Solid-State Electronics 49 (2005) 227232 www.elsevier.

com/locate/sse

Measuring the specic contact resistance of contacts to semiconductor nanowires

a,*

, Y. Wang b, M.A. Cabassi b, K.K. Lew a, S. Dey a, J.M. Redwing a, T.S. Mayer b

Department of Materials Science and Engineering and Materials Research Institute, The Pennsylvania State University, 109 Steidle Building, University Park, PA 16802, USA b Department of Electrical Engineering, The Pennsylvania State University, University Park, PA 16802, USA Received 19 March 2004; received in revised form 17 July 2004; accepted 13 August 2004 Available online 29 September 2004

The review of this paper was arranged by Prof. A. Zaslavsky

Abstract Ohmic contacts to semiconductor nanowires are essential components of many new nanoscale electronic devices. Equations for extracting specic contact resistance (or contact resistivity) from several dierent test structures have been developed by modeling the metal/semiconductor contact as a transmission line, leading to the development of equations analogous to those used for planar contacts. The advantages and disadvantages of various test structures are discussed. To fabricate test structures using a convenient four-point approach, silicon nanowires have been aligned using eld-assisted assembly and contacts fabricated. Finally, specic contact resistances near 5 104 X cm2 have been measured for Ti/Au contacts to p-type Si nanowires with diameters of 78 and 104 nm. 2004 Elsevier Ltd. All rights reserved.
PACS: 73.40.Cg; 73.40.C; 85.40.Ux; 73.40.N Keywords: Nanowire; Ohmic contact; Nanocontact; Specic contact resistance; Contact resistivity

1. Introduction Semiconductor nanowires are drawing increasing attention as building blocks in bottom-up assembly of nanoscale devices and circuits [1]. In fact, chemical and biological sensors [2], eld eect transistors [3], and logic circuits [4] have already been demonstrated. Ohmic contacts to semiconductor nanowires are essential for devices such as these, and research on ohmic contacts to semiconductor nanowires is of current interest.

Corresponding author. +1 814 863 0744; fax: +1 814 865 2917. E-mail address: mohney@ems.psu.edu (S.E. Mohney).

Important parameters in the study of metal/semiconductor contacts are the contact resistance Rc and specic contact resistance qc, also frequently referred to as the contact resistivity. The specic contact resistance is a particularly useful area-independent parameter for comparing contacts of dierent geometries, and it is typically expressed in units of X cm2. The contact resistance Rc is then given by qc/A, where A is the active area of the contact. What complicates the determination of qc is that current transport does not always occur uniformly in a contact; rather, it may take place primarily near the edges in certain geometries. Therefore, modeling is often required to determine the active area of the contact and extract qc. Using a variety of test structures, methods have long since been established for extracting the contact

0038-1101/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.sse.2004.08.006

228

S.E. Mohney et al. / Solid-State Electronics 49 (2005) 227232

resistance and specic contact resistance of planar contacts. One common situation is to have lateral contacts to a semiconductor layer that is on a non-conducting substrate or is electrically isolated from the substrate by a pn junction. A transmission line model (TLM) is often used to describe current transport in lateral contacts made to such semiconductor layers [5,6]. In this manuscript, we develop equations and propose multiterminal test structures for analogous methods of extracting the specic contact resistance of contacts to semiconductor nanowires. We then describe the advantages and disadvantages of these test structures. Finally, we demonstrate the use of the most experimentally convenient of these test structures to measure the specic contact resistance of contacts to a p-type Si nanowire.

Fig. 2. Model of the metal/semiconductor nanowire contact.

2. Nanowire transmission line model Fig. 1 shows a nanowire placed on a metal pad and then covered with another layer of the same metallization. This geometry is the one we are examining for our metal/semiconductor contacts. We assume that the entire circumference of the nanowire of radius r is in direct contact with the metallization along a length L of the nanowire. We also assume that the voltage is uniform throughout the entire metal contact and that it varies in the semiconductor only along the length of the nanowire. Finally, we assume that the resistivity and radius of the semiconductor nanowire beneath the contact is not altered by the fabrication of the contact. We may use a transmission line model to describe current transport at the metal/semiconductor interface, as shown in Fig. 2, where qs is the resistivity of the semiconductor, qc is the specic contact resistance, r is the radius of the wire, L is the length of the contact, I(x) is the current in the semiconductor nanowire, v0 is the

voltage drop between the metal contact and the semiconductor at x = L (beneath the end of the contact), and V(x) is the voltage drop in the semiconductor between x and x = L. The dierential resistance at the metal/semiconductor interface along a length of the nanowire dx is given by qc/(2pr dx), while the resistance of the semiconductor along a length dx is given by qs dx/ (pr2). In our model, the current is therefore distributed homogeneously throughout the cross-section of the semiconductor nanowire. To ensure that the current will ow vertically through the contact region into the crosssection of the nanowire, the resistance at the metal/ semiconductor interface must be high compared to that of nanowire segment beneath it, a condition that is met for the measurements we report later in the manuscript. An analogous requirement for applying the transmission line model to planar contacts was described previously by Berger [6]. For this situation, we can describe the current and voltage in the semiconductor using dV q s2 I dx pr and dI 2pr v0 V : dx qc Together these equations give d2 I 2qs I 0: dx2 rqc 3 2 1

For a contact of length L, I(x) grows from 0 at x = 0 to the total current i0 at the end of the contact (x = L). Therefore, Ix i0 sinhx=LT ; sinhL=LT 4

where the transfer length LT is given by r rqc LT : 2qs

Dierentiating Eq. (4) and inserting it into Eq. (2), we nd that

V x v0

qc i0 coshx=LT : 2prLT sinhL=LT

S.E. Mohney et al. / Solid-State Electronics 49 (2005) 227232

229

With V(L) = 0 and using Eq. (5), we nd that v0 qs i0 LT cothL=LT : 7 pr2 Since the contact resistance Rc is simply v0/i0, we obtain

Rc

qs LT cothL=LT : 8 pr2 In the limit of a long contact (L > 3LT), the expression simplies to Rc qs LT ; pr2 or using Eq. (5), qc Rc : 2prLT 9

10
Fig. 3. Multiwire test structure.

3. Test structures We can envision several dierent test structures that would allow us to measure contact resistance, specic contact resistance, and nanowire resistivity using the equations above. All of these test structures require that each contact in the structure have the same specic contact resistance as the others, which is a common requirement for test structures for both planar contacts and other geometries. Furthermore, we neglect the eects of any depletion or accumulation at the surfaces of the nanowires between the contacts, and we assume when we extract the resistivity of the nanowires that the current is carried by the full cross-sectional area. In the measurements reported later in this paper, the nanowires are very heavily doped, which is the situation that most closely allows us to meet this requirement. However, this issue becomes increasingly important for very small nanowires and nanowires that are not heavily doped. The rst structure is shown in Fig. 3. Pairs of contacts are separated by dierent distances (l1, l2, etc.), and a different nanowire is located between each pair of contacts. There is no way to separate the resistance of the contacts from the resistance of the semiconductor when only one of the nanowires is measured; however, with nanowires of varying length, there is an easy approach analogous to that commonly used for planar contacts [7]. As long as we have long contacts (L > 3LT), the exact length L of each contact is not important, and the total resistance RT between a pair of contacts separated by a segment of nanowire of length l is given by RT 2Rc or RT qs 2LT l: pr2 12
Fig. 4. Test structure with multiple contacts.

In this case, all that is necessary is to t the data for RT versus l to the equation for a straight line, yielding a slope of qs/(pr2), an intercept at RT = 0 of 2LT, and an intercept at l = 0 of 2Rc. Thus, qs is readily obtained from the slope and radius of the wire, and qc can be obtained from LT using Eq. (5), qs, and r. An advantage of this method is the simplicity of fabrication. However, a serious disadvantage is that we must assume that all the nanowires have identical radii and resistivity. This assumption is often not fully met with the nanowires under study today, since current fabrication methods typically yield nanowires of slightly varying radii [8], more than one dierent crystallographic orientation within the same batch of nanowires [8], and a fraction of nanowires that are bicrystals [9]. A second test structure makes use of multiple contacts along the length of a nanowire, as shown in Fig. 4. If contacts 15 are all long compared to LT, we may measure RT between pairs of contacts 1 and 2,

qs l; pr2

11

230

S.E. Mohney et al. / Solid-State Electronics 49 (2005) 227232

2 and 3, etc. Then we may again t data for RT as a function of l to Eq. (11) to extract qs and Rc and use LT to nd qc, as described previously for the multiwire approach. However, it is not unusual for LT of a metal/ semiconductor contact to be on the order of a micron; thus, the total length of the test structure could be on the order of 30 lm or greater. This approach therefore requires rather long nanowires. If the middle contacts 2, 3, and 4 are not long compared to LT, it is in principle possible to obtain data from only the identical middle contacts 2, 3, and 4 of length L and then t pairs of data points (RT, l) to   qs L RT 2 2LT coth l : 13 LT pr However, this approach is more cumbersome than simply tting data to a straight line, which is all that is required for the case of long contacts. A third test structure, which is particularly convenient for shorter nanowires, is shown in Fig. 5. This approach diers from those described previously in that the resistivity of the nanowires is measured using a four point probe technique, where the current I14 is sourced between the two end contacts, and the voltage drop between the two middle contacts V23 is measured. To avoid large measurement error, the input impedance of the voltmeter must be much greater than the resistance of the nanowire, which can be achieved using commercially available test equipment. The resistivity of the nanowire is then given by qs V 23 pr2 : I 14 l2 14

RT 2Rc

qs l1 : pr2

15

Finally, using Eqs. (9) and (10), we can extract qc. We assume in this analysis that the potential drop along the length of the nanowire is not altered by the presence of the two middle contacts, a condition that is well met for ohmic contacts with low barrier heights. To avoid large errors when using this measurement technique, it is also necessary for the two middle contacts to be short compared to LT, or else a signicant portion of the current may be transported through the middle contacts rather than the semiconductor nanowire when we measure RT between contacts 1 and 4, making inaccurate our use of l1 as the length of the semiconductor nanowire in Eq. (15). This error is further minimized if the middle contacts are very short compared to l1.

4. Example from experiments The four point test structure in Fig. 5 was used to measure qc for contacts to p-type Si nanowires, the conductivity type of which was veried by eld eect measurements reported elsewhere [11]. The nanowires were grown using the vaporliquidsolid technique with trimethylboron as the dopant. An estimate of the B concentration in excess of 1019 cm3 was obtained using secondary ion mass spectrometry (SIMS) for nanowires grown under the same conditions as those in our experiment [11], although the SIMS measurements contain added uncertainty due to the diculty of performing such measurements on nanowires. The nanowires were aligned on Ti (20 nm)/Au (80 nm) end contacts using a eld-assisted assembly technique [10]. Electron beam lithography was then performed to pattern all four top contacts shown in Fig. 5, and after the native oxide on the nanowire was stripped, 80 nm of Ti followed by 200 nm of Au was evaporated on the nanowire. A completed test structure is shown in the eld emission scanning electron microscope (FESEM) image in Fig. 6 for sample 2 from Tables 1 and 2. Table 1 provides the dimensions of two test structures, as measured using FESEM, along with resistance measurements. These data were used along with the equations developed for the four point approach to extract the resistivity of the nanowire, contact resistance, transfer length, and specic contact resistance, as listed for each test structure in Table 2. The length L of the contacts is much greater than 3LT for sample 1 and nearly 3LT for sample 2, so the long contact approximation is fairly well justied. Furthermore, the middle contacts are much shorter than LT, so the use of the four point approach is also reasonable. The values of qc extracted above are calculated assuming that the entire circumference of the nanowire

Then we can measure RT between the two end contacts and determine the contact resistance Rc using

Fig. 5. Four point test structure.

S.E. Mohney et al. / Solid-State Electronics 49 (2005) 227232

231

Fig. 6. FESEM image of an actual four point test structure.

Table 1 Measurements from two test structures Sample 1 2 r (nm) 52 39 2 (lm) 2.24 1.10 1 (lm) 5.86 3.81 V23/I14 (X) 1.93 106 2.60 105 RT (X) 5.85 106 1.41 106

instead of Eq. (2). The transfer length LT then becomes s 2rqc ; 17 LT 3qs and we obtain values of qc of 4.5 104 X cm2 and 5.0 104 X cm2 for samples 1 and 2, respectively. 5. Conclusions Methods for extracting the specic contact resistance of contacts to semiconductor nanowires have been described, and advantages and disadvantages of dierent approaches have been discussed. An example is described in which a four point technique is used to measure specic contact resistances near 5 104 X cm2 for Ti/Au contacts to p-type Si nanowires with diameters of 78 and 104 nm. Other contact metallizations and contacts to Si nanowires of other radii and doping densities are currently under investigation. Acknowledgment This work was supported by the National Science Foundation under grant DMR-0103068 and the Penn State Materials Research Science and Engineering Center (MRSEC) on Nanoscale Science. References
[1] Lieber CM. The incredible shrinking circuit. Sci Am 2001;285: 5864.

Table 2 Parameters extracted from the measurement of the four point test structures Sample 1 2 Rc (X) 4.0 105 2.6 105 qs (X cm) 0.73 0.11 LT (lm) 0.46 1.1 qc (X cm2) 5.9 104 6.7 104

is in contact with the metallization at the ends of the nanowires. This approximation is not perfectly met in our measurements. The nanowires were aligned on Ti/ Au contacts, with the Au in direct contact with the nanowires. Past work in our laboratory has revealed that negligible current is transported through the nanowires before the native oxide is stripped and the top Ti/ Au contact is deposited (this time with the Ti in direct contact with the nanowire). By performing cross-sectional transmission electron microscopy of segments of nanowires beneath other contacts, we have found that the metallization deposited on top of similarly sized nanowires is in direct contact with approximately 75% of the circumference of the nanowire. In this case, Eq. (1) still applies, while we have dI 3pr v0 V dx 2qc 16

232

S.E. Mohney et al. / Solid-State Electronics 49 (2005) 227232 [7] Schroder DK. Semiconductor material and device characterization. New York: John Wiley; 1990. [8] Lew K-K, Reuther C, Carim AH, Redwing JM, Martin BR. Template-directed vaporliquidsolid growth of silicon nanowires. J Vac Sci Technol B 2002;20(1):38992. [9] Carim AH, Lew K-K, Redwing JM. Bicrystalline silicon nanowires. Adv Mater 2001;13(19):148991. [10] Smith PA, Nordquist CD, Jackson TN, Mayer TS, Martin BR, Mbindyo J, et al. Electric-eld assisted assembly and alignment of metallic nanowires. Appl Phys Lett 2000;77(9):1399401. [11] Lew K-K, Pan L, Bogart TE, Dilts SM, Dickey EC, Redwing JM, et al. Structural and electrical properties of trimethylboron-doped silicon nanowires. Appl Phys Lett, in press.

[2] Cui Y, Wei Q, Park H, Lieber CM. Nanowire nanosensors for highly sensitive and selective detection of biological and chemical species. Science 2001;293:128992. [3] Cui Y, Zhong Z, Wang D, Wang WU, Lieber CM. High performance silicon nanowire eld eect transistors. Nanoletters 2003;3(2):14952. [4] Huang Y, Duan X, Cui Y, Lauhon LJ, Kim K-H, Lieber CM. Logic gates and computation from assembled nanowire building blocks. Science 2001;294:13137. [5] Murrmann H, Widmann D. Current crowding on metal contacts to planar devices. IEEE Trans Electron Dev 1969;16:10224. [6] Berger HH. Models for contacts to planar devices. Solid State Electron 1972;15:14558.