semiconductor ultraviolet detectors - antoni rogalski · semiconductor ultraviolet detectors ......

APPLIED PHYSICS REVIEWS

Semiconductor ultraviolet detectorsM. Razeghia) and A. Rogalskib)Center for Quantum Devices, Department of Electrical Engineering and Computer Science,Northwestern University, Evanston, Illinois 60201

~Received 7 September 1995; accepted for publication 12 January 1996!

In this review article a comprehensive analysis of the developments in ultraviolet~UV! detectortechnology is described. At the beginning, the classification of UV detectors and generalrequirements imposed on these detectors are presented. Further considerations are restricted tomodern semiconductor UV detectors, so the basic theory of photoconductive and photovoltaicdetectors is presented in a uniform way convenient for various detector materials. Next, the currentstate of the art of different types of semiconductor UV detectors is presented. Hitherto, thesemiconductor UV detectors have been mainly fabricated using Si. Industries such as the aerospace,automotive, petroleum, and others have continuously provided the impetus pushing the developmentof fringe technologies which are tolerant of increasingly high temperatures and hostileenvironments. As a result, the main efforts are currently directed to a new generation of UVdetectors fabricated from wide band-gap semiconductors the most promising of which are diamondand AlGaN. The latest progress in development of AlGaN UV detectors is finally described indetail. © 1996 American Institute of Physics.@S0021-8979~96!06110-7#

-

t

TABLE OF CONTENTS

I. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7434II. Classification of ultraviolet detectors. . . . . . . . . . . . 7435III. Ultraviolet photodetectors. . . . . . . . . . . . . . . . . . . . 7437

A. General theory of photodetectors. . . . . . . . . . . . . 7437B. Photoconductive detectors. .. . . . . . . . . . . . . . . . 7439

1. Current and voltage responsivity. . . . . . . . . . 74392. Sweep-out effects. . . . . . . . . . . . . . . . . . . . . .74403. Noise mechanisms in photoconductors. . . . . 74414. Quantum efficiency. . . . . . . . . . . . . . . . . . . . 74415. Influence of surface recombination. . . . . . . . 7442

C. p-n junction photodiodes. . . . . . . . . . . . . . . . . . . 74421. Ideal diffusion-limitedp-n junctions.. . . . . 74432. Other current mechanisms. . . . . . . . . . . . . . . 74453. Response time. . . . . . . . . . . . . . . . . . . . . . . . .7449

D. Schottky barrier photodiodes. . . . . . . . . . . . . . . . 74501. Schottky–Mott theory and its

modifications. . . . . . . . . . . . . . . . . . . . . . . . . .74502. Current transport processes. . . . . . . . . . . . . . 74513. R0A product and responsivity. . . . . . . . . . . . 7451

E. Comparison of different types of semiconductorphotodetectors. . . . . . . . . . . . . . . . . . . . . . . . . . . .7452

F. Semiconductor materials used for ultravioletdetectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7452

IV. Si ultraviolet detectors. . . . . . . . . . . . . . . . . . . . . . .7454A. Diffused photodiodes. . . . . . . . . . . . . . . . . . . . . .7454B. Si photodiodes for vacuum ultraviolet

applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7457

a!Electronic mail: [email protected]!Present address: Institute of Technical Physics, Military UniversityTechnology, 01-489 Warsaw, Kaliskiego 2, Poland.

J. Appl. Phys. 79 (10), 15 May 1996 0021-8979/96/79(10)/74

V. Schottky barrier ultraviolet detectors. . . . . . . . . . . . 7458VI. SiC ultraviolet detectors. . . . . . . . . . . . . . . . . . . . .7459

A. p-n junction photodiodes. . . . . . . . . . . . . . . . . . . 7459B. Schottky barrier photodiodes. . . . . . . . . . . . . . . . 7461

VII. III–V nitrides as a materials for ultravioletdetectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7462

A. Physical properties. . . . . . . . . . . . . . . . . . . . . . . .7463B. Ohmic contacts to GaN. . . . . . . . . . . . . . . . . . . . 7465C. Etching of III–V nitrides. . . . . . . . . . . . . . . . . . . 7466D. AlGaN ultraviolet detectors. . . . . . . . . . . . . . . . . 7467

1. Photoconductive detectors. . . . . . . . . . . . . . . 74672. Photovoltaic detectors. . . . . . . . . . . . . . . . . . 7468

VIII. Other materials for ultraviolet detectors. . . . . . . . 7470IX. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7470

GLOSSARY OF FREQUENTLY USED SYMBOLS

a lattice constantA detector areab electron to hole mobility ratioc speed of lightD,Da ,De ,Dh diffusion coefficient, ambipolar, of elec-

trons, of holesD* , DBLIP* detectivity, background limitedEg ,EF ,Et energy gap, Fermi energy, trap level en

ergyD f bandwidthG recombination rateg optical gainh,\ Planck’s constanth/2pI ,I e ,I h ,I ph current, of electrons, of holes, photocurrenI D ,IGR,I T ,I n diffusion, generation–recombination, tun-

neling, and noise currentI d ,I s dark, saturation current

of

743333/41/$10.00 © 1996 American Institute of Physics

o

n

d

f

d

,

a

f

thewill

thofoontra-ato-aceherlar. Inab-00ume tooddif-i-ul-ods,rat-umnlytainec-ersblehteri-arnderein-reaset-asandtan-

arenm.

o-

be-n is

orb-

io-ul-

J,Je ,Jh ,Jph current density, of electrons, of holes, phtocurrent density

k Boltzmann’s constantl lengthL,Le ,Lh diffusion length, of electrons, of holesm free electron massm* , mc* , mn* effective mass, of electrons, of holesn,ni electron and intrinsic carrier concentration0 ,nn ,np ,Dn equilibrium electron concentration, major

ity and minority electron concentration, excess free electrons

Na ,Nd concentration of acceptors, of donorsNEP noise equivalent powerp,p0 ,pp ,pn ,Dp hole concentration, equilibrium, majority

and minority hole concentration, excesfree holes

q electron charger surface reflectanceR,RL ,Rs resistance, load, seriesR0 zero bias detector resistanceRi ,Rv current and voltage responsivitys surface recombinationt thicknessT temperaturev,vd ,va ,v th velocity, drift, ambipolar, saturation, an

thermal carrierV,Vb ,Vbi ,Vn electrical voltage, bias, built-in and noisew widthx alloy composition, distance variablea absorption coefficientb coefficient of I–V diode characteristic«0 permittivity of space«,«0,«` dielectric coefficient, relative static and op

ticalh quantum efficiencyl,lp ,lc wavelength, peak wavelength, cuto

wavelengthn frequencym,me ,mh mobility of carriers, of electrons, of holest,te ,th ,tef carrier lifetime, of electrons, of holes an

effective lifetimetRC,td ,ts response time limited by RC, diffusion

and transit timet0 carrier lifetime in depletion junction regionFB ,Fs background, signal photon fluxf work functionC barrier potentialv angular frequency

I. INTRODUCTION

The ultraviolet ~UV! region is commonly divided intothe following subdivisions. These names are widely usedare recommended:near ultraviolet NUV 400–300 nmmid ultraviolet MUV 300–200 nmfar ultraviolet FUV 200–100 nmextreme ultraviolet EUV 100–10 nm

In addition to the above names, the following nameswavelength regions may be encountered:

7434 J. Appl. Phys., Vol. 79, No. 10, 15 May 1996

-

--

s

-

f

nd

or

vacuum ultraviolet VUV 200–10 nmdeep ultraviolet DUV 350–190 nmultraviolet-A UV-A 400–320 nmultraviolet-B UV-B 320–280 nm

Again, it must be emphasized that these appear to bemost common names and wavelength limits, but othersbe found.

Ultraviolet research began in the latter half of the 19century, when the invisible radiation beyond the blue endthe visible spectrum began to receive attention. It was srealized that the Earth’s atmosphere set limitations on ulviolet research. For solar and celestial observations, strspheric ozone limited the wavelengths reaching the surfof the Earth to about 300 nm. Spectrographs carried to higaltitudes on mountains gave intriguing evidence that soand stellar emissions continued to shorter wavelengthslaboratory spectrographs, atmospheric molecular oxygensorption limited the useful lower wavelengths to about 2nm, unless the spectrograph could be placed in a vacuchamber. The wavelengths shorter than 200 nm thus cambe called the vacuum ultraviolet. Because of the lack of govacuum pumps and associated technology, research wasficult and not widely done. In addition to atmospheric limtations, optical methods used in the visible failed in thetraviolet because of the lack of materials having gotransmissivity and reflectivity. In the beginning of the 1880Rowland and co-workers developed concave diffraction gings and discovered the Rowland circle mount for vacuspectrographs. This was a great step forward, since osingle reflection and no transmission were needed to obspectra on a photographic plate. Vacuum ultraviolet sptrography was pioneered by Schumann, Lyman, and othin the early decades of this century. The lack of suitawindows, filters, gratings, calibration standards, ligsources, and vacuum pumps resulted in difficulties in expmentation. The next large stimulus occurred after World WII, when it was possible to use first sounding rockets athen satellites to investigate the Earth’s upper atmosphand to make solar and astronomical observations withoutterference by the atmosphere. These applications aopened up the need for better instrumentation, including bter windows, gratings, filters, detectors, and light sourceswell as improved spectroscopy of the atoms, molecules,ions of planetary and stellar atmospheres and reliable sdards for calibrations.

The major constituents of the terrestrial atmospherestrong absorbers of radiation at wavelengths below 300Radiation of mid ultraviolet~MUV ! wavelengths between200 and 300 nm is absorbed primarily by ozone while mlecular oxygen is the major absorber at far ultraviolet~FUV!between 110 and 250 nm. Consequently, at wavelengthslow about 200 nm, the use of evacuated instrumentatiomandatory. At extreme ultraviolet~EUV! wavelengths of110 nm, atomic and molecular gases become strong absers.

Several books from the mid 1960s describe the ultravlet technology of this time. These books are on vacuumtraviolet spectroscopic techniques,1 the middle ultraviolet@near ultraviolet~NUV! and MUV, generally#,2 ultraviolet

Appl. Phys. Rev.: M. Razeghi and A. Rogalski

J. Appl. Phys., Vol. 79

FIG. 1. Classification of ultraviolet photon detectors.

n

hinc-is-ofthec-m-s aa-c-m-Theceg.ores,as

ndagegehic-y-ond

yx-

en-terleictaic

light sources,3 and spectroscopy.4 The historical develop-ment of spectroscopy is traced in the introductory chapterSamson’s book,1 also the chapter in Green’s book by Henneand Dunkleman,5 entitled Ultraviolet Technology, is anavailable source for the development of experimental tecniques of that time. Recently, Huffman has published twbooks about the ultraviolet spectral regions.6,7 The second ofthem7 supplements previously published books, since thvolume is a collection of reprints that present the milestoarticles in ultraviolet optics and technology.

The issue of UV detectors is treated in several mongraphs and reviews. An extensive examination of variodetector systems for both imaging and nonimaging applictions is presented by Carruthers.8 Timothy and Madden9

have restricted their review to photon detectors that are crently available for use at ultraviolet and x-ray wavelengthMore information about the history of the developmentUV detectors and their current status in astronomy canfound in an excellent book entitledLow light level detectorsin astronomyby Eccles, Sim, and Tritton.10 Another ofTimothy’s review articles is devoted mainly to detectors fooptical wavelengths,11 and a comparison of charge coupledevices~CCDs! to other optical detectors is given by Janesicket al.12 The reviews of the present and future technlogical concepts currently being considered in astrophysand astronomy are given by Welsh and Kaplan,13 Joseph,14

and by Ulmeret al.15 However, the UV detectors have alsfound terrestrial applications. They can detect UV emissiofrom flames in the presence of hot backgrounds~such asinfrared emission from the hot bricks in a furnace!. Thisprovides an excellent flame on/off determination systemcontrolling the gas supply to large furnaces and boiler sytems. Flame safeguard and fire control areas are just twothe various possible applications for the UV detectors.

In this article we will first present the classification oUV detectors and general requirements imposed on thesetectors. Further consideration will be restricted to modesemiconductor UV detectors, so the basic theory of phoconductive and photovoltaic detectors will be presented inuniform way convenient for various detector materials. Nethe current state of the art of different types of semiconducUV detectors will be presented. The main effort will be epecially directed to a new generation of UV detectors. Thgeneration of detectors is the product of five years of marials and device research, which resulted in the developmof high quality GaN layers.

, No. 10, 15 May 1996

ins

h-o

isne

o-usa-

ur-s.ofbe

rd-o-ics

ons

fors-of

fde-rnto-a

xt,tors-iste-ent

II. CLASSIFICATION OF ULTRAVIOLET DETECTORS

In general, UV detectors fall into two categories: photodetectors~also named photodetectors! and thermal detectors.In photon detectors the incident photons are absorbed witthe material by interaction with electrons. The observed eletrical signal results from the changed electronic energy dtribution. The photon detectors measure the rate of arrivalquanta and show a selective wavelength dependence ofresponse per unit incident radiation power. In thermal detetors, the incident radiation is absorbed and raises the teperature of the material. The output signal is observed achange in some temperature-dependent property of the mterial. In pyroelectric detectors a change in the internal eletrical polarization is measured, whereas in the case of boloeters a change in the electrical resistance is measured.thermal effects are generally wavelength independent sinthe radiation can be absorbed in a ‘‘black’’ surface coatinBecause of greater sensitivity, photon detectors are mcommonly utilized at UV wavelengths. Thermal detectorhowever, are sometimes employed at UV wavelengthsabsolute radiometric standards.

UV photon detectors~see Fig. 1! have traditionally beendevoted into two distinct classes, namely, photographic aphotoelectric. Photographic emulsion has the great advantof an image-storing capability and can thus record a laramount of data in a single exposure. However, photograpemulsion has a number of limitations: sensitivity is considerably lower than that of a photoelectric detector, the dnamic range is limited, the response is not a linear functiof the incident photon flux at a specific wavelength, anemulsion is sensitive to a very wide energy range~accord-ingly the elimination of background fog levels induced bscattered light and by high-energy charged particles is etremely difficult!.

Photoelectric detectors, on the other hand, are more ssitive, have a greater stability of response and provide betlinearity characteristics. In the last decade considerabprogress in the image-recording capability of photoelectrondevices has been observed. Recently developed photovolarray detectors@such as the charge coupled devices# and pho-toemissive array detectors@such as the microchannel arrayplates~MCPs!# for the first time combine the sensitivity andradiometric stability of a photomultiplier with a high-resolution imaging capability.

7435Appl. Phys. Rev.: M. Razeghi and A. Rogalski

hethebea

syons.hult-rbyhissi-

ted-ndings.e-torel-

k-

-

thee.toantent

ofseVl-

In the most commonly employed photoemissive UV dtectors, the photon is allowed to impact a solid surface reizing a photoelectron into the vacuum [email protected]~a!#. Applying a voltage between the photocathode surfand a positively biased anode causes a photoelectron cuto flow in proportion to the intensity of the incident radiatioSince these detectors make use of the external photoeleeffect, the wavelength range of sensitivity is defined primrily by the work function of the surface material.

In the semiconductor detectors, the photons are absoin the bulk of the semiconductor material producielectron-hole pairs which are separated by an electrical fiThese detectors make use of the internal photoelectric ewhere the energy of the photons is large enough to raiseelectrons into the conduction band of the semiconductorterial. In the case of photovoltaic detectors, the electron-hpairs are separated by the electrical field ofp-n junctions,Schottky barrier, or metal-insulator-semiconductor~MIS! ca-pacitors, which leads to an external photocurrent proptional to the number of detected photons@Fig. 2~b!#. Apply-ing a voltage across the absorbing region causes a curreflow in proportion to the intensity of the incident radiation

In the photoemissive detectors, the primary photoeltron can be multiplied by the process of secondary emissto produce a large cloud of electrons. The occurrence osingle photoelectron event then can be detected eitherrectly with conventional electronic circuits or by acceleratithe electron cloud to high energy and allowing it to impacphosphor screen. The emitted pulse of visible-light photcan then be viewed directly, or detected and recordedadditional photosensitive systems. Detectors operated inpulse-counting mode can provide the ultimate level of setivity set by the quantum efficiency of the photocathode.

FIG. 2. Principle operation of photoemissive~a! and semiconductor detectors ~b!.


e-al-

acerrentn.ctrica-

rbedngeld.ffectthema-ole

or-

nt to.ec-ionf adi-

ngt aonsbythisnsi-

There are a number of important differences between ttwo classes of detectors. In the photoemissive detectors,primary photoelectron produced by the photocathode canmultiplied by the process of secondary emission to producelarge cloud of electrons. The degree of multiplication icalled the gain of the detector. If the gain is sufficientllarge, the electron cloud generated by a single photoelectrcan be detected directly with conventional electronic circuitAlternatively, the electron cloud can be accelerated to higenergy and allowed to impact a phosphor screen. The resing pulse of visible-light photons emitted from the phosphocan be viewed directly or can be detected and recordedadditional photosensitive systems. Detectors operated in tpulse-counting mode can provide the ultimate level of sentivity at very low signal level. That form of intensificationimplies, however, high voltages and the inherent associadifficulties. Moreover, it is not possible in this mode of operation to store the detected events within the detector, athe detected signal must be integrated in a separate recordsystem. Table I compares these two types of UV detector

The importance of UV semiconductor detectors has rsulted in the recent meteoric expansion of the semiconducindustry, and second, the continuing emphasis on the devopment of low-light-level imaging systems for military andcivilian surveillance applications. These detectors should:

~1! not be sensitive to light at optical wavelengths~com-monly referred to as being solar blind!,

~2! have high quantum efficiency,~3! have a high dynamic range of operation,~4! have low backgrounds since noise arising from the bac

ground often dominates in faint UV observations.

Multiplying the number of primary charge carriers isgenerally not possible in semiconductor detectors~althoughthis deficiency may be offset by the very high internal quantum efficiency of the semiconductor material!. These detec-tors are thus less sensitive at the lowest signal levels thanphotoemissive detectors operating in pulse-counting modHowever, the semiconductor detectors have the abilitystore charge and integrate the detected signal for significperiods of time. Recently, there has been much developmof CCD detectors for use in the UV spectral ranges.7,8,12,16–18

On the contrary, for visible spectral range devices the useSi CCDs in the UV region is not yet well established becauof the many problems connected with the interaction of Uradiation with the materials typically used in silicon techno

-

ty

TABLE I. Comparison of photoemissive and semiconductor UV detectors.

Type Advantages Disadvantages

Photoemissive detectors Easy to operate Low quantum efficiencyHigh sensitivity Strong spectral dependence of responsiviSolar blind Sensitiveness to surface contaminations

Semiconductor detectors Broad spectral responsivity Induced aging effectsExcellent linearityHigh quantum efficiencyHigh dynamic range of operationLarge-format image arrays


a

r

o

g

hec-orgro-ct.fi-en-i-nd

e

rs,

c-for

n-

ease

r-

eu-ed

lyofluetouesice.

rent

ogy. For example, the relatively thin gate oxide layer~500–1200 Å! on the CCD front face is strongly absorbing of Uradiation, so that it is not possible to use front-illuminatCCDs for direct detection of this radiation. The electron-hpairs created in this layer are not separated by an intefield and will therefore not contribute to the photocurreThinned, back-illuminated CCDs also have some probleIn fact, after the device is thinned, a native back-oxide laforms naturally, and even if it is very thin, typically less th50 Å, this dramatically influences overall sensor perfmance. A number of solutions have been tried, ranging fthe deposition of a scintillator, such as coronene19 orlumigen,20 to the backside illumination of thinned devices,to avoid the absorption of radiation by the thick polysilicgate structure. Some methods have been developed to rethe surface state problems using ‘‘back-accumulatiotechniques.21 They are based on the hypothesis that the bsurface potential can be pinned by heavily populating it wholes which create an electric field that directs the photonerated charges away from the back surface. They cadivided into three categories: backside charging, flashtechniques, and ion implantation.8,21

A semiconductor array can also be used to directly

FIG. 3. Quantum efficiency of thinned backside illuminated CCDs~a! andsolar-blind efficiencies of various UV detectors~b!. The detective quantumefficiency~DQE! of a complete detector system takes into account anyof detected photons and the quantum efficiency of each stage of the syThe DQE can be expressed in terms of the signal-to-noise characteristthe input and output signals as the ratio DQE5~S/Nout!

2/~S/Nin!2 ~after Ref.

14!.

J. Appl. Phys., Vol. 79, No. 10, 15 May 1996

Vedolernalnt.ms.yernor-om

sonducen’’ackithge-n beate

de-

tect the photoelectron released from a photocathode. Tphotoelectron is accelerated to very high energy by an eletrostatic field and allowed to bombard the semiconductmaterial. If sufficient energy is released within the absorbinregion, enough electron-hole pairs can be produced to pvide a measurable signal from a single photoelectron impa

Figure 3 shows the current status of the quantum efciency of common detectors. The top plot in Fig. 3 shows thquantum efficiencies that can be obtained with various cofigurations of thinned, back-illuminated CCDs. Future antreflection coatings, depicted as a dot-dash line, may extethe range well into the UV range. Figure 3~b! shows thesolar-blind efficiencies obtained for various UV detectors.

III. ULTRAVIOLET PHOTODETECTORS

Ultraviolet semiconductor photodetectors work in threfundamental modes:

~1! photoconductive detectors,~2! photodiodep-n junctions, and~3! Schottky barrier detectors.

Below, after describing the general theory of photodetectothe basic theories of photoconductors,p-n junctions, andSchottky barriers are presented. In the last part of this setion, a short presentation of semiconductor materials usedthe fabrication of UV detectors is included.

A. General theory of photodetectors

The photodetector is a slab of homogeneous semicoductor with the actual ‘‘electrical’’ areaAe which is coupledto a beam of infrared radiation by its optical areaAo. Usu-ally, the optical and electrical areas of the device are thsame or close. The use of optical concentrators can incretheAo/Ae ratio.

The current responsivity of the photodetector is detemined by the quantum efficiencyh and by the photoelectricgaing. The quantum efficiency value describes how well thdetector is coupled to the radiation to be detected. It is usally defined as the number of electron-hole pairs generatper incident photon. The idea of photoconductive gaing wasput forth by Rose22 as a simplifying concept for the under-standing of photoconductive phenomena and is now wideused in the field. The photoelectric gain is the numbercarriers passing contacts per one generated pair. This vashows how well the generated electron-hole pairs are usedgenerate the current response of a photodetector. Both valare assumed here as constant over the volume of the dev

The spectral current responsivity is equal to

Ri5lh

hcqg, ~1!

wherel is the wavelength,h is Planck’s constant,c is thelight velocity, q is the electron charge, andg is the photo-electric current gain. Assuming that the current gains fophotocurrent and noise current are the same, the currnoise due to generation and recombination processes is22

I n252~G1R!AetD f q2g2, ~2!

lossstem.ics of


e-

i

r

f

e

b

ah

-u

or-

en

lueAton

ity

a

ind

n

r

onsthe

ve-nd2gth

whereG andR are the generation and recombination ratD f is the frequency band, andt is the thickness of the detector.

DetectivityD* is the main parameter characterizing nomalized signal-to-noise performance of detectors and candefined as

D*5Ri~AoD f !1/2

I n. ~3!

According to Eqs.~1!–~3!23

D*5l

hc SAo

AeD 1/2h@2~G1R!t#21/2. ~4!

For a given wavelength and operating temperature,highest performance can be obtained by maximizh/[ t(G1R)] 1/2 which corresponds to the condition of thhighest ratio of the sheet optical generation to the squareof sheet thermal generation–recombination.

The effects of a fluctuating recombination can frequenbe avoided by arranging for the recombination processtake place in a region of the device where it has little effedue to low photoelectric gain: for example, at the contactssweep-out photoconductors or in the neutral regions ofodes. In this case, the noise can be reduced by a factor oand detectivity increased by the same factor. The generaprocess with its associated fluctuation, however, cannotavoided by any means.

At equilibrium the generation and recombination ratare equal, and assuming thatAo5Ae we have

D*5l

2hc

h

t1/2G21/2. ~5!

The total generation rate is a sum of the optical athermal generation

G5Gth1Gop. ~6!

The optical generation may be due to the signal or thmal background radiation. For infrared detectors, usuathermal background radiation is higher compared to the snal radiation. If the thermal generation is reduced muchlow the background level, the performance of the devicedetermined by the background radiation@conditions forbackground limited infrared photodetector~BLIP!#. Thenoise equivalent power~NEP! in this approximation is givenby24,25

NEP5~AD f !1/2

D*5hgS 2AFBD f

h D 1/2, ~7!

whereFB is the total background photon flux density reacing the detector, andD f is the electrical bandwidth of thereceiver. The background photon flux density received bydetector depends on its angular view of the backgroundon its ability to respond to the wavelengths contained in tsource.

When photodetectors are operated in conditions whthe background flux is less than the optical~signal! flux, theultimate performance of detectors is determined by the sigfluctuation limit ~SFL!. It is achieved in practice with photomultipliers operating in the visible and ultraviolet region, b


s,

r-be

thengeoot

tlytoctindi-2.5tionbe

s

nd

er-llyig-e-is

h-

thendis

ere

nal

t

it is rarely achieved with solid-state devices, which are nmally detector-noise or electronic-noise limited. The~NEP!and detectivity of detectors operating in this limit have bederived by a number of authors~see, e.g., Kruseet al.24,25!.

The NEP in the SFL is given by

NEP52hcD f

hl~8!

when Poisson statistics are applicable. This threshold vaimplies a low number of photons per observation interval.more meaningful parameter is the probability that a phowill be detected during an observation period. Kruse25 showsthat the minimum signal power to achieve 99% probabilthat a photon will be detected in an observation periodto is

NEPmin59.22hcD f

hl, ~9!

whereD f is assumed to be 1/2to. Note that the detector aredoes not enter into the expression and that NEPmin dependslinearly upon the bandwidth, which differs from the casewhich the detection limit is set by internal or backgrounnoise.

Seib and Aukerman26 also have derived an expressiofor the SFL identical to Eq.~9! except that the multiplicativeconstant is not 9.22 but 23/2 for an ideal photoemissive ophotovoltaic detector and 25/2 for a photoconductor. This dif-ference in the constant arises from the differing assumptias to the manner in which the detector is employed andminimum detectable signal-to-noise ratio.

FIG. 4. Minimum detectable monochromatic power as a function of walength for composite of SFL and BLIP limit for two detector areas aelectrical bandwidths. Background temperature is 290 K and FOV ispsteradiants. Detector long wavelength limit is equal to source wavelen~after Ref. 25!.


s.ndd

en

the

-lowinadthetheolt-ge

re

-for

.d

n

yc-r-

It is interesting to determine the composite signal flutuation and background fluctuation limits. Figure 4 illustrathe spectral NEP over the wavelength range from 0.1 tomm assuming a background temperature of 290 K deteareas of 1 cm2 and 1 mm2 ~applicable only to the backgrounfluctuation limit!, a 2p steradian field of view~FOV! @appli-cable only to the background fluctuation limit#, and band-widths of 1 and 104 Hz. These additional values of parameters are specified because of different dependencies of~upon area and bandwidth! for signal fluctuation and background fluctuation limits. Note that the intersections of thrpairs of curves, for which the bandwidths of the signal abackground fluctuation limits are equal, lie between 1.0 a1.5 mm. To illustrate the composite, a detector with an aof 1 cm2 ~applicable to the background fluctuation limit! anda bandwidth of 1 Hz has been scored. For this pair, at walengths below 1.2mm the SFL dominates; the conversetrue above 1.2mm. The minimum detectable monochromaradiant power at 1.2mm is 1.5310218 W in a 1 Hz band-width. Below 1.2mm the wavelength dependence is smaAbove 1.2mm it is very large, due to steep dependence upwavelength of the short wavelength end of the 290 K baground spectral distribution.

The basic equations for dc analysis of detector permance are well known: two current-density equationselectronsI e and holesJb two continuity equations for electrons and holes, and Poisson’s equation which are coltively referred to as the Van Roosbroeck model27

Je5qDe

dn

dx2qmen

dC

dxcurrent density for electrons,

~10!

Jh5qDh

dp

dx2qmhp

dC

dxcurrent density for holes,~11!

1

q

dJedx

1~G2R!50 continuity equation for electrons,

~12!

1

q

dJhdx

2~G2R!50 continuity equation for holes, ~13!

d2C

dx252

q

«0« r~Nd2Na1p2n!, Poisson’s equation,

~14!

whereC is the electrostatic potential,Nd is the concentrationof donors,Na is the concentration of acceptors,De andDh

are the electron and hole diffusion coefficients,n andp arethe electron and hole densities, and«0« r is the permittivityof the semiconductor. A self-consistent iterative procedfor the solution of this mathematical model is well docmented in the literature.28 The method reformulates thmodel in terms of integral equations which incorporateboundary conditions and eliminateJe andJh as intermediatevariables. This allows the carrier densities to be compufrom the potential distribution.

The generation–recombination term (G1R) is associ-ated with the predominant recombination mechanisms. Thare three important generation and recombination mec


c-tes20ctord

-NEP-eendndrea

ve-istic

ll.onck-

for-for-lec-

ureu-ethe

ted

ereha-

nisms: Shockley–Read, radiative, and Auger mechanismThe Shockley–Read mechanism occurs via lattice defect aimpurity energy levels within the forbidden energy gap anis the dominant mechanism in wide band-gap materials.

B. Photoconductive detectors

The photoconductive detector~also named as the photo-conductor! is essentially a radiation-sensitive resistor. Thoperation of a photoconductor is shown in Fig. 5. A photoof energyhn greater than the band-gap energyEg is ab-sorbed to produce an electron-hole pair, thereby changingelectrical conductivity of the semiconductor.

In almost all cases the change in conductivity is measured by means of electrodes attached to the sample. Forresistance material, the photoconductor is usually operateda constant current circuit as shown in Fig. 5. The series loresistance is large compared to the sample resistance, andsignal is detected as a change in voltage developed acrosssample. For high resistance photoconductors, a constant vage circuit is preferred and the signal is detected as a chanin current in the bias circuit.

1. Current and voltage responsivity

We assume that the signal photon flux densityFs~l! isincident on the detector areaA5wl and that the detector isoperated under constant current conditions, i.e.,RL@R. Wesuppose further that the illumination and the bias field aweak, and the excess carrier lifetime,t, is the same for ma-jority and minority carriers. To derive an expression for voltage responsivity, we take a one-dimensional approachsimplicity. This is justified for a detector thickness,t, that issmall with respect to the minority carrier diffusion lengthWe also neglect the effect of recombination at frontal anrear surfaces.

The basic expression describing photoconductivity isemiconductors under equilibrium excitation~i.e., steadystate! is

I ph5qhAFsg, ~15!

whereI ph is the short circuit photocurrent at zero frequenc~dc!, i.e., the increase in current above the dark current acompanying irradiation. The photoconductive gain is dete

FIG. 5. Geometry and bias of a photoconductor.


for

ed

-ri-n-lown

al-m-

sr,al

e-hetsr-od-n-inis

e-s

ry

dreofsrityrgec-Itatr,ee

mined by the properties of the detector, i.e., by which detetion effect is used and the material and configuration of tdetector.

In general, photoconductivity is a two-carrier phenomenon and the total photocurrent of electrons and holes is

I ph5wtq~Dnme1Dpmh!Vb

l, ~16!

whereme is the electron mobility,mh is the hole mobility,Vb

is the bias voltage andn5n01Dn, p5p01Dp. ~17!

The termsn0 and p0 are the average thermal equilibriumcarrier densities, andDn andDp are the excess carrier concentrations.

Taking the conductivity to be dominated by electrons~inall known high sensitivity photoconductors this is found tbe the case! and assuming uniform and complete absorptioof the light in the detector, the rate equation for the exceelectron concentration in the sample is29

dDn

dt5

Fsh

t2

Dn

t, ~18!

wheret is the excess carrier lifetime. In the steady conditiothe excess carrier lifetime is given by the equation

t5Dnt

hFs. ~19!

Equating~15! to ~16! gives

g5tVhmeDn

lhFs~20!

and invoking Eq.~19! we get for the photoconductive gain

g5tmeVb

l 25

t

l 2/meVb. ~21!

So, the photoconductive gain can be defined as

g5t

t t, ~22!

wheret t is the transit time of electrons between ohmic cotacts. This means that the photoconductive gain is giventhe ratio of free carrier lifetime,t, to transit time,t t betweenthe sample electrodes. The photoconductive gain can bethan or greater than unity depending upon whether the dlength, Ld5vdt, is less than or greater than interelectrodspacing, 1. The value ofLd.1 implies that a free chargecarrier swept out at one electrode is immediately replacedinjection of an equivalent free charge carrier at the opposelectrode. Thus, a free charge carrier will continue to circlate until recombination takes place.

WhenRL@R, a signal voltage across the load resistoressentially the open circuit voltage

Vs5I phRd5I phl

qwtnme, ~23!

where Rd is the detector resistance. Assuming that tchange in conductivity upon irradiation is small comparedthe dark conductivity, the voltage responsivity is expressas


c-he

-

-

onss

n,

n-by

lessrifte

byiteu-

is

hetoed

Rv5Vs

Pl5

h

Iwt

lt

hc

Vb

n0, ~24!

where the absorbed monochromatic powerPl5FsAhn.This expression shows clearly the basic requirements

high photoconductive responsivity at a given wavelengthl:one must have high quantum efficiencyh, long excess car-rier lifetime t, the smallest possible piece of crystal, lowthermal equilibrium carrier concentrationsn0, and the high-est possible bias voltageVb .

The frequency dependent responsivity can be determinby the equation

Rv5h

lwt

ltefhc

Vb

n0

l

~ l1v2tef2 !1/2

, ~25!

wheretel is the effective carrier lifetime. Usually, the photoconductive detectors fabricated from wide band-gap mateals reveal strongly sub-linear responses and excitatiodependent response time is observed even at relativelyexcitation levels. This can be attributed to the redistributioof the charge carriers with increased excitation level.

The above simple model takes no account of additionlimitations related to the practical conditions of photoconductor operation such as sweep-out effects or surface recobination. These are specified below.

2. Sweep-out effects

Equation~25! shows that voltage responsivity increasemonotonically with the increase of bias voltage. Howevethere are two limits on applied bias voltage, namely: thermconditions~joule heating of the detector element! and sweepout of minority carriers. The thermal conductance of the dtector depends on the device fabrication procedure. If texcess carrier lifetime is long, we cannot ignore the effecof contacts and of drift and diffusion on the device perfomance. Present-day material technology is such that at merate bias fields minority carriers can drift to the ohmic cotacts in a short time compared to the recombination timethe material. Removal of carriers at an ohmic contact in thway is referred to as ‘‘sweep out.’’30,31 Minority carriersweep out limits the maximum applied voltage ofVb . Theeffective carrier lifetime can be reduced considerably in dtectors where the minority carrier diffusion length exceedthe detector length~even at very low bias voltages!. At lowbias the average drift length of the minority carriers is vemuch less than the detector length,I , and the minority carrierlifetime is determined by the bulk recombination modifieby diffusion to surface and contacts. The carrier densities auniform along the length of the detector. At higher valuesthe applied field, the drift length of the minority carriers icomparable to or greater than 1. Some of the excess minocarriers are lost at an electrode, and to maintain space-chaequilibrium, a drop in excess majority carrier density is neessary. This way the majority carrier lifetime is reduced.should be pointed out that the loss of the majority carriersone ohmic contact is replenished by injection at the othebut the minority carriers are not replaced. At high bias thexcess carrier density is nonuniformly distributed along thlength of the sample.


t

m

o

r

e

e

edyin-

inp-erns

ise.

oree to1/

eing1/atlysur-sla-

rs,e

/de

m-r ace

To achieve high photoelectric gain, low resistance alow surface recombination velocity contacts are requireThe metallic contacts are usually far from expectation. Tcontacts are characterized by a recombination velocity whcan be varied from infinity~ohmic contacts! to zero ~per-fectly blocking contacts!. In the latter case, a more intenseldoped region at the contact~e.g.,n1 for n-type devices orheterojunction contact! causes a built-in electric field tharepels minority carriers, thereby reducing recombination aincreasing the effective lifetime and the responsivity. Praccal realization of the device would require double epitaxigrowth and the subsequent removal of the heavy doped mterial from the active area of the device.

3. Noise mechanisms in photoconductors

All detectors are limited in the minimum radiant powewhich they can detect by some form of noise which maarise in the detector itself, in the radiant energy to which tdetector responds, or in the electronic system following tdetector. Careful electronic design, including low noise aplification, can reduce the system noise below that in toutput of the detector. This topic will not be treated here.

We can distinguish two groups of noise; the radiationoise and the noise internal to the detector. The radiatnoise includes signal fluctuation noise and background flutuation noise. Under most operating conditions the signfluctuation limit is operative for ultraviolet and visible detectors.

The random processes occurring in semiconductors grise to internal noise in detectors even in the absenceillumination. There are two fundamental processes respsible for the noise: fluctuations in the velocities of free cariers due to their random thermal motion, and fluctuationsthe densities of free carriers due to randomness in the ratethermal generation and recombination.32

Johnson–Nyquist noise is associated with the finitesistanceR of the device. This type of noise is due to thrandom thermal motion of charge carriers in the crystal anot due to fluctuations in the total number of these charcarriers. It occurs in the absence of external bias as a flucating voltage or current depending upon the method of msurement. Small changes in the voltage or current at theminals of the device are due to the random arrival of charat the terminals. The root mean square of the JohnsoNyquist noise voltage in the bandwidthD f is given by

Vj5~4kTRD f !1/2, ~26!

wherek is the Boltzmann constant andT is the temperature.This type of noise has a ‘‘white’’ frequency distribution.

At finite bias currents, the carrier density fluctuationcause resistance variations, which are observed as noiseceeding Johnson–Nyquist noise. This type of excess noisphotoconductive detectors is referred to as generatiorecombination~g–r! noise. G–r noise is due to the randomgeneration of free charge carriers by the crystal vibratioand their subsequent random recombination. Because ofrandomness of the generation and recombination procesit is unlikely that there will be exactly the same number ocharge carriers in the free state at succeeding instance


ndd.heich

y

ndti-ala-

ryhehe-

he

nionc-al-

iveofn-r-ins of

e-endgetu-a-ter-gen–

sex-in

n–

nstheses,fs of

time. This leads to conductivity changes that will be reflectas fluctuations in current flow through the crystal. Manforms of g–r noise expression exist, depending upon theternal properties of the semiconductors.

The rms g–r noise current for an extrinsicn-type photo-conductor with carrier lifetimet can be written32

I g–r2 5

4I 2DN2tD f

N2~11v2t2!, ~27!

whereN is the number of carriers in the detector. Usually,an extrinsic semiconductor there will be some counterdoing, i.e., electrons trapped at deep lying levels. If the numbof deep traps is small compared to the number of electro~electrons being the majority carriers!, then the varianceDN2

is equal toN32. The current flowing in the device isI5Nqg/t, hence

I g–r2 5

4qIgD f

11v2t2. ~28!

Generation–recombination noise usually dominates the nospectrum of photoconductors at intermediate frequencies

An additional type of noise, referred to as 1/f noise,because it exhibits an approximately 1/f power law spec-trum, is always observed. It is less understood than the mfundamental noise sources and is not generally amenablmathematical analysis. The general expression for thefnoise current is

I 1/f5SKI baD f

f b D 1/2, ~29!

whereK is a proportionality factor,I b is the bias current,a isa constant whose value is about 2, andb is a constant whosevalue is about unity.

In general, 1/f noise appears to be associated with thpresence of potential barriers at the contacts, surface trappphenomena, and surface leakage currents. Reduction offnoise to an acceptable level is an art which depends greon the processes employed in preparing the contacts andfaces. Up until now, no fully satisfactory general theory habeen formulated. The two most current models for the expnation of 1/f noise were considered: Hooge’s model,33 whichassumes fluctuations in the mobility of free charge carrieand McWhortel’s model,32 based on the idea that the frecarrier density fluctuates.

The UV semiconductor detectors usually exhibit 1fnoise at low frequency. At higher frequencies the amplitudrops below that of one of the other types of noise.

4. Quantum efficiency

In most photoconductor materials the internal quantuefficiencyh0 is nearly unity; that is, almost all photons absorbed contribute to the photoconductive phenomenon. Fodetector, as a slab of material, shown in Fig. 5, with surfareflection coefficientsr 1 and r 2 ~on the top and bottom sur-


r,

a

ina-iers

faces, respectively!, and absorption coefficienta, the internalphotogenerated charge profile in they direction is34

S~y!5h0~12r 1!a

12r 1r 2 exp~22at !@exp~2ay!

1r 2 exp~22at !exp~2ay!#. ~30!

The external quantum efficiency is simply the integral of thfunction over the detector thickness

h5E0

t

S~y!dy

5h0~12r 1!@11r 2 exp~2at !#@12exp~at !#

12r 1r 2 exp~22at !. ~31!

When r 1 and r 25r , the quantum efficiency is reduced to

h5h0~12r !@12exp~at !#

12r exp~2at !. ~32!

Intrinsic detector materials tend to be highly absorptive; soa practical well-designed detector assembly only the top sface reflection term is significant, and then


is

inur-

h'h0~12r !'12r . ~33!

By anti-reflection coating the front surface of the detectothis quantity can be made greater than 0.9.

5. Influence of surface recombination

The photoconductive lifetime in general provideslower limit to the bulk lifetime, due to the possibility ofenhanced recombination at the surface. Surface recombtion reduces the total number of steady-state excess carrby reducing the recombination time. It can be shown thattefis related to the bulk lifetime by the expression35

teft

5A1

a2LD2 21

, ~34!

where

the

A15LDaF ~aDD1s1!$s2@ch~ t/LD!21#1~DD /LD!sh~ t/LD!%

~DD /LD!~s11s2!ch~ t/LD!1~DD2 /LD

2 1s1s2!sh~ t/LD!

3~aDD2s2!$s1@ch~ t/LD!21#1~DD /LD!sh~ t/LD!%eat

~DD /LD!~s11s2!ch~ t/LD!1~DD2 /LD

2 1s1s2!sh~ t/LD!2~12e2at!G

and

DD5Dep0mh1Dhn0me

n0me1p0mh

is the ambipolar diffusion coefficient,s1 ands2 are the surface recombination velocities at the front and back surfaces ofphotoconductor, andLD5(DDt)1/2. Other marks have their usual meanings,De,h5(kT/q)me,h are the respective carrierdiffusion coefficients for electrons and holes.

If the absorption coefficienta is largee2at'0 ands1!aDD , Eq. ~34! is simplified to the well-known expression30

teft

5DD

LD

s2@ch~ t/LD!21#1~DD /LD!sh~ t/LD!

LD~DD /LD!~s11s2!ch~ t/LD!1~DD2 /LD

2 1s11s2!sh~ t/LD!. ~35!

for

is,-

oninemn-ityrs

Further simplification fors15s25s leads to

1

tef51

t12s

t. ~36!

C. p -n junction photodiodes

Photoeffects which occur in structures with built-in potential barriers are essentially photovoltaic and result whexcess carriers are injected optically into the vicinity of sucbarriers. The role of the built-in electric field is to cause thcharge carriers of opposite sign to move in opposite diretions depending upon the external circuit. Several structuare possible to observe the photovoltaic effect. These inclup-n junctions, heterojunctions, Schottky barriers, and met

-enhec-resdeal-

insulator-semiconductor~MIS! photocapacitors. Each ofthese different types of devices has certain advantagesUV detection, depending on the particular applications.

The most common example of a photovoltaic detectorthe abrupt p-n junction prepared in the semiconductorwhich is often referred to simply as a photodiode. The operation of thep-n junction photodiode is illustrated in Fig. 6.Photons with energy greater than the energy gap, incidentthe front surface of the device, create electron-hole pairsthe material on both sides of the junction. By diffusion, thelectrons and holes generated within a diffusion length frothe junction reach the space-charge region. Then electrohole pairs are separated by the strong electric field; minorcarriers are readily accelerated to become majority carrie


t

fes

-duit

gce

at

is

dn-

n

ns

ther

-

n

on the other side. This way a photocurrent is generawhich shifts the current–voltage characteristic in the diretion of negative or reverse current, as shown in Fig. 6~c!. Theequivalent circuit of a photodiode is shown in Fig. 6~b!.

The total current density in thep-n junction is usuallywritten as

J~V,F!5Jd~V!2Jph~F!, ~37!

where the dark current density,Jd , depends only onV andthe photocurrent depends only on the photon flux densityF.

Generally, the current gain in a simple photovoltaic dtector ~e.g., not an avalanche photodiode! is equal to 1, andthen, according to Eq.~14!, the magnitude of photocurrent isequal

I ph5hqAF. ~38!

The dark current and photocurrent are linearly independ~which occurs even when these currents are significant! andthe quantum efficiency can be calculated in a straightforwamanner.36

FIG. 6. Schematic representation of the operation of ap-n junction photo-diode: ~a! geometrical model of the structure,~b! equivalent circuit of anilluminated photodiode~the series resistance includes the contact resistaas well as the bulkp- andn-regions!, ~c! current–voltage characteristics forthe illuminated and nonilluminated photodiode.


edc-

e-

ent

rd

If the p-n diode is open circuited, the accumulation oelectrons and holes on the two sides of the junction producan open-circuit voltage@Fig. 6~c!#. If a load is connected tothe diode, a current will conduct in the circuit. The maximum current is realized when an electrical short is placeacross the diode terminals, and this is called the short-circcurrent.

The open-circuit voltage can be obtained by multiplyinthe short-circuit current by the incremental diode resistanR5(]I /]V)21 at V5Vb:

Vph5hqAFR, ~39!

whereVb is the bias voltage andI5 f (V) is the current–voltage characteristic of the diode.

In many direct applications the photodiode is operatedzero-bias voltage

R05S ]I

]VDuVs50

21

. ~40!

A frequently encountered figure of merit for a photodiodetheR0A product

R0A5S ]J

]VDuVs50

21

, ~41!

whereJ51/A is the current density.In the detection of radiation, the photodiode is operate

at any point of theI –V characteristic. Reverse bias operatiois usually used for very high frequency applications to reduce the RC time constant of the devices.

1. Ideal diffusion-limited p -n junctionsa. Diffusion current. Diffusion current is the fundamen-

tal current mechanism in ap-n junction photodiode. Figure6~a! shows a one-dimensional photodiode model with aabrupt junction where the spatial charge of widthw sur-rounds the metalographic junction boundaryx5t, and twoquasineutral regions~0,xn! and (xn1w,t1d) are homoge-neously doped. The dark current density consists of electroinjected from then side over the potential barrier into thepside and an analogous current due to holes injected fromp side into then side. The current–voltage characteristic foan ideal diffusion-limited diode is given by

I D5AJgFexpS qVkTD21G . ~42!

For a junction with thick quasineutral regions the saturation current density is equal to

Js5qDhpn0Lh

1qDenp0Ls

, ~43!

where36,37

ce

Js5qDhpn0Lh

g1ch~xn /Lh!1sh~xn /Ln!

g1sh~xh /Lh!1ch~xh /Lh!1qDenp0Le

g2ch@~ t1d2xn2w!/Le#1sh@~ t1d2xn2w!/Le#

g2sh@~ t1d2xn2w!/Le#1ch@~ t1d2xn2w!/Le#~44!


-

t

in which caseg15s1Lh/Dh , g25s2Le/De , pn0 , andnp0 arethe concentrations of minority carriers on both sides ofjunction,s1 ands2 are the surface recombination velocitiesthe illuminated~for holes inn-type material! and back pho-todiode surface~for electrons inp-type material!, respec-tively. The value of the saturation current density,Js , de-pends on minority carrier diffusion lengths (Le ,Lh),minority carrier diffusion coefficients (De ,Dh), surface re-combination velocities (s1 ,s2), minority carrier concentra-tions (pn0 , np0) and junction design (xn ,t,w,d).

For nondegenerated statistics,nn0pn05ni2, D5(kT/q)m

andL5(Dt)1/2 and then

Js5~kT!1/2ni2q1/2F 1

pp0S me

teD 1/21 1

nn0S mh

thD 1/2G21

,

~45!

whereni is the intrinsic carrier concentration,pp0 and nn0are the hole and electron majority carrier concentrations,


theat

and

te andth the electron and hole lifetimes in thep- andn-typeregions, respectively. Diffusion current varies with temperature asni

2.The resistance at zero bias can be obtained from Eq.~40!

by differentiation ofI –V characteristic

R05kT

qIn~46!

and then theR0A product determined by diffusion current is

~R0A!D5S dJDdV DuVs50

21

5kT

qJs. ~47!

b. Quantum efficiency.Three regions contribute to pho-todiode quantum efficiency: two neutral regions of differentypes of conductivity and the spatial charge region~see Fig.6!. Thus36,37

h5hn1hDR1hp , ~48!

hn5~12r !aLha2Lh

221 H aLh1g12e2axn@g1ch~xn /Lh!1sh~xn /Lh!#

g1sh~xn /Lh!1ch~xn /Lh!2aLhe

2axnJ , ~49!

hp5~12r !aLea2Le

221e2a~xn1w!H ~g22aLe!e

2a~ t1d2xn2w!2sh@~ t1d2xn2w!/Le#2g2ch@~ t1d2xn2w!/Le#

ch@~ t1d2xn2w!/Le#2g2sh@~ t1d2xn2w!/Le#1aLeJ , ~50!

mrn

hr

tte

t

bcu

g

-al

in

gs,aln–in

ro,ted

-r-

ngin

de

hDR5~12r !@e2axn2e2a~xn1w!#. ~51!

In the following we shall consider the internal quantuefficiency, neglecting the losses due to reflection of thediation from the illuminated photodiode surface. Obtainihigh photodiode quantum efficiency requires that the illumnated region of the junction be sufficiently thin so that tgenerated carriers may reach the junction potential barriediffusion.

Normally, the photodiode is designed so that most ofradiation is absorbed in one side of the junction, e.g., inp-type side in Fig. 6~a!. This could be achieved in practiceither by making then-type region very thin or by using aheterojunction in which the band gap in then region is largerthan the photon energy so that most of the incident radiacan reach the junction without being absorbed. If the bacontact is several minority carrier diffusion lengths,Le awayfrom the junction, the quantum efficiency is given by

h~l!5~12r !a~l!Le

11a~l!Le. ~52!

If the back contact is less than a diffusion length away frothe junction, the quantum efficiency tends to

h~l!5~12r !@12e2a~l!d#, ~53!

whered is the thickness of thep-type region. It has beenassumed that the back contact has zero surface recomtion velocity and that no radiation is reflected from the basurface. Thus, if the above conditions hold, a high quantefficiency can be achieved using an anti-reflection coating

a-gi-eby

hehe

ionck

m

ina-kmto

minimize the reflectance of the front surface and ensurinthat the device is thicker than the absorption length.

c. Noise. In comparison with photoconductive detectors, the two fundamental processes responsible for thermnoise mechanisms~fluctuations in the velocities of free car-riers due to their random motion, and due to randomnessthe rates of thermal generation and recombination! are lessreadily distinguishable in the case of junction devices, givinrise jointly to shot noise on the minority carrier componentwhich make up the net junction current. The random thermmotion is responsible for fluctuations in the diffusion rates ithe neutral regions of junction devices and generationrecombination fluctuations both in the neutral regions andthe depletion region. We will show later that for a junctiondevice at zero bias, i.e., when the net junction current is zethe resulting noise is identical to Johnson noise associawith the incremental slope of the device.

A general theory of noise in photodiodes, which is applicable at arbitrary bias and to all sources of leakage curent, has not been developed.38 The intrinsic noise mecha-nism of a photodiode is shot noise in the current passithrough the diode. It is generally accepted that the noisean ideal diode is given by

I n25@2q~ I D12I s!14kT~Gj2G0!#D f , ~54!

whereI D5I s@exp(qV/kT)21#, GJ is the conductance of thejunction andG0 is the low-frequency value ofGJ . In thelow-frequency region, the second term on the right-hand siis zero. For a diode in thermal equilibrium~i.e., without ap-


h

s

o

be

l

m

usoa

thnh

n

f

de-on-veideted.de-in-taladeact,h

inr,ally

thehensce.

re-

re-

ion

e-

ha-la-is,eofe’s-ri-m-

plied bias voltage and external photon flux! the mean-squarenoise current is just the Johnson–Nyquist noise of the ptodiode zero bias resistance (R0

215qLs/kT)

I n25

4kT

R0D f ~55!

and

Vn254kTR0D f . ~56!

Note that the mean-square shot noise in reverse biahalf the mean-square Johnson–Nyquist noise at zero bHowever, this predicted improvement in the diode noisereverse bias is not normally observed in practice due tocreased 1/f noise.

d. Detectivity. In the case of the photodiode, the photelectric gain is usually equal to 1 so, according to Eq.~1!, thecurrent responsitivity is

Ri5ql

hch ~57!

and detectivity@see Eq.~3!# can be determined as

D*5hlq

hc F A

2q~ I D12I s!G1/2. ~58!

In reverse biases,I d tends to2I s and the expression inbrackets tends toI s .

From the discussion it is shown that the performancean ideal diffusion-limited photodiode can be optimizedmaximizing the quantum efficiency and minimizing the rverse saturation current,I s . For a diffusion-limited photodi-ode, the general expression for the saturation current of etron from thep-type side is@see Eq.~44!#

I sp5A

qDenp0Le

sh~d/Le!1~s2Le /De!ch~d/Le!

ch~d/Le!1~s2Le /De!sh~d/Le!. ~59!

It is normally possible to minimize the leakage current frothe side which does not contribute to the photosignal. Tminority carrier generation rate and hence the diffusion crent can be greatly reduced, in theory at least, by increadoping or the band gap on the inactive side of the juncti

If the back contact is several diffusion lengths awfrom the junction, then Eq.~59! tends to

I s5qDenp0Ls

A. ~60!

As the back contact is brought closer to the junction,leakage current can either increase or decrease, dependiwhether the surface recombination velocity is greater tthe diffusion velocityDe/Le . In the limiting case whered!Le , the saturation current is reduced by a factord/Lerelative to Eq.~60! for s50 and increased by a factor ofLe/dfor s5`. If the surface recombination velocity is small theEq. ~59! can usually be written in the form

I s5qGVdiff , ~61!

whereG is the bulk minority carrier generation rate per unvolume andVdiff is the effective volume of material fromwhich the minority carriers diffuse to the junction. The e


o-

isias.inin-

-

ofy-

ec-

her-ingn.y

eg onan

it

-

fective volume isALe for Le!d and tends toAd for [email protected] p-type material, the generation rate is given by

G5np0te

5ni2

Ndt. ~62!

The discussion indicates that the performance of thevice is strongly dependent on the properties of the back ctact. The most common solution to this problem is to mothe back contact many diffusion lengths away to one sand to ensure that all the surfaces are properly passivaAlternatively, the back contact itself can sometimes besigned to have a low surface recombination velocity bytroducing a barrier for minority carriers between the mecontact and the rest of the device. This barrier can be mby increasing the doping or the band gap near the contwhich effectively isolates the minority carriers from the higrecombination rate at the contact.

2. Other current mechanisms

In the previous section, photodiodes were analyzedwhich the dark current was limited by diffusion. Howevethis behavior is not always observed in practice, especifor wide-gap semiconductorp-n junctions. Several addi-tional excess mechanisms are involved in determiningdark current–voltage characteristics of the photodiode. Tdark current is the superposition of current contributiofrom three diode regions: bulk, depletion region, and surfaBetween them we can distinguish:

1. Thermally generated current in the bulk and depletiongion

~a! diffusion current in the bulkp andn regions,~b! generation–recombination current in the depletion

gion,~c! band-to-band tunneling,~d! intertrap and trap-to-band tunneling,~e! anomalous avalanche current,~f! ohmic leakage across the depletion region,

2. Surface leakage current

~a! surface generation current from surface states,~b! generation current in a field-induced surface deplet

region,~c! tunneling induced near the surface,~d! ohmic or nonohmic shunt leakage,~e! avalanche multiplication in a field-induced surface r

gion.

3. Space-charge limited current.

Figure 7 illustrates schematically some of these mecnisms. Each of the components has its own individual retionship to voltage and temperature. On account of thmany researchers analyzing theI –V characteristics assumthat only one mechanism dominates in a specific regionthe diode bias voltage. This method of analysis of the diodI –V curves is not always valid. A better solution is to numerically fit the sum of the current components to expemental data over a range of both applied voltage and teperature.


,

a

rlrh

d

a

ois

of

hi-ge

nt

g

by-

f

lea-

e-ive

n

sateandd.ctrter

-

Below, we will be concerned with the current contribution of high quality photodiodes with highR0A productslimited by:

~1! generation–recombination within the depletion region~2! tunneling through the depletion region,~3! surface effects,~4! impact ionization,~5! space-charge limited current.

a. Generation–recombination current. The importanceof this current mechanism was first pointed out by Set al.,39 and was later extended by Choo.40 It appears thatone space-charge region generation–recombination curcould be more important than its diffusion current, especiaat low temperatures, although the width of the space-charegion is much less than the minority carrier diffusion lengtThe generation rate in the depletion region can be very mugreater than in the bulk of the material. In reverse bias, tcurrent can be given by an equation similar to Eq.~61!,

I5qGdepVdep, ~63!

wheregdep is the generation rate andVdep is the volume ofthe depletion region. In particular, the generation rate frotraps in the depletion region is given by the Shockley-ReaHall formula

Gdep5ni2

n1te01p1tn0, ~64!

where n1 and p1 are the electron and hole concentrationwhich would be obtained if the Fermi energy was at the trenergy andte0 and th0 are the lifetimes in stronglyn-typeand p-type material. Normally, one of the terms in the denominator of Eq.~64! will dominate and for the case of atrap at the intrinsic level holdsni andpi5ni , giving

Gdep5nit0

~65!

and then the generation–recombination current of the deption region is equal

JGR5qwni

t0. ~66!

FIG. 7. Schematic representation of some of the mechanisms by which dcurrent is generated in a reverse-biasedp-n junction ~after Ref. 31!.


-

h

entlyge.chhe

m–

sp

-

le-

The comparison of Eqs.~62! and ~65! indicates that thegeneration rate in the bulk of the material is proportional tni2, whereas the generation rate in the depletion regionproportional to ni for a midgap state. The generation–recombination current varies roughly as the square rootthe applied voltage (w;V1/2) for the case of an abrupt junc-tion, or as the cube root (w;V1/2) for a linearly gradedjunction. This behavior, in which the current increases witreverse bias, can be contrasted with a diffusion-limited dode, where the reverse current is independent of voltaabove a fewkT/q.

Space-charge region generation–recombination currevaries with temperature asni , i.e., less rapidly than diffusioncurrent which varies asni

2.The zero bias resistance can be found by differentiatin

Eq. ~66! and settingV50

~R0A!GR5S dJGRdV DuV50

21

52Vbt0qniw

, ~67!

where Vb5kT ln (NaNd/ni2). In evaluating Eq.~67!, the

term of greatest uncertainty ist0.b. Tunneling current. The third type of dark current

component that can exist is a tunneling current causedelectrons directly tunneling across the junction from the valence band to the conduction band~direct tunneling! or byelectrons indirectly tunneling across the junction by way ointermediate trap sites in the junction region@indirect tunnel-ing or trap-assisted tunneling~see Fig. 7!#.

The usual direct tunneling calculations assume a particof constant effective mass incident on a triangular or parbolic potential barrier.41 For the triangular potential barrier

JT5q2EVb4p2\2 S 2m*

EgD 1/2expF2

4~2m* !12Eg

3/2

3q\EG . ~68!

For an abruptp-n junction the electric field can be approxi-mated by

E5F 2q«0«sSEg

q6VbD np

n1pG1/2. ~69!

The tunnel current is seen to have an extremely strong dpendence on energy gap, applied voltage, and an effectdoping concentrationNef5np/(n1p). It is relatively insen-sitive to temperature variation and the shape of the junctiobarrier. For the parabolic barrier41

JT} expF2~pm* !1/2Eg

3/2

23/2q\E G . ~70!

Apart from direct band-to-band tunneling, tunneling ispossible by means of indirect transitions in which impuritieor defects within the space-charge region act as intermedistates. This is a two-step process in which one step isthermal transition between one of the bands and the trap athe other is tunneling between the trap and the other banThe tunneling process occurs at lower fields than direband-to-band tunneling because the electrons have a shodistance to tunnel~see Fig. 7!.

c. Surface leakage current.In a realp-n junction, par-ticularly in wide gap semiconductors and at low tempera

ark


is

c

a

t

c

s

T

r-recer-ce

e-

ri-

r-

or

in

ct

ar-de-ite

nlengr-

ingtiont,,

.

tures, additional dark current related to the surface occuSurface phenomena play an important part in determinphotovoltaic detector performance. The surface providediscontinuity which can result in a large density of interfacstates. These generate minority carriers by the ShockleRead–Hall~SRH! mechanism, and can increase both the dfusion and depletion region generated currents. The surfcan also have a net charge, which affects the position ofdepletion region at the surface.

The surface of actual devices is passivated in orderstabilize the surface against chemical and heat-induchanges as well as to control surface recombination, leakaand related noise. Native oxides and overlying insulatorscommonly employed inp-n junction fabrication which in-troduce fixed charge states which then induce accumulaor depletion at the semiconductor-insulator surface. We cdistinguish three main types of states on the semiconducinsulator interface, namely, fixed insulator charge, low suface states, and fast surface states. The fixed charge ininsulator modifies the surface potential of the junction.positively charged surface pushes the depletion region fther into thep-type side and a negatively charged surfapushes the depletion region towards then-type side. If thedepletion region is moved towards the more highly dopside, the field will increase, and tunneling becomes molikely. If it is moved towards the more lightly doped side, thdepletion region can extend along the surface, greatlycreasing the depletion region generated currents. Whenficient fixed charge is present, accumulated and invertedgions as well asn-type andp-type surface channels areformed ~see Fig. 8!. An ideal surface would be electricallyneutral and would have a very low density of surface statThe ideal passivation would be a wide gap insulator growwith no fixed charge at the interface.

Fast interface states, which act as generatiorecombination centers, and fixed charge in the insulacause a variety of surface-related current mechanisms.kinetics of generation–recombination through fast surfastates is identical to that through bulk SRH centers. The crent in a surface channel is given by@see Eq.~66!#

FIG. 8. Effect of fixed insulator charge on the effective junction spaccharge region:~a! flatband condition;~b! positive fixed charge~inversion ofthep side, formation of ann-type surface channel!; ~c! negative fixed chan-nel ~accumulation of thep side, field induced junction at the surface!; ~d!large negative fixed charge~inversion of then side, formation ofp-typesurface channel!.


rs.ngaey–if-acethe

toedge,re

ionantor-r-theAur-e

edreein-uf-re-

es.n

n–torheceur-

IGRS5qniwcAc

t0, ~71!

wherewc is the channel width andAc is the channel area.Apart from generation–recombination processes occu

ring at the surface and within surface channels, there aother surface-related current mechanisms, termed surfaleakage, with ohmic or breakdownlike current–voltage chaacteristics. They are nearly temperature independent. Surfabreakdown occurs in a region of high electric field@Fig. 8~c!,where the depletion layer intersects the surface; Fig. 8~d!,very narrow depletion layer#.

Band bending at thep-n junction surface can be con-trolled by a gate electrode overlaid around the junction primeter on an insulating film.

The dark current as a sum of several independent contbutions can also be written as follows:

I5I sFexpq~V2IRs!

bkT21G1

V2IRs

Rsh1I T , ~72!

whereRs is the series resistance andRsh is the shunt resis-tance of the diode. In the case of predominant diffusion curent the b coefficient approaches unity; but when thegeneration–recombination current is mainly responsible fcarrier transport,b equals 2.

d. Impact ionization, avalanche photodiode.When theelectric field in a semiconductor is increased above a certavalue, the carriers gain enough energy~greater than the bandgap! so that they can excite electron–hole pairs by impaionization. An avalanche photodiode~APD! operates by con-verting each detected photon into a cascade of moving crier pairs. The device is a strongly reverse-biased photodioin which the junction electric field is large; the charge carriers therefore accelerate, acquiring enough energy to excnew carriers by the process of impact ionization.

The avalanche multiplication process is illustrated iFig. 9. A photon absorbed at point 1 creates an electron-hopair. The electron accelerates under the effect of the stroelectric field. The acceleration process is constantly interupted by random collisions with the lattice in which theelectron loses some of its acquired energy. These competprocesses cause the electron to reach an average saturavelocity. The electron can gain enough kinetic energy thaupon collision with an atom, it can break the lattice bonds

e-FIG. 9. Schematic representation of the multiplication process in an APD


a

t

l

t

i

e

i

ron-

utaa-thegtheinsta-nter-ex-

ethe,re5.

c-

nt

creating a second electron-hole pair. This is called impionization ~at point 2!. The newly created electron and holboth acquire kinetic energy from the field and create adtional electron-hole pairs~e.g., at point 3!. These in turncontinue the process, creating other electron-hole pairs. Tprocess is therefore called avalanche multiplication.

The abilities of electrons and holes to impact ionize acharacterized by the ionization coefficientsae andah . Thesequantities represent ionization probabilities per unit lengThe ionization coefficients increase with the depletion layelectric field and decrease with increasing device tempeture.

An important parameter for characterizing the perfomance of an APD is the ionization ratiok5ah/ae . Whenholes do not ionize appreciably~i.e.,ah!ae , k!1!, most ofthe ionization is achieved by electrons. The avalanching pcess then proceeds principally from left to right~i.e., fromthe p side ton side! in Fig. 9. It terminates some time latewhen all the electrons arrive at then side of the depletionlayer. If electrons and holes both ionize appreciably~k'1!,on the other hand, those holes moving to the left create etrons that move to the right, which, in turn, generate furthholes moving to the left, in a possibly unending circulatioAlthough this feedback process increases the gain of thevice ~i.e., the total generated charge in the circuit per phocarrier pair!, it is nevertheless undesirable for several resons: it is time consuming and therefore reduces the devbandwidth, it is random and therefore increases the devnoise, and it can be unstable, thereby causing avalanbreakdown. It is therefore desirable to fabricate APDs fromaterials that permit only one type of carrier~either elec-trons or holes! to impact ionize. If electrons have the higheionization coefficient, for example, optimal behavior iachieved by injecting the electron of a photocarrier pairthe p edge of the depletion layer and by using a materwhose value ofk is as low as possible. If holes are injectedthe hole of a photocarrier pair should be injected at thenedge of the depletion layer andk should be as large as possible. The ideal case of single-carrier multiplicationachieved whenk50 or `.

In an optimally designed photodiode, the geometrythe APD should maximize photon absorption~for example,by assuming the form of ap- i -n structure! and the multipli-cation region should be thin to minimize the possibility olocalized uncontrolled avalanches~instabilities or microplas-mas! being produced by strong electric field. Greatelectric-field uniformity can be achieved in a thin regionThese two conflicting requirements call for an APD designwhich the absorption and multiplication regions are separaFor example, Fig. 10 shows a reach-throughp1 -p-p-n1

APD structure which accomplishes this. Photon absorptoccurs in the widep region ~very lightly dopedp region!.Electrons drift through thep region into a thinp-n1 junc-tion, where they experience a sufficiently strong electric fieto cause avalanching. The reverse bias applied acrossdevice is large enough for the depletion layer to reathrough thep andp regions into thep1 contact layer.

A comprehensive theory of avalanche noise in APD


ctedi-

his

re

h.erra-

r-

ro-

r

ec-ern.de-o-a-iceicechem

rsatial,

-s

of

f

r.inte.

on

ldthech

s

was developed by McIntyre.42 The noise of an APD per unitbandwidth can be described by the formula

^I n2&52qIph M &2F, ~73!

where I ph is unmultiplied photocurrent~signal!, ^M & is theaverage avalanche gain andF is the excess noise factowhich turns out to be related to the mean gain and the iization ratio by

F5k^M &1~12k!S 121

^M & D . ~74!

In p-n and p- i -n reverse-biased photodiodes withogain, ^M &51, F51 and the well-known shot noise formulwill indicate the device’s noise performance. In the avlanche process, if every injected photocarrier underwentsame gainM , the factor would be unity, and the resultinnoise power would only be the input shot noise due torandom arrival of signal photons, multiplied by the gasquared. The avalanche process is, instead, intrinsicallytistical, so that individual carriers generally have differeavalanche gains characterized by a distribution with an avage^M &. This causes additional noise called avalanchecess noise, which is conveniently expressed by theF factorin Eq. ~74!. As was mentioned above, to achieve a lowF, notonly mustae andah be as different as possible, but also thavalanche process must be initiated by carriers withhigher ionization coefficient. According to McIntyre’s rulethe noise performance of ADP can be improved by mothan a factor of 10 when the ionization ratio is increased to

e. Space-charge limited current.In the case of wideband-gapp-n junctions, the forward current–voltage charateristics are often described by equation

J} expS qV

bkTD , ~75!

where the diode ideality factorb.2. This value ofb doesnot fall within the range that results when diffusion curre~b51! or depletion layer current~b52! dominate the

FIG. 10. Reach-throughp1-p-p-n1 APD structure.


on

misc-

ted

n

-

t

beif-

he

-

allyheises

as

r-gk-ere

vs

forward-bias current conduction. This behavior is typicalan insulator with shallow and/or deep traps and thermgenerated carriers.

Space-charge limited~SCL! current flow in solids hasbeen considered in detail by Rose,43 Lampert, and Mark44,45

They loosely defined materials withEg<2 eV as semicon-ductors and those withEg>2 eV as insulators. It means, e.gthat GaN could be considered as an insulator.

When a sufficiently large field is applied to an insulatwith ohmic contacts, electrons will be injected into the buof the material to form a current which is limited by spaccharge effects. When trapping centers are present, theyture many of the injected carriers, thus reducing the denof free carriers.

At low voltages where the injection of carriers into thsemiinsulating material is negligible, Ohm’s law is obeyand the slope of theJ–V characteristics defines the resistiity r of the material. At some applied voltageVTH the currentbegins to increase more rapidly than linearly with applvoltages. HereVTH is given by

VTH54p31012qptt2

«, ~76!

where t is the thickness of the semiinsulating material, apt is the density of trapped holes, if one carrier~holes! space-charge limited current is considered. As the voltage is ctinuously increased beyondVTH additional excess holes arinjected into the material and the current density is given

J510213mh«uV2

t2, ~77!

whereu is the probability of trap occupation determined aratio of the densities of free to trapped holes. Hereu is givenby

u5p

pt5Nv

NtexpS 2

Et

kTD , ~78!

whereNv is the effective density of states in the valenband,Nt is the density of traps, andEt is the depth of trapsfrom the top of the valence band. When the applied voltafurther increases, the square-law region of Eq.~77! will ter-minate in a steeply rising current which increases untibecomes the trap-free SCL currents given by

J510213mh«V2

t2. ~79!

The density of trapsNt can be determined from the volageVTFL at which the traps are filled and the currents rsharply

VTFL54p31012qNt

t2

«. ~80!

The trap depthEt can also be calculated from the E~80! using the value ofNt . But if Eq. ~77! is measured as thfunction of temperature and a plot of ln~uT3/2! vs 1/T ispossible,Et andNt can be obtained directly from these dawithout referring to Eq.~80!.


forally

.,

orlke-cap-sity

eedv-

ied

nd

on-eby

s a

ce

ge

l it

t-ise

q.e

ta

For the purpose of illustration of the above phenomenin semiinsulating materials, consider the four logJ vs logVgraphs in Fig. 11. Figure 11~a! represents an ideal insulatorwhere I}V2, indicating SCL current flow. In other words,there are no thermally generated carriers resulting froimpurity-band or band-to-band transitions; the conductiononly within the conduction band as a result of carrier injetion. As shown in Fig. 11~b!, ohmic conduction is obtainedin trap-free insulators in the presence of thermally generafree carriers,n0. When the injected carrier densityninj ex-ceeds n0~ninj.n0!, ideal insulator characteristics areobserved (I}V2). Shallow traps contribute to anI}V2 re-gime at a lower voltage followed by a sharp transition to aideal insulator, square-law regime as shown in Fig. 11~c!.The sharp transition corresponds to an applied voltage.VTFLrequired to fill a discrete set of traps which are initially unoccupied. Figure 11~d! illustrates the case of a material withdeep traps which have become filled whenninj becomescomparable tono . The voltage at which this occurs isVTFL .Therefore, atV,VTFL ohmic conduction is observed and aV.VTFL SCL current flow dominates.

3. Response time

The upper-frequency response of a photodiode maydetermined by basically three effects: the time of carrier dfusion to the junction depletion region,t0; the transit time ofcarrier drift across the depletion region,ts ; and theRC timeconstant associated with circuit parameters including tjunction capacitanceC and the parallel combination of dioderesistance and external load~the series resistance is neglected!.

Photodiodes designed for fast response are generconstructed so that the absorption of radiation occurs in tp-type region. This ensures that most of the photocurrentcarried by electrons which are more mobile than hol~whether by diffusion or drift!. The frequency response forthe diffusion process in a backside illuminated diode hbeen calculated by Sawyeret al.47 as a function of diodethickness, diffusion constant, absorption depth, minority carier lifetime, and surface-recombination velocity. Assuminthat the diffusion length is greater than both the diode thicness and the absorption depth, the cutoff frequency whthe response drops by&, is given by31

FIG. 11. Schematic drawings of the logarithmic dependence of currentvoltage for ~a! an ideal insulator,~b! a trap-free insulator with thermallygenerated free carriers,~c! an insulator with shallow traps and thermal freecarriers, and~d! an insulator with deep traps and thermal-free carriers~afterRef. 46!.


a

t

o

i-n

e

sdtal

le

-

nri

of

i-

ser-

f diff52.43D

2pt2, ~81!

whereD is the diffusion constant andt is the diode thick-ness.

The depletion region transit time is equal to

t t5wdep

vs, ~82!

wherewdep is the depletion region width andvs is the carriersaturation drift velocity in the junction field, which hasvalue of about 107 cm/s. The frequency response of a transtime limited diode has been derived by Gartner.48

In practice the most serious limitation arises from thRC time constant, which for an abrupt junction is given b

tRC5ART2 S q«0«sN

V D 1/2. ~83!

RT5Rd(Rs1RL)/(Rs1Rd1RL) whereRs , Rd , andRL arethe series, diode, and load resistances. To reduce theRCtime constant, we can decrease the majority carrier conctration adjacent to the junction, increaseV by means of areverse bias, decrease the junction area, or lower eitherdiode resistance or the load resistance. Except for the apcation of a reverse bias, all of these changes reduce thetectivity. The trade-off between response time and detecity is then apparent.

D. Schottky barrier photodiodes

Schottky barrier photodiodes have been studied quextensively41 and have also found application as ultravioledetectors. These devices reveal some advantages overp-njunction photodiodes: fabrication simplicity, absencehigh-temperature diffusion processes, and high speed ofsponse.

1. Schottky –Mott theory and its modifications

According to a simple Schottky–Mott model, the rectfying property of the metal-semiconductor contact arisfrom the presence of an electrostatic barrier betweenmetal and the semiconductor which is due to the differenin work functionsfm andfs of the metal and semiconductorrespectively. For example, for a metal contact with ann-typesemiconductor,fm should be greater, while for ap-typesemiconductor it should be less thanfs . The barrier heightsin both these cases, shown in Figs. 12~a! and 12~b!, are givenby

FIG. 12. Equilibrium energy band diagram of Schottky barrier junctions:~a!metal-~n-type! semiconductor,~b! metal-~p-type! semiconductor.


it

ey

en-

thepli-de-iv-

itet

fre-

i-esthece,

fbn5fm2xs ~84!

and

fbp5xs1Eg2fm, ~85!

respectively, wherexs is the electron affinity of the semicon-ductor. The potential barrier between the interior of the semconductor and the interface, known as band bending, is giveby

cs5fm2fs ~86!

in both cases. Iffb.Eg , the layer of thep-type semicon-ductor adjacent to the surface is inverted in type and we hava p-n junction within the material. However, in practice thebuilt-in barrier does not follow such a simple relationshipwith fm and is effectively reduced due to interface stateoriginating either from surface states or from metal-inducegap states and/or due to interface chemical reactions of meand semiconductor atoms.

In the literature, there are numerous and considerabvariations among experimental data onfm .

49 Their analysisindicates an empirical relationship of the type60

fb5g1fm1g2 , ~87!

whereg1 andg2 are constants characteristic of the semiconductors. Two limit cases, namelyg150 andg151 indicativeof Bardeen barrier~when influence of localized surface statesis decisive! and ideal Schottky barrier, respectively, can bevisualized from such an empirical relation. It has also beepointed out by various workers that the slope parameteg15]fb/]fm can be used for describing the extent of Fermlevel stabilization or pinning for a given semiconductor. Theparametersg1 andg2 have been used by some workers forestimating the interface state density.

It was shown by Cowley and Sze50 that, according to theBardeen model, the barrier height in the case of ann-typesemiconductor is given approximately by

fon5g~fm2xs!1~12g!~Eg2f0!2Df, ~88!

whereg5« i(« i1qdDg). The termf0 is the position of theneutral level of the interface states measured from the topthe valence band,Df is the barrier lowering due to imageforces,d is the thickness of the interfacial layer, and«i itstotal permittivity. The surface states are assumed to be unformly distributed in energy within the band gap, with adensityDg per electron volt per unit area. If there are notsurface states,Dg50, and neglectingDf, Eq.~88! reduces toEq. ~84!. If the density of states is very high,g becomes verysmall andfbn approaches the valueEg2f0. This is becausea very small deviation of the Fermi level from the neutrallevel can produce a large dipole moment, which stabilizethe barrier height by a sort of negative feedback effect. ThFermi level is pinned relative to the band edges by the suface states.

A similar analysis for the case of ap-type semiconductorshows thatfbp is approximately given by

fbp5g~Eg2fm1xs!1~12g!f0 . ~89!


y

s

rf

t

ca

il

t

g

oweoreheten-ityier

,

are

-nssi-ty

to

p-andee

fheec-ntd-

t-theeeally

Hence if fbn and fbp refer to the same metal onn- andp-type specimens of the same semiconductor, we shohave

fbn5fbp>Eg ~90!

if the semiconductor surface is prepared in the same waboth cases, so thatd, «i , Dg , and f0 are the same. Thisrelationship holds quite well in practice. It is usually true thfbn.Eg/2, andfbn.fbp.

2. Current transport processes

The current transport in metal-semiconductor contactdue mainly to majority carriers, in contrast top-n junctions,where current transport is due mainly to minority carrieThe current can be transported in various ways underward bias conditions as shown in Fig. 13. The four procesare:49

~a! emission of electrons from the semiconductor overtop of the barrier into the metal,

~b! quantum mechanical tunneling through the barrier,~c! recombination in the space-charge region,~d! recombination in the neutral region~equivalent hole

injection from the metal to the semiconductor!.

The inverse processes occur under reverse bias. In ation, we may have edge leakage current due to a high elefield at the contact periphery or interface current due to trat the metal-semiconductor interface.

The transport of electrons over the potential barrier habeen described by various theories, namely: diffusion,51,52

thermionic emission,53 and unified thermionic emissiondiffusion.50 It is now widely accepted that, for high-mobilitysemiconductors with impurity concentrations of practicalterest, the thermionic emission theory appears to expqualitatively the experimentally observed I –Vcharacteristics.54 Some workers55 have also included in thesimple thermionic theory the quantum effects~i.e., quantum-mechanical reflection and tunneling of carriers throughbarrier! and have tried to obtain modified analytical expresions for the current–voltage relation. This, however, hessentially led to a lowering of the barrier height androunding off of the top.

The thermionic emission theory by Bethe53 is derivedfrom the assumptions that the barrier height is much larthankT, thermal equilibrium is established at the plane th

FIG. 13. Four basic transport processes in forward-biased Schottky baon ann-type semiconductor~after Ref. 49!.


uld

in

at

is

s.or-ses

he

ddi-tricps

ve

n-ain

hes-asa

erat

determines emission, and the existence of a net current fldoes not affect this equilibrium. Bethe’s criterion for thslope of the barrier is that the barrier must decrease by mthan kT over a distance equal to the scattering length. Tresulting current flow will depend only on the barrier heighand not on the width, and the saturation current is not depdent on the applied bias. Then the current density of majorcarriers from the semiconductor over the potential barrinto the metal is expressed as

JMSt5JstFexpS qV

bkTD21G , ~91!

where saturation current density

Jst5A*T2 expS 2fb

kTD ~92!

and A*54pqk2m* /h3 equal 120(m* /m) A cm22 K22 isthe Richardson constant,m* is the effective electron massandb is an empirical constant close to unity. Equation~91!is similar to the transport equation forp-n junctions. How-ever, the expression for the saturation current densitiesquite different.

It appears that the diffusion theory is applicable to lowmobility semiconductors and the current density expressioof the diffusion and thermionic emission theories are bacally very similar. However, the saturation current densifor the diffusion theory

Jsd5q2DeNc

kT Fq~Vbi2V!2Nd

«0«sG ~93!

varies more rapidly with the voltage but is less sensitivetemperature compared with the saturation current densityJstof the thermionic emission theory.

A synthesis of the thermionic emission and diffusion aproaches described above has been proposed by CrowellSze.50 They assumed Bethe’s criterion for the validity of ththermionic emission~the mean free path should exceed thdistance within which the barrier falls bykT/q from itsmaximum value! and also took into account the effects ooptical phonon scattering in the region between the top of tbarrier and the metal and of the quantum mechanical refltion of electrons which have sufficient energy to surmouthe barrier. Their combined effect is to replace the Richarson constantA* with A** 5 f pf qA* , where f p is the prob-ability of an electron reaching the metal without being scatered by an optical phonon after having passed the top ofbarrier, andf q is the average transmission coefficient. Thtermsf p and f q depend on the maximum electric field in thbarrier, the temperature, and the effective mass. Generspeaking, the productf pf q is on the order of 0.5.

3. R0A product and responsivity

Knowing Jst with Eq. ~92!, R0A can be calculated from

~R0A!MS5S dJMStdV DuV50

21

5kT

qJst5

k

qA*TexpS fb

kTD . ~94!

According to Eq.~1!, the current responsivity may bewritten in the form

rrier


ion

-asndi-ofars.ed

l as-

ier.

Ri5ql

hch ~95!

and the voltage responsivity

Rv5ql

hchR, ~96!

whereR5(dI/dV)21 is the differential resistance of the photodiode.

E. Comparison of different types of semiconductorphotodetectors

At this point it is important to discuss some importandifferences between photoconductive,p-n junction andSchottky barrier detectors.

The photoconductive detectors exhibit the important avantage of the internal photoelectric gain, which relaxes rquirements to a low noise preamplifier. The advantagesp-n junction detectors relative to photoconductors are: loor zero bias currents; high impedance, which aids couplito read-out circuits in focal plane arrays; capability for highfrequency operation and the compatibility of the fabricatiotechnology with planar-processing techniques. In compason with Schottky barriers, thep-n junction photodiodesalso indicate some important advantages. The thermioemission process in Schottky barrier is much more efficiethan the diffusion process and, therefore, for a given built-voltage, the saturation current in a Schottky diode is seveorders of magnitude higher than in thep-n junction. In ad-dition, the built-in voltage of a Schottky diode is smaller thathat of ap-n junction with the same semiconductor. However, high-frequency operation ofp-n junction photodiodesis limited by the minority-carrier storage problem. In othewords, the minimum time required to dissipate the carrieinjected by the forward bias is dictated by the recombinatilifetime. In a Schottky barrier, electrons are injected from thsemiconductor into the metal under forward bias if the semconductor isn type. Next they thermalize very rapidly~'1014 s! by carrier–carrier collisions, and this time is negligible compared to the minority-carrier recombination lifetime.

F. Semiconductor materials used for ultravioletdetectors

Different semiconductor compounds have been usedfabricate photodetectors with spectral responses in theregion. Figure 14 shows the spectral detectivity of opticdetectors responding in the 0.1–1.2mm region. Note thatdetectivity is notD* , but rather reciprocal of NEP for a 1 Hzbandwidth. This figure of merit is employed to include photomultipliers whose noise does not in all cases depend upthe square root of the photocathode area. Table II listsareas which Seib and Aukerman30 state are proper to thevarious detectors illustrated. The reader can convert toD* values appropriate to the photoconductive and photovtaic detectors by multiplying the detectivity value illustrate


-

t

d-e-ofwng-nri-

nicntinral

n-

rrsonei-

--

toUVal

-onthe

theol-d

by the square root of the detector area. The signal fluctuatlimit shown in Table II is independent of area~see Sec.III A !.

Historically, the development of a high quantum efficiency semiconductor detectors in the whole UV range hbeen hampered by the extremely strong absorption astrong radiation induced aging effects in most of the semconductor materials. It is especially visible in the casesilicon which, due to mature technology, is the most populsemiconductor material used for fabrication of UV detectorFor this semiconductor the photon penetration depth rangfrom less than 10 nm in the NUV and MUV to 100mm for10 keV x rays@see Fig. 15~a!#.8 The absorption lengths forthree semiconductor materials, Si, GaAs, and GaP as welfor SiO2 and Au~photoemissive gold diode is the most com

FIG. 14. Detectivity vs wavelength values of 0.1–1.2mm photodetectors.PC indicates a photoconductive detector and PM indicates a photomultiplDetector areas are given in Table II~after Ref. 26!.

TABLE II. Areas of detectors illustrated in Fig. 14.

Detector Area~cm2!

CdS photoconductor~PC! 1CdSe photoconductor~PC! 1Si Schottky barrier photodiode 0.03Si p-n junction photodiode 0.25Si photoconductor 0.25Si avalanche photodiode 0.07Ge photoconductor~PC! 0.20Ge ac bias photoconductor~PC! 2.431025

Photomultiplier~PM! 1.0


li-hletorpi-e-er-heinu-ofhcalate-el-ereVri-ntg-

eir

butalapr-sothegdNp-

r

leaNtelean

aN

mon transfer detector in the spectral range above 250 eV! areshown in Fig. 15~b! for photon energies between 30 eV an6 keV, corresponding to wavelengths between 40 andnm. Especially for photon energies below about 700 eVgreat part of the incident radiation is absorbed in the fiseveral hundred nanometers of a semiconductor detectorgarding the absorption lengths at higher photon energies,sensitive depth of a semiconductor detector should be oforder of at least 10mm to avoid the penetration of photonthrough this sensitive volume. For high-energy detectiusually the depletion layer is taken as sensitive volume.

Figure 14 shows that different semiconductors coverUV spectral range. However, the modern semiconductor

FIG. 15. Absorption lengths of silicon~a! and various materials~b! in theUV spectral region~after Refs. 8 and 56!.


d0.2, arst. Re-thetheson,

theUV

detectors are mainly fabricated using Si. Many of the appcations for UV detection involve hostile environments sucas in situ combustion monitoring and satellite-based missiplume detection, where the ruggedness of active UV detecmaterial is an important advantage. Other applications catalize on the sensitivity of wide band-gap semiconductor dtectors, such as air quality monitoring, gas sensing, and psonal UV exposure dosimetry. Industries such as taerospace, automotive, petroleum, and others have contously provided the impetus pushing the developmentfringe technologies which are tolerant of increasingly higtemperatures and hostile environments. In the field of optidevices, several trends are pushing research into new mrials. Considering the relatively advanced stage of SiC devopment, it appears that most high-temperature, high-powelectronic devices will be fabricated from that material. Thwide band-gap energy also permits fabrication of SiC Udetectors. A technological progress in fabrication of mateals based on the III–V nitrides together with their inherefavorable properties, make III–V nitrides the most promisinsemiconductor materials for application in the UV wavelength range.57–59 III–V nitride devices will be capable ofimproved high power and temperature operation due to thlarge band gaps.

GaN has many advantages over SiC for applications,lags in development. While its thermal conductivity, thermstability, chemical inertness, breakdown fields, and band gare all roughly comparable to SiC, its optoelectronic propeties will be superior. GaN has a direct band gap,momentum-conserving transitions connect states havingsamek values. Therefore, GaN is one of the most promisinmaterials for light-emitting devices in the blue, violet, anultraviolet spectral regions. Table III compares SiC and Gato the conventional semiconductors for power electronic aplications. The Johnson60 and Keyes61 figure of merit pro-vides an estimate of the suitability of a material for poweelectronics. Because of the wide band gap of GaN~Eg53.4eV! there is no responsivity to infrared radiation~long wave-length cutoff occurs at 365 nm! which is important in manyapplications whenever it is desirable to detect UV in a visiband infrared background. It should be expected that Gradiation resistance is superior to SiC. Initial results indicathat ohmic contacts to GaN will be superior to those possibwith SiC. Finally, the ease with which heterostructures c

TABLE III. Johnson ~Ref. 60! and Keyes~Ref. 61! figures of merit forelectronic devices fabrication in the Si, Ge, GaAs, 6H and 3C SiC, and Gsystems~after Ref. 58!.a

MaterialEvs/2p ~V/s!Johnson

st(cvs/4p«)1/2 ~W/deg!Keyes

Breakdown field~V/cm!

Si 231011 6.73107 43105

Ge 131011 1.53107 2.53105

GaAs 731011 2.73107 53105

6H–SiC 1.331013 353107 43106

3C–SiC 1.231013 393107 33106

GaN 1.631013 123107 on GaN 43106

30–403107 on SiC

aE0 breakdown field,vs saturation electron drift velocity,st thermal con-ductivity, « dielectric constant.


7454 J. App

TABLE IV. Comparison of important semiconductor properties for high-temperature electronics~after Refs. 57and 58!.

Parameter Si GaAs GaP3C–SiC

~6H–SiC! GaN Diamond

Lattice parameters~Å! a55.4301 a55.6533 a54.36 a53.189~a53.08! c55.185

~c515.12!Band gap~eV! 1.1 1.4 2.3 2.2 3.39 5.5

at 300 K ~2.9!Coefficient of thermal 3.59 6.0 5.59

expansion~104 K21! ~4.2! 3.17~4.68!

Maximum operating 600? 760? 1250? 1200 1400?temperature~K! ~1580!Melting point ~K! 1690 1510 1740 Sublimes Phase

.2100 changesPhysical stability Good Fair Fair Excellent Good Very goodElectron mobility~cm2/V s! 1400 8500 350 1000 900 2200

at 300 K ~600!Hole mobility ~cm2/V s! 600 400 100 40 150? 1600

at 300 KBreakdown voltage 0.3 0.4 ••• 4 5? 10

~106 V/cm!Thermal conductivity 1.5 0.5 0.8 5 1.3 20

~W/cm!Saturation electron drift 1 2 ••• 2 2.7 2

velocity ~107 cm/s!Static dielectric constant 11.8 12.8 11.1 9.7 9 5.5

f

eerppdoewn

rrho

lsf

cnn

d

an-enn-

be formed with AlN and lnN alloys is a great advantagemore complex device structures.

The wurtzite polytypes of GaN, AlN, and lnN form continuous alloy systems whose direct band gaps range fromeV for InN, to 3.4 eV for GaN, to 6.2 eV for AlN. Thus, thIII–V nitrides could potentially be fabricated into optical dvices which are active at wavelengths ranging from thewell into the UV. Table IV compares relevant material proerties of SiC and GaN with Si and GaAs, the two most polar semiconductor device technologies, and GaP andmond, two other contenders for high-temperature operatiDiamond has long been recognized as a promising matfor fabricating robust, solar-blind radiation detectors. Hoever, due to technological limitation, diamond is still potetial material for future technologies.13,62 The device maxi-mum operating temperature parameter given in Table IVcalculated as the temperature at which the intrinsic caconcentration equals 531015 cm23 and is intended as a rougestimate of the band gap limitation on device operatiMore important for the maximum operating temperaturethe physical stability of the material.

IV. Si ULTRAVIOLET DETECTORS

Silicon photodiodes, originally designed for the visibspectral range, can be also used in the UV region. Arraysilicon photodiodes are fabricated by a number of manuturers, e.g., the most commonly used arrays for astronomspectroscopy are fabricated by EG and Reticon.63 Building asilicon photodiode with a high ultraviolet quantum efficienhas not been the problem. The problem has been to maithat high quantum efficiency under ultraviolet irradiation a

l. Phys., Vol. 79, No. 10, 15 May 1996

or

-1.9

-ed-u-ia-ns.rial--

isier

n.is

eof

ac-ical

ytaind

over a period of years.Two types of silicon UV detectors are fabricated:

~1! diffused photodiode,~2! Schottky barrier photodiode.

In the diffusion photodiode a depletion layer is formebetween thep- andn-doped layers, in a Schottky diode it isformed at the semiconductor-metal interface.

A. Diffused photodiodes

Figure 16 shows a typical spectral dependence of qutum efficiency of the older types of silicon photodiodes. Thquantum efficiency of this device is considerably lower thathe theoretical value between 1400 and 2000 Å. Additio

FIG. 16. Quantum efficiency of older design silicon photodiode~after Ref.64!.


hh

o

an

aat-e

a

ld

re-n-ed

ffi-s-desre-areonsetorsnndon

o-etoeathe

idese-luebemall.s

eic

i

ally, this photodiode is almost dead to UV photons of wavlengths shorter than 1200 Å, presumably because of a tanti-reflecting silicon dioxide layer on the front surface. Tmajority of the oldest types of silicon photodiodes haddead region at their surface, so carriers generated byphotons were lost in this region before they could be clected by thep-n junction, leading to a loss of quantumefficiency. The susceptibility of the oxide–silicon interfacto high levels of UV radiation has been known about at lesince 1966.65 Even if a photodiode had 100% internal quatum efficiency, a thick silicon dioxide layer on the diodsurface would reduce the quantum efficiency in the 4001200 Å region drastically.

To overcome the above-mentioned difficulties, it hbeen proposed and demonstrated that natural inversion lphotodiodes66 should have an internal short-wavelengquantum efficiency of 100% due to the nature of the builtfield associated with a strongly inverted layer at the oxidsilicon interface. Figure 17~a! shows a schematic diagram othis photodiode. The inversion layer is contacted via then1

diffusion by means of the inner metal ring. The outer mering contacts thep1 guarding and thereby thep substrate.The substrate can also be connected via thep1 diffusion onthe backside. Hansen66 has fabricated the photodiodes o100V cm substrates, which corresponds to an acceptor ccentration of 1.431017 cm23. The substrates were orientealong the 111&-direction. The oxide, which also serves asanti-reflection layer, was grown in dry oxygen. The intrinsfixed positive charge within the SiO2 causes thep-type sub-strate to undergo inversion, and thus induces a very tn-type region adjacent to the SiO2Si interface, and forms therequiredp-n junction. Such devices have an oxiden1 pstructure and are available commercially, e.g., UDT-UV5068

FIG. 17. Schematic diagram of the silicon UV photodiodes:~a! inversionlayer type;~b! phosphorus diffused type~n on p!; and ~c! boron-diffusedtype ~p on n! ~after Ref. 67!.


e-ickeaUVl-

est-eto

syerhin–f

tal

non-dnic

hin

.

Since the junction creation process results in a built-in fiewhich is optimum for removing minority carriers from theoxide–silicon interface, these devices are less prone tocombination and trapping effects, and thus higher UV quatum efficiency can be achieved than in the case of diffusjunction photodiodes.

Figure 18 shows spectral dependence of quantum eciency of UDT standard UV 100 inversion layer photodiodewhich had 1200 Å anti-reflection layers of oxide. The absence of any dead region at the front surface of these diohas been verified by internal quantum efficiency measuments in the 250–500 nm spectral range. These diodesalso almost dead below 1200 Å because of the thick silicoxide anti-reflecting coating on the front surface. Becauthe inversion layer charge which causes the semiconducsurface inversion is situated within the first 100 Å of silicooxide from the semiconductor surface, any efforts to extethe diodes response below 1200 Å by reducing the silicoxide coating to a few tens of angstroms is undesirable.

Instabilities were also encountered with this type of phtodiode. Inversion layer diodes have a low linearity rang~the range over which the diode current is proportionalinput irradiance! due to the inherently high sheet resistancof an inverted semiconductor layer. Moreover, it appears ththe fixed oxide charge that induces the inversion layer in tdiode was neutralized as a result of UV irradiation.67 Sincethe inversion layer charge decreases with decreasing oxcharge, the diode inversion layer resistance, which is inries with the diode, would increase. Thus the threshold vaof forward bias needed to cause a nonlinearity coulddropped across the increased series resistance by a svalue of photocurrent, resulting in a loss of linearity range

The best quality Si photodiodes for UV applicationhave been developed by Korde and co-workers.64,67,69,70

They used different methods of junction fabrication, but thbetter results have been obtained using defect-free arsen69

and phosphorus67,70 diffusions.Figure 17~b! is a schematic diagram of diffused type S

photodiode. The starting material was a float-zone,^111& ori-ented,p-type, 100V cm, one side mirror polished, silicon

FIG. 18. Quantum efficiency of inversion layer silicon photodiodes~UDTstandard UV 100 photodiodes! compared to the efficiency of typical fluo-rescing coatings~after Ref. 64!.


-

,

era-

e

ep-

arc-

–toty

r

ion

wafer having 1 cm2 active area. After thep1 channel stopandn1 guard ring were formed by selective boron and phophorus diffusions, respectively, arsenic was predepositethe active area of the device using a planar arsenic souThe arsenosilicate glass formed during the predepositionetched in 10% hydrofluoric acid and the final passivatiSi2O anti-reflection coating with a thickness of about 600was grown in dry oxygen. A standard photolithographic pcess was used to open windows at the desired locations indiffused regions, and aluminum was vacuum evaporatedthese windows for ohmic contacts.

Since UV radiation is absorbed in the first few hundrnanometers of silicon, the doped diffusions were kept shlow to minimize the possibility of recombination betweethe oxide–silicon interface and the junction. Lowtemperature dopant deposition and drive in were adopteget a shallow junction, about 0.2mm. The carrier concentration profiles of diffused photodiodes are shown in Fig. 19

It appears thatn-type impurities such as phosphorus aarsenic tend to pile up in the silicon during the oxide growcreating a built-in electric field near the oxide–silicon inteface, which eliminates interface recombination.71,72So a highminority carrier lifetime in the diffused region results i100% internal quantum efficiency, which is shown in Fi20. The increase in internal quantum efficiency below 3nm is caused by impact ionization involving energetic phtogenerated carriers, while the decrease in internal quanefficiency above 600 nm is caused by the recombinationsome of the minority carriers created in thep-type siliconbeyond the depletion region.

As possible causes of instability of quantum efficienof silicon UV photodiodes, we can distinguish:67 unsatisfac-tory silicon–oxide interfaces, latent recombination centersthe diffused layers, and moisture absorption by the devMeticulous care as described in Ref. 67 is necessary duthe fabrication of devices. A dry, clean oxide process durthe growth of the final anti-reflection coating is essentialachieve stability against UV exposure. According to Kor

FIG. 19. Carrier concentration profiles of boron diffusion inp-on-n photo-diodes, and of phosphorus diffusion inn-on-p photodiode~after Ref. 67!.


s-d inrce.wasngÅro-theon

edal-n-d to-.ndth,r-

ng.50o-tumof

cy

inice.ringingtode

and Geist the most important steps in obtaining a UVenhanced photodiode are:67

~1! extremely high-purity chemicals for wafer cleaning,~2! a clean oxide process, such as described by Botchek73

~3! special annealing procedure74 to prevent UV-inducedcharge neutralization in the oxide, as well as to rendthe oxide–silicon interface less susceptible to UV radition,

~4! careful procedure of deposition of ohmic contacts~byelectron-beam evaporation!.

N on p photodiodes are inherently more stable than thcommonly used boron-diffused devices based onn-type sili-con. Boron migrates from the silicon surface into the oxidduring the final thermal oxidation step. The resulting accetor profile, which is depleted in the vicinity of the oxide–silicon interface, creates a built-in electric field perpendiculto the interface, directed toward the bulk. The resulting eletric field attracts the photogenerated minority carriers~elec-trons! created in the diffused region towards the oxidesilicon interface, greatly enhancing the carrier loss dueinterface recombination. The results of a long-term stabilitest ~baking for several months at 110 °C! shown in Fig. 21

FIG. 20. Internal quantum efficiency as a function of wavelength fophosphorus-diffused photodiode~after Ref. 67!.

FIG. 21. Change in spectral responsivity at 400 nm as a function of duratof exposure to 110 °C~after Ref. 67!.


a-e

-gehesey

-ede

ofr

m

yseg,-se

re

eofc-

indicate little or no change in responsivity of the phosphous diffused and inversion layer devices, while the borodiffused devices fabricated using the standard process exited a monotonic decrease in responsivity during tprolonged baking.

Further improvement in the fabrication of Si UV photodiodes has been achieved by reduction of the silicon dioxlayer to a thickness between 40 and 80 Å.64,70 The thinneroxide should lead to photodiodes with higher quantum eciency stability because of the smaller volume availablegeneration of electron-hole traps which are invariably cauby the diode exposure to humidity and UV radiation. Prodution of the electron-hole traps in the passivating siliconoxide layer is one of the major causes for quantum efficienchanges in silicon photodiodes. Figure 22 shows considable improvement of quantum efficiency in a short wavlength region for photodiodes with 46 Å of oxide layer. Rduction of the oxide thickness is limited by a criticathickness below which device behavior is unacceptabl70

The performance of the 28 Å sample is seriously degradNot only is the efficiency of the 28 Å sample reduced, butexposure stability is quite poor.

Improvements in the performance of Si UV photodiodstill require improved oxide formation techniques. NovelUV photodetector structures are continuously designedfabricated.75,76 Also the amorphous Si UVp- i -n photo-diodes have been fabricated successfully.77,78 EnhancementUV response of amorphous SiC/Si heterojunctions in200–400 nm wavelength region has been achieved by uhigh optical gapa-SiC:H as the front layer, by the reductioof the front-layer thickness, and by using low-level phophorus doping in thei -a-Si:H layer.78

B. Si photodiodes for vacuum ultraviolet applications

Detectors for the vacuum ultraviolet~VUV ! region mustusually be operated in a clean, ultrahigh vacuum envirment, and must be free from interactive degradation inpresence of radiation. Further, they should possess a

FIG. 22. Measured quantum efficiency of Si photodiodes with 46 and 2of oxide ~after Ref. 70!.


or-n-hib-he

-ide

ffi-forsedc-di-cyer-e-e-le.ed.its

esSiand

thesingns-

on-thelow

background noise level to assist in the detection of a reltively weak source. Si photodiodes for VUV technology artypically made with a very thin passivating silicon dioxidesurface layer, as described previously.

The mean energy,w, required for electron-hole pair cre-ation depends on the band gap of a semiconductor.79 Forexample, experimental values forw are 3.61 for Si,80 4.2 forGaAs,80 and 6.54 for GaP.81 If these pairs are created in thedepletion layer or if they diffuse into the space-charge region, they are separated by the internal electric field givinrise to an external photocurrent. Figure 23 shows that thquantum efficiency of Si photodiodes nearly approaches tpredicted slope of 3.63 eV per electron hole created. Thetypes of photodiodes can also be optimized for the soft x-raregion.82

One significant shortcoming of Si photodiodes for certain VUV applications is the inherent broadband responsextending from x rays to the near infrared. This broadbancharacteristic is often undesirable in many applications. Thaddition of a thin film of a suitable filtering material to thesource of such a photodiode can accomplish the restrictionthe sensitivity of the silicon to a much narrower band, obands, in the VUV.83

A schematic of the optically filtered photodiode is shownin Fig. 24. The photodiodes were fabricated on 100-nm-diap silicon wafers that have ap-type epitaxial layer about 5mm thick. Before the active area formation, a deepp-n junc-tion, a p1 channel stop, and ann1 guard ring were createdby diffusion. The purpose of the side junction is to improvethe degree of discrimination against unfiltered radiation bisolating carriers resulting from radiation entering the sideof the structure from the main detection circuit. The activarea of 1 cm2 was produced by phosphorous/arsenic dopinfollowed by the thermal growth of a 15-nm-passivating surface oxide layer. The front surface of the photodiode wacoated with the filter materials. Useful bandpasses in thVUV were selected: Al/C, Al/C/Sc, Ti, Sn, and Ag. Thefilters have bandpasses ranging from 10 to 50 nm and acommonly used in the region 0.1–80 nm.

8 Å

FIG. 23. Quantum efficiency of As- and P-doped photodiodes with oxidthickness as indicated. The straight line represents the quantum efficiencySi based on a value of 3.63 eV per electron-hole pair, not including refletion losses~after Ref. 70!.


el/areserthsses

regvelyityups-

s-

ahe

to-p-aro-ntg

org

asesantnir-

mar-ofsedononnenofsisndr-

ieryt of-

The spectral efficiency of the coated photodiodesshown in Fig. 25. Also shown are the calculated photodioefficiencies based on the conversion efficiency of silicon athe transmittance of the thin-film coatings, including the 1nm-thick passivating silicon oxide layer. It can be seen tthere is reasonable agreement between the resulting calated values and the measured device efficiencies. Typdark current values measured before and after coatings wabout 100 fA. Noise variations of this current were aboutorder of magnitude less.

The photodiodes coated with appropriate filtering marials may find applications in diverse VUV and EUV discplines, such as plasma diagnostics, solar physics, x-raythography, x-ray microscopy, and materials science, andreduce the dependence on relatively fragile unsupportedtallic filters for spectral discrimination.

V. SCHOTTKY BARRIER ULTRAVIOLETPHOTODIODES

In a Schottky barrier photodiode, a thin metal layerdeposited onto a semiconductor. The transmission thro

FIG. 24. Schematic of optically filtered VUV photodiode~after Ref. 83!.

FIG. 25. Measured~points! and calculated~curves! efficiencies of siliconphotodiodes coated with several VUV filtering materials~after Ref. 83!.


isdend5-hatlcu-icalerean

te-i-li-

canme-

isugh

the thin metal contact is relatively high, and thus a largfraction of the UV radiation is absorbed close to the metasemiconductor interface where photogenerated carriersefficiently collected by the depletion layer field. This givethe device an acceptable quantum efficiency, albeit lowthan for the best photodiodes. Usually at shorter wavelengthe collection efficiency of the generated carriers decreafor Schottky barriers. The responsivity also decreases.84

For many applications Schottky barrier photodiodes asuperior to other detectors. They are particularly promisinas calibrated standard detectors, because they are relatiinsensitive to surface contaminations, show a higher stabilin intense UV radiation, and posses a quantum efficiencyto four orders of magnitude higher than that of photoemisive detectors in the VUV spectral range.

Schottky barrier UV detectors have been fabricated uing different semiconductors.85–91 A straightforward way tomake a highly selective~blindless! UV detector is to use awide band-gap semiconductor, e.g., ZnO86 or diamond.90,91

Semiconductor diamond is of considerable interest frombasic physics, as well as an engineering, point of view. Thigh thermal conductivity, small dielectric constant~«'5.5!,and large band-gap energy~Eg55.4 eV! combine to makediamond an ideal candidate for many electronic and opelectronic applications, particularly for high-temperature oeration. Unfortunately, these wide band-gap materials ofgood enough quality are not readily available, and the apppriate technology is far from being mature. However, receprogress in diamond thin-film technology is very promisinfor future applications.

Recently Zhaoet al.91 have reported photoresponsecharacteristics of boron doped hot-filament chemical vapdeposition~HF-CVD! diamond based Schottky diodes usinsemitransparent Al contacts. The polycrystalline 7-8-mm-thick diamond films were deposited onp-type ~100! siliconsubstrates using hydrogen and methane as the reaction gand acetone vapor as the carrier gas for the boron dop~trimethyl borate!. The acetone was also an additional carbosource. After surface chemical cleaning of samples, the ccular Schottky contacts~with an area of 4.5 mm2! were fab-ricated by thermal evaporation of semitransparent aluminucontacts. However, these Schottky diodes were not solblind detectors. Figure 26 shows the spectral dependencethe quantum efficiency of the photodiode. The photoresponof these diodes to visible and near UV radiation is attributeto several mechanisms including photoelectronic emissibetween metal and semiconductor, extrinsic photoexcitatiof impurities, as well as crystal defects. Without correctiofor surface reflection, the quantum efficiencies were betwe5% and 10% for unbiased diodes. When a reverse biasover 1 V was applied, the quantum efficiency of 30% waobtained at 500 nm. The increase of quantum efficiencycaused by a broadening of the junction depletion region amore efficient collection of photocarriers. The Schottky barier height between Al andp-type polycrystalline diamondwas determined to be 1.15 eV. It appears that the barrheight of HF-CVD diamond Schottky diodes is controlled ba high density of defect states, and hence is independenthe metal work function. The Schottky barrier is pinned ap


o

it

r

r

-ide-on989.m-.

to-

to a

rra-

onThehown

proximately 0.2Eg from the top of the valence band. Theorigin of the pinning has not been identified, but is likely tbe due to damage introduced during growth.92

The conventional method to make a selective UV detetor is to combine a conventional broadband UV-enhancphotodiode with a UV filter. An example of this kind ofcombination is a Ag–GaAs Schottky photodiode with a thsilver contact.85 The UV selectivity of this detector is a resulof the narrow transmission ‘‘window’’ of silver films nearthe wavelength of 322 nm. A disadvantage of this approais a deterioration of the responsivity also in the UV region

Schottky barrier photodiodes have been mainly fabcated for VUV applications. Investigations carried out bKrumrey et al.87,88 indicate that GaAsP and GaP Schottkphotodiodes show remarkable stability and high quantumficiency. The stability measurements of the efficiency weconducted with a monochromatic photon flux of;1010

photons/s into an area of 1 mm2 at a photon energy of 124eV. Typical behavior of a silicon and GaAsP Schottky photodiode is shown in Fig. 27. In the case of silicon photodode, the quantum efficiency decreases from the initial valof 27% by 0.25%/s. The decrease of efficiency shows saration behavior, but even at 8% a further decrease of torder of 0.01% is observed. In contradiction, the quantuefficiency of GaAsP and GaP Schottky diodes~HamamatsuG1127-02 and G1963! is stable at least within 0.05%~whichis the uncertainty of the measurement! during an exposure ofmore than 1 h. The electrical and optical properties of GaASchottky diodes in the spectral range below 6 eV have beinvestigated in detail by Wilson and Lyall.89 The GaAsPphotodiodes can be used in the hole spectral region fromeV up to at least 3.5 keV~Fig. 28!. Near the visible regionthe efficiency is comparable to that of photoemissive diodebut it is higher by three to four orders of magnitude in thregion above 1 keV.

VI. SiC ULTRAVIOLET DETECTORS

The inherent material properties of the 6H polytype oSiC are very attractive for UV applications. In comparisowith Si ~see Table IV!, SiC is characterized by a highe

FIG. 26. Quantum efficiency vs wavelength for AJ-HFCVD diamonSchottky diodes without the correction of reflection~after Ref. 91!.


c-ed

n

ch.i-yyef-re

-i-uetu-hem

sPen

2

s,e

fn

breakdown field that permits much smaller drift regions~i.e.,much lower drift region resistances!, a higher thermal conductivity that enables superior heat dissipation, and a wband-gap energy~2.9 eV! that permits higher junction operating temperature. Due to the fact that reproducible siliccarbide wafers and epilayers were unavailable before 1SiC carbide device fabrication technology is relatively imature compared to silicon device fabrication technology

A. P-n junction photodiodes

The earlier published research on 6H–SiC UV phodiodes utilized diffusion of Al inton-type substrates.93 Thediffusion process at temperatures of about 2000 °C led

d

FIG. 27. Stability of three different semiconductor photodiodes when idiated with 1010 photons/~s mm2! at 124 eV~after Ref. 87!.

FIG. 28. Quantum efficiency of a GaAsP Schottky diode, determineddifferent monochromators by comparison with photoemissive detectors.quantum efficiency of the photoemissive reference diodes are also s~after Ref. 87!.


ffi-tionorp-re-op

nn

esand

-er

edas

as

nc-

usde

nts,

m-tthwn,

her

-ro-

ivi-een

structural decomposition of surface layers. This processsulted in high-leakage devices with low quantum efficienc

Glassowet al.94 improved upon this initial design byutilizing N implantation to form a very shallow~0.05 mm!n1-p junction in a 5-mm-p-type epitaxial layer grown on ap-type substrate. The successive steps of the manufactuprocess of a planar photodiode are shown in Fig. 29. Furnannealing and rapid isothermal annealing~RIA! were em-ployed to activate the implanted nitrogen.I –V characteris-tics and spectral quantum efficiency were investigated intemperature range from 296 to 826 K and 295 to 676respectively. This study resulted in devices with a maximuquantum efficiency of 75%~with an anti-reflection coating!at a peak wavelength of 280 nm at room temperature. Tscatter of the maximum value of the quantum efficiency wexplained by differing surface recombination velocitiewhich were caused by differing surface qualities. Threverse-bias dark-current density was excessive, on the oof 1025 A/cm2 at 210 V and room temperature. Probablycontact sintering at high temperatures could induce diffusiof the contact metal through crystalline defects in thinn1

layer and thereby increase diode leakage current.Figure 30 shows a mesa SiC photodiode structure fab

cated from commercial 6H–SiC wafers, 1 in. in diameter,which a p-n junction was grown epitaxially.95 A heavilyN-dopedn-type epitaxial layer of 0.2 or 0.3mm thick wasutilized to form ann1-p junction. The concentrations ofimpurities was 5–831017 cm23 in p-type epitaxial layers,and 5–1031018 cm23 in n1 cap layer. The device mesa wapatterned and etched using reactive ion etching~RIE! and anNF2O2 gas mixture. Device passivation was accomplishedgrowing a thin~0.05mm! layer of SiO2. The contact to thetop n1 layer was Ni sintered in Ar at temperatures betwe900 and 1000 °C. Optical responsivity with a peak at abo

FIG. 29. SiC UV photodiode processing steps~after Ref. 94!.


re-y.

ringace

theK,m

heasserder,on

ri-on

s

by

enut

270 nm between 150 and 175 mA/W with a quantum eciency of about 70%–85% was measured. As the detecwavelength is decreased, a larger proportion of the abstion occurs near the semiconductor surface and surfacecombination becomes more important. By thinning the tn1 layer outside of the mesa contact~see Fig. 30!, a respon-sivity of 50 mA/W was achieved at 200 nm. The diffusiolength of electrons in thep-type layer was varied betwee1.4 and 2.2mm, and was smaller than the estimate of 3mmgiven in Ref. 94. This diffusion length determination donot depend on any assumptions about carrier lifetimemobility as was made by Glassowet al.94

Edmond and co-workers96 have demonstrated a considerable improvement of 6H–SiC UV photodiode quality ovprevious efforts.93–95Wafers ofn- andp-type 6H–SiC hav-ing a diameter of 25.4 mm with typical resistivities in thrange of 0.02–0.04 and 1–10V cm, respectively, were useas substrates for epitaxial growth. An epitaxial junction wproduced by first growing a predominantly Al-dopedp-typelayer followed by a predominantly N-dopedn-type layer ona p-type substrate. The doping in the background layer wcontrolled between~1–5!31016 cm23 with a thickness of;3mm. Then-type layer was heavily doped to;1019 cm23. Alldevices were fabricated using a mesa geometry. The jutions were passivated with thermally grown SiO2. Ohmiccontact materials for thep-type side andn-type side weresintered Al and Ni, respectively.

Figure 31~a! shows the reverse-bias dark current versvoltage as a function of temperature for a UV photodiowith square active junction with an area of 0.04 cm2. Thephotodiode has demonstrated extremely small dark curreas little as 10211 A at 21 V at 473 K. At the highest voltagemeasured,210 V, the dark current density increased fro;1029 to ;331028 A/cm2 when increasing the temperatures from 473 to 623 K. Figure 31~b! shows the dark currendensity at210 V as a function of temperature together wipreviously reported data for comparison purposes. As shothe devices reported by Edmond and co-workers96 exhibitorders of magnitude lower leakage current than any otpublished data in 6H–SiC. The forwardI –V characteristicscorrespond to theb values ranging from 1.5 to 2 which indicate both diffusion and generation–recombination pcesses.

The same devices exhibited near unity peak responsties between 268 and 299 nm in temperature range betw

FIG. 30. SiC photodiode cross section~after Ref. 95!.


J. Appl. Phys.,

FIG. 31. Dark current density vs~a! reverse voltage and~b! inverse temperature for 6H–SiC photodiodes~after Ref. 96!.

e

n

o-

Clrsne,o-gs2of

0unveof

ge

al

wbe

-

e

223 and 623 K~see Fig. 32!. The responsivity of the photo-diode at a higher temperature shifts to longer wavelengbecause of band gap narrowing. The indirect band gapcreases from;3.03 eV at 223 K to;2.88 eV at 623 K,corresponding to a rate of 3.831024 eV/K.94

Recently, Hirabayashiet al.97 have fabricatedp-n junc-tion 3C–SiC photodiodes by atmospheric chemical vapdeposition. However, these photodiodes were inferiorcomparison with those previously described. The maximuvalue of responsivity was 72 mA/W at 250 nm and the quatum efficiency was 36%~cubic SiC energy gap is 2.2 eV aroom temperature!.

Finally, it should be noticed that to produce large arand low-cost UV detectors an innovative family of devicefabricated of hydrogenated amorphous SiC/Si heterojutions have been described.98,99

B. Schottky barrier photodiodes

Schottky barrier 6H–SiC photodiodes are of interest fmany applications because of their relatively simple fabric

FIG. 32. Temperature-dependent responsivity of a SiC photodiode. Ttemperatures tested were:~h! 223 K. ~l! 300 K, ~s! 498 K, and~j! 623K ~after Ref. 96!.

Vol. 79, No. 10, 15 May 1996

thsde-

orinmn-t

asc-

ora-

tion process compared top-n photodiodes. Frojdhet al.100,101 have fabricated finger shaped Schottky photdiodes by evaporating Ti on substrates with differentn-typeandp-type doping levels. The starting material was 6H–Siwafers with an epitaxial layer. The doping of the epitaxialayers is specified in Table V. Prior to processing the wafewere degreased by subsequent dipping in trichloroethyleacetone, and 2-propanol and finally dipped in hydrogen fluride ~HF!. Next, a pattern with a number of shaped openinwas formed on the wafer with finger widths in the range 1–mm. Each diode is surrounded by an opening with a width10 mm ~see Fig. 33.! Four different photodiode areas wereused: 1003100 mm, 75375 mm, 50350 mm, and 25325mm.

As a barrier, titanium with resulting layer thickness 20nm was evaporated on the wafers using an electron gevaporator. A standard lift-off procedure was used to remoexcess metal. A back contact was made by evaporation500 nm of Al on the backside of the wafer~see Fig. 34.!

Most of the Schottky barriers showed good rectifyinbehavior with an ideality factor generally below 1.2. Thbest results were obtained for diodes on thep-type materialfor which the reverse leakage current at210 V was below 1pA and for a reverse voltage of up to270 V the leakagecurrent remained below 500 pA. Also, results of spectrresponsivity measurements indicate that thep-type materialhas the highest sensitivity~see Fig. 33!. Sample N1 showsvery low responsivity which can be explained by the shalloepitaxial layer. The low response at short wavelengths canexplained by the influence of surface recombination.

Generally, thep-type material shows a much higher photoresponse than then-type material. This can be explainedby two effects:

~1! the built-in voltage in thep-type material is much higher~about factor of 2! than the built-in voltage in then-typematerial,

~2! the diffusion length for electrons is much larger than thdiffusion length for holes.

he


7462 J. App

TABLE V. Characteristics of 6H–SiC Schottky barrier photodiodes fabricated with different materials~afterRef. 100!.

Sample Sample characterization Nd ,Na ~cm23!Fb ~eV!C–V

Fb ~eV!I –V b

I revat 210 V

N1 n-type, t51.6mm,no visible defects

7.531015 0.98 0.97 1.08 ,10 pA

N2 n-type, t'10 mm 2.931016 0.89 0.92 1.14 0.4 nAN3 n-type, t'5 mm 2.731012 0.94 0.79 1.23 0.5 mAN4 n-type, t'5 mm,

high density of pinholes1.031017 0.94 0.84 1.17 18mA

P1 p-type, t'10 mm 1.631015 2.43 1.94 1.03 ,1 pA

s

i

iru

m

v

-sireentilinhete’shistoa-efi-ssed

-

yy

sseseN

The second reason also explains the larger photoresponp-type material at long wavelengths~see Fig. 35!.

The better results forn-type 6H–SiC Schottky barriershave been obtained by Anikinet al.102The Au–SiC Schottkybarriers were formed on the surface ofn-type epitaxial layerswith uncompensated donor concentration~5–10!31016

cm23, grown by an open sublimation method onn-type 6H–SiC substrates oriented along the~0001! Si plane. Frontsemitransparent contacts were made by vacuum deposof a thin Au layer. Before evaporation of the barrier contaa crystal of SiC was etched in molten KOH, washed in dionized water and in organic solvents, and finally annealin a vacuum at 500 °C.103 The height of the potential barriewas 1.4–1.63 eV, depending on the structure. The structwith an area 331023 cm2 showed low leakage currents;10210 A, up to a breakdown voltage of 100–170 V at rootemperature, and could be operated up to 573 K whereleakage current rose to;1028 A. The monochromatic re-sponsivity in the best samples at wavelength of 215 nreached 0.15 A/W, which corresponds to the quantum eciency of 80%.

It is expected that the best quality 6H–SiC UV detectoshould be fabricated usingp-n junctions.104,105A consider-ably good progress in fabrication ofp-n junction usingchemical vapor deposition has been recently observed.devices displayed excellent rectification characteristics atevated temperatures as well as avalanche breakdownages of 2000 V at room temperature.

FIG. 33. The finger area of a photodiode.

l. Phys., Vol. 79, No. 10, 15 May 1996

e of

tioncte-ng

res,

the

m,ffi-

rs

Theel-olt-

VII. III–V NITRIDES AS A MATERIALS FORULTRAVIOLET DETECTORS

The III–V nitrides are the next group of wide gap semiconductors which are very promising for device applicationin the UV wavelengths in much the same manner that thehighly successful As-based and P-based cousins have bexploited in the infrared, red, and green wavelengths. Unnow, efforts to develop semiconductor device technologythe UV spectral region have been far unsuccessful. TIII–V nitrides are excellent materials to cover importantechnologically band, which occurs in the 240–280 nm rang~;4.75 eV!, where absorption by ozone makes the earthatmosphere nearly opaque. Space communications in tband would be secure from earth, although vulnerablesatellite surveillance. On the earth, shielded by the sun’s rdiation, imaging arrays operating in this band would providextremely sensitive surveillance of objects coming up out othe atmosphere. Due to their wide band gap, the III–V ntrides are expected to exhibit superior radiation hardnecharacteristics compared to Si. GaN is by far the best-studiIII–V nitride semiconductor.

The main obstacle in the development of the nitride UVphotodetectors was significant difficulties in obtaining highquality material. GaN has highn-type background carrierconcentration resulting from native defects commonlthought to be nitrogen vacancies. In films having relativelsmall background electron concentrations,p-type doping isdifficult to achieve. In the past several years much progrewas made in developing modern epitaxial growth techniquand in understanding the properties of this material. Throom temperature background electron concentration of Gafilms has been considerably reduced andp-type material has

FIG. 34. Cross section of 6H–SiC Schottky barrier photodiode~after Ref.100!.


t

p-andri-

c-thatthe

eim-eralack-

bil-ntion

on-

-

onzedha--era-onby

been reported.106,107 At present the metal-organic chemicvapor deposition~MOCVD! grown GaN layers with the AlNbuffer layer haven-type conductivity with an electron concentration of about 1015 cm23 at room temperature and thelectron mobility of about 500 cm2/V s Mg-dopedp-typelayers have hole concentrations above 1012 cm23 and holemobilities of 10–50 cm2 V s at room temperature.108,109

A number of reviews concerning III–V nitrides havbeen published. Morkocet al.57–59,110and Davis111 recentlypublished several reviews which present a comprehenhistory of the developments in the III–V nitrides, growtechniques, structural, optical, and electrical properties. Athe present-day achievements and future prospects in dient devices structures are described.

FIG. 35. Spectral responsivity of 6H–SiC Schottky barrier photodiodes~af-ter Ref. 100!.


al

-e

e

sivehlsoffer-

In this chapter, after a brief review of the physical proerties of mainly GaN, we focus on the achievementsfuture prospects of III–V nitrides as the materials for fabcation of different types of UV detectors.

A. Physical properties

GaN normally crystallizes in the wurtzite crystal struture, although researchers have known for some timeGaN also has a zinc blende polytype. Table VI listsknown properties of wurtzite GaN.

Control of the electrical properties is still one of thgreatest challenges facing nitride researchers. Usingproved crystal-growth techniques, researchers in sevleading laboratories have succeeded in reducing the bground electron concentration to 1015 cm23. Nakamuraet al.112 have reported the highest GaN bulk electron moity of 600 and 1500 cm2/V s at 300 and 77 K, respectively, ian unintentionally doped sample having carrier concentra431016 cm23. Theoretical modeling of Chinet al.113 carriedout for T577 and 300 K demonstrates that the carrier ccentration dependence of Hall and drift mobilities~importantfor device modeling! is a significant function of the compensation ratios. These calculated results@see Fig. 36~a!# agreewell with available experiments only for high compensatiratio. Piezoelectric acoustic phonon scattering and ioniimpurity scattering are the two dominant scattering mecnisms at temperaturesT<200 K, and the polar optical phonon scattering is the most significant scattering at temptures T>200 K. It has also been found that electrmobilities in GaN correlated well with those calculatedMonte Carlo stimulations. Joshii114 using this method has

TABLE VI. Properties of wurtzite GaN, AlN, and InN~after Ref. 57!.

Parameter GaN AlN InN

Band-gap energy~eV! 3.39 ~300 K! 6.2 ~300 K! 1.89 ~300 K!3.50 ~1.6 K! 6.28 ~5 K!

Band-gap temperature coefficient~eV/K! 26.031014 21.831024

~for T.180 K!Band-gap pressure coefficient~eV/kbar! 4.231023

~T5300 K!Lattice constant~nm! a50.3189 a50.3112 a50.3548

~T5300 K! c50.5185 c50.4982 c50.5760Coefficient of thermal expansion~K21! Da/a55.5931026 Da/a54.231026

~T5300 K! Dc/c53.1731026 Dc/c55.331026

Thermal conductivity~W/cmK! 1.3 2Electron mobility~cm2/V s! 900

~T5300 K!Hole mobility cm2/V s 150?

~T5300 K!Index of refraction 2.33~1 eV! 2.15 2.85–3.05

2.67 ~3.38 eV!Dielectric constant e059.5 e058.5

e`55.35 e`54.68Saturation electron drift velocity~107 cm/s! 2.7 2Electron effective mass 0.20 0.11Phonon modes~cm21! A1~TO!5532 TO5667 TO5478

~T5300 K! E1~TO!5560 E25665 LO5694E25144, 569 LO5910A1~LO!5710E1~TO!5741


hl

e

i

nt

reole

-t ise

rsolen-

ned

a-

-Ned

f-

at

or

obtained good agreement with experimental data115 whenspace-charge scattering was taken into account@see Fig.36~b!#.

One of the long-standing problems in GaN researchbeen the search for a shallowp-type dopant. Most potentiadopants have been observed to compensate GaN, produhighly resistive material.P-type conduction in Mg-dopedGaN has been realized for the first time by low-energy eltron beam irradiation~LEEBI! treatment116 of metal-organicvapor phase epitaxy~MOVPE!-grown Mg-doped films. Lateron, Nakamuraet al.117 achieved thep-type conduction inGaN by thermal annealing of the MOCVD-grown Mg-dopefilms, more recently;p-type GaN has been obtained bplasma-enhanced molecular beam epitaxy~MBE! withoutthe need for a postannealing procedure.118,119These resultssupport the findings and suggestions of Nakamuraet al.106

that acceptor H neutral complexes are formed in the Mdoped GaN during CVD as a result of the H2 carrier gas anddiluent during the growth. These complexes, in turn, effetively reduce the number of acceptor states which couldoccupied by electrons from the valence band and resultcompensated material. A decrease of H-atom concentrawas observed after the thermal treatment in Ar gas, supp

FIG. 36. The electron mobility in GaN:~a! the electron drift~solid curves!and Hall~dashed curves! mobility at 300 K as a function of carrier concentration with compensation ratios 0.00, 0.15, 0.30, 0.45, 0.60, 0.75, and 0the experimental data are taken from different articles~after Ref. 113!; ~b!comparison of Monte Carlo results with experimental mobility data for bun-type GaN. Ref. 115. The solid curves were obtained by-using threeferent internal potential variations~after Ref. 114!.


as

cing

c-

dy

g-

c-ben ationort-

ing the hypothesis that H-atom extraction plays an importarole in obtaining low-resistivityp-type conduction.107 Theuse of gas source MBE, in which the only H is derived fromthe background, permits the direct deposition ofp-type ma-terial without the need for postdeposition processing. Figu37 shows the relationship between room-temperature hconcentration and mobility in GaN films grown by variousgroups using both MOVPE and MBE. The circles correspond to the Mg-doped samples, and the squares to whaperceived to be unintentionally doped C samples. While thMg-doped layers needed activation by the N2 heat treatment,or by a low-energy beam irradiation, the C-doped layeneeded none. A statistical averaging suggests that the hmobility decreases in general with increasing carrier concetration.

Recently, Rubin and co-workers121 have succeeded inmakingp-type GaN using MBE assisted by a direct nitrogeion beam together with a variety of doping techniques usin conventional integrated circuit processing~ion implanta-tion, diffusion, and coevaporation of Mg!. The undoped filmshad hole concentrations of 531011 cm23 and hole mobilitiesof over 400 cm2/V s at 250 K. From the slope of the carrierconcentration versus temperature shown in Fig. 38~a!, an ac-tivation energy of 0.29 eV is obtained. The strong temperture dependence of the mobility@see Fig. 38~b!# indicates thedominant contribution of a phonon scattering mechanism.

A conversion of unintentionally dopedn-type GaN lay-ers top-type material was performed by the diffusion process of Mg in a sealed ampoule for 80 h at atmospheric2pressures at 800 °C. After this, Hall measurements indicata hole concentration of 231016 cm23 and a mobility of 12cm2/V s.

P-type inversion conductivity has also been obtained ater Mg implantation with large doses~231014 cm2!. How-ever, high-energy implants~80–100 keV! introduced strainand defects into GaN films, and even after annealing800 °C for 30 min in N2, the x-ray peaks did not return totheir initial positions. Films implanted with Mg ions at lowerenergies~40–60 keV! recovered their original lattice param-

-.90,

lkdif-

FIG. 37. The 300 K Hall mobility vs hole concentration for GaN fromvarious groups using both QMVPE and MBE, the solid squares are fC-doped samples and the solid circles are for Mg-doped samples~after Ref.120!,


tyi-he

n

arets

ffas-Ausengs-uin,ofve

itheheheier-

c--ndc-

eter after annealing at 800 °C and hot-point probe tests cfirmed that these lower energy implants producedp-typeconduction.

For selective area doping or isolation of GaN sampthe use of ion implantation is a critical requirement for avancement of device technology. Peartonet al.122 have re-ported achievement of bothn- and p-type GaN using Si1

and Mg11P1 implantation, respectively, and also the cration of semi-insulating~>53109 V/h! regions in initiallyn- andp-type material using multiple energy N1 implanta-tion and subsequent annealing.

Recently Wilsonet al.123 have reported results of measurements of depth profiles and stability against redistrition with annealing up to 800 or 900 °C, for implanted BCl, Mg, Si, S, Zn, Ge, and Se as dopants in GaN. They foexcellent thermal stability for acceptors~Be, Mg, Zn, and C!and for group IV donors~Si and Ge! for temperatures up to900 °C. In the case of the group VI donors, S redistribute600 °C, and Se at 800 °C. It means also that diffusionmost elements into GaN from an external source doesappear to be a feasible approach for doping GaN sampl

In spite of achievements in the fabrication of GaN mterial, this material is still far from ideal.p-type andn-typesamples exhibit photoconductivity responses at photon egies far below the band-gap energy.124 It may be largelyrelated to the deep tail states created by the high grotemperature, the large stress arising from the lattice m

FIG. 38. Hall carrier concentration~a! and Hall mobility~b! vs temperaturefor undopedp-type GaN~after Ref. 121!.


on-

lesd-

e-

-bu-e,und

s atofnotes.a-

ner-

wthis-

match to the substrate, impurity incorporation, and impuricompensation. Understanding their origin in order to elimnate the defect states will be an important endeavor in tnear future for device development based on GaN films.

B. Ohmic contacts to GaN

Early results for GaN indicate that ohmic contacts caeasily be formed to bothn- and p-type material. The mostpopular contacts for many years for Hall measurementssoldered ln contacts. The first investigation of GaN contacwas carried out by Foresi and Moustakas.125 The Au and Alcontacts were deposited on then-type GaN layers by thermalevaporation and patterned using photolithography and liftotechniques. Current–voltage characteristics show that thedeposited Al contacts are ohmic while the as-depositedcontacts are rectifying and form Schottky barriers. TheSchottky barriers are leaky due to tunneling effects arisifrom the high carrier concentration of the films and the posible existence of an interfacial native oxide layer. The Acontacts become ohmic after annealing at 575 °C for 10 ma result attributed to gold diffusion in GaN. Measurementsthe contact resistivities of Al and Au annealed contacts gavalues in the range of 1027 –1028 V m2.

For most of the important semiconductors, including Sand GaAs, the dependence of the metal barrier height onwork function difference has not been observed due to texistence of surface states which pin the Fermi level at tinterface. Figure 39 shows the dependence of the barrheight on the work function difference correlated to the ionicity of the semiconductor by Kurtinet al.126 In this Fig. 39,the vertical axis is a parameterS which is defined as thechange in barrier height over the change in metal work funtion (S5dFb/xm) and the horizontal axis is the electronegativity difference between the components of a compouwhich is a measure of the compound’s ionicity. The eletronegativity difference for GaN is 1.87 eV.127 This puts

FIG. 39. Dependence ofS5dFb/xm ~change in barrier height over thechange in metal work function! on the electronegativity difference betweenthe components of the compound~after Ref. 126!.


e-s

nalth,llatn-

lo-

old

asnghentin

nefuld

is-

ofst

,ur-

so

0

n-scts,

i-

GaN above the knee of the curve in Fig. 39 implying thSchottky barriers on GaN should have barrier heights whdepend directly on the work function difference betweenmetal and GaN. It indicates that the surface Fermi energunpinned which greatly reduces the complication of creatohmic contacts to GaN as it is only necessary to determmetals with appropriate work functions.

The work function of GaN has been measured to beeV.128 Therefore, any metal with a work function equal olower than that of GaN should form an ohmic contactn-type material, and any metal with a higher work functioshould form a rectifying contact ton-type GaN. The workfunction of Al is 4.08 eV129putting Al in the ohmic category.In the case of Au, the work function is 4.82 eV which puAu in the rectifying category. The above theoretical predtions have been confirmed by the experimental resultsForesi and Moustakas.125

Theoretical predictions indicate that Al can form ohmcontacts ton-type and Au can form ohmic contacts top-typeGaN. Au as a contact to GaN requires a thin interlayer~Ti orCr! to prevent a diffusion of Au in the GaN.

Lin et al.130 have reported the results of an ohmic cotact ton-GaN ~;1017 cm23! study of four separate metallization schemes: Au, Al, Ti/Au, and Ti/Al. The metal layewere deposited via e-beam evaporation. Bilayer contaconsisted of a 20 nm Ti layer deposited directly on Gafollowed by a 100 nm capping layer of Au or Al. Prior tannealing only the Ti/Al contact exhibited near linear I–characteristics. The other contacts exhibited nonlinear Icharacteristics even for small currents. It is probably duethe formation of rectifying Schottky contacts. Then differeannealing procedures were used to improve contact chateristics: annealing at 500 °C in N2 ambient; further anneal-ing using the rapid thermal annealing~RTA! method at 700°C for 20 s. Two samples, Al/Ti and Au/Ti were furtheannealed to 900 °C for 30 s. Device quality ohmic contawith contact resistivity values of 831026 V cm2 have beenobtained when Ti/Al contacts were annealed at 900 °C fors. The nature of reactions responsible for the low resistacontacts has not been well clarified. The most likely mecnism for the low contact resistance is solid-phase epitaxyTiN2 via N outdiffusion. In this case the surface GaN woube highly n-type due to the preponderance of N vacancand effective tunneling through a surface layer is possibl

A novel scheme of nonalloyed ohmic contacts on Gusing a short-period superlattice composed of GaN androw band-gap InN, sandwiched between the GaN chanand an InN cap layer, has been demonstrated by Linet al.131

A contact resistance as low as 631025 V cm2 with GaNdoped at about 531018 cm23 has been obtained without anpostannealing.

C. Etching of III–V nitrides

GaN is characterized by its exceptional stability at hitemperatures and under chemical attack. GaN is insolublH2O acids or bases at room temperature, but dissolves inalkali solutions at very low rate.57 Low quality material hasbeen observed to etch at reasonably high rates in NaH2SO4, and H3PO4. These etches are useful for identifyin


atichthey isingine

4.1rton

tsic-of

ic

n--rsctsN,oV–Vtontrac-

rcts

30nceha-ofldiese.aNnar-nel

y

ghe inhot

OH,g

defects and estimating their densities in GaN layers. A rcently published article stated that good quality GaN filmetch slowly in 30%–50% NaOH/H2O solutions~rate of;2nm/min! while HF, H3PO4, and HNO3/HCl mixtures are in-effective in producing any etching.132 This chemical stabilityprovides a challenge of GaN processing when conventiointegrated circuit processing techniques are used. Smoocontrolled dry etching is necessary for fabrication of smageometry devices in a reproducible fashion. It means thetch rates must be achieved at a low dc bias where ioinduced damage is minimized.

In the last two years considerable progress in the expration of different methods for RIE of III–V nitrides hasbeen observed.132–143An attractive approach to this is the useof electron cyclotron resonance~ECR! discharges which pro-duce high ion densities~>531011 cm23! at low pressurewith ion energies below the displacement damage threshfor III–V materials ~i.e.,<25 eV!.136

Adesidaet al.134 reported a RIE of GaN in silicon tetra-chloride plasmas~SiCl4, 1:1 SiCl4:Ar, and 1:1/SiCl4 SiE4! inthe pressure range between 20 and 80 m Torr. For all gmixtures, each rate increased monotically with increasiplasma self-bias voltage exceeding 50 nm/min at 400 V. Tslight overcut in the profiles was attributed to the significaphysical component in the etching process. A wet etchdilute HF was needed to clear the Si~in the form of SiO4!embedded in the surface of GaN.

Based on an analogy to other III–V semiconductors, owould expect Cl-biased discharge chemistries to be usefor GaN and AlN because the III chlorides are volatile, anthe addition of H2 should remove the nitrogen as NH3 Cl-based etches have worked well on GaN using ECR dcharges of CH4/H2, Cl2/H2 and BCl3/Ar, while the additionof F to the discharges dramatically reduces the etch rateAlN, what can be used for selecting etching of AlN. Film;200 nm thick were lithographically patterned with Hun1182 photoresist to give openings of 1–100mm in variousshapes. Controlled rates of;20 nm/min were obtained atmoderate dc biases~<2200 V!, low pressure~1 mm Torr!,and low microwave power~200 W!. Higher etch rates of;70 nm/min at2150 V and 1 m Torr for high microwavepowers ~1000 W! have been obtained by Peartonet al.136

with the Cl2/H2 chemistry. In this case a lithographic maskAZ 1350J photoresist, was spun onto the samples. The sface of III–V nitrides remain stoichiometric without requir-ing a wet-etch cleanup. A success in RIE of GaN has albeen achieved with BCl3-etch rate 8.6 A/s which was ob-tained for a plasma power of 200 W and pressure of 1m Torr.135

Alternative chemistries which have been proven advatageous in improving etch rates or selectivities in variouapplications include those based the group III etch produin iodine plasmas. A comparison of etch rates of GaN, AlNand InN in HI/H2 high ion density~;531011 cm23! plasmawith results obtained from chlorine-based chemistries indcates that a HI/H2/Ar plasma chemistry produces slightlyfaster etch rates for GaN and AlN than comparable Cl2-basedmixtures, while the rates for InN are significantlyimproved.139By contrast, HBr/H2/Ar yields slower etch rates


b

d

p

h

65heed

ofe

-ised0cyy/

d

upur-e-therap/c-

sre-re-eenbeth

Neeeetheesseshea-

ci-

than the chlorine chemistries for all three binary band gnitrides. Both types of chemistries produce smoothhighly anisotropic etched features. RIE of GaN usiCHF3/Ar and C2CIF5/Ar chemistries has also beereported.143 The etch rates which varied between 60 and 4Å/min, were found to increase linearly with rf plasma powand to decrease with chamber pressure.

Recently, Peartonet al.138,140,141 have studied theAr1-ion milling characteristics of III–V nitrides. It appearthat the surface morphologies of all nitrides~GaN, InN, AlN,and InGaN! remain smooth after ion milling at 500 eV anthere is no preferential sputtering of the nitrogen. Themilling rate for the nitrides are less than those of maskmaterials such as SiO2, SiNx , and photoresist, so this technique is useful only for shallow-mesa applications. Dry etcing methods involving an additional chemical componentmore practical alternatives for device patterning, e.g., theof an Ar ion beam and HCl gas in the chemically assistedbeam etching of GaN may be suitable for the fabricationlaser facets and mirrors.142

D. AlGaN ultraviolet detectors

1. Photoconductive detectors

Development of high quality III–V nitride semiconductor layers in the last several years has resulted in the facation of a new generation of UV detectors. Due to this eastage of development only a few articles concerning UVtectors have been published.109,144–153

The first high quality GaN photoconductive detecto~using insulating material! were fabricated by Khanet al.144

The active region of detector was a 0.8-mm-thick epitaxiallayer of insulating GaN deposited over a 0.1-mm-thick AlNbuffer layer using metal-organic chemical vapor depositover basal-plane sapphire substrates. Interdigitated, elecal, 5000-Å-thick gold contacts were fabricated on the eayers for the photoconductive response measurements.interdigitated electrodes were 3mm wide, 1 mm long, andhad a 10mm spacing. To define the electrode pattern, a stdard photolithographic procedure and a lift-off techniqwere used.

Figure 40 shows the spectral responsivity of GaN p

FIG. 40. Normalized spectral responsivity for a GaN photoconductivetectorn~after Ref. 144!,


aputngn70er

s

dioning-h-areuseionof

-bri-rlye-

rs

ionctri-il-The

an-ue

o-

toconductive detectors. It reaches its maximum value at 3nm and then remains nearly constant down to 200 nm. Tpeak responsivity is about 1000 A/W, and the estimatgain/quantum efficiency product for the detectors is 4.83103.The response of GaN detectors is linear over five ordersincident radiation power, which is shown in Fig. 41. Thdetector response time is about 1 ms.

The performance ofn-type GaN photoconductive detectors made from films grown by molecular beam epitaxyinferior.151 The following parameters have been measurfor interdigitated device with Ti/Al ohmic contacts and 2mm inter-electrode spacing: the gain/quantum efficienproduct of 600 at 254 nm and at a bias of 25 V, the mobilitlifetime product of 9.531025 cm2/V and current responsivityof 125 A/W at 254 nm.

The responsivity ofp-type GaN:Mg layers on Si sub-strate is lower in comparison with insulating layers, anachieved a maximum value of 30 A/W at 14 V bias.149 Theresponsivity increases nearly linearly with applied voltageto 8 V, then the increase slows toward saturation. Photocrent decay for a pulsed input shows a hyperbolic type rsponse. It is hypothesized that holes are captured at eicompensated Mg deep acceptor sites or Mg-related trrecombination centers, resulting in a greatly prolonged eletron free carrier lifetime.

Recently Kunget al.150 have studied the kinetics of thephotoconductivity ofn-type GaN photoconductive detectorby the measurements of the frequency dependent photosponse and photoconductivity decay. Strongly sub-linearsponse and excitation-dependent response time have bobserved even at relatively low excitation levels. This canattributed to the redistribution of the charge carriers wiincreased excitation level.

Figure 42 shows the voltage responsivity of the Gadetector as a function of the modulation frequency for thrdifferent excitation power densities. From Fig. 42 we can sthat the responsivity remains frequency dependent overwhole range of frequencies applied. This dependence is lpronounced for higher excitation levels. At high frequenciwhen the period of radiation power oscillations exceeds tlifetime, the frequency response curves for different excit

de-FIG. 41. GaN photoconductive detector responsivity as a function of indent power under 254 nm excitation~after Ref. 144!,


calob-

-on

inn-,

Fig-ide

n-1ebar-ed5.ncedl

di-

tion level tends to have the same frequency dependence.is because at high frequencies the photogenerated exconcentration becomes dependent only on the excitationriod rather than the lifetime of the majority carrier lifetimA consequence of the redistribution of the charge carrwith increased excitation level is the decreasing of the ptoconductive gain with increasing in UV intensity, which hbeen recently observed by Goldenberget al.153 ~see Fig. 43!.

The observed kinetics of photoresponse have imporconsequences for practical applications. The measuredtimes are too long for many applications resulting in tfrequency dependent responsivity and nonlinear respoThis can be changed in a number of ways including delibate introduction of recombination centers to stabilizesponse time at a required level and by reducing the lifetin short photoconductors using contact phenomena. Furwork is necessary to understand the origin and propertiethe deep centers in GaN.

2. Photovoltaic detectors

The first Schottky GaN barriers were reported overyears ago.154 However, due to a fabrication procedure th

FIG. 42. GaN detector responsivity as a function of chopping frequencythree different excitation power densities of the He–Cd laser light~325 nm!.The solid line shows the fit to conventional dependenceRv5Rvo~114p2f 2t2)1/2 ~after Ref. 150!.

FIG. 43. The variation of photoconductive gain of GaN detector with intsity of light at 350 nm~after Ref. 153!.


Thiscesspe-e.iersho-as

tantlife-hense.er-re-imethers of

20at

was not well controlled, an anomalous behavior of electriproperties of these Zn/GaN Schottky barriers has beenserved.

More recently, Khanet al.145 have reported the fabrication and characterization of Schottky barrier photodiodesp-type GaN layers. As-grown 2-mm-thick material~grownover basal plane sapphire using MOCVD! was ann-typeconductivity with residual carrier concentration of 1017 cm23

and a room-temperature mobility of about 400 cm2/V s. Post-growth, p-type conductivity was enhanced using a 20 m650 °C N2 ambient anneal. After this process, the hole cocentration was about 731017 cm23. Subsequent to growthdoughnut-shaped Cr/Au~2000 Å thick! ohmic contacts weredeposited using standard photolithographic techniques.ure 44 shows the resulting device configuration. The insdiameters for the doughnuts were ranged from 200mm to 1mm. For improving the contact resistivities, the ohmic cotacts were annealed in flowing forming gas at 350 Å formin. As a Schottky barrier 1500-Å-thick Ti/Au metals werdeposited. The estimated responsivity of these Schottkyriers was 0.13 A/W at wavelength 320 nm for a normalizarea of 1 mm2. The spectral responsivity is shown in Fig. 4The response time of detectors is limited by the resistacapacitance (RC) time constant of the measured circuit anwas around 1ms. Other geometry of detectors with a Ti/ASchottky barrier is shown in Fig. 46~a!.

for

of

en-

FIG. 44. Schematic device configuration of GaN Schottky barrier photoode ~after Ref. 145!.

FIG. 45. Spectral responsivity of GaN Schottky barrier photodiode~afterRef. 145!.


ni

-

o

-

oa

ne

e

ae

h

t

f

o

ra-t-

ofl-is

i-mi-

oveeted

iv-.

m,o-

It should be noticed that much information concerniGaN Schottky barriers is under discussion. The barrier hedepends on a procedure used for diode fabrication. Forample, Linet al.130 used Ti/Al and Ti/Au metallization andsubsequent annealing to receive ohmic contacts ton-typeGaN. Instead, Binariet al.155 have fabricated Schottky barriers using a Ti layer and a Au overlayer depositedelectron-beam evaporation. The barrier height of Tin-type GaN has been measured to be 0.58 and 0.59 eVcapacitance–voltage (C–V) and current–voltage (I –V)techniques, respectively. The barrier height of Au onn-typeGaN was determined to be 1.19 and 1.15 eV fromI –V andC–V measurements155 which is somewhat lower than reported previously by Hackeet al.156 ~0.84 eV by I –V and0.94 eV byC–V!. Also, results presented by Guoet al.157

for Pt and Pd barriers confirm that the metal work functishould not be the only factor affecting the Schottky charteristics.

The first information aboutp-n junction GaN photovol-taic detectors is included in Ref. 109. Zhanget al.158 haveestimated by theoretical modeling of the spectral resposhape ofp-n junction that the diffusion length of holes in thn-type GaN region is about 0.1mm. Next, Chenet al.152

have studied two detector structures grown by low-pressMOCVD, one n-on-p and the otherp-on-n. The p-on-nsample consisted of the deposition over basal plane sappof a 1-mm-thick unintentionally dopedn-GaN layer followedby a 0.5-mm-thick Mg-dopedp-GaN layer. Similarly, then-on-p structure consisted of a 1mm p-GaN layer followedby a 0.5mm n-GaN layer. Large area mesa structures wdefined using reactive ion etching@Fig. 46~b!#. Ohmic con-tacts were formed using sputtered W for n- and e-beevaporated Ni/Au bilayer for thep contacts. These electrodpatterns were formed by standard lift-off processes. Tspectral responsivity data are plotted in Fig. 47. The mamum responsivity~0.09 A/W! is comparable to that of aUV-enhanced Si detector. The rise and fall times of the ptoresponse were measured to be;0.4 ms for a detector areaof 16 mm2. However, these times are significantly greathan the value of 1ms for the estimatedRC time constant.More work is needed to identify the factors responsiblethe decrease in the photodiode speed.

More sophisticated structures are also used for the detion of UV radiation. Recently, Khanet al.147,148 have re-ported the GaN/AlGaN heterostructure field-effect transist

FIG. 46. GaN photovoltaic detector structures:~a! Schottky barrier and~b!p-n junction ~after Ref. 109!.


gghtex-

bynby

nc-

se

ure

hire

re

m

hexi-

o-

er

or

tec-

rs

~HFETs! as gated visible blind photodetectors. These 0.2mmgate HFETs operated as microwave amplifiers to a tempeture of 300 °C with maximum oscillation frequency and cuoff frequency of'70 MHz and'22 GHz, respectively, atroom temperature. Figure 48 shows the schematic diagramthe short gate GaN/AlGaN HFET. The thickness of the AGaN barrier layer is approximately 25 nm, and its dopingon the order of 431018 cm23. The GaN layer is unintention-ally doped. The gated photodetectors are backside illumnated. The electron-hole pairs are generated in the seinsulating i -GaN buffer layer and in the activen-GaNregion. The photogenerated holes move towards thei -GaN/sapphire interface, and the photogenerated electrons mtowards the channel where they are quickly driven into thdrain by the high channel field. The measured and calcularesponsivity of a backside illuminated GaN/AlGaN HFETphotodetector is shown in Fig. 49. The maximum responsity of 3000 A/W was achieved at a gate voltage of about 1 VThe spectral responsivity~see Fig. 50! falls sharply by twoorders of magnitude for wavelengths larger than 365 nwhich demonstrates the visible blind operation of gated phtodetector.

FIG. 47. Spectral responsivity of GaNp-n detectors:~a! light from thenside; ~b! light from thep side; ~c! spectral responsivity of UV-enhanced Sip-n junction ~after Ref. 152!.

FIG. 48. Schematic diagram of the GaN/AlGaN HFET~after Ref. 159!.


o

n

-c--

airhe

-n-

re-orEx-gs--fu-

ro-ofli-

eedide

VIII. OTHER MATERIALS FOR ULTRAVIOLETDETECTORS

A straightforward way to make a highly selective~blind-less! UV detectors is to use wide band-gap semiconductHowever, due to technological difficulties with those marials, the UV detectors are fabricated also with Si and GaTheir principal advantage lies in the high crystalline qualof the materials and their well-known processing methoRecently, Constantet al.160 have presented GaAs photocoductive detectors with static responsivity values of 107–108

A/W in UV spectral range. Some applications require tmeasurement of UV radiation in the presence of visible ligThe final design of the devices can include additional bloing filters or dielectric coating to fit them for the wavelengof interest.

Misra et al.161 have presented preliminary results cocerning fabrication and characterization of BNxP12x photo-conductive detectors. This alloy system is an attractive mterial for UV application because its band gap can be tu

FIG. 49. Measured~solid line! and calculated~dashed line! responsivity of0.2 mm gate GaN/AlGaN HFET photodetector backside illuminated wHe–Cd laser~l5325 nm!. Drain-to-source voltage 10 V~after Ref. 147!.

FIG. 50. Measured spectral responsivity of 0.2mm gate GaN/AlGaN HFETphotodetector~after Ref. 147!.


rs.te-As.ityds.n-

heht.ck-th

n-

a-ed

in the UV region~between 200 and 400 nm! by adjusting theboron phosphide~BP! to boron nitride~BN! ratio in the ter-nary compound.

Also recently, for the first time, to the best of our knowledge, the dramatic improvement in diamond photocondutive characteristics~106 higher response to 200 nm than visible wavelengths,,0.1 nA dark current! has been achievedfollowing methane-air treatment of the polycrystallinematerial.162 Figure 51~curve 3! shows the dramatic changein the response~for wavelengths less than 225 nm! after700 °C treatment in a methane ambient~1 h! and secondtreatment at 400 °C in air~1 h!. The improvement in thedevice characteristics indicates an active role for methane-gases in modifying the surface, or near surface, region of tthin-film polycrystalline diamond.

IX. CONCLUSIONS

Currently, the old detection techniques used in UV technology~photographic films, detectors based on gas photoioization, and photoemissive detectors! are replaced by semi-conductor photon detectors where higher sensitivity, moquantitatively accurate photometric information, or more immediate data availability are required. The semiconductdetectors offer small size, ruggedness, and ease of use.pansion of the semiconductor industry and the continuinemphasis on the development of low-light level imaging sytems for military and civilian surveillance applications ensure the dominant position of these detectors in the nearture. Manufactured using proven wafer-scale silicon or III–Vprocessing techniques, they are inherently cheaper and pvide high quantum efficiency. At present, a new classvideo-rate imagers based on thinned, back-illuminated sicon charge coupled devices~CCDs! with improved UV sen-sitivity ~to 50% quantum efficiency at 200 nm! is available toreplace conventional image intensifiers.162 However, Si pho-todetectors require bulky band filters to block the visiblsolar radiation background and are intolerant of elevattemperatures and caustic environments. With the use of wband-gap semiconductors~like GaN or diamond!, the need

ith

FIG. 51. Spectral responsivity~arbitrary units! of diamond photoconductivedevice: ~1! as-fabricated,~2! following methane annealing, and~3! aftermethane and air annealing~after Ref. 162!.


-

.

ri,

.

.

for these filters will be largely eliminated simplifying thedesign of the spectroscopic monitoring equipment. Morover, these devices are potentially free from the mentionshortcomings; they are capable of operating at high tempetures and in hostile environments.

Interest in a new generation of wide band-gap semicoductor detectors stems in part from the fact that the Earthatmosphere is opaque at wavelength below 300 nm. To taadvantage of this optical window, Earth to space communcation will need detectors that are UV sensitive but blindthe ambient visible radiation. Another area where these dtectors can find applications is in the combustion monitorinof gases where UV emission is a normal by-product.

Semiconductor diamond is of considerable interest fUV detectors. Unfortunately, diamond of a good enougquality is not readily available, and appropriate technologyfar from being mature. However, recent progress in diamothin-film technology is promising for future applications.

Unlike SiC, which is another wide band-gap semiconductor with demonstratedn- andp-type doping and excellentpower device performance, one notable advantage of III–nitrides is that they form direct band-gap heterostructurehave better ohmic contacts and heterostructures. Becausthis, it is anticipated that III–V nitrides may eventually proveto be more promising than SiC for optoelectronic devices.

The future direction in research and developmentGaN-based UV detectors~mainly photodiodes! can bejudged from the remarkable progress in their performanattained over the last few years. A late newspaper at anternational Electron Device Meeting announces the first oservation of stimulated emission from a current injecteInGaN/AlGaN double-heterostructure diode.163,164The majorobstacle prohibiting the sort of improvement needed is tlarge concentration of defects, although some attribute soof the shortcomings to processing-related problems. Larconcentrations of defects~107–1010 cm22! in GaN structuresmanifest themselves as short carrier lifetimes and reducphotoluminescence intensity. Further progress will molikely be due to the use of better matched substrates. A grdeal of effort is presently being expended in laboratories eerywhere to search for a better substrate or compliant basedecrease the defect generation, and thus increase quanefficiency and photoconductive gain of detectors.165

1J. A. R. Samson,Techniques of Vacuum Ultraviolet Spectroscopy~Wiley,New York, 1967!.

2The Middle Ultraviolet: Its Science and Technology, edited by A. E. S.Green~Wiley, New York, 1966!.

3L. R. Koller, Ultraviolet Radiation~Wiley, New York, 1965!.4A. N. Zaidel and E. Ya. Shreider,Vacuum Ultraviolet Spectroscopy~Humphrey, Ann Arbor, MI, 1970!.

5J. Hennes and L. Dunkleman,The Middle Ultraviolet: Its Science andTechnology, edited by A. E. S. Green~Wiley, New York, 1966!, Chap. 15.

6R. E. Huffman,Atmospheric, Ultraviolet Remote Sensing, InternationalGeophysics Series~Academic, New York, 1992!, Vol. 52.

7R. E. Huffman,Ultraviolet Optics and Technology, SPIE Milestone Series~SPIE Optical Engineering, Bellingham, Washington, 1993!, Vol. MS 80.

8G. R. Carruthers,Electro-Optics Handbook, edited by R. W. Waynant andM. N. Ediger ~McGraw-Hill, New York, 1994!, Chap. 15.

9J. G. Timothy and R. P. Madden,Handbook on Synchrotron Radiation,edited by E. E. Koch~North-Holland, Amsterdam, 1983!, pp. 315–366.

10M. J. Eccles, M. E. Sim, and K. P. Tritton,Low Light Level Detectors inAstronomy~Cambridge University Press, Cambridge, England, 1983!.


e-edra-

n-’skei-toe-g

orhisnd

-

Vs,e of

of

ceIn-b-d

hemege

edreeatv-totum

11J. G. Timothy, Publ. Astron. Soc. Pac.95, 810 ~1983!.12J. R. Janesick, T. Elliott, S. Collins, H. Marsh, M. Blouke, and M. Freeman, Proc. SPIE501, 2 ~1984!.

13B. Y. Welsh and M. Kaplan, Proc. SPIE1743, 452 ~1992!.14C. L. Joseph, Wisconsin Astrophy. No. 562, 1~1995!.15M. P. Ulmer, M. Razeghi, and E. Bigan, Proc. SPIE2397, 210 ~1995!.16J. R. Janesick, D. Campbell, T. Elliott, and T. Daud, Opt. Eng.26, 825

~1987!.17R. A. Stern, R. C. Catura, R. Kimble, A. F. Davidsen, M. Winzenread, MM. Blouke, R. Hayes, D. M. Walton, and J. L. Culhane, Opt. Eng.26, 875~1987!.

18G. Naletto, G. Tondello, G. Bonanno, R. Di Benedetto, and S. ScudeOpt. Eng.38, 2544~1994!.

19M. M. Blouke, M. W. Cowens, J. E. Hall, J. A. Westphal, and A. B.Christensen, Appl. Opt.19, 3318~1980!.

20M. W. Cowens, M. M. Blouke, T. Fairchild, and J. A. Westphal, Appl.Opt: 19, 3727~1980!.

21J. R. Janesick, T. Elliott, G. Fraschetti, S. Collins, M. Blouke, and BCorrie, Proc. SPIE1071, 153 ~1989!.

22A. Rose, Concepts in Photoconductivity and Allied Problems~Inter-science, New York, 1963!.

23J. Piotrowski,Infrared Photon Detectors, edited by A. Rogalski~SPIEOptical Engineering, Bellingham, Washington, 1995!, Chap. 11.

24P. W. Kruse, L. D. McGlauchlin, and R. B. McQuistan,Elements of In-frared Technology~Wiley, New York, 1962!.

25P. W. Kruse,Optical and Infrared Detectors, edited by R. J. Keyes~Springer, Berlin, 1977!, pp. 5–69.

26D. H. Seib and L. W. Aukerman,Advances in Electronics and ElectronPhysics, edited by L. Morton~Academic, New York, 1973!, Vol. 34, pp.95–221.

27W. Van Roosbroeck, Bell Syst. Tech. J.29, 560 ~1950!.28X. Zhang, P. Kung, D. Walker, J. Piotrowski, A. Rogalski, A. Saxler, andM. Razeghi, Appl. Phys. Lett.67, 2028~1995!.

29R. A. Smith,Semiconductors~Cambridge University Press, Cambridge,England, 1978!.

30E. S. Rittner,Photoconductivity Conference at Atlantic City 1954, editedby R. Breckenbridge, B. Russel, and E. Hahn~Wiley, New York, 1956!,pp. 215–268.

31C. T. Elliott, Handbook on Semiconductors, edited by C. Hilsum~North-Holland, Amsterdam, 1993!, Vol. 4, pp. 841–936.

32A. Van der Ziel, Fluctuation Phenomena in Semiconductors~Butter-worths, London, 1959!.

33F. N. Hooge, Phys. Lett.29 A, 123 ~1969!.34R. D. Nelson, Opt. Eng.16, 275 ~1977!.35V. Gopal, Infrared Phys.25, 615 ~1985!.36H. J. Hovel,Semiconductors and Semimetals, edited by R. K. Willardsonand A. C. Berr~Academic, New York, 1975!, Vol. 11.

37T. S. Moss, G. J. Burrel, and B. Ellis,Semiconductor Optoelectronics~Butterworths, London, 1973!.

38M. J. Buckingham and E. A. Faulkner, Radio Electron. Eng.44, 125~1974!.

39C. T. Sah, R. N. Noyce, and W. Shockley, Proc. IRE45, 1228~1957!.40S. C. Choo, Solid-State Electron.11, 1069~1968!.41S. M. Sze,Physics of Semiconductor Devices~Wiley, New York, 1981!.42R. J. McIntyre, IEEE Trans. Electron DevicesED-13, 164 ~1966!.43A. Rose, Phys. Rev.97, 1538~1955!.44M. A. Lampert, Phys. Rev.103, 1648~1956!.45M. A. Lampert and P. Mark,Current Injections in Solids, edited by H. G.Booker and N. DeClaris~Academic, New York, 1970!.

46J. A. Edmond, K. Das, and R. F. Davis, J. Appl. Phys.63, 922 ~1988!.47D. E. Sawyer and R. H. Rediker, Proc. IRE46, 1122~1958!.48W. W. Gartner, Phys. Rev.116, 84 ~1959!.49E. R. Rhoderick, IEE Proc.129, 1 ~1982!.50A. M. Cowley and S. M. Sze, J. Appl. Phys.36, 3212~1965!.51W. Schottky and E. Spenke, Wiss. Veroff. Siemens-Werken.18, 225

~1939!.52E. Spenke,Electronic Semiconductors~McGraw-Hill, New York, 1958!.53H. A. Bethe, MIT Radiation Laboratory Report No. 43-12, 1942.54C. R. Crowell and S. M. Sze, Solid-State Electron.9, 1035~1966!.55G. Ottaviani, K. N. Tu, and J. M. Mayer, Phys. Rev. Lett.44, 284~1980!.56M. Krumrey and E. Tegeler, Nucl. Instrum. Methods A288, 114 ~1990!.57S. Strite and H. Morkoc, J. Vac. Sci. Technol. B10, 1237~1992!.58H. Morkoc, S. Strire, G. B. Gao, M. E. Lin, B. Sverdlov, and M. Burns, JAppl. Phys.76, 1363~1994!.


a

C

a

C

o,

.

-

s.

ys.

l.

pl.

e,

A.

tz,

.

ys.

es.

ng.

59S. N. Mohammad, A. A. Salvador, and H. Morkoc, Proc. IEEE83, 1306~1995!.

60E. O. Johnson, RCA Rev.26, 163 ~1965!.61R. W. Keyes, Proc. IEEE60, 225 ~1972!.62M. Marchywka, S. C. Binari, P. E. Pehrsson, and D. Moses, Proc. S2282, 20 ~1994!.

63Y. Talmi and R. W. Simpson, Appl. Opt.19, 1401~1980!.64R. Korde, L. R. Canfield, and B. Wallis, Proc. SPIE932, 153 ~1988!.65E. Kooi, IEEE Trans. Electron DevicesED-13, 238 ~1966!.66T. E. Hansen, Phys. Scr.18, 471 ~1978!.67R. Korde and J. Geist, Appl. Opt.26, 5284~1987!.68United Detector Technology Inc., Data Sheet~1980!.69R. Korde and J. Geist, Solid-State Electron.30, 89 ~1987!.70L. R. Canfield, J. Kerner, and R. Korde, Appl. Opt.28, 3940~1989!.71S. G. Chamberlain, J. Appl. Phys.50, 7228~1979!.72H. Ouchi, T. Mukai, T. Kamei, and M. Okamura, IEEE Trans. ElectrDevicesED-26, 1965~1979!.

73C. M. Botchek,VLSI Basic MOS Engineering~Pacific Technical, Sara-sota, FL, 1984!, Vol. 1, p. 275.

74Z. A. Weinberg, D. R. Young, J. A. Calise, S. A. Cohen, J. C. DeLuand V. R. Deline, Appl. Phys. Lett.45, 1204~1984!.

75R. S. Popovic, K. Solt, U. Falk, and Z. Stoessel, Sensors and ActuaA21–A23, 553 ~1990!.

76R. F. Wolffenbuttel, Sensors and ActuatorsA21–A23, 559 ~1990!.77W. Huang, C. Salupo, L. F. Szabo, G. P. Ceaser, and W. Iavurek, MRes. Soc. Symp. Proc.118, 411 ~1988!.

78Y. K. Fang, S. B. Hwang, K. H. Chen, C. R. Liu, M. J. Tsai, and L.Kuo, IEEE Trans. Electron DevicesED-39, 292 ~1992!.

79R. C. Alig, S. Bloom, and C. W. Struck, Phys. Rev. B22, 5565~1980!.80E. Sakai, Nucl. Instrum. Methods196, 121 ~1982!.81T. Kabayashi, Appl. Phys. Lett.21, 150 ~1972!.82L. R. Canfield and J. Kerner, Proc. SPIE1344, 372 ~1990!.83L. R. Canfield, R. Vest, T. N. Woods, and R. Korde, Proc. SPIE2282, 31

~1984!.84M. Lavagna, J. P. Pigue, and Y. Marfaing, Solid-State Electron.20, 235

~1977!.85R. D. Baertsch and J. R. Richardson, J. Appl. Phys.40, 229 ~1969!.86H. Fabricius, T. Skettrup, and P. Bisgaard, Appl. Opt.25, 2764~1986!.87M. Krumrey, E. Tegler, J. Barth, M. Krisch, F. Schafers, and R. WoAppl. Opt.27, 4336~1988!.

88E. Tegeler and M. Krumrey, Nucl. Instrum. Methods A282, 701 ~1989!.89A. D. Wilson and H. Lyall, Appl. Opt.25, 4530~1986!.90M. Marchywka, J. F. Hochedez, M. W. Geis, D. G. Socker, D. Moses,R. T. Goldberg, Appl. Opt.30, 5010~1991!.

91G. Zhao, T. Stacy, E. M. Charlson, E. J. Charlson, M. Hajsaid, R. Rchoudhury, A. H. Khan, J. M. Meese, Z. He, and M. Prelas, Mater. RSoc. Symp. Proc.339, 191 ~1994!.

92M. C. Hicks, C. R. Wronski, S. A. Grot, G. Sh. Gildenblat, A. R. BadziaT. Badzian, and R. Messier, J. Appl. Phys.65, 2139~1989!.

93R. B. Campbell and H. Chang, Solid-State Electron.10, 949 ~1967!.94P. Glasow, G. Ziegler, W. Suttrop, G. Pensl, and R. Helbig, Proc. S868, 40 ~1987!.

95D. M. Brown, E. T. Downey, M. Ghezzo, J. W. Kretchmer, R. J. Saia,S. Liu, J. A. Edmond, G. Gati, J. M. Pimbley, and W. E. Schneider, IETrans. Electron DevicesED-40, 325 ~1993!.

96J. A. Edmond, H. S. Kong, and C. H. Carter Jr., Phys. B185, 453~1993!.97Y. Hirabayashi, S. Karasawa, K. Kobayashi, S. Misawa, and S. YoshSensors and Actuators A43, 164 ~1994!.

98Y. K. Fang, S. B. Hwang, K. H. Chen, C. R. Liu, M. J. Tsai, and L.Kuo, IEEE Trans. Electron DevicesED-39, 292 ~1992!.

99G. Cesare, F. Irrera, F. Palma, M. Tucci, E. Jannitti, G. Naletto, andNicolosi, Appl. Phys. Lett.67, 335 ~1995!.

100C. Frojdh, G. Thungstro¨m, H. E. Nilsson, and C. S. Petersson, Phys. ST 54, 169 ~1994!.

101C. Frojdh, G. Thungstro¨m, H. E. Nilsson, and C. S. Petersson, MateRes. Soc. Symp. Proc.339, 215 ~1994!.

102M. M. Anikin, A. N. Andreev, S. N. Pyatko, N. S. Savkina, A. M. Strechuk, A. L. Syrkin, and V. E. Chelnokov, Sensors and Actuators A33, 91~1992!.

103M. M. Anikin, A. N. Andreev, A. A. Lebedev, S. N. Pyatko, M. GRastegaeva, N. S. Savkina, A. M. Strelchuk, A. L. Syrkin, and V.Chelnokov, Sov. Phys. Semicond.25, 198 ~1991!.

104L. G. Matus, J. A. Powell, and C. S. Salupo, Appl. Phys. Lett.59, 1770~1991!.


PIE

on

ca,

tors

ter.

.

lf,

nd

oy-es.

n,

PIE

Y.EE

ida,

.

P.

cr.

r.

l-

.E.

105P. G. Neudeck, D. J. Larkin, J. A. Powell, L. G. Matus, and C. S. SalupAppl. Phys. Lett.64, 1386~1994!.

106S. Nakamura, T. Mukai, M. Senoh, and N. Iwasa, Jpn. J. Appl. Phys.31,L139 ~1992!.

107Y. Ohba and A. Hatano, Jpn. J. Appl. Phys.33, L1367 ~1994!.108I. Akasaki and H. Amano, J. Electrochem. Soc.141, 2266~1994!.109M. A. Khan, Q. Chen, C. J. Sun, M. S. Shur, M. F. Macmillan, R. PDevaty, and J. Choyke, Proc. SPIE2397, 283 ~1995!.

110S. Strite, M. E. Lin, and H. Morkoc, Thin Solid Films231, 197 ~1993!.111R. F. Davis, Proc. IEEE79, 702 ~1991!.112S. Nakamura, Y. Harada, and M. Senoh, Appl. Phys. Lett.58, 2021

~1991!.113V. W. Chin, T. L. Tansley, and T. Osotchan, J. Appl. Phys.75, 7365

~1994!.114R. P. Joshi, Appl. Phys. Lett.64, 223 ~1994!.115M. A. Khan, J. N. Kuznia, J. M. Van Hove, D. T. Olson, S. Krishnankutty, and R. M. Kolbas, Appl. Phys. Lett.58, 526 ~1991!.

116H. Amano, M. Kito, K. Hiramatsu, and I. Akasaki, Jpn. J. Appl. Phys.28,L2112 ~1989!.

117S. Nakamura, N. Iwasa, M. Senoh, and T. Mukai, Jpn. J. Appl. Phys.31,1258 ~1992!.

118T. D. Moustakas and R. J. Molnar, Mater. Res. Soc. Proc.281, 753~1993!.

119M. E. Lin, G. Xue, G. L. Zhu, J. E. Greene, and H. Morkoc, Appl. PhyLett. 63, 932 ~1993!.

120D. K. Gaskill, L. B. Rowland, and K. Doverspike, inProperties of GroupIII Nitrides, edited by J. H. Edgar~INSPEC, IEE, London, 1994!, pp.101–116.

121M. Rubin, N. Newman, J. S. Chan, T. C. Fu, and J. T. Ross, Appl. PhLett. 66, 64 ~1994!.

122S. J. Pearton, C. B. Vartuli, J. C. Zolper, C. Yuan, and R. A. Stall, AppPhys. Lett.67, 1435~1995!.

123R. G. Wilson, S. J. Pearton, C. R. Abernathy, and J. M. Zavada, ApPhys. Lett.66, 2238~1995!.

124C. H. Qiu, C. Hoggatt, W. Melton, M. W. Leksono, and J. I. PankovAppl. Phys. Lett.66, 2712~1995!.

125J. S. Foresi and T. D. Moustakas, Appl. Phys. Lett.62, 2859~1993!.126K. Kurtin, T. C. McGill, and C. A. Mead, Phys. Rev. Lett.22, 1433

~1969!.127C. M. Wolfe, N. Holonyak, Jr., and G. S. Stillman,Physical Properties ofSemiconductors~Prentice-Hall, Englewood Cliffs, NJ, 1989!.

128J. I. Pankove and H. E. P. Schade, Appl. Phys. Lett.25, 53 ~1974!.129K. W. Boer,Survey of Semiconductor Physics~Van Nostrand, New York,1990!.

130M. E. Lin, Z. Ma, F. Y. Huang, Z. F. Fan, L. H. Allen, and H. Morkoc,Appl. Phys. Lett.64, 1003~1994!.

131M. E. Lin, F. Y. Huang, and H. Morkoc, Appl. Phys. Lett.64, 2557~1994!.

132S. J. Pearton, C. R. Abernathy, F. Ren, J. R. Lothian, P. W. Wisk, andKatz, J. Vac. Sci. Technol. A1, 1772~1993!.

133S. J. Pearton, C. R. Abernathy, F. Ren, J. R. Lothian, P. Wisk, A. Kaand C. Constantine, Semicond. Sci. Technol.8, 310 ~1993!.

134I. Adesida, A. Mahajan, E. Andideh, M. A. Khan, D. T. Olsen, and J. NKuznia, Appl. Phys. Lett.63, 2777~1993!.

135M. E. Lin, Z. F. Fan, Z. Ma, L. H. Allen, and H. Morkoc, Appl. Phys.Lett. 64, 887 ~1994!.

136S. J. Pearton, C. R. Abernathy, and F. Ren, Appl. Phys. Lett.64, 2294~1994!.

137S. J. Pearton, C. R. Abernathy, and F. Ren, Appl. Phys. Lett.64, 3643~1994!.

138S. J. Pearton, C. R. Abernathy, F. Ren, and J. R. Lothian, J. Appl. Ph76, 1210~1994!.

139S. J. Pearton, C. R. Abernathy, and C. B. Vartuli, Electron. Lett.30, 1985~1994!.

140S. J. Pearton, C. R. Abernathy, F. Ren, and J. R. Lothian, Mater. RSoc. Symp. Proc.339, 179 ~1994!.

141S. J. Pearton, C. B. Vartuli, R. J. Shul, and J. C. Zolper, Mater. Sci. EB 31, 309 ~1995!.

142A. T. Ping, I. Adesida, and M. A. Khan, Appl. Phys. Lett.67, 1250~1995!.


,

s.

pl.

d

.

nd

c.

143H. Lee, D. B. Oberman, and J. S. Harris, Appl. Phys. Lett.67, 1754~1995!.

144M. A. Khan, J. N. Kuznia, D. T. Olson, J. M. Van Hove, M. Blasingameand L. F. Reitz, Appl. Phys. Lett.60, 2917~1992!.

145M. A. Khan, J. N. Kuznia, D. T. Olson, M. Blasingame, and A. RBhattarai, Appl. Phys. Lett.63, 2455~1993!.

146M. A. Khan, Microwave J.67 ~November 1993!.147M. A. Khan, M. S. Shur, Q. Chen, J. N. Kuznia, and C. J. Sun, ElectroLett. 31, 398 ~1995!.

148M. S. Shur and M. A. Khan, Proc. SPIE2397, 294 ~1995!.149K. S. Stevens, M. Kinniburgh, and R. Beresford, Appl. Phys. Lett.66,3518 ~1995!.

150P. Kung, X. Zhang, D. Walker, A. Saxler, J. Piotrowski, A. Rogalski, anM. Razeghi~to be published!.

151M. Misra, T. D. Moustakas, R. P. Vaudo, R. Singh, and K. S. Shah, PrSPIE2519, 78 ~1995!.

152Q. Chen, M. A. Khan, C. J. Sun, and J. W. Yang, Electron. Lett.31, 1781~1995!.

153B. Goldenberg, J. D. Zook, and R. J. Ulmer, inTopical Workshop onIII –V Nitrides Proc.~North Holland, Nagoya, 1995!.

154Y. Morimoto and S. Ushio, Jpn. J. Appl. Phys.13, 365 ~1974!.


,

.

n.

d

oc.

155S. C. Binari, H. B. Dietrich, G. Kelner, L. B. Rowland, K. Doverspikeand D. K. Gaskill, Electron. Lett.30, 909 ~1994!.

156P. Hacke, T. Detchprohm, K. Hiramatsu, and N. Sawaki, Appl. PhyLett. 63, 2676~1993!.

157J. D. Guo, M. S. Feng, R. J. Guo, F. M. Pan, and C. Y. Chang, ApPhys. Lett.67, 2657~1995!.

158X. Zhang, P. Kung, D. Walker, J. Piotrowski, A. Rogalski, A. Saxler, anM. Razeghi, Appl. Phys. Lett.76, 2028~1995!.

159M. A. Khan, J. N. Kuznia, D. T. Olson, W. Schaff, J. Burm, and M. SShur, Appl. Phys. Lett.65, 1121~1994!.

160M. Constant, D. Loridant, J. C. Camart, S. Meziere, L. Boussekey, aM. Chive, Proc. SPIE2367, 229 ~1995!.

161M. Misra, C. Zhou, P. B. Bennett, M. R. Squillants, and F. Ahmad, ProSPIE2282, 49 ~1994!.

162R. D. McKeag, S. S. M. Chan, and R. B. Jackman, Appl. Phys. Lett.67,2117 ~1995!.

163G. M. Williams and M. M. Blouke, Laser Focus World129 ~September1995!.

164Compound Semiconductor8 ~Nov./Dec. 1995!.165Compound Semiconductor9 ~Nov./Dec. 1995!.


semiconductor ultraviolet detectors - antoni rogalski · semiconductor ultraviolet detectors ......

Documents