VOLUME 50, NUMBER 25 PHYSICAL REVIEW LETTERS 20 JUNE 1985

Theory and Application for the Scanning Tunneling Microscope

J. Tersoff and D. R. Hamann
Bell Laboratories, Murray Hill, Neu Jersey 07974
(Received 17 March 1983)
A theory is presented for vacuum tunneling between a real solid surface and a model
probe with a locally spherical tip, applicable to the recently developed "scanning tunnel-
ing microscope.
"
Calculations for 2&& 1 and 3& 1 reconstructions of Au(110) are in ex-
cellent agreement with recent experimental results, if an effective radius of curvature of
9 g is assumed for the tip.
PACS numbers: 68.20. +t, 07.80. +x, 73.20.-r, 73.40. 0k

One of the most fundamental problems in sur- The tunneling current by first-order perturba-
face physics is the determination of surface struc- tion theory is
ture. Recently a new and uniquely promising
technique, the scanning tunneling microscope"
I =(2me/I)g„„f (E„)[1 —f(E, + eV)]
(STM), was introduced. ' ' This method offers, x )Mq„~ 6(Ep-E„), (1
for the first time, the possibility of di~ect, ~eal-
sPace determination of surface structure, includ-
where f (E) is the Fermi function, V is the applied
voltage, M„, is the tunneling matrix element be-
ing nonperiodic structures. A small metal tip is
tween states g„of the probe and g, of the surface,
brought near enough to the surface that the vacu-
and E„ is the energy of state f„ in the absence of
um tunneling resistance between surface and tip
tunneling. Since the experiments are performed
is finite and measurable. The tip scans the sur-
at room temperature or below and at small volt-
face in two dimensions, while its height is ad-
justed to maintain a constant tunneling resistance. age (-10 meV for metal-metal tunneling), we
take the limits of small voltage and temperature,
The result is essentially a contour map of the
surface. I=(2&/ )e'VQp„)M~„~ 5(E„-EF)5(E~-EF), (2)
The one-dimensional tunneling problem (i.e. ,
where EF is the Fermi level. The essential prob-
through two-dimensionally-uniform barriers) has
'
been treated extensively, and field emission
lem is to calculate M„„. Bardeen' has shown that
from a tip is well understood. The usefulness of
STM stems from the fact that it is neither one-
.
M, =-(@'/2m) Jd~ (0, *~4. 0.&4 *)— , (3)
where the integral is over any surface lying en-
dimensional nor field emission, but is instead
sensitive to the full three-dimensional structure
tirely within the vacuum (barrier) region separat-
of the surface. Little is known about tunneling in
'
this case. Here we present the first quantitative
theory for the scanning tunneling microscope, and
apply the theory to Au(110). Results are in ex-
cellent agreement with experiment. We infer that
under actual experimental conditions of Ref. 4,
the tip was roughly 6 A from the surface and had
an effective radius of curvature of about 9 A.
Unfortunately, little is known about the struc-
ture of the tunneling probe tip, which is at pres-

reproducible manner. ''

ent prepared in a relatively uncontrolled and non-
Similar results have
been obtained with W, Mo, and stainless steel
'
tips, so that details of the tip electronic struc-
ture do not appear to be important. We model
the tip as a locally spherical potential well where FIG. 1. Schematic picture of tunneling geometry.
Probe tip has arbitrary shape but is assumed locally
it approaches nearest to the surface, as illus- spherical with radius of curvature R, where it ap-
trated in Fig. l.A is the local radius of curva- proaches nearest the surface (shaded). Distance of
ture about the center located at r„and d is the nearest approach is d. Center of curvature of tip is
distance of nearest approach to the surface. labeled ro.

1998 1983 The American Physical Society

VOLUME 50, NUMBER 25 PHYSICAL REVIEW LETTERS 20 JUNz 1983

ing the two sides. The quantity in parentheses is where D, is the density of states per unit volume
simply the current operator. of the probe tip. Note that (8) does not imply that
To evaluate M„„, we expand the surface wave the value of the surface wave function g, at r, is
function in the form physically relevant. The matrix element is deter-
' )' mined by an integral entirely within the gap re-
f =Q Qoao exp[ —(k + I k~(+ GI &]
gion. However, because of the analytic proper
x exp[i(k~~+ 6) x], ties of (4) and (5), the formal evaluation of P„at
~
(4)
distance R + d correctly describes the lateral
which is a completely general expression for g in averaging due to finite tip size.
the region of negligible potential. Here ~, is The spherical-tip approximation entered only in
sample volume, k = & '(2m+) "'
is the inverse the normalization of (5). The crucial approxima-
decay length for the wave functions in vacuum, y tion was evaluating the matrix element only for an
is the work function, kII is the surface Bloch s-wave tip wave function. The q dependence of
wave vector of the state, and G is a surface re- b(q) in (7) then cancelled that of the & derivative
ciprocal-lattice vector. For a nonperiodic sur- in the matrix element (8), so that (9) involved
face the sum over G becomes an integral. The only undistorted wave functions of the surface.
first few ~& are of order unity. For tip wave functions with angular dependence
The wave functions of the tip are, in general, (I &0), it is sufficient to include the m =0 term
not known. In the region of interest, however, (other m give a node towards the surface). In that
they have the asymptotic spherical form case the terms in the Fourier expansion of g, are
=0 '"ckRe (kIr-r I) 'e '' ' weighted by a factor -(1+@'/k')'" in the matrix
tP (5)
element, which for relevant values of q can be
where , is the probe volume, k is defined as .
neglected for small I (In the example below the
above, and R is the radius of curvature. (We as- relevant q'/k'= 0.1.) The tip model therefore be-
sume for simplicity that the work function p for comes less accurate for large R, where higher l
the tip is equal to that of the surface. ) The form values become more important. A more exact
is chosen to be correctly normalized when the treatment would probably be f ar less useful, since
parameter c, (which is determined by the tip it would require more specific information about
geometry, detailed electronic structure, and tip- the tip wave functions, and would not reduce to an
vacuum boundary condition) is of order 1, with explicit equation such as (9) or (10) and (11) be-
the assumption R»k '.
We have neglected the low.
possible angular dependence of g„, which intro- Substituting typical metallic values into (9), one
duces some quantitative modifications discussed obtains for the tunneling conductance
below.
o = 0.1R'e'
p(r, ;EF), (10)
We expand the tip wave function (5) in the same
form as the surface (4) using the fact that p(r„E) =El 4.(r.)l'5(E. -E),
(kr) 'e = Jd'qb( q) exp[- (k'+ q')"'I zI]
x exp(i q x), ~
(6)
where 0 is in , distances are in atomic units,
and energy in electron volts. Since P Pr, ) '
I I

' ~ exp[2k(R + d)], we see from (10) that v ~e '""

b(q) =(2&) 'k (1+ q /k ) (7)
as expected. Because of the exponential depen-
The matrix element is then almost trivial to eval- dence on distance, it is not essential that the co-
uate. Substituting the surface and the tip wave efficient in (10) be every accurate.
functions in (3) and evaluating the expansion term By consideration of general aspects of the be-
by term in G, one finds havior of p(r;EF), it is possible to draw some
conclusions concerning the sensitivity of STM. In
M„,=(S /2m2)4&k '0, ' 2kRe P„(ro), (8) particular, the suppression of higher Fourier
where r, is the position of the center of curvature components in (4) and (9) is equivalent to an in-
of the tip. On substitution into (2) the desired re- strumental resolution of roughly [ 2k '(R + d)]"',
sult is at least for components q'«20. Since 2k ' = 1.6 A,
if R + d =15 A (as in the example below), the reso-
I =327J@ e yq7 D (EF}R2k 4e2~
lution is about 5 A, so that the 8 A periodicity of
x+I p„(r,) I'&(E„-EF), (9) Au(110) (2x 1) is resolved. It is important to note
that for periodic structures the measured corru-
1999
 Vpi. UME ) +UM&ER 25 PHYS ICAL REVIEW LETTERS 20 JvNE 1983

—&o-"
16 -„o-~e
—)o-L
—14-«-" -15
LLJ

4 1O-'
M
CI
2 ~~

J~.
1g . i
oli i 1 (in lan)ad q (out of plane). Coontours
s ale. Pecul n
e — ar of the basis in d rg the exponent' ia 1 d eeay I.n-
roug s. Center of eurv 'ne.
ws dashed lin

'
gation amplitude e d ecreases exPonentially
o with error from thisis f'finite-interval a r
approximation can
'
resolution. Thisis Debye-Wailer-li
D ex- e
bee estimated n i at'ively and s
quantit

"'"""1 z e off why in Ref. 2

plains the puzzle4
h't
'
so ution, non p eri odic structur es
an 1 io h ere. ) Des p'ite some difficult yin
res
obtaining conv ergence with respect to the plane-
re
were clearl y resolved on Au(110), our asis set for the op c-
t' was not.
but the periio di c reconstruuction tul uni cells we umer-
The Au(110)
t t
u
't
0', surface normall
a missing-row
re
d tllb
o e quite ade u
tt er f or the 2~1 e e
ruc ion has also be e tl t' pproximate meth d f
e a . reported hiigh-resolution treating m ore comp]. icated strue,
'
e
STM measuremmen t s for an Au~1100) surface with
regions of both 2&&1 and 3&&1 stru t "ll' 'nd
e ~ 1 structure
ana ogous to thee 2~
STM corrugation
d 3'1
~ 1. Measured
ions were 0.45 and 1 k. -. -..- ince the actual ti e
si d er a tip radius
corrugation of
nelin t ' 10'
Th en is found to be 6
W h 1 ltd ( r.'E
, F) for ce u nucj. ei to e lp
ces using a recentl y d eveloped linear- ot ~ti l 11 (.i.e. , the shell at wave
ized augmentede p 1ane-wave method
m described fu es ecayin g iin character). Thi s ~

elsewhere surfaaee we used a sa

F or the 2&&1 sur 1 b ex e '
sis ent with experimental esti t
geometry of th ree complete la of d based on
1 [ lt t e (110) rows m' g i er , (o) A for
geometry suuggested by Binniigeet (31) surface in excellen
ce ent agreement w'th
al. ' was emplooye; d' an as mm periment.
o complete la ers s above d was determined
ln the anal ysis d rath
sing row, and a fo y wi1 tw 0 directly by (10) since the deependence
e re t
of curre
see Fig. 2). The S uupon R largel y cancels as not o e d abov
ov r
xl- was i db ftt
i ing the exp er'imental corr-
ru-
mated by the ch within 0.5 eV of E g ati on, which dep en d s onR+d , and subtractining d.
i
inite interval width
ivided by thee finit .
of 1 eV. (Thee
'
Thhe corrugation n is m
more suseeptibli e to errors
2000
V&&UME 50, NUMBER 25 PHYSICAL REVIEW LETTERS 20 JUNE 198$

both experimental and theoretical, than is the cur- valuable suggestions.

rent. Moderate errors (-20%) in either the calcu-
lated or measured corrugation amplitude have
little eff ect on our conclusions. Nevertheless,
since this is the first such calculation for STM, 'G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel,
we believe it would be premature to rule out a tip Appl. Phys. Lett. 40, 178 (1982).
consisting in effect of one or two atoms. For a 'G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel,
sufficiently small cluster of atoms, the effective Phys. Rev. Lett. 49, 57 (1982).
value of R depends on the precise geometry. 3G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel,
We conclude that a relatively simple model for Phys. Rev. Lett. 50, 120 (1983).
4G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel,
the tip, in conjunction with detailed calculations Bull. Am. Phys. Soc. 28, 461 (1983), and to be pub-
for the surface, gives excellent agreement with lished.
experimental results of STM and provides in- 'C. B. Duke, Tunneling in Solids (Academic, New
sight into the method's resolution and sensitivity. York, 1969).
With improved characterization of the probe tip 6H. Rohrer, private communication.
structure in the future, more precise compari- 7J. Bardeen, Phys. Rev. Lett. 6, 57 (1961).
sons will become possible, furthering the detailed I. K. Robinson, Phys. Rev. Lett. 50, 1145 (1983),
and references therein.
understanding of STM.
W. Moritz and D. Wolf, Surf. Sci. 88, L29-34 (1979);
We are grateful to H. Hohrer for making re- Young Kuk, Bull. Am. Phys. Soc. 28, 260 (1983), and
sults available prior to publication, and to him private communication.
and A. Baratoff for stimulating discussion and ' D. R. Hamann, Phys. Rev. Lett. 46, 1227 (1981).

2001