Fabrication and characterization of semiconductor nanowires for thermoelectric application 4

Thermal Conductivity of GeNw and SiNw 72 Chapter 4 Thermal conductivity characterization of Ge Nanowires 4.1 Introduction Germanium (Ge) has potential to be used as a thermoelectric material because of its higher electron mobility (or lower electrical resistivity) and lower thermal conductivity compared to silicon (Si). The electrical resistivity of bulk Ge is typically about 100 !"cm, about 1000 times less than that of bulk Si. The thermal conductivity of bulk Ge is 60.2 Wm-1K-1, which is less than half of that of bulk Si. It has been predicted that semiconductors such as Si and Ge can have a thermal conductivity that is two orders lower than its bulk value when shrunk to nanoscale dimensions. This work will investigate the extent to which the thermal conductivity (k) can be decreased in Ge nanowires produced by the VLS process. The investigation will also show the feasibility of using the 3! method, which was used in characterizing the thermal conductivity of carbon nanotubes [7], to measure the k value of the Ge nanowires. Si nanowires were also fabricated for comparison with the result from Ge nanowires. 4.2 Verification of the 3! method setup with platinum microwire When the thermal conductivity of the platinum microwire was initially tested with a home-made current source circuit with op-amp TL074, capacitors and a few resistors, there was some problem in obtaining a thermal conductivity that was close to the theoretical value. Since other researchers who published on the 3# method also used Thermal Conductivity of GeNw and SiNw 73 platinum microwire of the same diameter to test their setup, the first priority was to make sure that the setup can measure the thermal conductivity of a platinum microwire that was close to the theoretical value. Only after the home-made op-amp current source circuit was replaced with the Keithley 6221 AC and DC current source, the 3! characterization setup was then able give results of thermal conductivity of the platinum microwire that match with the theoretical value. Figure 31 shows the platinum sample that was placed into a CERDIP package. The contacts were made by silver paste. The whole package was put in an oven and kept at 90"C for 30 minutes to drive out the moisture. !"#$%&'()'*+,-".$/'/"0%12"%&'3,/4+&'".','56789*'4,0:,#&;' Thermal Conductivity of GeNw and SiNw 74 Listed below are some theoretical values of platinum which are needed for the thermal conductivity measurement: Stefan–Boltzmann constant ! = 5.67*10-8 Emissivity " = 1 (for true blackbody object, ?@@A'.7%78&1&%'7B7/;C&%D' From Figure 38, a slightly non-linear graph was observed from 0 to 0.5V. The electrodes contacting the GeNw are Cr/Au (10/100 nm). This is because contacts to homostructure nanowires are usually dominated by Schottky barriers in both room temperature and low temperature measurements [94, 95]. Wang stated that the conductance of nanowire FETS are dominated primarily by the modulation of the Schottky barrier formed between the nanowire and source/drain contacts, which depends on dopant type and concentration [95] [Quote ref. by Wang]. Thermal Conductivity of GeNw and SiNw 85 Verification for leakage To verify that the conductivity is due to the Ge Nw sample and not due to a leakage path through the substrate, a control sample with 4 electrodes contacting the substrate, but without any GeNw, was probed and tested. As shown in Figure 39, only current at the background noise level of picoampere or tenths of picoamperes could be detected under the sweeping voltage bias. !"#$%&'()' ' *'+,'-'./01'02'13&'4051%0/',67./&'850'56509"%&:'93&%&'13&';'&/&41%0'?&@9A Thermal Conductivity of GeNw and SiNw 86 !"#$#%&'(")*+)%,")-"./)0"1#2#)3245)-") ) All possible permutations of two point contacts on the Ge Nw sample was investigated to ensure that all 4 electrodes have good electrical contact to the GeNw. The two-point resistivity of the GeNw were measured and extracted accordingly as below. Electrode 1 to 2, 2 to 3 and 3 to 4 : 0.349 ! m Electrode 1 to 3 and 2 to 4 : 0.413 ! m Electrode 1 to 4 : 0.482 ! m Literature value for resistivity of Ge (bulk) : 0.305 ! m The two-point resistivity values measured were slightly higher than the theoretical value from literature as these included the effects of contact resistance and the resistance of the GeNw. The measured two-point resistivity generally matches with the literature value of the resistivity of bulk Ge, thus verifying that there are no high resistance issues for the four electrode contacts to the GeNw. For the 3! thermal conductivity measurement, the 4-point probe measurement technique will be used so that the effect of the contact resistance is reduced. 4.3.3 Effect of the process time on GeNw sample Unlike Si, the oxidation of Ge is not self limiting, which means that the GeNw could be fully oxidized after it is left in air for a long time. An experiment was done to test how long a duration was allowed before the GeNw sample becomes fully oxidized. Figures 40(a) and 40(b) shows the SEM images of the GeNw sample that was annealed Thermal Conductivity of GeNw and SiNw 87 after 60 hours and 24 hours respectively from the completion of nanowire fabrication. The SEM image clearly shows that the Ge Nw in Figure 40(a) had disappeared after the annealing while the Ge Nw sample in Figure 40(b) still remained after the annealing. As discussed in the experimental details chapter, annealing is capable of activating the chemical reaction that transforms GeO2 to GeO followed by the complete desorption of GeO [91]. This was why the GeNw sample in Figure 40(a) had disappeared as it is likely the nanowire had been substantially oxidized to GeO2 after being left for 60 hours after fabrication. The subsequent annealing of the nanowire will cause the GeO2 to react with remaining Ge to form GeO, which is then desorbed. To further confirm that there was no nanowire contacting the 4 electrodes, the sample in Figure 40(a) was also tested with the parameter analyzer and no conductivity was detected between pairs of electrodes. Since GeNw oxidized so readily, the total process time from the completion of nanowire fabrication to the device integration to form the 4-electrode contact to the dispersed nanowire and final testing is critical. It is best to keep the total process and measurement time down to within 24 hours after completion of the nanowire fabrication. Thermal Conductivity of GeNw and SiNw 88 !"#$%&'()'*+,'-./'"0+#&'12'34&'5&67'8+09:&'+;;&+:&,' -./'"0+#&'12'34&'5&67'8+09:&'+;;&+:&[...]... to within 24 hours after completion of the nanowire fabrication Thermal Conductivity of GeNw and SiNw 88 !"#$%&'()'*+,'-./'"0+#&'12' 34& '5&67'8+09:&'+;;&+:&,' -./'"0+#&'12' 34& '5&67'8+09:&'+;;&+:& ... roughness of the nanowires and act as a form of phonon-scattering elements at several length scales If precipitates are incorporated carefully, the k of Ge Thermal Conductivity of GeNw and SiNw 77 nanowires. .. resistivity of the GeNw were measured and extracted accordingly as below Electrode to 2, to and to : 0. 349 ! m Electrode to and to : 0 .41 3 ! m Electrode to : 0 .48 2 ! m Literature value for resistivity of. .. before the GeNw sample becomes fully oxidized Figures 40 (a) and 40 (b) shows the SEM images of the GeNw sample that was annealed Thermal Conductivity of GeNw and SiNw 87 after 60 hours and 24