Optical efficiency of far-infrared photoconductors
J.-Q. Wang, P. L. Richards, J. W. Beeman, N. M. Haegel, and E. E. Haller
We report an experimental and theoretical study to optimize the geometry of far-IR photoconductive
detectors with diffraction-limited throughput. Factors considered in this optimization include internal
optical path relative to measured absorption length, photoconductive gain, uniformity of illumination, cosmic
ray cross section, and compatibility of the design with the requirements of 1- and 2-D arrays. A rod-shaped
detector geometry with square cross section, electrodes on the lateral faces, and a beveled backface to trap the
radiation by total internal reflection was found to have nearly equal responsivity to the best detectors in
integrating cavities.
1. Introduction
Photoconductive detectors made from germaniumdoped with acceptors such as gallium and berylliumare used when sensitive far-IR measurements aremade over the wavelength range from 30 to 240 im.
The most stringent requirements on detector sensitiv-ity are encountered in connection with space astrono-my experiements which employ cooled optics. Pro-jects such as the proposed NASA Space InfraredTelescope Facility (SIRTF) require both 1- and 2-Darrays of carefully optimized extrinsic Ge photocon-ductors with diffraction-limited throughput.
An overall optimization of all parameters of suchphotoconductive detectors is a very complicated task.The detector size, shape, materials parameters, oper-ating temperature, and bias must be selected to opti-mize the required combination of responsivity, detec-tive quantum efficiency, dark current, and cosmic raycross section. In this paper we restrict our attention toan optimization of the size and shape of these detec-tors.
To proceed with this more limited goal, we mustidentify the ways in which the important figures ofmerit depend on detector dimensions. We will showthat the optimization of detector size and shape can becarried out independently of other material and oper-ating parameters once the required throughput andabsorption length for IR in the material are known.
All authors are with University of California, Berkeley, California
94720; J.-Q. Wang and P. L. Richards are in the Physics Department,
the other authors are in the Department of Materials Science.Received 6 August 1986.0003-6935/86/224127-08$02.00/0.
To minimize limitations due to amplifier noise atvery low illumination it is necessary to maximize theresponsivity, which is the ratio of photocurrent I to theincident photon rate N. The responsivity can be writ-ten as a product of the electronic charge e, the photo-conductive gain g, and the responsive quantum effi-ciency ?1R:
I/N= egnR. (1)
The photoconductive gain is defined as the ratio ofthe carrier lifetime to the transit time between theelectrodes which are separated by the distance d. Thedependence of g on detector dimensions arises fromthe transit time. If parameters such as doping, tem-perature, and bias field are kept fixed, the simplest(uniform field) models of detector operation predictthat gain varies as d- 1 . More realistic models, whichpredict a nonuniform field distribution in the directionof current flow,1 also suggest that g will increase withdecreasing d.
The responsive quantum efficiency qR = Eoec can bewritten as the product of the optical efficiency e0 withwhich incident photons are absorbed and the efficien-cy Cc with which absorbed photons liberate mobilecarriers. The optical absorption efficiency Ec dependson detector dimensions in a straightforward way. Itcan be made large by minimizing reflection loss at theentrance to the detector and by selecting detector di-mensions so that the optical path inside the detector islonger than the absorption length. The efficiency ofgeneration of free carriers depends only on materialparameters such as the cross section for the generationof free carriers compared with the total cross sectionwhich also includes contributions from the excitationof electrons into bound states and from the generationof phonons. These parameters influence the selectionof the best acceptor density and thus help to determinethe absorption length but do not otherwise affect ouroptimization.
In many detector applications, amplifier noise canbe reduced below photon noise and intrinsic detectornoise, so the detective quantum efficiency nD becomesan important figure of merit. The noise equivalentphoton rate in a photoconductive detector can be writ-ten in terms of this parameter:
NET = (4N/,&D)"2 - (2)
The detective quantum efficiency 7D is equal to theresponsive quantum efficiency f7R if there are no noisemechanisms in addition to photogeneration of carriersand the resulting recombination. If there are addi-tional noise mechanisms, E7D < 1R- We have alreadydiscussed how R enters our optimization. Detectordimensions must be selected to minimize excess noisemechanisms. Mechanisms to be considered includepossible contributions to the dark current due to sur-face currents and noise phenomena which arise fromnonuniform illumination2 or nonuniform electric fielddistributions transverse to the direction of currentflow. 2
In space experiments, cosmic rays can interfere sig-nificantly with the desired performance of photocon-ductive detectors. High energy photons and chargedparticles create electron-hole pairs in detector materi-al which cause both noise and an enhancement indetector responsivity.3 -5 Detector size should be min-imized to avoid these effects.
Several other factors also favor small detector size.Decreasing the interelectrode distance increases thephotoconductive gain. Uniform illumination is moreeasily achieved in small detectors. However, detectorsize can only be reduced subject to the need to acceptand absorb the incident photons. Also, there can bepractical limitations to the smallest detectors that canbe fabricated.
The acceptance area of the detector is determinedby the required throughput and the focal ratio of thefeed optics. For projects such as the multichannelimaging photometer for SIRTF, the angular diameterof the pixels is -X/D (X/2D when the focal plane mustbe fully sampled to obtain superresolution). For afocal ratio of f/13, for example, pixels for X = 100 mhave diameters of 300 ym (or 150 gm). Focal ratiosshould be chosen large to minimize detector dimen-sions, subject to the constraints set by optical aberra-tions and the ability to fabricate small detectors.
In addition to the pixel diameter, optimum detectorgeometry depends on the absorption length for infra-red in the detector material. The dominant absorp-tion mechanism is proportional to the product of theabsorption cross section and the density of the major-ity impurities. The majority impurity density shouldbe large subject to the constraint that dark currentmust be avoided. In Ge:Ga, for example, an acceptorconcentration of NA 2 X 1014 cm- 3 is safely below thethreshold for a significant dark current due to hoppingconduction.6 7 The possibility of other dark currentmechanisms and their dependence on NA has not beenadequately explored. As we will show, this acceptorconcentration yields an absorption length of mm at
100 cm-', which is the peak of the photoconductiveresponse of Ge:Ga in zero stress. Optical paths of thisorder in the detector are required to obtain efficientabsorption. The optimum absorption length is notwell known for Ge:Be or stressed Ge:Ga detectors butcould be significantly different.
The responsivity of a photoconductive detector de-pends on the bias voltage. The optimum bias voltageis in turn sensitive to the acceptor concentration NA,the compensation ratio K,"6 7 and the amount ofbreakdown noise that can be tolerated.' This compli-cated behavior does not influence the optimization ofthe size and shape of detectors made from a givenmaterial, except through nonlinearities in the depen-dence of photoconductive gain and breakdown voltageon the interelectrode distance.' In principle, thetrade-off between optical efficiency and photoconduc-tive gain could lead to different optimum interelec-trode distances for low background applications whereamplifier noise is important and high background ap-plications where amplifier noise is not important. Inpractice, these dependences are weak; so our optimiza-tion experiments are carried out for a single interelec-trode distance to avoid confusion in selecting compara-ble bias field conditions for detectors with differentseparations.
We conclude from this analysis that if we are giventhe required pixel diameter and the optimum absorp-tion length, we can find an optimum shape and size forfar-IR photoconductive detectors without reference toother material or operating parameters. In this paperwe explore this optimization for 0.5- X 0.5-mm2 pixelsand an absorption length of 5 mm. Our results can bescaled to other values of these parameters. To avoidchaotic phenomena,", 2 we choose to consider only ge-ometries for which the photocurrent is uniformly dis-tributed. That is, we require that the detector materi-al be nearly uniformly illuminated and that thedetector cross section parallel to the electrodes be con-stant.
Detectors made from materials such as extrinsic Si,for which the absorption length is comparable to or lessthan the pixel diameter, often make use of IR illumina-tion through a transparent front electrode. Despitemajor advantages for the fabrication of arrays, thisgeometry does not appear competitive for materialswith long absorption lengths because of the large sacri-fice in photoconductive gain.
We have compared two generic detector geometries.One is a flat rectangular detector in a metallic cavitywith electrodes on the broad faces, such as the oneshown in Fig. 1(a). The other is a rod-shaped detectorwith square cross section which has electrodes on twoof the lateral faces, such as is shown in Fig. 1(b). Thisdetector is illuminated on one end; the other end isbeveled so as to trap the radiation by total internalreflection. We use the term end-fire to describe thisdetector. Both detector designs are well known to thefar-IR detector community, but there is no consensusas to which is preferable. We have not been able toidentify the origins of either of these designs.
.1. ~Ge:GaIIIAII~i~ii~~iii.~~ large (100) stress6 21i|_ ~~~~d =-1.45 mm
I I - .I
t~~~~~~~8 mO
r 20 40 60Frequency (cm-')
Fig. 2. Far-infrared transmittance at 1.2 K of 1.45 mm of Ge:Ga
material 1 with a large uniaxial stress along a (100) axis.
(b)
Fig. 1. Two generic types of far-IR detector compared experimen-
tally: (a) cross section of 2- X 2- X 0.5-mm detector in a metal cavity.
(b) Perspective view of 2.5- X 0.5- X 0.5-mm end-fire detector.
Table I. Acceptor Concentration and Compensation of Detector Materials
NASample Material (cm-3) K
(1) LBL 82-head Ge:Ga 1.2 X 1014 -10-2(2) LBL 108-14.6 Ge:Ga 1.8 X 1014 ,40-2
(3) LBL 728-4.9 Ge:Be 7.5 X 1014 -
E
x
0
.L
.2
D
3aE
II. Measurements of Absorption Length
The absorption length, or its reciprocal, the frequen-cy-dependent absorption coefficient a(v), is an impor-tant parameter for detector material. We have usedmeasurements of far-IR transmittance to determinethe absorption coefficients in Ge:Ga with a large (100)stress, unstressed Ge:Ga, and Ge:Be detector materi-als. In the stressed sample the direction of propaga-tion of unpolarized radiation was perpendicular to theaxis of the stress. Since the dominant contribution tocx(V) in the frequency range of interest is proportionalto the acceptor concentration NA, it is different indifferent detector materials. In Table I we list thevalues of NA and compensation ratio K for our samplesas determined from Hall effect measurements as afunction of temperature.
Our measurements of a (v) can be easily scaled to anyavailable detector or detector material by measuringthe room temperature conductivity to determine NA.
For Ge:Ga we have the relation NA(cm- 3 ) = 2.8 X 10'5a300(0 cm)-'. The origin of this convenient relation
can be understood by expressing the conductivity asthe product of the hole mobility g, the electroniccharge e, and the hole density p, ar = gep. The mobil-ity of free carriers in Ge is limited by phonons at 300 Kin the useful range of concentration, so is independentof NA. Since the impurity states are ionized at 300 K,but excitation across the band gap can be neglected, p
Frequency (cm-1)
Fig. 3. Ratio of frequency-dependent absorption coefficients a(v)
to acceptor concentration NA for (a) Ge:Ga sample 1 with infrared
propagating perpendicular to a large (100) uniaxial stress, (b) Ge:Ga
sample 1 with no stress, and (c) Ge:Be sample 3. Typical estimatederrors are shown. In the region covered by dashed lines, the absorp-
tion is due primarily to transitions to bound states. The solid linesshow absorption due primarily to the generation of free carriers.
The application of a large stress in the (100) direction increases thepeak value of a by a factor 3.
= NA. Thus we have NA = a/ge. The same argumentcan be used for Ge:Be, except that the double acceptorBe yields two mobile holes at 300 K. As a result, NA =
1.4 X 10+15 0-300 for Ge:Be.Infrared absorption coefficients were determined
from measurements of normalized sample transmit-tance made at 1.2 K with a Fourier transform spec-trometer and bolometric detector. An example of thetransmittance of Ge:Ga (with uniaxial stress) is shownin Fig. 2. Interference fringes which arise from paral-lel sample faces are clearly visible. The transmittancescale contains uncertainties due to the partial satura-tion of the bolometer when the sample was removed toobtain a reference spectrum. Changes in the opticalpath when a high-index medium was inserted or re-moved caused additional uncertainties. The most
precise value of the scale factor was obtained as fol-lows: The index of refraction n = (2vd)'- was mea-sured from the fringe separation v and sample thick-ness d. The measured value n = 4.0 is very close to thezero frequency value for pure Ge and is independent offrequency, impurity content, and stress. The dielec-tric reflectance R = [(n - 1)/(n + 1)]2 = 0.36 was thencomputed. The experimental value of the transmit-tance averaged over the interference fringes at lowfrequencies, where absorption can be neglected, wasthen adjusted to equal the theoretical value T = (1 -R)/(1 + R) = 0.47.
The frequency-dependent absorption coefficientcx(v) was computed from the measured T(v), averagedover interference fringes, and the computed R from theequation
This equation can be derived by adding the intensitiesof multiple reflections or alternatively by averagingthe full expression8 over interference fringes in thelimit k/n << 1. In this limit, R is independent offrequency, even in the lossy region above threshold.
The measurements of a(v) shown in Fig. 3 werededuced from transmittance measurements on thesamples listed in Table I. Relatively thin d = 1.4-mmsamples were used to show clearly the interferencefringes as in Fig. 2. Thicker samples were also used toemphasize the absorption. In each plot, a()/NA isgiven as a dashed line over the frequency range inwhich the absorption is dominated by transitions be-tween bound states which do not cause a photocurrentand by a solid line above the threshold for photocon-ductivity where the cross section is dominated by tran-sitions from the acceptor ground state to the band.
The transition region between these two limits isquite complicated and depends on NA through thebroadening of bound state levels by interactions be-tween acceptor centers. Measured values of a/NA inunstressed Ge:Ga for NA = 1.2 and 1.8 X 10'4 cm- 3 arethe same above 110 cm-' but vary by 15% at 90 cm-'.The peak in the photoresponse occurs at 100 cm-'.The useful frequency range for Ge:Be detectors ex-tends above 300 cm-' where phonon absorption (thatis not proportional to NA) becomes important. 9
The large change in (v) with stress deserves somecomment. A uniaxial stress applied along the (100)axis of Ge lifts the degeneracy at the center of thevalence band and reduces the binding energy of shal-low acceptors such as Ga by a factor' 0 "'1 approaching 2.The utility of the stressed Ge detector arises from theresulting shift of the photoresponse to lower frequen-cies. The shallow acceptor state for the nondegener-ate band is analogous to the hydrogen atom, and thecross section falls rapidly with frequency above thethreshold.' 2 The cross section of the acceptor state forthe degenerate band of unstressed Ge:Ga evidentlyvaries more slowly with frequency.
The increase in a with stress arises from an increasein the absorption cross section which can be estimated
from an increase in the area of the effective Bohrorbits. In zero stress the average value of Irl weightedby the probability density in the ground state 3 is a =75 A. In a large (100) stress the wave function spreadsand elongates in the direction of the stress' 4 so that a"= 129 A and a' = 93 A. This picture leads us to expectan increase in a by a factor of -2 for light propagatingperpendicular to the stress axis, which is comparableto the observed effect.
Our measurements give the relationship between aand NA. They do not tell us the optimum extinctionlength unless the optimum value of NA is known. Un-stressed Ge:Ga detectors typically have NA _ 2 X 10'4cm 3 . Since the hopping contribution to the darkcurrent depends on the wave function overlap in thedirection of current flow, we can anticipate that theoptimum NA will be somewhat smaller for stressedGe:Ga detectors. We have used Ge:Ga with NA = 1.2X 10"4 cm-3 for stressed detectors, but the dark currenthas not yet been measured. Successful Ge:Be detec-tors with very low dark current' 5 have been made usingBe concentrations >7.5 X 10'4 cm- 3 .
111. Design of Test Detectors
To determine the detector design with the best opti-cal efficiency, it is necessary to compare optimizedversions of each competing detector type. Since dif-fraction effects limit the validity of theoretical cal-culations, we must rely on a limited number of experi-mental comparisons. Comparisons between the re-sponsivities of different detectors are meaningful indi-cators of optical efficiency only if the detectors aremade from the same materials, have the same inter-electrode distance and the same kind of contacts, andare operated at the same temperature and bias. Toachieve these conditions, we use detectors with a singleinterelectrode separation d = 0.5 mm, which were cutfrom a single wafer of Ge:Ga whose acceptor concen-tration is very homogeneous, as determined from mea-surements of the room temperature conductivity.The entire wafer was implanted with B+ ions, metal-lized, and annealed as described in Ref. 6 to insurecontact uniformity. A portion of the wafer was leftunmetallized to make a detector with transparent elec-trodes.
Two generic detector geometries are commonly usedwhen the extinction length is an order of magnitudelonger than the required pixel diameter. In one ap-proach the detector is located in a metallic integratingcavity whose entrance aperture is determined by therequired pixel diameter. The idea is that photonswhich are not absorbed by the detector on the first passare given further opportunities after reflection fromthe cavity walls. The use of a cavity is thought toprovide good absorption efficiency in a smaller detec-tor with more uniform illumination. The perfor-mance of such cavities can be analyzed quantitativelyin the geometrical optics approximation by computerray tracing or by a statistical approach' 6 which as-sumes a random distribution of photons in the cavity.Accurate predictions are not possible with either ap-
proach for diffraction-limited pixel diameters, howev-er, because diffraction effects are important. In prac-tice, many favorite cavity designs exist, and there islittle evidence to support preferences between them.
Figure 1(a) shows a 0.5- X 2- X 2-mm detector withelectrodes on the large faces located in a brass cavitywhich has axial symmetry about the detector post.The cavity is designed to avoid parallel faces and tokeep the area of the metal walls small. Experimentsshowed that the best response was obtained with thevertical position of the detector adjusted to be in linewith the entrance aperture and the detector rotated sothat reflections from its front face could not escapedirectly out of the aperture. A smaller 0.5- X 1- X 1-mm detector was installed in a cavity with similarproportions but smaller diameter.
The large cavity detector with metallized electrodessuffers from the fact that only a third of the detectorsurface area can accept photons. Truly transparentelectrodes should improve the optical efficiency or re-duce the required detector dimensions. Semitrans-parent electrodes made by B+-ion implantation with-out metallization absorb17 significantly more thancavity walls or metallic electrodes. Measurementswere made of the IR transmittance of a 0.5-mm waferof sample 2 which was implanted on both sides andannealed according to the prescription in Ref. 6 but notmetallized. The measured transmittance was fitted toa theory' 8 of the transmittance of a dielectric slab withthe known value of n and a(v) and with thin conductingfilms on both sides. The fit gave a sheet conductancefor each implanted layer of 360 W/. The IR transmit-tance of a single surface is 43%. The absorptance for aray passing from vacuum to detector is 11% and for aray passing from detector to vacuum is 45%. A large(2- X 2-mm) cavity detector was made from this mate-rial.
The second generic detector is rod shaped with asquare cross section as shown in Fig. 1(b). The trans-verse dimensions are set by the pixel size, and thelength is chosen to obtain efficient absorption. Tokeep the interelectrode distance small, the electrodesare located on lateral faces. The loss of photons fromthe sides of such detectors is small because of totalinternal reflection from the open Ge sides and the highreflectivity of the metallic electrodes. Losses in inter-nal reflection from an implanted and metallized con-tact are expected to be much smaller than for the sameimplanted layer in a transparent contact because themetal boundary creates a null in the parallel infrared Efield close to the location of the implanted layer. Forgrazing incidence the loss is zero for a E field perpen-dicular to a conducting sheet or surface.
The detector length required to absorb efficientlycan be reduced by cutting the back face at an anglen 20° as shown in Fig. 1(b) and Fig. 4. Rays propagat-ing parallel to the optic axis are then totally reflectedat the back face. As shown in Fig. 4 these rays make atotal of either three or four passes along the detectorbefore radiation can escape. The detailed dependenceof the average number of passes on detector length is
(a)
(b)
Fig. 4. Two different optical paths through an end-fire detectorshowing (a) three and (b) four passes before the beam can escape.
Table II. Results of Detector Comparisons
OpticalDimensions Pixel Area Responsivity Efficiencyd
Detector (mm3 ) (mm2 ) (A/W)
Cav-1a 2 x 2 x 0.48 0.24 9.3Cav-2b 1.94 X 1.95 X 0.48 0.24 6.4Cav-3b 0.92 X 0.96 X 0.49 0.25 4.5Ef-1 2.39 X 0.43 x 0.48 0.21 5.8 (8.7)c 0.59 (0.89)Ef-2 0.91 X 0.43 X 0.48 0.21 3.8 (5.0)c 0.43 (0.57)Rect-1 0.87 X 0.44 X 0.48 0.21 2.1 (3.0)c 0.16 (0.23)Cube-i 0.48 X 0.48 X 0.48 0.23 1.5 (2.1)c 0.10 (0.14)
a Semitransparent electrodes.b Metallized electrodes.c (Estimated responsivity with an
front surface.)d Calculated from Eqs. (4) and (5).
antireflection coating on the
Fig. 5. Cross section of the test apparatus used to compare opticalresponsivities of far-IR detectors: a, light pipe; b, filter; c, roofmirror; d, apertures; e, end-fire photoconductive detector; f, black-
quite complicated. To a good approximation, howev-er, the average optical path inside the detector with a20° bevel is 4.1 X the average detector length.
Because the index of refraction n = 4 of Ge is quitelarge, this analysis remains valid for significant anglesof incidence. Infrared incident with a focal ratio f/3outside the detector, for example, forms an internalcone of rays with a half-angle of only 2.40 about thedetector axis. Since the cutoff angle for total internalreflection is 14.50, all such rays are totally reflected atthe 200 back face.
The optical efficiency e of an end-fire detector ofaverage length I and absorption coefficient a(v) can beestimated from the approxmate equation
c = (1 -R)1 - exp[-ar(v)L]j/$ - R exp[-a(v)L]j, (4)
where L = 4.11. The same equation is exact for adetector with a perpendicular backface, except that L= . Again, because of the large index for Ge, thereflection loss R = 0.36 from the detector surface isimportant. Antireflection coatings can be used to re-duce this loss without any change in the cutoff anglefor total internal reflection at the front face.
When an ideal antireflection coating is used on thefront surface of the detector, Eq. (4) can be used forend-fire detectors if we set R = 0. When detectorswith a perpendicular back face are antireflection coat-ed on the front face, the optical efficiency is given by
E = [1- exp(-aL)] [1 + R exp(-aL)]. (5)
IV. Detector Fabrication
Test detectors were prepared from a 0.5-mm thickslice of sample 2 whose properties are listed in Table I.The wafer was lightly etched, ion-implanted, gold-metallized, and annealed on both sides using the pre-scription in Ref. 6. The interelectrode separation was0.48 mm. Square samples with nominal dimensions of1 X 1 and 2 X 2 mm were cut from this wafer for cavitydetectors. A 2 X 2-mm cavity detector was also cutfrom a section of the wafer that was implanted, but notmetallized, so as to give semitransparent electrodes.Rods 0.5 mm wide were cut from the metallized materi-al. These rods were cut to 0.5 and 1-mm lengths withperpendicular end faces, and both 1- and 2.5-mm aver-age lengths with back faces beveled at 200.
To avoid excessive handling of detectors after etch-ing, the actual detector sizes were determined opticallybefore etching, and the size reduction during etchingwas estimated from witness samples. Values oflength, width, and interelectrode separation are listedin Table II. Each detector was mounted on a 0.34-mmdiam gold plated steel rod using Epo-tek H20E con-ducting epoxy19 for the metallized detectors and Insolder for the unmetallized detector. An 0.13-mmcopper lead was soldered with In to the other detectorcontact.
V. Experimental Detector Comparisons
The test apparatus shown in Fig. 5 was designed toprovide accurate comparisons of the spectral responsi-
vity of pairs of detectors. The f/11.5 IR beam from aFourier transform far-IR spectrometer is chopped at10 Hz and enters the apparatus through a metal lightpipe labeled a, which contains a black polyethylene lowpass filter and a 1% NiCr neutral-density filter at b.These filters were selected to give a photon rate of 5 X10'0 s-1 from 100 to 300 cm-' for an optical throughputof AR = 10-4 cm2 . The metal roof mirror c divides thebeam symmetrically, reflecting it horizontally to leftand right. Two apertures labeled d are separated by acavity coated with Ames 24E20 optical black. The firstaperture, which has a diameter of 1.7 mm and is 7.1 mmin front of the entrance to the detector, limits the fieldof view at the detector to f/4.2 or 0.045 sr. With pixeldimensions of 0.25 mm2 this gives AO = 2 at X = 106,gm. The second blackened aperture serves to reducestray light. In the case of the cavity detectors, the holein the cavity serves to define the detector acceptancearea. It is located immediately behind the secondblackened aperture. When endfire detectors weremeasured, a metal foil with a square hole etched in itwas used immediately in front of the detector to definethe acceptance area. Endfire detectors measuredwithout this aperture gave -35% larger signals becauseof absorption on the sides of the detector. This ab-sorption is quite efficient because the angles of inci-dence approach the Brewster angle. Figure 5 shows anendfire detector e mounted at the right backed by ablackened cavity f used to absorb stray light. Thisapparatus was equipped with thermometers and wassurrounded with a blackened light-tight metal can im-mersed in LHe in a glass Dewar. This can could beeither evacuated or filled with He exchange gas.
This system was used for several different detectorcomparisons. The spectral responsivity of each of ourphotoconductive Ge:Ga detectors was compared withthat of the detector Cav-2. Detectors were inter-changed on several occasions to evaluate the small(-5%) asymmetry of the apertures, and repeated mea-surements were made to test the reproducibility,which was within 5%.
The relative spectral responsivity of the photocon-ductive detectors was measured by replacing one de-tector by a composite bolometer with a metal filmabsorber,21 as shown at g in Fig. 5. The far-IR trans-mittance of the composite bolometer has been mea-sured and shown to be consistent with that expectedfor a 200-Q/3 film which has an absorptivity of 50%independent of wavelength.22 The Ge:Ga photocon-ductor was operated in a vacuum at 3 K, and then thetemperature was lowered to 1.2 K for a bolometricmeasurement with the same spectrometer output.The bolometer signal was used to normalize the spec-tral response of the photoconductor.
Measurements of the absolute responsivity of pho-toconductive detectors were made as follows. Theresponsivity of the bolometer was determined usingthe standard procedure from the dc load curve and ameasurement of the responsivity as a function of chop-ping frequency.23 This procedure is not as accurate asone involving a separate heater on the bolometer2 2 but
has been shown to produce results with an accuracy of±20% when used with bolometers with ohmic con-tacts.22 Once the electrical responsivity of the bolom-eter and its absorptivity were known, it was used tocalibrate the responsivity of the other detectors byusing a suitable mask to define the acceptance area ofthe bolometer.
VI. Summary and Conclusions
The results of our detector comparisons are shown inTable II. The largest responsivity was observed forCav-1 which was a -2- X 2-mm2 cavity detector withsemi-transparent electrodes. The detectors Cav-2and Cav-3, which were -2 X 2- and - 1 X 1-mm 2 cavitydetectors with metallized electrodes, gave progressive-ly smaller responsivities. This is presumably becausethe absorption in the Ge becomes small compared withthe other losses.
Two end-fire detectors Ef-1 and Ef-2 with beveledback faces were tested along with two detectors Rect-1and Cube-1 with perpendicular back faces. All thesedetectors suffer a 36% reflection loss. We believe thata suitable antireflection coating can recover most ofthis loss but have not yet fabricated one. We, there-fore, list both the measured responsivity and (in paren-theses) the responsivity corrected to the case of noreflection at the front surface. The -2.5-mm Ef-1 hasessentially the same corrected responsivity as Cav-1.The value of a beveled backface is clearly indicated bya comparison of Ef-2 and Rect-1, both of which are -1mm long.
The relative accuracy of these responsivity measure-ments should be better than ±10%. The absoluteaccuracy, however, is lower, perhaps ±30%. The abso-lute responsivity measurements were obtained with anoperating temperature of 3.0 K and a bias voltage of 50mV, where the responsivity was varying as the squareof the bias voltage. The breakdown voltage was 170mV. Although no noise measurements were made, itis believed that this bias is representative of that usedin many experiments.
The optical efficiency E0 of end-fire, rectangular, andcubic detectors was calculated from Eqs. (4) and (5)using a value of a = 2.27 cm-' for sample 2 at 100 cm-1.The calculated efficiencies given in Table II show thesame qualitative trend as the measured responsivities.The ratio of the measured responsivity to the calculat-ed optical efficiency is a measure of the success of thetheoretical model. It is 9.3 ± 0.5 for the end-firedetectors and 14 + 1 for the detectors with perpendicu-lar back faces. This suggests that only two-thirds ofthe predicted benefit of the wedged back face are beingrealized.
We conclude that the long end-fire detector Ef-1with antireflection coating will have a responsivityequal to our best cavity detector. It has a significantlysmaller cross section for cosmic rays and is much easierto use with uniaxial stress and to assemble into close-packed 1- or 2-D detector arrays. Our choice for theoptimum detector is thus the antireflection coated
end-fire geometry with a length equal to 0.3-0.5 timesa.
This work was supported in part by the Director,Office of Energy Research, Office of Basic Energy Sci-ences, Materials Sciences Division, and in part byNASA contract W-14,606 under Interagency Agree-ment, Office of Health and Environmental Research ofthe U.S. Department of Energy under contract DE-AC03-76SF00098.
All authors also work in the Lawrence Berkeley Lab-oratory.
References1. R. W. Westervelt and S. W. Teitsworth, "Nonlinear Transient
Response of Extrinsic Ge Far-Infrared Photoconductors," J.Appl. Phys. 57, 5457 (1985).
2. S. W. Teitsworth, R. N. Westervelt, and E. E. Haller, "NonlinearOscillation and Chaos in Electrical Breakdown in Ge," Phys.Rev. Lett. 51, 825 (1983).
3. H. Aumanm, B. Brown, F. Gillett, W. Irace, D. Langford, P.Mason, and R. Salazar, IRAS In-Orbit Performance Assess-ment, Jet Propulsion Laboratory Report 871 (1983), p. 19.
4. J. Wolf and D. Lemke, "Tests of Low-Background InfraredDetectors," in Proceedings, Second ESOInfrared Workshop, A.F. M. Moorwood and K. Kjar, Eds. (European Southern Obser-vatory, Garching bei Munchen, 1982), p. 191.
5. N. Oda, D. Lemke, and J. Wolf, "Radiation Effects on a Ge:GaPhotoconductive Detector," Int. J. Infrared Millimeter Waves 5,1499 (1984).
6. E. E. Haller, M. R. Hueschen, and P. L. Richards, "Ge:GaPhotoconductors in Low Infrared Backgrounds," Appl. Phys.Lett. 34, 495 (1979).
7. N. M. Haegel, "Performance and Materials Aspects of Ge:Beand Ge:Ga Photoconductors for Far Infrared Detection," MSThesis, U. California, Berkeley (1983), Lawrence Berkeley Lab-oratory Report 16694.
8. F. Abeles, Ed., Optical Properties of Solids (North-Holland,Amsterdam, 1972), p. 26.
9. R. J. Collins and H. Y. Fan, "Infrared Lattice Absorption Bandsin Germanium, Silicon and Diamond," Phys. Rev. 93, 674(1954).
10. A. G. Kazanskii, P. L. Richards, and E. E. Haller, "Photoioniza-tion of Acceptors in Uniaxially Stressed Germanium," SolidState Commun. 24, 603 (1977).
11. A. G. Kazanskii, P. L. Richards, and E. E. Haller, "Far-InfraredPhotoconductivity of Uniaxially Stressed Germanium," Appl.Phys. Lett. 31, 496 (1977).
12. H. S. W. Massey, Electronic and Ionic Impact Phenomena(Oxford U.P., London, 1969), Vol. 2, p. 1073.
13. A. Baldereschi and N. 0. Lipari, "Spherical Model of ShallowAcceptor States in Semiconductors," Phys. Rev. B 8, 2697(1973).
14. F. H. Pollak, "Effect of Uniaxial Compression on ImpurityConduction in P-Germanium," Phys. Rev. A 138, 618 (1965).
15. N. M. Haegel, E. E. Haller, and P. N. Luke, "Performance andMaterials Aspects of Ge:Be Photodetectors," Int. J. InfraredMillimeter Waves 4, 945 (1983).
16. W. E. Lamb, Jr., "Theory of a Microwave Spectroscope," Phys.Rev. 70, 308 (1946).
17. V. Hadek, D. M. Watson, C. A. Beichman, and M. D. Jaek, "Far-Infrared Transmittance of Boron-Implanted Germanium atLiquid-Helium Temperatures," Phys. Rev. B 31, 3630 (1985).
18. G. I. Hoffer, "Superconducting Junction Bolometers," Ph.D.Thesis, U. California, Berkeley (1975), Lawrence Berkeley Lab-oratory Report 3759.
20. Black coating AMES-24E obtained from J. M. Smith, NASAAmes Research Center.
21. N. S. Nishioka, P. L. Richards, and D. P. Woody, "CompositeBolometers for Submillimeter Wavelengths," Appl. Opt. 17,1562 (1978).
22. A. E. Lange, E. Kreysa, S. E. McBride, P. L. Richards, and E. E.Haller, "Improved Fabrication Techniques for Infrared Bolo-meters," Int. J. Infrared Millimeter Waves 4, 689 (1983).
23. R. C. Jones, "The General Theory of Bolometer Performance,"J. Opt. Soc. Am. 43, 1 (1953).
Of Optics(.Oltilued from page 4101
of the Emperor penguin. They also will measure the averagemetabolic rate for the birds at the colony. With these dataand those obtained in the King penguin study, the teams willdevelop better models for the impact penguins have on thesouthern oceans, calculate how efficiently growing chicksconvert food into energy, and better understand how thesespecies exploit the harsh environment in which they live.Oceanography
One of the earth's largest sea-ice environments, the Wed-dell Sea, will be studied by a Woods Hole OceanographicInstitution scientist for information on the flow, origin, andseasonality of organic and inorganic particles in the open sea.The Weddell Sea, a deep embayment in Antarctica's coast-line, has an area of about three million square miles and liessoutheast of the southern tip of South America.
Although recent research has helped clarify the distribu-tion of carbon and other organic matter and sedimentationprocesses in the Antarctic continental shelf, such processesin open areas of the southern oceans are not well understood.As the ice edge retreats each summer, a set of unique circum-stances is generated that may control sedimentation. Oneimportant question is whether particulate organic materialproduced in the open sea remains suspended or is dissolvedunder the ice during the austral winter.
This Weddell Sea Sediment Trap Experiment project isone of eight studies of that body of water to be conducted byU.S. and German oceanographers this season aboard thePolarstern, a West German icebreaker.
Oceanographers have worked since 1979 to describe thesediments that blanket the Antarctic sea floor and to relatethe different types to glacial and oceanographic conditions.A Rice University scientist this season will lead an effortfocusing on the McMurdo Sound. This area was selectedbecause it includes a variety of glacial marine environmentsin a relatively small and accessible shelf basin. The re-searchers are also interested in the origin of marine canyonson the Western side of the sound and the role they play insediment transport to the ocean floor.
Working from the U.S. Coast Guard icebreaker Glacierduring February 1987, researchers will take piston cores ofsediments and conduct seismic profiling to examine the gla-cial history of the region and compare Quarternary Periodglacial records from McMurdo Sound and the adjacent RossSea shelf. The Quarternary Period covers a time span goingback 600,000 years.
Sediment cores will be made available to other investiga-tors through the Antarctic Marine Geology Research Facilityat Florida State University.
Atmospheric SciencesA team from NOAA's Environmental Research Laborato-
ries will conduct long-term measurements of trace atmo-spheric constituents that may affect global climate. Re-searchers will measure CO2 and surface ozone levels, windvelocities, air pressure and temperatures, and trace constitu-tions using a clean air facility at Amundsen-Scott South PoleStation. The project hopes to determine the rate at whichconcentrations of these factors change and to examine theirsources and sinks.
Working with climate modelers, data will be used to deter-mine how the rate of change in concentrations affects cli-mate. Personnel at Palmer Station in the Antarctic Penin-sula area will assist by gathering CO2 samples for the NOAAgroup.
The physical processes associated with the transfer ofsolar wind energy to the earth's magnetosphere and upperatmosphere will be studied by a team from the University ofMaryland. At the Amundsen-Scott South Pole Station, thescientists will focus on the polar cusp, a window throughwhich solar wind plasmas can enter the earth's atmosphere.During the 1986-87 austral summer, the researchers willconstruct a 100-by-100-ft phased array antenna at the SouthPole for imaging data gathered by a riometer, an instrumentthat provides continuous information on precipitation ofenergetic particles and perturbations of the atmosphere.
GeologyA scientist from New Mexico Institute of Mining and
Technology will seek a better understanding of the nature,behavior, and evolution of the magma (molten rock) cham-ber of Mount Erebus, a 12,000-ft active volcano on RossIsland. The volcano, which has been active for at least 106
years, contains the world's only easily accessible convectinglava lake. This lake-the top of the volcano's magma cham-ber-is a window through which the processes operating inthe chamber system can be studied.
In September 1984, the volcano was convulsed by a seriesof large eruptions, believed to be the largest since Erebus wasdiscovered in 1841. Based on observations made in Decem-ber 1984, geologists believe the volcano has begun a neweruptive cycle. Observations made last year indicate that alava lake 45 ft in diameter exists in a site similar to that of theformer lake. Scientists will collect samples of materialsejected from the volcano and try to document changes inErebus's activity. Older lava flows and rock material brokeninto fragments by volcanic or igneous action will be mappedand examined. The research also will determine the tem-perature of the lava lake, measure the emission rate of sulfurdioxide and particulate matter, and examine changes in thecomposition and crystallization of magma by analyzing re-cently ejected materials.
Scientists from the U.S. Geological Survey will map, de-scribe, and interpret the geology of previously unexploredregions of the southern Antarctic Peninsula. Working withthe British Antarctic Survey, the U.S. researchers will studyan area called the Antarctic Peninsula-eastern EllsworthLand plate. This is the largest and most exposed part of theearth's crust of any area in Antarctica and has the greatestpotential for containing economic metallic mineral deposits.Because they will be working in an unexplored region, thescientists will develop various topical studies in the field.The geology of this area will be compared with that of nearbypreviously studied regions.
