358 IEEE TRANSACTIONS ON NUCLEAR SCIENCE June

THE FIELD-EFFECT TRANSISTOR-

ITS CHARACTERISTICS AND APPLICATIONS*
V. Radeka
Brookhaven National Laboratory
Upton, N. Y.

SUMMARY most important features in the context

of nuclear instrumentation. Figures
The characteristics of the junction of merit are derived in terms of the
gate field-effect transistor and the physical parameters and their limits
basic relations among its parameters are are discussed. Some applications are
described. Some figures of merit are given.
derived in terms of its physical
parameters and the limitations considered. THE PRINCIPLE AND CHARACTERISTICS
Applications in nuclear instrumentation
are discussed. The field-effect transistor is based
on voltage control of majority carrier
Some advantages of field-effect flow through a semiconductor channel.
transistors are in high input impedance, The effective width, and with it the
low noise at very low and medium conductance, of the channel is varied
frequencies, and a wide operating by the reverse potential of the gate-
temperature range (-200°C to +1000C). In channel junction. A planar field-effect
addition, the field-effect transistor transistor structure is illustrated in
represents a new circuit element - Fig. 1, and its dc-characteristics are
voltage (charge) controlled conductance shown, qualitatively, in Fig. 2. The
with fast response. gate is the control electrode. It is
usually assumed that the gate doping
Field-effect transistors are (charge density) is large compared with
particularly suitable for low noise the channel doping, so that the depletion
amplifiers at low and medium frequencies, region spreads mostly into the channel.
for high input impedance amplifiers The two channel ends are referred to as
(as for integrators), for charge- drain and source. The drain is that
sensitive amplifiers, and for analogue channel end which is at higher reverse
multipliers. They can also be applied
as controllable feedback elements.
P N DEPLETION CONDUCTING
INTRODUCTION (N) (P) REGION CHANNEL

The unipolar field-effect transistor

with junction gate was proposed by Xd DRAIN
Schockleyl in 1952. Recently there has
been considerable effort expended in
the development of this device which has Z TX W Vd Woo (Vdif + Vg)
-

resulted in a number of commercially L X Vd Woo - (Vdif + Vg)

available types. This development was
possible due to advances in semiconductor
technology, and it was stimulated by GATE
the need to complement the bipolar
transistor in applications where high
input impedance is required. Its unique
principle has already created some new
applications.
In this paper the characteristics
of the field-effect transistor and basic
relations among its parameters are
summarized in an attempt to present its o
* This work was done under the auspices Fig. 1-Schematic representation of a planar junction
of the U. S. Atomic Energy Commission. gate field-effect transistor.
1964

(a)
Vd <O
VdO
>

V >0

(c)

/
Vd +

CHFtNNEL

L - Lo 4,hL
(b)

Ud
, V30

Fig. 2-Characteristics of the field-effect transistor

a) depletion region and conducting channel for
low drain voltage
b) general current-voltage characteristics
c) current-voltage characteristics for low
drain voltage
DO
_3= 0
,, O
V9-0

-~vd
Woo

~g
RADEKA: THE FIELD-EFFECT TRANSISTOR

vacuum pentode or bipolar transistor,

Fig. 2b.

d
The current-voltage characteristics
depend on the carrier density profile
of the gate-channel junction. However,
there is a quite general relation among
the following parameters: the trans-
conductance, the output conductance,
and the channel conductance for zero
drain voltage. These can be expressed
in terms of charge (carrier) density
as follows,2
Transconductance:

gm

voltage:

Gdo
where:
,u

Q(x) =

p(x)
=
av

Output conductance:

=V
d

= carrier mobility

=
0
x
Q(x )

[Q(a)
Channel conductance for zero drain

[Q(a) - Q(x8)
- Q(x, )j

- Q(xd)J

fp(x)dx = the charge from 0

charge density
to x per unit length
in y direction
359

(l:)

(2)

(3)

(The parameters are per unit width of

the gate-channel junction.) These
voltage with respect to the gate. expressions indicate that the trans-
(There is no structural difference conductance is equal to the conductance
between drain and source.) of the rectangular section of the channel
bounded by x8 and xd,, irrespective of
There are two distinctly different the charge density profile. It also
modes of operation of the field-effect follows that the maximum transconductance,
transistor. First, for zero or small which is reached in saturation (pinch-
voltages across the channel, where the off) for xd = a, is equal to the sum
conductance of the channel is not of gm and gd in the nonsaturation
markedly changed by the current flow, region, and also to the channel conduct-
and second, in saturation (or pinch-off) ance for zero drain voltage,
where the channel conductance is affected
by the current flow so that the current gm + gd = gms i Gdo.
becomes virtually independent of the
drain voltage. In the first mode the For an abrupt gate-channel junction and
device can be considered as a passive uniform charge density the drain current
element variable conductance
- -
as a funition of the gate voltage is
controlled by the gate voltage, Fig. 2a given byl
and 2c. In the second, it appears as
a three-terminal active device with
characteristics similar to those of d 00 [d- g- (wd/-3/)
t5
360 IEEE TRANSACTIONS ON NUCLEAR SCIENCE June

where: Field-effect transistors made by

ppa2 diffusion processes may have a charge
00 0 (6) density distribution which decreases
with x as or as
internal satura- exp[- (X/X 0)2]
WO V
00 dif
-
tion or pinch-off
potential, i.e.
junction barrier
the error function of Then the x/x0.
potential re- transconductance and the channel
quired to de- conductance have approximately linear
crease the channel dependence on the gate voltage3
width to zero

normalized gate- gs =Gdo = G0 (l -

wg) , (8)
w = (Vdif+Vg ) -
d -w(00 dif g source (internal)
potential and consequently, the saturation drain
current has square law dependence
Wd =WL(Vdif+Vg+d) - normalized gate-
drain (internal)
potential I d= I do (l - w )2 . (9)
V00 - externally (The laws (8) and (9) represent exactly
measured satura- exp(-x/xo) charge distribution.)3
tion or pinch-
off potential The capacitance of the gate-channel
junction reduces, in saturation, to the
Vg - gate terminal gate-source capacitance, since the drain
voltage with voltage does not affect the depletion
respect to region width significantly. This
source capacitance as a function of the gate
voltage for uniform charge distribution
is given by5
Vd - drain terminal
voltage with
respect to source
l+wl/2
Vdif - diffusion poten- 0gs = 3 p0aL
tial gss = 2 v00 (1+2wg
K - dielectric
constant (l±w /2)
= KLa 3 v (10)
a - channel width ( 1+2w1/2
g )2
L - channel length
(The drain-gate capacitance is determined
All potentials are assumed positive when only by contact and lead capacitances.)
reverse biasing the sate-channel junc-
tion. Expression (5) was derived for
the nonsaturated region and applies to NOISE
saturation as a limit for wd = 1. The The basic noise source in field-
transconductance in saturation, and effect transistors is the thermal noise
therefore the channel conductance for of the conducting channel modified by the
Vd = 0, is effect of channel width modulation. It
can be represented as a current noise
gMs 9GdoG = Wtpa
L w1/2
(l1-w/)( (7 generator in4parallel with the conduct-
ing channel,

A practical maximum for the transconduct- i= 4KTAf

= * g H (w;} wd) (11)
ance is for V = 0 (wg should not be
much less tha§ Vdif
I as this would where g, is the transconductance in the
00
saturat58 part of the characteristics
for the given bias wg, and H1 (wg, wd)
forward bias the gate-channel junction). is a function of the operating point,
1964 RADEKA: THE FIELD-EFFECT TRANSISTOR 361

usually less than unity. Under pract In the case Cgs = Cin (zero external
operating conditions H1 (wg, wd) is ab capacitance), Rn would be increased by
0.7. The channel thermal noise can b 1/3.
expressed as an equivalent noise
resistance in series with the gate. These results are derived from the
From the condition nonsaturated region model and are
applied to the saturated region as a
id = 4KTAf R .2 and (11) it boundary value for wd = 1. While
there is no noise theory for the
follows that saturated region so far, measurements
show that the noise does not change
Rn 0 .7 significantly in the region 1 < wd < 2
n
ms which is of interest for small signal
applications. These results apply for
Comparing this with (7) or (8) for a frequencies lower than the cutoff
typical gate bias wg = 1/5 tO1/4, it frequency wo of the field-effect
follows that the equivalent noise transistor.
resistance is approximately equal to
the channel dc resistance. Other noise sources, not inherent
to the field-effect transistor principle,
In addition, the channel thermal are present. One is the shot noise due
noise introduces a noise current whic to thermally generated reverse (leakage)
flows in the input circuit. This cur current It in the gate-channel junction,
is due to the coupling of thermal noi which is (assuming negligible gate-
voltages developed along the channel source forward current) given by
the gate by the gate-channel junction
capacitance. It may be represented a i2 2eIt f (15)
an equivalent noise conductance in
parallel with the input (gate-source) ''V Tl -here is also "1/f noise" in the
channel current, which can be
g R lw2Ca * H2 (w) (13) represented as an equivalent gate
noise voltage
where C g is gate-source junction
capacit nce, Hs(w ) m 0.35 for practi ~cal v3
f = A f ftf
T- (16)
bias conditions. The effect of this
noise source depends on the impedance where Af is 1/f noise constant in V.
of the input circuit. In a charge- A contribution to 1/f noise can be
sensitive amplifier, where the input explained in terms of carrier-density
impedance is capacitive, it results i fluctuations due to trapping in the
a noise voltage which is independent
frequency, space-charg9 region
transistor.
of the field-effect
These noise sources are
markedly dependent on transistor
4KTAfg 1 technology and differ considerably
n = in 1
= . gn
among presently made types.
in W2in

At low temperatures the gate leakage

current noise decreases with the leakage
34KTAf 1
RR( tgA current. (The thermal noise decreases
n
linearly with temperature.) At
I

in

temperatures where the carriers in the

channel are only partly activated,
where Cin is total input capacitance there will be an additional noise due
including Cgs. Therefore, the effect to fluctuations in carrier density. For
of the "induced" gate noise in a Si devices this would take place at
charge-sensitive amplifier is to temperatures below 1000K.
increase the equivalent thermal noise
resistance in series with the gate to In presently available devices the
the value gate-leakage current noise would be
negligible for most applications. The
predominant noise in the low frequency
RI R( 1 + ( (14) region would be l/f noise. The coytant
Af for better devices is about lO-' Va.
362 IEEE TRANSACTIONS ON NUCLEAR SCIENCE June

FIGURES OF MERIT The terms fl,. f2, f3 which are functions

of the gate bias are equal to 1 at
Two generally used figures of merit W = 0. f2 and f3 have a broad maximum
can be defined for the field-effect a about wg = 1/4. These figures of
transistor, the ratio gms/Ids, i.e. merit were derived for uniform charge
the relative current change per unit distribution. However, these relations
voltage, and the cut-off frequency fo. are generally valid for any continuous
As a special figure of merit the ratio charge distribution (po is then
gms/C3 can be defined as a measure replaced by average charge density p
of nose performance in charge-sensitive Functions fl, f2, f3 do not vary at any
amplifiers (equivalent mean square given bias more than by a factor of 2
noise charge is proportional to for th
tions .
,ostextreme charge distribu-
02
R nC2gs = 0,7 -gs ). These figures of Expression (18) shows that for a
m low typical value of WOO of about 2V
merit are useful in assessing a field- the ratio gm/Ids > 2 V-1 (which
effect transistor and in relating the
device characteristics to its physical corresponds to that of electron tubes).
parameters. For most presently available field-effect
transistors this ratio is about 1 V-1.
The cut-off frequency is defined
as the frequency at which the trans- The cut-off frequency is determined
conductance decreases to 1/./2 of its by the dielectric relaxation time
dc value. It can be derived by two K/(tpo ) and by the width-to-length ratio
approaches. In the first, the field- of the channel. However, for a given
effect transistor is considered as a saturation potential WOO fo is
distributed RC transmission line. The determined only by the channel length.
time required to affect a change in Figures of merit (19) and (20) emphasize
transconductance is calculated, and the fact that for a high cut-off
from this the cut-off frequency. In frequency and low noise the channel
the second, the carrier transit time length-to-width ratio should be as small
through the channel is determined, and as possible. For low noise the ratio
from this f0. It can be shown that fO/C should be high. For a typical
both approaches lead to approximately mediR capacitance device with
the same result, relating fo to the Cgs = 20 pF, gms = 1 mA/V, WOO = 3-4V9
lumped transconductance and capacitance fo t17 Mc/sec. Assuming carrier t

as mobility of 1500 cm2s-V-lVl the channel

length would be of the order of 10-2 cm,
so that for higher frequencies extremely
fo = 53 gs (17 small dimensions are required. (A
gs detailed consideration of field-effect
transistor limits should take into
where 0 is a constant dependent on account the dependence of carrier
charge density profile, equal to about mobility on the electric field.)
1/3 for uniform charge distribution.
Using the relations (5), (6), (7), In a comparison of the gM/C2 ratio
and (10), the figures of merit can be of field-effect transistors with that
expressed as of electron tubes a factor of 3.5
should be taken in favor of the former
(the equivalent noise resistance for
Ids -
ids go
*L fl(wg) (18 electron tubes Rn ET t 2-5/g while
m
for field-effect transistors
Rn FET s ° 7/gm' so that for equal
gms _ .tp 0 0
0 K
0

-5 f2 (w9 ) transconductance Rn ET/Rn FET t 35).

gs
(19)
-La_ 2P f2 (w) APPLICATIONS
g9
)

123
Field-Effect Transistors as Active Devices
The low gate leaka e current
gms ( a 3. 1 f (w ) - (20) (typically 10-11 to 10-A) and low noise
C2 (Rn A 500-1000) make field-effect
gs
1964 RADEKA: THE FIELD-EFFECT TRANSISTOR 363
cantly due to the decrease in carrier
transistors particularly suitable for mobility and in number of activated
charge-sensitive amplifiers. A number carriers) is liquid nitrogen temperature
of such applications has recently been (770K). gms and Ids at 770K are about
reported.O,99,l0 An example of a 2/3 of the room temperature value
general purpose charge-sensitive (for Si devices). This makes them
amplifier for a wide range of particle suitable for use with cooled semiconduc-
energies is given in Fig. 3. The noise tor detectors. Integrated detector-
field-effect transistor systems offer
better performance due to lower circuit
capacitance, and also due to longer
time constants, which are made possible
by decreased detector (and transistor)
leakage currents.
Other applications for signal
amplification include high input
impedance amplifiers for integrators
and hi§h input impedance input stages
using source followers" (with "boot-
strapping" for low input capacitance).
GUARD VOt-AIE P
Some field-effect transistors with
very low 1/f noise (2N2500) are suitable
RAH6E TOTA TOTAL ENERGY OUTPUT
-E C, (p o (Wa) RAN6E (AEV) AMPLITUDE v)
50 500 0 5.0-10
for very low frequency amplifiers (in
K 2 2.2 220 -

X 5 10 10 5--&o 22 -220
X" I 44'- Z2O
10-1000 cps region).
\5 2 50 - 25-0

C 020 J, 200000, 2NZ608 RL -050-106.) AC ORPON

-. IRC -TYPE 0TFR P-CHANNLFLET rO DErECTOR LeAKA6NE
O WA T 6. / O
10X -1500 hS CURRENT
r2 - ZN28S0
r3 - 22ZZS5
r'- 20N706
Field-Effect Transistors as Passive Devices
One application that employs field-
effect transistors as variable conduc-
Fig. 3-Charge sensitive amplifier. tances is analogue multiplication.lll2
Analogue multiplication is based on the
performance to be expected with a proportionality of the channel
typical medium capacitance field-effect conductance to the g ate voltage
transistor (2N2500, Texas Instr., or (relations (7) and (8) ). A four-
2N2608, Siliconix) with g * 1 mA/V quadrant pulse amplitude multiplier,
and Cgs t 20 to 30 pF is RHM (3 to 6) suitable for use in E * AE particle
keV + (0.1 to 0 . 2) keV/pF at integrating identification systems, is shown in
and differentiating time constant Fig. 4. The minimum solution time of
T = 1 psec. Somewhat better performance such a multiplier is about 20 nsec.
can be expected at longer time constants,
T t 5 ,usec, if these are not prohibited R3
by high detector leakage currents. x
Good noise performance is restricted to il
lower capacitance detectors. An ll
~ XF
XF- I
-

optimum field-effect transistor can be

chosen for given detector capacitance
if a line of field-effect transistors )6v
with equal scale factors for g and
cgs is available. From the coaition
that the equivalent ms noise charge8

1 (Cg8 + Cd)2 should be

minimum, it follows that Cg5 should be

equal to Cd (Cd = detector and input
circuit capacitance). Such a line of
field-effect transistors is 2N3113,
2N2606,7,8,9 Siliconix, with Cgs from CURRENT AMPLIFIER
1 pF to about 50 pF.
m (x-y)
The lowest practical temperature
at which field-effect transistors can
be operated (before the current and Fig. 4-Four-quadrant analogue pulse amplitude
the transconductance decrease signifi- multiplier. 11,12
364 IEEE TRANSACTIONS ON NUCLEAR SCIENCE June

For some special applications 4. Van der Ziel, A., "Thernal Noise in
field-effect transistors can be used as Field-Effect Transistors," Proc. IRE,
controllable feedback elements and as 50 (1962), 1808-1812
gain control elements in amplifiers.
5. Van der Ziel, A., "Gate Noise in
As there is no internal potential Field-Effect Transistors at Moderately
between drain and source (no offset High Frequencies," Proc. IEEE, 51 (1963),
potential) field-effect transistors can 461-467
also be used as low-level choppers.1
6. Van der Ziel, A., "Carrier Density
Fluctuation Noise in Field-Effect
CONCLUSIONS Transistors," Proc. IEEE, 51 (1963),
The advantages of field-effect
1670-1671
transistors are in their high input 7. Richer, I., "Basic Limits on the
impedance, low noise, and a wide Properties of Field-Effect TransistorsI,"
operating temperature range Solid State Electronics 6 (1963),
(-2000C to +1000C). In addition, the 539-542
field-effect transistor represents a
new circuit element - voltage (charge)
controlled conductance with a fast 8. Radeka, V.,p "Field-Effect Transistors
response. in Charge Sensitive Amplifiers," Proc.
NAS Conference on "Instrument Techniques
Field-effect transistors are in Nuclear Pulse Analysis" Monterey,
particularly suitable for low noise California, 1963, to be published
amplifiers at low and medium frequencies,
for high input impedance amplifiers 9. Blalock, T. V.,, Pierce, J. F .,
(as for integrators) and for charge- "Application of Field-Effect Transistors
sensitive amplifiers. In these in Low Noise Wideband Voltage and Charge
applications they complement bipolar Sensitive Preamplifiers,," University of
transistors in replacing vacuum tubes. Tennessee, El. Eng. Dpt., Scientific
As variable conductances, they are Report No. 5, 1963
suitable for analogue multipliers and
as controllable feedback elements. 10. Smith, A. J., "The Behavior of Field-
Effect Transistors in a Charge Sensitive
They are limited, at present, by Amplifier," AERE-R4379, Harwell,
the difficulties in making extremely Berkshire
small geometries, to frequencies below
about 50 Mc/sec and in transconductance 11. Miller, G. L., Radeka,, V.,, "Analogue
to values below 5 mA/V (at reasonable Multiplication with Field-Effect
capacitance values). Transistors," Proc. NAS Conference on
"Instrument Techniques in Nuclear Pulse
There is another class of unipolar Analysis" Monterey, California, 1963,
to be published
devices, so called "insultteg gate '
field-effect transistorsl.1l or dielec- 12. Radeka, V., "Fast Analogue
tric triodes.15 They are, at present, Multipliers with Field-Effect Transistors,"
limited by very high 1/f noise and IEEE Trans., NS-l1 (1964 ), 302-307
they have a low gm/I ratio. However,
they are very promising with respect to 13. "Field-Effect Low-Level Choppers,"
high input impedance and cut-off Siliconix, Application Note, Nov. 1963
frequency.
14. Wallmark, J. T., "The Field-Effect
RE FERENCES Transistor - A Review," RCA Rev. 24
(1963), 641-660
1. Shockley, W., "A Unipolar "Field-
Effect" Transistor,," Proc. IRE, 40 15. Wright, G. T., "Space-Charge-
(1952), 1365-1376 Limited Solid-State Devices," Proc.
IEEE, 51 (1963), 1642-1652
2. Bockemuehl, R. R., "Analysis of
Field-Effect Transistors with Arbitrary 16. Borkan, H., Weimer, P. K., "An
Charge Distribution.," IEEE Trans., ED-10 Analysis of the Characteristics of
(1963), 31-34 Insulated-Gate Thin-Film Transistors,"
RCA Rev. 24 (1963), 153-165
3. B6senberg, W. A., "Transfer Charac-
teristics of Field-Effect Transistors,"
RCA Rev., 24 (1963), 688-704