Solid-State Electronics Pergamon Press 1967. Vol. 10, pp. 39-43.

Printed in Great Britain

CALCULATION OF AVALANCHE BREAKDOWN

VOLTAGES OF SILICON p-n JUNCTIONS
w. FULOP

Brunel University, Woodlands Avenue, London W.3

(Received 21 June 1966)

Abstract-An empirical
formula for the ionization coefficient given by cc = CE’ (C and g constants,
E electric field) and which is considered as a common effective value for both holes and electrons
yields tractable expressions for the breakdown voltages of abrupt and graded silicon pn junctions.
Numerical values compare favourably with breakdown voltages obtained from expressions which,
though derived from 0: values of a more theoretical lineage are considerably more involved and not so
readily incorporated in comprehensive device design calculations.

R&urn&Une formule empirique pour le coefficient d’ionisation donnee par a = CEP (C et g sont
des constantes et E le champ Blectrique) et consideree comme une valeur effective commune pour les
trous et electrons produit des expressions traitables pour les tensions de coupure des jonctions de sili-
ciumpn abruptes et graduees. Les valeurs numeriques sont en bon accord avec les tensions de rupture
obtenues des expressions qui, malgre que ddrivdes des valeurs d’a d’ordre plus theoriques, sont
beaucoup plus compliquees et ne peuvent Ctre facilement inclues dans les calculs plus detaillts
des constructions de dispositifs.

Zusaxnmenfassung-Die empirische Formel a = CEV fiir den effektiven Ionisierungskoeffizienten

GLvon Elektronen und LGchern (C und g konstant, E elektr. Feld) ergibt praktikable Ausdrticke
fiir die Durchbruchspannungen von abrupten und linear gradiertenpn-Uberg;ingen in Silizium. Die
numerische Auswertung liefert Zihnliche Durchbruchspannungen wie man sie such aus theoretisch
besser begriindeten Am&en bekommen kann. Letztere ftihren aber xu ziemlich verwickelten Aus-
drticken, deren Handhabung beim Entwurf von Halhleiterelementen unniitige Schwierigkeiten
bereitet.

THE THEORP of avalanche breakdown in semi- this initial electron current is multiplied by a
conductors closely follows-at least from a factor M, > 1 due to ionization then
phenomenological point of view-that of the
Townsend discharge in gases. Similar parameters
are defined, the principal one being a the ioniza- 1 -(l/M,) = i a,exp[ - i (a,-a,)dx’]dx
tion rate per unit distance of travel of the ionizing 0 0
(charged) particle. Much of the past work has been
(1)
directed to determine a as a function of the electric
field E either empirically from measurements of where the integration extends over the width W
breakdown voltage in p-n junctions(lW6) or from a of the space charge region of the p-n junction. An
mixed”) or purely theoretical approach.@) analogous expression holds for M, the multiplica-
In general different values of a for electrons and tion of an initiating hole current. Avalanche condi-
holes were obtained and denoting these as a, and tion obtains when M,,, --f co. Since over a wide
ap respectively it has been shownC5) that if the range of electric fields a,, and aP differ by con-
initiating current is due to electrons only and if siderably less than an order of magnitude (see
39
40 W. FULOP

10

Ionization
coefficient
a cm-’

1 This paper Equn (5)

2 Chynowetht2) Holes
3 Chynoweth(*)
Electrons
4 Maserjian c4) Equns( 3)
(4)

7 2 3 4 5
Electric field E in lo5 volts/cm

FIG. 1. CC
vs. E for silicon.

Fig. 1) no serious error is obtained by putting obtained differing values for dc, and uP and these
a, M dcP or alternatively the common tc thus are plotted in Fig. 1. MASERJIAN’~) also used
defined can be considered as an effective ionization equation (3) for tc but combined electron and hole
coefficient ae,f(3*4) with avalanche occurring when values into a common a,., with breakdown being
given by equation (2). For silicon he found the
W
values of a and b for best fit of experimental data
aef/ dx = 1 (2)
s to be
0
a = 9 x lo5 cm-l
which is the equation employed in this paper. (4)
b = 1.8 x 10” V/cm.
LEE et ~1.~~) have discussed the case when both
electrons and holes act as initiating charge carriers Calculated dependence of u,~, as a function of E
and obtain equations for which equation (1) based upon these values of a and b are also shown
represents a limiting case. in Fig. 1.
There is considerable experimental@~‘) and MYASERJIAN’~) then employed the form of GC given
theoretical(‘*s) evidence that ionization coefficients in equation (3) to calculate analytically the break-
have the form down voltage V, of abrupt and graded junctions.
The expressions obtained, which are shown’4)
cc = a exp( -b/E) (3) plotted over a useful range of parameters and which
where E is the electric field. CHYNOWETH(~) are discussed below, are however not very tractable
CALCULATION OF AVALANCHE BREAKDOWN VOLTAGES OF SILICON p-n JUNCTIONS 41

for ready application to device design, particularly orders of magnitude from the N,, then most of the
so for the graded junctions. space charge region will be in the N1 region and
For device design calculations a form for zeff the following relationships hold in this region.
which has given very useful results is For the electric field E (V/cm)
u e1l = CEg cm-l (5) E = Eo[l-(x/W)] (6)
where for silicon C = 14 x 10-35, g = 7 and where x = 0 is given by the position of the step in
E is in V/cm. Evidence for this form for cceff is impurity concentration E, = maximum field,
purely empirical(1*3) and there are admittedly no W = penetration of the space charge into the N1
theoretical grounds for such a function. However, region. Similarly for the potential Ir (V)
values of aef, when plotted according to equation
v = Vc[l -(x/W)]2 (7)
(5) (Fig. 1) show reasonable correspondence with
other values of cc and furthermore yield par- V, = maximum potential across the junction
ticularly useful expressions for V, of abrupt and which includes the built in voltage,
graded junctions, which, when compared below
V,, = (qN,W2/24 and E, = (2v,,,/ W) (8)
to values of I’, calculated from more accurate
values, indicate only minimal numerical departure. E = &s K = 12 for silicon
e0 = 8.85 x lo-l4 F/cm,
ABRUPT JUNCTIONS Q = electronic charge = 1.6 x lo-l9 C.
If one of the junction is moderately doped at Performing now the integration of equation (2)
constant concentration of NI cmm3 and the with aelf given by equation (5) and with E taken
change to the opposite type of impurity occurs from equations (6) and (8) one obtains on putting
abruptly in form of a step function with very high v= v,
doping level No such that N, differs by many
VB = Hk+1)/Cl 2/k7+1)(4NI/p)(l-
d/(1+g). (9)

--x This Paper EqunUO

- - -e Miller(6) Equn (111

--------A Maserjion

15 16
10 -3 10
NI cm

FIG. 2. Vs vs. N, (impurity concentration of the lowly doped region) for abrupt junctions
in silicon.
42 W. FULOP

Inserting the appropriate constants gives Putting X = E-E0 and f = -E. the integral
V, = 5.3 x 1013N,‘-3’4
can be thrown into the form
(10)

with Nr in cm- 3. Equation (10) is plotted in Fig. 2 -Eof7(_1)‘[1-(X/5)17dX

together with MILLER’S data given by - (17)
s Xl 12
V, = 2.9 x 1012N,-o’66 (11)
0

Equation (10) is similar to SHIELDS(~) equations Expanding binomially and integrating term by
for V,. term and inserting limits, equation (16) results in
In contrast MASERJIANc4) obtains, by an approxi-
mate integration of zeff given by equation (3), the
c(g2cg![yq5= 1 (18)
expression

N, =
W2/2d O-318 being the algebraic sum of binomial terms.
(12) From this follows
V, ln2[2aVa/b]

which, with the values of a, b and E for silicon

v, = ;[ (;)2$z]1’5. (19)
reduces to
2.029 x 1018
N, = (13) This shows that
VJ? log102[ VBi V, = ~~‘-215)
(20)
This is also shown plotted in Fig. 2. All these
plots for V, show fairly close numerical agreement. (where B is a constant) and illustrates in a simple
manner the direct dependence of V, upon s, the
GRADED JUNCTIONS impurity concentration gradient.
For linearly graded junctions the following In Fig. 3, V, is plotted vs. s and curve 1 has
relationships hold been calculated from equation (19) with the
appropriate values of E and C for silicon.
MASERJIAN’S(~) values for V, shown as curve 2
E = $x2 - xo2), E, = - ST2 in Fig. 3 have been calculated from the expression
(14) for V, which, when corrected for a printing error
in his equation (7), is given by
( W,b)3’2
where s = impurity profile gradient in cmb4, ‘I3 = ln3/2[(3 ra2 W,/8b)3’4 V,] (21)
x0 = semiwidth of space charge, xT = total width
of space charge = 2x0, E. = maximum field and and which is based on the more accurate equation
V, = sum of applied and built in potential. (3) for aeff. WI is defined as (12e/qs)1’3 which,
Putting the integral of equation (2) in the form from equation (14) can readily be seen to be the
of space charge width for one volt total bias (including
+ 20 + so the built in voltage).

s
- to
as,, dx = 1 = 2
s
0
~efr dx (15)
As shown in Fig. 3 the discrepancy between the
two curves is small (about 10 per cent in V, over
the most important region of s). Equation (21)
-the latter form of the integral being due to the which is implicit in s and V, does not lend itself to
assumed symmetry-inserting the value for ael, direct calculations and a solution must be ap-
from equation (5) and changing the integration proached by a series of approximations. It is there-
variable there follows fore not so readily applicable to device design cal-
0 culations as the form for VB of equation (19). In

C-
2E II2

[1
4s s E’dE
(E_Eo)lE
+%I
= I. (16)
any case the discrepancy between the two curves
of Fig. 3 could be even further reduced by match-
ing the curves at their point of closest approach,
CALCULATION OF AVALANCHE BREAKDOWN VOLTAGES OF SILICON p-n JUNCTIONS 43

1000

l-
19
10 1020 1021
Impurity concentration gradient 3 in cme4

FIG. 3. Breakdown voltage vs. concentration gradients for linearly

graded junctions in silicon.

which is near V, = 100 V and which can be accurate equation (19) and differs from it with
effected by adjustment of the constant C of values of V, lower by approximately 10 per cent.
equation (5).
Finally, an approximation of the integral of Acknowledgements-The author is indebted to Prof.
equation (2) which has often been resorted to in C. A. HOCARTH for a critical reading of the manuscript
the literature on V, calculations, is to take half the and would like to thank the Directors of Standard
TeIecommunication Laboratories Ltd. for permission to
value of aelf at the maximum field E, in the
publish this paper.
junction, multiply this by the space charge width
and equate this to unity. This amounts to a rough REFERENCES
averaging of Q~,,. Performing then this calculation 1. B. M. WUL and A. R. SHOTOV, Multiplication of
Electrons and Holes in p-n junctions. In Congress
on Solid State Physics, Brussels, 1958, Academic
(22) Press, London.
2. A. G. CHYNO~ETH, Phys. Rev. 109,1537 (1958).
3. J. SHIBLDS, J. Electron. Control 6, 130 (1959).
where xr = total space charge width as before. 4. J. MASERJIAN, J. appl. Phys. 30,1613 (1959).
Using equation (14) for E, and xr and putting 5. S. L. MILLER, Phys. Rev. !W,l234 (1955).
V,, = V, there results 6. S. L. MILLER, Phys. Rev. 105,246 (1957).
7. J. L. MOLL and R. VAN OVERSTRAETENS,Solid-St.

1
a 4 115 Electron. 6, 147 (1963).
v,-_’ E (23) 8. G. A. BARAFF, Phys. Rev. 128,2507 (1962).
3 [O ; c 9. C. A. LEE, R. A. LOGAN, R. L. BATDORF, J. J.
KLEIMACK and W. WIEGMANN, Phys. Rev. 134,
which shows close correspondence with the more A761 (1964).