Low dark current far infrared detector with an optical cavity architecture

Size: px

Start display at page:

Download "Low dark current far infrared detector with an optical cavity architecture"

Hester Palmer
5 years ago
Views:

1 Solid-State Electronics 45 (2001) 87±93 Low dark current far infrared detector with an optical cavity architecture A.L. Korotkov a, A.G.U. Perera a, *, W.Z. Shen a,1, H.C. Liu b, M. Buchanan b a Department of Physics and Astronomy, Georgia State University, Atlanta, GA 30303, USA b Institute for Microstructural Sciences, National Research Council, Ottawa K1A 0R6, Canada Received 25 August 2000; accepted 25 September 2000 Abstract Here we report designs for performance improvements of homojunction interfacial workfunction internal photoemission (HIWIP) detectors for di erent far infrared regions. A design is given to reduce dark current to about 10 e/s for a 300 lm cut-o detector at 1.3 K, at a bias eld of 500 V/cm by adjusting the thickness of the intrinsic layer to eliminate tunneling component. The intrinsic region thickness and the bottom contact are used to obtain an optical cavity e ect thus increasing the absorption to almost 100%. Increased responsivity due to the optical cavity e ect is experimentally veri ed by using three detector structures, with three di erent cavity structures. Ó 2001 Elsevier Science Ltd. All rights reserved. Keywords: Far infrared detector; Homojunction interfacial workfunction internal photoemission; Cavity e ect High performance far infrared (FIR) (40±200 lm) semiconductor detectors as well as large focal plane arrays are required for space astronomy applications, such as NASA's airborne mission, Stratospheric Observatory for Infrared Astronomy (SOFIA), and the ESA's Far-infrared and Sub-mm Telescope (FIRST) programs. Si or GaAs homojunction interfacial workfunction internal photoemission (HIWIP) FIR detectors can compete with extrinsic Ge photoconductors (unstressed or stressed) and Ge block-impurity-band (BIB) detectors due to the material advantage of Si or GaAs over Ge [1]. The detection mechanism [1] of HIWIP detectors involves absorption in the highly-doped emitter layers mainly by free carriers followed by the internal photoemission of photoexcited carriers across the junction barrier and then collection. The cut-o * Corresponding author.fax: address: uperera@gsu.edu (A.G.U. Perera). 1 Present address: Department of Applied Physics, Shanghai Jiao Tong University, 1954 Hua Shan Road, Shanghai , People's Republic of China. wavelength k c is determined by the interfacial barrier height between the emitter layers and undoped intrinsic layers. Signi cant progress has already been achieved in the development of p-gaas HIWIP FIR detectors [2], resulting in a responsivity p of 3.1 A/W and detectivity of 5: cm Hz /W, with kc as long as 100 lm. Experimental and calculated results were presented [3] for GaAs HIWIP detectors with relatively thin (W i ˆ A) i-regions. Typical current noise spectra demonstrated the 1=f behavior at frequencies below 1.5 khz. It is shown that the noise power density (S i ) in HIWIP depends strongly on the dark current (I d ) through the detector: S i f ˆCId 2=A jfn is, where C 0:1, A j is the device area, and N is ±10 11 cm 2 is the interface state density [4]. The HIWIP structures with the thin i-region had relatively high dark current. For example, a 1000 A thin i-region, and a 100 lm cut-o detector with area of cm 2, will have 10 6 A dark current at 4.2 K under a eld of 500 V/cm [5]. Reducing the temperature did not lower the dark current indicating the dominance of tunneling. Introducing thicker i-regions to reduce I d will shift the barrier maximum position away from the interface due to the space /01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved. PII: S (00)

2 88 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 charge (SC) e ects [6]. Hence a tradeo between dark current and the quantum e ciency is expected due to increased i-region thickness of the detector. This article studies the e ect of layer thickness in reducing dark current, and increasing the absorption using an optical cavity architecture. A typical p ±i (or n ±i HIWIP structure consists of a heavily doped emitter layer, the intrinsic (or lightly doped) layer, the bottom contact layer, having thicknesses W e, W i, and W b, with corresponding doping concentrations N e, N i, and N b. Several units of emitterintrinsic layers could be present (for the structures studied so far), as seen in Fig. 1. For the structure with number N of i±p repeats and for the thickness of the top emitter layer W te the total thickness is W T ˆ W bc W bi N W e W i W te. It is known that the resonant cavity enhanced (RCE) devices [7] utilize the re ection of the incoming radiation from the bottom mirror to increase the optical eld intensity in the active region of the detector, increasing the absorption leading to improved quantum e ciency. It is an attractive idea to use a bottom contact layer as a ``mirror'' re ecting the radiation which is not absorbed by the emitter/absorber in the rst pass through giving rise to a simple cavity architecture. Increasing the thickness and doping concentration of the bottom contact layer will increase the re ection from it. The FIR absorption is calculated from the complex permittivity of each layer by matching the electric and magnetic elds at the interfaces. Far away from the reststrahlen region the frequency-dependent permittivity of the highly doped emitter layer can be considered as [8] ˆ s x 2 p 1 x x ix 0! Eq. (1) describes the free-carrier absorption in the frame of Drude model. Here x is the optical frequency, x 0 ˆ 1=s is the free-carrier damping constant with relaxation time s (which is independent of frequency in semiclassical transport theory), x p ˆ N e q 2 = 0 s m 1=2 is the plasma frequency, and s is the low frequency dielectric constant of an intrinsic semiconductor, m ˆ 0:5m 0 is the heavy-hole e ective mass in GaAs (in one-band model), m 0 is the free electron mass, and q is the magnitude of the electron charge. The carrier concentration N e ˆ cm 3 can be estimated from the doping level, while the relaxation time s s was measured for a sample with similar doping level [9]. Reststrahlen absorption can be taken into account by modifying Eq. (1) with speci c term [8]. The substrate and the i-regions were considered transparent faraway from the reststrahlen region with permittivity s. The permittivity of the bottom contact layer can be calculated by Eq. (1) with using the appropriate concentration N b. By using the above expressions, absorption, re ection, and transmission coe cients for GaAs p -i structures have been calculated. As expected, the absorption maxima for the structure with one emitter are realized at wavelengths corresponding to k s 1 2n =4 distance from the bottom contact layer where k s is the wavelength inside the media, with n ˆ 0; 1; 2;... Optical eld distribution inside the structure for a normal incident electro-magnetic wave with wavelength corresponding to the rst absorption maximum (n ˆ 0), is shown in Fig. 2. It is seen that the absorption maximum is reached 1 Fig. 1. The schematics of the tested multilayer HIWIP detector. p is the contact layer, p the emitter layer, and i the undoped layer. A window is opened on the top side for frontside illumination. The projection shows the N-multilayers, replaced by a thicker single i- region.

3 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 89 at W T k s =4 condition, when the top emitter layer is in the maximal optical electric eld. The absolute value of absorption coe cient depends on the optical thicknesses of the emitter and bottom contact layers. The doping concentration is chosen to give rise to an appropriate detector interfacial work function D, and W e should be of the order of inelastic scattering length L z 200±300 A (for p-gaas) [10] for optimum collection e ciency. The interaction with light depends strongly on the carrier concentration and the thickness of the active layers. In the FIR range, GaAs with a doping density p of cm 3 has a skin depth d ˆ c=xim of the order of 1 lm. Hence a typical 4000±7000 A, bottom contact region doped with cm 3 will re ect back a signi cant fraction of the incident radiation to form a standing wave greatly reducing the electric eld at the bottom contact. Since the emitter layers are optically thin (i.e. W e k s ; d), the absorption depends on the amplitude of electric eld and hence on the distance from the bottom contact layer which plays the role of the ``mirror''. Several detector structures were designed, grown (by MBE), and tested successfully. Measured re ection from the etched structures, nos. 9605, 9601, and 9604, at 300 K, and calculated re ection using Eq. (1) modi ed with the ``reststrahlen'' term, [8] is shown in Fig. 3. Parameters for these three devices are given in Table 1. The tting parameters s used were: 1: and 1: s for contact concentrations 2: and 1: cm 3, respectively. For emitter regions s was 2: ,2:010 14, and 1: s for emitter concentrations 5: , , and 8: cm 3, respectively. All the structures demonstrated signi cant resonant cavity e ects due to re ection from the bottom contact layer with spectral features corresponding to their geometrical parameters. The almost periodic structure of a re ection coe cient as a function of wave number is seen in Fig. 3. The period is scaled approximately as 1=W T, where W T is the total growth thickness or the distance from the top highly doped region (absorber region) to the bottom contact layer (re ection surface). The period is shorter for structures with greater numbers of internal emitters, because W T increases with N. Re ection from the bottom contact layer will be weaker for lower doping concentrations, and hence the interference e ect will be weaker. This is seen in the weaker oscillations for no structure with a bottom contact concentration of 1: cm 3, compared to the other structures, nos and 9604, with a bottom contact concentration of 2: cm 3. Fig. 2. Optical electric eld distribution in the p-gaas HIWIP structure illuminated from the top. The IR wavelength k ˆ 68 lm corresponds to the maximum absorption in the emitter layer. The structure is shown above the graph with the arrows indicating the 300 A thick top emitter region. Calculations for the 2: cm 3 p-doped emitter layer with a 4 lm thick intrinsic region, and a 7000 A thick 2: cm 3 p-doped bottom contact layer. Optimal eld at the top emitter is reached when the structure thickness W T is close to k s =4. The standing wave in the substrate behind the bottom contact is formed due to re ection from its back side (indicated by the right dashed vertical line). Increasing the optical thickness of the bottom contact is necessary to decrease its transparency and in turn to increase the absorption in the emitter.

4 90 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 Fig. 3. Calculated (- - -) and measured (Ð) re ection for Be-doped multilayer detector [2] structures: (a) no with only top emitter, (b) no with 10 periods, and (c) no with 20 periods of i-region-emitter. Re ection faraway from the reststrahlen region is almost periodic function of wave number due to Fabry±Perot resonator formed by the bottom contact layer and the top highly doped region. This period is inverse proportional to the thickness W T and increases with reducing the number of i±p+ repeats (from the bottom to the top). Table 1 Parameters for three p-gaas HIWIP FIR detector structures with thin i-regions. Measured (from spectrum) interfacial work-function D, peak quantum e ciency g p, noise equivalent power NEP, peak responsivity R p, and cut-o wavelength k c. The valence-band-edge o set is given by DE v ˆ D E F. Here, W i, W e, and W bi are the thicknesses of the intrinsic (i), emitter (p ), and bottom intrinsic (i), respectively. N e and N c are the doping concentrations of the emitter and contact layers, respectively. W e was 150 A. The thickness of the top contact (p ), the top emitter (p ) and bottom contact (p ) layers, W tc, W te and W bc were 3000, 3000, and 4000 (3500 for no. 9604) A, respectively, for three samples. Then the top contact layer was etched out completely, and the top emitter layer was also partially etched out. The thickness of the top emitter in the etched structures was 600 A. W T is the total thickness. Sample number Number of repeats N W T (lm) W i ( A) W e ( A) N e (10 18 cm 3 ) D(meV) g p % NEP p (10 12 R p A=W k c (lm) W/ Hz ) ± ± The increase in the response with the number of emitter layers which was experimentally demonstrated for 1, 11, 21 emitter (N ˆ 0, 10, 20) structures with a total thickness (W T ) of 0.91, 1.71, and 2.16 lm. These structures had measured responsivity (at 50 lm) of 0.12, 1.51, and 3.10 A/W, showing that the response increases with the number of layers [2]. The detector with the lowest W T value shows a lower responsivity than

5 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 91 Fig. 4. Calculated absorption in emitter layer and re ection from the structure as a function of wavelenght for two di erent intrinsic region thicknesses, and the e ect of a top coating to reduce the intrinsic region thickness. (a) 4 lm thick intrinsic region, (b) 14 lm thick intrinsic region, (c) 4 lm thick intrinsic region and 10 lm thick dielectric (e ˆ 13) coating above the top emitter. 300 A thick 2: cm 3 p-doped emitter layer with a 5 lm thick bottom contact W b region. Increasing the optical thickness of bottom contact results in a reduction of absorption in it increasing the absorption in active (emitter) layer. A thick, highly doped bottom contact can be formed on the substrate surface by ion-implantation before MBE growth. Dielectice coating as well as an increase of the i-region thickness result in the shift of absorption peak to longer wavelengths. A top coating for a structure with 4 lm i-region with a 10 lm dielectric layer moves the absorption maximum to the same position as a 14 lm i-region uncoated structure with reduced absorption. Fig. 5. Dark current (thin lines with the left vertical axis) and internal quantum e ciency (thicker dashed lines with the right vertical axis) for the 4 lm i-region GaAs HIWIP. For a detector with a doping concentration in the emitter N e ˆ 2: cm 3, a quantum e ciency of 2.0% can be achieved for 100 lm IR radiation with a bias eld of 500 V/cm, with a 1.3 K calculated dark current A (less than 100 e/s). For the same parameters, 150 and 200 lm quantum e ciency will be 0.7% and 0.3%, respectively.

6 92 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 expected from the calculation indicating the e ect of bias eld redistribution inside the illuminated multilayer structures [11]. For multilayer samples with thicker i-regions (W i k s =4) it is expected that each emitter layer will give a response maxima at its own resonant wavelength due to the distance from the bottom contact layer. For example, 300 A thick, 2: cm 3 p-doped emitter regions at distances 14, 9, and 4 lm from the bottom contact, will have corresponding absorption maxima at 180, 160, and 70 lm, respectively. This will require i-region thicknesses of 14, 9, and 4 lm, which can be grown by MOCVD or CBE, improving absorption and also giving a broader wavelength spectrum. A highly doped bottom contact can be formed by ion-implantation at the top surface of the substrate before MBE/MOCVD/CBE growth or by using a highly doped substrate wafer. It will allow the reduction of the total epitaxial growth thickness while increasing the absorption in the emitter layer. Further reduction of the total growth thickness is possible by using a dielectric coating above the top emitter layer. As it is shown in Fig. 4, such a coating shifts the absorption maximum to longer wavelengths as similar to increasing the i-region thickness W i. Although increasing W i is more e ective (absorption at 200 lm is80% compared to 40% in the case of coating), the dielectric coating is a simple way to tune the spectral response and to design an array for multiwavelength operation without growing very thick i-regions. As indicated in our previous work [6] the i-layer, (unintentionally doped p-type) contains fully ionized compensating donors. If the i-region is rather thick, the ionized acceptor concentration (N ai ) is constant and equal to the compensating donor concentration (N di )at low temperatures. In the emitter vicinity some ionized acceptors will be neutralized by the excess holes coming from p layer, resulting in a distribution of xed positive space charge. The dark current calculation is based on a simple model using uniform space charge (constant N di ). The parameters used are: N di ˆ cm 3, acceptor activation energy E a ˆ 30 mev, and junction area A j ˆ cm 2. As can be seen in Fig. 5, 10 ± 100 e/s dark current can be obtained for a 4 lm thick i-region at 1.3 K. Despite the low absolute value of g, high performance of GaAs HIWIPs can be reached with almost Fig. 6. (a) Calculated spectra for three p-gaas single emitter detectors (shown in Fig. 1a) with the same i-region thickness W i ˆ 4 lm. and di erent emitter doping concentrations N e : cm 3 ; 2 1: cm 3 ; and (3) 2: cm 3. Other parameters are the same for all three case: W e ˆ 100 A,W s ˆ 350 lm, W b ˆ 7000 A,N b ˆ 2: cm 3, bias electric eld F ˆ 500 V/cm. (b) Calculated spectra for three p-gaas detectors with N e ˆ 2: cm 3, 500 V/cm electric eld, giving a 300 lm k c, with di erent i- region thicknesses W i : (1) 4, (2) 6 and 3 14 lm. Only the rst resonance peaks corresponding to the longest wavelenghts are shown. (c) Calculated noise equivalent power at 500 Hz for the 4 lm i-region at (1) T ˆ 1:7 K, and (2) T ˆ 1:3 K, N e ˆ 2: cm 3 ; W e ˆ 30 nm, junction area A j ˆ cm 2 ; F ˆ 500 V=cm.

7 A.L. Korotkov et al. / Solid-State Electronics 45 (2001) 87±93 93 complete absorption and operating at temperatures where very low dark current is expected. For example, a detector with k c ˆ 300 lm with a 100 e/s dark current operating at 1.3 K will have noise equivalent power p (NEP) associated with internal noise, of W/ Hz. Further improvements could be obtained by controlling the barrier shape which is not addressed here. Calculated responsivity for the structure shown in the projection in Fig. 1 (N multilayers replaced with a single intrinsic region) with di erent N e and W i is shown in Fig. 6 a and b. As N e is increased from to cm 3, k c can be increased from 100 to above 300 lm. k c can also be in uenced by the applied electric eld (F) due to the Schottky e ect. For N e ˆ cm 3, k c increases from 100 to 300 lm as bias eld F increases from 100 to 500 V/cm. The NEP using calculated noise density for a GaAs HIWIP with a 4 lm thick i-region is shown in Fig. 6c. It is considered that noise power density is determined by the dark current as S i f ˆ CId 2=A jfn is [4]. Parameters used in the calculations are: N b ˆ cm 3, m ˆ 0:47m 0, x 0 ˆ s 1, W e ˆ 300 A, W b ˆ 7000 A, W s ˆ 350 lm, and s ˆ 13. In summary, the HIWIP detector characteristics, such as responsivity, quantum e ciency, and NEP, have been investigated in multilayer p-gaas HIWIP FIR detectors with di erent distances W T from the top emitter to the bottom contact layer. The detector with maximal W T ˆ 2.16 lm, which is still less then k s =4 had the best performance with R p ˆ 3.1 A/W, and NEP ˆ W = Hz p at 4.2 K, for k c from 80 to 100 lm. It is shown that the use of thicker i-regions in the detector design will result in an increase in absorption due to the optical cavity e ect and dark current reduction. For a p-gaas single emitter detector with N e ˆ cm 3, W i ˆ 4±14 lm absorption in the emitter layer could reach the value of 35±80%, with a very low dark current I d AatT ˆ 1:3 K. The use of a thicker i-region shifts the absorption maximum to longer wavelengths. k c can be extended up to 200±300 lm with the emitter layer concentration in the order of cm 3. Acknowledgements This work was supported in parts by the NASA under contract No. NAG5-4950, and DND. Authors acknowledge the contribution made by W. J. Scha at Cornell in sample growth. References [1] Perera AGU. In: Francombe MH, Vossen JL, editors. Physics and novel device applications in semiconductor homojunctions in physics of thin lms, vol. 21. NY: Academic Press; p. 1±75. [2] Perera AGU, Shen WZ. GaAs homojuction interfacial workfunction internal photoemission (HIWIP) far infared detectors. Opto-Electron Rev 1999;7:153. [3] Perera AGU, Yuan HX, Gamage SK, Shen WZ, Francombe MH, Liu HC, Buchanan M, Scha WJ. GaAs multilayer p ±i homojunction far infrared detectors. J Appl Phys 1997;81:3316. [4] Shen WZ, Perera AGU. Low frequency noise and interface states in GaAs homojunction far infrared detectors. IEEE Trans Electron Dev 1999;46:811. [5] Shen WZ, Perera AGU, Francombe MH, Liu HC, Buchanan M, Scha WJ. E ect of emitter layer concentration on the performance of GaAs p ±i homojunction far infrared detectors: A comparison of theory and experiment. IEEE Trans Electron Dev 1998;45:1671. [6] Yuan HX, Perera AGU. Space charge analysis of Si n ±i structures with application to far infrared detectors. Solid- State Electron 1996;39:621. [7] Unlu MS, Strite S. Resonant cavity enhanced photonic devices. J Appl Phys 1995;78:607. [8] Blakemore JS. Semiconducting and other major properties of gallium arsenide. J Appl Phys 1982;53:R123. [9] Huberman ML, Ksendzov A, Larsson A, Terhune R, Maserjian J. Optical absorption by free holes in heavily doped GaAs. Phys Rev B 1991;44:1128. [10] Maserjian J. Long-wave infrared (LWIR) detectors based on III-V materials. Proc SPIE 1991;1540:127. [11] Ershov M, Liu HC, Buchanan M, Wasilewski ZR, Ryzhii V. Photoconductivity nonlinearity at high excitation power in quantum well infrared photodetectors. Appl Phys Lett 1997;70:414.

Effects of a p n junction on heterojunction far infrared detectors

Infrared Physics & Technology 50 (2007) 274 278 www.elsevier.com/locate/infrared Effects of a p n junction on heterojunction far infrared detectors S.G. Matsik a, *, M.B.M. Rinzan a, A.G.U. Perera a, H.H.