Photogating-assisted tunneling boosts the responsivity and speed of heterogeneous WSe2/Ta2NiSe5 photodetectors

Photogating effect is the dominant mechanism of most high-responsivity two-dimensional (2D) material photodetectors. However, the ultrahigh responsivities in those devices are intrinsically at the cost of very slow response speed. In this work, we report a WSe2/Ta2NiSe5 heterostructure detector whose photodetection gain and response speed can be enhanced simultaneously, overcoming the trade-off between responsivity and speed. We reveal that photogating-assisted tunneling synergistically allows photocarrier multiplication and carrier acceleration through tunneling under an electrical field. The photogating effect in our device features low-power consumption (in the order of nW) and shows a dependence on the polarization states of incident light, which can be further tuned by source-drain voltages, allowing for wavelength discrimination with just a two-electrode planar structure. Our findings offer more opportunities for the long-sought next-generation photodetectors with high responsivity, fast speed, polarization detection, and multi-color sensing, simultaneously.

incident angle between the polarization directions of the incident light and x-axis of the crystal, was varied from 0 to 360°.The different Raman modes intensities change periodically with the variation of rotation angle under the parallel-polarized configurations, indicating a strong inplane vibration anisotropy (Supplementary Fig. 2d).The characteristic Raman peaks of Ta2NiSe5 were identified from the spectrum.The Raman active modes of 96.6, 121.1, 147, 176,   190, and 214 cm -1 are assigned to A g 1 to A g 6 , respectively, which is consistent with previous reports. 3,4 s shown in Supplementary Fig. 2e, the peak intensity can be written as: 2 () + 2 +     2 (), the fact that the intensity of Raman peaks changed periodically with θ from 0 to 180° is noteworthy, and the same changing trend was confirmed in multiple samples.The A g 1 mode is minimized when the direction of laser polarization is parallel to the y direction (i.e., the long axis of the flake), which corresponds to a-axis of the crystal, according to previous reports. 1 verify the difference in conductivity between the two orthogonal directions, we also fabricated a multi-electrode device based on the Ta2NiSe5 crystal, as shown in Supplementary Fig. 2f.The crystallographic direction of the flake is confirmed by angle-resolved polarized Raman spectroscopy.We found that the current along the a-axis direction (≈ 2×10 -3 A at Vds=1 V) is significantly higher than that along the c-axis direction (≈ 5×10 -4 A at Vds=1 V).According to above measurements, a convenient way to determine the crystal orientation is by inspecting the crystal shape, and the crystal is generally elongated along the a axis of the crystal.
The coupling effect and charge transfer at WSe2/Ta2NiSe5 interface can be verified via optical characterization.Supplementary Fig. 3a shows the non-polarized Raman spectra collected from isolated WSe2, Ta2NiSe5, and the overlapped heterojunction region, respectively.
The characteristic peaks of both WSe2 and Ta2NiSe5 were observed in the overlapped heterostructure region.The corresponding Raman mapping images were measured at the characteristic Raman peaks of Ta2NiSe5 (121 cm −1 ) and WSe2 (260 cm −1 ), respectively, as shown in Supplementary Fig. 3b,c.The Raman mapping images exhibit a good homogeneity of each component, indicating high quality of the heterostructure after the exfoliation and target-transfer processes.We also collected the PL spectra from different areas of the heterostructure with a laser illumination of 532 nm (Supplementary Fig. 3d).Isolated Ta2NiSe5 shows no obvious PL signal in the measured wavelength range due to its narrow optical bandgap. 3,5 wo obvious peaks at ~780 nm (corresponding to ≈1.6 eV) and 860 nm (≈1. 2 eV) are observed for WSe2.These two PL peaks are related to the A excitons (A) and indirect transitions (I), which is in accordance with the PL spectra of multilayer WSe2. 6,7 ably, the PL intensity at the heterojunction is much weaker than that of isolated WSe2.][10] In addition, Supplementary Fig. 3f shows the UV-vis-NIR absorption spectra of Ta2NiSe5, WSe2 and WSe2/Ta2NiSe5, respectively.The peak at ~760 nm of WSe2 in the absorption spectrum is attributed to its strong excitonic absorption. 11In contrast, multilayer Ta2NiSe5 has an obvious broad spectral absorption because of its narrow bandgap.4.0x10 -9

Supplementary Note 4. Electrical characterization of Ta2NiSe5 and
WSe2 field effect transistors.
To investigate the electrical polarity of Ta2NiSe5 and WSe2 respectively, we fabricated Ta2NiSe5 and WSe2 field effect transistors (FETs) on Si/SiO2 substrate and then performed the electrical measurements at room temperature.Supplementary Fig. 7a, e shows the optical images of the back-gated Ta2NiSe5 and WSe2 with Ti/Au source/drain electrodes, respectively.
The WSe2 and Ta2NiSe5 flakes were cleaved by the mechanical exfoliation and transferred on the Si/SiO2 (300 nm) substrate.The morphology and thickness of the flakes were characterized using AFM, and it can be observed that the surface of the sample is very flat (Supplementary Fig. 7b, f).As is shown in Supplementary Fig. 7c, g, the thickness of the Ta2NiSe5 and WSe2 flakes were determined to be ~19 nm and 86 nm, respectively.Supplementary Fig. 7d shows the transfer curve of the Ta2NiSe5 FET with a linear plot of the source-drain current (Ids) versus back gate voltage (Vgs) when sweeping from −5 to +5 V at a fixed source-drain voltage (Vds = 1 V).The transfer curve suggests the weakly n-type semiconducting property of Ta2NiSe5.The carrier mobility (μFE) of Ta2NiSe5 and WSe2 were further extracted from the transfer curves, using a relation given as where L and W represent the channel length and width of the device, ( V Supplementary Note 7. Photocarrier tunneling at positive bias voltage under light illumination. The Schottky barrier in a Schottky junction essentially determines the junction's behavior by governing electron/hole transport.Depending on the height and width of the barrier, charge transport through the barrier may occur through thermionic emission (TE) or tunnelling, thereby we analyzed the device characteristics by using these two models in the following.
TE model is widely used to explain the electron transport mechanism of Schottky junction and to extract the Schottky barrier height. 15,16 ccording to TE theory, density of current through the barrier is determined as where A* is the known Richardson constant in the previous study, which is 27.6A cm −2 K −2 for the WSe2. 17,18  is temperature, k is the Boltzmann constant, q is the electronic charge, V is the bias voltage applied.The saturation current is described as where  0 is the barrier height at zero bias.Experimentally, the saturation current can be estimated from the intersection of a linear approximation of ln J(V) and the Y-axis.So,  0 is determined by equation In real structures an I-V dependence deviates from the ideal theory, and an ideality factor  is introduced: According to the above formula, it is found that the barrier height ( 0 ) and ideal factor  of Schottky junction can be extracted from the intercept and slope of the ln J(V) curve in semi-logarithmic coordinates. 19 According to formulas ( 5) and ( 6), the calculated values of  and  0 are 11.2 and 0.618 eV for the device, respectively (Supplementary Fig. 11a).However, it is found that the ideal factor  is far greater than 1, indicating that there was an inaccuracy between the TE model and the experimental results.
Thereby, tunneling may dominate the transport of our device.It is noted that the Ids-Vds curve at positive bias voltages can be well modeled by a tunneling barrier with the Simmons approximation (Fig. 2f and Supplementary Fig. 11b, c).The dominant tunneling occurs with direct tunneling (DT) at low bias voltage and Fowler-Nordheim tunneling (FNT) at high voltage.The DT and FNT can be expressed by [20][21][22][23]   ∝  (− Where d, m*, ϕ, h is the tunneling thickness, effective electron mass, tunneling barrier, and the Plank constant, respectively.The fitting plot of ln (I/V 2 ) versus 1/V shows linear dependence with a negative slope for the FNT under larger Vds, and rises exponentially for the DT under small Vds under light illumination, as demonstrated in the figures.Similar phenomenon is also observed for other devices (Supplementary Fig. 11d-f, the device 3 as described in Supplementary Note 9).We also extract the corresponding parameters of device 3, and the calculated values of  and  0 are 21.4 and 0.564 eV for the device respectively (Supplementary Fig. 11d).We found a similar phenomenon as Supplementary Fig. 11a, where the n value is far greater than 1.It is further verified that our device is not dominant by TE transport.The tunneling-mediated transport in the device 3 is also confirmed with a temperature-dependent measurements in Supplementary Fig. 11e-f.The plot of ln (I/V 2 ) versus 1/V displays small variations as the temperature decreases from 300 K to 128 K, with the linear region of curves exhibiting nearly the same slope.The above observations indicate that the tunneling-dominated transport of charge carriers under positive biases.
-6 -4 -2 0 2 4 6 0.0 2.0x10 6 4.0x10 6 6.0x10 6 8.0x10 6 1.0x10 7 J dark (nA/cm 2 ) V ds (V) here I0 denotes the saturation photocurrent, A and B are two constants, τd1 and τd2 are the fast and slow time constants for the decaying photocurrents, respectively.Based on equation ( 2), the values of τd1 and τd2 under light illumination at different bias conditions were obtained, respectively (Supplementary Fig. 12).The relatively fast process (τd1) is attributed to the recombination of the carriers that occurs when the light is turned off, while the slow process (τd2) is caused by detrapping of the carriers at defects/traps, which is corresponding to the carrier lifetime, .
Then, the value of carrier transit time   can be extracted according to   =  / .The carrier transit time is 61 µs and 3.57 ns at -1 V and +1 V bias voltage under 785 nm illumination.
It can be clearly seen that the carrier transit time of the device under positive bias is about four orders of magnitude faster than that under negative bias.Similar results are observed under incidence of other wavelengths .
Supplementary Fig. 12 The extracted carrier lifetime under light illumination at different bias conditions.a, 635 nm, Vds = -1 V; b, 635 nm, Vds = 0 V; c, 635 nm, Vds = 1 V; d, 532 nm, Vds = -1 V; e, 532 nm, Vds = 0 V; f, 532 nm, Vds = 1 V; g, 785 nm, Vds = -1 V; h, 785 nm, Vds = 0 V; i, 785 nm, Vds = 1 V.  8x10 -9 8x10 -9 Power density (μW/mm 2 ) where   is the dark current, e is the elementary charge, B is the bandwidth, T is the temperature, and  Ω is the shunt resistance of device (The value of  Ω can be obtained by Power density 4.0x10 -9 6.0x10 -9 8.0x10 -9 1.0x10 -8 1.2x10 -8 fitting the I-V experimental data under weak voltage).The  ℎ is thermal noise,  ℎ is shot noise,  1/ is 1/f noise, and  − is generation-recombination noise, respectively.The 1/f noise and g-r noise are dominant at low frequencies, ascribing to the interface traps or defects. 30,33 Asis shown in Supplementary Fig. 17d,e, when a bias voltage of -1 V and 0 V are applied, since the noise spectrum is almost frequency independent at the bandwidth of 1 Hz, white noise is the primary source of noise current.Therefore, the noise current (  ) can be expressed as: where √2   is short noise and √ 4  Ω is thermal noise.Due to the ultra-large value of RΩ in our device (RΩ > 7.9 GΩ), the  ℎ 2 is far less than  ℎ 2 ( ℎ 2 ≪ ( ℎ /10) 2 ), so the noise current can be estimated as   = √2  .It is obvious that the noise mainly comes from the dark current.According to the noise spectral density extracted at the bandwidth of 1 Hz, the calculated D* for Device 1 as listed in Supplementary Note 9 is 2.18×10 12 Jones (Vds = -1 V) and 9.6×10 10 Jones (Vds = 0 V) at 785 nm with a power density of 0.05 μW/mm 2 , respectively.
However, under the bias voltage of +1 V, we found that the noise current has a weak frequency dependence at around the bandwidth of 1 Hz.Thus, the specific detectivity can be expressed as: 29 where R is responsivity and A is device active area.According to the formula (3), the D* is 8.3×10 13 Jones at 785 nm for Device 1 with a power density of 0.05 μW/mm 2 , which is very close to the detectivity (1.5×10 14 Jones) calculated assuming the dark current dominants the noise current.Thereby, to simplified, we derived the specific detectivity from the formula D* = RA 1/2 /(2qIdark) 1/2 for the Device 1 to 6 as listed in above Supplementary Note 9.
Supplementary Fig. 20d-f show the optoelectronic response of the device under negative bias (Vds= -1 V).Under the illumination of 1064 nm light, the R of the device is 368 mA/W under light illumination of 0.002 mW/mm 2 (Supplementary Fig. 20d), which is three orders of magnitude less than that under forward bias.The corresponding D* is 1.12 × 10 11 Jones.The device has a longer response time with a rise/fall time of 5.11/5.55ms (Supplementary Fig. 20f).Compared with the cases under positive bias voltage, the response speed is reduced by about 10 times.

Fig. 1
Schematic of the crystal structures of a, Ta2NiSe5, and b, WSe2.Supplementary Fig. 2 Anisotropic structural characterization of Ta2NiSe5 crystals.a, Schematic of the anisotropic structure of Ta2NiSe5.b-c, Optical microscope images of the exfoliated Ta2NiSe5 flakes.The white arrow represents the direction of polarization of light, the angle between the white arrow and the x-axis is defined as the polarization angle θ. d, The angle-resolved polarized Raman spectra of Ta2NiSe5.e, The peak intensity (A g 1 mode) as a function of polarization angle.f, The anisotropic electrical transport property of Ta2NiSe5.Inset: Top-view optical image of the fabricated multi-electrode Ta2NiSe5 device.

Fig. 3
Optical characterization of the WSe2/Ta2NiSe5 heterojunction.a, The non-polarized Raman spectra collected from isolated WSe2, Ta2NiSe5, and the overlapped heterojunction region.b-c, The corresponding Raman mapping images measured at the characteristic Raman peaks of (b) Ta2NiSe5 (121 cm −1 ) and (c) WSe2 (260 cm −1 ).The blue and white dashed lines represent the regions of WSe2 and Ta2NiSe5, respectively.d, PL spectra of isolated WSe2, Ta2NiSe5, and the WSe2/Ta2NiSe5 heterojunction.The dashed line corresponds to the position of the exciton absorption peak of WSe2.e, The corresponding PL mapping image of the WSe2/Ta2NiSe5 heterojunction.f, UV-vis-NIR absorption spectra of Ta2NiSe5 ).The results give strong signals of Ta, Ni, Se, and W, Se, respectively, and basically agree with the stoichiometric ratio with slight selenium vacancy.Supplementary Fig. 5 Characterization of the WSe2/Ta2NiSe5 heterojunction.a, SEM image taken at the heterostructure region.The blue and white dashed lines represent the regions of WSe2 and Ta2NiSe5, respectively.b-c, The corresponding EDS results of Ta2NiSe5 and WSe2, respectively.Supplementary Fig. 6 Output characteristics (Ids-Vds) of the WSe2/Ta2NiSe5 heterostructure device.a, linear scale and b, log-scale of the curve.

Furthermore, we found
that Ta2NiSe5 has a near ohmic contact with the metal electrode as shown in the inset of Supplementary Fig.7d.On the other hand, the transfer curve of WSe2 transistors at Vds = 1 V displays a typical hole-dominated p-type semiconducting property (Supplementary Fig.7h).The output curves at different back-gate voltages are shown in the inset of Supplementary Fig.7h.The output current gradually decreases as the gate voltage changes from negative to positive, further proving that holes are the majority carriers in the material.Moreover, the Ids-Vds curve under small gate bias shows that a Schottky contact was formed between WSe2 and Ti/Au electrode in our experiments.
curves, Cbg (∼ 115 aF/μm 2 ) is the gate capacitance.Thus, the μFE of Ta2NiSe5 and WSe2 were estimated to be 12.43 and 4.64 cm 2 /V•s, respectively.Supplementary Fig 7 Electrical characterization of Ta2NiSe5 and WSe2 FETs.a, Optical image of the back-gated Ta2NiSe5 FET.b, AFM topography of the Ta2NiSe5 FET.c, Thickness measurement taken along the red solid line in b. d, Transfer curve (Ids-Vgs) of the Ta2NiSe5 FET.Inset: output curve (Ids-Vds) of the Ta2NiSe5 FET.e, Optical image of the back-gated WSe2 FET.f, AFM topography of the WSe2 FET.g, Thickness measurement taken along the red solid line in f. h, Transfer curve (Ids-Vgs) of the WSe2 FET.Inset: output curve (Ids-Vds) of the WSe2 FET.
image that reveals the surface potential difference of the WSe2/Ta2NiSe5 heterojunction.b, Surface potential difference taken along the red line in a. c-d, Band diagrams of the heterostructure device (c) before and (d) after contact.The arrow represents the direction of carrier transport.e-g, The spatially resolved photocurrent mapping images at (e) Vds = 0 V, (f) Vds = -1 V, and (g) Vds = 1 V. Scale bar: 6 μm.The illumination wavelength is 633 nm.The white, blue and yellow dashed lines represent the regions of WSe2, Ta2NiSe5 and metal electrodes, respectively.
10 Photoresponse of the WSe2/Ta2NiSe5 heterostructure.a-b, Output curves (Ids-Vds) in dark and under illumination with different power densities.The incident light is at 635 and 785 nm wavelength, respectively.Vgs = 0 V. c-d, The corresponding transfer curves (Ids-Vgs) of the device at Vds= 1 V. ∆Vg and the shaded area represent the range of the change of the charge neutrality point.

Supplementary Note 8 . 25 𝐼
Photodetection gain of a photodetector can be estimated from the device responsivity, following the below equation

Supplementary Table 1 :
bias voltage is obviously higher than that of the negative bias voltage.For instance, Device 6 shows a rise time of 23.6 µs under positive bias in Supplementary Fig. 14i, which is around 10× faster than that of negative bias, indicating the excellent capability of the device to follow ultrafast switching light signals.We then measured the relative response with optical modulation.The 3 dB bandwidth of devices are extracted according to the dependence of photocurrent on the optical modulation frequency.As shown in Supplementary Fig. 15, the 3 dB cutoff frequency measured for device 4-6 can reach up to 130-195 kHz under 785 nm illumination.The response time of the device is estimated to be 2.7 µs -1.8 µs by the equation: f3dB = 0.35/tr, where tr is the response time of the device.Thickness parameters of heterojunction devices Supplementary Fig. 13 Photoresponse of WSe2/Ta2NiSe5 heterostructures with thicker WSe2 (which is corresponding to devices 1, 2 and 3).a,d,g, Responsivity and detectivity at , ) 1/ 2 + (, ) − 2

Fig. 19
Broadband infrared photoresponse of the WSe2/Ta2NiSe5 heterostructure.a-d, The current−voltage (Ids−Vds) curves measured both in dark and under 1064 nm to 2200 nm light illuminations.e-h, The time-dependent photoresponse under 1064 nm to 2200 nm light illuminations.Vds = 1 V, Vgs = 0 V.

characterization under different light illumination at Vds = 1 V
and -1 V. a, Responsivity and detectivity at 1064 nm, 1310 nm, 1550 nm, 2200 nm at Vds = 1 V. b, Photocurrent as a function of incident light powers at Vds = 1 V. c, Single magnified response curves at Vds = 1 V. d, Responsivity and detectivity at 1064 nm, 1310 nm, 1550 nm, 2200 nm at Vds = -1 V. e, Photocurrent as a function of incident light powers at Vds = -1 V. f, Single magnified response curves at Vds = -1 V.
photocurrents under illumination of different wavelengths from 785 to 1550 nm at Vds = -1 V.