1466 PROCEEDINGS OF THE IEEE, VOL. 58, NO.

10, OCTOBER 1970

Photodetectors for Optical

Communication Systems
HANS MELCHIOR, MEMBER, IEEE, MAHLON €3.
FISHER, MEMBER, IEEE,
AND FRANK R. ARAMS, FELLOW, IEEE

Abstract-The characteristics of high-sensitivity photodetectors 1) large response to the incident optical signal,
suitable for wide bandwidth optical communication systems are sum-
marized. Photodiodes, photomultipliers, and photoconductive detec- 2) sufficient instantaneous bandwidth to accommodate
tors for wavelengths from 0.3 pm to 10.6 pm are covered. The use of the information bandwidthof the incoming signal
internal current gain by means of avalanche and electron multiplica- 3) minimum of noise added by the demodulationprocess.
tion and by means of optical heterodyne detection to increase sensi-
tivity of high speed photodetectors is discussed. The application t o
Involved is the optimization of the entire detection system
visible and infrared laser communication systems is reviewed.
with respect to speed of response and sensitivity to weak
light signals [1]-[4]. Attention must not only be given to
I. INTRODUCTION
the choice of the optimum detector and to the design of its

D EMODULATION of signals in optical communica- load circuit and associated low noise amplifier but also to
tion systems requires photodetectors and detection other system parameters, such as desiredfield-of-view
systems which combine wide instantaneous band- [SI-[7], optical bandwidth, possible relative motion be-
width with high sensitivityto weak light signals. This paper tween transmitter and receiver (possiblyleading to Doppler
will discuss the basic requirements of photodetection sys- shift and/or a requirement for the generation of pointing-
tems suitable for the demodulation of laser signalsand pre- error information [8]), and interference due to the sun or
sent a reviewof the relevant characteristics ofdifferent other radiation.
types of high speed photodetectors. Photodetectors convert the absorbed optical radiation
Vacuum and Si photodiodes with response in the ultra- into electrical output signals. Theyare squarelaw detectors
violet, visible, and near-infrared part of the spectrum upto that respondto the intensity of light averaged over a number
about 1 pmwillbe described. Photomultipliers and av- of optical cycles [5], [9].This is because the speed of response
alanche photodiodes with wide instantaneous bandwidths is determined by carrier transport and relaxation processes
that operate in the same wavelengthregion but provide in- within the photodetector. These processes donot have
ternal gain for the photocurrent are discussed. It will be sufficiently short time constants to reproduce fieldvari-
shown that internal current gain as well as optical hetero- ations which occur at optical frequencies. The general ex-
dyne detection sigmficantly improvethe sensitivity of pression for the conversion of an incident photon stream
photodetection systems to weak light signals.A comparison of average power Pop* and optical frequency v into a primary
of the characteristics of heterodyne and direct detection photocurrent Zph is
systems will be presented. The most important infrared
photodetectors with response to a few micrometers includ-
ing Ge, I d s , and InSb solid-state photodiodes with and
without internal current gain willbe mentioned. The ex- with PopJhv= average numberof incident photons per unit
trinsic and intrinsic photoconductors andthe mixed crystal time, Zpdq=average number per unit time of electrons
photodetectors whichmaybeutilized as detectors for emitted from the photocathode,or electron-hole pairs
10.6-pm radiation will be discussed. collected across the junction region of photodiode, or
11. GENERAL REQUIREMENTS FOR PHOTODETECTORS IN mobile electrons and holes excited within the photocon-.
OPTICAL COMMUNICATION SYSTEMS ductor, andT,I =conversion or quantumefficiency.
Proper design of a demodulation system for optical sig- Utilization of wide information bandwidths for optical
nals requires speciallydesigned photodetectors that are communication systems-bandwidths of several hundred
efficient and fast. Themajorrequirements imposed on MHz or even GHz are underconsideration-requires
photodetectors and detection systems for optical commu- demodulation systems with correspondingly wide instan-
nication applications thus include taneous bandwidths. The information bandwidth of the
communication system sets the longest time constants that
are permissible in the photodetector and associated output
Manuscript received June18,1970. amplifier and load circuit. Circuit considerations play a
H. MelchioriswiththeBellTelephoneLaboratories, Inc., Murray key role in determining the bandwidthand sensitivity of a
Hill, N.J. 07922. detection system. Detector shunt capacitance, series resis-
M. B. Fisher is with Sylvania ElectronicProducts, Seneca Falls, N. Y.
F. R.Arams is with AIL, a divisionof Cutler-Hammer, Melville,N. Y. tance, and circuit parasitics must be considered. Because
MELCHIOR er a/.: PHOTODETEmORS FOROPTICAL COMMUNICATION 1467

of the capacitance inherent inphotodetector

anyoutput
and T - -
e -

circuit, the load or amplifier input resistance required to

I1 I1 (1
achieve a sufficiently large bandwidth is, with the exception
of certain
photoconductors and cooled long wavelength
photodiodes, much smaller than the internal shunt resis-
i,M(w) 0
- O@
tance of the photodetector. As a consequence, the sensitiv-
ity to weak light signals can be quite low for large bandwidth - -- --
direct detection systems. In the absence of a gain mechanism Fig. 1 . Equivalent circuit of generalized photodetection system showing
principal signal and noise sources of photodetector and load circuit.
for the photosignal ahead of or within the photodetector,
the thermal noise of the load resistance or the amplifier
noise dominates over the quantum or shot noise of the op- especially foravalanche photodiodes. For photoconductors
tical signal, thus limiting the sensitivity for demodulation and unbiased photodiodes, F = 2 at low frequencies because,
of weak signals. as will be shown in Section III-C, the magnitude of the radi-
An improvement in sensitivity is possible through the ation induced generation-recombination noise is twice as
use of wide-band output transformers such as the helical large as the shot noise.
coupling structure of traveling wave phototubes [lo]. How- From (2) the minimum signal that can be detected for a
ever, the most significant improvement in sensitivity for given bandwidth and signal-to-noise ratio can be de-
the demodulation of weak light signals with wide band- termined.
widths is brought about by the introduction of gain for the With a direct detection system, the highest sensitivity is
photosignal before it reaches the detectoroutput. Prac- reached if the minimum detectable signal is limited by
tically useful gain mechanisms are 1) current gain within fluctuations in the average signal current itself. In this limit,
photomultipliers, avalanche photodiodes, and photocon- the peak value of the minimum detectable optical power
ductors, and 2) optical mixing, that is, heterodyne detec- becomes
tion [ 1 13-[ 181. In addition, optical preamplification [ 191-
[22] and parametric upconversion [23]-[28] constitute po- Psmin
= mPop,= 4hv
tential means for obtaining high detection sensitivity.
where (3) is the quantum noise limited sensitivity of direct
A . Direct detection.
The signal-to-noise ratio and the minimum detectable In practical broad-band direct photodetection systems
signal are majorcriteria by which the sensitivity of detection without internal current gain, the minimum detectable
systems to weak optical signals is judged. The signal-to- signal is usually limited by the thermal noise of the detector
noise ratio at the outputof a generalized direct photodetec- and load resistance and by amplifier noise
tion system that comprises a photodetector with internal
current gain A4 and aload circuit as shown in Fig. 1 is given
by [11, ~ 9 1

(thermal or amplifier noise limited sensitivity).

An appreciable improvement in sensitivity over the ther-
mal or amplifier noise limited case (4)is possible through the
useof photodetectors with internal current gain such as
In this equation m = modulation index of the light, photomultipliers, avalanche photodiodes, and photocon-
M(o)=current gain or multiplication - factor within the ductors. The ideal quantum noise limited sensitivity of (3)
photodetector, 2q(Zph+ ZB + ID)B= i l = mean-square shot- can, however, not be fully reached because the current gain
noise current, ZB = background radiationinduced photocur- in photomultipliers and especially in avalanche photodiodes
rent [3], [ 5 ] , ZD=dark current component thatis multiplied provides excess noise ( F ( M ) )and because practical quantum
within photodetector, B = electrical bandwidth of detection efficiencies are smaller than unity. In a practical photo-
system, F ( M )=factorthat accounts for the increase in detection system with internal current gain the highest
noise induced by the internal currentgain process. The term sensitivity will be reached when the shot noise ; which is
4k(T,,,/Re,)B represents the effective mean-square thermal induced by the average light signal, the background radi-
noise current, with Re, being the equivalent resistance of ation, and the leakage currents ; is multiplied to a level
the photodetectorandoutput circuit [l], and T,,, is an comparable to the thermal or amplifier noise [l], [30]. For
effective noise temperature that takes intoaccount thermal this optimum current gain Mop*,the minimum detectable
noise due to the detector and load resistor, if any, and the signal will be lower by approximately a factor of 2/M0,, as
following amplifier. compared to the thermal noise limited case without gain
The factor F ( M ) accounts for the increase in noise that is (4) [l 1. If more gain than this optimum value is used in
induced by the current gain process. While this noise factor photomultipliers and especially in avalanche photodiodes,
is unity in reverse-biased solid-state, and vacuum photo- the sensitivity decreases again because of the excess noise
diodes, it is larger than unity for photomultipliers, and associated with carrier multiplication.
1468 PROCEEDINGS OF THE IEEE, OCTOBER 1970

Sufficient current gain as determined by the previously

-
TOPTICAL

i\
SUMMATION
mentioned noise considerations cannot be reached in all DEVICE
wide bandwidth detection systems becausegain-bandwidth
limitations ( M . B ) of the current amplification mechanism
set an upper limit to the maximum achievable gain M,. ELECTRICAL
OUTPUT AT
In this case, the sensitivity increases only by a factor M,. OPTICAL D ~ ~ ~
DIFFERENCE ~
SIGNAL Vs FREOUENCY
In detection systemswith moderate bandwidths, both
those with and without currentgain, the sensitivity is often
not limited by thermal or signal induced noise but by noise
generated by background radiation (IB)[5], [6] or by dark
current (ID).Noise dueto leakage currents areof importance
in photoemittersand photodiodes. Noise due to background
radiation often sets the sensitivity limit in direct detection
systems that operate at infrared wavelengths [4]-[6]. How-
r-l
- OPTICAL
LOCAL
OSCILLATOR

Fig. 2. Opticalheterodynedetectionsystem.
ever, by limiting the size, spectral acceptance bandwidth,
and field of view of the detector and through cooling, these . generation-recombination noise duetobackgroundand
currents can be usually lowered considerably. dark current, becomes small. This is the case of a well-
designed heterodyne receiver in which noise is dominated
B. Optical Heterodyne Detection by local-oscillator induced shot or generation-recombina-
In optical mixing or heterodyne detection [2E[4], [lo], tion noise.
Heterodyne detection thus allows realization of optical
[11]-[18], a coherent optical signal P, is mixed with a laser
receivers with wide bandwidths and high sensitivity almost
local oscillator PLo at the input of a photodetector as shown
equal to the theoretical quantum noise limit (hvB).The
schematically in Fig. 2. The mean value of the photocurrent
availability of high conversion gain eases the design re-
generated at the intermediate or differencefrequencyis
quirements for photodetectors considerably. High quan-
then [ll], [16], [31] for PLo>>Ps
tum efficiency and sufficient speedof response of the photo-
i,, = 14
hv
,/a. ( 5 )
detector and the outputcircuit are necessary, but the inter-
nal current gain is not. Operation at higher temperature
than direct detection may also be possible.
For the generalized photodetection circuit of Fig. 1, the Practical application of heterodyne detection is most
heterodyne transducer gain GT defined [32] as the ratio of successful at infrared wavelengths because the alignment
actual IF powerdelivered to the output amplifierwith requirements are easier to maintain. For 10.6 pm, where
input resistance R,, divided by the available optical signal strong and stable single line local oscillators exist, hetero-
power Ps, is dyne detection systems with 1-GHz base-bandwidth and
sensitivities that approach the ideal limits havebeenre-
ported [171.
Heterodyne detection systems have a diffraction-limited
field of view [33] and small spectral acceptance bandwidth.
This equation indicates that high conversiongain is possible This makes them relatively insensitiveto background radia-
using sufficient local oscillator power. tion and interference effects. Optical mixing preserves the
For proper mixing action, and for best response at the frequency and phase information of the input signal and
intermediate frequency, the polarizations of the signal and can thus be used for intensity, phase, and frequency modu-
the local oscillators should be the same[33]-[35]. The signal Iation.
and thelocal oscillator must both be normally incident and
maintain parallel wavefronts over the entire sensitive area 111. REVIEW OF DETECTOR CHARACTERISTICS
of the detector, the dimensions of which should be mini- A . Photoemissive Devices
mized to ease the alignment requirement. The minimum
detectable signal is then as determined from (2) replacing The external photoelectric effectwhichdescribes the
emission of electrons into a vacuum froma material which
m Z , J f i by ilF, so that m=2(PdPL,)* has absorbed optical radiation provides a basic mechanism
forthe detection of modulationonan optical carrier.
Photoemission fromphotoelectric materials, that is photo-
cathodes [36], [37], is used in both vacuum photodiodes
and photomultipliers for the conversionof optical radiation
into an electron current. The time constants involved in
the photoemissionprocess are sufficiently short [9] that the
(sensitivity of heterodyne detection). (7) intensity modulation present on theincoming radiation
For sufficient LO power, the second term representing is converted to a similar modulation of the electron current
thermal noise of detector, load and amplifier, and shot or emitted from the photocathode for modulation frequencies
MELCHIOR et nl.: PHOTODETEmORS FOR OPTICAL COMMUNICATION 1469

TABLE I
B m GAP ENERGY, ELECTRON AFFINITY,
AND QUANTUM EEFICIENCY
FOR
SOME SEMICONDUCTORS
WITH HIGHPHOTOEMISSIVE EFFICIENCY

::;1: ;I Number
Energy-
Electron
Quantum
;:; E
?;

- FERMILEVEL
TOP OF
VALENCE BAND
(Cs)Na2KSB

Cs,Sb 1.6
1 S-20

S-17
1 .o -0.55

0.45
40

30
1371

1371

GaAs-Cs,O 1.4 -0.55 3 5 a t 0 . 4 ~ [38]-

ISEMICONDUCTOR 15at0.8pm [40],
[431
+ \CESIUM

Fig. 3. Idealized energy band diagram of semiconductor photocathode. In0.16Ga0.84'4s-Cs2O 1.1 1 at[42]
0.9pm
Photoexcitation of electrons from valence band over energy bandgap 1.1 -0.25
0.8at 1 . 0 6 ~[43]
E , and electron afFinity EA into vacuum is indicated. ~nA~,,,,P,.,,-C~2O

extending into the microwave range. In a vacuum photo- 1

diode the .modulatedelectron.current is h e c t l y collected
at an anode and is passed through an external load resis-
tance to generate the output signal. In a photomultiplier,
---
however, the electron current is first amplified in achain of
secondary emission dynodes.
The characteristics of-vacuum photodiodes and-photo-
multipliers which are of importance in the performanceof
optical receivers are 1) the efficiency with which the photo- w
cathode converts optical radiation to electron current, 2)
time dispersion effects within the device which may limit
bandwidth, 3) internal gain mechanisms which may be used
to amplify the primary photocurrent, and4) internal noise
sources which may limit the sensitivity of the system. These
characteristics will nowbe discussed withparticular empha-
sis upon the factorswhich affectthe performanceof optical
communication systems.
The net efficiency with which the incident photons are
converted to emitted electrons, termed the external quan-
tum efficiency of the photocathode, is dependent upon the 0.4 0.6 a8 I
incident photon energy hv, the effective diffusion length of WAVELENGTH,MICRONS
electrons within the photosensitive material, and the work Fig. 4. Wavelength dependence of responsivity and quantum. efficiency
function of thephotosurface [37]. Most highefficiency for several high e5ciency photosurfaces.
photocathodesaresemiconductors [37] andan energy
diagram of a typical semiconductor photocathodeis shown
in Fig. 3 where the parameters of interest are the bandgap in sensitivity at near infrared wavelengths, where the GaAs
energy EG and the electron affinity EA. The work function and the Nd:YAG lasers operate, appearpossible.
of the photosurface E,, which defines the long wavelength The wavelength variation of the quantum efficiencyis
threshold for photoemission, is then the sum of the bandgap dependent upon the reflection and absorption properties
energy and the electron affinity. Representative values of of the photosurface as well as on theescape probability of
bandgap energy, electron a h t y , and maximum quantum the photoexcited electron in the material [37]. The typical
efficiency for several high efficiencyphotosurfaces are given variation of the current responsivity (A/W) and the quan-
in Table I. The newest photoemitters characterized by zero tum efficiency with wavelength for several photosurfaces is
or negative electron affinity[38] such as GaAs[Cs,O] shown in Fig. 4 together with some of the more important
[39]-[42] and 1&4~.15P&85[Cs20]1431 are of particular laser wavelengths.
importance both because of high quantum efficiency and As Fig. 4 and (1) showthe magnitudeof the current from
extended wavelength response. The longwavelength thresh- a photocathode is .typicaIly very small. For example, the
old for these -surfaces .is 'essentially determined by ahe -current from an S-20snrtkce illuminated .by one micro-
bandgap of the photosensitive material rather than the watt of argon - hser -power at 4880 isaDurcnrimatelv
work fnnction of the surface and substantial improvements 7 x amperes. The output .signal.generated in a simple
1470 PROCEEDINGS OF THE IEEE, OCTOBER 1970

vacuum photodiode by passing this photocurrent directly emission gain as a function of the voltage between the
through aload resistor is thus very small, particularly in the stages [36]. Although this peak israther broadit still results
case of wide bandwidth devices which require small load in abandwidth that is notindependent of the gain.
impedances. Thus the signal-to-noise ratio of the photo- Additional shot noiseis introduced due to the statistics
diode system will be limited by noise associated with the involved in the secondary emission process which amplifies
preamplifier unless some means of amplifying the primary both signal and primaryshot noise currents [47]. The
photocurrent prior totheload resistor is utilized. In a mean-square shot noise of the primary current leaving the
photomultiplier the process of secondaryelectron emission photocathode ?=2q(Ip,+ IB + ID)& with components due
provides such an internal current gain with a very low noise to signal, background, and dark current, increases through
factor. Thetypical photomultiplier geometry utilizes reflec- this multiplication process to
tion dynodes in which the secondary electrons are emitted -
from the same side of the dynode struck by the primary i2 = 2q(1,, + zB + I,)M;FB
beam. In most commonlyused devicesthe electron beam is
where Mo =FN =the average total multiplication, N = the
focused through theseries of dynodes by a suitably shaped
number of stages, F = the average gain per stage, and 6 is
electrostatic field. Magnetic focusing has been used in some
assumed to have a Poisson distribution. The noise factor F
very wide bandwidth photomultipliers.
of the photomultiplier is given by [47]
sN
The gain of an electron multiplier is given by where 6
is the averagegain per stage and N is the number of stages.
Commonly used secondary emission surfaces such as ce-
siumantimonide, magnesium oxide, or beryllium oxide
provide gains per stage in the range of 3 to5 [36]. Recently or, for 6 >> 1 which is the case for any practical secondary
developed negative electron affinity surfaces such as emission dynode material
GaP[Cs] [44] or GaAs[Cs,O] can provide gains per stage
6
of 20 to 50. An overall gain of lo6 to 10' can be achieved Fx-. (11)
using any ofthese surfaces although obviously a larger 8-1
number ofstages will be required in the case ofconventional Thus it may be observed that a photomultiplier has a very
low gain dynodes. Thechoice of dynode material must pri- low noise factor, a typical example being F x 1.5 for F = 3.
marily be made pn the basis ofdesired bandwidth andnoise The new high gain dynodes result in even lower noise fac-
factor rather thanoverall gain as will now be described. tors. Since the electron multiplier structure is in effect a
The factors which limit the bandwidth of photomulti- cascaded series of amplifiers, the gain of the first stage is
pliers are time dispersion effects within the secondary elec-more important in determiningexcess shot noise than the
tron multiplier and capacitance associated with the anode gain of the remainingstages.
output circuit [45], [46]. The time dispersion within the The optimum photomultiplier for an optical communi-
electron multiplier is caused by nonuniform electron transit cation system operating at a specificwavelength should
times for 1) electrons emitted from different positions on have the following characteristics based upon the previous
the dynode surface because of spatially varyingelectric. discussion. The photosurface should be chosen for maxi-
fields, and 2) electrons emitted from the same position on mum quantum efficiency at the operating wavelength using
the dynode surface but with finite initial velocity distribu-antireflection coatings and multiple-pass techniques to
tions. The effect of nonuniform electric fields can be mini- enhance optical absorption wheneverpossible[36],[37].
mized by proper device geometry. The timedispersion due High gain dynodes suchas GaPshould be utilized, particu-
to the unavoidable initial velocity distribution can only be larly for the first stage, since excess shot noise can be mini-
minimized by employing high interstage voltages and re- mized in this manner. These high gain dynodes typically
ducing the numberof stages. For example, the upper half- require high dynode voltages for maximum gain and the
power frequency limit of a rather idealized electron mul- resulting high electric field reduces transit time dispersion.
tiplier in which the electric field is assumed
to have no spatial
In addition,their high gain reduces the numberof required
variation is given by [45] stages with a resultant further reduction in transit time
dispersion and increase in bandwidth. Finally, the anode
lead geometry or output couplingcircuit must be designed
to provide a bandwidthconsistent with that of the commu-
nications system. Relatively narrow bandwidth photomul-
where E =electric fieldbetween dynodes, rno = electron tipliers employ simple wire connections from the anode to
mass, ij = most probable initial velocity, and N = number of the output header pin, while in wide bandwidth devices, the
stages in the electron multiplier. Thus the bandwidth is output current must be focused into a coaxial anode and
proportional to the dynode voltage (for a given dynode output transmission system. Factors such as photocathode
spacing) and inversely proportional to the square root of area and dynamic range mustbe considered in addition. It
the number of stages. The dynode voltage, however, also is desirable to minimize the cathode area to reduce the dark
determines the secondary emission gain since most com- current noise, consistent with optical and mechanical design
monly used dynode materials exhibit a peak in secondary requirements. The dynamic rangeis limited at the low end
MELCHIOR er a / . : PHOTODETECTORS FOR OiTICAL COMMUNICATION 1471

by noise and at the h g h end by saturation of the output

current or by space charge effects that originate in the
dynode chain.
The dynamic range of photomultipliers to beused in
wide bandwidth systems must be carefully considered with
respect to the desired S ( N ratio. The SIN ratio measured
at the output of the photomultiplier in a direct detection
system with sufficient gain to achieve shot noiselimited TlON
operation is, after (2), given by
s - )(mZp,M)ZR,
_
N - 2qZp,MZFBR,,’
However, if the output current mZFM becomes limited by COATING METAL LAYER
saturation effects to ZSat, the SIN ratio has alimiting value of
METAL
thv fCONTACT,
which might not besufficient to reachsignal shot noise
limited sensitivity. . , , ,

Of the commercially available photomultipliers, the con-

ventional electrostatically focused multipliers have band- AND LIGHT REFLECTION
widths up to about 100 MHz. If a third high voltage elec- (e) (0
trode isplacedbetween each pair of dynodes to reduce Fig. 5. Construction of different high speed photodiodes. (a) p n diode.
transit times, bandwidths of several hundred MHz result (b) pi-n diode (Si optimized for 0.63 pm).(c) pi-n diode with illum-
[49]. Improved designs that still maintain the conventional ination parallel to junction. (d) Metal-semiconductor diode. (e)
dynode chain, but use high efficiencyphotoemitters and the Metal-i-n diode. (0Semiconductor point contact diode.
new high gain dynodes have been reported with subnano-
second response times [50]. Magnetically focused crossed- RS

field photomultipliers can provide a bandwidth of 6 GHz by

using a well-matched coaxial output line [46].

B. Solid-state Photodiodes and Avalanche Photodiodes

1) Photodiodes: Efficienthighspeed photodiodes, and
avalanche photodiodeswith internal currentgain, have been
Fig. 6 . Equivalent circuit of photodiode.
fabricated by using specially designed p-n, p-i-n or metal-
semiconductor junctions as shown in Fig. 5. These diodes
are usually operated in the reverse bias region. In general, The ac characteristics of photodiodes can be described
photon excited electrons and holes that are generated 1) in terms of an equivalent circuit [15], [29], [31], [52], [54]
within the high fieldregion of the junction [51] and 2) in the that contains a photocurrent generator ,,i diode capaci-
bulk region and then diffuse to the junction [52], are col- tance C , series resistance R,, and shunt resistance Ri, as
lected as photocurrentacross the high field region. However, shown in Fig. 6. The shunt resistance Ri is very high for
diffusion processes are slow compared to the drift of carriers diodes operating in the visible but is included to account for
in the high fieldregion. Therefore, in high speed
photodiodes, the relativelylow leakage resistances of infrared photo-
the carriers have to be excited withinthe high field regionof diodes.
the junction or so close to the junction that diffusion times In thevisible and nearinfrared well-designed diodes with
are shorter than, or at least comparable to, carrier drift high quantum efficiency, fast speed of response, low dark
times [15], [29]. Carriersare then collected across the currents, and low series resistances can be obtained. High
junctionat scattering limited velocities. thatare of the quantum efficiency requires minimization of the light re-
order of usat= lo6 to lo7 cm/s. Under these circumstances, flection at the diode surface and placement of the junction
the peak ac photocurrent is given by [51] (for photoexcita- in such a way that most photons are absorbed within or
tion of carriers at the junction
edge) close to the high field region of the junction. Solid-state
photodiodes are usually designed for light incidence normal
to the junctionplane. The quantumefficiency for aparticu-
lar photodiode depends on wavelength the of operation and
where T,=drift transit time. As space charge layer widths on the depth and widthof the junctionregion. The spectral
(wi) can be in the order of micrometers, carrier transit dependence of thequantum efficiency foranumber of
times T,= wJu,,, in the subnanosecond region are possible. photodiodes is shown in Figs. 7 and 8.
1472 PROCEEDINGS OF THE IEEE, OCTOBER 1970

Fig. 7. Wavelength dependence of quantum efficiency and responsitivity for several high speed photodiodes.

pleted space charge layers or in p+-n diodes with an epi-

taxial base. For operation with baseband widths up into
the GHz region, high speed photodiodes (and avalanche
photodiodes) are commonly mounted into especially de-
signed coaxial headers [54], [56].
The choice of a photodiode depends mainly the on wave-
length of operation (Table 11). Because of their developed
technology, silicon diodes are preferentially used in the
ultraviolet, visible, and near infrared part of the spectrum
up to about1 pm. With germaniumdiodes the response can
be extended to over 1.5 pm. The geometry of both Si and
0 I 2 3 4 5 Ge photodiodes can be optimized for particular applica-
W A V E L E N G T H [prn] tions. At short wavelengths, where light is absorbed close
Fig. 8. Quantum efficiency and responsivity of photodiodes to the semiconductor surface it makes sense to use metal-
that operate between 1 and 6 pm. semiconductorphotodiodes with thin semitransparent
metal layers [56)-[58]. Carriers are then separated in the
The speed of response of photodiodes is determined by highfield region close to the surface thus yieldinghigh
RC time constants [29], [54], and the already mentioned quantum efficiencies andshort response times.Several
carrier diffusion or drift transit times. The character- metal-semiconductor photodiodes [Fig. 5(d) and (e)] with
istic cutoff frequencyof a photodiode is givenby response at short wavelengths have been fabricated from
+
fc = 1/27cR,C( 1 (R$R,)).Practical detection systems have, Si [57], GaAs [59], and ZnS [60] as indicated in Fig. 7 and
however, lower cutofffrequencies because ofthe finite load Table 11. For the visible rangeof the spectrum, where
resistance andtheadditional capacitance (and induc- light penetrates a few micrometersinto silicon, diffused
tances) of the load or amplifier circuit. Fast photodiodes p n [61] and pi-n junctions [53], [54], [62] are used [Fig.
have usually planar junctions that aresmall, 50 to lo00 pm 5(a) and (b)]. Diffused junctions exhibit somewhat smaller
in diameter, in order to keep the diode capacitance and reverse dark currents than metal-semiconductor junctions.
leakage currents small [53]. Light has to be focused onto For longer wavelengths close to the bandedge of the diode
these diodes. The highest speeds of response andthe material, light penetrates, deeply into the material. High
smallest capacitances might be obtained from point con- quantum efficiency thus requires wide space charge layers.
tact diodes [Fig. 5 ( f ) ] [55] but the focusing of the light onto This leads to relatively long carriertransit times. For these
the small light sensitive areas is more difficult. Low series diodes a tradeoff exists between quantum efficiency and
resistances in small area photodiodes are possible if the speed of response [63]. As an example Fig. 7 shows the
thickness of undepletedsemiconducting bulk regions is quantum efficiency at long wavelengths around 1 pm for
kept small, as in pi-n or metal-i-n diodes with fully de- Si pi-n diodeswith various spacecharge layer widths
 MELCHIOR et al.: PHOTODETECTORS FOR OPTICAL COMMUNICATION 1473

TABLE I1
PERFORMANCE CHARACTERISTICS
OF PHOTODIODES

Wave- Peak
length Efficiency (a,) Sensitive Series Response Operating
Diode Range or Respon- Area Capacitance Resistance Time Dark Current
Temperature Comments Ref.
@m) sivity (cm2) (PF) (R) (seconds) ( K)

Silicon n+-p 0.61 40 2x lo-’ 0.8 at -23 V 6 130 ps50 pA at - 10V 300 avalanche photo- [61]
with 50-0 diode
load

Silicon pi-n 0.6328 >90 2x10-5 <I -1 100

ps < A at-40 V 300 optimized for [62]
with 5 0 4 0.6328 A
load

Silicon pi-n 0.4-1.2 >90at 5x lo-’ 3 at -200 V <I ns

70.2 P A at -30 V 300 [651
0.9 pm
> 70 at 3 at -200 V <1 7 ns
1.06 pm

Metal-i-nSi 0.38-0.8 270 3x lo-’ 15 at - l00V lOnswith 2 x 10-’Aat -6V 300 [57 1
5 0 4 load

0 0.6328 Au-nSi 2 < 500 ps 300 Schottky barrier. [56]

antireflection
coating

PtSi-nSi 0.354.6 -40 x 2 lo-’ <1 120 ps 300 Schottky barrier[58]

avalanche photo-
diode

Ag-GaAs < .36 50 300 [591

Ag-ZnS 70< .35 300 [601
Au-ZnS < .35 50 300 [601
n+-p
Ge 0.41.55 50 2 x lo-’ 0.8at
-16V < 10 120
ps
2 x 10-8 300 Germainum [30]
uncoated avalanche photo-
diode

pi-nGe 1-1.65 60 2.5 x 10-5 3 25 ns at 71 illumination [65]

500 v entering
from
[66]
side

GaAs point 0.6328 40 0.027 30

contact

InAs p n 0.5-3.5 >25 3.2~ 3-5at V 12 < 10-6 77

InSb p n 0.65.5 >.25 x5 7.1 at -0.2

V 18 5x 77 ~ 5 1
[291
~7 I
InSb p n 2-5.6 5 X 10-4 1MR shunt 77 Reverse break [68]
resistance down voltages
30 V
Pbl -,Sn,Te
x=0.16
9.5 pm 45
VjW
q=60
4 x lo-’ - 10-9 77 shunt resistance
R i =10 R
[72]
[741

Pb, -,Sn,Se
X = 0.064
11.4 pm 3.5 V/W
q=15
7.8 X 10-3 - 10-9 77 shunt resistance
Ri= 2.5 R
[73]
[741

Hg,-xCd,Te 15 pm q - 10-30 4 x 10-4 8 < 3 x 10-9 77 shunt resistance [70]

x=0.17 R , > lOOQ ~711,
~361

[a], [65]. The corresponding carrier transit times or currents. Reduced dark currents and thushgher sensitivity
diode response times and bias voltages can be determined can be obtained in both Si and Ge diodes by cooling. The
from Fig. 9 [63]. Germanium diodes [53] are used at wave- problems encountered in the design of Ge diodes that are
lengths below 1 pm to beyond 1.5 pm. However, due tothe fast and efficient for 1.53 (HeNe laser) and 1.54 pm (Erbium
narrower bandgap, germanium diodes exhibit larger dark laser) are similar to those encountered with Si diodes for
 I414 PROCEEDINGS OF THE IEEE, OCTOBER 1970

efficiency curves of Fig. 8. For maximum sensitivity these

diodes have to be operated at temperatures between 77 and
135°K.
For longer wavelengths in the 6- to 30-pm range and es-
pecially for 10.6 pm, where the COz laser operates, mixed
crystal photodiodessuch as Hg,-,Cd,Te [69&[71] and
Pb,-,Sn,(Te,Se) [72]-[74] are under development. Dy-
namic shunt resistances in Hg, -,Cd,Te in the range loz
to lo5 ohms at 77°K have been obtained [70], [71], so that
photodiodes for 10.6 pm that are fast and efficient are
becoming available for operation at77°K or even 110°K.
Recently, (Hg, Cd) Te photodiodes with wider.bandgaps
have been investigated for use at shorterwavelengths down
to 1-2 pm [69].
2) Avalanche Photodiodes: Current gain in solid-state
photodiodes is possible through avalanche carrier multi-
plication [75], observed at high reverse bias voltages where
carriers gain sufficient energy to release new electron-hole
pairs through ionization. Substantial current gains have
beenachievedby this process even at microwavefre-
quencies [61],[75]. Despite the fact that excess noise is
Fig. 9. Carrier transit times for Si p-i-n diodes with different. introduced by the multiplication process [W], [61], [76]-
depletion layer widths (after McIntyre[63], [MI). [78], significant improvements inoverall sensitivity are pos-
sible using silicon or germanium: avalanche photodiodes
1h y IhV for photodetection systems with wide instantaneous band-
widths.
Similar criteria apply for avalanche photodiodes, with
respect to quantum efficiency and speed of response, as for
P+ P'
conventional nonmultiplying photodiodes. Additional at-
(a) (b) tention must, however,be given to the current gain and its
h' ANTIREFLECTION limitations and tothe noise properties of avalanche photo-
COATING
diodes.
In the design of avalanche photodiodes, special precau-
tions must be taken to assure spatial uniformity of carrier
multiplication over the entire light sensitive diode area.
Microplasmas, i.e.,small areaswith lower breakdown
voltages than the remainder of the junction, andexcessive
leakage at the junction edges can be eliminated through the
use of guard ring structures [61], [79] as indicated in Fig.
10. The selection of defect-free material and cleanliness in
\ I / processing allows fabrication of microplasma-free diodes.
I L Highly uniform carrier multiplications in excess of lo4 and
'DEPLETION'
200 have been reached at room temperatures in small area
silicon [61] and,germanium [30] diodes, respectively. In
large area diodes that are free of microplasmas, thespatial
Fig. 10.Construction of dserent avalanche photodiodes.(a) Siguard uniformity of carrier multiplication is limited either by
ring structure.(b) Ge mesa structure with guard ring. (c) Meta-semi- doping inhomogeneities of the starting material or by in-
conductor structure with guard ring. (d) Planar p-n' structure with.
low fields at junction edges. (e) Si-n+-pn-p+ structure. (f) Si struc- homogeneities in the diffusion profle. Typical variations
ture with wide depletion region and slant surfaces. can be 20 to 50 percent at anaverage multiplication of lo3.
The highest current gains are observed if the diodes are
0.9 and 1.06 pm. Better results might be achieved if light biased to the breakdown voltage. This is illustrated by
is allowed to penetrate from the side [66], parallel to the Fig. 1 1 . where the voltage dependence of the dark current
junction a s shown in Fig. 5(c). This may result in a very of a Ge avalanche photodiode as well as the response to
narrowsensitive area. pulses from a phase locked 6328 A laser are shown. Atlow
For wavelengths beyond 1.5 pm, up to 3.6 pm and 5.6 reverse bias voltages, no carrier multiplication takes place.
pm, respectively, InAs and InSb photodiodes [29], -[32], As the reverse bias voltage is increased, carrier multiplica-
[67], [68] are suitable as can be seen from the quantum tion sets in asindicated by the increase in the currentpulse.
MELCHIOR er al.: PHOTODETECTORS FOR OITICAL
COMMUNICATION 1475

GERMANIUM
PHOTODIODE
HE-NE
LASER * 6 3 2 8 H Tp--.6ns through the multiplication region long after the primary
carrier has left the junction.
The currentgain-bandwidth product is inversely propor-
tional to the average transit time of the carriers through the
multiplication regon [80]-[83], [85] and depends on the
ratio between the electron and hole ionization rates [84].
HIGHEST For Ge, the current gain-bandwidth product is independent
MULTIPLICATION
OF PHOTOCURRENT of excitation because the ionization coefficients are almost
\ equal, but for Si, higher current gain-bandwidth products
result if the avalanche is initiated with electrons because
they have a hgher ionization rate. If one canrealize carrier
NO CARRIER
MULTIPLICATION -
multiplication in solid-state diodes in which only one type
of carrier ionizes, then the current gain would not be limited
by a current gain-bandwidth product. The response time
at high multiplications would then be about twice as long
as without multiplication. Although indications exist that
6 8 IO 12 14 I 6 V, only electrons cause ionizations in InSb [86] and Schottky
VOLTAGE [ V I
barrier GaAs [87] photodiodes, no speed of response mea-
Fig. 1 1 . Current voltage characteristic and pulse response surements have been reported to date.
of germanium avalanche photodiode.
Excess noise in avalanche photodiodes is induced by
fluctuations in the carrier multiplication process [76]-[79].
The maximum pulse amplitudeand multiplication of For practical diodes, the shot noise of the average photo
photocarriers is obtained at the breakdown voltage. At and dark current hasbeen found [30],[61],[78],[79] to
voltages above the breakdown voltage, a self-sustained increase faster than the square of the carrier
multiplication
avalanche current flows that becomes less and less sensitive M , approximately as
to photon-excited carriers.
The maximum gain of an avalanche photodiode is limited
either by current induced saturation effects [30] or by a
current gain bandwidth product [80E[83],[85]. Current in reasonably close agreement with theoretical calculations
induced saturation of the carrier multiplication is observed based on a spatially uniform avalanche ionization region
because the carriers that emerge from the multiplication [77]. The noise factor F =M" depends on the ratio of the
region reduce the electric field within the junction and cause ionization rates and on the type of carriers that initiate the
voltage drops across the series and load resistance of the avalanche [l], [30], [77]-[79]. For germanium diodes
diode [30]. This leads to a current dependent reduction of [30], where the ionization coefficients for holes and elec-
the carrier multiplication factor. This saturation manifests trons are about equal, the excess noise factor increases as
itself for high light intensities, at which the multiplied cur- the carrier multiplication (x = 1). This relatively high noise
rent will only increase as the square root of the photocur- is due to the fact that for equal ionization rates, a relatively
rent instead of being proportional to the photocurrent as low number of carriers is present within the multiplication
at low light intensities. For low light intensities, the dark region at any one time so that fluctuations in the ionization
current will set a limit to the average value for the maximum events have a large influence on the avalanche process.
carrier multiplication that can be achieved at low frequen- Much lower noise is observed in junctions withhighly
cies. The aforementioned maximum multiplication of 200 unequal ionization rates if the avalanche is initiated with
for germanium diodes [30] is due to the h g h dark current. carriers of the higher ionization rate. Initiation of avalanche
Lowering of thetemperaturein diodes with sufficiently carrier multiplication by electrons leads to low excess noise
high breakdown voltages ( > 20 volts for Si and Ge) so that in Si ( F = M 0 . 4 ) [78], [79], InSb [86], and Schottky barrier
no internal field emission takes place, decreases the dark GaAs [87] diodes. If only one type of carrier can ionize,
current and leads to higher values for the maximum carrier theoretical calculations show that the noise factor does not
multiplication. follow (15), but would be equal to two at high multiplica-
At high modulation frequencies or for fast optical pulses, tions [77]. The results for InSb andfor certain GaAsdiodes
the currentgain-bandwidth product sets a limit to the maxi- indicate that only electron ionization is of importance in
mum achievable gain [l], [80]-[85]. For silicon [61] and these materials.
germanium [30] n+-p diodes gain-bandwidth products of The different types of avalanche photodiodes that have
100.and 60 GHz have been reported as indicated in Table been constructed to date are shown schematically in Fig.
I n . This current gain-bandwidth product is due to the fact 10 and their performance data are listed in Table 111. As
that both electrons and holes excite electron-hole .pairs Table I11 shows, various Si n+-p guard ring avalanche
throughionization in silicon and germanium diodes. photodiodes have been developed with diameters of the
Secondary electrons and holes thus travel back and forth light sensitive area between 40 and 200 pm [61], [79], [89]-
1476 PROCEEDINGS OF THE IEEE,OCTOBER 1970

TABLE I11
CHARACTERIS~CS
OF AVALANCHE
PHOTODIODES

Current
Wave- Sensitive Avalanche Multiplication Gain- Capaci- Series
Diode Construc-
length Area Breakdown
Dark
Maximum
Bandwidth
Noise tance Re& Ref,
tion Range Current Voltage Gain iz Mz+I
(CmZ) (volts) Product (p~) tan=
Olm)
(GH4 (a)
Silicon n -p
+ Fig. 10(a) 0.4-1 2x 50 23
pA at 104 ..;-0.5 100 0.8 at 6 W I
-10 v -23 V
Silicon n+-pn-p+ Fig. 10 (e) 0.5-1.1 2x 200 - 88 1-0.4 high

Silicon n+-i-p+ Fig. 1 0O0. S l . 1 200 to low not very

2000 high
Psi-nSi Fig. l q c ) 0.35-0.6 4x --1ov
1 nA at 50 400 x - 1 for
visible
40for UV
excitation
<l 1581

illumination

Pt-GaAs Fig. 1O(c) 0.40.88 very

-60 > 100 low >SO ~ 7 1

Germanium n+-p Fig. (lob) 0.4-1.55 2 x lo-' 2 x IO-' A at250

16.8 x- 1 60 0.8 at < 10 [30]
- 16 V and 300°K > 1 O4 -16V
300'K at 80°K

Germanium n+-p Fig. 10(a) for 1.54 150

InAs 0.5-3.5

InSb at 77:K 0.5-5.5 a few very 10 low

[91]. These are the simplest avalanche photodiodes. They coefficients for holes are not much lower than theionization
operate at relatively low voltages and are useful between coefficients for electrons [84].
about 0.4 and 0.8 pm. Similar Ge n+-p diodes have been Silicon avalanche photodiodes withhigh current gain
constructed [30] with aguard ring that terminates in a and relatively low excess noise are thus available for the
mesa structure in order to reduce the surface leakage cur- wavelengthrange between 0.4 and 0.9 pm.Germanium
rent. These Gediodes are useful as fast diodes with gain at avalanche photodiodes havehigher excess noiseand higher
wavelengths between 0.5 and 1.5 pm. Both the small area leakage current if not cooled, but they are excellent detec-
Si and Ge avalanche photodiodes canresolve optical pulses tors for 1.06 pm and can be used up to over 1.5 pm as well
with widthsof 130 ps between their 50-percent rise and fall [93].As an example, the application of a Ge avalanche
time points. The current gain-bandwidth products are 100 diode to the demodulation of small signals at 6 GHz on
and 60 GHz, respectively, thus indicating that currentgains a 1.15-pm laser beam is illustrated in Fig. 12. As can be
of 100 and 60 are possible in a detection system with 1-GHz seen, the signal from the diode increases as M 2with mul-
bandwidth,. tiplication and the noise from the avalanchediode as
Two different structures have beendeveloped for Si M 3 . The best operating point (M,,J with the largest SIN
avalanche photodiodes for operation at wavelengths be- ratio and the highestsensitivity to weaklightsignalsis
tween 0.8and aboutl pm. In order to achieve highquantum reached if the carrier multiplication is adjusted so that the
efficiencies these diodes have wide space charge layers. One noise of the avalanche diodeis about equal to theamplifier
structure, the n+-v-p+ structure [89] [see Fig. lO(f)] has a noise.
wide high field regon in which carrier multiplication can Results for GaAs and InSb diodes indicate that avalanche
take place. Because the electric fields are relatively low in photodiodes with almost no excess noise (only a factor of
this case, a very low hole ionization coefficient results [84]. two) and high gain that is not limited by current gain-
These diodes exhibit low excessnoise [77] but will have only bandwidth products will be forthcoming in the foreseeable
a low current gain-bandwidth product [81],[82]. In the future. The InSb avalanche diodes will extend the range
n+-p-z-p+ structure [90],[91][Fig.lO(c)], the multipli- of solid-state diodes with internal currentgain up to5.6 p.
cation is concentrated to the narrow n+-p region whereas Avalanche carrier multiplication has further been ob-
the wide K region acts as a collection region for photon served in a number of additional materials, among them
excited carriers. This narrow multiplication region will re- InAs [88], but no practical devices have been reported.
sult in a high gain-bandwidth product but results also in Extension of avalanche photodetection to wavelengths
higher excess noise. The noise is higher because, at the high beyond the bandedgeof a semiconductor material is possi-
fields in this narrow multiplication region, the ionization ble in metal-semiconductor junctions as photoemission of
MELCHIOR et a/.: PHOTODETECTORS FOR OPTICAL COMMUNICATION 1477

-60 I , i
I I
I I

GE - AVALAHCHE
+ + + I
-

-
T I
Fig. 13. Small signal equivalent circuit of photoconductive detector.

the speed of response T is indicated by the existence of a

current gain-bandwidth product [96], [97]

-100 - -
RECEIVER
The maximumvalue of this current gam-bandwidth product
is independent of the carrier lifetime .z and limited as indi-
1 2 5 IO 20 50 1 0 0 200 500 cated by the dielectric relaxation time rrel= PE [95], where
DC
MULTIPLICATION
E = permitivity and p = resistivity of the photoconductor.
Fig. 12. Signalandnoisepoweroutput of germaniumphotodiodein Evaluation of an optical detection system that utilizes
1-MHz band at 3 GHz. Optimum operating point (M,) for best S / N
ratio and sensitivityis indicated. photoconductors requires in addition consideration of
the differential conductance G and capacitance C of the
photoconductor andof the different noise sources indicated
carriers from the metalcontactcan be combined with in the small signal equivalent circuit of Fig. 13.The differen-
ionization of carriers within .the high field region of the tial conductance is given by [98]
junction. Only small quantum efficiencies can, however,
be expected. For palladium-silicon Schottky barrier diodes
quantum efficiencies of 0.75 percent and avalanche multi- Go = -
"yn1if determined by doping
plications >200 have been observed at 1.3 pm [94].
C . Photoconductive Detectors -
if determined by illumination (18)
Electron-hole pairs created in a bulk semiconductor by
the absorption of optical radiation increase the conductiv- where ( 4+I& -number of photons absorbed per unit
ity of the material during the lifetime of the pairs. This time dueto both signal (I&) and background (IB) radiation.
effect, termedphotoconductivity [95], provides -a basic The principal mean-square noise current generators of a
mechanism for the detection of optical signals, since upon photoconductor are [95]
application of a bias voltage this conductivity modulation 1) generation-recombination noise[99], [ l o ]
can be translated into a modulation of the current that
flows through the output circuit.
Fora simpledcbiased photoconductor with ohmic
contacts, whose conductance is dominated by one type of
with I, = ( T / T ) ( l p h + I B ) being the dc light induced output
excited carrier, the optically induced small signal peak ac
current is [16], [31] current of the photoconductor.
2) thermal noise due to the differential conductance
Go of the photoconductor at temperatureTG
-
ii = 4kTGGOB. (20)
where T = mean carrier lifetime, T,= Lz/pV = carrier transit
time, p = carrier mobility, L =interelectrode spacing V The SIN and sensitivity considerations of Section I1 can
=bias voltage, and 4 =tan- '(or). be applied to detection systemswith photoconductors.
The factor M=(T/T,) exp ( - J ~ ) ( l + 0 2 ~ Zis) -the
1/2 The generation-recombinationnoise of (19) can be brought
photoconductive gain and indicates the number of carriers into the form &(I,,, i- . B with (kf(W)I2
I~)Ihf(0)1~F = kf;/
which cross the photoconductor for each absorbed photon. (l+o%') and F = 2 . The thermal noise due to the output
Practical application of photoconductors generally requires circuit with a load conductance GL,if any, at temperature
this gain to bemaximized consistent withfrequencyre- T L and amplifier input conductance G , and effective

sponse requirements by operating the photoconductor at amplifier noise temperature T,

the highest bias voltage compatible with breakdown and -
i i = 4kTLGLB
extraneous noise considerations. The fundamental tradeoff
in phatoccmchcters -between the-current gain M= r / T , and and
1478 PROCEEDINGS OF THE IEEE, OCTOBER 1970

can be combined with the thermal noise of the photocon-

ductor into the single thermal noise source used in (2H7).
The equivalent circuit [31] of Fig. 13 takes account of the
main features of the photoconductor. It should, however,
be kept in mind, thatin practical photoconductors material
parameters often change with bias voltage, operating
temperature, and optical power. For example, if the differ-
ential conductance Go of a photoconductor is too low, the
carrier transit time T, becomes shorter than the dielectric
relaxation time zrelat relatively lowbias voltages. This limits
the current gain-bandwidth product [95] and leads to space
BIASFIELD IN V K M
charge limited currents [95]. Saturation of the current gain Fig. 14. Response time of photoconductive Ge:Cu and Ge: Hg(Sb)
is often observedat high bias voltages [l 1 1 1. This is because (After Buczek and picus [102]).
the carrier velocities cease to increase as u =pE with electric
field E but saturate at a scattering limited velocity U , ~thus
leading to avoltage independent transit time T,=L/uSatand the lifetime should not be shorter than thatdictated by the
gain M = u,,z/L. In several extrinsic photoconductors an bandwidth, in order to optimize the current gain. Once the
increase of the carrier lifetime z (as shown in Fig. 14 for value of 7 is established the M B product can be optimized b~
Ge :Cu and Ge:Hg(Sb) [102]) and of the p product [lo41 minimizing the transit time, which is accomplished by the
and impact ionization have been observed with rising bias use of high dc bias voltages and small detector dimensions
fields. Nonohmic contacts lead to carrier depletion regions in the direction of current flow. In addition, sufficient speed
and low current gain [95]. In certain materials, such as in of response must be provided by the outputcircuit to satisfy
CdSe [105], the current gain and the carrier lifetime depend the condition Re&< 1/27rB. The value of C includes the
strongly onlight intensity, typicallydecreasinglike the capacitance of the photoconductor, its mount, and the out-
square rootof the optical power at high lightintensities [ 161. put circuit whichmust thus be kept as low as possible.
In many photoconductors bothtypes of excited carriers Given the value of B and C, a value of RL is chosen equal to
will contribute to the photoconductivity and be able to or lower than the photoconductor resistance to satisfy the
move out through the contacts thus leading to a much previously mentioned condition. For widesignal band-
lower photoconductive gain than indicated by(16). For widths this loading leads to reducedsensitivity in direct
this case and for nonohmic contacts, microwave bias has detection systems since the photoconductive gain is now
been explored as a potential means to increase the current insufficient to overcome receiver thermal noise. However,
gain and the gain-bandwidth product [106]. The effect of by the use of heterodyne detection, photoconductors can
RF bias with a sufficiently high frequency is to effectively yield high sensitivity and wide bandwidth in combination
"trap" carriers in the bulk of the photoconductor, so that by using sufficient local oscillator power to overcome the
they are not swept out to theelectrodes for recombination, thermal noise contributions.
but instead move back and forth contributing to conduc- 1) Characteristics of Infrared Photoconductors: The
tivity, until volume recombination takes place. Although spectral sensitivity for a number of photoconductors (and
impressive results have been reported [loll, [107], practical photovoltaic detectors) is shown in Fig. 15. Direct detection
application of ac biasing might be somewhat limited be- sensitivity is given in termsof detectivity D*,the commonly
cause of complicated circuitry and because of the noise that used figureof merit for infrared detectors [5], D* is inunits of
is inducedthrough instabilities in the microwavebias - ',
cm Hz''2 . W- it is normalized to asensitive area = 1 cm2
source. and a postdetection bandwidth B = 1 Hz. D* is related to
In the design of photoconductive direct detection sys- the noise equivalent power NEP (SIN= 1 and m = 1 in (4))
temsone strives to reach the generation-recombination by D* = A'/2B'/2/NEP. The values for D* given in Fig. 15
noise limited sensitivity. This is obtained by the useof are only representative and are given mostly for a 60" field
high gain to overcome the thermal noise of the photocon- of view. The highest values of D* are usually reached by the
ductor and output circuit. Once sensitivityislimited by use of high purity materials with long time constants and
background radiation induced g-r noise, a further increase high resistances.
in detection sensitivity is still possible through the useof D* is commonly measuredat audiomodulation frequen-
aperturing, which narrows the field of view of the photo- cies usinga matched load resistance. This is satisfactory for
conductor, and by spectral filtering and cooling [5]-[7]. radiation sensing applications of infrared detectors, where
. To obtain a flatfrequency response and to minimize speed isusually secondary to sensitivity. The adventof opti-
phase distortion within the bandwidth of operation B, the cal communications and other laser applications gave im-
photoconductive lifetime should be as short as T I 1/27cB. petus to more detailed investigations of detector speed and
Operation in the rolloff regon ( 5 ) 1/27rB) is not acceptable to the development of infrared detectors with wide band-
in mast communications applications. On the other hand, width capabilities.
MELCHIOR er a / . : PHOTODETECTORS FOR OPTICAL COMMUNICATION 1479

J
I 2 4 IO 20 40

WAVELENGTH IN MlCRWETERS
Fig. 15. Spectral dependence of detectivity for photoconductive and
photovoltaic infrared detectors. Representative values are given [ 5 ] ,
[61, [1191.
TABLE IV
INFRARED
PHOTOCONDUCTIVE
DETECTORS

Approximate

Material Reference

\r-.,
lperatron

InAs 3.3
195 3.6 - 3 x 103 3 x IO" 5 x lo-'

lnSb 1 IO 77 5.6 5.3 -3 x103 0.54.8 103-104 6 x IOLa

-1 x IO" 5 x 10.6

Ge:Au 60 77 9 6 - 2 x 0.2-0.3
4 IO' 3 x 10~-10'0 3 x IO-@

Ge:Au(Sbl 60 6 77 9 106 6 x IO9 1.6 x 10-9

7 x IO'
-4 x 10'0 3 x 10-8
Ge: Hg 35 4.2 14 11 - 3 0.24.6 1-4 x IO' 4 x IO'" -10-9 [108], [113]-[115], [119], [I201
27 14 10.5 - 4 0.62 1.2 x 10s [I151

Ge: Hg(Sb) 35 4.2 14 II 5 x 105 1.8 x IO'" 3 x I O - ~ Q -x ~1 0 - 9

3 x 10-10-3 x 10-9

Ge: Cu 17 4.2 27 23 - 4 0.2-0.6 2 X 104 2 4 x 10'0 3x10-9-10-*

20 4x -1.3 x l o - '

Ge: Cu(Sb) 2717 4.2 23 2 x 105 2 x 10'0 <2.2 x 10-9

Hg, ..Cd,Tt 12 77 14 -10' 0.054.3 6&4m 10'0 410-6

x =0.2 20-200 6 x IOLo < 4 x la-.)

Pb, .Sn,Te 71 I1 IO -10' 42 3 x 10' 1.5 x 10.6

x=0.174.2 4.2 15 14 52 1.7 x IO'O 1.2 x 10-63

Table IV liststhe properties of infrared photoconductors excitation of holes from impurity centers such as Au[llO],
in further detail. For wavelengths up to 3.5 to 5.7 p,the [111],Cu [4], [112], [113],Hg [113]-[115],orCd [116],into
use of cooled intrinsic InAs and InSb is indicated. For the the valence band of a p-type semiconductor. Ge isused
longer wavelengths, specifically 10.6 pm, extrinsic and in- principally as the host crystal but Ge-Si [lo81 and Si [117]
trinsic photoconductorsare available and willnow be have also been investigated. In order to keep competing
discussed. thermal excitation low, extrinsic photoconductorsare
a) Extrinsic infrared photoconductors: Extrinsic infra- cooled to temperatures between 77 and 4.2"K (see Table
red photoconductors [4]-[6],[108], [ l o ] rely on optical IV). Since optical absorption constants in the wavelength
1480 PROCEEDINGS OF THE WEE,OCTOBER 1970

8 d
1
I
lo13 ION
DOWOR CONCENTRATIW IN CM3
Fig. 16. Dependence of hole lifetime on antimonycompensation in
Ge:Au (after Bratt [128]). The bottom point is from [120].

TABLE V
SWY OF RESULTS ON DETECTION
INFRARED HETERODYNE

Measured Measured Calculated

Material ~ %zT Laser

Wavelength
Noise
Equiv
(pm) FactortPowert
Quantum
Noise
Quantum
Eficiency
Noise
Equiv
Power
Temper-
atnre
(OK)
Response
Time
(Seconds)
Band-
width
Dynamic
Resistance
(a)
Ref,
(WiHz) (dB) (WiW
I

InAs I pv 3.39 1.25 x 10-19 3.3 300

- 3 x 10-19
0.25 7.1 300

InSb 0.5 77 10-8

I 0.6328i
3.39
Ge: Au

&:Hg
II PC
3.39

10.6 1.73 x 9.7

0.08

0.5 7.5 x lo-''

77

4.2
10-8

3.3~ 50 MHz
~431

~41,[1441

Ge: Hg(Sb) PC 3.39 4.2 <3.7x lo-'' >450 MHz [41, [lo21

i
~

Ge:Cu PC 10.6 10-19S

1.38.4
x 0.5 7 . 5 lo-''
~ 4.2 2 x 10-9

1i
x7 10-20Q 5.7

Ge:Cu(Sb) PC 10.6 7.5 x lo-'' 6 0.56 6.7~

lo-'' 4.2 lo-''
x 2 1.5 GHz 1.2 x lo3 [4], [17],
[I461

Hg -,Cd,Te PC 10.6x 7 77 5.7 lo-'

OS-2.5~ 1-10 MHz 200 [133]
~

Hg, -,Cd,Te 1 PV 10.6 2.210.7

x 10-19 0.09 2.1 x10-19 77 >50 MHz 190 [138]

Pb,-,Sn,Se ~
PV 10.6 2.5 x 1 0 - 1 9 ~ 11.2 0.085 2 2 x lo-'' 77 2 x 10-8
1.0 x 10-'8S 17.3

* PV = photovoltaic; PC =photoconductive
t Quantum noise limit h v = 1.87 x and 5.85 x lo-'' W/Hz at 10.6 and 3.39 pm,respectively.
Q = sensitivity calculated using LO-generated noise only.
S = measured overall receiver sensitivity.

range of extrinsic carrier excitation are generallylow peak quantumefficiency in the 0.2 to 0.6 range [102], [113F
(typically of the order of 1 to 10 cm-'), [127], the require- [115],
[119],
[120]. The usefulness of Ge:Au (and
ment for high quantum efficiency leads to detector dimen- Ge :Au(Sb)) at 10.6 pm is limited by lowquantum efficiency
sions in the direction of light incidence of several milli- (< at this wavelength but it is used in the laboratory
meters. because it operates at liquid N2 temperature.
Although Ge :Hg and Ge :Cu have drastically different For wide-band applications, short carrier lifetimes can be
cutoff wavelengths of 14 and 27 pm, respectively, their D* obtained in impurity-activated germanium by compensa-
at 10.6 pm are close [113]. Time constants for uncompen- tion with antimony donors [102], [112], [119], [128], [129].
sated samples are typically lo-' to 3 x seconds,with The measuredshortening of the hole lifetime withincreasing
 MELCHIOR et ai.: PHOTODETECTORS €OR OPTICAL COMMUNICATION 1481

compensation by donors in p-type Ge :Au is shown in Fig. Although the type of the detector chosen for each system
16 [128]. Compensation initially has only a small effect on will depend upon many factors, the characteristics of well-
D* as the data in Table IV for Ge :Au(Sb), Ge :Hg(Sb), and suited detectors for each system will now be discussed.
Ge :Cu(Sb) show. Compensation increases the resistance of
the detector [129], thereby reducing the direct detection A. Argon and Doubled YAG Luser Systems
circuit bandwidth unless a low R , with aconcomittant Optical communication systems utilizing argon or
reduction in sensitivity is used. The usable degree of com- doubled YAG lasers are likely to use photomultipliers in a
pensation has a further limit in that quantum efficiency direct detection mode of operation because of high current
decreases in proportion to ( N A- ND), where NA, N,=ac- gain and low darkcurrent. These wavelengths are very
ceptor anddonor concentrations, respectively [129]. N,, near to the peak response of S-20 or S-25 photosurfaces.
typically has avalue ofthe order of 1016cm-3 [108].The use- Slight changes in the process used to form these photosur-
fulness of compensation is indicated in Table IV (and in faces can yield small changes in the wavelength of maximum
Table V) where measured lifetimevalues as short as response, thus allowing photomultipliers to be optimized
3 x 10-lo and 2 x 10-lo seconds are given for Ge:Hg(Sb) for these systems.
and Ge: Cu(Sb), respectively. [4], 1171, [102]. Parameters affecting the type of photomultiplier to be
b) Intrinsic infrared photoconductors: Intrinsic photo- employed aremodulationbandwidthand optical signal
conductors utilize band-to-band excitation. Consequently, power available at the receiver. Systems designed for modu-
the light absorption coefficient is high, typically lo3 to lo4lation frequencies up to 100 MHz may employ conven-
cm- l , so that the dimension in the direction of light inci- tional electrostatically focused photomultipliers with the
dence need only be a few micrometers. Intrinsic .infra- number of secondary emission stages determined by the
red photoconductors are also listed in Table IV. In addition received signal power and desired (SIN) ratio. Systems with
to the 111-V photoconductors InAs and InSb with response a bandwidth of several hundred MHz will require a photo-
to a few micrometers, narrow bandgap intrinsic detectors multiplier in which both transit time dispersion in the elec-
have been fabricated from mixed crystals Hg,-,Cd,Te tron multiplier and output capacitance have been mini-
[4],[119], [123] and Pb,-,Sn,Te [4], [74],[126]. These mized. The use of a high-voltage accelerating grid between
detectors can be tailored to the desired wavelength byvary- each pair of dynodes, together with a coaxial anode struc-
ing the molar content x 141, [130]. Hg,-,Cd,Te detectors ture such as employed in the RCA C70045 photomultiplier
have been fabricated for operation at wavelengths from 1 to is one method for achieving the necessary modulation
30 pm [131], [132]. Emphasis has been on the 8- to 13-pm bandwidth. The use of lugh gain GaP dynode material re-
atmospheric window. Here their technical importance duces both the excess shot noise and required number of
comes about because they operate at higher temperatures, stages and it would be desirable for this application.
that is 77" to beyond 110°K[4], [125] than the extrinsic Veryhigh data rate systems with modulationband-
photoconductors. As Tables IV and V show, response times widths of several GHz impose the most difficult require-
as short as 50 and 15 ns have been obtained in photoconduc- ments upon the photomultiplier. High interstage voltages
tive Hg,-,Cd,Te and Pb,-,Sn,Te, respectively [4], [131]- must be used in the electron multiplier to minimize transit
[133]. time dispersion due to initial velocity distributions and the
Background limited (BLIP) detectivity has been achieved spatial variation of electric fields must beminimized to
in Hg,-.Cd,Te [125]. Although the possible design of fastachieve uniform interstage electron transit times. The mag-
photoconductive Hg, -,Cd,Te detectors has been consid- netically focused crossed-field photomultiplier, such as
ered in detail [134], [135] photovoltaic Hg,-,Cd,Te and the Sylvania 502, can provide a useful bandwidth of several
Pb, -,Sn,Te detectors are demonstratinggreater capability GHz because of both extremely low transit time dispersion
in meeting the requirements of MHz bandwidth optical and a well-matched signal output circuit. The use of high-
communication systems [8]. Bandwidths of 30 MHzto gain dynodes would again be verydesirable to reduce excess
greater than 50 MHz have been obtained in both direct de- shot noise.
tection [136], [137] and heterodyne detection [138] modes. Silicon avalanche photodiodes are also useful at these
wavelengths. Their quantum efficiencyishigher than in
IV. APPLICATIONS OF PHOTODETECTORSTO VARIOUS photomultipliers and theresponse and currentgain mecha-
OPTICAL COMMUNICATION SYSTEMS nism are sufficiently fast. However, unless cooled, noise and
dark current are larger thus limiting their sensitivityfor
The most important parameter in the choice of a photo-
direct detection of weak signals. The higher quantum effi-
detector is the wavelength of the optical communication
ciency of silicon diodes would make them useful in hetero-
system. The current state of the art in high efficiency lugh
dyne detection systems where the preservation of optical
power lasers leads to four principal possible system wave-
frequency and phase information and rejection of inter-
length regions :
ference may beof importance.
1) 4880 A-Argon, and 5300 A-doubled YAG
2 ) 850CL9000 A-GaAs B. GaAs Systems
pm-YAG
3) 1.06 Optical communication systems utilizing GaAs the
4) 10.6 pm-CO,. (840CL9000 A) laser operate at wavelengths where both
1482 PROCEEDINGS OF THE IEEE, OCTOBER 1970

photomultipliers and solid-state photodiodes may be util- marized in Table V. Results are listed for 10.6 pm (CO,)
ized. Small light-weight systems designed for short range and 3.39 pm (HeNe). Sensitivities within 3 to 10 dB of the
communication may employ silicon photodiodes, whle quantum noiselimit (hv= 1.87 x W/Hz at 10.6 pm)
systems requiring maximum sensitivity will utilize photo- havebeen demonstrated.Experimental sensitivityvalues
multipliers or silicon avalanche photodiodes. in TableV are marked as to whether they take into account
Photomultipliers with an S-25 or one of the new photo- the total receiver noise (9,or only the LO-generatednoise
surfaces such as GaAs: Cs20offer reasonable sensitivity (Q). With improved heterodyne receiver design, the quan-
at these wavelengths. Wide bandwidths may be achievedby tum noise limit willbe more closely approached in the
incorporating these cathodes into fast photomultiplier future. In the case of photovoltaic mixed crystal materials,
structures. For silicon photodiodes and avalanche photo- particularly HgCdTe,heterodyne sensitivityisbeing in-
diodes, a tradeoff exists between quantum efficiency and creased as improved quantum efficiency elements become
bandwidth as mentioned in Section 111-B. However, quan- available.
tum efficiencies of 90percent have been achievedfor speeds As Table V shows, receiversensitivity near W/Hz
of response of 10 ns [65]. Extensive development activity simultaneously with a bandwidth exceeding 1 GHz has
in many laboratories on both GaAs :Cs20 and other type been obtained using highly compensated photoconductive
111-V photoemitters and on silicon avalanche photodiodes Ge: Cu(Sb) witha time constant of200 ps [5]. The genera-
is expected to result in practical photomultipliers and av- tion-recombination noise spectrum of t h s mixer element
alanchephotodiodeswith -high sensitivity in this wave- measured upto 4 GHz is shown inFig. 17, when illuminated
length region. with sufficient LO power to depress mixer resistance from
5 x lo5 to 1.2 x lo3 ohms. Due to the available conversion
C. YAG Systems gain, heterodyneoperation is obtained well beyondthe
Photomultipliers are presently not veryuseful for the 750-MHz rollover frequency. The measuredvariation of the
demodulation of 1.06-pm Nd: YAG laser radiation. This heterodyne noise equivalent power (SIN= 1) with dc bias,
is becauseeven the best developmental infrared photo- and hence conversiongain, isshown in Fig. 18..The behavior
cathodes have quantum efficiencies of only 0.8 percent at shown is consistent with (7), which for a photoconductive
1.06 pm [43]. Siliconphotodiodes which combine quantum mixer withhigh dark resistance and low backgroundirradi-
efficiencies and response times of 75 percent and 25 m. or ation canbe written [17]
45 percent and 10 ns are available [65]. P h o t o m u l t i p k
and Si avalanche photcadigdes have recently been cornpazed
theoretically as detectam for laser pulses of lo-’ twlO-’
duration at 1.06 pm [I391 and it has been found that Si
avalanche diodes can detect one to two orders of magnitude where the available conversion gain due to the local
lower pulse energies than the presently available photomul- oscillator is given by
tipliers. However, avalanche photodiodes with optimized
properties for 1.06 pm are not yet available. Detection sys-
tems for 1.06 pm with signal bandwidths in excess of 100
MHz would thus use germanium avalanche photodiodes Note thatfor this photoconductive case G is independent of
which, in order toreduce dark current,might be cooled to LO power variations since mixer conductance is directly
about 250°K. proportional toPLo (18). Expressions similar to the preced-
ing ones but for the photodiode heterodynecase are given
D . CO, Laser Systems .. in [138].
Bothphotuconductiveand photovoltaic detectors are. For systems with bandwidths up to 100 MHz, uncom-
useful for the detection of 10.6-pm radiation. Their charac- pensated extrinsic photoconductors have demonstrated
teristics were given in Tables I1 and IV. Direct detection is sensitive heterodyne performance- [dl, [18],[144], 11451.
useful in 10.6-pmsystemswheresimplicity but not high PhotoconductiveHgCdTe has given performance to 10
sensitivity are desired. Heterodyne detection is preferred MHz [133]. Photovoltaic HgCdTe has emerged as a strong
because it combines high sensitivity with widebandwidths. competitor offering incombinationgoodfrequency re-
Heterodyne detection is most practical atthe longer sponse, operation at moderatecryogenic temperatures, and
wavelengths due to such factors as: sensitivity improve- low power dissipation [ 1381. As problems in mixed crystal
ment due to reduced quantum noise (see (7)), easier align- fabrication are overcome, the use of 10.6-pm photodiode
ment of signal and local oscillator, and greater diffraction- mixers will be extended to higher modulation frequencies.
limited acceptance angle for a given aperture size. The CO, Heterodyne operation requires the use of a laser local
laser has provencapable ofsingle-frequencystabilized oscillator which is offset bythe I F frequency, and frequency-
operation as is desired for the LO [133]. stabilized by means of a circuit such as Fig. 19(a). The LO
Sensitivities in coherent detection approaching the theo- could be a CO; or possibly a PbSnTe laser [142]. In systems
retical quantum noise limit have been demonstrated in a where s i d c a n t Doppier shifts occur between transmitter
variety of photodetectors. The results obtained are sum- and receiver, the use of the receiver conliguration of Fig.
 MELCHIOR et a/.:
FOR PHOTODETECrORS OPTICAL COMMUNICATION 1483

-80 erty has been usedto combine infrared heterodyne elements

1
m
to form spatially coherent arrays at 10.6pm with directional
0

L
- 90 properties exactly analogous to those obtained with multi-
K
W feed microwave antennas [146]. These coherent arrays are
-100 capable of providing a multiplicity of resolution elements
Y
-
0 in the object field ofthe optical receiver for such purposes as
7 -110 spatial acquisition and tracking [8], [ 1521.
Y
-> Pyroelectric detectors with response times as short as a
"
E -120
few nanoseconds have been reported [153]. They can be
u.
W operated at room temperature, but such thermal detectors
-130
have limited sensitivity in both direct detection [4], [154],
1 0 0 200 300 400 500600 800 Kxx)
FREOUENCr IN MHz [ M I and heterodyne detection [156]-[158] modes. High
Fig. 17. Measured g-r noise spectrum of Ge:Cu(Sb)mixer [17]. speed detection using metal point contacts with a time con-
stant of 3 x seconds has also recently been reported
[159].
2.8 a 1 0 - l ~
I
V. CONCLUSIONS
Z 2.4 x 10.~9 It is evident that there is a large number of photodetectors
4

L 2.0 a ~ 0 - 1 9
and detection techniques available for use in optical com-
i!
z
1.6 x IO-''
munication systems. The correct choice of a particular de-
tection technique and photodetector depends mainly on
Y
the wavelength of operation, the information bandwidth
1.2 x 10-l'
8 of the signal, and on the sensitivity desired for the de-
W
$j 8 x10-20
modulation of weak signals. Direct detection is preferred for
wavelengths at which photomultipliers and avalanche
4 a 10-20
RELATIVE BIAS
POWER
photodiodes with internalcurrent gain, wide electrical
Fig. 18. Measured dependence of sensitivity of 10.6 pm heterodynere- bandwidth, and low excess noise are available.
ceiver on dc bias power (after [4] and [17]. (Relative bias power has Photomultipliers and silicon avalanche photodiodes are
linear scale.) As bias power is increasedgeneration-recombination useful in the ultraviolet, visible, and near infrared to about
noise swamps out all other receiver noise contributions.
1 pm, germanium avalanche diodes extend this range to be-
OPTICAL
SlGNAL , - , INFRARED
SIGNAL
OUTPUT
w
yond 1.5 pm. Photomultipliers provide high-gain and low-
excessnoise.Specially constructed photomultipliers have

I
MIXER
I F WPL achieved gains of lo4 for baseband widths up to several

TUNABLE
LASER LO
- - Low-pAss

AND DRIVER
LIMITER-
DISCRIMINATOR
*
GHz. Avalanche photodiodes also combine gain and high
speed of response as manifested by current gain-bandwidth
products of up to 100 GHz. The multiplication process in
avalanche diodes, however, results in higher excess noise.

OPTICAL
SIGNAL 1 INFRAREO 'IDEBAND
*
-
.
NARROWBAND
IF AMPL
OUTPUT
-
SIGN& But despite t h s excess noise, current gain in both photo-
multipliers and avalanche photodiodes allows realization
of direct detection systems with wide bandwidths and con-
LOW-NOISE
I F AMPL
-+ I F MIXER 4)
siderably higher sensitivities to weak light signals as com-
pared to detection systems without gain that are limited by
I I
thermal noise of the detector and output circuit. Although

.".( LASER L 0

(b)
LOW- PASS
FILTER
AND M V E R
quite desirable, the development of avalanche photodiodes
with response at longer wavelengths from materials like
InAs, InSb, andeven mixedcrystals is stillat anearly stage.
Photoconductors with wide bandwidths have been de-
Fig. 19. Methods of frequency tracking in optical heterodyne receivers. veloped for theinfrared region, but the lack of sufficient in-
(a) Optical tracking. (b) IF tracking. ternal current gain limits their performance in direct de-
tection systems.
19(b) is useful, and has been demonstrated at 10.6 pm [148] Heterodyne detection in bothphotoconductors and
in a quantumnoise limited 1.3-GHz bandwidth heterodyne photodiodes is most successfully used at infrared wave-
receiver. Here frequency tracking is accomplished at IF and lengths where no fast detectors with sufficient internal cur-
the information bandwidthestablished in the narrow-band rent gain are available. For 10.6 pm, sensitivitiesclose to the
second I F amplifier. theoretical quantum noise limit (hvB)have been realized in
The optical heterodyne detector behaves simultaneously heterodyne detection systems with bandwidths approach-
as a receiver and an antenna [33], [149]-[151]. This prop- ing 1 GHz.
1484 PROCEEDINGS OF THE IEEE, OCTOBER 1970

ACKNOWLEDGMENT parametric processes. I,” Phys.Rev., vol. 124, pp. 1646-1654,

December 1961.
Theauthors wish tothank D. Bode and P. Bratt of [24] J. E. Midwinter and J. Warner, “Up-conversion of near infrared to
Santa Barbara Research Center; E. Sard, B. Peyton, and visible radiation in lithium-metaniobate,” J . Appl. Phys., vol. 38,
pp. 519-523, February 1967.
F. Pace of AIL; for supplying useful data and suggestions [25] J. Warner, “Photomultiplier detection of 10.6 pm radiation using
for this paper. optical up-conversion in proustite,” Appl. Phys. Lett., vol. 12, pp.
222-224, March 1968.
[26] G. D.Boyd, T.J. Bridges, and E. G . Burkhardt, “Up-conversion of
REFERENCES
10.6 p radiation to the visible and second harmonic generation in
L. K. Anderson and B. J. McMurtry, “High-speed photodetectors,” HgS,” IEEE J . Quantum Electron., vol. QE-4, pp. 515-519, Septem-
Proc. IEEE, vol. 54, pp. 1335-1349, October 1966. ber 1968.
W. R.Pratt, Laser Communication Systems. New York: Wiley, [27] D. A. Kleinman and G. D. Boyd, “Infrared detection by optical
1969. mixing,” J . Appl. Phys.,vol. 40,pp. 546-566, February 1969.
M. Ross, Laser Receivers. New York: Wiley, 1967. [28] Y. Klinger and F. Arams, “Infrared 10.6 micron CW up-conversion
A. Beer and R. Willardson, Eds., Semiconductors and Semimetab, in proustite using an Nd: YAGlaser pump,” Proc. ZEEE, vol. 57,
vol. 5. Infrared Detectors. The following chapters are of special in- pp. 1797-1798, October 1969.
terest: P. Kruse, “Indium antimonide photoconductive and PEM [29] G. Lucovsky and R. B. Emmons, “High frequency photodiodes,”
detectors,”; D. Long and J. Schmit, “Hg,-,Cd,Te and closely re- Appl. Opt.,vol. 4, pp. 697-702, June 1965.
lated alloys as intrinsic IR detector materials,”; I. Melngailis and [30] H. Melchior and W. T. Lynch, “Signal and noise response of high
T. Harman, “Single crystal lead-tin chalcogenides,”; R. Keyes and speed germanium avalanche photodiodes,” IEEE Trans. Electron
T. Quist, “Low-level coherent and incoherent detection in the in- Devices, vol. ED-13, pp. 829-838, December 1966.
frared,”; M. Teich, “Coherent detection in theinfrared,”; F. [31] M. DiDomenico, Jr., and 0. Svelto, “Solid-state photodetection:
Arams, E. Sard, B. Peyton, and F. Pace, “Infrared heterodyne de- A comparison between photodiodes and photoconductors,” Proc.
tection with gigahertz IF response,”; E. Putley, “Pyroelectric de- IEEE, vol. 52, pp. 136-144, February 1964.
tectors,”; H. Sommers, “Microwave biased photoconductive [32] G. Lucovsky, R. B. Emmons, B. Harned, and J. K. Powers, “De-
detector.“ tection of coherent light by heterodyne techniques using solid state
P. W. Kruse, L. D. McGlauchlin, and R.B. McQuistan, Elements photodiodes,” in Quantum Electronics III. P. Grivet and N. Bloem-
of Infrared Technology. New York: Wiley, 1962. bergen, Eds. New York: Columbia University Press, 1964, pp.
R. Hudson, InfraredSystem Engineering. New York: Wiley, 1969. 1731-1738.
T. M. Ouist, “Comerdomd germanium detectors,” Proc. IEEE [33] A. E. Siegman, “The antenna properties of optical heterodyne re-
p
.i
(Letter$, vol. 56, 1212-1213, July 1968. ceivers,” Proc. ZEEE, vol. 54, pp. 1350-1356, October 1966.
[8] G. McElroy, McAvoy, H. Richard, T. McGunigal, and G. ScWner, [34] A. J. Bahr, “The effect of polarization selectivity on optical mixing
“Carbon dioxide laser communication systems for near earth ap- photoelectric surfaces,” Proc. IEEE (Correspondence), vol. 53, p.
plications,” to be published, and H. Plotkin and J. Randall, “Sys- 513, May 1965.
terns aspects of optical space communications,” to be published. [35] 0. DeLange, “Optical heterodyne detection,” ZEEE Spectrum, vol.
[9] A. M. Johnson, “Square law behavior of photocathodes at high 5, pp. 77-85, October 1968.
light intensities and high frequencies,” IEEE J. Quantum Electron., [36] A. H. Sommer and W. E.’Spicer, Photoelectronic Materials and De-
vol. QE-1, pp. 99-101, May 1965. vices. Princeton, N. J.: Van Nostrand, 1965.
[IO] B. J. McMurtry and A. E. Siegmann, “Photomixing experiments [37] A. H. Sommer, Photoemissive Materials. New York: Wiley, 1968.
with a ruby optical maser and a traveling-wave microwave photo- [38] J. J. Scheer and J. van Laar, “GaAs-Cs-a new type of photo-
tube,” Appl. Opt.,vol. 1, pp. 51-53, January 1962. emitter,” Solid-State Commun., vol. 3, pp. 189-193, August 1965.
[ l l ] B. M. Oliver, “Signal-to-noise ratios in photoelectric mixing,” [39] A. A. Turnbilland G.B. Evans, “Photoemission fromGaAs-0-0,”
Proc.IRE (Correspondence), vol. 49, pp. 1961961, December Brit. J . Appl. Phys., ser. 2, vol. 1, pp. 155-160, February 1968.
1961. [40] J. J. Uebbing and R. L. Bell, “Improved photoemitters using GaAs
[12] H. A. Hans, C. H. Townes, and B. M. Oliver, “Comments on noise and InGaAs,” Proc. IEEE (Letters), vol. 56, pp. 1624-1625, Sep-
in photoelectric mixing,’’ Proc. IRE(Letters), vol. 50, pp. 1544-1 546, tember 1968.
June 1962.
[41] R. L. Bell and J. J. Uebbing, “Photoemission from InP-Cs-0,”
[13] B. M. Oliver, “Thermal and quantum noise,” Proc. IEEE, vol. 53,
Appl. Phys. Lett.,vol. 12, pp. 76-78, February 1968.
pp. 436454, May 1965.
[42] B. F. Williams, “InGaAs-CsO, a low work function (less than 1.0
[14] S . Jacobsand P. Rabinowitq “Optical heterodyningwitha CW g w -
eV) photoemitter,” Appl. Phys. Lett., vol. 14, pp. 27S275, May
ous laser,” in Quantum Mechanics ZII, P. Grivet and N. Bloember-
gen Eds. New York: Columbia University Press, 1964, pp. 481 1969.A l s o , B. F. Williams et al., Appl. Phys. Letters,
to be published.
487. [43] H. Sonnenberg, “InAsP-Cs,O, a high efficiency infrared photo-
[15] G. Lucovsky, M. E. Lasser, and R. B. Emmons, “Coherent light cathode,” Appl. Phys. Lett., vol. 16, pp. 245-246, March 1970.
detection in solid-state photodiodes,’’ Proc. IEEE, vol. 51, pp. 166- [44] R. E. Simon, A. H. Sommer, J. J.Tietjen, and B. F. Williams, “New
172, January 1963. high-gain dynode for photomultipliers,” Appl. Phys. Lett., vol. 13,
[16] 0. Svelto, P. D. Coleman, M. DiDomenico, and R. H. Pantell, pp. 35S357, November 1968.
“Photoconductive mixing in CdSe single crystals,” J . Appl. Phys., [45] R. C.Miller and N. C. Wittwer, “Secondary emission amplification
vol. 34, pp. 3182-3186, November 1963. at microwave frequencies, IEEE J . Quantum Electron. vol. QE-1,
[I71 F. R. Arams, E. W. Sard, B. J. Peyton, and F. P. Pace, “Infrared pp. 49-59, April 1965.
10.6 micon heterodynedetection with gigahertz IF capability,” (461 M. B. Fisher and R. T. McKenzie, “A traveling-wave photomulti-
IEEE J . Quantum Electron., vol. QE-3, pp. 484-492, November piier,” IEEE J . Quantum Electron., vol. QE-2, pp. 322-327, August
1967; see also IEEE Spectrum, vol. 5, p. 5, June-July 1968. 1966.
[18] M. C. Teich, “Infrared heterodyne detection,” Proc. IEEE, vol. 56, [47] W. Shockley and J.R. Pierce, “A theory of noise for electron multi-
pp. 3 7 4 , January 1968. pliers,” Proc. IRE, vol. 26, pp. 321-332, March 1938.
[19] J. E. Geusic and H. E. D. Scovil, “A unidirectional traveling-wave [48] R. M. Matheson, “Recent photomultiplier developments at RCA,”
opticalmaser,” BeNSyst. Tech.J., vol. 41, pp. 1371-1397, July 1962. IEEE Tram. Nucl. Sci., vol. NS-11, pp. 64-71, June 1964.
[20] H. Kogelnik and A. Yariv, “Considerations of noise and schemes [49] J. R. Kerr, private communication.
for its reduction in laser amplifiers,” Proc. IEEE, vol. 52, pp. 165- [50] D. E. Persyk, “New photomultiplier detectors for laser applica-
172, February 1964. tions,” Laser J . , pp. 21-23, November-December 1969.
[21] F. Arams andM. Wang, “Infrared laser preamplifier system,” [51] W. W. Gaertner, “Depletion-layer photoeffects in semiconductors,”
Proc. IEEE (Correspondence), vol. 53, p. 329, March 1965. Phys. Rev., vol. 116, pp. 84-87, October 1959.
[22] G. C. Holst and E. Snitzer, “Detection with a fiber laser preampli- [52] D. E. Sawyer and R. H. Rediker, “Narrow base germanium photo-
fier at 1.06 p,” IEEE 1. Quantum Electron., vol. QE-5, pp. 319-320, diodes,”Proc. IRE, vol. 4 6 , pp. 1122-1 130, June 1958.
June 1969. (531 R. P. Riesz, “High-speed semiconductor photodiodes,” Rev. Sei.
[23] W. H. Louise11 and A. Yariv, “Quantum fluctuations and noise in Instr., vol. 33, pp. 994998, September 1962.
MELCHIOR er al.: PHOTODETECTORS FOR OPTICAL COMMUNICATION 1485

[54] L. K. Anderson,“Photodiode detection,” Proc. I943 Symp. on J. J. Chang, “Frequency response of PIN avalanche photodiodes,”
Optical Masers, Polytechnic Press, pp. 549-563. IEEETrans. Electron Devices, vol. ED-14, pp. 139-145, March
(551W. M. Sharpless, “Evaluation of a specially designed GaAs 1967.
Schottky-barrier photodiode using 6328 A radiation modulated at C. A.Lee, R. A. Logan, J. J.Kleimack, and W. Wiegrnann, “Ioniza-
4 GHz,” Appl. Opt.,vol. 9, pp. 489494,February 1970. tion rates of holes and electrons in silicon,” Phys. Rev., vol. 134,
[56] M. V. Schneider, “Schottky barrier photodiode with antireflection pp. A761-A773, May 1964.
coating,” Bell Syst. Tech. J., vol. 45, pp. 1611-1638, November S. Donati and V. Svelto, “The statistical behavior of the avalanche
1966. photodiode,” Alta Freq., vol. 37, pp. 476486, May 1968.
[57] United Detector Technology (manufacturers data). R. D. Baertsch, “Noise and multiplication measurements in InSb
[58] H. Melchior, M. P. Lepselter, and S. M. Sze, “Metal-semiconductor avalanche photodiodes,” J. Appl.Phys. vol. 38, pp. 42674274,
avalanche photodiode,’’ presented at IEEESolid-State Device Res. October 1967.
C o d , Baulder, Colo., June 1968. W. T. Lindley, R. J. Phelan, C. M. Wolfe, and A. G. Foyt, “GaAs
[59] R. D. Baertsch and J. R.Richardson, “An Ag-GaAs Schottky bar- Schottky bamer avalanche photodiodes,” Appl. Phys. Lett.,vol. 14,
rier ultraviolet detector,” J.Appl. Phys.,vol. 40, pp. 229-235, Janu- pp. 197-199, March 1969.
ary 1969. G. Lucovsky andR. B. Emmons, “Avalanche multiplication in
[60] J. R. Richardson and R. D.Baertsch, “Zinc sulfide Schottky bamer InAs photodiodes,” Proc. IEEE (Correspondence), vol. 53, p. 180,
ultraviolet detectors,” Solid-State Electron., vol. 12, pp. 393-397, February 1965.
May 1969. R. J. Locker and G. C. Huth, “A new ionizing radiation detection
[61] L. K. Anderson, P. G. McMullin, L. A. D’Asaro, and A. Goetz- concept which employs semiconductor avalanche amplification and
berger, “Microwave photodiodes exhibiting microplasma-free car- the tunnel diode element,” Appl. Phys. Lett., vol. 9, pp. 227-230,
rier multiplication,” Appl. Phys. Lett., vol. 6, pp. 62-64, February September 1966.
1965. H. Ruegg, “An optimized avalanche photodiode,” IEEE Trans.
[62] E. Labate, private communication. Electron Devices, vol. ED-14, pp. 239-251, May 1967.
[63] R. J. McIntyre and H. C. Sprigings, “Multielement silicon photo- J. R. Biard and W. N. Shaudeld, “A model of the avalanche
diodes for detection at 1.06 microns,” presented at the C o d . on photodiode,’’ IEEE Trans. Electron Decices, vol. ED-14, pp. 233-
Prep. Control of Electron. Mat., Boston, Mass., August 1966. 238, May 1967.
[64] H. C.Sprigings and R. J. McIntyre, “Improved multielement silicon W. N. Shaunlield and J.R.Biard, “A high-speed germanium photo-
photodiodes fordetection of 1.06pm,” presented at the Int. Electron detector for 1.54 microns,” presented at the Solid-state Sensors
Devices Meet., Washington, D. C., October 1968. Symp., Minrreapolis, Minn., September 1968.
[65] RCA, Montreal, Canada (manufacturers data). W. T. Lynch, “Elimination of the guard ring in uniform avalanche
[66] D. P. Mathur, R. J.McIntyre, and P. P. Webb, “A new germanium photodiodes,’’ IEEE Trans. Electron Deaices, vol. ED-15, pp. 7 3 s
photodiode with extended long-wavelength response,” presented 741, October 1968.
at the Int. Electron Device Meet., Washington, D. C., October 1968. F. D. Shepherd, Jr., A. C . Yang, and R. W. Taylor, “A 1 to 2 pm
[67] B. R. Pagel and R. L. Petritz, “Noise in InSbphotodiodes,” J. silicon avalanche photodiode,” Proc. IEEE (Letters), vol. 58, pp.
Appl. Phys., vol. 32, pp. 1901-1904, October 1961. 1160-1 162, July 1970.
[68] H. Protschka and D. C. Shang, “InSb photodiodes with high re- A. Rose, Concepts in Photoconductivity and AppliedProblems.
verse breakdown voltage,” presented at the Int. Electron Device New York: Interscience, 1963.
Meet., Washington, D. C . , October 1967. A. Rose, “Maximumperformance of photoconductors,” Helc.
[69] A. Kohn and J. Schlickman, “1-2 micron (HgCd)Te photodetec- Phys. Acta, V O ~30, . pp. 242-244, August 1957.
tors,” IEEE Trans. Electron Decices, vol. ED-16, pp. 885-890, R. W. Reddington,“Gain-bandwidthproduct of photoconduc-
October 1960. tors,” Phys. Rec., vol. 115, pp. 894896, August, 1959.
[70] M. Rodot, C. Vtrit, Y. Marfaing, J. Besson, andH. Lebloch, P. D. Coleman, R. C. Eden, and J. N. Weaver, “Mixing and detec-
“Semiconductor lasers and fast detectors in the infrared (3 to 15 tion of coherent light in a bulk photoconductor,” IEEE Trans.
microns),” IEEE J. Quantum Electron., vol. QE-2, pp. 586-593, Electron Decices, vol. ED-11, pp. 488-497, November 1964.
September 1966. R. E.Burgess, “The statistics of charge carrier fluctuations in semi-
C. Vtrie and J. Ayas, “Cd,Hg,-,Te infrared photovoltaic detec- conductors,” Proc. Phys. Soc. London, (Gen.), vol. 69, pp. 1020-
tors,” Appl. Phys. Lett., vol. 10, pp. 241-243, May 1967. 1027.Qctober 1956.
I. Melngailis and A. R. Calawa, “Photovoltaic effect in Pb,Sn, -,Te [loo] K. M. Van Vliet, “Noise in semiconductors and photoconductors,”
diodes,” Appl. Phys. Lett., vol. 9, pp. 304-306, October 1966. Proc. IRE,vol. 4 6 , pp. 10041018, June1968.
J. F. Butler, A. R. Calawa, I. Melngailis, T. C. Harman, and J. 0. [loll H. Sommer and E. K. Gatchell, “Demodulation of low-levelbroad-
Dimmock, “Laser action and photovoltaic effect in Pb,-.&,Se band optical signals with semiconductors,” Proc.IEEE, vol.54,
diodes,” Bull. Am. Phys. Soc.,vol. 12, p. 384, March 1967. pp. 1553-1568, November 1966.
I. Melngailis, “Laser action and photodetection in lead-tin chalco- [I021 C. Buczek and G. Picus, “Far infrared laser receiver investigation,”
genides,” J. de Physique, vol. 29, colloque C4, supplement au no. Hughes Res. Labs., Malibu, Calif., A. F. Rept. AFAL-TR-68-102,
11-12, pp. C4-8444-94, November-December 1968. May 1968.
K. M. Johnson, “High-speed photodiode signal enhancement at [lo31 A. Yariv, C. Buczek, and G . Picus, “Recombination studies of hot
avalanche breakdown voltage,” IEEE Trans. Electron Decices, vol. holes in mercury-doped germanium,” Proc.Int. Conf Phys. of
ED-12, pp. 55-63, February 1965.
A. S. Tager, “Current fluctuations in a semiconductor (dielectric) Semicond. (Moscow), vol. 9-10, p. 500, July 1968.
under conditions of impact ionization and avalanche breakdown,” [lo41 R. L. Williams, “Response characteristics of extrinsic photocon-
SOU. Phys.-SolidState, vol. 6, pp. 1919-1925, February 1965. ductors,” J. Appl. Phys.,vol. 40, pp. 184-192, January 1969.
R. J. McIntyre, “Avalanche multiplication noise in semiconductor [lo51 M. DiDomenico and L. K. Anderson, “Microwave signal-to-noise
junctions,” IEEE Trans.Electron Devices, vol. ED-13, pp. 164-175, performance of CdSe bulk photoconductivedetectors,” Proc.
January 1966. IEEE, vol. 52, pp. 815-822, July 1964.
H. Melchior and L. K. Anderson, “Noise in high speed avalanche [I061 H. Sommer and W. Teutsch, “Demodulation of low-level broad-
photodiodes,’’ presented at the Int. Electron Device Meet., Wash- band optical signals with semiconductors; pt. I1 analysis of photo-
ington, D. C., October 1965. conductive detector,” Proc. IEEE, vol. 52, pp. 1 6 1 5 3 , February
R. D.Baertsch, “Noise and ionization rate measurements in silicon 1964.
photodiodes,” IEEE Trans.Electron Devices (Correspondence), vol. [lo71 C . Sun and T.E. Walsh, “A solid-state microwave-biased photo-
ED-13, p.987, December 1966. conductive detector for 10.6 p n , ” IEEE J. Quantum Electron., vol.
W. T. Read, “A proposed high-frequency negative resistance QE-5, pp. 320-321, June 1969.
diode,” Bell Syst. Tech. J., vol. 37, pp. 401-446, March 1958. [lo81 H. Levinstein, “Extrinsic detectors,” Appl. Optics, vol. 4, pp. 639-
R. B. Emmons and G. Lucovsky, “The frequency response of ava- 647, June 1965.
lanching photodiodes,” IEEE Trans. Electron Devices, vol. ED-13, [lo91 E. Putley, “Far infrared photoconductivity,” Phys. Status Solidi,
pp. 297-305, March 1966. vol. 6, pp. 571-614, September 1964.
R. B. Emmons,“Avalanche-photodiode frequency response,” J. [110] Johnson and H. Levinstein, “Infrared properties of gold in ger-
Appl. Phys., vol. 38, pp.3705-3714, August 1967. manium,” Phys. Rev., vol. 117, pp. 1191-1203, March 1960.
 PROCEEDINGS OF THE IEEE, OCTOBER 1970
1486

[l 11] W. Beyer and H. Levinstein, “Cooled photoconductive infrared de- [135] D. Long, “On generation-recombination noise in infrared detection
tectors,” J . Opt. SOC. Am., vol. 49, pp. 686-692, July 1959. materials,” Infrared Phys., vol. 7, pp. 169-170, September 1967.
[112] G. S. Picus, “Carrier generation and recombination processes in [136] A. Sorrentino, “IR detectors, developments at SAT,” presented at
copperdoped germanium photoconductors,” J . Phys. Chem. Infrared Symp. Royal Radar Establishment, Malvern, England,
Solids, vol. 23, pp. 1753-1761, 1962. April 1969.
11131 D. Bode and H. Graham,“Comparison of performance of copper- [137] E. D. Hinkley and C. Fried, private communication.
doped germanium and mercury-doped germanium detectors,” In- [138] B. Peyton, E. Sard, R. Lange, and F. Arams, “Infrared 10.6-pm
frared Phys., vol. 3, pp. 129-137, September 1963. photodiode heterodyne detection,” this issue, pp. 1769-1770.
[114] S. Borrello andH. Levinstein, “Preparation and Properties of [139] R. J. Mcintyre, “Comparison of photomultipliers and avalanche
mercury doped germanium,” J . Appl. Phys.,vol. 33, pp. 2947-2950, photodiodes for laser applications,” IEEE Trans. Electron Devices,
October 1962. V O ~ .ED-17, pp. 347-352, April 1970.
[115] Y. Darviot, A. Sorrentino, B. Joly, and B. Pajot, “Metallurgy and [140] J. Hanlon and S. Jacobs,“Narrow-band optical heterodyne de-
physical properties ot’mercury-doped germanium related to the per- tection,” Appl. Opt., vol. 6, pp. 577-578, March 1967.
formance of the infrared detector,”Infrared Phys., vol. 7, pp. 1-10, [141] F. E. Goodwin and M. E. Pedinoff, “Application of CCI, and
March 1967. CCI, :CC1, ultrasonic modulators to infrared optical heterodyne
[116] P. Bratt, W. Engeler, H. Levinstein, A. Mac Rae, and J. Pehek, “A experiments,” Appl. Phys. Lett., vol. 8, pp. 60-61, February 1966..
status report on infrared detectors,” InfraredPhys., vol. 1, pp. [142] E. N. Fuls, “Optical frequency mixing in photoconductive InSb,”
27-38, March 1961. Appl. Phys. Lett., vol. 4, pp. 7-8, January 1964.
I117 R. Soref, “Extrinsic IR photoconductivity of Si doped with B, Al, [143] R. A. Wood, “Golddoped germanium as an infrared high-fre-
Ga, P, As or Sb,” J . Appl. Phys., vol. 38, pp. 5201-5209, December quencydetector,” J . Appl. Phys.,vol. 36, pp. 1490-1491, April 1965.
1967; see also, “Coherent homodyne detection at 10.6pm with an ex- [144] C. Buczek and G . Picus, “Heterodyne performance of mercury-
trinsic photoconductor,” Electron.Lett., vol. 2, pp. 410-411, doped germanium,” Appl. Phys. Lett.,vol. 11, pp. 12S126, August
November 1966. 1967.
[118 F. D. Morten and R. E. J. King, “Photoconductive indium anti- [145] M. Teich, R. Keyes, and R. Kingston, “Optimum heterodyne de-
monide detectors,” Appl. Opt., vol. 4, pp.659463, June 1965. tection at 10.6 microns in photoconductive Ge:Cu,” Appl. Phys.
[119 Manufacturers’ Data (Santa Barbara Research Center, Texas In- Lett., vol. 9, pp. 357-360, November 1966.
struments, Honeywell, Philco-Ford, MullardRaytheon, Socittt [146] F. Pace, R. Lange, B. Peyton, and F. Arams, “Infrared heterodyne
Anonyme de Ttlhmmunications). receiver array with wide IF bandwidth for CO, laser systems,” to
I1201 D. Bohe and P. Bratt, private communication. be published.
[121] T. Bridges, T. Chang, and P. Cheo, “Pulse response of electrooptic [147] E. Hinkley, T. Harman, and C. Freed, “Optical heterodyne detec-
modulators and photoconductivedetectors at 10.6 pn,” Appl. tion at 10.6 microns of the beat frequency between a tunable
Phys. Lett.,vol. 12, pp. 297-300, May 1968. PbSnTe diode laser and a C 0 2 gas laser,” Appl. Phys. Lett.,vol. 13,
[122] J. T. Yardley and C. B. Moore, “Response times 0fGe:Cu infrared pp. 49-51, July 1968.
detectors,” Appl. Phys. Lett., vol. 7, pp. 311-312, December 1965. [148] W. Chiou, T. Flattau, B. Peyton, and R. Lange, “High-sensitivity
[123] J. Schlickman, “Mercury cadmium telluride intrinsic photodetec- infrared heterodyne receivers with gigahertz IF bandwidth Pt. I11 :
tors,” Proc. Electroopt. Syst. Conf., Industr. Sci. Conf. Mgmt, Inc., Packaged 10 pm heterodyne 1.3 GHz bandwidth trackingreceiver,”
(Chicago, Ill.), pp. 289-309, September 1969. IEEE Spectrum, vol. 7, p. 5, January 1970 (Advertisement for AIL,
[124] P. W. Kruse, “Photon effects in Hg,-,Cd,Te,” Appl. Opt., vol. 4, Div. Cutler-Hammer, Deer Park, N. Y.).
pp. 687692, June 1965. [149] A. E. Siegman, “A maximum-signal theorem for the spatially co-
[125] B. Bartlett, D. Charlton, W. Dunn, P. Ellen, M. Jenner, and M. herentdetection of scattered radiation,” IEEETrans. Antennas
Jervis, “Background limited photoconductiveHgCdTedetectors Propagat., vol. AP-15, pp. 192-194, January 1%7.
for use in the 1 1 4 micron atmospheric window,” Infrared Phys., [150] V. J. Corcoran, “Directional characteristics in optical heterodyne
vol. 9, pp. 3536, 1969. detectionprocess,”J. Appl. Phys.,vol.36, pp. 1819-1825, June 1965.
[126] I. Melngailis and T. Harman, “Photoconductivity in singlecrystal [151] W. Read andD. Fried,“Optical heterodyning and noncritical
Pb,-,Sn,Te,” Appl. Phys. Lett., vol. 13, pp. 180-183, September angular alignment,” Proc. IEEE(Letters), vol. 51, p. 787, December
1968. 1963.
[127] W. Kaiser, R. Collins, and H. Fan, “Infrared absorption in p-type [152] H. A. Bostick, “A carbon dioxide laser radar system,” IEEE J .
germanium,” Phys. Rec., vol. 91, pp. 1380-1381, September 1953. Quantum Electron., vol. QE-3, p. 232, June 1967.
[128] P. Bratt,“Photoconductivity in impurity-activated germanium,” [153] A. Glass, “Investigation of electrical properties of Sr,-,Ba,Nb,O,
Ph.D. dissertation, University of Syracuse, Syracuse, N. Y., January with special reference to pyroelectric detection,” J . Appl. Phys.,vol.
1965. 40, pp. 4699-4713, November 1969.
[129] T. Vogl, J. Hansen, and M. Garbuny, “Photoconductive time con- [154] M. Kimmit, J. Ludlow, and E. Putley, “Use of pyroelectric detector
stants andrelated characteristics of p-type gold-doped germanium,” to measure Q-switched C 0 2 laser pulses,” Proc. IEEE (Letters),
J . Opt ...Sot. Am., vol. 51, pp. 70-75, January 1961. vol. 56, p. 1250, 1968.
[130] J. Schmit and E. Stelzer, “Temperature and alloy compositional [155] A. Hadni, R. Thomas, and J.Perrin, “Response of triglycine sulfate
dependences of energy gap of Hg, -,Cd,Te,” J . Appl. Phys.,vol. 40, pyroelectric detector to high frequencies (300KHz),’’ J . Appl.
pp. 4865-4869, November 1969. Phys., vol. 40, pp. 274&2745, 1969.
[131] W. Saur, “Long wavelength mercurycadmium telluride photocon- [156] E. Leiba, “Heterodynageoptique avec un detecteur pyroelec-
ductive infrared detectors,” InfraredPhys., vol. 8, pp. 25S258, trique,” C. R. Acad. Sei. (Paris), vol. 268, ser. B, pp. 31-33, January
1968. 1969.
[132] A. Kohn and J. Schlickman, “1-2 micron (Hg, Cd)Te photodetec- [157] R. L. Abrams and A. M. Glass, “Photomixing at 10.6 p with stron-
tors,” IEEETrans.Electron Daices, vol. ED-16, pp. 885890,
October 1969. tium barium niobate pyroelectric detectors,” Appl. Phys.L e t t . , vol.
[r33] H. Mocker, “A 10.6 p optical heterodyne communication system,” 15, pp. 251-253, October 1969.
Appl. Opt., vol. 8, pp. 677484, March 1969. [158] S. Eng and R. Gudmundsen, “Theory of optical heterodyne detec-
[134] D. Breitzer, E. Sard, B. Peyton, and J. McElroy, “G-R noise for tion using pyroelectric effect,” Appl.Opt., vol. 9, pp. 161-166,
auger band-to-band processes,” to be published; see also: J. H. January 1970.
McElroy, S. C. Cohen, and H. E. Walker, “First and second s u m - [159] V. Danen, D. Sokoloff, A. Sanchez, and A. Javan, “Extension of
mary design reportATS-F laser communication experiment in- laser harmonic-frequency mixing techniques into the9p region
frared mixer and radiationcooler subsystem,” Goddard Space with an infrared metal-metal pointcontract diode,” Appl. Phys.
Flight Center, Greenbelt, Md., Rep. X-524-69-227, 1969. Lett., vol. 15, pp. 398-401, December 1969.