Extrinsic Voltage Control of Carrier Lifetime in Polycrystalline PbSe Mid-wave IR Photo Detectors for Increased Detectivity
Samiran Ganguly, Tang Xin, Sung-Shik Yoo, Philippe Guyot-Sionnest, Avik W. Ghosh
EE XTRINSIC V OLTAGE C ONTROL OF C ARRIER L IFETIME IN P OLYCRYSTALLINE
P bSe M ID - WAVE
IR P
HOTO D ETECTORSFOR I NCREASED D ETECTIVITY
A P
REPRINT
Samiran Ganguly, Avik W. Ghosh
Charles L. Brown Dept. of Electrical and Computer Engineering, University of Virginia, Charlottesville, VATang Xin, Philippe Guyot-SionnestDept. of Chemistry and Dept. of Physics, University of Chicago, Chicago, ILSung-Shik YooNorthrop Grumman Systems Corp., Rolling Meadows, ILJuly 7, 2020 A BSTRACT
Polycrystalline
P bSe for mid-wave IR (MWIR) photodetector is an attractive material option dueto high operating/ambient temperature operation and relatively easy and cheap fabrication process,making it candidate for low-power and small footprint applications such as internet-of-thing (IoT)sensors and deployment on mobile platforms due to reduced/removed active cooling requirements.However, there are many material challenges that reduce the detectivity of these detectors. In thiswork, we demonstrate that it is possible to improve upon this metric by externally modulating thelifetime of conducting carriers by application of a back-gate voltage that can control the recombinationrate of generated carrier. We first describe the physics of
P bSe detectors, the mechanisms underlyingcarrier transport, and long observed lifetimes of conducting carriers. We then discuss the voltagecontrol of these inverted channels using a back-plane gate resulting in modulation of the lifetime ofthese carriers. This voltage control represents and extrinsic “knob” through which it may be possibleto open a pathway for design of high performance IR photodetectors, as shown in this work.
Mid-wave Infrared (IR) atmospheric detection window ( − µm of EM spectrum) has been useful for applicationsin astronomy and earth sciences [1, 2] as well as military and surveillance [3]. Integration of imaging sensors incommercial civilian applications such as self-driving automotives [4] may expand the range of application of suchsensors many-fold. These new classes of sensors will need to have high resolution and performance without activecooling to reduce energy and production cost, and increase compactness and integrability with other systems.Therefore, there is a need for novel designs of photo-detectors that can meet such targets. P bSe is a century old material[5, 6] that only has a niche market in mid-wave IR, compared to HgCdTe detectors, even though its cost of productionis low and is easily integrable with a Si based read-out integrated circuit (ROIC) [7]. This is due to relatively lowerperformance and higher device-to-device variability in a large focal plane array. Better controlled fabrication protocolshelps in mitigating these issues to an extent, however, it is expedient to explore extrinsic electrical control that a ROICcan provision to control these factors.In these detector the central working principle is modification of the detector film resistivity in presence of incidentphotons through generation of new charge carriers in the film [8] and the “signal” depends on the difference betweenthe background doping (dictates dark current) and the resulting carrier concentration (dictates lit current). The numberof photo-generated carrier above the background doping can be improved in a few ways. One approach is throughplasmonic enhancement of absorption (see e.g.[9, 10, 11]) which can improve the quantity of photo-generated current.Another approach is through appropriate choice of blocking contact material [12] that can recycle the trapped electrons a r X i v : . [ phy s i c s . a pp - ph ] J u l PREPRINT - J
ULY
7, 2020over many holes which enhances the total number of photo-generated carriers seen by the ROIC. In this work, wediscuss another technique focusing purely on enhancing the lifetime τ of the photo-generated holes by reduction oftheir recombination in the film through a back-gate voltage.In this paper we first briefly discuss the detector fabrication and characterization setup. These were performed atthe Northrop Grumman Corp. and University of Chicago. Then we discuss the physics of these detectors and theconnection between the carrier lifetime with detector performance. We also provide a benchmark against our analyticalmodel for the device against the fabricated device’s electrical characteristics. Then we build a semi-analytical modelfor lifetime vs. the back-gate voltage and benchmark against the experimental device. In doing so we also build asemi-empirical model of carrier lifetime vs. temperature in these detectors. Finally we provide analysis of performanceimprovement seen by the experimental detector due to back-gate voltage and design optimizations that can translatethe small signature of performance improvement in the experimental device into substantial increase in well-designeddetectors.Figure 1: a. Schematic of the back-plane gated P bSe photodetector. The gate voltage is applied through the Si | SiO substrate, while source and drain contacts carry dark and photo-generated current to a ROIC. b. Plan view image of theactual patterned detector with Au leads. c. Surface micrograph of the film showing its polycrystalline nature. P bSe films were fabricated using the standard liquid epitaxy method commonly called the chemical bath deposition,where the organic precursors containing
P b and Se are allowed to chemically combine and precipitate on the substrate.This film is taken through two further steps of annealing, once in an Oxygen environment, and then in an Iodineenvironment. These steps increase the film’s light sensitivity significantly and Hall measurements find the film to haveundergone carrier inversion and significant reduction in mobility. The top half of the film also converts to P bI whichact as a natural capping layer, but also enables high photo-sensitivity as it has been observed and has been part of P bSe detector fabrication for nearly four decades [13].The sensitized
P bSe films on
SiO | Si substrate were then patterned into squares to define the active sensing area(fig. 1b). Photoresist AZ703 was used to define and cover the sensing area by UV lithography, followed wet chemicaletching in hydrogen peroxide bath for ∼ min . After patterning, the residual photoresists were removed by rinsing inacetone and isopropanol. Electrodes were defined by photolithography and deposition of nm T i and nm Au bye-beam evaporation. Lift-off process was conducted in acetone bath for min , dried by nitrogen blow.Blackbody source with controlled temperature was used as the light source in current-voltage ( I d − V d ) and fieldeffect transistors ( I d − V g ) measurements. The photocurrent was first amplified with a trans-impedance amplifier(DLPCA-200, Femto) and then amplified by a voltage amplifier. The applied bias voltages can be tuned by the built-inbias of the transimpedance amplifier. The amplified photocurrents were digitalized and recorded by a data acquisitioncard. 2 PREPRINT - J
ULY
7, 2020
Previously we [8] and others [14, 15] have discussed the critical role that is played by the interfacial depletion fieldbetween
P bSe and
P bI in enhancing the lifetime of the conducting holes in the photo-sensitive P bSe detector films.These depletion film formed at the interface of the
P bSe | P bI separate the photo-generated excitons in to electronsand holes. The electrons are trapped within the grain boundaries rich in P bO ++ species, while holes are available forconduction. In these polycrystalline films the transport of the conducting carriers, in this case the holes, can be modeledas the hopping from one conducting site to the other over the barriers. We also derived an expression for the currentdensity-applied voltage relationship through these films in the photo-conductive mode given as: | J | = 2 q ( p + Gτ /t film ) µ p exp( − qφ b k B T ) sinh( qV d N b k B T ) φ b t b (1)In this equation J, p , G, τ, t film , φ b , V d , N b , µ p , t b , T respectively are the current density magnitude, intrinsic carrierconcentration, photo-carrier generation rate, hole lifetime, film thickness, average inter-crystalline barrier height, appliedvoltage bias, total number of barriers presented by the film in the transport direction, average barrier thickness, holemobility, and temperature of the detector. q, k B are the usual physical constants they represent. The parameters N b , t b can be obtained from imaging techniques such as scanning electron microscopy where average crystallite dimensionsand inter-crystallite distances can be estimated.It can be seen from eq. 1 that the photo-generated carriers increase the conductivity of the film and therefore it presentsa lower resistance, that can be measured by the ROIC by measuring the RC delay of a network formed using thedetector and a controlled capacitor. The generation rate G depends on the quantum efficiency, i.e. the efficiency ofconversion of light into carriers and the light absorption.Two important metrics for photo-detectors are responsivity and specific detectivity, which measure photo-signal toelectrical-signal transduction and signal-to-noise ratio and are given by: R = I photo P optical ∝ GµτP optical (2) D ∗ ∝ I photo I noise ∝ GµτI noise (3)Where I photo , P optical , G, I noise stand for generated photo-current (given by: I photo = I lit − I dark ), incident opticalpower, carrier generation rate, and noise current in the detector. It can be seen that all these three metrics depend on thelifetime of the transporting carriers. Therefore, increasing the carrier lifetime in critical to obtain higher performancefrom such detectors.In ref. [8] we developed a simple model of the carrier lifetime ( τ ) in terms of partial derivative of the total recombinationrate R with electron and hole concentrations ( n, p ), and the fraction of depleted P bSe layer r surf : τ − ≈ ∂R∂ ∆ p + (1 − r surf ) ∂R∂ ∆ n (4)where the total recombination rate is sum of the partial rates from the recombination phenomena in the film, viz.direct/radiative ( P bSe is a direct bandgap material), Shockley-Reed-Hall (SRH), and Auger. The surface depletionfrom the
P bSe | P bI interface and its effect on the lifetime through suppression of non-conducting carrier, which iselectrons in this, and hence the factor r surf appears with the partial derivative w.r.t. electrons. We argue that by buildinga back gate into the detector structure, we can electrostatically control r surf and therefore control the lifetime of thedetector which results in increase in the metrics of performance: responsivity and detectivity.In fig. 2a we show the experimental electrical characteristics and noise in the fabricated detectors, and the analyticalmodel (eq. 1) as a circuit model also accounting for the contact resistances (fig. 2b) . It can be seen that we can veryclosely match the I − V characteristics (fig. 2c), film resistance (fig. 2d) as well as the noise magnitude (fig. 2e) usingthe model (parameters listed in tab. 1) and is close to those used in ref. [8]. We measure the modulation of dark and lit current as a function of gate voltage for a constant drain bias (fig. 3a,b) anddevelop a numerical capacitor-divider based model (fig. 3c) for the electrostatics of the detector stack, by considering3
PREPRINT - J
ULY
7, 2020Figure 2: Benchmarking of detector characteristics a. Measured I d − V d characteristics, R , I noise b. Circuit model forthe detector as a parallel combination of controlled resistor and noise current source, with source and drain resistancesalso accounted for. c. Modeled I d − V d characteristics c. Modeled R − V d characteristics d. Modeled I noise − E d characteristics. Parameter Value Parameter Value
Area ( L × W ) × µm t P bSe . µmt P bI µm t ch ;0 nmt b nm w b . µmt ox . µm T − Kp . × /cm E g . eVE b meV µ (295 K ) 0 . cm /V sG . × /cm s P opt µW/cm ρ SRH (250 K ) 170 µs ρ Aug (250 K ) 155 µsκ SRH κ Aug . Table 1: Parameter list used in the back-gated PbSe detector model.a series network of the back-gate oxide,
P bSe , depleted layer, and
P bI layer. We then incorporate the effect ofthe applied gate voltage on the ratio of the depletion layer thickness to the P bSe layer thickness, assuming uniformdepletion along the plane of the detector film. This ratio gives us the r surf parameter for the eq. 4, thereby yielding asemi-analytical formulation for the lifetime of the carriers.However, the lifetime is also a function of temperature as at different temperature range different recombinationmechanisms dominate. In particular we focus on multi-particle processes such as Shockley Reed Hall (SRH) and Augerprocesses which are mediated by trapping centers and phonons respectively. At low temperatures, the trapping centersbecome “deeper”, as the thermal motion reduce which reduces the likelihood of a trapped particle to escape. Similarlyat high temperatures the phonons increase in number (by Bose-Einstein statistics) which increase their likelihood tocause Auger recombination. From experimental measurements of lifetime we observe that it shows a non-monotonicbehavior with temperature that can be explained by the heuristic presented above. We can therefore model this usingthe following phenomenological equation: τ SRH = ρ SRH T κ SRH (5) τ Aug = ρ Aug T − κ Aug (6)4
PREPRINT - J
ULY
7, 2020 τ = 1 τ SRH + 1 τ SRH (7)Where the ρ SRH , ρ
Aug , κ
SRH , κ
Aug are fitting parameters to match the experimental trends, τ SRH , τ
Aug , τ arelifetime contributions from SRH and Auger mechanisms as the total lifetime at gate-bias. We see a good fit (fig. 3d)with the observed non-monotonic trend of the lifetime with temperature. Using the calculated τ from the above model,we can recast the eq. 4 in the following form: τ = τ − [ t ch − t ch ;0 t PbSe − t ch ;0 ] γ (8)Where t ch , t ch ;0 , t P bSe are the depletion width at an applied gate-bias, built-in depletion width, and the
P bSe layerthickness respectively, γ is a fitting exponent that ranges from . to . in the range of temperatures measured. Thedepletion layer thickness can be calculated using the electrostatic model as discussed above.Figure 3: Benchmarking of back-gating effect a. Measured I d − V g characteristics at different temperatures. b. Measured I d − V g characteristics for lit and dark conditions at K . c. Measured vs. Modeled τ − T emp. characteristics. d.Modeled I photod − V g characteristics at different temperaturesAs per the eq. 4, the exponent γ should ideally be , but needs slight adjustment to account for simplifications built intothe model, e.g. ignoring direct recombination, and the effect of lifetime due to modulation of hole density (consideredto be 0) as holes are majority conducting-carriers in the depletion region and their concentration does not changesignificantly. Using 8 we can capture the experimentally measured trends of lifetime on back-gate bias quite closely (fig.3e). This demonstrates the validity of the back-gating approach in improving the lifetime of the carriers extrinsically. As noted in eq. 3 performance metric like D ∗ , benefit from increased carrier lifetime, as they enhance the signal. Whilewe did not measure the D ∗ itself directly, it is easy to estimate that it will be modulated by − over the voltage5 PREPRINT - J
ULY
7, 2020range shown in fig. 3e directly due to the lifetime modulation on the experimental detector, depending on the operatingtemperature ( K − K ).Figure 4: Projection of back-gating effect on performance a. Optimized detector design, introduction of a transparenttop ground-plane, thinning of gate oxide, and P bI layers. b. Modeled D ∗ − V g characteristics for experimentaldetector to optimized detector design.In fig. 4a we show the schematic of the optimized design. The oxide layer is thinned significantly to nm , thethickness of P bI layer is halved ( nm to nm ), and a top surface full coverage ground plane assumed (can befabricated from mid wave IR transparent contacts such as ITO or graphene). It can be seen from fig. 4b that for thedetector at K , the improvement of D ∗ between zero back-gate bias to the high back-gate bias of − V is inexperimental device while it is in the optimized detector. Other possible optimizations, which we do not performhere, include modulating the doping of the P bSe film and optimizing the film thickness of
P bSe layer, though thefilm cannot be thinned significantly as this reduces the absorption of light in the detector which harms the signal. Thisdemonstrates that extrinsic gate control of lifetime can be an effective means in improving the performance of
P bSe mid-wave IR photo-detectors.
Acknowledgments and Data Availability
Research funded by DARPA/MTO under the WIRED program contract no. FA8650-16-C-7637. The authors thank Dr.Justin Grayer for useful discussions. The data that supports the findings of this study are available within the article.
References
Recent Development in Optoelectronic Devices ,December 2017. Publisher: IntechOpen.[5] Antoni Rogalski. Infrared detectors: an overview.
Infrared Physics & Technology , 43(3):187–210, June 2002.[6] A. Rogalski. History of infrared detectors.
Opto-Electronics Review , 20(3):279–308, 2012.[7] Gunther Roelkens, Utsav Dave, Alban Gassenq, Nannicha Hattasan, Chen Hu, Bart Kuyken, Francois Leo, AdityaMalik, Muhammad Muneeb, Eva Ryckeboer, Sarah Uvin, Zeger Hens, Roel Baets, Yosuke Shimura, FedericaGencarelli, Benjamin Vincent, Roger Loo, Joris Van Campenhout, Laurent Cerutti, Jean-Baptiste Rodriguez, EricTourni´e, Xia Chen, Milos Nedeljkovic, Goran Mashanovich, Li Shen, Noel Healy, Anna C. Peacock, XiaopingLiu, Richard Osgood, and William Green. Silicon-based heterogeneous photonic integrated circuits for themid-infrared.
Optical Materials Express , 3(9):1523–1536, September 2013.[8] Samiran Ganguly, Moon-Hyung Jang, Yaohua Tan, Sung-Shik Yoo, Mool C. Gupta, and Avik W. Ghosh. Amultiscale materials-to-systems modeling of polycrystalline PbSe photodetectors.
Journal of Applied Physics ,126(14):143103, October 2019.[9] Vladimir Wittenberg, Michael Rosenblit, and Gabby Sarusi. Surface plasmon enhanced SWIR absorption at theultra n-doped substrate/PbSe nanostructure layer interface.
Infrared Physics & Technology , 84:43–49, August2017. 6
PREPRINT - J
ULY
7, 2020[10] Justin Grayer, Samiran Ganguly, and Sung-Shik Yoo. Embedded surface plasmon resonant disc arrays for improvedMWIR sensitivity and increased operating temperature of PbSe photoconductive detectors. In
Plasmonics: Design,Materials, Fabrication, Characterization, and Applications XVII , volume 11082, page 1108224. InternationalSociety for Optics and Photonics, September 2019.[11] Hossein Rabiee-Golgir, Soroush Ghandiparsi, Ekaterina Ponizovskaya Devine, Ahmed S. Mayet, Cesar Bartolo-Perez, Priyalal S. Wijewarnasuriya, Nibir K. Dhar, and M. Saif Islam. Ultra-thin super absorbing photon trappingmaterials for high-performance infrared detection. In
Infrared Technology and Applications XLV , volume 11002,page 110020T. International Society for Optics and Photonics, May 2019.[12] Samiran Ganguly and Sung-Shik Yoo. On the Choice of Metallic Contacts with Polycrystalline PbSe Films and ItsEffect on Carrier Sweepout and Performance in Mid-wave Infrared (MWIR) Photodetectors.
Journal of ElectronicMaterials , 48(10):6169–6175, October 2019.[13] T. H. Johnson. Lead Salt Detectors And Arrays PbS And PbSe. volume 0443, pages 60–95. International Societyfor Optics and Photonics, December 1983.[14] Jijun Qiu, Binbin Weng, Zijian Yuan, and Zhisheng Shi. Study of sensitization process on mid-infrared uncooledPbSe photoconductive detectors leads to high detectivity.
Journal of Applied Physics , 113(10):103102, March2013.[15] Lihua Zhao, Jijun Qiu, Binbin Weng, Caleb Chang, Zijian Yuan, and Zhisheng Shi. Understanding sensitizationbehavior of lead selenide photoconductive detectors by charge separation model.