Letter pubs.acs.org/NanoLett
Quantitative Thermopower Profiling across a Silicon p−n Junction with Nanometer Resolution Byeonghee Lee,† Kyeongtae Kim,‡ Seungkoo Lee,§ Jong Hoon Kim,§ Dae Soon Lim,§ Ohmyoung Kwon,*,‡ and Joon Sik Lee† †
School of Mechanical and Aerospace Engineering, Seoul National University, Seoul 151-744, Korea Department of Mechanical Engineering, Korea University, Seoul 136-701, Korea § Department of Materials Science and Engineering, Korea University, Seoul 136-701, Korea ‡
S Supporting Information *
ABSTRACT: Thermopower (S) profiling with nanometer resolution is essential for enhancing the thermoelectric figure of merit, ZT, through the nanostructuring of materials and for carrier density profiling in nanoelectronic devices. However, only qualitative and impractical methods or techniques with low resolutions have been reported thus far. Herein, we develop a quantitative S profiling method with nanometer resolution, scanning Seebeck microscopy (SSM), and batch-fabricate diamond thermocouple probes to apply SSM to silicon, which requires a contact stress higher than 10 GPa for stable electrical contact. The distance between the positive and negative peaks of the S profile across the silicon p−n junction measured by SSM is 4 nm, while the theoretical distance is 2 nm. Because of its extremely high spatial resolution, quantitative measurement, and ease of use, SSM could be a crucial tool not only for the characterization of nano-thermoelectric materials and nanoelectronic devices but also for the analysis of nanoscale thermal and electrical phenomena in general. KEYWORDS: Scanning Seebeck microscopy, diamond thermocouple probe, quantitative profiling, thermopower, carrier density, nanometer resolution he thermoelectric figure of merit ZT is defined as S2σT/k, where S is the Seebeck coefficient or thermopower, σ the electrical conductivity, k the thermal conductivity, and T the mean temperature. In order to compete with conventional refrigerators and generators, ZT should be higher than 3. However, even the best ZT value of bulk thermoelectric materials such as Bi2Te3 and its alloys still remains at around 1 near 300 K.1 Commercial thermoelectric devices based on these bulk materials also have ZT lower than 1. Because of the low efficiency, the use of thermoelectric materials has been quite limited. However, after the discovery that the ZT value can be enhanced up to 2.4 in nanostructured materials, an enormous amount of research has begun in this field.1−5 For bulk materials, due to the couplings between material properties (k−σ, S−σ), ZT has been saturated as noted above. However, for nanostructured materials, the boundary scattering of phonons can lower k without suppressing σ as much, while the quantum confinement can increase S without reducing σ too much, and these combined effects can enhance ZT.2,3 Hence, the quantitative profiling of S as well as k with nanometer resolution6 is critical for enhancing ZT through the nanostructuring of materials. In addition, since the S value of a semiconductor is a function of its charge carrier density, the quantitative profiling of S can also be useful in the quantitative carrier density profiling of
T
© 2012 American Chemical Society
nanoelectronic devices. Though such tools as scanning spreading resistance microscopy (SSRM)7 and scanning capacitance microscopy (SCM)8 are available, SSRM is not quantitative and SCM loses its sensitivity at the nanoscale. Even with the demands mentioned above, the quantitative profiling of S at the nanoscale has never been achieved yet. For example, the potential Seebeek microprobe9 can profile S quantitatively, but its spatial resolution is only 10 μm. For nanoscale S profiling, Lyeo et al. developed scanning thermoelectric microscopy (SThEM).10 The distance between the positive and negative peaks of the S data measured point-bypoint by SThEM is 8 nm.10 Although the data are scientifically significant, SThEM is not widely used for the following reasons. First, it cannot control the tip position and measure the thermoelectric signal simultaneously. As a result, the measurement steps are extremely difficult and cumbersome, making it very time-consuming and impractical. Second, because SThEM cannot measure the temperature of the tip−sample contact, quantitative measurement is impossible. Third, because SThEM uses an atomically sharp metal tip, it is not applicable for silicon, which is arguably the most important engineering Received: April 10, 2012 Revised: August 9, 2012 Published: August 13, 2012 4472
dx.doi.org/10.1021/nl301359c | Nano Lett. 2012, 12, 4472−4476
Nano Letters
Letter
maximized at the junction, the Joule heating with a frequency of 2ω becomes most intense there.13,14 Then, the periodic heat flux through the tip−sample contact establishes a periodic temperature gradient in the sample, which leads to the generation of a thermoelectric voltage of frequency 2ω, VS, which is
material. In order to establish a stable ohmic contact with the silicon surface, the tip should be able to endure contact stresses higher than 10 GPa.11 Finally, because SThEM is a DC measurement technique, the measured signal is subject to dc noise due to the photoionization across the p−n junction, the complete removal of which is quite difficult. In order to profile S across a silicon p−n junction and get rid of the dc noise, Kim et al. developed an ac measurement technique using a diamond-like carbon (DLC) coated probe.12 However, because the tip radius increases as much as the thickness of the DLC coating and the data are still obtained in a point-by-point manner, its spatial resolution deteriorates and the measured profile fails to show the sharp variation of S near the depletion region. In order to overcome the limitations of the currently available methods, we develop SSM and batch-fabricate diamond thermocouple probes, the core part of SSM. The principle of SSM and the structure of the tip of the diamond thermocouple probe are illustrated in Figure 1. To heat the tip
VS = S(x , y)ΔT
(1)
where S(x,y) is the local S in the sample and ΔT is the periodic temperature amplitude of the contact in the sample. In eq 1, the Seebeck coefficient of the metal film is ignored because it is much smaller than that of the semiconductor sample. The temperature gradient in the exposed diamond tip is also ignored because the thermal conductivity of diamond is extremely high and the height of the exposed part is very small (
