In this letter we report the UV detection characteristics of an epitaxial graphene (EG)/SiC based

Transcription

1 Evidence of minority carrier injection efficiency >90% in an Epitaxial Graphene/SiC Schottky Emitter Bipolar Junction Phototransistor for Ultraviolet Detection Venkata S. N. Chava 1, Sabih U. Omar 2, Gabriel Brown 1, Shamaita S. Shetu 2, J. Andrews 1, T.S. Sudarshan 1, M.V.S. Chandrashekhar 1 1 Department of Electrical Engineering, University of South Carolina, Columbia, SC 29208, U.S.A. 2 Intel Corporation, Hillsboro, Oregon 97124, U.S.A. Abstract In this letter we report the UV detection characteristics of an epitaxial graphene (EG)/SiC based Schottky emitter bipolar phototransistors (SEPTs) with EG on top as the transparent Schottky emitter layer. Under 0.43 µw UV illumination, the device showed a maximum common emitter current gain of 113, when operated in the Schottky emitter mode. We argue that avalanche gain and photoconductive gain can be excluded, indicating minority carrier injection efficiency, γ, as high as 99% at the EG/p-SiC Schottky junction. This high γ is attributed to the large, highly asymmetric barrier, which EG forms with the p-sic. The maximum responsivity of the UV phototransistor is estimated to be 7.1 A/W. The observed decrease in gain with increase in UV power is attributed to recombination in the base region which reduces the minority carrier lifetime. Key words: Epitaxial graphene (EG), Silicon Carbide (SiC), EG/SiC heterojunction, minority carrier injection, phototransistor, UV detector Corresponding author vchava@ .sc.edu 1

2 Introduction UV photodetectors have a wide variety of applications in defense, and also as biological and chemical sensors. Solid state UV detectors based on Si, SiC, GaN, AlGaN, InGaAs and GaAs are popular due to their reliability and light weight. Since Si based UV detectors are sensitive to visible light, visible blind wide bandgap devices based on SiC could be a better choice due to their radiation hardness and high breakdown fields. In the past many groups have reported SiC UV detectors based on Schottky, MSM and p-i-n structures [1, 2]. EG/SiC phototransistors are of particular interest due to the advancement in the growth technology of both epitaxial graphene and SiC and, recent research related [3] to the understanding of the interface and transport characteristics of EG/SiC junctions. Here [3], the authors reported a non-ideal EG/p-SiC Schottky diode due to two carrier transport whereas the EG/n-SiC generally shows a near ideal thermionic emission behavior. Additionally, the advantage of using EG on top as an emitter layer is that EG which is 2-3 atomic layers thick (absorption ~0.6%/monolayer on SiC [4]) does not strongly absorb high energy UV photons which would otherwise be absorbed at the surface layer of a pn junction, typically ~100nm thick. UV light can therefore be effectively transmitted through the EG to reach the base region and create e-h pairs in the base and the B-C region depending on the energy of the incident photons. As Schottky devices are typically majority carrier devices they offer advantages like fast switching time due to fast recombination in the metal, and minimal series resistance in the emitter, as has been shown in bipolar mode Schottky devices [5]. Moreover, the large Schottky barrier height to p-sic (which is ~2.7 ev) in our device would result in reduced reverse leakage current [6], potentially breaking the tradeoffs in speed and leakage current between unipolar and bipolar devices in certain applications. 2

3 When a phototransistor is illuminated using a light source, the incident light passes through the transparent EG layer and gets absorbed in the base, B-C junction and collector regions and thus creating e-h pairs, and due to the applied E-C bias voltage, the holes are swept towards the base region and electrons are swept towards the collector terminal. However these holes in the base region see a large potential barrier near the EG/p-SiC interface and therefore gets accumulated near the valence band edge of the base. To maintain the charge neutrality in the base region, there has to be electron injection from the EG emitter layer into the base layer, thereby resulting in a bipolar action. Therefore, it is of interest to understand the injection mechanism and also how this carrier injection affects phototransistor characteristics such as gain and responsivity. In this letter, we describe the device fabrication and characteristics of an EG/p- SiC/ n + -SiC Schottky emitter bipolar phototransistor (SEPT). Here it is worth pointing out the interface characteristics of EG/p-SiC interface, which is the E-B junction, in our study. The presence of a buffer layer composed of C atoms at the EG/SiC interface causes the Fermi level to pin close to the SiC conduction band regardless of doping of the SiC. A low potential barrier of 0.5 ev is reported for n-sic and a high barrier of ~2.7 ev to p-sic on the SiC (0001) surface, also known as the Si face [7]. Fig. 1(b) shows the energy band diagram of the EG/ p-sic/ n + -SiC SEPT. As shown, the interface between EG/ p- SiC is a Schottky barrier, the carrier transport across the interface is expected to occur by thermionic emission of the majority carriers. Experimental details As shown in fig. 1(a), a bipolar transistor is fabricated with an EG emitter, p-sic base and n + - SiC as collector. Epitaxial growth of p-sic on commercially available 4 0 off-cut n + -SiC (0001) 3

4 substrate is done using CVD growth. A low C/Si ratio of 1.2 is maintained in the source gas during the CVD process. The pressure and temperature are maintained at 300 Torr and C respectively. The growth rate during this process is observed to be 20 μm/hr [8]. The resultant p- doping is found to be 3x10 14 cm -3 using a mercury probe C-V method [8]. The epi-layer thicknesses (i.e., the base width, W B ) as measured by FTIR was ~30 μm [9] and is not mesa isolated. The epitaxial graphene (EG) was grown by thermal sublimation of the p-sic epi-layer surface in vacuum at C, and is described in greater detail elsewhere [10]. The presence of graphene was confirmed using Raman spectra obtained using a Horiba JY spectrometer with an excitation line of 631 nm. The D/G ratio was estimated to be <0.06. The XPS measurements showed EG thicknesses of 2-3 monolayer in similar growth conditions [11]. The 250 μm diameter circular graphene mesa was defined using an O 2 reactive ion etching (RIE) through a photoresist mask. An RF-sputtered Ti/Au film was used to form a large area ohmic contact on the back of the SiC substrate. For phototransistor operation, the graphene emitter was held at a negative bias with respect to the n + -SiC emitter layer by directly contacting the graphene layer with a tungsten probe. The base here is floating and the base photocurrent was provided by optical excitation from an Omnicure S1000 Hg-vapor lamp with variable intensity as the source in a microscope with 10x objective. The UV light intensity of this lamp is further attenuated by placing a SiC wafer at the inlet, along with an additional aperture to limit the spot size to <200 µm i.e., to confine it to the limits of the device. The lamp spectrum provided three excitation lines at 312, 334 and 365 nm wavelengths which are above the bandgap for SiC. The optical power at the device was measured using a Si photodiode sensor PM 100D from Thorlabs. 4

5 Results and discussion: Fig. 2(a) shows typical collector current (I c ) vs collector-emitter voltage (V CE ) characteristics for the EG/p-SiC/n + -SiC SEPT under different levels of UV illumination. Fig. 2(b) shows the photocurrent (I c ) ph vs bias voltage, where (I c ) ph is obtained by subtracting the dark current from I c. In our present study, the UV illumination is varied from 0.43 µw to 7.87 µw, corresponding to the measured short circuit current values 43.5 pa to 50.4 pa. The relatively large dark current is attributed to the lack of mesa isolation between the p-sic base and n + -SiC collector, as it is difficult to etch large depths of SiC selectively, 30 µm in this situation ensures adequate absorption of the 365 nm UV light. It is clear that the photocurrent at the collector increases with the UV light power. Under constant UV illumination, the collector current increases with an increase in the emitter-collector bias voltage. This increase in the collector current may be explained by considering the built-in potential at the EG/SiC interface which gets lowered by applying a forward bias at this E-B junction, resulting in the injection of electrons from the EG to p-sic. These electrons are collected by the collector terminal due to the applied reverse bias voltage near the B-C region. This injection can occur in 2 ways: i) Thermionic emission for carriers over the Schottky barrier: This is considered the dominant current component in Schottky junctions, and corresponds to the majority carriers (holes, in this case) in typical situations. In most Schottky barriers, the barrier height of the majority carrier is smaller than the corresponding barrier for the minority carriers i.e. Φ b <E g /2. However, in this EG/p-SiC junction, the situation is reversed, with the barrier for minority electrons, Φ n << Φ p the barrier for majority holes (Fig. 1. (b)). Thus, one may expect that on the basis of thermionic emission 5

6 alone, significant minority carrier injection would be observed. A similar situation was also observed in Schottky point contacts by Bardeen et al. [12], where the majority barrier Φ n to the n-base was >E g /2, leading to efficient minority carrier injection, and consequently, bipolar gain. ii) Diffusion over the barrier by reduction of the built-in potential: As always, there will also be a diffusion component to the overall current, both from the electrons and holes. There will be electron diffusion due to the reduction of the built-in voltage from the applied bias. There will also be hole injection from the p-sic base into the EG, although this is expected to be much smaller due to the much smaller hole concentration in SiC (~ cm -3 ) vs. the electron concentration in EG (>10 19 cm -3 ). We note here that the built-in voltage ~2.4 V is smaller than the Schottky barrier height to p-sic (2.7 ev), meaning that diffusion of electrons can overwhelm thermionic emission of holes. Thus, because of the large Φ p >E g /2, it is possible for electron injection into the base to be larger than hole injection back the other way. These injected minority electrons reach the collector region through base layer by diffusion, assuming that the back BC junction is strongly recombining [13], a point we will revisit later. This recombination leads to a concentration gradient of minority carrier electrons in the base region, which drives the diffusion. Here, since holes are the majority carriers for EG/p-SiC, we would not measure bipolar gain if the current transport across the E-B junction is only due to majority holes instead of minority electrons as is the case in n-p-n transistors. Further below, we exclude the possibility of gain from avalanche processes, as well as persistent photoconductivity, only leaving bipolar gain as a possibility. Thus, we argue that there has to be significant electron injection from EG to p- 6

7 SiC to observe gain in the device. This is corroborated by a recent report [3], where they show evidence for minority carrier injection from EG into the SiC. From the measured collector current, the peak value of electron concentration at the interface of the base and E-B depletion region, is estimated to be 5x10 13 cm -3 by assuming that the concentration of electrons at the back edge of the collector is zero (recombination velocity >10 5 cm/s at the back interface [13]). For the estimation of the electron concentration, we used Fick s law of diffusion in the neutral base region: qd n dn dx = I n (1) Here D n is the diffusion coefficient of electrons and q is the charge of the electron. Since epi layer (base) doping is 3x10 14 cm -3 > 5x10 13 cm -3, indicating low injection level. This low injection level is valid for the highest collector current obtained for 7.87µW illumination at 60V. In this SEPT, the optically generated carriers are multiplied by the transistor common emitter gain and thus enhances the collector current depending on how this gain changes. Optical gain (h FE ) [14] is given by: h FE = (I C) ph I b 1 (2) Where (I c ) ph is the measured collector current after subtracting the dark current and I b is the base current. The subtraction of unity in equation (2) accounts for the photogenerated current, which should not be double counted. The base photocurrent (I b ) [14] is calculated by: I b = I Ph = P Opt 1 exp ( α abs W b ) E Ph (3) Here P opt is the incident UV power, W b is the neutral base width, α abs is the absorption coefficient, approximately ~ 80 cm -1 [15] for 365nm wavelength photons in 4H-SiC, E ph is the 7

8 energy of UV photons corresponding to 365nm. We assumed the reflectance of our device is the same as that of the Si photodiode, so that the measured P opt by Si photodiode represents the number of UV photons passing through the device. All calculations are done by assuming the UV light as a monochromatic UV radiation since the 365 nm line is dominant in the UV lamp spectrum. Fig. 3 show the variation in gain calculated using equations (1) and (2), with UV power at 60V. Maximum gain of 113 is obtained when the illumination level is set to be 0.43 µw. The gain values are decreasing steadily by increasing the incident UV power. The collector current is a net result of the diffusion of injected minority carrier electrons and this diffusion current depends on how many of these injected electrons diffuse through the base without recombining with holes before they get collected by the collector terminal. Hence, even if there is electron injection from the EG into the p-sic base, the e-h recombination in the base region is a limiting factor for the collector current. Thus, as the current increases, the total number of injected electrons and holes in the base also increase, leading to faster recombination of carriers in the base, which is proportional to the np product [14]. This suggests that base recombination is the limiting mechanism of the gain in this device, similar to that observed in GaN heterojunction bipolar devices [16]. The device characteristics of other devices fabricated on the same substrate with same dimensions also found to show similar I-V characteristics and the gain values are also in the same range. In general, the gain in photodetector can occur by any of three mechanisms namely i) photoconductive [17], ii) avalanche [18] and iii) bipolar gain [19] depending on the device structure and applied bias. To understand the gain mechanism in SEPT, we calculated the transit 8

9 time of electrons at low injection in the base region (τ) by diffusion [19], and is estimated to be 180ns, using: τ transit = W B 2 (4) 2D n Also, the lower bound for the recombination time, τ recombination, of electrons diffusing from base to collector is estimated by assuming the recombination velocity of electrons at the p/n+ SiC BC junction terminal as >10 5 cm/s as measured by Kimoto [13], and its value is estimated to be <30ns for a 30µm base. Therefore, since τ recombination < τ transit, the carrier recombines as soon as it makes one pass through the long 30µm base. Thus, we exclude the possibility of photoconductive gain here. In avalanche photodetectors the photocurrent and dark current are increased by a multiplication factor M [18] given by, M = 1 1 ( V CB V B ) (5) Here, V CB is the collector-base voltage and V B is the breakdown voltage of the B-C junction. For 4H-SiC, V B is calculated [20] using: V B = 3 X X(N D ) 3/4.. (6) The breakdown voltage of the B-C junction is found to be ~ 10 4 V for the base doping, N D = 3x10 14 cm -3. From eq. (5), M=1. Therefore, we do not expect avalanche effect in our device, and this is evident from Fig. 2, where no sharp increase in the current is observed within a small range of applied bias voltage, confirming that the current gain is due to transistor action alone. Using the gain values at 60 V, and common base gain, α=β/(β+1) [14], we calculated the emitter injection efficiency (γ), using the equation (7) [14]: 9

10 α = γα T δ.... (7) Here, α T is the base transport factor and δ is the recombination factor. Assuming that α T and δ are unity, we estimate a lower bound for the minority carrier emitter injection efficiency (γ) to be 99.12% for the highest gain, decreasing to 91.7%. We have previously attributed this decrease to recombination in the base region (Fig.3). Therefore our assumption that α T is unity may not be true, indicating that γ may be even higher. The high γ for a Schottky junction is higher than the 10-15% observed in n-schottky emitter transistors [21]. Our high γ may be due to the fact that this is a p-schottky junction, so that minority carriers, electrons, are more mobile than majority carriers, leading to a gain proportional to D n/ /D p ~ 8 for p-sic, rather than D p /D n ~ 0.2 for n-si, as discussed by Chuang [22]. This observation of strong minority carrier action is consistent with Bardeen et al. [12]. The responsivity of our SEPT device at 365nm at zero bias voltage, calculated as the ratio of (I c ) ph and incident UV power, is 0.25 ma/w for 0.43 µw UV illumination. This responsivity has increased with increase in the bias voltage and reached a maximum value of 7.1 A/W at 60 V. It should be noted that the UV responsivity of our device is better than the recently reported EG/SiC UV detector [23] and graphene/sic MSM photodetector [24] which showed a maximum responsivity of 0.2 A/W at 310 nm and 2 ma/w a 365 nm in the respective order. Further, the responsivity showed a decreasing trend with increasing UV illumination at 60 V which again might be a result of recombination in the base region. From table I, it is clear that our device shows better responsivity compared to other devices in the near UV regime. 10

11 S. No. UV detector Responsivity (A/W) at 365 nm 1 EG/SiC SEPT EG/SiC pn diode 0.2 ( at 310 nm) [23] 3 EG/SiC MSM [24] 4 4H-SiC p-i-n [25] 5 4H-SiC APD 0.01 [26] 6 GaN p-i-n diode 0.15 [27] 7 GaN pn diode 0.1 [28] Table I. Comparison of near UV responsivities of different UV detectors. Since our detector operates as a bipolar phototransistor, high gain values can be achieved even at low bias values by optimizing the design of the detector such as epi-layer thickness and doping. The dark current can also be reduced significantly by mesa isolating the base. Conclusions: Vertical heterojunction bipolar phototransistors are fabricated with EG/ p- SiC/ n + -SiC as emitter, base and collector layers in the respective order. The current-voltage characteristics of this device is tested under UV light for its application in UV detectors. From these I-V characteristics, gain values are estimated for different UV illumination powers at 365nm. The highest gain value is found to be 113 and also responsivity values as high as 7.1 A/W are observed under 0.43 µw illumination at 60 V bias voltage. We argue that the gain in this bipolar phototransistor device is a result of a two carrier transport across the EG/SiC junction, contrary 11

12 to the general assumption that as a Schottky junction, it is expected to show thermionic emission due to majority carriers. High responsivities of our device under low UV illuminations will help researchers to further explore devices based on EG/SiC heterojunctions. Acknowledgments: This research work was partially supported by the Office of Naval Research (ONR), grant no. N The authors deeply appreciate the continued encouragement and support from program manager Dr. H. Scott Coombe. The authors would also like to acknowledge the support from National Science Foundation (NSF), ECCS-EPMD Award No under the supervision of program director Dr. Anupama Kaul. Finally, the authors would like to acknowledge the support from Department of Energy (DOE), Nuclear Energy University Program (NEUP) Award No under the supervision of program director Dr. Michael Miller. References 1 E. Monroy, F. Omnes, and F. Calle, Semicond. Sci. Technol. 18, R33 (2003). 2 M. Razeghi, Proc. IEEE 90, 1006 (2002). 12

13 3 T.J. Anderson, K.D. Hobart, L.O. Nyakiti, V.D. Wheeler, R.L. Myers-Ward, J.D. Caldwell, F.J. Bezares, G.G. Jernigan, M.J. Tadjer, E. A. Imhoff, A. D. Koehler, D.K. Gaskill, C.R. Eddy, and F.J. Kub, IEEE Electron Device Lett. 33, 1610 (2012). 4 J.M. Dawlaty, S. Shivaraman, J. Strait, P. George, M. Chandrashekhar, F. Rana, M.G. Spencer, D. Veksler, and Y. Chen, Appl. Phys. Lett. 93, 2013 (2008). 5 Y.Amemiya, and Y.Mizushima, IEEE Trans. Electron Devices 31, 35 (1984). 6 K.J. Schoen, J.M. Woodall, J. a. Cooper, and M.R. Melloch, IEEE Trans. Electron Devices 45, 1595 (1998). 7 C. Coletti, S. Forti, A. Principi, K. V. Emtsev, A. A. Zakharov, K.M. Daniels, B.K. Daas, M.V.S. Chandrashekhar, T. Ouisse, D. Chaussende, A. H. MacDonald, M. Polini, and U. Starke, Phys. Rev. B 88, (2013). 8 Haizheng Song, Tawhid Rana, M.V.S. Chandrashekhar, Sabih U. Omar, and T.S.Sudarshan, ECS Trans. 58, 97 (2013). 9 M.F. Macmillan, A. Henry, and E. Janzeni, J. Electron. Mater. 27, 300 (1998). 10 B.K. Daas, K.M. Daniels, T.S. Sudarshan, and M.V.S. Chandrashekhar, J. Appl. Phys. 110, (2011). 11 B.K. Daas, S.U. Omar, S. Shetu, K.M. Daniels, S. Ma, T.S. Sudarshan, and M.V.S. Chandrashekhar, Cryst. Growth Des. 12, 3379 (2012). 12 J. Bardeen and W.H. Brattain, Phys. Rev. 74, 230 (1948). 13

14 13 T. Kimoto, T. Hiyoshi, T. Hayashi, and J. Suda, J. Appl. Phys. 108, (2010). 14 S.M. Sze, Physics of Semiconductor Devices, 2nd Ed (1981). 15 S.G. Sridhara, R.P. Devaty, and W.J. Choyke, J. Appl. Phys. 84, 2963 (1998). 16 W. Yang, T. Nohava, S. Krishnankutty, R. Torreano, S. McPherson, and H. Marsh, Appl. Phys. Lett. 73, 978 (1998). 17 B.E.A. Saleh, M.C. Teich, Fundamentals of Photonics (Wiley Series in Pure and Applied Optics) (1991). 18 J.C. Campbell, A. Dentai, G.J. Qua, and J. Ferguson. IEEE J. Quantum Electron. 19, 1134 (1983). 19 Kunihiro Suzuki, IEEE Trans. Electron Devices 38, 2512 (1991). 20 B.J. Baliga, Fundamentals of Power Semiconductor Devices (2008). 21 A.Y.C. Yu and E.H. Snow, Solid. State. Electron. 12, 155 (1969). 22 C.T.Chuang, IEEE Trans. Electron Devices 30, 700 (1983). 23 Travis J. Anderson, Karl D. Hobart, Jordan D. Greenlee, David I. Shahin and and F.J.K. Marko J. Tadjer, Eugene A. Imhoff, Rachael L. Myers-Ward, and Aris Christou, Appl. Phys. Express 8, (2015). 24 Erdi Kusdemir, Dlice Ozkendir, Volkan Firat, and Cem Celebi, J. Phys. D. Appl. Phys. 48, (2015). 14

15 25 X. Chen, H. Zhu, J. Cai, and Z. Wu, J. Appl. Phys. 102, (2007). 26 H. Zhu, X. Chen, J. Cai, and Z. Wu, Solid. State. Electron. 53, 7 (2009). 27 D. Walker, A. Saxler, P. Kung, X. Zhang, M. Hamilton, J. Diaz, and M. Razeghi, Appl. Phys. Lett. 72, 3303 (1998). 28 E. Monroy, E. Muñoz, F.J. Sánchez, F. Calle, E. Calleja, B. Beaumont, P. Gibart, J. A Muñoz, and F. Cussó, Semicond. Sci. Technol. 13, 1042 (1999). 15

16 Figure and Table Captions: Fig. 1 (a). Schematic of EG/p-SiC/n + -SiC Schottky emitter bipolar phototransistor (SEPT) under UV illumination with EG in emitter mode. (b) Band diagram of the Schottky emitter bipolar phototransistor (SEPT) under illumination. (No charge is collected at forward biased E-B junction and charge collection occurs at reverse biased B-C junction) Fig. 2 (a) Current-voltage characteristics and (b) photocurrent-voltage characteristics of a typical EG/ p- SiC/n + - SiC SEPT under various UV light illumination levels. (Inset in fig. 2 (a) shows full range of measured current-voltage (I c vs V CE ) characteristics) Fig. 3 Variation of current gain (h FE ) of EG/ p-sic/ n + - SiC SEPT with UV illumination level at 60 V. Table I. Comparison of near UV responsivities of different UV detectors. 16

17 Figure 1. 17

18 Figure 2. 18

19 Figure 3. 19

20 Table I. S. No. UV detector Responsivity (A/W) at 365 nm 1 EG/SiC SEPT EG/SiC pn diode 0.2 ( at 310 nm) [23] 3 EG/SiC MSM [24] 4 4H-SiC p-i-n [25] 5 4H-SiC APD 0.01 [26] 6 GaN p-i-n diode 0.15 [27] 7 GaN pn diode 0.1 [28] 20

21

22

23