Date

1999-10-15

Description

Comment: 19 pages total, 4 figures; accepted for publication in Phys. Rev. B

The resistance R of a high density network of 6 nm diameter Bi wires in

porous Vycor glass is studied in order to observe its expected semiconductor

behavior. R increases from 300 K down to 0.3 K. Below 4 K, where R varies

approximately as ln(1/T), the order-of-magnitude of the resistance rise, as

well as the behavior of the magnetoresistance are consistent with localization

and electron-electron interaction theories of a one-dimensional disordered

conductor in the presence of strong spin-orbit scattering. We show that this

behaviour and the surface-enhanced carrier density may mask the proposed

semimetal-to-semiconductor transition for quantum Bi wires.

porous Vycor glass is studied in order to observe its expected semiconductor

behavior. R increases from 300 K down to 0.3 K. Below 4 K, where R varies

approximately as ln(1/T), the order-of-magnitude of the resistance rise, as

well as the behavior of the magnetoresistance are consistent with localization

and electron-electron interaction theories of a one-dimensional disordered

conductor in the presence of strong spin-orbit scattering. We show that this

behaviour and the surface-enhanced carrier density may mask the proposed

semimetal-to-semiconductor transition for quantum Bi wires.

Type

Database

Link to record

Show preview

Hide preview

1
Electronic Transport in a Three-dimensional Network of 1-D Bismuth Quantum Wires

T. E. Huber, Howard University, Washington, D.C. 20059

and

M. J. Graf, Department of Physics, Boston College, Chestnut Hill, MA 02467

The resistance R of a high density network of 6 nm diameter Bi wires in porous Vycor

glass is studied in order to observe its expected semiconductor behavior. R increases from 300 K

down to 0.3 K. Below 4 K, where R varies approximately as ln(1/T), the order-of-magnitude of

the resistance rise, as well as the behavior of the magnetoresistance are consistent with

localization and electron-electron interaction theories of a one-dimensional disordered conductor

in the presence of strong spin-orbit scattering. We show that this behaviour and the surface-

enhanced carrier density may mask the proposed semimetal-to-semiconductor transition for

quantum Bi wires.

2

I. INTRODUCTION

The latest experimental advances in preparation of nanowires have given give rise to both

new applications and fundamental science. One of the most exciting prospects is that of an ideal

quantum wire of diameter d less than the Fermi wavelength l F . Profound changes in the density

of state (DOS) and deep modifications of the transport properties are expected, as seen, for

example in single wall carbon nanotubes. Enormous effort has been applied to the study of

carbon nanotubes since its discovery in 1991 with considerable success.1 Comparatively little

attention has been given to the semimetal Bi, that has the smallest effective mass among all

known materials. Due to this property, quantum confinement in Bi can be observed in nanowires

of larger diameter than those of any other material. Here we present a study of the electronic

transport in a network of Bi wires of 6 nm- diameter, much smaller than the bulk electronic

Fermi wavelength l F of 25 nm, and we are able to confirm some theoretical predictions.

The quantum-confinement driven semimetal-to semiconductor (SMSC) transition has

been studied recently in carbon nanotubes2 Bulk Bi is also a semimetal whose three-conduction

band minima at the L-points overlap the valence-band maximum at the T-point by about 40 meV

and a SMSC transition has also been predicted for Bi. Sandomirskii3 pointed out that as a result

of the electrons'zero-point energy, the band overlap decreases for thin films and if the film is thin

enough a gap is formed and the semimetal turns into a semiconductor. The critical thickness tc

for this transition for Bi is expected to be 30 nm. Despite many transport and optical

investigations of quantum phenomena in thin films the experimental evidence is not conclusive.

These films' resistivity is weakly T-dependent, unlike a typical intrinsic semiconductor that

exhibits a thermally activated behavior. This has been interpreted in terms of a very short

effective mean free path l ~ t, and a surface-enhanced carrier density.4,5 Since the majority carrier

concentration then becomes insensitive to the introduction of an energy gap one should expect no

3 abrupt changes at the transition point. The alternative view, that the SMSC does not occur

because the boundary condition of vanishing electron wavefunction is spoiled by disorder, also

has proponents. Experiments, however, are lacking. Gurvitch7 and Brandt et al8 have shown that

200 nm diameter Bi wires display metallic behavior down to 2 K and exhibit quantum size

effects. More recently, Liu, Chien and Searson presented electrical transport measurements on

thick wires (d > 200 nm>>dc , l F). 9 There is renewed interest in the electronic properties of

quantum wires, of diameter d where d< l F, and where l F is the Fermi wavelength equal to 25

nm for the electrons (7 nm for holes) in bulk Bi, because theory indicates that they may have

outstanding thermoelectric properties.10,11 Recently, Zhang, Sun, Dresselhaus, Ying and

Heremans12 reported a reversal of the magnetoresistance temperature dependence for Bi wires

with 65 nm diameter relative to those with 110 nm diameter. The phenomenon is due to the

quantization of the transverse momentum of the carriers which results in the SMSC transition. A

nanowire of diameter d<<dc, the critical diameter, is a simpler case. However, to our knowledge

the electronic transport of a wire of d<dc,l F , has not been addressed experimentally. One

approach is that of confining Bi in a microporous solid insulator. In this study we take such an

approach and we use porous Vycor glass (PVG) because it is a unique prototype monolithic

nanoscale porous material. The use of PVG as a confining host has been particularly fruitful, for

example, in studying the properties of superfluid helium,13 in a restricted geometry. The

characterization of PVG, particularly the surface, is intrinsically difficult but the basic network

structure is reasonable well known from adsorption isotherm and small-angle X-ray scattering

studies.14 Many aspects of the network structure have been clarified by studying the temperature

dependence of the the critical field of In injected in PVG. 15

Confinement introduces scattering and it is known that electronic transport in disordered

conductors can be ballistic, diffusive or in a localized regime. Relevant lengthscales set by are

the carriers' elastic le and inelastic pathlength l f . In disordered systems l f can exceed le by

several orders of magnitude. The coherent superposition of the scattered electron wave results in

4 back-scattering. This causes a weak localization correction to the classical resistance which is

composed of the Drude and electron-phonon contributions. Such corrections appear in many

metallic systems of restricted geometry (wires and films) which show dr /dT <0 at low enough

temperatures. The resistance rise can be expressed as d R/R ~ L /L, where L is the length of a

conductor having the characteristic quantum resistance h/(2e2) = 12.9 k W . L has been derived in

the case of localization,16 where L =l f

for H=0, and also for electron-electron interaction17

theories. The application of these ideas to Bi18 two-dimensional and three-dimensional films is

well known. Bi, being a heavier metal, has a large spin-orbit interaction, characterized by a

lengthscale lso, which can change the character of the low temperature correction to one of "

antilocalization" (dr /dT >0). We show for the first time that 6 nm-diameter Bi wires exhibit

only weak localization and only at the lowest temperatures (T<2 K) .

Progress in the study of true one-dimensional quantum wire systems has been slow due to

the difficulty of fabricating such materials. On the other hand, the quasi-1D case has been around

for 20 years and is well within the realm of microelectronic technology. Mohanty, Jariwala, and

Webb19 have reviewed many studies of this type. All papers cited consider the quasi-1D case l F

<W<l j

where W is the wire width. A more recent case is presented by Khavin et al.20 In this

case l F = 10 nm and W is 50 nm. Also, a common technique of confining the two-dimensional

electron gas is to pattern a gate. Applying a negative voltage to the gate with a split gate

geometry depletes the electron gas beneath it results of a narrow conducting channel; the

conducting width can be tuned down to zero by making the gate voltage more negative.

However, the actual induced potential is not known directly, and theoretical studies indicate that

it has a saddle shape. In the constriction, electrons are confined in the lateral direction and also

retarded by the presence of a potential barrier in the wire direction.21 As we will show, our

experiment presents the (real) 1D case, a network of wires of diameter d< l F <lj . In our

experiment the confinement is well defined by the wire geometry and the network structure. The

physics underlying the two experimental cases should not be confused.

5 The plan of the paper is as follows. In Sec. II we discuss briefly the sample preparation

method, sample characterization and other experimental issues. In Sec. III we analyze the

experimental results.

II. SAMPLE PREPARATION AND CHARACTERIZATION

The samples are prepared by melting pure Bi (99.9999 %) and injecting it into porous

Vycor glass (PVG) by applying hydrostatic pressures of 5 kilobars. We have used this technique

previously for In and other materials.15 The PVG used has an average pore diameter d of 6 nm.

The interconnected network of pores occupies approximately 30% of the total volume. The shiny

black samples have approximately 80 % of the pore volume filled with Bi. While the silica

backbone structure is complex and interconnected, one can consider it being made up of silica

particles of a characteristic size of 26 nm, and this structure is unchanged by the injection

process.22 X-ray diffraction (XRD) from the Bi-PVG composite shows that the Bi retains its

rhombohedral (trigonal) structure but with shrinkage of its unit cell. For example, the (102)

planes in the hexagonal indexing system, that are separated by 0.328 nm in bulk Bi are separated

by 0.320 nm in the composite. This corresponds to a lattice linear contraction of approximately

2.5%. This is likely a result of the injection process since Bi expands on solidification by 3.3 %

An estimate for the average Bi crystallite size D can be obtained from the widths of the XRD

peaks through Scherrer's equation.23 A value D= 9 nm was determined from the .017 FWHM of

the peak corresponding to the (102) planes, comparable to the PVG average pore diameter d=6

nm. A scanning electron micrograph image of the composite is shown in Fig. 1. The Bi in the

composite was exposed by etching with HF. We have observed that these samples are not very

stable and that samples larger than a few millimeters crumble when subject to stress. Still, we

were able to select samples that were robust and therefore suitable for our experiments.

6 The sample resistance was measured using four terminal d.c. and a.c. (f = 100 Hz)

techniques. Electrical contact was made via brass wires attached with silver epoxy to gold pads

deposited in a vacuum evaporator. At 300 K the resistivity of the composite is roughly 21

m W -cm, and cooling to 4 K increases this to 27 m W -cm. The ratio of the room temperature

resistivity of the composite to that of the bulk is r = r Bi-PVG/r bulk = 175. A simplified structural

model of the composite is that of a 3D simple-cubic lattice of 6-nm diameter nanowires where

the periodicity of the network lattice is s ~ 26 nm, roughly the size of the silica particles. A

value of r ~ 12 is estimated by considering that Bi occupies 80% of the pore volume and that

about 1/3 of the metal wires are oriented along the current flow. For example, r is 20 for In-filled

Vycor at room temperature, where the electronic mean free path le << d and surface scattering is

not very important.15 Therefore, r for Bi-PVG is roughly one order of magnitude larger than

expected from structural considerations, indicating that the resistivity is much larger than the

bulk at room temperature (RT). Accordingly, the RT resistivity, r Bi, of Bi in the nanowires is

estimated to be 1.8 m W -cm. We will show later that this value is consistent with results obtained

for thin Bi films

It is obvious that Bi is subject to large stresses in the pores of PVG and that these might

induce a phase change. A hydrostatic pressure of 20 kBar induces the transition from semimetal

Bi(I) to the metallic Bi(II). We can calculate the stresses present in our samples in the following

manner: from a linear contraction of 2.5 % in the [102] direction and taking Young's modulus to

be 5x1011 dyn/cm2,24 we find an stress of 1.3x109 dy/cm2 or 13 kBar. This pressure is insufficient

to promote the Bi(I)-(II) transition.25 Also, the effect of compression in thick (500 nm) Bi films

has been investigated and is known that a pressure of 10 kBar increases the film resistivity by

roughly 7%. Therefore, we believe the stresses in our samples are not directly relevant to the

high resistivity of Bi in the composite.

7 III. RESULTS

The temperature-dependent resistivity of Bi-PVG is shown in Fig. 2. The inset displays

the resistivity over a wide temperature range exhibiting a negative temperature coefficient. In

comparison, the bulk Bi resistivity obeys a T2 law at low temperatures and is roughly

proportional to T for T > 100 K.26 The expected semiconductor activation energy can be obtained

as follows. The presumed thin film's gap dependence was found theoretically2 and

experimentally4 to obey roughly D = 40 meV ((tc/t) 2-1). Assuming the behavior of a nanowire of

d= 6 nm to be that of a thin film of the same thickness we obtain D = 1 eV. Clearly the Bi wire

network fails to display the thermally activated behavior expected for an intrinsic semiconductor

with such a large band gap. This can indeed be understood in terms of a surface-enhanced carrier

density. In thin films,3,4 it was found that p=pi+ps/t where pi is the effective carrier intrinsic

concentration resulting from band overlap (pi = 2.5x10 17/cm3 for thick films) and ps = 3x10

12/cm2

is the effective density per unit film area. At low temperatures (kB T<D ) the majority carrier

concentration becomes temperature independent. For our nanowires, considering the number of

surface states is proportional to the surface area, the effective bulk concentration can be

estimated to be 5x1018/cm3. This concentration is consistent with our results. From r Bi=(nqm ) -1

we obtain m =103 cm3 /(sec volt), in good agreement with thin Bi film results for t = 20 nm. More

study is needed to understand this issue and, in particular, the impact of the surface states on the

expected thermoelectric properties.7,8 In order to exhibit the intrinsic regime, the Bi samples

should have lower net doping levels, which may be obtained by doping or surface modification.

Due to the close analogy between the phenomena exhibited by Bi films and nanowires with

8 regard to the temperature dependence of the resistivity we expect that the two subjects will

develop in close association.

Superimposed on the slowly rising r (T) as T decreases, a much sharper temperature rise

takes place at temperatures below 4K. This is the main focus of this paper and is shown in more

detail in Fig. 3. The variation is seen to be approximately logarithmic although the data is also

roughly consistent with a T-1/2 dependence. This particular temperature dependence is indicative

of one-dimensional weak localization in wires. It is useful to compare the order-of-magnitude of

our results with those of other workers. Williams and Giordano27 studied Au wires of 8 nm

diameter and found that the resistance fractional rise [(R(1.5 K)/R(4 K))-1] is 1.5x10-2 and had a

T1/2 temperature dependence. If this number is multiplied by the factor 64/36 to take into account

the smaller diameter of our wires, the extrapolated temperature rise is 2.6x10-2, in numerical

agreement with our result which is 1.7x10-2. This suggests that our composite sample closely

resembles a network of nearly one-dimensional wires. It is interesting that if a simple-cubic

three-dimensional Bravais lattice of wires is assumed, the resistance of each 30 nm long section

of 6 nm diameter wire is 9 kW at 4 K. This estimate is consistent with the characteristic quantum

resistance h/(2e2) ~12.9 K W for which weak localization effects are expected to be important.16

Alternatively, we can look at the low temperature anomaly as a manifestation of disorder in a 3D

sample. The Yoffe-Regel-Mott criterion28 for localization in 3D indicates that the Drude theory

is inappropriate for samples with r > 2 m W l F(in nm). Since the Fermi wavelength l F can be

estimated to be 25 nm for electrons and 7 nm for holes, our sample resistivity of 21 mW .cm is

clearly problematic.

Williams et al27 and also White et al29 observed that their metallic wires exhibit

negligible magnetoresistance (MR). In contrast, our samples show a sizable one (Fig. 4). It

should be noted that the MR is positive and that it increases for decreasing temperatures. The

transverse30 MR of the Bi-PVG is quite different from that of the bulk Bi which exhibits a B2

9 dependence at low magnetic fields and a B1.6 dependence31 in the high magnetic field range. The

behavior shown by Bi-PVG is reminiscent to that of thin Bi films where the Coulomb anomaly

and the weak localization anomaly occur simultaneously.4,18

The MR of quasi-1D wires is well studied. Altshuler and Aronov32 calculated the

localization contribution to the magnetoresistance of wires by treating the magnetic field as a

perturbation. This approximation actually improves for decreasing wire diameter. Taking spin

orbit effects into account33 D R(T,H) µ 3(1/l f

2 + 1/lso 2 +1/lH

2)-1/2 -(1/l f

2 +1/lH 2)-1/2, where lH is the

magnetic field dephasing length which is roughly 1.73 f 0/H and f 0 = 2p hc/2e. The best fit of the

MR data at 2.3 K is shown in Fig. 4., using l f

= 80 nm and lso = 30 nm. In comparison, l f is 100

nm for thin Bi films at the same temperature.18 These parameters satisfy the condition lso ,l f >>

d, where d is the wire diameter. However, since our wires are short and lso,l f >> s = 26 nm, the

magnetoresistance is also modulated by the mesoscopic network structure. For example,

quantum interference in the electronic path around a silica particle should lead to a

magnetoresistance oscillation with a period D H = 4 f 0/( p s 2) ~ 3T. The effect of the network in

shaping the magnetoresistance, the relative orientation between wire and magnetic field, as well

as the evaluation of the various phase-breaking mechanisms34 is outside the scope of this paper.

It should be noted that weak localization effects were not considered in the interpretation of their

Bi nanowire magnetoresistance measurements by the previous workers.7-11 Because these studies

were for larger diameter wires at higher temperatures we can infer that weak localization effects

appear in Bi at low temperatures and only for very small diameters wires.

We now discuss some structural considerations. We have already noted that large Bi-

PVG samples are not stable and crumble under severe stress. The cracks fill with bulk Bi. We

have also observed that samples that appear to be whole will undergo small changes from one

experimental run to another. Upon aging of several months the composite resistance becomes

slightly smaller, i.e, the resistance at 4 K changes from 26 mW .cm to 25.2 m W cm. This decrease

10

is accompanied by a shoulder at 2 K in the temperature dependence of the resistivity. The I(V) is

non-linear near the bump as indicated by different R(T) readings under d.c. and a.c. excitation. In

comparison, the low temperature (T ~ 4 K) I=I(V) curve of "new" samples is linear up to a

current density of 0.6 A/cm2. This contrast is consistent with the expectation that the sample is

becoming more "granular", and consisting of thinner wires, upon aging. This is expected to be a

result of aging since such changes would minimize the surface energy of the composite. We

have observed that, upon aging, the magnetoresistance becomes larger. For example, D R/R is

0.05 at 3 T and 0.35 K, while it is 0.02 of the "new" samples. This is consistent with weak

localization since thinning down of wires would tend to increase the magnetoresistance.33 It is

observed that the room temperature resistance changes very little in this process, actually

decreasing for older samples. In a first approximation, the shortened mean free path (le ~ d) in

smaller wires is compensated by an increased number of majority carriers (n ~ d-1).

Weitzel and Micklitz35 have studied granular systems built of well-defined rhombohedral

Bi nanoclusters (d~4 nm) in a variety of matrices (Ge, O2, H2, and Xe). They found that many of

these composites exhibit superconductivity at low temperatures (Tc~4 K). This behavior

contrasts with the localization behavior of our sample. However, the onset of superconductivity

may be related to the bump observed by us at 2 K in "aged" or granular samples.

IV. SUMMARY

We present results of an experimental study of the temperature and magnetic-field

dependence of the resistivity of a network of 6 nm diameter Bi wires where d<<l F. The wire

diameter, 6 nm, is much smaller than the critical diameter for the SMSC transition estimated to

be 30 nm. However, we do not observe the expected temperature-activated behavior typical of an

11

intrinsic semiconductor in the temperature range investigated. This can be interpreted in terms of

surface-enhanced carriers. Also, we do not observe strong localization and the composite is a

basically a good conductor. The observed temperature rise at low temperature and associated

magnetoresistance are interpreted in terms of weak localization effects due to disorder. The 80

nm dephasing length scale as inferred from the data is somewhat larger than the network

periodicity and the wirelength s and, therefore, the magnetoresistance could be modulated by the

network mesoscopic structure. Undoubtedly, further study of long, straight nanowires will help

understand better the role of localization in electronic transport in networks of 1-D wires; we are

currently undertaking such studies, utilizing parallel nanochannels insulating host materials.36

The work of T.E.H. was supported by the Army Research Office through DAA H04-95-

1-0117 and by the National Science Foundation through DMR-9632819. M.J.G. is supported in

part through Research Corporation Grant RA0246.

12

REFERENCES

1. S. Tans et al, Nature 386, 474 (1997).

2. G. Lopinski, V.I. Merkulov, and J.S. Lannin, Phys. Rev. Lett. 80, 4641 (1998).

3. V.B. Sandormirskii, JEPT 25, 101 (1967).

4. Y.F. Komnik et al, Sov. J. Low Temp. Phys. 1, 51 (1975).

5. C.A. Hoffman et al, Phys. Rev. B51 5535 (1995).

6. H.T. Chu, Phys. Rev. B51 5532 (1995).

7. M. Gurtvich, J. Low Temp. Phys. 38, 777 (1980).

8. N.B. Brandt, D.V. Gitsu, A.A. Nikolaeva, and Ya.G. Ponomarev, Soviet JETP, 24 273

(1976).

9. K.Liu et al, Phys. Rev. B58, R14681 (1998).

10. E.N. Bogacheck, A.G. Scherbakov, and U. Landman, in "Nanowires", edited by P.A.

Serena and N. Garcia (Kluwer, Dordrecht, 1997).

11. L.D. Hicks and M.S. Dresselhaus, Phys. Rev. B47, 16631 (1993).

12. Z. Zhang, X.Sun, M.S. Dresselhaus, J. Ying, and J. Heremans, Appl. Phys. Lett. 73 1589

(1998).

13. See, for example, C. Lie-Zhao, D.F. Brewer, C. Cirit, E.N. Smith, and J.D. Reppy, Phys.

Rev. B33 106 (1986)

14. For a review, see for example, A. Ch. Mitropolous et al Phys. Rev. B52, 10035 (1995).

15. M.J. Graf et al, Phys. Rev. B45, 3133 (1992). Also, F. Dong et al, Sol. State Comm. 101,

929 (1997).

16. D.J. Thouless, Phys. Rev. Lett. 39, 1167 (1977). P.W. Anderson, D.J. Thouless, E.

Abrahams, and D.S. Fisher, Phys. Rev. B22, 3519 (1980).

17. B.L. Altshuler, D. Khmel'nitzkii, A.I. Larkin, and P.A. Lee, Phys.

Rev. B22, 5142 (1980).

13

18. Y.F. Komnic et al, Sov. J. Low. Temp. Phys. 7, 656 (1982). Y.F. Kommik et al, Solid

State Comm. 44, 865 (1982).

19. P. Mohanty, E.M.Q. Jariwala, and R.A. Webb, Phys. Rev.Lett. 78 3366 (1997), Table I.

20. Y.B.Khavin et al, Phys. Rev. Lett. 81 , 1066 (1998).

21. C.W.J. Beenakker and H.van Houten in Solid State Physics, edited by H. Ehrenreich and

D. Turnbull (Academic Press, New York, 1991)

22. T.E. Huber, P.W. Schmidt, J.S. Lin, and C.A. Huber, to be published.

23. C.A. Huber in "Handbook of Nanophase Materials", edited by A. Goldstein (M. Dekker,

New York, 1996).

24. Y.A. Burenov et al, Sov. Phys. Sol. State 14, 215 (1972).

25. F.P. Bundy, Phys. Rev. 110, 314 (1958).

26. V.N. Lutskii, JEPT 25, 101 (1967).

27. G.D. Williams and N. Giordano, Phys. Rev. 33 8146 (1986).

28. Y. Imry in "Introduction to Mesoscopic Physics" (Oxford University Press, Oxford,

1997).

29. A.E. White, M. Tinkham, W.J. Skocpol, and D.C. Flanders, Phys. Rev. Lett. 48 1752

(1982).

30. Although the macroscopic current is transverse to the magnetic field, the nanowires are

oriented at random. Since the MR has different H-dependence in the parallel and

transverse geometries (Ref. 17) our sample geometry will affect the observed

H-dependence of the composite MR.

31. L.A. Falkovskii, Soviet Phys. Uspekii 94, 1 (1968).

32. B.L. Altshuler and A.G. Aronov, JEPT Lett. 33, 515 (1981).

33. S. Wind, M.J. Rooks, V. Chandrasekhar, and D.E. Prober, Phys. Rev. Lett. 57, 633

(1986).

34. M.E. Gershenson et al, Phys. Rev. B51 10256 (1995).

35. H.B. Weitzel and H. Micklitz, Phys. Rev. Lett. 66, 385 (1991).

14

36. T.E. Huber, M.J. Graf, and C.A. Foss, in Thermoelectric Materials- The Next Generation

Materials for Small-Scale Refrigeration and Power Generation Applications edited by T.

M. Tritt, H.B. Lyon, Jr., G. Mahan, and M.G. Kanatzidis (Materials Research Society,

Pittsburg, 1999), p.227.

15

FIGURE CAPTIONS

Fig. 1. Scanning electron micrograph of the Bi-PVG composite. Dark

areas correspond to silica.

Fig. 2. The resistivity from 0.3 K to room temperature.

Fig. 3. Low temperature resistivity of Bi-injected PVG.

Fig. 4 Magnetoresistance of Bi-injected PVG at 0.3 K. The solid line is a fit

to the theory of Ref. 33.

16

Fig. 1

17

Fig. 2

21

23

25

27

29

0 100 200 300

r

( m

W

-c m

)

Temperature (K)

18

Fig. 3

27.0

27.4

27.8

28.2

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

r

( m

W

-c m

)

ln T

19

Fig. 4

T. E. Huber, Howard University, Washington, D.C. 20059

and

M. J. Graf, Department of Physics, Boston College, Chestnut Hill, MA 02467

The resistance R of a high density network of 6 nm diameter Bi wires in porous Vycor

glass is studied in order to observe its expected semiconductor behavior. R increases from 300 K

down to 0.3 K. Below 4 K, where R varies approximately as ln(1/T), the order-of-magnitude of

the resistance rise, as well as the behavior of the magnetoresistance are consistent with

localization and electron-electron interaction theories of a one-dimensional disordered conductor

in the presence of strong spin-orbit scattering. We show that this behaviour and the surface-

enhanced carrier density may mask the proposed semimetal-to-semiconductor transition for

quantum Bi wires.

2

I. INTRODUCTION

The latest experimental advances in preparation of nanowires have given give rise to both

new applications and fundamental science. One of the most exciting prospects is that of an ideal

quantum wire of diameter d less than the Fermi wavelength l F . Profound changes in the density

of state (DOS) and deep modifications of the transport properties are expected, as seen, for

example in single wall carbon nanotubes. Enormous effort has been applied to the study of

carbon nanotubes since its discovery in 1991 with considerable success.1 Comparatively little

attention has been given to the semimetal Bi, that has the smallest effective mass among all

known materials. Due to this property, quantum confinement in Bi can be observed in nanowires

of larger diameter than those of any other material. Here we present a study of the electronic

transport in a network of Bi wires of 6 nm- diameter, much smaller than the bulk electronic

Fermi wavelength l F of 25 nm, and we are able to confirm some theoretical predictions.

The quantum-confinement driven semimetal-to semiconductor (SMSC) transition has

been studied recently in carbon nanotubes2 Bulk Bi is also a semimetal whose three-conduction

band minima at the L-points overlap the valence-band maximum at the T-point by about 40 meV

and a SMSC transition has also been predicted for Bi. Sandomirskii3 pointed out that as a result

of the electrons'zero-point energy, the band overlap decreases for thin films and if the film is thin

enough a gap is formed and the semimetal turns into a semiconductor. The critical thickness tc

for this transition for Bi is expected to be 30 nm. Despite many transport and optical

investigations of quantum phenomena in thin films the experimental evidence is not conclusive.

These films' resistivity is weakly T-dependent, unlike a typical intrinsic semiconductor that

exhibits a thermally activated behavior. This has been interpreted in terms of a very short

effective mean free path l ~ t, and a surface-enhanced carrier density.4,5 Since the majority carrier

concentration then becomes insensitive to the introduction of an energy gap one should expect no

3 abrupt changes at the transition point. The alternative view, that the SMSC does not occur

because the boundary condition of vanishing electron wavefunction is spoiled by disorder, also

has proponents. Experiments, however, are lacking. Gurvitch7 and Brandt et al8 have shown that

200 nm diameter Bi wires display metallic behavior down to 2 K and exhibit quantum size

effects. More recently, Liu, Chien and Searson presented electrical transport measurements on

thick wires (d > 200 nm>>dc , l F). 9 There is renewed interest in the electronic properties of

quantum wires, of diameter d where d< l F, and where l F is the Fermi wavelength equal to 25

nm for the electrons (7 nm for holes) in bulk Bi, because theory indicates that they may have

outstanding thermoelectric properties.10,11 Recently, Zhang, Sun, Dresselhaus, Ying and

Heremans12 reported a reversal of the magnetoresistance temperature dependence for Bi wires

with 65 nm diameter relative to those with 110 nm diameter. The phenomenon is due to the

quantization of the transverse momentum of the carriers which results in the SMSC transition. A

nanowire of diameter d<<dc, the critical diameter, is a simpler case. However, to our knowledge

the electronic transport of a wire of d<dc,l F , has not been addressed experimentally. One

approach is that of confining Bi in a microporous solid insulator. In this study we take such an

approach and we use porous Vycor glass (PVG) because it is a unique prototype monolithic

nanoscale porous material. The use of PVG as a confining host has been particularly fruitful, for

example, in studying the properties of superfluid helium,13 in a restricted geometry. The

characterization of PVG, particularly the surface, is intrinsically difficult but the basic network

structure is reasonable well known from adsorption isotherm and small-angle X-ray scattering

studies.14 Many aspects of the network structure have been clarified by studying the temperature

dependence of the the critical field of In injected in PVG. 15

Confinement introduces scattering and it is known that electronic transport in disordered

conductors can be ballistic, diffusive or in a localized regime. Relevant lengthscales set by are

the carriers' elastic le and inelastic pathlength l f . In disordered systems l f can exceed le by

several orders of magnitude. The coherent superposition of the scattered electron wave results in

4 back-scattering. This causes a weak localization correction to the classical resistance which is

composed of the Drude and electron-phonon contributions. Such corrections appear in many

metallic systems of restricted geometry (wires and films) which show dr /dT <0 at low enough

temperatures. The resistance rise can be expressed as d R/R ~ L /L, where L is the length of a

conductor having the characteristic quantum resistance h/(2e2) = 12.9 k W . L has been derived in

the case of localization,16 where L =l f

for H=0, and also for electron-electron interaction17

theories. The application of these ideas to Bi18 two-dimensional and three-dimensional films is

well known. Bi, being a heavier metal, has a large spin-orbit interaction, characterized by a

lengthscale lso, which can change the character of the low temperature correction to one of "

antilocalization" (dr /dT >0). We show for the first time that 6 nm-diameter Bi wires exhibit

only weak localization and only at the lowest temperatures (T<2 K) .

Progress in the study of true one-dimensional quantum wire systems has been slow due to

the difficulty of fabricating such materials. On the other hand, the quasi-1D case has been around

for 20 years and is well within the realm of microelectronic technology. Mohanty, Jariwala, and

Webb19 have reviewed many studies of this type. All papers cited consider the quasi-1D case l F

<W<l j

where W is the wire width. A more recent case is presented by Khavin et al.20 In this

case l F = 10 nm and W is 50 nm. Also, a common technique of confining the two-dimensional

electron gas is to pattern a gate. Applying a negative voltage to the gate with a split gate

geometry depletes the electron gas beneath it results of a narrow conducting channel; the

conducting width can be tuned down to zero by making the gate voltage more negative.

However, the actual induced potential is not known directly, and theoretical studies indicate that

it has a saddle shape. In the constriction, electrons are confined in the lateral direction and also

retarded by the presence of a potential barrier in the wire direction.21 As we will show, our

experiment presents the (real) 1D case, a network of wires of diameter d< l F <lj . In our

experiment the confinement is well defined by the wire geometry and the network structure. The

physics underlying the two experimental cases should not be confused.

5 The plan of the paper is as follows. In Sec. II we discuss briefly the sample preparation

method, sample characterization and other experimental issues. In Sec. III we analyze the

experimental results.

II. SAMPLE PREPARATION AND CHARACTERIZATION

The samples are prepared by melting pure Bi (99.9999 %) and injecting it into porous

Vycor glass (PVG) by applying hydrostatic pressures of 5 kilobars. We have used this technique

previously for In and other materials.15 The PVG used has an average pore diameter d of 6 nm.

The interconnected network of pores occupies approximately 30% of the total volume. The shiny

black samples have approximately 80 % of the pore volume filled with Bi. While the silica

backbone structure is complex and interconnected, one can consider it being made up of silica

particles of a characteristic size of 26 nm, and this structure is unchanged by the injection

process.22 X-ray diffraction (XRD) from the Bi-PVG composite shows that the Bi retains its

rhombohedral (trigonal) structure but with shrinkage of its unit cell. For example, the (102)

planes in the hexagonal indexing system, that are separated by 0.328 nm in bulk Bi are separated

by 0.320 nm in the composite. This corresponds to a lattice linear contraction of approximately

2.5%. This is likely a result of the injection process since Bi expands on solidification by 3.3 %

An estimate for the average Bi crystallite size D can be obtained from the widths of the XRD

peaks through Scherrer's equation.23 A value D= 9 nm was determined from the .017 FWHM of

the peak corresponding to the (102) planes, comparable to the PVG average pore diameter d=6

nm. A scanning electron micrograph image of the composite is shown in Fig. 1. The Bi in the

composite was exposed by etching with HF. We have observed that these samples are not very

stable and that samples larger than a few millimeters crumble when subject to stress. Still, we

were able to select samples that were robust and therefore suitable for our experiments.

6 The sample resistance was measured using four terminal d.c. and a.c. (f = 100 Hz)

techniques. Electrical contact was made via brass wires attached with silver epoxy to gold pads

deposited in a vacuum evaporator. At 300 K the resistivity of the composite is roughly 21

m W -cm, and cooling to 4 K increases this to 27 m W -cm. The ratio of the room temperature

resistivity of the composite to that of the bulk is r = r Bi-PVG/r bulk = 175. A simplified structural

model of the composite is that of a 3D simple-cubic lattice of 6-nm diameter nanowires where

the periodicity of the network lattice is s ~ 26 nm, roughly the size of the silica particles. A

value of r ~ 12 is estimated by considering that Bi occupies 80% of the pore volume and that

about 1/3 of the metal wires are oriented along the current flow. For example, r is 20 for In-filled

Vycor at room temperature, where the electronic mean free path le << d and surface scattering is

not very important.15 Therefore, r for Bi-PVG is roughly one order of magnitude larger than

expected from structural considerations, indicating that the resistivity is much larger than the

bulk at room temperature (RT). Accordingly, the RT resistivity, r Bi, of Bi in the nanowires is

estimated to be 1.8 m W -cm. We will show later that this value is consistent with results obtained

for thin Bi films

It is obvious that Bi is subject to large stresses in the pores of PVG and that these might

induce a phase change. A hydrostatic pressure of 20 kBar induces the transition from semimetal

Bi(I) to the metallic Bi(II). We can calculate the stresses present in our samples in the following

manner: from a linear contraction of 2.5 % in the [102] direction and taking Young's modulus to

be 5x1011 dyn/cm2,24 we find an stress of 1.3x109 dy/cm2 or 13 kBar. This pressure is insufficient

to promote the Bi(I)-(II) transition.25 Also, the effect of compression in thick (500 nm) Bi films

has been investigated and is known that a pressure of 10 kBar increases the film resistivity by

roughly 7%. Therefore, we believe the stresses in our samples are not directly relevant to the

high resistivity of Bi in the composite.

7 III. RESULTS

The temperature-dependent resistivity of Bi-PVG is shown in Fig. 2. The inset displays

the resistivity over a wide temperature range exhibiting a negative temperature coefficient. In

comparison, the bulk Bi resistivity obeys a T2 law at low temperatures and is roughly

proportional to T for T > 100 K.26 The expected semiconductor activation energy can be obtained

as follows. The presumed thin film's gap dependence was found theoretically2 and

experimentally4 to obey roughly D = 40 meV ((tc/t) 2-1). Assuming the behavior of a nanowire of

d= 6 nm to be that of a thin film of the same thickness we obtain D = 1 eV. Clearly the Bi wire

network fails to display the thermally activated behavior expected for an intrinsic semiconductor

with such a large band gap. This can indeed be understood in terms of a surface-enhanced carrier

density. In thin films,3,4 it was found that p=pi+ps/t where pi is the effective carrier intrinsic

concentration resulting from band overlap (pi = 2.5x10 17/cm3 for thick films) and ps = 3x10

12/cm2

is the effective density per unit film area. At low temperatures (kB T<D ) the majority carrier

concentration becomes temperature independent. For our nanowires, considering the number of

surface states is proportional to the surface area, the effective bulk concentration can be

estimated to be 5x1018/cm3. This concentration is consistent with our results. From r Bi=(nqm ) -1

we obtain m =103 cm3 /(sec volt), in good agreement with thin Bi film results for t = 20 nm. More

study is needed to understand this issue and, in particular, the impact of the surface states on the

expected thermoelectric properties.7,8 In order to exhibit the intrinsic regime, the Bi samples

should have lower net doping levels, which may be obtained by doping or surface modification.

Due to the close analogy between the phenomena exhibited by Bi films and nanowires with

8 regard to the temperature dependence of the resistivity we expect that the two subjects will

develop in close association.

Superimposed on the slowly rising r (T) as T decreases, a much sharper temperature rise

takes place at temperatures below 4K. This is the main focus of this paper and is shown in more

detail in Fig. 3. The variation is seen to be approximately logarithmic although the data is also

roughly consistent with a T-1/2 dependence. This particular temperature dependence is indicative

of one-dimensional weak localization in wires. It is useful to compare the order-of-magnitude of

our results with those of other workers. Williams and Giordano27 studied Au wires of 8 nm

diameter and found that the resistance fractional rise [(R(1.5 K)/R(4 K))-1] is 1.5x10-2 and had a

T1/2 temperature dependence. If this number is multiplied by the factor 64/36 to take into account

the smaller diameter of our wires, the extrapolated temperature rise is 2.6x10-2, in numerical

agreement with our result which is 1.7x10-2. This suggests that our composite sample closely

resembles a network of nearly one-dimensional wires. It is interesting that if a simple-cubic

three-dimensional Bravais lattice of wires is assumed, the resistance of each 30 nm long section

of 6 nm diameter wire is 9 kW at 4 K. This estimate is consistent with the characteristic quantum

resistance h/(2e2) ~12.9 K W for which weak localization effects are expected to be important.16

Alternatively, we can look at the low temperature anomaly as a manifestation of disorder in a 3D

sample. The Yoffe-Regel-Mott criterion28 for localization in 3D indicates that the Drude theory

is inappropriate for samples with r > 2 m W l F(in nm). Since the Fermi wavelength l F can be

estimated to be 25 nm for electrons and 7 nm for holes, our sample resistivity of 21 mW .cm is

clearly problematic.

Williams et al27 and also White et al29 observed that their metallic wires exhibit

negligible magnetoresistance (MR). In contrast, our samples show a sizable one (Fig. 4). It

should be noted that the MR is positive and that it increases for decreasing temperatures. The

transverse30 MR of the Bi-PVG is quite different from that of the bulk Bi which exhibits a B2

9 dependence at low magnetic fields and a B1.6 dependence31 in the high magnetic field range. The

behavior shown by Bi-PVG is reminiscent to that of thin Bi films where the Coulomb anomaly

and the weak localization anomaly occur simultaneously.4,18

The MR of quasi-1D wires is well studied. Altshuler and Aronov32 calculated the

localization contribution to the magnetoresistance of wires by treating the magnetic field as a

perturbation. This approximation actually improves for decreasing wire diameter. Taking spin

orbit effects into account33 D R(T,H) µ 3(1/l f

2 + 1/lso 2 +1/lH

2)-1/2 -(1/l f

2 +1/lH 2)-1/2, where lH is the

magnetic field dephasing length which is roughly 1.73 f 0/H and f 0 = 2p hc/2e. The best fit of the

MR data at 2.3 K is shown in Fig. 4., using l f

= 80 nm and lso = 30 nm. In comparison, l f is 100

nm for thin Bi films at the same temperature.18 These parameters satisfy the condition lso ,l f >>

d, where d is the wire diameter. However, since our wires are short and lso,l f >> s = 26 nm, the

magnetoresistance is also modulated by the mesoscopic network structure. For example,

quantum interference in the electronic path around a silica particle should lead to a

magnetoresistance oscillation with a period D H = 4 f 0/( p s 2) ~ 3T. The effect of the network in

shaping the magnetoresistance, the relative orientation between wire and magnetic field, as well

as the evaluation of the various phase-breaking mechanisms34 is outside the scope of this paper.

It should be noted that weak localization effects were not considered in the interpretation of their

Bi nanowire magnetoresistance measurements by the previous workers.7-11 Because these studies

were for larger diameter wires at higher temperatures we can infer that weak localization effects

appear in Bi at low temperatures and only for very small diameters wires.

We now discuss some structural considerations. We have already noted that large Bi-

PVG samples are not stable and crumble under severe stress. The cracks fill with bulk Bi. We

have also observed that samples that appear to be whole will undergo small changes from one

experimental run to another. Upon aging of several months the composite resistance becomes

slightly smaller, i.e, the resistance at 4 K changes from 26 mW .cm to 25.2 m W cm. This decrease

10

is accompanied by a shoulder at 2 K in the temperature dependence of the resistivity. The I(V) is

non-linear near the bump as indicated by different R(T) readings under d.c. and a.c. excitation. In

comparison, the low temperature (T ~ 4 K) I=I(V) curve of "new" samples is linear up to a

current density of 0.6 A/cm2. This contrast is consistent with the expectation that the sample is

becoming more "granular", and consisting of thinner wires, upon aging. This is expected to be a

result of aging since such changes would minimize the surface energy of the composite. We

have observed that, upon aging, the magnetoresistance becomes larger. For example, D R/R is

0.05 at 3 T and 0.35 K, while it is 0.02 of the "new" samples. This is consistent with weak

localization since thinning down of wires would tend to increase the magnetoresistance.33 It is

observed that the room temperature resistance changes very little in this process, actually

decreasing for older samples. In a first approximation, the shortened mean free path (le ~ d) in

smaller wires is compensated by an increased number of majority carriers (n ~ d-1).

Weitzel and Micklitz35 have studied granular systems built of well-defined rhombohedral

Bi nanoclusters (d~4 nm) in a variety of matrices (Ge, O2, H2, and Xe). They found that many of

these composites exhibit superconductivity at low temperatures (Tc~4 K). This behavior

contrasts with the localization behavior of our sample. However, the onset of superconductivity

may be related to the bump observed by us at 2 K in "aged" or granular samples.

IV. SUMMARY

We present results of an experimental study of the temperature and magnetic-field

dependence of the resistivity of a network of 6 nm diameter Bi wires where d<<l F. The wire

diameter, 6 nm, is much smaller than the critical diameter for the SMSC transition estimated to

be 30 nm. However, we do not observe the expected temperature-activated behavior typical of an

11

intrinsic semiconductor in the temperature range investigated. This can be interpreted in terms of

surface-enhanced carriers. Also, we do not observe strong localization and the composite is a

basically a good conductor. The observed temperature rise at low temperature and associated

magnetoresistance are interpreted in terms of weak localization effects due to disorder. The 80

nm dephasing length scale as inferred from the data is somewhat larger than the network

periodicity and the wirelength s and, therefore, the magnetoresistance could be modulated by the

network mesoscopic structure. Undoubtedly, further study of long, straight nanowires will help

understand better the role of localization in electronic transport in networks of 1-D wires; we are

currently undertaking such studies, utilizing parallel nanochannels insulating host materials.36

The work of T.E.H. was supported by the Army Research Office through DAA H04-95-

1-0117 and by the National Science Foundation through DMR-9632819. M.J.G. is supported in

part through Research Corporation Grant RA0246.

12

REFERENCES

1. S. Tans et al, Nature 386, 474 (1997).

2. G. Lopinski, V.I. Merkulov, and J.S. Lannin, Phys. Rev. Lett. 80, 4641 (1998).

3. V.B. Sandormirskii, JEPT 25, 101 (1967).

4. Y.F. Komnik et al, Sov. J. Low Temp. Phys. 1, 51 (1975).

5. C.A. Hoffman et al, Phys. Rev. B51 5535 (1995).

6. H.T. Chu, Phys. Rev. B51 5532 (1995).

7. M. Gurtvich, J. Low Temp. Phys. 38, 777 (1980).

8. N.B. Brandt, D.V. Gitsu, A.A. Nikolaeva, and Ya.G. Ponomarev, Soviet JETP, 24 273

(1976).

9. K.Liu et al, Phys. Rev. B58, R14681 (1998).

10. E.N. Bogacheck, A.G. Scherbakov, and U. Landman, in "Nanowires", edited by P.A.

Serena and N. Garcia (Kluwer, Dordrecht, 1997).

11. L.D. Hicks and M.S. Dresselhaus, Phys. Rev. B47, 16631 (1993).

12. Z. Zhang, X.Sun, M.S. Dresselhaus, J. Ying, and J. Heremans, Appl. Phys. Lett. 73 1589

(1998).

13. See, for example, C. Lie-Zhao, D.F. Brewer, C. Cirit, E.N. Smith, and J.D. Reppy, Phys.

Rev. B33 106 (1986)

14. For a review, see for example, A. Ch. Mitropolous et al Phys. Rev. B52, 10035 (1995).

15. M.J. Graf et al, Phys. Rev. B45, 3133 (1992). Also, F. Dong et al, Sol. State Comm. 101,

929 (1997).

16. D.J. Thouless, Phys. Rev. Lett. 39, 1167 (1977). P.W. Anderson, D.J. Thouless, E.

Abrahams, and D.S. Fisher, Phys. Rev. B22, 3519 (1980).

17. B.L. Altshuler, D. Khmel'nitzkii, A.I. Larkin, and P.A. Lee, Phys.

Rev. B22, 5142 (1980).

13

18. Y.F. Komnic et al, Sov. J. Low. Temp. Phys. 7, 656 (1982). Y.F. Kommik et al, Solid

State Comm. 44, 865 (1982).

19. P. Mohanty, E.M.Q. Jariwala, and R.A. Webb, Phys. Rev.Lett. 78 3366 (1997), Table I.

20. Y.B.Khavin et al, Phys. Rev. Lett. 81 , 1066 (1998).

21. C.W.J. Beenakker and H.van Houten in Solid State Physics, edited by H. Ehrenreich and

D. Turnbull (Academic Press, New York, 1991)

22. T.E. Huber, P.W. Schmidt, J.S. Lin, and C.A. Huber, to be published.

23. C.A. Huber in "Handbook of Nanophase Materials", edited by A. Goldstein (M. Dekker,

New York, 1996).

24. Y.A. Burenov et al, Sov. Phys. Sol. State 14, 215 (1972).

25. F.P. Bundy, Phys. Rev. 110, 314 (1958).

26. V.N. Lutskii, JEPT 25, 101 (1967).

27. G.D. Williams and N. Giordano, Phys. Rev. 33 8146 (1986).

28. Y. Imry in "Introduction to Mesoscopic Physics" (Oxford University Press, Oxford,

1997).

29. A.E. White, M. Tinkham, W.J. Skocpol, and D.C. Flanders, Phys. Rev. Lett. 48 1752

(1982).

30. Although the macroscopic current is transverse to the magnetic field, the nanowires are

oriented at random. Since the MR has different H-dependence in the parallel and

transverse geometries (Ref. 17) our sample geometry will affect the observed

H-dependence of the composite MR.

31. L.A. Falkovskii, Soviet Phys. Uspekii 94, 1 (1968).

32. B.L. Altshuler and A.G. Aronov, JEPT Lett. 33, 515 (1981).

33. S. Wind, M.J. Rooks, V. Chandrasekhar, and D.E. Prober, Phys. Rev. Lett. 57, 633

(1986).

34. M.E. Gershenson et al, Phys. Rev. B51 10256 (1995).

35. H.B. Weitzel and H. Micklitz, Phys. Rev. Lett. 66, 385 (1991).

14

36. T.E. Huber, M.J. Graf, and C.A. Foss, in Thermoelectric Materials- The Next Generation

Materials for Small-Scale Refrigeration and Power Generation Applications edited by T.

M. Tritt, H.B. Lyon, Jr., G. Mahan, and M.G. Kanatzidis (Materials Research Society,

Pittsburg, 1999), p.227.

15

FIGURE CAPTIONS

Fig. 1. Scanning electron micrograph of the Bi-PVG composite. Dark

areas correspond to silica.

Fig. 2. The resistivity from 0.3 K to room temperature.

Fig. 3. Low temperature resistivity of Bi-injected PVG.

Fig. 4 Magnetoresistance of Bi-injected PVG at 0.3 K. The solid line is a fit

to the theory of Ref. 33.

16

Fig. 1

17

Fig. 2

21

23

25

27

29

0 100 200 300

r

( m

W

-c m

)

Temperature (K)

18

Fig. 3

27.0

27.4

27.8

28.2

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

r

( m

W

-c m

)

ln T

19

Fig. 4

Comments