,A.N:ovel Avalanche Free Sing Photon D etector

0G. Memis, S.C. Kong, A. Katsnelson, M.P. Tomamichel, and H. MohseniDepartment of Electrical Engineering and Computer Science

Northwestern UniversityEvanston, IL 60208, USA

Email: hmohlseni@ece.northwestern.edu

Abstract- We have conceived a novel single photon detector for CarriersIR wavelengths above 1 jm. The detection mechanism is basedon carrier focalization and nano-injection. Preliminary measuredAdata from unpassivated devices show a very high internal gain aand low dark current at 1.55 gm at room temperature

words- single photon detector ifrared detector, type-II,finite element method, simulatiot(

INTRODUCTION _± A Oxidized Layer Absorptin Lyr

Infrared detectors have been known and developed for Nanoinjecor Hole annelmany Years. However a wide range of applications requiresingle photon detectors with lower noise, better signal-to-noiseratio, faster response, and operating wavelength beyond gin. p rThese include biophotonics tomography, non-destmctive SubtmteaXmterial inspection, astronomy quantum key distribution,qua-nttum imaging, and homeland security. (b)

Existing devices for infrared detection are photodiodes Figure 1. The schematics ofFOCUS, showing (a) lateral focalization and (b)(PIN and avalanche), phototransistors, quantum well (QW]EP)) vertical transport.and quantun dot infared photodetectors (QDIP). However, notall of these are suitable for single-photon detection. in order tobe used as a single-photon detector an infrared detector needs created to deliver better performance and higher SNR: whento be extremely low noise vhile creating enough signal optically generated electron-hole pairs are attracted and trappedamplitude: generally this is oly possible with quantum at the no-jector. Ths creates a high concenion atconfinement, such as a single-electron transistor based reduces the potential bamfer of the injector and leads to thedetectorfs [l, or an internal gainlmechntism, suh as avaanche injection of many electrons to the absorption layer (Fig.2). Thebased detectors [21. PIN photodiodes are the simplest and most strtice features shoret tansit time for electrons and higmature of all IR photon detectors. However, the have no li or holes n te no-ijctor, whch acconts or thieinternal gain, which linit their applicatioi for single photon highgin. Furtemr, e thic absotlayer and effitdetection. Silicon based singe photon detectors show good collection of holes results in high quantum efficiency.detection performance belowvI m. Unfortunately, at telecomwavelengths (1.3gm, 1.55gm) silicon does not absorbefficlenly. Therefore, researchers have focused on othermaterials for this important part of JR spectra. inGaAs/InPavalanche photodetectors have been the main candidates forsingle photon detection to -1.7 gim. However, they generatelarge dark curIents that reduce effective single photon detection PhotonSfor many apllcations. In particular, they cannot be used innon-synchronized applications such as imaging systems [3, 4]. Figure 2. Electron injection resulted from entrapment of a hole.

Here we present a novel avalanche-free single photon To improve carier injection, FOCUS is utilizing type-IIdetezctorV based on carrier focalization and augmentation. The band alignmElent in the lattce miatched InP/aAsSb/InGaAsprinciple of operation of the so-called Focalizeod Carnr ma.terial system. Ht electron injection is easily achieved in thsaUgmented Senlsor (FOCUS) is showvn in Fig. 1 schematically. system [5, 6], anld is used to enhance the device gainl. Unlike

other low dimensional detectors, FOCtUS contans a larg

This wrk is partially sponlsored by DAR:PAthrSough grat # HROO1 1-05-1-0058

742:

1-4244-0078-3/106/820.00X (c) 2006 IEEER

absorption medium to increase quantum efficiency, and adds nonlinear effects or forced boundaries), then using that solutionthe benefit of same smiall sensing volume of a quantum dot. as initial condition increased convergence and stabilized theThe small volume of the nano-injector produces an ultra-low simulation.capacitance that leads to a large change of potential even withentpment of a single hole (Fig.3). The change of potential Oher than convergence nissues, a modeling ofreduces the barrier between the emitter and collector, and surfacesneededspecialattenion:The nano-inJectorregionwasproduces a large electron injection. Our simulation results show extremely sensitive to any surface effect due to its high surface-a gain of lO64 is easily achievable. The structure is also to-volume ratio. It was important to accurately depict surfacedesigned to reduce recombination of injected electrons with the recombination physics Further, to improve the results of FEMphoto-generated holes by separating the electron and hole simulation, weak Constralnts were used in current calculations.currentpaths. Weak constraints ensured that predicted currents are correctand free of the inherent boundary flux calculation problem in

Enegy levj FEM.

-3.05

-3.05tRESULTS-4U5 Electrons Simulation results show a high gain and low dark current,-4.41-5 wwhich correspond to a hig SNR in this device. Companng to-4- JInGaAs phototransistor and photodiode, although the galn of-5.05- < X - phototransistor is hier and the dark current of photodiode is

lower than FOCUS, SN ratio of both is still lower thanFOCUS. The current components within the device are also as

25A15 expected: the electron injection is maiuly under the nano-injector, and hence recombination with photo-generated holes

D A3 UAG4 .A? 11.49 0.51 0.5;3 9.5b M.7!Gri0 40. 0itio.50mic0rons5 is mininized (Fig. 5) Currnt-voltage characteristics of a 5 p.m

Figure 3. The band diagram at the center of device (vertical). wide device at different optical powers are shown in Fig.6. Thedevice is capable of detecting ver weak optical powers even at

SIMULATION MODEL room temperature, while demonstrating a respectable ittemalgain of about 4x103 even with a small multiplication region.

A simulation tool was created using Comsol Multiphysics[7J, which uses Finite Element Method (FEM) to simulatephysical environment. The model incorporates the Poisson'sequation along with the diffusion-drift equations usingappropnate boundary conditions and cylindrical symmetry. Toimprove performance and lower the comnplexity, the simuilationof a radial cross-section of FOCUS is perfonned instead of thewhole 3-dimensioal volume (Fig.4). The model incororatesseveral nonflinear effects such as incomplete ionization ofdopanits, surface recombination, impact ion°ization, nonlinearmobility and hot electron effects.

Figure 5. Simulated electron injection density (arrow) and electrostaticpotential (color) ofthe device at Th300 K anid under a bias value of-,O.5 volts.

Figure 4. The radial cross section used in FEM simulation

As the device Contains higW nonlinear dynamics, stabilityof the mnodel was a problem. Consequently, there wereconvergence issues which needed to be addressed: First, forcedboundaries in the form of ohmic or Schottky contads (i.e.Dirichlet boundary conditions) imposed infinite recombinationand created stress on the domains they are attached to. Inadditionl to this, the heterojunctions in the device resulted injulmps in thLe carir concentrations, which made currfentcontinluitx harder to sulstain. The most effective method toovercome thiese pro;blems was to build up the simulation instepss. Stating withi a less constrained muodel (i.e. without

743

1-4244-0078-3/06/&$20.OO (c) 2006 IEEE

10 r.IIIII9 TSOO T=300 K

_D-5pm 2 _ _ R=5FmD=Spm 1-4 -- - 600 -Runpassivated 5OpW1700-nm1 rD lw 50 su rfaces----50surfaces-

1700nm 00 500 fW*~400

*_0 0.1. 0.2.3..4050. 01 0.80 1 ou-

C ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~5 r I- w- || |L 4 _ _ .r- - f- r t 0 1 0. 0.4 .6 0.

L 1S tf211:> {= ~~~~~~~~~~~~~~~300 -

~~~~Bia (V ltl) Bia Votae (V)!, P|450~~ ~ ~ ~ ~ ~ ~ ~

the1radialdiffusion length compaRedtotheaxiald--Rdii.d-20 Alf) 3100 - -------_

lli .1 O,Z :e#.3 OA 0.5 Dl£ 0et.ir Dl*X :@,91 10.2 0.4 0.6 0.8 1

Bias {VoltA ~~~~~~~~~~BiasVoltage()Fire 7. Measured cunent-voltageedataeatudierentilluminationpowers of

Firer6. Curent voltae characterioticsofa5mwide deviceatT302K50mdiameterdevicewithoutysurfacepassivation(T=300K)anddifferentoptical powers.The deice showsaninternal gain of400.

The resultstshowthatCsligar increasinojthe randus ofnadio- mindector isqureseffeicteinof1the 4s im afn 100 jm sid 100hltengthwertse pTteusslaion phsotthograttherutinoftRl5gmthe camere then rednarrow nano-onjfechor region reuces Nwee

used |Jas hadms fo ethig 0IIh * Il*

sheo ratal thffecsion lencll compared toiseax1ilarel tusion F 5. Measured h o atl

the device size, on average. The tested devices300have--rrelativelyI

len. This redubtiontwas sxpected ae to than congestion ofcamers aod the center ofFOCUS device. 250 - - - -i-- - - -a- - - -e u- - -c n- - - va-- - - p-- - - -eth

ElXPERAIMENTAL RESULTS 2:00 I -4We have started processing andfabcatedW heza no gown ve150s-n-g-l-e-

on InP substrates. Circudar . I ripasaife25 om atpa5 am rrge -and sqtwe shaped devices of 10 WfiB 4g0 m w 1g00Wttbrieo 100a-y-e- I F-assevera f to r e n

length were pattemed using hotoliio hyl The metal a h g T de w creatcontacts were then formed using lift--off technique and weire 50x t i -into -account the fabrican ad p csi- gused as hfra-maufor oehhinge e

Thetested deie shodA ver Iih gai fmoe,ha 0.4 0.5 0.6 0.7 0.8 0.9 1 .i 1.2

ez~~~~ ~ ~ ~ ~ ~ ~~~~~h model enables us to se and compare performancesamon

1n000) ei fr smaller deics Dar cfn clcultion Bias Voltage (V)show tht theW efictiv coillecti AVseon radius is 1.5 tm larer dmn Figure 8. Measured bandWXidth 6 5 gm dia4cfer device atdifferent voltagEestheL devioe size, on average. The teLsted deviOes hSae relativlyhigh band-vvidth of imore tian 70kHzd reaching to 400ikrz at Note that ofe mpaho ed deviceswera not passivateds We1.2 volts. believe thata sigereiicait reduch tionrr im rv tcan be

achieved ulsiig conventional passivation n-fictods.

CONCLUSION

Wew have designed and realized a novel single photonXdetector. It incorporates a nao-inector regionl ont top of a lareabsorption layer. It has gevferal katurezs to reduce thie noiselevel while maintaimiig a high gain. The designi was createdWhile takting into acGounlt the fabricati:on and processingtechnology limnitations.

FLu-tenmore, we created a thee-dimnensional non-linear

744

1-4244-0078-31061&$20.00 (c) 2006 IEEE

the simtulation: we are cturrentl adding quantum effects andMonte Carlo analysis to the exsting model.

The optimized epitaxial layer structire and device geometry,provided by the simulation tool, was used for epitiaal growthald processing mask design. First batch of devices have beenfabricated and tested. These unpassivated devices show a highilternal gain at small bias values, high firequency bandwidth,and low dark current at room temperature.

REFERtENCES

[1] P. W. Li, W. M. Liao, David M. T. Kuo, and S. W. Lin, "Fabrication ofa germanium quantum-dot single-electron transistor with largeCoulomb-blockade oscillations at room temperature", App. Phys. Lett.S5(9): 1532-1534 (2004)

[2] A. Rogalski, "Infrared detectors, status and trends", Pro ess inQuanttum Elec. 27: 59-2 10 (2003).

[3] A. Lacaita, F. Zappa, S. Cova, and P. Lovati "Si'ngle-photon detectionbeyond 1 jim: performance of commercially available InGaAs@InPdetectors", Applied Optics. 35 (16):2986-2996 (1996).

[4] F. Zappa, A. L. Lacaita, S. D. Cova and P. Lovati, "Solid-state single-photon detectors", Opt Eng. 35(4): 938-945 (1996)

[5] J. Hu, N. 0. Xui, J. A. H. Stotz, S. P. Watkins, A. E. C:urzon, M. L. W.Thewalt, N. Matine, C. R. Bolognesi, "Type II photoluminescence andconduction band offsets of GaAsSb/InGaAs and GaAsSb/InPheterostrucures grown by metalorganic vapor phase epitaxy", Appl.Phys. Lett. 73(19): 2799-2801 (1998)

[6] R. Sidhu, N. Duian, J. C. Campbell, A. L. Holmes,"A long-wavelengthphotodiode on InP usin:g lattice-matched GaInAs-GaAsSb type-I1quantum wells", IEEE Photonics Tech. Lett. 17(12): 2715-2717 (2005)

[7] Comsol Multiphysics, Comsol, IIc. httbp:/ T.cosol.com

745

1-4244-0078-3/06/&$20.OO (c) 2006 IEEE