High‐Speed Waveguide‐Integrated InSe Photodetector on SiN Photonics for Near‐Infrared Applications

On‐chip integration of two‐dimensional (2D) materials holds immense potential for novel optoelectronic devices across diverse photonic platforms. In particular, indium selenide (InSe) is a very promising 2D material due to its ultra‐high carrier mobility and outstanding photoresponsivity. Herein, a high‐speed photodetector based on a multilayer 90 nm thick InSe integrated on a silicon nitride (SiN) waveguide is reported. The device exhibits a remarkable low dark current density of ≈40 nA μm−2, a photoresponsivity of 0.38 A W−1, and an external quantum efficiency of ≈48.4%. Tested under ambient conditions at near‐infrared 976 nm wavelength, it exhibits an absorption coefficient of 0.11 dB μm−1. Additionally, the photodetector demonstrates a 3‐dB radiofrequency bandwidth of 85 MHz and an open‐eye diagram at data transmission of 1 Gbit s−1. Based on these exceptional optoelectronic advantages, integrated multilayer InSe enables the realization of active photonic devices for a range of applications, such as short‐reach optical interconnects, LiDAR imaging, and biosensing.

the group IIIA-VIA layered semiconductors (MX, M = Ga and In, X = S, Se, and Te), [12] exhibits a tunable bandgap dependent on its thickness.11b,13] The carrier mobility of InSe nanosheets surpasses that of transition metal dichalcogenides (TMDs) and is comparable to black phosphorus. [14]arge-scale growth of InSe thin films on different substrates has also been achieved using chemical vapor deposition (CVD), [15] atomic layer deposition (ALD), [16] and pulsed laser deposition (PLD) [17] methods, demonstrating the feasibility of large-scale integration of InSe devices.
Photodetectors based on few-layered InSe sheets exhibit broadband response from visible to near-infrared wavelengths with high photoresponsivity. [18]Extensive reports have demonstrated various types of InSe detectors, including photoconductive (PC), phototransistors (PT), self-powered photodetectors (SPPD), and avalanche photodetectors (APD). [18,19]However, most of these studies are limited to vertically coupled free-space devices in the visible range, where InSe exhibits the highest absorption coefficient.Recently, Xiaoqi Cui et al. demonstrated an on-chip integrated InSe photodetector operating at 520 nm wavelength with a responsivity of 0.115 A W À1 , although the study focused on the static response. [20]Despite its promising performance, the lack of InSe studies in the NIR range for integrated systems calls for an in-depth investigation of this material on the SiN platform.
In this work, we demonstrate integrated InSe photodetectors on the SiN platform operating in the NIR region.The hybrid SiN/InSe waveguide photodetectors are fabricated by mechanical exfoliation of InSe flakes on top of SiN waveguides.The device operates under bias and exhibits a low dark current density of a few nano-amps.Furthermore, a remarkable photoresponsivity of 0.38 A W À1 and low noise-equivalent power (NEP) of 4.7 nW Hz 1/2 are achieved at 976 nm wavelength.The devices exhibit a 3-dB radiofrequency (RF) bandwidth of 85 MHz and open-eye diagrams are demonstrated for on-off modulated signals from 50 to 1 Gbit s À1 .The achieved excellent merits offer a new avenue for building and designing active photonic components on the SiN platform using InSe.

InSe Integration into SiN Photonics
Figure 1a shows an optical microscopy image of the fabricated unloaded SiN waveguide chip.In this design, a 400 nm top silicon nitride layer and a 2 μm buried oxide are used.The waveguide is air-cladded and has a width of 1.1 μm.Light is coupled in and off the SiN chip using lensed fibers.The photodetector's contacts are formed using a prepatterned metal electrode of Cr/Au (10 nm/120 nm). Figure 1b shows a schematic representation of the 3D cross-section view representation of the photodetector, respectively.Note that the metal pads and the SiN waveguide are not on the same plane and there is a 270 nm height difference between the metal and waveguide top surfaces.
A deterministic dry transfer procedure is used to transfer the mechanically exfoliated multilayered InSe on top of the SiN waveguide, laid directly on top of the metal electrodes.It is worth noting that a recent study has reported an anisotropic nature of InSe where the optical absorption is highly affected by the crystal orientation.10a,21] A scanning electron microscopy (SEM) image of a fabricated device is shown in Figure 1c.The metal electrodes are symmetrically deposited on the sides of the waveguide spaced with a gap fixed at 8 μm (distance between metal edges), which minimizes optical absorption losses from metal absorption in the electrodes.In the fabricated device, the overlapping length of the detector is %30 μm along the waveguide.The thickness of the InSe is measured using an atomic force microscope which is determined to be %90 nm (see Figure 1d).As can be seen in this figure, the flake adheres conformably to the photonic waveguide structure beneath it.

Guided Light Interaction with Multilayer InSe
The material characteristics of bulk and transferred InSe are confirmed by energy dispersive spectroscopy (EDS), Raman spectroscopy, X-Ray diffraction spectroscopy (XRD), and ellipsometry.To confirm its stoichiometry, EDS of the grown single crystal InSe is performed on the selected area depicted by the highlighted pink color box, as shown in the SEM image in Figure 2a.The spectra indicate only peaks for the indium (In) and selenium (Se) elements with an atomic ratio about 1:1 (In:Se).Additionally, the SEM image and elemental composition of the waveguide integrated InSe are shown in Figure S1 (Supporting Information), confirming a controlled 1:1 (In:Se) stoichiometry after exfoliation and transfer.To confirm the crystallographic structure and the material phase, room temperature XRD on the bulk crystal is carried out.Results are shown in Figure 2b.All the observed diffraction peaks agree well with the previously reported patterns of β-phase InSe, confirming the high purity of the sample. [18,22]Figure S2 (Supporting Information) shows the schematic representation of the β-phase InSe crystal structure.Furthermore, Raman spectra of the waveguide-integrated InSe also confirm the β-InSe phase as depicted in the supplementary Figure S3 (Supporting Information).All observed active phonon modes are in agreement with other reports on β-phase InSe. [23]As shown in Figure S4 (Supporting Information), high-resolution spectroscopic ellipsometry performed on InSe flakes on 300 SiO 2 /Si substrate is used to determine its refractive index (n) and extinction coefficient (k).These data are fed into optical simulations.
To study the device's optical characteristics, numerical investigations of the hybrid InSe/SiN structure's effective refractive index (n eff ), mode confinement, and light absorption loss (κ) are carried out for a 90 nm InSe thick flake.The modal and propagation properties are simulated using an eigenmode solver and finite differences in the time domain (FDTD) (see Experimental Section).Figure 2c (top panel) shows the electric-field profiles (|E| 2 ) of the fundamental transverse electric field modes (TE 0 ) of the unloaded SiN waveguide and loaded structures.Table S1 (Supporting Information) encloses simulated information about the waveguide parameters.Since the InSe flake has a higher refractive index (n = 3.1) than the SiN waveguide (n = 2.7); the optical mode is more confined to the InSe flake, leading to higher optical absorption in that layer.
The evanescent field coupling in the hybrid structure over a propagation length is simulated using FDTD, as shown in Figure 2c  propagation losses for the loaded SiN waveguide is 0.27 dB μm À1 .It is evident that an interaction length of 30 μm and a flake thickness of 90 nm result in near-complete absorption, offering a reduced device footprint.The absorption of the on-chip integrated InSe photodetector depends on the polarization of the propagating light.FDTD simulations revealed that the TE polarization is more confined into the InSe compared to the TM polarization (see Figure S5 and Table S1, Supporting Information).Therefore, we performed all the experiments under the TE polarization.Moreover, it is important to note that the crystal axis orientation rules the absorption in InSe, [24] and given that the flakes are randomly transferred, this effect was not controlled in our experiments.Furthermore, the optical losses before and after the integration of multilayer InSe are measured by coupling the laser light into a reference waveguide and monitoring the output power using an optical spectrum analyzer (OSA).The total coverage area of the InSe on waveguide is 93 μm, and this value was not selected but measured on the fabricated device.This length results from two flakes that are simultaneously deposited on top of two different regions of the same waveguide (57.49 and 37.5 μm long respectively, as shown in Figure S6, Supporting Information).Figure 2d depicts the introduced losses, calculated from measured transmissions of loaded and unloaded waveguides (the inset shows a schematic of the setup).This result is based on an InSe thickness of 90 nm and an interaction length of %93 μm at 976 nm.The measured optical loss due to the absorption in the loaded device is 0.11 dB μm À1 , which is in the same range as the simulated losses.The discrepancy between the calculated and measured propagation losses can be attributed to variability and the dependence on the absorption of the InSe crystal's orientation, as discussed earlier.

Static Device Response
The current-voltage characteristics of the integrated photodetector are tested in the dark state and under monochromatic laser excitation of 976 nm.As previously stated in Section 2.2, the refractive index contrast between InSe and SiN induces high optical confinement within the active InSe layer (see Figure 2c).In this case, InSe absorbs photons via direct band-to-band transitions, which is confirmed by the measured photoluminescence of InSe (see Figure S7, Supporting Information).It also leads to an increase in electrical conductivity due to the electron-hole pairs generation.Additionally, the photoexcited carriers drift using the lateral electric field applied between the two in-plane metal electrodes.
Figure 3a shows the I-V characteristics of the photodetector under dark and at different coupled light intensities, in a range between 0.21 and 340 μW.As shown in the inset of this figure, the devices exhibit a low dark current density %4.5 nA μm À2 at 1 V and increase to about %40 nA μm À2 at 5 V bias (device area %240 μm 2 ).Note that due to the anisotropic transport behavior of InSe, the dark current values also depend on the crystal orientation.This effect has been studied by Guo et al., demonstrating an angle-dependent dark current in InSe with a ratio of 3.76 (V ds = 1 V) between two orthogonal orientations. [23]At a coupled light intensity of 0.21 μW, an open-circuit voltage (V OC ) and short-circuit current (I SC ) of 0.19 V and 63 pA, are measured, respectively.Increasing the light intensity further to 340 μW leads to a V OC and I SC of 0.61 V and 73 nA, respectively.
It is worth noting that the work function difference between Au (5.1 eV) and InSe (4.6) results in a Schottky junction. [25]Hence, the fabricated device is comprised of back-to-back Schottky diodes.However, it is clear that the device exhibits an asymmetric transport behavior.This indicates that the Schottky barrier height at the two Au-InSe junctions is not the same.This can be attributed to defects and trap states at the interface between metal and semiconductor, which are difficult to control and can result in Fermi-level pinning. [26]Additionally, the asymmetric contact geometries, defined as the difference in the contact area between the InSe flake and the metals on both sides, may result in a rectifying behavior despite the symmetric metal electrode configuration. [27]19a,19b] The generated photocurrents (I ph = I light -I dark ) under different powers as a function of the applied voltage are depicted in Figure 3b.It shows a high photogenerated current which can be attributed to the InSe's direct bandgap nature and the extended interaction length with the optical guided mode.The photocurrent versus incident optical power curve (see Figure S8a, Supporting Information) indicates a nonlinear behavior for small optical powers, which may be primarily attributed to the presence of long-lived traps in the InSe. [28]he performance of the photodetectors is also assessed by measuring their responsivity (R = I ph /P).It is calculated as the ratio of the photocurrent (I ph ) to the optical power (P) received by the detector.The optical losses were calibrated using a test structure which consists of bare reference SiN straight waveguides integrated on the same chip.We used a laser operating at 976 nm to edge-couple transverse electric (TE) polarized into the reference waveguide using a lensed fiber.The output response was also collected by another lensed fiber and detected using a power meter.The total measured insertion loss of the reference SiN waveguide (P r ) is 14.5 dB.The initial power which is launched into the SiN/InSe device is (P in ).The total power (P) delivered to the integrated device is used to determine the EQE and responsivity calculation is deduced by subtracting the P in -P r .
Figure 3c shows a 3D plot of responsivity as function of applied voltage and laser power.A photoresponsivity of 0.38 A W À1 at 5 V is measured.This corresponds to an external quantum efficiency (EQE = 1.24Â R/λ) of 44%.10b,18] We can express the EQE = R (hν/e) = [tα] τ l/ τ t .Here, α is the absorption coefficient of InSe at the photon energy hν, t is the thickness of the InSe layer, and τ l /τ t is the ratio of the minority carrier lifetime (τ l ) and transit time (τ t ) of electrons in InSe, [29] which indicates EQE is mainly governed by the ratio τ l/ τ t .Therefore, the decrease of EQE with increasing power suggests a decrease of τ l and/or an increase of τ t .Under low laser excitation powers, the minority carriers are trapped which enhances the lifetime of the minority carriers (τ l ), which increases the EQE of the device.However, at high excitation powers, the defect/trap states are saturated, and no further minority carriers are trapped, which decreases the lifetime of the carriers, resulting in low EQE.Based on the obtained results, the photocurrent increases with the applied voltage, unlike the bolometric effect where the photocurrent has an opposite sign to the applied bias.Additionally, since the detector has a symmetric electrode configuration, hence, the photo-thermoelectrical effect (PTE) can be ruled out.Therefore, our photodetector's performance is dominated by the photoconductive effect.
Furthermore, the photodetector's noise equivalent power is calculated to assess its minimal detectable power (noise equivalent power NEP).The latter is defined as i n /R, where i n is the noise current and R is the responsivity.Therefore, the NEP indicates the excitation power needed to produce a photocurrent equal to the noise current.There are typically three noise sources present: 1=f noise, shot noise, and Johnson noise, where f is the frequency.However, for high-frequency signals f > 1 kHz, the total noise current is determined by the Johnson noise (i nJ ¼ ð4k B TΔf Þ=R 0 , where k B is the Boltzmann constant, T is the temperature, R 0 is the channel resistanceÞ, and shot noise (i ns = 2eðI d þ I ph ÞΔf , where q is the electric charge, I D is the dark current, Δf is the device bandwidth). [30]The measured Johnson noise, shot noise, and NEP at 5 V bias are 0.17 nA Hz À1/2 , 1.68 nA Hz À1/2 , and 4.7 nW Hz 1/2 , respectively (See Figure S8b, Supporting Information).

High-Speed Photo Response
To test the dynamic performance of the photodetector, its frequency response (RF) is measured.A continuous wave laser at 976 nm is modulated by a commercial optical high-speed modulator.A DC source is used to bias the photodetector using a bias tee.The RF signal is then fed into the network spectrum analyzer (ESA).The schematic of the experimental setup is shown in supplementary Figure S9 (Supporting Information).
Figure 4a shows the average frequency response of the InSe photodetector measured at 10 V bias.The device exhibits a robust reproducible response with a 3-dB cutoff frequency of 85 MHz at 10 V bias.The extracted rise time τ rise % 0.35=f 3dB is approximately 4.1 ns, where τ rise is the time between the 10% and 90% transient response, equivalent to a time constant of τ = 1.87 ns.To the best of our knowledge, this InSe-based photodetector has the smallest recorded response time, several orders of magnitude smaller than earlier investigations at the NIR band, and particularly 5-6 orders of magnitude smaller than other InSe studies (see below Table 1).The achieved short response time is attributed to the high mobility nature of the InSe.It is usually related to the carrier mobility of the material, the channel's width, the applied bias, and the mechanism of photoelectric detection.Recent research by Bandurin et al. reported a few-layer InSe (6-10 layers) with carrier mobilities exceeding 10 3 and 10 4 cm 2 V À1 s À1 at room and liquid-helium temperatures, respectively. [31]Other studies on InSe multilayers (>20 layers) demonstrated mobilities higher than 2000 cm 2 V À1 s À1 achieved by suppressing carrier scattering at the dielectric interface. [13,32]he aforementioned studies reveal the superior mobility characteristics of InSe.
However, high-mobility InSe has not yet explored on the waveguide-integrated platform.Moreover, the small size of the waveguide-integrated devices results in low absolute dark current.The integration of photodetectors with waveguides can overcome a trade-off problem between the efficiency and bandwidth of the photodetector by having a photon-absorption path and a carrier-collection path perpendicular to each other. [33]erefore, compared to their free-space counterparts, waveguide-integrated InSe detectors not only facilitate planar photonic integration, but also offer a significant performance advantage in terms of speed and signal-to-noise ratio (SNR).
The temporal response of an InSe device is studied using an arbitrary waveform generator and high-speed optical modulator, and the output signal shape is monitored using a high-speed oscilloscope.The complete setup is shown in the supplementary Figure S10 (Supporting Information) and the voltage-dependent f 3dB is depicted in Figure S11 (Supporting Information).The measured and smoothed responses at V ds = 10 V are shown in Figure 4b,c using a modulated sinusoidal signal.The devices were tested under different bias conditions.Poor signal-to-noise ratios (SNR) were observed at low bias voltages.To achieve high data rates and a high SNR, we used 10 V bias.Further, under high bias voltage (10 V), the transit time is improved and a better SNR (photocurrent to dark current) can be achieved.The signalto-noise ratio of the temporal responses is evaluated to be 5.8 and 3.4 at 100 MHz and 1 GHz, respectively.It is important to note that the signal can be clearly discerned at 1 GHz.The temporal width at half-maximum amplitude (FWHM) is a measure of the pulse window width, which is the relationship between the full window temporal FWHM and frequency for different modulation frequencies (100 MHz and 1 GHz).The FWHM values of the temporal responses are measured to be 4.3 and 0.41 ns at 100 MHz 1 GHz, respectively.The photodetector is further tested for an on-off modulated pseudorandom binary sequence (PRBS) data transmission.It is evident that there is room for enhancing the bandwidth of the InSe device.One can improve the performance of the integrated InSe device by the following methods: 1) shorten the InSe channel length to further increase the bandwidth.Photodetectors with a channel length of %10 μm, need mobility of %1000 cm 2 V À1 s À1 to reach GHz range bandwidth; 2) In the current study, InSe flake is transferred onto predeposited metal electrodes, which results in a higher contact resistance compared to the buried devices; 3) InSe exhibits a polarization-dependent anisotropic ratio of dark current to photocurrent, therefore, a control over the InSe flake orientation during transfer can enhance the device performance; 4) Creating a heterojunction with graphene or using graphene as electrodes; and 5) Encapsulation of the InSe with hexagonal boron nitride (hBN) can yield higher carrier mobility.Note that no passivation before the transfer of InSe on the SiN waveguide is used.It is well established that a transfer of hBN before the integration can passivate the waveguide and reduce roughness.

Conclusion
In summary, we demonstrated an InSe/SiN waveguideintegrated photodetector operating in the near-infrared (976 nm) regime.The fabrication process is simple compared to other integrated photodetectors.The as-assembled devices exhibited an extrinsic responsivity of 0.38 A W À1 (EQE of %44%), a low dark current density of 40 nA μm À2 at 5 V, and a fairly low NEP of 4.7 nW Hz À1/2 .Additionally, a high-index contrast of InSe compared to SiN waveguide of %0.4, it results in improved optical absorption.The measured absorption coefficient for a 90 nm thick flake is 0.11 dB μm À1 .Additionally, the device showed a 3 dB cutoff frequency of 85 MHz at 10 V and open eye diagrams at a bit rate of 0.5 up to 1 Gbit s À1 .This InSe-based photodetector exhibits high bandwidth and outperforms similar reported devices by more than 5 to 6 orders of magnitude.We believe that the bandwidth can be further improved by increasing the carrier mobility, enhancing the waveguide-InSe interface, and reducing the circuit capacitance.This study provides a realization of high-performance photodetectors using a simple architecture with potential applications in a wide range of functions including lab-on-a-chip devices, LiDAR system imaging, and short-reach interconnects for data centers.

Experimental Section
Materials and Fabrication: Crystal Growth: The InSe single crystals were produced using a 99.999% pure molar combination of In (52.4% weight percentage) and Se (47.6%) compounds that were purchased from Sigma-Aldrich.Conical quartz ampoules evacuated to 10 À4 Pa were used to produce single-crystalline InSe flakes.A horizontal furnace was used to homogenize the batches and create the InSe flakes for 48 h at 550 °C.The Bridgman vertical method was used to grow the mixed crystals.The melt-filled ampoules were heated at 850 °C for 24 h before pulling; once the melt had filled the ampoule's tip, the ampoules were lowered via a 1 °C temperature gradient at a rate of 0.1 mm h À1 .The InSe crystals that were formed had dimensions of 3 and 1.2 cm.
Si 3 N 4 Platform: The passive SiN photonic devices were fabricated with standard 220-nm-SOI processes in Applied Nanotools Inc. Metal pads in the devices were patterned by electron-beam lithography followed by the deposition of metal pads (Cr/Au, 4/100 nm) by electron-beam evaporation and a standard lift-off process.After depositing the metal pads, InSe is transferred with the help of a PDMS film.InSe Transfer: A few layers InSe flakes were exfoliated using scotch tape and transferred onto the PDMS substrate.By using the controllable dry transfer method, the selected InSe flakes were transferred onto the SiN waveguides.
Simulations: The electric field profile in the silicon nitride waveguide and the beam propagation were calculated using the MODE Solutions eigenmode solver and FDTD simulation in Lumerical.The optical parameters used for InSe were those extracted from ellipsometry.
Material Characterization: Spectroscopic Imaging Ellipsometer: The optical parameters of multilayer InSe are determined by Accurion's Imaging Ellipsometry (https://accurion.com/company).This system combines optical microscopy and ellipsometry for spatially resolved layer-thickness and refractive index measurements.The tool is highly sensitive to ultrathin single-and multilayer films, ranging from monoatomic or monomolecular layers (sub-nm regime) up to thicknesses of several microns.Additionally, imaging ellipsometers can perform layer thickness measurements with a spatial resolution down to 1 μm.The ellipsometric parameters (Psi (ψ) and Delta (Δ)) were fitted using EP4 model software.
Scanning Electron Microscopy: A (FEI) Quanta 450 field emission scanning electron microscope with an electron energy of 10 KV was used to image the photonic chips while they were placed on an SEM stub using carbon tape.
Atomic Force Microscopy: The tapping mode of the WITec atomic force microscope (AFM) module was used to determine the thickness of the transferred InSe flake.The cantilever tip (Scanasyst-air) had a radius of 7 nm, a force constant of 0.2 N m À1 , and a resonance frequency of 14 kHz.
Device Characterization: Optical Characterization: The optical transmission was performed using edge coupling the light into the photonic chips through lensed fiber and a tunable laser operating at 976 nm.The output response from the devices was collected by an output lensed fiber and detected by a power meter.The output optical power intensities were calibrated before testing the device using a standard photodiode power sensor.
DC Measurements: The steady-state performance of the InSe photodetectors was tested by measuring their dark current and responsivity.A transverse electric-polarized (TE) of 976 nm light was edge-coupled via lensed optical fiber to the devices.A curve tracer/power device analyzer/(Agilent B1505A) was used to control the biases and measure the I-V characteristics in the dark and upon light coupling via a pair of standard DC electrical probes.

Figure 1 .
Figure 1.Hybrid integration of InSe into SiN photonic circuit: a) Optical image of the SiN unloaded waveguide chip.b) 3D cross-section representation of heterogeneous InSe/SiN photodetector.c) An SEM image of the transferred InSe on SiN waveguide and interaction length of %30 μm.d) An AFM image scan, the dashed yellow line shows a thickness of %90 nm InSe.

Figure 2 .
Figure 2. Material characterization and guided light interaction of 90 nm thick InSe layer with a length of 30 μm at 976 nm wavelength: a) SEM scan with EDS maps captured from the multilayer InSe.The scale bar is 50 μm.b) XRD spectrum of the bulk β-InSe.c) Electric-field profiles (|E| 2 ) of TE modes of unloaded SiN waveguide and 90 nm InSe on SiN at 976 nm (top panel).d) FDTD simulation of beam propagation (bottom panel) measured optical propagation losses of InSe.Inset shows a schematic representation of the measurement setup.

Figure 5
shows the measured eye diagrams of the photodetector at 10 V bias.A clear and open eye

Figure 4 .
Figure 4. Dynamic performance characteristics of the InSe photodetector.a) Normalized frequency response at 10 V. A total of 50 measurements (bluescattered points) and average (red line) data are plotted.A 3-dB cutoff frequency of 85 MHz is measured.b,c) The temporal response recorded in the time domain for 100 MHz and 1 GHz modulated optical signals, the curves are smoothed by using 200 points moving average.
diagram is obtained for non-return-to-zero (NRZ) data rates 0.1 Gbit s À1 .Additionally, an open-eye diagram is recorded at a bit rate of 0.5 and partially open one at 1 Gbit s À1 .The performance of the SiN/InSe photodetector is compared to other waveguide-integrated photodetectors in FigureS12(Supporting Information).Notably, most integrated photodetectors are reported for telecom bands, and this study's f 3dB is lower compared to graphene or black-phosphorous (BP)-based photodetectors.From the capacitance-resistance (RC) measurements, the measured values are %0.1 pF, and 0.1 MΩ at 1 MHz frequency, respectively, which corresponds to RC limited bandwidth of %100 MHz.

Figure 5 .
Figure 5. Measured eye diagrams at data rate from 100 Mbit s À1 to 1 Gbit s À1 NRZ modulation at 10 V bias.

Table 1 .
Comparison of the performance of InSe/SiN photodetector with other NIR photodetectors.