Uncooled long wavelength infrared photon detectors - Antoni Rogalski

17 Pages·2004·0.5 MB·English

InfraredPhysics&Technology46(2004)115–131
www.elsevier.com/locate/infrared
Uncooled long wavelength infrared photon detectors
J. Piotrowski a, A. Rogalski b,*
aVigoSystemLtd.,3S(cid:1)wietlik(cid:1)owStr.,01-397Warsaw,Poland
bInstituteofAppliedPhysics,MilitaryUniversityofTechnology,2KaliskiegoStreet,4900-908Warsaw,Poland
Availableonline6May2004
Abstract
Infrared(IR) photondetectors operating in the middle (3–5lm)and long wavelength (8–14 lm) infrared spectral
rangerequirecryogeniccoolingtoachieveusefulperformance.TheneedforcoolingisamajorlimitationofIRphoton
detectors what prevents more widespread use of IR technology. At present, uncooled thermal detector focal plane
arraysaresuccessfullyusedinstaringthermalimagers.However,theperformanceofthermaldetectorsismodest,they
sufferfromslowresponseandtheyarenotveryusefulinapplicationsrequiringmultispectraldetection.Inthepaper,a
numberofconceptstoimproveperformanceofphotondetectorsoperatingatroomtemperaturearepresented.Several
types of detector materials are considered: HgCdTe, Sb-based III–V ternary alloys, and type-II InAs/GaInSb super-
lattice. Initial efforts were concentrated on photoconductorsand photoelectromagnetic detectors. Recently, advanced
heterojunction photovoltaic detectors havebeen developed. Itis shownthatuncooled HgCdTephotovoltaic detector
canachievedetectivityof109cmHz1=2W(cid:1)1atthe8–9lmrange.Potentiallythedevicescanbeassembledinlargefocal
plane arrays. This will enable obtaining of NEDT of less than 0.1 K for staring thermal imagers operating with f/1
optics and30s(cid:1)1 frame rate.
(cid:1) 2004Elsevier B.V. Allrightsreserved.
1. Introduction eras, require cost-effective detectors that operate
without cooling or, at least, at temperatures com-
Progress in infrared (IR) photon detector tech- patible with long-life, low power and low cost
nology is mainly connected with semiconductor coolers.
infrared detectors. The most important are pho- A number of concepts to improve performance
todetectors fabricated with narrow-gap semicon- of the near room temperature IR photodetectors
ductors. To obtain high performance detectors, have been proposed [1–3]. Apart from photocon-
temperaturesmuch lower than 300 K are required ductive detectors and photodiodes, three other
(typically 80–200 K). Cooling requirements sig- types of IR photodetectors can operate at near
nificantly increase price and size of IR systems, room temperature; photoelectromagnetic (or
thereforeoneofthemainresearchgoalsisincrease PEM) detectors, magnetoconcentration detectors
in working temperature of photodetectors. and Dember effect detectors [3]. While significant
Affordable IR systems, such as IR imaging cam- improvements have been obtained by suppression
of Auger thermal generation in excluded photo-
conductors [4], extracted photodiodes [4–6], and
*Correspondingauthor.Tel./fax:[phone]. magnetoconcentration effect detectors [7,8], these
1350-4495/$-seefrontmatter (cid:1) 2004ElsevierB.V.Allrightsreserved.
doi:10.1016/j.infrared.2004.03.016
116 J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131
non-equilibrium devices require significant bias
1013
currents and exhibit excessive low frequency 1=f N =1014 cm-3
d
noise that extends up to MHz range. The noise, 1012
whichreducesthe800Hzdetectivityby2–3orders W) T=80 K
oqfuemncaigensiatubdoevectohmep1a=rfedknteoet(h(cid:2)a1t–m10eaMsuHrezd),aastwfreel-l 1/2Hz/1011 BLIP
as the high current requirements is severe obstacle cm1010 120 K
to Htheerirewwideesrperveieawd apprpolgicraetsisonisn. development of ectivity ( 109 200 K
uncooled long wavelength infrared (LWIR) pho- et
D
tondetectors.DifferenttypesofIRphotodetectors 108 300 K
especially photoconductors, photoelectromagnetic
107
(PEM) detectors, Dember effect detectors and 2 4 6 8 10 12 14 16 18 20
photodiodeswithmajoremphasisondevicesbased Cutoff wavelength (µm)
on Hg Cd Te (HgCdTe) alloys are considered.
1(cid:1)x x Fig. 1. Calculated performance of Auger generation-recombi-
Special attention is put on optimization of the nationlimitedHgCdTephotodetectorsasafunctionofwave-
devices for room temperature operation. lengthandtemperatureofoperation.BLIPdetectivityhasbeen
The review is mainly based on the authors calculated for 2p FOV, background temperature TBLIP ¼300
K,andquantumefficiencyg¼1.
experience in developing near room temperature
HgCdTe detectors at the Institute of Applied
Physics Military University of Technology and at figure of merit of any material intended for
Vigo System Ltd. An approach to reduce photo- infrared photodetectors. This figure should be
detector cooling requirements based of non-equi- used to assess any potential material. Analysis
librium mode of operation [4–6,9] and proposed shows, that the narrow gap semiconductors are
by British workers is omitted in our consider- more suitable for high temperature photodetec-
ations. tors in comparison to competing technologies
such as extrinsic devices, Schottky barrier, and
QWIP devices [11]. The main reason for high
performance of intrinsic photodetectors is high
2. Fundamental performance limits of near room
density of states in the valence and conduction
temperature photodetectors
bands, which results in strong absorption of
infrared radiation.
The performance of infrared photodetectors is
The Auger mechanism is likely to impose fun-
limited by statistical nature of generation and
damental limitations to the LWIR HgCdTe
recombination of charge carriers in the semicon-
detector performance. Fig. 1 shows the calculated
ductor. Thermal processes dominate at near room
detectivity of Auger generation-recombination
temperatureinmaterials usedformiddleandlong
limited HgCdTe photodetectors as a function of
wavelength devices. The detectivity of an opti-
wavelength and temperature of operation. The
mizedinfraredphotodetectorislimitedbythermal
calculations have been performed for doping level
processes in the active region of the device. It can
be expressed as [10] equal Nd ¼1014 cm(cid:1)3, as the lowest donor doping
level, which at present is achievable in a control-
k (cid:1)a(cid:2)1=2
D(cid:3) ¼0:31 k ; ð1Þ lable manner in practice. We can see that
hc G
where 16k62, and k is dependent on the con- • liquidnitrogencoolingpotentiallymakesitpos-
tribution of recombination and backside reflec- sible to achieve BLIP performance in the wide
tion. 2–20 lm range,
The ratio of the absorption coefficient to the • 200 K cooling, which is achievable with Peltier
thermal generation rate, a=G, is the fundamental coolers, would be sufficient for BLIP operation
J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131 117
in the middle and short wavelength IR regions
1013
(below 5 lm).
1012
It is interesting to consider the performance W)
rteecqtuoirrsemfoerntthseormf anleacarmroeorams.tTehmeprmeraalturreesopluhtoiotondoe-f 1/2Hz/ 1011
m
infrared thermal systems is usually characterized c1010
by the noise equivalent temperature difference vity (
(NETD). It can be shown, that [12] cti 109
e
et
D
4F2Df1=2(cid:4)Z kb dM (cid:5)(cid:1)1 108
NETD¼ D(cid:3)ðkÞdk ; ð2Þ
A1d=2top ka dT 107
2 4 6 8 10 12 14 16 18 20
whereF istheopticsf-number,Df isthefrequency Cutoff wavelength (µm)
band, A is the detector area, t is the optics
d op
Fig. 2. Detectivity needed to obtain NEDT¼0.1 K in a
transmission, M is the spectral emittance of the
photoncounterdetectorthermalimagerasafunctionofcutoff
blackbody described by the Plancks law.
wavelength.
As Eq. (2) shows, the thermal resolution im-
proves with an increase in detector area. Increas-
ing detector area results in reduced spatial
resolution, however. Hence, a reasonable com- of the uncooled photodetectors is not sufficient to
promise between the requirement of high thermal achieve the required 0.1 K thermal resolution.
and spatial resolution is necessary. Improvement The thermal resolution below 0.1 K is achieved
of thermal resolution without spatial resolution for staring thermal imagers containing thermal
worsening may be achieved by detector focal plane arrays. Generally, the per-
formance of thermal detectors are modest, they
• increase of detector area combined with corre- suffer from slow response, and they are not very
spondingincreaseoffocallengthandtheobjec- useful in applications requiring multispectral
tive aperture, detection.
• improved detector performance,
• increase in number of detectors.
3. Ways to improve detector’s performance without
Increase of aperture is undesirable because it cooling
increases sizes, mass and price of an infrared sys-
tem.Itismorepropertouseadetectorwithhigher There are some ways to improve the perfor-
detectivity. Another possibility is the application mance of the photodetectors without cooling. A
of multi-elemental sensor, what reduces each ele- moderate p-type doping of the absorber detector
ment bandwidth proportionally to the number of region is widely used for some suppression of the
elements for the same frame rate and other Auger mechanisms [1,2,4]. More efficient sup-
parameters. pression can be obtained with non-equilibrium
Fig. 2 shows the dependence of detectivity on depletion of the semiconductor [4]. However, the
cutoff wavelength for a photon counter detector non-equilibrium mode devices suffer from a high
thermal imager with a resolution of 0.1 K. De- level of flicker noise that make them useless for
tectivities of 1.9·108, 2.3·108, and 2·109 most of practical applications that require detec-
cmHz1=2/W are necessary to obtain NEDT¼0:1 tion of IR radiation in the low and moderate fre-
K for 10, 9, and 5 lm cutoff wavelength photon quency range. An example is thermal imaging.
counter detectors, respectively. The above esti- Hence, non-equilibrium mode devices are omitted
mations indicate, that the ultimate performance in our considerations.
118 J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131
A possible way to improve the performance of possible to improve collection efficiency in photo-
infrareddetectorsistoreducethephysicalvolume diodes since thickness can be made smaller than
ofthesemiconductor,thusreducingtheamountof the diffusion length. Making use of interference
thermal generation. This must be done without effects, highly non-uniform absorption can be
decrease in quantum efficiency, optical area, and achievedevenforlongwavelengthradiationwitha
field of view (FOV) of the detector. low absorption coefficient, which is important for
some devices such as photoelectromagnetic and
3.1. Reduction of thickness Dember effect devices [1,3].
It should be noted that the interference effects
Thickness of the active region can be signifi- strongly modify spectral response of the device.
cantly reduced by enhanced absorption of radia- The gain due to the optical cavitycan be achieved
tion. This can be achieved by using multiple pass only in narrow spectral regions. This may be an
of radiation with a backside reflector. Even more important limitation for some applications that
efficientistheuseofinterferencephenomenatoset require wide spectral band sensitivity. In practice,
up a resonant cavity within the photodetector [1]. infraredsystemsusuallyoperateinaspectralband
Various optical resonator structures have been (e.g. atmospheric windows) and the use of reso-
proposed as shown in Fig. 3. nant cavities may yield significant gains. Another
In the simplest method, interference occurs be- limitationcomesfromthefactthatefficientoptical
tween the waves reflected at the rear, highly resonance occurs only for near-perpendicular
reflective surface and at the front surface of incidenceandislesseffectiveforobliqueincidence.
semiconductor. The thickness of the semiconduc- This limits the use of the devices with fast optics.
torisselectedtosetupthethermalstandingwaves
in the structure with peaks at the front and nodes 3.2. Reduction of physical area of the detector
at the back surface. The quantum efficiency oscil-
lateswiththicknessofthestructure,withthepeaks A possible way to improve the performance of
atathicknesscorrespondingtoanoddmultipleof infrared photodetector is an increase in apparent
k=4n, where n is the refractive index of the semi- ‘‘optical’’ size of the detector in comparison with
conductor. The gain in quantum efficiency in- the actual physical size using a suitable concen-
creases with n. Higher gain can be obtained in the trator which compresses impinging infrared radi-
structures shown in Fig. 3(b). There are some ation [1,2,13–15]. This must be achieved without
additional gains achieved in particular types of reductionoftheacceptanceangle,oratleast,with
infrared detectors. Increased absorption makes limitedreductiontoanglesrequiredforfastoptics
hν
Semiconductor hν
(a)
Reflector n ,d
1 1
n ,d
2 2
n ,d
3 3
(c)
hν Semicondutor
n
1
Semiconductor
(b) n
2
Reflector
Fig. 3.Schematicstructureofinterferenceenhancedphotodetectors:(a)thesimpleststructure,(b)structureimmersedbetweentwo
dielectriclayerssuppliedwithbacksidereflector,and(c)structureimmersedbetweentwophotoniccrystals.
J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131 119
A
o
A
o
A
e A
e
(a) (b)
A
A o
e
Immersion
lens
Fig. 5.Schematiccross-sectionofanexperimentalmonolithic
Objective
optically immersed Dember detector (after Vigo System data
(c) sheets,Warsaw,1993).
Fig. 4.Operationofhemispherical(a),hyperhemispherical(b)
structure of a monolithic optically immersed
lens, and radiation running in objective/hemispherical lens
opticalsystem(c). Dember detector.
Monolithic optical immersion results in signifi-
cant improvement in detector parameters. The
of infrared systems. Various types of suitable gain factors achieved with hyperhemispherical
optical concentrators can be used, including opti- immersion are substantially higher compared to
cal cones and conical fibres. An efficient possible those for hemispherical immersion. The hyper-
way to achieve an effective concentration of radi- hemispherical immersion may restrict the accep-
ation is to immerse the photodetector to the tanceangleofthedetectorandrequiremoresevere
hemispherical or hyperhemispherical lenses (see manufacturing tolerances. These restrictions de-
Fig.4)[1,2].Theimmersionlensplaystheroleofa pend on the refraction coefficient of the lens. For
field lens which increases the field of view of the CdZnTe, however, they are not so severe as for
optical system. The hyperhemisphere used as an germanium lenses, and have no practical impor-
aplanaticlensresultsinapparentincreaseinlinear tance in many cases. For example, the numerical
detector size by a factor of n4. aperture of the main optical system is limited to
Several factors must be taken into account in about 1.4 by the CdZnTe immersion lens.
practical realization of the optically immersed
detectors. Use of immersion technology has been
limitedduetoproblemswithmechanicalmatching 4. Classical photon detectors for high temperature
of detector and lens materials as well as severe operation
transmission and reflection losses. Another limi-
tationisduetothelimitedacceptanceangleofthe The class of photon detectors is sub-divided
devicesasaresultofthetotalreflectionatthelens- into different types depending on how the electric
glue interface. or magnetic fields are developed (as shown in
The problems of matching the detectors to Table 1). Initial efforts to produce near room
immersion lenses have been solved by the use of temperature LWIR photon detectors have con-
monolithic technology developed at Vigo System centrated mainly on optimization of photocon-
[1]. The technology is based on epitaxy of ductors and PEM detectors [1–3].
HgCdZnTeonCdZnTesubstrate.TheHgCdZnTe
serves as the sensitive element, while the immer- 4.1. Photoconductors
sion lens is formed directly in the transparent
CdZnTe (n¼2:7) or GaAs (n¼3:4) substrates. Considerations carried out in Ref. [1] indicate
For example, Fig. 5 shows schematically the thatforoptimallydopedphotoconductoratahigh
1
2
0
Table 1
HgCdTeuncooledphotodetectors
Modeof Schematicof Advantages Disadvantages Performance(optically
operation detector immersed,k¼10:6lm)
Photoconductor • lowcosttechnology • biasrequired at300K:
Signal
RL • highresponsivity • D(cid:3) upto2·108 cmHz1=2/W
Incident radiation • responsetime 61ns J.
w P
Ohmic tcontact l •atD2(cid:3)20upKt:o3·109 cmHz1=2/W iotrow
• responsetime 610ns sk
i,
A
PEMdetector • nobiasrequired • bulky at300K: .
Incident radiaStioignnal z • noflickernoise • lowresistance • D(cid:3) upto1·108 cmHz1=2/W Ro
t - + w x • vtiemryeshortresponse • responsetime 610ns galsk
Ohmic contactMagnetilc field y i/I
n
fr
Demberdetector • nobiasrequired • lowresistance at300K: ar
Incident radiation • noflickernoise • requirelowresistance • D(cid:3) upto1.5·107 cmHz1=2/W ed
Contact contacts • responsetime 61ns Ph
y
s
ic
r1,s1 s
t r2,s2 &
T
e
c
l w Contact hn
o
lo
g
Photodiode • fastresponse • verylowdynamic at220K(multielementarray): y
Incident radiation Contact • noflickernoise resistance • D(cid:3) upto2·108 cmHz1=2/W 46
• lowquantumefficiency • responsetime 63ns (2
0
t p-type Depletion region • influenceofseries 04
n+-type resistance )1
1
Contact 5–
1
3
Stackedmulti- • highdynamicresistance • influenceofseries • generation-recombinationlimitof 1
heterojunction OP+-hHmgicC cdoTnetacts • goodquantumefficiency resistance performancecanbeachievedin
p-HgCdTe absorber
photodiode Substrate N-+HgCdTe • noflickernoise practiceatanywavelength
• extremelyfastresponse • practicalimplementationrequires
Incident radiation
• canbeusefromDCtoveryhigh wellestablishedepitaxialtechnology
frequencies
J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131 121
bias, the detectivity is limited by generation-
recombination noise and is equal to
k
D(cid:3) ¼0:45 y1=2; ð3Þ
hc
where
asðzþ1=zÞ
y ¼ : ð4Þ
n
i
In the above equations k is the wavelength of
radiation, h is the Planck’s constant, c is the light
velocity, z¼p=n, n is the intrinsic concentration,
i i
p is the hole concentration, a is the absorption
coefficient,andsisthecarrierlifetime.Atlowbias
when thermal noise prevails, the detectivity in- Fig. 6. Examples of uncooled and thermoelectrically cooled
creases with bias voltage. single element photodetectors manufactured by Vigo System,
Warsaw.
FromEq.(3)resultsadvantagetofabricatenear
room temperature photoconductors from p-type
materials, since absorption coefficient and recom-
bination time achieve their maximum value at designed to minimize parasitic impedances as the
z(cid:8)2. A further increase in doping results in de- ambienttemperaturephotoconductorsarecapable
creaseofy,mainlyduetotherapiddecreaseinthe of high frequency operation with a typical band-
Auger recombination lifetime. width of (cid:2)200 MHz, achieving (cid:2)1 GHz in high-
To achieve maximum performance, very good frequency optimized devices.
bias power dissipation isrequired,what isdifficult
to realize in practice. The highest measured detec- 4.2. PEM detectors
tivityofuncooled10.6lmphotoconductorsathigh
frequencyisabout6·107 cmHz1=2W(cid:1)1,thisvalue The PEM effect has been traditionally used for
isafactorof(cid:2)2.5below thetheoretical limit. The roomtemperaturedetectorsinthemiddle-andfar-
commerciallyavailabledeviceshavedetectivitiesof infraredband.CoolingofthePEMdetectorsposes
about1·107cmHz1=2W(cid:1)1(seewww.vigo.com.pl). problems associated with the use of a magnet and
A moderate cooling, which can be conveniently results in little improvement in device perfor-
done with one or several stage thermoelectrical mance. Advantages of the PEM detector are lack
coolers, can improve the performance. The ofbias andlow resistance,enablingtheuseoffast
improvement in performance is, however, accom- electronics.
paniedbyanincreaseinaresponsetime.Thepeak For a long time InSb has been the most com-
wavelength of 10.6 lm optimized device is shorter mon material for photoelectromagnetic detectors.
than 10.6 lm for which the device was optimized. TheirpropertieshavebeenreviewedbyKruse[16].
The shift increases with decreasing thickness, and InSbwithmaximumperformanceatabout6.5lm
in commercial devices with low bias power, the exhibit no response in the 8–14 lm atmospheric
peak responsivity is at6–7 lm. window and relatively modest performance in the
Typical structure of uncooled photoconductor 3–5lm window, however.Hg Cd Te (HgCdTe)
1(cid:1)x x
issimilartocooleddevice.Theoptimumthickness and closely related Hg Zn Te (HgZnTe) and
1(cid:1)x x
of the active elements (a few microns) depends Hg Mn Te (HgMnTe) alloys made possible to
1(cid:1)x x
upon the temperature of operation and is smaller optimize performance of PEM detectors at any
in uncooled devices. The sensitive elements are specific wavelength [2,3].
housedinvariouspackages,theexamplesofwhich The PEM effect is caused by diffusion of
are shown in Fig. 6. The packages are usually photogenerated carriers due to the photo-induced
122 J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131
carrierconcentrationgradientandbydeflectionof
electron and hole trajectories in opposite direc-
tions by the magnetic field (see Table 1). If the
sample ends are open-circuit in the x -direction, a
space charge builds up giving rise to an electric
field along the x axis (open-circuit voltage). If the
sample ends are short-circuited in the x-direction,
acurrentflowsthroughtheshortingcircuit(short-
circuit current).
Theoretical analysis indicates that the maxi-
mum voltage responsivities of PEM detectors can
Fig. 7.Schematicofthehousingofathermoelectricallycooled,
be reached in relatively strong magnetic fields optically immersed PEM detector (after Vigo System data
(B(cid:8)1=l ). For materials such as Hg Cd Te, sheets,Warsaw,1993).
e 1(cid:1)x x
where l =l (cid:9)1, the resistance of the detector
e h
reaches its maximum value at the point
p=n (cid:8)ðl =l Þ1=2 andthehighestvalueforvoltage active elements are mounted in a housing which
i e h
responsivity is reached for lightly doped p-type incorporates a miniature two-element permanent
material. The acceptor concentration in narrow magnet and pole pieces. Magnetic fields
gap semiconductors is adjusted to a level of about approaching 2 T are achievable with the use of
(2–3)·1017 cm(cid:1)3 [1–3]. At room temperature, the modern rare-earth magnetic materials for the
ambipolar diffusion length in narrow gap semi- permanent magnet and cobalt steel for the pole
conductors is small (several lm) while the pieces.
absorptionofradiationisrelativelyweak(1/a(cid:8)10 The best uncooled HgCdTe 10.6 lm PEM
lm). In such cases the radiation is almost uni- detectors exhibited measured voltage responsivity
formlyabsorbedwithinthediffusionlength.Thus, ofabout0.1V/W(widthof1mm)anddetectivities
a low recombination velocity at the front surface of about1·107 cmHz1=2W(cid:1)1 which is by a factor
and a high recombination velocity at the back of (cid:2)3 below the predicted ultimate value.
surface are necessary for a good PEM detector Both the theoretical and the measured perfor-
response [3].Insuchdevices, thepolarityofsignal mance of PEM detectors are inferior to that of
reverses with a change of illumination direction photoconductive detectors. PEM detectors have,
from the low to the high recombination velocity however,additionaladvantages,whichmakethem
surface, while the responsivity remains almost useful in many applications. In contrast to
unchanged. photoconductors, they do not require electrical
The preparation of PEM detectors is similar to biasing. The frequency characteristics of a PEM
that of photoconductive ones, with the exception detector is flat over a wide frequency range,
of the back surface preparation which in PEM starting from constant current. This is due to lack
devices is subjected to a special mechanochemical of low frequency noise and a very short response
treatment to achieve a high recombination time. The response time, which is determined by
velocity. This procedure is not mandatory when effective recombination time is less than the bulk
using graded gap structures. The wafers are recombinationtime.Asshortresponsetimes,as50
thinned to final thickness of about 5 lm and cut ps, are easily achievable in devices optimized for
into bars sized (cid:2)1·2 mm2. Electrical contacts high frequency operation, with little expense in
are usually made by gold electroplating to the responsivity. The resistances of PEM detectors do
ends of the bars, to which gold wires are then not decrease with increasing size what makes
attached. possible to achieve potential performance in large
Fig. 7 shows schematically the housings of size devices. With the resistance typically close to
PEMdetectors,whicharebasedonstandardTO-5 50X,thedevicesareconvenientlycoupleddirectly
or, for larger elements, TO-8 transistor cans. The to wideband amplifiers.
J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131 123
Magnetoconcentration effect detector is a rela- Dember detector is not biased and the noise volt-
tivelynewdevice[3,7,8]thatcanberelatedtonon- ageisdeterminedbytheJohnson–Nyquistthermal
equilibrium mode devices. The geometry of the noise.AlsotheresponsetimeofDemberdetectors
device is the same as that of conventional PEM isdeterminedbyprocessesanalogoustothoseused
detectors, but the device is biased with an electric for PEM detectors.
field perpendicular to the magnetic field and ori- The calculated detectivity of 200–300 K Dem-
ented in such a way that the Lorentz force, acting ber detectors is higher compared to that of
on the current carriers, directs them tothe surface photoconductors operated under the same condi-
of high surface recombination velocity. In effect, tions. However, very low resistances, low voltage
the concentration of electrons and holes in the responsivities and noise voltages well below the
bulk of the material is diminished and thus the noise level of the best amplifiers pose serious
regionsbecomedepleted.Thisleadstosuppression problems in achieving the potential performance.
of thermal generation and recombination pro- For example, uncooled 10.6 lm devices with size
cesses and in consequence in reduction of ther- 50· 50 lm2 will have a resistance of (cid:8)0.15 X and
mally generated noise. However, the high bias noise voltage of (cid:8)50 pV/Hz1=2. This explains why
current requirements and significant 1=fnoise in the conventional Dember detectors cannot com-
this type of detector have been severe obstacles to pete with other types of photodetectors [3].
their widespread applications. Therearesomepotentialwaystoovercomethis
difficulty. One is the connection of small area
4.3. Dember effect detectors Demberdetectorsinseries.Forexample,toobtain
reasonableresistance((cid:8)50X)ina1mm2 totalarea
Dember effect detectors are a type of photo- sensitiveelementofanuncooled10.6lmdetector,
voltaic devices, based on bulk photodiffusion it is necessary to divide its area into 400 50·50
voltageinasimplestructurewithonlyonetypeof lm2 elements, connected in series with low resis-
semiconductor doping. The use of the Dember tance interconnections, which is feasible with
effect for high temperature long-wavelength modern fabrication.
detectors has been suggested by Piotrowski et al. The other possibility is to use optical immer-
[1,17]. When radiation is incident on the surface sion. Detectivities as high as (cid:8)1.7·109 and
of a semiconductor, a potential difference is usu- 1.5·1010 cmHz1=2/W are expected at 300 and 200
ally developed in the direction of the radiation Kinoptimizedopticallyimmersed10.6lmdevices
(see Table 1). This is known as photodiffusion or [17].
the Dember effect. Two conditions are required Due to ambipolar diffusion of excess carriers
for generation of the photovoltage: the distribu- the response time of a Dember detector is shorter
tion of photogenerated carriers should be non- thanthebulk recombination timeand can bevery
uniform and the diffusion coefficients of electrons shortfordeviceswiththicknessmuchlessthanthe
and holes must be different. The gradient usually diffusion length [1]. The response time of opti-
results from non-uniform optical generation and mizeddevices isabout1ns, shorteningwithp-type
from different surface recombination velocities at doping, with some expense in performance. The
the front and back surfaces of the device. The high frequency optimized Dember detectors
Dember effect electrical field restrains the elec- should be prepared from heavily doped material
trons with higher mobility, while holes are accel- and the thickness should be kept low. For exam-
erated, thus making both fluxes equal. The best ple, for detectors with N ¼5(cid:10)1017 cm(cid:1)3 and
a
performance is achievable for a device thickness t¼2 lm, s is (cid:8)50 ps. Another possible limitation
of the order of a diffusion length, low recombi- of response time, the RC constant, is not impor-
nation velocity at the front and high at the back tantduetoalowresistivityandlowcapacityofthe
surface. Dember detector.
As in the case of the PEM detector, the best Manufacturing of Dember detectors from
performance is achieved in p-type material. The Hg-based narrow gap semiconductors requires
124 J.Piotrowski,A.Rogalski/InfraredPhysics&Technology46(2004)115–131
well-established technology of devices from bulk The voltage responsivity of a homojunction
crystalsorepilayers.Fig.5showsaschematicofa photodiode with photoelectrical gain of unity is
monolithic optically immersed uncooled Dember
gk
effect detector manufactured by VIGO. The sen-
R ¼ R ; ð5Þ
sitive element has been prepared from an v hc d
Hg Cd Te epitaxial layer grown with the hy-
1(cid:1)x x where R ¼dI=dV is the diode differential resis-
perhemispherical lens formed directly in the d
tance. Assuming negligible effect of background
transparent CdZnTe substrate. The surface of the
radiation, the zero bias noise is determined by the
sensitive element was subjected to a mechano-
Johnson–Nyquist noise and detectivity of zero
chemical treatment to produce a high recombina-
biased photodiode is
tionvelocityandcovered withreflectinggold.The
‘‘electrical’’ size of the sensitive element is 7·7 gk (cid:6)R A(cid:7)1=2
lm2 while, due to the hyperhemispherical immer- D(cid:3) ¼ 0 : ð6Þ
2hc kT
sion, the apparent optical size is increased to
(cid:8)50·50 lm2. The measured voltage responsivity As Eq. (6) shows,for thebestperformance, the
was (cid:8)100 V/W, which is about one order of value gðR AÞ1=2 should be optimized for given
0
magnitude below the calculated value for an wavelengthandtemperature.Simpleconsideration
optimized device. indicatethatduetohighmobilityofelectrons,the
Large active area devices (>100 lm square) are nþ-p structure photodiodes offer the best perfor-
produced by connections of small element in ser- mance (better R A product and quantum effi-
0
ies. Although they are less sensitive, they can ciency).
successfully compete with photoconductors in The photovoltaic device of conventional design
applications requiring very low frequency or/and suffers from
very high frequency operation.
Recently,thevolumeDembereffectiscombined • poor quantum efficiency, and
with photovoltaic effect at heterojunctions result- • low differential resistance.
ing in better response.
Onlychargecarriersthatarephotogeneratedat
4.4. Photodiodes distance shorter than the diffusion length from
junctioncanbecollected.Theabsorptiondepthof
As early as in 1975, Koehler and Chiang [18] long wavelength IR radiation (k>5 lm) islonger
reported nþ-p 10.6 lm photodiodes operated at thanthediffusionlength.Therefore,onlyalimited
170 K. Shanley et al. [19] showed a capability of fraction of the photogenerated charge contributes
achieving bandwidths of 475–725 MHz, quantum tothequantumefficiency.Consideranexampleof
efficiency of 78%, an R A product of 0.006 Xcm2 an uncooled 10.6 lm photodiode. Calculations
0
and fairly good heterodyne sensitivity of 1.3·1019 show that the ambipolar diffusion length is less
W/Hz at a temperature of 145 K. Extension of than 2 lm while the absorption depth is (cid:8)13 lm.
operational temperature to 185–200 K was sug- This reduces the quantum efficiency to (cid:8)15% for
gested with a proper adjustment of bandgap and single pass of radiation through the detector.
doping level. Gordon [13] discussed possible The resistance of the p–n junction is very low
operation of a 9.5 lm photodiode array using duetoahighthermalgeneration.Inmaterialswith
optical immersion with microlenses at tempera- a high electron to hole mobility ration, the resis-
tures 150–245 K. The practical 128-element array tance is additionally reduced by ambipolar effects
based on the loophole technology with silicon [8].Asaresult,thepreamplifiernoiseandnoiseof
microlenses has been reported by Jones et al. parasitic resistances may exceed the thermal gen-
[14]. At 192 K and wavelength of 9 lm eration-recombination noise. As a result, the per-
detectivities of 4·109 cmHz1=2/W have been formance of conventional devices is very poor, so
achieved. they are not usable for practical applications.