Chapter 1 An Introduction to the nBn Photodetector
Transcript of Chapter 1 An Introduction to the nBn Photodetector
1
Chapter 1 An Introduction to the nBn Photodetector
Infrared detector systems are integral to a variety of surveillance and remote sensing
applications for the military. They are anticipated to become prevalent in civilian markets
supplying equipment used to monitor environmental factors and the weather, make
astronomical observations, meet the demands of law enforcement, and perform medical
diagnostics. These systems are frequently built around focal plane arrays, which consist of
as many as millions of individual photodetectors.1,2
Future infrared photodetector systems must possess reduced cooling requirements, consume
less power, exhibit longer lifetimes, and have improved manufacturability. The highest
performing infrared photodetectors are currently cooled to cryogenic temperatures, which
enables optimal operation by decreasing internal detector noise.3 Warmer temperature
operation will enable the use of lower cost and less bulky coolers with decreased power
requirements. This is anticipated to permit more extensive field use and result in a decrease
of the, sometimes critical, wait between powering on the imaging system and being able to
use it. Increasing the yields and simplifying the fabrication requirements of the detector
arrays are projected to result in substantial cost savings.2,4,5
The nBn photodetector6, the subject of this thesis, is a recently introduced class of midwave
infrared photodetectors developed as an alternative to those photodetectors currently
incorporated into infrared detector systems. The nBn photodetector architecture specifies an
n-type absorption layer, a Barrier layer, and an n-type contact layer. The barrier layer is
constrained to have both a negligible valence band energy offset with the absorption layer,
which allows the free passage minority carrier holes, and a large conduction band energy
offset with the absorption layer, which blocks the flow of majority carrier electrons. Nowhere
in the device is the Fermi level near the middle of the bandgap. This suppresses the
Shockley-Read-Hall (SRH) generation current that is a primary noise source in cooled7,8,9, as
well as some uncooled10, infrared p-n junction photodiodes; the depletion region is the
primary source of SRH generation current in p-n junction photodiodes, which is a
consequence the Fermi level in the depletion region being located in the middle of the
bandgap. In addition, the unpassivated nBn photodetector exhibits negligible surface
leakage current.6 The surface leakage current path is disrupted by the barrier layer, which is
not etched during processing. Surface leakage current is of considerable concern in p-n
1.1 Focal Plane Arrays
2
junction photodiodes,7 particularly if they are fabricated from semiconductor materials with
narrow bandgap energies. It is common to passivate p-n junction photodiodes during
fabrication to reduce surface leakage current.2,7,11 The nBn photodetector, through the near-
elimination of SRH and surface leakage currents, requires less cooling to operate optimally
and has simpler processing requirements than the p-n junction photodiodes. The nBn
architecture may be extended to different material systems, but the focus of this work is on
the InAs-based nBn photodetector intended for midwave infrared operation.6
1.1 Focal Plane Arrays: A Brief Background
Infrared imaging systems trace their origins to 1800 when William Herschel,12 using a prism
and a thermometer, discovered infrared radiation. The earliest infrared radiation detection
systems were based on thermometers, thermocouples, and bolometers.13-16 Subsequent
work led to achievements including lead salt photon-based infrared detectors,17,18 but it was
not until the introduction of the transistor in the late 1940s that the development of modern
infrared detectors became possible. Research conducted through the 1950s, spurred by the
transistor, led to improved semiconductor purification and material growth techniques. During
this time, the III-V semiconductors were also identified, and infrared detectors based on
extrinsic germanium, InSb, and variable-gap HgCdTe were fabricated.19 The III-V
semiconductor family became of particular interest to researchers working with molecular
beam epitaxy (MBE), which was introduced in the early 1970s and permitted the fine control
of semiconductor growth conditions not possible with contemporary thin film deposition
techniques.20 In the mid-1970s, silicon emerged as an important material for imaging with the
advent of charge coupled devices (CCDs), which had a substantial impact on the industry
due to the integration of detection and readout functions on the same chip.19
Photolithographic processing techniques, adapted from the printing industry, revolutionized
semiconductor device processing by enabling complex device and chip designs.19 Before
photolithography, CCDs, and other charge transfer devices (CTDs) became available, the
first infrared detector systems were limited to single elements and sparsely-populated linear
arrays of elements.2 These elements and modestly sized arrays are rastered along both the
x and y axes to create an image of the scene. The scans are complex to implement, and the
detector systems possess limited sensitivity as a consequence of short integration times.
Photolithography, combined with etching techniques, advanced the state of the art by
1.1 Focal Plane Arrays
3
facilitating the fabrication of two-dimensional matrices of infrared detectors. The production
of the first forward looking infrared (FLIR) detector arrays occurred in the 1960s.21
Beginning in the 1970s, advances in CTD technologies supported the development of two-
dimensional detector arrays with thousands of elements.2,7 The earliest were 16 x 16 and 32
x 32 arrays, which are stepped over the entire field of view to form an image. These small
two-dimensional scanned arrays led to larger staring arrays, also called Focal Plane Arrays
(FPAs). It is not necessary to raster staring arrays, as these detector arrays contain enough
elements to image the full scene.2 Focal plane arrays are considered critical to advanced
infrared imaging systems. Staring arrays, by avoiding inefficiencies associated with
mechanical scanning, reduce the complexity and cost of infrared detection systems.7 Larger
arrays also offer better sensitivity, as a result of significantly increased integration times.
However, they have nontrivial signal processing requirements and place difficult demands on
the imaging optics. There are also producibility concerns, due to low yields, for some
material systems.2
FPAs can be monolithic or a hybrid combination of detector elements and multiplexing
readout circuitry.7 A monolithic structure, in which the detector and multiplexing functions are
integrated, is not always an option. Despite research into multiplexers produced from
narrow-gap semiconductors, silicon remains the only material system producing marketable
monolithic FPAs. For this reason FPAs are frequently hybrids, meaning the detector array
and the associated multiplexers are fabricated from different materials and are electrically
and physically joined using a technology such as indium bumps. Hybridization has the
benefit of allowing the detector and multiplexer to be optimized independently of one another.
In the case of a hybrid, once the detector array and multiplexer are produced, indium bumps
are deposited on both, the two are aligned, and then they are pressed together to make
electrical contact.1,2 The detectors may be illuminated from the frontside, with the radiation
passing through the multiplexer, or from the backside, with the radiation passing through the
substrate underpinning the detector array. The latter is usually chosen to avoid transmission
losses due to the opaque metallization on the multiplexer. If the substrate is not transparent
to the radiation of interest, it must be thinned to minimize undesirable absorption. It may also
be thinned to improve the resolution and response of the FPA.7
Uniformity in the performance of the constituent detectors is key to the overall performance of
FPAs.2 The performance of a photon detector is strongly dependent on the temperature of
the device; a small variation in operating temperature across an array substantially impacts
1.1 Focal Plane Arrays
4
the quality of the image. Achieving lower operating temperatures consumes more energy, as
multiple cooling stages are required. Four are not uncommon. It is common for dewars to
cost more than the FPA and to be more labor intensive to produce.2,22 An accepted goal of
infrared detector system research is the reduction of cooling requirements enabling long-lived
and inexpensive coolers with modest power consumption, such as thermoelectric coolers
permitting operating temperatures of 200 K and higher, to be used.3
1.2 Emission and Atmospheric Transmission of Infrared Radiation
Infrared detectors monitor thermal radiators, and these may be approximated as blackbody
sources. The greater the thermal energy of the object, the higher the energies of the emitted
photons. The spectral radiant exitance of a blackbody source, Meλ(λ,T) is obtained from the
Planck radiation law (see for example Reference 23), and may be expressed as:
Me,λ λ,T( )=2πhc2
λ5 exp hcλkT
−1
−1
(1.2.1)
where λ is the wavelength, h is Planck’s constant, c is the velocity of light, and k is
Boltzmann’s constant. As plotted in Figure 1.2.1, the peak power emitted by hotter objects
occurs at lower wavelengths, while the power emitted at all wavelengths increases. The
spectral window spanning 2 to 12 microns coincides with a majority of the radiation emitted
by objects near room temperature. Infrared systems designed to monitor earth-bound
objects target this wavelength range.1,23
Infrared photodetection systems are often used as remote sensors, which has led to infrared
detectors commonly being designed to operate in one of two atmospheric transmission
windows as shown in Figure 1.2.2: the 3-5 micron medium wavelength infrared (MWIR)
window or the 8-14 micron long wavelength infrared (LWIR) window. Operation in the MWIR
atmospheric window is typical when the application relies on contrast resolution more than
detector sensitivity, comparatively hot objects are monitored, and operation occurs during
clear weather. High humidity conditions do not affect the atmospheric transmission of
radiation in this window, however the LWIR range is better suited for foggy, hazy, dusty, or
misty conditions.1
1.2 IR Emission and Transmission
5
Figure 1.2.1 Emission spectra of blackbody radiators at selected temperatures.
Figure 1.2.2 Atmospheric transmission spectrum of radiation. Transmission through 6000 feet at sea level.1 Reproduced with kind permission from Rogalski, A. and Chrzanowski, K., (2002). “Infrared devices and techniques,” Opto-Electronics Review, 10 (2), pp. 111-136.
1.3 Noise in MWIR Photodiodes
6
1.3 Noise in Semiconductor-Based MWIR Photodetectors
Noise in infrared detection systems is attributable to statistical fluctuations in the detected
signal as well as fluctuations in the flow of spurious charge carriers created through means
other than by the absorption of a signal photon, such as through thermal processes.24
Thermal processes are particularly important for narrow-bandgap semiconductor-based
photodetectors operating near room temperature, as the energy of the charge carriers, kT, is
comparable to the transition energy.3 Some common types of internal noise in infrared
detectors are Johnson-Nyquist, 1/f, generation-recombination, diffusion, tunneling, and
surface leakage.19,23 Noise may also arise through external sources such as background
radiation, cooling irregularities, mechanical issues, and incident optical sources.23,25
Random motion of carriers in a semiconductor results in fluctuations of the open circuit
voltage, Vj, called thermal or Johnson-Nyquist noise,24
Vj = 4kTR∆f (1.3.1)
with ∆f representing the electrical bandwidth, and R the real part of the impedance.
More precisely written
1/ f α , with α having a value of approximately unity, 1/f noise is
sometimes correlated with surface effects and is important at lower frequencies. It has not
been modeled precisely.24,25
Shot noise has a white spectrum at low frequencies and arises as a consequence of the
quantum fluctuations of the optical field, as well as statistical variations in the emission,
recombination, and flow of charge carriers. The mean square shot noise current, In, is
modeled as24
In = 2qI ∆f (1.3.2)
with I the current, and q the electronic charge.
The components of dark current important to this work are due to surface leakage, generation
processes, and those produced by the detection of background radiation. These may be
1.3 Noise in MWIR Photodiodes
7
modeled as shot noise.19 Recombination processes are not significant for photodiodes and
nBn photodetectors. Recombination processes become important in the presence of excess
carriers, as in the case of photoconductors and when carriers are injected into a device.
Generation processes, and not recombination processes, are considered in devices without
excess carriers, which is the case for devices that extract carriers, such as photodiodes.19,26
The characteristics of generation noise depend on the constituent detector material, device
structure, and device temperature.27 The three most important types of generation
mechanisms in narrow-bandgap semiconductors are: Shockley-Read-Hall, Auger, and
radiative.7 Equation (1.3.3) is a general formalism applicable to variety of semiconductor
generation processes. The generation rate, Gk, of a particular mechanism, k, is related to the
square of the intrinsic carrier concentration, ni, through a proportionality term, gk:7
Gk = gk ni2( ). (1.3.3)
The net generation rate, G, is the sum of all Gk. The square of the intrinsic carrier
concentration is the product of the electron, no, and hole, po, concentrations under conditions
of thermal equilibrium:
ni2 = no po. Non-equilibrium carrier densities are n = no + δn, and p =
po + δp. The general expressions for the electron, n, and hole, p, concentrations are:28
n = NcF1/ 2 −Ec − E f( )
kT
(1.3.4a)
p = NvF1/ 2 −E f − Ev( )
kT
(1.3.4b)
with the Fermi-Dirac integral, F1/2[η], defined using
F j η( )=1
Γ j +1( )x jdx
1+ exp x −η( )0
∞
∫ , (1.3.5)
where Γ represents the gamma function, and the effective densities of states in the
conduction and valence bands, Nc and Nv respectively, are
1.3 Noise in MWIR Photodiodes
8
Nc = 2 me*kT
2πh2
3 / 2
(1.3.6a)
Nv = 2 mh*kT
2πh2
3 / 2
(1.3.6b)
with electron and hole effective masses, me* and mh*. In this work, when treating the case of
degenerate semiconductors, the Fermi-Dirac integral is approximated as29
F1/ 2 η( ) ≈ exp −η( )+3π1/ 2 /4
η4 + 33.6η 1− 0.68exp −0.17 η +1( )2[ ]{ }+ 50
3 / 8
−1
(1.3.7a)
which valid over
−∞ < η < ∞ . In the case of non-degenerate semiconductors the form
F1/ 2 η( ) ≈ exp η( ) (1.3.7b)
is used.
The net generation rate of the process k, expressed in Equation (1.3.3), becomes, under
small-signal conditions and expressed in terms of the lifetime of the process, τk,7
Gk =δnτ k
(1.3.8)
with the overall recombination lifetime the parallel sum
1τ
=1τ kk
∑ . (1.3.9)
1.3 Noise in MWIR Photodetectors
9
1.3.1 Shockley-Read-Hall Noise
Shockley-Read-Hall generation-recombination processes30,31 are extrinsic, as they utilize
energy levels within the bandgap resulting from impurities and material defects. Electrons
and holes may be both captured by and emitted from these trap states, which impacts excess
carrier lifetimes. Semiconductor material with better crystalline quality exhibits less SRH
generation current. The net generation rate, which takes into account all four processes, is,
after Reference 7,
GSRH =σ nσ pvthni
2NDL
σ n n + n1( )+ σ p p + p1( ) (1.3.10)
with σn and σp the electron and hole capture cross sections, vth the carrier thermal velocity,
NDL the density of the traps, and the nondegenerate electron, n1, and hole, p1, concentrations
corresponding to a Fermi energy level coincident with the energy of the traps, ET, are
n1 = Nc exp −Ec − ET( )
kT
(1.3.11a)
p1 = Nv exp −ET − Ev( )
kT
. (1.3.11b)
The most efficient generation of SRH current occurs when the energies of the trap states and
the Fermi level coincide near the middle of the bandgap. This is found by varying the Fermi
energy and trap energy to maximize GSRH. The trap energy required to maximize GSRH is that
which maximizes 1/ (n1 + p1). Figure 1.3.1 is a normalized plot of this fraction for the case of
InAs at 200 K. The peak occurs when ET = 0.22 eV above the valence band edge energy.
This coincides with the intrinsic Fermi energy, Efi, and is near the middle of the bandgap,
which is 0.19 eV greater than the valence band energy.32 The proximity of the intrinsic Fermi
energy and the energy at the middle of the bandgap leads to the common and convenient
approximation of Efi ~ Eg/2.
It is of particular interest to note the Fermi energy lies in the middle of the bandgap in the
depletion region of p-n junction photodetectors, and SRH generation current is a primary
noise source for cooled photodiodes.
1.3 Noise in MWIR Photodetectors
10
Figure 1.3.1 Determination of the trap energy required to maximize GSRH. The trap energy is measured from the top of the valence band edge and the computed fraction is normalized.
It is useful to simplify Equation (1.3.10) by approximating terms n, p, n1, and p1 for cases in
which the Fermi level is located at mid-gap in the depletion region of a semiconductor device.
The approximation substitutes –Eg/2kT for the arguments in the exponentials of each of these
four terms. Doing so allows n, p, n1, and p1 to be easily summed and GSRH to be expressed,
as
GSRHDepl , in terms of the intrinsic carrier concentration and a lifetime, τo:
GSRHdepl ≈
ni2
τ o NcNv exp −Eg /2kT
=ni
τ o
(1.3.12)
where
τ o =σ nσ pvthNDL NcNv
2σ nNc + 2σ pNv
−1
(1.3.13)
1.3 Noise in MWIR Photodetectors
11
The relation between the bandgap energy, temperature, and
GSRHdepl is
GSRHdepl ∝T1/ 2ni ∝T2 exp −
Eg
2kT
, (1.3.14)
which takes into account the T1/2 temperature dependence of the thermal velocity 24 in the
lifetime term.
1.3.2 Auger Noise
Multiple types of band-to-band Auger processes are possible in semiconductor materials,33
but only one is relevant to this work. In direct bandgap n-type semiconductors, such as the
unintentionally doped InAs composing the absorption layer of the InAs-based nBn
photodetector, the most significant Auger process is known as Auger 1 or CHCC.7,34 It
involves two transitions: one between the conduction and heavy-hole bands, and the other
between two energy states in the conduction band. According to the Auger 1 process,
generation involves two electrons and one heavy hole; collision with an energetic conduction
band electron excites an electron from the heavy hole valence band into the conduction
band. The generation rate for the Auger 1 process is
GA = GA1nni2 (1.3.15)
with
GA1 =8 2π( )5 / 2q4mo
h3 4πεoεs( )2
me* /mo( )F1F2
2
1+ µ( )1/ 2 1+ 2µ( )
1ni
2kTEg
3 / 2
exp −1+ 2µ1+ µ
Eg
kT
(1.3.16)
µ = me* / mh* is the electron and hole effective mass ratio, εs is the relative static dielectric
constant, εo is the vacuum permittivity, mo is the rest mass of the electron, and F1 and F2 are
the overlap integrals of the periodic portion of the wave functions of the electrons, with 0.1 <
|F1F2| < 0.3. It is common to approximate Equation (1.3.16). One approximation comes from
recognizing narrow-bandgap semiconductor materials such as InAs have µ << 1, so that the
argument in the exponential term becomes –Eg/kT. Another approximation can be made
1.3 Noise in MWIR Photodetectors
12
while considering n-type InAs. The difference between the Fermi and conduction band
energies for unintentionally doped n-type InAs with a carrier concentration of 1 x 1016 cm-1 are
calculated to vary from 0.06 to 0.02 eV for temperatures from 300 K to 150 K. Then, for n-
type InAs, it is reasonable to disregard the temperature dependence of the exponential
portion of the n term in Equation (1.3.15). The value of Eg can also be considered to be
approximately temperature independent. GA is then found to be proportional to the square of
the intrinsic carrier concentration,
GA ∝ ni2 ∝T3 exp −
Eg
kT
. (1.3.17)
When the first two approximations are not made, but the temperature dependence of Eg is
neglected, a more accurate proportional relationship for the case of InAs is:
GA ∝T3 exp −1.08Eg + Ec − E f( )
kT
. (1.3.18)
The equation for the lifetime associated with the Auger 1 process is not reproduced here, but
it can be found in Reference 7. Figure 1.3.2 includes a curve corresponding to the
calculated Auger 1 lifetime in InAs.
The other two Auger processes recognized as important in semiconductors with one
conduction band and heavy hole and light hole conduction bands are the Auger 7, in which
one carrier transitions between the conduction and heavy hole bands and another between
the heavy hole and light hole bands (CHLH), and the Auger S, in which one carrier transitions
between the conduction and heavy hole bands and another between the split-off and heavy
hole bands (CHSH). These processes may be disregarded in n-type material but should be
considered for p-type material. The Auger 7 may be the dominant Auger process in a p-type
material if the spin split-off energy is much larger than the bandgap energy of the material.7
1.3 Noise in MWIR Photodetectors
13
1.3.3 Radiative Recombination
Radiative recombination occurs when a photon is emitted upon the recombination of an
electron and a hole. In direct bandgap semiconductor-based light emitting diodes, the
intrinsic radiative recombination process occurring across the bandgap is important.
Radiative recombination is not considered a limiting factor in practical photodetectors. Most
photons arising through radiative decay are reabsorbed, which lengthens the measured
radiative lifetime. An expression for the radiative lifetime, τR, is7
τ R =ni
2
GR no + po( ). (1.3.19)
1.3 Noise in MWIR Photodetectors
14
Figure 1.3.2 Carrier lifetimes of Auger and radiative recombination processes in InAs. Lifetimes denoted τA1 for Auger 1, τR for radiative recombination, τA7 for Auger 7, and τAS for Auger S processes. Lifetimes are plotted with respect to normalized doping concentrations in InAs at 200 and 300 K.7,34-36 Reproduced with kind permission from Rogalski, A., Adamiec, K., and Rutkowski, J., (2000). Narrow-Gap Semiconductor Photodiodes, Bellingham, Washington: SPIE-The International Society for Optical Engineering, p. 99.
Auger and radiative recombination are intrinsic processes for which universal curves of the
associated lifetimes, such as those shown in Figure 1.3.2, may be generated. Universal
curves may not be produced for the lifetimes of SRH processes, as these depend on the
density of traps. Figure 1.3.2 plots the calculated lifetimes for the radiative, Auger 1, Auger 7,
and Auger S processes for InAs material of various doping densities at 200 K and 300 K.
Auger 1 is the dominant process in n-type InAs.7
1.4 Figures of Merit
Figures of merit quantify the performance of an infrared detection system; they contrast the
numbers of carriers created by incident signal photons with the numbers generated from
noise processes. The most useful figures of merit allow comparison of infrared detection
1.3 Noise in MWIR Photodetectors
15
systems with different architectures, constituent material systems, and manufacturers. While
some figures of merit are broadly applicable, none is universally appropriate. In some cases,
experimental conditions influence the reported numerical values.2
Most of the figures of merit applied to infrared detection systems use the signal to noise ratio,
SNR, as a basis, and are therefore related to one another. One expression of SNR,26
SNR =Is
In
(1.4.1)
is the ratio of the signal current, Is, and the noise current, In.
The responsivity, Ri, is the frequency dependent ratio of the output photocurrent, Is, to the
incident optical power, Popt,24,26
Ri =Is
Popt
=ηqhν
= η qλhc
(1.4.2)
in which ν is the frequency of light, and the quantum efficiency, η, is defined as the number of
electron-hole pairs generated for each incident photon. The value of the quantum efficiency
strongly depends on the absorption coefficient of the material, α(λ).24 Responsivity is a more
attractive metric when the ratio is linear, but it is not as useful as some other figures of merit
as it contains no information about the minimum strength a signal must have to be
detectable.37
Noise equivalent power, NEP, is the ratio of the signal noise to the responsivity, and the
detectivity, D, is the reciprocal of the NEP:26,37
NEP =1D
=In
Ri
. (1.4.3)
A detector with superior sensitivity has a lower noise equivalent power and, consequently, a
higher detectivity. NEP and D are functions of detector area, A, and electrical bandwidth.
1.4 Figures of Merit
16
When D is inversely proportional to
A∆f , as is frequently the case, the spectral detectivity
D*, is defined37
D* = D A∆f . (1.4.4)
The value of D* is both wavelength and temperature dependent. Photon-based detectors
have a often have a more pronounced dependence on D* than thermal detectors.7
D* also depends on the field of view (FOV) when the current arising from the detection of
background radiation is the primary contributor to noise current,7,19,23
In2 = 2q qηAφB[ ]∆f (1.4.5)
where φB is the background photon flux density,19
φB =π sin2 θ /2( )
π2πc
λ4 exp hc /λkTB( )−1[ ]0
λc
∫ dλ . (1.4.6)
and the FOV is θ. The value of D* under these conditions is,
DBLIP* =
ηqλhc
2q2ηφB[ ]−1/ 2∝
1sin FOV /2( )
. (1.4.7)
When a single value is quoted for D*, it generally refers to the peak value of detectivity. Plots
of spectral detectivity curves for infrared photoconductive and photovoltaic detectors are
widely available, and they are typically plotted against ideal curves computed assuming peak
detectivity at each wavelength.1,2 An example is shown in Figure 1.4.1.7
1.4 Figures of Merit
17
Figure 1.4.1 D* for a selection of commercially available infrared photodetectors.1 Reproduced with kind permission from Rogalski, A. and Chrzanowski, K., (2002). “Infrared devices and techniques,” Opto-Electronics Review, 10 (2), pp. 111-136.
The noise equivalent temperature difference (NETD, NE∆T), which is expressed in units of
temperature, is the minimum discernible temperature difference present in a scene. NETD is
that temperature difference producing a system output signal equal to the noise of the
camera,1,37
NETD =∆T
I∆s / In
(1.4.8)
with
I∆s = η Is
∆T
∆T . (1.4.9)
Here I∆s is the differential current associated with the ∆T change in temperature, and Is is the
current from the detector element. Two pixels can be distinguished from one another when
one pixel images a portion of the scene that differs by the NETD from the portion of the scene
imaged by another pixel. This basic equation can be cast to more explicitly show influences
of different elements, including D*, on the NETD.2,7 The equation may be stated in terms of
the f-number of the optics, f# the focal length divided by the lens diameter, as1
1.4 Figures of Merit
18
NETD =4 f#
2 ∆ftop M* A
(1.4.10)
with the top the transmission of the optics, and M* the spectrally-dependent figure of merit,
M* =∂L∂T
λ
tatDλ*dλ
0
∞
∫ , (1.4.11)
which incorporates the partial derivative of the emitted radiance as a function of temperature,
∂L/∂T , and the atmospheric transmission tat.. This figure of merit is especially useful for
thermal imagers, which convert the temperature differences in an imaged scene into a visual
image. Temperature resolution is therefore a common figure of merit for these systems.2
However, as this work focuses on detector elements and not infrared systems, D* is the
preferred figure of merit for this work.
The RoA product of a photodiode, which is computed by multiplying the area by the
resistance computed for zero-bias voltage, is a figure of merit frequently quoted for
photodiodes. It may be calculated directly from the current density as a function of voltage
data: it is the reciprocal of the slope of the curve taken at a zero voltage bias:19
RoA=∂J∂V
−1
V= 0
(1.4.12)
It is desirable for RoA to be as large as possible; this is advantageous when mating the
detector to a readout, and it identifies a detector with less noise and lower leakage current.2
Device cooling is used to improve RoA, but it cannot be increased to an arbitrary value. The
sensitivity of RoA to temperature becomes negligible under conditions including surface
leakage current dominating the dark current, tunneling through the depletion region, and the
detection of background radiation.7,39 The RoA of a photodiode is frequently evaluated for the
case of zero voltage bias and no background radiation.39
Background limited infrared photodetection, also known as background limited in
performance, (BLIP) operation can be defined for specific background radiation conditions.
1.4 Figures of Merit
19
BLIP operation occurs when the internally-generated and temperature-dependent dark
current of a photodetector has been reduced, through cooling, to equal the magnitude of the
current produced by the detection of background radiation. Equations (1.4.5) through (1.4.7)
assume BLIP operation. Reducing the temperature of the device below the temperature for
which BLIP operation is achieved does not increase performance, as the primary noise
component is now contributed by an external stimulus.23,40
1.5 Classes of Infrared Photodetectors
Infrared detectors dominating the commercial marketplace are solid state devices that
convert infrared radiation into an electrical signal, and these are broadly classified as either
photon or thermal detectors. Thermal detectors experience changes in electrical properties
when the temperature of the constituent material changes after absorbing infrared radiation.
In photon-based detectors, incident photons interact directly with electrons. This results in a
modification of the electronic distribution in the material and the alteration of an electronic
property.1,41
Microbolometers, which are members of the thermal class of detectors, and p-n junction
photodiodes, which are photon-based detectors, are currently the most widely used MWIR
detectors, however no one type of detector is well suited to all applications.1,2 Operating
conditions, minimum required response time and detectivity, the wavelength range of interest,
and budget dictate the best choice. Whether the detector will monitor space or Earth-bound
targets substantially impacts selection. The latter assumes targets are imbedded against a
background of objects at 300 K, while background radiation in the former case may be
consistent with blackbodies having temperatures lower than a couple tens of kelvin. Whether
or not the application supports detector cooling is also a factor. Cooling is most easily
implemented in fixed laboratories and large vehicles, such as tanks. Cooling is generally not
available and/or practical for applications requiring portability, such as in the case of imaging
systems carried by personnel in the field or employed in temporary installations.
1.5.1 Thermal Detectors
1.5 IR Photodetector Classes
20
The three types of thermal detectors most important to infrared radiation sensing are
bolometers, pyroelectric detectors, and sensors based on the thermoelectric effect such as
thermocouples. Electrical resistance changes with temperature in bolometers, differences in
the spontaneous electrical polarization are monitored in pyroelectric detectors, and variable
voltage generation is the basis of thermocouples. These, like all thermal detectors, return a
signal based on the radiant power incident on the detector and are largely insensitive to the
spectral content of the radiation. The majority of thermal detectors are not cooled, which
makes them inexpensive, easy to use, and favored for use in space-based and field
operations. (Thermal photodetectors based on superconductors, which are cooled to tens of
kelvin, are a significant exception.) Benefits associated with uncooled operation are offset by
slow response times, which is a consequence of the necessary wait while the material heats
and cools, sensitivities well below those of photon-based detectors, and historically large
center-to-center spacing of the elements.2,19
Micromachined bolometers, termed microbolometers, have been fabricated and fashioned
into uncooled FPAs with monolithic and hybrid readout circuits.2,5 Infrared detector systems
based on microbolometer FPAs are becoming increasingly important for LWIR detection for
applications such as missile warning, surveillance, night vision, and thermal sights.
Microbolometer arrays with 640 x 480 elements and pixels with 17 micron side lengths are
commercially available.42
1.5.2 Photon Detectors
The class of photon-based detectors, in which photodetection relies on charge carriers
interacting directly with incident photons, includes semiconductor-based infrared detectors.
Charge carriers may be bound to lattice or impurity atoms, or they may be free. Interband
transitions have been leveraged in photoconductive and photovoltaic detectors, among
others. Carriers generated in photoconductive detectors alter the conductivity of the material,
while electron-hole pairs created in photovoltaic detectors result in a photocurrent or a
voltage difference detectable across two electrodes. Energy states residing in the bandgap
of the host semiconductor may be created by doping a semiconductor with impurity atoms.
The extrinsic absorption of a photon involves exciting a carrier from such a state into an
energy band of the semiconductor, and these transitions are used in some photoconductive
detectors. Internal photoemission, which referrs to the photon-induced ejection of an
1.5 IR Photodetector Classes
21
electron, is the basis of Schottky barrier detector operation. Quantized energy levels may be
created with a quantum well structure, and transitons involving these levels have been used
in photoconductive and photovoltaic detectors.7,43
Unlike the majority of thermal detectors, photon-based detectors exhibit pronounced
wavelength dependence. For photon-based detectors there exists an abrupt long wavelength
cutoff and a typically more gradual drop off in responsivity for shorter wavelengths.7,24
Detection occurs when the photons composing the incident radiation possess enough energy
to transition carriers into different energy states. In the case of intrinsic photoconductive and
photovoltaic detectors, carriers are excited across the bandgap of the semiconductor. The
long wavelength cutoff,
λc =hcEg
, (1.5.1)
occurs for photons with energies less than the bandgap energy; these photons are not able to
effect a transition and are so minimally absorbed that the material is considered transparent
to them. For ideal photoconductors and photodiodes operating at wavelengths below the
cutoff, η = 1. However, responsivity for high-energy photons is actually poor as a
consequence of the large value of the absorption coefficient at these lower wavelengths. At a
penetration depth of x, the transmitted radition intensity, Iv, is reduced from the incident
intensity, Ivo.24,43
Iv x( )= Ivo exp −αx( ). (1.5.2)
Short wavelength photons are absorbed close to the surface of the semiconductor. These
are prone to quickly recombine, which prevents them from contributing to the measured
electrical signal.
Although the spectral response of photon-based detectors is more narrow than their thermal-
based counterparts, photon-based detectors possess a number of advantages.
Semiconductor material systems of most consequence to the field of infrared detectors
possess direct energy gaps, doping flexibility, high electron mobilities, and low dielectric
constants.7 This endows them with fast response times, high sensitivities, and high signal to
noise ratios.1,45
1.5 IR Photodetector Classes
22
Photon-based detectors are frequently cooled for better performance. Detectors operating in
the MWIR are generally cooled to temperatures between 200 and 77 K, and it is common to
cool those operating in the LWIR to temperatures 77 K or lower. Cooling supresses the
thermal generation of carriers, which competes with optical generation processess and
obsures the signal. More agressive cooling is required for LWIR photodetectors, as the
smaller energy transitions are more easily triggered by thermal processes. Fulfilling cooling
requirements presents an important obstacle to the more widespread adoption of photon-
based infrared systems. There is a focus on developing devices having optimal or near-
optimal performance at room temperature or, baring that, at temperatures greater than 200 K,
which are acheivable with thermoelectric coolers. Thermoelectric coolers are lighter-weight
and reliable performers with modest power requirements.2,45
1.6 Significant Semiconductor-Based Infrared Photodetectors
A variety of semiconductor materials are used in photon-based infrared photodetectors, with
Hg1-xCdxTe and InSb being the most significant to commercial detection systems. While Hg1-
xCdxTe is used in both MWIR and LWIR systems, InSb-based systems are exclusive to the
MWIR. Other materials, such as extrinsic semiconductors and the lead salts, have not been
widely adopted for use in the MWIR and LWIR.5 Lead salts are used in a limited fashion in
the MWIR, often with minimal cooling and in applications for which high resolution is not
critical. Photodetectors based on the lead salts are inexpensive and straight-forward to
produce, but they degrade in humid and acidic environments and must be shielded from
bright visible light.2 Extrinsic silicon and germanium are featured prominently in systems
detecting wavelengths beyond 20 microns, and these detectors must be cooled below 4.2 K.5
1.6 Semiconductor-based MWIR Photodetectors
23
1.6.1 Mercury Cadmium Telluride (Hg1-xCdxTe)
Hg1-xCdxTe, of the II-VI semiconductor material system, occupies the dominant position in
infrared detector systems and is the most commonly used material in forward looking infrared
(FLIR) detection systems for the military. It is found in systems intended for both high and
low background radiation applications. No material system has been developed more
extensively for infrared photodetection, which is traceable to the enormous freedom it confers
during system design. The undisputed advantage of using Hg1-xCdxTe is the ability to tune
the wavelength sensitivity over an extensive range, from 1 to 30 microns and beyond, during
manufacture by adjusting the composition of the ternary. It also experiences notably small
changes in lattice constant with changes in composition; the 0.3% change in lattice constant
between HgTe and CdTe permits layered and graded-gap structures to be grown as high-
quality crystals. Hg1-xCdxTe does not lack for competitors, however these have not been able
to claim better performance or, with the exception of thermal detectors, operation at higher
temperatures. Instead, most rivals claim to be more manufacturable.2,5,7
The manufacturability of Hg1-xCdxTe is a significant issue. For commercial viability, it is
necessary for semiconductor materials possessing low defect densities and good uniformity
to be grown reproducibly on large-diameter wafers,. Growing these crystals with uniform
compositions is a significant challenge, as a high mercury pressure during crystal growth
complicates the regulation of the stoichiometry.7 Compositional variations across the wafer
translate to a range of cutoff wavelengths, which is broader for longer wavelength cutoffs. As
responsivity is dependent on wavelength and FPAs require uniform responsivity across the
array, it is necessary to tightly control the composition of the Hg1-xCdxTe during growth. The
difficulties in doing so at longer wavelengths, where a 5% variation in composition
corresponds to variations in cutoff wavelength of several microns, have limited the cutoff
wavelengths of commercially produced large-area FPAs to around 11 microns.2,7
Post-growth processing of Hg1-xCdxTe, a typically proprietary and onerous procedure that
includes the deposition of a passivation layer, is required to yield the high-performance
photovoltaic devices incorporated into FPAs. Passivation is critical to performance, as it
reduces the otherwise high levels of surface leakage current, suppresses tunneling, and
prevents the decomposition of the material. Without passivation, mercury evaporates from
and alters the properties of the material. Passivation also reduces surface recombination
1.6 Semiconductor-based MWIR Photodetectors
24
velocity and increases RoA.2,7 Elimination of the passivation step would reduce costs and
increase yields.6
Infrared photodetectors fabricated from Hg1-xCdxTe exhibit quantum efficiencies of 50-60
percent without anti-reflective coatings and 70-80 percent with them. FPAs composed of
photovoltaic devices operating in the MWIR typically exhibit D* of 1 x 1011 cm·Hz1/2·W-1 at 100
K. Uniformities are strongly influenced by the cutoff wavelength of the alloy; arrays operating
at shorter wavelengths are more uniform, and uniformities are better for smaller arrays.2
Nonuniformities of less than 4 percent have been reported for a 256 x 256 array operating at
4.9 microns.46 Uncorrected nonuniformities for LWIR arrays are typically 10 to 20 percent,
and the best corrected uniformities are an order of magnitude lower.2
The RoA of MWIR photodiodes, reported to be 4 x 107 ohm·cm2 at 77 K for a 4.0 micron cutoff
wavelength, is limited by generation-recombination current originating in the depletion
region.48-50 The RoA drops to approximately 30 ohm·cm2 at 193 K, where currents originating
from radiative and Auger mechanisms are prevalent.47,48,51,52 The RoA for average devices
with cutoff wavelengths between 9-11 microns at 78 K is 300 ohm·cm2,53 and it is around 650
ohm·cm2 for the best.54 Under BLIP conditions at 40 K, the RoA is greater than 6 x 105
ohm·cm2.53
1.6.2 Indium Arsenide Antimonide (InAs1-xSbx)
InAs and InSb, which are both narrow-bandgap binaries of the III-V semiconductor family with
similar physical properties, are common MWIR photodetector materials. InSb is the more
prevalent as it has a smaller bandgap, 0.22 eV32 and a cutoff wavelength of 5.6 microns,
which is better suited to cover the whole 3 to 5 micron wavelength range of the MWIR
atmospheric transmission window. InAs possesses a larger bandgap of 0.35 eV and a
consequently shorter cutoff wavelength of 3.6 microns.32 InAs is not currently used or under
consideration for use in FPAs.2 FPAs based on InAs1-xSbx alloys are under development, as
these promise to extend to the cutoff wavelengths beyond those of the binaries; InAs0.40Sb0.60
with a cutoff wavelength of 7.0 microns has been demonstrated. An infrared detector
composed of InAs0.10Sb0.90 has the potential for 77 K operation at 9.0 microns, which is a
cutoff wavelength no other III-V semiconductor can match.55,56
1.6 Semiconductor-based MWIR Photodetectors
25
InSb is a mature material, and it is being positioned as a less-expensive alternative to
HgCdTe. Photodetectors produced from InSb and HgCdTe perform equally well in the
MWIR, InSb is easier to manufacture than HgCdTe, and detectors fabricated from both have
similar power requirements. InSb is also easier to grow uniformly; nonuniformities across the
array, which are incurred during the manufacturing process after crystal growth, are typically
one to two percent. InSb is frequently seen as the best choice for MWIR applications
demanding high sensitivity, good corrected uniformity, and ease of manufacture. Despite the
benefits of InSb, it is a fragile material and fabrication requires specialized equipment and
training. InSb use is mostly confined to the niche FPA market.2
InSb photodetectors may be photovoltaic or photoconductive, and both monolithic and hybrid
arrays have been produced. Hybridized arrays are common; producing monolithic arrays
results in detectors having shorter cutoff wavelengths than when they are manufactured
separately as well as nonlinearities in the CMOS readout.2 While free carrier absorption is
reduced when the back surfaces of the detector arrays are thinned to 100 microns, they are
frequently polished to a 10 micron thickness to achieve higher quantum efficiency.5,7 The
diffusion length of minority carriers in n-type InSb is approximately 30 microns at 80 K,57 and
detector arrays with thicker substrates suffer from a loss of resolution and reduced
responsivity.2 Once thinned, the back surfaces are passivated and coated with antireflection
film.2,7 A report of an 128 x 128 FPA claims a D* of 7.56 x 1011 cm·Hz1/2·W-1 for illumination
conditions which have a predicted theoretical BLIP D* of 9.4 x 1011 cm·Hz1/2·W-1.59 Other
figures of merit for FPAs include RoA of 2 x 106 ohm·cm2 at 77 K,58 and quantum efficiencies
approaching 90 percent.2
InAs1-xSbx has attracted interest as it can potentially produce detectors operating out to 9
microns, but the challenge of fabricating high-quality material is substantial. The lack of
substrates with suitable lattice constants is a key issue; devices grown from alloys closely
lattice matched to and grown on GaSb substrates perform best. This material must be
passivated to suppress the otherwise high levels of surface leakage current, and the higher
levels of generation current, arising from an increased dislocation density in the strained
material, is significant. To date, InAs0.91Sb0.09 photodetectors have seen acceptance in
optical fiber communication systems operating in the 2-4 micron wavelength range.2,7
1.6 Semiconductor-based MWIR Photodetectors
26
1.6.3 Schottky Barrier Photodiodes and the Silicides
Schottky barrier photodiodes have been under sustained development as alternatives to p-n
junction photodiodes. They consist of a thin film, which is on the order of 10-20 Angstroms
thick and contains heavy metal atoms, in contact with a semiconductor crystal. Palladium,
platinum, and iridium are popular dopants, with cutoff wavelengths of 3.2, 5.7, and 10
microns, respectively. Rectification creates a potential barrier between the metal and the
doped silicate layers. Incident signal photons are absorbed in the metal-containing film, and
those possessing enough energy excite majority carriers over the potential barrier.2 Majority
carriers are typically the carriers of signal current. Carriers may also tunnel through the
barrier, and carrier recombination may occur in the space charge and neutral regions.24,43
Platinum silicide Schottky barriers have enjoyed considerable commercial interest.2
Schottky barrier photodiodes have a number of advantages as compared with p-n junction
photodiodes. The manufacturing process for the former is notably less complex, they are not
subject to diffusion processes at high temperatures, and they have faster response times that
allow higher frequency operation.7 Forward biasing a Schottky barrier photodiode based on
an n-type semiconductor crystal injects electrons into the metal-containing film.24,43 There,
collisions between carriers result in thermalization occurring on a time scale of approximately
10-14 seconds.60 Forward biasing a p-n junction photodiode injects carriers that must instead
dissipate through recombination, and minority carrier recombination lifetimes are typically on
the order of 0.5 microseconds. The difference in lifetimes results in Schottky barrier
photodiodes having the potential for higher frequency operation.7
Schottky barriers are well suited to applications in which high resolution is required, the
background radiation is high, and cooling to temperatures below 80 K is acceptable. Material
uniformity is excellent. Uncorrected uniformity is better than 99 percent for Pt:Si, which
allows for the fabrication of large arrays.2 Arrays of Pt:Si devices with 1040 x 1040 pixels are
available.61 These detectors must be cooled to temperatures below those required by p-n
photodiodes and other detectors, due to the necessity of quelling the thermal emission of
carriers over the barrier. However, once appropriately cooled, typically to 77 K, the sensitivity
of Schottky barrier photodetectors is largely independent of temperature, which is a noted
advantage over other types of photon detectors.2,7 Sensitivity is low, with scattering
mechanisms interfering with the transport of photogenerated carriers.62 Quantum efficiencies
are 4 percent and 0.5 percent for Pt:Si at 2.5 and 4 microns, respectively.2
1.6 Semiconductor-based MWIR Photodetectors
27
This work does not consider Schottky barrier photodiodes to be direct competitors to p-n
junction photodiodes or the nBn photodetector, as they have very low sensitivities and
operate at liquid nitrogen temperatures. They are not considered further in this work.
1.6.4 Photoconductors
Photoconductors are essentially optically variable resistors, in which conductivity increases
when absorbed photons create free carriers.19,24,43 In the case of intrinsic semiconductor
material, the cutoff wavelength of the absorbed photons is determined by the bandgap
energy. For extrinsic material, it is the difference between the energy of the donor state and
the band edge. Photoconductors are structurally simple photodetectors, consisting of a width
of semiconductor material sandwiched between two ohmic contacts.
Figure 1.6.1 Sketch of a Photoconductor. The physical structure is illustrated at left. The schematic at right shows electron-hole pair generation across the bandgap. Ec is the energy at the bottom of the conduction band and Ev is the energy at the top of the valence band.
During operation, an electric field is applied across the contacts. A current always flows in
these devices, due to the lack of barriers and the sustained voltage bias, but the magnitude
of the current increases in the presence of a signal. The conductivity is expressed as
σ = q µnn + µp p( ), (1.6.1)
where q is the charge on the carrier, µn and µp are the electron and hole mobilities, and n and
p are the electron and hole densities.
L
hν
Ev Ec
hν
1.6 Semiconductor-based MWIR Photodetectors
28
Performance may be quantified in terms of quantum efficiency, gain, response time, and
detectivity. The internal quantum efficiency approaches unity, as an absorbed photon nearly
always increases the conductivity. The external quantum efficiency can exceed 0.9 when the
illuminated surface of the detector is coated with an anti-reflection coating. The gain,
G =Ip
Iph
=τtr
=τ
L/vd
, (1.6.3)
where L is the length of the semiconductor and vd is the drift velocity, is the ratio of the carrier
lifetime, τ, and the carrier transit time, tr, as well as the ratio of the measured photocurrent, Ip,
to the primary photocurrent,
Iph = q ηPopt
hν
. (1.6.2)
The carrier lifetime is a critical parameter to both the gain, which may range from 1 to 106,
and the response time, which varies from 10-3 to 10-8 seconds and is longer than that of
photovoltaic photodetectors.24 A benefit of gain includes lessened dependence on low-noise
preamplifiers. However, a longer response time can be a liability; photoconductive
photodetectors are limited to lower frequency operation than photovoltaic photodetectors.2,7
The primary internal noise sources in photoconductive photodetectors are Johnson-Nyquist
noise noise, given in Equation (1.3.1), generation-recombination noise, and 1/f noise, for
which there is no exact analytical model.19 Both generation and recombination processes are
active in photoconductive photodetectors, while only generation processes are significant in
the nBn photodetector and the p-n junction photodiode. As a consequence, photoconductive
photodetectors are constrained to have at least
2 times more noise than photodetectors, in
which only the generation process contributes significant noise.19,26 There are a variety of
expressions for generation-recombination noise in photoconductive detectors; each reflects
different compositional and operational characteristics. One applicable to a nearly intrinsic
photoconductor is,63
Vgr =2Vb
Atµe + µh
µen + µh pnp
n + p
1/ 2τ∆f
1+ ω 2τ 2
1/ 2
, (1.6.3)
1.6 Semiconductor-based MWIR Photodetectors
29
where Vb is the bias voltage, t is the thickness of the absorbing material, and ω is the
modulation frequency. An expression for the voltage responsivity,
Rv =Vb
hcµe + µh
noµe + poµh
ηAt
, (1.6.4)
is applicable when sweep-out, surface recombination, edge effects, and absorption of
background radiation may be disregarded. Further assuming 1/f has been made negligible
during device manufacture and Johnson-Nyquist and generation-recombination noise are
dominant, the expression for D* becomes19
D* =Rv A∆fVj
2 + Vgr2
. (1.6.5)
The advantages of photoconductive photodetectors include high gain and ease of
manufacture. This is particularly true of those based on extrinsic silicon and germanium. In
the HgCdTe material system photovoltaic photodetectors, with microwatt power dissipation,
are generally preferred over the photoconductive, with milliwatt power dissipation. Arrays of
extrinsic silicon photoconductive photodetectors are modest in size, usually not larger than
128 x 128, highly sensitive, typically useful into the LWIR, and require cooling to
temperatures in the neighborhood of 20 K to reduce dark current to acceptable levels.
HgCdTe based photoconductive photodetectors are either operated singly or as groupings of
a few elements.2 Photoconductive photodetectors are not considered significant competitors
to the p-n junction photodiode and nBn photodetector.
1.6.5 p-n Junction Photodiodes
The p-n junction photodiode incorporates a barrier to the flow of majority carriers absent from
the photoconductive detector. A p-n junction is formed at the interface of p-type and n-type
materials. Diffusion of majority charge carriers across the metallurgical junction creates
space charge regions surrounding the junction that are depleted of free charge carriers. This
diffusion continues until the electric field arising from the exposed and fixed ions is sufficiently
large to discourage the further net migration of the majority charge carriers. In thermal
1.6 Semiconductor-based MWIR Photodetectors
30
equilibrium, the Fermi energy, Ef, is uniform throughout the device. In the absence of an
externally applied field and under steady-state conditions, the built-in potential barrier, Vbi, is
used to define the difference between the intrinsic Fermi energies in the n-type and p-type
regions.43
Figure 1.6.2 Sketch of a p-n junction. Top: The space charge region exists at the interface of p-type and n-type materials. Bottom: Energy diagram of a p-n junction showing Auger (band-to-band) and SRH generation (trap-assisted) processes. Direction of free carrier diffusion is indicated. (After Reference 7.)
The majority of photogenerated carriers in MWIR photodiodes are produced in the material
adjoining the depletion region, rather than in the depletion region itself. Photoexcitation
occurring across the bandgap is the preferred process. Arranging for the majority of the
absorption to occur in the p-type region is typical, as electrons have a higher mobility than
holes.19 Carriers generated within a diffusion length may diffuse into the space-charge
region, where the electric field selectively sweeps the minority carriers through the depletion
region. Gain is taken to be unity, except in the case of avalanche photodiodes. Photodiodes
1.6 Semiconductor-based MWIR Photodetectors
31
may be operated under a negative voltage bias, in the photoconductive mode, or with no
voltage bias, in the photovoltaic mode.24,28
Dark current in infrared, narrow-bandgap, semiconductor-based photodiodes arises through
mechanisms including band-to-band generation (diffusion current), generation through trap
states (SRH generation current), Johnson-Nyquist noise, band-to-band tunneling across the
depletion region, trap-assisted tunneling through the depletion region, and surface leakage.19
In the following, only diffusion currents, SRH generation currents, and Johnson-Nyquist noise
are treated.
The shot noise model describes the intrinsic noise in infrared photodiodes.19 When operating
in the low-frequency regime, the mean square noise current is
In2 = 2q Isat exp qV
kT
+ Isat
∆f , (1.6.6)
where Isat is the saturation current and V is the applied voltage bias. Under the additional
constraints of thermal equilibrium, no applied voltage, and no incident radiation, the mean
square noise current consists of only Johnson-Nyquist noise,
In2 =
4kTRo
∆f . (1.6.7)
Comparison of Equations (1.6.6) and (1.6.7) gives the saturation current,
Isat =kTqRo
. (1.6.8)
Diffusion current in p-n junctions is
ID = Isat exp qVkT
−1
, (1.6.9)
1.6 Semiconductor-based MWIR Photodetectors
32
and SRH generation current, in which Wdep is the width of the depletion region and with the
assumptions that the Fermi and trap energies coincide and the electron and hole lifetimes are
both equal to τo, is19,43
ISRH =qWdepni
2τ o
A. (1.6.10)
An expression for D* is obtained using Equations (1.4.2) through (1.4.4),
D* =ηqλhc
A∆fIn
. (1.6.11)
Using Equation (1.6.11), two equivalent expressions for the maximum value of D* for a p-n
junction photodiode under zero voltage bias and in the absence of incident radiation are,19
Dpeak* =
ηqλc
2hcRoAkT
1/ 2
(1.6.12a)
and
Dpeak* =
ηλc
2hcq
Jsat
1/ 2
, (1.6.12b)
in which Jsat is the saturation current density.
Infrared photodetection systems, particularly those employing large FPAs, frequently
incorporate p-n junction photodiodes, which have a number of benefits over photoconductive
photodetectors.2 Stringent noise requirements can be more easily met. Only generation,
rather than both generation and recombination, processes are significant in the p-n junction
photodiode. This difference alone results in the noise in p-n junction photodiodes being
better than that in photoconductive photodetectors by a factor of
2 .19,26 In addition, bias
currents in p-n junction photodiodes are low or negligible, which results in lower thermal
power dissipation, and coupling to the input state of a silicon CCD benefits from the naturally
high impedance of the devices. Photodiodes also permit higher-frequency operation, biasing
is simpler, and responsivity can be more accurately predicted.2,7
1.6 Semiconductor-based MWIR Photodetectors
33
The preceding discussion of dark current, noise current, and D* does not consider surface
leakage current, however the primary contributors to the noise current in narrow-bandgap
and cooled p-n junction photodiodes are surface leakage and SRH generation currents.
Surface leakage current has an ohmic current-voltage relationship, and it is approximately
temperature independent. It is enabled by a uniformly n-type layer that covers the surface of
narrow-bandgap semiconductors, regardless of the doping in the bulk. This surface layer
arises due to the presence of surface charges. Surface leakage current negatively impacts
the performance of the p-n junction photodiode, as charge carriers are able to use surface
leakage channels, which run parallel to the depletion region, to bypass the potential barrier.
The surfaces of p-n junction based devices are routinely passivated, through the application
of native oxides or other insulators, to reduce surface leakage current and surface
recombination and to protect the devices against undesired environmental effects.7
Passivation is a time consuming, and hence undesirable, process, and one suitable for
commercial InAs devices has not yet been developed.
While it is possible to mitigate the effects of surface leakage current, SRH current remains a
primary source of dark current in cooled photodiodes. The bulk of the SRH current is
generated within depletion region, and, as seen in Equation (1.6.10), the magnitude is
proportional to both the volume of the depletion region and, through the presence of ni, the
generation rate, GSRH. The key to minimizing SRH current is removing the Fermi level from
the middle of the bandgap. The Fermi level may be moved from the middle of the bandgap
by designing a photodetector composed of n-type or p-type material and without a depletion
region. The nBn detector, discussed in the next section, is designed to suppress SRH
current by displacing the Fermi level from the middle of the bandgap.6
1.6.6 The nBn Photodetector
The InAs / AlAsSb nBn photodetector is shown in Figure 1.6.3 as a representative of the
class of nBn photodetectors.6 The name of the nBn photodetector class derives from the n-
type absorption layer, the Barrier layer, and the n-type contact layer. Signal photons
interacting with the absorption layer result in band-to-band excitations, and signal current is
generated when the resulting holes travel to the contact. The nBn photodetector design
stipulates that there be no energy discontinuity in the valence band, and consequently no
barrier to the free flow of minority carrier holes. However, the intentionally large conduction
1.6 Semiconductor-based MWIR Photodetectors
34
band energy discontinuity, which is much larger than kT, blocks the flow of the majority
carrier electrons. In the InAs-based nBn photodetector, both the zero valence band and the
large conduction band energy discontinuities may be achieved by an AlAsxSb1-x barrier layer
with a specific composition, x, which is found to be 0.14 < x < 0.17. The composition of the
AlAsSb barrier layer is the subject of Chapter 2. A pixel is defined by etching through the
contact layer; the barrier layer acts as an etch stop. Gold metal contacts are deposited on
top of the contact layer. The nBn photodetector operates under a voltage bias, and the bias
applied to the device shown in Figure 1.6.3 is referred to as a reverse bias.
Figure 1.6.3 Sketch of the voltage-biased nBn photodetector. The fabricated device structure is shown at the upper left, and the energy diagram is beneath. The contacts are gold, the substrate, absorption, and contact layers are n-type InAs, and the barrier layer is AlAsSb. In the energy diagram, the contact layer is located to the right of the barrier layer.
The design of the nBn photodetector suppresses SRH generation current and reduces
surface leakage current to negligible levels.6 Nowhere does the Fermi energy coincide with
the middle of the bandgap, which significantly reduces the SRH generation rate. Surface
leakage current is essentially eliminated by the inclusion of the wide bandgap barrier layer. It
1.6 Semiconductor-based MWIR Photodetectors
35
breaks the surface conductivity path that would otherwise exist between the absorbing and
contact layers, which are composed of narrow bandgap semiconductors in MWIR
photodetectors. The wide bandgap barrier layer offers no shunt path for mobile majority
carriers to use to bypass the large conduction band energy discontinuity. Measurements of
temperature-dependent dark current presented in Chapter 3 show diffusion current, which
results from the thermal generation of carriers across the bandgap in the absorption region,
dominates the dark current of the lattice-matched InAs-based nBn photodetector. As a
consequence, the cooled nBn photodetector has significantly lower levels of dark current than
the cooled p-n junction photodiodes tested for comparison.
Figure 1.6.4 Temperature dependent D* of SRH generation and diffusion currents. Calculations of ISRH (blue) and Idiff (maroon) assume unintentionally doped InAs with good crystalline quality. D* is proportional to (ni)-1 and (ni
2/Nd)-1 respectively, and Nd =1x1016cm-3.
Auger 1 generation processes are the primary sources of dark current in nBn photodetectors.
The diffusion current may be expressed as
1.6 Semiconductor-based MWIR Photodetectors
36
Idiff ∝qpo1
τ diff
AL = q ni2
Nd
1τ diff
AL, (1.6.20)
where Idiff is proportional to ni2 as expected, τdiff is the carrier lifetime, Nd is the n-type doping
density, and L is the width of the neutral region.6,28 The equations for the SRH generation
current, Equation (1.6.10), and the diffusion, or Auger, noise current, Equation (1.6.20), may
be directly compared. The lifetime tdiff is reported to be approximately 1 microsecond in
unintentionally doped InAs,7 and tSRH, is reported to be approximately 200 ns in high quality
InAs0.91Sb0.09.64 This factor of five difference and the roughly order of magnitude difference
between lengths L and Wdep are not considered significant to this analysis.6,24,28,34 Given this,
the SRH generation current, ISRH, which is the primary source of dark current in p-n junction
photodiodes, may be taken as being proportional to ni, and Idiff to ni2/Nd. This results in the D*
of the nBn photoconductor exceeding that of the p-n junction photodetector over a wide range
of temperatures, as is seen in Figure 1.6.4. The reduced noise in nBn photodetectors
benefits applications that support cooling for optimal performance and those requiring
operation at warmer device temperatures.6
The potential to increase the D* of a photodetector through cooling exists only if the
internally-generated components of dark current are greater than the current produced by the
detection of the incident background radiation. When the internally-generated and
temperature-dependent dark current of a photodetector has been reduced, through cooling,
to a magnitude equal to that arising from detection of the background radiation, the detector
has achieved BLIP operation. Reducing the temperature of the device below the BLIP
temperature does not increase performance. For temperatures cooler than the BLIP
temperature, the primary component of current is a product of the detection of the
background radiation, which originates from an external stimulus and is not dependent on the
temperature of the photodetector.40
While it is routine to cool photodetectors when logistics permit, reducing cooling requirements
remains a priority. Photodiodes and nBn photodetectors operating at BLIP temperatures
deliver similar performance, but the BLIP temperature of the nBn photodetector is higher than
that of the photodiode.6 The difference in the two BLIP temperatures derives from the
performance of the photodiodes being limited by SRH dark current while the performance of
the nBn photodetectors is limited by diffusion current.23 Conventional MWIR photodiodes are
usually cooled to below 200 K,3 and the BLIP temperatures of at least some commercial InAs
1.6 Semiconductor-based MWIR Photodetectors
37
MWIR photodiodes are below 188 K.22 It is typical to operate InSb MWIR photodiodes at 77
K. Focal plane arrays consisting of these photodiodes are known to suffer from unacceptable
levels of dark current at temperatures in excess of 80 K.2 This is in contrast with InAs-based
nBn photodetectors, which have been observed to achieve BLIP operation when cooled to
only 230 K.6
1.7 Purpose of Current Research
Initial measurements of the InAs-based nBn photodetector have been published,6 but the
device has not been fully characterized. The purpose of this research is to better
characterize this MWIR infrared photodetector. A series of InAs / AlAsSb nBn photodetectors
are fabricated from sample crystal growths performed using molecular beam epitaxy.
Measurements, which include taking current-voltage (I-V) traces of selected pixels as a
function of temperature, are made on these photodetectors. These data are analyzed,
compared with the results of mathematical modeling, and contrasted with similar data taken
for p-n junction photodiodes. Conclusions are made on the subjects of the barrier
composition of AlAsSb (Chapter 2), surface leakage current (Chapter 3), the effects on dark
current in lattice-mismatched InAs-based nBn photodetectors possessing absorption layers
with high densities of dislocations (Chapter 4), and the temperature-dependent lateral
diffusion lengths in the absorption layers of various InAs-based nBn photodetectors (Chapter
5). Chapter 6 contains a summary of this research and suggestions for future work.
Molecular beam epitaxy (MBE)20,65 is a technique for growing crystalline epitaxial layers on
prepared substrates. In the Molecular Beam Epitaxy Laboratory, crystal growths are
performed on Riber 32P and Varian Gen II MBE systems. The crystalline epilayers consist of
III-V semiconductor compounds, and the work presented here is based exclusively on the
arsenide-antimonides. The growths are conducted under ultra-high vacuum conditions,
which are better than 10-9 Torr, and at growth rates of approximately 1 micron/hour. The
ultra-high vacuum conditions limit the levels of residual background gas, and thus the rate at
which contaminants may be incorporated into the growing layer. Ultra-high vacuum
conditions limit the growth rate of contaminants a factor of 10-5 times the growth rate of the
desired semiconductor compound. Vacuum is maintained on the Riber and Gen II through a
combination of ion pumps and cryopumps, as well as shrouds filled with liquid nitrogen.
Advantages of using MBE for crystal growth include the ability to tightly constrain the growth
1.7 Purpose of Current Research
38
conditions, the repeatable production of high-purity epilayers with finely controlled
compositions, and the ability to monitor and alter the growth conditions while the growth is in
progress.
Figure 1.7.1 Cut-away illustration of the growth chamber of the Riber 32P.
Figure 1.7.1 shows a cut-away illustration of the growth chamber of the Riber 32P MBE
machine.66 There are differences between the Riber 32P and Varian Gen II MBE systems,
but they are fundamentally similar. With the exception of a clearly identified commercially
sourced photodiode, all devices evaluated in this work are products of either the Riber 32P or
the Varian Gen II as described in this section.
The Riber accepts up to 3” wafer substrates, while the Gen II is limited to 2” and smaller
substrates. One eighth to one quarter pieces of 2” wafers are used as substrates for the
majority of crystal growths conducted for this research. The substrates are acquired epi-
ready from outside vendors and are loaded into the machines without being subjected to
1.7 Purpose of Current Research
39
additional chemical treatments. Indium-free mounting of the substrates on the substrate
mounting blocks is an option in both systems, but all crystal growths conducted as part of this
work are on substrates mounted with indium; the substrates are mounted on a silicon carrier
wafer when the growths are conducted in the Riber and on a solid molybdenum block when
crystal growth occurs in the Gen II. In Figure 1.7.1 the substrate is orange, circular in shape,
and located in the center of the growth chamber. The mounting block is depicted by the thin
cylindrical ring behind it, and this assembly is held by the L-shaped manipulator arm. In the
image the substrate faces an array of effusion cells, which contain the precursor materials for
the growth.
Both systems are similarly equipped with source material. Valved cracker sources are
capable of delivering As2 and As4; this work uses As2 exclusively. Conventional effusion cells
supply the rest of the source materials, including Sb4 and the group III elements. All source
material is elemental, with the exception of the GaTe precursor, which supplies the n-type
dopant Te. Beryllium is used as the p-type dopant. The source materials are maintained at
precisely controlled temperatures by the combination of the effusion cells and temperature
controllers. Molecular beams evaporate or sublimate from the contents of the effusion cells
and overlap at the growing surface of the substrate. The molecular, or atomic, flux
distribution is highest towards the center of the beam and decreases on either side at a rate
determined by the geometry of the cell. High vacuum conditions, 10-5 to 10-9 Torr, are
sufficient for maintaining the beam nature of the evaporated precursor materials.65
Crystal growth conditions are specified by controlling source and substrate temperatures and
the position and rotation of the substrate in the growth chamber. Growth conditions depend
on the status of the MBE machine, and they evolve as conditions in the machine change.
The temperatures of the sources control the rate of flux of the precursor elements, which
affects the growth rate and quality of the growing crystal. The temperature of the substrate is
monitored in a few ways, and all are used together to ensure repeatable growths. A
pyrometer provides an estimated temperature reading of the surface of the substrate, and a
thermocouple located on the manipulator arm approximates the temperature of the mounting
block. Reflection high energy electron diffraction (RHEED)20,65 is used to monitor the surface
of the growing crystal. In Figure 1.7.1 the RHEED gun is found at the lower left, protruding
from the growth chamber at a position just above the ‘Riber’ label. The electron beam
emitted by the RHEED gun intersects the growing crystal surface at a glancing angle. The
resulting forward-scattered diffraction pattern is incident on a phosphor screen, which is
1.7 Purpose of Current Research
40
shown positioned in the circular flange located to the right of the substrate. The diffraction
pattern on the phosphor screen is visible from the outside of the machine, and the pattern
changes as the surface reconstruction of the growing epilayer changes. As the surface
reconstruction changes at certain temperatures and under various growth conditions, the
forward-scattered RHEED diffraction pattern contains information about both the temperature
of the substrate and the quality of the growth.
Chapter 2 investigates AlAsxSb1-x as a barrier layer in the InAs-based nBn photodetector.
Investigations of the composition, x, which possesses the required energy band offsets with
InAs are conducted and compared with the composition of AlAsxSb1-x lattice matching InAs.
Maximum and minimum practical thickness of the barrier layer are explored, using criteria
which include: not exceeding the pseudomorphic critical thickness and minimizing electron
tunneling through the barrier.
Chapter 3 compares an InAs-based nBn photodetector with two InAs-based photodiodes as a
means of investigating the temperature-dependent dark currents in the two types of devices.
It is posited and confirmed that the performance of the photodiodes are limited by surface
leakage currents, while the nBn photodetector contains undetectable levels of surface
leakage current. Conclusions are made about the effects of the composition of dark current
on the D* figure of merit.
Chapter 4 uses temperature-dependent I-V data, measured for InAs-based nBn
photodetectors grown on lattice-mismatched substrates, to investigate the impact of an
absorption layer containing high densities of dislocations on the performance of the nBn
photodetector. The primary contributor to the temperature-dependent dark current of these
devices is the focus of this chapter.
Chapter 5 uses the temperature-dependent I-V data measured for several different InAs-
based photodetectors to estimate the values of the temperature-dependent lateral diffusion
lengths of the minority carriers in the absorbing layer.