稀有金属(英文版) 2019,38(08),713-720

Contact mode thermal sensors for ultrahigh-temperature region of 2000-3500 K

Sheng-Yong Xu Zhen-Hai Wang Li-Jiang Gui

Key Laboratory for the Physics and Chemistry of Nanodevices,Department of Electronics,Peking University

Department of Micro-Nano Fabrication Technology,Institute of Electrical Engineering,Chinese Academy of Sciences

作者简介：*Sheng-Yong Xu is now working in the Key Laboratory for the Physics and Chemistry of Nanodevices, and Department of Electronics, Peking University. He received his Bachelor degree in Peking University,Physics Department,1988, and Doctor degree in physics in National University of Singapore, 1999. In 2001-2006, he worked in the Department of Physics and Center for Nanoscience, Pennsylvania State University,USA. From 2006 until now,he is a Professor in Institute of Electronics Engineering and Computer Science, Peking University. His research has been focused on nano-structure and materials. He has participated in the work of more than 130 technical publications in total. e-mail: xusy@pku.edu.cn;*Li-Jiang Gui received the Ph.D. degree from Beihang University in 2014.He is currently postdoctorate at the Institute of Electrical Engineering, Chinese Academy of Sciences. His research interests aremicro-/nano-sensors. e-mail: lijianggui@mail.iee.ac.cn;

收稿日期：11 December 2018

基金：financially supported by National Key R&D Program of China(Nos.2016YFA0200802 and2017YFA0701302);

Contact mode thermal sensors for ultrahigh-temperature region of 2000-3500 K

Sheng-Yong Xu Zhen-Hai Wang Li-Jiang Gui

Key Laboratory for the Physics and Chemistry of Nanodevices,Department of Electronics,Peking University

Department of Micro-Nano Fabrication Technology,Institute of Electrical Engineering,Chinese Academy of Sciences

Abstract：

In this article,we reviewed several existing techniques that were capable of detecting local temperatures in the range of 2000-3500 K in a contact manner.These techniques included several non-standard thermocouples,Seger cones and ultrasonic meters.In particular,ultrasonic meters made of tungsten(W)wires were proven to be working well in nuclear plant for detecting the central temperature of reaction zone.We also presented two alternative approaches.One of them was a kind of singlemetal-based thermal sensor made from W,Mo and Ta wires,which utilized the size effect of Seebeck coefficient and theoretically applicable in contact mode measurement up to 3500 K.The other was a kind of detectors with micro-/nano-patterns,which utilized the size effect of surface melting point of bulky materials.This work should shed light on measurement approaches for ultrahigh temperatures in a variety of practical applications.

Keyword：

Thermocouple; Ultrasound; Ultrahigh temperature; Contact mode; Size effect; Seebeck effect; Surface melting point;

Received： 11 December 2018

1 Introduction

Temperature (T)is a characteristic physical parameter that cannot be measured directly,because it is a term for describing the level of chaos at atomic and molecular scales.Tis usually proportional to the average kinetic energy of atomic particles (atoms,molecules,ions,etc.),ε_k,in a system under test,i.e.,ε_k=c.k_B·T,where c is a systemdependent constant and k_R is the Boltzmann constant,1.38×10^-23 J.K^-1.

Many optical techniques have been developed to measure high temperatures above 2000 K,namely optical fiber pyrometer,infrared thermometer,ratio pyrometer and radiation thermometer [ 1, 2, 3, 4] .These sensing techniques are based on the physical mechanism of irradiation,such as the black body irradiation,Wien's law and Stefan-Boltzmann law,and they work well with high accuracy.Unfortunately,in some unique situations,there might be no room for optical measurement instruments.For example,for measuring the surface temperature of a spaceship in its reentry process into the atmosphere,for monitoring and controlling the temperature of reaction zone of a nuclear plant,or for measuring the inner surface temperature of rocket nozzle,the surface temperature of ultrahigh speed aircrafts,etc.,contact mode measurement could be theonly choice.

To measure temperatures over 2000 K with a contact lmode ren,ains a technical challenge.Contact mode measurement is based on the so-called the zeroth law of thermodynamics,1.e.,when a sensory subject is firmly connected to the subject under test for an efficiently long time,these two subjects would share the same temperature.And,to complete this task is an extreme technical challenge.The major reason of this challenge is the limited choices of materials for the thermal sensors that are sustainable in this temperature range of 2000-3500 K.For example,many conventional thermocouples will melt far below 2000 K.A secondary reason is that one should find a physical process or mechanism that works at the high temperature range and find one or two measurable parameters that change with the temperature.However,many physical properties of materials at this ultrahightemperature range are usually not well known.

2 Current contact mode thermal sensing techniques for ultrahigh temperatures

Currently,only a few kinds of techniques were reported for the temperature measurement with a contact manner at ultrahigh-temperature region of 2000-3500 K,i.e.,non-standard thermocouples [ 5, 6] ,Seger cones [ 7, 8, 9] ,and ultrasonic techniques [ 10] .These methods are partially workable,depending much on the testing environments.

Based on Seebeck effect,thermocouples are extensively applied in temperature measurements [ 11, 12, 13, 14, 15, 16] .The Seebeck effect refers to the phenomenon that a thermally induced stable and static voltage (ΔV) is measured across two ends of a piece of conductor or semiconductor when a temperature difference (ΔT) is established between the two ends,i.e.,ΔV=S·ΔT,where S is the Seebeck coefficient,which is usually determined by the distribution of state density near Fermi level of the material [ 17, 18] .A thermocouple usually consists of two metallic or semiconducting beams that have a measurable difference in their Seebeck coefficients (ΔS).The sensitivity and stability of a thermocouple rely onΔS,as well as the stability ofΔS in different temperature regions.To measure ultrahigh temperatures over 2000 K,it also requires that the materials of the two leads could sustain in the testing process without melting or losing their thermopower.

As early as in 1960's,tungsten-rhenium (W-Re) thermocouples,in particular,W3Re/W25Re and W5Re/W26Re,were evaluated for use in measuring temperatures in the range of 1900-3300 K in vacuum [ 19] .The results showed that W3Re/W25Re and W5Re/W26Re thermocouples showed a thermal shift of±50 K after 2900 K for15 h.In another report,W/Re thermocouples were tested in the range of 2300-2800 K in furnaces with graphite heaters and an inductive heater [ 6] .Now,among commercialized non-standard thermocouples,several W-Re-based products have realized the repeatable and reliable measurement range to 2000 K and pushed the applicable up limit to2800 K [ 20] .For Ir-Rh thermocouples,the applicable up limit is 2100 K [ 21] .At temperature region higher than1000 K,the reported measurement accuracy for these products is±1%of the measured value.

Besides the thermocouples,an interesting technique called“Seger cone”was developed [ 9] .As shown in Fig.1,these sensors are made of ceramic materials consisting of glass components.They are fabricated into small cones and put vertically with a slight tilting angle (80°to horizon level,for instance) beside the subject undergoing heating processes.When the environment temperature increases to the melting point of the cones,their mechanical strength is reduced and the whole cone structure starts to bend in the gravitation field.Calibration is performed beforehand.By checking the bending angle and shape of the cones after cooling down,one will obtain the highest temperature of the system under test.Currently,commercial Seger cone arrays are capable of measuring temperatures up to 2300 K.

As shown in Fig.2,ultrasonic technique has been proven to be a unique approach for real-time measurement of ultrahigh temperature in some extreme working environments,such as the core of a nuclear reaction zone [ 22] .When an ultrasonic pulse propagated along a thin rod,its velocity was affected by the environmental temperature.The velocity of ultrasonic wave v(T) could be expressed as:

where T is the absolute temperature,E(T) is the Young's modulus of the sensor material andρ(T) is the density of the material.The environment temperature can be deduced from the measured value of velocity along the thin rod.An ultrasonic thermometry system is composed of a pulse generator,a polarizing magnet,a magnetos trie tive stub,a thin-wire (e.g.,W) sensor,a sheath,a receive coil,an amplifier,an oscilloscope and a start/stop counter,as schematically shown in Fig.2.When a pulse generator produces a short-time pulse,the magnetostrictive rod transforms it into a narrow-band ultrasonic wave.The ultrasonic pulse propagates along the sensing element.When it encounters notches pre-built on the rod,a portion of its energy is then reflected.The reflected echoes are received by the received coil as part of the feedback signal and are transformed into an electrical voltage spectrum.The signals are amplified and transferred to an oscilloscope and assessed in the start/stop counter.Measuring the time interval between two adjoining echoes will give the average ultrasonic wave velocity of corresponding area,which offers the average temperature of the same area under test.When several discontinuities are pre-made along the probe,the temperature profile of the sensing element can be revealed.

Fig.1 Schematic illustrations of working mechanism of Seger cones:a side view,b top view,c installation position and d after heating processes

Fig.2 A schematic diagram of an ultrasonic thermometry measure-ment system [22]

A variety of sensor materials can be chosen according to various environmental conditions and certain temperature ranges.The maximum operation temperature of titanium (Ti) was reported to be 1080 K.Stainless steel and nickel-iron could be good candidate materials when the temperature was up to 1580 K.The ordinary operation temperature range of Re was from 1780 to 2880 K [ 23] .But the transmutation effect from Re to Os limited the application of Re as a sensor material.In current state of the art,thoriated W had been deemed as the only feasible material in the temperature range from 2880 to 3380 K [ 24] .

The ultrasonic thermometry sensors have advantages of simple structure,high stabilization,prompt response and high radiation resistance.Figure 3 presents one of the testing results.They could be applied to measure the temperature in harsh environments (e.g.,in nuclear reactors) [ 25] .In a nuclear reactor,the ultrasonic thermometry sensors (UTS) frequently encountered the problems of decalibration due to the transmutation of sensor materials.The measurement error,generated by the structure changes (e.g.,crystal dislocations) of materials under radiation,was diminished by the annealing effect at high temperature.Several methods had been developed to circumvent the problem of spurious echoes,which was led by the“sticking effect”of the wire sensor and its protection tube [ 26] .

Fig.3 Some testing results of ultrasonic thermometry [22]

Over the past decades,ultrasonic thermometry was still paid much attention and developed.Figure 4 illustrates an alternative ultrasonic thermometry measurement system recently developed by Chinese researchers [ 21] .Here,the magnetostrictive stub was substituted by piezoelectric transducers as the ultrasonic source and the iridium-rhodium (60%Ir-40%Rh) alloy served as the ultrasonic sensor material in this system.The experiments showed that the operation temperature range of this system was from 800 to 2500 K in oxidizing environments.

Fig.4 A schematic view of an alternative ultrasonic thermometry measurement system [21]

3 Size effect of Seebeck coefficient and its potential application for sensing ultrahigh temperatures

In recent years,an interesting phenomenon was reported.A size effect of Seebeck coefficient (S) was observed in a variety of metallic materials,from nano-sized thin-film stripes to micro-sized thin-film stripes,up to millimeter diameter wires,where the absolute S values of more than ten kinds of metals were found to reduce slightly from those of their counterparty bulks [ 27, 28, 29] .Here,we definedΔS as the absolute difference of Seebeck coefficient of the thin-film stripe or thin wire samples to that of the bulky value,i.e.,ΔS=|S_sample-S_bulk|.In 80-500-nm-wide thinfilm stripes,theirΔS values were measured to be0.1-2.0μV·K^-1,depending on different kinds of original metals.In 30-100-μm-wide thin-film stripes,the measuredΔS values were in the range of 0.01-0.10μV·K^-1.Even in metallic wires with diameters from 0.1 up to 1.0-2.0 mm,an unexpectedΔS at the order of 1-100 nV.K^-1 was measured.Figure 5 plots the major experimental data.

These results indicated that the size effect of Seebeck coefficient was a universal phenomenon.But the length scales of 0.1-1.0 mm were larger than the mean free length(usually<1μm) of electrons in these materials by a few orders of magnitude;therefore,it could not be well explained with current solid-state physics.Thus,this unexpected finding caused arguments [ 30, 31] .Yet it was repeatedly observed by several groups and has been applied in various thermal sensing devices [ 32] .It could be an overlooked phenomenon and needed further theoretical investigation.

Figure 6 presents a typical measurement result of a device made of two Mo wires,one with diameter of0.07 mm and the other with 0.70 mm.According to brown fitting line in Fig.5 which showed the general trend,the expectedΔS value was around 0.05-0.07μV·K^-1.However,in experimental data,the slope shown in Fig.6was around 0.23μV·K^-1,enhanced by 4 times.We believed that there were two effects co-played in this device.One was the size effect of Seebeck coefficient.The other was a slight difference in elemental components of the two Mo wires.Indeed,as shown in red markers,in devices made of two wires of different diameter,such enhancedΔS values were repeatedly obtained in many samples.The results indicated that Seebeck coefficient was very sensitive to the impurity and defects of a metallic wire.This leaves a big room for making tailored wire couples as thermal sensing devices.

Fig.5 Experimental results for size (φ,width of stripes or wires)effect of Seebeck coefficient measured in a variety of metallic thin-film stripes and wires

Fig.6 Measurement results of thermal power from a Mo-Mo dual-wire device

A thermocouple-like device with a thermopower (sensitivity)ΔS in the order of 0.01-0.10μV·K^-1 is too small for normal thermal sensors,and the output is hard to measure.But whenΔT is in the order of 1000 K,the outputΔV becomes measurable,in the order of 10-100μV,asΔV=ΔS·ΔT.It leads to a straightforward proposal for practical application of the slight size effect of Seebeck coefficient in thermal sensors.

For example,the sensor could be made of two tungsten(W) wires with diameters of 1.0 mm for W1 and 0.1 mm for W2,respectively,and we can tailor tiny difference in elemental composition of different wires so that they could have roughly the same melting point but enhanced difference in Seebeck coefficient.These two wires join at the junction region,which serves as the hot end in the measurement.For instance,when the temperature difference is2000-3500 K between the two ends of the W1-W2 thermocouple sensor,the output of this sensor will read10-17μV,withΔS of 0.05 UV·K^-1.This read is high enough for measurement with a commercial nanovoltmeter.

Figure 7a presents an illustration of such a thermal sensor sealed in a Th case.Such a W-Th device structure has been tested in the reaction zone of a nuclear plant,proven to be workable for ultrasonic sensors and sustainable in irradiation of extremely high intensity [ 22] .The difference is,for ultrasonic sensor,it needs a signal generator close to the core W wire,while for the proposed W-W sensor,a passive sensor,one only needs to measure the voltage output at the cold ends.In particular,the length of the two W wires is not limited.The W1 and W2 wires serve as both the body and leads of the sensor,as shown in Fig.7b.

Fig.7 a Proposed W-W thermal sensor for ultrahigh-temperature measurement and b schematic image for application in a nuclear plant

For different application environments,many materials could be listed as the core for the sensor shown in Fig.7,such as Mo,Ta,Re,Th and U.

4 Size effect of surface melting point and its potential application for sensing ultrahigh temperatures

We also propose here another alternative contact method for the measurement of ultrahigh temperatures.This method utilizes the geometric size effect of melting point for a solid material.It is well known that the surface melting point is lower than the bulk melting point because of less bonding connections of a surface atom to its neighboring atoms.Since the ratio of surface area over body volume is extremely high in nanomaterials,when a solid is made into nanoscale,the decrease in surface melting point becomes prominent and measurable [ 33, 34, 35] .When the size of a solid nanoparticle is reduced to be less than 10 nm,its melting point decreases dramatically,as typically shown in Fig.8.In an in situ transmission electron microscopy (TEM) study,solid Pt nanoparticles of average diameter 6-7 nm started to melt at the temperature range of 700-800 K [ 36] ,as shown in Fig.9,in spite of the fact that bulky Pt has a melting point of around 2040 K.

This interesting phenomenon could be applied in recording the highest temperature point of a heating process.The sensor could be made of a high-melting-point material,as one of those shown in Tables 1 and 2.An array of patterns in micro-and nano-sc ales could be pre-built on the surface of such a sensor.We thus expect that when the measured temperature is close to the melting point of the bulk material,there will be a series of changes remained in the micro-/nano-array of the structural patterns,both in the view of physical and chemical aspects,and these“traces”may offer valuable hints for the exact heating processes that have undergone.In this way,the sensor serves as a“black box”to the heating processes at ultrahigh-temperature zone.Since it is feasible to make the sensor from a series of materials whose bulk melting points span in the range of interest,we may therefore record more information with a multi-material sensor.In addition to the recording of the highest heating temperature,a bunch of interesting and fundamental problems may occur in the investigation.

Fig.8 Measured and calculated temperature (T) results for size (D,diameter of bulk) effect of melting points of gold bulks (m.p.bulk) [34]

In Tables 1 and 2,we present not only the melting points of 24 elements,but also those for W-Ta,W-Mo and Re-Rh alloys.For instance,the melting point of W-Ta alloy covers a wide range of 3300-3700 K.This fact remarkably enlarges the choice range of materials for the proposed ultrahigh-temperature sensors.

Fig.9 A study on melting phenomenon of 6-7 nm Pt nanoparticles with a transmission electron microscope [36]

  下载原图

Table 1 Melting point of high-melting-point materials at 1.01×10⁵ Pa (K)

  下载原图

Table 2 Melting point temperature range of several alloys (K)

Different measurement techniques may be suitable for different applications conditions.The proposed singlemetal,dual-beam sensors,may serve as well as the ultrasonic sensors in measuring the temperature of reaction zone,or measuring the surface temperature of a near-sun probe,where plenty of room is available for mounting the sensors,leads and measurement instruments.But for measuring the temperature of inner surface of a rocket nozzle,probably the proposed sensors working with the effect of melting point dependent on micro-/nano-shape of material may be more suitable.

Except the thermal effect,high-temperature-induced chemical effects may have huge influence on the physical performance and lifetime of all kind of contact mode sensors working at the ultrahigh-temperature range.In some cases,high pressure and intensive collisions may occur on the surface,for example,when measuring the surface temperature of a high-speed aerocraft.This adds more difficulty in the design,fabrication and measurement of the thermal sensors.

5 Conclusion

In this paper,we reviewed several available sensors for measuring ultrahigh temperatures in the region of2000-3500 K with a manner of contact mode.They were a few non-standard thermocouples,Seger cones and ultrasonic sensors.The up limit of measurable temperatures was around 2800,2300 and 3400 K,respectively.We proposed two alternative methods,i.e.,single-metal dual-beam sensors working with the mechanism of size effect of Seebeck coefficient,and geometric shape sensors working with the mechanism of size effect of melting point.Theoretically,these two novel methods may be applicable to temperatures up to 3500 K or higher,i.e.,the up limit of bulk melting point of available materials on the earth.Investigation on these nexw sensor techniques may greatly push the development of space sciences,ultrahigh-tcmpcrature physics and material sciences.

Acknowledgements This work was financially supported by National Key R&D Program of China (Nos.2016YF A0200802 and2017YFA0701302).

参考文献

[1] Fairchild C,Hoover W.Disappearance of the filament and diffraction effects in improved forms of an optical pyrometerl.JOSA.1923;7(7):543.

[2] Coates P,Lowe D.The Fundamentals of Radiation Thermometers.Boca Raton:CRC Press;2016.75.

[3] Morozhenko V.Infrared Radiation.Rijeka:InTech;2012.85.

[4] Skorupa W,Schmidt H.Subsecond Annealing of Advanced Materials:Annealing by Lasers,Flash Lamps and Swift Heavy Ions.Cham:Springer;2013.211.

[5] Ulanovskiy A,Edler F,Fischer J,Oleynikov P,Zaytsev P,Pokhodun A.Features of high-temperature calibration of W/Re thermocouples.Int J Thermophys.2015;36(2-3):433.

[6] Ulanovskiy A,Medvedev V,Nenashev S,Sild YA,Matveyev M,Pokhodun A,Oleynikov P.Thermoelectric characteristic of high-temperature thermocouples W5%RE/W20%RE.Int J Thermophys.2010;31(8-9):1573.

[7] Beerman H.Calibration of pyrometric cones.J Am Ceram Soc.1956;39(2):47.

[8] Basegio T,Haas C,Pokorny A,Bernardes A,Bergmann C.Production of materials with alumina and ashes from incineration of chromium tanned leather shavings:environmental and technical aspects.J Hazard Mater.2006;137(2):1156.

[9] Rea RF.Temperature measuring cones.J Am Ceram Soc.1938;21(3):98.

[10] Kazys R,Zukauskas E,Rekuviene R.Modeling of ultrasonic measurement systems with waveguides.Elektronika ir Elektrotechnika.2012;123(7):61.

[11] Zhao XH,Wang YR,Chen YZ,Jiang HC,Zhang WL.Enhanced thermoelectric property and stability of NiCr-NiSi thin film thermocouple on superalloy substrate.Rare Met.2017;36(6):512.

[12] Li G,Wang Z,Mao X,Zhang Y,Huo X,Liu H,Xu S.Real-time two-dimensional mapping of relative local surface temperatures with a thin-film sensor array.Sensors.2016;16(7):977.

[13] Choi H,Li X.Experimental investigations of laser micromachining of nickel using thin film micro thermocouples.J Manuf Sci Eng.2008;130(2):021002.

[14] Gregory OJ,You T.Ceramic temperature sensors for harsh environments.IEEE Sens J.2005;5(5):833.

[15] Chu D,Wong WK,Goodson KE,Pease RFW.Transient temperature measurements of resist heating using nanothermocouples.J Vac Sci Technol B Microelectron Nanometer Struct Process Meas Phenom.2003;21(6):2985.

[16] Chu D,Bilir DT,Pease RFW,Goodson KE.Submicron thermocouple measurements of electron-beam resist heating.J Vac Sci Technol B Microelectron Nanometer Struct Process Meas Phenom.2002;20(6):3044.

[17] Chopra K,Bahl S,Randlett M.Thermopower in thin-film copper—constantan couples.J Appl Phys.1968;39(3):1525.

[18] Justi E,Kohler M,Lautz G.(U|¨)ber die differentielle Thermokraft d(u|¨)nner Metallschichten.Zeitschrift f(u|¨)r Naturforschung A.1951;6(8):456.

[19] Asamoto R,Novak P.Tungsten-rhenium thermocouples for use at high temperatures.Rev Sci Instrum.1967;38(8):1047.

[20] Bradley D,Matthews K.Measurement of high gas temperatures with fine wire thermocouples.J Mech Eng Sci.1968;10(4):299.

[21] Wei Y,Wang G,Gao Y,Liu Z,Xu L,Tian M,Yuan D,Ren H,Zhou H,Yang L.A measurement system of high-temperature oxidation environment with ultrasonic Ir0.6Rth0.4 alloy thermometry.Ultrasonics.2018;89:102.

[22] Laurie M,Mag.allon D,Rempe J,Wilkins C,Pierre J,Marquie C,Eymery S,Morice R.Ultrasonic high-temperature sensors:past experiments and prospects for future use.Int J Thermophys.2010;31(8-9):1417.

[23] Carlson G,Plein HG.Refractory metals for ultrasonic thermometry application.1978;10(9):10448109.

[24] Tasman HA.Nuclear applications of ultrasonic thermometry.In:Proceedings of the Ultrasonics Symposium,New Orleans;1979.380.

[25] Lynnworth LC,Carnevale EH,Mcdonough MS,Fam SS.Ultrasonic thermometry for nuclear reactors.Nucl Sci IEEE Trans.1969;16(1):184.

[26] Arave A,Christensen J,Panisko F.High-temperature ultrasonic thermometer for in-reactor fuel rod centerline temperature measurements.Trans Am Nucl Soc.1972;15(2):766.

[27] Liu H,Sun W,Xu S.An extremely simple thermocouple made of a single layer of metal.Adv Mater.2012;24(24):3275.

[28] Huo X,Liu H,Liang Y,Fu M,Sun W,Chen Q,Xu S.a nano-stripe based sensor for temperature measurement at the submicrometer and nano scales.Small.2014;10(19):3869.

[29] Sun W,Liu H,Gong W,Peng LM,Xu SY.Unexpected size effect in the thermopower of thin-film stripes.J Appl Phys.2011;110(8):083709.

[30] Szakmany GP,Orlov AO,Bernstein GH,Porod W.Comment on“Unexpected size effect in the thermopower of thin-film stripes”.J Appl Phys.2014;115(23):236101.

[31] Huo X,Sun W,Liu H,Peng L,Xu S.Response to comment on'Unexpected size effect in the thermopower of thin-film stripes'.J Appl Phys.2014;115(23):236102.

[32] Szakmany GP,Orlov AO,Bernstein GH,Porod W.Single-metal nanoscale thermocouples.IEEE Trans Nanotechnol.2014;13(6):1234.

[33] Xia Y,Yang P,Sun Y,Wu Y,Mayers B,Gates B,Yin Y,Kim F,Yan H.One-dimensional nanostructures:synthesis,characterization,and applications.Adv Mater.2003;15(5):353.

[34] Buffat P,Borel JP.Size effect on the melting temperature of gold particles.Phys Rev A.1976;13(6):2287.

[35] Lai S,Guo J,Petrova V,Ramanath G,Allen L.Size-dependent melting properties of small tin particles:nanocalorimetric measurements.Phys Rev Lett.1996;77(1):99.

[36] Wang ZL,Petroski JM,Green TC,El-Sayed MA.Shape transformation and surface melting of cubic and tetrahedral platinum nanocrystals.J Phys Chem B.1998;102(32):6145.