Sie sind auf Seite 1von 14

Solid-State Electronics Pergamon Press 1966. Vol. 9, pp. 1035-1048.

Printed in Great Britain

CURRENT T R A N S P O R T IN M E T A L - S E M I C O N D U C T O R
BARRIERS
C. R. CROWELL t and S. M. SZE
Bell Telephone Laboratories, Inc., Murray Hill, New Jersey, U.S.A.

(Received 1 June 1966)

A b s t r a e t - - A theory for calculating the magnitude of majority carrier current flow in metal-
semiconductor barriers is developed which incorporates Schottky's diffusion (D) theory and Bethe's
thermionic emission (T) theory into a single T - D emission theory, and which includes the effects
of the image force. A low electric field limit for application of this theory is estimated from
consideration of phonon-induced backscattering near the potential energy maximum. A high electric
field limit associated with the transition to T - F emission is estimated from calculations of the
quantum-mechanical transmission of a Maxwellian distribution of electrons incident on the barrier.
The theory predicts a wide range of electric field ~ 2 x 102 to 4 x 105 V/cm over which the T - D
theory may be applied to metal-n-type Si barriers at 300°K. The corresponding range for metal-
n-type GaAs barriers is 9 x 103 to 8 x 10 ~ V/cm at 300°K. The decreased upper limit is due mainly
to the smaller electron effective mass in GaAs, the increased lower limit to a small optical-phonon
energy and a shorter e!ectron-optical-phonon mean-free path. The theory predicts Richardson
constants of 96 and 4.4 A[cm~/OK 2 for metal-n-type Si and metal-n-type GaAs barriers respec-
tively. Experimental measurements on both metal-Si and metal-GaAs barriers are in general
agreement with the theory. Values of the barrier n [ ~ ( q / k T ) ( d V / d In J)] appreciably greater than
unity are predicted for the field-dependent barrier height which occurs when an interface layer of
the order of atomic thickness exists between the metal and the semi-conductor. A field dependence
of the barrier height is shown to have no first order effect on the derivative of the 1[C 2 vs. V rela-
tionship for the barrier. The intercept of a 1/C 2 vs. V plot is shown to yield the barrier height
extrapolated linearly to zero field in the semiconductor. Experimental evidence for the existence
of interface layers in near-ideal Schottky barriers is also presented.

R 6 $ u r n 6 - - O n d6veloppe une th6orie pour calculer la grandeur du courant porteur majoritaire dans
les barri~res m6tal-semiconducteur qui comprend la thdorie de diffusion de Schottky (D) et la th6orie
d'dmission thermionique de Bethe (T) combindes dans une seule; cette th6orie d'dmission T - D
comprend les effets de la force d'image. Une faible limite de champ dlectrique pour rapplication de
cette thdorie est estim6e en consid6rant les dparpillements arri~res induits par des phonons pr6s du
maximum d'6nergie potentielle. Une forte limite de champ dlectrique associde h la transition de
l'dmission T F est estimde par des calculs de transmission quanta-m6caniques d'une distribution
maxwellienne d'dectrons incidents h la barri~re. La thdorie prddit une large gamme de champs
61ectriques ~ 2 x 102 /t 4 x 105 V/cm le long de laquelle la th6orie T D peut ~tre appliqude aux
barri~res mdtal-Si de type n h 300°K. La gamme correspondante pour les barri6res m6tal-AsGa de
type n est de 9 x 10 a ~ 8 × 10 ~ V/cm h 300°K. La r6duction de la limite sup6rieure est due princi-
palement/t la masse d'dlectrons effectivement plus petite dans l'AsGa et la r6duction de la limite
infdrieure h une plus petite 6nergie phonon-optique et un plus petit parcours moyen libre 61ectron-
optique-phonon. La th6orie pr6dit des constantes de Richardson de 96 et de 4,4 A/cm2/°K 2 pour
les barri6res m6tal-Si de type n e t m6tal-AsGa de type n respectivement. Des mesures expdri-
mentales sur ces deux barri~res sont en accord gdndral avec la th6orie. Des valeurs de barri~re n
[ ~ ( q / k T ) ( d V / d In J)] beaucoup plus grandes que l'unit6 ont dr6 prddites pour la hauteur de barribre
it d6pendance de champ qui se produit quand une couche d'interface ayant des dpaisseurs d'ordre
atomiques existe entre le m&al et le semiconducteur. On ddmontre que la ddpendance de champ de

t Present address: E.E. Dept., Univ. of Southern California, Los Angeles, California, U.S.A.
1035
1036 C. R. C R O W E L L and S. M. SZE

la hauteur de la barri6re ne poss~de pas d'effets de premier ordre sur la relation--deriv6e de 1/C a
en fonction de V--de la barri~re. On d6montre que l'interception de l'axe clans le graphique 1]C 2 en
fonction de V donne la hauteur de barribre extrapol6e lin6airement au champ z6ro dans le semi-
conducteur. On pr~sente aussi des preuves exp6rimentales pour l'existence de couches d'interface
dans les barri~res Schottky quasi-id6ales.

Zusammenfassung--Eine Theorie fiir die Berechnung des Majoritiitstr~igerstromes in Metall-


Halbleiter-~bergiingen wird entwickelt, welche Schottkys Diffusionstheorie (D) und Bethes
thermionische Emissionstheorie (T) zu einer TD-Theorie kombiniert und die Auswirkungen der
Bildkraft Beriicksichtigt. Die Anwendharkeit der Theorie ist dutch eine untere und eine obere
Grenze ftir das elektrische Feld beschr/inkt. Die untere Feldgrenze wird aus der Riickstreuung der
Elektronen infolge Elektron-Phonon-Wechselwirkungin der Niihe des Maximums der potentiellen
Energie abgeschiitzt. Die obere Grenze ffir das elektrische Feld ist durch den Einsatz der thermion-
ischen Feldemission (TF) gegeben und wird fiber die Tunnelwahrscheinliehkeiteiner Maxwellschen
Elektronenverteilung abgesch~itzt. Fiir Metall-nSi-t~bergiinge bei 300°K ergibt sich ein grosser
Feldst~rkenbereich, in dem die TD-Theorie anwendbar sein sollte, n~imlich2 x 102 bis 4 x 105 V/cm.
Der entsprechende Bereich flit Metall--nGaAs-l~berg~nge bei 300°K erstreckt sich dagegen nur yon
9 x 10 s bis 8 × 105 V/cm. Die Abnahme der oberan Feldgrenze ist hier haupts/ichlich auf die
geringere effektive Masse yon GaAs gegenfiber Si zurtickzufiihren und die Zunahme der unteren
Grenze auf die kleinere Anregungsenergie fiir optisehe Phononen sowei die kfirzere mittlere freie
Wegl~ingeder Elektronen bei Wechselwirkung mit optischen Phononen. Die Theorie sagt Richard-
sonkonstanten yon 96 und 4,4 A/cma/°K ftir Metall-nSi bzw. Metall-nGaAs-~lbergiinge voraus.
Experimentelle Ergebnisse sowohl an Metall-Si-als auch an Metall-GaAs-~bergiingenstimmen
im aUgemeinen mit der Theorie fiberein. Eine monoatomare Grenzschicht zwischen Halbleiter und
Metall hat eine felst~irkenabhiingigePotentialschwelle zur Folge. In diesem Fall liefert die Rechnung
fiir die Kennzahl n [ ~ (q]kT)(dV]d In J)] erheblich grSssere Werte als 1. Es wird gezeigt, dass sich
eine feldabh~ngige Potentialstufe in erster N~herung nicht auf die Kurve 1 / C 2 gegen V auswirkt.
Der auf/iussere Spannung Null extrapolierte Achsenabschnitt einer Kurve 1]C 2 gegen V steht im
Zusammenhang mit der Potentialstufe des betreffenden ~berganges. Experimentelle Hinweise ffir
die Existenz yon Grenzschichten in nahezu idealen Schottky-~berg/ingen werden ebenfalls
gebracht.

1. INTRODUCTION and collector. (12) This has prompted us to apply


THE EXISTENCEof many of the phenomena affect- the same theoretical considerations to a prediction
ing majority carrier current flow in Schottky of the I - V characteristic of a Schottky diode to
barriers, namely diffusion, thermionic emission, explore the range of applied electric field over which
tunnel emission, space charge effects, and image current flow should be described by SCHOTTKY'S
force lowering of the barrier has long been appreci- diffusion theory, (13) BETHE'S thermionic emission
ated.(1) Until the advent of procedures for obtaining theory, (14'Is,z) or a theory considering tunnel
reproducible near-ideal diodes, (2-5) however, the emission.
identification of these phenomena has been ex- I n Section 2 the factors affecting the potential
tremely difficult, and correlation between experi- energy vs. distance for a Schottky barrier are dis-
m e n t and theory has been of necessity somewhat cussed. Current transport limitations imposed by
qualitative. Since photo-excited electrons can be the diffusion of carriers across this barrier are
collected over a Schottky barrier, and in fact used treated under circumstances where it is assumed
to measure the barrier height, (~) a reasonable frac- that a thermionic recombination velocity boundary
tion of the excited electrons incident on a metal- condition can be applied near the metal-semicon-
semiconductor barrier do not recombine at the ductor interface. T h e result is a synthesis of the
interface. The current transfer observed in semi- thermionic (T) and diffusion (D) approaches, a
c o n d u c t o r - m e t a l - semiconductor (hot-electron) T - D model. T h e use of such a recombination
transistors further demonstrates this. (7-~°) I n velocity is considered in Sections 3 and 4. I n Sec-
addition the current gain limitations(1°) in these tion 3 the role of electron optical-phonon scattering
structures can be explained in terms of q u a n t u m - in the region between the barrier potential energy
mechanical reflection at the collector, (11) and maximum and the metal is estimated. T h e results
electron-optical-phonon scattering in the emitter in this section predict a low-field limit for applying
CURRENT TRANSPORT IN METAL-SEMICONDUCTOR BARRIERS 1037

the thermionic emission model, i.e. for assuming of $) and the attractive image force experienced by
that the metal acts essentially as a sink for carriers an electron when it approaches the metal. As
which cross the potential energy maximum in the drawn, the applied voltage difference V between the
direction of the metal. In Section 4 the effect of metal and the semiconductor bulk would give rise
quantum-mechanical reflection and tunnelling on to a flow of electrons into the metal. T h e imref cp
the recombination velocity is considered. This associated with the electron current density J in
treatment predicts the high-field limit of validity of the barrier is also shown schematically as a function
the thermionic emission model and the onset of of distance in Fig. 1. Throughout the region
thermionic-field (T-F) emission. This calculation between x m and w,
is notable for the fact that the W.K.B. approxima- d9~
tion is not used, and that the shape of the image J = --ql~ne dxx (1)
force barrier is considered explicitly. In Section 5
the relationship of the theoretical predictions to where/~ is the electron mobility and n e the electron
existing experimental results is considered. density at the point x.
n e = N c exp[(q/kT)(~-~)] (2)
3_
A9 8 where Arc is the effective density of states in the
conduction band, and T is the electron tempera-

I
ture. We will assume that the region between x m
and w is isothermal and that the electron tem-
Cn perature is equal to the lattice temperature.
Equations (1) and (2) will not be applicable be-
tween Xr~ and the metal surface since there the
potential energy changes rapidly in distances com-
FERMI [
: E~k\6 \ w x-,-
parable to the electron mean free path. In this
. . . Xm
region the distribution of carriers thus cannot be
METAL SEMICONDUCTOR described by an imref nor be associated with an
effective density of states. If this portion of the
FIO. 1. Electron potential energy (~) and imref 9 as a barrier acts as a sink for electrons, however, we can
function of distance in a Schottky barrier.
describe the current flow in terms of an effective
recombination velocity v R at the potential energy
2. CARRIER DIFFUSION IN maximum (this concept will be refined in Sections
SCHO'FrKY BARRIERS
3 and 4):
Since the diffusion of carriers is strongly affected J = q ( n m - no)vR (3)
by the potential configuration in the region
through which the diffusion occurs, consider first where n m is the electron density at xr~ when the
a typical electron potential energy (q~) vs. distance current is flowing, n o is a quasi equilibrium elec-
curve as shown in Fig. 1 for a metal-semiconductor tron density at x m, the density which would
barrier. The origin of the barrier itself has been occur if it were possible to reach equilibrium
considered by many authors and is due mainly to without altering the magnitude or position of the
a combination of the effects of surface states in the potential energy maximum. It is convenient to
semiconductor and the metal work function. (16-1s) measure both ~0 and ~ with respect to the Fermi
We will consider the case where the barrier is high level in the metal. Then
enough that the charge density between the metal ~(w) = V,
surface and x = w is essentially that of the ionized
donors, i.e. w is the edge of the electron depletion no = N c e x p ( - q ¢ ~ 8 / k T ) ,
layer. T h e rounding of ~bnear the metal-semicon- and
ductor interface is due to the superimposed effects
nm = Nc exp{q[~(xm) - ~B]/kT) (4)
of the electric field associated with the ionized
donors (which is shown by the dotted extrapolation where 4B is the barrier height.
1038 C. R. C R O W E L L and S. M. SZE

If n~ is eliminated from equations (1) and (2) semiconductor near the boundary. The standard
and the resulting expression for cp integrated Schottky result would then be obtained. (la) To
between x m and w, include image force effects in the calculation of va,
exp[(q/k T)¢( Xm) ] -- exp(q V/ k T) the appropriate expression for ~ in equation (7) is
¢ = ~B+ACB--gx--q/(16rrEX), (9)
w

where e is the image force permittivity, (1) and A~s


J fexp(q+/kT) dx. (5) is the image force lowering of the barrier. ~1)
I~NckT
Zm
( q5~ ~1j2
ACs = k ~ 7 ] ' (10)
Then from equations (3), (4) and (5),
qNcvR if ~ is constant for x < Xm.f If the electron
J = - exp[-q(~B/kT)]
1 + vR/v a mobility is constant, and the electric field is con-
stant for x ~ x m (l+2kT/qACs), then from
x [exp(qV/hT) - 1], (6) equation (7),
where va ~ ~#//3, (11)
w
where
va = q e x p [ _ ( q / k T ) ( ~ b B _ ¢ ) ] dx} -1 (7) co

Zm - - - d r. (12)
/3 q~jexp --q 2 k T ( 1 + 7 )
0
is an effective diffusion velocity associated with
the tranport of electrons from the edge of the "/ .~ (X--Xm)/X m. /3 is the approximate factor by
depletion layer at w to the potential energy maxi- which v a is reduced due to the modification of the
mum. If the electron distribution is Maxwellian barrier by the image force. Figure 2 shows/3 as a
for x > xm, and if no electrons return from the function of qA~h~/2kT. It is apparent that when
metal other than those associated with the current the Schottky lowering is comparable to or less than
density qnoVr, the semiconductor acts as a therm- kT/q,/3 is virtually unity.+ This is typically the case
ionic emitter. T h e n for Schottky barriers near room temperature.
A* T 2 Even at high field the effect of the image force
v~ = - - - - , (8) should not appreciably affect the current flow
qNo
where A* is the effective Richardson constant for t Image force effects only occur when an eIectron is
the semiconductor surface orientation. A* has present. If it were possible to determine the force on an
been determined by CROWELL(~5) for a variety of infinitesimal charge, it could be demonstrated that an
electric field actually exists at the potential energy
semiconductor tensor effective masses. At 300°K, maximum for electrons. This field is associated with the
v n is 7.0 x 106, 5.2 × 105 and 1.0 × 107 cm/sec for ionized donors in the depletion layer.
( 1 1 1 ) oriented n-type Ge, (111), n-type Si ;~ For £1~b~~ kT/q,
and n-type GaAs respectively. SCHULTZ(19) has
presented a similar derivation of equation (6)
assuming a free electron mass in the calculation
of v R and neglecting all image force effects. If the exact result for a parabolic potential distribution (ef.
Fig. 2). This result is similar to Landsberg's for diffusion
v a ~ vn, the pre-exponential term in equation (6) on both sides of the potential energy maximum of a
is dominated by v R and the thermionic theory most symmetricaI parabolic barrier. (~I) His result analogous
nearly applies. If, however, va 4 vR, the diffusion to equation (6) can be obtained by setting Vn equal to
process is dominant. I f we were to neglect image the value of va obtained from equation (11) and (13)
force effects, and if the electron mobility were (i.e. by assuming identical effects of diffusion on both
sides of the potential energy maximum). This treatment,
independent of the electric field, (2°) v a would be however, considerably distorts the effect of the image
equal to t~g", where 5° is the electric field in the force on the diffusion of carriers.
CURRENT TRANSPORT IN METAL-SEMICONDUCTOR BARRIERS 1039

5 to show not so much the quantitative corrections


PARABOLIC arising from the backscattered electrons as to
>-~4 POTENTIAL
determine the adequacy of the assumption that

~ LINEAR
BA ..J the region between the potential energy maximum
and the metal surface acts as a sink.
The approach used here parallels the authors'
-- POTENTIAL / // calculation of collector and emitter efficiencies in
DISTRIBUTION hot electron transistors. (12) The calculations are
based on the assumption that backscattering is due
t'~n." ~ I , I 1 I I II to optical phonon absorption and excitation, that
o.I I.O I0 this scattering is essentially isotropic, and that an
qA~ B effective mass approximation for the energy
2kT dependence of the relaxation time as used by
FIG. 2. Diffusion velocity reduction factor due to image CONWELJU(z2) for n-type Ge is appropriate. Scat-
force as a function of image force lowering measured in tering by acoustic phonons has been neglected.
units of 2kT/q. The equations determining the mean free paths
normal to the potential energy maximum, ha and
because with materials of moderately high mobility Ao, for absorption and generation respectively of an
the condition va > vn can still be satisfied in spite optical phonon are
of the fact that/3 is greater than unity.
In summary, equation (6) gives a result which is 1 _ 1 (E,~+Et+Eo) 1/2 1
a synthesis of Schottky's diffusion theory and A-a ;0 E-~ [exp( EoTk T) - 1]'
Bethe's thermionic emission theory, and which
predicts currents in essential agreement with the (13)
thermionic emission theory if tzd~(xra) > v R. This and if
criterion is roughly equivalent to and more easily E,~ + E t > E o,
estimated than Bethe's condition d~(Xm) > kT/qA,
where h must be an average carrier mean fl'ee path. 1 ( E n + E t - E o ~ 1~2 1
Since as will be discussed in the following section,
the carrier mean-free paths can be a strong func- k 0 \" < ] [1-exp(-Eo/kT)]"
tion of energy near the potential energy maximum, (14)
a might even be defined as tzkT/qv•. The following
If
sections will consider effects not treated in either
Bethe's or Schottky's theory.
e.+E~ < Eo, 1/a~ = 0.
E n is the energy associated with momentum normal
3. EFFECT OF P H O N O N SCATTERING to the barrier (and is a function of position), E t is
O N THERMIONIC EMISSION I N the energy associated with transverse momentum,
SCH.OTTKY BARRIERS Eo is the optical phonon energy, and Ao is the mean
In the previous section a recombination free path for generation at large values of E~ and
velocity v R associated with thermionic emission low temperatures. For electrons in 8i, Ge and
was introduced as a boundary condition to des- GaAs the values of Ao from avalanche multiplica-
cribe the collecting action of the metal in a tion measurements are 76, 105 and 58 A respec-
Schottky barrier. In many cases there is an ap- tively.(2a) When averaging the transmission proba-
preciable probability that an electron which crosses bility over the Maxwellian distribution, the effects
the potential energy maximum will be back- of transverse momentum were considered to first
scattered with a subsequent reduction in the net order by assigning to E t the effective value k T .
current over the barrier. Provided the back- The effects of multiple scattering were neglected,
scattered electrons are a small fraction of the total since simiple one- and two-collision theories give
electron flux, this effect can be viewed as a small very similar results. (2~) Only the backscattering
perturbation. The following estimates are designed which would result from first collisions occurring
2
1040 C. R. C R O W E L L and S. M. SZE
1.0

0.9

0.8

0.7

0.6

0.5
Ge si o ooA~
0.4 Xo=IO5;, 76A 58A

0.3

0,2

0.1

0 I I , I i f I
io3 io 4 io 5 io6 io~ 104 tO5 tO6 103 104 I05 I06
ELECTRIC FIELD (V/CM)
FIG. 3. Phonon-induced probability of emission, fp, averaged over a Maxwellian distribution of electrons
incident on the potential energy maximum of Ge, Si and GaAs Schottky barriers as a function of electric field at the
metal-semiconductor interface.

between the potential energy maximum and the comparison with leakage, generation, and recom-
metal are considered here. In principle additional bination currents. (2s)
backscattering might be expected from the metal, The above treatment is not expected to be valid
but the large ratio of the Fermi kinetic energy to in the low field limit, especially when k T >~ E o.
k T makes the solid angle of acceptance negligible. In this temperature range the electron energy dis-
The probability of emission, fe, averaged over tribution is not appreciably altered by the optical
a Maxwellian distribution of electrons at the poten- phonon scattering. In addition, in the low-field
tial energy maximum of metal-Si, metal-Ge and limit, about half the carriers return towards the
metal-GaAs Schottky diodes is shown in Fig. 3 potential energy maximum no matter where the
[a value offp less than unity implies that v R should scattering occurs. The approximations used in
be replaced by a smaller recombination velocity obtaining the results in Fig. 3 should thus give a
fevR in equations (3) and (6)]. low field limit of 0.5 for fp. Values o f f p less than
0.7 are thus indicative of failure of the ther-
oo mionic emission boundary condition, t The values
(t5) of fp are greatest for Si for a given electric field.
:p - f fp¢E, T) exp(-E/kT) dE/kT This reflects the fact that since k T ,~ E o at room
0 temperature, the majority of electrons generate
optical phonons when scattered and thus lose
f e increases slightly with decreasing temperature enough energy that subsequent passage back over
and increasing electric fields. The predicted
changes shown in Fig. 3 are too small to be directly t For fields low enough that fp cannot be reliably
measurable because the current variation due to determined by the above method, a diffusion analysis
changes in f e is small in comparison with changes extending beyond the potential energy maximum might
be used instead of the thermionic boundary condition,
caused by the image force lowering of the barrier. but the results should not be expected to be comparable
The latter effect in turn is difficult to measure in with the standard Sehottky result.
CURRENT TRANSPORT IN METAL-SEMICONDUCTOR BARRIERS 1041

the potential energy maximum cannot occur. In in excess of 2 x l0 s, 2 × 10 2 and 9 x 10 s V/cm.


the case of Ge the optical-phonon energy is much for Schottky barriers on Gee Si and GaAs, res-
nearer k T , but the mean free path is sufficiently pectively.
large that the majority of electron-phonon col-
lisions occur near enough to the metal surface that 4. QUANTUM-MECHANICAL TRANSMISSION
the solid angle for backscattering over the barrier In addition to effects of phonon scattering esti-
is much reduced. The collecting action of the metal mated in the previous section, the energy distribu-
also tends to be enhanced in Ge in comparison with tion of carriers should be further distorted from a
the case of Si and GaAs because the image force Maxwellian distribution because of quantum-
dielectric constant is larger. This makes the poten- mechanical reflection of electrons which have
tial energy maximum closer to the metal surface enough energy to surmount the barrier, and also

I00

n-GoAs
80 -

~ 6O - i ~ '-F=4X10
5V/cm
N
Z

0 l I I I
-.04 -.02 0 .02 .04 .06 .08 .IO
ELECTRON ENERGY {eV)

FIG. 4. Q u a n t u m - m e c h a n i c a l transmission coefficient for A u - G a A s


Schottky barriers as a function of electron energy relative to the
barrier height for selected electric fields.

for a given electric field (Ge, Si and GaAs have because of tunnelling of electrons through the
high frequency dielectric constants of 16.0, 11.7 barrier.
and 11.7 respectively). For GaAs the relatively CROW~LL and SzE (11) have outlined an approach
short mean free path and the fact that k T ,.~ Eo whereby the quantum-mechanical transmission
place more stringent lower limits on the electric (QMT) of carriers incident on Schottky barriers
field. The result for GaAs is the least rigorous of may be calculated as a function of the carrier
the calculations since the expressions for Ag and Aa energy and effective mass, the barrier height,
do not take into the account the polar nature of the electric field at the metal-semiconductor interface,
lattice vibrations which introduces additional image force dielectric constant, and shape of the
energy dependent terms into the expressions for potential barrier in the neighbourhood of the point
the mean-free path. where the conduction band edge in the semicon-
In summary, the above approach predicts that ductor merges into the conduction band edge in the
at room temperature the T - D results [equation (6) metal. The MacColl model is considered repre-
with v~ multiplied b y f v ] is valid for electric fields sentative of one of the more abrupt transitions
1042 C. R. CP, O W E L L and S. M. SZE

which might occur and the T h o m a s - F e r m i model Note the following interesting features of the
representative of one of the smoother transi- results.
tions.(11) 1. At a given temperature and below a threshold
Predictions of the electron Q M T coefficient, ~, electric field, )co does not vary rapidly with the
for Au-GaAs Schottky barriers are shown in field. T h e field at which fo starts to rise rapidly
Fig. 4 as a function of electron energy, E, relative marks the transition between thermionic and
to the barrier height for selected electric fields. It tllermionic-field (T-F) emission. Because of the
is apparent that appreciable tunnelling is predicted weighting produced by the Maxwellian distribu-
at fields in excess of 104 V/cm. In Fig. 5, fQ, the tion, this field decreases as the temperature is
lowered. The predicted transition to T - F emission
I.O I / depends only slightly on the barrier shape assumed.
,T:.OOO /, / 7//
l/ /~oooK // / I..~
"°' si I/ I /
] -- THOMAS- FERMi / ] I /
,6 I~'/ / % 300OK .... co . I/ //
o L \,oooK ...i~
I ,-.,.../ \lOOOK
oK iI

.4 f GaAs
•~ ~THOMAS-FERMI
.... M AC COLE
0 1 1 1 llllll I I I Illlll t I I fllll .E
10 3 IO 4 I0 5 IO 6
ELECTRIC FIELD (V/CM)
0 i t t i trill ~ i I t lilt] t t t i iltq
FIG. 5. Q u a n t u m - m e c h a n i c a l t r a n s m i s s i o n coefficient, to3 io 4 to 5 ios
fQ, averaged over a Maxwellian distribution of electrons ELECTRIC FIELD (V/CM)
incident on the potential energy maximum of an Au-
GaAs Schottky barrier as a function of electric field for FIC. 6. Q u a n t u m - m e c h a n i c a l t r a n s m i s s i o n coefficient,
selected temperatures. fQ, averaged over a Maxwellian distribution of electrons
incident o n the potential energy m a x i m u m of an A u - S i
Scbottky barrier as a f u n c t i o n of electric field for
ratio of the total current flow predicted considering selected t e m p e r a t u r e s .
tunneling and quantum-mechanical reflection to
the thermionic current flow predicted neglecting 2. There is no predicted periodic deviation of
these effects as shown as a function of electric A** as is found for thermionic emission into
field for the two barrier shapes investigated. vacuum. This result follows in spite of a large
reflection from both the metal-GaAs interface and
the potential energy maximum. The existence of a
fo =- i "c e x p ( - E / k T ) dE/'kT (16) periodic deviation requires as a rough approxima-
--00
tion that the distance from the metal to the poten-
tial energy maximum vary an amount in excess of
On examination of equation (3) it is apparent an electron wavelength as the electric field is varied
that the effective recombination velocity must be (el. HERRING and NICHOLS(26)). The presence of
fQfevn if we retain equation (8) as a definition of an image force dielectric constant appreciably in
v n. The complete expression for the J - V character- excess of unity and the small electron effective
istic taking into account fQ is thus mass cause the potential energy maximum to
occur within an electron wavelength of the metal
] = A** r 2 exp( - q~be/kT)[exp(q V/k T) - 1] for the fields in Fig. 5. At lower fields where x m
(17) is larger, lattice scattering (cf. Section 3) would
presumably break up the coherent scattering neces-
where sary to obtain a periodic deviation in A**.
A** = fffoA*/(1 +fpfovR/Va) (18) T h e results for Si (Fig. 6) are similar to those
C U R R E N T T R A N S P O R T IN M E T A L - S E M I C O N D U C T O R BARRIERS 1043

for GaAs: f o is somewhat smaller than that for this particular valley. (iS) The effective mass for
GaAs at low fields, and the predicted fields for the other valleys should be much smaller. The
T - F emission are larger, mainly because an effec- overall Q M T in Ge should be comparable to that
tive mass of 0.30 (the component of the effective in Si.
mass tensor in the <111 > direction) has been used In summary the product fpfo ~ 0.5 for barriers
in the calculation. This effective mass is consider- on Ge, Si and GaAs over the range of electric fields
ably greater than the effective mass of 0.072 for where the T - D model is applicable.
conduction band electrons in GaAs. The case of
Si is also more complex because electron emission 5. C O M P A R I S O N OF EXPERIMENT
in the <111 ) direction comes from six equivalent A N D THEORY
valleys in the band structure rather than a single The results of the above theory are summarized
valley in the case of GaAs. The theoretical ap- in Table 1 where the low-field and high-field
proach used here for Si does not consider either predictions for the limits of the T - D approach
the possibility of diffuse reflection (which would are listed as well as the values of A** for GaAs, Si
then permit an electron to be reflected into one of and Ge at 300°K. It is predicted that at room
six valleys) nor a rigorous treatment of the role of temperature the thermionic emission model should
the anisotropic effective mass. (lz) be valid for Si over a much wider range of electric

1.0
[ Table 1. Field limitations and A** values for the
THOMAS-FERMI I
I thermionic emission model at 300°K
.8 f . . Ge
. . MAC COLL I
I
.6 T =I00° K " " ~ ] 8D-r 8r-v A**
V/cm A/cm~/°K2
.4 . ii 25OOOK'x..~ GaAs 9 x 103 8 x 10¢ 4"4
Si 2 x 102 4 x 105 96
Ge 2 x 103 ~ 4 x 105 ~50
.2

0 , I , I
I03 I04 [0 5 IO6
ELECTRIC FIELD (V/CM] field than for GaAs, and that at low temperatures
the I - V characteristic of GaAs surface barriers will
FIG. 7. Quantum-mechanical transmission coei~eient,
.to, averaged over a Maxwellian distribution of electrons be dominated by T - F emission essentially as
incident on the potential energy maximum of an Au-Ge reported by PADOVANIarid STRATTON.(28) Especi-
Schottky barrier as a function of electric field for ally near room temperature the present theory
selected temperatures. provides a more rigorous approach than that of
Padovani and Stratton who have not considered
The case of a <lll)-oriented barrier in Ge is either the effects of the image force or the reflec-
still more complex since three equivalent and one- tion of electrons which have enough energy to
non-equivalent minima at the Brillouin zone clear the barrier without tunneling. These omis-
boundaries are involved. The results shown in sions can be expected to be important mainly in
Fig. 7 are for an effective mass of 1.59. These defining the transition from the T-D region to the
results, coupled with those of Figs. 5 and 6, illus- T - F region. The present theoretical calculations
trate the reduced transmission at low field and are in essential agreement with the experimental
the extension of the T - D model to higher fields results for Schottky barriers on heavily doped
with increasing effective mass. There is also some GaAs. (28) The small field range for application of
evidence of a periodic deviation in the Schottky the thermionic emission model may also explain
effect. This should, however, be very difficult to the difficulties experienced by Padovani and Strat-
measure since only ~ 10 percent of the electron ton in predicting I - V characteristic of surface
current in a Schottky barrier is associated with barriers on lightly doped GaAs.
1044 C. R. C R O W E L L and S. M. SZE
"Ideal" barriers Since d#B/d T may be of the order of 5 x 10- ~ V/°K,
We will discuss in detail only the experimental without an accurate independent measurement of
measurements on Si Schottky barriers. Three ~ba, the above approach cannot be expected to pro-
levels of analysis of the J - V characteristic are vide a check on the theoretical values of A**.
open: 2. Provided the current flow is due solely to
1. If the saturation current density Jsat (J majority carriers, the Richardson plot can be used
extrapolated to V = 0) from a plot of log J vs. V with less ambiguity if log (Jsat/T 2) is plotted as a
in the forward direction) is measured as a function function of Ca for a variety of different metal-Si

A**=96Arnp/cm;Z/°K
z
T=297°K

Ag P/t Au

.7!
I ........ I
.5 .6 .8
~c (w)
Fro. 8. Comparison of barrier height in the absence of the
image force deduced from I-V measurements and ~b, obtained
from C-V measurements : Cu-, Ag--, Pd-, Pt-, Au-Si: ATALLA
et al.,~°>W-Si! CROWELLet al. ~4~
of temperature and log (Jsat/T 2) plotted as a func- barriers at one temperature. A variant of this ap-
tion of 1/2, this Richardson plot will give approxi- proach is used in Fig. 8 where ~be is the barrier
mate values for A** and ~bB. A strong temperature height deduced by calculating ~bB from Js~t and the
dependence of 4JB in Au-Si barriers has been theoretical value of A** [via equation (12)], and
m e a s u r e d photo-electrically, (28) but no direct then adding A~B (at zero bias) to ~b~. I f the theor-
correlation has been simultaneously obtained be- etical A** is the correct value, 4Je and ~bc,the capaci-
tween Jsa~ and T. In the absence of explicit tively measured barrier height, should be com-
measurements of ~B as a function of temperature, parable. T h e data are consistent with the theoreti-
the Richardson plot yields cal value of A**, but a value of ~ 70 A/cm2/°K 2
provides somewhat better agreement between the
A**exp( kq dCB~ two sets of data.
3. There are many barriers (including those
(19) referred to in Fig. 8) which give 4Je values within
C U R R E N T T R A N S P O R T IN M E T A L - S E M I C O N D U C T O R BARRIERS 1045

~-, 50 mV of ~,, but whose J-V characteristics at This can only occur if there are some interface
first sight are not in detailed agreement with effects.
equation (17). Such barriers frequently exhibit
n values, Interface effects
q dV If the metal and semiconductor are separated
n --= , (20) by an interface layer of thickness As and permit-
k T d In Jcv>skrta) tivity E~, and if the surface state density at the
in excess of 1.05. The exceptions in Fig. 8 are the semiconductor is nsc near the Fermi level,
Pt-Si and W-Si barriers which have n values of
1.02. Since both A** and ~bB are field dependent, dee qn~ + (25)
some small deviations of n from unity should occur: do~ I_ %
from equations (17 and (20),
(cf. CROWELLet al.Cls)). Relatively large deviations
k T d In A** of n from unity may occur if d ce/d o~ is appreciable
n=[1- m
2q V e d l n f f in comparison with the width of the semiconductor
depletion layer [cf. equation (21): ~ / 2 V e ~ w-l].
A$B Equation (25) neither specifies the value of nsc nor
(21) A~[E~for a given d ~be[d~, nor, for that matter does
4 ~ + df~
it specify all the sources of potential effects of
where field on ~be.
o~ = ( 2 q N a Ve/%)~ 12, (22) The existence of a field dependence of~e would,
of course, represent a departure from what is
and normally considered an ideal Schottky barrier.
kT Let us, however, consider some of the resulting
Ve = ~b~- V - - - [1 + In (Nc/Na) ]. (23) effects. Equations (17), (22) and (23) still hold if
q
the interface layer is thin enough that electrons
Na is the donor density, and % the semiconductor can tunnel through it easily enough that they can
permittivity. rapidly adjust the population of the semiconductor
d lnfp d lnfo surface states to the Fermi level in the metal.
Figure 9 shows the ~e vs. ~ relationships which
d In A** d In oa d In o*
follow when these equations are used to analyse
d I n d' f~fQv R the J-V characteristics of Au-Si and Pt-Si barriers
1 + - - investigated by ATaLLA et al. (s) These barriers
Vd
have been produced on chemically cleaned surfaces.
d In Va The n of the Pt-Si barrier is 1.02, and that of the
Au-Si barrier 1.15 at low current densities de-
d I n o~
+ (24) creasing to 1.04 at high current densities. The
Pt-Si barrier is very nearly an ideal Schottky
1 + ~ barrier: d~be/d~ >~ 15A. This value could be
fdo*~ caused by either a high surface-state density
The contribution of field dependence of A~B (nsc ~ 10x4 states/cm2/eV), or a very thin interface
or A** to n normally will not exceed 1-2 percent layer (A ~ 1 A if %/E~"~ 12). Similarly ideal
unless measurements are taken very close to the barriers have been reported for chemically-grown
flat band conditions (where d In A**/d In ff ~ 1, W on Si, <4) and at liquid nitrogen temperature for
since v a ~ tzo~) or unless T - F esmision occurs (in Au-Si barriers examined photoelectrically.C2s)
which case d In f o l d In op may be very large). If The results for the Au-Si barrier in Fig. 9 are also
we are to explain the J-V characteristic of a barrier of interest. ATALLA et al. ¢s) have reported two
in terms of majority carrier flow when n > 1.02, properties of similar barriers which suggest that the
there must thus be an appreciable contribution J-V characteristic is due mainly to majority
from field dependence of 5be [cf. equation (21)]. carrier flow.
1046 C.R. C R O W E L L and S. M. SZE


.860
A" ~ = 9 6 Arnp/cm2/°K ?-
T=297°K
.840 St

~.820
"9.,
iO Pt- St
.800 . . . . x ~ %x-- x ~X~x.._. W . _ . x.
X
.780

! I I I 1 1
.760
1.4 2.0 3.0 4.0 4.2
ELEC'rRIC FIELD (XIO 4 V/cm)
Fla. 9. Barrier height ~b, vs. electric field E for Pt-Si and Au-Si
Schottky barriers.

1. When such barriers are used as emitters in where ¢0 is the barrier height extrapolated linearly
hot-electron transistor structures, the collector to zero electric field. T h e n
current can be proportional to the emitter current i 2 [ kT
even when n is as large as 1.2. C--~= q%Na" [ ¢0 V----q (l+lnNc/Na)
2. T h e eapacitively-measured barrier height of
the Pt-Si barrier is less than that of the Au-Si
d2 ( d ~ e ~ 2 + . . . ] (28)
barrier on similar material, yet the saturation
reverse current density of the Pt-Si barrier is less. 4Ve \ d E ]
T h e reverse current of the Au-Si barrier is thus i.e. 1/C 2 is independent of first order terms in
presumably not due to carrier generation in the d Ce/d ~, and a plot of 1/Cs vs. V will give infor-
depletion layer. mation about interface effects only if they are large
Both the above observations can be shown to be enough to produce a non-linearity in the plot.~
consistent with the existence of an interface layer From Fig. 9 it appears that the Au-Si barrier is not
of appreciable thickness. T h e first has a straight- as high as that of the Pt-Si barrier at appreciable
forward interpretation if the interface layer is not reverse bias. When ~be is extrapolated to the case
thick enough that tunneling of electrons into the of zero electric field, however, it also appears that
metal is appreciably inhibited and if the barrier n the Au-Si barrier height will exceed that of the
can subsequently be explained. T h e second obser- Pt-Si barrier in agreement with the capacitance
vation requires an initial consideration of the C - V measurements. T h e Z-shaped ~be-d~ curve for the
relationship of the barrier. Since the charge per Au-Si barrier can be explained if there are regions
unit area, Q, stored in the depletion region is in the Si bandgap of high ( ~ 1 0 la states/cm2/eV)
Q = e s d°, and C=-dQ/dV, (26)
t Equation (28) is, of course, a general result inde-
1 2 Ve pendent of the origin of d~b~/dg. If, however, the barrier
is biased far enough in the forward direction, C will
C2 q%Na[ d~d,tb_e]2 become comparable to e,/a. If the frequency at which
1 2V~ d E J the capacitance measurement is made is comparable to
Now the relaxation time for charging the interface states, a
study of the frequency dependence of C should yield
Ce -- ~bB+ ACB = ¢o + d~bk N, (27) information about both the interface thickness and the
dg surface state density.
CURRENT TRANSPORT IN METAL-SEMICONDUCTOR BARRIERS 1047

surface-state density commencing ~ 35 mV on


(32)
either side of the Fermi level at zero bias, and an
interface thickness A~ ~ 45 A (if%/q ~ 12). These
parameters should, however, only be accepted as When V > 3 kT/q,
semi-quantitative. An interface of this thickness J ~ Jsatexp(q V/nk T)
should decrease fo, possibly by an order of mag-
nitude. T h e corresponding decrease in A** would When - V > VD
also decrease the calculated values of ~e, d ~e/d ~,
and ft. A photoelectric measurement of ~bB as a J - - Jsatexp[2(nnl) q T ( - VVD) 1/2]
function of bias would provide a more reliable check
on such interface effects and would also permit a We thus see the typical lack of saturation of the
direct measurment of A** at any bias. reverse characteristic of a Schottky barrier as a
T h e present results suggest that the major direct consequence of interface effects and as
reason for deviation from the ideal J - V character- having a quantitative relationship to the departure
istic is modulation of the barrier height by the of the forward n from unity.
applied voltage through effects independent of
the image force. T h u s from equations (17) and 6. CONCLUSIONS
(20), It appears that the J - V characteristics of t h e
dSB n -- 1
- (29) Schottky barriers referred to in Fig. 9 are essenti-
dV n ally due to majority carrier current flow (one of the
major desirable properties for Schottky barrier
at any operating point such that V > 3 kT/q. applications) even if interface effects produce
T h e value of n should also be reflected in the marked deviations of n [ = - (q/kT)(dV/d In J)] from
reverse characteristic: since d Ce/d d ° is constant if unity. Even an n of near unity does not necessarily
nsc is constant and the contribution of the image imply an extremely thin interface layer since the
force to n is typically < 1 percent, when n ~> 1.02,
barrier may also be pinned by a high density of
a logical approximate form for the J - V character- interface states at the semiconductor surface. T h u s
istic should be in principle the values of the Richardson constant
(A**) deduced in Sections 2-4 are predicted upper
J _- J s ~ t e x p [ - q ( d ~ - g ~ o ) dCB-/kT] limits applicable in the case of thin interface
d6 ° J
layers.+*
[exp(q V/k T) - 1], (30) T h e control of the J - V characteristic of a
Schottky barrier is thus intimately related to
where fro is the value offf at zero bias. If we neglect
control of the interface properties, especially states
the effect of variation of ¢s on ff~,
at the semiconductor surface. I n this respect the
d~bB 2 ( n - - 1) [esVD] 1/2 Schottky barrier resembles the M O S structure.
d--if----- n [2--~aJ ' (31) T h e major point of differentiation between the
two structures lies in the difference in thickness of
where VD is the intercept of the 1/C 2 vs. V plot. the oxide or interface layer. In the M O S structure
Then the oxide layer is thick enough that the Fermi
2(n-1) q level in the interface states is closely coupled to the
J = Jsa~exp - Fermi level in the semiconductor, while in the
n kT Schottky barrier, the interface layer is thin enough

x VD[(1 -- V/VD) 1:2 - 1]] :~ If the data in Fig. 8 were presented in the form of a
3 plot of $o (the barrier height extrapolated to zero electric
field) vs. $c (the capacitively-determined barrier height),
"~The derivation of the equivalent of equations (31) the need for a smaller experimental value of A** would
and (32) is considerably more complex if this approxi- become apparent. This is further evidence of the pre-
mation is not made. Multiplying VD by n 2](2n- 1) yields sence of interface layers of appreciable thickness in t h e
the exact result. measured barriers.
1048 C. R. C R O W E L L and S. M. S Z E

14. H. A. BETHE, M I T Radiation Lab. Report 43/12,


that the F erm i level of the interface states is closely
coupled to the Fermi level in the metal. (1942).
15. C. R. CROWELL,Solid-St. Electron. 8, 395 (1965).
16. J. BAm~EEN,Phys. Rev. 71, 717 (1941).
17. A. M. COWLEYand S. M. Sz~, J. appl. Phys. 36,
REFERENCES 3212 (1965).
1. H. K. HENISCI~,Rectifying Semiconductor Contacts, 18. C. R. CROW~LL,H. B. Srlom~ and E. E. LABATE,J.
Clarendon Press, Oxford (1957). appl. Phys. 36, 3843 (1965).
2. D. KAHNO, Bell Syst. tech. Y. 43, 215 (1964). 19. W. ScHULTZ, Z. Physik. 138, 598 (1954).
3. M. IV[. ATALLA,R. W. Sos~Ea, R. C. LvcAs, D. A. 20. It is probable that the high electric field-low imref
REID and V. M. Dowlma, Investigation of Hot gradient value of the mobility appropriate here is
Electron Emitter, Scientific Report No. 3, not the same as either the low field-low imref
AFCRL-63-113, February, 1963. gradient mobility or the high field-high imref
4. C. R. CROWELL,J. C. SARACEand S. M. Sz~-, Trans. gradient mobility. C. GOLDBERa, private com-
Metall. Soc. A.I.M.E. 233, 478 (1965). munication.
5. F. A. PADOVANIand G. G. SUMNER, J. appl. Phys. 21. P. T. LANDSBERO,Proc. Roy. Soc. A. 206, 463 (1951).
36, 3744 (1965). 22. E. M. CONW~LL,Phys. Rev. 135, Al138 (1964).
6. W. G. SelTZER,C. R. CROWELLand M. W. ATALLA, 23. C. R. CROWELLand S. M. SzE, Temperature De-
Phys. Rev. Lett. 8, 57 (1963). pendence of Avalanche Multiplication in Semi-
7. M. M. ATALLA and D. KAHNG, I.R.E. Trans. conductors, IEEE Solid State Device Research
Electronic Devices 9, 507 (1962). Conference, Northwestern Univ. Evanston, Ill.,
8. D. V. GEPPERT,Proc. Inst. Radio Engrs. 50, 1627 June, 1966.
(1963). 24. C. R. CaOWrLL and S. M. SzE, Solid-St. Electron.
9. C. R. CROWELLand S. M. SzE, Phys. Rev. Lett. 15, 8, 673 (1965).
659 (1965). 25. S. M. SzE, C. R. CROWELLand D. KAI~NO,J. appl.
10. S. M. SzE, C. R. CROWELL,G. P. Campy and E. E. Phys. 35, 2534 (1964).
LABaT~, J. appl. Phys. 37, 2690 (1966). 26. C. HmmiNa and M. H. NICHOLS, Rev. Mod. Phys.
11. C. R. CRGWELL and S. M. SzE, J. appl. Phys. 37, 21, 185 (1949).
2683 (1966). 27. F. A. PADOV,~',CZ and R. STRATTON, Solid-St.
12. C. R. CROWrLL and S. M. SzE, Solid-St. Electron. Electron. 9, 695 (1966).
8, 979 (1965). 28. C. R. CROWrLL, S. M. SzE and W. G. SelTZER,
13. W. SCHOTrKV,Phys. Z. 32, 833 (1931). Appl. Phys. Lett. 4, 91 (1964).

Das könnte Ihnen auch gefallen