8 IEEE TRA.P:SACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO.

1, JANUARY 1986
A Generalized Theory of an Electrolyte-Insulator-
Semiconductor FielLd-Effect Transistor
Absfmct-A model of surface ionization and complexation of surface
hydroxyl groups on the gate insulator surface is adapted in conjunction
with electronic device physics to arrive at a generalized theory for the
current-voltage characteristics of an electrolyte-insulator-semiconduc-
tor field-effect transistor (EISFET) in electrolyte solutions. EISFETs
that employ thermally grown silicon dioxide were tested in simple elec-
trolytes that contain Na+, K+, and Li+ ions titrated in apH range from
2 to 9. Experimental results show good agreement with the theory. The
model successfully explains pH sensitivity, as well as the ion interfer-
ence effect, of the EISFET working as a pH sensor. From this model,
it is conluded that, among all the electrolyte parameters associated with
an EISFET, the surface site density of the hydroxyl groups N, and the
separation of surface ionization constants ApK are the primary factors
to consider when employing EISFETs as pH sensors. For high sensitiv-
ity and good selectivity, large N, and small ApK values are required.
I. INTRODUCTION
I
N PAST YEARS, much work has been done to dem-
onstrate the feasibility of using ion-sensitive field-effect
transistors (ISFETs) [I], [2] for measuring pH and other
ions in the electrolyte. For the category of ISFETs that
use an ion-selective membrane on top of the gate insulator
film, the device response can be satisfactorily explained
by the well-established theory of ion-selective electrodes,
whereas the principle of ISFETs that employ only an in-
organic insulator film without an ion-selective membrane
is not well understood.
When Bergveld first introduced the ISFET [3], [4] the
device was operated without a reference electrode. How-
ever, later work by other investigators [5], [6] indicates
that proper operation of the ISFET requires a reference
electrode to establish the electrolyte potential with respect
to the semiconductor substrate. This type of ISFET, which
is operated with a reference electrode and consists of a
gate structure of electrolyte-insulator-semiconductor
(EIS) interfaces, will be referred to as the electrolyte-in-
sulator-semiconductor field-effect transistor (EISFET) [7].
Early attempts to model the response of EISFETs were
based upon an empirical approach that combines the glass-
electrode potential with the device physics of a metal-ox-
ide-semiconductor field-effect transistor (MOSFET) 1181,
Manuscript received February 14, 1985; revised July 18, 1985. This work
was supported by the NIH under Grant RR 02024.
C . D. Fung and W. H. KO are with the Electronics Design Center and
the Department of Electrical Engineering and Applied Physics, Case West-
ern Reserve University, Cleveland, OH 44106.
P. W. Cheung is with the Electrical Engineering Department, University
of Washington, Seattle, WA 98195.
IEEE Log Number 8405705.
[9] and were not able to account for many of the experi-
mental observations on the output sensitivity to hydrogen
ions and the output selectivity for hydrogen ions over the
other ions in the solution. In particular, it failed to predict
an often observed non-Nernstian response with pH. Dif-
ferent mechanisms were suggested to identify the origin
of the EISFET response in solutions. The influence on the
interface state density at the insulator-semiconductor in-
terface was once speculated to be responsible for the op-
eration of EISFETs [lo]. However, at a later time, ex-
perimental results indicated that it was unlikely for such a
phenomenon to cause the response of an EISFET [ll],
[12]. The behavior of EISFETs is primarily attributed to
an interfacial potential developed at the insulator-electro-
lyte interface [12], [13]. An experiment was designed to
study the effect on EISFET response by modifying the
EISFET at the surface of its gate insulator during the fab-
rication process [13], [ 141. In the experiment, the com-
posite gate insulator layer consisting of silicon nitride on
silicon dioxide was oxidized using wet oxygen at 1050C
to convert the silicon nitr!de surface into a silicon oxide
layer approximately 150 A in thickness. The response of
the surface-oxidized nitride-gate EISFET degraded to that
of a silicon-dioxide-gate EISFET, with lower pH-sensitiv-
ity and decreased selectivity. When the oxidized surface
layer was removed, the response of the device recovered
to that of the nitride-gate EISFET. This experiment estab-
lished the surface-dominant behavior of the device. Siu et
al. [15] used the site-binding model by Yates et aE. .[ 161 to
describe the ionic absorption processes at the electrolyte-
SiOz interface and to examine the flat-band voltage of the
electrolyte-Si02-Si system in response to pH. A similar
approach was taken by Fung et al. [7], [14] and Bousse et
al. [17] to study the characteristics of EISFETs and the
operation of chemically sensitive field-effect sensors, re-
spectively. From this point of view, the EISFET can be
regarded as a transistor that measures the phenomena oc-
curring at the insulator-electrolyte interface through a
field effect.
Based on the concept originally presented in [7], this
paper describes a full theoretical development of a gener-
alized model with a detailed discussion of experimental
and theoretical results. A model of surface ionization and
complexation by Yates et al. [16] and Davis et al. [18] is
merged with the physics of a MOSFET to arrive at this
generalized theory for the EISFET. The theory is general
in the sense that 1) it can explain quantitatively the EIS-
0018-9383/86/0100-0~B08$01.00 O 1986 IEEE
FUNG et al . : ELECTROLYTE-INSULATOR-SEMICONDUCTOR TRANSISTOR 9
FET response to hydrogen ions, as well as other ions pres-
ent in electrolyte solutions; 2) the derivation of the EIS-
FET current-voltage characteristics relies on the solution
of a general set of equations that describe the EIS system.
(These equations can be readily modified or extended for
a different or more sophisticated system.) Based on this
theory, a detailed discussion of the EISFET's electro-
chemical parameters with regard to the device's perfor-
mance as a pH sensor is then presented.
11. THEORETICAL MODEL OF DEVICE CHARACTERISTICS
In this section the surface interaction at the electrolyte-
insulator interface will be described first, followed by a
study of the charge and potential distributions across an
ideal EIS structure. The drain current-voltage character-
istics of an EISFET is then developed based on conven-
tional MOSFET theory.
A. Surface Interaction and Ideal EIS Structure
At a blocked insulator-electrolyte interface involved in
an EISFET, ions are distributed only between the solution
and the surface of the solid with neither electron nor ion
exchange across the boundary of the two phases. The elec-
trical double layer of oxide suspensions has been a subject
extensively studied by colloid and surface scientists. In
explaining interactions of electrolyte ions with oxide sur-
faces, some researchers have emphasized the structural
and the physical aspect of the distribution of solutes. Oth-
ers have stressed the specific chemical interactions of sol-
utes with oxide surfaces and solution. However, no com-
prehensive model existed for simultaneously estimating
adsorption density, surface charge, stoichiometry of sur-
face reactions, and diffuse layer potentials at the oxide-
electrolyte interface until the work of Yates et al. [ 161and
Davies et al. [18] appeared. Their work resulted in major
improvements in the double-layer model by including the
formation of surface complexes in addition to simple ion-
ization of surface hydroxyl groups by interactions with hy-
drogen and hydroxyl ions. The model is reasonably com-
plete in that both the physical interaction in controlling the
distribution of ions and the chemical interaction of the ions
with the oxide simultaneously considered. Their model can
be readily generalized to account for other blocked insu-
lator-electrolyte interfaces with similar surface sites.
For insulators containing hydroxyl groups, using silicon
dioxide as an example, the chemical reaction at the inter-
face can be described as follows [ 181:
K+
SOH,' e SiOH +H l
SiOH e SiO- +H l
The amphoteric dissociation constants for the above equi-
libria are given by
K-
[SiOH] [H+],
[SiOH:]
K+ =
[SiO-] [H+],
[SiOH]
K- =
Since H+ and OH- are the predominant ions in deter-
mining the charges and, in turn, the potentials at the elec-
trical double layer, they are referred to as the potential-
determining ions for this 'interface. The other ions in the
electrolyte are referred to as supporting-electrolyte ions
or, simply, electrolyte ions.
In addition to the amphoteric dissociation, the support-
ing electrolyte can also form ion pairs with the charged
surface sites. This process will be referred to as surface
complexation and is illustrated below using NaCl as a typ-
ical univalent supporting electrolyte.
r-
SiO- - Na+e SiO- +Nal
SiOHl - C1- * SiOHl +Cl,.
K+
The equilibrium constants can again be written as
[SiO-] [Na'],
[SiO- - Na+]
[SiOH:] [Cl-1,
KL =
K' =
+ [SiOH: - Cl-1'
While the surface complexation depends on the existing
SiO- and SiOHz+sites, the formation of surface com-
plexes also readjusts the acid-base equilibrium and affects
the surface charge.
The charge density and electrostatic potential at differ-
ent locations of the electrical double layer at the interface
are shown in Fig. 1, where the outer Helmholtz plane
(OHP) is the locus of the centers of the hyrated ions with
the closest approach to the solid, and the inner Helmholtz
plane (IHP) is the plane of the adsorbed ions which form
pairs with the charged surface sites [ 181. According to the
Boltzmann distribution, the concentration of an ion spe-
cies X at a location i in the electrical double layer [XIi as
related to the bulk concentration [X] is given by
[XIi =[X] exp [ -q$i/kT].
Therefore, (1) may be rewritten as
[SiOH] [Hf]
[SiOH:]
K+ =
exp ( -&om ( 5)
where [Hf] is the bulk concentration of Hf and [H+], =
[H+] exp (-qrl/&T) is used.
Likewise, (2) is rewritten as
Since [Na+], =[Na'] exp ( - q$B/ kT) and [Cl-1, =
[Cl-] exp (qrl/@T) from (3) and (4) and by replacing
[Na+] and [Cl-] with the electrolyte bulk concentration C
IEEE TEANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 1, JANUARY 1986
Electrolyte Insulator
Qd 6 Qo
Layer I I I I
Di f f u s e -1
(b)
Fig. 1. Schematic diagram showing (a) the charge distribution and (b) po -
tential profile at the insulator-electrolyte interface with surface charge11
sites.
The fractions of the surface sites are introduced by defin-
ing
0; =[SiOH: - C1-1 nau/ANs
81 =[SiO- - Na+] nau/ANs
8, =[SiOH] nav/ANs
8; =[SiOH:] nau/ANs
e? =[S~O-I n,,/ms
where nu, is Avogadros number (number per mole), A is
the surface area available in solution (square centimetels
per liter) [ 181, [SiOH] and similar bracketed surface spe-
cies are the equivalent concentration of surface species in
moles per liter, and N, is the total number of available sites.
Therefore
8, +e? +e; +e: +er, =I . ( 5 1
With the fractions defined above, from ( 5) and (6)
-2.303 ApH =q$,/kT +1/2 In (8$/0!.) (1Cj
where ApH =pH - pH,,, and pHpzc is defined by pH,,,
=- 1/2 log ( K+K- ) . One can see from (1) and ( 2) that
pH,,, corresponds to the pH at the point of zero charge on
the surface, i.e., [SOH+] =[SiO-1.
Also from ( 5) and (6)
K-/K+ =e:e%/e;. (1 1)
- exp (- q$p/kT) =0
e!.
(12)
K e-
- exp (q$@/kT) =-.
er,
Kr, 0:
Up to this point, equatians have been established for the
compact layer of the electrified interface. The remainder
of the electrified interface, the diffuse layer which extends
from the OHP to the bulk of the electrolyte, can be de-
scribed by the Gouy-Chapman theory 1161, 1181, which is
based on the Poisson and Boltzmann equations. The dif-
fuse layer charge Qd in a column of unit cross-sectional
area extending from the boundary of the diffuse layer to
the bulk of the electrolyte is related to the mean potential
$d at the OHP as
Equations (7) and (8) can be rewritten as
C
C
(13)
Qd =-(8EEOnkT)2 Sinh (q$&!kT) ( 14)
where E is the dielectric constant of the solution, E , is the
permittivity of free space, and n is the bulk number den-
sity of the electrolyte ions.
By knowing the double-layer structure at the insulator-
electrolyte interface, it is possible to consider the EIS sys-
tem as a whole. First, the theory regarding the charge and
potential distributions in an ideal EIS structure will be
formulated with the following assumptions:
1) In the electrolyte the ionic concentration is suffi-
ciently low that the Boltzmann distribution law for ions
applies. Therefore, the diffuse part of the electrical double
layer at the insulator-electrolyte interface can be de-
scribed by the Gouy-Chapman theory.
2) At the surface of the insulator, hydroxyl groups exist
which can be ionized by the hydrogen ions or hydroxyl
ions to produce charged sites, which subsequently form
complexation with the supporting electrolyte ions.
3 ) The bulk of the insulator is perfect in the sense that
there is no transport of charge carriers or uncharged spe-
cies.
4) There are neither charges within the insulator nor
surface state charges at the semiconductor-insulator in-
terface. Interface states at the semiconductor-insulator
interface are also neglected.
5) The semiconductor is nondegenerate, and therefore
the Boltzmann distribution for the charge carriers is ap-
plicable.
Several commonly used insulator films, such as silicon
dioxide, aluminum oxide, silicon oxynitride, and silicate
glass, contain Si-0-Si or A1-O-A1 structures which pro-
duce surface hydroxyl groups upon hydrolyzing in the
electrolyte solution. Assumption 2 may be extended to ac-
FUNG et al.: ELECTROLYTE-INSULATOR-SEMICONDUCTOR TRANSISTOR 11
Electrolyte I Insulator Ip-Semiconductor
I I t i
Qs
I
I
O'HP IHP
(a)
" d 'B ' 0
V
(b)
Fig. 2. (a) Charge and (b) potential distributions in an ideal electrolyte-
insulator-semiconductor structure.
count for other interactions that might occur at a specific
electrolyte-insulator interface.
The justification for assumption 3 relies critically on the
properties of the insulator films. Because of variations in
the fabrication processes, it is quite possible that imper-
fect insulator films will result; therefore, the transport of
charge carriers or uncharged species may occur during
immersion in aqueous solutions. This situation will be dis-
cussed later.
In assumption 4, for simplicity, charges within the in-
sulator or at the semiconductor-insulator interface are not
considered. In practical situations, these charges exist and
are process-dependent. As in standard MOS theory [ 191,
these charges can be taken into account by a flat-band
voltage. This effect will be included in the derivation of
the theoretical model of the EISFET in a later section.
Fig. 2 shows the charge and potential distribution of an
ideal EIS structure with a p-type semiconductor substrate.
In Fig. 2, the charge at the surface of the insulator Qi is
shown to be negative, which is the case for pH >pH,,,.
The p-type semiconductor is shown to have an inverslon
layer at the surface. For convenience in later derivation of
the EISFET theory, the potentials in Fig. 2 are designated
by two sets of notations. The potentials &, $p, $o, and $s
are the electrostatic potentials at the OHP, the IHP, the
insulator surface, and the semiconductor surface, respec-
tively, in reference to the electrolyte bulk potential; like-
wise, the potentials denoted by 4 ' s are the corresponding
potentials with respect to the semiconductor bulk poten-
tial.
The potential difference +G between the bulk of the
electrolyte and that of the semiconductor is established
through a reference electrode immersed in the electrolyte
solution. Since the liquid junction potential is minimized
in a reference electrode, the overall potential of the ref-
erence electrode can be treated as a constant. Therefore,
if a gate voltage V, is applied at the reference electrode
with respect to the semiconductor substrate, the potential
4, will follow V,, but deviate from it by a constant de-
termined by the particular reference electrode in use.
Equations (9) through (13), which determine the surface
interaction at an insulator-electrolyte interface, and (14),
which describes the diffuse layer of the double layer, can
be adapted for the ideal EIS system. The relation between
the semiconductor charge Q, and semiconductor surface
potential 4s is expressed [20] by
+- [exp(u,) - us -
n0
Po
11] 1'2
where us =q$,/kT, no and po are the equilibrium concen-
tration of electrons and holes, respectively, and Lo is the
extrinsic Debye length for holes defined by
Lo =(2kT~,~0/'042) 1'2.
In (15), the charge within the semiconductor Qs, is taken
as positive for u, <0, and negative for us >0.
The charge neutrality of the system requires that
Qs +Qo +Qp +Qd =0 (16)
and the charges and potentials at the interfaces are related
by the following equations.
$0 - $p =(Qo +QJC1 (17)
$0 - $d =- Qd / c 2 (18)
$#- $0 =QJ Co (19)
where C, and C2 are the unit-area capacitances of the in-
ner and outer parts of the compact layer, respectively, and
Co is the unit-area capacitance of the insulator film.
The system of equations that describe an ideal EIS sys-
tem can be summarized as follows. Equations (9) through
(14) represent the reaction at the electrolyte-insulator in-
terface and the charge potential relationship in the elec-
trolyte diffuse layer, whereas equation (15) describes the
charge potential relation on the semiconductor surface.
The requirement of charge neutrality is expressed by (16).
The link between the different regions of the system is
contained in (17) through (19). In an EIS system, the
semiconductor surface charge Q, plays a role in determin-
ing the interfacial properties, as indicated by (17) and (19).
On the other hand, measurement of Q, at the semiconduc-
tor surface provides a means of obtaining the surface po-
tential Go at the electrolyte-insulator interface.
In a later derivation for the current-voltage character-
istics of an EISFET, a strong inversion approximation [ 191
at the semiconductor surface will be employed. This ap-
proximation assumes that, once the semiconductor surface
is strongly inverted, the surface potential will remain con-
stant at a value approximated by 4s =24F. where 4F is
the Fermi potential of the semiconductor. The approxi-
12
IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-32, NO. 1, JANUARY 1986
Electrolyte
Solution
I
I
Reference
Electrode
I
q50 will be taken as a constant value along the direction of
the channel and is independent of the location y .
The potential +o is related to the semiconductor surface
charge and potential by
where C$o(FB) is the flat-band potential defined as the value
of 4o to achieve the flat-band condition within the semi-
conductor, i.e., +o(FB) =Qss/Co, where Qss is the effective
surface state charge. This flat-band potential represents
the effect of charges due to surface states at the semicon-
ductor-insulator interface, as well as the charges within
the insulator. It is assumed that the flat-band potential
&(FB) is a parameter that depends on the fabrication pro-
cess, but is not influenced by the surface field at the semi-
conductor nor by the interaction occurring at the insula-
tor-electrolyte interface.
Following the standard deviation for the drain current-
voltage characteristics of MOSFET [19], with the gate
voltage replaced by the insulator surface potential 40, one
can obtain the drain current-voltage characteristics of an
EISFET in the nonsaturation region as
1 I 1
I
L
(b)
Fig. 3. (a) Cross-sectional view and (b) potential distribution of an EIS-
FET.
mation of strong inversion, when it is written in terms of
t+bs, can then be expressed as q5G s -$s +2 4 F .
E. Drain Current-Voltage Characteristics of an ElSFET
IR this section, first the current-voltage characteristics
of the EISFET in an electrolyte solution are derived; then
the EISFET response to electrolyte solutions under the
condition of a constant drain current and drain voltage is
formulated.
Consider the n-channel EISFET with the source con-
nected to the substrate at a ground potential as shown in
Fig. 3. Because of the close association (1 to 1.5 A) of the
adsorbed ions in the IHP with the charged sites on the
insulator surface [ 181, the lateral field in the y direction
due to the applied drain voltage is negligible compared to
the normal field in the x direction at the inner zone of the
double layer. The estimated value of the latter is three or-
ders of magnitude higher than the value of the former.
Therefore, it is reasonable to assume that the lateral field,
due to the drain-to-source voltage, will not affect the in-,
sulator surface potential 40. This has also been substan-.
tiated by an experimental investigation [21] which shows
no detectable effect of the lateral fringe field on the re-
sponse of the EISFET. Therefore, the insulator potential
where L is the channel length, Wis the channel width, and
p; is the effective electron mobility. In the saturation re-
gion, as the conduction channel is pinched off, the drain
current can be approximated by a constant value which is
the drain current at the pinchoff voltage.
For practical purposes the characteristics can be ap-
proximated by (21) and (22).
Nonsaturption region:
10 =P [ ( ~ O - 4 T ) v D s - (v,.J ~/~I (21)
Saturation region:
P
I - - (40 - 4d2
O - 2
(22)
where 4, =(Qss/Co) +2 4 F - @,/eo); P is a constant
and Q, is the charge in the depletion region of the semi-
conductor.
It should be reemphasized that the above derivation of
the drain current-voltage characteristics is based on the
surface potential (bo at the insulator. The threshold voltage
4T is a quantity determined by the semiconductor sub-
strate, the insulator capacitance, and the fabrication pro-
cess which gives rise to the c + ~ ( ~ ~ ) term. Since the thresh-
old voltage of a MOSFET is
FUNG et al.: ELECTROLYTE-INSULATOR-SEMICONDUCTOR TRANSISTOR 13
therefore
VT =4 T +4 M S
where is the work function difference between the
metal and the semiconductor. It is possible to estimate the
value of 4T by measuring the threshold voltage of a mon-
itor MOSFET on the same chip. While by definition &- is
not associated with the behavior at the electrolyte-insu-
lator interface, the EISFET does respond to the electro-
lyte solution through the term 40.
In the operation of an EISFET, a common method is to
bias the device at a constant drain current and drain volt-
age. This is achieved by using a feedback circuit which
varies the gate voltage to restore the drain current as the
electrolyte parameters are changed. Therefore, the gate
voltage measured in such a feedback circuit provides in-
formation about the electrolyte solution.
Working in either the nonsaturation or the saturation re-
gion when the EISFET is biased at a constant drain cur-
rent and voltage, the surface potential & of the gate in-
sulator is fixed according to the drain current-voltage
characteristics. Since the electrolyte bulk potential 4, only
deviates from the gate voltage V, by a constant value de-
termined by the particular type of reference electrode
used, the problem of finding the response of the EISFET
in terms of the gate voltage-becomes a situation of solving
the EIS system to obtain +G as a function of the electrolyte
condition.
At first glance, solving +G to find the response of the
EISFET in electrolyte solutions appears to be a two-di-
mensional problem. A closer examination of the EIS in-
terfaces in Fig. 3 reveals that, although the surface poten-
tial c j ~ ~ of the semiconductor varies along the channel and
the substrate caused by the voltage drop due to the drain
current, the potential in the electrolyte is only a function
of the x axis, with 4 =c # ~ ~ at the insulator surface and 6
=4G in the electrolyte bulk. Therefore, to solve for &,
one can choose any plane of a constant y coordinate be-
tween the source ( y =0) and the drain ( y =15). Never-
theless, it is advantageous to select the plane at y =0
because the semiconductor surface potential 4s at that
point is maintained at its equilibrium value of 2cPF due to
the absence of reverse bias between the channel and the
substrate.
Therefore, to find the response of EISFET's in electro-
lyte solutions operated at a constant drain current and
drain voltage, the procedure is first to calculate the poten-
tial 40, determined by the drain current-voltage charac-
teristics, (21) or (22). The set of simultaneous equations
for the EIS system using the strong inversion approxima-
tion is then solved for +G, assuming a constant 4o deter-
mined from the drain current-voltage characteristics. This
approach has been taken to arrive at theoretical calcula-
tions which are then compared with experimental results
described in the next section.
111. EXPERIMENTAL METHODS AND RESULTS
The structure and geometry of the EISFET's used in
the experiments have been reported elsewhere [71. The
G
"6
- ?
Devi ce
under
t es t
Fig. 4. Schematic diagram of the measurement circuit.
EISFET's were n-channel depletion-mode FET's built on
p-type (100) silicon substrates with a resistivity of 10 Q
cm. A thermally grown layer of 880-A silicon dioxide
served as the gate insulator. The channel length is 20 ,urn
with a large channel width-to-length ratio of approxi-
mately 90: 1 achieved by an interdigitated configuration.
The diffused drain and source regions are extended to one
end of the chip to facilitate lead attachment and packag-
ing.
The EISFET's, which used a calomel reference elec-
trode as the gate, were connected to a measurement cir-
cuit with a digital voltmeter to provide output readings.
The experiments were carried out at a room temperature
of 23 "C. During titration measurement, the pH value was
monitored by a pH meter (Orion 901). The schematic dia-
gram of the measurement circuit is shown in Fig. 4. The
circuit allows the drain voltage and drain current to be set
individually by R1 and R 2 , respectively. The operational
amplifier A2 provides a feedback so that the gate voltage
VG is varied during the titration measurement. In this way
the drain current of the EISFET is maintained at a con-
stant value set by the variable resistor R2. The gate voltage
V, serves as the output in response to electrolyte solutions
being tested. The drain voltage was set at 3 V and the
drain current at 100 pA to 1 mA. As far as the relative
output voltage is concerned, no difference was observed
within experimental error for different drain currents.
Simple 1 : 1 valent electrolyte solutions, including KC1,
NaCl, and LiC1, were used. These solutions did not con-
tain any buffer substances and were titrated with the ap-
propriate hydroxides (e.g., NaOH for NaCl solutions) and
hydrochloric acid solutions.
The devices were tested in a standard pH 7 buffer so-
lution, both before and after titration experiments, to pro-
vide calibration points for the devices. All the devices re-
sponded to titrations within several seconds. However,
they exhibited a long-term drift of 5 to 10 mV/h and a
hysteresis on the order of 10 mV.
Typical experimental results for the silicon dioxide EIS-
FET's in NaCl solutions are shown in Fig. 5 indicated by
data points. For low concentrations M), titration was
only carried out toward the acidic side to avoid adding
NaOH, which could alter the cation ion concentration of
the solutions at high pH value. In this low concentration
14 IEEE TRAhSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 1, JANUARY 1986
400
3
ApK = 8.2
200
10% NaCl
A lo- NaCl
100
0
1
1 I 1 I I I
2 13 4 5 6 7 8 9
I
%zc pH
Fig. 5. Experimental result of silicon dioxide EISFET by titration.
solution, addition of hydrochloric acid can alter the anion
concentration; however, as long as the device is in the
range of pH >pH,,,, the anions will not affect the de-
vices response. The response for the M concentra-
tion is rather linear and much steeper than that for the lo-
M. The latter is flattened in the lower pH range, where
there is a predominance of the electrolyte ions on the dif-
fuse layer, but it turns steeper in the higher pH range due
to the enhancement of the surface ionization by the elec-
trolyte ions in forming complexes with the charged sites.
In Fig. 5, the solid lines indicate the results of theoret-
ical calculation. The constants N, , C, , C, , K:, and KL
were chosen based on the work of Davis et al. [ 181, Abend-
roth [22], and Dugger et al. [ 2 3 ] . The ApK value was
adjusted to fit the experimental data. (ApK =pK- - pK, .)
It was found that ApK =8.2 provided a good fit of the
theoretical calculations and the experimental results. The
ApK value is in good agreement with the value of 8.4 given
by Davis et al. In order to match the theoretical curves
with the experimental data, the point of ApH =0 occurs
at 2.8 in the pH axis. This value of pH,,, =2.8 is con-
sistent with the published values [24] of 1.5 to 3.0 for sil-
icon dioxide.
A comparison of the effects between the lithium, so-
dium, and potassium ions is shown in Fig. 6. Data points
for an NaCl solution are reproduced from Fig. 5. The the-
oretical curves indicate a 6-mV difference at ApH =6
between KC1 and LiCl solutions. Since the reproducibility
of the silicon dioxide EISFET is limited to approximately
10 mV due to hysteresis and slow drift, it is not possible
to distinguish the difference between the solutions as
shown in the figure.
IV. DISCUSSION
It is apparent that good correlation between the exper-,
imental results and the theoretical calculations was ob-
400
300
200
0 KC1
A NdCl P
0 Li Cl
10- 5
L k l
I f I 1 1 I
2 3 4 5 6 7 a
PH
Fig. 6. Comparison of the effect between Li+, Kt, .and Na+ions on the
silicon dioxide EISFET.
served. It should be emphasized that the relative response
of the EISFETs is independent of the operation point,
provided the drain current is reasonably low so that the
variation of the semiconductor charge AQ, due to varied
drain current is negligible compared to the total insulator
surface charge AQo. For the operating drain current from
1 pA to 1 mA, the response is found to be essentially
independent of the operating current. Only minor differ-
ences on the order of a few millivolts appear at ApH =0
for very low electrolyte concentrations, e.g., lo- M. Un-
der constant current with feedback, variation of the de-
vices electrical parameters, such as @, &, and r pF, will
cause a variation of 40. In effect, these parameters will
vary the operating point but do not play a role as far as
the relative response of the EISFETs is concerned. How-
ever, the electrochemical parameters of the insulator-elec-
trolyte interface are significant in determining the perfor-
mance of the EISFET as a pH sensor.
The pertinent electrochemical parameters of the insu-
lator-electrolyte interfaces can be obtained from electro-
chemical measurement of the interfacial properties. Al-
though the data on this subject are rather limited, some
are available for certain oxide-electrolyte interfaces. The
main purpose of the following sections is to discuss the
effect of these electrochemical parameters on the behavior
of EISFETs in response to the electrolyte pH values and
ionic concentrations. Among these parameters, the ratio
of the surface ionization constants, K-/K+, is of particular
importance. To facilitate discussions of the response of
EISFETs, a constant ApK is defined as ApK =pK- -
pK, or ApK =-log(K_/K+). Typical values for the elec-
trochemical parameters are taken from the literature [ 161,
r 181.
FUNG et al.: ELECTROLYTE-INSULATOR-SEMICONDUCTOR TRANSISTOR
15
400. I
I I I I I
N~ = 5 x 1 0 ~ ~ ~ ~ - ~
K: = . 5
K; = .317
C1 = 140 pF/cm
10% - -
2 IO- -
300
- C2 = 20 pF/cmZ
-
APK= a A
-1 0 1 2 3 4 5 6
AhpH
Fig. 7. The effect of the inner layer capacitance C, on the response of an
EISFET.
A. The Compact Zone Capacitances
Both Yates et al. [16] and Davis et al. [ 181 agreed to
use 20 pF/cm2 for the outer layer capacitance C2 of the
compact zone for the electrical double layer. This is a rea-
sonable value to adopt from the classical value of the com-
pact layer capacitance of Hg and AgI 12.51, 126). There-
fore, the same value of C, is used in the present study.
However, there is some disagreement about the values of
the inner-layer capacitance C1 based on how close the ad-
sorbed ion species can reach the surface of the insulator.
Yates et al. [16] use a constant inner-layer capacitance Cl
of 140 pF/cm2 for all oxides and electrolytes, while Davis
et aE. [18] claim 140 pF/cm2 is an appropriate value for
lithium ions, but a smaller value, i.e., 100 to 125 pFlcm2,
should be used for potassium ions because they are larger
than the lithium ions. To a certain degree, through
the effect of complexation, the inner-layer capacitance C,
will affect the charges at the insulator; however, it was
found that the EISFETs response is very insensitive to
the variation of Cls in a wide range as shown in Fig. 7.
In this figure, C1 varies from 50 to 220 pF/cm2, and the
response for the EISFET is plotted for electrolyte concen-
trations of lop5 and lo- M. For electrolyte concentra-
tions of lop5 M, the curves for the relative EISFET gate
voltages at different C,s overlap each other due to the
lack of surface complexation for low electrolyte concen-
trations. For the higher electrolyte concentrations of 10-1
M, although minor differences do appear at higher ApH
because of surface complexation, only a variation of ap-
proximately 10mV is obtained at ApH =6 between C1
=50 and 220 pF/cm2. There is essentially no difference
between the curves as ApH approaches zero.
Fig. 8. The effect of surface complexation constant K- on the response of
an EISFET.
The above discussions, based on Fig. 7, are for a high
ApK value of 8. For lower ApK values, the EISFETs re-
sponse is expected to be even less sensitive to the variation
of els because of the less important role of surface com-
plexation at lower ApK values.
B. Surface Complexation Constants
Other than the inner.layer capacitance C1, the effect that
various supporting electrolyte ions have on each other is
mainly determined by the surface complexation constants,
i.e. , K1_ and K;. The K; value, which determines the ad-
sorption of anions to form complexes with positive sites,
is found to have essentially no effect on the devices re-
sponse in the region of ApH >0, possibly due to the lack
of positively charged sites.
On the other hand, the KL value, which determines the
adsorption of cations to form complexes with negatively
charged sites, has a certain effect on the response for high
electrolyte concentrations in the region of ApH >0. In
Fig. 8 it is shown that, with decreasing K values, the
response for lo- M moves toward the response for
M.
While the inner-layer capacitance C1 is related to the
distance between the adsorbed charge plan (IHP) and the
surface charge plane, the surface complexation constants
K> and KL indicate the tendency of the ions to form com-
plexes with the charged sites. One can recall from (9) and
(10) the smaller the K[+ and KL values, the stronger the
tendency for surface complexation. As shown by Fig. 8 in
the region of ApH >0, as KL decreases the flattened por-
tion of the lo- M EISFET response at low ApH tends to
become steeper and the M curve moves toward the
M curve. This effect may be attributed to a higher
16 IEEE TR LANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 1, J ANUARY 1986
w
(1
0
m
L)
7
0
>
B
Y
1)
(1
400 1 I 1 I
1
ApK = 8
K I = .5
K; = ,317
--
1 O-%
l o- ,y
I
300
-
200
-
100
-
c1 =
c2 =
140 pF/crn2
NS=5x10 I
20 yF/cm2 ,
- 1 /
O - 1 I 0 1 2 3 4 5 6
A pH
Fig. 9. The effect of the number of surface sites N, on the response of a1\
EISFET.
degree of surface complexation which enhances the sur-
face ionization.
The above discussions have been on the positive side ad
the ApH axis; a similar conclusion can be drawn for the
negative side of the ApH axis with the roles of K!. and
K: interchanged.
C. Surface Site Density
The total surface site density N, is also a factor of in,-
terest due to its important property of providing available
sites for surface interaction. As can be seen from Fig. 9,
a large N, results i n higher pH sensitivity and smaller in-
terference effect for the electrolyte ions over the hydrogen
ions in terms of a steeper curve and smaller separation
between the and the lo- M curves. However, unless
N, is extraordinarily high, in the normal range of 1014to
l OI 5 cm-2, the effectiveness of ionizing the sites also has
to be considered.
D. Surface Ionization Constants
The separation of the surface ionization constants ApK
is a parameter of paramount importance in determining
the response of EISFETs in terms of application as a pH
sensor. Toillustrate the general trends regarding the effect
of ApK on the response of EISFETs, Fig. 10 shows the
response of EISFETs as a function of ApH for different
ApK values at electrolyte concentrations of low5 and 10-1
M. As shown in Fig. 10, a lower ApK results in higher
pH sensitivity and smaller interference effect for the elec-
trolyte ions over the hydrogen ions.
For more quantitative comparisons, two quantities are
defined as shown in the insert in Fig. 11. The first quantity
is the slope rn, defined by the difference of the relative gate
-
-
>
>
W
0
>
- 1 0 1 2 3 4 5
6
APH
Fig. 10. Effect of the separation of the surface ionization constant ApK on
the reponse of an EISFET.
E
40 -

m
5x10
Total Site Density, N, (an-*)
0 )
Fig. 11. Effect of N, and ApK on the p H performance of an EISFET. (a)
Slope versus density, (b) deviation versus density.
FUNG et al.: ELECTROLYTE-INSULATOR-SEMICONDUCTOR TRANSISTOR 17
TABLE I
ELECTROCHEMICAL PARAMETERS OF Two OXIDE-ELECTROLYTE SYSTEMS
___
SYSTEM
POINT OF ZERO CHARGE SURFACE SITE DENSITY lONIZRTION CONSTANTS
PHPX NS(1014cm-T) PK, PK- QPK
__-____
A l 2O3/ NaC1 1271,[281 8.5 8 1291 7.68-7.9 9.1 1.2
Si0,jKCl 1181,1221 1 - 3 4. 6 - 5 [3Ol -1.2 7.2 8. 4
voltage between ApH =4 and 3 for the electrolyte con-
centration of lo- M. The slope m is expressed in milli-
volts per pH and represents the pH sensitivity of the de-
vice. The second, quantity is the deviation d, defined by
the difference of the relative gate voltage between electro-
lyte concentrations of and lo- M at ApH =3. The
deviation is expressed in millivolts and represents the in-
terference effect of the device as a pH sensor.
The slope m and the deviation d are plotted in Fig. ll(a)
and (b), respectively, as functions of N, with different ApK
values. Generally speaking, as in the case for most of the
oxides studied, N, varies over a relatively narrow range
from 5 X 1014to 1.2 X cm-2 [lS], whereas ApK can
vary in a large range from 1.2 to 8.4. Therefore, fur nor-
mal insulator surfaces, it is expected that the dominant
factor in determining the pH performance of the EISFET
is the ApK value. This is true unless certain surface mod-
ification methods are available that drastically increase the
number of surface sites enormously to enhance the pH
performance regardless of the ApK value.
In summary, the EISFET is a device that measures the
phenomena occurring at the insulator-electrolyte inter-
face. The pH sensitivity of the device arises from the ion-
ization by hydrogen ions (Hf) or hydroxyl ions (OH-) of
the surface hydroxyl groups. These surface ionization
processes are influenced by adsorbed electrolyte ions
present on the charged sites to form surface complexes.
The site density N, of the surface hydroxyl groups is a
factor of interest in the sense that it is the ultimate limit
for the number of charged sites. Higher N, values favor
the EISFETs acting as a pH sensor. However, the sepa-
ration of the surface ionization constants, expressed by a
ApK value, is the decisive factor in determining the pH
performance of the device. The response curve of the EIS-
FET with a large ApK value tends to be flattened at ApH
close to the zero, and the degree of flatness increases with
electrolyte concentration. It is evident that surface hy-
droxyl groups with small ApK values are preferable forpH
sensor applications.
V. CONCLUSIONS
A good correlation between the experimental results and
the theoretical calculations was observed for an EISFET
using silicon dioxide as the gate insulator. The generalized
theory serves very well to explain the primary character-
istics of an EISFET. For an EISFET to work as apH sen-
sor, the surface layer of the gate insulator, in the presence
of an electrolyte solution, should possess a low ApK value.
In addition, high surface site density, N, is preferable. Ta-
ble I summarizes the pertinent electrochemical parameters
of two oxide-electrolyte systems. It is apparent that alu-
minum oxide is an excellent candidate for the application
of the EISFET as a pH sensor due to the relatively large
N, and small ApK value. Indeed, good performance has
been observed in recent literature for EISFETs using alu-
minum oxide [31]. The pH at zero point of charge pH,,,
for the different systems is also included in Table I. It is
desirable to select insulator materials with a pHpzc value
far away from the pH range of application. However,
among the three parameters, ApK is the prime factor in
determining the performance of the EISFET when it is
used as a pH sensor.
Secondary effects, such as long-term drift and hyster-
esis, are not considered in this theory. These phenomena
can result from the slow progressive hydration of the in-
sulator film, as well as from the transport of certain spe-
cies throughout the insulator film, which affects the semi-
conductor-insulator interface. In a practical situation,
aluminum oxide film with its passivation capability to re-
sist hydration, as well as migration of ions or hydrogen
molecules, exhibits less hysteresis and long-term drift. On
the other hand, silicon dioxide film, comparatively, is
much more subject to hydration and ion migrations and,
therefore, results in.less reproducible device performance.
It is likely that the mechanism proposed by Lauks [32]
suggesting a surface ion-conducting membrane on the in-
sulator film, as well as the possible diffusion of species
through the insulator film, suggested by de Rooij et aE.
[lo], should be integrated with the present model to
achieve a theory which can completely describe the be-
havior of an EISFET.
REFERENCES
[l] P. W. Cheung, D. G. Fleming, W. H. KO, and M. R. Neuman, Eds.,
Theory, Design and Biomedical Application of Solid Stare Chemical
Sensors. Boca Raton, FL: CRC Press, 1978.
[2] J. J anata and R. J. Huber, Ion sensitive field effect transistors and
related devices, Ion Select. Electrode Rev., vol. 1, pp. 31-79, 1979.
[3] P. Bergveld, Development of an ion-sensitive solid-state device for
neurophysiological measurements, IEEE Trans. Biomed. Eng., vol.
[4] P. Bergveld, Development, operation, and application of the ion-sen-
sitive fidd-effect transistor as a tool for electrophysiology, IEEE
Trans. Biomed. Eng., vol. BME-19, pp. 342-351, 1972.
[51 S. D. Moss, J . J anata, and C. C. J ohnson, Potassiumion-sensitive
field-effect transistor, Anal. Chem., vol. 47, pp. 2238-2242, 1975.
[6] R. G. Kelly, Microelectronic approaches to solid state ion selective
electrode, Electrochimica Acta, vol. 22, pp. 1-8, 1977.
[71 C. D. Fung, P. W. Cheung, and W. N. KO, Electrolyte-insulator-
semiconductor field-effect transistor, in IEDM Tech. Dig., pp 689-
692. 1980.
BME-17, pp. 70-71, 1970.
18 EEE TR
ANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 1, J ANUARY 1986
[SI M. Esashi and T. Matsuo, Integrated micro multi ion sensor using
field effect of semiconductor, ZEEE Trans. Biomed. Eng., vol. BME-
[9] S. D. Moss, C. C. J ohnson, and J. J anata, Hydrogen, calcium, and
potassiumion-sensitive FET transducers: A preliminary report, IEEE
Trans. Biomed. Eng. , vol. BME-25, pp. 49-54, 1978.
[lo] N. F. deRooij and P. Bergveld, The influence of the pH on the sur-
face state density at the SiOn-Si interface, in The Physics tf SiO,
and Its Interfaces, S. T. Patelides, Ed. New York: Pergamon. 1978,
1111P. R. Barabash and R. S. C. Cobbold, Dependence of interface state
properties of electrolyte-Si0,-Si Structures on pH, IEEE Trans.
Electron Devices, vol. ED-29, pp. 102-108, 1982.
[12] A. Prolonge, P. Gentil, and G. Kamarinos, Study of the working of
ion sensitive FET, in IEDM Tech. Dig., pp. 147-150, 1979.
1131C. D. Fung, P.W. Cheung, A. S. Wong, J. A. Topich, and W. 11. KO,
Investigation of insulator materials on the response of ion-sellsitive
field-effect transistors (ISFETs), in Extelzded Abstracts Eb?ctro-
chem. SOC. Meeting, pp. 197-199, 1978.
[14] C. D. Fung, Characterization and theory of electrolyte-insu ator-
semiconductor field-effect transistor, Ph.D. dissertation, Case ?Vest-
ern Reserve University, 1980.
[15] W. M. Siu and R. S. C. Cobbold, Basic properties of the electrolyte-
SO,-Si system: Physical and theoretical aspects, IEEE Trans. ,?lee-
tron Devices, vol. ED-26, pp. 1805-1815, 1979.
[I61 D. E. Yates, S. Levine, and T. W. Healy, Site-binding model of the
electrical double layer at the oxide/water interface, J. Chem. SQC.
Faraday I , vol. 70, pp. 1807-1818, 1974.
1171L. Bousse, N. F. DeRooij, and P. Bergveld, Operation of chemic:ally
sensitive field-effect sensors as a function of the insulator-electrdyte
interface, IEEE Trans. Electron Devices, vol. ED-30, pp. 1263-1.!70,
1983.
1181 J. A. Davis, R. 0. J ames, and J. 0. Leckie, Surface ionization and
complexation at the oxide/water interface, I: Computation of electrical
double layer properties in simple electrolytes, J. Colloid Znteqace
Sci., vol. 63, pp. 480-499, 1978.
[19] S. M. Sze, Physics of Semiconductor Devices. New York: Wiey,
1969.
[20] R. H. Kingston and S. F. Neustadter, Calculation of the space cha.rge,
electric field and free carrier concentration at the surface of a se~ni-
conductor, J. Appl. Phys., vol. 26, pp. 718-720, 1955.
[21] W. H. KO, J. M. Lee, C. D. Fung, and I? W. Cheung, The frir ge
field effect of VDs on ISFET pH sensors, Sensors and Actuators, vA.
[22] R. P. Abendroth, Behavior of a pyrogenic silica in simple electlo-
lytes, J. Colloid Interface Sci., vol. 34, pp. 591-596, 1970.
[23] D. L. Dugger, J. H. Staton, E?. N. Irby, B. L. McConell, W. W. Cum-
mings, and R. W. Maatmen, The exchange of twenty metal ions with
the weakly acidic silanol group of silica gel, J. Phys. Chem., vcl.
1241S. M. Ahmed, Electrical double layer at metal oxide-solution inter-
faces, in Oxides and Oxide Films, vol. 1, J. W. Diggle, Ed. NeN
York: Marcel Dekker, 1972.
1251W. Stumm, C. P. Huang, and S. R. J enkins, Specific chemical i r-
teractions affecting the stability of dispersed systems, Croat. Chem.
[26] J. Lyklema and J. T. H. G. Overbeek, Electrochemistry of silver
iodide, the capacity of the double layer at the silver iodide-water in-
terface, J. Colloid Sei., vol. 16, p. 595, 1961.
[27] H. Sadek, A. K. Hielmy, V. M. Sabet, and Th. F. Tadros, Adsorp.
tion of potential determining ions at the aluminumoxide-aqueous in.
25, pp. 184-192, 1978.
pp. 433-437.
3, pp. 91-98, 1982/83.
68, pp. 757-760, 1964.
Acta, VOI. 42, pp. 223-224, 1970.
terface and the point of zero-charge, J. Electroanal. Chem:, vol. 27.
pp. 257-266, 1970.
[2S] C. P. Huang and W. Stumm, Specific adsorption of cations on hy-
drous y-Al,O,, J. Colloid Integace Sci., vol. 43, pp. 409-420, 1973.
1291I. B. Peri, Infrared and gravimetric study of the surface hydration
of Alumina, J. Phys. Chem., vol. 69, pp. 211-219, 1965.
[30] C. G. Armistead, A. J. Hambleton, S. A. Mitchell, and J . A. Hockey,
The surface hydroxylation of silica, J Phys. Chem., vol. 73, pp.
[31] H. Abe, M. Esashi, and T. Matsuo, ISFETs using inorganic gate
thin films, IEEE Trans. Electron Devices, vol. ED-26, pp. 1939-
1944, 1979.
1321I. R. Lauks, pH measurement using polarizable electrodes, IEEE
Trans. Electron Devices, vol. ED-26, pp. 1952-1959, 1979.
3947-3953, 1969.
*
Seattle. His research
and instrumentation f
itoring.
il
01
Peter W. Cheung (S83-M84) received the B.S.
degree in chemistry fromOregon State Univer-
sity, Corvallis, in 1968, the M.S. degree in chem-
istry fromthe University of Puget Sound, Tacoma,
WA, in 1969, and the Ph.D. degree in electrical
engineering fromthe University of Washington in
1973.
He was an Associate Professor at Case Western
Reserve University, Cleveland, OH. He is cur-
rently a Professor of Electrical Engineering and
Bioengineering at the University of Washington,
xterests are in the areas of biomedical transducers
physiological research and continuous patient mon-
*
Wen H. KO (S55-M59-SM75-F78) was born
in China. He received the B.S.E.E. degree from
the Chinese National University of Amoy in 1946
and the M.S. and Ph.D. degrees in electrical en-
gineering fromCase Institute of Technology, Case
Western Reserve University, Cleveland, OH, in
1956 and 1967, respectively.
He was an Assistant and Associate Professor at
Case Institute of Technology from1954 to 1967,
and has been a Professor in the Electrical Engi-
.neering Department and the Biomedical Engineer-
ing Department since 1967. Some of his areas of interest are solid-state
transducers, solid-state devices and technology, and biomedical instrumen-
tation, especially implant electronics.