Carrier density modulation and photocarrier transportation of graphene/InSb heterojunction middle-wavelength infrared photodetectors

Abstract. The photoresponse mechanism of graphene/InSb heterojunction middle-wavelength infrared (MWIR) photodetectors was investigated. The devices comprised a graphene/InSb heterojunction as a carrier-injection region and an insulator region of graphene on tetraethyl orthosilicate (TEOS) for photogating. The MWIR photoresponse was significantly amplified with an increase in the graphene/TEOS cross-sectional area by covering the entire detector with graphene. The graphene-channel dependence of the MWIR photoresponse indicated that the graphene carrier density was modulated by photocarrier accumulation at the TEOS/InSb boundary, resulting in photogating. The dark current of the devices was suppressed by a decrease in the graphene/InSb carrier-injection region and the formation of the heterojunction using an n-type InSb substrate. The results indicate that photocarrier transportation was dominated by the formation of a Schottky barrier at the interface of the graphene/InSb heterojunction and a Fermi-level shift under bias application. The high-responsivity and low-dark-current photoresponse mechanism was attributed to the graphene/InSb heterojunction diode behavior and the photogating effect. The devices combining the aforementioned features had a noise equivalent power of 0.43  pW  /  Hz1/2. The results obtained in our study will contribute to the development of high-performance graphene-based IR image sensors.


Introduction
Graphene-based infrared (IR) photodetectors are promising devices that take advantage of the unique optoelectronic properties of graphene such as broadband light absorption, 1-3 high carrier mobility, 4 high thermal conductivity, 5-7 gate-tunable plasmons, 8,9 and strong nonlinear optical response, [10][11][12] as well as its excellent chemical stability. Graphene exhibits broadband light absorption over wavelengths ranging from the ultraviolet to the terahertz region and a fast photoresponse that provides a GHz-order bandwidth. 13,14 In addition, graphene can be fabricated at a low cost via nontoxic processing, which is more advantageous than quantum-type IR detector materials. However, conventional graphene field-effect transistors (FETs) have drawbacks for IR photodetector applications in which high-photoresponsivity and low-noise characteristics are required to be consistent. The responsivity of graphene devices suffers because a single layer of the graphene absorbs only 2.3% of light. 15 In addition, the dark current of conventional graphene FETs is extremely high because of the intrinsic nature of graphene with a zero-bandgap structure. 4,16 Meanwhile, the dark current should be suppressed to improve signal-to-noise performance. To develop high-performance graphene IR photodetectors, these deficiencies should be improved.
Here, we report on a detailed mechanism of the devices and design principle of structures to improve the responsivity and dark-current characteristics in parallel. InSb, which is well known photomaterial for middle-wavelength IR (MWIR) detection, was applied. The responsivity enhancement was assessed by comparing devices entirely and partially covered with graphene. The graphene-channel dependence of the MWIR photoresponse enhancement was investigated to assess the photogating. A low-dark-current bias region was compared between devices with distinct graphene/InSb heterojunction areas and dopants of InSb substrate. The MWIR photoresponse performance of the devices that combined the features above was evaluated. Figure 1 shows a schematic of the graphene/InSb heterojunction photodetector. The devices consist of a graphene and p/n-doped InSb heterojunction and a graphene/tetraethyl orthosilicate (TEOS) region. A 600-μm-thick InSb substrate with a 100-nm-thick TEOS insulator layer was prepared. The carrier concentrations of the p/n-doped substrate are 2 × 10 14−15 and 1 to 3 × 10 15 cm 3 , respectively. The drain electrode consisted of 10-nm-thick Cr and 50-nm-thick Au layers and was sputtered on the TEOS layer. The TEOS layer on the center region of the devices was etched using buffered hydrogen fluoride to form a graphene/InSb region. Graphene was fabricated by chemical vapor deposition and was transferred onto the surface of the devices using a conventional graphene-transfer method. 59,60 The graphene channel was formed through a conventional photolithography process and oxygen plasma etching. To assess the graphene cross-sectional area dependence of the MWIR photoresponse enhancement, the devices were entirely or partially covered with graphene, as shown in Figs. 1(b) and 1(c). Figure 2 shows the Raman spectrum obtained for the graphene channel using a 512 nm excitation laser. The spectrum has the typical characteristics of graphene, including a G peak at 1580 cm −1 and a D peak at 2700 cm −1 , which corresponds to the bond stretching and secondorder breathing modes of sp 2 carbon atoms, respectively. The spectrum indicates that a monolayer graphene channel was successfully formed. 61,62 The photoresponse characteristics were investigated by current measurement. The devices were set in a vacuum probe chamber with a cooler at 10 −3 Pa and 77 K. The backside of the substrate was electrically grounded. Current measurements and voltage application were conducted using a device analyzer (B1500A, Keysight). A quantum cascade laser (QD4550CM1, Thorlabs) with a wavelength of 4.6 μm was used as the light source.  Figure 3(b) shows the current-bias characteristics of the devices under various laser powers. The photocurrent showed a linear increase with variation of the laser light power and V d application. The maximum MWIR light responsivity of the devices was calculated to be 4.68 A∕W at a V d of 0.5 V. We also confirmed that the devices exhibited photoresponses under visible, near-IR, and MWIR around 3 to 5 μm.

Graphene Channel Dependence of Photogating
The MWIR photoresponse of the devices was investigated, and Fig. 4 shows a comparison of the photocurrent characteristics of entirely and partially graphene-covered structures. The graphene/ InSb contact region was 50 × 50 μm 2 in both devices. The entirely graphene-covered device indicated a significant increase in the photocurrent, where the maximum photocurrent at a V d of 0.5 V reached 67.63 AE 4.03 μA, whereas that of the partially graphene-covered device exhibited a maximum photocurrent of 9.44 AE 0.23 μA.
The photoresponse of the devices is mainly affected by the graphene/TEOS region because the photogating occurs in this region. Under positive V d application, photogenerated electrons and holes are separated in the p-InSb photosensitizer, and electrons in the vicinity of the graphene/InSb interface are injected into the graphene. On the other hand, photoelectrons excited in the InSb under the graphene/TEOS contact region accumulate at the TEOS/InSb interface owing   2 Raman spectrum of graphene channel in the TEOS insulator layer of the devices in Fig. 1(b). Fukushima  to the depletion layer formed by V d application, which changes the graphene's surface carrier density. As a result, the photocurrent of the devices can be amplified. This phenomenon is referred to as photogating. The remaining TEOS region can provide additional photogating by the photocarriers generated in the TEOS/InSb depletion layer, although it may be less effective for photoresponse enhancement because the carrier diffusion distance of the InSb is short, at around a few micrometers, 41 and the effective photocarrier region is situated only in the vicinity of the graphene/TEOS region. The TEOS layer may also decrease the MWIR incident light power that penetrates to the InSb substrate. TEOS-SiO 2 has an absorption peak around 9 to 10 μm; therefore, the effect of this decrease is negligible. The results show that photogating in the graphene/TEOS region plays a dominant role in the MWIR photoresponse.
To clarify the influence of the graphene/TEOS region on the responsivity enhancement, the MWIR photoresponse characteristics obtained with different graphene/TEOS cross-sectional areas were compared. Figure 5(a) shows a schematic diagram of a device structure. In the region with graphene, S CH , L, and W are defined as the graphene/TEOS region, length, and width, respectively. As shown in Fig. 5(b), the photoresponse increased with an increase in S CH . Next, the effects of the graphene channel length L with a fixed W of 100 μm and W with a fixed L of 200 μm were investigated. Figures 5(c) and 5(d) show that the photocurrent increased as L and W increased.
The photoresponse in graphene FETs with photogating is dependent on the graphene-channel aspect ratio because the devices operate under the same principles as metal-oxide-semiconductor  FETs. 42,44 By contrast, the devices with graphene/InSb heterojunction structures exhibit greater photocurrent when the graphene-channel W, L, and S CH increase. An entirely covered graphene channel enables the injected photocarriers in the graphene/InSb heterojunction region to spread to the drain electrode in all directions. This increases the effective region of the photogating, which multiplies the photocurrent. In addition, the photocurrent is nonlinearly increased with W. Although an increase in L increases only the graphene/TEOS region for the photogating effect, an increase in W increases both the graphene/TEOS region and graphene/InSb heterojunction region. The heterojunction region size does not affect the graphene/InSb Schottky barrier height but changes the flow rate of the photocarriers between graphene and InSb. Both the increase in photocarriers injected into graphene and the improvement of the responsivity due to photogating cause a nonlinear increase.

Dark-Current Reduction
Next, the dark-current dependence on the device structures was investigated. Figures 6(a) and 6(b) show the V d -dependent dark-current characteristics with different graphene/InSb heterojunction cross-sectional areas and dopant types of the InSb substrate. The devices were fabricated to be entirely covered with a graphene channel, with S CH and the graphene/n-InSb contact region (S CI ) set as 50 × 50 μm 2 , 100 × 100 μm 2 , and 200 × 200 μm 2 , respectively, as shown in Fig. 6(c). The maximum dark current decreased with decreasing S CI , and the bias region of low dark current within 5 μA expanded from 62 to 73 and 85 mV in graphene/p-InSb devices and from 54 to 708 and 711 mV in graphene/n-InSb devices according to decreasing S CI . These results indicate that S CI significantly affects the photoresponse and operating characteristics of both dopant types.
The dark-current behavior was investigated using a band model of a graphene/InSb heterojunction, as shown in Figs. 6(d) and 6(e). A reverse bias decreases the Schottky barrier height (Φ B ), which corresponds to the work function difference between the Fermi level of graphene and the valence band of n-InSb, and causes a leak current for both graphene/p-InSb and graphene/n-InSb structures. The work function of InSb is around 4.77 eV 63 and is very close to that of the graphene at around 4.5 to 5.0 eV. [64][65][66] Since the work function of graphene is strongly affected by the surface condition 45,67 in which the adhesion of moisture in the atmosphere or residue causes hole doping, 68 the Fermi level E F of graphene is lowered from the neutral point. This lower shift of the Fermi level in graphene increases Φ B and suppresses the dark-current leak at the graphene/n-InSb heterojunction.

Performance Evaluation of Improved Structured Device
The MWIR photoresponse performance of the devices that combines the aforementioned features was evaluated. The devices were fabricated as fully covered with a graphene channel of 20 × 20 μm 2 as the MWIR light irradiation area, and the graphene/n-InSb contact region S CI was decreased to 7 × 7 μm 2 , as shown in Fig. 7(a). Figure 7 Figure 7(c) shows the MWIR performance of the devices with the dark current of −0.96 nA AE 0.58 pA and photocurrent of −8.67 AE 0.08 nA at a V d of −1 mV, which corresponds to a noise equivalent power (NEP) of 0.43 pW∕Hz 1∕2 . The performance of the graphene photodetectors is more dependent on the device fabrication process and the graphene synthesis method than on the device structure. The NEP value was significantly improved from 94.5 nW∕Hz 1∕2 in our previous work, 17 where the graphene and other device elements were prepared in the same manner.
The devices exhibited a high responsivity that exceeded 100% of the external quantum efficiency under a large bias application. The responsivity reached 6.14 A∕W at V d at −0.2 V and 12.6 A∕W at V d at 0.1 V. Figures 7(d) and 7(e) show band models of the graphene/n-InSb heterojunction under reverse and forward bias. Under reverse-bias negative V d application, the Fermi level is upshifted in the graphene. Φ B is further decreased with the upshift of the Fermi level by photogenerated electrons in the graphene from E F to E Fphoto . The photogenerated electrons can obtain enough energy to generate an avalanche multiplication process at the depletion layer formed in the graphene/n-InSb heterojunction. Since the carrier transportation has not alienated the negative V d region, the dark current is also increased with the photocurrent. By contrast, a positive V d application and photogenerated holes decrease Φ B , and the photogenerated electrons in the InSb and the holes in graphene are transported between each other. A further increase of positive V d promotes the recombination of the photogenerated electrons and injected holes in n-InSb, and the photocurrent decreases. The results obtained here indicate that the high performance of the devices stands on the formation of Schottky barriers at the interface of the graphene/n-InSb heterojunction and photocarrier transportation.

Conclusion
We investigated graphene/InSb heterojunction MWIR photodetectors that can achieve high responsivity and low-dark-current characteristics. The MWIR photoresponse of the devices indicated that the injected photocarriers from the substrate to graphene were amplified by the photogating induced in the graphene/TEOS region. The photocurrent characteristics for various graphene channel sizes indicated that the photocurrent changes linearly with the cross-sectional area of the graphene/TEOS region, in which the photogating contributes large photocurrent modulation. In addition, the graphene shape entirely covers the device area, thus enhancing the photogating. It was also demonstrated that the graphene-InSb substrate heterojunction region has a significant influence on the dark-current performance, and the dark current was suppressed by a decrease in the graphene/InSb carrier-injection region and formation of the heterojunction using an n-type InSb substrate. The voltage-dependent current characteristics indicated that photocarrier transportation was dominated by the formation of a Schottky barrier at the graphene/ InSb heterojunction and a Fermi-level shift of the graphene under bias application. The devices that combined these features exhibited NEP of 0.43 pW∕Hz 1∕2 . The results obtained in this study will contribute to the development of high-performance graphene-based IR image sensors.