integrated optical detectors
TRANSCRIPT
-
8/2/2019 Integrated Optical Detectors
1/9
Integrated Optical Detectors
Detectors for use in integrated-optic applications must have highsensitivity, short
response time, large quantum efficiency and low power consumption
[1]. In this
chapter, a number of diffierent detector structures having theseperformance characteristics
are discussed.
17.1 Depletion Layer Photodiodes
The most common type of semiconductor optical detector, used inboth integrated
optic and discrete device applications, is the depletion-layer photo-
diode. Thedepletion-layer photodiode is essentially a reverse-biased
semiconductor diode in
which reverse current is modulated by the electron-hole pairs
produced in or near
the depletion layer by the absorption of photons of light. The diode isgenerally
operated in thephotodiode mode, with relatively large bias voltage,
rather than inphotovoltaic mode, in which the diode itself is the electrical generator
and no bias
voltage is applied [2].
17.1.1 Conventional Discrete Photodiodes
The simplest type of depletion layer photodiode is the p-n junction
diode. The
energy band diagram for such a device, with reverse bias voltage Va
applied is shownin Fig. 17.1. The total current of the depletion layer photodiode
consists of two components:
a drift component originating from carriers generated in region (b)and a
diffusion component originating in regions (a) and (c). Holes and
electrons generated
in region (b) are separated by the reverse bias field, with holes being
sweptinto the p-region (c) and electrons being swept into the n-region (a).
Holes generated
in the n-region or electrons generated in the p-region have a certain
probabilityof diffusion to the edge of the depeletion region (b), at which point
they are swept
across by the field. Majority carriers, electrons in (a) or holes in (c) are
held in theirrespective regions by the reverse bias voltage, and are not swept
across the depletion
layer.
Fig. 17.1 Energy band
diagram for a p-n junctiondiode under application of a
reverse bias voltage Va
In order to minimize series resistance in a practical photodiode while
still maintaining
maximum depletion width, usually one region is much more heavilydoped
than the other. In that case, the depletion layer forms almost entirelyon the more
lightly doped side of the junction, as shown in Fig. 17.2. Such a device
is called ahigh-low abrupt junction. In GaAs and its ternary and quaternary
alloys, electron
mobility is generally much larger than hole mobility. Thus, the p-
region is usually
made thinner and much more heavily doped than the n-region, so that
the device
-
8/2/2019 Integrated Optical Detectors
2/9
will be formed mostly in n-type material, and the p-region then serves
essentiallyjust as a contact layer.
For a device with the high-low junction geometry indicated in Fig.
17.2, it can
be shown that the total current densityJtot is given by [3]Jtot = q0
_
1 eW
_1 + L p__
+ qpn0
DpL p
(17.1)
where 0 is the total photon flux in photons/cm2s, Wis the width of
the depletion
layer, q is the magnitude of the electronic charge, is the opticalinterband absorption
coefficient, Lp is the diffusion length for holes, Dp is the diffusion
constant forholes, and Pn0 is the equilibrium hole density. The last term of (17.1)
represents the
reverse leakage current (or darkcurrent), which results from
thermally generated
holes in the n-material. This explains why that term is not
proportional to the photon
flux 0. The first term of (17.1.1) gives the photocurrent, which is
proportionalFig. 17.2 Energy band
diagram for a p+-n (high-low)
junction diode underapplication of a reverse bias
voltage Va
to 0, and includes current from both the drift of carriers generated
within the depletion
layer and the diffusion and drift of holes generated within a diffusion
length Lp
of the depletion layer edge. The quantum efficiency q of the detector,
or the number
of carriers generated per incident photon, is given by
q = 1 eW_1 + L p_
, (17.2)
which can have any value from zero to one. It should be noted that(17.1) and (17.2)
are based on the tacit assumption that scattering loss and free carrier
absorption are
negligibly small. The effect of these loss mechanisms on the quantum
efficiency,when they are not negligible. is discussed in Section 17.1.3.
It can be seen from (17.2) that, in order to maximize q. it is desirable
to make
the products Wand Lp as large as possible. When Wand Lp are
large enoughso thatq is approximately equal to one, the diode current is then
essentially proportional
to 0, because the dark current is usually negligibly small.
If the interband absorption coefficientis too small compared toWand
Lp. many
of the incident photons will pass completely through the active layersof the diode
into the substrate, as shown in Fig. 17.3. Only those photons absorbed
within the
depletion layer, of thickness W, have maximum effectiveness in carrier
generation.
-
8/2/2019 Integrated Optical Detectors
3/9
Photons absorbed at depths up to a diffusion length Lp from the
depletion layeredge are somewhat effective in generating photo-carriers, in that holes
can diffuse
into the depletion layer. Photons that penetrate to as depth greater
than (W+ Lp)before being absorbed are essentially lost to the photo-generation
process because
they have such a very low statistical probability of producing a hole
that can reachthe depletion layer and be swept across. Within the semiconductor,
the photon flux(x) falls off exponentially with increasing depthxfrom the surface, asshown in
Fig. 17.4. Thus, ifis not large enough, many photons will penetrate
too deeply
before being absorbed, thus producing carriers that (on average) will
recombinebefore diffusing far enough to reach the depletion layer.
Interband absorption is a strong function of wavelength in a
semiconductor. Theabsorption coefficientusually exhibits a special response curve that
rises sharply at
the absorption-edge (band-edge)wavelength and then saturates at
awavelength that is
Fig. 17.3 Diagram of a
conventional mesa-geometry
photodiode with p+-n doping
profile showing photonpenetration
Fig. 17.4 Optical absorption
versus depth from the surface
in a conventional mesa
photodiodeslightly shorter than the bandgap wavelength, increasing slowly for
yet shorter wavelengths.
Thus, it is impossible to design a diode with an ideal Wfor all
wavelengths.For wavelengths near the absorption edge, the long-wavelength
response of a diode
is limited by excess penetration of photons into the substrate, as
shown in Figs. 17.3and 17.4; its short wavelength response can be limited by too strong
an absorption of
photons in the p+ layer near the surface, where recombinationprobability is large.
Aside from the reduction of quantum efficiency that results from poor
matching
of, Wand Lp, there are some other limitations to depletion layer
photodiode performancethat are also important. Since Wis usually relatively small (in the
range
from 0.1 to 1.0 m), junction capacitance can limit high-frequencyresponse through
the familiar R-C time constant. Also, the time required for carriers to
diffuse from
depths between Wand (W+ Lp) can limit the high frequency response
of a conventional
photodiode. The waveguide depletion layer photodiode, which is
discussed in
the next section, significantly mitigates many of these problems of theconventional
photodiode.
17.1.2 Waveguide PhotodiodesIf the basic depletion layer photodiode is incorporated into a
waveguide structure, as
-
8/2/2019 Integrated Optical Detectors
4/9
shown in Fig. 17.5, a number of improvements in performance are
realized. In thiscase, the light is incident transversely on the active volume of the
detector, rather than
being normal to the junction plane. The diode photocurrent density is
then given byJ= q0 _1 eL _, (17.3)
where L is the length of the detector in the direction of light
propagation. Since W
and L are two independent parameters, the carrier concentrationwithin the detector
volume and the bias voltage Va can be chosen so that the depletion
layer thickness Wis equal to the thickness of the waveguide, while L can be made as long
as necessary
to make L>>1. Thus 100% quantum efficiency can be obtained for
any value of,
by merely adjusting the length L. For example, for a material with therelatively small
value of= 30 cm1, a length ofL = 3 mm would give q = 0.99988.
(Again, ithas been tacitly assumed in (17.3) that scattering loss andfree-carrier absorption are
negligible.)
Because a waveguide detector can be formed in a narrow channel
waveguide, the
capacitance can be very small. even ifL is relatively large. For example,
for a material
with a relative dielectric constant = 12, such as GaAs, a 3 mm long
detectorformed in a 3m wide channel waveguide has a capacitance of only
0.32 pF. This
capacitance is about a factor of ten less than that of a typicalconventional mesa
photodiode. Hence, the high frequency response can be expected to be
correspondingly
improved. Experimentally demonstrated bandwidth of 5 GHz and
quantumefficiency of 83% have been obtained with waveguide detectors on
GaAs substrate
material [4], and InGaAs waveguide photodetectors on InP substrates
also haveexhibited a 5 GHz bandwidth for light in the wavelength range of 1.3
1.6m [5].
Computer simulation of waveguide photodetectors in AlGaInAs
GaInAs, designedfor ultrawide-band operation at 60 and 100 GHz, predict internal
quantum efficiencies
as high as 94% and 75%, respectively, at 1.55m wavelength [6].Because all of the incident photons are absorbed directly within the
depletion
layer of a waveguide photodetector, not only is q improved, but also
the time
delay associated with the diffusion of carriers is eliminated. This resultis a further
improvement in high frequency response.
Due to the many improvements in performance inherent in thetransverse structure
of the waveguide detector, as compared to the axial geometry of the
conventional
mesa photodiode, waveguide detectors should be considered for use in
discrete-device applications, as well as in optical integrated circuits. At
the present
time, waveguide detectors are not commercially available as discrete
devices. However,they can be fabricated with relative ease in many laboratories. Hence
availability
should not long be a problem.17.1.3 Effects of Scattering and Free-Carrier Absorption
The relations given by (17.1), (17.2) and (17.3) neglected the effects of
free-carrier
-
8/2/2019 Integrated Optical Detectors
5/9
absorption and photon scattering on the quantum efficiency of the
detector. Becauseboth of these mechanisms result in the loss ofphotons without the generation of
any new carriers, they tend to reduce quantum efficiency. In many
cases they can
be neglected, and (17.1), (17.2) and (17.3) will give accurate
predictions. However,
when the free carrier absorption coefficientFC and the scattering
loss coefficients
are not negligible as compared to the interband absorption coefficientIB, a more
sophisticated expression for q is required. Such an expression can be
derived asfollows. The photon flux at any point located a distancexfrom the
surface of the
detector on which the photons are first incident is assumed to have the
form given
by(x) = 0ex , (17.4)
where in general the loss coefficientis given by
= IB + FC + S. (17.5)The hole-electron pair generation rate G(x) is given by
G (x) = IB0ex , (17.6)
since only IB results in carrier generation. Thus the photocurrent
density is given
by
J= q L
0
G (x) dx(17.7)or
J= q0
IBIB + FC + S
_1 e(IB+FC+S)L _ . (17.8)
Comparing (17.8) with (17.3), it is obvious that the effect of additional
lossesdue to scattering and free-carrier absorption is to reduce the quantum
efficiency by
a factor ofIB/, even when L is large enough to maximizeq.
Ifs and FC are small compared to IB, as is generally true, (17.8)
reduces to
(17.3). However, if the waveguide is inhomogeneous or is unusually
rough, or if the
detector volume is heavily doped so thats and FC are not negligible,then (17.8)
must be used.
17.2 Specialized Photodiode StructuresThere are two very useful photodiode structures that can be fabricated
in either a
waveguiding or conventional, nonwaveguiding form. These are the
Schottky-barrier
photodiode and the avalanche photodiode.
17.2.1 Schottky-Barrier Photodiode
The Schottky-barrier photodiode is simply a depletion layer
photodiode in which the
p-n junction is replaced by a metal-semiconductor rectifying(blocking) contact. For
example, if the p-type layers in the devices of Figs. 17.3 and 17.5 were
replaced by
a metal that forms a rectifying contact to the semiconductor, Schottky-
barrier photodiodes
would result. The photocurrent would still be given by (17.1) and
(17.3),
and the devices would have essentially the same performance
characteristics as their
p+-n junction counterparts. The energy band diagrams for a Schottky-
barrier diode,
-
8/2/2019 Integrated Optical Detectors
6/9
under zero bias and under reverse bias, are given in Fig. 17.6. It can be
seen thatthe depletion region extends into the n-type material just as in the case
of a p+-
n junction. The barrier heightB depends on the particular metal-
semiconductor
combination that is used. Typical values for B are about 1 V.
In conventional mesa devices, a thin, optically transparent Schottky-
barrier contact
is often used (rather than a p+-n junction) to enhance short-wavelength response,
by eliminating the strong absorption of these higher energy photons
that occurs inthe p+ layer. In a waveguide photodiode, a Schottky-barrier contact is
not needed for
improved short-wavelength response because the photons enter the
active volume
transversely. However, case of fabrication often makes the Schottky-barrier photodiode
the best choice in integrated applications. For example, almost
anymetal (exceptfor silver) produces a rectifying Schottky-barrier when evaporated
onto GaAs or
GaAlAs at room temperature. Gold, aluminum or platinum are often
used. Transparent
conductive oxides such as indium Tin oxide (ITO) and cadmium Tin
oxide
(CTO) can also be used to eliminate the photon masking effect of the
contacts andthereby improve the quantum efficiency, as described in Section
17.2.4. Photoresist
masking is adequate to define the lateral dimensions duringevaporation, and no
careful control of time and temperature is required, as in the case of
diffusion of a
shallow p+ layer.
A detailed discussion of the properties of Schottky-barrier diodes isbeyond the
scope of this text so that the interested reader should refer to the
information available
elsewhere [7].
17.2.2 Avalanche Photodiodes
The gain of a depletion layer photodiode (i.e. the quantum efficiency),
of either the
p-n junction or Schottky-barrier type, can be at most equal to unity,under normal
Fig. 17.6
conditions of reverse bias. However, if the device is biased precisely at
the point of
avalanche breakdown, carrier multiplication due to impact ionization
can result in
substantial gain in terms of increase in the carrier to photon ratio. Infact, avalanche
gains as high as 104 are not uncommon. Typical current-voltage
characteristics for
an avalanche photodiode are shown in Fig. 17.7. The upper curve is for
darkenedconditions, while the lower one shows the effects of illummination. For
relatively
low reverse bias voltages, the diode exhibits a saturated dark current
Id0 and a saturated
photocurrentIph0. However, when biased at the point of avalanche
breakdown,
carrier multiplication results in increased dark currentId, as well as
increased photocurrent
Iph. It is possible to define a photomultiplication factor Mph, given by
Mph Iph
Iph0, (17.9)
-
8/2/2019 Integrated Optical Detectors
7/9
and a multiplication factor M, given by
M Iph + Id
Iph0 + Id0
, (17.10)
An exact equation for the current-voltage curve is difficult to obtain in
the region
of bias in which avalanche breakdown occurs. However, Miller [8] has
represented
the functional form of the photomultiplication factor by the expression
Mph = 11 (Va/Vb)n . (17.11)
where Vb is the breakdown voltage, and n is an empirically
determined exponentdepending on the wavelength of light, doping concentration, and, of
course, the
semiconductor material from which the diode is fabricated. For the
case of large
photocurrentIph0 >> Id0 Melchior and Lynch [9] have shown that themultiplication
faSpecialized Photodiode Structures 353
M= 11 _ VaI R
Vb
_n , (17.12)
where Iis the total current, given by
I= Id + Iph, (17.13)
R being the series resistance of the diode (including space-charge
resistance if significant).
The derivation of (17.12) assumes thatIR
-
8/2/2019 Integrated Optical Detectors
8/9
carefully fabricated and are hermetically sealed into adequate
packages, mean timeto failure as high as 105 h at 170 C has been observed [1, p. 80],
which projects to
about 109 h at room temperature.
17.2.3 p-i-n Photodiodes
In Section 17.1.1. it was pointed out that conventional photodiodes
must be designed
so as to have a large Wproduct in order to maximize q; but one
doesnt havecomplete control over either the depletion width W, which depends on
dopant concentrations,
or the absorption coefficient, which depends mostly on the bandgap.In
the p-i-n photodiode, a very lightly doped intrinsic layer is formed
between the p
and n sides of the diode. This layer generally has a carrier
concentration of less thanctor is given by
1014/cm3, but it is compensated by a balance of p- and n-type
dopants rather than
being truly intrinsic. Because of the low carrier concentration, the
depletion layer
in a p-i-n diode extends completely through the i layer so that the total
thickness of
the active layer is the sum of the i-layer thickness Wi and the
deplection width on
the lightly doped (n) side of the junction. Thus the device designer can
adjust thetotal depletion width to produce a large Wproduct by varying the
thickness of the
i-layer. The presence of the relatively thick i-layer also reduces thejunction capacitance
and increases the R-C cutoff frequency of the diode, p-i-n photo-diodes
are
widely used as detectors in optical systems because of their high
quantum efficiency(responsivity) and wide bandwidth. For example, Kato et al. [12] have
reported a
waveguide p-i-n photodiode operating at 1.55m wavelength with a
quantum efficiency
of 50% and a 3 dB bandwidth of 75 GHZ.
17.2.4 Metal-Semiconductor-Metal Photodiodes
Metal-semiconductor-Metal (MSM) photodiodes are surface-oriented
devices thatfeature interdigitated, finger-like, Schottky barrier contacts formed on
the surface of
a thin semiconducting layer on a semi-insulating substrate. A typicalMSM photodiode
structure is shown in Fig. 17.8. Carriers generated by the absorption of
photons
in regions of the semiconducting layer between the contacts are swept
by the fringingelectric field and collected by the contacts. Holes are collected by the
cathode
and electrons by the anode. Spacing of the contact fingers must be lessthan the
diffusion length of the carriers in order to produce a high collection
efficiency.
Because the contact fingers are very narrow and closely spaced
(1m) the
capacitance is relatively low and the transit times of carriers are short.
Hence, widebandwidth
operation is possible. Hsiang et al. have made Si MSM diodes with0.2m width and spacing with a full-width-half-maximum pulse
response of 3.7 ps,
corresponding to a 3 dB bandwidth at 110 GHz [13]. The Schottkyelectrodes of the
MSM photodiode are essentially identical to the gate metalization of
field-effect
-
8/2/2019 Integrated Optical Detectors
9/9
transistors, which facilities their monolithic integration with FETs.
For example,Mactaggart et al. [14] have made a fully-integrated 400 Mb/s burst-
mode data
OEIC receiver for application as a phased-array antenna controller.
Approximately
350 source-coupled FET logic gates are present on the GaAs chip,
along with a
780 nm wavelength MSM photodiode. MSM photodiodes have also
been integrated
monolithically with High-Electron-Mobility field effectTransistors
(HEMTs) to
produce OEIC receivers with bandwidths larger than 14 GHz [15, 16].
HEMTs
have also been integrated with p-i-n photodiodes to produce OEIC
receivers with a
bandwidth of 42 GHz [17].The most significant disadvantage of MSM detectors is their inherent
low responsibility
because the metallization for the surface electrodes shadows the
active
light-collecting region. However, this problem can be mitigated by
using a transparent
conducting material for the contact electrodes. Gao et al. [18] have
fabricated
InGaAs MSM photodiodes with transparent Cadmium Tin Oxide (CTO)
electrodes.
The responsivity of these devices to 1.3m light was 0.49 A/W, ascompared to
0.28 A/W for identical control samples with conventional Ti/Au
electrodes.Another approach to improve the overall responsivity of an m-s-m
photodiode is
to monolithically integrate it with an amplifier. For example, Cha et al.
[19] haveintegrated an m-s-m photodiode with a high-electron-mobility
transistor (HEMT)
on an InP substrate, with an InGaAsP buffer layer (g = 1.3 m). They
measured a
responsivity of 0.7 A/W at a wavelength of=1.3 m, and the 1.5100
m2 gate
HEMT hadft andfmax of 18.7 and 47 GHz, respectively.
17.3 Techniques for Modifying Spectral ResponseThe fundamental problem of wavelength incompatibility, which was
encountered
previously in regard to the design and fabrication of monolithiclaser/waveguide
structures in Chapter 14, is also very significant with respect to
waveguide detectors.
An ideal waveguide should have minimal absorption at the wavelength
being used.However a detector depends on interband absorption for carrier
generation. Hence,
if a detector is monolithically coupled to a waveguide, some meansmust be provided
for increasing the absorption of the photons transmitted by the
waveguide within
the detector volume. A number of different techniques have proven
effective in this
regard.