1. Introduction

Recently, research on two-dimensional (2D) material-based photodetectors has drawn significant attention by virtue of their broad sensing capabilities (UV-THz), their ability to be integrated onto photonic chips, and their high-speed performance [1]. In particular, 2D materials with high carrier mobility, short-wavelength infrared (SWIR) light detectability, and stable ambient performance are in great demand for future telecom applications [2]. In this regard, graphene is the first on the list where several experimental studies revealed its superior performance in the SWIR region (76 GHz), however, its semi-metal nature and atomic layer thickness limits its responsivity (1 mA/W) [1,3]. On the other hand, graphene/quantum dots heterostructure demonstrated a high responsivity (10⁷ A/W) but showed a low-frequency response (< 10 Hz) [4]. This tradeoff poses a big challenge for graphene-based photodetectors [5]. In addition to graphene, black phosphorous (BP) also gained a lot of attention for the SWIR region due to its small bandgap and high carrier mobility [6–8]. However, BP is highly unstable at ambient conditions. Additionally, its performance exhibits a tradeoff between responsivity (0.13 A/W - 82 A/W) and bandwidth (3GHz-10 kHz) [9–11]. Thus, it is important to explore photodetectors based on newly emerging stable 2D material which can simultaneously achieve high responsivity and bandwidth.

A recent theoretical study on GaGeTe crystals reported very promising graphene-like electronic properties along with a narrow bandgap electronic structure [12–16]. The crystals belong to the 2D layered materials group of MXTe (M = Al, Ga; X = Si, Ge, Sn), which comprises stacks of monolayers with weak van der Waals interactions among them [13]. Experimental transmittance and reflectance studies on GaGeTe observed broadband peaks at 1.1 eV and 0.56 eV [17]. Another experimental study on GaGeTe reported hole mobility and photoresponsivity of 0.45 cm²/Vs and 3.6 A/W (at 855 nm), respectively [18]. Despite the theoretical prediction, so far, experimental studies were limited to the near-infrared wavelength (at λ = 855 nm) while studies in the SWIR region are limited. In this work, we present a detailed study of multi-layered (ML) GaGeTe photodetectors fabricated on SiO₂/Si substrate. To the best of our knowledge, this is the first demonstration that ML-GaGeTe is capable of SWIR detection with relatively high response speed. Good reproducibility and stability of the measured devices were also observed.

2. Material characterization

The GaGeTe crystal belongs to the R-3m trigonal crystal structure group [18,19]. As depicted in the top and side views of a monolayer (Fig. 1(a)), a single hextuple-layer in the Te–Ga–Ge–Ge–Ga–Te sequence forms a 2D film consisting of tightly bound germanene sandwiched between two-atom thick GaTe fragments [15,19]. Figure 1(b) shows a high-resolution TEM (HRTEM) image of GaGeTe and the corresponding selective area electron diffraction (SAED) pattern. The measured thickness of the single layer is 0.8 nm as obtained from a magnified version of the HRTEM image, as shown in Fig. 1(c). The observed interlayer distance (1.17 nm) is close to the value reported in the literature [19,20]. Figure 1(d) shows the Energy-dispersive X-ray spectroscopy (EDS) spectrum of the GaGeTe crystal which confirms the presence of all three elements and their corresponding elemental mapping indicates the uniform nature of the crystal. As shown in Fig. 1(e), the measured EDS spectrum indicates a stoichiometry of 1:1:1 (Ga:Ge:Te). Figure 1(f) depicts the UV-Vis-NIR absorbance of liquid exfoliated GaGeTe flakes which exhibits a broadband wavelength absorption characteristic from 400 nm to 1800 nm. Recently reported electronic band structure calculations and optical measurements on GaGeTe predicted three optical transitions existing at ∼0.2 (A), ∼1.1 eV (B) and 0.5 eV (C) [13,17]. The bandgap (A) is observed for the out-of-plane component, whereas the latter B is due to the in-plane component. The A bandgap corresponds to interband optical transitions in the region of the T point. The second band (B) opening can be attributed to interband optical transitions near the Γ point. Whereas bandgap (C) is attributed to intraband transitions in the valance band. Based on the aforementioned transitions, the possible optical band transitions were schematically presented in Fig. S1 (see Supplement 1).

 

Fig. 1. (a) Schematic representation of the side and top views of the GaGeTe crystal. (b) Cross-sectional HRTEM image obtained from the prepared GaGeTe lamella and corresponding selected area diffraction pattern (SADP) and (c) a magnified section from the area in the red box, in figure (b) where the measured single layer thickness was 0.8 nm. (d) SEM image of the original bulk GaGeTe crystal and EDS elemental mapping of Te, Ga, and Ge, showing a homogeneous distribution for all. The scale bar is 20μm. (e) Corresponding EDS spectrum obtained from the same region where the HRTEM image was taken. The star indicates the Cu signal originated from the Cu tape. (f) Absorbance at various concentrations of the GaGeTe crystal flakes suspended in IPA solvent measured from 400 nm to 1800 nm.

Download Full Size | PDF

3. Device fabrication

To determine the photodetector’s performance, we have fabricated a standard Si back-gate-based field-effect transistor (FET). A schematic representation of the fabricated few-layered GaGeTe device structure is shown in Fig. 2(a) and an optical image of a typical device is depicted in Fig. 2(b). By using atomic force microscopy (AFM), the fabricated device’s topography and thickness (65 nm) were measured (see Supplement 1, Fig. S2).

 

Fig. 2. (a) Schematic illustrations of a 2D GaGeTe back-gated FET. (b) Optical microscopy image of the fabricated device. (c) I_ds-V_ds curves were obtained for V_gs values sweeping from -30 to 30V indicating a p-type semiconductor and good ohmic contact between the GaGeTe and the deposited metal pads. (d) The transfer curve (log (I_ds) vs V_gs) of a few-layered GaGeTe FET was measured by scanning V_gs from −30 V to 30 V at V_ds = 0.5 V.

Download Full Size | PDF

The output characteristics (I_ds vs V_ds) were measured from -2 V to 2V while applying a gate voltage from -30 to 30 V. Results are depicted in Fig. 2(c). A good linear relationship between the drain current (I_ds) and source-drain voltage (V_ds) confirms an ohmic contact between the electrodes and the GaGeTe flake. In addition, we also measured the I_ds vs V_ds curves of other GaGeTe devices to confirm the ohmic contact nature between the channel and the deposited metal pads (see Supplement 1, Fig. S3(a-b). Furthermore, we also measure the temperature dependence of the I_ds vs V_ds curves to investigate the thermal stability of the devices. The substrate temperature is varied between 25 to 80 °C degrees (see Supplement 1, Fig. S3(c)). The obtained curves exhibited a constant slope over the bias voltage. Figure 2(d) shows the impact of the gate voltage (V_gs) on the transfer characteristics of the GaGeTe device by sweeping it at a fixed V_ds bias of 0.5 V. The results are plotted on a logarithmic scale. It is observed that the drain current decreases when a positive gate voltage is applied while it increases with negative gate bias, indicating an intrinsic p-type carrier characteristic, which could be attributed to intermixing of Ga/Ge [21]. The hole mobility can be extracted from the linear region in the transfer curve from $\mu_h = ({L/W{C_{0x}}{V_{ds}}} )\left( {\frac{{{\varDelta I_{ds}}}}{{{\varDelta V_{gs}}}}} \right),\; $ where ΔI_ds/ΔV_gs is the maximum slope of the linear region of the transfer curve, L and W are the length and width of the channel, V_ds is the source-drain voltage, and C_ox is the capacitance of the dielectric. C_ox = ε₀ε_r/d, ε_0, and ε_r denote the vacuum and relative permittivity of SiO₂, respectively, and d is its thickness (285 nm). The two-terminal hole mobility is calculated to be 8.6 cm² V⁻¹ s⁻¹, this value can be further improved by careful electrode/ channel band engineering [22].

Additionally, the transfer characteristics of the other GaGeTe devices were measured to get further insight into the mobility range (see Supplement 1, Fig. S4 (a-b)). The mobility of the GaGeTe devices is ranging between 8.6 -20 cm² V⁻¹ s⁻¹. It is well known that, due to the inclusion of the contact effect, the standard two-terminal FET configuration tends to underestimate the field-effect mobility compared to the intrinsic value obtained by four-terminal measurements [23]. The measured mobility values in this study were achieved without contact optimization or dielectric engineering (the thickness of SiO₂ is 285 nm). Note that, the two-terminal GaGeTe mobility values are comparable to the other 2D semiconductors [24]. In general, the mobility of two-dimensional (2D) materials is greatly influenced by defects and phonons. The defects can be intrinsic, trap states at the flake/substrate interface, or due ambient absorbent. Whereas the phonon limited mobility is attributed to strong scatterings by the longitudinal optical phonons, out-of-plane acoustic phonons, and the higher density of electronic states [24]. In line with other layered 2D materials, we believe it is possible to enhance the mobility of the GaGeTe by encapsulating it between hexagonal boron nitride (h-BN) layers, dielectric thickness engineering, and the use of top gate FET structures. [25,26].

4. Optical measurements

To investigate the photoresponse of the fabricated GaGeTe photodetector, first, we measured its source-drain bias voltage (V_ds) vs source-drain current (I_ds) under dark and illumination conditions. A 1310 nm laser is used to illuminate the device while sweeping V_ds between -2 V to 2V. The excitation is performed with different incident light intensities (P) ranging from 0.056 mW/cm² to 238.82 mW/cm², as shown in Fig. 3(a). We show the upper set of the V_ds range given the high dark current response that dominates over the overall regime. Figure 3(b) shows the plot between the photocurrent (I_ph = I _light-I _dark) as a function of incident laser power. The measured photocurrents at low optical powers (0.056 and 1.15 mW/cm²) are two orders of magnitude lower than the photocurrents measured at high powers. Hence, it was difficult to distinguish the two photocurrent curves. For better readability, the photocurrent curves at low powers (black and red curves) were replotted separately (see Supplement 1, Fig. S5). The corresponding photocurrents plotted in Fig. 3(c), at V_ds= 2V show, unlike normal bulk photodetectors, a non-linear relation with the incident light power before the absorption is saturated. This dependence can be described with a simple power-law I_ph ∝ P^α, fitted to I_ph ≈ P^0.58 in this case for the data shown in Fig. 3(c). The value of the constant α is close to 1 in common thin-film defect-free photodetectors and between 0 and 1 in most two-dimensional photodetectors [27,28]. This can be attributed to the photocarrier transitions, trap states, Ga/Ge intermixing, defects, and the influence of the ambient atmosphere on the transition of the photocarriers from the channel to the electrodes [21,22,27]. The measured results show that the photocurrent flowing through the device doesn’t increase proportionally with the incident light intensity. We expect that the main mechanism that accounts for the generated photocurrent reported in this work is the photoconductive gain. Generally, this mechanism is reduced by recombination. Under laser excitation, the minority carriers (electrons) are trapped, and the majority carriers (holes) have a faster transit time (τ_t) across the device’s channel which enhances the lifetime of the minority carriers (τ_l). Due to the external applied bias, the majority of the carriers reach the metal contact and exit the device while new ones get injected from the opposite contact to maintain charge neutrality. The injected majority carriers travel in the channel until they recombine with minority carriers. This leads to a photoconductive gain (G) defined as G =$\frac{{{\tau _l}}}{{{\tau _t}}}$ . As mentioned earlier, the mechanism is consistent with the observed photocurrent increases in our measurement as the applied voltage increases, see Fig. 3(b). The higher bias voltages lead to higher drift velocities for the majority carriers, hence the carrier transit time τ_t is reduced [27,29].

 

Fig. 3. Optoelectronic characterization of a multi-layered (ML) GaGeTe photodetector. (a) Typical output curves of a ML-GaGeTe photodetector acquired in the dark and with1310 nm illumination at various excitation powers at V_gs = 0 V. (b) The plot between the photocurrent as a function of incident laser power while applying the bias voltage (V_ds) between -2V to 2V. (c) The measured photocurrent (black solid circles) and responsivity (red solid squares) as a function of incident power intensity at V_ds = 2 V and V_g = 0 V under 1310 nm excitations. The solid black and red lines correspond to the photocurrent and responsivity fitted to a power law, respectively. (d) Photocurrent curves of a few-layered GaGeTe photodetector obtained with 1310,1064, 852 785, and 520 nm illuminations at V_gs = 0 V.

Download Full Size | PDF

Photoresponsivity is one of the important performance parameters of a photodetector. The responsivity is expressed as $R = \frac{{{I_{ph}}}}{{P\ast S}}$, where I_ph is the generated photocurrent, P is incident optical power density and S is detector area, and the calculated R vs incident power is shown in Fig. 3(c). We measured a maximum responsivity of 57 A/W observed under 1310 nm illumination at laser power of 0.49 µW (I_ph = 6 nA, P = 0.056 mW/cm², S = 1850 µm², radius of the incident spot size is 520 µm²) at a bias of V_ds = 2V. The decrease in responsivity with increasing power indicates the presence of traps for minority carriers (electrons) in the device, which is similar to the previously reported results on other 2D-based photodetectors [30,31]. The photoresponsivity mainly depends on the ratio of photocarrier lifetime (τ_l) to the photocarrier transit time (τ_t) [31].

At high incident optical powers, most of the traps are already filled in and further illumination power cannot effectively increase the photogain, which yields a lower responsivity at higher induced laser powers [28]. The measured responsivity also exhibited a nonlinear power-law dependence where R ≈ P^−0.42 was determined by fitting the measured data (see Fig. 3(c)). Additionally, we investigated the spectral sensitivity of our device under different laser illumination for λ = 520 nm (input laser power of 124 mW/cm²), 785 nm (189 mW/cm² mW), 852 nm (163 mW/cm²), 1064 nm (109 mW/cm²) and 1310 nm (147 mW/cm²), by applying a V_ds from 0 to 2V; results are shown in Fig. 3(d). It was observed that the measured photocurrent under 520 nm illumination is 25 μA (124 mW/cm²) which decreases to 2 μA (147 mW/cm²) under 1310 nm illumination. The photocurrent follows the wavelength absorption dependence trend of GaGeTe from visible to SWIR wavelengths (see Fig. 1(f)). Additionally, we explore the dependence of the current profile on the back-gate voltage (V_gs) in the GaGeTe channel. The device performance under dark and 1310 nm laser illumination for gate voltage between -30 V to +30 V at a fixed applied bias of V_ds = 3 V were measured (see Supplement 1, Fig. S6). Under illumination, the device exhibits a p-type semiconductor behavior and shows higher current conductivity under the negative gate voltage.

The time response is also a key factor in determining the photodetector performance. It demonstrates the ability of a device to follow a fast-varying optical signal. The device was illuminated under a 1310 nm wavelength (0.5 Hz) at an applied V_ds bias voltage of 1 V. As shown in Fig. 4(a), the photocurrent increased to 0.6 µA under 1310 nm, when the laser was turned on, and it then sharply decreased to its initial value as the laser was turned off. The rise and fall time of GaGeTe devices under 1310 nm illumination was measured as 0.53s and 0.66 s, respectively (see Supplement 1, Fig. S7(a-b)). After that, the device's temporal response dependence on the wavelengths (633nm and 1550 nm) was also investigated. The measured rise (0.53 s) and fall time (0.67 s) of the device under 1550 nm excitation (0.5 Hz) were similar to the 1310 nm results (see Supplement 1, Fig. S7(c-d)). However, it was observed that, under 633 nm excitation (2 Hz), the device rise time (28 ms) and fall time (83 ms) were improved (see Supplement 1, Fig. S7(e-f)). This could be attributed to the higher absorption or higher penetration depth of the 633 nm laser. It is worth mentioning that these measurements were limited by our Agilent system resolution of 0.1ms. At such a low frequency, the photocurrent generation mechanism is dominated by photoconductivity and traps induced gain mechanism [28]. Under laser excitation, the minority carriers (electrons) are trapped at trap states which enhances the lifetime of the majority carriers (τ_l). The observed slow response time is mainly governed by the longer lifetime of the photogenerated carriers [28]. On the other hand, the trapping states do not respond to high frequencies [10]. This reduces the lifetime of the photogenerated carriers and increases the response speed. To characterize the dynamic performance of the photodetector, the normalized frequency response was also measured (see the measurement setup schematic in Fig. 4(b)). For this purpose, the GaGeTe device was illuminated by a modulated and amplified 1310 nm continuous-wave laser. The laser is modulated by a commercial optical modulator (MOD) driven by a reference signal from a lock-in amplifier. A 9 V bias voltage was applied through the RF probe where the electrical signal from the device was connected to a lock-in amplifier through bias tee (BT) and trans-impedance amplifier. As shown in Fig. 4(b), the response of the photodetector was recorded while sweeping the frequency. The measured 3 dB cutoff frequency is ∼ 0.9 MHz. It is worth noting that the device shows a response up to 100 MHz, which is higher than the early-stage results of photogain based BP photodetector’s (speed ∼ kHz) [5,10,32]. In Supplement 1, Table S1 compares the performance of the GaGeTe device with other 2D-based SWIR photodetectors and commercial InGaAs diodes. Furthermore, we analyzed the influence of the bias voltage on the response speed of the device. Figure 4(c) shows the device frequency response at the different bias voltages. It was observed that the device's 3dB cutoff frequency increases from 0.38 MHz to 0.42 MHz when the applied bias voltage changes from 3V to 7V as shown in Fig. 4(c). This indicates the transit time of the charge carriers was governed by the applied voltage. Next, we investigated the role of applied gate voltage on the response speed of the device (see Supplement 1, Fig. S8(a)). We found that the device's 3dB cutoff frequency is sensitive to the applied back gate voltages. However, the small change in speed with gate voltage was limited by the dielectric thickness, thick flake-induced electrostatic screening, and interface defects between the dielectric and device channel. The frequency response of photodetectors can be further improved by controlling the contact type (Schottky), the thickness of the photodetector, device configuration (vertical device), applied voltage, and spacing of the electrodes [33,34]. Hence, we believe that there is still ample room for improving the speed of GaGeTe based phototransistors.

 

Fig. 4. Transient photoresponse of GaGeTe devices. (a) A test of the photo-switching stability for the GaGeTe device in response to a chain of pulsed illumination at λ = 1310 nm at V_ds = 1 V. (b) Dynamic performance characteristics of the GaGeTe photodetector. The relative radio frequency (RF) signal response is a function of the light intensity modulation frequency. A 3-dB cutoff frequency of 0.9 MHz is observed. The inset shows the measurement setup. (c) The measured device frequency response at different bias voltages. (d) The measured NEP (black circles) and specific detectivity (red solid triangles) as a function of applied bias voltage.

Download Full Size | PDF

Next, we probed the sensitivity of the GaGeTe detectors for SWIR illuminations. The sensitivity of the GaGeTe detector is limited by the total noise (i_n) of the device. The noise in a photodetector arises mainly from the thermal noise (Johnson–Nyquist noise) (i_nJ $= \frac{{4{k_B}T\Delta f}}{{{R_0}}}$), shot noise (i_ns = $2e({{I_d} + {I_{ph}}} )\Delta f$), and the 1/f noise (i_1/f =${k_1}\frac{{{I^b}\Delta f}}{{{f^a}}}$). where k_B is the Boltzmann constant, T is the temperature, $\Delta f$ is the bandwidth, and R₀ is the resistance of the device, k₁, a, and b are coefficients related to the fabrication process, the material properties, and the specific photodetector geometry [29]. In our case, the shot and thermal noises are dominant due to the large dark current and due to the absence of a junction in the channel (metal-semiconductor-metal), respectively. However, at low frequencies, the 1/f-noise dominates which is generally attributed to a large number of charges being trapped and de-trapped [10]. At high frequencies, the noise level is close to the shot noise limit since trap states do not respond to such high frequencies [10]. Due to the high measured dark current, the operation bandwidth, and based on the device configuration, both the shot and thermal noises are the most prevalent noise sources in our devices. For any photodetector, the detection limit is assessed by the noise-equivalent-power (NEP=$\sqrt {total\; noise} /R\; $), i.e. the incident power at which the signal power is equal to the dark current noise. The calculated NEP from measurements at various bias voltages is shown in Fig. 4(d). The NEP is found to be ∼0.15 nW/√Hz at V_ds=2 V, indicating that SWIR radiation in the nano-watt range can be detected. For normal-incident PDs, the specific detectivity D* is also widely used. The latter is defined as ${D^\ast } = \frac{{\sqrt A }}{{NEP}}$ with a unit of Jones (1 Jones = 1 cm Hz^1/2 W^-1), where A is the detector illumination area. The measured ${D^\ast }$ over the applied bias voltage V_ds was also shown in Fig. 4(d).

Finally, we tested the stability of the fabricated GaGeTe devices. The devices exhibit good stability even after 90 days of fabrication. The electrical and optical measurements were repeated on the same devices. The obtained results were similar to earlier measurements (see Supplement 1, Fig. S9(a-b)). As shown in Fig. S9(a), the measurements after 90 days indicate that the photocurrent is reduced from 60 nA to 40 nA under 0.056mW/cm² illumination, while it was increased under 1.15mW/cm² illumination. The 3dB cutoff frequency was also changed from 0.9 MHz to 0.46 MHz, as shown in Fig. S9(b). We believe that the small variation observed in the performance can be partially attributed to the damage in the contact metal pads during repetitive probing. The optical images of the devices are shown in the inset of Fig. S9(b). The top and bottom insets represent the device on day 1 and day 90, respectively. It is worth mentioning that the devices are kept at ambient conditions. Furthermore, the performed photoswitching stability test (> 1 week) exhibits the good stability nature of the material (see Supplement 1, Fig. S10(a-b)). Moreover, the top surface of the GaGeTe retains its intrinsic properties where the color contrast of the fabricated device channel is similar after 90 days, which rules out the oxidation of the top surface.

5. Conclusion

In summary, we have fabricated and investigated the optoelectronic properties of a photodetector based on mechanically exfoliated GaGeTe. We demonstrate stable devices at room ambient conditions with ultra-broadband wavelength photoresponse from 520 nm to 1550 nm with good reproducibility. Our devices exhibited a responsivity of 57 A/W at 2V under 1310 nm wavelength illumination. Further, we showed that GaGeTe has the potential to operate as a fast photodetector with a 3dB cutoff frequency of 0.9 MHz. Further, the influence of the bias and back gate voltages on the response speed of the device was also investigated. We believe the reported operating bandwidth can be enhanced further by redesigning the device and improved contacts properties.

6. Experimental method

6.1 Material characterization

The GaGeTe crystal flakes were suspended in isopropyl alcohol (IPA) using sonication and spectroscopy measurements were carried using an Ultraviolet-Visible-Near-Infrared Spectrophotometer (Agilent Cary 5000). For TEM analysis, a thin GaGeTe lamella was prepared using a standard in situ lift-out procedure by using a dual-beam FIB system (Thermo Fisher Scientific, Helios 650). Following that, the Pt capping layer was deposited on the top surface of GaGeTe, using the precursor gas (CH₃)₃Pt(CpCH₃)) in the dual-beam system, to protect from the ion beam induce damage. The prepared lamella was used for cross-sectional views of the crystal. The HRTEM images were obtained by an image corrected TEM system (Thermo Fisher Scientific, Titan G2) operating at 300 kV.

6.2 Device fabrication

A Si substrate with 285 nm thick SiO₂ is cleaned by ultrasonication in acetone with subsequent rinsing in isopropanol and deionized water. Then, scotch tape is used to exfoliate bulk GaGeTe (bulk crystals, 2Dsemiconductors) where the flakes are transferred on top of the Si/SiO₂ substrate. An optical microscope is used to identify suitable flakes for the device fabrication. The samples are then spin-coated with a photoresist (AZ5214 E, 4000 rpm for the 40s) and pre-baked at 115°C for 3 minutes. To fabricate a multilayer GaGeTe photodetector, standard photolithography (SUSS MicroTec) and electron-beam evaporation (Kurt J. Lesker PVD) were used to define and deposit the source-drain electrodes (Cr (15 nm)/Au (70 nm)), respectively. Where the channel length (L) is 23 μm and the channel width (W) is 12 µm. The heavily-doped Si acts as the bottom gate.

6.3 Electrical and photocurrent measurement

A probe station with three micromanipulators was used to apply bias and gate voltages. A curve tracer/power device analyzer / (Agilent B1505A) was used to control the biases and measure the channel current. The incident laser sources are continuous wave laser diodes, the visible wavelength lasers were acquired using a 4-channel fiber-coupled laser source (MCLS1 Thorlabs) set at different drive currents. The incident light output was guided to the top of the sample surface with a single-mode pigtailed GRIN fiber collimator (Thorlabs). The temporal photoresponse was measured using an on/off cycle of 2 seconds.