitude forπΉ
2case more thanπΉ
1
case. Thus we will find the phonon amplitude atπΉ
2is smaller thanπΉ
1. AsπΉfurther increases, the larger energy dumped into the system will compensate the damping loss and the phonon amplitude will slowly increase until it reaches the boundary between the two valleys again at πΉ = πΉπ
2. Then the new reversal occurs and the phonon amplitude will drop again. Therefore, the QOCP amplitude will show a sudden change at eachπΉπ,π and thus depicts an oscillatory behavior as a function of πΉ.
Also note that in the minimal model we neglect the intrinsic energetic difference between ππ₯ π¦ and the nearly degenerate ππ₯ π§/π¦ π§-dominated states, which arises from noncubic environment induced by the octahedron distortion and the planar cornered- shared lattice structure. We can include the tetragonal splitting by phenomeno- logically adding a linear term ππΈππΈ π to tilt the PES so that the ground state of compressedπ§βaxis is uniquely favored. Based on this assumption, the JT effect will become pseudo-JT effect due to the energetic nondegeneracy in the parent phase (Bersuker, 2006). After including the tetragonal splitting, we find that the original ππ₯ π¦-dominated valley will almost always be reached eventually after about 3 ps in-
1.0
0.8
0.6
0.4
0.2
0.0
FFT Amplitude (a.u.)
4.0 3.0 2.0
FFT Frequency (THz) QEΞΈ sim.
Ξ±-peak probe exp. 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00
FFT Amplitude (a.u.)
4.5 4.0 3.5 3.0 2.5 2.0
FFT Frequency (THz) 280K
220 K
130 K 80 K
Ξ²-peak probe exp. 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00
4.5 4.0 3.5 3.0 2.5 2.0
FFT Frequency (THz) g = 0
g = 0.5g0
g = 0.85g0
g = g0
QEΞ΅ sim.
(a) (b)
Figure 5.8: Illustration of the 2.5 THz hump. (a) Comparison of FFT spectra of QOCP between experiment and simulation with twoππ₯ π¦βππ₯ π§/π¦ π§-peak probes (0.69, 0.92 eV) at 80 K. b. Comparison of FFT spectra of QOCP between experiment and simulation with 1.55 eVππ₯ π§/π¦ π§βππ₯ π§/π¦ π§-peak probe at various temperatures. All the experimental data were taken at a pump fluence of 15 mJ/cm2. The simulated FFT spectra are calculated by changingπin a form ofβοΈ
1βπ/πππ. The relative value of πat each temperature is denoted. The data are scaled and offset vertically for better comparison.
dependent of πΉ, but a transient switch between different QOs still occurs. We note that πΉπ in the presence of tetragonal splitting is larger than that without tetragonal splitting becauseππ₯ π§andππ¦ π§valleys are elevated and thus harder to reach. However, despite the discrepancy, the fluence dependence of QOCP amplitude and frequency is qualitatively identical to the case with tetragonal splitting ignored.
We can also simulate the temperature dependence of QOCP amplitude and πΉπ by assuming that microscopically the JT coupling constant π, which determines the spacing and depth of different valleys, decreases as temperature increases in an order-parameter formβοΈ
1βπ/πππ. By introducing a temperature dependence ofπ, our model confirms a primary order-parameter-like behavior βοΈ
1βπ/πππ of both QOCP amplitude andπΉπ, in perfect agreement with the experimental data in Figure 5.5(b). These results also underpin thatπΉπ can be used to characterize QO.
We note that in both the experimental and simulational results, a hump shows up at the low-frequency shoulder of the QOCP peak. According to previous Raman spectroscopy results (Rho et al., 2005), all theπ΄
1πmodes have been identified with energies all above 3.7 THz. Since our isotropic reflectivity measurement can only probe π΄
1π phonons according to symmetry, we can rule out the possibility that this damped mode is a phonon. Also, this mode exists even aboveππ but disappears at
πππ [Figure 5.6(b)], indicating that it is not a magnon but should be related to QO.
We thus attribute this hump feature to the non-harmonicity of the PES. As SOC smoothens the barrier, the local curvature at the boundary between the two valleys decreases, rendering the PES a local anharmonicity. Therefore, when the system is transiently switched between the valleys, a low energy component emerges from the reduced curvature around the anharmonic part of PES. If this hypothesis is true, we would expect that this low frequency hump should emerge in bothππΈ π andππΈ π only atπ < πππ and πΉ > πΉπ. This is indeed corroborated by both our simulation and experiments with good agreement [Figure 5.8]. We would also expect that the amplitude of the hump should increase withπΉ. Our current experimental setup can hardly resolve the dynamics of this mode unambiguously over the entire fluence and temperature range due to unideal SNR. This speculation thus calls for detailed measurements with better SNR in the future.