Ch6半导体光电子器件

6Semiconductor Detectors
6. Semiconductor Detectors
A large variety of semiconductor materials, structures and devices are used as photodetectors in optical receivers.
The most important for communications are:
pn p i n and Schottk Barrier Photodiodes
•pn, p-i-n and Schottky Barrier Photodiodes
•Avalanche Photodiodes
•Metal-Semiconductor-Metal(MSM)Photodiodes
Metal Semiconductor Metal (MSM) Photodiodes
•Photoconductors
Equally important optical devices, but structurally completely q y p p,y p y different and not used for optical communications include:
•Charge-Coupled Devices (CCDs)
•CMOS Imagers
•Photocathodes
•Solar Cells
Solar Cells
Optical Absorption Optical Absorption
Optical Absorption in Semiconductors
p p g g
The photon flux passing through
an absorbing medium is
Since the carrier collection
regions are ≤1 µm, the
absorption coefficient needs to
b101hi hi h
be ~104cm-1to achieve high
efficiency, which only occurs for
direct bandgap materials near
di t b d t i l
the bandgap. Basically want
identical thermal and photon
identical thermal and photon
energies for generation.
Photocurrent-The Mechanism Optical absorption creates extra pairs of electrons and holes
q
in excess of the thermal equilibrium concentration. If this is in the depletion region, then under the built-in field, or adding to that with reverse bias, the carriers are swept out by the electric field to give a reverse current of one electron for every generated electron-hole pair. Because electrons and holes have opposite charge, they move in opposite directions and there is only one particle passing any given point, so there is current of only one electronic charge, not two.
This drift of charge increases the nominal reverse current of the diode in the short circuit condition or pushes the diode into th di d i th h t i it diti h th di d i t forward bias if in the open circuit condition. The latter is the operating region for photovoltaic or solar cell operation. operating region for photovoltaic or solar cell operation
Optical Responsivity and
Quantum Efficiency
Quantum Efficiency
The optical response of a photodetector is characterized by
either quantum efficiency η, or responsivity, R
Quantum efficiency can be external,
ηext or internal, ηint. External is the
or internal External is the
number of electrons of current per
p
incident photon.
Internal is the number of electrons of
current per absorbed photon
Responsivity, R, is the photocurrent per
i i i h h
unit incident power (amps per watt)
At 1.24µm, η = 100% corresponds to R = 1A/W
Photogeneration and Photocurrent in pn Diodes
i Di d
p-i-n Photodiode
p
The primary disadvantage of the pn homojunction is that with moderate doping concentrations in the conducting
i h d d i i i h d i regions for low resistance, the resulting depletion region
is quite thin (e.g., 0.1-.2µm). This causes two problems:
i it thi(012)Thi t bl
1) low efficiency since relatively little absorption occurs in a thin depletion region (d ~ 2 ) and
thin depletion region(d2α)and
2) relatively high depletion capacitance, which decreases device speed (RC time constant).
device speed(RC time constant)
A general rule is that only carriers generated within the depletion region are efficiently and rapidly collected as photocurrent. The goal is to create a diode with a wide
p g
depletion region
Photogeneration and Photocurrent in p-i-n Diodes
i Di d
Schottky Photodiode
Analogous to a p-n junction,
only the built-in field is created
by surface Fermi level pinning.
Photo generated electrons are Photo generated electrons are
accelerated toward to the
metal-semiconductor junction
j by the built-in field and transfer
into the metal, creating a
h
photocurrent.Th l ilib i d l t t ti t t ti l Thermal equilibrium and electrostatics create a potential barrier, φbp for holes in p-type material or φbn for electrons in n type material With low doping analogous M i n or M i p material. With low doping, analogous M-i-n or M-i-p
diodes can be realized.
Differences Between
Schottky and pn Diodes
S h k d Di d
Schottky barriers behave straightforwardly as photodetectors Schottky barriers behave straightforwardly as photodetectors Photons absorbed in the depletion region near metal produce a p,j p j
drift photocurrent due to the surface field, just as in pn junction. However, the current is not controlled by recombination times and the diode cannot be used as a light emitter.
No biasing configuration of a Schottky diode produces substantial N bi i fi i f S h k di d d b i l overlapping populations of electrons and holes in same place because there is no injection of carriers from the metal into the because there is no injection of carriers from the metal into the semiconductor, hence no minority carrier injection into the semiconductor.
Any hole collection from the semiconductor into metal "recombines" very rapidly by non-radiative electron-electron scattering within the electron gas, hence no usable minority carrier
i i hi h l h bl i i i density (holes) in metal.
Diode Depletion Capacitance
p p
Diode capacitance determines the speed of response of optoelectronic devices e.g. RC time constant
It also determines the sensitivity of detectors since a smaller capacitance gives larger voltage swing for same number of photogenerated electrons
The diode depletion region is analogous to a dielectric, thus
Th di d d l i i i l di l i h
the capacitance is viewed as a dielectric parallel plate capacitor with depletion region width w d, depletion capacitance capacitor with depletion region width depletion capacitance is, C j
13, C j 1.15 x 10 F/cm 0.1fF/µm for w d 1 µm, and,
for~1µm,and,ε~13,~1.15x10-82=0.1fF/µm2
r
Frequency Limitations Frequency Limitations
The diode frequency response is limited by two factors: The diode frequency response is limited by two factors: 1) RC charging time, which for a 50 W load is ~100 GHz for a 50 µm diameter diode.
GHz for a50µm diameter diode
2) Transit time, t = w d/v sat which is ~ 10-11sec or 100 GHz for a 1 µm depletion width diode.
GH f1d l i id h di d
Actual diode parameters are seldom the frequency limitation, but parasitics associated with bonding and interconnects. Integration with the amplifier is key to high frequency receiver performance.
Avalanche Photodiodes
At low voltages, the maximum quantum efficiency in a diode
is a current of one electron per absorbed photon (100%).
However, with increasing reverse
bias voltage and at a higher electric
field, it becomes possible for an
electron (or hole) to be accelerated
electron(or hole)to be accelerated
so that its kinetic energy exceeds
g p gy
the bandgap energy and it can
create an additional electron-hole
pair through impact ionization--the
inverse of Auger recombination
i f A bi i
(both are 3 particle processes). Such
a process must exist from detailed
a process must exist from detailed
balance in thermodynamics.
() Avalanche Photodiodes (2)
It is possible to collect more than one electron of photocurrent per absorbed photon with very high bias across the depletion per absorbed photon with very high bias across the depletion region--each photoelectron (hole) generates additional electrons (and holes) when the electron energy exceeds the electrons(and holes)when the electron energy exceeds the bandgap energy in an exponential growing process called avalanche gain or multiplication. avalanche"gain"or"multiplication"
Impact ionization Coefficients
We describe the impact ionization process through impact ionization coefficients (or rates), αn and αp
p g p
These represent the strength of the processes for electrons and holes, respectively.
1/αn will correspond to the average distance for which an will correspond to the average distance for which an electron is accelerated before it creates an electron-hole pair by impact ionization, and similarly for 1/p for holes
b impact ioni ation and similarl for1/αfor holes
p() Impact ionization Coefficients (2)αn and αp are proportional to exp (-C/E), where C is a constant for a particular material and carrier type, and E is the electric for a particular material and carrier type and E is the electric field. For electric fields ~ 3 x 105V/ cm, for example, the
µ
distance between ionization events is ~ 1 µm in GaAs, which means relatively thick avalanche regions are
required to achieve even
moderate avalanche
d t l h
gains, e.g., ~3-4
microns. This results in
microns This results in
quite high bias voltages,
~ 100 V, not a desired
range for CMOS
systems architectures
and slower device
d l d i
response (~20 GHz).
Multiplication Noise and Gain Bandwidth Product
Impact ionization
by a single carrier
(electrons) G = 8
in this case, and
τt = w/v
e-sat
+ w/v h-sat
I t i i ti b Impact ionization by both electrons and holes1<G<but holes 1< G < ∞ but w/v e-sat< τt< ∞
p p Avalanche photodiode problems
Problems -"excess noise" and non-uniform avalanche multiplication occur when both carriers can initiate impact ionization events occur when both carriers can initiate impact ionization events Three consequences
1) We have much larger variability in the avalanche gain (literally anything from 1 to ∞), causing an additional source of variability in the resulting detected signal because the gain process is now a sum of (a) an average of M successive electron impact ionizations, plus of(a)an average of M successive electron impact ionizations plus (b) (an average of) M+1 electron impact ionizations and 1 hole impact ionization (the initial electron creates a hole that creates an electron that starts the process all over again), plus ... .
l t th t t t th ll i)l
2) Overall response is slowed down because electrons generate holes, which generate electrons, which generate holes, etc.
,g,g,
3) We can have a "run-away" process where the avalanche gain becomes infinite, with electrons impact ionizing to give holes that impact ionize to give electrons that impact ionize to give holes, and impact ionize to give electrons that impact ionize to give holes and so on.
Solutions to Avalanche Photodiode Problems
These problems become worse as αn and αp, become closer to one another, which is unfortunately the case for most III-V materials. The ratio between αn and αp is large ONLY in silicon, but silicon does not absorb beyond 1.1µm. One solution is to make optical absorption and avalanche gain regions out of different materials
Impact-Ionization Impact-Ionization Engineered APDs
hotodiode Noise Avalanche
Metal-semiconductor-metal (MSM) photodiode
Form two Schottky diodes close to one another on the same Form two Schottky diodes close to one another on the same (doped) semiconductor surface. Bias the resulting structure with some d.c.voltage and one of the diodes becomes reverse
g
biased, forming a depletion region that will tend to sweep out photocarriers. The other diode becomes forward biased, allowing the collected photocurrent to flow out just as if we had ll i h ll d h fl j if h d formed an Ohmic contact
Desirable to keep the
Desirable to keep the
distance between the
two metal electrodes
small to achieve high
speed, which leads to
the choice of an
th h i f
interdigitated structure
Metal-semiconductor-metal (MSM) photodiode (2)
MSMs have several attractive features
y p g yp q,
• only one doping type semiconductor is required,
• only one kind of semiconductor is required
• fabrication of interdigitated MSM photodetectors is quite •fabrication of interdigitated MSM photodetectors is quite simple and compatible with integrated circuit processing, which allows processes with fine (e.g., 1µm wide) lines required for ll ith fi(1id)li i d f dense interdigitation and high speed
• devices can also have very low capacitance for high speed
CCD(charge coupled device) CCD (charge coupled device) Common form of photodetector arrays
and readout method employed in most
and reado t method emplo ed in most
small camcorder TV cameras.
Concept is reminiscent of the
Concept is reminiscent of the
Schottky photodiode, except
there is the additional presence
of the high-bandgap oxide as a
current blocking insulator.
Band edges for the CCD cell (a)
Band edges for the CCD cell(a)
with initial bias but no photoinjected
charge, (b) with bias,
after photogenerated electrons have moved in the depletion region towards the positive electrode, (c) the thermal equilibrium situation that would exist with bias after any excess charge densities had leaked away
g()
Charge coupled device (2)
Pockets of charge moved serially out of the structure by "bucketbrigade" method like the example three-phase clocking method
This can be done
with 2-phase
ith2h
clocking if there is
an asymmetry in
an asymmetry in
the oxide thickness
under the gates
CMOS image sensors
g
Advanced CMOS technology has led to an effort to utilize it directly for image sensors (Moore’s Law in action)
directly for image sensors(Moore’s Law in action)
If it can be done without modifying the CMOS process, it leads to low cost image sensors. These can also be combined with
low cost image sensors These can also be combined with silicon digital (and analog) electronic signal processing
CMOS image sensors generally employ silicon photodiodes or CMOS image sensors generally employ silicon photodiodes or variants in which the charge is created directly in a polysilicon g
transistor gate
The photogenerated charge is read out by sequentially turning
on switches to read out the charge or voltage on each photodetector in turn rather than by the "bucket brigade" analog shift register of the CCD.
In general, less expensive, but lower resolution.
Photoconductors
A piece of semiconductor material with two Ohmic contacts, and a voltage is applied between them. The semiconductor is most
lt i li d b t th Th i d t i t likely doped and thus conducting, hence there is some current flowing even without light shining on the material(a dark current) flowing even without light shining on the material (a dark current) If we shine light on the material,
electron-hole pairs will be
electron-hole pairs will be
generated and the carrier
concentration is increased in
concentration is increased in
the material, thus the
y
conductivity of the material
increases, giving larger current
Photoconductors (2)
Photoconductors differ from the photodiodes in several important ways y
1) Current is carried both by minority and majority carriers in the photoconductor
2) Current continues flowing in the photoconductor until all the excess electrons and holes recombine, but the majority carriers do not recombine at the electrodes. For every electron in ndoped material that leaves the structure by passing into the
contact region, another electron is injected at the other contact to maintain charge neutrality
Minority carriers (holes in this example) do recombine when they Mi it i(h l i thi l)d bi h th reach the electrode. Thus the time to turn off photoconduction is governed by either by minority carrier lifetime inside the material governed by either by minority carrier lifetime inside the material or transit time of the minority carriers to the electrodes
g Photoconductive gain
If the majority carrier transit time is long compared to the effective minority carrier lifetime (transit or bulk, whichever is )y y y p shorter), then an electron may effectively make many passes through the material before recombining.
,p y
As a result, it is possible to have many electrons of current flow through the structure for one absorbed photon, i.e., quantum efficiency greater than one, a phenomenon known as
y g,p
photoconductive gain
Photoconductors (3)
Less desirable features
1) Can be relatively slow; unless they are made very small, transit
times can be long
times can be long
2) Use of photoconductive gain occurs also at the expense of
speed s ce e jo y c e us s esse y g pc es speed since the majority carrier must transit essentially g times
through the structure for a photoconductive gain of g pc.
3) Dark current can contribute significantly to noise and it is
difficult to detect a small photocurrent in the presence of a larger
dark current.
Can make a photoconductor fast by arranging for a very short Can make a photoconductor fast by arranging for a very short minority carrier lifetime in the material, but then, the responsivity
y p g
can be low because only a corresponding fraction of an electron’s
worth of current flows through the circuit.
Very fast photoconductors can be made by “killing” the lifetime
and this is used to make very fast "switches" triggered by short
(femtosec) laser pulses。