This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

This content was downloaded by: ryang

IP Address: 128.101.142.152

This content was downloaded on 29/08/2014 at 15:24

Please note that terms and conditions apply.

Separation of the piezotronic and piezoresistive effects in a zinc oxide nanowire

View the table of contents for this issue, or go to the journal homepage for more

2014 Nanotechnology 25 345702

(http://iopscience.iop.org/0957-4484/25/34/345702)

Home Search Collections Journals About Contact us My IOPscience

Separation of the piezotronic andpiezoresistive effects in a zinc oxidenanowire

Ren Zhu and Rusen Yang

Department of Mechanical Engineering, University of Minnesota, Minneapolis, MN 55455, USA

E-mail: yangr@umn.edu

Received 14 May 2014, revised 30 June 2014Accepted for publication 3 July 2014Published 7 August 2014

AbstractThe strain-induced band structure change in a semiconductor can change its resistivity, known asthe piezoresistive effect. If the semiconductor is also a piezoelectric material, strain-inducedpolarization charge can control the current transport at the metal-semiconductor contact, which iscalled a ‘piezotronic effect’. Piezotronic effect is intertwined with piezoresistive effect in thestudy of present piezotronic nanowire devices. Decoupling those effects will facilitate thefundamental study on the piezotronic devices and simplify the data analysis in real applications.Here, we report a general method to separate the piezotronic and piezoresistive effects in thesame nanowire, based on modified four-point measurements. Current transport characteristics ofeach contact was extracted and showed different responses to the strain. The piezoresistive effectwas measured in zinc oxide nanowires for the first time, and the result confirmed the dominantrole of piezotronic effect in the strain-induced change of transport characteristics in apiezoelectric semiconductor. This study validates the assumption made in present piezotronicdevices and provides a guideline for further investigation.

Keywords: piezotronic, piezoresistive, nanowire

(Some figures may appear in colour only in the online journal)

1. Introduction

The phenomenon that a mechanical strain can affect thecharge carrier transport plays important roles in various fieldsincluding artificial skin [1], human-machine interface [2],strain engineering [3], etc. For a homogeneous material withelectrodes, three mechanisms may alter its transport char-acteristics—the geometric change, the piezoresistive effectthat changes the material resistivity [4], and the piezotroniceffect that changes the interfacial barrier between the materialand the electrode [5, 6].

Piezoresistance roots in the bending/splitting of electro-nic band structures under strain and is significant in semi-conductors like silicon nanowires [7, 8]. Unlikepiezoresistance, piezotronic effect only appears in semi-conductors with a non-centrosymmetric crystal structure, suchas wurtzite zinc oxide (ZnO) and gallium nitride (GaN). Inthose materials, piezoelectric charges control the barrier

height at the metal-semiconductor interface and modulate thecarrier transport [9–11]. The single crystal ZnO nanowire isthe most studied piezotronic material, due to its simplesynthesis [12, 13], flexibility [14], and robustness [15, 16].Piezotronic nanowire devices demonstrate superior strainsensitivity and have wide applications in strain-gated tran-sistors [17–19], tactile sensor array [20], logic circuits [21],memories [22], etc.

A typical piezotronic device consists of a ZnO nanowireand two metal electrodes on the ends. Consequently, thechange of transport characteristics under strain combines thepiezotronic effect at two contacts and the piezoresistive effectin the nanowire (the geometric influence is relatively smalland can be neglected). Previous studies assumed that thepiezotronic effect dominated the current transport because ofthe asymmetric current change at forward and reverse biases[20, 23]. However, there is no report on the piezoresistance ofZnO nanowires, and quantitative comparison of the

Nanotechnology

Nanotechnology 25 (2014) 345702 (6pp) doi:10.1088/0957-4484/25/34/345702

0957-4484/14/345702+06$33.00 © 2014 IOP Publishing Ltd Printed in the UK1

piezotronic and piezoresistive effects is still lacking. It isessential to develop a general method which can separatethose effects and reveal the strain-dependent transport char-acteristics of individual contacts and the nanowire itself.

2. Experimental section

Separation of the two effects is based on a multi-terminalnanowire device. We first synthesized ultra-long ZnO nano-wires with diameters between 400 and 800 nm through che-mical vapor deposition [24, 25], and then transferred ananowire onto a polyethylene terephthalate(PET) substrate, asillustrated in figure 1(a). Prior to the electrodes deposition, a200 nm SU-8 layer was spin-coated on the substrate andproduced a non-conformal coverage around the nanowire,

figures 1(b) and (c). The SU-8 was thicker on the two sides ofthe nanowire than on the top, so a short oxygen plasma ashingcould expose the top part of the nanowire while the SU-8 ontwo sides remained, figure 1(c). The SU-8 film, called a liftinglayer [26, 27], planarized the step between the nanowire andthe substrate for electrodes deposition (figure 1(d)). Moreimportantly, it could anchor the nanowire to the substrate andtransfer the strain from the substrate to the nanowire duringthe test. The lifting layer process is self-adaptive, as the samecoating and ashing parameters worked successfully for wireswith different diameters. This simple technique can be appliedto similar studies on flexible nanowire electronics.

The electrodes were patterned on the ZnO nanowire witha photolithography liftoff process. It is worth noting that fullycross-linked SU-8 lifting layer is a permanent epoxy and willnot be attacked by the liftoff solvent. We deposited 100 nmtitanium, 200 nm copper, and 10 nm gold through evapora-tion. The titanium layer formed the electric contact with thenanowire, the thick copper layer improved the mechanicalstrength of the electrodes, and the thin gold layer preventedthe electrodes oxidation. After the device fabrication, thevoltage–current curves of nanowires were obtained with acurrent source (Keithley 6221) and a nanovoltmeter (Keithley2182A).

3. Results and discussion

Figure 1(e) shows a scanning electron microscopy image ofthe as-synthesized nanowires, and figure 1(f) is an atomicforce microscopy (AFM) image of a thick wire surrounded bya lifting layer on the substrate. The inset of figure 1(f) plots across section of the AFM scan, in which the exposed part ofthe wire and the two SU-8 slopes are distinguishable. Thefinal device has a series of electrodes on a secured ZnOnanowire, figure 1(g).

Ideally, metal electrodes on a ZnO nanowire form eitherrectifying Schottky contact or linear Ohmic contact, solelydetermined by their work functions. In reality, the contactdepends on the surface condition, crystallinity, and chemicalreaction as well, and thus the contact can have more varieties[28]. By using titanium as the electrodes, we obtained bothlinear Ohmic contacts and nonlinear barrier contacts. Weconfigured a four-point connection (figure 2(a)) to test thepiezoresistance in nanowires with Ohmic contacts. The strainwas applied to the nanowire by bending the substrate, asillustrated in figure 2(b). Because the nanowire stays on thesurface of the substrate and its dimension is much smallerthan the substrate, it has mainly tensile/compressive strainwith minimal shear strain. Since the strain, current density,and electric field are all along the c-axis of the ZnO nanowire,this setup characterizes the longitudinal piezoresistive effect[29], which is the most relevant for piezotronic nanowiredevices.

Figure 2(c) shows typical voltage–current (V–I) curvesfrom the four-point measurement. The resistance decreaseswhen the nanowire is stretched, and increases when it iscompressed. The histogram in figure 2(d) represents the

Figure 1. Fabrication of the multi-terminal device with a ZnOnanowire. (a) Schematic diagram of a nanowire transferred onto apolymer substrate. (b) An SU-8 layer is spin-coated on the substrate.(c) An oxygen plasma ashing exposes the top part of the nanowire.(d) Electrodes are deposited on the exposed nanowire. (e) Scanningelectron microscopy image of as-synthesized ZnO nanowires; thenanowire grows along its c-axis. (f) Atomic force microscopy imageof a nanowire in the SU-8 lifting layer with the top part exposed; theinset shows a section view of the scan. (g) Optical microscopy imageof a nanowire with multiple electrodes.

2

Nanotechnology 25 (2014) 345702 R Zhu and R Yang

relative change of resistance under 0.25% tensile strainamong 27 nanowires. All the nanowires have lower resistanceunder the tensile strain, with the ΔR/R value between −0.3%and −6.8%. Since a pure geometric change increases theresistance by 0.4% at 0.25% strain, the resistivity of nano-wires must have decreased here. For the first time we showthat the c-orientated n-type ZnO nanowire has a negativelongitudinal piezoresistive coefficient.

The variance of the ΔR/R value among nanowires cancome from two factors. First, piezoresistance depends onsemiconductor parameters such as the carrier concentration[30]. Based on the resistance and geometry of those ZnOnanowires, we found that the resistivity can be more than anorder of magnitude different, suggesting great variance in thecarrier density among nanowires. Second, the 0.25% strain isthe strain on the surface of the flexible substrate. The actualstrain in the nanowire depends on its diameter (discussedlater) and the adhesion between the SU-8 and the nanowire,and thus can deviate to different extents from the 0.25%value.

Variation of the four-point measurement, as illustrated infigure 3(a), can introduce the contact 1 and contact 2 inaddition to the nanowire resistance. At a certain current,subtracting Va from Vb gives the potential difference acrosscontact 1 (ΔV1). Similarly, subtracting Va from Vc reveals thepotential difference across contact 2 (ΔV2). When the Ti/ZnOcontact has a barrier, the change of ΔV1 and ΔV2 under strainrepresents the piezotronic effect. Therefore, measurements in

figure 3(a) can separate the piezotronic and piezoresistiveeffects.

Figure 3(b) compares the V–I curves of nanowire resis-tance, contact 1, and contact 2 in a nanowire with barriercontacts. The curves of contact 1 and contact 2 are obtainedafter the subtraction of Va from Vb and Vc in figure 3(a).Under the same current, much higher voltage across contact 1and contact 2 than that across the nanowire indicates that theinterfacial barriers are more significant than the nanowireresistance. More importantly, V–I curves of the two contactshave prominent change under 0.25% tensile strain while thechange of nanowire resistance is negligible. Thus the piezo-tronic effect at the contacts is much greater than the piezo-resistive effect in the nanowire, which supports directly theassumption made in previous studies [6].

Moreover, decoupling the piezotronic effects at the twocontacts enables us to extract detailed information at theindividual contact. Figure 3(b) shows that contact 1 behavesas a rectifying barrier under zero strain, and the tensile strainlowers the barrier and increases the reverse current. Contact 2has a lower barrier height than contact 1 and is close to anOhmic contact, suggesting that the tunneling may dominatethe current transport [31]. The strain further reduces thebarrier height. In terms of strain sensing, contact 2 has highersensitivity than contact 1. Such distinctive transport char-acteristic of contact 1 and contact 2, when lumped in a two-terminal device, complicates the analysis and makes it chal-lenging to predict the device behavior. In fact, most ZnOpiezotronic devices in the literature formed different electric

Figure 2.Measurement of the piezoresistive effect in devices with Ohmic contacts. (a) Schematic diagram of the four-point measurement. (b)Schematic diagram of the experimental setup used to stretch the ZnO nanowire; the polymer substrate is deflected by a programmable linearmotor. (c) Voltage–current curves of a nanowire under different strains, obtained with the four-point configuration in (a); the inset is a zoom-in view of the curves at the high current end. (d) Histogram of the relative change of resistance, measured from 27 nanowire segments undertensile strain.

3

Nanotechnology 25 (2014) 345702 R Zhu and R Yang

contacts at two ends even with the same metal, so it isnecessary to analyze the two contacts independently.

Figure 3(c) shows the V–I curves of another multi-terminal device. While the piezotronic effect still dominates,this nanowire has an unusual piezoresistive effect. The tensilestrain increases nanowire resistance with ΔR/R as high as23%. The positive piezoresistive coefficient opposes the resultin figure 2(d) where all the nanowires with Ohmic contactshave a negative coefficient. Such a phenomenon has beenobserved in several devices with barrier contacts, and here wepropose a possible explanation. Although the four-pointmeasurement excludes the contacts, the effect of contacts canextend into the thin nanowire. Numerical calculation showsthat when ZnO nanowire forms Schottky contacts with elec-trodes, the region near the contact has a lower carrier densitythan the rest part of the nanowire. Strain-induced polarizationcharge at the contact further changes the carrier distribution

[10]. Therefore, a strain could alter the carrier profile withinthe nanowire, which in turn affects the overall resistance. Themeasured piezoresistive effect depends on the contacts, and itmay be more properly called an ‘apparent’ piezoresistiveeffect.

Finally, we discuss a feature of the multi-terminal pie-zotronic device, i.e., the electrodes are on the side surface ofthe nanowire instead of its two ends. The piezotronic effect iscaused by the polarization charge [10, 11]. For a ZnOnanowire under uniform tensile/compressive strain, thecharge only appears at the two ends. Here, the piezotroniceffect with the electrode on the side surface is ascribed to thenon-uniform strain along the nanowire. We used COMSOLMultiphysics to simulate the elastic and piezoelectric behaviorof a ZnO nanowire embedded in a polymer matrix. Forsimplicity, we assume ZnO is an insulator and there is noslipping between the nanowire and the polymer. Two ends of

Figure 3. Comparison of the piezotronic effect and the piezoresistive effect in devices with barrier contacts. (a) Schematic diagram of threedifferent connections to extract the nanowire resistance and two contacts. (b) Voltage–current (V–I) measurement of a typical device; the V–Icurve of nanowire resistance is directly obtained, while the V–I curves of contact 1 and contact 2 are derived by subtracting Va from Vb and Vc

in (a). (c) Voltage–current curves of another device obtained with the similar methods; the measured nanowire resistance increases under thetensile strain.

4

Nanotechnology 25 (2014) 345702 R Zhu and R Yang

the polymer are subject to uniform tensile loads, as shown infigure 4(a), and the strain is transferred from the polymer tothe nanowire. Figure 4(b) shows the normal strain along thecentral axis of nanowires with different diameters. The strainnear the two ends of the nanowire is much lower than that inthe polymer far away from the nanowire (indicated by thedashed line). As the diameter becomes larger, the averagestrain in the nanowire is lower. The strain distribution resultsfrom the difference between the elastic moduli of the ZnO andthe polymer. Figure 4(c) shows the electric polarization alongthe central axis of the nanowires, which has the similarcharacteristics as the normal strain. By calculating thedivergence of the polarization field, we found that thepolarization charge density is the highest at the two ends, butnon-zero on the side surface. Positive charge appears near the(0001) end of the nanowire, and the charge density decaysfurther away from the end surface. The (0001̄) end has asimilar distribution of negative charge.

The charge distribution from the ends to the middle of thenanowire can produce a rich variety of piezotronic effects,depending on the positions of the electrodes. When thecharges near the two electrodes have the same polarity,

piezotronic effects raise or reduce the barrier height of bothelectrodes, e.g., the device in figure 3(b). If the charges havedifferent polarities, the piezotronic effects are opposite at thetwo contacts, such as the device in figure 3(c). When theelectrodes are in the region where the polarization charge isnegligible and the change of the barrier height is insignificant,the device will not show significant piezotronic effects. If thestrain distribution can be under good control, it is not onlyfeasible but also beneficial to put contacts on the side surfaceof a nanowire, in order to have multiple piezotronic effectswith desired properties.

4. Conclusions

The objective of this work is to independently study thepiezotronic effect and the piezoresistive effect in a ZnOnanowire. With a four-point connection, we extracted thepiezoresistive effect in ZnO nanowires for the first time andfound that a tensile strain reduces the resistivity. However, abarrier contact could influence the apparent piezoresistiveeffect even with the four-point connection. Modified four-point measurements revealed the piezotronic effect at a barriercontact, which proves to be more significant than the piezo-resistive effect. In addition, the feasibility and advantage ofusing piezotronic effects on the side surface of nanowires hasbeen discussed. The general methods developed in this workcontribute to the fundamental understanding of the piezo-tronic effect and provide guideline for the practicalapplications.

Acknowledgments

The authors are truly grateful for the financial support fromthe Department of Mechanical Engineering and the College ofScience and Engineering of the University of Minnesota.Research is also supported by NSF (ECCS-1150147). Theelectron microscopy image was obtained in the Character-ization Facility, University of Minnesota, which receivespartial support from NSF through the MRSEC program. Thedevice fabrication was performed in the Minnesota NanoCenter, a part of NSF-funded National NanotechnologyInfrastructure Network.

References

[1] Someya T, Sekitani T, Iba S, Kato Y, Kawaguchi H andSakurai T 2004 Proc. Natl Acad. Sci. USA 101 9966–70

[2] Kim D-H et al 2011 Science 333 838–43[3] Sun Y, Thompson S E and Nishida T 2007 J. Appl. Phys. 101

104503[4] Barlian A A, Park W-T, Mallon J R, Rastegar A J and

Pruitt B L 2009 Proc. IEEE 97 513–52[5] Wang X, Zhou J, Song J, Liu J, Xu N and Wang Z L 2006

Nano Lett. 6 2768–72[6] Wu W, Pan C, Zhang Y, Wen X and Wang Z L 2013 Nano

Today 8 619–42

Figure 4. Simulation of the non-uniform strain in ZnO nanowires. (a)Schematic diagram of a nanowire embedded in a polymer matrix. (b)Normal strain along the central axis of nanowires with the samelength but different diameters; the dashed line indicates the normalstrain in the polymer far away from the nanowire. (c) Electricpolarization along the central axis of nanowires with the same lengthbut different diameters.

5

Nanotechnology 25 (2014) 345702 R Zhu and R Yang

[7] He R and Yang P 2006 Nat. Nanotechnology 1 42–6[8] Cao J X, Gong X G and Wu R Q 2007 Phys. Rev. B 75 233302[9] Yu E T, Dang X Z, Yu L S, Qiao D, Asbeck P M, Lau S S,

Sullivan G J, Boutros K S and Redwing J M 1998 Appl.Phys. Lett. 73 1880–2

[10] Zhang Y, Liu Y and Wang Z L 2011 Adv. Mater. 23 3004–13[11] Shi J, Starr M B and Wang X 2012 Adv. Mater. 24 4683–91[12] Xu S and Wang Z 2011 Nano Res. 4 1013–98[13] Huang M H, Wu Y, Feick H, Tran N, Weber E and Yang P

2001 Adv. Mater. 13 113–6[14] Hoffmann S, Östlund F, Michler J, Fan H J, Zacharias M,

Christiansen S H and Ballif C 2007 Nanotechnology 18205503

[15] Gao Z, Ding Y, Lin S, Hao Y and Wang Z L 2009 Phys. StatusSolidi RRL 3 260–2

[16] Li P et al 2014 Nano Lett. 14 480–5[17] Zhou J, Gu Y, Fei P, Mai W, Gao Y, Yang R, Bao G and

Wang Z L 2008 Nano Lett. 8 3035–40[18] Han W, Zhou Y, Zhang Y, Chen C-Y, Lin L, Wang X,

Wang S and Wang Z L 2012 ACS Nano 6 3760–6[19] Pradel K C, Wu W, Zhou Y, Wen X, Ding Y and Wang Z L

2013 Nano Lett. 13 2647–53

[20] Wu W, Wen X and Wang Z L 2013 Science 340 952–7[21] Wu W, Wei Y and Wang Z L 2010 Adv. Mater. 22 4711–5[22] Wu W and Wang Z L 2011 Nano Lett. 11 2779–85[23] Hu Y, Klein B D B, Su Y, Niu S, Liu Y and Wang Z L 2013

Nano Lett. 13 5026–32[24] Zhu G, Zhou Y, Wang S, Yang R, Ding Y, Wang X,

Bando Y and Wang Z L 2012 Nanotechnology 23 055604[25] Zhu R, Zhang W, Li C and Yang R 2013 Nano Lett. 13 5171–6[26] Zimmler M A, Stichtenoth D, Ronning C, Yi W,

Narayanamurti V, Voss T and Capasso F 2008 Nano Lett. 81695–9

[27] Storm K, Halvardsson F, Heurlin M, Lindgren D,Gustafsson A, Wu P M, Monemar B and Samuelson L 2012Nat. Nanotechnology 7 718–22

[28] Brillson L J and Lu Y 2011 J. Appl. Phys. 109 121301[29] Mason W P and Thurston R N 1957 J. Acoust. Soc. Am. 29

1096–101[30] Harkey J A and Kenny T W 2000 J. Microelectromech. Syst. 9

226–35[31] Sze S M and Ng K K 2007 Physics of Semiconductor Devices

(New York: Wiley)

6

Nanotechnology 25 (2014) 345702 R Zhu and R Yang