PROC.

20th INTERNATIONAL CONFERENCE ON MICROELECTRONICS (MIEL'95), VOL 2, N15, SERBIA, 12-14 SEPTEMBER, 1995

Numerical Calculation of Photodetector Response

Time Using Ramo's Theorem
Z. Subjar, Z. Djuri6, M. Smiljanie, 2. Lazi6

Absrrw- The work presents a methodologyfor numerical valley with electric field increase. These effects are signi-
calculation of response time of p-i-n photodiodes using the ficant if the response time is about 5ps. Averaged, these
corpuscular approach, i.e. the Ramo's theorem. The calcula- effects can also be taken into account by using empirical
tions were performed for the Si p-i-n photodiode homostruc-
dependencies of mobility on electric field [5,6,8,9].
ture which is now already standard and for a high-speed
M-InGaAs-InP photodetector heterostructure. The results Another approach to the calculation of response time in
show a good agreement with the results obtained using the the depletion layer is the corpuscular one, using the Ra-
usual collective approach, however in our calculation the mo's theorem [2-4]. M e r a number of contradictory state-
numerical calculation is significantly simplified, and the phy- ments [3,4] it may be taken as proved that the Ramo's
sical presentation of the process is extremely simple. theorem is applicable for determination of impulse respon-
se in depletion layer. In [2] a method was presented for
I. INTRODUCTION determination of impulse response of photodetectors and
applied to the case when it was assumed that the velocity
It is well known that the performance of modern photo- of carriers in the depletion layer was constant.
detectors is determined by three basic parameters: sensiti- In this work the Ramo's theorem is generalized and
vity, response time and noise equivalent power (NEP). In applied to the case when mobility depends on electric field.
practical situations it is almost always necessary to achieve The basic advantages of this approach are a very clear
maximum product of sensitivity and reciprocal response physical presentation of the process and an extremely
time. simple numerical procedure.
Optical communication applications require very fast de-
tectors (>lGHz), therefore they are fabricated in materials 11. THEORY
with high carrier mobility (GaAs, InGaAs, etc.). The res-
ponse time in the depletion layer of a photodetector is The structure of a p-i-n photodiode is shown in Fig. 1.
much shorter than the response time in the quasineutral
pi Si or pi I ~ P n Si orn InP
regions. These photodetectors are therefore designed to
achieve maximum of optically generated carriers in the
depletion layer.
A number of works handles the problem of determina-
tion of response time [l-91. All the applied methods use
Y$.+jTL
(window)

n-Si or n-InGaAs
(Buffer)

two approaches, the collective and the corpuscular appro- 0 d

ach. The collective approach [1,6-91 uses the usual system

Fig. 1. Schematical presentation of a p-i-n structure.
of semiconductor kinetic equations and the Poisson's
equation. The solutions of this system for different optical According to the generalized Ramo's theorem [2], a mo-
excitations determine time dependence of some of its para-
ving charge q k generated in the coordinate %(f) in the
meters, usually the short circuit current. Generally, it is depletion layer causes in the moment t the appearance of
assumed here that carrier mobility depends on electric the charge Q k at the edge of the depletion layer, given by,
field.
An especially interesting consideration of this problem
for the case of two-valley semiconductor was published in
[7], where the collective approach was generalized to take
into account the nonstationary effects of electron transfer
where AVk is the voltage drop at the depletion layer due to
from the higher mobility valley into the lower mobility
generation of the charge Qk.The short circuit current
(A&=O) is obtained by simple differentiation of (1)
2. SuSnjar, 2. DjuriE and 2. L a i d are with IHTM - Institute of
Microelectronic Technologies and Single Crystals, the Univer-
sity of Belgrade, NjegoSeva 12, 11000 Belgrade, Yugoslavia,
E-mail eirntOOl@yuss21.bg.ac.yu

0-7803-2786-1/95/$4.00 0 1995 IEEE

71 7
 Total short circuit current is calculated using the super- where qb and qb denote the carrier package given by
position principle:

(3) A is the photodiode area, q is electron charge, a is the ab-

sorption coefficient, Axo is the width of the package, and
Here v&) denotes the effective velocity caused by electric Fo the number of photons falling onto a unit area of
field and diffusion. The diffusion term in velocity can photodetector. To determine the position of qkn or qkp for a
usually be omitted because it becomes neglectable as soon package generated in the point XkO it is necessary to find
as the potential across the region becomes higher than the package coordinate in the moment Q=to+jAt.The re-
kT/q [8]. Thus the drift velocity in (3) becomes current relation for determination of the electron package
position is

xw = X k n ( j - l ) + v n [ ~ ( X k n ~ j - ~ ~ ) ] ~ (7)

where pk(E) is mobility, and E the electric field. with the initial condition
To consider the influence of capacitance of the depletion
region on the response time we will suppose that the
resistor R is connected to the photodetector (R includes the
built-in detector resistance). The voltage change due to
and analogously for the position of hole packages. In this
carrier generation is then AV=R-i,thus using (1) we obtain
way (3) becomes

(5)

where io(t) is given by (3). In obtaining (3) we neglected

the dependence of the depletion layer width on time.
Let us consider now the response from the depletion
layer to the impulse light excitation F&(t). The charge 111. NUMERICAL,
CALCULATION

carrier distribution immediately &er the illumination To perform calculation using (8) it is necessary to know
(go) is shown in Fig. 2a, while Fig. 2b shows the charge the explicit dependence of electric field on position and the
carrier distribution for a given P O .
dependence of the drift velocity on electric field.
In this work we consider two cases, the standard silicon
p-i-n structure for which we assumed that the electric field
is given by the expression [13

d2W
E=
.(x-W eff ) j"; =
d<W
(9)

Fig. 2. Distribution of charge carriers after illumination FoS(t)e"

in depletion layer, (a) t=O and (b) 0 0 . . For the drift velocity of electrons and holes for silicon we
used the empirical relation [111:
Assuming that the number of optically generated carriers
remains small enough not to perturb the existing electric
field E which can be obtained by solving the Poisson's
equation, using (3) we may write

where ~ = 1 3 3 0 c m ~ N sb=495cm2Ns,
, Ecn=8000 Vlm,
Ecp=19500 Vlm, p,,=2, Pp=l and FU-I-U. is the s u m of
the applied voltage and the contact potential difference.

71 8
 Another structure considered here belongs to the p-i-n
structures [8] used for new high bit rate optical communi-
cafion systems. It consists of an intrinsic absorbing layer of
InGaAs between two InP doped layers, p’ and n,
respectively. The determination of the expression for elec- where p,,=10.500cm2Ns, h=420cm2Ns, v,p6.106cm/s,
tric field in such a structure is much more complex than in v , ~ = ~ . S * ~ O ~p=7.4.10-’0(V/~~)-22.s,
C~/S, ~2.5.
homostructures. Fig. 3 shows the distribution of charge
carriers and electric field in the mentioned structure. The numerical calculations were performed for a Si
p-n--n structure, where concentration in the weakly doped
n- region was N~10’4cm’3, and the region width
d=l50pm. The time dependence of normalized short circu-
it current was calculated first for the impulse excitation
F&(t); then the simple integration was applied to obtain
response for the Heaviside unit excitation. Fig. 4 shows the
calculation results for three different values of potential t + ~
at the wavelength of Nd:YAG laser radiation. The intrin-
sic absorption coefficient was a=13.25cmm1[13].

1.o

Fig. 3. Schematic presentation of charge carrier and electric field 0.8

distribution in Irs-InGaAs-EP p-n--nheterostructure.

E Od
.3
For the case of depletion of the absorption region with $ 0.4

the width d, the electric field is given by [SI: 3

0.2

0.00 1 2 3 4 5
--x <x<o t Ins1
P

Fig. 4. Normalized short circuit current for silicon photodiode,

d=150 pm, N~=l0”~ m ’N~~, = l o ”cm-3

The rise time was calculated (0.1-0.9) using the Heavi-

side response. It is shown in Fig. 5 for different voltages.
7

where 8

T 5
.5

.-
.-E 4

.c 3

0 20 40 80 80 100 120 140 180

Y IVl
Fig. 5. Rise time dependence on bias for silicon photodiode,
d=150 pm N~=lo’~ ~ m -N~=l0’*
~, ~ m -N.=1017
~, ~m-~.
EF13-Eo [lo], ~=12.4.~0[12].
The empirical relations for electron and hole velocities In the curves in Fig. 5 two characteristic parts can be
in InGaAs are [SI seen., The first part concerns voltages for which the
depletion region width is smaller than the weakly doped
region width d, and the other part denotes voltages higher
than depletion voltage. In our case the depletion voltage
was 40V.

71 9
 Similar calculations were performed for another (I*- IV. CONCLUSION
InGaAs-InP) photodiode structure, where the weakly
doped region was 3pm wide, N+-10'5cm-3, N~=lO'*cm-~ Contrary to the highly complex mathematical models
and ND=10'7cm'3. The absorption coefficient at h=1.3pm [6-91 based on the collective approach in determination of
was 1.16.1O6m-*[lo]. Fig. 6 shows separately response dynamic response, in this work we used the simple
time of electrons, t, (dashed line), holes, tp (dotted) and corpuscular approach to arrive at the same results. We
resultant response time, tc (solid). confirmed it by comparison of numerical calculations for
the case when the dependence of carrier velocity on
electric field was taken into account for two basic structu-
res: a Si p-i-n structure and a 10s-InGaAs-InP p-i-n hete-
rostructure.

REFERENCES

[l] Z. DjuriC, B. RadjenoviC, "Rise Time of Silicon p-i-n Pho-

todiodes", SoIid St. Electronics, 26, 12, pp. 1143-1149,
1983
[2] Z. Djurib, M. SmiljaniC, B. RadjenoviC, The Application
of Ramo's Theorem to the Impulse response Calculation of
a Reach-Through Avalanche Photodiode", Solid St. Electro-
0 2 4 6 8 1 0 1 2 1 4
\Y [VI nics, 27, 10, pp. 833-835, 1984
[3] P. De Visschere, "The Validity of Ramo's Theorem", Solid
St. Electronics, 33,44, pp. 455-459, 1990
Fig. 6. Rise time dependence on bias for InP-InGaAs-InPphoto- [4] W. Dabrowski, 'Transport Equations and Rarno's Theorem.
diode, d=3pn. Applications to the Impulse Response of a Semiconductor
Detector and to the Generation-RecombinationNoise in a
It can be seen that the resultant response time has a Semiconductor Junction", Progress in Quantum Electro-
minimum for low bias, which was presented in a different nics, 13,3, pp.233-266,19?9
manner in [SI and [7]. [ 5 ] M. SmiljaniC, Z. DjuriC, Z. LaziC, "Electron Transit Time
To compare our calculation to other authors' results [8], Through Depletion Layer of G W s p a Junctions", Elec-
tronicsLetters, 25,2, pp. 150-152,1989
we determined the normalized frequency response using
[6] J. W. Lee, D. M. Kim, "Analytic Time Domain Charac-
the Fourier transform of the impulse response. In our terization of p-i-n Photodiodes: Effect of Drift,Diffusion,
calculation we included the influence of the RC constant, Recombination and Absorption", J. Appl. Phys, 71, 8, pp.
according to ( 5 ) (R=50Q C=lOO€F). The width of the 2950-2958,1987
weakly doped region d was parameter (Fig. 7). We can see [7] J. B. Radunovit, D.M. Gvozdit, 'Nonstationary and Non-
that the agreement between the results presented in [8] linear Response of p-i-n Photodiode Made of a Two-Valley
(Fig. 8) and ours in Fig. 7 is very good. Semiconductor",IEEE Trans. Electron Devices, 40, 7, pp.
1238-1244,1993
[8] R. Sabella, S. Merle, "Analysis of InGaAs P-I-N Photo-
diode Frequency Response", IEEE J. Quantum Electronics,
29,3, pp. 906-916, 1993
1.o
[9] G. George, J. P. Krusius, ''Dynamic Response of High-
0.8 d=l.
,
= . . ~ ~
.....
Speed PIN and Schottky-Barrier Photodiode Layers to
Nonuniform Optical Illumination", J. Lightwave Technol.
12, 8, pp. 1387-1393, 1994
0.6
E [lo] J. B.D. Soole, H. Schumacher, H. P. LeBlanc, R. Bhat,
I
M. A. Koza, 'IHigh-Frequency Performance of InGaAs Me-
0.4
tal-Semiconductor-MetalPhotodetectors at 1.55 and 1.3 pm
Wave1engths"Appl.Fhys. Lett 55 (S), 21 August 1989
0.2
[l 11 M. Kurata "Numerical Analysis for Semiconductor Devi-
ces" Lexington Books, 1982 by D. C. Heath and Company
0.0
20 40 80 100 Lexington.
f [GH~'
[12] S. M. Sze "F'hysics of Semiconductor Devices", John
Wiley & Sons 1981
[13] P. P. Webb, R. J. Mchture, and J. Conradi "Properties of
Fig. 7. InP-InGaAs-InP photodiode frequency response for dif- Avalanche photodiodes" RCA Review Vol35,234,1974
ferent widths of I region d=0.7, 1.0 and 1 . 4 p , F 5 V .

720