Solid-State Electronics Pergamon Press 1965. Vol. 8, pp. 211-226.

Printed in Great Britain

SURFACE CAPACITY OF OXIDE COATED

SEMICONDUCTORS
R. HALL and J. P. WHITE

Radio Corporation of America, Electronic Components and Devices Division, Somerville, New Jersey

(Received 28 April 1964; in revisedform 16 July 1964)

Abstract-Determination of the surface capacity presents a convenient method for investigating the
oxide-semiconductor interface. An exact analysis of the surface state distribution incorporates
capacity measurements made over a wide range of frequencies. When the states are assumed to be
uniformly distributed across the forbidden region, the analysis is simplified and quantitative informa-
tion concerning the type, density and frequency response of the states is readily obtained.
Measurements made on silicon and gallium arsenide reveal that at a test frequency of 1 MC/S the
surface state capacity is relatively independent of the surface voltage. State densities of the order of
101s cm-s V-1 were observed for thermally grown silicon dioxide on silicon and pyrolytically
deposited silicon dioxide on gallium arsenide. The state densities were slightly higher on gallium
arsenide and their response was significantly faster.

R&sum&La determination de la capacite de surface presente une methode convenable a l’examen

de l’interface oxyde-semiconducteur. Une analyse exacte de la distribution de surfaces d’etats
comprend des mesures de capacites faites sur une gamme de frequences tres large. Quand on assume
que les &tats sont distribues le long de la region defendue, l’analyse est simplifiee et les don&es
quantitatives concernant le type, la densite et la reponse de frequence des Ctats sont facilement
obtenus.
Des mesures faites sur le silicium et l’arseniure de gallium demontre qu’a une frequence d’epreuve
d’l MHz la capacite de surface d’etats est relativement independante de la tension de surface. Les
densites d’dtats de l’ordre de lOi cm-2 V-1 ont CtC observees dans le silicium developpe thermique-
ment sur le silicium et dans le bioxyde de silicium depose pyrolitiquement sur l’arseniure de gallium.
Les densites d’etats Ctaient legerement superieures sur l’arseniure de gallium et leur reponse etait
nettement plus rapide.

Zusammenfassung-Die Bestimmung der Oberflachenkapazit~t bietet eine geeignete Methode

zur Untersuchung der Grenzfllche zwischen Oxid und Halbleiter. Eine genaue Analyse der
Verteilung der Oberfllchenzustiinde umfasst Kapazitatsmessungen in einem ausgedehnten Fre-
quenzbereich. Bei der Annahme einer gleichfijrmigen Verteilung der Zustande fiber die verbotene
Zone vereinfacht sich die Analyse, und man erhalt leicht quantitative Information iiber die Art,
die Dichte und die Frequenzansprache der Zustande.
Messungen auf Silizium und Galliumarsenid zeigen, dass bei einer Testfrequenz von 1 MHz die
Kapazitat der OberflPchenzustXnde relativ unabhangig von der Oberfllchenspannung ist. Man
beobachtete Zustandsdichten von der Griissenordnung 1012 cm-sV-l an thermisch geziichtetem
Siliziumdioxid auf Silizium und pyrotechnisch abgelagertem Siliziumdioxid auf Galliumarsenid.
Die Zustandsdichte war auf Galliumarsenid etwas hiiher und ihr Ansprechen erfolgte bedeutend
schneller.

1. INTRODUCTION the quality of the oxide-semiconductor interface.

THE OXIDE coated semiconductor is an integral An investigation of the interface may be achieved
part of many recent solid state devices. These in- by measuring the surface capacity of the oxide-
elude the silicon planar and MOS transistors,(lls) coated semiconductor. This capacity can be
the cadmium sulphide thin film transistorc 3~4) and measured by evaporating a metal electrode over the
the gallium arsenide MOS transistor.(s) Successful oxide and using this electrode and the semi-
operation of these devices is largely dependent on conductor as the parallel plates of a capacitor
211
212 R. I-IALI, and
whose dielectric is the oxide. The structure was
first proposed as a variable capacity device by
PFANN and GARRETT.(~)
The basic theory of the charge relations at the
semiconductor surface under thermodynamic
equilibrium was developed by GARRETT and
BRATTAN and was applied by LINUKER@) to
characterize silicon-silicon dioxide R’IOS varactors.
He attributed the discrepancies between theoretical
and experimental curves to a donor surface state
density of 5 x 1011 cm- 2. Surface states at the
silicon-silicon dioxide interface were investigated
extensively by TERMAN,@) by measuring the
characteristics of an KIOS varactor over a wide
range of frequencies. He determined the location,
densit! and frequency response of discrete
trapping levels at the interface. Recently, LE HOVW
and SI.OBOII)SROY(~~) have derived equivalent
circuits for the ll’IOS system in which they consider
deviations from Boltzmann distributions.
The complexity associated with the intcr-
pretation of capacity measurements makes it a
formidable task for the device engineer to study the
effects of processing on the surface state para-
meters. In this paper the analysis of capacity
measurements is simplified so that information
concerning the type, density and frequency re-
sponsc of the interface states may be readily
obtained. This simplification is accomplished by
assuming that the distribution of interface states
in the forbidden gap is a continuous function of
energy and that at equilibrium these states arc
filled up to the bulk Fermi level. It may be shown
that the semiconductor surface capacity is the
parallel combination of the space charge and
surface state capacities, which in general are both
functions of thk surface voltage.
However, if the surface states are assumed to be
distributed uniformly the surface state capacity is
independent of the surface voltage. This makes it
possible to easily obtain approximate values of the
surface state parameters from surface capacity
measurements. A proccdurc is formulated using
this assumption and applied to surface varactors
fabricated on silicon and gallium arsenide. The
error in the assumption of a uniform distribution
of surface states may be seen by using the cal-
culated parameters to plot the corresponding
characteristic. It is shown that for both gallium
arsenide and silicon a reasonable agreement is
J. I?. \VHITE
achieved between the experimental curves and
those plotted. Thus, although no claim is made
that the distribution of surface states is uniform,
the usefulness of this method to obtain cluantitnti\.e
information is established.
2. GENERAL CONSIDERATIONS
It is known that there arc available states for the
electron in the proximity of the semiconductor-
oxide intcrfacc, within the energy range of the
forbidden gap. ‘I’hesc states arc normally divided
into slow anti ‘fast'. The slow states arc associated
with the trapping of charge within the oxide, and
because of their nature respond to applied fields
in the order of minutes to months. l’hc type and
densit.y.of these states arc dependent on the oxide
depositlon or growth, and they cause a built-
in field at the semiconductor surface. ‘l‘he ctfect
of these states on C-V characteristic of surface
varactors is merclg a displacement of the voltage
axis. They cause no change on the shape of the
curve. For this reason the effect of slow states is
neglected in this paper. ‘1%~ fast states arc thought
to exist at or near the interface and arc closely
associated with the free charges of the bulk
material. They have response times of seconds to
microseconds and arc greatly ctfccted bv surface
treatment, direction of crystal oricn&on, etc.
The distribution of fast states over the forbidden
gap may be assumed to bc filled up to a level $,+
Equilibrium is reached between the bulk and the
surface by an interchange of electrons which cause
hand bending at the surface. ‘l?he direction of band
bending depends on the relative positions of ds
and 40, the Fermi level in the bulk [see Figs. l(a),
(b), (c) and (d)]. It Ilecomes established when the
Fermi energy of electrons in the surface states is
equal to the Fermi level in the bulk. The amount
of band bending is determined by the surface state
distribution and by the potential difference
between +s and 40, and is represented by a potential
$so at the oxide-semiconductor interface.
Since the charges in the fast surface states and
those in the bulk material near the surface may be
displaced by the application of an electric field,
then there must bc a capacity associated with both
the space charge region and the surface states.
Separation of these capacities in the C-V character-
istic of the surface varactor allows the surface state
parameters to be determined.
 SURFACE CAPACITY OF OXIDE COATED SEMICONDUCTORS 213

----__
(al n-lype semiconductor c#, >c#B~

NOTATION Permittivity of semiconductor

Explanation Potential of filling level from mid-band
Potential of fermi level in bulk from mid-band
Surface state density per volt
Potential of surface with respect to bulk
Measured capacitv of surface varactor
Surface potential at minimum surface capacity
Normalised capacity
Surface potential with no applied field
Minimum normalised capacity
Angular frequency
Maximum normal&d capacity
Angular test frequency
Oxide capacity
Space charge capacity
Minimum space charge capacity 4. SURFACE VARACTOR CAPACITANCE
Space charge capacity for level bands The capacity of the device in F-cm-a is given by
Surface state capacity the expression
Field
Frequency factor
c meas dQP
-,_r
Uniform surface state density = (1)
Bulk density of electrons “VaPP
Density of ionised acceptors or donors where Q, is the charge on the plate in C/cm2 and
Bulk density of holes
Electronic charge
Vam is the applied voltage.
Charge on plate The net charge in the system is zero
Total space charge
Total space charge with no applied field Qp+Qsst+Qst = 0
Total space charge at minimum capacity therefore,
Charge in surface states
Applied voltage Qp = - (Qsst+ Qd (2)
Voltage across oxide
where Qsst is the total charge, per cm”, in the
Applied voltage to produce Cntmax)
Applied voltage to produce Cntmrn) surface states and Qst is the total charge, per ems,
qlk T in the space charge region.
214 R. HALL and J. P. WHITE

The applied voltage neglecting the difference

e?l--
in contact potential is the sum of the voltage across
the oxide and the change in voltage across the
semiconductor, thus

V aPP = Vm+(~s-#so) (3)

where lclS is the potential of the band edge at the
surface with respect to its value in the bulk and is
taken positive for bands bending in the direction
c.x L,t-J
of increasing hole potential. FIG. 2. RlOS surt’ace varactor equivalent circuit
Applying Gauss’s theorem assuming that there is (omitting resistances).
no charge in the oxide we obtain using equation (2)
(i) Space charge capacity
FoIlon-ing the approach taken by Garrett and
Brattain, the field at any point in the space charge
region is given by
Differentiating equations (2) and (4) with respect
to +S yields

ws + no(exp(B$) -P#- I)]’ 2 (8)

and
d Vaiapl, c,, + CS
___ = !+---
(6)
dh c OX
where Applying Gauss’s theorem in the space charge
dQsst
- - dQst region.
c,, = ___ and C,s = -,
d#s d$s Qst = -EE($,) (9)

and are the capacities associated with the surface

states and the space charge region. Equation (1)
can be written as
Because the sign of Qst is always opposite to #,, T
dQ&#s
c is replaced by the dimensionless factor -#S/i#Sl.
meaS = dVa’,,,/d#, Figure 3 illustrates Qst plotted as a function of #S
Thus, if equation (5) is divided by equation (G), for 1016 p-type silicon and gallium arsenide. The
the following expression is obtained curve monotonically decreases for all values of tiS.
It decreases rapidly in the enhancement region

c,, = = _ es,+ c,
FcgE (7)
and then goes through the origin and decreases
slowly up to gL, z 240, (the bulk Fermi level). For
OX css
+cs+ cox larger values of #,, the charge decreases much
where C’oz has been brought to the right side to more rapidly.
make the equation dimensionless. The capacity associated with the space charge
The form of this expression indicates that the region is defined as C, = (- dQ,t/d#,), i.e. as the
oxide capacity is in series with the parallel com- negative of the slope of the Qst vs. (crS curve.
bination of the surface state capacity and the space Therefore, the capacity of the surface goes through
charge capacity. The equivalent circuit omitting a minimum at the point of inflection of the Qst vs.
the resistance through which these capacitors are z/s curve. The value of the space charge at
charged and discharged is shown in Fig. 2. The minimum capacity, Qstm, and the corresponding
expressions for the space charge capacity and the value of surface voltage, #,, are plotted in Figs. 4
surface state capacity will now be derived. and 5 as a function of concentration.
 SURFACE CAPACITY OF OXIDE COATED SEMICONDUCTORS 215

KY cme3p- type material

arsenide

inversion

FIG. 3. Space charge vs. surface voltage.

ICF Positive for n-type

Negative for p-type

0 I 1 I I I
Id” Id5 IO’” lOI’ ICF Id9
Concentration, cme3
Concentration, cmm3

FIG. 4. Space charge at minimum surface capacity vs. FIG. 5. Surface voltage at minimum surface capacity vs.
bulk concentration. concentration.
216 R. HALL and J. P. WHITE

analytical expression the space should be noted that a displacement of charge

capacity may obtained by from the surface is not necessary to have a
equation (lo), differential space charge capacity.) This can be
readily obtained, using equation (1 l), as

(11)

Figure 6 illustrates Cs as a function of *S, for 1016 = [&PO+ no)]l!“. (13)

p-type silicon and gallium arsenide (complete
ionization was assumed). As predicted from the
variation of space charge as a function of surface

, 10l6p-type silicon

I/’ lo’6p-
gallium
type
arsenide

I I I I
IOr’ lot6 d’ ION8 10’9

Concentration, cm.3

If C(y5) is the total state density per \.olt, in-

cluding all states at the oxide -semiconductor
interface, and 4 range over the forbidden gap and is
, O,ll , , , / , , , L
measured from midgap with positive voltage in the
-02 0 0.2 0 4 06 0.8 I,0 I.2 lb4 direction of increasing hole potential, then the
+$. ”
total charge in the surface states, Q)s,Yt, when the
potential at the surface is *,5 will bc

voltage, the space charge capacity reaches a

minimum value, CSm, at a surface voltage, #,,,
just less than 21~5”.The minimum value of capacity
is a function of concentration; it may be obtained
by setting the differential of equation (11) equal
to zero. For tiS > 0.3 V the minimum capacity
can be approximated by
This expression can he written as
(12)
gw = Qw(40- l/is)- Qss(Th).
Figure 7 illustrates C,,, as a function of con- The capacity of the surface states
centration for silicon and gallium arsenide. ills0
illustrated in Fig. 7 is the space charge capacity
C,, when the bands are level (i.e. #S = 0). (It
 SURFACE C.4PACITY OF OXIDE COATED SEMICONDUCTORS 217

Therefore, using equation (15), this expression

becomes
d&(+o - &)
C,S(lclS) =
d(+o - h>
or
C.sS(~~) = $(+0-M (16)

which indicates
corresponding
that the surface
to a surface voltage
state capacity
(l/IS) is the
x q$Jp
I+lcls-*so. (19)

surface state density per volt at a voltage, +o-$~. The quantity #,,, which is the relative potential
The above analysis assumes that all the surface at the surface when VaBD= 0 and is caused by a
states are able to follow the a.c. signal used to redistribution of charge between the bulk and the
measure the capacity. In general, this assumption surface states, can be determined from equation
is not valid but equation (16) can be modified to (2), using equations (10) and (14) with #S replaced
include the effect of finite surface state charging by #S0. The following equation for I/J,, results in
times by introducing the dimensionless multiplying
factor f(w, lclS).
Equation (16) becomes

CSS($S) = qf(w, MC($o-$S). (17)

where 0 < f(w, $S) < 1 and is a function of This equation relates #,, to the surface state
frequency and surface voltage. At very high distribution and filling level, and to the bulk pro-
frequencies, none of the surface states will be able perties of the semiconductor.
to follow the a.c. signal, therefore, If the state density, C($), the filling level, $s, and
the frequency factor f(~, t,!~~), are known, the
hmf(W’ $S) = 0 and CSS(&) = 0. capacity-vs.-applied-voltage curves can be cal-
culated, using equations (lg), (19) and (20).
At very low frequencies all states will follow the However, the problem is to determine these
signal and functions from measured curves. An approach
was indicated by Terman. If the frequency of the
a.c. signal is made high enough, the surface state
capacitance in equation (18) approaches zero and
5. SURFACE VARACTOR CHARACTERISTICS
the measured capacity is just the oxide capacity
The method used for measuring the character- and space charge capacity in series. The oxide
istics of the surface varactor was to apply a d.c. capacity is a constant, equal to the maximum
bias to the metal electrode and to measure the measured capacity, and the functional relation
capacity with a capacity bridge, which applies a between the space charge capacity and $J~ (equa-
small a.c. signal. The measured capacity can be tion (11)) is known. From the measured capacity
expressed in terms of I,/I~and w, using equations curve, the relation between C, and VapD can be
(7), (11) and (17), as obtained. Therefore, with the aid of equation (ll),
$J~ can be related to Vapp. This relationship is in-
dependent of frequency.
Applying equation (19) and the relationship
between $J~and V,,, to the varactor characteristics
measured at various frequencies, the function
f(w, $S)C(&-~,) can be determined. At very low
frequencies, f(w, z+!J~)+1 and
$a is related to VaPP through equation (4) which
can be expressed, employing equations (10) and
218 K. HALL and J. P. WHITE

giving the overall state distribution. When this is For different bulk concentrations, the value ot
known, it is easy to determinef(w, #S). The filling equilibrium surface voltage may be calculated from
level +,? can then be determined by solving equation (23).
equation (20).
To perform this analysis, it is necessary to
accurately measure the capacity voltage character- qK(C-cmZV ‘1

istics of the varactor over a wide range of fre- IO 6

quencies. Many measurements are required and IO7
the calculations necessary to evaluate a single
varactor are numerous. Terman used a frequency
range from 0 to 100 MC/S.
lO-8 02 04 06 08

Uniform surface state distribution -,O -08 -06-04-02

In this section, theoretical capacity vs. voltage

curves are calculated for the case where the surface
states are assumed distributed uniformly in (alp-type material

energy across the forbidden region, and the

frequency factor is not a function of lclS.Using these
curves it is possible, from a single capacity vs.
voltage characteristic measured at an arbitrary
frequency wg, to obtain readily the following
quantitative information about the surface:
(a) the total number of surface states cm-‘-V-l;
(b) the value of the frequency factorf(wo) ;
(c) the surface voltage under no applied field
ti s0, and
(d) the surface state filling level, qSS.
Assuming that C(4) = K, where K = surface (b) n-type material
state density cm-2-V-‘, equations (18), (19), and
(20) become FIG. X. Plot of equilibrium surface voltage vs. d,-00.

Figure 9 illustrates the normalized capacity of a

surface varactor, as a function of applied d.c.
voltage, when no surface states are present. The
(21) typical case is taken for silicon and gallium arsenide,
with a p-type concentration of 1016 cm-a. Although
the curves are plotted for zero surface states, the
effect of surface states with a density K < 101’
cm-2 V-1 are negligible. As the oxide of the
varactor is made thinner, the measured capacity
follows more closely the variation of capacity of the
semiconductor surface. ,41so, the potential
difference across the oxide is reduced, so that less
f ) . F(&)1/‘2. (23) applied voltage is needed to attain a specified field.
This is seen by comparing the curves for 1000 !I
In Fig. 8, $,0 is plotted for different values of K, and 3500 -4 thick silicon dioxide.
as a function of 4S - 40, for both 1016 cm-3 silicon As shown earlier, the space charge capacity of a
and gallium arsenide. The curves are plotted for semiconductor is a function of bulk concentration.
values of gK = 10-6, 10-7 and 10-s C-cm-” V-1. This is reflected in the normalized capacity curve.
 SURFACE CAPACITY OF OXIDE COATED SEMICONDUCTORS 219

For a given oxide thickness, the lower carrier con- This creates a self depletion region at the semi-
centrations both increase the change in capacity, conductor surface for both n and p-material and
and reduce the voltage spread. causes the curve to shift so that the minimum
capacity lies nearer to the zero applied voltage.

(a) K = 6 x10” crri2 V-’

9, =O

I I I I I I I I I I I I I I I I I !
-4 -2 0 2 4 6 I3 IO -3 -2 -I I23456

VOPP’ V V0PP’ V

FIG. 9. Plot of normalised capacity as a function of

applied voltage for different oxide thicknesses.

The effect of surface states on the normalized

capacity curve is shown in Figs. 10(a) and 10(b).
The curves are plotted for 1Ol6 cm-s P-type (b) K = 6 x IO” cm-’ V-’

material, with an oxide thickness of 1000 _A.

The curves are plotted for surface state densities +s =O

qk FZ 10-s and 10-7 C-cm-s V-1, for a filling level

of dS = 0, and for extremevaluesoff(ws) equal to
0 and 1. It should be noted that the normalized
capacity curves have been plotted with identical
vertical axis; however, the voltage axis for the
high surface state density has been magnified to -20 -15 -10 -5 5 IO 15 20 25

give the complete curve. The curves show that the

VOPP’ V
capacity generally reaches a minimum for positive
applied voltage, for P-type material. For n-type FIG. 10. Normalised capacity curve showing the effect of
material, the minimum is obtained for a negatively surface states.
applied voltage.
The effect of increasing the surface state density If the filling level is not at the center of the
does not effect the minimum capacity, whenf = 0. forbidden gap, the shift of the n and p-type curves
However, it reduces the change in capacity for are not identical. For example, if the filling level
f > 0, as is shown for the case,f = 1. lies in the lower half of the forbidden gap, opposite
The curves are drawn for the case where the the Fermi level for p-type material, then under a
filling level is at the center of the forbidden gap zero applied field, the energy bands of the p-type
(i.e. +S = 0). For p-type material, this causes material will not be bent and no charged carrier will
holes to become trapped in the surface states for no be present in the surface states. However, for
applied field, reducing the hole concentration at the n-type material, a larger number of electrons will
surface of the semiconductor. The effect on n-type be trapped in the surface states and the bands will
material is to trap electrons in the surface states. be bent to a p-type surface. The effect on the
220 R. HALL and J. P. WHITE

capacity curve is to shift its position horizontally. approximate value of C,, illustrated in this figure
When 4, = - I&l, the reverse effect results, the and equation (24), the value offK can be obtained:
n-type material having no states filled, and the
p-type material having a large number of holes in
the surface states.

‘To separate f and K, it is necessary to obtain

(ii) Analysis of measured capacity curve another known point on the normalised capacity
The general shape of the normalized capacity curve. IVhen the energy hands are level, #,$ = 0.
curve is shown in Fig. 11. However, determining The norrnalised capacity can be calculated for this
the values of the parameters K, f(wo), z,h,,and I$~ point by using equation (7), the value of qfKfrom
(assuming that the concentration in the crystal is equation (25) and the approximate value of C‘,,
known) poses a problem. Equation (7) can be re- from Fig. 7, yielding
written for a uniform state density as
cl,, + qfK
C (26)
‘L(max)
= eso+qfK+ C,,’
(24)
The value of J j,, necessary to produce level
bands ( VmRX) can be found from this calculated
where COXis the maximum measured capacity and
value of (Cn)(m~x~, using Fig. 11. The general ex-
C, is the normalized capacity = C,,,,/C&. pression for 16r11,, which is given in equation (22)
The normalized capacity and the space charge may he written as
capacity reach a minimum at the same point
J’min, C,z(min).

At the minimum normalised capacity point,

#s = $,,,, and Qst = QstrrL. Values of Q!stl,L and
$sm are plotted, in Figs. 4 and 5, as a function of
concentration.
The voltage applied to produce this minimum is
then

_ _ Observed turve~

When the bands arc level #I~ = 0 and Q,t = 0.

Under these conditions, the applied voltage may be
expressed as

V OPP

FIG. 11. General shape of normalised capacity curve. Subtracting equation (29) from (2X) yields

For gallium arsenide varactors it was found that

this minimum is not well defined. However, the
analytical method described may still be used, This equation can he solved for K, as,
because the slope of the normalized capacity curve
changes rapidly prior to the minimum, and a good
estimate of (C’n)(min) can he obtained.
The minimum space charge capacity is plotted,
in Fig. 7, as a function of concentration. Using the Since I/;nin and J,,,, are known and #s,,, and Qstnl
 SURFACE CAPACITY OF OXIDE COATED SEMICONDUCTORS 221

can be found from Figs. 4 and 5, K can cal- 6. MEASUREMENTS

culated. Substituting this value of K in equation Surface varactors were fabricated on chemically
(25) yields f(wo). polished wafers of n and p-type silicon and gallium
Using equation (27) with Vapp = 0 and arsenide. Both crystals were cut on the (111) plane;
Q&t = Qsto, and equation (29) the following ex- the arsenic face of the gallium arsenide was used.
pression may be obtained for #,,. The capacitor dielectric was a thin insulating layer
(31) (N2000 A) o f si 1icon dioxide, and the plate of the
Vm,&,, + (Co, + qK)&s, = Qsto.
varactor was a 9 mil silver disc, which was formed
Equation (31) can be normalised by dividing by by vacuum evaporating the silver through a metal
z/N. Thus, the equation becomes relatively in- mask. Prior to the oxide deposition operation the
dependent of material and concentration in the wafers were cleaned in a mixture of H202, HzS04

n-type material .+_

FIG. 12. General space charge curves.

enhancement and depletion regions, i.e. and H20. Two methods of oxide deposition were
employed. On silicon the oxide was thermally
VIIl,&JX C,, + qK grown in a dry oxygen atmosphere at a temperature
--- z/N *so = +;. (32)
dN + of 1000°C. The other method, used on gallium
arsenide, consisted of pyrolytically depositing
Figure 12 is a plot of the right side of the equation, the oxide on the wafers at a temperature of 730°C.
as a function oft/,,. Since V,,, and qK are known, After metalisation the wafers were pelletised and
equation (32) may be solved by plotting its left individual units were mounted on a gold plated
side on Fig. 12 and by finding the intercept that header. Gold wires were then thermocompression-
determines the values of #,, and Qsto. With this bonded to the silver, completing the devices.
information and equation (29), +s can now be Capacity measurements were made on the
found. Thus varactors at a test frequency of 1 MC, using a 1 mV
signal. The capacity was measured at various d.c.
bias voltages. The measured points were normal-
ized with respect to the oxide capacity and plotted.
The above procedure to analyze a measured curve Then, following the procedure outlined in
is summarised in Appendix 1. Appendix 1, the values of K, f(wg), phso, and &
2
222 R. HALL and J. P. WHITE

were calculated and theoretical curves were plotted surface. The total density of surface states for the
using these parameters. n-type material is close to that for silicon but the
Normalized capacity curves of silicon surface response of the states is faster. The higher
varactors fabricated on 10 R-cm n-type and frequency response of states is also shown for the
7 Q-cm p-type material are shown in Figs. 13 and p-type curve which has a total density of states of
14. A table of the surface state parameters and an 1.4 x lOI cm-2 V-r.
equilibrium band diagram are also included. The
capacity of both the n and p-type surface varactors 7. DISCUSSION AND CONCLUSIONS
have well defined minima, and exhibit the expected In the model presented for the surface varactor,
characteristic except for the region where the a number of assumptions were made. The in-
space charge capacity is dependent on the response sulating oxide used to fabricate the device was
of minority carriers (inversion region). In this treated as a perfect dielectric, i.e. the permittivity
region the capacity is limited by the ability of the of the oxide is independent of the applied voltage
minority carrier density to follow the applied field and the d.c. resistance is infinite. The effect of
at the test frequency. The frequency response finite resistance would be represented by a parallel
due to thermal generation and recombination is resistance across the equivalent circuit shown in
very low (normally less than 50 c/s for silicon). Fig. 2. When the resistance is very high, its effect
However it has been recently shown(la) that it is on the capacity measurement may be neglected.
possible to couple to an external supply of minority There is a possibility that an ionic drift or an
carriers and observe a significant increase in ionised impurity level may cause the oxide to
frequency response. develop a space charge region under the effect of
The theoretical curves are seen to be a fairly an applied field. However, capacity measurements
good representation of the measured points for would detect this condition as a variation of
both the n- and p-type material except in the in- maximum measured capacity. A condition which
version region. This implies that the surface state would go undetected is a built-in ionised charge
capacity at the test frequency is relatively in- that is not influenced by an applied voltage. This
dependent of the surface voltage. The band dia- would cause the capacity curve to be biased along
grams show that both materials have a positive the applied voltage axis, but would not affect the
equilibrium surface voltage which increases the calculated value of state density. A similar effect
electron concentration and reduces the hole con- would be caused by a difference in contact potential
centration at the surface. In the g-type silicon a between the semiconductor and the metal elec-
density of surface states of 0.7 x 1012 crne2 V-1 was trode. However, this has been minimised by using
observed, most of which did not respond to the silver as the metal electrode which has a work
applied signal. On the n-type silicon the total function within a few tenths of a volt of silicon and
number of surface states was 50 per cent higher gallium arsenide. The possibility of non uniform
with half of this increase being accounted for by bulk impurity concentrations in the vicinity of the
‘fast’ states (i.e. those which responded to the surface, because of absorbed ions, is minimised
1 MC/~ signal). by suitable precleaning treatment. For the oxide
Normalised capacity curves of gallium arsenide thicknesses used, variations in concentration within
surface varactors fabricated on 0.04 Q-cm n-type about 100 x of the interface, may be included in the
and 0.2 !&cm p-type material are shown in Figs. effect of surface states.
15 and 16. The capacity of these varactors do not By assuming that the surface state distribution is
exhibit a minimum. This implies that there is no uniform in energy the mathematical analysis bc-
response of the minority carriers to the applied comes simplified to a point where quantitative in-
signal in the inversion region. Except for this formation concerning the surface state parameters
discrepancy the theoretical curves are seen to give can be readily obtained from a single capacit?
a good fit to the measured points. VS. voltage curve. Although the treatment is
The band diagrams for gallium arsenide show not exact, the parameters calculated have been
that the equilibrium surface voltage has such a consistent. The surface state densities observed
polarity to reduce the majority carriers at the for silicon are in good agreement with the results
SURFACE CAPACITY OF OXIDE COATED SEMICONDUCTORS 223

-I 0

/I.
/.-•

-0 9

-0 8

cn

-0 7

values
-0 6

I I

FIG. 13. Silicon (jz-type)-thermal oxide varactor characteristics.

C”
-*-*-a IO Parameter WIN3

1
A.

BEI
\
. P” 2x IO I5 cd

-0-9 C 2.27 x Id8 F- Crn~Z

\ cx

.
K 0.70 x IO” cm-’ V-l

Theoretical curve - fbd,) 0.07

\ O-6
$3, +0.21 V

+ 0.02 V
9,

1. . 07

*Represents experimental oe l * -----

values

‘\,
-05
i I ’ 1 I 1 I , ,
-10 -6 -6 -4 -2 0 2 4 6 8 IO

V C9P’ V

FIG. 14. Silicon (p-type)-thermal oxide varactor characteristics.

224 R. HALL and J. P. WHITE

1neorencai

1 _
curve--
rl.0

/
./
0-0

*Reoresents
values
exDerimentol 11
;I
-“‘g5
. . l
. l
II/
i
03.

I 1 , I 1 , ,
-25 -20 -15 -10 -5 5 IO 15

V OPP’ V

FIG. 15. Gallium arsenide (n-type)-silicon dioxide varactor characteristics.

FIG. 16. Gallium arsenide (P-type)-silicon dioxide varactor characteristics.

 SURFACE CAPACITY OF OXIDE COATED SEMICOKDUCTORS 225

obtained by other experimenters.@~sJs) Although the oxide-semiconductor interface. The procedure

there has been little published data on the density outlined facilitates the interpretation of this data
of surface states on gallium arsenide the results to extend its use to a wide variety of problems.
obtained are comparable to those obtained for Combined with the measurements of the character-
silicon. This low density of surface states on istics of semiconductor devices in which the oxide-
gallium arsenide was achieved only after suitable coated surface is an integral part, a clearer physical
surface treatments had been established and the understanding of the operation of these devices will
oxide deposition process optimised. These results be possible.
are supported by measurements on gallium
arsenide MOS transistors@) which have also Acknowledgements-This work was supported by the
indicated surface state densities of the order of Molecular Electronics Branch of the Avionics Labora-
tory, U.S. Air Force Systems Command, under Contract
1012 cm-2 V-1. It was found that for the curves
AF33(657)-11402. The authors would like to thank
analyzed, the values of I were between 0.07 Mr. D. BERGMAN and Mrs. D. JAC~KOWICZfor their
and 0.79, which lie within the theoretical limits. assistance in the fabrication and measurement of
These values, obtained at 1 MC/S, clearly demon- samples.
strate the significant difference in the frequency
REFERENCES
response of the surface states on silicon and
I. H. R. J. IHANTOLA, Design Theory of Surface
gallium arsenide. For silicon, only a small fraction
Field-Effect Transistor Stanford Res. Tech. Rept.,
of the states responded whereas for gallium 1661-1, Stanford University, August (1961).
arsenide a response time of lpsec seems typical. 2. S. R. HOFSTEINand F. P. HEIMAN, Pvoc. Inst. Radio
The equilibrium surface voltage (#s0) for silicon Engrs. 51,119O (1963).
was found to be positive, indicating an increase in 3. P. K. WEIMER, Proc. Inst. Radio Engrs. 50, 1462
(1962).
electron density at the surface. This effect, if
4. R. R. HAERING, Solid-State Electron. 7, 31 (1964).
exaggerated, could account for the built-in in- 5. H. BECKE, R. HALL and J. WHITE, Gov. I.T.D.R.
version region observed in silicon MOS transistors. 3 Contract AF33(657)-11402, February (1964).
For gallium arsenide I/I,,, was found to be de- 6. W. G. PFA~N and C. G. B. GARRETT, Proc. Inst.
pendent on the impurity type of the bulk con- Radio Engrs. 47,201l (1959).
7. C. G. B. GARRETT and W. H. BRATTAIN, Phys.
centration. Although for both gallium arsenide
Rev. 99, 376 (1955).
and silicon the energy bands are significantly 8. R. LINDER, Bell Tech. J. 41, 803 (1962).
distorted at the surface under equilibrium, in no 9. M. TERMAN, Solid-State Electron. 5, 285 (1962).
case is the surface of the semiconductor in the state IO. K. LEHOVEC and A. SLOBODSKOY, Solid-State
Electron. 7, 59 (1964).
of degeneracy.
11. E. L. JORDAN,J. Electrochem. Sot. 108,478 (1961).
The parameter ds which was introduced to 12. S. R. HOFSTEIN,K. H. ZAININGER,and G. WARFIELD
account for the magnitude and sign of the charges Proc. Inst. Elect. Electron Engrs., S.S.R.C. July
in the surface states under no applied field did not, (1964).
in general, fall within the forbidden gap. This may 13. J. F. DELORD, T. HUTCHINS and G. STRINGER,
Electron Device Meeting, Washington, D.C.,
be simply interpreted as surface states lying outside
(1963).
the forbidden region, but could also be caused by
other effects which would not be distinguished APPENDIX I
using this analysis. It is possible that a high density Procedure for analyzing MOS varactors
of very slow surface states exists and strongly (a) Plot a graph of the normalised capacity, C,, as a
influences the equilibrium surface voltage. The function of applied d.c. voltage, Vapp.
(b) Locate or estimate, for gallium arsenide, the point of
effect of these states, would be to cause the capacity
minimum normalised capacity, Cnfmin), and note the
readings to drift slowly with time. In fact some drift corresponding voltage, Vmin.
was observed but appeared to come to equilibrium (c) From Fig. 7, read off the minimum space charge
within a few seconds. However, states with response capacity (C,,,) for the bulk concentration of the semi-
times of greater than about a minute are difficult conductor being used.
(d) Calculate
to detect and were neglected.
Measurement of the capacity of the surface
varactor is a convenient method for investigating
-Gn.
226 R. HALL and J. P. WHITE

(e) From Fig. 7, read off the capacity for level bands (j) Calculatef(wo) using the results of steps (d) and (i).
(C,,) for the corresponding bulk concentration. (k) Plot a straight line graph of
(f) Calculate

(g) Locate the applied voltage (I’,,,) on the normalised

capacity curve corresponding to Cn(msx). as a function of I/I~on Fig. 12. It intersects the approxi.
(h) Read off Qatm from Fig. 4, and #sm from Fig. 5. mate curve at $SO.
(i) Calculate (1) Calculate