ISSN 0005−1144

ATKAAF 43(1−2), 47−53 (2002)

Julijana Divkovi} Puk{ec

Recombination Processes and Holes and Electrons Lifetimes1

UDK 621.38.032
IFAC IA 4.0.1
Original scientific paper

In the semiconductor with indirect band gap, such as silicon, recombination on a deep center determines the
lifetime of electrons and holes. In this article lifetime is calculated in dependence of both recombination processes,
Shockley-Read Hall and Auger. The calculations of lifetime are made for gold in silicon, taking into account both
deep levels and neglecting one of them. It is found that in the most cases gold, although having two deep levels,
will act as a single level deep impurity. Exceptions are high injection levels where both deep energy levels have in-
fluence on recombination process. According to the measured values of lifetime it is confirm that the capture co-
efficients are temperature dependent and that the both recombination processes, Shockley-Read-Hall and Auger
have significant influence on a lifetime.
Key words: recombination, deep impurity, lifetime

1 INTRODUCTION In this work we will focus only on the Shockley-

-Read-Hall and Auger recombination processes. The
The lifetime depends on a recombination pro- lifetimes of minority and majority carriers are cal-
cess. The recombination and it's opposite – a gene- culated, assuming single level deep impurity and
ration process – are processes where electron goes impurity with two deep levels. In all calculations I
from conduction to valence band and vice versa.
have used simulation program of my own design.
An electron can travel directly from one band to
The numerical results are compared with the experi-
another, or its transition can be supported by deep
mentally obtained lifetimes.
energy level introduced into the band gap. In the
semiconductor with indirect band gap, such as Si,
Ge, GaP, the probability of direct transition be- 2 SHOCKLEY-READ-HALLAND AUGER
tween valence and conduction bands is very small, RECOMBINATION
because it is phonon participated. In such a semi-
conductor the deep impurity must be added. The In recombination and generation processes the
deep impurity having energy layer deep in the band certain amount of energy is involved. In the Shock-
gap acts as a recombination center The aim of such ley-Read-Hall process we do not have to take into
centers, often called traps, is to enable fast recom- consideration how required energy is provided,
bination and short electrons' and holes' lifetime. whereas in the Auger process we do.
When moving from one band to another electron In the Shockley-Read-Hall recombination process
changes its energy; for generation an electron must only two particles are considered, an electron and a
receive a certain amount of energy while in recom- hole. The Auger recombination is such a process
bination the energy will be taken away from it. The where three particles are involved; two of them in
most popular and the simplest model of recombina- the process of generation or recombination, and the
tion is Shockley-Read-Hall, but it did not take into third that takes or gives some energy. The probabil-
consideration the energy involved in the process. ity of a recombination is proportional to the con-
There are several models of recombination proces- centrations of all involved particles; in the case of
ses; the difference between them is in the way of Shockley-Read-Hall recombination the probability
removing the excess energy: the main are Auger, that the process will take place is proportional to
optical, multiphonon and cascade [1]. the product of concentrations of the electrons and
the holes n ⋅ p, while in the case of the Auger pro-
cess probability is proportional to n2 ⋅ p or n ⋅ p2.
1) This work is part of the research project »036001 Re- Accordingly, it can be concluded that the Auger
search on VLSI/ULSI Semiconductor Structures« suppor- process will be more significant as concentrations of
ted by Ministry of Science of the Republic of Croatia. electrons and holes become greater.

AUTOMATIKA 43(2002) 1−2, 47−53 47

Recombination Processes and Holes and Electrons Lifetimes Julijana Divkovi} Puk{ec

If only Shockley-Read-Hall recombination is as- ∆n ∆p

sumed four processes are possible: τn = τp = (1)
dn dt d p dt
1. the electron from conduction band can be trap-
ped by an empty trap, ∆n, ∆p – excess number of electrons and holes,
2. a filled trap can emit an electron in the conduc- dn/dt and d p/d t – rate of recombination of elec-
tion band, trons or holes.
3. an empty trap can emit a hole in the valence The rate of recombination of electrons and holes
band (in fact, an empty trap will attract an elec- depends on the magnitude of the disturbance and
tron from the valence band) and on the recombination process.
4. a trap filled with an electron can trap a hole
(which means such a trap will emit an electron 3.1 The deep impurity with one deep energy level
in the valence band). The recombination and generation processes de-
In an Auger recombination process we have all pend on the capability of a recombination center to
four mentioned processes, but in each of these cer- capture or emit an electron. These capabilities are
tain amount of energy will be given to, or taken described by capture and emission rates. The elec-
from, an electron or a hole. So, in this case, there tron and hole capture in a Shockley-Read-Hall pro-
are eight Auger processes. cess are described by an electron cn (cm3s −1) or a
hole capture rate cp (cm3s −1) [2, 3]. In the Auger
Those events are shown in Figure 1. process [4, 5] those capture rates will have two ad-
ditional subscripts – A is for Auger and the third
letter determines the third particle involved in the
process; e. g. cAnp (cm6s −1) is the capture rate for an
electron while a certain amount of energy is given
to the hole. The electron and the hole emission
from the trap are described by the emission rate
en, ep (s −1) for the Shockley-Read-Hall recombina-
tion and eAnx, eApx (cm3s −1) for the Auger recombi-
nation.
In the steady-state conditions net change of the
electrons in the conductance band and the holes in
the valence band equals zero; from that conditions
a) Shockley-Read-Hall recombination connection between capture and emission rate can
be obtained [3]:
for Shockley-Read-Hall recombination:
en = cn ⋅ nT ep = cp ⋅ pT (2)

for Auger process:

e Ann = c Ann ⋅ n T e Anp = c Anp ⋅ n T (3a)

e Apn = c Apn ⋅ p T e App = cApp ⋅ p T . (3b)

Concentrations nT and pT are concentrations of

electrons or holes if the Fermi level is equal to the
deep impurity energy level EF = ET.
b) Auger recombination The mathematical calculations for obtaining the
recombination rate are described in details in [3, 4,
Fig. 1 Recombination processes
6]. A deep energy level can have an electron or can
be empty; if there is an electron it can be send to
3 THE CARRIERS' LIFETIME
the conduction band, or if the level is unoccupied it
can capture an electron from the conduction band.
For the certain disturbance from equilibrium in Shortly, the net change of the electrons' density in
carrier density ∆n or ∆p, the recombination and ge- the conduction band can be given as a sum of
neration processes will tend to revert back the semi- emission process (electrons go to the conduction
conductor into equilibrium condition. The lifetimes band) minus capture process (electrons go from the
of electrons τn and of holes τp are defined as [2, 3]: conduction band), using relations (2) and (3). The

48 AUTOMATIKA 43(2002) 1−2, 47−53

Julijana Divkovi} Puk{ec Recombination Processes and Holes and Electrons Lifetimes

same is valid for a net change of the holes' density will be neutral when its energy level is occupied by
in the valence band. an electron and will be ionized positively when the
level is empty. The atom of acceptor type is neutral
The rate of the disappearance of electrons and when its level is unoccupied or it is negatively
holes by recombination in the steady state condi- charged when the level is filled with an electron. In
tion is determined as: (6) it will be:
dn d p n ⋅ p − ni2
= = . (4a) ∆ NT− ≠ 0 and ∆ NT+ = 0 if a deep impurity is
dt d t τ p0 ⋅ ( n + n1 ) + τ n0 ⋅ ( p + p1 ) acceptor type, or
NT is the concentration of the deep impurity. ∆ NT− = 0 and ∆ NT+ ≠ 0 if a deep impurity is
donor type .
The times τn0 and τp0 are defined as
1 1 To obtain electrons or holes lifetime we must
τ n0 = and τp 0 = .
( cn + c An ) NT ( c p + c Ap ) NT solve equations (4) and (6). Using (4) and (6) the
solution for ∆ n or ∆ p is obtain in the form of a
(4b) polynom of 2th degree if only Shockley-Read-Hall
recombination is taken into account, or 3th degree
The capture coefficients cn and cp that have been
if both, Shockley-Read-Hall and Auger recombina-
described earlier, are related to the Shockley-Read-
tions are considered. In the calculation the minori-
-Hall recombination process, while the capture co-
ty excess concentrations is proposed, and as a solu-
efficients related to the Auger process are:
tion we obtain the excess concentration of majority
c An = c Ann ⋅ n + c Anp ⋅ p c Ap = c Apn ⋅ n + c App ⋅ p. carriers. The solution is obtained using simulation
program of our own rendered in Fortran.
(4c)
n, p – nonequillibrium concentrations of electrons 3.2 The deep impurity with two deep energy levels
and holes.
If the deep impurity introduces more than one
The times τn0 and τp0 are the minimum possible deep energy level situation will be much more com-
lifetimes for the electrons in the p-type of semicon- plex. The most common deep impurity used as a
ductor, and for the holes in the n-type. As it can be recombination center is gold. Gold in silicon ex-
seen, if only Shockley-Read-Hall recombination hibits two deep energy levels, one of a donor type
exists those minimum possible lifetimes depend on- (neutral when occupied), and the other which be-
ly on the concentration of recombination center NT haves as an acceptor (neutral when empty). The
(and its capture capacity), while in the case of mathematical procedure is the same as the one de-
Auger recombination, according to equation (4), scribed in detail in [3, 6], and the obtained expres-
those lifetimes depend on the concentrations of sion for an excess concentration for majority carri-
electrons n and holes p as well. ers is of the 5th degree if both recombination pro-
cesses, Shockley-Read-Hall and Auger, are taken
3.1.1 Calculation of the lifetime into account. This expression can not be solved ana-
lytically. If only Shockley-Read-Hall recombination
According to the expressions (1) and (4) lifetime
is considered the expression is of the 3th degree.
will be determined when excess concentration of
electrons and holes are known. All nonequilibrium
concentrations n, p, NT are given as a sum of equi- 3.3 The dependence of lifetime on the injection level
librium n0, p0, NT 0, and excess concentrations ∆n,
∆p, ∆NT: 3.3.1 Low injection level

n = n0 + ∆ n , p = p0 + ∆ p , N T = N T0 + ∆ N T . (5) The semiconductor is in the low injection level

condition if the non equilibrium concentration of
minority carriers is lower than the equilibrium con-
Those expressions for excess carriers concentra-
centrations of the majority carriers. If we presup-
tions must be inserted in (4). We shall presume the
pose a single level recombination center, in the n-
condition of a neutrality on a semiconductor with
-type of semiconductor under the low injection will
the resulting condition:
be valid:
∆ n + ∆ NT− = ∆ p + ∆ N T+ . (6)
p = p0 + ∆ p Ñ ∆ p << nT , pT << n = n0 + ∆ n Ñ n0 .
The deep impurity can be either of a donor type In this case the lifetime of a holes, a minority
or of an acceptor type. If it is of a donor type it carriers, will be equal to:

AUTOMATIKA 43(2002) 1−2, 47−53 49

Recombination Processes and Holes and Electrons Lifetimes Julijana Divkovi} Puk{ec

1 Gold in silicon introduces two deep energy levels,

τ p ≈ τ p0 ≈ . (7a)
NT ⋅ ( cp + cApn ⋅ n 0 ) one is acceptor type and the other is donor type.
Their positions and a capture coefficients at T =
The lifetime of the majority carriers is longer = 300 K are [3, 7]:
than the lifetime of the minority carriers.
EA = E c − 0.54 eV,
 ( cp + cpn ⋅ n 0 ) NT 
τn ≈ τ p 1 + . (7b) cm3 cm3
( cn + cnn ⋅ n 0 ) n 0 cn = 5 ⋅ 10−9 , cp = 1 ⋅10−8 ,
  s s
In the both cases, calculation and experiment ED = E v + 0.35 eV,
confirm that if the deep impurity introduces two
energy levels, at the low level injection only one cm3 cm3
cn = 3.5 ⋅10−9 , cp = 3 ⋅10 −8 .
level is important for a recombination process, and s s
the minority carriers' lifetime will be equal to the
τp0 or τn0 [3, 6, 8]. Ev and Ec are the energies of a bottom and a
top of a bandgap.
The minority carriers' lifetime as a function of a
gold density is calculated and shown in Figure 2. Calculations of lifetimes presented in Figure 2
The lifetimes are considered first using both deep have been obtained presupposing the low injection
levels introduced by gold in silicon band gap and level and Shockley-Read-Hall process only. As it
then only one of the two. can be seen if the density of a deep impurity is
lower than the shallow impurity density, only one
level is important for recombination process, the
one closer to the Fermi level of a semiconductor.
In the p-type recombination goes through the
donor level ED, the lower one, while in the n-type
the important deep level is higher one, the acceptor
level EA.

3.3.2 High injection level

Under the high injection level the concentrations

of minority and majority carriers become equal:
p Ñ ∆ p Ñ n Ñ ∆ n >> nT , pT .

The lifetimes of electrons and holes tend to be

equal τn ≈ τp0 too. For recombination through one
deep energy level lifetimes will be [3, 6]
a) p-type
τn ≈ τ p ≈ τ n0 + τ p 0 =

1  1
10−3 =  ( c + c ⋅ ∆ n + c ⋅ ∆ p) +
NT  n nn np

1 
+
( cp + cpn ⋅ ∆ n + cpp ⋅ ∆ p) 
.
(8a)

For recombination center with two energy levels

ET1 and ET2 those lifetimes will be somewhat shor-
ter then in the case of one energy level and are ex-
pressed as:
τn01 ⋅ τn02 + τ p01 ⋅ τn02 + τ p02 ⋅ τn01
τn ≈ τ p = . (8b)
τ p01 + τn02
b) n-type The lifetime at the high injection level is longer
Fig. 2 Lifetime of a minority carriers in a silicon doped with shal- than at the low injection. That is valid in both cas-
low impurity and gold es, taking into account both deep energy levels or

50 AUTOMATIKA 43(2002) 1−2, 47−53

Julijana Divkovi} Puk{ec Recombination Processes and Holes and Electrons Lifetimes

only one of them, as can be seen according to 5 EXPERIMENTALLY AND THEORETICALLY

equations (7) and (8). This has also be demonstrat- OBTAINED LIFETIME
ed on Figure 3, where the lifetime as a function of
a disturbance of a minority carriers' density ∆n or 5.1 Description of the experiment
∆p are shown. In the experiment described in [8] the power
p +pnn + diode was used. On the base of a given po-
sition of a Fermi level as EF = Ec − 0.35 eV, the shal-
low donor concentration in an n-region was esti-
mated to be ND =⋅ 1014 cm−3. This was done using re-
lation
Nc
EF = Ec − kT ⋅ ln
n
with T = 300 K, Nc = 3.81 ⋅1019 cm−3 [13] and n =⋅ ND.
To such a structure various deep impurities were
added. Because gold is the best described deep im-
purity, it is used here for comparison between theo-
retically and experimentally obtained lifetimes. In [8]
at low injection level measured lifetime was τp ≈ 1.8
µs; using (4b), neglecting Auger recombination the
gold concentration is estimated as NT =⋅ 6 ⋅ 1013 cm−3.

a) p-type
The measurement of the lifetime was made as a
function of a current density in the interval from
5 mA/cm2 to 0.5 A/cm2. To connect current with
carriers’ density the Shockly relation was used:
∆p
J p = q ⋅ Dp ⋅
Lp
where:
kT
Dp = ⋅ µp
q
Lp = Dp ⋅ τ p .

τp is lifetime of holes and µp its mobility.

Using T = 300 K, ND =⋅ 1014 cm−3, holes mobility
µp =⋅ 460 cm2/Vs [13], measured lifetime τp (between
≈ 2 – 3 µs), after a rough calculation results in an in-
jection level going from ∆p =⋅ 1013 cm−3 to ∆p =⋅ 1015
b) n-type cm−3.

Fig. 3 The lifetimes of a minority carriers as a function of the in- In Figure 4 the lifetime of a holes as a function
jection level of an injection level is shown. Measured values are
compared with the calculated ones for three diffe-
rent temperatures. The calculated values of lifetimes
Calculated lifetimes given in Figure 3 are for
are obtained inferring that only the acceptor level
gold in silicon taking into account first the both
of gold will take a role in the recombination
and than only the one deep level. As it can be
process in an n-type of silicon and only the Shock-
clearly seen from Figure 3 at the low level injection
ley-Read-Hall recombination is considered.
only one level is important for recombination
process, as it was already shown in Figure 2. The According to the behavior of experimentally ob-
lifetime rises with the rise of injection level. At the tained lifetime given in Figure 4 we can see that
very high injection levels both energy levels have the lifetime will be longer as injection level be-
influence on recombination process. The lifetime is comes higher, which was predicted in equation (8).
shorter when both levels are considered than when Further on, it is obvious that for all injection levels,
only one is taken into account. as temperature raises the lifetime rises as well. This

AUTOMATIKA 43(2002) 1−2, 47−53 51

Recombination Processes and Holes and Electrons Lifetimes Julijana Divkovi} Puk{ec

is the consequence of the temperature dependence

of capture coefficients [8]. According to measured
lifetimes a capture coefficient must decrease as
temperature increases. The exact temperature de-
pendence of a capture coefficient is not known, but
it is predicted in [9, 10] for the Shockley-Read-Hall
recombination:
32
c = c (300) ⋅ 
300 
 . (9)
 T 

Fig. 5 The lifetime as a function of an injection level at T = 17 °C

It can be seen that for low injection the slope of

a lifetime vs. injection level is almost the same for
measured and calculated values. This is obtained us-
ing in (10) parameter a = 105.
The difference between measured and calculated
lifetimes at high injection levels still exists, but is
smaller than it was predicted by the Shockley-Read-
-Hall recombination (compare Figures 4 and 5).
This can be attributed to the influence of a second
deep energy level introduced by gold in silicon.
Fig. 4 Silicon doped with gold density of 6 ⋅ 1013 cm−3
6 CONCLUSION
Looking at Figure 4 it can be seen that the dif-
The deep impurity added into a semiconductor
ference among the measured and the calculated
in order to support the recombination process will
lifetimes becomes greater as injection level rises.
have, according to (4) shorter lifetime as its density
Calculated values rise faster then measured, which
is higher. But, the density of deep impurity must
implies that the capture coefficients depend on the
not be too high, in comparison with the shallow im-
injection level, that means on to the carriers' densi-
purity, if we wish to avoid its influence on to the
ty, and the Auger recombination must be conside-
semiconductor resistivity [11, 12].
red too. Involving the Auger recombination into the
calculation it becomes very complex. To make it Owing to the measured values presented in [8] it
simpler it is usual to predict that the Auger capture has been confirmed that the multilevel deep impu-
coefficients for an electron when the energy is ex- rity, such as gold in silicon, can be treated as a sin-
change with other electron is the same as in the gle level. This is valid until the very high injection
case when the third particle is a hole. The same is level is reached when both levels become important
valid for capture coefficients of a hole. Those coef- for recombination process. Than, of course, two le-
ficients are expressed as a part of the capture coef- vels are more effective than only one and the exact
ficients for the Shockley-Read-Hall recombination lifetime is shorter than those calculated with only
according to [4, 5]: one level, as can be seen in Figures 3, 4 and 5.
cn The comparison between measured and calculated
cAnn = cAnp = (10a) values, given in Figure 4 gave us more interesting
a ⋅ ni
information about the recombination process. As it
cp is already mentioned, only the Shockley-Read-Hall
cApn = cApp = . (10b)
a ⋅ ni recombination has been taken into account. Bearing
this in mind the obtained difference between mea-
In Figure 5 the measured lifetimes at a room tem- sured and calculated lifetimes implies that the re-
perature (17°°C) are shown and compared with the combination processes depend on the carriers' con-
calculated values obtained considering both types of centrations, which means that the Auger type of re-
recombination, Shockley-Read-Hall and Auger. combination has certain influence on to the recom-

52 AUTOMATIKA 43(2002) 1−2, 47−53

Julijana Divkovi} Puk{ec Recombination Processes and Holes and Electrons Lifetimes

bination process. The lifetime in the semiconductor [3] J. L. Moll, Physics of Semiconductors. McGraw-Hill, 1964.
must be modeled taking into account the Shockley- [4] D. A. Evans, P. T. Landsberg, Recombination for Auger Ef-
-Read-Hall and the Auger recombination with the fects with Applications to p-n Junctions. Solid-State Elec-
aim to achieve calculated value close to the real tronics, 1963, vol. 6, pp. 169–181.
one, as it has been done and present in Figure 5. [5] J. G. Fossum, R. P. Martens, D. S. Lee, J. F. Nijs, Carrier
Recombination and Lifetime in Highly Doped Silicon. So-
The measured values also confirm that the re- lid-State Electronics, 1983, vol. 26, No. 6, pp. 569–576.
combination process is temperature dependent. The
[6] J. Divkovi} Puk{ec, Influence of Deep Impurities on Elec-
lifetime is longer as the temperature is higher. trical Characteristic of Semiconductors Devices. Ph.D.
In the end it must be pointed out that the re- Thesis, University of Zagreb, 1996. (in Croatian).
combination of free carriers in a semiconductor [7] A. G. Milnes, Deep Impurities in Semiconductors. J. Wi-
trough the recombination centers has never been ley&Sons, New York 1973.
the subject of a greater interest. As it was stated 20 [8] V. Benda, Using Carrier Lifetime Dependece on Tempera-
years ago [9] »...The problem is a difficult one and ture and Current Density in Diagnostic of Silicon Structu-
res. EPE-MAVEP, pp. 0–065 to 0–068, Firenca 1991.
the understanding was restricted to a global and
heuristic interpretation...«. Today, the situation is al- [9] H. J. Queisser, Recombination at Deep Traps. Solid-State
most the same. That is why there are no data about Electronics, 1978, vol. 21, pp. 1495–1503.
capture coefficients for deep impurities, with the [10] J. Divkovi} Puk{ec, The Influence of Temperature on Life-
exception of gold. To find out the capture coeffi- time. Proceedings of Melecon-98, vol. 1. pp. 349–353, Tel
Aviv 1998.
cients' temperature dependence, the values of
Auger coefficients and all relevant data for recom- [11] B. I. Boltaks, G. S. Kulikov, R. Sh. Malkovich. The Effect of
bination process, many more experiments like those Gold on Electrical Properties of Silicon, Soviet Physics So-
lid State, vol. 2., No. 2, pp 181–203, February 1960.
described in [8] must be done.
[12] M. Valdinoci, L. Colalongo, A. Pelegrini, M. Rudan, Analy-
sis of Conductivity Degradation in Gold/Platinum-Doped
REFERENCES Silicon. IEEE Tr. on Electron Devices, vol. 43., pp 2269–
[1] N. F. Mott, Recombination: a Survey. Solid-State Electro- 2275, December 1996.
nics, 1978, vol. 21, pp. 1275–1280. [13] J.Š[ribar, J. Divkovi} Puk{ec, Physics of Semiconductor De-
[2] J. S. Blakemore, Solid State Physics. Saunders, Philadelphia vices: Solved problems with Theory. Vol. I, Element, Za-
1974. greb, 1994. (in Croatian).

Rekombinacijski procesi i vremena `ivota {upljina i elektrona. Poluvodi~ima s indirektnim zabranjenim poja-
som, kakav je silicij, dodaju se duboke primjese s ciljem da bi se postiglo odre|eno vrijeme `ivota elektrona i
{upljina. U ~lanku je razmatrano vrijeme `ivota uzimaju}i u obzir dva osnovna tipa rekombinacijskih procesa,
Shockley-Read-Hallov i Augerov. Pri prora~unu je kao duboka primjesa uzeto zlato, koje u silicij unosi dvije
duboke energetske razine. Ra~unski je pokazano, a i eksperimentom potvr|eno, da je u ve}ini slu~ajeva za rekom-
binacijski proces bitna samo jedna duboka razina i to ona koja je bli`a Fermijevoj razini poluvodi~a. Iznimka je rad
poluvodi~a pri visokoj injekciji kada se rekombinacija obavlja preko obje razine. Eksperimentom je potvr|eno da
koeficijenti zarobljavanja ovise o temperaturi, te da oba rekombinacijska procesa, Shockley-Read-Hallov i Augerov,
treba uzeti u obzir pri prora~unu vremena `ivota. Pri prora~unima je kori{ten vlastiti program, kojim je mogu}e,
osim prora~una vremena `ivota uzeti u obzir i ostale efekte koje duboka primjesa ima na elektri~ka svojstva polu-
vodi~a, kao npr. utjecaj na vodljivost poluvodi~a, na {irinu i na kapacitivnost osiroma{enog sloja.

Klju~ne rije~i: rekombinacija, duboka primjesa, vrijeme `ivota

AUTHORS ADDRESS:
Asst. prof. Julijana Divkovi} Puk{ec, Ph. D.
Faculty of Electrical Engineering and Computing,
University of Zagreb
10000 Zagreb, Unska 3, Croatia
Received: 2002−10−05

AUTOMATIKA 43(2002) 1−2, 47−53 53