Investigating the effect of number of metal electrodes on performance parameters of AlGaN MSM photodetectors

AlGaN MSM photodetectors (PDs) are good options for quick optical detection, wide bandwidth and low-noise-based communication applications. The major advantage of the MSM detector over other detector types is that it utilises metal electrodes that serve as Schottky contacts; thus, the difficulty of realising ohmic contacts is avoided here. Moreover, due to lateral, planar structure, these are suitable for the monolithic integration with other electronic components (Monroy et al., 1999). The band gap energy of Al_xGa_1-xN alloy can be varied from 3.43 eV (GaN) to 6.2 eV (AlN) by changing Al mole fraction. Hence it is possible to alter the absorption wavelength from 360 nm to 200 nm (Reverchon et al., 2004), thus enabling visible-blind and solar-blind devices. Good quality Schottky contacts for MSM detectors with high Al content require appropriate metal–semiconductor interface parameters (Reverchon et al., 2004). Recently efficient detection of UV light has become progressively more important in several areas, such as optical communication, UV emitter calibration, combustion monitoring, flame detection, astronomy, chemical analysis and a wide range of ground and space-based switching applications. GaN or AlGaN based semiconductors have attracted much attention for such applications due to high saturation velocity, high breakdown field and high-temperature tolerance (Miller et al., 2018).

Most of the researchers have studied fixed area MSM PDs with equal (S = W) electrode dimensions for performance analysis. It had been reported that larger electrode widths and spacings result in lower responsivity and longer response time (Zhao and Donaldson, 2018a, 2018b; Zhao, 2018). As per past studies, front-illuminated MSM devices result in Quantum Efficiency (QE) loss due to the shadowing effect. Since responsivity depends on the device fill factor (ratio of illuminated area/total area) of the detector. Prior research studies based on fixed area AlN MSM detector with circular electrodes inspected that fill factor can be increased by decreasing the number of fingers. Therefore, if finger width is kept constant then the numbers of fingers can be decreased by increasing the spacing between fingers (Barkad et al., 2010). Schottky Barrier Height (SBH) for metal contacts has significant effect on MSM performance. It can be enhanced by increasing Al composition in Al_xGa_1-xN alloy because the value of electron affinity decreases for large band gap (Sang et al., 2013). In addition the quality of the AlGaN layer could be improved with the increase of Al mole fraction from 0 to 50% together with the use of AlN nucleation layer between the active layer and Sapphire substrate. As low defect density of the active layer results in high responsivity (Mello et al., 2008). Various research groups reported that photoconductive gain is directly related to the electric field between electrodes. The cause of gain in AlGaN MSM PDs is reported as due to threading dislocations and hole trapping regions. Besides this, when holes at the trap regions accumulate, electron injection at the cathode terminal increases (Ozbay et al., 2004; Gökkavas et al., 2007; Xie et al., 2019). As per prior research studies, it was remarked that dark current and sensitivity of MSM detectors significantly affected by finger gap/finger width ratio. It was also suggested that by increasing the finger gap, capacitance of detector could be lowered at the cost of transit-time bandwidth (McAdoo, 2000). For Al_0.25Ga_0.75N (3 μm thick) based MSM detector with a fixed number of fingers, it was remarked in Li et al. (2010) that low dark current and high photocurrent could improve the detectivity and Signal to Noise (S/N) ratio for microwave applications. Other research group illustrated that ultra-low dark current could be obtained for large area AlGaN/AlN/sapphire-based MSM PD with high SBH (Xie et al., 2012). Earlier investigations showed that thermally annealed transparent electrodes could provide high responsivity than metal contacts. However, these electrodes might result in smaller 3-dB bandwidth and a slower operation due to photoconductive gain (Su et al., 2002). Simple design rules and optimum electrode dimensions had already been proposed for high-speed operation based fixed area MSM detector. It was reported that the capacitance of the contact system could be decreased by decreasing the number of fingers (Averine et al., 2001). Studies indicate that maximum QE of PD can be related to material selection and absorption layer thickness. However, if the width of the electrodes is small at constant spacing higher spectral response can be obtained. Otherwise, increasing electrode width may result in the degradation of noise characteristics due to the enhancement of dark current (Alhelfi et al., 2016). For GaN-based MSM PDs, it had been demonstrated that a superlinear enhancement in the photo-responsivity could be obtained because when finger pitch and width are reduced, linearity and UV/visible ratio got improved (Palacios et al., 2002). Moreover, at low bias, front-illuminated MSM PDs provide greater photo-responsivity but high dislocation density leads to significant recombination loss. In addition, light absorption efficiency is considered the dominant factor determining the QE of MSM PD (Wang et al., 2013), necessitating optimising its geometry. Some studies demonstrated that utilising a higher number of electrode pairs or processing higher electrode pair density on thick absorber layers, a significant increase in external QE for top-illuminated PDs can be achieved (Brendel et al., 2017). MSM detectors generally exhibited higher dark current and nonlinear photocurrent due to dislocation defects and crystallographic imperfections (Reverchon et al., 2004; Schühle and Hochedez, 2013). It was also revealed that the detector's response time could be determined with carrier mobility and electrical field in the finger spacing. Hence decrease in finger spacing could increase dark current and device capacitance while decreasing the carrier transit time (Zhao, 2018). Various groups had analysed high responsivity and high bandwidth for MSM PDs by utilising different opacity materials for cathode and anode terminals (Padhy, 2015). It had been demonstrated that the superlattices based detectors lower dark current as they can provide high internal gain for low optical power and rapid time response for high optical power (Xie et al., 2019).

Various research reports have demonstrated the effect of equal but varying finger dimensions on responsivity/photocurrent and speed of operation in Palacios et al. (2002), Ozbay et al. (2004) and Zhao and Donaldson (2018b). Equal width and spacing dimensions for electrodes have been utilised for the majority of MSM structures so as to comply with the geometry optimisation rule (that is S W) and to get lower transit time of photogenerated carriers (Li et al., 2010; Wang et al., 2013). Moreover, the performance of fixed area detectors using circular interdigitated electrode design has also been investigated with varying spacing (at fixed width) via changing the number of fingers. However, responsivity was reported to decrease with the fill factor because of fixed width (Barkad et al., 2010). Variable area detectors with varying width and spacing have been studied by researchers in Gökkavas et al. (2007) and Zebentout et al. (2011). Therefore, as per the literature survey, there are limited reports on comparative performance analysis between two (S = W and S = 2W) strategies based AlGaN/AlN/Sapphire MSM design types.

In this paper, the numbers of metal electrode fingers have been varied for both W = S and S = 2W based MSM detectors. The influence of electrode dimensions and number of fingers has been investigated for photocurrent, dark current density, light current density and transient response. Comparative analysis has been carried out to select an appropriate fingers based structure as per required performance parameters for particular application.

Section II details material and methods. Section III presents the simulation approach for performing investigation. Section IV presents the discussion of simulation results. Section V concludes the study.

AlGaN MSM PDs are extensively studied because of their usefulness in low noise and high-speed communication applications (Li et al., 2010). In the present work, 1 μm thickness of Al_0.5Ga_0.5N layer has been utilised for all MSM structures considered. As doping concentration is the critical parameter for the light-absorbing layer and if doped uniformly, then the electron concentration may reach 10¹⁸ cm⁻³ (Li et al., 2018). Silicon is preferred as the n-type dopants for III-Nitrides (Tsao et al., 2018). Therefore in the present simulation work doping level of 10¹⁸ cm⁻³ has been considered for the Al_0.5Ga_0.5N layer.

Various research studies reported that the nucleation layer must be suitably selected for providing good optical and electrical qualities to the active layer. As studied, the AlN buffer layer could provide an excellent solution to get good optical properties (Reverchon et al., 2004). Therefore, AlN buffer layer of thickness 500 nm has been utilised between the active layer and Sapphire substrate. The structural design Al_0.5Ga_0.5N/AlN/Sapphire MSM detector with two fingers is shown in Figure 1. The horizontal and vertical dimensions of the MSM structure are considered as 40 μm.

Figure 1

The absorption coefficient and thickness of the photo-absorbing layer play an essential role in the performance of MSM PDs. Since the carriers generated by incident light reach the photo-absorbent layer in the case of front-illuminated detectors, the responsivity of the MSM UV sensor remains higher if the contact electrodes are transparent (Su et al., 2002) and as thin as possible so that they can transmit maximum possible light. That's why Gold (Au) layer thickness for metal contacts has been considered as 20 nm for all MSM structures under the present analysis. The absorption coefficient per unit length (α) of the Al_0.5Ga_0.5N layer has been defined in the Atlas program. It can also be calculated using equation (1) mentioned in Bütün (2011): (1) α=4πk/λ \alpha = 4\pi k/\lambda where k is the imaginary part of the refractive index, and λ is the wavelength of the incident light. Adequate material properties of all structure layers have been defined in Atlas-Silvaco program that include band gap, electron affinity, permittivity and optical constants.

As per past studies, TCAD Altas Silvaco numerical simulation helps to solve Poisson's and carrier continuity equations related to semiconductor physics so that the performance of semiconductor devices can be predicted accurately (Lophitis et al., 2018). Fermi Dirac statistics, fixed carrier lifetimes, Standard concentration and electric field dependent mobility and Selberherr impact ionisation models have been proposed in Atlas (2005) and Zhao (2018) for the simulation work. The thermionic emission concept was imitative by Bethe for high-mobility semiconductors; however, Schottky derived the diffusion concept for low-mobility semiconductors (Padhy, 2015). The mobility in AlGaN is mainly limited by impurity and phonon scattering (Atlas, 2005; Tsao et al., 2018). In our simulation work, low electric field mobility values have been considered for investigating I–V and current density characteristics. The electrical characteristics of all MSM structures under analysis have been extracted using SRH, optical and Auger recombination processes. Materials quality like trap/defects density is set to be ideal. Hence TCAD simulation tool can provide help as how design parameters can influence the device performance.

Table 1 shows the electrode dimensions of the MSM detectors considered in the simulation work. For calculating the fill factor, the ratio of the total active area to the total detector area is considered. The calculated fill-factor value is less than 50% for (S = W) based structures. However, double spacing structures provide better light conversion efficiency (Averine et al., 2001) because more than 50% of the total detector is exposed to incident light.

Table 1 shows that both electrode finger width and spacing tend to decrease with an increase of fingers for both structures. The responsivity of MSM detectors depends on the ratio of active optical area to the total detector area. It also depends on the area of SCR (Space Charge Region) associated with the cathode and the anode terminals (Palacios et al., 2002). The excitation power of 1 W/cm² at a wavelength of 270 nm has been utilised for incident light in the simulation work.

In MSM PDs, the photocurrent is associated with the generation of minority carriers around the SCR of Schottky contacts when UV light is incident on it. These carriers are collected by the metal electrodes when drifted by electric field (Palacios et al., 2002). Since every structure is RB at 50 V, so constant electric field has been observed for most of the MSM structures under investigation. The electric field of 145 kV/cm has contributed for collecting photogenerated carriers in the active layer.

Figure 2 shows Reverse biased I–V characteristics of 5-Fingers based S = W design. Red line is representing dark current and blue line representing photocurrent at 1 W/cm² light intensity.

Figure 2

Figure 2 shows that dark current remains very low even at 50 V. Ultra-low dark current has been observed for each structure in our set of analyses, indicating high SBH for Au contacts. The photocurrent starts to increase at around (−45 V) and then tends to increase with the increase in R.B as seen in Figure 2. As mentioned in Mello et al. (2008), the dependence of photocurrent or responsivity on applied voltage shows the existence of photoconductive gain in the MSM detector. The cause of gain is the tunnelling of electrons due to hole accumulation and image carrier lowering at cathode terminal. Around 16.5 mA photocurrent has been observed at 50 V of R.B in Figure 2.

Using I–V characteristics and current density plots, the performance parameters of both the designs up to 6 fingers have been compared in Table 2.

Results in Table 2 reveal that 6-Fingers based (S = W) MSM structure shows the highest PDCDR and light current density values. Past research studies concluded that thermally stable contacts with low dark current are desirable for higher QE, PDCR, and responsivity (Miller et al., 2018). The least dark current density is observed for 5-Fingers based (S = 2W) design but with lower PDCDR. In both the structures, the light current density tends to increase with the increase in the number of fingers up to 5 and then it decreases with more increase in fingers.

The plots related to Dark current density and photocurrent for (S = W) and (S = 2W) based designs are shown in Figures 3 and 4, respectively. It has been observed that there is a significant effect on dark current density with the variation in number of fingers for both design types.

Figure 3

Figure 4

It can be seen in Figure 3a that dark current density is relatively low for 14-Fingers based (S = 2W) MSM design. As with the increase in number of fingers, finger width tends to decrease thus dark current density reduces. However, for 14-Fingers based (S = W) MSM structure, the dark current density is comparatively higher as shown in Figure 3b because of the larger width results in a high dark current (Zebentout et al., 2011; Zhao, 2018). We have calculated total electrode width and total spacing for all MSM detectors using equation (2) and equation (3), respectively. (2) Total width = Sum of width dimension of all electrodes Total\,width\, = \,Sum\,of\,width\,dimension\,of\,all\,electrodes (3) Total spacing = Sum of spacing dimension of all electrodes Total\,spacing\, = \,Sum\,of\,spacing\,dimension\,of\,all\,electrodes

The SW relation must be preserved for electrode dimensions in case of MSM detectors (Averine et al., 2001). Lower dark current density has been obtained for S = 2W designs in comparison to S = W based designs of the same number of fingers. The width dimension of all structures under S = 2W is lower as compared to S = W designs therefore as the width of electrodes decreases, dark current gets decreased as mentioned in Alhelfi et al. (2016).

It can be seen in Figure 3a that with the increase in number of fingers, dark current density tends to decrease. This is because width decreases with the increase in fingers. However, dark current density increases abruptly for 3- and 6-Fingers based MSM detectors. Higher dark current density for 3-Fingers design is due to high electric field. Moreover for 6-Fingers and 5-Fingers designs, total spacing and total width values are coming out to be almost the same. However, for the same total width and spacing values, dark current density remains higher for 6-Fingers due to overlapping of space charge regions (Palacios et al., 2002).

A similar decrease in dark current density with the increase in fingers has also been observed in Figure 3b. However, for 5- and 14-fingers, there is an abrupt increase in dark current density values. Since total calculated width and spacing values are coming out to be almost the same for both 5-Finger and 6-Finger designs. Therefore we are getting a large dark current density for 5-Fingers. For all S = W based designs, the total spacing dimension remains lower than total width. However, for 14-Fingers design the total spacing is coming out to be higher than width; thus, dark current density has increased.

Recombination due to defects in the depletion region has been identified as the primary carrier loss mechanism in Nitride devices as per research reports (Allam et al., 2013). Recombination loss due to dislocation density in AlGaN layer can be minimised by applying a higher electric field, which helps to minimise the carriers' transit time within the depletion region (Wang et al., 2013). A large and almost constant electric field has been observed in all proposed MSM structures, which leads to promote the separation of photogenerated carriers for collection by electrodes.

It can be seen from (Figure 4a and b) that photocurrent/responsivity increases with an increase in the number of fingers. This is because spacing between the fingers decreases with increase in number of fingers and transit time of carriers gets decreased as expected from the theory of the current transport in the semiconductor devices (Mello et al., 2008). Carrier collection ability of electrodes/responsivity improves when the spacing between contacts decreases (Reverchon et al., 2004; Schühle and Hochedez, 2013) because illumination area or fill factor increases subsequently, increasing illumination current. In addition, it has been stated in Palacios et al. (2002), Zhao and Donaldson (2018a) and Zhao (2018) that when finger pitch and width are reduced, enhancement in photocurrent can be obtained. That is why a large photocurrent has been observed for S = W design when plots are shown in Figure 4a and b have been compared. With the 4- and 5-Fingers, it seemed to remain almost constant, however, it would keep on increasing with further increase in number of fingers.

In both (Figure 4a and b), photocurrent keep on increasing with the increase in fingers because spacing dimension has been decreasing with increase in number of fingers as seen in Table 2. It is mentioned in Mello et al., (2008) and Zebentout et al. (2011) that responsivity tends to increase when spacing between contacts decreases due to the effect of transit time. In Figure 4a, almost the same values of photocurrents have been obtained for both 4- and 5-Fingers based designs because almost the same total spacing area has been exposed to incident light. If width of electrodes is more than spacing even then responsivity decreases as compared to S = W based design (Barkad et al., 2010). In Figure 4b slight increase in photocurrent has been obtained for 5-Fingers as compared to 4-Fingers because a slightly higher value of total finger spacing has been exposed to light.

The comparison of present work with existing similar MSM detectors has been performed using Table 3. Existing MSM detectors have large number of fingers, however, based on S = W electrode dimensions.

Using Table 3, it can be seen that in the present work, lower dark current density at 50 V (for S = 2W 5-Finger design) has been obtained as compared to the dark current density mentioned in Gökkavas et al. (2007). However, dark current density in the present work at 20 V remains higher in comparison with the values mentioned for Xie et al. (2012) and Wang et al. (2013) because major defect states could not be compensated in simulation work in contrast to experimental work. Moreover, photocurrent in our simulation work at 20 V could be approximated in good agreement with the experimental work performed for similar AlGaN/AlN/Sapphire MSM structure in Xie et al. (2012). Since the defects in the active layer are considered ideal in the present work which influence the life time of carriers, recombination rate and responsivity or photocurrent of detector (Mello et al., 2008). Therefore, in our work we are getting slightly lower value of photocurrent at 20 V as specified in Table 3 for Xie et al. (2012) and Wang et al. (2013).

Figure 5 shows lowest dark current density of 9.5e-11 A/cm² at 50 V for (S = 2W) detector with 5-Fingers. The cut-line plot of conduction current density has been derived along the x-dimension of the device structure.

Figure 5

Response time is another critical consideration in the performance of the device. Response time of PD (which is related to bandwidth) is quantified 10% to 90% as rising time and 90% to 10% as decay time of photocurrent (Miller et al., 2018). AlGaN MSM structures enable temporal response in ps (picosecond) with a very low dark current (Zhao, 2018). Photocurrent response follows an exponential decay when recombination time remains shorter than carriers' transit time (McAdoo, 2000). A long decay tail arises with the arrival of released carriers when photogenerated carriers get trapped in the contact regions. Another factor responsible for slowing the photocurrent mechanism is the carrier diffusion component, which originates from low field areas closer to the centre of the structure (Ozbay et al., 2004; Tsao et al., 2018). Therefore, as per past studies, the response speed is mainly limited by the RC time constant of the interdigitated electrode geometry and the transit time of optically generated carriers (Averine et al., 2001). Exponential decay in photocurrent using transient analysis has been observed in the present work when incident light was switched OFF from high light intensity. For the largest number of finger-based designs, the least value of fall time is observed in transient response as expected.

A cut-off wavelength of 400 nm was observed for all MSM structures using another set of spectral analyses. This observation is another evidence for the suggested UV detection. Therefore, low dark current density and reasonable photocurrent of AlGaN MSM PD design structures could be used in optical communication, microwave and UV detection applications with low noise and higher sensitivity.

The effect of electrode dimensions by varying the number of fingers on the performance factors of fixed area Al_0.5Ga_0.5N/AlN/Sapphire MSM detector is analysed. Two MSM detectors based S = W and S = 2W have been simulated and compared using Atlas TCAD software. I–V characteristics, current density and transient plots have been taken into consideration for comparative analysis. Designs based on S = W exhibited large photocurrent or responsivity as compared to S = 2W based detectors. All the devices under investigation delivered excellent performance in terms of low dark current and good linearity between photocurrent and incident power density. The trend of variation in dark current density with the increase in the number of fingers has been observed different for both structures. For ultra-low dark current density, S = 2W designs can be preferred. The largest number of fingers based structures exhibited fast transient response in terms of least fall time as expected. A sharp cutoff wavelength at 400 nm was obtained at 270 nm incident light wavelength. Therefore by selecting an appropriate number of fingers with a suitable MSM detector structure, reasonable photocurrent and better speed performance at low dark current density could be obtained for low noise UV detection and optical communication applications.