Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
~sa~o
AVALANCHE PHOTODIODES AND METHODS FOR THEIR MANUFACTURE
Field o~ the Invention
This invention relates to avalanche photodiodes and
methods for their manufacture.
Backqround of the Invention
Avalanche photodiodes (APDs) are commonly used in
optical fiber telecommuncations systems to convert optical
signals to electrical signals. APDs used for this purpose
should be capable of very high speed operation since optical
fiber telecommunications systems may be operated at very high
data rates. Such APDs should also have a high internal gain
for conversion of weak optical signals to electrical signals
of usable amplitude. Moreover, such APDs should conduct very
little current when no optical signal is present since such
"dark current" is a noise component which limits the usable
sensitivity of the APD.
The most basic APD structure comprises a pn
junction formed in a semicondu~tor material having a band gap
which is less than the energy of the photons to be detected.
The pn junction is reverse biased to set up a large electric
field at the junction. Photons which are absorbed in the
semiconductor material generate carriers which are swept
`~ through the large electric field at the junction and detected
; as a photocurrent. If the reverse bias and the resulting
electric field at the junction are sufficiently large, the
photogenerated carriers acquire enough energy while drifting
through the pn junction region to generate further carriers.
The generation of further carriers provides a photocurrent
gain mechanism which is called "avalanche multiplication".
The avalanche multiplication gain mechanism also
`` operates on leakage currents which result directly from the
reverse bias applied to the pn junction. If the reverse bias
is sufficiently large, the pn junction will "break down" and
conduct very large avalanche multiplied leakage currents. As
~ 35 these leakage currents do not depend on photogeneration of
; carriers, multiplied leakage currents are "dark currents"
which flow even in the absence of an optical signal and which
~k
~98640
therefore degrade the usablé sensitivity of the ~PD. As the
leakage currents increase with increasing reverse bias of the
pn junction, the reverse bias must be made large enough to
support avalanche multiplication of the photocurrent, but
small enough to ensure that the avalanche multiplied leakage
currents do not swamp the avalanche multiplied photocurrent.
In particular, the reverse bias must be less than the reverse
bias required to support avalanche breakdown of the pn
junction.
Unfortunately, the large electric fields which are
required to support avalanche multiplication in the
semiconductor materials which have band gaps appropriate for
absorption at wavelengths commonly used for optical fiber
transmission also tend to cause "tunneling" of carriers at
the reverse biased pn junction. The current due to this
tunneling is a leakage current which is not dependent on
photogeneration of carriers. Thus the tunneling current is a
dark current which degrades the usable sensitivity of the
APD.
Improved APD structures employ separate absorption
and multiplication regions to reduce the dark current due to
tunneling. The absorption region comprises a layer of
semiconductor material which has a band gap appropriate for
absorption at wavelengths commonly used for optical fiber
transmission. The multiplication region comprises a pn
junction formed in a semiconductor material having a band gap
which is wider than the band gap of the semiconductor
material forming the absorption region. The pn junction is
reverse biased to set up a large electric field at the pn
junction. Carriers which are photogenerated in the
absorption region drift into the multiplication region where
they are swept through the reverse biased pn junction by the
large electric field and detected as a photocurrent. If the
reverse bias and the resulting electric field are
sufficiently large, avalanche multiplication of the
photocurrent occurs near the pn junction. The wider band gap
semiconductor which forms the multiplication region is less
prone to tunneling than the narrower band gap semiconductor
~L29~
of the absorption region, even at the relatively larger
electric fields required to support avalanche multiplication
in the wider band gap semiconductor.
In devices having separate absorption and
multiplication regions, the multiplication region can be made
by forming a doped well of one conductivity type in a
semiconductor layer of an opposite conductivity type. The
doped well has finite lateral dimensions and therefore
defines a pn junction having a central portion which extends
generally parallel to the major dimensions of the
semiconductor layer in which the doped well is formed, and
peripheral portions which extend generally perpendicular to
the major dimensions of the semiconductor layer in which the
doped well is formed. The peripheral portions of the pn
junction meet the central portion at "corners" or "edges".
When this pn junction is reverse biased, the electric field
is greater at the edges than at the central portion of the pn
junction. When the electric field is large enough to support
a desired level of avalanche multiplication at the edges of
the pn junction, it may not be large enough to support
avalanche mul~iplication at the central portion of the pn
junction, and the overall photocurrent gain will be limited.
Moreover, when the electric field is large enough to support
avalanche multiplication of the photocurrent at the central
portion of the pn junction for large overall gain, the
electric field at the edges of the pn junction may be large
enough to cause avalanche breakdown at the edges of the pn
junction. Thus, while separation of the absorption and
multiplication regions reduces dark current due to tunneling
at the pn junction, it does not eliminate dark current due to
premature breakdown at the edges of the pn junction.
Moreover, the electric field in the absorption region beneath
the edges of the pn junction may be large enough to support
tunneling in this part of the absorption region, and this
localized tunneling is a further source of dark current.
The concentration of electric field at the edges of
the pn junction has been reduced by etching the
multiplication region around the central region of the pn
~29~;4~
junction to remove the peripheral portions and edges of the
pn junction, leaving a mesa structure with a plane pn
junction extending across the mesa. Unfortunately, the
etched surfaces are difficult to passivate reliably, and the
required etching steps complicate integration of mesa type
APDs with other electronic and optoelectronic devices.
Moreover, the etched mesas generally have sloped sidewalls,
and charge balancing requirements at the pn junction near the
sloped sidewalls cause some electric field concentration near
the sloped sidewalls even though the edges of the pn junction
have been removed. Such electric field concentration may
cause premature avalanche breakdown near the sloped sidewalls
of the mesa when the APD is reverse biased for avalanche
multiplication at a central portion of the pn junction.
Alternatively, the concentration of electric field
at the edges of the pn junction has been reduced by forming
one or more a~nular doped wells at the periphery of the doped
well which defines the pn junction. These annular doped
wells are called "guard rings" and effectively round the
edges of the pn junction by providing a less abrupt doping
junction. Unfortunately, the guard rings must be very
accurately placed relative to the doped well which defines
the central portion of the pn junction, and this is very
difficult to achieve in practice.
P.P. Webb et al have reported another APD structure
employing separate absorption and multiplication regions
(SPIE, Vol. 839, Components for Fiber Optic Applications II
(1987)). In this structure, the multiplication region is
very lightly doped except for a heavily doped well which
defines a pn junction as described above, and a highly doped
charge sheet which is located between the central portion of
the pn junction and the absorption region. The charge sheet
does not extend between the edges of the pn junction and the
absorption region, so that the multiplication reglon has a
relatively high concentration of carriers per unit area
between the central portion of the pn junction and the
absorption region due to the charge sheet, and a relatively
low concentration of carriers per unit area between the edges
~9~o
of the pn junction and the absorption region. As will be
explained in greater detail below, this distribution of
carriers in the multiplication region between the pn junction
and the absorption region can be arranged so as to ensure
that there is a larger electric field at the central portion
of the pn junction than at the edges of the pn junction when
the pn junction is reverse biased. Consequently, the pn
junction can ~e reverse biased for good avalanche gain near
the central portion of the pn junction without causing
avalanche breakdown at the edges of the pn junction.
Webb et al note that the charge sheet in
combination with the light background doping of the
multiplication region is e~fective in eliminating dark
current due to avalanche breakdown at the edges of the pn
junction only if the curvature of those edges is fairly
gradual. Webb et al used junction depths exceeding 3
micrometers ts achieve the required gradual curvature at the
edges of the pn junctions.
While the distribution of carriers in the
multiplication region of the Webb et al APD ensllres that
there is a larger electric field at the central por~ion of
the pn junction than at the edges of the pn junction, this
carrier distribution also ensures that there is a larger
electric field in peripheral portions of the absorption
region which are aligned with the edges of the pn junction
than in a central portion of the absorption region which is
aligned with the central portion of the pn junction and the
charge sheet. In particular, the electric field in the
peripheral portions of the absorption region may be large
enough to cause tunneling in these portions when the reverse
bias on the pn junction is larye enough to support good
avalanche gain near the central portion of the pn junction.
Such tunneling is a source of dark current which degrades the
usable sensitivity of the APD.
Webb et al also disclose a method for making the
APD structure which is described above. According to the
disclosed method, an n~ InGaAs absorption layer is grown on
- an n-type InP substrate, and a thin n~ InP layer is grown on
~98~
the InGaAs layer. Si is implanted into a central portion of
the thin InP layer to define an n+ charge sheet in the thin
InP layer. Further n~ InP is then grown to thicken the InP
layer, and Cd is diffused into the InP layer to define a p~
doped well in the thickened InP layer over the Si implant.
The doped well defines a pn junction which, when reverse
biased, provides the large electric fields necessary to make
the InP layer a multiplication layer. The pn junction has a
central portion which extends generally parallel to the major
dimensions of the InP layer and-peripheral portions which
extend generally perpendicular to the major dimensions of the
InP layer and which meet the central portion at "corners" or
"edges". The doped well is sized and placed so that the
central portion of the pn junction is over the implanted
charge sheet, but the edges of the pn junction are not over
the implanted charge sheet.
Summar~v of the Invention
This invention provides an APD design which permits
further reduction of dark currents due to tunneling~ This
invention also provides novel methods for making APDsO
One aspect of the invention provides an avalanche
photodiode having separate absorption and multiplication
; regions, said photodiode having a first charge sheet located
between the absorption region and a central portion of a pn
junction of the multiplication region, and a second charge
sheet located between the absorption region and edges of the
pn junction, said second charge sheet having a lower doping
concentration per unit area than said first charge sheet.
The use of two charge sheets facilitates
independent control of the electric fields at central and
peripheral portions of the multiplication region and
absorption region for optimized design. In particular, the
use of two charge sheets facilitates the design and
manufacture of an APD in which good avalanche gain can be
achieved in the central portion of the multiplication region
without avalanche breakdown in the peripheral portions of the
multiplication region or tunneling in the peripheral portions
of the absorption region.
Stated in more detailed structural terms, one
aspect of the invention pxovides an avalanche photodiode
comprising an absorption layer of a semiconductor having a
first band gap and a first conductivity type and a
multiplication layer of a semiconductor having a second band
gap exceeding the first band gap. The multiplication layer
contains a first region of the first conductivity type which
has a first doping concentration per unit volume. The
multiplication region also contains a second region of a
second conductivity type opposite to the first conductivity
type which is surrounded by the first region. The first and
second regions together define a pn junction having a central
portion which extends generally parallel to major dimensions
of the multiplication layer and peripheral portions which
extend generally perpendicular to the major dimensions of the
multiplication layer. The peripheral portions meet the
central portion at edges of the pn junction. The
multiplication layer further contains a third region of the
first conductivity type having a third doping concentration
per unit volume which exceeds the first doping concentration
per unit volume and a first doping concentration per unit
area. The third region is located between the edges of the
pn junction and the absorption layer. The multiplication
layer also contains a fourth region of the first conductivity
type which has a fourth doping concentration per unit volume
which exceeds the first doping concentration per unit volume
and a second doping concentration per unit area which exceeds
the first doping concentration per unit area. The fourth
region is located between the central portion of the pn
junction and the absorption layer. The avalanche photodiode
further comprises an electrical contact to the second region
of the multiplication layer and an electrical contact to the
absorption layer.
Another aspect of the invention provides a method
for making an avalanche photodiode according to the
invention. The method comprises the steps of forming an
9L;29~6~
absorption layer of a semiconductor having a first band gap
and a first conductivity type, forming a multiplication layer
of a semiconductor having a second band gap exceeding the
first band gap, with the multiplication layer containing
first, second, third and fourth regions as described above,
and forming electrical contacts to the second reyion of the
multiplication layer and to the absorption layer.
The third doping concentration per unit volume may
equal the fourth doping concentration per unit volume, and
the fourth region may be thicker-than the third region, so
that the second doping concentration per unit area exceeds
the first doping concentration per unit area.
Another aspect of the invention provides a method
for ma~ing an avalanche photodiode which comprises the ~teps
of forming an absorption layer of a semiconductor having a
first band gap and a first conductivity type and forming a
multiplication layer of a semiconductor having a second band
gap exceeding the first band gap. The multiplication layer
is formed by forming a first sublayer of the semiconductor
having the second band gap and a first doping concentration
per unit volu~e of the first conductivity type and
preferentially removing a peripheral portion of the first
sublayer. A second sublayer of semiconductor having the
second band gap and a second doping concentration per unit
volume of the first conductivity type is formed over a
remaining portion of the first sublayer. The second doping
concentration per unit volume is less than the first doping
concentration per unit volume. A doped well of a second
conductivity type which is opposite to the first conductivity
type is formed in the second sublayer. The doped well has
lateral boundaries which are disposed over a region where the
first sublayer was preferentially removed, and a central
portion which is disposed over a region where the first
sublayer was not preferentially removed. The method further
comprises the step of forming electrical contacts to the
absorption region and to the doped well.
In one embodiment of this method, a peripheral
portion of the first sublayer is preferentially removed to
~L~9~
leave a central portion of the first sublayer surrounded by a
thinner peripheral portion of the first sublayer. The doped
well is formed in the second sublayer with lateral boundaries
disposed over the thinner peripheral portion of the first
sublayer and a central portion disposed over the central
portion of the first sublayer.
The semiconductor sublayers comprising the
multiplication layer may be formed directly on an absorption
layer of semiconductor having the first band gap and the
first conductivity type. Preferably however, a gra~ing layer
is formed on the absorption layer, and the multiplication
layer is formed on the grading layer. The grading layer,
which may comprise one or more carefully selected
semiconductor layers facilitates movement of carriers between
the absorption layer and the multiplication layer.
Brief Description of the Drawinas
Embodiments of the invention are described below by
way of example only. Reference is made to the accompanying
drawings, in which:
Figure 1 is a cross-sectional view of an APD as
disclosed by Webb et al;
Figure 2 is a series of plots showing the doping
profile, fixed charge profile and electric field profiles for
the APD of Figure 1 along section line II-II of Figure 1 when
the APD is reverse biased;
Figure 3 is a series of plots showing the doping
profile, fixed charge profile and electric field profiles for
the APD of Figure 1 along section line III-III of Figure 1
when the APD is reverse biased;
Figure 4 is a cross-sectional view of an APD
according to a first embodiment;
Figure 5 is a series of plots showing the doping
profile, fixed charge profile and electric field profiles for
the APD of Figure 3 along section line V-V of Figure 4 when
the APD is reverse biased; and
1298~4~
Figure 6 is a series of cross-sectional views of
the APD of Figure 4 at successive stages of its manufacture.
Detailed Description of Embodiments
Figure 1 is a cross-sectional view of an APD 100 as
disclosed by Webb et al. The APD 100 comprises an n+ InP
substrate 110 including an n+ InP buffer layer 112, an
absorption region in the fcrm of an n~ InGaAs layer 120 on
the buffer layer 112, and a multiplication region in the form
of an InP layer 130 on the absorption layer 120. The
absorption layer 120 and the multiplication layer together
define a heterojunction 125 at their interface.
The multiplication layer 130 contains a p-type
region in the form of a doped well 132 which is surrounded by
an n-type region 133 of the multiplication layer 130. The p-
15 type doped well 132 and the surrounding n-type region 133
together define a pn junction 135 which has a central portion
136 extending generally parallel to major dimensions of the
multiplication layer 130, and peripheral portions 137
extending generally perpendicular to the major dimensions of
20 the multiplication layer 130. The peripheral portions 137
meet the central portion 136 at edges 138 of the pn junction
135. An n+ region or "charge sheet" 139 is located in the
multiplication layer 130 between the central portion 136 of
the pn junction 135 and the absorption layer 120.
The APD further comprises an annular electrical
contact 140 to the doped well 132 and an electrical contact
150 to the absorption layer 120 via the substrate 110 and
buffer layer 112.
In operation of the APD 100, a reverse bias is
applied to the pn junction 135 v~a the electrical contacts
140, 150, and an optical signal is introduced into the
absorption layer 120 through the opening defined by the
annular contact 140. Photons which are absorbed in the
absorption layer 120 generate electrons and holes. Under the
influence of electric fields set up by the reverse bias, the
photogenerated electrons drift to the lower electrical
contact 150 via the buffer layer 112 and substrate 110, while
~ ;~9~40
11
the photogenerated holes drift into the multiplication layer
130 and across the pn junction 135 to the upper electrical
contact 140. Thus, the photogenerated carriers are detected
at the contacts 140, 150 as a photocurrent.
If the reverse bias and the resulting ~lectric
field at the pn junction 135 are large enough, the
photogenerated holes acquire enough energy while drifting
through the junction region to generate further holes and
electrons, which also drift to the upper and lower contacts
140, 150 respectively. This generation of further carriers
provides a photocurrent gain mechanism which is called
"avalanche multiplication".
Fiyure 2(a) illustrates the doping profile of the
APD 100 along line II-II in Figure 1 which passes through the
central portion 136 of the pn junction 135. The pn junction
135 and the heterojunction 125 are modeled as abrupt
junctions for the purposes of this discussion. In practice
they will be less abrupt than this illustration suggests.
Figure 2(b) illustrates the charge density profile
of the APD 100 along line II-II in Figure 1 when a reverse
bias is applied to the APD 100. This charge density profile,
which is determined by the doping profile and the applied
reverse bias, sets up an electric field profile as
illustrated in Figure 2(c). The maximum electric field
occurs at the pn junc~ion 135. The total area under the
electric field profile corresponds to the applied reverse
bias.
The reverse bias is selected so as to set up an
electric field at the pn junction 135 which is at least EM,
the minimum electric field required to support avalanche
multiplication, but less than EB, the minimum electric field
at which the pn junction 135 breaks down.
Figures 3(a), (b) and (c) illustrate the doping
profile, charge density profile and electric field profile 35 for the APD 100 taken along a line III-III in Figure 1 which
passes through an edge 138 of the pn junction. As
illustrated in Figure 2(f), the curvature of the pn junction
135 at the edges 138, and the absence of the charge sheet 139
~z9~
in the region of the multiplication layer 130 which is
between the edges 138 of the pn junction 135 and the
absorption layer 120 affect the electric field profile. In
particular, the electric field at the edges 138 of the pn
junction 135 ls smaller than the electric field at the
central portion 136 of the pn junction 135, so that the
electric field at the edges 138 of the pn junction 135 is
less than both EB, the minimum electric field required to
support avalanche breakdown at the pn junction 135, and EM,
the minimum electric field requi-red to support avalanche
multiplication when the electric field at the central portion
136 of the pn junction 135 is EM. However, the electric
field in a peripheral portion of the absorption layer 120
beneath the edges 138 of the pn junction 135 is larger than
the electric field in a central portion of the absorption
layer 120 beneath the central portion 136 of the pn junction
135. Thus, the electric field in the peripheral portion of
the absorption layer 120 beneath the edges 138 of the pn
junction 135 may be larger than ET, the minimum electric
field required to support tunneling in the absorption layer,
even though the electric field in the central portion of the
absorption layer 120 beneath the central portion 136 of the
pn junction 135 is less than ET. The resulting dark current
degrades the usable sensitivity of the APD 100.
: 25 Figure 4 is a cross-sectional view of an APD 200
according to a first embodiment of the invention. The APD
200 is similar to the Webb et al APD 100, except that an n+
; region or "charge sheet" 210 is located in the multiplication
layer 130 between the edges 138 of the pn junction 135 and
the absorption layer 120. The n+ region 210 has a lower
doping concentration per unit area than the n+ region 139
which is located in the multiplication layer between the
central portion 136 of the pn junction 135 and the absorption
layer 120.
The APD 200 according to the first embodiment also
includes a grading layer 220 between the absorption layer 120
and the multiplication layer 130. The structure of the
grading layer 220, which facilitates movement of holes
~9~69~
between the absorption layer 120 and the multiplication layer
130, is discussed in greater detail below.
As the doping profile for the APD 200 along line
II-II in Figure 4 is identical to the doping profile along
the line II-II for the Webb et al APD 100, the corresponding
charge density profile and electric field profile for the
modified APD 200 should also be identical to the
corresponding charge density profile and electric field
profile for the Webb et al APD 100 under identical reverse
- 10 bias conditions. In particular, the doping profile, charge
density profile ancl electric field profile for the modified
APD 2û0 along the line II-II of Figure 4 are as shown in
Figures 2(a), (b) and tc~ respectively if the same reverse
bias is applied to the APD 200 as is applied to the APD 100.
Consequently, the same reverse bias should provide the same
avalanche gain at the central portion 136 of the pn junction
135 for both the Webb et al APD 100 and the APD 200 according
to the first embodiment.
Figure 6(a) illustrates the doping profile for the
APD 200 taken along a line V-V in Figure 4 which passes
~: through an edge 138 of the pn junction 135. As this doping
profile differs from the doping profile shown in Figure 3 (a)
for the Webb et al APD 100, the corresponding charge density
- profile and electric field profile for the APD 200 willdiffer from those shown in Figures 3 (b) and (c) respectively
if the same reverse bias is applied to the APD 200 as is
applied to the Webb et al APD 100. In particular, the APD
200 will have a somewhat higher electric field at edges 138
of the pn junction 135 than the Webb et al APD 100 when they
are reverse biased so as to cause the same amount of
avalanche multiplication across the central portiDn 136 of
the pn junction 135. This higher electric field will cause
breakdown at the edges 13~ of the pn junction 135 only if it
exceeds EB, and will cause significant avalanche gain at the
edges 138 of the pn junction 135 only if it approaches EM.
Because of the n+ region 210, the APD 200 will also have a
lower electric field in the absorption layer 120 beneath the
edges 138 of the pn junction 135 than the Webb et al APD 100
~;~98~4~9
14
when they are reverse biased so as to cause the same amount
of avalanche multiplication across the central portion 136 of
the pn junction 135. With appropriate selection of the
doping concentration per unit area for the region 210, the
electric field in the absorption region 120 beneath the edges
138 of the pn junction 135 can be brought below ET, the
minimum electric field for significant tunneling in the
absorption layer 120 without raising the electric field at
the edges 139 of the pn junction 136 above EB, the minimum
electric field for avalanche breakdown at the pn junction, or
EM, the mlnimum electric field required to support avalanche
multiplication at the pn junction 135. Thus, inclusion of
the n~ region 210 may permit reduction in dark currents due
to tunneling in the absorption layer 120 without causing dark
currents due to avalanche breakdown, and, in any case,
provides a further degree of design freedom for minimization
of dark currents.
Figures 6(a) to 6(e) illustrate the APD 200
according to the first embodiment at successive stages in its
manufacture.
The n InP buffer layer 112, n InGaAs absorption
layer 120, grading layer 220, and an n+ InP sublayer 510 are
successively grown on the n+ InP substrate 110 by
MetalOrganic Chemical Vapour Deposition (MOCVD) to provide
the structure shown in Figure 6(a).
The n+ substrate 110 has a Sn doping concentration
of approximately 1019 cm~3. The buffer layer 112 is
approximately 1.3 micrometers thick, and is not intentionally
doped, but has a background doping concentration of
approximately 1015 cm~3. The absorption layer 120 is
approximately 2.5 micrometers thick, has an In to Ga ratio of
0.53 to 0.47, and is not intentionally doped, but has a
background doping concentration of approximately 1015 cm~3.
The grading layer 220 comprises a series of thin
alternating sublayers of InP and InGaAs. The InP layers are
grown progressively thicker from a thickness of 5 Angstrom
units to a thickness of 45 Angstrom units over 18 sublayers,
and the InGaAs sublayers are grown progressively thinner from
~186~L~
a thlckness of 45 Angstrom units to a thickness of 5.5
Angstrom units over 18 sublayers to provide a graded
interface between the InGaAs absorption layer 120 and the InP
multiplication layer. The graded interface smooths band edge
discontinuities which can impede movement of photogenerated
; holes across the interface.
The n+ InP sublayer 510 has a Si doping
concentration of approximately 2.6xl017 cm~3 and is
approximately 0.1 micrometers thick.
A central portion 139 of the n+ InP sublayer 510 is
masked with photoresist and an unmasked peripheral portion of
the layer 510 is preferentially etched by wet etching or
reactive ion etching to remove approximately 0.07 micrometers
of the thickness of the peripheral portion, leaving the
central portion 139 surrounded by a thinner peripheral
portion 210. The structure which remains after the etching
step is shown in Figure 6(b). The central portion 139 of the
remaining n+ InP sublayer 510 defines the n+ region or
"charge sheet" 139 which is between the central portion 136
of the pn junction 135 and the absorption layer 120 in the
finished APD 200. The thinner peripheral portion 210 of the
remaining n+ InP sublayer 510 defines the n+ region or
"charge sheet" 210 which is between the edges 138 of the pn
junction 135 and the absorption layer 120 in the finished APD
200. Because the central and peripheral portions 210, 139
have the same doping concentration per unit volume, but the
peripheral portion 210 is thinner than the central portion
139, the peripheral portion 210 has a lower doping
concentration per unit area than the central portion 139.
The central portion 139 is approximately 0.1 micrometers
thick and approximately 26 micrometers wide, and has a doping
concentration per unit volume of approximately 2.6x1017 cm~3
for a doping concentration per unit area of approximately
2.6x1012 cm~2. The peripheral portion 210 is approximately
0.03 micrometers thick for a doping concentration per unit
area of approximately 7.8xlOll cm~2.
An n~ InP sublayer 520 approximately 2.5
micrometers thick is then grown on the remaining central and
~98~
16
peripheral portions 139, 210 of the n+ InP sublayer 510 by
MOCVD to form the structure shown in Figure 6(c). No n-type
dopant is intentionally added to the InP during growth of the
n~ sublayer 520, but unavoidable impurities provide an n~
doping of approximately 1015 cm~3 in the grown InP. The n~
InP sublayer 520 has a lower doping concentration per unit
area than the central and peripheral portions 139, 210 of the
n+ InP sublayer 510 which are intentionally doped during
growth.
The upper surface of the structure shown in Figure
6(c) is masked with Si3N4 530 deposited by Electron Cyclotron
Resonance (ECR). The Si3N4 530 is patterned using standard
photolithographic techniques, and Zn is diffused through an
opening in the Si3N4 mask 530 to form a p+ doped well 132
approximately 2 micrometers deep and approximately 44
micrometers wide in the n~ InP sublayer 520, as shown in
Figure 6(d). The opening in the Si3N4 mask 530 is aligned
with the central portion 139 of the n+ InP sublayer 510 so
that the doped well 132 has lateral boundaries disposed over
the peripheral portions 210 of the n+ InP sublayer and a
central portion disposed over the central portion 139 of the
n+ InP sublayer.
The upper surface of the structure shown in Figure
6(d) is then passivated with an antireflective coating of
Si3N4 540, an annular contact hole is opened in the Si3N4
passivation 540 over the doped well 132, and Cr/Au is vacuum
evaporated onto the InP surface which is exposed by the
contact hole to form a contact 140 to the doped well, as
shown in Figure 6(e). Cr/Au is also vacuum evaporated onto
the back surface of the n+ InP substrate 110 to form a
contact 150 to the absorption layer 120 via the substrate 100
; and buffer layer 112.
Numerous variations of the APD structure and
fabrication method described above will be apparent to those
familiar with the design and manufacture of APDs. For
example, other crystal growth techniques such as Molecular
Beam Epitaxy (MBE) and Liquid Phase Epitaxy (LPE) could be
used instead of MOCVD to form the semiconductor layers 112,
'
~g~6~0
120, 220, 510, 520. Other semiconductor materials systems
such as the GaAs/AlGaAs system could be used instead of the
InP/InGaAs system. Other p and n type dopants could be used,
the doping polarities could be reversed, or the doping
concentrations and device dimensions could be altered. For
example, the central charge sheet 139 could be disposed
between 0.2 micrometers and 0.5 micrometers below the pn
junction 135. Other dielectrics could be used for
passivation and diffusion masks 530, 540, and other metal
systems could be used for the electrical contacts 140, 150.
The lower contact 150 could be made annular instead
of the upper contact 140 to permit illumination of the
absorption layer through the lower contact 150 instead of the
upper contact 140.
The "charge sheets" 139, 210 could be formed by
implantation into undoped or lightly doped semiconductor,
although this approach would require additional masking,
implantation and annealing steps.
In the growth and etch back method described above,
the peripheral portion 210 of the sublayer 510 could be
entirely removed to leave a single charge sheet 139 under the
central portion 136 of the pn junction 135. This variation
provides a novel method for making the Webb et al APD 100.
In one advantageous variation of the growth and
etch back method described above, the sublayer 510 is formed
as three successively grown strata. The first of these
strata is 0.04 micrometers thick and has a doping
concentration per unit volume of 1017 cm~3. The second
stratum is 0.08 micrometers thick and has a doping
concentration per unit volume of 5X1016 cm~3. The third
stratum is 0.18 micrometers thick and has a doping
concentration per unit volume of 1017 cm~3. A peripheral
portion of the sublayer is etched back 0.2 micrometers to
define a central unetched portion having a doping
concentration per unit area of 2.6x1012 cm~2 and peripheral
etched portion having a doping concentration per unit area of
8Xloll cm~3 Because the etching is terminated in the
relatively lightly doped second stratum, any errors in the
~986~
18
8X1o11 cm~3 Because the etching is terminated in the
relatively lightly doped second stratum, any errors in the
etching depth will have a relatively small effect on the
doping density per unit area of the peripheral portion.
Moreover, because the doping concentrations per unit volume
of all three strata are smaller than in the fabrication
method described above, the etching depth required to define
the desired doping concentrations per unit area is larger.
The larger etching depth provides a taller and more visible
mesa, which is more readily used for alignment of masks in
subsequent processing steps.
Selective epitaxy could also be used instead of
growth and etch back techniques to form the charge sheets
139, 210. For example, a first stratum of the sublayer 510
could be grown across the entire device to form the entire
peripheral charge sheet 210 and part of the thickness of the
central charge sheet 139. A second stratum of the sublayer
510 could then be selectively grown only where the central
charge sheet 139 is desired to complete the central charge
sheet 139. This variation avoids the need for etching back
the sublayer 510.
Other known forms of grading layer could be used in
place of the grading layer 220 described above, or the
grading layer 220 could be omitted, though omission of the
grading layer would likely degrade the high speed performance
of the resulting APD.
These and other variations are within the scope of
the invention as claimed below.
