1228 PROCEEDINGS OF THE IRE September

Carrier Generation and Recombination in P-NV

Junctions and P-N Junction Characteristics*
CHIH-TANG SAHt, MEMBER, IRE, ROBERT N. NOYCEt, MEMBER, IRE AND
WILLIAM SHOCKLEYt, FELLOW, IRE

Summary--For certain p-n junctions, it has been observed that It has been pointed out by Shockley and Read4 that
the measured current-voltage characteristics deviate from the ideal generation of the current carriers in the space charge
case of the diffusion model. It is the purpose of this paper to show region or the transition layer of a p-n junction may be
that the current due to generation and recombination of carriers from
generation-recombination centers in the space charge region of a extremely high. The essential features of the reverse
p-n junction accounts for the observed characteristics. This phe- characteristics of a silicon p-n junction can be under-
nomenon dominates in semiconductors with large energy gap, low stood in terms of this phenomenon by using a model of
lifetimes, and low resistivity. This model not only accounts for the single energy level uniformly distributed Shockley-
nonsaturable reverse current, but also predicts an apparent Read-Hall recombination centers. Pell and Roe,' and
exp (qV/nkT) dependence of the forward current in a p-n junction.
The relative importance of the diffusion current outside the space Kleinknecht and Seiler' have independently used this
charge layer and the recombination current inside the space charge model to account for the reverse characteristics of sili-
layer also explains the increase of the emitter efficiency of silicon con p-n junctions.
transistors with emitter current. A correlation of the theory with ex- Shockley has suggested that the recombination of the
periment indicates that the energy level of the centers is a few kT carriers may also dominate in the space charge region
from the intrinsic Fermi level.
when a p-n junction is biased in the forward direction.'
I. INTRODUCTION However, the importance of this effect on the p-n j unc-
r HE VOLTAGE current characteristics of p-n tion and junction transistor characteristics has not been
junctions have been studied by many authors. realized until recently.7
The ideal theory of a p-n junction of Shockley In this paper, we present a physical theory of p-n
accounts for the electrical characteristics of a ger- junctions taking into account the recombination and
manium p-n junction quite well at room temperatures.' generation of the carriers in the space charge region or
However, it is generally observed that at room tem- the transition region.
peratures the measured electrical characteristics of sili- II. SHOCKLEY-READ-HALL RECOMBINATION STATISTICS
con p-n junctions deviate considerably from that predi-
cated by the ideal theory. For example, Shockley's ideal A brief derivation is given in this section for the
theory predicts a saturation current for the p-n junction steady-state recombination statistics of electrons and
under large reverse bias and a simple dependence of the holes in semiconductors to illustrate the method and
forward current on the applied bias through the Boltz- physical principle involved. For a complete and detailed
mann's factor of the full applied voltage when the for- treatment of this subject, the reader is referred to
ward bias is several kT/q. Usually, at room temperature Shockley anid Read.4
the current in a silicon p-n junction does not saturate at The recombination and generation of electrons and
large reverse bias and increases much slower than the holes in semiconductors may take place at some type of
Boltzmann's factor to the full applied voltage under recombination-generation centers or traps. These sites
forward bias. Discrepancy has also been observed by may be crystal lattice dislocations, impurity atoms lo-
many workers between the ideal theory for p-n junction cated interstitially or substitutionally in the crystal
transistors and the observed characteristics of silicon lattice, or surface defects. Recombination may also oc-
p-n junction transistors. In particular, the ideal theory
predicts a nearly unity value for the transistor current 4 W. Shockley and W. T. Read, Jr., "Statistics of recombinations
amplification factor, alpha, at low emitter current of holes and electrons," Phys. Rev., vol. 87, pp. 835-842; Septemnber,
1952.
densities, while the observed alpha usually becomes very R. N. Hall, "Germanium rectifier characteristics," Phys. Rev.,
small at low emitter current densities.2'3 vol. 83, p. 228; July, 1951.
R. N. Hall, "Electron-hole recombination in germanium," Phys.
Rev., vol. 87, p. 387; July, 1952.
5 E. M. Pell and G. M. Roe, "Reverse current and carrier life-
*Original manuscript received by the IRE, March 23, 1957; time as a function of temperature in germanium junction diodes,"
revised manuscript received, May 13, 1957. J. Appi. Phys., vol. 26, pp. 658-665; June, 1955.
8 H. Kleinknecht
t Shockley Semiconductor Lab., Mountain View, Calif. and K. Seiler, "Einkristalle und pn Schicht-
1 W. Shockley, "The theory of p-n junctions in semiconductors kristalle aus Silizium," Z. Physik, vol. 139, pp. 599-618; December
and p-n junction transistors," Bell Sys. Tech. J., vol. 28, pp. 435- 20, 1954.
489; July, 1949. 7 R. N. Noyce, C. T. Sah, and W. Shockley, "Carrier generation
2 J. L. Moll, et al., "P-N-P-N transistor switches," PROC. IRE, and recombination in the space charge region of a p-n junction,'
vol. 44, pp. 1174-1182; September, 1956. Bull. Amer. Phys. Soc. II, vol. 1, H9, p. 382; December 27, 1956.
8 M. Tanenbaum and D. E. Thomas, "Diffused emitter and base C. T. Sah, and W. Shockley, 'Interpretation of silicon p-n junc-
silicon transistors," BeU Sys. Tech. J., vol. 35, pp. 1-22; January, tion current-voltage characteristics," Bull. Amer. Phys. Soc. II, vol.
1956. 1, H1O, p. 382; December 27, 1956.
1957 Sah, Noyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1229

cur directly with the emission of light or by the three (or TABLE I
more) particle process (Auger process) with the third SYMBOLS
carrier carrying away the energy. The radiative process n = density of electrons in the conduction band = ni exp (F,,-Ei) /k T
is rather improbable. At present, there is no sufficient p = density of holes in the valence band =ni exp (Ei - F5)/kT
information about the Auger process in semiconductors. E,= energy of the highest valence band level
E= energy of the lowest conduction band level
Thus, these processes will not be considered. E,=intrinsic Fermi level =-q
Under steady-state conditions, a single energy level Et= energy level of the recombination-generation centers or traps
recombination center is characterized by three numbers: eg=(Et-Ei)/kT+ln V/rpO/7nO
I =electrostatic potential =-E,q
the capture cross section for electrons, the capture 'D=diffusion or built-in voltage in a p-n junction
cross section for holes, and the energy involved in these on= quasi-Fermi electrostatic potential or imref for electrons
transitions. The cross sections are inversely proportional 4,=quasi-Fermi electrostatic potential or imref for holes=- F/q
to the lifetimes of electrons and holes respectively, and n=-density of electrons or holes in an intrinsic specimen
N0 =effective density of levels for conduction band
the transition energy may be measured from one of the N= effective density of levels for valence band
edges of the energy gap of the semiconductor.8 Nt=density of the recombination-generation centers or traps
ftp=fraction of traps occupied by holes
There are four basic processes involved in the carrier ft = fraction of traps occupied by electrons=1 -ftp
n,=density of electrons in the conduction band when the Fermi
generation and recombination through the traps.4 If a level falls at Et=ni exp (Et-Ei)/kT
trap is occupied by a hole, an electron may drop into the pt= density of holes in the valence band when the Fermi level falls
trap from the conduction band and recombine with the at Et = ni exp (Ei-Et)/k T
mrO=lifetime for electrons injected into highly p-type specimen
hole, or the trap may emit the hole to the valence band. tro=lifetime for holes injected into highly n-type specimen
If the trap is initially filled with an electron, the trapped q = magnitude of electronic charge
k = Boltzmann constant
electron may be emitted to the conduction band or a T= absolute temperature
valence band hole may move into the trap and recom- U0. = net electron capture rate
Uc,,= net hole capture rate
bine with the trapped electron. These four processes are U= steady-state electron or hole recombination rate
illustrated in Fig. 1. The direction of the arrow indicates Jr=recombination-generation current density in the space charge
layer
CONDUCTION BAND JD -diffusion current density outside the space charge layer

E
cEI where a is a proportionality factor which includes the
trap density, the total number of empty electronic states
in the conduction band and the probability of electron
emission from the traps. The expression for a can be
t
D- obtained by a detailed balance argument for the system
under the thermal equilibrium condition. Under this
condition, the electron emission rate must be equal to
the electron capture rate, i.e., (1) and (2) are equal.
Thus, if the occupancy of the traps is expressed in terms
of a quasi-Fermi level4 (or imref) for traps Ft, then the
v imref for electrons, Fn, must fall at Ft at thermal equi-
librium. Assuming that the semiconductor is nonde-
(a) (b) (c) generate, and using9
VALENCE BAND
Fig. 1-The basic processes of carrier generation and recombination ft = (1 + exp (Et -Ft)kT)-
through traps, (a) electron capture, (b) electron emission, (c) hole
capture, (d) hole emission.
then from equating (1) and (2) we obtain
the direction of transition for either a conduction or a
valence band electron as the case may be. a = ni/rnO (3)
Consider the electron capture process indicated by The net capture rate for electrons by the traps under
Fig. l(a). The rate that the electron in the conduction nonequilibrium conditions can then be written as
band will drop into an empty trap is
nfepl/rno. (1) Ucn = (nft, - nift)/7.o. (4)
(See Table I for the meaning of the symbols.) An entirely similar treatment can be carried out for
The electron emission rate indicated by Fig. l(b) can holes leading to the following equation for the net rate
be written as of capture for holes under nonequilibrium conditions.
aftI (2)
Usp = (pft - piftp)/rpo. (5)
The steady-state recombination statistics involving centers
8
with more than one energy level has also been considered by the
authors. (To be published.) I Electron spin degeneracy is included in E,.
1230 PROCEEDINGS OF THE IRE September
The rate of recombination for nonequilibrium but EC E @c ss ED
steady-state conditions is obtained by requiring that the
net rate of capture of electrons be equal to that of holes.
I ~~~~~~~~~~~IIc
Fn
This condition leads to El [3 ° ° Elt Et 2 a E

U = Ucn
-
Ucp
- (pn -
n2)/[(n + nl)Tlo + (p + pl)rnOj (6) Ev .1
+

for the steady-state recombination rate for electrons or (a) (b)

holes.
The result of these statistics is applied to the current
carriers in the transition region of a p-n junction and in
the region outside of the transition region.
III. IDEALIZED MODEL 8
An idealized p-n junction is considered in this section
so that the physical processes can be readily visualized.
It is assumed that there are single-level, uniformly-
distributed recombination-generation centers located at
the intrinsic Fermi level. Thus Pi = nli = n. It is further
assumed that the lifetimes, mobilities, and the densities (c)
of the minority carriers on opposite sides of the junction Fig. 2-Recombination and generation processes in semiconductor,
(a) p region, (b) n region, (c) intrinsic with swept field.
are equal. The recombination and generation processes
are considered separately in the p region, the n region, the current which must flow in the external circuit to
and the transition region of a p-n junction. replenish the diffusing minority carriers is
In the p region shown in Fig. 2(a), the traps are
mostly empty and ready to trap the injected electrons. JdA = - 2qn,LoA./To, (9)
Subsequently, the holes would move into these occupied
traps and recombine with the trapped electrons. In order where np-Pn and rTo=r1po=rO have been used. Lo, the
to preserve electrical neutrality, holes must be re- diffusion length of the minority carrier, is given by
plenished through electron current flowing in the ex- (Doro)"2, and DO=Dn=D, is the diffusion constant of
ternal circuit. Thus, with p =pp>>p1, n1 and n, and the minority carriers. A is the area of the junction.
n = np +An, (6) reduces to Let us next consider the space charge region or the
transition region of a reversely biased p-n junction. Both
U = (n - n')/-rno. (7) types of carriers are rapidly swept out of this region by
the large electric field, thus their densities are small com-
Similarly, the recombination rate in the n region, shown pared to ni. The situation is shown in Fig. 2(c). For this
in Fig. 2(b), is given by case, the generation rate from (6) is
U = (p pn)I/rO. -
(8) -U = ni/2To, (10)
These are the usual linear recombination laws which and is approximately constant over the entire space
hold when the injected minority carrier density is much charge layer. It may be noted that the recombination
smaller than the equilibrium carrier density. effect is again unimportant here because of the presence
If the minority carriers are rapidly moving out of a of the large electric field. It may also be noted that the
certain region of the semiconductor by electric field, the centers are about half empty and half filled in this
generation rates in this region would be region.
- U = nv/rno, (7a) Since for every pair of carriers generated, one electron
must flow in the external circuit, the current flowing in
for p-type region, or the external circuit due to this effect is
- U = Pn/rpO (8a) J,,A = -
qWniA/2ro, (1 1)
for n-type region. The edges of the space charge layer where W is the width of the space charge layer.
of a reversely biased p-n junction are precisely the re- The currents generated in the p region and the n re-
gions where (7a) and (8a) prevail. On the average, a gion and the current generated in the space charge layer
minority carrier generated in the n region or the p region may be compared. The ratio of these two components
within one diffusion length from the edge of the space given by (9) and (11) is
charge layer would diffuse to the edge and slide down the
potential hill of the space charge region rapidly.' Thus, Jrgl/Jd =
(n./4n) W/Lo. (12)
1957 Sah, Noyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1231
A number of conclusions regarding the relative im- The sum of the diffusion current densities which flow
portance of these two currents can be deduced. Eq. (12) in the p region and the n region is given by'
indicates that the current generated in the space charge
region may be extremely large compared with the dif- JD = (2qnpLo/ro) exp (qV/kT) (17)
fusion current for semiconductors with short lifetimes, for applied forward bias of several kT/q. The two cur-
low resistivities, large energy gap and at low tempera- rents can again be compared, giving a current ratio of
tures, even if W<<Lo.
For a typical example, consider a 2 ohm-cm material Jt./JD= (ni/np)(W/2Lo)[kT/q(JD - V)]
of one microsecond lifetime. Assuming a one micron *exp (-qV/2kT). (18)
space charge layer width, at room temperature the cur- Using the same data as in (12) with a space charge layer
rent ratio is about 3000 for a silicon p-n junction while width of 0.1 micron, the recombination current becomes
it is only about 0.1 for a germanium p-n junction, Thus, equal and greater than the diffusion current at applied
the generation-recombination current in the space bias of less than 10 kT/q.
charge layer would greatly influence the characteristics The total current per unit area is given by the sum of
of silicon semiconductor devices. the diffusion and the recombination-generation current
If the current due to generation in the space charge densities. The comparison given above indicates that
layer is important under the reverse bias condition, it the recombination-generation current is much greater
would be expected that the reverse process, namely, re- than the diffusion current, and the total current would
combination in the space charge layer, will be important vary slower than the ideal rectifier formula of
under forward bias. However, the extension of the argu- exp (qV/kT).
ment for deriving the current produced in the space
charge layer must take a slightly different form from IV. THEORETICAL CALCULATION OF THE
that applied to the reverse-bias case. CURRENT-VOLTAGE CHARACTERISTICS
Under forward bias, the carrier concentrations at the
center of the space charge region are equal and vary as In the last section and in the literature," 4 an effective
lifetime, which is a function of the carrier densities, has
n = p = ni exp (qV/2kT), (13) been used to obtain the characteristics of p-n junction
since np = n12 exp (qV/kT). The recombination rate at devices. The effective lifetime is usually obtained by
this p-n boundary is then experimental measurements. In order to obtain a com-
plete theoretical relation between the current and the
U = ni exp (qV/2kT)/2r0 (14) applied voltage, it is necessary to start with the exact
for V greater than several kT/q, and falls exponentially expression of the steady-state recombination rate of the
with distance on either side of the p-n boundary with a carriers given by (6).
characteristic length of kT/qE where E is the electric Substitution of the following relations
field at the junction. Thus, the recombination current p= ni exp (E,- Et)/kT
density is n= ni exp (Et - Ei)/kT
J, = 2(kT/qE)qni exp (qV/2kT)/2ro, (15) p ni exp ($p - 'I)q/kT
where the factor 2 comes from the contribution on the n -n exp (T - On)q/kT
two sides of the p-n boundary. The exact expression for
this idealized case involves an additional factor of 7r/2, Ei= (E + E - kT ln N/N,,)/2 = -q'I (19)
which takes into account that the recombination rate into (6) gives the following expression for the steady-
falls slower than exp (-qEx/kT) near x=0. state recombination rate:

ni
4\/TpoTrno
sinh qk
2kT
(4p - q5n)
(20)
cosh /Et-Ei + ln
cosh q +T _p 2 _
In - 1 + exp L2-q (p-,)i /TpO
L\T 2 /rnOJ L_2kT (4 J \kT n

Using a linear potential variation across the junction, The notations are identical to those used by Shockley
the electric field can be written as and Read,4 and are listed in Table I.
E = ('D- V)/W (16) The recombination rate given above is sketched for
where 'D is the built-in voltage. several limiting conditions in Fig. 3.
1232 PROCEEDINGS OF THE IRE September
log I

4
Fig. 3-Variation of the recombination rate in the p-n junction transi-
tion region; (a) forward bias (Op-4,0)q/2kT>>1, (Et-EO)/kT
+j ln r,o/r.ol, (b) reverse bias(4,%-41,)q/2kT>»1, (Et-E,)/kT +3
3-3
+* In rpo/,r.o , (c) small bias (O,,-0)q/2kT-1, (E,-E,)
Ik Ti In rps/raoI=4.

The recombination-generation current in the space

charge layer is obtained by integrating (20) over the
space charge layer. For the one-dimensional case, the
total recombination-generation current density in the
space charge layer is given by
Fig. 4-p-n junction under large reverse bias.
Jrg = qf Udx, (21)
where W is the space charge layer width and the genera-
where the integration is taken over the space charge tion rate is given by

ni[2V/rporno cosh E(E1T+3,n

layer.
The exact solution of this problem requires a knowl-
edge of the imrefs 4, and 4n and the electrostatic poten- -U = kT

±2 rno-
) (23)
tial I as a function of position in the p-n junction. This
is a difficult problem involving the solution of simul- This approximation slightly overestimates the genera-
taneous nonlinear differential equations.' In this treat- tion current since the generation rate drops exponen-
ment, we shall follow a self-consistent approach given tially to zero near the edges of the space charge layer.
in Appendix I. It is convenient to consider three regions The result obtained above indicates that the traps are
of the applied bias, the large reverse-bias region, the most effective as generation centers if they are located at
small bias region, and the large forward-bias region. the intrinsic Fermi level when the lifetimes of the car-
riers are equal. In general, the traps are most effective
Large Reverse-Bias, - q(k 4>)/kT>»1 if (El-Ei)/kT+(1/2)ln rpo/rno=O. The above result
For this case, the potential energy diagram for holes also shows that the generation current cannot saturate.
is shown in Fig. 4 together with the current distribution For the linear-graded junction this current is propor-
in the p-n junction. The hole current J, shown in the tional to ij power of the applied voltage and for a step
lower part of Fig. 4 can be deduced from the recombina- junction it is proportional to a power of the applied
tion-generation rate and (21). Under large reverse-bias, voltage since the space charge layer width varies ac-
the first term in the denominator of (20) is small com- cordingly.
pared with the second term, since 4O.-4O, is several Although it was shown that the shapes of the imrefs
kT/q inside the space charge layer. Thus, the recombi- have negligible effect on the calculation of the genera-
nation rate is approximately constant over the entire tion-recombination rate, it is instructive to deduce the
transition region as illustrated in Fig. 3(b). The hole shapes of the imrefs. From the relation between the hole
current then is proportional to the distance and de- current and the imref for holes,
creases monotonically. The total recombination-gen- Jp = - qniwp exp Rop- 4)q/kT]d4p/dx, (24)
eration current for this case is then
it can be concluded that in the transition region, the in-
Jra9 =
qUW, (22) crease of the slope of 4, must be slower than the decrease
1957 Sah, Noyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1233

of the hole density p or the factor exp (k, -*)q/kT in

order to give a monotonic variation of J,. A similar con-
sideration can be given to the electron imref variation in
the transition region. These arguments lead to the po-
tential diagram shown in Fig. 4. The drop of the imrefs
is almost complete inside the space charge layer. A
numerical calculation in Appendix I for a silicon junc-
tion also gives the same conclusion.
It is interesting to note that the product of the junc-
tion capacity and the space charge layer generated cur-
rent is a constant under large reverse-bias and is inde-
pendent of the shape of the electrostatic potential dis-
tribution inside the junction. This relation can be
written as
CJ7g = eoKqU, (25)
where e0 is the permittivity of free space and K is the
dielectric constant. The constancy of the product given
above holds only at intermediate range of applied volt-
age. At reverse-bias near the avalanche breakdown, the
product given by (25) increases. The lifetimes r,,o and
r,,o may be field dependent and cause additional varia-
tion of the product with the reverse-bias.
The avalanche mechanism is considered in detail in 3
u
Appendix II. The effect of the avalanche multiplication
is to multiply the recombination-generation current in
the same way as if it were injected into the space charge
layer if the electron and hole ionization rates are equal. Fig. 5-p-n junction under forward bias.
Thus, for large reverse-bias, the recombination-genera-
tion current is given by
I. Outside the space charge region the imrefs vary
Jr= qUWM, (22a) linearly with distance. The approximate potential en-
ergy diagram and the current distribution across the
where M is the avalanche multiplication factor given in junction is shown in Fig. 5 from these conclusions. The
Appendix II. current shown there is for a case where the recombina-
The thermal activation energy or the energy level of tion current is about 2.5 times higher than the total
the recombination-generation centers can be obtained diffusion outside the space charge layer.
readily by temperature measurement of the reverse From Fig. 5, it can be seen that the electrostatic po-
characteristics. The magnitude of the trap energy level tential can be written approximately
from the intrinsic Fermi level is given by the difference
of half the zero temperature energy gap width and the 4pv +4O_n D -(f41p -n)
'D n W W
activation energy obtained from the slope of a plot of -
2
- =. -
W
x -
2
< x < -,
2
(26)
ln J, - (5/2) ln kT vs (kT)-1. The detailed calculation
is given in Appendix III. where W is the total space charge layer width, p, -On iS
Small Applied Bias approximately the applied voltage and kD is the built-in
voltage or the difference of the intrinsic Fermi levels on
For this case, both terms in the denominator of the ex- the two sides of the junction. An exact expression for the
pression for U are important and one has to consider electrostatic potential variation could be obtained.'
the shape of the imrefs in order to obtain a good ap- However, a linear approximation given by (26) will
proximation to 'F-(0 +4n) /2 in the expression for U. give a maximum of 50 per cent error in the slope of 4'
A similar argument for the shape of the imrefs as that and will put the integral of (21) into a more tractable
given to the large reverse-bias case can be made here. form. 10
This leads to the conclusion that the variation of the
imrefs in the space charge layer is very small, contrary
Dr. Ruth F. Schwarz of Philco Corporation has kindly sent us
to the result of the large reverse-bias case. This is also her 10more exact calculation for step junctions. The linear approxima-
confirmed by a numerical calculation given in Appendix tion of (26) is quite good for grown or diffused type graded-junctions.
1234 PROCEEDINGS OF THE IRE September
Performing the integration, the recombination cur-

rentcan be written as

qni
V/TPOTnO
W 2 sinh
(TD -
(4Op 4,)q/2kT
4)p
-

+ 4n)q/kT
(27)

where
rZ2 dz
f(b) = 2 + 2bz 1

b = exp [- (op 0))q/2kT]

-

-Et - E
* cosh + (1/2) In (TpO/nrfo)

and the integration limits are

Fig. 6-The function f(b) and its slope.
Z1,2 =
(rpO/rnO) 12 exp [ T (TD- 4p + kOn)q/2kT].
2. 0
The integral given by f(b) can be evaluated exactly. ° 9 O.Dr/kT-20
However, at applied voltages several kT/q less than the
built-in voltage, the integration limits can be extended 1. 5
from zero to infinity with small error. The value of the
integral and its slope is plotted in Fig. 6 for the case of
Zi =0 and Z2 = o. The variation of the slope of log Jr, is 2
=/
also plotted in Fig. 7. 1. 0

There are several regions of the current-voltage char-

acteristics for this case which need special discussion. I~ =10I
1) Very Small Bias: When the applied bias is less 0. 5

than kT/q, the recombination-generation current follows 0 2 4 6 8 10 12 14 16 18 20

essentially the ohmic law. The recombination conduct- k %ry(p- ~fn)---

ance at zero bias is given by Fig. 7-Slope of In Jr, vs forward applied bias.
Rrg-7 =
qniWf(b)l/A/TpOTnO!De (28)
Large Forward-Bias
A comparison of this resistance with the diffusion re- At forward-bias about or greater than the built-in
sistance gives the same conclusions for the relative im- voltage, the width of the space charge layer is small. The
portance of the two currents as that obtained in recombination current in the space charge layer is
Section II. negligible compared with the diffusion current due to
2) Medium Forward-Bias with Deep Traps:When the injected carriers outside the space charge layer. We shall
traps are located near the intrinsic Fermi level, the consider the case for which the injected carrier density
function f(b) increases slightly when the forward-bias is much higher than the fixed impurity charge concen-
decreases. Thus the recombination current will vary tration in the n region. One may assume for this case
slightly faster than exp (qV/2kT) as can be seen from that n ; p. The boundary condition at the junction be-
(27). This is in agreement with the result obtained in comes
Section II from intuitive arguments.
3) Small or Medium Forward-Bias with Shallow n(O) =
p(O) =
ni exp (qV/2kT), (29)
Traps: A region of small or medium bias with values of b
greater than 10 may exist, if the traps are quite shallow,
where V is the applied voltage less the ohmic drop. The
i.e., the effective trap level is about lOkT from the in-
usual diffusion formula applies and the total current is
to the hole current at the junction. Thus
trinsic Fermi level. The integral given by f(b) will vary nearly equal
approximately as 1/b and the recombination current will Jp(O) 2(qDpnt/Lo) exp (qV/2kT),
J = (30) =

vary with the applied voltage approximately as exp

(qV/kT)-1 shown in Fig. 7. where Lo is an effective diffusion length given in Ap-
The result given by (27) would reduce to that given pendix IV. A more detailed treatment of the diffusion
by (22) and (23). However, (27) is not accurate in the current is given there also. We may note that the region
transition region between small and large reverse bias of exp (q V/2kT) occurs at higher applied voltage for the
because of the error in estimating the shape of the imrefs. diffusion current than the same case for the recombina-
1957 Sah, Noyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1235
tion current. The case for the diffusion current has been
observed for certain P+ IN diodes with very low ohmic L AVALANCHE
BREAKDOWN
LOG V
drops. 11 AVALANCHE
At still higher current density or applied bias, the MULTIPLICATION REGION
current will eventually be limited by the ohmic contact
resistance and the bulk resistance of the material. The SLOPE = 3
current voltage characteristics become linear in this
range.
To summarize the theoretical calculation, the char-
acteristics of a linearly graded junction is sketched in
Fig. 8. Fig. 8(a) shows the reverse characteristics and LINEAR
LOG I
Fig. 8(b) shows the forward characteristics including the --
diffusion current outside the space charge layer at large
forward bias. For an actual case the regions of distinct (a)
but different slopes shown in Fig. 8(b) would merge and
a varying slope of between 0.5 and 1 would be ob-
served. From these considerations, we may also con- / LINEAR
clude here that the minority carrier concentrations at OHMIC
the boundary cannot vary faster than exp (qV/kT). SLOPE = 0.5 (diffusion
n= 2 current)
This is in agreement with statistical mechanics con-
siderations. SLOPE =
It=i
V. THE EMITTER EFFICIENCY AND THE CURRENT AM- SLOPE S Q. 5
PLIFICATION FACTOR OF A p-n JUNCTION TRANSISTOR nl' 2 recombination in
the space charge layer)
Carrier recombination in the space charge layer of a
forward-biased emitter may play an important role in SLOPE s
the determination of the emitter efficiency of a silicon IL=
junction transistor at low emitter current. This effect
thus imposes a lower limit on the emitter current density
for high emitter efficiency operation. qQ V
?.7T.
k v
From the theoretical considerations given in the 2 4 6 8 10 12 14 16 18
previous sections, the recombination process is domi- (b)
nant in a forward-biased silicon emitter junction. The
Fig. 8-Characteristics of a linearly graded p-n junction,
portion of the emitter current due to this mechanism (a) reverse bias, (b) forward bias.
consists of carrier recombination in the space charge
layer and thus is not available for transistor action. At
high emitter current, the diffusion current dominates made in the previous section that the dc transistor alpha
and the emitter efficiency increases. This phenomenon given by (31) approaches zero at low emitter current
has been observed for silicon transistors.2'3 The recombi- density since Jd<<Jr. At large forward bias the alpha
nation in the space charge layer which causes increasing increases toward unity since Jd>>Jr, Jd'. At still higher
transistor alpha with current is also the basis for the emitter current densities, Jd, Jd'>>JTg the emitter effi-
operation of low current p-n-p-n transistor switches.2 ciency or the alpha decreases as pointed out by Webster.'2
The dc transistor current amplification factor, alpha, The proximity of the collector junction increases the
can be calculated by the following expression at reverse diffusion current component and increases the alpha of
collector bias when the recombination in the emitter the device at low emitter current densities.
space charge layer is taken into consideration. Fig. 9 shows the current distribution in a p-n junc-
tion transistor under various emitter bias conditions.
a = sech(Wb/Lb) { 1+ [(Jr+fJd')/Jdtanh(Wb/Lb) ]'. (31) Fig. 10 shows several calculated transistor alpha using
(31) and the calculated result for the diode shown in the
Jd and Jd' are the injected current densities flowing into next section. These theoretical curves follow closely the
the base and the emitter region respectively. Lb is the experimental alpha of Moll2 and Tanenbaum."3
minority carrier diffusion length in the base and Wb is
the base layer width. It is evident from the calculations 12 W. M. Webster, "On the variation of junction-transistor cur-
rent amplification factor with emitter current," PROC. IRE, vol. 42,
pp. 914-920; June, 1954.
11 R. N. Hall, "Power rectifiers and transistors," PROC. IRE, vol. 13 We are indebted to J. L. Moll for pointing out a numerical
40, pp. 1512-1519; November, 1952. error.
 1236 PROCEEDINGS OF THE IRE September
The following data are obtained from measurements
of a typical diffused junction and from calculations.
Bulk concentration ND = 2.2 X 1015 cm-3
Surface concentration of boron = 4 X 1021 cm-3
(Dr)"12 diffusion length of boron= 1.8 microns
Junction depth = 11.8 microns from surface
Breakdown voltage= 120 volts (calculated from re-
J
n sults in Appendix II)
Ji
Built-in voltage = 0.47 volt (see Shockley,' with linear
graded junction approximation)
I t Area of junction 0.620 mm2.
In order to calculate the theoretical p-n junction char-
0

Ji_ SMALL EMITTER BIAS acteristics we shall use the linear-graded junction ap-
proximation. The trap level Et-Ei and lifetimes are
obtained by matching the theoretical formula and the ex-
J2 perimental data at three points. It is most convenient
to use two points at relatively small applied bias where
the current from the space charge layer is dominant and
'i
one point at large forward bias where the diffusion cur-
rent dominates. We choose these points to be at applied
bias of 0=qV/kT= -4.6, 4.6, and +18.3. The last point
corresponds to the built-in voltage at room temper-
0
LARGE EMITTER BIAS ature. The currents at these voltages are 2.96X 10-10,
3.75X10-9, and l.OX10-4 amperes respectively from
Figs. 11 and 12.
10
J
Using (27) for the first two points and the formula for
n diffusion current given in Appendix IV we obtain the
following results:
p
IaI
rpo = 1.2 X 10-8 sec,
-nu . I I I

VERY LARGE EMITTER BlAS 7rnO= 4.3 X 10-6 sec,

Fig. 9-Current in junction transistor with large carrier Et-Ei = 4.6kT or 1.3 kT,
recombination in the emitter junction.
where we have used the following constants :14
ni = 10-10 cm-3
Dp= 1. cm2/sec,
Dn= 2.77Dp,
E,= 1.21ev at 0° K.
The value of Et-Ei cannot be uniquely determined
from this experiment since the results only give cosh
[(Et-Ei)/kT+ (1/2) ln(rpo/rno)] which makes the
sign uncertain. However, the value of Et-Ei obtained
is in good agreement with the result of Pell and Roe.5
The theoretical calculations and the experimental
I measurements are presented in Figs. 11 and 12. Good
EMITTER CURILENT DENSITY "/-r.2 agreement is obtained over the entire range of more
Fig. l1-p-n-p silicon transistor dc alpha. than 9 decades of junction current. In addition to the
three-point-match, we have also taken into considera-
tion the contact resistance drop which accounts for
VI. EXPERIMENTAL MEASUREMENTS
The experimental p-n junctions were produced by dif- 14 Morin and Maita, "Electrical properties of silicon containing
fusion of boron from the gaseous phase into an n-type arsenic and boron," Phys. Rev., vol. 96, pp. 28-35; October 1, 1954.
M. B. Prince, "Drift mobility in semiconductors. II. Silicon,"
silicon of 2.4 ohm-cm resistivity at 1250°C for 80". Phys. Rev., vol. 93, pp. 1204-1206; March 15, 1954.
1957 1957 ~Sah, Noyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1237
13
Zs the slow increase of the large forward current. The
10 - zr -- - - - -
avalanche multiplication near the avalanche breakdown
voltage is also considered. In this region, the experi-
__calculat(oals /j mental measured currents were found for all cases to be
*experiments-tJ always greater than the calculated values. This may be
-6 due to the dependence of the lifetimes on the electric
field and due to surface breakdown.
-k- Additional data are shown in Figs. 13(a) and (b), on
the next page, for silicon diodes of different base resis-
-1511 tivities and surface concentrations. Table II shows the
14.4 4 TD+~~~~~~DJ calculated lifetimes and trap energy levels from the ex-
-Z
S--
perimental data. The energy level of the recombination-
generation center or trap is at about -f4kT from the in-
trinsic Fermi level. This is in fair agreement with the
temperature data shown in Fig. 14.
TABLE II
/I
-8--oPonz Unit No. 352-10 354-3 354-4 3151 [2 317A18
Type P+N P+N P+N N+.I N+P
p(Q-cm) 2.4 2.4 2.4 21 21
Co surface cm-' 5 X102' 5 X102' 5 x1021 1022 3 X10'8
-Tpo (jsec) 0.035 0.035 0.039 0. 05 0.15s
toL 11 / 0 -I.
01/ 0Q-I-% -I-
0.3 =4"10
0.4. --l-
Q5A -.",
0C6
Tn (jLsec)
cosh et
0.65
6.3
4.0
0.69
5.3
3.9
0.69
6.0
3.9
0.07
2.2
1.3
0.012
5.5
3.7
Voltage vo/t -1.1 -0.9 -1.0 6 -1.2

Fig. 11-Forward characteristics of a silicon p-n junction.

Under certain conditions, the surface leakage current
may be important in silicon p-n junctions.","5 If the sur-
face charge is smaller than the fixed charge of the im-
purities, it can be shown that the surface current, pro-
duced by recombination-generation centers or surface
states at the circumferential surface of the junction, will
follow the same voltage dependence, namely, V'12 or
V's3, under the large reverse bias condition.
In order to separate the recombination-generation
current due to surface centers from that due to volume
centers we have made experimental measurements on
p-n junctions with extremely large and small ratios of
junction area to circumference. The geometries are
shown in Fig. 15. The results indicate that the surface
current is negligible compared with bulk current for
freshly etched junctions.
VII1. CONCLUSION
In this paper we have used a model of the single level
uniformly distributed Shockley-Read recombination-
generation centers to explain the p-n junction character-
istics. This model explailns both the nonsaturable re-
verse characteristics and the apparent exp (q V/nkT)
dependence of the forward current of typical sil'icon
1f, M. Culter and H. M. Bath, "Surface leakage current in silicon
fuised junction diodes," PROC. IRE, vol. 45, pp. 39-43; January,
1957.
1 10 109 10
-
16 W. T. Eriksen, H. Statz, and G. A. DeMars, 'Excess surface
currents on germanium and silicon diodes," J. Appi. Phys., vol. 28,
Fig. 12-Silicon p-n junction characteristics. pp. 133-139; January, 1957.
1238 PROCEEDINGS OF THE IRE September

(a) Fig. 14-Activation energy of recombination-generation

centers in silicon p-n junctions.

(a)

X. i
io_4 X0 _3 o1-Z lo-l
E -1,r 10
-N

(b) (b)
Fig. 13-(a) Experimental forward characteristics of silicon p-n Fig. 15-Geometry for separating surface and bulk currents; (a) large
junctions. (b) Experimental reverse characteristics of silicon p-n junction area to circumference ratio geometry; (b) small junction
junctions. area to circumference ratio geometry.
1957 Sah, NVoyce and Shockiey: Carrier Generation and Recombination in P-N Junctions 1239

p-n junctions. The recombination of the carriers in the Thus, for a forward current of J= 2 amp/cm2, W= 10-6
space charge layer also explains the variation of the cur- cm and ,up = 500 for Si, the maximum drop of the qutasi-
rent amplification factor of a silicon transistor at low Fermi potential for holes is
emitter current density.
APPENDIX I sinh qSOp/2kT = 0.4 exp - [¢() 2kT]
VARIATION OF THE QUASI-FERMI LEVELS IN THE
SPACE CHARGE LAYER The exponential factor on the right-hand side cannot be
The hole current can be written as more than 1. This is easily seen from Fig. 5 if the refer-
ence potential is at (2) [0,(x) +kn(x) ].==o as was implied
Jp = - qipp(d4p/dx). (32) in (36). In addition, a current less than J should be used
Substituting the expression for p from (19) into (32) as shown in Fig. 5. Thus the upper limit of the drop of
the following differential equation can be obtained. the quasi-Fermi level in the space charge layer is about
(kT/q) and is indeed very small.
d
- exlp-
q
(4o-I
q /d1\ q_ p- 4-')
-(-Iexp-(4 For the large reverse-bias case, 01 > 5, the situation is
ax kT kT dx- kT quite different. Eq. (37) can be approximated by
=--Jr_ . (33)
qMpni sinh 2-
2(W)1,T
The solution of this equation is
exp [() + V( )] 2kT}
exp (qOp/kT) = C- .exp (T
\k T J)dx
/qApnj
(34)
ni, 8e1201
To calculate the variation of the quasi-Fermi potential qp 012 (39)
variation across the space charge layer we use the linear
approximation for the hole current density Jp, and for Thus, for J= 10-6 amp/cm2, W= 10-4 cm and 01= 100,
the electrostatic potential variation. Thus, for-2 W the drop of the quasi-Fermi potential is
x<2W
Jp = J(1 -2x/W) (35) sinh (q5p/2 kT) =
exp [80.7 -
44(P ) q/2kT].
and
' = ('D - 6b) X/W, (36) By successive approximation it can be shown that the
drop of the quasi-Fermi potential for holes is approxi-
where "D is the built-in voltage, &k is the applied volt- mately equal to the applied voltage as indicated in
age and W is the space charge layer width. By the evalu- Fig. 4.
ation of the integral of (34) with these approximations, For small bias, 0, is equal to the built-in voltage. For
the following relation is obtained for the quasi-Fermi 01 20, J= 10-8 amp/cm2, and W= 10-5 cm the drop of
=
potentials for holes at the edges of the space charge the quasi-Fermi potential for holes is approximately
layer.
sinh &tpq/2kT = exp (-12.5 - &kq/kT).
exp kT[ (2 (L)
Thus the drop is appreciable only when the reverse-bias
is greater than about 12.5kT/q.
*sinh qk
2kTL /WP
(2 P
2)2
APPENDIX II
= (JW/qpni) [inh(01/2) - exp (-1/2)]] /1 (37) AVALANCHE MULTIPLICATION OF THE GENERATED
CARRIERS IN THE SPACE CHARGE LAYER
where 01= (ID-- &)q/k T. Suppose that the p-n junction shown in Fig. 16 is
For forward bias, 01 is small and (37) reduces to biased in the reverse direction. Hole current density per
unit area Jf(x) flows toward the left, and electron current
exp [ 2 + P 2 2kT density per unit area Jn(x) flows toward the right. There
is also a net steady-state generation of the hole-electron
sinh(2P ) - 2)]) = JW/2qivni. (38) pairs in the space charge region of U per unit volume per
unit time from the Shockley-Read-Hall generation-
1240 PROCEEDINGS OF THE IRE Septemnber
The solutions are
(x) + dJp (x) IX(X)
_Of Jn(x) = eJ(a-0()dz
J0(X) -_ _ J,x)+dJ,(X) \ N-TYPE f (pJ + qU)eJ (-dzdx + J,no (45)-

.N q It dx
hole
-4-- q LI dx
electron
and

Jp(x) = e fx (a- )drf (aiJ+ qU)ef (a-p)dx dx + J (46)

w
I \, l _ow
Thus, the total current is
x= 0 x x+ dx x= W

Fig. 16-Avalanche multiplication of electron and hole currents J = Jp(x) + Jh(x)

in the space charge layer of a reverse biased p-n junction. J w
JPO + Jno + e-Jx(a-#)dx J qUef"(a-B) dxd x
recombination centers. We definie, as done by McKay
and Miller17 the ionization coefficients as follows: I J efx(a-)dxdx -J Waefo(a-l)dxdx
. (47)
-
rs ww
a = no. of electron-hole pairs produced per cm travel
of an electron. For a = 3, J reduces to
3=no. of electron-hole pairs produced per cm travel rw
of a hole. Jpo + Jno + fWqUdx
J -.
We assume that a and 3 are functions of the local elec- rw
tric field only and neglect the past history of the carriers. 1 fadx
We also neglect the space charge effect or the interaction
between the carriers. The change of electroni current in
dx per unit area is
=
MYJPO + Jno +- w
qUdx), (48)

dJ0(x) =
aJ0(x)dx + 3J,(x)dx + qUdx, (40) where M is the current multiplication factor given by

where the first term on the right comes from the multi-
/ W \-1
M I )x
plication of the electron current going into dx from the
side x, and the second term comes from the electrons
Eq. (48) indicates that for =,B the generated current
a

produced by holes going into dx from the side x+dx. in the space charge layer, fqUdx, acts as if it is injected
Similarly, the change of the hole current in dx per unit
into the layer as far as avalanche multiplication is con-
area is
cerned.
-dJp(x) =
aJ0(x)dx + IJp(x)dx + qUdx, (41)
APPENDIX II I
and the sum of the two incremental currents satisfies the ACTIVATION ENERGY OF THE TRAPS
condition that the total current is space constant,
namely The thermal activation energy of the traps is obtained
from the slope of a plot of the log of the reverse current
J = J,(x) + Jn(x) = independent of position. (42) as a function of the reciprocal of temperature at a con-
stant reverse-bias.
To solve the two simultaneous differential equations Consider the range of -6 > 5 where is the normalized
we use the following boundary conditions. At x = 0, applied bias ('p-0-n)q/kT. The function f(b) for this
J= J,o and at x = W, J, = Jpo. Substituting (42) into case can be approximated by (log 2b)/b and the recom-
(40) and (41) the electron and the hole current satisfies bination-generation current can be written as
the following equations:
log [Jrg/Jo(kT) 5/2]
dJn(x)/dx = (a -
$)Jn(x) + 3J + qU (43) _
-log ni
log (E -E -1 (49)
and (t In
6(kT) III cosh +
dJp(x)/dx (a O)Jp(x) aJ qU. (44) L kT 2 rn

Since n, is proportional to T312 exp (-E0o/2kT) and

17 K. G.McKay, "Avalanche breakdown in silicon," Phys. Rev., b>10 for the restriction imposed on the reverse-bias,
vol. 94, pp. 877-884; May 15, 1954. the activation energy can be obtained by differentiating
S. L. Miller, 'Ionization rates for holes and electrons in silicon,"
Phys. Rev., vol. 105, pp. 1246-1249; February 15, 1957. (49) and is
1957 Sah, lNvoyce and Shockley: Carrier Generation and Recombination in P-N Junctions 1241
1 namely, .J= qD,,N grad On, wrhere J is the total current
E, = --Ego-(Et-E,)
2 density. This correction of the voltage at the junction
can then be made.
*tanh (Et - E + ln ,P), (50) The boundary condition follows from the second as-
.kT TnO sumption and may be written as
where Ego is the energy gap width at zero temperature. pn = ni2 exp (O.. - nb)q/kT = ni2 exp (qV/kT), (52)
Thus for deep traps, Et-Ei and the trap energy cannot where V is V (applied) - &fn, and 6q5, is the electron
be obtained exactly unless measurements are done at imref drop in the N region.
low temperatures. For shallow traps, accurate energy We shall start from the following relations:
level can be obtained in the room temperature range.
If Keo div E = q(p-n + N),1
Jp = qDp(qEp/kT - grad p),
( -E+ i
In n = qDn(qEn/kT + grad n),
div Jp = -qU,
is greater than about 2, the last term of (50) reduces div Jn= qU, (53)
to the absolute value of Et -Es. The energy of the traps
can be obtained from where the notations are defined before.
For the one-dimension case with uniform doping in
± (Et - Ei) = Ea + Ego/2. (51) the n region the above equations can be reduced to
d2p/dx2 = (Dn + Dp) U/2DpDn (54)
and
APPENDIX IV
J - qD,(b -l)dp/dx
CURRENTS OUTSIDE SPACE CHARGE LAYER qE/kTD= qDp[p(1 + b) + bN] (55)
In this appendix we shall obtain the diffusion current
flowing in the n region of a forward biased P+N junc- viously where b = Dn/Dp differs from the same symbol used pre-
tion. The approach here is slightly different from that from (6) inbecomes the text. The recombination-generation rate
given in the text but is similar to that generally used in
the literature.'8 We take into consideration both the U = (rpo + TnO)'l
nonlinear recombination rate and the high-level bound-
ary condition. [P + N-_ po _ Po(N -po) + ni] (56)
The assumptions of this analysis are: p + Po
1) The space charge due to the divergence of the where
electric field E is small compared with the fixed charge
N due to the impurities. This assumption implies that Tponl + TnOpl + TpoN
n=p+N. Po = (57)
-TPO + 'rnO
2) The imrefs are approximately constant in the
transition region and their difference is only slightly less We shall use the following boundary conditions de-
than the applied voltage. rived from our assumptions.
The second assumption above is a very good one at At x = 0, the edge of the space charge layer on the n-
not very large current densities. At very high current region side
densities the ohmic resistance drop in the N region be-
comes important. This ohmic drop can be calculated p(O)n(O) = nj2 exp (qV/kT)
from the total drop of the electron imref in this region, and
n(O) = p(O) + N
18 E. S. Rittner, "Extension of the theory of the junction transis- so that
tor," Phys. Rev., vol. 94, pp. 1161-1171; June 1, 1954.
T. Misawa, 'A note on the extended theory of the junction tran- p(O) = [A/1 + (4n,2/N2) exp (qV/kT) - 1IN/2. (58)
sistor," J. Phys. Soc. Japan, vol. 11, pp. 728-739; July, 1956.
W. T. Read, Jr., "Theory of the swept intrinsic structure," Bell
Sys. Tech. J., vol. 35, pp. 1239-1284; November, 1956. Atx= o, p( cI) =pn and let grad p=O, then
D. A. Kleinman, "The forward characteristic of the PIN diode,"
Bell Sys. Tech. J., vol. 35, pp. 685-706; May, 1956. qE(oo) i
K. B. Tolpygo, 'Emissioni capability of an abrupt p-n transition (59)
and its effect upon the conductivity of a semiconductor," J. Tech.
Physics, vol. 1, pp. 287-305 (English translation, Am. Inst. of Phys.)
kT qD,[N + pn(1 + b)/b]
J. A. Swanson, "Diode theory in the light of hole injection,"
J. Appi. Phys., vol. 25, pp. 314-323; March, 1954. Contribution from the traps is neglected.
19
1 242 PROCEEDINGS OF THE IREI KSeptemb)er
and tioIn of negligible charge due to the divergence of electric
field is less restrictive than the usual one.
J,(cJp
_ Jpn
(1 + b)pn+ bN
=
( 60) For a one ohm-cm material with 1-microsecond life-
time the maximum range of validity of the above quan-
tities are listed below for germanium and silicon at room
The gradient of the hole density can now be solved. temperature.
Eqs. (54) and (56) give
d2p/dX2 = (2L02)-j Jp(0)maX Vmax
amp/cm2 volts
po- po(N-po) + n,,21 (61) Ge 0.16 0.15
.

.pN-
p+ N- Po-
+ p+ Po -X Si 0.25 0.57

where 2) Medium-Level Case p(O) > N

Lo-2 = (Dn-' + Dp-1)(-r-o + TnO). (62) For this case the current densities cannot be reduced
Eq. (61) can be easily integrated to give to simple expressions. However, the hole gradient at
x = 0 can be written as
(dp/dx) 2 Lo-2 [p2/2- pn2/2 + (N -Po) (P - pn) grad p(0) = [p2(o) + 2(N - po)p(O)]/2 Lo.
- (po(N - po) + n,2) In (p + PO)/(Pn + PO)]. (63)
The electric fields can be obtained approximately as
We are interested in the solution at x =0 since at this qE(oo)/kT = 3/(2bLo)
point the relation between the total current and the and
hole current can be written as qE(0)/kT 1/(2Lo)=
J = JP(O) + Jrg, (64) thus the electric field is relatively constant over the
entire n region.
if the electron current in the p region is neglected. Using For a one ohm-cm material with 1-microsecond life-
these results we can deduce the following relations for time, the following list summarizes the minimum cur-
the total current and the hole current at x = 0. rent density and minimum applied voltage for this range
Jrg [(1 + b)p(o) +bN] -qD, [2p(0) + N]dp(0)/dx at room temperature.
J=- -o(65)
b[p(0) + N]
and
Jrgp(O) - qbDp[2p(O) + N]dp(0)/dx
JP(O) - b[p(°) + N]J (66)
3) High-Level Case p(O) _ JON
Thus, (58), (65), and (66) represent the complete solu-
tion. For this case the boundary condition becomes
The asymptotic forms of the above results will be p(O) = ni exp (qV/2kT)
considered. The regions to be considered are the low-
level case p(0) _ N/10, the medium-level case p(0) . N, and the current densities are
and the high-level case p(O) _ 10N. J = Jrg(l + b)/b + \'2qDpp(0)/Lo
and
1) Low-Level Case, p (O) . N/JO
For this case the total current is given by Jp(O) = Jrg/b + \V2qDpp(O)/Lo.
J = Jr + qp,Lpo.\I(l + b-')/2[exp (qV/kT) - 1]/Tpo It has been shown in the text that for this case the re-
combination current in the space charge layer is negli-
and gible compared with the diffusion current. Thus the
total current is essentially equal to the hole current at
Jp(O) = qpnLpoV/(1 + b-1)/2[exp (qV/kT) - 1]/rpo x=O, and varies as exp (qV/2kT). This region is at a
where L0 = V/Dpr,o. The hole current varies exponen- much higher current density than the similar region con-
tially with distance. However, the expression for the hole sidered for the recombination-generation current in the
current at x =0 given above differs from the ordinary space charge layer.
low-level relation by a factor V(l+b-1)/2. This comes We need to examine the electric field and its diver-
from the difference of the assumption made. In the or- gence at the boundaries to verify the validity of our
dinary low-level theory it is assumed that not only the assumptions. At the lowest applied voltage for this case
charge neutrality holds but the electric field is also zero of large bias the electric fields can be obtained by using
everywhere Qutside the junction. The present assump- (58) and (59).
1957 Tou: Digital Compensation for Control and Simulation is4'3

qE(0)/kT (1/A/2Lo)
= We shall give again a list indicating the limiting
values of current density and voltage for this range.
qE( oo)/kT = (1/V2Lo) 2p(O)/bN.
Jnxin Vmin
Thus the electric field varies about a factor of 10 at least. amp/cm2 volt
The gradient of the electric field can be obtained by dif- 27 0.34
ferentiating (55). At x = oo, dE/dx is zero and at x = 0 43 0.75
.I
the electric field is about
ACKNOWLEDGMENT
(q/kT)(dE/dx) = I/2LM2 The authors wish to acknowledge the helpful discus-
at the lowest injection limit. This corresponds to about sions with their colleagues at the Shockley Semicon-
25 v/cm for a 1-microsecond lifetime material at room ductor Laboratory. The diffused silicon materials were
temperature. The quantity (Keo/q)dEdx can now be prepared by D. Grunwald, and some of the data were
compared with p-n+N from (53). Using the values taken by Dr. S. M. Fok and M. Asemissen.
given above we obtain (Note added in proof: The forward characteristics of
germanium p-n junctions at low temperatures have re-
p - n + N>>7.6 X 1010 cm-3, cently been investigated independently by M. Bernard
using the same model. "Mesures in fonction de Ia tem-
which is usually satisfied since N is of the order of IO'l perature du Courant," J. Electronics, vol. 2, pp. 579-
cm-3 for silicon of a few ohm-cm resistivity. 596; May, 1957.)

Digital Compensation for Control and Simulation*

JULIUS TOUt, SENIOR MEMBER, IRE
Summary-This paper describes a technique for improving the
performance of digital feedback control systems and operational digi-
tal simulators by making use of the computer to perform information
programming or data processing. The system stability can be im-
proved and the system error can be reduced by digital programng
of the input and error signals together with digital correction com-
putations. This technique of digital compensation is illustrated by an
example.
Fig. 1-A typical digital control system or a channel of an oper-
INTRODUCTION tional digital simulator.
AMONG the many performance criteria for feed-
back control systems, the stability and accuracy consists of a digital computer, a decoder, an encoder and
considerations are of primary importance. Con- the control element and controlled system. The com-
siderable amount of work has been done for improving puter performs error detection and system compensa-
the performance of feedback control systems and a num- tion. The decoder converts signals in digital or sampled
ber of methods have been published in the literature. form into continuous form. The encoder converts con-
With the rapid advance of computer design techniques tinuous data into sampled form or digital code. The
and their applications, there arose a new type of system stability and accuracy of such control systems can be
for control and simulation which involves a digital com- improved by designing a suitable program for the com-
puter. Such systems are often called digital control sys- puter. The technique of digital programming com-
tems and operational digital simulators. pensation will be described in this paper.
A typical digital control system or a channel of an One of the most powerful tools for analyzing and de-
operational digital simulator is shown in Fig. 1, which signing digital and sampled-data feedback control sys-
tems and simulators is the Z-transform method.",2
*
Original manuscript received by the IRE, August 13, 1956; Since this method of approach will be employed in this
revised manuscript received, June 11, 1957. This article is based
upon the paper 'High Accuracy Operational Digital Simulation"
presented by the author at the first Natl. Simulation Conf., in Dallas, 1 R. H. Barker, 'The pulse transfer function and its application
Tex., January, 1956. to sampling servo systems," Proc. IEE, vol. 99, pt. IV, pp. 302-317;
t Purdue Univ., Lafayette, Ind. Formerly with Moore School of December, 1952.
Elec. Eng., Univ. ofPennsylvania and the Philco Corp., Philadelphia, 2J. R. Ragazzini and L. A. Zadeh, 'The analysis of sampled-
Pa. data systems," AIEE Trans., vol. 71, pp. 225-232; November, 1952.