Simple, Reversible Gradient Seebeck Coefficient Measurement System for 300-600K
Soumya Biswas, Aditya S Dutt, Nirmal Sabastian, Vinayak B Kamble
SSimple, Reversible Gradient Seebeck Coefficient Measurement System for300 - 600K
Soumya Biswas, a) Aditya S Dutt, a) Nirmal Sabastian, and Vinayak B Kamble b) School of Physics, Indian Institute of Science Education and Research, Thiruvananthapuram, Kerala,India 695551 (Dated: 16 August 2019)
An in-house Seebeck coefficient measurement system has been developed which can measure the thermoemf (Seebeckcoefficient) of the sample, under large temperature difference, in the temperature range 300-600 K. Unlike majorityof reported instrumental designs, the system does not have a hot walled chamber and hence is much closer to realtime thermoelectric applications conditions. The system consists of two brass blocks supported heaters. These heatersare placed on either side of the sample through silver caps, thus allows individual temprature control . A reversibletemperature gradient is applied across the sample and the measurement is carried out in quasi-static direct currentmode. Hence, a more accurate Seebeck coefficient measurement is obtained. By virtue of its design the sample holderensures a minimum thermal and electrical contact resistance during a measurement cycle. The combination of metalsused for measurement (Ag and Cu) shows negligible junction contribution. The variance upto ±
2% and accuracyupto 7% at high temperature has been obtained using calibration samples’ reference data of state of the art commercialsystem.
I. INTRODUCTION
Thermoelectric effect is a phenomenon by which a temper-ature gradient generates an electric potential or vice versa .It does have a great relevance now, since we are overly de-pending on non-renewable sources of energy, which shall notbe replenished over time. Thus, by using an engine havingan added thermoelectric power generation unit, it can convertthis dissipated energy into usable electrical energy, which inturn increases the efficiency of the engine. However, in orderto tap its complete potential, a high efficiency thermoelectricmaterial is highly desirable for a practical applications . The“Figure of merit” is the quantity, which represents the effi-ciency of a thermoelectric material. It is represented as "zT"and the expression for the same is as shown by Eq. (1) zT = S σ T κ (1)Where,S is the thermopower (also known as Seebeck coefficient); σ is the electrical conductivity; κ is the total thermal conductivity andT is the absolute temperature.The quantity in the numerator S σ , is a measure of poweroutput and is called as power factor. Thus, for the estimationof thermoelectric efficiency, one has to measure Seebeck co-efficient, electrical conductivity and thermal conductivity ofthe given material. Although, there are several of commer-cial systems available for the Seebeck measurement in diversetemperature range, they are fairly expensive and need a peri-odic maintenance to keep them in working condition. Besides,thye cannot be customized as per the experimental require-ments such as with magnetic field or gas ambience. Hence, a) These two authors contributed equally b) Electronic mail: [email protected] we have developed a very low cost system (about 20% of thecost of the commercial system), which offers similar accuracyas that of the prior in a limited temperature range. Thus, theobjective of this study is to design, fabricate and calibrate asetup that can measure the Seebeck coefficient of the samplein a given temperature range (300 - 650 K), which is goodenough for study of a range of class of compounds such aschalcogenides, oxides and other alloys. A number of designshave been reported in literature till date.
However, manyof them are quite complex or involve a hot zone furnace etc increasing the cost of the instrument as well as the powerrequirements. Moreover, such designs require high tempera-ture stable refractory materials such as Alumina and Platinumfor electrically conducting leads. Here, because of the smallersize of the heaters, the hot zone does not extend much furtheraway from the sample and does not impose stringent high tem-perature stability requirements for many of the components.Further, the entire system is enclosed in stainless steel cham-ber which allows to control the environment.
II. MATERIALS AND METHODSA. Design of the Apparatus
Fig. 1 shows the schematic design of the system for measur-ing the Seebeck co-efficient of a sample. A home built systemis designed which can measure the thermoemf by applyinga small temperature difference across the sample. Here, thesample is sandwiched between two heaters, which are placedon two solid brass cylinders as shown in the Fig. 1. It hastaken a significant effort to optimize the design of the setup.The system is calibrated and interfaced for data acquisitionsystem described below. The system is designed to work fortemperature range of RT - 350 o C, which is ideal for mostchalcogenide and a large family of oxides and alloys. a r X i v : . [ phy s i c s . i n s - d e t ] A ug FIG. 1: (a) The schematic diagram of the Seebeck set-up with enlarged view of the assembly.
B. Control and Data Acquisition System
Here, the sample (blue colored object in the schematic di-agram shown in Fig. 2) is sandwiched between a pair ofbrass cylinders having silver cap electrodes. Heating wires(nichrome) are wound on the brass cylinders and followed byheater insulation in order to minimize the heat loss and allow-ing greater control over temperature. These heaters provideFIG. 2: A schematic diagram and a digital photographshowing thermocouple and voltage lead junctions in contactwith the sample. the temperature difference across the sample, which are mea-sured as well as controlled using two “K” type thermocou-ples inserted inside these brass cylinders centrally as seen infigure. The thermocouples are electrically insulated from thebrass rods using mica sheets. While, the voltage produced, ismeasured between two copper wires drawn from either silvercaps attached to each brass rod. Silver was chosen for its highelectrical (6.25 × Ω − . m ) and thermal conductivity (406W/K.m) . The assemblly is held with the help of two longSS screws anchored form a CF100 flnage. All the wires aresoldered to a circular connector at the center of the mountingflange allowing complete enclosure. One of the brass rod ismounted on to a spring loaded arrangement as shown in Fig.1. The entire assembly is inserted into a a vacuum chamberconnected to a vacuum pump, allowing control over the en-vironment i.e. either in ambient or ultra high vacuum (10 − torr).The heaters were controlled using Lakeshore 336 temper-ature controller having four control outputs and two thermo-couple controls. Two K type thermocouples (30 guage) wereconnected to the Lakeshore temperature controller. A tem-perature difference of about ± o C is given across samplereversibly. As a result of this temperature difference, a smallvoltage (usually a few mili volts) is produced, which is mea-sured with the help of a keithley 2700 digital multimeter hav-ing a 7700 multiplexure card of 20 channels.FIG. 3: (a) The typical measurement procedure of Seebeckmeasurement at two different temperatures i.e. 350 and 400K. (b) The variation of temperature profile of the two heatersacross the sample with temperature difference
III. RESULTS
Two K type thermocouples are placed in either silver capsthrough insulation, to measure the temperature. Owing to highthermal conductivity of silver, it is assumed that the sampleand silver cap’s flat surface are in thermal equilibrium. Oneof the block is attached to a spring, so that it allows easy place-ment and removal of sample, at the same time, ensure a goodelectrical and thermal contact between the metal blocks andthe sample thorugh pressure. The K type thermocouples arecomposed of alumel and chromel junction which has its ownseebeck of 40 µ V/K and which is linear across the entire mea-surement range.The copper wires were inserted into a hole in the silvercap, which served as voltage leads. The voltage leads and thethermocouples are connected to the digital multimeter. The7700 scanner card has 20 channels allowing to do simultane-ous measurements of 20 physical quantities such as resistance,temperature, voltage, current etc. FIG. 4: The comparison of the seebeck data of Bi Te andHalf heusler alloy with reference data. A. Measurement procedures
In order to use the above configuration for measuring thevoltage, it is mandatory to confirm that all the junctions areohmic in nature i.e. they allow reversible charge flow. A setof I-V measurement, were taken at each of the stage to en-sure the proper electrical contact. The measurement is carriedout in quasi-steady state differnetial fasion. This method istime efficient and practical. The Fig 3(a) shows the profileof temperatures across the samples for both the heaters. Thisdemonstrates either polarities of the temperature gradient andhence more reliable data. The chamber was evacuated usingrotary pump and then pressurized to near atmospheric value(760 Torr) using dry nitrogen gas before beginning the mea-surement. The sample was brought to measurement tempera-ture by heating form either side to a desired value (for instanceat 350 K as shown in Fig 3(a)). Subsequently, a small tem-perature difference is applied. In a typical measurement theheater 2 is held at a constant measurement temperature (forinstance T K) and the temperature of heater 1 is varied fromT + 10 K. The Heater 1 is then set to T - 10K from Heater 2.This allows the sample to cool naturally. The data of voltageproduced is recorded during cooling ensuring a steady statemeasurement. The typical measurement interval of the systemi.e the time interval between measurement of temperature dif-ference and the voltage, is about 100-150 mS. However, mostof hot walled measurement systems like ZEM-3, first heats thesample to a desired temperature either by resistive heating orInfra Red radiation and then give a small temperature differ-ence on one side. The major drawback of commercial systemdesign is, it needs a tall sample of at least 6 mm in height foraccurate measurements. Whereas, here the data of Bi Te re-ported has been performed on sample of 2 mm thickness. TheFIG. 5: (a) The Seebeck Coefficient values of Cu as afunction of temperature from literature and (b) Thecomparison of voltage measured before and after correctiongthe wire seebeck using eq 4 at 500K.horizontal dimensions of the sample in our design can be any-where between 3 to 12 mm.The two samples chosen for calibration have very different Svalue (by an order of magnitude) as well as the trend withtemperature. The data so obtained are compared with thosemeasured on state of the art systems from reputed laboratoriesworking in thermoelectrics from the country. The calibrationwas done with two samples of known Seebeck value (p-typebismuth telluride and half heusler alloy sample) from standardsystems as shown in Fig 4. It may be seen the values obtainedare found to be consistent with the commercial ZEM-3 sys-tem. B. Error estmation
The voltage produced between two leads of the voltmetercan be expressed as − ∆ V = (cid:90) Term Term E . dl = (cid:90) Term Term S ( T ) . dT . (2) Where, E is the electric field developed over length dl due tothermoemf (S).This may be written as- − ∆ V = (cid:90) T T a S Cu dT + (cid:90) T T S Ag dT + (cid:90) T T S ( T ) . dT + (cid:90) T T S Ag dT + (cid:90) T b T S Cu dT (3)Here, T − T (cid:28) T , and T a = T b (The temperature of con-nector) Besides, the temperature within the silver blocks doesnot change and S Ag also does not vary significantly over ∆ T then we can write , − ∆ V = (cid:90) T T S ( T ) dT + (cid:90) T T S wire dT (4)where (cid:90) T T S wire dT = (cid:90) T T b S Cu dT (cid:48) − (cid:90) T b T S Cu dT (cid:48)(cid:48) (5)Since, dT’ and dT" are the temperature difference between T b and T and T b and T respectivelys such that, dT = | dT (cid:48)(cid:48) − dT (cid:48) | (6)Thus, the uncertainly in the measured voltage and actual sam-ple voltage is evaluated by correcting with the seebeck coeffi-cient of the wire material i.e. Cu at measurment temperatureand the corrosponding temperature difference. The values ofseebeck coefficient have been obtained from literature andplotted at shown in Fig 5. Here, the advantage of using steadystate differential method is evident that it does not imposes therequirement of curves intersecting the ordant, i.e V=0 at ∆ T =0. This offset might arise from thermocouple inhomogeni-ties and possibility of non homogenous contact interfaces. .However, the presence of offset does not affect the slope of thedata and hence is ignored. (it is also found in data measuredusing Commercial Zem-3 system at high temperatures .Thus, the close agreement in values obtained between thesystem developed and that of the commercial system signifiesthe goodness of the data obtained and hence can be used forcharacterizing new samples.The statistical varition in the data shown, is normally sameas the size of the legend in Fig 4, however, is barely seen incase Bi Te due to high seebeck value. This is evaluated bystatistical mean of several measurements of same sample atsame temperatures and it has been observed that the varianceof data obtained is upto ±
2% of the full scale. Further, theimportant parameter of error has been evaluated by estimat-ing the difference between measured average value and therreference value ( ∆ S ) normalized with the reference value (S).This percentage error has been shown in Fig 6 as a functionof sample temperature. It may be noted that the value is max ±
7% at high temperatures, which is similar to uncertainty inZEM-3 system where it arises predominently due to coldfinger effect of thermocouples and electrical leads. C. Optimization of system
The system has undergone several changes in order to op-timize the data accuracy. Some were empirical, while manywere leart from literature like Martin et al. and others .Here, the following changes were made for optimzations. • The copper blocks were used intially which were foundto corrode/oxidize at high temperature in ambient at-mosphere. Hence, brass blocks were used which showshigher melting temprature, thus better thermal as wellas chemical stability. Besides, brass are softer andhence easily machinable. • However, the junction of brass block and copper wirewas found to introduce significant voltage and hencethe silver cap was used which acted as electrode andalso has same seebeck value as that of copper . • When brass blocks were longer than the heater area,there used to instability in the tempearature due to con-tinuous cooling of exposed brass surface. Hence, thesame was covered with thermal insulation for retainingthe heat, which also improved temperature stability. • When thermocouples are not electrically insulated fromthe metal block in which they are inserted, it leads todiscrepency in the thermocouple voltage due to anotherjunction. Hence, insulating thermocouples with highthermal conductivity material like mica avoids this spu-rious error. • Besides, some samples like Bi Te and oxides showsa change in seebeck value when meausured in oxy-gen atmosphere (i.e. ambient) and inert atmosphere.Hence, introducing the controlled atmosphere has beenbeneficial for more accurate measurement. How-ever, some robust samples like the half heusler alloy( Zr Ti NiSn Si ) measured in this study didFIG. 6: The estimated error of the seebeck coefficient ofBi Te and Half Heusler (HH) alloy( Zr Ti NiSn Si ) with reference data. not show noticable change when measured in ambientor inert atmosphere. IV. SUMMARY
The simple, low cost, Seebeck co-efficient measurementsystem has been made in-house and calibrated at higher tem-perature (300 - 600 K) with a reference samples measured inthe commercial machine. The sample values were in goodagreement with the commercial Seebeck ZEM-3 system byULVAC RIKO. The measurement were conducted under suf-ficiently large, reversible temperature gradient ( about ± ±
7% ).
V. NOTE
The current affiliations of AD is TU Dresden and that ofNS is IISc Bangalore. The work was performed during theiraffiliation at IISER Thiruvananthapuram.
ACKNOWLEDGMENTS
The authors would like to thank Prof. Satish Vitta fromMEMS, IIT Bombay for providing the half heusler alloy sam-ple and its data measured on ULVAC Riko system. Moreover,we are also thankful to Prof. Arun Umarji, Mr Rajasekhar ofIISc Bangalore for the Bi Te sample and its data. T. M. Tritt and M. Subramanian, MRS bulletin , 188 (2006). J. R. Sootsman, D. Y. Chung, and M. G. Kanatzidis, Angewandte ChemieInternational Edition , 8616 (2009). G. D. Mahan, APL Materials , 104806 (2016). J. He and T. M. Tritt, Science , eaak9997 (2017). K. A. Borup, J. De Boor, H. Wang, F. Drymiotis, F. Gascoin, X. Shi,L. Chen, M. I. Fedorov, E. Müller, B. B. Iversen, et al. , Energy & Envi-ronmental Science , 423 (2015). C. Wood, A. Chmielewski, and D. Zoltan, Review of scientific instruments , 951 (1988). S. Singh and S. K. Pandey, Measurement , 26 (2017). A. Mishra, S. Bhattacharjee, and S. Anwar, Measurement , 295 (2015). A. Burkov, A. Heinrich, P. Konstantinov, T. Nakama, and K. Yagasaki,Measurement Science and Technology , 264 (2001). T. Dasgupta and A. Umarji, Review of scientific instruments , 094901(2005). Q. Zhu, H. S. Kim, and Z. Ren, Review of Scientific Instruments ,094902 (2017). J. Martin, T. Tritt, and C. Uher, Journal of Applied Physics , 14 (2010). S. Iwanaga, E. S. Toberer, A. LaLonde, and G. J. Snyder, Review of Scien-tific Instruments , 063905 (2011). M. Gunes, M. Parlak, and M. Ozenbas, Measurement Science and Tech-nology , 055901 (2014). A. Kumar, arXiv preprint arXiv:1906.08023 (2019). V. Ponnambalam, S. Lindsey, N. Hickman, and T. M. Tritt, Review ofscientific instruments , 073904 (2006). A. P. Kharote and D. Ramachandran, ECS Journal of Solid State Scienceand Technology , N3001 (2017). N. Cusack and P. Kendall, Proceedings of the physical society , 898(1958). A. Guan, H. Wang, H. Jin, W. Chu, Y. Guo, and G. Lu, Review of ScientificInstruments , 043903 (2013). J. Mackey, F. Dynys, and A. Sehirlioglu, Review of Scientific Instruments , 085119 (2014). J. De Boor and E. Müller, Review of scientific instruments84