PHYSICAL REVIEWB VOLUME 45, NUMBER 4 15JANUARY 1992-II
Potential
oscillations near
a
barrier
in
the
presence
of
phase-breaking
scattering
E.
Tekman andS.
CiraciDepartment
of
Physics, Bilkent University, Bilkent 06533, Ankara, Turkey(Received 31 July 1991;revised manuscript received 30September 1991)
Using the Green s function method for nonequilibrium processes, we study the potential oscillations near a barrier inboth coherent- and incoherent-transport regimes. In the fully coherent regime the local electrochemical potential oscillates near the barrier, due to interference ofthe incident and reflected
waves. The inclusion ofphase-breaking scattering leads tosuppression ofthese oscillations as a result of increasing contribution from the incoherent processes. As one goes away from the barrier, the
ampli-tude ofoscillations isfound todecay with a decay length equal to the phase-scattering length.
Structures across which the electron wave function re-tains its phase are commonly denoted as mesoscopic sys-tems. ' In recent years, experiments on electron waveguides and quantum-point contacts unambiguously showed that at such small length scales electron transport can thoroughly be understood only by use
of
the quantum-mechanical wave function and especially in-cluding the effectsof
its phase. In basic quantum mechanics a particle is assumed to perform wave motion throughout space without suffering from scattering events which randomly change its phase (i.e., phase-breaking scattering). Since the early daysof
studying mesoscopic systems it has been stressed that one has to go beyond this naive approach for a complete interpreta-tionof
the experimental data.To
this end, a phase-coherence length L&, which is the average distance trav-eled by electrons without having a phase-breaking scattering, has been introduced. Accordingly, the in-terference effects are present only when the sizeof
the de-vice is on the orderof
L&, but diminish as the size gets larger.The effects
of
phase-breaking scattering have been in-cluded in the ensemble-averaged properties using di-agrammatic techniques based on the Kubo formalism. The studyof
dynamics and kineticsof
the system in the presenceof
phase breaking, on the other hand, is only re-cent. Datta employed the Dyson equation in the Kel-dysh formalism to derive asteady-state quantum kinetic equation in the presenceof
inelastic scattering. His for-mulation is applicable to the phase-breaking scattering as well, for which energy transfer between the electron and its environment is not necessary. Resultsof
Datta showed that an excitation travels a distance on the orderof
L&before losing its phase information, indicating that the Keldysh formalism is appropriate for studying trans-port in the mesoscopic regime. D'Amato and Pastawski studied the effectsof
dephasing using the multiprobe gen-eralizationof
the Landauer formula as proposed by Buttiker. According to Buttiker's approach the voltage probes in the circuit are sourcesof
inelastic scattering since they are connected to macroscopic reservoirs and thus the incoming and outgoing fluxes are totally in-coherent. Reversing this statement one finds that inelas-tic scattering in a device can be modeled by using fictitious voltage probes. In fact, the equivalence
of
the mul-tiprobe approach to dephasing has recently been shown. In their work, D'Amato and Pastawski focused attention on the conductanceof
a disordered linear chain forwhich elastic scattering and localization are as important as in-elastic scattering. Recently Flores and Anda' also used the Keldysh formalism to study the conductanceof
a linear chain by including inelastic-scattering events.The subject matter
of
this paper is amore direct studyof
phase coherence in mesoscopic systems.It
isknown" that the local chemical potential oscillates near a barrier in the presenceof
transport due to interferenceof
in-cident and reflected waves.Biittiker"
examined the relevanceof
Landauer formulas by studying these poten-tial oscillations. Oscillationsof
the same nature were shown' to exist in narrow constrictions as well.It
was argued that the potential oscillations may also affect the noninvasive measurements by the scanning-tunneling mi-croscope. ' In this paper we study the effectof
phase-breaking scattering on the potential oscillations. First, we apply the Keldysh formulation for nonequilibrium processes to calculate carrier density and current in a mesoscopic system. We use a strictly one-dimensional (1D) model and a local self-energy operator, and find that the chemical potential oscillations are suppressed as aresultof
phase-breaking scattering. Moreover, weshow that the strengthof
such interference effects decay as exp(I/L&),
whe—re lis the total distance to be traveled by the electron wave in orderto
give rise to interference (i.e.,twice the distance from the pointof
measurement to the barrier forthe present problem).We first follow the approach developed by Datta, and westart with the Dyson equation in the Keldysh formula-tion, ' in terms
of
the Hamiltonianof
the noninteractingsystem Ho, Green's function G, and self-energy X ma-trices. Assuming a steady-state condition, the quantum-kinetic equation forthe Green's function
[E
Ho(r)]G(r, r',
E)—
=5(r
—
r')I+
f
dr"X(r,
r";E)G(r",r';E)
(1)is obtained from the Fourier transform
of
the Dyson equation. Furthermore, this differential equation can be transformed into an integral equation1920 BRIEFREPORTS 45
G(r,
r', E)=GO(r,
r';E)+
J
dr"dr"'Go(r,
r";E)
XX(r",
r'";E)G(r"',
E)
(2) by using the Green's function for the noninteracting system Go(r,r',
E),
which satisfies the equation(E
—
Ho)GO=I.
The electron densityn(r;E),
hole densi-ty p(r;
E
),and current densityj
(r;E
) (all per unit ener-gy), in turn, can be found in termsof
the Green's func-tion. Detailsof
the derivation are the same as those found inRef. 4. For
an arbitrary self-energy function,Eq.
(2) is difficult to solve and does not have any advan-tages over the quantum-kinetic equationEq.
(I).
Howev-er, with the following assumptions it can be simplified to agreat extent. First, we restrict our attention to a strict-ly 1D system. Second, we include the elastic-scattering events into X together with phase-breaking scattering. This wayGo(x,
x',
E)
can be calculated directly from a1D free-electron approximation. While doing so we as-sume that the device is connected
to
two reservoirs atx=+Oo
having chemical potentials pL andpz
for the left- and right-hand-side reservoir, respectively. Clearly,we do not impose any boundary conditions on the Green's function on a boundary specified apriori. This is consistent with the open nature
of
the system at hand. ' That is, the system is driven by external agents and the responseof
the system to this external excitation can be found only after completely solving the problem. Lastly we assume that scattering events (both elastic and phase breaking) take place at discrete and uncorrelated scatter-ing center.To
this end, one can writeX(x,
x';E)=5(x
—
x')
g
X;(E)5(x
—x;),
(3)as
where
X;(E)
denotes the self-energy contribution by the ith scatterer located atx;.
Substituting
Eq.
(3) intoEq.
(2) one can solve G in termsof
Gp. Then, the electron density and current den-sity are calculated in termsof
the noninteracting Green's function and the reducible self-energy matrixX„(x,
x';E)
=
g
[X„(E)];J5(x
—
x;}5(x'
—
xj
)n(x;E)=no(x;E)+
—
1 Img
Go(x,
x;;E)[X„"(E)]~iGO
(x~,x;E)
l,J
g
Go
(x,
x;;E)[X„(E)];,
Go"(x,
,x;E),
2
(4a) fi
dG"
(x,
x;;E
}j(x;E)=jo(x;E)
—
Reg
[X„(E)]JGO
(x,
x;E)
2am Bx
1,J
Re
g
G"
(x,
x;;E)[[X,(E}];
BG"(
x,
x;E)IB
x+[
X"„(E)];
BG(x,
x;E)IBx]
27Tm
7
(4b)
Here superscripts
(,
),
A, and Rstand for the electron, hole, advanced, and retarded functions, respectively. Note that, the reducible self-energy matrix satisfies the Dyson equationX,
=X+XGpX„.
In order to study potential oscillations in the presence
of
phase-breaking scattering, we assume that X;of
all the scatterers has a phase-breaking part, but only self-energiesof
the scatterers lying in the barrier region have an elastic-scattering part. The phase-breaking part is given byX;
(E)=[ihlr;(E)]p(x;;E},
X;
(E)=[
iklr;(E)]n—
(x;;E)
.
(5a) (5b) As shown in
Ref.
4 these self-energy functions are con-sistent with the golden-rule expression for a systemof
electrons interacting with a reservoir
of
oscillators at thermodynamic equilibrium, with the restriction that the scattering is not inelastic (i.e., the energyof
the electron is conserved). In addition, one can show that the self-energies given byEq.
(5) yield a conserved current throughout the device. This point can also be reached starting from the multiprobe approach and yields %'ard identities. Note that Eq. (5) has a simple physical inter-pretation as well. Takingr,
(E)
as the average time for an electron or a holeof
energyE
to suffer a phase-breaking scattering at site i,assuming that the state(elec-I
tron or hole state) isinitially filled, X;
(E)
andX;
(E
)be-come the average times between these scattering events including the effects
of
exclusion. One other important point tonotice isthat the self-energy functions depend on the electron and hole densities. Therefore, Eqs. (4) and (5) have to be solved self-consistently. The elastic partof
the self-energies, on the other hand, is given by
o;(E)I,
and is independentof
the carrier densities.Finally, we define a local electrochemical potential within the linear-response approximation,
i.
e., by neglect-ing all the energy dependences and calculating everything at the Fermi levelEF.
To
this end we consider a reser-voir at thermodynamic equilibrium with chemical poten-tialp;,
connected to the ith scattering site by an ideal wire ' (i.e., without any internal structure ). Assumingthat pL
)
p~ one finds that the current into the reservoir is proportional to (pL—
p, )n(x,;EF),
and the current outof
the reservoir isproportional to(p,
—
p~ )p(x,;E„),
the proportionality constants being the same. Thus, in order to have zero net current into the reservoir its chemical potential has tobe given bypL
n(x,
;EF
)+pzp(x, ;EF
) Pi n(x,;EF
)+p(x,
;
EF
)This definition
of p;
is reminiscentof
the counting argu-mentof
Landauer. ' However, in the present study theEP«TS
1921p.p
—10.p —5.p O.p
I
5.p
FIG.
1y
ariation ofthe lo potentialp«f,
lf
lectrpchem; 0.1;c,05'
y 1. /L~ values pf a Q. e,2. Dotted . , 'b, position pfthe lines indicate th e Potential barr et e d g.=0.
06Xf'k
rri«with k1=2
~ . F/m.»ce
is kFL=2Q.
elength ofthe d 5kF. The curve &o scatterers larity. Inset h es are offset byQ or e, the crpss-shaded o& ows the
struc-place.
p e-breaking scattering tages
-1.
0 -10.0 0.0 krX 10.0 0.00: 'a 0 ~W V Q —0.05 O —0.10 0.00 0.05 0.10FIG.
2. XF/L„ED(x'E
) co(a) Percent change in the
ffset by 1. (b) Deca e p i ation ampli-various l a ering stren th F 1 d L T F
he solidi linine denotes k,x
carrier enslties~ ~ are det mechanical cal
etermined from ties i, i6
"'"
ons and hav
m ull
~~~~tu~
As shown b ave interference correctedt
y Buttiker, 'E
e proper-1 d th o be e gth is short r than orhN
hatp;
is not areal 5.0 at 'electrochemi-(a)
3 0R
bV V
10.pcalpotential since itdoes no
h d 1 tion
s.
ynamic e u' ' At h'',
i quilibrium condi-t is point, it is in order p d 1i0
w mtegral equationE
reen's functio.'on. This pr 'd equation d a set
of
matrix't
t''G
g reen's funct'
ducible 1f- tio
.
Thtnes and nurnericall
h
y e in this
cur ent throu h t d cessary to intr unction
6,
so d tothe 'on, t eartifa uristic extern p e eliminate th p e modelf
ns
ite within thec
rrent sources ms may be em-e probem-es is m
e,
we believe raey.
an has to be ' The local electr
h e investi-work po eectrochemical
as
method have d'ff defines the 1 1 1 hern1odyna 1od
amc
e uili th loca1electrochem' 1 there is no p he present definition m d 1f
'
'"-'""""
e noninvasive The m o1to
d'
binding mod ores and Anlb
of
dff
equilibrium Green' two system a ove hs.
Ino
at there , asnon-1922 BRIEFREPORTS 45
equilibrium distribution
of
carriers for the noninteracting system and we calculate Go accordingly. The multiprobe approaches, ' on the other hand, are aiming only at thecalculation
of
the conductance with the helpof
artificial dephasing probes. Thus, they may not be appropriate for more detailed analyses. More recently, McLennan and co-workers' also used the formalism developed by Datta tostudy voltage drop in mesoscopic systems. Their work and ours''
independently arrived at the same con-clusion that the phase-breaking scattering leads to decay-ing potential oscillations going away from an obstacle.We applied the present method to the structure shown in the inset to
Fig. 1.
This structure is characterized by three parameters: the phase-breaking time ~, elastic-scattering strength o (both are calculated atEF
and are the same for all scatterers), and the densityof
scatterersp„.
The barrier extends from—
I/2
to+1/2,
and the phase-breaking scatterers are uniformly distributed be-tweenL
/2 a—nd+L/2.
The phase-breaking length L& is given byL~=A'kF&/m
p„)
(7)which is the Fermi velocity times the phase-breaking time averaged over the unit length. In Fig. 1 the local
electrochemical potential profile is shown for the varying
L/L&
ratio. Clearly, forL&~
~
one obtains the phase-coherentresult"
and potential oscillations have the same oscillation amplitude independent from the distance to the barrier. In the other extreme,i.
e.
,forL&((L,
phase coherence is lost and electrons act as classical particles. Thus, outof
the barrier regionp;
varies linearly with po-sition. In the intermediate regime potential oscillations are still present, but their amplitudes get smaller as they go away from the barrier. That is, a transition from the classical regime to the quantum regime takes place as afunction
of L/L&.
One can show by using Eqs. (3)and (2) that the periodof
oscillations iskF/2
provided that Lp»kF
We develop a more quantitative approach to these de-caying oscillations by considering the total density
of
states per unit energy,D(x;E)=n(x;E)+p(x;E)
. (8)Note that in the absence
of
the barrier and phase-breaking scattering,D(x;EF)
isjust equal to the densityof
states Do(EF)=m/vrA kF. In the limit L&«Az,
on the other handD(x;EF)
satisfies a diffusion equation, and hence exponentially decays away from the barrier. The percent change in the densityof
statesb
D(x;EF
)=
[D(x;EF
)Do(E—
F)]/Do(EF
),
corresponding to the potential profiles in Fig. 1,isshown in Fig. 2(a). The envelope
of
bD(x;E~)
may be com-pared to the retarded Green's function as found by Dat-ta, since it represents the propagationof
the excitation created by the barrier. In order to analyze the interfer-ence effects away from the barrier, inFig.
2(b) decayof
magnitude
of
AD(x;EF)
per period is shown as a func-tionof
A,z/L&.
Clearly, L&dependenceof
decay isgiven by exp( A,~/L&).
—
This is due to the fact that in orderfor incident and
rejected
waves to interfere a distance d away from the barrier, the wave has totravel atotal dis-tanceof
2d without suffering aphase-breaking scattering. Consequently, the strengthof
the interference decays as exp(—
2d /L& ). This result verifies the phenomenological exponential dependenceof
the phase-coherence effects on L&, and corroborates that a diffusion approach is ap-propriate for studying mesoscopic systems with phase-breaking scattering.Forareview, see Nanostructure Physics and Fabrication, edited
by M.A.Reed and W.P.Kirk (Academic, New York, 1989)~
B.
L.Al'tshuler, A.G.Aronov, and D.E.
Khmel'nitskii, Solid State Commun. 39, 619 (1981);J.
Phys. C 15, 7367 (1982). 3P.A.Leeand A.D.Stone, Phys. Rev.Lett. 55, 1622(1985);B.
L.Al'tshuler and
B.
I.
Shklovskii, Zh. Eksp. Teor. Fiz. 91, 220 (1986) [Sov. Phys. JETP64, 127(1986)];P.A. Lee,A.D. Stone, and H.Fukuyama, Phys. Rev.B35,1039(1986). 4S. Datta, Phys. Rev. B 40, 5830 (1989);J.
Phys. Condens.Matter 2,8023(1990)~
5L.V.Keldysh, Zh. Eksp. Tear. Fiz.47, 1515(1964) [Sov. Phys. JETP20, 1018(1965)].We use the notation of
G.
D. Mahan, Phys. Rep. 145,251(1987)~J.
L.D'Amato and H. M. Pastawski, Phys. Rev. B41, 7411 (1990).7M. Buttiker, Phys. Rev.Lett. 57,1761(1986).
~H. L. Engquist and P.W. Anderson, Phys. Rev. B24, 1151 (1981);M.Buttiker, ibid. 33,3020(1986).
S.Hersh6eld, Phys. Rev.B43,11586(1991).
'
F.
Flores andE.
V,Anda (unpublished)."M.
Buttiker, Phys. Rev.B40, 3049(1989).P. L.Pernas, A.Martin-Rodero, and
F.
Flores, Phys. Rev. B 41,8553(1990).'
J.
R.
Kirtley, S.Washburn, and M.J.
Brady, IBMJ.
Res. De-velop. 32, 414 (1988); C. S.Chu andR.
S.Sorbello, Phys. Rev. B42, 4928(1990).'
R.
Landauer, Z. Phys. B 68, 217 (1987), and references therein.'~This has not been taken as a proper multiprobe measurement,
since the inclusion ofthe ideal wire is assumed not todisturb
the quantum-mechanical properties ofthe system.
Neverthe-less, some aspects ofinvasive measurements can beanalyzed using such anexternal probe (Ref.17).
M.Buttiker, IBM
J.
Res.Develop. 32, 317 (1988).E.
Tekman, PhD. thesis, Bilkent University, 1990.M.McLennan, Y.Lee,and S.Datta, Phys. Rev. B 43, 13 846 (1991).