- Email: [email protected]
Development of a two dimensional scanning Seebeck coefficient measurement system by a micro-probe method
Intermetallics 32 (2013) 233e238
Contents lists available at SciVerse ScienceDirect
Intermetallics journal homepage: www.elsevier.com/locate/intermet
Development of a two dimensional scanning Seebeck coefficient measurement system by a micro-probe method Go Nakamoto a, *, Yuuji Nakabayashi b a b
School of Materials Science, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa 923-1292, Japan Center for Nano Materials and Technology, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa 923-1292, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 5 April 2012 Received in revised form 2 July 2012 Accepted 17 August 2012 Available online 9 October 2012
A Seebeck micro-probe measurement system with high spatial resolution has been developed to examine two-dimensional spatial distribution of Seebeck coefficient for thermoelectric materials at room temperature. A contact area of 10 mmf is realized by using a Copper probe tip fabricated by a mechanical machining process. The two-dimensional spatial distribution of the Seebeck coefficient for the Bismuth eTelluride and ZinceAntimonide systems has been measured. The sign inversion of the Seebeck coefficient from p- to n-type is clearly detected along the crystal growth direction in the BismutheTelluride system. On the other hand, the anisotropic Seebeck coefficient reflecting grain distribution is conspicuously observed in the ZinceAntimonide system. Ó 2012 Elsevier Ltd. All rights reserved.
Keywords: A. Intermetallics, miscellaneous B. Thermoelectric properties PACS: 72.20.Pa
1. Introduction
2. Seebeck micro-probe measurement system (SMP-MS)
A micro-probe method has been developed to examine local physical properties and/or distribution of physical quantities originating from inhomogeneity of chemical composition and crystal structure of materials [1e4]. Furthermore, the micro-probe method has been applied to use a high-speed combinatorial examination of thermoelectric properties for various thermoelectric materials [5e7]. Recently, the development of new measurement methods for physical properties at a micro- or nano-scopic area becomes so important and is highly desirable associated with the recent development of functionally graded materials and nano-structured materials. The purpose of this study is to develop a two dimensional scanning Seebeck coefficient measurement system by a micro-probe method with high spatial resolution and to examine the two-dimensional distribution of the Seebeck coefficient for the BismutheTelluride and ZinceAntimonide thermoelectric systems.
2.1. Fabrication of Seebeck micro-probe tips
* Corresponding author. Tel.: þ81 761 51 1558; fax: þ81 761 51 1149. E-mail address: [email protected] (G. Nakamoto). 0966-9795/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.intermet.2012.08.012
Fig. 1 shows photographs of two different types of Seebeck micro-probe tips fabricated in this study. Copper was selected as a probe-tip material because of the high-thermal and electrical conductivities and low Seebeck coefficient as well as an easy machining process. A contact area of 10 mmf was obtained by a mechanical machining process using a microscope (Fig. 1 (a)). The dimension of the probe tip is 4 mm in diameter and 4 mm in length with flat surfaces on both sides by spark-erosion cutting to attach heaters on the flat surfaces as explained later. An angle of relief of 120 for the probe tip is adapted to secure mechanical strength of the probe tip and to lower thermal resistance between the tip and a sensor position in the probe. A Teflon coated AlumeleChromel thermocouple of 0.076 mmf was selected as a sensor and attached to a drilled hole with a diameter of 0.2 mmf located 0.5 mm above from the probe tip by a silver paste as shown in Fig. 1 (b). Two strain gauges (Kyowa KFL-1-120-C1-16) connected in a serial way with a total electrical resistance of 240 U were used as a heater to generate temperature difference between top and bottom surfaces of a sample. These heaters are attached on both side walls of the probe tip by a wanish.
234
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 1 (c) shows a photograph of a whole figure of the Seebeck micro probe. The Copper probe tip is attached to a metal shaft made by stainless steel mediated by an insulating acryl spacer to suppress heat leaking from the Copper tip to an upper part of the probe. A coil spring with a spring constant of 10 gf/mm is equipped in an upper acryl cylinder to make good thermal and electrical contacts between the probe and sample. A typical contact pressure is estimated to be of the order of 100 kgf/mm2 when a contact area of 10 mmf and a spring stroke of 1 mm are assumed. Any remarkable effect of the contact pressure on Seebeck coefficient is not observed. Seebeck coefficient does not depend on the contact pressure when the pressure is changed by five times by increasing the spring stroke from 1 to 5 mm. As shown later, the measured values of the Seebeck coefficient for the BismutheTelluride and ZinceAntimonide systems are in good agreement with those measured by other commercially available apparatus. The AlumeleChromel thermocouple and lead wires of the heater are fixed on a probe arm made by Aluminum through an acryl wire-guide plate and connected with twisted Copper wires at a thermal anchor located on a thermal bath made by Copper. 2.2. Constitution of the SMP-MS and a measurement procedure
Fig. 1. Two different types of the Seebeck-micro probe tips fabricated by a mechanical machining process (a) and (b), and the whole figure of the Seebeck micro probe with the mechanical machined tip (c).
Fig. 2 shows a schematic drawing of the Seebeck micro-probe measurement system (SMP-MS) developed in this study. The system consists of two digital nano-voltmeters (Agilent 34420A), a dc current source (Agilent E3647A), a function generator (nf DF1906), XY and Z stages (SURUGA SEIKI KS202-70 and PZG413L05AG), a stepping motor controller for the stages (SURUGA SEIKI D223) and a personal computer. All the instruments are controlled by the PC via GPIB-USB interface (Agilent 82357B) with a full-automatic measurement program composed by LabView. Data analysis software is also developed in this study. The measurement condition such as shape and size of a measurement area, a magnitude of heater current, trigger frequency and points for data acquisition, a probe stroke and temperature equilibrium condition are input as initial parameters of the measurement.
Fig. 2. Schematic drawing of the Seebeck micro-probe measurement system.
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
A two-probe differential method was employed to measure Seebeck coefficient. A plated sample with a thickness of about 2 mm is fixed on to the thermal bath made by Copper using a silver paste. An AlumeleChromel thermocouple is also attached to the thermal bath by a silver paste. The Seebeck micro-probe tip is touched down to the sample surface and the heater attached on the both side walls of the probe tip gives a temperature difference DT between the top and bottom surfaces of the sample. The temperature difference was about 0.5e2 K when 12e24 mA of the heater current was supplied for 5 s in the present study. The generated Seebeck voltages VA and VC between the Alumele Alumel wires and between the ChromeleChromel wires of both thermocouples at the probe and thermal bath, respectively, were measured in a heating process. It is noted that a function generator is used to give trigger signal two digital nano-voltmeters to measure a timely-synchronized Seebeck voltages VA and VC accurately and quickly. The typical trigger frequency and points are 4 Hz and 20 points, respectively, giving 20 measurement points per 5 s. The Seebeck voltages VA and VC between two pairs of the wires were plotted against the temperature difference DT. The Seebeck voltages VA and VC are given by VA ¼ (S SA)DT and VC ¼ (S SC)DT, where S, SA and SC represent the Seebeck coefficients of the sample, Alumel and Chromel, respectively. Therefore, the Seebeck coefficient of the sample at the probe position was obtained from the slopes of the VeDT plot by a least square fitting. The data tables for Seebeck coefficient and Seebeck voltage of the Alumel and Chromel wires are composed based on the NIST data [8]. Two Seebeck coefficients calculated from the Seebeck voltages of the Alumel and Chromel wires coincide with each other within the order of 103 mV/K. The thermal bath and micro-probe are mounted on motor-driven XY and Z stages, respectively, and the XY stages are able to move with a spatial resolution of 1 mm. Thus, the spatial distribution of Seebeck coefficient can be measured at a selected area by moving the stages. A two-dimensional contour map and line profiles along the X and Y directions of Seebeck coefficient are displayed on the measurement screen. The measurement is started after stabilizing temperature change within 1% of the temperature difference DT in a relaxation process, taking about 100 s. Therefore, the average total measurement time for one point is about 120 s at the present. The measurement system is covered with a box sealed by Aluminum foils to keep temperature constant and cut a highfrequency noise during a measurement. Each connection part of lead wires is also covered with a heat-insulating material to eliminate temperature difference among connection points which causes an extra Seebeck voltage. An actual temperature fluctuation in a sample space is found to be less than 1 K during a measurement for about five days in a temperature controlled room by an air conditioner. The extraction of line profile data and combination of the more than two data for different measurement areas can be possible by the data analysis software developed in this study.
235
The sample was fixed on to the thermal bath using a silver paste. An area of 4 4 mm2 was measured 30 times at 297 K with a step of 1 mm for both X and Y directions. Thus, 480 measurements were performed for the small area of 4 4 mm2 in total. Temperature fluctuation during the measurements was within 1 K. Fig. 3 shows the histogram of the Seebeck coefficient for the standard Ni. The average value of the Seebeck coefficient is 12.51 mV/K. The Seebeck coefficient of Ni at room temperature has been reported previously [9,10]. The values are around 19 mV/K. The absolute measurement value by the SMP-MS is smaller than those values by about 6e7 mV/K. The difference in the Seebeck coefficient of Ni between the present value and reference ones may be caused by the temperature difference between the tip and the sensor position in the probe. Since the thermal conductivity of Ni is only about four times smaller than that of Cu, temperature difference between the tip and the sensor arising from thermal resistance ratio between the sample and probe tip should not be ignored. When we consider the thermal circuit of the SMP-MS, the correct Seebeck coefficient of the sample S is given by a following equation;
S ¼
DTm Sp DT 0 Sm DT DTm
(1)
where DT, DTm, Sm, Sp and DT0 represent the actual temperature difference in the sample, the measured temperature difference between the bottom face of the sample and upper sensor, the measured Seebeck coefficient, the Seebeck coefficient of the probe tip and the extra temperature difference in the probe tip (between the contact point and sensor position in the probe tip), respectively. Thus, DTm is equal to DT þ DT0. Now, when we assume the thermal conductivity of 4000 and 910 mW/Kcm and the Seebeck coefficients, and 19.1 and 1.71 mV/K for Ni and Cu, respectively, the ratio of thermal resistance between the Cu probe tip and the measured Ni is estimated to be about 3:7. The temperature difference in the probe tip DT0 becomes 30% of the total DTm for the present case. On contrary, thermoelectric materials have extremely small thermal conductivity of about the order of 10 mW/Kcm, which is four hundred times smaller than that of Cu. Therefore, the thermal resistance of the probe tip is compared to be extremely smaller than that of the sample. DT0 is estimated to be so small less than 1% of DTm and can be ignored in this case. In fact, the measured Seebeck coefficients of BismutheTelluride and ZinceAntimonide are in good agreement with those by other measurement apparatus as
3. Results and discussion 3.1. Standard Ni In order to confirm the accuracy of the measurement for the SMP-MS, Ni has been measured as a standard material. A Ni rod in purity of 4 N with a diameter of 8 mmf was cut into a parallel piped shape with a dimension of 2 2 17 mm3. The sample was sealed in a silica tube in vacuum and annealed by an electric furnace to remove strain.
Fig. 3. Histogram of the Seebeck coefficient for standard Ni. The solid curve represents the result of Gaussian fitting with the average value of 12.51 mV/K and the half width of half maximum of 0.65 mV/K.
236
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 4. The cross section cut parallel to the crystal growth direction for the Bismuthe Telluride system taken by an optical polarized microscope. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. The detailed distribution of the Seebeck coefficient in the BismutheTelluride system measured for a rectangular area near the sign inversion boundary surrounded by red lines in Fig. 4 with measurement steps of X and Y ¼ 50 mm.
shown later. On the other hand, the half width of half maximum of the distribution of the Seebeck coefficient is 0.65 mV/K. The scattering of the data is small enough within 5% of the average value. Thus, we have judged that the SMP-MS can be enough to be used for measurements of Seebeck coefficient for thermoelectric materials.
bottom part of the ingot exhibits p-type conduction with S ¼ 250 mV/K. In the p-type region, S has a nearly constant value. The sign reversal observed around X ¼ 0 mm (5 mm from the top edge). The upper part of the ingot shows n-type conduction of S ¼ 250 mV/K. These Seebeck coefficients for both the p- and ntype areas are consistent with those observed in the previous study. A transition width (the width between Smax , p and Smax , n) is estimated to be about 5000 mm. No remarkable enhancement of the Seebeck coefficient near the boundary is observed in both the pand n-type conduction areas. More detailed result is shown in Fig. 6. A measurement was done near the boundary area of 2000 5000 mm2 with a 50 mm step surrounded by red lines in Fig. 4. The sign inversion from p- to n-type is observed more obviously. The sign inversion from p to n type along the crystal growth direction observed in this system is attributable to segregation of excess Te at the upper part of the ingot. The excess Te as a donor is wiped out at the upper part of the ingot during crystal growth. As a result, the bottom part exhibits p type, whereas the upper one including larger amount of the Te shows n-type conduction. In fact, the position at which the sign inversion takes place strongly depends on the excess Te concentration d and crystal growth rate. The sign inversion is observed at the upper position for larger Te concentration and faster growth rate.
3.2. BismutheTelluride system The BismutheTelluride system (Bi,Sb)2Te3þd by partial substitution of Bi with Sb and addition of excess Te was grown by a onedirectional solidification method [11]. In the previous study, it has been found that the conduction type changes in an ingot depending on growth rate and excess amount of Te [11]. The details of the inversion of the conduction type in the system have been examined by the SMP-MS. Fig. 4 shows a photograph of the cross section cut parallel to the growth direction for the BismutheTelluride crystal taken by an optical polarized microscope. Typical dimension of the grown crystal is 30 mm in length and 8 mm in diameter. Several large and long grains are found along the crystal growth direction, indicating that a preferred orientation exists along the crystal growth direction. The similar grain distribution is observed on the opposite side. A line profile of the Seebeck coefficient along the crystal growth direction (along the yellow line in Fig. 4) is displayed in Fig. 5. The
Fig. 5. The line profile of the Seebeck coefficient in the BismutheTelluride system along the crystal growth direction indicated by a yellow line in Fig. 4.
3.3. ZinceAntimonide system The ZinceAntimonide system has been known as a thermoelectric material with the high figure of merit of 1.3 in the moderate temperature range [12]. Strong anisotropy of the thermoelectric properties is expected because of the rhombohedral crystal structure. However, it has not been succeeded to grow the large single crystal enough to examine the anisotropy of the thermoelectric property since many structural phase transitions exist between the melting point and room temperature [13]. The SMP-MS is able to measure local Seebeck coefficient in a small area of about several 10 mm. Therefore, it is also expected that the SMPMS is one of the powerful tools to examine anisotropic physical properties of materials that are obtained only as polycrystalline samples. The ingot of ZinceAntimonide was also grown by a onedirectional solidification method with a dimension of 30 mm in length and 10 mm in diameter [14]. The disk plate sample with a thickness of 2 mm was taken out of the ingot by slicing perpendicular to the crystal growth direction and used for the SMP measurements. Fig. 7 shows the grain distribution of the sample taken by an optical polarized microscope. Cracks and grain boundaries are clearly observed in the figure. It is confirmed that both the sides of the disk plate sample have the same grain distribution. Typical grain size is found to be of the order of about
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 7. The grain distribution of the surface of the disk sample perpendicular to the crystal growth direction in the ZinceAntimonide system. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
1 mm2. The SMP measurements have been carried out for a square area of 1 1 mm2 with steps of X ¼ Y ¼ 50 mm indicated by red lines in Fig. 7. The spatial distribution of the Seebeck coefficient is clearly observed as shown in Fig. 8. It is revealed that the distribution of the Seebeck coefficient strongly resembles that of the grain. Therefore, the distribution of the Seebeck coefficient originates from the anisotropy reflecting the difference in crystal orientation of each grain. The distribution of the Seebeck coefficient ranges from 100 to 120 mV/K except the crack region. These values are consistent with those measured by other measurement apparatus. Judging from the haziness of the grain boundaries and cracks in the distribution of the Seebeck coefficient, it is suggested that an effective contact area of the SMP measurement is larger than the actual contact area by a few ten percent. Fig. 9 shows the histogram of the Seebeck coefficient for the measurement area except the crack region displayed by Fig. 8. Solid curves represent the result of a least square fitting assuming Gaussian function. It is found that the profile consists of at least two components with the average values of 102 and 112 mV/K with the half width of half maxima of 2.50 and 3.17 mV/K, respectively. The former peak with the lower value corresponds
237
Fig. 9. The histogram of the Seebeck coefficient in the ZinceAntimonide system for the square area except the crack region indicated by red lines in Fig. 7. The solid curves represent the results of a least square fitting assuming Gaussian distribution.
to the fringe area, whereas, the latter one to the island area at the center. 4. Conclusion The two dimensional scanning Seebeck coefficient measurement system by a micro-probe method has been developed with high spatial resolution. A Seebeck micro-probe with a contact area of 10 mm is fabricated by a mechanical machining process. The full automatic measurement program and data analysis software are also composed by LabView. The distribution of the Seebeck coefficient has been measured for the BismutheTelluride and ZinceAntimonide systems. The inversion from p- to n-type conduction along the crystal growth direction is clearly observed in the BismutheTelluride system, caused by a gradient of Te concentration along the crystal growth direction. On the other hand, the anisotropic Seebeck coefficient reflecting the grain distribution is obviously found in the Zince Antimonide system. These results indicate that the Seebeck micro-probe measurement system developed in the present study is one of very powerful tools to examine local Seebeck coefficient of materials, reflecting local difference in a chemical composition and a crystal structure of materials. It is also expected that this method can be applied widely to examine physical properties of nano-structured and functionally graded materials by realizing the smaller probe size and addition of a temperature control system. Acknowledgment This work was supported by Grant-in-Aid for Scientific Research (C), Japan Society for the Promotion of Science (No. 23560363). References Schilz J, Kaysser WA. Mater Sci Eng 2003;A362:17. Müller E, Drasar C, Ni HL, Zhao XB, Karpinski G, Müller E. J Mater Sci 2005;40:605. Satou K, Yamashita O, Odahara H, Tomiyoshi S. Appl Phys 2006;A84:103. Nakamoto G, Kurisu M. J Electron Mater 2009;38:916. Yamamoto A, Kukuruznyak D, Ahmet P, Chikyow T, Ohuchi FS. Mater Res Soc Proc 2003;804:3. [6] Ikeuchi S, Shimada K, Takahashi Y, Ishii Y, Yamamoto A. ULVAC Tech J 2008; 69. [7] Kosuga A, Kurosaki K, Muta H, Stiewe C, Karpinski G, Müller E, et al. Mater Trans 2006;47:1440. [1] [2] [3] [4] [5]
Fig. 8. The spatial distribution of the Seebeck coefficient in the ZinceAntimonide system for the square area indicated by red lines in Fig. 7.
238
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
[8] Croarkin M, Guthrie W, Burns G, Kaeser M, Strouse G. Temperature-electromotive force reference functions and tables for the letter-designated thermocouple types based on the ITS-90. Natl Inst Stand Technol Monogr 1993; 175:630. Available online at: http://srdata.nist.gov/its90/main/. T.C. tables NIST. [9] Blatt FJ, Rowe V, Schroeder PA, Cox JE. Phys Rev Lett 1967;18:395.
[10] Laubitz MJ, Matsumura T, Kelly TJ. Can J Phys 1976;54:92. [11] Huong NT, Setou Y, Nakamoto G, Kurisu M, Kajihara T, Mizukami H, et al. J Alloys Compd 2004;368:44. [12] Caillat T, Fleurial JP, Borshchevsky A. J Phys Chem Solids 1997;58:1119. [13] Nakamoto G, Kinoshita K, Kurisu M. J Alloys Compd 2007;436:65. [14] Souma T, Nakamoto G, Kurisu M. J Alloys Compd 2002;340:275.
Contents lists available at SciVerse ScienceDirect
Intermetallics journal homepage: www.elsevier.com/locate/intermet
Development of a two dimensional scanning Seebeck coefficient measurement system by a micro-probe method Go Nakamoto a, *, Yuuji Nakabayashi b a b
School of Materials Science, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa 923-1292, Japan Center for Nano Materials and Technology, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa 923-1292, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 5 April 2012 Received in revised form 2 July 2012 Accepted 17 August 2012 Available online 9 October 2012
A Seebeck micro-probe measurement system with high spatial resolution has been developed to examine two-dimensional spatial distribution of Seebeck coefficient for thermoelectric materials at room temperature. A contact area of 10 mmf is realized by using a Copper probe tip fabricated by a mechanical machining process. The two-dimensional spatial distribution of the Seebeck coefficient for the Bismuth eTelluride and ZinceAntimonide systems has been measured. The sign inversion of the Seebeck coefficient from p- to n-type is clearly detected along the crystal growth direction in the BismutheTelluride system. On the other hand, the anisotropic Seebeck coefficient reflecting grain distribution is conspicuously observed in the ZinceAntimonide system. Ó 2012 Elsevier Ltd. All rights reserved.
Keywords: A. Intermetallics, miscellaneous B. Thermoelectric properties PACS: 72.20.Pa
1. Introduction
2. Seebeck micro-probe measurement system (SMP-MS)
A micro-probe method has been developed to examine local physical properties and/or distribution of physical quantities originating from inhomogeneity of chemical composition and crystal structure of materials [1e4]. Furthermore, the micro-probe method has been applied to use a high-speed combinatorial examination of thermoelectric properties for various thermoelectric materials [5e7]. Recently, the development of new measurement methods for physical properties at a micro- or nano-scopic area becomes so important and is highly desirable associated with the recent development of functionally graded materials and nano-structured materials. The purpose of this study is to develop a two dimensional scanning Seebeck coefficient measurement system by a micro-probe method with high spatial resolution and to examine the two-dimensional distribution of the Seebeck coefficient for the BismutheTelluride and ZinceAntimonide thermoelectric systems.
2.1. Fabrication of Seebeck micro-probe tips
* Corresponding author. Tel.: þ81 761 51 1558; fax: þ81 761 51 1149. E-mail address: [email protected] (G. Nakamoto). 0966-9795/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.intermet.2012.08.012
Fig. 1 shows photographs of two different types of Seebeck micro-probe tips fabricated in this study. Copper was selected as a probe-tip material because of the high-thermal and electrical conductivities and low Seebeck coefficient as well as an easy machining process. A contact area of 10 mmf was obtained by a mechanical machining process using a microscope (Fig. 1 (a)). The dimension of the probe tip is 4 mm in diameter and 4 mm in length with flat surfaces on both sides by spark-erosion cutting to attach heaters on the flat surfaces as explained later. An angle of relief of 120 for the probe tip is adapted to secure mechanical strength of the probe tip and to lower thermal resistance between the tip and a sensor position in the probe. A Teflon coated AlumeleChromel thermocouple of 0.076 mmf was selected as a sensor and attached to a drilled hole with a diameter of 0.2 mmf located 0.5 mm above from the probe tip by a silver paste as shown in Fig. 1 (b). Two strain gauges (Kyowa KFL-1-120-C1-16) connected in a serial way with a total electrical resistance of 240 U were used as a heater to generate temperature difference between top and bottom surfaces of a sample. These heaters are attached on both side walls of the probe tip by a wanish.
234
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 1 (c) shows a photograph of a whole figure of the Seebeck micro probe. The Copper probe tip is attached to a metal shaft made by stainless steel mediated by an insulating acryl spacer to suppress heat leaking from the Copper tip to an upper part of the probe. A coil spring with a spring constant of 10 gf/mm is equipped in an upper acryl cylinder to make good thermal and electrical contacts between the probe and sample. A typical contact pressure is estimated to be of the order of 100 kgf/mm2 when a contact area of 10 mmf and a spring stroke of 1 mm are assumed. Any remarkable effect of the contact pressure on Seebeck coefficient is not observed. Seebeck coefficient does not depend on the contact pressure when the pressure is changed by five times by increasing the spring stroke from 1 to 5 mm. As shown later, the measured values of the Seebeck coefficient for the BismutheTelluride and ZinceAntimonide systems are in good agreement with those measured by other commercially available apparatus. The AlumeleChromel thermocouple and lead wires of the heater are fixed on a probe arm made by Aluminum through an acryl wire-guide plate and connected with twisted Copper wires at a thermal anchor located on a thermal bath made by Copper. 2.2. Constitution of the SMP-MS and a measurement procedure
Fig. 1. Two different types of the Seebeck-micro probe tips fabricated by a mechanical machining process (a) and (b), and the whole figure of the Seebeck micro probe with the mechanical machined tip (c).
Fig. 2 shows a schematic drawing of the Seebeck micro-probe measurement system (SMP-MS) developed in this study. The system consists of two digital nano-voltmeters (Agilent 34420A), a dc current source (Agilent E3647A), a function generator (nf DF1906), XY and Z stages (SURUGA SEIKI KS202-70 and PZG413L05AG), a stepping motor controller for the stages (SURUGA SEIKI D223) and a personal computer. All the instruments are controlled by the PC via GPIB-USB interface (Agilent 82357B) with a full-automatic measurement program composed by LabView. Data analysis software is also developed in this study. The measurement condition such as shape and size of a measurement area, a magnitude of heater current, trigger frequency and points for data acquisition, a probe stroke and temperature equilibrium condition are input as initial parameters of the measurement.
Fig. 2. Schematic drawing of the Seebeck micro-probe measurement system.
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
A two-probe differential method was employed to measure Seebeck coefficient. A plated sample with a thickness of about 2 mm is fixed on to the thermal bath made by Copper using a silver paste. An AlumeleChromel thermocouple is also attached to the thermal bath by a silver paste. The Seebeck micro-probe tip is touched down to the sample surface and the heater attached on the both side walls of the probe tip gives a temperature difference DT between the top and bottom surfaces of the sample. The temperature difference was about 0.5e2 K when 12e24 mA of the heater current was supplied for 5 s in the present study. The generated Seebeck voltages VA and VC between the Alumele Alumel wires and between the ChromeleChromel wires of both thermocouples at the probe and thermal bath, respectively, were measured in a heating process. It is noted that a function generator is used to give trigger signal two digital nano-voltmeters to measure a timely-synchronized Seebeck voltages VA and VC accurately and quickly. The typical trigger frequency and points are 4 Hz and 20 points, respectively, giving 20 measurement points per 5 s. The Seebeck voltages VA and VC between two pairs of the wires were plotted against the temperature difference DT. The Seebeck voltages VA and VC are given by VA ¼ (S SA)DT and VC ¼ (S SC)DT, where S, SA and SC represent the Seebeck coefficients of the sample, Alumel and Chromel, respectively. Therefore, the Seebeck coefficient of the sample at the probe position was obtained from the slopes of the VeDT plot by a least square fitting. The data tables for Seebeck coefficient and Seebeck voltage of the Alumel and Chromel wires are composed based on the NIST data [8]. Two Seebeck coefficients calculated from the Seebeck voltages of the Alumel and Chromel wires coincide with each other within the order of 103 mV/K. The thermal bath and micro-probe are mounted on motor-driven XY and Z stages, respectively, and the XY stages are able to move with a spatial resolution of 1 mm. Thus, the spatial distribution of Seebeck coefficient can be measured at a selected area by moving the stages. A two-dimensional contour map and line profiles along the X and Y directions of Seebeck coefficient are displayed on the measurement screen. The measurement is started after stabilizing temperature change within 1% of the temperature difference DT in a relaxation process, taking about 100 s. Therefore, the average total measurement time for one point is about 120 s at the present. The measurement system is covered with a box sealed by Aluminum foils to keep temperature constant and cut a highfrequency noise during a measurement. Each connection part of lead wires is also covered with a heat-insulating material to eliminate temperature difference among connection points which causes an extra Seebeck voltage. An actual temperature fluctuation in a sample space is found to be less than 1 K during a measurement for about five days in a temperature controlled room by an air conditioner. The extraction of line profile data and combination of the more than two data for different measurement areas can be possible by the data analysis software developed in this study.
235
The sample was fixed on to the thermal bath using a silver paste. An area of 4 4 mm2 was measured 30 times at 297 K with a step of 1 mm for both X and Y directions. Thus, 480 measurements were performed for the small area of 4 4 mm2 in total. Temperature fluctuation during the measurements was within 1 K. Fig. 3 shows the histogram of the Seebeck coefficient for the standard Ni. The average value of the Seebeck coefficient is 12.51 mV/K. The Seebeck coefficient of Ni at room temperature has been reported previously [9,10]. The values are around 19 mV/K. The absolute measurement value by the SMP-MS is smaller than those values by about 6e7 mV/K. The difference in the Seebeck coefficient of Ni between the present value and reference ones may be caused by the temperature difference between the tip and the sensor position in the probe. Since the thermal conductivity of Ni is only about four times smaller than that of Cu, temperature difference between the tip and the sensor arising from thermal resistance ratio between the sample and probe tip should not be ignored. When we consider the thermal circuit of the SMP-MS, the correct Seebeck coefficient of the sample S is given by a following equation;
S ¼
DTm Sp DT 0 Sm DT DTm
(1)
where DT, DTm, Sm, Sp and DT0 represent the actual temperature difference in the sample, the measured temperature difference between the bottom face of the sample and upper sensor, the measured Seebeck coefficient, the Seebeck coefficient of the probe tip and the extra temperature difference in the probe tip (between the contact point and sensor position in the probe tip), respectively. Thus, DTm is equal to DT þ DT0. Now, when we assume the thermal conductivity of 4000 and 910 mW/Kcm and the Seebeck coefficients, and 19.1 and 1.71 mV/K for Ni and Cu, respectively, the ratio of thermal resistance between the Cu probe tip and the measured Ni is estimated to be about 3:7. The temperature difference in the probe tip DT0 becomes 30% of the total DTm for the present case. On contrary, thermoelectric materials have extremely small thermal conductivity of about the order of 10 mW/Kcm, which is four hundred times smaller than that of Cu. Therefore, the thermal resistance of the probe tip is compared to be extremely smaller than that of the sample. DT0 is estimated to be so small less than 1% of DTm and can be ignored in this case. In fact, the measured Seebeck coefficients of BismutheTelluride and ZinceAntimonide are in good agreement with those by other measurement apparatus as
3. Results and discussion 3.1. Standard Ni In order to confirm the accuracy of the measurement for the SMP-MS, Ni has been measured as a standard material. A Ni rod in purity of 4 N with a diameter of 8 mmf was cut into a parallel piped shape with a dimension of 2 2 17 mm3. The sample was sealed in a silica tube in vacuum and annealed by an electric furnace to remove strain.
Fig. 3. Histogram of the Seebeck coefficient for standard Ni. The solid curve represents the result of Gaussian fitting with the average value of 12.51 mV/K and the half width of half maximum of 0.65 mV/K.
236
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 4. The cross section cut parallel to the crystal growth direction for the Bismuthe Telluride system taken by an optical polarized microscope. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. The detailed distribution of the Seebeck coefficient in the BismutheTelluride system measured for a rectangular area near the sign inversion boundary surrounded by red lines in Fig. 4 with measurement steps of X and Y ¼ 50 mm.
shown later. On the other hand, the half width of half maximum of the distribution of the Seebeck coefficient is 0.65 mV/K. The scattering of the data is small enough within 5% of the average value. Thus, we have judged that the SMP-MS can be enough to be used for measurements of Seebeck coefficient for thermoelectric materials.
bottom part of the ingot exhibits p-type conduction with S ¼ 250 mV/K. In the p-type region, S has a nearly constant value. The sign reversal observed around X ¼ 0 mm (5 mm from the top edge). The upper part of the ingot shows n-type conduction of S ¼ 250 mV/K. These Seebeck coefficients for both the p- and ntype areas are consistent with those observed in the previous study. A transition width (the width between Smax , p and Smax , n) is estimated to be about 5000 mm. No remarkable enhancement of the Seebeck coefficient near the boundary is observed in both the pand n-type conduction areas. More detailed result is shown in Fig. 6. A measurement was done near the boundary area of 2000 5000 mm2 with a 50 mm step surrounded by red lines in Fig. 4. The sign inversion from p- to n-type is observed more obviously. The sign inversion from p to n type along the crystal growth direction observed in this system is attributable to segregation of excess Te at the upper part of the ingot. The excess Te as a donor is wiped out at the upper part of the ingot during crystal growth. As a result, the bottom part exhibits p type, whereas the upper one including larger amount of the Te shows n-type conduction. In fact, the position at which the sign inversion takes place strongly depends on the excess Te concentration d and crystal growth rate. The sign inversion is observed at the upper position for larger Te concentration and faster growth rate.
3.2. BismutheTelluride system The BismutheTelluride system (Bi,Sb)2Te3þd by partial substitution of Bi with Sb and addition of excess Te was grown by a onedirectional solidification method [11]. In the previous study, it has been found that the conduction type changes in an ingot depending on growth rate and excess amount of Te [11]. The details of the inversion of the conduction type in the system have been examined by the SMP-MS. Fig. 4 shows a photograph of the cross section cut parallel to the growth direction for the BismutheTelluride crystal taken by an optical polarized microscope. Typical dimension of the grown crystal is 30 mm in length and 8 mm in diameter. Several large and long grains are found along the crystal growth direction, indicating that a preferred orientation exists along the crystal growth direction. The similar grain distribution is observed on the opposite side. A line profile of the Seebeck coefficient along the crystal growth direction (along the yellow line in Fig. 4) is displayed in Fig. 5. The
Fig. 5. The line profile of the Seebeck coefficient in the BismutheTelluride system along the crystal growth direction indicated by a yellow line in Fig. 4.
3.3. ZinceAntimonide system The ZinceAntimonide system has been known as a thermoelectric material with the high figure of merit of 1.3 in the moderate temperature range [12]. Strong anisotropy of the thermoelectric properties is expected because of the rhombohedral crystal structure. However, it has not been succeeded to grow the large single crystal enough to examine the anisotropy of the thermoelectric property since many structural phase transitions exist between the melting point and room temperature [13]. The SMP-MS is able to measure local Seebeck coefficient in a small area of about several 10 mm. Therefore, it is also expected that the SMPMS is one of the powerful tools to examine anisotropic physical properties of materials that are obtained only as polycrystalline samples. The ingot of ZinceAntimonide was also grown by a onedirectional solidification method with a dimension of 30 mm in length and 10 mm in diameter [14]. The disk plate sample with a thickness of 2 mm was taken out of the ingot by slicing perpendicular to the crystal growth direction and used for the SMP measurements. Fig. 7 shows the grain distribution of the sample taken by an optical polarized microscope. Cracks and grain boundaries are clearly observed in the figure. It is confirmed that both the sides of the disk plate sample have the same grain distribution. Typical grain size is found to be of the order of about
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
Fig. 7. The grain distribution of the surface of the disk sample perpendicular to the crystal growth direction in the ZinceAntimonide system. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
1 mm2. The SMP measurements have been carried out for a square area of 1 1 mm2 with steps of X ¼ Y ¼ 50 mm indicated by red lines in Fig. 7. The spatial distribution of the Seebeck coefficient is clearly observed as shown in Fig. 8. It is revealed that the distribution of the Seebeck coefficient strongly resembles that of the grain. Therefore, the distribution of the Seebeck coefficient originates from the anisotropy reflecting the difference in crystal orientation of each grain. The distribution of the Seebeck coefficient ranges from 100 to 120 mV/K except the crack region. These values are consistent with those measured by other measurement apparatus. Judging from the haziness of the grain boundaries and cracks in the distribution of the Seebeck coefficient, it is suggested that an effective contact area of the SMP measurement is larger than the actual contact area by a few ten percent. Fig. 9 shows the histogram of the Seebeck coefficient for the measurement area except the crack region displayed by Fig. 8. Solid curves represent the result of a least square fitting assuming Gaussian function. It is found that the profile consists of at least two components with the average values of 102 and 112 mV/K with the half width of half maxima of 2.50 and 3.17 mV/K, respectively. The former peak with the lower value corresponds
237
Fig. 9. The histogram of the Seebeck coefficient in the ZinceAntimonide system for the square area except the crack region indicated by red lines in Fig. 7. The solid curves represent the results of a least square fitting assuming Gaussian distribution.
to the fringe area, whereas, the latter one to the island area at the center. 4. Conclusion The two dimensional scanning Seebeck coefficient measurement system by a micro-probe method has been developed with high spatial resolution. A Seebeck micro-probe with a contact area of 10 mm is fabricated by a mechanical machining process. The full automatic measurement program and data analysis software are also composed by LabView. The distribution of the Seebeck coefficient has been measured for the BismutheTelluride and ZinceAntimonide systems. The inversion from p- to n-type conduction along the crystal growth direction is clearly observed in the BismutheTelluride system, caused by a gradient of Te concentration along the crystal growth direction. On the other hand, the anisotropic Seebeck coefficient reflecting the grain distribution is obviously found in the Zince Antimonide system. These results indicate that the Seebeck micro-probe measurement system developed in the present study is one of very powerful tools to examine local Seebeck coefficient of materials, reflecting local difference in a chemical composition and a crystal structure of materials. It is also expected that this method can be applied widely to examine physical properties of nano-structured and functionally graded materials by realizing the smaller probe size and addition of a temperature control system. Acknowledgment This work was supported by Grant-in-Aid for Scientific Research (C), Japan Society for the Promotion of Science (No. 23560363). References Schilz J, Kaysser WA. Mater Sci Eng 2003;A362:17. Müller E, Drasar C, Ni HL, Zhao XB, Karpinski G, Müller E. J Mater Sci 2005;40:605. Satou K, Yamashita O, Odahara H, Tomiyoshi S. Appl Phys 2006;A84:103. Nakamoto G, Kurisu M. J Electron Mater 2009;38:916. Yamamoto A, Kukuruznyak D, Ahmet P, Chikyow T, Ohuchi FS. Mater Res Soc Proc 2003;804:3. [6] Ikeuchi S, Shimada K, Takahashi Y, Ishii Y, Yamamoto A. ULVAC Tech J 2008; 69. [7] Kosuga A, Kurosaki K, Muta H, Stiewe C, Karpinski G, Müller E, et al. Mater Trans 2006;47:1440. [1] [2] [3] [4] [5]
Fig. 8. The spatial distribution of the Seebeck coefficient in the ZinceAntimonide system for the square area indicated by red lines in Fig. 7.
238
G. Nakamoto, Y. Nakabayashi / Intermetallics 32 (2013) 233e238
[8] Croarkin M, Guthrie W, Burns G, Kaeser M, Strouse G. Temperature-electromotive force reference functions and tables for the letter-designated thermocouple types based on the ITS-90. Natl Inst Stand Technol Monogr 1993; 175:630. Available online at: http://srdata.nist.gov/its90/main/. T.C. tables NIST. [9] Blatt FJ, Rowe V, Schroeder PA, Cox JE. Phys Rev Lett 1967;18:395.
[10] Laubitz MJ, Matsumura T, Kelly TJ. Can J Phys 1976;54:92. [11] Huong NT, Setou Y, Nakamoto G, Kurisu M, Kajihara T, Mizukami H, et al. J Alloys Compd 2004;368:44. [12] Caillat T, Fleurial JP, Borshchevsky A. J Phys Chem Solids 1997;58:1119. [13] Nakamoto G, Kinoshita K, Kurisu M. J Alloys Compd 2007;436:65. [14] Souma T, Nakamoto G, Kurisu M. J Alloys Compd 2002;340:275.