Diameter-Dependent Thermal Transport in Individual ZnO Nanowires and its Correlation with Surface Coating and Defects

Transcription

1 full papers Thermal Transport Diameter-Dependent Thermal Transport in Individual ZnO Nanowires and its Correlation with Surface Coating and Defects Cong Tinh Bui, Rongguo Xie,* Minrui Zheng, Qingxin Zhang, Chorng Haur Sow, Baowen Li,* and John T. L. Thong* A systematic study of the thermal transport properties of individual single-crystal zinc oxide (ZnO) nanowires (NWs) with diameters in the range of 21 nm is presented. The thermal conductivity of the NWs is found to be dramatically reduced by at least an order of magnitude compared to bulk values, due to enhanced phonon-boundary scattering with a reduction in sample size. While the conventional phonon transport model can qualitatively explain the temperature dependence, it fails to account for the diameter dependence. An empirical relationship for assessing diameter-dependent thermal properties is observed, which shows an approximately linear dependence of the thermal conductivity on the cross-sectional area of the NWs in the measured diameter range. Furthermore, it is found that an amorphous-carbon layer coating on the NWs does not perturb the thermal properties of the NW cores, whereas 3 kev Ga + ion irradiation at low dose ( cm 2 ) leads to a remarkable reduction of the thermal conductivity of the ZnO NWs. C. T. Bui, [+] Prof. C. H. Sow, Prof. B. Li, Prof. J. T. L. Thong NUS Graduate School for Integrative Sciences and Engineering National University of Singapore Singapore, 11746, Republic of Singapore phylibw@nus.edu.sg; elettl@nus.edu.sg Dr. R. Xie, [+] M. Zheng, Prof. C. H. Sow, Prof. B. Li Department of Physics National University of Singapore Singapore, 11742, Republic of Singapore phyxrg@nus.edu.sg Dr. R. Xie, Prof. B. Li Centre for Computational Science and Engineering (CCSE) National University of Singapore Singapore, 11746, Republic of Singapore Dr. R. Xie, Prof. J. T. L. Thong Department of Electrical and Computer Engineering National University of Singapore Singapore, 11776, Republic of Singapore C. T. Bui, Dr. Q. Zhang Institute of Microelectronics 11 Science Park Road, Singapore Science Park II Singapore 1768, Republic of Singapore [+] These authors have contributed equally to this work. DOI: 1.12/smll Introduction Thermal transport in low-dimensional materials has attracted considerable attention in recent years, due to its enormous practical implications in applications ranging from thermal management to thermoelectric energy conversion. [1 3] The development of microfabricated suspended devices to measure the thermal conductance of low-dimensional materials [4 6] has paved the way for a number of significant studies that have advanced the understanding of thermal transport at the nanoscale. [7 16] For example, thermal rectification [7] and the breakdown of Fourier s law [8] have been observed in carbon nanotubes and boron nitride nanotubes, while significant thermoelectric performance has been achieved in rough Si nanowires (NWs) as a result of enhanced phononboundary scattering. [9] However, this experimental field is still in its infancy, and systematic studies of size effect on the thermal properties of nanostructures are still rare. Indeed, only Si NWs and carbon nanotubes have been extensively studied so far. [6 13] Thus the understanding of the low-dimensional thermal transport problems has largely been relegated to the domain of theoretical studies due to the scarcity of relevant experimental data. 738 wileyonlinelibrary.com 212 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 212, 8, No.,

2 Thermal Transport in ZnO Nanowires and its Correlation with Surface Coating and Defects Apart from the experimental challenge in placing suspended samples, the measurement of thermal conductance of an individual NW generally suffers from two problems. The primary one lies with the unknown thermal contact resistance between the sample and substrate, especially for short/thick NWs with large thermal conductance. Unlike electrical contact resistance whose effect can be completely eliminated by adopting a four-point measurement configuration, a solution to the thermal contact resistance problem remains elusive due to the lack of robust tools to probe it. This raises the question of whether the thermal characteristics being measured are those of the sample, the contact interface, or a combination of both. Another problem is the surface coating, which might introduce unexpected surface interactions with the NW cores. Transmission electron microscopy (TEM) imaging has shown that many NWs are generally covered with native oxide layers or surface contamination with thicknesses typically on the order of a few nanometers. Several theoretical simulations have shown that the presence of a thin layer of amorphous coating can strongly modify the thermal properties of the NW cores. [16 19] Experimental studies have also demonstrated that an asymmetric surface coating of Pt C composite onto nanotubes can give rise to thermal rectification effects. [7] More recently, it was found that the substrate and polymer residues dramatically affect the thermal conductivity of graphene. [2 22] Therefore, the influence of such surface coatings on NWs has yet to be elucidated in interpreting the intrinsic thermal properties measured from such NWs. In this paper, we report a systematic and comprehensive study of the diameter dependence of the thermal transport properties of individual ZnO NWs in the temperature range of 77 4 K using a suspended micro-electro thermal system (METS). ZnO NWs were selected for study because they are of great interest in many applications ranging from piezoelectric power generation to field-effect transistors, [23] and knowledge about the thermal properties of the ZnO NWs in these devices is valuable, but experimental data is not available so far. In addition, unlike Si and many other semiconductor NWs, ZnO NWs are free from the problem of a surface layer of native oxide, thus making them an ideal system for the study of intrinsic thermal properties and their diameter dependence. Particularly, we have made significant advances in the measurement approach to address the problem of contact thermal resistance. We have observed that the thermal conductivity of individual ZnO NWs is dramatically reduced compared to bulk values, and decreases much more significantly than predicted by the conventional phonon transport model. Furthermore, we have also studied the effects of surface coating and ion irradiation on the thermal properties of the ZnO NWs. Figure 1. a) SEM image of a METS device with an individual ZnO NW bonded on the heater (left) and sensor (right). b) SEM images of three ZnO NWs with different diameters measured in this study. c) Low-magnification TEM image of a METS device with an individual ZnO NW (Inset: its SAED pattern). d) High-resolution TEM image of the ZnO nanowire. The scale bars shown in a c are 2 μm. 2. Results and Discussion One important feature of our measurement approach is that we can directly examine the microstructure of the same NW that was used in thermal transport measurements and determine its diameter, surface roughness and contamination, crystalline orientation, and defects, thus directly correlating the transport properties with the microstructure of the same individual NW. This was made possible because the METS devices were micromachined with a through-substrate hole (Figure 1 and Figure S1 in the Supporting Information (SI)) and can be mounted onto a custom-built TEM holder for microstructure analysis in the TEM. Figure 1a shows a typical scanning electron microscopy (SEM) image of a METS device consisting of an individual ZnO NW bonded on the heater and sensor using electron-beam-induced deposition (EBID). TEM observation showed that during the sample preparation and bonding process, amorphous carbon (a-c) layers were inevitably coated on the surface of the NWs by EBID of hydrocarbon residues from the SEM chamber. To remove this surface contamination, the NWs were cleaned before thermal transport measurements using an Evactron radiofrequency (RF) plasma cleaner. Figure 1b shows the cleaned samples with diameters of 1, 12, and 29 nm. A low-magnification TEM image of a METS device with an individual ZnO NW is shown in Figure 1c. The selected-area electron diffraction (SAED) pattern and high-resolution image show that all the ZnO NWs studied in this work are single crystalline with wurtzite structure and that they are grown along a small 212, 8, No., Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 739

3 full papers C. T. Bui et al. c-axis orientation. The surface of the cleaned NWs is seen to be free from an amorphous layer and is naturally rough with root-mean-square (rms) roughness value of 1. nm. A second important feature of our measurement approach is that we have used a noncontact spatially-resolved electronbeam heating technique in combination with the METS device to probe the contact thermal resistance as well as the internal thermal resistance of the heater and sensor. As described previously, [22] a focused electron beam (e-beam) in the SEM was used to induce localized heating along the sample. By scanning the e-beam from the heater to the NW, and recording the corresponding temperature rise of the heater, ΔT hi, and the sensor ΔT si, the cumulative thermal resistance R i from the e-beam heating position to the heater can be spatially resolved as R i = R b (a a i )/(1 + a i ), where a is the ratio of the temperature rise ΔT h /ΔT s when the e-beam is focused on the heater, a i is the ratio, ΔT hi /ΔT si, when the e-beam is scanned at position i, and R b is the thermal resistance of the connecting beams which was measured using a method described by L. Shi et al. [4] Figure 2a shows R i as a function of position from the heater to the contact area of the NW, which shows that the value of R i increases slowly upon scanning the e-beam to the contact area. Two steeper slopes are clearly observed from the R i profile when the e-beam traverses the two gaps without Pt metallization due to the low thermal conductivity of the supporting SiN x membrane. When the e-beam scans from the contact area towards the NW (Figure 2b), R i increases very slowly at the contact, indicating a very low contact thermal resistance. It then increases linearly with a much steeper slope when the e-beam scans along the fully suspended part of the NW. The linear dependence of thermal resistance with length suggests that the thermal transport in ZnO NW is purely diffusive, in contrast to ballistic or superdiffusive transport. Measurements of the thermal contact resistance of many samples bonded with Pt C composite at different temperatures show that the thermal contact resistance is negligible compared to the total measured thermal resistance of the NW, suggesting the deposition of Pt C composite at the contact is an effective way to minimize the contact resistance. However, an intercept of K/W in R i is observed at the starting point of suspended NW segment at 3 K (Figure 2b). This intercept includes the distributed internal thermal resistance of the heater island and the contact resistance, where the dominant contribution comes from the distributed internal thermal resistance (Figure 2a), which is consistent with thermal modeling results. [22] As shown in Figure 2c, the overall thermal resistance from the heater to the contact area is relatively small (< K/W) at all the measured temperatures, but it can introduce significant measurement error if not accounted for, especially for short/thick NWs with low thermal resistance (or high thermal conductance). We have made the suspended segment of all the samples sufficiently long (. μm), so that the intrinsic thermal resistance R NW of the NW (on the order of K/W) is dominant in the measured value. Moreover, R NW was extracted by subtracting the distributed internal thermal resistances and the contact resistances of both heater and sensor from the measured total thermal resistance R, further improving the accuracy in the subsequent thermal conductivity calculations. Figure 3a shows the measured thermal conductivity, κ, as a function of temperature for the ZnO NWs with diameters of 7, 84, 12, 166, and 29 nm. For all the samples, as temperature increases, the thermal conductivity first increases to a maximum then decreases. Compared to the thermal conductivity of bulk single-crystalline ZnO (inset of Figure 3a), there are two important features that are common to all the NWs we measured. Firstly, the measured thermal conductivities are dramatically reduced, by more than one order of magnitude a) R i (K/W) 1x1 6 8x1 6x1 4x1 2x Position (µm) b) R i ( K/W) c) R (K/W) 1.2x1 7 9.x1 6 6.x1 6 3.x Position 1.6x1 6 (µm) 1.4x x1 6 1.x Figure 2. a) Thermal resistance profile scanned from the heater to the contact area of the NW at 3 K. b) Thermal resistance profile scanned along the NW starting from the contact area at 3 K. c) Temperature dependence of the cumulative thermal resistance from the heater to the contact area Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 212, 8, No.,

4 Thermal Transport in ZnO Nanowires and its Correlation with Surface Coating and Defects a) b) nm 166 nm 12 nm 84 nm 7 nm model (29 nm) model (12 nm) model (7 nm) ~ T -1 Figure 3. a) Temperature dependence of thermal conductivity of ZnO NWs of different diameter. Inset shows the thermal conductivity of bulk ZnO from modeling. [24] b) Log log plot in the temperature range of 16 4 K, showing κ T α with the exponent α is in the range of ; the two curves T 1. and T 1 are shown to guide the eyes. compared to that of bulk ZnO ( 1 W m 1 K 1 at 3 K) [24] and, with decreasing diameter, the corresponding thermal conductivity is reduced over the entire measured temperature range. This clearly indicates that enhanced phonon-boundary scattering exerts a strong suppression effect on phonon transport in ZnO NWs. In normal bulk crystal, boundary scattering is important only at very low temperatures (typically below 3 K) due to the scattering of long wavelength phonons. In contrast, the reduction of the thermal conductivity is observed for the NWs over the entire experimental temperature range, suggesting that boundary scattering is also significant even at high temperatures. Secondly, the NWs show a maximum of thermal conductivity at 12 K to 1 K, which is much higher than the peak temperature of bulk ZnO (3 3 K). [24] The shift of the peak suggests that boundary scattering is dominant until 12 K, and Umklapp scattering, which reduces the thermal conductivity with temperature, becomes important and comes to dominate over boundary scattering at higher temperatures (>1 K) ~ T -1. ZnO Bulk Modeling (Ref. 24) The dramatic reduction in thermal conductivity with diameter and the shift of the peak to higher temperature has also been observed for Si NWs. [9 11] However, the thermal conductivity of ZnO NWs decreases very rapidly beyond the peak with temperature as T α (the exponent α is in the range of ) (Figure 3b). This is in sharp contrast to the temperature dependence observed for Si NWs of similar diameters, which shows only a slight decline in thermal conductivity beyond the peak. [9 11] For bulk pure single crystal, the thermal conductivity varies with temperature as T 1 where the main scattering mechanism is Umklapp scattering. The fall-off sharper than T 1 suggests that, besides boundary and Umklapp scattering, some other scattering mechanisms must also play important roles. From TEM observation, the NWs are single crystalline without bulk defects, such as dislocations and voids. Thus phonon scattering by bulk defects is unlikely. The photoluminescence spectrum of the as-grown ZnO NWs shows a broad visible emission band centered at 3 nm due to deeplevel recombination, indicating the presence of intrinsic point defects (oxygen vacancies or Zn interstitials) in the ZnO NWs (see Figure S2 and S3 in the SI). Furthermore, for ZnO, the natural abundance Zn comprises three isotopes (48.6% 64 Zn, 27.9% 66 Zn, and 18.8% 68 Zn, mass variance: ), and the natural abundance of O is 99.8% 16 O. The large isotopic disorder can affect the thermal conductivity dramatically. [2] In contrast, Si (92.23% 28 Si, 4.67% 29 Si, and 3.1% 3 Si) has much less isotopic disorder (mass variance: ). The significantly larger isotopic disorder along with the presence of intrinsic point defects (oxygen vacancies or Zn interstitials) might be the major reasons for the sharper fall-off in the thermal conductivity at intermediate and high temperatures. To gain a more quantitative understanding of the dia meter-dependent thermal transport in ZnO NWs, we used a conventional Callaway model to analyze the experimental data, taking into account the effect of phonon-boundary scattering as follows: [26] κ = k B 2π 2 v (k BT h )3 θd x 4 e x (e x 1) 2 dx AT 4 + (B 1 + B 2 )T x 2 (1) + v/l where x = ħω/k B T, T is absolute temperature, θ D is Debye temperature (399. K for ZnO), [24] v is the average sound velocity of phonon (384 m/s in bulk ZnO), L is the characteristic length of material, ω is circular frequency, k B is the Boltzmann constant, ħ is the reduced Planck constant, and A, B 1, and B 2 are the relaxation parameters. In this equation, three scattering process are considered as expressed by three terms in the denominator: i) impurity scattering, including point defects (oxygen vacancies or Zn interstitials) and isotope scattering, in which relaxation time is proportional to ω 4, which is independent of temperature; ii) threephonon normal scattering and Umklapp scattering, in which relaxation time is proportional to (ω 2 T 3 ) 1 ; and iii) boundary scattering, described by a constant relaxation time L/v based on the Casimir model. [27] In the present study, L would be represented by the diameter (d) of the NWs. We used the thermal conductivity of the NW with largest diameter (L = 29 nm) to adjust the phonon scattering parameters, and A and B 1 + B 2 were numerically optimized as K 4 s 1 and small 212, 8, No., Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 741

5 full papers C. T. Bui et al K 8K 1x1 4 2x1 4 3x1 4 4x1 4 d 2 (nm 2 ) Figure 4. Diameter dependence of thermal conductivity of the ZnO NWs at 8 and 3 K. The thermal conductivity increases linearly with cross-sectional area ( d 2 ) in the measured diameter range. The dasheddotted lines are shown to guide the eyes..23 K s 1, respectively. Then L was varied based on the measured values, while the other parameters were fixed. It can be seen that the fitting results obtained by varying L do not match the thermal conductivities for different diameters. Attempts were also made to use different forms of Umklapp scattering, which also yielded results that do not agree well with the experimental data. The largest discrepancy between experimental data and the conventional phonon transport model is that the measured thermal conductivity decreases much more rapidly with diameter than the conventional model, especially at high temperature. This suggests that the conventional phonon-boundary scattering model based on the Casimir limit, [27] which assumes that boundary scattering is frequency independent and temperature independent, cannot account well for the diameter-dependent thermal transport properties of ZnO NWs. Large discrepancies between experimental data and the conventional phonon transport model have also been observed for Si NWs. [1,11] In light of the systematic experimental data for Si NWs, several theoretical approaches have been proposed. [11,28] For example, Chen et al. have suggested that phonon-boundary scattering is highly frequency dependent, and ranges from nearly ballistic to completely diffusive, which can explain the temperature dependence observed for thin Si NWs. [11] Mingo has put forward a complete phonon dispersion approach to account for both temperature dependence and diameter dependence of Si NWs. [28] However, currently there is no exact theoretical model for ZnO NWs below 4 K. Our systematic experimental data should stimulate further theoretical investigations of this intriguing diameter-dependent thermal transport in ZnO NWs. As theoretical studies on the thermal transport properties of ZnO NWs are not yet widely available, we plotted the thermal conductivity as a function of d 2 in Figure 4, which may provide a generally useful although approximate rule of thumb for assessing the size effect. Strikingly, it can be seen that the thermal conductivity scales approximately linearly with the cross-sectional area ( d 2 ) in the measured diameter range at both low temperature (8 K) and room temperature (3 K). Moreover, the thermal conductivity increases more rapidly with diameter at 8 K than at 3 K. This can be understood because phonon-boundary scattering in limiting phonon mean free path is much more important at lower temperatures. To elucidate the effect of surface coating on the thermal transport properties of ZnO NWs, we intentionally coated an a-c layer onto the NW surface using EBID after the thermal conductance measurement on the pristine NWs. The ZnO NWs were intentionally exposed to an e-beam for a prolonged period ( 1 min). A thin layer of a-c coated the surface of the NWs in a shell-like manner as a result of EBID from hydrocarbon residues in the SEM chamber. The schematic of the coating process and a typical TEM image of a ZnO NW with an a-c shell are shown in Figure a,b. The thickness of the a-c shell is typically 1 nm for all samples, depending on the exposure time. Figure c shows the thermal conductance as a function of temperature for the ZnO NW before (d 84 nm) and after coating with the a-c shell (d 12 nm) (see SEM image in Figure c). As can be seen, after the a-c shell coating, the thermal conductance shows only a slight increase, while the shape of the curve follows the same trend as the one without the a-c shell. A repeat of the thermal conductance measurement was also carried out after the carbon contamination had been removed using RF plasma cleaning, which recovers the curve of the pristine NW. Based on the difference of the thermal conductance and the thickness of the shell measured in TEM, we extracted the thermal conductivity of a-c (<.8 W m 1 K 1, inset of Figure ), which is in good agreement with the values reported previously. [29] Similar results were obtained for all the other ZnO NWs of different diameters. These observations suggest that the a-c shell generally acts as an independent parallel thermal path alongside the ZnO NW, and does not introduce detectable perturbation to the thermal transport in the NW core. Thus for practical devices where the NWs are generally supported on amorphous dielectrics or polymers, the presence of the substrate should not adversely affect their thermal conductivity. Of course, when the NW diameter shrinks down to the range where phonon confinement occurs, the effect of the surface layer and substrate will become important, as predicted by theoretical atomistic simulations. [16 19] Such intriguing crossover calls for further experimental investigations. Furthermore, we have studied the effect of ion-irradiationinduced defects on the thermal transport properties of ZnO NWs. We prepared another set of ZnO NWs with diameters of 1, 9, and 126 nm, and then used ion-beam-induced deposition (IBID) to bond both ends of the NWs under Ga + ion beam irradiation at 3 kv and a beam current of 1 pa. During inspection of the devices, the NWs were inevitably exposed to the ion irradiation. Precautions were taken to keep the ion dose on the suspended part of the NWs as low as possible (~ cm 2 ). TEM images show that even a slight exposure can cause significant damage to the ZnO lattice. As shown in the insets of Figure 6a,b, after ion irradiation, the surface was significantly roughened, and the crystalline lattice of ZnO near the surface was partially Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 212, 8, No.,

6 Thermal Transport in ZnO Nanowires and its Correlation with Surface Coating and Defects a) b) c) Thermal Conductance (nw/k) ZnO NW core Thermal conductivity (W/m.K) a-carbon shell κ of a-carbon Figure. a) Schematic of ZnO nanowire with and without a-c shell coating. b) TEM image of a ZnO NW core coated with a-c shell; scale bar: 2 nm. c) Thermal conductance measurement of a ZnO NW with and without a-c shell. Top-right insets: SEM images of the ZnO NW with and without a-c shell. Bottom-left inset: Extracted thermal conductivity of a-c. transformed into an amorphous structure. Additional defects, such as vacancies, interstitials, and Ga impurities, were also produced in the ZnO NW under ion irradiation. The SAED pattern confirmed that the NW remains single crystalline (inset, Figure 6b). Note that an nm layer of an a-c shell was precoated the NW during the bonding process. Compared to the non-irradiated samples of similar diameters, the thermal conductivities of the irradiated samples, all show a notable reduction, especially in the lower temperature range (Figure 6a). A key feature caused by ion irradiation is that the temperature-dependence curves are substantially flattened and the peaks are shifted to higher temperatures. The reason can be attributed to the enhanced defect scattering which comes to dominate at intermediate temperatures, as the intermediate frequency phonons are most efficiently scattered by the defects. This phenomenon is the most serious for the thinnest NW, which shows an almost flat curve over a wide temperature range. Because a focused-ion beam is often used to deposit and create metallic contacts to nanodevices, precautions should be taken to keep the active regions away from the ion exposure as it may result in significant degradation of their transport properties. a-carbon shell ZnONW core With a-carbon shell Without a-carbon shell Without C With C 84 nm 12 nm 3. Conclusion In summary, we have systematically studied the diameter-dependent thermal conductivity of ZnO NWs in the temperature range of 77 4 K using METS devices. It is found that the thermal conductivity is strongly suppressed due to enhanced phonon-boundary scattering, and the suppression is much stronger than predicted by the conventional phonon transport model based on the Casimir model to account for phonon-boundary scattering. An empirical relationship for assessing diameter-dependent thermal properties is presented, which shows an approximately linear dependence of the thermal conductivity on the crosssectional area ( d 2 ) of the NWs in the measured diameter range. Temperaturedependent measurements show that beyond the low-temperature maximum, the thermal conductivity decreases with temperature as T α (where α is in the range of ), indicating strong impurity scattering (due to point defects, isotopes, etc.) and Umklapp scattering at intermediate and high temperatures. Furthermore, the thermal transport properties of the ZnO NWs are found to be insensitive to the surface a-c contamination, while they can be greatly degraded by ion irradiation even at low dose. These findings are useful for the development of nanostructure-based thermoelectric devices, as well as for efficient thermal management in NW-based electronic and optical devices. Furthermore, the diameter dependence observed in this study is expected to stimulate further theoretical investigations into the intriguing effect of size on the thermal transport in nanostructures. 4. Experimental Section ZnO NWs were synthesized via vapor transport process in a sealed horizontal tube furnace (Carbolite CTF 12/7/7). [3] The thermal transport measurements were carried out using a METS device in vacuum (<1 mbar) in the temperature range of 77 4 K. Similar to the devices described previously, [22,31] the METS device consists of a 3 nm thick SiN x layer as mechanical support, integrated with a Pt resistive loop that serves as both heater and temperature sensor. Each of the heater/sensor islands is supported by Pt leads on SiN x beams (2 μm in width and 4 μm in length) from the substrate (see Figure S1 in the SI). A nanomanipulator (Kleindiek MM3A-EM) in a scanning electron microscope (Philips XL3 FEG) was used to pick up and place individual NWs onto the METS devices. To reduce thermal contact resistance, two ends of the NWs were bonded onto the Pt electrodes by Pt C composite small 212, 8, No., Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 743

7 full papers a) b) nm 1nm 9nm 126nm Figure 6. a) Temperature dependence of thermal conductivity of ionirradiated ZnO NWs with different diameters. Inset: Low-magnification TEM image of an ion-irradiated ZnO NW. b) High-resolution TEM image of an ion-irradiated ZnO NW. Inset: its SAED pattern. using EBID (FEI Quanta 2-3D). During the pick-up and bonding process, amorphous carbon layers were inevitably coated on the surface of the NWs by EBID of the hydrocarbon residues from the SEM chamber. To remove this surface contamination, the NWs were cleaned using an Evactron RF plasma cleaner attached to the SEM chamber at 14 W in.4 mbar of air for 2 h. The thermal resistance of the samples was measured by the previously described method. [4,22,31] A direct current I h was passed through the heater loop for the heating. The four-terminal electrical resistance of the Pt loops R h and R s were acquired using lock-in amplifiers by passing a very small alternating current ( na, 217 Hz for the heater, and 1917 Hz for the sensor) that superimposes on I h. The heater and sensor temperatures (T h and T s ) were then obtained based on R h and R s, which were calibrated against the substrate temperature of the METS (T sub ) which sat on a Janis ST-4 Cryostat. Based on the temperature changes and the heating power P at the heater, the thermal resistance R (the reciprocal of thermal conductance G) can be determined from the relation [4] C. T. Bui et al. T R = R b ( h 1 T s ) (2) where T h,s = T h,s T sub and the thermal resistance of the connecting beams is R b = T h + T s (3) P The measured R is the overall thermal resistance from the heater to the sensor, which includes the distributed internal thermal resistance of the heater and sensor islands, the contact resistance, and the thermal resistance of the suspended NWs. To extract the intrinsic thermal resistance (R NW ) of the suspended NW segment, an electron-beam heating technique was employed for the spatially resolved measurement of the cumulative thermal resistance. Instead of using Joule heating from the heater to establish a temperature gradient for thermal conductance measurement, a focused e-beam ( kv,.1 na) in an SEM (FEI Nova NanoSEM 23) was used to induce localized heating along the sample, while both the heater and sensor (in the previous configuration) acted as temperature sensors. A Gatan C12 nitrogen-cooled cold stage was used to control the substrate temperature of the sample. By scanning the e-beam from the heater to the contact area of the ZnO NW, the cumulative thermal resistance including distributed internal thermal resistance of the heater and the contact resistance was obtained. R NW was extracted by subtracting the distributed internal thermal resistances and the contact resistances at both heater and sensor from the measured total thermal resistance R. The e-beam scanning along the sample does not introduce a detectable change in the thermal resistance of the sample; after the e-beam scanning, repeated measurement of total thermal resistance R produces essentially an identical value. After the thermal conductance measurement on these pristine samples, the NWs were then intentionally coated with an a-c layer, and the thermal conductance of the NWs were measured again to investigate the effect of surface coating on the thermal transport properties. To obtain the direct correlation between the transport properties with the microstructure of the same NW under study, the METS device was mounted onto a custom-built TEM holder (Nanofactory SA2) for microstructure analysis of the nanowires in the TEM (JEOL TEM 21F). Some of the NWs had broken just before TEM observation after the thermal measurement, and for these samples the diameters were determined from earlier SEM images (FEI Nova NanoSEM 23). The thermal conductivity was then calculated using κ = L/R NW a, where L is the length of the suspended NW segment, and a is the cross-sectional area calculated using the NW diameter as measured by either TEM or SEM. The error bars in the thermal conductivity values include both the calibration error of temperature, and the uncertainty in the determination the dimensions of the NWs. The dominant uncertainty for thin NWs comes from the determination of the NW diameter measured by SEM, which has a typical resolution of nm based on multiple measurements of different sections of the same NWs calibrated in TEM. Four-point electrical measurements were also conducted on some samples, which showed that the resistances of the NWs are typically larger than 1 MΩ, which allows us to neglect chargecarrier contributions to thermal transport. Low-voltage (2 kv) SEM observations indicated that buried charge arising from e-beam irradiation is negligible in the ZnO NW Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 212, 8, No.,

8 Thermal Transport in ZnO Nanowires and its Correlation with Surface Coating and Defects Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements We would like to thank Dr. Minggang Xia at Xi an Jiaotong University for helpful discussions. This work is supported by research grants R and R from the National University of Singapore. [1] D. Cahill, W. Ford, K. Goodson, G. Mahan, A. Majumdar, H. Maris, R. Merlin, S. Phillpot, J. Appl. Phys. 23, 93, [2] A. Majumdar, Science 24, 33, 777. [3] M. S. Dresselhaus, G. Chen, M. Y. Tang, R. Yang, H. Lee, D. Wang, Z. Ren, J. Fleurial, P. Gogna, Adv. Mater. 27, 19, [4] L. Shi, D. Y. Li, C. H. Yu, W. Y. Jang, D. Y. Kim, Z. Yao, P. Kim, A. Majumdar, J. Heat Transf. 23, 12, [] K. Schwab, E. A. Henriksen, J. M. Worlock, M. L. Roukes, Nature 2, 44, [6] K. Hippalgaonkar, B. Huang, R. Chen, K. Sawyer, P. Ercius, A. Majumdar, Nano Lett. 21, 1, [7] C. W. Chang, D. Okawa, A. Majumdar, A. Zettl, Science 26, 314, [8] C. W. Chang, D. Okawa, H. Garcia, A. Majumdar, A. Zettl, Phys. Rev. Lett. 28, 11, 793. [9] A. Hochbaum, R. Chen, R. Delgado, W. Liang, E. Garnett, M. Najarian, A. Majumdar, P. Yang, Nature 28, 41, [1] D. Li, Y. Wu, P. Kim, L. Shi, P. Yang, A. Majumdar, App. Phys. Lett. 23, 83, [11] R. Chen, A. Hochbaum, P. Murphy, J. Moore, P. Yang, A. Majumdar, Phys. Rev. Lett. 28, 11, 11. [12] P. Kim, L. Shi, A. Majumdar, P. McEuen, Phys. Rev. Lett. 21, 87, 212. [13] M. T. Pettes, L. Shi, Adv. Funct. Mater. 29, 19, [14] A. Moore, M. Pettes, F. Zhou, L. Shi, J. App. Phys. 29, 19, [1] F. Zhou, J. Szczech, M. Pettes, A. Moore, S. Jin, L. Shi, Nano Lett. 27, 7, [16] N. Mingo, L. Yang, Phys. Rev. B 23, 68, [17] D. Donadio, G. Galli, Nano Lett. 21, 1, [18] M. Hu, K. P. Giapis, J. V. Goicochea, X. Zhang, D. Poulikakos, Nano Lett. 211, 11, [19] J. Chen, G. Zhang, B. Li, J. Chem. Phys. 211, 13, 148. [2] J. Seol, I. Jo, A. Moore, L. Lindsay, Z. Aitken, M. Pettes, X. Li, Z. Yao, R. Huang, D. Broido, N. Mingo, R. Ruoff, L. Shi, Science 21, 328, [21] M. T. Pettes, I. Jo, Z. Yao, L. Shi, Nano Lett. 211, 11, [22] Z. Wang, R. Xie, C. T. Bui, D. Liu, X. Ni, B. Li, J. Thong, Nano Lett. 211, 11, [23] Z. Wang, Mater. Sci. Eng. R 29, 64, [24] J. Alvarez-Quintana, E. Martínez, E. Pérez-Tijerina, S. A. Pérez-García, J. Rodríguez-Viejo, J. Appl. Phys. 21, 17, [2] N. Yang, G. Zhang, B. Li, Nano Lett. 28, 8, [26] J. Callaway, Phys. Rev. 199, 113, 146. [27] H. B. G. Casimir, Physica 1938,, 49. [28] N. Mingo, Phys. Rev. B 23, 68, [29] J. Bullen, K. E. O Hara, D. Cahill, O. Monteiro, A. von Keudell, J. Appl. Phys. 2, 88, [3] S. Deng, H. Fan, M. Wang, M. Zheng, J. Yi, R. Wu, H. Tan, C. Sow, J. Ding, Y. Feng, K. Loh, ACS Nano 21, 4, 49. [31] R. Xie, C. T. Bui, B. Varghese, Q. Zhang, C. Sow, B. Li, J. Thong, Adv. Funct. Mater. 211, 21, Received: September 29, 211 Published online: December 9, 211 small 212, 8, No., Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 74