高比探测率和高速石墨烯/n-GaAs复合结构的光电探测器

田慧军，刘巧莉，岳 恒，胡安琪 ，郭 霞

（1. 北京工业大学 材料与制造学部，北京 100124；2. 北京邮电大学 电子工程学院，北京 100876）

1 Introduction

Graphene with atomic layer thickness has attracted worldwide attention because of its unique electronic, optoelectronic, mechanical properties[1,2].Recently, hybrid graphene/semiconductor phototransistors have been increasingly investigated for optoelectronic applications because of their ultrahigh responsivity[3-5]. In such structures, photogenerated carriers separate at the interface of the graphene/semiconductor junction. Because of the photogating effect, the responsivity obtained via such photoconductivity-based devices, whose photocurrent was measured from source-drain electrodes, can be as high as 1010A/W at 0.19V[6]. The corresponding specific detectivity (D*), which characterizes the sensitivity of a photodetector, has been reported to be as high as 1.4×1012Jones[6]. However,based on a detailed comparison of the magnitude of three types of noise[6,7], it can be found that the D*of such phototransistors is mainly limited by 1/f noise[6], not the thermal noise or the shot noise,which play dominant roles in conventional photovoltaic-type semiconductor photodiodes[7-8]. The 1/f noise in this system mainly originates from the carrier trapping and detrapping processes at the graphene/semiconductor interface[9]. In other words,the D*of photoconductivity-based devices ranges from ～108Jones to ～1014Jones, which brings great enhancement in responsivity[3-6]. A high density of trapping centers at the interface also results in severe persistent photoconduction, and a relatively long transient response time. Generally, the reported transient response time of such hybrid phototransistors is in the order of seconds or even milliseconds[5,6]. These photodetectors need to work under external bias voltage, whose applications are usually limited because of the large 1/f noise. To overcome these limitations, new types of photodetectors are required[10].

For the direction perpendicular to the graphene/semiconductor junction, whose photocurrent is measured by source-gate electrodes, interface traps play only a small role in the entire current transport loop. The source of 1/f noise is lessened and the persistent photoconductive effect can be relieved,which induces a high D*and high response speed[11].These photodetectors with low power consumption,low cost and high sensitivity can satisfy practical applications’ requirements such as in visible light imaging and large-scale integrated circuitry[12,13]. For photovoltaic-type devices, the D*ranges from ～1010Jones to ～1013Jones, which originates from there being lower noise under zero voltage[10]. Moreover,the reported transient response time of photovoltaictype devices is in the order of microseconds[11]. In this paper, the photocurrent measured from the source-gate electrodes of a hybrid graphene/n-GaAs photodiode demonstrated improved D*and response speed. This work is promising for developing low power consumption visible-light photodetectors with high sensitivity that are ideal in applications like imaging.

2 Materials and methods

Si doped with a GaAs substrate with an electron concentration of ～1×1018cm−3was selected for this study. Au/Ge/Ni/Au (10/10/5/100nm) electrodes were deposited on the back of the n-GaAs substrate and then treated via rapid thermal annealing at 430°C for 35 s in an N2environment to form ohmic contact[14]. The graphene was synthesized via chemical vapor deposition on Cu foil and then transferred to the front of the GaAs substrate using polymethylmethacrylate (PMMA) as the supporting film. Thus, a hybrid graphene/n-GaAs photodiode was obtained after metal deposition. The Raman spectrum was tested using a Renishaw Invia Raman microscope with a 514-nm laser source. Electrical measurements were carried out using a Keithley 4 200. Illumination was provided by a laser at 650nm with a series of neutral density filters under ambient conditions. A Si photodiode (Hamamatsu S2387)was used to calibrate the power.

3 Results and discussion

Figure 1(a) demonstrates the schematic diagram of the graphene/n-GaAs photodiode studied in this work. The optical and electrical performances were measured from source-gate electrodes at room temperature. Figure 1(b) illustrates the energy band diagram of the graphene/n-GaAs heterojunction, where a Schottky barrier forms at the interface. Because of the different work functions between graphene(～4.8 eV) and n-GaAs (～4.1 eV)[15-16], the energy band bends upward and a built-in electric field is formed at the interface between graphene and n-GaAs with a barrier height of ～0.7eV. The photoexcited electron-hole pairs separate under the function of the built-in field when light illuminates the sample. Due to the high surface state density of the GaAs material, the photoexcited holes fill up its surface states first because they accumulate near the interface, as illustrated in Figure 1(b). The Raman spectrum of the graphene is as shown in Figure 1(c).The G and 2D bands are located at 1 582.9 cm−1and 2683.8 cm−1, respectively. The absence of a D band indicates that there are few defects[17]. The intensity ratio of the 2D band to the G band is above 3, which causes monolayer characteristics of graphene[18].Figure 1(d) shows the spectral response of the device. Under a zero bias voltage, the spectral responsivity of the photodiode was shown for an incident light power of ～200nW over the range of 300～1 100nm. The responsivity exceeded 75 mA/W in the visible light region (380～760nm). The cutoff wavelength was around 870-nm which corresponded to the GaAs edge of the energy band.

Fig. 1 (a) Schematic diagram of the graphene/n-GaAs photodetector. The optical and electrical performances were measured by source-gate electrodes. (b) Energy band diagram of the graphene/n-GaAs heterojunction with the Schottky barrier height ΦB of the graphene/GaAs junction of ～0.7eV. The interface states are depicted at the interface, illustrating that the carrier trapped photons at the graphene/GaAs interface during carrier transport through the junction. (c) Measurement result of the Raman spectrum of graphene on the GaAs substrate. (d) Spectral response of the photodiode under zero bias voltage.

Figure 2(a) (Color online) shows the currentvoltage (I-V) measurement results of the photodiodes illuminated under a 650-nm laser with different light intensities. All the curves demonstrate good rectification behavior due to the heterojunction formed between graphene and n-GaAs[19]. According to thermionic emission theory, current-voltage curves are expressed as below:

whereqis the charge,Vis the bias voltage,kis the Boltzmann constant,Tis the temperature,nis the ideality factor,I0is the reverse saturation current,Ais the photosensitive area of the photodiode,A* is the effective Richardson constant of n-GaAs(12.1 A·K−2cm−2)[18], andΦBis the barrier height of the Schottky junction[20-21]. According to the fitting results,ΦBandnwere respectively extracted to be 0.65 eV and 1.89 for the dark curve, which is almost consistent with the above theoretically determined value. The extracted data ofn, which is much larger than 1, indicates that the recombination process with the assistance of interfacial states dominates the carrier transport process at the graphene/GaAs interface.

Fig. 2(a) Current versus voltage curves of the device under different light powers. (b) The relationship between photocurrent and photovoltage (Voc) with the incident light’s power in the self-driven mode. (c) Responsivity and D* versus illumination power under a zero bias voltage. (d) Illustration of the carrier trapping and detrapping processes at the interface of the graphene/GaAs interface, which is the main source of 1/f noise.

The photocurrent (Iphoto) increases linearly from 49.7nA to 2.93μA as the illumination power increased from 492nW to 136μW at a zero bias voltage, as shown in Figure 2(b). The photovoltage(Voc) increased from 0.12to 0.275V when the power increased from 492nW to 136μW.Figure 2(c) shows the responsivity versus the power of the incident light of the device. The responsivity(R) of the photodiode is defined by the equation

wherePinis the incident light power. Under a zero bias voltage, the highest responsivity was 95 mA/W.It can be seen that the responsivity of the photodi-ode slightly decreased with an increase in power.This can be attributed to a reduction in the built-in field when there is an increasing number of photogenerated carriers, which forms an electric field that opposes the built-in field. When light illuminates the graphene/n-GaAs photodetector, GaAs absorbs the photons with energy larger than its bandgap and induces electron-hole pairs, which are separated by the function of the built-in field. Photo-induced holes move towards graphene, while photo-induced electrons move towards GaAs. The photo-induced holes fill up the surface states first because of the high surface state density of the GaAs material.Then the other photo-induced holes can be driven into graphene and collected by the electrode to generate the photoresponse.

D*is one of the important parameters for a photodetector, which is defined as

where A is the area (0.01 cm2) of the device, B is the electrical bandwidth, and PNis the noise’s equivalent power. PNis expressed as

where R is the responsivity and SIis the meansquare of the noise’s current in the dark. According to the theory of noise, SIis the total sum of the 1/f noise ( SI(1/f)), the shot noise ( SI(shot)), and the thermal noise ( SI(thermal))[22], which can be calculated using

At a modulation frequency of 1 Hz, the SI(1/f)is ～10−35A2·Hz−1, which can be represented by the following equation:

where I(f) is the discrete Fourier transform of the d ark current waveform I(t), FSis the sampling rate,and N is the number of data points. The SI(shot) is calculated to be ～1.04×10−27A2·Hz−1by using

where q is the elemental charge and Idis the dark current of the device. SI(thermal) is calculated to be～1.66×10−27A2·Hz−1at room temperature by using Nyquist’s equation, represented as

where k is the Boltzmann constant, T is the temperature, and RSis the differential resistance of the device in the dark. As shown in Figure 2(c), D*at 0V was observed to decrease with increasing light power because of the reduction of responsivity. The maximum D*of the device was 1.82×1011Jones,which is 562times higher than the source-drain signal of a graphene/GaAs phototransistor that was fabricated in the same experiment conditions[11]. The obvious improvement of D*is attributed to the screen effect from the Schottky barrier. Comparing the two types of carrier transport processes, as illustrated in Figure 2(d) and 1(b), the 1/f noise dominates the mean square noise SIfor the source-drain photocurrent because of the trapping and detrapping processes at the interface during carrier transport, which is related to the carrier number and mobility fluctuation[9]. The 1/f noise of the source-gate measurement, as illustrated in Figure 1(b), was cut off because the interfacial states were filled by photoexcited holes first. Only then were the other photoexcited holes allowed the transport electrodes. The trapping barrier can be lowered only at elevated temperatures and bias. Because the source of noise was changed from well-known 1/f noise to thermal noise, D*improved greatly.

Figure 3 shows the transient response measurement of the device under a laser power of 136μW at 0V at room temperature[23]. It should be noted that there is a peak when the light turns on, which can be repeated. We attributed it to the pyroelectricity effect caused by the electrical response to a sudden tiny temperature change when the light was turned on[24]. The rise time (τr) and decay time (τf) were measured to be 4 ms and 37ms, respectively. The relatively long decay time after switching off the light indicates that it took more time for carriers to be transported from the GaAs semiconductor to the depletion region via diffusion and then be injected into graphene and recombined. The fast response was attributed to the quick separation of photogenerated carriers by the built-in electric field at the interface. Compared with the results measured by source-drain electrodes, which respectively had a rise and decay time of 270ms and 28.5 s[11], the transit response performance also improved by ～2 orders of magnitude.

Fig. 3 The response time of the photodiode at a zero bias voltage under a laser power of 136μW where τr is～4 ms and τf is ～37ms.

4 Conclusion

In this work, the optoelectronic performance of the graphene/n-GaAs structure measured by sourcegate electrodes was described. Due to the trapping and detrapping processes, there was a maximum specific detectivity of 1.82×1011Jones with a rise time of 4 ms and a decay time of 37ms, achieved at 0V at room temperature. We attribute the ～2orders of magnitude improvement in specific detectivity and its corresponding response time to the screening of interfacial states given by the Schottky barrier.