Ballistic
transport
through
a
quantum
point
contact: Elastic
scattering
by
impurities
E.
Tekman andS.
CiraciDepartment
of
Physics, Bilkent University, Bilkent, 06533Ankara, Turkey(Received 5March 1990;revised manuscript received 15 May 1990)
The effects ofelastic scattering due to impurities in a quasi-one-dimensional constriction are
in-vestigated with an exact calculation ofthe conductance. It isfound that the quantization of con-ductance is distorted owing toscattering by asingle impurity which exists inan infinite constriction. The extent ofdeviation from quantized values depends on the strength, position, and lateral range ofthe scattering potential. The resonance structure due tointerference ofcurrent-carrying waves is still apparent for a constriction offinite length containing an impurity. However, both the
magni-tude and position ofthese oscillations in the resonance structure are affected as a result ofelastic scattering. A resonant tunneling effect isfound due to a state bound tothe attractive impurity po-tential.
I.
INTRODUCTIONUsing high-mobility two-dimensional (2D) electron gas (EG) and split-gate structures, van Wees et al.' and Wharam et al. fabricated quantum point contacts (QPC) with length scales smaller than the electron mean free path and comparable with the Fermi wavelength A,
F.
They observed that the two-terminal conductance
of
the QPC is quantized in unitsof
2e /h as a functionof
the widthof
the constriction m. Recently, assuming that the transport is ballistic, several groups developed theories to explain the quantizationof
conductance. Furthermore, they predicted resonances superimposed on the quantized plateaus. The deviations from exact quant-ization and the lackof
the resonance structure in the ex-perimental results' have become important issues, and were attributed to various effects.For
example, Glazman et a1. showed that the current-carrying states evolve adi-abatically without reflection and without intersubband scattering in certain hornlike connections to the reser-voirs. The authors, on the other hand, showed that sharply quantized conductance G (w)devoidof
resonancestructure can occur even ifthe adiabaticity requirements are not satisfied for certain QPC geometries. The elastic scattering by impurities in a ballistic channel can also affect the above-mentioned quantization
of
conductance and the resonance structure. Earlier works ' ' haveindi-cated such a possibility, but a clear understanding
of
im-purity effects on experimentally relevant systems is not fully developed yet. Therefore, scattering by an impurity in a QPC has several interesting features which deserve further study.In this paper we investigate the effects
of
elastic scattering by an impurity in a ballistic channel. Using a Green's-function technique, we obtain the expression for conductance for an infinite quasi-1D constriction with a single impurity represented by a model potential. The formof
the model potential is realistic and enables us to obtain exact solutions for scattering events. Moreover, it is appropriate to carry out a systematic analysis on theeffects
of
the position and lateral extentof
the impurity. The formalism developed for an infinite constriction is further extended to treat a finite-length QPC with asingle impurity. Our results are in overall agreement with the resultsof
the earlier studies, ' ' which were obtainedby using completely different approaches. Present study in-vestigates several aspects
of
scattering by an impurity in a QPC (which were not treated earlier) by using more real-istic scattering potentials and boundary conditions. InSec.
II
we describe the methodof
calculation and intro-duce the model potential. A critical comparisonof
our method with the earlier ones isalso presented in this sec-tion. InSec.
III
we present the results obtained by using this formalism for the infinite and finite constrictions and discuss the similarities and differences with thoseof
the earlier ones. Important aspectsof
our study are stressed by wayof
conclusions in Sec.IV.
II.
METHODWe first consider an infinite constriction, for which z is the propagation direction and y is the transverse direc-tion as described by the inset in
Fig. 1.
We also assume that the confinement in thex
direction is complete. The eigenstates for such a uniform quasi-1D constriction (electron wave guide) in the presenceof
a scattering po-tential ut(y, z)canbe written asg,
(y,z)=e
'
P(y)+
J
dy'Jdz'g(z
—
z',
y,y')
Xut(y',
z')@ (y',z')
. The first term on the right-hand side represents the in-cident wave, which is the unperturbed solution for thejth
subband with the wave function P (y), the eigenenergyc,
and the corresponding wave vectory~
=2m*(E
—
EJ)/A' along the z direction. Details forthe unperturbed solutions for the current-carrying states and the variation
of
conductance calculated thereof for the uniform and tapered quasi-1D constrictions can be found in Refs. 5 and7.
The above expression inEq.
(1)iswhere
T„(k)
is given by(2)
T„(k)=V„(k
—
y )+
g
f
dk'V„(k
—
k')G (k')T
(k')
. (3) Note that the Fourier transformsof
g, vi, and t are ma-trices Cx (diagonal), V, andT,
respectively. An elementof
such matrices are calculated from the integral de-scribed by the following expression:F„(k)=
f
dy t))„(y)P (y)f
dz e'"'f
(y, z) . (4) By solving Eq. (3)forT,
one obtains the solution for the the well-known Lippmann-Schwinger equation adapted to quasi-1D systems with the retarded Green's functiong.
The exact solutionof
Eq.(I)
can be written using the t operator asg,
(y, z)=e
'
P,
(y)+
g
P„(y)
f
dke'"'G„(k)T„(k),
scattering problem for a right-going incident wave in the
jth
subband. The solutionP,
(y,z)isfound similarly for a left-going incident wave.It
isimportant to note that in the present study we cal-culate the conductanceof
the constriction by using a two-terminal geometry. That is, two reservoirs are con-nected to the endsof
the constriction (or the 2DEG
for the finite constriction) so that the voltage difference be-tween the reservoirs isjust the differenceof
the electro-chemical potential deep in the reservoirs (which is taken to be infinitesimal). The conductance G„(
w)of
an infinite constriction is then calculated from the expecta-tion valueof
the momentum operator,2 OCC
G„(to)=
g
(g,
lP,l (l,)
.j
7JThe solution
of
Eq. (3)for a general potential ui(y, z) is complicated and may require extensive computations. In order to obtain an analytical solution which leads to a clear pictureof
the effectsof
elastic scattering, we use the following model potential for a scat terer located at (yi,z,
):2
UI(y,
z)=
exp(—
qly—
yil)5(z—
zi),
m*
(6)//I/ /////r /////I//3 ///// / / / /////Ir/ j/
V=Q Z X(y,z ) 12- //rl/rr////r/I /////// rr////// /rr///r y)
-0
q =10&F 6 -II3I 0.2f
p4
0.6 cv p60-
1.0 I/I =0.6kF 8 - q =10XF8-
yI05
I 0./. a33 0.25 0Wl
AFFIG.
1. The conductanceG„vs
the width w ofan infiniteconstriction containing an impurity. (a)yl
=0
and q =10k,~',the strength pl varying (in units of kF). (b) rpr
=
0. 6kr andq
=
10XF', the position y, varying (in units of A,F). Solid(dashed) curves correspond to repulsive (attractive) impurities,
and are vertically offset by an amount 1.5X(2e /h) for clarity. The geometry ofthe channel isdescribed by the inset.
which is aDirac 5 function in the zdirection, and has the exponentially decaying form in the y direction with a de-cay length
of
q'.
The strengthof
this potential isset bythe magnitude
of P,
which may be both attractive(P(0)
and repulsive(P)
0). For
this formof
the potential, Eq. (3)isexactly solvable and theT
matrix is given by277
where
Q=u(I'+iu)
'I
withf',
=$,
y,
andu,&=13
f
dy p;(y)pj(y)exp( qly yil ) .The conductance for an infinite constriction containing an elastic scatterer as described in
Eq.
(6)is expressed in termsof
these matrices asIm[(Q),
]
Re[(Q I
'Q).
,
]
G„=
—g
I+2
'
+
h y XJ (c &E~.j j (9)It
should benoted that the effectof
the evanescent waves with c,)
Ez
is included in the above formalismof
G„.
This is provided through the intersubband coupling in
6
and yields novel effects described inSec.
III.
These effects do not exist in strictly 1Dsystems.To
calculate the conductance for aQPCof
finite length d, we furthermore assume that the impurity potentialUr(y,z) is zero outside the QPC region
O~z
~d.
Thus, the solutionof
the Schrodinger equation in the 2DEG
(z
~
0
and z~
d) is a linear combinationof
plane waves, each plane wave being a solutionof
the 2DEG
reser-voirs. This assumption simplifies the solution since elas-tic scattering takes place only in the constriction, andthus the use
of
Green's function in the 2DEG
is not necessary. Note that the inodel potential in Eq. (6) satisfies this conditionif
the impurity is located in the constriction (i.e.,0
zI1).
The solution in the constric-tion isexpressed in termsof
PJ andg
as%'(y,
z)=
g
[A,
g,
(y,z)+B
g,
(y,z)]
.J
(10)
The boundary conditions at
z=0
and z=d
are used to find the coefficients A andB.
Next we express tp(y,z)in termsof
a linear combinationof
exponentials either for0&z
&zI orz&&z&d
asG~=
I
dg [(Q~Re{I
IQ~—
4
Re{I I4)
irh
-iF
k,
(~)+2Im(O Im{I
I4)]
. (12)In Eq.(12) the coefficients Qand
4
depend on the param-etersof
the impurity, namely0
andzI,
as well as the pa-rametersof
the constriction. In the numerical studies presented inSec.
III
we used an infinite-well confinement in the transverse direction. Nevertheless, Eqs. (1)and (3) have general validity, and Eqs. (9) and (12) are valid for the impurity potential given by Eq. (6).At this point it is in order to compare our model with the earlier ones. ' ' Haanapel and van der Marel used
the tight-binding method to analyze the effects
of
an im-purity in or near the constriction for short QPC's. They argued that the presenceof
the impurity in or near the constriction prevents the quantizationof
conductance. However, since the potentialof
the impurity was taken as a 2D Dirac 5function, their study was not able to reveal the scattering effects in detail. Recently Chu and Sorbel-lo calculated the conductanceof
an infinite constriction in the presenceof
an impurity by using a scattering theoretical formulation. They provided an exact analytic solution for the conductance in termsof
phase shifts pointing out interesting featuresof
impurity scattering. However, the applicabilityof
their analysis made by the isotropic (s-like) scatterer in an infinite wave guide is lim-ited for an experimentally relevant system. Masek and co-workers employed the Anderson model to analyze the conductanceof
a disordered quasi-1D conductor. Their results may be significant forensemble-averaged effectsof
impurities. Although their results are in agreement with those obtained by other methods, the microscopic aspects
of
scattering due to a single impurity cannot beextracted from that study. The present model provides exact and partly analytical solution for the conductance. As seen, the formof
the potential and thus the formalism is versa-tile and enables us to study various parameters such as the position, lateral extent, and strengthof
the impurity. The weaknessof
the model potential used in this study is that itis highly anisotropic. Consequently, adirect quan-titative comparison with the experimental systems may'p(y,
z)=
gp,
(y)(e'
8,
+e
'
4,
) . JFinally, the conductance Gz(w) is expressed in terms
of
these vectors
of
coefficients Q and4
as described else-where:not be straightforward.
Finally, we comment on the effects
of
the self-consistent potential and inelastic scattering. Earlier, Lan-dauer' argued that self-consistent charge due to none-quilibrium electrons is accumulated near the impurity, which yields corrections to the conductance. This is closely related to the questionof
which Landauer formu-la, G—
T
or G—
T/8
(T
and R being transmission and reflection probabilities, respectively), has to be applied. An extensive discussionof
this issue is beyond the scopeof
our work, however. Relevant references, which present comprehensive reviewsof
several efforts and de-bates, are given inRef.
10.
It
becomes clear now that a different Landauer formula applies to different measure-ment geometry.For
the system we are considering, the voltage difference is measured between the reservoirs. As stated above, this is a two-terminal geometry. That is, expressing in termsof
relevant Landauer formula, ' the conductance is given by G—
T
[for multichannelcase"
Tr(t t
),t
being the transmission matrix]. In the present approach the finite temperature effects are also neglected. Despite this, we think that progress towards a better un-derstandingof
elastic scattering in a ballistic channel is made by the present work. Moreover, our findings have close bearings to the resonant tunneling, especially in scanning tunneling microscopy.III.
RESULTS AND DISCUSSION A. Infinite constrictionThe variation
of
G„(w)
for an infinite constriction having a single impurity is shown in Figs. 1 and2.
As seen, the ideal quantization isdistorted in the presenceof
the scatterer.
If
the potentialof
the impurity is weak [e.g., ~p~~
0.
5kF for q=10AF
' inFig.
1(a)],G„(w)
stillreflects a staircase structure with smoothed steps and with plateaus very close to the quantized values, 2e N,/h (N, being the number
of
subbands belowEz).
Another observation is that for weak scatterers the signof
the po-tential does not have a pronounced effect on the conduc-tance. This result is in compliance with the first-order Born approximation, since the lowest-order correction to the conductance is proportional top
in the perturbative treatmentof
the impurity. Therefore, both repulsive(p&0)
and attractive(p(0)
impurities have the same effects on the transport. In order for the Born approxi-mation to be valid, and thus for only a single scattering event to take place, the velocity or equivalently the wave vectorof
electrons has to be large. In the quasi-1D sys-tem under investigation the related wave vector is the propagation constant y and is equal to zero whenever a new subband dips the Fermi leve1,i.
e., m=
Ã,
A,F/2.
Thus, the Born approximation fails for m values just above
X,
XF/2 and it isnecessary toinclude the multiple scattering events.For
relatively stronger impurities[0.
5k+5
~p~~
kz forq
=
10k,F' inFig.
1(a)]not only the steps are smoothed, but also the plateaus exhibit deviations from the quan-tized values 2eX,
/h. The most remarkable effect ob-served in this rangeof p
is the difference between theat-12 y(
-0
q-K„
0.24-06
r—
0.8 0'0
I/I
=0.6kF q =KF 8 8 -~y 0.6 0.~ 0.33r
-0.25 / FFIG.
2. The conductanceG„vs
the width m ofan infiniteconstriction containing an impurity. (a)y,
=0
and q=kF,
the strength ~P~ varying (in units of kF). (b) ~P~=0.
6kF andq
=
A,F,
the position yz varying (in units ofA,F). Solid (dashed)curves correspond to repulsive (attractive) impurities, and are vertically offset by an amount 1.5X (2e /h) for clarity.
tractive
(p(0)
and repulsive(p)0)
scatters. Also for strong scatterers the Born approximation begins to fail for the whole rangeof
w. As seen, for attractive impuri-ties the dips in the G versus w curves appear below the steps. The conductances at these minima are approxi-mately equal to 2e (N,—
I)lh
for laterally confined im-purities (i.e.,large q),and there are sharp rises to the next quantized value above these dips. One important point we notice is that dips do not occur below allof
the steps.To
analyze this, we calculatedG„(w)
for different trans-verse positions (yr)of
the impurity as shown inFig.
1(b), and consider6"
given byEq.
(8).For
a laterally confined impurity (i.e.,large q),6
is approximately proportional to ~PJ(yl)~ .To
a first approximation, the effectof
the impurity is large on the N,th plateau when ~PN (yI)~ is aC
maximum, but is small when it is negligible.
For
exam-ple, for yl=0
the deviations from the quantized values will be large on the odd-numbered plateaus and small on the even-numbered ones. On the other hand, the size, width, and existenceof
the dips below theX,
th step for the attractive impurities' are determined by magnitudesof 6;z
for i(X,
.
Analyzing these dips in detail we findC
that they originate from the enhancement
of
backscatter-ing due to the intersubband scattering. 'For
the strictly 1D problem total backscattering is not allowed since the boundary condition at z=zI
for the derivativeof
the wave function cannot be satisfied.For
the quasi-1D casethere are subbands which may be coupled in the presence
of
the impurity. Therefore, the total backscattering can occur in a subband by inclusionof
the evanescent states in aquasi-1D system. Since the first-order Born approxi-mation employs the equivalent 1Dproblem for each sub-band, the dips cannot be obtained perturbatively. The backscattering effect isvisible in Fig. 2(a).For
an impuri-ty positioned at the centerof
the channel, even- and odd-numbered subbands are completely decoupled. There-fore, there is no dip below the second step, and the dip below the third step is due toenhanced backscattering in the first subband caused by the evanescent third-subband state. In the presenceof
a large numberof
impurities, all the subbands are mixed and it is possible to observe dips below allof
the steps.For
laterally spread impurity potentials with small q the deviations from the quantized steps [seeFig.
2(a)]are enhanced compared to those with large q.For
example, the dips do not have conductance 2e (N,—
1).
This is due to the large integrated strength-plq.
Note that6
given in Eq. (8)is determined by this integrated strength and not solely by the strengthp.
Another observation is that for attractive impurity potentials the dips are shifted to valuesof
w which are smaller than N,A.Fl2
and appeartogether with peaks. Since the impurity potential influences a wide range
of
the constriction, the wave function evaluated at yl cannot give an idea about the effectof
the scatterers. Although the deviations from the quantized plateaus vary with yI[Fig.
2(b)], this effect is not asdrastic as it was forlarge q.Comparing these results with those obtained by Chu and Sorbello and Masek et
al.
, it isconcluded that the present model potential is more appropriate to analyze the transport in a ballistic channel with a single impurity. Although the dips were also found by those authors, 'all
of
the steps were alike in the results given by Chu and Sorbello since the positionof
the scatterer is chosento
yield coupling
of
allof
the subbands. In additionto
that, their approach does not allow one to vary the strength and the integrated strength independently. Therefore the results presented inFig.
2 are unique to the present study. Another important advantageof
the present ap-proach is that it enables the controlof
the intersubband coupling.For
large valuesof
q the scatterer looks like a 5function, which enhances the intersubband interactions. In contrast, the potential becomes flat in the lateral direc-tion and the intersubband interaction vanishes for small valuesof
q. In this case the dips disappear.B.
Finite constrictionHaving discussed elastic scattering due to a single im-purity in an infinite constriction, we next consider the sit-uation in a QPC
of
finite length d. Using the formalism described in Sec.II,
we calculated the conductance Gd(w). The results are summarized inFig. 3.
As for the impurity-free constriction, the main effectof
finite length is to smooth out the sharp changes in G„(
w) (or its first derivative) due to inclusionof
evanescent states. This effect isof
major importance for short constrictions (d ~A,z).
For
longer constrictions the effectof
evanes-
d-J///Jd/////s rf y yr X /&/ / //// ///'/ / /'/' i 2 DEG; ', 2DEG p =0.6 kF 1 yI d 2 1 Lll 0 P=-06
q=
6-
y= 1cent states decreases, but a new feature due to interfer-ence
of
left- and right-going waves arises, namely the res-onance structure. Since the effectsof
only elastic scatter-ing by asingle impurity are taken into account, neither a phase breaking due to an inelastic event nor a phase averaging due to a large numberof
scatterers can take place. In other words, the system we are investigating here is the quasiballistic regime, which still contains well-defined interference effects leading to the resonance structure inFig. 3.
The dramatic effectof
the impurity is revealed by comparing conductancesof
finite (neglecting the contribution due to tunneling) and infinite constric-tions.For
an impurity-free channel the conductanceof
the finite constriction is smaller than that
of
an infinite constriction (i.e., smaller than the ideal quantized steps)for all w. In contrast, for aconstriction with asingle im-purity Gd(ttr) may be larger than
G„(w).
This isa resultof
the combined scattering from the impurity and the endsof
the constriction(z=O
and z=d).
That is, scattering from the ends may depress the effectof
scatter-ing by the impurity.Clearly the main features
of
Gd(ttr) shown inFig.
3,in particular the heights and positionsof
the resonances and antiresonances, are strongly dependent on the positionof
the impurity along the z direction. That is, moving the impurity along the channel will give rise to oscillations in the conductance. The magnitude and period
of
the oscil-lations are related tothe length and widthof
the constric-tion, as well as the propertiesof
the impurity. A similar effect is observed by moving defects in a metallic nano-constriction. ' In Fig. 4 the resistanceof
a typical QPC is shown when an impurity is present in the constriction. Clearly, for large q the deviation from the quantizedvalues is approximately constant for a given
X,
and de-creases with increasing N,.
This result closely resembles the experitnental observationof
Wharam etal.
' These examples show that it is possible to observe the effectsof
elastic scattering in the channel. However, additional ex-perimental studies are still needed
to
fully exploit this conclusion.Finally, we wish to point out a novel feature
of
attrac-tive impurities.For
short constrictions with an attractive impurity placed near their center (zr=d/2)
the conduc-tance curve Gd(ttr) has sharp peaks just below the steps[Fig.
3(b)]. The widthsof
these peaks decrease with in-creasing d, and for very long constrictions the peaks can-not even be resolved. Moving the impurity away from the centerof
the constriction (by changing either the po-sitionof
the impurity z~ or the lengthof
the constrictiond) has the same effect. Similarly increasing the strength or integrated strength
of
the impurity causes the peaks to shift the lower w values. A detailed analysisof
these re-sults shows that these peaks are associated with resonant tunneling through quasi-OD states bound to the impurity. The propertiesof
this resonant-tunneling effect are analo-gous to those obtainable from the double-barrier reso-nance tunneling structures. Hence, similar to formationof
quasi-OD states due to geometrical effects (local wideningof
the constriction) in an impurity-free ballistic channel, it is possible toobtain bound states in a constric-tion in the presenceof
an attractive impurity potential. A final remark about these resonances is that the peaks2-2
~
06
0 C CL0.
2/=06
kF. y)=0 d-PF
z(-0 2PF q =10FIG.
3. The conductance Gd vsthe width wofafinite lengthconstriction for q
=
El.-',yi
=0,
and (a)/
=0.
6k~, (b)P=
—
0.6kF. The length of the constriction d is varying (in units of XF}. Solid (dashed j curves denoting zI=
0.2A,F(zl
=0.
5A,F),and are vertically offset by an amount (2e'/h) for clarity. The geometry ofthe channel isdescribed by the inset.W/ AF
FIG.
4. The resistance ofaQPC oflength d =AFcontainingan impurity at (yl, z,
)=(0;0.
2)k~ with r(3=0.6kF. q is varying (in units of A,F').
The dotted lines indicate the valueappear exactly at the same positions with the peaks above the dips in
G„(w).
This isdue to the presenceof
two or-thogonal solutions, one being a quasi-OD state and the other the current-carrying state with unity transmission. AlthoughG„(w)
is calculated by including only the current-carrying states, Gz(w) has contributions from bothof
the above states. Therefore the effectsof
both quasi-OD state and current-carrying states are visible inFig. 3.
The resonant-tunneling effect is usually depressed for laterally confined impurities (i.e., large q) since the resonance peaks and steps are very close to each other, yielding the overlapof
corresponding features in G&(ta). An important remark is about the differenceof
the evanescent states leading to the dips and peaks. Al-though the dips are formed as a resultof
the enhanced backscattering stimulated by the intersubband scattering, the peaks are related tobound (or resonance) states local-ized around the impurity. By decreasing q it is possible to turn off the intersubband scattering and thus to dis-card the dips. On the other hand, the resonance states become real bound states in the absenceof
subband mix-ing. Thus, while the dips are specific to quasi-1D sys-tems, the peaks are due to resonant tunneling and are achievable forall dimensions.IV. CONCLUSION
We investigate the effects
of
elastic scattering by an im-purity in a ballistic channel. Using a model potential weobtained exact expressions for the conductance both for finite and infinite constrictions. We summarize the im-portant findings
of
this study as follows. (i)In agreement with the earlier studies, ' 'we found that the presence
of
an impurity in the ballistic channel distorts the quantiza-tion
of
conductance. The deviation from quantized values increases with increasing strength or increasing in-tegrated strengthof
the impurity potential. (ii)For
at-tractive impurity potentials the dips in the conductance curve form as a resultof
complete reflection. The posi-tion and strengthof
the impurity determines the struc-tureof
these dips. (iii)For
finite constrictions the effectof
the impurity may be depressed by combined scattering from the endsof
the constriction. The resonance struc-ture due to interferenceof
current-carrying states is still visible in the conductance curve. (iv) A resonant-tunneling event takes place for attractive impurity poten-tials. This is a resultof
formationof
quasi-OD states bound tothe impurity.ACKNOWLEDGMENTS
This work is partially supported by a joint project agreement between
IBM
Zurich Research Laboratory and Bilkent University. We acknowledge stimulating dis-cussions withDr. M.
Buttiker,Dr.
P.
Gueret,Dr
B.
J.
van Wees, and
Dr.
L. P.
Kouwenhoven, and we wish to thank toDr. C. S.
Chu andDr.
R.
S.
Sorbello for provid-ing us with their results prior topublication.'B.
J.
van Wees, H.van Houten, C.W.J.
Beenakker,J.
G.Wil-liams, L. P.Kouwenhoven, D.van der Marel, and C.
T.
Fox-on, Phys. Rev. Lett. 60,848(1988).~D. A. %'haram,
T.
J.
Thornton,R.
Newbury, M. Pepper, H. Rithcie, and G.A.C.Jones,J.
Phys. C21, L209 (1988). G.Kirczenow, Solid State Commun. 68, 715 (1988); A.Szaferand A.D. Stone, Phys. Rev. Lett. 60, 300 (1989);L.Escapa and N. Garcia,
J.
Phys. Condens. Matter 1,2125(1989). 4E.G.Haanapel and D.van der Marel, Phys. Rev. B39,5435(1989).
5E.Tekman and S. Ciraci, Phys. Rev.B39,8772(1989). L.
I.
Glazman, G.B.
Lesorik, D.E.
Khmelnitskii, andR.
I.
Shekhter, Pis'ma Zh. Eksp. Teor. Fiz. 48, 218 (1988) [Sov. Phys.
—
JETP Lett. 48, 238(1988)]; A. Yacoby and Y.Imry, Phys. Rev.B41,5341(1990).7E.Tekman and S.Ciraci, Phys. Rev.B40, 8559 (1989). 8C.S.Chu and
R.
S.Sorbello, Phys. Rev.B40,5941(1989).J.
Masek, P. Lipavsky, and B. Kramer,J.
Phys. Condens. Matter 1, 6395 (1989).'
R.
Landauer, IBMJ.
Res. Dev. 32, 306 (1988); M. Buttiker,ibid. 32, 317(1988); A.D.Stone and A. Szafer, ibid. 32,384 (1988);
R.
Landauer,J.
Phys. Condens. Matter. 1, 8099 (1989).'D.
S.Fisher and P.A. Lee, Phys. Rev. B23, 6851 (1988);Y. Imry, in Directions in Condensed Matter Physics, edited by G.Grinstein and G. Mazenko (World Scientific, Singapore, 1986).
' Chu and Sorbello pointed out this feature in afootnote ofRef.
8. However, they did not provide a quantitative analysis of the effect ofyI on G
„(
w).Recently these dips were associated with the quasibound
states in P.W. Bagwell, Phys. Rev.B 41,10354 (1990). How-ever, these states are not resonance orquasibound states since the propagating part ofthe wave function is as important as the evanescent part. Our definition for the quasibound states
is given inSec.
III B.
K.
S.Rails andR.
A. Bukrman, Phys. Rev. Lett. 60, 2434 (1988).D. A.Wharam, M. Pepper, H. Ahmed,