Sei sulla pagina 1di 6

This article was downloaded by: [Stanford University Libraries]

On: 09 August 2012, At: 19:42


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Philosophical Magazine
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tphm19

Electrical resistance of disordered one-dimensional


lattices
Rolf Landauer
a

IBM Thomas J. Watson Research Center, Yorktown Heights, N.Y.

Version of record first published: 13 Sep 2006

To cite this article: Rolf Landauer (1970): Electrical resistance of disordered one-dimensional lattices, Philosophical
Magazine, 21:172, 863-867
To link to this article: http://dx.doi.org/10.1080/14786437008238472

PLEASE SCROLL DOWN FOR ARTICLE


Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.

Electrical Resistance of Disordered One-dimensional Lattices


By ROLF LANDAUER
IBM Thomas J. Watson Research Center, Yorktown Heights, N.Y.
[Received 4 July 1969 and in final form 21 November 19691

Downloaded by [Stanford University Libraries] at 19:42 09 August 2012

ABSTRACT
T h e electrical resistance of a Fermi gas in a completely disordered
one-dimensional lattice is evaluated. The wavefunctions of the electrically
unperturbed lattice are used to evaluate a diffusion coefficient, and the
Einstein relation then leads to the resistance. An ensemble averaging process
is invoked, in which the diatancea between adjacent obatocles are varied independently of one anolher with an equal probability for all possible phase
rolationships between adjacent obstacles. The result is a resistance growing
exponentially, rather than linearly with the number of obstacles. This is
believed to be relatad to earlier results by others observing localizing wavefunctions due t o the prevalence of exponentially decaying wavefunctions.

THEelectronic states of disordered one-dimensional lattices have been


extensively studied. Recent reviews by Mott (1967) and Hori (1968)
survey this work, and demonstrate the predominant concern of the work
with the density of states. It is the purpose of this note to show that
it is possible in fact to calculate the average resistivity for an ensemble
of completely disordered arrays very easily. The method to be used is
derived from concepts originally explained by Landauer (1967), in a paper
treating the metallic residual resistivity problem from a viewpoint in
which the current flow into the specimen is taken as the causal agent and
the electric field is built up as the consequence of a continued flow of
charges against scattering centres. The duality of currents and fields as
source agents is, of course, very familiar in electrical engineering, but
has not become a prevailing viewpoint in transport theory. I n the
residual resistivity problem, Landauer found that his method yields an
electric field, whose space average equals that found through the usual
momentum space balance approach, but also that the field is in fact
spatially very non-uniform and concentrated about the scattering centres.
Erd6s (1965 a, b) extended the treatment of fluxes m causative agents, to
thermal transport problems. Landauer (1958, 1960) has refined and
extended his initial discussion of residual resistivity, and the present
calculation is in fact a simplification of one first cited in the latest of
these references. The calculation has, since its original appearance, also

Downloaded by [Stanford University Libraries] at 19:42 09 August 2012

864

Correspondence

been greatly refined and extended by Erdas (1968) and by Erdas and
Herndon (1969). The simplicity of our present approach still seems to
make this presentation worth while.
Consider electrons incident on an array of obstacles, and for conceptual
clarification assume that there is a space free of obstacles at each end
of the array. Let the reflection probability for the whole array be R .
Then for a carrier incident from the left, the relative particle density in
the free space to the left of the array will be 1 + R (after averaging over
a space of several wavelengths), since both the incident and the reflected
stream are present. To the right of the obstacles, the density will be
1 - R ; only the transmitted flux exists there. The density gradient
across an array of length L will be -2RIL. I n actuality, the gradient
is non-uniform with the wavefunction amplitude changes occurring in
the immediate vicinity of each obstacle, a point effectively made in the
closely related discussions of 8 2.3 of Motts (1967) review article. The
non-uniformity of the density gradient is, however, not important to us
in this particular discussion. We will, in a subsequent section,
return to the relationship of our analysis to Motts discussion. The
current, j, of particles associated with this gradient is v(1- R ) , where v
is the electron velocity in the absence of scatterers. Using the diffusion
equation, j= - D grad n , this situation leads to a diffusion coefficient :

vLI-R
2 R

D = --

. . . . . . . .

(1)

If we had, instead, taken a,particle incident from the right, we would


have obtained the same result. Therefore, any incoherent superposition
of these two states will also be characterized by this same diffusion
coefficient. An incoherent superposition can be presumed, coherence
between waves emanating from the two independent black-body
reservoirs at the two ends of the chain is not plausible. These reservoirs
exert a phase randomizing influence on waves incident on them, and thus
destroy phase memory before the particle can re-emerge from a black
body into the obstacle array. (Without this phase memory destruction,
our system would have no source of irreversibility, and the sign of the
diffusion coefficient would be undetermined.) It remains to be pointed
out that eqn. (1) gives the diffusion coefficient for a degenerate Fermi
gas, since the diffusion coefficient in that case is determined by the
behaviour at the Fermi surface. The diffusion coefficient can, in turn,
be used to find the resistivity through the Einstein relation :

in the form discussed by Kubo (1957), for example. I n eqn. ( 2 ) , p is


the chemical potential for a density of n carriers. The effects of array
disorder on
can be expected to be secondary, if we are far from a

Correspondence

865

zone boundary, and not changing the average potential in the ensemble
averaging process. The total resistance of the lattice, 52, is then :

. . . .

Downloaded by [Stanford University Libraries] at 19:42 09 August 2012

R = p L --2 -lRR($),/e2v.

The remainder of the discussion will be aimed at finding an ensemble


average of the factor R / ( l - R), occurring in eqn. (3). I n our ensemble
we will consider non-overlapping obstacles, and the distances between
adjacent obstacles will be varied independently of one another. The
chain, therefore, does not have a fixed length. We will also assume that
the phase shifts of the free-electron wave between adjacent obstacles
are varied uniformly over integral multiples of 27~. This is admittedly
a somewhat special ensemble, since the choice of the ensemble depends
upon the wavelength of the electrons in the interobstacle space, and
therefore upon the Fermi energy. At energies slightly below the Fermi
energy, for example, the uniformly varied phase shifts will cover a range
somewhat smaller than the integral multiple of 2n involved at the Fermi
level. (The author is indebted to the referee for pointing out this
limitation.) Alternatively we could have chosen a Gaussian probability
distribution, exp - (za/da),for the interobstacle distances, z. In the
limit as d becomes very large this Gaussian distribution also favours all
phases equally, and does so for all energies. Unfortunately in the process
of increasing d we are continually spreading out our obstacle chain, and
this is perhaps a less plausible ensemble than the uniform distribution
of phases.
Consider two adjacent obstacles, characterized respectively by reflection
probabilities (not wave amplitude coefficients) r1 and T 2 . Elementary
considerations lead to a characterization of the combined obstacles with
a reflection probability r, which obeys :
1
1 + r l r , + 2(rlr,)1laCOB 4
-- ,

1-r,

-rt- - rl

. . . . .

(4)

. . . . .

(5)

(l--r1)(1-rz)
ra 2 ( ~ ~ rcos~ 4) ~ ' ~

+ +

1-rt
(1 -r1)(1 -rz)
where is an angle which adds the phase shift between the obstacles and
also terms which depend on phase shifts associated with reflection and
transmission by the obstacles. If we now identify r1 with r, the reflection
probability of one additional obstacle added to a chain of (n - 1) obstacles,
and r2 with the remaining chain of (n- 1 ) obstacles, and average over a
range of interobstacle distances, and therefore 4, which lets the cosine
terms in eqns. (4) and (5) vanish, then we find :

866

Correspondence

Denoting (1-r,,)-l by x, and rn(1-rn)-l by yn we can write an


algebraically equivalent set to (6) and (7) :

2Zn- 1,

. . . . . . .

(8)

Y n = zyn-1+ ?Pn-1.

. . . . . . .

(9)

2, = yyII-l

These equations have already been averaged over the range of


possible spacings between the nth obstacle and the other (n-1)
obstacles. Averaging over the remaining phase differences yields a
~ ) the
true recursion relation between the vector ( ( z , , ) . ~( ~Y, ~ ) ~and
vector ((Zn-I)Av, (Yn-l)Av),
:
(~n)nv=~(~n-i)~v+~(~n-i)~v,

(Yn)Av=2(Yn-l)hv+Y(2n-l)nv.

Downloaded by [Stanford University Libraries] at 19:42 09 August 2012

. (10)
*

(11)

Such a recursion relationship can be solved by finding the characteristic


vectors of the transformation. The initial vector (zl, yl) must then be
decomposed into a sum of two characteristic vectors, each of which is
multiplied by the appropriate characteristic value, at every subsequent
step of the recursion. This yields :

The somewhat surprising element in this result is the exponential


growth, with n, of the resistance. An exponential growth of resistance
has, however, also been found by Erdos (1968), and by Erdos and Herndon
(1969) both in their analysis as well as in Monte-Carlo computer simulations
of specific disordered lattices. If we had taken a sequence of classical
obstacles, each of which simply intercepts a portion T , of the incident
beam, we would have instead obtained :
B/(l.-B)=nr/(l-~),

. . . . . .

(13)

leading to the intuitively expected linear growth of resistance with length.


We will not attempt to dwell on the physical interpretation of the
anomalous growth of reativity, except to point out the apparent relationship to the exponential decay of wavefunctions, in a disordered lattice,
discussed in connection with localized states, by Mott and Twose (1961),
and Borland (1961,1963). It is, however,worth while to note the analytical
relationship between Motts (1967) discussion, based on Borlands (1963)
analysis, and our present ensemble averaging. Mott, in his Q 2.3, points
out that the particle probability density increases by a factor 1 + $Oa at
each obstacle. O is a measure of the obstacle strength. Motts discussion
applies to weak delta function scatterers, and applies only for Oa Q 1.
In this limit, Motts factor 1 + *Oa is equal to the factor (1 + r ) / (1 - r ) in
our eqn. (12). Aside from the trivial distinction, arising from Motts
intentional tutorial simplifications (i.e. restriction to weak 6 functions),
the primary difference between our analysis and his arises from the fact

Downloaded by [Stanford University Libraries] at 19:42 09 August 2012

Correspondence

867

that we are concerned with a complex, current carrying state, whereas


Borland and Mott emphasize real wavefunctions. Mott, of course,
elsewhere in the same Q 2.3 considers complex wavefunctions and points
out that they also are characterized by exponential growth, and thus
makes all the basic physical points contained in our present ensemble
averaging calculation.
It must be borne in mind that an ensemble average of the resistance
can give undue weight t o a small proportion of ensembles with particularly
high resistance. If the ensemble were not a very dispersive one, and
consisted of very similarly behaving members then the exponential rise
in resistance would be mirrored by an exponential drop in conductance.
By illustrative contrast the integral for the ensemble average for the
conductance can actually be shown to diverge, for a chain of two or
more obstacles ; there are too many ensemble members which afford
close to complete transmission.

ACKNOWLEDGMENTS
The author is particularly indebted to G. Lasher who, in an earlier
version of the ensemble averaging process, provided the analytical
procedure to demonstrate the exponential behaviour of the resistance
growth. The author is also indebted to John M. Blatt, and to P. Erdos
for their serious and early interest, at a time when a widespread acceptance
for the exponential behaviour of wavefunctions in a disordered lattice had
not yet come into existence.

REFERENCES
BORLAND,
R. E., 1961, Proc. phys. Soc., 78,926 ; 1963, Proc. R . Soc. A, 274,529.
ERDOS,P., 1965 a, Phys. Rev., 138, A1200 ; 1965 b, Ibid., 139, A1249 ; 1968
(unpublished paper).
ERDOS,P., and HERNDON,
R. C., 1969, Bull. Am. phys. Soc., 14, 29.
HORI,J., 1968, Spectral Properties of Disordered Chains and Lattices (New York :
Pergamon Press).
KUBO,R., 1957, J . phys. SOC.Japan, 12, 570.
LANDAUER,
R., 1967, IBM J . Res. Dev., I, 223 ; 1958, Report to the Sponsors of
the International Conference of the Electronic Properties of Metals at Low
Temperatures (Geneva, N.Y.), p. 173 ; 1960, Talk at the Conference
on Statistical Mechanics and Irreversibility (Queen Mary College),
Localized Scattering, available as unpublished paper from the author.
MOTT,N. F., 1967, Adv. Phys., 16, 49.
MOTT,N. F., and TWOSE,
W. D., 1961, -4dv. Phys., 10, 107.

Potrebbero piacerti anche