Electrically Tunable Space-Time Metasurfaces at Optical Frequencies
Electrically Tunable Space-Time Metasurfaces at Optical Frequencies
Article https://doi.org/10.1038/s41565-024-01728-9
Metasurfaces are two-dimensional arrays of subwavelength scatterers gradient configurations but scatter light at the incident wavelength. In
that offer compact wavefront control by introducing abrupt phase this mode, momentum can be imparted to light, enabling shaping of
changes at an interface. By tuning the geometry of the constituent the spatial wavefront, but the optical wavelength cannot be changed.
subwavelength elements, metasurfaces can impart spatial momentum Time-varying media offer an opportunity to overcome this
to incoming light. This phenomenon, described by the generalized limitation. Rapid modulation of the refractive index of substituent
form of Snell’s law1, has enabled the development of diverse optical metasurface scatterers creates a temporal interface: an analogue to
components, including those for anomalous reflection2–4, flat lenses5–7, the previously discussed spatial interface. At a temporal interface,
polarization plates8–10 and more. While these works illustrate a vast the spatial momentum of light is conserved, but the frequency is
range of applications, they are all limited to passive operation where altered through energy exchange between the incident wave and the
their functionality is fixed after fabrication. time-varying medium22. Temporal control of electromagnetic waves
To address this constraint, there has been substantial research has led to the demonstration of exotic optical phenomena such as
effort to achieve active, reconfigurable metasurfaces11–13. Active meta- frequency mixing23, harmonic beam shaping24 and breaking of Lorentz
surfaces consist of periodic arrays of scatterers that exhibit an optical reciprocity25–27.
resonance that can be tuned by an external stimulus (for example, Recent experiments demonstrated that combining metasurface
electrical biasing14,15, optical pumping16, thermal control17, mechanical spatial and temporal control can offer many unique optical proper-
motion18,19 or electrochemical reaction20). Through independent bias- ties27–32. Spatiotemporal modulation of dielectric media interacting
ing of active metasurface arrays, researchers have demonstrated an with an electromagnetic wave has been studied for decades and has
electrically reprogrammable optical response, achieving multiple func- produced non-reciprocal radio frequency (RF) devices capable of func-
tions in a single device21. While all active metasurfaces can change their tions such as optical isolation33, pure frequency mixing34 and amplifica-
properties in time, their response is often slow, limiting operation to the tion of travelling waves35. More recently, digital coding architectures,
‘quasi-static’ regime where they switch between different spatial phase in which biasing of positive-intrinsic-negative (PIN) diodes alters the
1
Thomas J. Watson Laboratories of Applied Physics, California Institute of Technology, Pasadena, CA, USA. 2These authors contributed equally:
Jared Sisler, Prachi Thureja. e-mail: [email protected]
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
a
b
α=0
4
Spatial phase profile Far-field intensity α = π/4
Amplitude (offset)
α=π
3
φ | {space}|2 I
2
x 1
θ
0
0 1 2 3
Temporal phase profile Frequency spectrum
Time (µs)
c
| {Time}|2 60
φ I α=0
50 α = π/4
α=π
Intensity (dB)
t f 40
30
20
ωinc
10
Phase (π)
1.0
0.5
Fig. 1 | Electrically controlled space–time metasurfaces. a, Top: the magnitude frequency of 1 MHz. Each waveform is identical, but has a different initial phase,
squared of the Fourier transform of a spatial phase gradient (left) provides the α. c, Magnitude squared of the Fourier transform of each waveform in b. The
far-field intensity profile (right). Middle: the magnitude squared of the Fourier full spectrum is plotted for α = 0, while the intensity at each integer frequency is
transform of a temporal phase gradient (left) provides the frequency spectrum plotted for α = π⁄4 and π for ease of comparison. d, Phase of the Fourier transform
of reflected light (right). Bottom: 3D visualization of the two-electrode space– of each waveform in b plotted at each harmonic of 1 MHz.
time diffracting metasurface. b, Three example sawtooth waveforms with a
phase response of scattered light, have realized a more general spa- Design principle
tiotemporal platform for frequency-multiplexed beam steering and A conceptual illustration of space–time modulation is shown in Fig. 1a.
shaping at microwave frequencies24,36,37. To achieve space–time meta- The top two plots illustrate how a binary spatial phase profile can cre-
surfaces at optical frequencies, optical pumping of passive phase gra- ate two intensity peaks in the far field corresponding to the ±1st dif-
dient metasurfaces consisting of nonlinear optical materials has been fraction orders (higher orders are cut off in the plot). Analogously, the
demonstrated22,38,39. These systems, however, require the integration middle plots of Fig. 1a show how square wave time modulation will
of high-power pump pulses, posing challenges for many applications. generate the ±1st sideband harmonics. By combining these two mod-
Supplementary Section 1 discusses the challenges associated with real- ulation schemes, a space–time metasurface is realized. The bottom
izing space–time metasurfaces at optical frequencies. image of Fig. 1a shows a schematic of the gate-tunable reflective space–
In this work, we use electrically tunable metasurfaces operating time metasurface used in this demonstration consisting of a
at 1,530 nm and modulated at 1 MHz to demonstrate spatiotemporal one-dimensional array of nanoantennas, electrically connected with
control over the scattered light wavefront and spectrum. We start by two independent electrodes contacting alternating groups of 12 nano-
collectively driving the metasurface with a time-varying voltage signal antennas. The metasurface accepts incoming light at a frequency, ωinc,
to generate frequency sidebands shifted from the incident laser light. and all antennas are modulated with an identical waveform at a fre-
Next, we time-delay the applied electrical signal in space to create quency, ωmod, time-delayed by half a period between the two antenna
a frequency-dependent spatial phase profile, enabling diffraction groups. As a result, reflected light is diffracted and consists of a span
of selected frequencies. Experimentally measured reflectance as a of harmonic frequencies given by ωinc ± nωmod, where n is an integer.
function of frequency and angle confirms that the frequency-shifted To design a spatiotemporal diffraction response, we start by
light is diffracted while the central frequency signal reflects normally, obtaining a relationship between the reflected amplitude and phase
resulting in excellent beam directivity for frequency-shifted light. of light at a single wavelength as a function of applied voltage. For
We further demonstrate, through the implementation of a real-time quasi-static operation, this relationship would be directly used as a
experimental optimization, that the frequency spectrum of reflected lookup table to determine the required voltage that must be applied
light can be arbitrarily constructed, and each generated frequency of across the metasurface. During spatiotemporal modulation, however,
interest can be independently manipulated in space. multiple frequencies are generated, and the amplitude and phase at
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
a b c
d e f
z 50 24.5
λ = 1,530 nm
y
x Ey 40
kz 24.0
Reflectance (%)
Reflectance (%) 30
Increasing 23.5
tHAOL wAnt bias
20
tITO tAnt
23.0
tAu 10
0 22.5
p 1,400 1,450 1,500 1,550 1,600 –4 –2 0 2 4
Wavelength (nm) Voltage (V)
Fig. 2 | ITO-based plasmonic metasurface. a, Camera image of chip p, and extends 120 μm in the x axis. The metasurface is illuminated from the
mounted and wire-bonded to a printed circuit board consisting of ten top with a normally incident plane wave polarized in the y axis. The dimensions
individual metasurfaces. Scale bar, 5 mm. b, Marked region from a showing labelled are wAnt = 220 nm, tAnt = 40 nm, tHAOL = 16 nm, tITO = 17 nm, tAu = 80 nm,
optical microscope image of a single metasurface. The centre region creates p = 400 nm. e, Measured reflectance spectra of metasurface under varying
a 120 × 120 μm2 active area of nanoantennas which are contacted by the applied bias from −4 to 4 V. The dotted line corresponds to the operation
interconnects on the side of the image. Scale bar, 100 μm. c, Marked region from wavelength of 1,530 nm used throughout this work. f, Measured reflectance
b showing SEM image of individual nanoantennas. Scale bar, 5 μm. d, Schematic modulation of the metasurface as a function of applied bias at an operation
of metasurface unit cell. The structure is periodic in the y axis with a period, wavelength of 1,530 nm.
each frequency must be considered. This information is obtained by Active metasurface characterization
taking the Fourier transform of the time-varying scattered field. Sup- In this work, we use an indium tin oxide (ITO)-based plasmonic meta-
plementary Section 2 describes this process in more detail. surface operating in reflection14,21,41,42. A camera image of ten devices
To illustrate the design principle, Fig. 1b shows three time-varying wire-bonded to a printed circuit board (PCB) is shown in Fig. 2a.
responses of a metasurface exhibiting pure amplitude modulation. Figure 2b,c shows a single metasurface under optical microscope and
Each waveform shape is identical, while the initial phase, α, is varied. scanning electron microscope (SEM), respectively, and the metasurface
Figure 1c,d shows the magnitude squared and phase, respectively, of cross-section is provided in Fig. 2d. ITO, the active material, is modu-
the temporal Fourier transform of each waveform in Fig. 1b. As all three lated in a metal-oxide-semiconductor capacitor configuration. Here,
signals in Fig. 1b have the same shape, the intensity (amplitude-squared) ITO acts as a highly doped semiconductor, HAOL—a nanolaminate of
spectra are identical. However, the phase spectrum for each waveform alumina (Al2O3) and hafnia (HfO2) (ref. 43)—is the gate dielectric, and
is frequency dependent, given by φ(n) = nα, where n is the harmonic the gold antennas act as the top metal contact. The bottom gold acts
number. If the α = 0 and α = π waveforms were applied to the two inde- as a backreflecting layer as well as the ITO electrical contact. Under
pendent electrodes shown in Fig. 1a (α1 = 0 and α2 = π), there would be applied voltage, carriers are accumulated in ITO at the ITO–HAOL
a binary 0–π phase grating with constant amplitude for odd harmon- interface, resulting in modulation of the ITO refractive index. When
ics, resulting in diffraction of those frequencies. This highlights the illuminated with transverse electric polarized light, a gap plasmon
unique capability of space–time metasurfaces to decouple phase and resonance confines light in the ITO and HAOL layers, increasing the
amplitude. The shape of the time-varying signal dictates the frequency optical modulation. Due to the highly confined optical resonance and
spectra (that is, intensity at each frequency) and the time delay of accumulation of charges, these devices have been shown to demon-
this signal provides a linear frequency-dependent phase shift. In this strate local behaviour such that their optical response does not depend
example, a metasurface with varying amplitude and constant phase on the number of unit cells that are electrically addressed21.
was transformed into a metasurface with constant amplitude and 2π Figure 2e shows the measured metasurface reflectance under
phase coverage at each generated harmonic. It is also important to applied bias. The 1,475 nm resonance at −4 V shifts to 1,460 nm at 4 V.
note that the phase at 0 MHz is identical for all values of α, meaning any The operation wavelength throughout this work is 1,530 nm, and Fig. 2f
unmodulated light will not receive a shift in spatial momentum. The shows the voltage-dependent reflectance at this wavelength. The phase
ratio of unmodulated to modulated light depends on the accessible shift at 1,530 nm is negligible and is shown in Supplementary Section 4.
phase and amplitude of the metasurface and the selected time-varying While the relative modulation is larger on resonance, 1,530 nm is the
signal31,32,40. Additionally, given the experimental methods used in shortest wavelength accessible in our experimental setup. It is note-
this work, the unmodulated light is not measured and, thus, is absent worthy, however, that a device with limited relative reflectance modu-
from all future plots in this manuscript. Supplementary Section 3 lation (~8%) and near-zero phase shift, can be used to demonstrate
provides detailed calculations of the fraction of unmodulated light in space–time diffraction with excellent directivity, as will be presented
our measurements. in the following sections.
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
a
AWG
Beam
Laser splitter
ωinc ~
Erbium-doped ωinc, ωinc ± ωmod Metasurface
fibre amplifier
Photodiode Optimization
algorithm
Spectrum analyser
b 40
c d
Sine wave 10 Sine wave 1.02 Sine wave
Optimized wave Optimized wave Optimized wave
5
Voltage (V)
0 0.98
–20 0 0.96
0.94
–40
–5
0.92
–60
–10 0.9
2 4 6 8 10 –2 –1 0 1 2 –2 –1 0 1 2
Frequency (MHz) Time (µs) Time (µs)
Fig. 3 | Time modulation and waveform optimization. a, Schematic of metasurface to produce the frequency spectra in b. d, Measured time response
experimental setup. b, Reflected frequency spectra when metasurface is driven of the reflectance from the metasurface when driving with the sine (blue) and
with a 5 Vp-p sine wave (blue) and the final optimized waveform (red). c, Sine optimized (red) waveforms from c.
(blue) and optimized (red) driving voltage waveforms that were applied to the
Temporal waveform optimization signal. To address this, we incorporate an experimental feedback loop
To study the temporal response of the metasurface, we first modulate to continuously update the driving waveform until a desired frequency
all antennas in-phase and measure the reflected frequency spectrum spectrum is measured (Fig. 3a). This technique is particularly useful
using the experimental setup in Fig. 3a. The metasurface is illuminated as it compensates for the nonlinearities of the metasurface as well as
with 1,530 nm light and modulated at megahertz frequencies with an other elements such as the metasurface interconnects, PCB, waveform
arbitrary waveform generator (AWG). A fast photodiode connected to a generator and photodiode. A genetic algorithm optimization was used
spectrum analyser captures reflected light. Because the time responses to fit the amplitude and phase of the first 20 terms in a Fourier series
of the photodiode (~10 MHz) and spectrum analyser (~1 GHz) are many to generate the final driving voltage waveform, using 1 MHz as the
orders of magnitude slower than the oscillation of the incident laser fundamental harmonic.
light (~200 THz), only the megahertz frequency sidebands imparted As a first demonstration, the applied waveform is optimized to
on the 1,530 nm light are measured. Therefore, the 1,530 nm laser light generate only the 1 MHz signal. The top blue curve in Fig. 3b shows the
acts as a carrier wave to study the modulation dynamics of the metas- measured frequency spectrum when the metasurface is driven with a
urface, and the intensity of any unmodulated reflected 1,530 nm light 1 MHz sine wave. As desired, there is a strong signal at 1 MHz but there is
is not detected. Because the device used in this work operates using unwanted power at 2 MHz. To suppress this harmonic, the optimization
amplitude modulation and modest phase modulation, there is substan- is run with a figure of merit defined to maximize the power at 1 MHz
tially more unmodulated light than frequency converted light, which while suppressing all other frequencies. Supplementary Section 6
limits our total diffraction efficiency to ~0.0036%. This low efficiency discusses details of the optimization algorithm and figure of merit
is a result of the device used and is not a fundamental limitation of the selection. The optimization produces a new spectrum (Fig. 3b, red
method presented in this work. Supplementary Section 3 presents an curve) that maintains power in the first harmonic while greatly reduc-
analysis of the total unmodulated light from our metasurface and pro- ing the second harmonic. Figure 3c shows the corresponding voltage
vides insights into how the total efficiency can be increased. It should waveforms sent to the AWG for the frequency spectra in Fig. 3b. The
be noted that ITO-based modulators have demonstrated modulation optimized waveform consists of a maximum frequency component
speeds up to the gigahertz range44,45 and, thus, it is the photodiode, not of 20 MHz, which is greater than the 3 dB cut-off of the photodiode
our metasurface, that limits this work to operation at megahertz fre- (10 MHz) and, thus, is partially filtered out in experiment. However, it
quencies. Further discussion about frequency limitations of ITO-based was found that including these higher-frequency components resulted
metasurfaces is presented in Supplementary Section 5. in improved higher-order harmonic suppression.
The reflected frequency spectrum can be manipulated by applying Figure 3d plots the measured time-domain reflectance while mod-
standard time-varying waveforms to the metasurface (for example, ulating with the single-frequency sine wave and then optimized wave
sine, square and sawtooth). However, our metasurface exhibits a non- from Fig. 3c. As expected, the reflectance when driving with a sine wave
linear reflectance–voltage relationship (Fig. 2f), so the reflected fre- is considerably distorted. The optimized wave is much closer to a sine
quency spectrum will not match the spectrum of the driving electrical wave but is still not perfect because the optimization algorithm can only
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
Intensity (dB)
1st order
2.0
0
0
–20
Frequency (MHz)
–40
1.5
–1
2 4 6 8 10
Frequency (MHz)
b Out-of-phase
V(t) 1.0
Diffraction order
+1 40
0th order
Intensity (dB)
20 1st order
0 0.5
0
–20
–40
–1
V(t – T/2) 2 4 6 8 10 –10 0 10 –10 0 10 –10 0 10
Frequency (MHz) Polar angle (°)
Fig. 4 | Diffraction of a single harmonic. Throughout this figure, the Right: measured frequency spectra when photodiode is placed at the 0th (blue)
metasurface is modulated with the red optimized voltage waveform from Fig. 3c. and +1st (red) spatial orders. c, Metasurface is modulated in the ‘out-of-phase’
a, All metasurface antennas are modulated in-phase. Left: schematic of condition from b. Calculated (left) and measured (centre) intensity of reflected
metasurface response. Right: measured frequency spectra when photodiode light as a function of frequency and angle, plotted in log scale. Right: spatial
is placed at the 0th (blue) and +1st (red) spatial orders. b, Driving waveform is intensity cross-sections from centre figure, plotted in linear scale at 1 MHz (blue)
applied to all metasurface nanoantennas, time-delayed by half a period between and 2 MHz (red).
adjacent nanoantenna groups. Left: schematic of metasurface response.
suppress higher-order harmonics into the noise floor of the frequency power is in the ±1st diffractions orders at the frequency of interest
measurements. To approach a perfect sinusoidal reflectance response, (1 MHz). This is a notable increase in performance compared with the
one could integrate the frequency measurements for a longer time, achievable quasi-static diffraction for this metasurface. Further discus-
providing better signal-to-noise ratio of each higher-order harmonic. sion comparing quasi-static and space–time diffraction is presented
in Supplementary Section 9.
Diffraction of a single harmonic
Space–time metasurfaces can spatially manipulate the wavefront for Arbitrary frequency generation and diffractive
the generated frequency spectrum. Figure 4a,b illustrate two driving switching
schemes under study. In both cases, the metasurface is driven with We also demonstrate waveform design to generate arbitrary frequency
the optimized waveform from Fig. 3 with each antenna group in-phase spectra with individual spatial control over each frequency of interest.
(Fig. 4a) and half a waveform period (π) out-of-phase (Fig. 4b). The First, we target the simultaneous generation of the third and fifth har-
resultant measured frequency spectra at the 0th and +1st diffraction monics at equal intensity while suppressing all others. Figure 5a plots
orders are shown for each situation. For the in-phase case, there is no the measured frequency spectrum for this case, generated using the
spatial phase profile, and nearly all the reflected 1 MHz signal is in the top blue driving voltage signal in Fig. 5b. In the waveform optimization,
0th order. There is a small 1 MHz signal in the +1st diffraction order that the fundamental frequency remains at 1 MHz (that is, the time-domain
is likely caused by the two independent antenna groups having a slightly signal has a period of 1 μs) and the algorithm is left to find a suitable
different amplitude/voltage response. See Supplementary Section 7 solution (see Supplementary Section 6 for details). Next, the waveform
for an in-depth discussion on this point. In the out-of-phase case, the applied to one antenna group is offset by half a period (Fig. 5b, bottom
1 MHz intensity at the +1st diffraction order is ~12 times greater than the red line). By doing this, the phase accumulated at the nth harmonic is
0th order. The small unwanted signal at the 0th order can be attributed equal to φ(n) = nπ (Fig. 1d). Since both excited harmonics are odd, the
to a change in the average reflectance of the entire metasurface in time. phase profiles at 3 and 5 MHz obtain the desired binary phase grating
Supplementary Section 7 also addresses this effect. for symmetric diffraction. Figure 5c shows the calculated (left) and
To further investigate the space–time performance of our metas- measured (centre) response of the metasurface using the two driving
urface, we employ the driving scheme shown in Fig. 4b while scanning waveforms in Fig. 5b. The right plot in Fig. 5c shows the measured spatial
the photodiode through space, measuring the frequency spectrum cross-sections at each harmonic.
at each point, to generate an intensity map of reflected light as a func- Next, we generate the first five harmonics with equal power while
tion frequency and angle (Fig. 4c). Supplementary Section 8 shows a suppressing all others. Figure 5d shows the measured frequency spec-
schematic of the optical setup used to obtain these measurements. trum when the metasurface is driven with the top blue waveform in
Figure 4c (left) shows the predicted response obtained by taking the Fig. 5e. In this case, to simultaneously diffract all generated frequency
two-dimensional Fourier transform of the electric field across the meta- harmonics, one cannot time-delay the driving waveform of one antenna
surface in space and time (see Supplementary Section 2 for details of group by half a period because the even harmonics will accumulate
calculation). Figure 4c (middle) shows the measured data, and Fig. 4c a phase shift equal to a multiple of 2π, resulting in a constant spatial
(right) shows a spatial cross-section of the measured intensity at each phase profile. However, since the optimized driving waveform is con-
integer harmonic of the modulation frequency. While the total amount structed from a Fourier series, the phase of each harmonic term can
of diffracted light is limited (Supplementary Section 3), 94% of the be easily controlled. A new waveform is created for the second group
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
a d g
Intensity (dB)
Intensity (dB)
Intensity (dB)
40 40 40
20 20 20
0 0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Frequency (MHz) Frequency (MHz) Frequency (MHz)
b e h
Voltage (V)
Voltage (V)
10
Voltage (V)
0 1 2 3 0 1 2 3 0 1 2 3
Time (µs) Time (µs) Time (µs)
5 5 5
4 4 4
Frequency (MHz)
Frequency (MHz)
Frequency (MHz)
3 3 3
2 2 2
1 1 1
Measurement Measurement Measurement
Calculation Measurement Calculation Measurement Calculation Measurement
cross-sections cross-sections cross-sections
Fig. 5 | Space–time modulation for arbitrary control over the spectral and frequencies of interest. The blue curve is applied to the entire metasurface in
spatial properties of light. a–c, Target functionality: generate third and fifth phase to generate frequency spectra in a, d and g. c, f and i show calculated (left)
harmonics with equal amplitude, diffract third and fifth harmonics. d–f, Target and measured (centre) intensity in frequency and space when the metasurface
functionality: generate first through fifth harmonics with equal amplitude, is driven with the two waveforms in b, e and h, plotted in log scale. Right plots:
diffract first through fifth harmonics. g–i, Target functionality: generate first linear scale spatial cross-sections of measured intensity (centre) at each
through fifth harmonics, diffract second and third harmonics. a, d and g show harmonic. Note: for c and f, cross-sections are normalized to the maximum
measured frequency spectra after optimizing driving waveform for each target intensity across all frequencies. For i, cross-sections are normalized to the
functionality of interest. b, e and h show driving voltage waveforms applied maximum intensity at each frequency.
to each metasurface electrode used to generate space–time diffraction of
of antennas by offsetting the phase of the first five terms in the driv- light. This function can expand the capabilities of optical communi-
ing waveform Fourier series by π. This new waveform is plotted in the cation and sensing systems by increasing the information/channel
lower red curve in Fig. 5e. Using the two waveforms in Fig. 5e, Fig. 5f capacity of a single active device. While this work serves as a proof
compares the calculated and measured data. As desired, all five gener- of concept for space–time control at optical wavelengths using a
ated harmonics are symmetrically diffracted. two-electrode device, the space–time active metasurface concept can
Finally, we reuse the optimized waveform for the first five harmon- be extended to a metasurface with multiple individually addressable
ics and arbitrarily decide to diffract only the second and third har- nanoantennas allowing for steering of frequency harmonics to arbitrary
monics. The frequency spectrum and driving waveform are repeated angles21,46–48. Furthermore, by integrating multiple resonant elements
in Fig. 5g and Fig. 5h, respectively. To generate the waveform for the in a single period to form a supercell, this platform could be extended
second electrode, we offset the phase of the second and third harmon- to accept multiple input wavelengths of light that can be modulated
ics by π, providing the bottom red waveform in Fig. 5h. The results of independently in space and time, allowing for a multi-input, multi-
space–time diffraction using these two waveforms are shown in Fig. 5i. output device49.
As desired, the second and third harmonics are diffracted while the The ability to independently control the spatial and spectral prop-
first, fourth and fifth are reflected normally. It should be noted that erties of light additionally provides a pathway towards realizing the
the number of frequencies generated and the spatial control over violation of Lorentz reciprocity. Non-reciprocity is typically achieved
these harmonics will ultimately be limited by the time resolution or by applying a large magnetic field that requires a bulky ferromagnet
bandwidth of the waveform generator. and limits possible applications. Spatiotemporal modulation, however,
can achieve non-reciprocal functionality in a more compact system
Outlook and conclusions due to the breaking of time reversal symmetry. To observe large spatial
This work illustrates the unique capability of space–time metasur- separation of light (which is required for optical isolation), the modu-
faces to independently control the spatial and spectral properties of lation frequency should be close to the frequency of incident light.
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
Because of this, non-reciprocal space–time metasurfaces have mostly 12. Thureja, P. et al. Toward a universal metasurface for optical
been demonstrated in the RF domain33,50–52. Recent optical metas- imaging, communication, and computation. Nanophotonics 11,
urfaces have demonstrated electrical modulation in the gigahertz 3745–3768 (2022).
regime53 and could potentially be extended using the modulation 13. Yang, J., Gurung, S., Bej, S., Ni, P. & Howard Lee, H. W. Active
schemes outlined in this work to yield small spatial offsets. However, optical metasurfaces: comprehensive review on physics,
novel optical designs would be required to realize larger spatial and mechanisms, and prospective applications. Rep. Prog. Phys. 85,
spectral separation for functionalities such as optical isolation. In this 036101 (2022).
context, optical pumping of space–time metasurfaces has been shown 14. Huang, Y. W. et al. Gate-tunable conducting oxide metasurfaces.
to be an effective technique to achieve large spatial separation26,38. Nano Lett. 16, 5319–5325 (2016).
In conclusion, we have demonstrated experimental operation of 15. Wu, P. C. et al. Dynamic beam steering with all-dielectric
near-IR metasurfaces electrically modulated at megahertz frequencies electro-optic III–V multiple-quantum-well metasurfaces.
to generate arbitrary harmonic spectra and individually diffract these Nat. Commun. 10, 3654 (2019).
frequencies in space. This result is enabled by the recent improve- 16. Shcherbakov, M. R. et al. Ultrafast all-optical tuning of direct-gap
ments in the modulation speed and degree of spatial phase gradient semiconductor metasurfaces. Nat. Commun. 8, 17 (2017).
control that is possible in active metasurfaces at optical frequencies. 17. Klopfer, E., Delgado, H. C., Dagli, S., Lawrence, M. & Dionne, J. A.
We employed a two-electrode ITO-based plasmonic metasurface to A thermally controlled high-Q metasurface lens. Appl. Phys. Lett.
first design a time-varying driving voltage signal to excite frequencies 122, 221701 (2023).
of interest and suppress unwanted frequencies. Next, each generated 18. Ou, J. Y., Plum, E., Jiang, L. & Zheludev, N. I. Reconfigurable
frequency was manipulated in space by introducing phase delays to photonic metamaterials. Nano Lett. 11, 2142–2144 (2011).
specific frequency components of the driving waveform being applied 19. Arbabi, E. et al. MEMS-tunable dielectric metasurface lens.
to each electrode. The ability to generate and steer frequencies in a Nat. Commun. 9, 812 (2018).
single chip-scale device has potential to find many applications in 20. Eaves-Rathert, J. et al. Dynamic color tuning with
optical communication and sensing. electrochemically actuated TiO2 metasurfaces. Nano Lett. 22,
1626–1632 (2022).
Online content 21. Shirmanesh, G. K., Sokhoyan, R., Wu, P. C. & Atwater, H. A.
Any methods, additional references, Nature Portfolio reporting sum- Electro-optically tunable multifunctional metasurfaces.
maries, source data, extended data, supplementary information, ACS Nano 14, 6912–6920 (2020).
acknowledgements, peer review information; details of author contri- 22. Tirole, R. et al. Double-slit time diffraction at optical frequencies.
butions and competing interests; and statements of data and code avail- Nat. Phys. 19, 999–1002 (2023).
ability are available at https://doi.org/10.1038/s41565-024-01728-9. 23. Wu, Z. & Grbic, A. Serrodyne frequency translation using
time-modulated metasurfaces. IEEE Trans. Antennas Propag. 68,
References 1599–1606 (2020).
1. Yu, N. & Capasso, F. Flat optics with designer metasurfaces. 24. Zhang, L. et al. Space–time-coding digital metasurfaces.
Nat. Mater. 13, 139–150 (2014). Nat. Commun. 9, 4334 (2018).
2. Sun, S. et al. High-efficiency broadband anomalous reflection by 25. Cardin, A. E. et al. Surface-wave-assisted nonreciprocity in
gradient meta-surfaces. Nano Lett. 12, 6223–6229 (2012). spatio-temporally modulated metasurfaces. Nat. Commun. 11,
3. Wong, A. M. H. & Eleftheriades, G. V. Perfect anomalous 1469 (2020).
reflection with a bipartite Huygens’ metasurface. Phys. Rev. X 8, 26. Guo, X., Ding, Y., Duan, Y. & Ni, X. Nonreciprocal metasurface with
011036 (2018). space–time phase modulation. Light. Sci. Appl. 8, 123 (2019).
4. Wang, X. et al. Simultaneous realization of anomalous reflection 27. Shaltout, A., Kildishev, A. & Shalaev, V. Time-varying metasurfaces
and transmission at two frequencies using bi-functional and Lorentz non-reciprocity. Opt. Mater. Express 5, 2459 (2015).
metasurfaces. Sci. Rep. 8, 1876 (2018). 28. Hadad, Y., Sounas, D. L. & Alu, A. Space–time gradient
5. Khorasaninejad, M. et al. Metalenses at visible wavelengths: metasurfaces. Phys. Rev. B 92, 100304(R) (2015).
diffraction-limited focusing and subwavelength resolution 29. Taravati, S. & Eleftheriades, G. V. Generalized space–time-periodic
imaging. Science 352, 1190–1194 (2016). diffraction gratings: theory and applications. Phys. Rev. Appl. 12,
6. Chen, W. T. et al. A broadband achromatic metalens for focusing 024026 (2019).
and imaging in the visible. Nat. Nanotechnol. 13, 220–226 (2018). 30. Wang, X., Asadchy, V. S., Fan, S. & Tretyakov, S. A. Space–time
7. Arbabi, A., Horie, Y., Ball, A. J., Bagheri, M. & Faraon, A. metasurfaces for power combining of waves. ACS Photonics 8,
Subwavelength-thick lenses with high numerical apertures 3034–3041 (2021).
and large efficiency based on high-contrast transmitarrays. 31. Salary, M. M. & Mosallaei, H. Time-modulated conducting oxide
Nat. Commun. 6, 7069 (2015). metasurfaces for adaptive multiple access optical communication.
8. Deng, Y., Wu, C., Meng, C., Bozhevolnyi, S. I. & Ding, F. IEEE Trans. Antennas Propag. 68, 1628–1642 (2020).
Functional metasurface quarter-wave plates for simultaneous 32. Sabri, R., Salary, M. M. & Mosallaei, H. Quasi‐static and time‐
polarization conversion and beam steering. ACS Nano 15, modulated optical phased arrays: beamforming analysis and
18532–18540 (2021). comparative study. Adv. Photonics Res. 2, 2100034 (2021).
9. Devlin, R. C., Ambrosio, A., Rubin, N. A., Mueller, J. P. B. & Capasso, 33. Taravati, S., Chamanara, N. & Caloz, C. Nonreciprocal
F. Arbitrary spin-to–orbital angular momentum conversion of electromagnetic scattering from a periodically space–time
light. Science 358, 896–901 (2017). modulated slab and application to a quasisonic isolator. Phys.
10. Balthasar Mueller, J. P., Rubin, N. A., Devlin, R. C., Groever, B. & Rev. B 96, 165144 (2017).
Capasso, F. Metasurface polarization optics: Independent phase 34. Taravati, S. Aperiodic space–time modulation for pure frequency
control of arbitrary orthogonal states of polarization. Phys. Rev. mixing. Phys. Rev. B 97, 115131 (2018).
Lett. 118, 113901 (2017). 35. Tien, P. K. Parametric amplification and frequency mixing in
11. Hail, C. U., Michel, A. U., Poulikakos, D. & Eghlidi, H. Optical propagating circuits. J. Appl. Phys. 29, 1347–1357 (1958).
metasurfaces: evolving from passive to adaptive. Adv. Opt. Mater. 36. Zhang, L. & Cui, T. J. Space-time-coding digital metasurfaces:
7, 1801786 (2019). principles and applications. Research 2021, 9802673 (2021).
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
37. Zhang, L. et al. Breaking reciprocity with space–time‐coding 48. Xu, B. H. et al. Multichannel metasurfaces with frequency‐
digital metasurfaces. Adv. Mater. 31, 1904069 (2019). direction multiplexed amplitude and phase modulations. Adv.
38. Karimi, M., Alam, M. Z., Upham, J., Reshef, O. & Boyd, R. W. Opt. Mater. 11, 2301299 (2023).
Time-varying gradient metasurface with applications in all-optical 49. Sabri, R., Forouzmand, A. & Mosallaei, H. Multi-wavelength
beam steering. Nanophotonics 12, 1733–1740 (2023). voltage-coded metasurface based on indium tin oxide:
39. Lustig, E. et al. Time-refraction optics with single cycle independently and dynamically controllable near-infrared
modulation. Nanophotonics 12, 2221–2230 (2023). multi-channels. Opt. Express 28, 3464 (2020).
40. Salary, M. M., Farazi, S. & Mosallaei, H. A dynamically modulated 50. Taravati, S. & Eleftheriades, G. V. Microwave space–time-
all‐dielectric metasurface doublet for directional harmonic modulated metasurfaces. ACS Photonics 9, 305–318 (2022).
generation and manipulation in transmission. Adv. Opt. Mater. 7, 51. Hadad, Y., Soric, J. C. & Alu, A. Breaking temporal symmetries
1900843 (2019). for emission and absorption. Proc. Natl Acad. Sci. USA 113,
41. Kafaie Shirmanesh, G., Sokhoyan, R., Pala, R. A. & Atwater, H. A. 3471–3475 (2016).
Dual-gated active metasurface at 1550 nm with wide (>300°) 52. Taravati, S. & Eleftheriades, G. V. Full-duplex nonreciprocal
phase tunability. Nano Lett. 18, 2957–2963 (2018). beam steering by time-modulated phase-gradient metasurfaces.
42. Amin, R. et al. Heterogeneously integrated ITO plasmonic Mach– Phys. Rev. Appl. 14, 014027 (2020).
Zehnder interferometric modulator on SOI. Sci. Rep. 11, 1287 (2021). 53. Benea-Chelmus, I.-C. et al. Gigahertz free-space electro-
43. Lu, B., Lv, H., Zhang, Y., Zhang, Y. & Liu, C. Comparison of HfAlO, optic modulators based on Mie resonances. Nat. Commun. 13,
HfO2/Al2O3, and HfO2 on n-type GaAs using atomic layer 3170 (2022).
deposition. Superlattices Microstruct. 99, 54–57 (2016).
44. Gui, Y. et al. 100 GHz micrometer-compact broadband monolithic Publisher’s note Springer Nature remains neutral with regard to
ITO Mach–Zehnder interferometer modulator enabling 3500 times jurisdictional claims in published maps and institutional affiliations.
higher packing density. Nanophotonics 11, 4001–4009 (2022).
45. Amin, R. et al. Sub-wavelength GHz-fast broadband ITO Mach– Springer Nature or its licensor (e.g. a society or other partner) holds
Zehnder modulator on silicon photonics. Optica 7, 333 (2020). exclusive rights to this article under a publishing agreement with
46. Sabri, R., Salary, M. M. & Mosallaei, H. Broadband continuous the author(s) or other rightsholder(s); author self-archiving of the
beam-steering with time-modulated metasurfaces in the accepted manuscript version of this article is solely governed by the
near-infrared spectral regime. APL Photonics 6, 086109 (2021). terms of such publishing agreement and applicable law.
47. Castaldi, G. et al. Joint multi‐frequency beam shaping and
steering via space–time‐coding digital metasurfaces. Adv. Funct. © The Author(s), under exclusive licence to Springer Nature Limited
Mater. 31, 2007620 (2021). 2024
Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01728-9
Methods (J.S., P.T., M.G., R.S., I.H. and H.A.A.) and DARPA EXTREME (H.A.A.).
Device fabrication J.S. acknowledges the support of the Natural Sciences and
Two-inch silicon wafers with a 1 μm thermally grown oxide were first Engineering Research Council of Canada (NSERC) through the
cleaned using a Radio Corporation of America 1 and Radio Corporation Postgraduate Scholarship – Doctoral program. P.T. acknowledges
of America 2 clean. The back gold layer was patterned using photo- support from Meta Platforms, Inc., through the PhD fellowship
lithography and liftoff in Remover PG. The deposition consisted of #C-834952. We are grateful to Caltech colleagues O. Painter and
electron beam (e-beam) evaporation of 5 nm Ti and 80 nm Au. Next, the A. Emami for support with waveform generation instrumentation.
ITO was patterned by photolithography and liftoff in Remover PG. The We further thank C. U. Hail, M. D. Kelzenberg and P. R. Jahelka for
17 nm ITO deposition used an RF sputterer using a 90/10 wt% In2O3/SnO2 technical support and discussions.
target at 3 mTorr and 100 W while flowing 20 sccm Ar and 4 sccm Ar/O2
in a 90/10 wt% mix. The HAOL was deposited via thermal atomic layer Author contributions
deposition at 150 °C using a shadow mask. The deposition consisted J.S., P.T., M.Y.G., R.S. and H.A.A. conceived the ideas for this
of four repetitions of 1 nm/3 nm Al2O3/HfO2 without breaking vacuum research project. J.S. and P.T. fabricated and characterized
to accumulate a total of 16 nm. The nanoantennas were patterned via the active metasurfaces. J.S. and M.Y.G. built the experimental
e-beam lithography using a bilayer polymethyl methacrylate (PMMA) setup for time-modulation. J.S. and P.T. performed quasi-static
resist and lifted off in Remover PG. The nanoantennas were deposited and time-modulated measurements. J.S. built the setup for and
using the e-beam evaporation of 2 nm Ge and 40 nm Au. Finally, the performed the space–time measurements. J.S. created the real-time
top contact pads were patterned using photolithography and lifted waveform optimization algorithm, with inputs from P.T. J.S. and
off in acetone. The contact pads were deposited using e-beam evapo- M.Y.G. performed the space–time calculations. I.H. and P.T.
ration of 20 nm Ti and 200 nm Au. The final devices were then diced designed the electrical PCB. All authors contributed to the writing
into 1 cm × 1 cm chips and wire-bonded to specifically designed PCBs. of the manuscript.
Nature Nanotechnology