Simulation of GaN micro-structured neutron detectors for improving electrical properties
Geng Xin-Lei1, Xia Xiao-Chuan1, †, Huang Huo-Lin2, Sun Zhong-Hao1, Zhang He-Qiu1, Cui Xing-Zhu3, Liang Xiao-Hua3, Liang Hong-Wei1, ‡
School of Microelectronics, Dalian University of Technology, Dalian 116024, China
School of Optoelectronic Engineering and Instrumentation Science, Dalian University of Technology, Dalian 116024, China
Institute of High Energy Physics, Chinese Academy of Sciences (CAS), Beijing 100049, China

 

† Corresponding author. E-mail: xiaochuan@dlut.edu.cn hwliang@dlut.edu.cn

Project supported by the National Natural Science Foundation of China (Grant Nos. 11675198, 11875097, 11975257, 61774072, 61574026, and 61971090), the National Key Research and Development Program of China (Grant Nos. 2016YFB0400600 and2016YFB0400601), the Fundamental Research Funds for the Central Universities, China (Grant No. DUT19LK45), the China Postdoctoral Science Foundation (Grant No. 2016M591434), and the Science and Technology Plan of Dalian City, China (Grant No. 2018J12GX060).

Abstract

Nowadays, the superior detection performance of semiconductor neutron detectors is a challenging task. In this paper, we deal with a novel GaN micro-structured neutron detector (GaN-MSND) and compare three different methods such as the method of modulating the trench depth, the method of introducing dielectric layer and p-type inversion region to improve the width of depletion region (W). It is observed that the intensity of electric field can be modulated by scaling the trench depth. On the other hand, the electron blocking region is formed in the detector enveloped with a dielectric layer. Furthermore, the introducing of p-type inversion region produces new p/n junction, which not only promotes the further expansion of the depletion region but also reduces the intensity of electric field produced by main junction. It can be realized that all these methods can considerably enhance the working voltage as well as W. Of them, the improvement on W of GaN-MSND with the p-type inversion region is the most significant and the value of W could reach 12.8 μm when the carrier concentration of p-type inversion region is 1017 cm−3. Consequently, the value of W is observed to improve 200% for the designed GaN-MSND as compared with that without additional design. This work ensures to the researchers and scientific community the fabrication of GaN-MSND having superior detection limit in the field of intense radiation.

1. Introduction

Due to the shortage and rising cost of 3He gas, the application of 3He proportional counters for detecting neutrons is restricted.[1] Si-based semiconductor neutron detectors have been utilized extensively as an alternative method for detecting neutrons during the past decade.[25] Nevertheless, they suffer radiation damage due to the small bandgap and displacement energy of Si.[6] With the development of wide bandgap semiconductors, GaN is considered to be the potential candidate for neutron detection under strong irradiation and high temperature conditions[7,8] due to its wide band gap (3.39 eV), high critical breakdown electric field (3 × 106 V/cm), large displacement energy (22 eV), high density (6.2 g/cm−3) and good thermal stability (melting point: 2500 °C).[911] However, the detection performance such as detection efficiency and energy resolution of the neutron detector based on GaN is still not comparable to those based on Si,[1214] which is due mainly to two reasons. (i) The planar structure is commonly employed in GaN neutron detectors, which cannot avoid the low detection efficiency restricted by the conflict between the long neutron mean free path and the short-charged particle range in neutron conversion layer.[15] (ii) The unintentionally doped GaN is usually of n-type with the background carrier concentration of ∼ 1016 cm−3. Additionally residual shallow donors generated during material growth in the metal–organic chemical vapor deposition (MOCVD) or the hydride vapor phase epitaxy (HVPE),[15,16] bring difficulties for the enlargement in the width of depletion region (always only few micrometers).[17,18] Thus, it is urgently needed to find a solution for GaN-based neutron detectors that could overcome the shortcomings of the planar architecture and achieve a larger depletion region at the same background carrier concentration.

A novel GaN micro-structured neutron detector (GaN-MSND) is proposed in this paper. The advantages of GaN-MSND over the planar structure are analyzed and the significance of the wide depletion region for the detection performance is discussed in detail. Moreover, Sentaurus-TCAD software is utilized to investigate three methods such as the method of modulating trench depth, the method of introducing dielectric layer and p-type inversion region to improve the width of depletion region of GaN-MSND. The mechanism and comparison of these methods are analyzed. It is expected that the GaN-MSND could enhance the detection efficiency of the planar neutron detector and achieve the superb detection performance with the improvement of the depletion region width. Consequently, such a novel strategy will be used in future to make the neutron detectors capable to cope with the extreme conditions.

2. Structures of GaN-MSND and simulation model

Figure 1 shows the schematic cross-section of the proposed GaN-MSND. The neutron converter of GaN-MSND is filled in the etched trenches and most of the reaction products generated at various depths of the neutron converter can enter into the adjacent detector. The neutron converter of planar neutron detector is commonly coated on the surface of GaN. The thick neutron converter can raise the chance of capturing neutron, but the entrance probability of the reaction products produced by the interaction between the neutron and the conversion material into the detector is reduced. This contradiction greatly limits the detection efficiency of planar neutron detector.[15] In contrast, the GaN-MSND has a larger amount of neutron converter filling and a wider contact area of neutron converter with the detector. Thus, the detection efficiency of GaN-MSND can be improved significantly.[19]

Fig. 1. Schematic drawing of GaN-MSND.

When the detector is operated at a reverse bias, the depletion region (near the p-GaN/i-GaN junction) is expanded obviously due to the large difference in concentration between p-GaN and i-GaN as marked in Fig. 1. The width of depletion region (W) has a significant importance for the detector because its value determines the detection performance of GaN-MSND such as charge collection efficiency and energy resolution.[2023] A large width of depletion region can deposit more particle energy and improve the charge collection efficiency. Besides, the incomplete charge collection noise is reduced, which will result in the improvement of the energy resolution as can be seen from Eqs. (1) and (3).[21] Therefore, the method of creating a relatively large W is essential for obtaining GaN-MSND with superb detection performance. Operating the detector at a higher reverse bias is considered to be an easy way of increasing the W value.[24] However, an excessive high voltage will also lead to a high leakage current which can affect the system electronic noise and further degrade the detection performance such as energy resolution according to Eqs. (1) and (2).[21] The methods discussed in the next section focus on enhancing the operation voltage tolerance, by which a large depletion region can be achieved.

where FWHM is the energy resolution; ΔE1, ΔE2, and ΔE3 are the system electronic noise, incomplete charge collection noise, and carrier fluctuation noise, respectively; ΔE0 is the electronic noise of preamplifier; e is 2.71828…; q is the electron charge; ε is the average ionizing energy; τ is the shaping time; Id is the leakage current of the detector; rs is the input impedance; Cd is the capacitance of the detector; C0 is the capacitance of the gate–source field-effect transistor in the preamplifier; a is 0.5 for GaN; γ is the charge collection efficiency; Er is the energy of the particle.

The simulations are performed by using the Sentaurus-TCAD. The doping-dependence and high-field-saturation models are employed for mobility. The Shockley–Read–Hall model is utilized to calculate the effect of charging and discharging of trap. To investigate the avalanche breakdown effect in the detector, the Selberherr’s impact ionization model is used to simulate the reverse characteristics.

Figure 2(a) shows the simulation results of the electric field distribution in a specific GaN-MSND. The structural parameters used in simulation are listed in Table 1. The anode and cathode of the device are formed as Ohmic contacts on p-GaN and n-GaN. The maximum working voltage (Vr) is defined at the current density I = 1 × 10−6 mA/mm, where the leakage level is considered to be an acceptable level as shown in many researches.[18,25] Figure 2(b) shows the electric field profile extracted along the x-cut marked in Fig. 2(a). The electric field reaches a maximum value at the p-GaN/i-GaN junction and decreases linearly to 0 V/cm in the i-GaN. Parameter W is defined as the distance between the p-GaN/i-GaN interface and the 0 V/cm position. Thus, the specific GaN-MSND without additional design has a W value of 3.9 μm at Vr of 150 V.

Fig. 2. (a) Electric field distribution in GaN-MSND at Vr = 150 V, and (b) electric field profile of GaN-MSND extracted along vertical direction (x-cut).
Table 1.

Structural parameters of GaN-MSND adopted in the work.

.
3. Results and discussion
3.1. Influence of modulating trench depth on W of GaN-MSND

The trenches filled with neutron convert material 6LiF which plays a role of transferring the neutrons into the charged particles are crucial for the GaN-MSND. It should be noted that the modulation of the trench depth d (as shown in Fig. 1) is an effective way to improve Vr and W of the device. The plots of simulated reverse IV characteristics and the corresponding results of the Vr and W versus/d in a range from 0.5 μm to 7.5 μm in GaN-MSND are illustrated in Fig. 3. The increase of d can lead both W and Vr (defined at the current reaches 106 mA/mm) to be improved, but the additional etching depth to have no obvious effect when d reaches 5.5 μm. Considering the feasibility of the device fabrication, d is set to be 5.5 μm, by which Vr and W can reach 226 V and 4.7 μm, respectively.

Fig. 3. (a) Reverse characteristics of GaN-MSND versus voltage for various d values, and (b) W and Vr extracted from the simulation results versus d.

To explain the changing trend of Vr, the distribution of electric field of the device is analyzed. Figure 4 demonstrates the electric field profile extracted by taking a y-cut along the bottom of their trenches with the applied reverse voltage of 60 V. The peaks of the electric field occur at the trench corners, which are caused by the large curvature of electric field line. In addition, these values become lower as the distance between trench corner and the p-GaN/i-GaN junction increases, which contributes to the improvement of Vr.[26] However, when d reaches 5.5 μm, the trench corners are located in the non-depletion region without high electric field, which has no effect on the electric field distribution near the junction. Therefore, neither Vr nor W will continue to change with d.

Fig. 4. Profiles of electric field versus x for GaN-MSND for d values of 0.5 μm, 1.5 μm, and 2.5 μm at the depth of respective trench bottom when the reverse voltage is 60 V.
3.2. Influence of dielectric layer on W of GaN-MSND

Based on the optimized results, a possible method to further improve Vr and W is to introduce a dielectric layer on the trench surface as shown in Fig. 5. Single-layered SiO2 or Al2O3 with a thickness of 0.5 μm is adopted and their effects on Vr and W of GaN-MSND are compared in the simulation. The density of fixed charges Qf at the dielectric/GaN interface is calibrated by the reported value in the literature: for the interface of SiO2/GaN, the positive Qf is set to be 3 × 1011 cm−2;[27,28] for Al2O3/GaN, the unusual negative Qf at the interface due to the O-rich region near the interface caused by Al vacancy and O interstitial has been reported.[29] And its value of 2 × 1012 cm−2 is employed in this work.[30,31]

Fig. 5. Schematic diagram of GaN-MSND with trench surface covered by dielectric layer.

Figure 6 shows the reverse IV characteristics of GaN-MSND with SiO2 or Al2O3 layer, and the corresponding Vr is extracted to be 167 V and 270 V, respectively. Compared with the optimized results above, the Vr of GaN-MSND with SiO2 decreases while that with Al2O3 increases. The negative Qf at the Al2O3/GaN interface leads the conduction band to bend up, where the electrons are blocked. The electrons blocking region can drive the electrons far from the Al2O3/GaN interface as well as reduce the leakage current along the trench sidewalls of GaN-MSND as shown in Fig. 7(a).

Fig. 6. Reverse characteristics of GaN-MSND with SiO2 with a positive fixed charge density of 3 × 1011 cm−2 on GaN/SiO2 interface and with Al2O3 with a negative fixed charge density of 2 × 1012 cm−2 on GaN/Al2O3 interface.
Fig. 7. Current density distribution for GaN-MSND with (a) Al2O3 and (b) SiO2.

Conversely, the electron accumulation region is formed near the trench surface due to the positive Qf at the SiO2/GaN interface, which is the main factor of increasing the surface leakage along the sidewalls and reducing the Vr as shown in Fig. 7(b). Therefore, we can conclude that the introduction of 0.5 μm-Al2O3 as the dielectric layer of GaN-MSND can increase Vr to 270 V and the width of depletion region W to 6.2 μm as shown in Fig. 8.

Fig. 8. Electric field distribution for GaN-MSND with Al2O3 at Vr of 270 V.
3.3. Influence of p-type inversion region on W of GaN-MSND

The introducing of p-type inversion layer is another way to increase Vr and W as shown in Fig. 9. The thickness of p-type inversion region is set to be 0.5 μm. The reverse characteristics of GaN-MSND with a p-GaN carrier concentration Np of 1 × 1016 cm−3, 3 × 1016 cm−3, 5 × 1016 cm−3, 7 × 1016 cm−3, 9 × 1016 cm−3, and 1 × 1017 cm−3 are illustrated in Fig. 10(a). The Vr values are obtained as follows: 306 V, 436 V, 595 V, 782 V, 1000 V, and 1112 V, which indicates a significant enhancement in Vr compared with the former results. Besides, with the increase of Np, the improvement of W is significant as well, as demonstrated in Fig. 10(b). When Np is increased to 1 × 1017 cm−3, W is enhanced up to 12.8 μm, which is more than twice the W of GaN-MSND with Al2O3 layer.

Fig. 9. Schematic drawing of GaN-MSND with p-type inversion region.
Fig. 10. (a) Reverse characteristics of GaN-MSND with Np of 1 × 1016 cm−3, 3 × 1016 cm−3, 5 × 1016 cm−3, 7 × 1016 cm−3, 9 × 1016 cm−3, and 1 × 1017 cm−3 and (b) relationship between Vr and Np, and W and Np for GaN-MSND with p-type inversion region.

Figure 11(a) shows the potential distribution of GaN-MSND with p-type inversion region. The p-type inversion region and the i-GaN form new p/n junctions and due to the continuity of potential, the potential is re-arranged and extended to the initially un-depleted region in i-GaN,[32] thus improving the width of depletion region. Besides, a lower electric field peak exists at the main junction with the assistance of the extension of depletion region as demonstrated in Fig. 11(b), which suggests an enhancement in Vr as well as W. The inset of Fig. 11(b) shows the relationship between Np and electric field at the main junction. It should be noted that the electric field peak decreases as Np increases, and hence, Vr increases accordingly.

Fig. 11. (a) Potential distribution for GaN-MSND with Np of 1 × 1017 cm−3 at reverse voltage of 1000 V, and (b) electric field profile for GaN-MSND with d of 0.5 μm and Np of 3 × 1016 cm−3 at x = 32.5 μm when reverse voltage is 200 V. The inset shows the main junction electric field varying with Np at reverse voltage of 200 V for GaN-MSND.

The GaN-MSND with p-type inversion region can significantly improve Vr and W, but the vertical p-type inversion region on the sidewall of GaN-MSND is not easy to realize in the device process. Considering the achievability of the process, the sidewalls are further etched into trapezoids so that the p-type region could easily be formed by ion implantation. The GaN-MSND, of which Np is 1 × 1017 cm−3, with the trapezoidal sidewalls is simulated and its Vr and W are improved to 1342 V and 13.6 μm as shown in Fig. 12, which may be attributed to the larger contact area between p-type inversion region and i-GaN.

Fig. 12. Reverse characteristic of GaN-MSND with trapezoidal p-type inversion region of which Np is 1 × 1017 cm−3, with the inset showing its potential distribution at reverse voltage of 1342 V.
3.4. Comparison among three methods to improve W

The transport distance of the charged particles in GaN depends on their energy. Upon absorbing a neutron, the 6LiF fissions into two reaction products, i.e. an alpha particle (2.06 MeV) and a 3H particle (2.73 MeV).[12] Since the energy loss per unit distance of alpha particle in GaN is larger than that of 3H, only the alpha particles incident into the detector is considered, while 3H is supposed to go throughout the device. The electron–hole pairs are generated as the alpha particles pass through the depletion region and drift towards respective electrodes by the electric field. The alpha particles with 2.06 MeV have a transport distance of approximately 5 μm in GaN. Therefore, for a given d of 5.5 μm, W should be at least 10.5 μm so that the detector can collect the alpha particles generated at the bottom of the trenches. As mentioned above, the GaN-MSND without additional design has a maximum W value of 4.7 μm by modulating the trench depth, while adding Al2O3 layer could increase W value to 6.2 μm. In contrast, the GaN-MSND with p-type inversion region has a largest W value of 12.8 μm and thus, it could be the most appropriate method to fabricate the detector with high detection performance.

4. Conclusions

In this work, three methods of improving W value of GaN-MSND is proposed and analyzed. For GaN-MSND with d value of 5.5 μm, W value increases to 4.7 μm at a Vr value of 226 V. The W value is further improved to 6.2 μm for GaN-MSND with the Al2O3 layer at Vr of 270 V. Moreover, GaN-MSND with the p-GaN inversion region can have a W value of 12.8 μm when Np is 1 × 1017 cm−3 at Vr of 1112 V. This study not only provides a series of pathways to obtaining GaN-MSND with better detection performance through the enhancement of the depletion region but also points out the GaN-MSND applications in the field of neutron detection in intense radiation environment.

Reference
[1] Ochs T R Bellinger S L Fronk R G Henson L C Huddleston D E Lyric Z I Shultis J K Smith C T Sobering T J McGregor D S 2017 IEEE Trans. Nucl. Sci. 64 1846
[2] McGregor D S Bellinger S L Shultis J K 2013 J. Cryst. Growth 379 99
[3] Shao Q Voss L F Conway A M Nikolic R J Dar M A Cheung C L 2013 Appl. Phys. Lett. 102 063503
[4] Huang K C Dahal R Lu J J Q Danon Y Bhat I B 2013 Appl. Phys. Lett. 102 152107
[5] Voss L F Reinhardt C E Graff R T Conway A M Nikolić R J Deo N Cheung C L 2010 J. Electron. Mater. 39 263
[6] Millan J Godignon P Perpina X Perez-Tomas A Rebollo J 2014 IEEE Trans. Power Electron. 29 2155
[7] L Zhang J C Xue J S Ma X H Zhang W Bi Z W Zhang Y Hao Y 2012 Chin. Phys. 21 037104
[8] Mulligan P Qiu J Wang J H Cao L R 2014 IEEE Trans. Nucl. Sci. 61 2040
[9] Shur M 2019 Solid-State Electron. 155 65
[10] Xu K Wang J F Ren G Q 2015 Chin. Phys. 24 066105
[11] Sellin P J Vaitkus J 2006 Nucl. Instrum. Methods Phys. Res., Sect. 557 479
[12] Mulligan P Wang J H Cao L 2013 Nucl. Instrum. Methods Phys. Res., Sect. 719 13
[13] Melton A G Burgett E Xu T M Hertel N Ferguson I T 2012 Physica Status Solidi 9 957
[14] Atsumi K Inoue Y Mimura H Aoki T Nakano T 2014 APL Mater. 2 032106
[15] Wang J H Mulligan P Brillson L Cao L R 2015 Appl. Phys. Rev. 2 031102
[16] Morkoc H 2001 Mater. Sci. Eng. R-Rep. 33 135
[17] Vaitkus J Cunningham W Gaubas E Rahman M Sakai S Smith K M Wang T 2003 Nucl. Instrum. Methods Phys. Res., Sect. 509 60
[18] Owens A Barnes A Farley R A Germain M Sellin P J 2012 Nucl. Instrum. Methods Phys. Res., Sect. 695 303
[19] Nikolic R J Cheung C L Reinhardt C E Wang T F 2005 Optoelectronic Devices: Physics, Fabrication, and Application II 24 October, 2005 Boston, USA 36 10.1016/j.nima.2011.09.003
[20] Xu Q Mulligan P Wang J H Chuirazzi W Cao L 2017 Nucl. Instrum. Methods Phys. Res., Sect. 849 11
[21] Wang G Fu K Yao C S Su D Zhang G G Wang J Y Lu M 2012 Nucl. Instrum. Methods Phys. Res., Sect. 663 10
[22] Ye X Xia X C Liang H W Li Z Zhang H Q Du G T Cui X Z Liang X H 2018 Chin. Phys. 27 087304
[23] Lu M Zhang G G Fu K Yu G H 2010 Chin. Phys. Lett. 27 052901
[24] Sze S M Ng K K 2006 Physics of semiconductor devices 3 Hoboken John Wiley & Sons 83
[25] Zhu Z F Zhang H Q Liang H W Tang B Peng X C Liu J X Yang C Xia X C Tao P C Shen R S Zou J J Du G T 2018 Nucl. Instrum. Methods Phys. Res., Sect. 893 39
[26] Xie G Xu E Hashemi N Zhang B Fu F Y Ng W T 2012 Chin. Phys. 21 086105
[27] Truyen N X Taoka N Ohta A Makihara K Yamada H Takahashi T Ikeda M Shimizu M Miyazaki S 2018 Jpn. J. Appl. Phys. 57 04FG11
[28] Ohta A Truyen N X Fujimura N Ikeda M Makihara K Miyazaki S 2018 Jpn. J. Appl. Phys. 57 06KA08
[29] Cho Y J Chung K B Chang H S 2018 Thin Solid Films 649 57
[30] Jackson C M Arehart A R Cinkilic E McSkimming B Speck J S Ringel S A 2013 J. Appl. Phys. 113 204505
[31] Biswas D Torii N Fujita H Yoshida T Kubo T Egawa T 2019 Semicond. Sci. Technol. 34 055014
[32] Udrea F Deboy G Fujihira T 2017 IEEE Trans. Electron Dev. 64 713