
2
Furthermore, it has been recently pointed out its pos-
sible role in the superconducting transition in electron-
doped samples [23–25], and the possibility of realizing
dynamical multiferrocity upon driving it with circularly
polarized THz electric fields [26, 27]. Here we show that
besides the well-studied electronic Kerr effect (EKE), due
to off-resonant electronic transitions in wide-band insu-
lating STO [28] (Fig. 1a), a ionic contribution, associated
with the second-order excitation of the TO1phonon, is
present (Fig. 1b). Such IKE manifests itself with a siz-
able temperature dependence of the Kerr response, which
is unexpected for the EKE in a wide-band insulator. In
contrast, the IKE rapidly disappears by decreasing tem-
perature, due to the phonon frequency softening far be-
low the central frequency of the pump field. We are able
to clearly distinguish between EKE and IKE thanks to
a detailed theoretical description of the THz Kerr sig-
nal, which we retrieve experimentally as a function of
the light polarization and of the pump-probe time de-
lay tpp. So far, the regime of large lattice displacements
has been mainly investigated to exploit the ability of
infrared-active vibrational modes, driven by strong THz
pulses, to anharmonically couple to other phononic ex-
citations [17, 29–32]. Our work demonstrates that non-
linear phononic represents also a suitable knob to ma-
nipulate the refractive index of the material, providing
potentially an additional pathway to drive materials to-
wards metastable states which may not be accessible at
thermal equilibrium [33].
I. EXPERIMENTAL SETUP
Measurements are performed on a 500 µm-thick
SrTiO3crystal substrate (MTI Corporation), cut with
the [001] crystallographic direction out of plane. Broad-
band single-cycle THz radiation is generated in a DSTMS
crystal via optical rectification of a 40 fs-long, 800 µJ
near-infrared laser pulse centered at a wavelength of 1300
nm. The near-infrared pulse is obtained by optical para-
metric amplification from a 40 fs-long, 6.3 mJ pulse at
800 nm wavelength, produced by a 1 kHz regenerative
amplifier. As schematically shown in Fig. 2, the broad-
band THz pulses are filtered with a 3 THz band-pass
filter, resulting in a peak frequency of Øp/2π= 3 THz
(with peak amplitude of 330 kV/cm), and focused onto
the sample to a spot of approximately 500 µm in di-
ameter. The time-delayed probe beam, whose polariza-
tion is controlled by means of a nanoparticle linear film
polarizer, is a 40 fs-long pulse at 800 nm wavelength
normally incident onto the sample surface. A 100 µm-
thick BBO crystal (β-BaB2O4, Newlight Photonics) and
a shortpass filter are used for converting the probe wave-
length to 400 nm and increasing the signal-to-noise ra-
tio in temperature-dependent measurements. The probe
size at the sample is approximately 100 µm, substan-
P
𝑥
𝑦
𝜃
S
HWP
S
P
THz filter
Wollaston
Photodetectors
a
𝐸!
"(𝑡)
Figure 2. Schematics of the experimental setup. The THz
pulse is in red while the optical probe is in blue. Inset a:
schematic of the polarization geometry. The sample is rotated
in the (S, P ) plane, such that the ycrystallographic direction
forms a variable angle θwith respect to S. Here the probe
(blue arrow) polarization is fixed along S, while the pump
pulse (red arrow) is polarized along either Sor Pfor linearly
polarized light and in both directions for circularly polarized
light. Inset b: typical time-trace of ∆Γ at fixed θ= 67.5◦
(blue line) compared with the intensity profile of the linear
pump pulse (red line).
tially smaller than the THz pump. The half-wave plate
(HWP) located after the sample is used to detect the po-
larization rotation of the incoming field induced by the
non-linear response, and a Wollaston prism is used to
implement a balanced detection scheme with two pho-
todiodes. The signals from the photodiodes are fed to
a lock-in amplifier, whose reference frequency (500 Hz)
comes from a chopper mounted in the pump path before
the DSTMS crystal. Data are collected as a function of
the tpp time-delay as well as a function of the angle θ
that the probe polarization direction forms with respect
to the main crystallographic axes. For convenience, here
we assume that the probe beam is kept fixed along the S
direction, such that Epr(t)=(Epr(t) sin θ, Epr(t) cos θ),
while the pump field Epcan be either linearly polarized
(along Sor Pdirection) or circularly polarized [26], by
placing a quarter-wave plate after the band-pass filter
(see Appendix A), such that in general
Ep(t) = Ep,S (t) sin θ+Ep,P (t) cos θ
Ep,S (t) cos θ−Ep,P (t) sin θ.(1)
In the presence of Kerr rotation, the probe field trans-
mitted through the sample ˜
E(t) acquires a finite Pcom-
ponent. The half-wave plate rotates ˜
E(t) by 45◦with
respect to Pdirection and the outgoing signal reaching
the two photodetectors (Γ1,Γ2) reads [34]
Γ1
Γ2∝1 1
1−1˜
ES
˜
EP.(2)