Nanosecond Ferroelectric Switching of Intralayer Excitons in Bilayer 3R-MoS2

through Coulomb Engineering

Jing Liang1,2#, Yuan Xie1,2#, Dongyang Yang1,2#, Shangyi Guo1,2, Kenji Watanabe3, Takashi
Taniguchi4, Jerry I. Dadap1,2, David Jones1,2, Ziliang Ye1,2*

1
Quantum Matter Institute, The University of British Columbia, Vancouver, BC V6T 1Z4, Canada
2
Department of Physics and Astronomy, The University of British Columbia, Vancouver, BC V6T
1Z1, Canada
3
Research Center for Functional Materials, National Institute for Materials Science, 1-1 Namiki,
Tsukuba 305-0044, Japan
4
International Center for Materials Nanoarchitectonics, National Institute for Materials Science,
1-1 Namiki, Tsukuba 305-0044, Japan

# These authors contributed equally to this manuscript.

* Correspondence: [email protected]

1 / 21
Abstract

High-speed, non-volatile tunability is critical for advancing reconfigurable photonic devices used
in neuromorphic information processing, sensing, and communication. Despite significant
progress in developing phase change and ferroelectric materials, achieving highly efficient,
reversible, rapid switching of optical properties has remained a challenge. Recently, sliding
ferroelectricity has been discovered in 2D semiconductors, which also host strong excitonic effects.
Here, we demonstrate that these materials enable nanosecond ferroelectric switching in the
complex refractive index, largely impacting their linear optical responses. The maximum index
modulation reaches about 4, resulting in a relative reflectance change exceeding 85%. Both on and
off switching occurs within 2.5 nanoseconds, with switching energy at femtojoule levels. The
switching mechanism is driven by tuning the excitonic peak splitting of a rhombohedral
molybdenum disulfide bilayer in an engineered Coulomb screening environment. This new
switching mechanism establishes a new direction for developing high-speed, non-volatile optical
memories and highly efficient, compact reconfigurable photonic devices. Additionally, the
demonstrated imaging technique offers a rapid method to characterize domains and domain walls
in 2D semiconductors with rhombohedral stacking.

Main

Switchable optical materials can enable non-volatile tunability in photonic devices with
applications that include neuromorphic computing and artificial intelligence, quantum information
processing, optical communications, and optical sensing1-6. Unlike volatile tuning, no power is
needed to maintain a switched state in a non-volatile device, which can therefore serve as an optical
memory/memristor7,8. To this end, much effort has been focused on the large refractive index
tuning in chalcogenide-based phase change materials, which requires pico- to nano-joules of
thermal energy to switch9,10. Recently, barium titanate thin films have been developed as a
promising ferroelectric material to achieve a switchable Pockels effect with record-low switching
energy and a refractive index modulation up to 10-3 under a constant electric field11. Here we

2 / 21
present a new scheme for non-volatile switching of the complex refractive index associated with
the excitonic resonances in two-dimensional (2D) semiconductors. The modulation in both the
refractive index and extinction coefficient is about 4, resulting in a relative reflectance change
exceeding 85%. Since no heating is involved in our switching mechanism, the switching energy
can be reduced to femtojoule levels, with both on and off switching occurring in less than 2.5 ns,
making it the fastest non-volatile tunability reported for optical materials7. Our scheme relies on
the dynamic tuning of the Coulomb screening of strong excitonic effects in 2D transition metal
dichalcogenides (TMDs) that are rhombohedrally stacked to enable sliding ferroelectricity.

Sliding ferroelectricity is a hysteretic phenomenon in 2D van der Waals materials with specific
stacking orders, where an electric field induces one layer of the material to slide relative to the
other due to an interfacial polarization arising from the interlayer coupling12-15. This effect can
occur in traditionally non-ferroelectric materials and has been electrically probed,16-21 with atomic
structure changes confirmed by scanning probe microscopy22 and electron microscopy23,24.
Initially found in artificially stacked 2D materials with marginal twists, interfacial polarization and
its switching have recently also been observed in chemically synthesized rhombohedral (3R)
TMDs25-30, where preexisting domain walls play a key role in lowering the coercive field. Since
atomically thin TMDs are semiconductors with much reduced Coulomb screening and
extraordinary excitonic effects31-34, the switch in the stacking configuration can also be reflected
in the excitonic properties as optical contrasts, which have been observed between intermediate
states in thick layers with multiple interfaces26,35. Nevertheless, in an intrinsic 3R bilayer with
only one interface, the stable stackings before and after the switch form a mirror image pair (i.e.,
AB and BA), rendering them indistinguishable with conventional linear optical probes. Optically,
the AB and BA stackings have therefore only been resolved by photoluminescence spectroscopy
under an external electric field26,36, which are incompatible with the applications discussed above.

On the other hand, the mirror symmetry between AB and BA stacking configurations can be
broken by introducing an asymmetric dielectric environment. Previous work has shown that the
Coulomb interaction in atomically thin TMDs can be modified by engineering the local

3 / 21
environment37-40. Large Coulomb screening can significantly reduce the quasiparticle bandgap
and exciton binding energy, and such a screening effect has a distance sensitivity down to a single
monolayer37. Consequently, when a TMD bilayer is placed on a substrate, the lower layer will
experience greater screening than the upper layer that is not in direct contact with the dielectric
substrate (Fig. 1a). If the intrinsic excitonic responses in two layers are identical, such asymmetric
screening will result in no net contrast between the two stackings. However, the interlayer coupling
in a bilayer 3R-TMD naturally breaks the layer symmetry, enabling significant non-volatile tuning
of the complex refractive index near the excitonic resonance.

Stacking-dependent excitonic resonances

In a rhombohedral MoS2 bilayer, the two layers experience different chemical environments
and can have distinct exciton peaks36,41. Here we define the A layer as having the molybdenum
atom aligned with the sulfide atom in the B layer (Fig. 1b). Under this definition, the molybdenum
atom in the B layer is not aligned with the sulfide atom in the A layer. Consequently, the lowest
excitonic resonance in the monolayer MoS2 splits into two peaks in a rhombohedral bilayer, each
peak originating from an intralayer exciton composed of an electron and a hole located within the
same layer. We have previously confirmed that the high-energy peak of this pair originates from
the K/K’ point transition in the A layer (X ) while the low-energy peak arises from the B layer
(X )41. In a symmetric dielectric environment, the energy splitting between the two exciton peaks,
𝛿 , is determined by the interlayer coupling strength, independent of the stacking configuration.
However, this degeneracy is lifted when the dielectric environments of the upper and lower layers
differ. For example, Coulomb screening from a substrate, as depicted in Fig. 1b, modifies the
energy splitting. In one stacking configuration where the B layer is at the bottom of the A layer
and is on top of the substrate, the low-energy X experiences a redshift (∆) and the peak splitting
increases (𝛿 𝛿 ∆). Conversely, when the A layer is at the bottom of the B layer, the high-
energy X experiences a redshift, resulting in a decrease in energy splitting (𝛿 𝛿 ∆). As a
result, when the stacking is switched through sliding, the energy splitting between the two excitons
changes accordingly. In real materials, the top layer may also experience some degree of screening

4 / 21
from the substrate, but the peak separation contrast between the two stackings should persist.

Following this picture, we studied a 3R-MoS2 bilayer directly exfoliated onto a SiO2/Si
substrate. The coexistence of both stacking configurations is reflected in the surface potential
contrast, mapped by electrostatic force microscopy (EFM) (Fig. 2a). Due to the polarization-
induced interlayer potential, the surface potential in the AB domain is higher than that in the BA
domain42. Subsequently, we conducted optical reflectance contrast (RC) spectroscopy at 1.6 K to
investigate the excitonic responses of the two differently stacked domains, marked by orange and
blue dots. Fig. 2b presents the RC spectra and their second energy derivatives taken at these two
locations, respectively. The AB domain clearly exhibits two distinguishable peaks while the BA
domain displays an effective single peak due to the unresolved peak splitting, as illustrated in the
lower right panel of Fig. 1b.

To quantify the asymmetric screening effect, we performed a mapping of the RC across the
sample (Fig. 2c). Although optical imaging has lower resolution due to the diffraction limit, the
reflectance at 1.919-eV photon energy shows a contrast similar to the EFM map. The statistics of
the excitonic peak splitting among AB domains yield an average of 𝛿 35 3 meV (Figs. 2d,
e). In a separate experiment, we fabricated a switchable 3R-MoS2 bilayer device in a symmetric
dielectric environment (Fig. S1). The excitonic peak splitting in both stackings is about 15 meV,
which is the intrinsic energy splitting (𝛿 ) of the rhombohedral bilayer. Consequently, we conclude
that the extra redshift ∆ caused by the SiO2 substrate is approximately 20 meV. This value suggests
𝛿 is about 5 meV smaller than the excitonic linewidth in our unencapsulated sample (~40 meV),
which explains the unresolved peak splitting in Fig. 2b. We found that the excitonic contrast
induced by the asymmetric Coulomb screening persists up to room temperature, enabling the direct
visualization of the AB and BA domains using an optical microscope with a suitable band pass
filter tuned to the excitonic resonance energy (Fig. 2f and S2). Such an optical contrast can be used
to rapidly screen the domain structure in chemically synthesized films on a wafer scale, benefiting
the fasting growing 3R-TMD crystal growth community43,44.

5 / 21
Dynamic tuning of excitonic peak splitting

Next, we show that the excitonic peak splitting can be dynamically tuned in a device that can
switch the stacking configuration through sliding ferroelectricity (Fig. 3a). The device consists of
a MoS2 bilayer, with the lower layer adjacent to a few-layer graphene (FLG) and the upper layer
next to an h-BN layer, forming an asymmetric Coulomb screening environment (see the optical
image of the device in Fig. S3). An additional FLG is placed on top to serve as an electrode for
applying a vertical electric field. The entire stack is placed on a SiO2/Si substrate. The bilayer
contains a preexisting domain wall (DW), and as we have previously demonstrated, an out-of-
plane electric field can induce an in-plane pressure on this DW since the neighboring domains
have opposite interfacial polarizations26. When the induced free energy difference across the DW
exceeds the local pinning potential, the DW can be released and sweeps through a large area until
trapped by another pinning center. This stacking configuration switching is analogous to
polarization poling in conventional ferroelectrics, except that the generation of DWs is challenging
in sliding ferroelectrics and most switching relies on the motion of preexisting DWs26. Since
excitonic peak splitting is enhanced in one stacking and reduced in the other, we can optically
probe the stacking configuration by measuring the RC spectrum after electrical poling. Here the
RC spectroscopy measurements are used to characterize the steady configurations. The switching
dynamics are measured using a single wavelength laser resonant with the exciton energy as
discussed in the next section.

The electric poling field (E ) dependent RC spectra are measured at the center of the device

using a broadband white light source focused down to a micron-sized spot (Fig. 3b). When the
poling field is scanned from positive to negative, an unresolved single peak at ~1.909 eV is

observed at the positive limit, indicating the initial stacking is BA. When E exceeds the coercive

field in the negative direction (𝐸 0.042 V/nm), a transition in the spectrum occurs, revealing
two excitonic peaks with a 22-meV splitting. Such a transition is consistent with the substrate-only
experiment and indicates that the stacking configuration has switched to AB. The energy splitting

6 / 21
differences between the two types of asymmetric dielectric environments are attributed to
variations in Coulomb screening, which depend on both the absolute dielectric constant values and
the contrasts between those of the substrate and superstrate37. Furthermore, polarization-switching-
induced free carrier density changes in the FLG layer are estimated to have a negligible impact (~
45
1 meV, see details in the Supplementary Information) on 𝛿 and 𝛿 . In contrast, no spectral
change is observed in the device with symmetric hBN encapsulation (Fig. S1). Similarly, a forward

E scan shows an opposite stacking switch from AB to BA when the external field exceeds the

positive coercive field (𝐸 0.038 V/nm). The coercive fields are not symmetric since they are
mostly determined by the trapping potential of the initial and final pinning centers. As the
reflectance mapping below will show, the intermediate state is caused by the pinning centers within
the focus spot, which can trap DWs under a small poling field with variations across devices (Fig.
S4-S5). These intermediate states can be engineered for multi-bit optical memristors in the future7.
Compared to the case of the SiO2 substrate, δ is smaller in the encapsulated device because the
peak separation is determined by the relative screening strength between the substrate and the
capping layer.

At the photon energy of 1.909 eV, which is near the center of the unresolved peak and the
middle of the double peak, a large reflectance contrast is observed before and after the switch. The
BA state yields a reflectance, RBA, of 8.56%, which is over six times higher than that of the AB
state (RAB=1.23%), corresponding to a relative change, (RBA-RAB)/RBA, of 86%. (The
unnormalized difference spectrum is plotted in Fig. S6). The switch is non-volatile with a clear
hysteretic dependence on the poling field (Fig. 3c). Given the device geometry and the observed
coercive field, the electrostatic energy needed to switch the stacking in our nonoptimized device
is less than one picojoule, indicating its potential for reconfigurable photonic applications. The
switching energy can be further lowered by reducing the device capacitance in future designs.

Additionally, we retrieved the complex refractive index change in the 3R-MoS2 bilayer by
fitting the reflectance spectra in the AB and BA stackings (Fig. 3d). The spectra near the excitonic
resonance are fitted with a dielectric function model composed of two Lorentz peaks and a

7 / 21
background dielectric constant, while the multilayered environment is taken into account via a
transfer-matrix model46-48 (see details in the Supplementary Information). The retrieved refractive
index (n) and extinction coefficient (k) are plotted in Fig. 3e. Near the excitonic resonance, a
maximum modulation of about 4 is observed in both optical constants (Fig. S6). The sum of the
two oscillator strengths remains nearly constant during the switching event. To explore the
compatibility of our scheme with various applications, we also investigated the switchability of
the device at room temperature (Fig. 3f and S7), where the peak splitting is no longer resolvable
due to thermal broadening. Since the merged peaks have very different effective linewidths, a sharp
change can still be observed in the reflectance spectra when the poling field exceeds the coercive
field. At room temperature, the reflectance changes from 1.42% to 1.95% at the resonance peak,
resulting in a relative change of 27%. With the light currently propagating vertically through a 1.4-
nm thin active layer, we expect the modulation depth can be further improved by integrating the
material into an optimized photonic device. Recent studies have shown that sliding ferroelectricity-
based memory can be retained for months and endure over 10 switching cycles20,21, making it
suitable for most reconfigurable photonic devices.

Switching speed

The large reflectance contrast between the AB and BA stackings can also reveal how the
domain and DW evolve with the poling field (Fig. 4a). We focused a 648.2-nm continuous wave
laser on a diffraction-limited spot on the device and measured its reflected intensity using a high-
speed avalanche photodiode (APD) while the sample is raster-scanned. Initially, the device is poled
with a large negative field and most areas exhibited low reflectivity, indicating a uniform AB
stacking, except for some highly reflective regions attributed to bubbles, wrinkles, and flake edges
(see the optical image in Fig. S3). After a poling voltage of 0.55 V is applied to the device, a small
BA domain with higher reflectance emerged near the edge of the FLG electrode. Since it is very
challenging to generate new DWs at such a low field, we attribute the new domain to the release
and leftward movement of a preexisting DW pinned at the edge. As the poling field increased, the
BA domain expanded further, eventually reaching the bubble region. Finally, at 0.9 V, the DW

8 / 21
moved downward and stopped at the lowest point, converting half of the sample into a
homogeneous BA domain. The spectra in Fig. 3, measured in the middle of this switched BA
domain (marked in Fig. 4a and S3), reflects the propagation of this single domain wall. When a
negative poling field exceeding the negative coercive field was applied, the BA domain shrank as
the DW returned to its initial edge. We note that all states remained non-volatile as the reflectance
mapping was carried out with the electric poling field turned off.

Finally, we performed the first real-time optical measurement of the switching dynamics in
sliding ferroelectrics. We continuously monitored the reflected laser beam intensity (top blue panel
in Fig. 4b) while applying a bipolar square wave voltage to the device through an arbitrary
waveform generator (AWG) (bottom red panel). The measurement spot is the same as marked in
Fig. 4a. The electric field level was chosen to be well above the coercive field but small enough to
keep the Fermi level in the MoS2 bandgap, so that the doping effect is minimized in such a single-
gate device (Fig. S8). In addition, since the electric field used in the speed measurement does not
match the resonance condition, interlayer excitons do not affect the transient measurements (Fig.
S8). Although the square wave voltage is not optimized due to a limited AWG bandwidth as well
as impedance mismatch, the two signals displayed clear correlation with distinct high/low contrast.
The measured dynamic contrast is similar to the contrast between the steady on and off states at
zero electric field, confirming a negligible contribution from the field-induced doping. Both rise
and fall times in the optical reflectance signal are about 2.5 ns (shown in the insets of Fig. 4b),
limited by the bandwidth of the APD and oscilloscope employed (SI). Interestingly, the step in the
optical reflectance is significantly shorter and cleaner than that in the electric field, indicating that
the abrupt change was caused by a fast-moving DW released by an electric field that exceeds the
pinning threshold, and the duration spent by the DW within the probing spot was very short. The
intrinsic switching speed is thus determined only by how quickly the DW sweeps across the focus
spot. Our real-time measurement results therefore set a lower bound on the DW velocity of
approximately 500 m/s. Assuming a free DW can propagate at the speed of sound20,21, the intrinsic
switching time for a diffraction-limited focus spot is expected to be on the order of 100s of ps.

9 / 21
 In summary, we have shown that nanosecond ferroelectric tuning of the complex refractive
index with a magnitude of approximately four is achievable near the excitonic resonance of a 3R-
MoS2 bilayer by introducing an asymmetric Coulomb screening environment. Without further
optimization of the photonic environment, a large non-volatile reflectance change was observed in
our device with a high energy efficiency. Compared to the contrast observed between intermediate
states in thick layers with multiple interfaces26,35, the engineered optical contrast between AB and
BA stackings in the bilayer with a single interface is larger and more robust. In the future, one can
improve the scalability of the device by engineering the domain walls and pinning center in
chemically synthesized films. The integration of this phenomenon into photonic waveguides can
also bring about new functionalities for integrated photonics. Finally, our optical imaging method
enables the exploration of domain wall dynamics and rapid characterization of the domain
distribution in 3R-TMDs with a high throughput.

Acknowledgement

We acknowledge support from the Natural Sciences and Engineering Research Council of
Canada, Canada Foundation for Innovation, New Frontiers in Research Fund, Canada First
Research Excellence Fund, and Max Planck–UBC–UTokyo Centre for Quantum Materials. Z.Y.
is also supported by the Canada Research Chairs Program. K.W. and T.T. acknowledge support
from JSPS KAKENHI (Grant Numbers 19H05790, 20H00354 and 21H05233).

Author Contributions

Z.Y. conceived and supervised the project. J.L., S.G., Y.X., D.Y., K.W. and T.T. prepared the
materials. J.L. conducted the EFM characterization. J.L. and S.G. performed the optical
spectroscopy in bare flakes. J.L. and Y.X. fabricated the switchable device and performed the
measurements with assistance from D.Y. and D.J.. Data analysis was carried out by J.L., Y.X., D.Y.,
and Z.Y.. J.L., J.D., D.J., and Z.Y. wrote the manuscript with input from all co-authors. J.L., Y.X.,
and D.Y. contributed to this work equally.

10 / 21
Competing interests

The authors declare no competing interests.

11 / 21
Figures and captions:

a b
Side view
A B

B A

substrate
Exciton binding

upper

lower

Photon energy

Figure 1 ︱Tuning the excitonic peak splitting through asymmetric dielectric screening. (a)
Schematic illustration of the screening effect of a dielectric substrate on the intralayer exciton in
the lower layer of a TMD bilayer. The exciton binding energy is lowered due to the reduced
Coulomb interaction. (b) Schematic of the stacking-dependent excitonic peak splitting in a 3R-
MoS2 bilayer on a dielectric substrate. Black arrows indicate the direction of spontaneous
polarization. The left and right halves illustrate the AB and BA stackings, respectively. The
intrinsic peak splitting and screening-induced redshift are denoted by δ and Δ, respectively.

12 / 21
Figure 2︱Excitonic contrast between AB and BA domains. (a) Electrostatic force microscopy
(EFM) map of a 3R-MoS2 bilayer with mixed AB and BA domains. The flake is directly exfoliated
onto a SiO2/Si substrate. The EFM contrast is caused by the surface potential difference between
AB and BA stackings. Scale bar: 4 μm. (b) Reflectance contrast (RC) spectra and their second
energy derivatives for the BA and AB domains. The measurement spots are marked by orange and
blue dots in panel a. Two distinct exciton peaks are observed in the AB domain (XA and XB), while
the BA domain exhibits only one unresolved peak. Optical measurements are taken at 1.6 K. (c)
Optical reflectance mapping (hν 1.919 eV) shows a similar contrast as in the red box in panel a.
Scale bar: 2 μm. (d, e) Peak energy statistics of XA and XB indicate a splitting energy of 35 3
meV. (f) Optical images of the 3R-MoS2 bilayer captured without any filter (top), and with a band-
pass filter centered at 670 nm with a bandwidth of 10 nm (bottom). Scale bar: 4 μm.

13 / 21
Figure 3︱Non-volatile switching of excitonic resonance. (a) Schematic of a 3R-MoS2 bilayer
screened asymmetrically by few-layer graphene (FLG) below and an h-BN layer above. The top
FLG serves as an electrode to apply a vertical electric field. The bilayer MoS2 has a preexisting
domain wall (DW) that can be released when the electric field surpasses the pinning threshold.
Black arrows indicate the direction of spontaneous polarization. (b) Local reflectance spectrum
under a broadband white light source focused on a μm-sized spot, as a function of the poling field
in both negative (left) and positive (right) scan directions. An electric poling field is applied for
one second and then turned off, followed by a reflectance measurement. All measurements are
performed at 4.5 K unless specified otherwise. (c) Reflectance at 1.909 eV as the poling field is
scanned in negative (blue) and positive (red) directions. A hysteretic loop with a large reflectance
contrast is observed and the origin of the intermediate steps is discussed in the main text. (d, e)
The refractive index (n) and extinction coefficient (𝑘) for AB and BA stackings near the excitonic
resonances are extracted by fitting the reflectance spectra. The maximum modulations in both
optical constants are about 4. (f) At room temperature, the excitonic peak splitting is not resolvable
due to thermal broadening, but the reflectance at the resonance peak can be switched between 1.42%
and 1.95%, giving rise to a modulation depth of 27%.

14 / 21
Figure 4︱Nanosecond reflectance switching via domain wall propagation. (a) Optical
reflectance mapping (hν 1.904 eV) after different poling fields were applied shows that the DW
was initially pinned at the upper right edge and propagated leftward and downward after release.
The initial, intermediate, and final states are repeatable and non-volatile, as DW is trapped by the
same local pinning centers. Other strongly reflective features in the map match the locations of the
bubbles, wrinkles, and flake edges in the device. The circle marks the measurement spot for data
presented in panel b and Fig. 3. Scale bar: 3 μm. (b) A clear correlation is observed in the time-
domain between the optical reflectance and the applied electric field at both the fall and rise edges.
The insets show a transient time of about 2.5 ns, limited by the measurement instruments. The
abrupt step in the APD signal is significantly shorter than the electric field transition, confirming
that the switching is caused by the sudden release and rapid propagation of the single domain as
seen in panel a.

15 / 21
Methods

Material and device fabrication: The 3R-MoS2 bilayer device featuring an asymmetric screening
environment was fabricated using a layer-by-layer dry transfer technique under ambient conditions.
Single crystals of 3R-MoS2 were purchased from HQ Graphene.

EFM characterization: Electrostatic force microscopy (EFM) was conducted using a Molecular
Vista atomic force microscope (AFM) in the tapping mode41. Gold-coated AFM tips (Tap300GB-
G) with a first mechanical resonance of ~ 300 kHz, a second resonance of ~ 1500 kHz, and a force
constant of ~ 40 N/m were used. For topographic information, the cantilever was driven by the
piezoelectric actuator at its second resonance with an oscillation amplitude close to 2 nm. To
measure the surface potential contrast, the AFM tip was grounded and an AC voltage oscillating
at the fundamental resonance with an amplitude of 0.2 V was applied to the highly conductive Si
substrate. The resulting electrostatic force was measured by the oscillation amplitude of the
cantilever at its fundamental resonance.

Optical spectroscopy: Reflection spectroscopy was measured at the base temperature of either an
Attodry-2100 (1.6 K) or CryoAdvance-50 (4.5 K) closed-cycle optical cryostat. A spatially filtered
broadband tungsten halogen lamp was focused on a 𝜇m-sized spot on the device via a microscope
objective (N.A. is 0.65 for Attodry-2100 and 0.5 for CryoAdvance-50). The reflected light was
spectrally resolved using a spectrometer (Princeton Instruments) equipped with a thermoelectric-
cooled CCD camera. The reflectance contrast spectra in the bare flake are normalized to the
reference spectrum collected outside the flake. The reflectance spectra in the encapsulated device
are normalized to the reference spectrum from a silver mirror.

Switching speed measurement: The switching speed was measured in the reflection geometry
shown in Fig. S9. The emission from a temperature-controlled diode laser with a wavelength of
648.2 nm was focused onto a diffraction-limited spot of about 1.3-μm diameter on the device,
which was cooled to the base temperature of CryoAdvance-50. The focus intensity was kept below
100 kW/cm2 to avoid saturation. The reflected light intensity was detected by a silicon avalanche

16 / 21
photodetector with a 400-MHz bandwidth (Thorlabs APD430A). The APD voltage was recorded
by a 4-GHz digital oscilloscope (Tektronix MSO64B) using the single trigger mode. The square
wave output by an arbitrary waveform generator (Stanford Research DS345) was used to switch
the device and trigger the oscilloscope. The APD-oscilloscope combination provides a time
resolution of about 2.5 ns, as reflected in the calibration of the instrument response function using
a femtosecond laser (Fig. S10). The lower bound of the domain wall velocity was estimated from
the ratio of the focus diameter to the instrument response time.

References

(1) Bogaerts, W., Pérez, D., Capmany, J., Miller, D. A. B., Poon, J., Englund, D., Morichetti, F. &
Melloni, A. Programmable photonic circuits. Nature 2020, 586, 207-216.

(2) Shen, Y., Harris, N. C., Skirlo, S., Prabhu, M., Baehr-Jones, T., Hochberg, M., Sun, X., Zhao,
S., Larochelle, H., Englund, D. et al. Deep learning with coherent nanophotonic circuits.
Nature Photonics 2017, 11, 441-446.

(3) Zhang, Y., Fowler, C., Liang, J., Azhar, B., Shalaginov, M. Y., Deckoff-Jones, S., An, S., Chou,
J. B., Roberts, C. M., Liberman, V. et al. Electrically reconfigurable non-volatile metasurface
using low-loss optical phase-change material. Nature Nanotechnology 2021, 16, 661-666.

(4) Wang, Y., Landreman, P., Schoen, D., Okabe, K., Marshall, A., Celano, U., Wong, H. S. P.,
Park, J. & Brongersma, M. L. Electrical tuning of phase-change antennas and metasurfaces.
Nature Nanotechnology 2021, 16, 667-672.

(5) Shastri, B. J., Tait, A. N., Ferreira de Lima, T., Pernice, W. H. P., Bhaskaran, H., Wright, C. D.
& Prucnal, P. R. Photonics for artificial intelligence and neuromorphic computing. Nature
Photonics 2021, 15, 102-114.

(6) Wu, C., Yu, H., Lee, S., Peng, R., Takeuchi, I. & Li, M. Programmable phase-change
metasurfaces on waveguides for multimode photonic convolutional neural network. Nature
Communications 2021, 12, 96.

(7) Youngblood, N., Ríos Ocampo, C. A., Pernice, W. H. P. & Bhaskaran, H. Integrated optical
memristors. Nature Photonics 2023, 17, 561-572.

(8) McMahon, P. L. The physics of optical computing. Nature Reviews Physics 2023, 5, 717-734.

17 / 21
(9) Wuttig, M., Bhaskaran, H. & Taubner, T. Phase-change materials for non-volatile photonic
applications. Nature Photonics 2017, 11, 465-476.

(10) Zhang, Y., Ríos, C., Shalaginov, M. Y., Li, M., Majumdar, A., Gu, T. & Hu, J. Myths and truths
about optical phase change materials: A perspective. Appl Phys Lett 2021, 118.

(11) Geler-Kremer, J., Eltes, F., Stark, P., Stark, D., Caimi, D., Siegwart, H., Jan Offrein, B.,
Fompeyrine, J. & Abel, S. A ferroelectric multilevel non-volatile photonic phase shifter. Nature
Photonics 2022, 16, 491-497.

(12) Wu, M. & Li, J. Sliding ferroelectricity in 2D van der Waals materials: Related physics and
future opportunities. Proceedings of the National Academy of Sciences 2021, 118,
e2115703118.

(13) Li, L. & Wu, M. Binary Compound Bilayer and Multilayer with Vertical Polarizations: Two-
Dimensional Ferroelectrics, Multiferroics, and Nanogenerators. ACS Nano 2017, 11, 6382-
6388.

(14) Ji, J., Yu, G., Xu, C. & Xiang, H. J. General Theory for Bilayer Stacking Ferroelectricity.
Physical Review Letters 2023, 130, 146801.

(15) Zhou, B. T., Pathak, V. & Franz, M. Quantum-Geometric Origin of Out-of-Plane Stacking
Ferroelectricity. Physical Review Letters 2024, 132, 196801.

(16) Yasuda, K., Wang, X., Watanabe, K., Taniguchi, T. & Jarillo-Herrero, P. Stacking-engineered
ferroelectricity in bilayer boron nitride. Science 2021, 372, 1458-1462.

(17) Vizner Stern, M., Waschitz, Y., Cao, W., Nevo, I., Watanabe, K., Taniguchi, T., Sela, E.,
Urbakh, M., Hod, O. & Ben Shalom, M. Interfacial ferroelectricity by van der Waals sliding.
Science 2021, 372, 1462-1466.

(18) Wang, X., Yasuda, K., Zhang, Y., Liu, S., Watanabe, K., Taniguchi, T., Hone, J., Fu, L. &
Jarillo-Herrero, P. Interfacial ferroelectricity in rhombohedral-stacked bilayer transition metal
dichalcogenides. Nature Nanotechnology 2022, 17, 367-371.

(19) Deb, S., Cao, W., Raab, N., Watanabe, K., Taniguchi, T., Goldstein, M., Kronik, L., Urbakh,
M., Hod, O. & Ben Shalom, M. Cumulative polarization in conductive interfacial ferroelectrics.
Nature 2022, 612, 465-469.

(20) Yasuda, K., Zalys-Geller, E., Wang, X., Bennett, D., Cheema, S. S., Watanabe, K., Taniguchi,
T., Kaxiras, E., Jarillo-Herrero, P. & Ashoori, R. Ultrafast high-endurance memory based on
sliding ferroelectrics. Science 2024, 385, 53-56.

(21) Bian, R., He, R., Pan, E., Li, Z., Cao, G., Meng, P., Chen, J., Liu, Q., Zhong, Z., Li, W. et al.

18 / 21
 Developing fatigue-resistant ferroelectrics using interlayer sliding switching. Science 2024,
385, 57-62.

(22) Molino, L., Aggarwal, L., Enaldiev, V., Plumadore, R., I. Fal´ko, V. & Luican-Mayer, A.
Ferroelectric Switching at Symmetry-Broken Interfaces by Local Control of Dislocations
Networks. Advanced Materials 2023, 35, 2207816.

(23) Weston, A., Castanon, E. G., Enaldiev, V., Ferreira, F., Bhattacharjee, S., Xu, S., Corte-León,
H., Wu, Z., Clark, N., Summerfield, A. et al. Interfacial ferroelectricity in marginally twisted
2D semiconductors. Nature Nanotechnology 2022, 17, 390-395.

(24) Ko, K., Yuk, A., Engelke, R., Carr, S., Kim, J., Park, D., Heo, H., Kim, H.-M., Kim, S.-G.,
Kim, H. et al. Operando electron microscopy investigation of polar domain dynamics in twisted
van der Waals homobilayers. Nature Materials 2023, 22, 992-998.

(25) Meng, P., Wu, Y., Bian, R., Pan, E., Dong, B., Zhao, X., Chen, J., Wu, L., Sun, Y., Fu, Q. et al.
Sliding induced multiple polarization states in two-dimensional ferroelectrics. Nature
Communications 2022, 13, 7696.

(26) Yang, D., Liang, J., Wu, J., Xiao, Y., Dadap, J. I., Watanabe, K., Taniguchi, T. & Ye, Z. Non-
volatile electrical polarization switching via domain wall release in 3R-MoS2 bilayer. Nature
Communications 2024, 15, 1389.

(27) Yang, D., Wu, J., Zhou, B. T., Liang, J., Ideue, T., Siu, T., Awan, K. M., Watanabe, K.,
Taniguchi, T., Iwasa, Y. et al. Spontaneous-polarization-induced photovoltaic effect in
rhombohedrally stacked MoS2. Nature Photonics 2022, 16, 469-474.

(28) Wu, J., Yang, D., Liang, J., Werner, M., Ostroumov, E., Xiao, Y., Watanabe, K., Taniguchi, T.,
Dadap, J. I., Jones, D. et al. Ultrafast response of spontaneous photovoltaic effect in 3R-MoS2-
based heterostructures. Science Advances 2022, 8, eade3759.

(29) Yang, T. H., Liang, B.-W., Hu, H.-C., Chen, F.-X., Ho, S.-Z., Chang, W.-H., Yang, L., Lo, H.-
C., Kuo, T.-H., Chen, J.-H. et al. Ferroelectric transistors based on shear-transformation-
mediated rhombohedral-stacked molybdenum disulfide. Nature Electronics 2024, 7, 29-38.

(30) Roux, S., Fraunié, J., Watanabe, K., Taniguchi, T., Lassagne, B. & Robert, C. Optical
Detection of Sliding Ferroelectric Switching in hBN with a WSe2 Monolayer. Nano Letters
2025, 25, 321-326.

(31) Ye, Z., Cao, T., O’Brien, K., Zhu, H., Yin, X., Wang, Y., Louie, S. G. & Zhang, X. Probing
excitonic dark states in single-layer tungsten disulphide. Nature 2014, 513, 214-218.

(32) Ugeda, M. M., Bradley, A. J., Shi, S.-F., da Jornada, F. H., Zhang, Y., Qiu, D. Y., Ruan, W.,
Mo, S.-K., Hussain, Z., Shen, Z.-X. et al. Giant bandgap renormalization and excitonic effects

19 / 21
 in a monolayer transition metal dichalcogenide semiconductor. Nature Materials 2014, 13,
1091-1095.

(33) Chernikov, A., Berkelbach, T. C., Hill, H. M., Rigosi, A., Li, Y., Aslan, B., Reichman, D. R.,
Hybertsen, M. S. & Heinz, T. F. Exciton Binding Energy and Nonhydrogenic Rydberg Series
in Monolayer WS2. Physical Review Letters 2014, 113, 076802.

(34) Wang, G., Chernikov, A., Glazov, M. M., Heinz, T. F., Marie, X., Amand, T. & Urbaszek, B.
Colloquium: Excitons in atomically thin transition metal dichalcogenides. Reviews of Modern
Physics 2018, 90, 021001.

(35) Liang, J., Yang, D., Wu, J., Xiao, Y., Watanabe, K., Taniguchi, T., Dadap, J. I. & Ye, Z.
Resolving Polarization Switching Pathways of Sliding Ferroelectricity in Trilayer 3R-MoS2.
Nature Nanotechnology 2025, 20, 500-506.

(36) Sung, J., Zhou, Y., Scuri, G., Zólyomi, V., Andersen, T. I., Yoo, H., Wild, D. S., Joe, A. Y.,
Gelly, R. J., Heo, H. et al. Broken mirror symmetry in excitonic response of reconstructed
domains in twisted MoSe2/MoSe2 bilayers. Nature Nanotechnology 2020, 15, 750-754.

(37) Raja, A., Chaves, A., Yu, J., Arefe, G., Hill, H. M., Rigosi, A. F., Berkelbach, T. C., Nagler, P.,
Schuller, C., Korn, T. et al. Coulomb engineering of the bandgap and excitons in two-
dimensional materials. Nature Communications 2017, 8, 15251.

(38) Raja, A., Waldecker, L., Zipfel, J., Cho, Y., Brem, S., Ziegler, J. D., Kulig, M., Taniguchi, T.,
Watanabe, K., Malic, E. et al. Dielectric disorder in two-dimensional materials. Nature
Nanotechnology 2019, 14, 832-837.

(39) He, M., Cai, J., Zheng, H., Seewald, E., Taniguchi, T., Watanabe, K., Yan, J., Yankowitz, M.,
Pasupathy, A., Yao, W. et al. Dynamically tunable moiré exciton Rydberg states in a monolayer
semiconductor on twisted bilayer graphene. Nature Materials 2024, 23, 224-229.

(40) Chaves, A., Azadani, J. G., Alsalman, H., da Costa, D. R., Frisenda, R., Chaves, A. J., Song,
S. H., Kim, Y. D., He, D., Zhou, J. et al. Bandgap engineering of two-dimensional
semiconductor materials. npj 2D Materials and Applications 2020, 4, 29.

(41) Liang, J., Yang, D., Wu, J., Dadap, J. I., Watanabe, K., Taniguchi, T. & Ye, Z. Optically Probing
the Asymmetric Interlayer Coupling in Rhombohedral-Stacked MoS2 Bilayer. Physical Review
X 2022, 12, 041005.

(42) Liang, J., Yang, D., Xiao, Y., Chen, S., Dadap, J. I., Rottler, J. & Ye, Z. Shear Strain-Induced
Two-Dimensional Slip Avalanches in Rhombohedral MoS2. Nano Letters 2023, 23, 7228-7235.

(43) Qin, B., Ma, C., Guo, Q., Li, X., Wei, W., Ma, C., Wang, Q., Liu, F., Zhao, M., Xue, G. et al.
Interfacial epitaxy of multilayer rhombohedral transition-metal dichalcogenide single crystals.

20 / 21
 Science 2024, 385, 99-104.

(44) Zhang, Z., Hocking, M., Peng, Z., Pendharkar, M., Courtney, E. D. S., Hu, J., Kastner, M. A.,
Goldhaber-Gordon, D., Heinz, T. F. & Mannix, A. J. Phase-Selective Synthesis of
Rhombohedral WS2 Multilayers by Confined-Space Hybrid Metal–Organic Chemical Vapor
Deposition. Nano Letters 2024.

(45) Tebbe, D., Schütte, M., Watanabe, K., Taniguchi, T., Stampfer, C., Beschoten, B. & Waldecker,
L. Tailoring the dielectric screening in WS2–graphene heterostructures. npj 2D Materials and
Applications 2023, 7, 29.

(46) Li, Y. & Heinz, T. F. Two-dimensional models for the optical response of thin films. 2d
Materials 2018, 5, 025021.

(47) Kravets, V. G., Wu, F., Auton, G. H., Yu, T., Imaizumi, S. & Grigorenko, A. N. Measurements
of electrically tunable refractive index of MoS2 monolayer and its usage in optical modulators.
npj 2D Materials and Applications 2019, 3, 36.

(48) Li, M., Biswas, S., Hail, C. U. & Atwater, H. A. Refractive Index Modulation in Monolayer
Molybdenum Diselenide. Nano Letters 2021, 21, 7602-7608.

21 / 21