Thermoelectric Properties and Morphology of Si/Sic Thin-Film Multilayers Grown by Ion Beam Sputtering

coatings

Article Thermoelectric Properties and Morphology of Si/SiC Thin-Film Multilayers Grown by Ion Beam Sputtering

Corson L. Cramer 1,2,* ID , Casey C. Farnell 1, Cody C. Farnell 1, Roy H. Geiss 3 and John D. Williams 1 1 Department of Mechanical Engineering, Colorado State University, Fort Collins, CO 80523, USA; [email protected] (Ca.C.F.); [email protected] (Co.C.F.); [email protected] (J.D.W.) 2 Manufacturing Demonstration Facility, Oak Ridge National Laboratories, Oak Ridge, TN 37831, USA 3 Department of Chemistry, Colorado State University, Fort Collins, CO 80523, USA; [email protected] * Correspondence: [email protected]; Tel.: +1-719-661-9530

Received: 6 February 2018; Accepted: 6 March 2018; Published: 19 March 2018

Abstract: Multilayers (MLs) of 31 bi-layers and a 10-nm layer thickness each of Si/SiC were deposited on silicon, quartz and mullite substrates using a high-speed, ion-beam sputter deposition process. The samples deposited on the silicon substrates were used for imaging purposes and structural verification as they did not allow for accurate electrical measurement of the material. The Seebeck coefficient and the electrical resistivity on the mullite and the quartz substrates were reported as a function of temperature and used to compare the film performance. The thermal conductivity measurement was performed for ML samples grown on Si, and an average value of the thermal conductivity was used to find the figure of merit, zT, for all samples tested. X-ray diffraction (XRD) spectra showed an amorphous nature of the thin films. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) were used to study the film morphology and verify the nature of the crystallinity. The mobility of the multilayer films was measured to be only 0.039 to 1.0 cm2/Vs at room temperature. The samples were tested three times in the temperature range of 300 K to 900 K to document the changes in the films with temperature cycling. The highest Seebeck coefficient is measured for a Si/SiC multilayer system on quartz and mullite substrates and were observed at 870 K to be roughly −2600 µV/K due to a strain-induced redistribution of the states’ effect. The highest figure of merit, zT, calculated for the multilayers in this study was 0.08 at 870 K.

Keywords: thin-ﬁlm thermoelectric; Si/SiC multilayers; giant Seebeck

1. Introduction Thermoelectric generator (TEG) devices are of interest in waste heat recovery and as primary heat-to-electricity sources. One promising material concept of interest to achieve high-energy conversion efficiency is a superlattice, where each leg in a module is composed of hundreds of nanometer-scaled thick crystalline bi-layers. In the current research, the term superlattice is not appropriate because we do not have full crystallinity in our thin films, and thus, we refer to them as multilayers. Even though the films are not fully crystalline, the defect doped behavior should be close to that of the crystalline material since the bands and electronic structure are similar to a crystalline material, and even purely amorphous material is capable of semiconductor behavior and quantum well effects [1–5]. Theoretical work of Hicks and Dresselhaus showed that nano-scaled superlattices (SLs) would display a quantum well effect that increases the carrier effective mass, which increases the Seebeck coefficient while leaving the mobility unchanged and increases the electrical conductivity. This theory suggested that zT values could be ~13-times higher than the bulk material values [6]. In later work,

Coatings 2018, 8, 109; doi:10.3390/coatings8030109 www.mdpi.com/journal/coatings Coatings 2018, 8, 109 2 of 10

Chen et al. [7] and Ezzahri et al. [8] both showed that the electrical conductivity increased, and the thermal conductivity decreased in a superlattice structure. Consequently, these studies show that all three properties of the figure of merit are affected such that the zT is increased with a superlattice structure. In related and more recent work, Lee et al. [9] showed the enhanced Seebeck coefficient for SrTiO3/SrTi0.8Nb0.2O3 superlattices, and Liu et al. [10] showed the promising in-plane thermoelectric (TE) properties on Si/Ge SLs. These more recent studies show increases of the figure of merit by an order of magnitude over bulk material. In addition, recent data from superlattices of GaN/AlN/AlGaN grown by chemical vapor deposition at high temperature showed that enhanced zT values could be obtained over bulk properties at room temperature [11]. In further theoretical work, testing done by Ghamaty and Elsner [12] and Balusu and Walker [13] showed the decoupling of the Seebeck coefficient and the electrical conductivity due to the quantum well effect, which increases zT [14]. Early n-type Si/Si0.8Ge0.2 and p-type B4C/B9C superlattices were deposited using molecular beam epitaxy [15] and magnetron sputtering [16], but a potentially faster and less expensive method is ion beam sputtering since the ion energies and sputter yields are higher [17]. Furthermore, excellent thermoelectric properties have been reported for Si/SiC, Si/Si0.8Ge0.2 and B4C/B9C superlattices using magnetron sputtering. A module zT near 4 at 523 K was reported for these materials [18–20], and more recent data for Si/SiC superlattices on silicon substrates made by magnetron sputtering [21] and ion beam sputtering [22] indicate even higher zT in high temperature applications. For the above studies, there is a significant doubt about the performance data due to the poor reproducibility with the resistivity measurements done with Si/SiC on silicon substrates as originally demonstrated [21]. In the current research, insulating substrates are used to get a true measure of the electrical properties, and the lattice thermal conductivity of the ML film was measured using the MEMS technique performed at the University of Denver where our ML was deposited on Si, and the Si thermal conductivity was subsequently measured to subtract from the total thermal conductivity yielding a measure of only the ML [22]. Several studies have shown that superlattices with a layer thickness of 5 to 10 nm have thermal conductivity of less than 5 W/mK. Some of them are the studies done by Aubain et al. [23] on the lattice thermal conductivity and the study done by Mazumder et al. [24] on the total thermal conductivity at different temperatures, specifically for Si/SiC films. These thermal conductivity measurements include the 3-ω transient method and the thermal reflectivity method, which are both well-accepted techniques.

2. Materials and Methods Before installing the silicon and the quartz substrates in the vacuum system, they were cleaned and rinsed with acetone, isopropyl alcohol and deionized water and dried. For the mullite (3Al2O3·2SiO2) substrates, no pre-cleaning preparation was performed. The configuration of the sputter deposition process is shown in Figure1. Sintered Si and SiC sputter targets were used. Inside the chamber evacuated to 3 × 10−6 Torr, a stage moved targets under a stationary ion beam to sputter different materials onto the substrates. The targets were mounted at forty-five degrees relative to the ion beam and the substrate holder as shown in Figure1. A second stage was used to position a mask over the substrate to control layer thickness. During deposition, the temperature of the substrates was held constant at about 773 K, and the partial pressure from Ar from the ion beam source is 5 × 10−3 Torr. The deposition rate is slowed to 3 nm/min by lowering the ion beam energy to 300 eV during the first 1.0 nanometers of growth of each layer to ensure sharp interfaces and less mixing at the interfaces. The remainder of each ML layer is grown more quickly at 10 nm/min using 1000 eV ions. A total of 31 bi-layers of silicon and silicon carbide were deposited to form multilayers structures. Each layer consists of 10 nm of Si followed by 10 nm of SiC. Coatings 2018, 8, 109 3 of 10 Coatings 2018, 8s 2018, 8, x

Coatings 2018, 8s 2018, 8, x

(a) (b) (c)

Figure 1. Configuration for the sputtering chamber: (a) is a schematic; (b) is a photograph showing Figure 1. Configuration for the sputtering chamber: (a) is a schematic; (b) is a photograph showing the the direction of target movement where the substrate holder is behind the mask; (c) is a photograph direction of target movement(a) where the substrate holder(b is) behind the mask; (c) is a photograph(c) of the of the substrate holder viewed perpendicularly to the target motion. The quartz crystal micrograph substrate holder viewed perpendicularly to the target motion. The quartz crystal micrograph (QCM) is Figure(QCM) 1. isConfiguration used to measured for the film sputtering thickness. chamber: (a) is a schematic; (b) is a photograph showing used to measured film thickness. the direction of target movement where the substrate holder is behind the mask; (c) is a photograph of Afterthe substrate the ML holderfilms are viewed deposited perpendicularly onto the substrates, to the target they motion. are installed The quartz into crystal a separate micrograph vacuum Aftertest(QCM) facility the ML is used where films to measured are the deposited Seebeck film thickness. coefficient onto the substrates, and the electrical they are resistivity installed were into a measured separate over vacuum a test facilitytemperature where the range Seebeck from coefficient 300 K to 900 and K. theThe electricalSeebeck coefficient resistivity is measured were measured in plane over by heating a temperature up one side and keeping the other side within 20 degrees of the hot side while measuring the voltage range fromAfter 300 the K ML to films 900 K.are The deposited Seebeck onto coefficient the substrates, is measured they are installed in plane into by a heating separate up vacuum one side across the samples and the temperature across the sample. The resistivity is measured in plane as andtest keeping facility the where other the side Seebeck within coefficient 20 degrees and of the electricalhot side while resistivity measuring were measured the voltage overacross a temperaturewell by measuring range from the electrical 300 K to resistance 900 K. The and Seebeck knowing coefficient the geometry. is measured A photograph in plane of the by testheating setup up the samplesis shown and in Figure the temperature 2. Two substrates across can the be sample.mounted Theonto resistivitythe sample holder is measured and tested in planeat the same as well by one side and keeping the other side within 20 degrees of the hot side while measuring the voltage measuringtime. theThe electricalsubstrate resistanceholder is made and knowingof boron nitride. the geometry. The tungsten A photograph wire-based of heater the test was setup used isto shown across the samples and the temperature across the sample. The resistivity is measured in plane as in Figurecontrol2. Twothe temperature substrates of can each be side mounted such that ontoa small the temperature sample holder difference and is established tested at the across same the time. well by measuring the electrical resistance and knowing the geometry. A photograph of the test setup substrate. The maximum possible achievable mean temperature was 870 K. Resistance temperature Theis substrate shown in holder Figure is 2. made Two substrates of boron nitride.can be mounted The tungsten onto the wire-based sample holder heater and was tested used at to the control same the detectors (RTDs) (Omega Engineering Inc., Norwalk, CA, USA) were used rather than thermocouples temperaturetime. The ofsubstrate each side holder such is that made a small of boron temperature nitride. The difference tungsten is wire established-based heater across was the used substrate. to for measuring temperatures. SEM of the MLs was done with a JEOL-JSM 6500F (JEOL Ltd., Tokyo, The maximum possible achievable mean temperature was 870 K. Resistance temperature detectors controlJapan )the in temperaturesecondary electron of each mode. side such TEM that of the a small MLs temperaturewas done with difference a JEOL JEM is established-2100F (JEOL across Ltd., the (RTDs)substrate.Tokyo, (Omega Japan)The Engineering maximum with the possibleultra Inc.,-high Norwalk,achievable resolution CA,mean (UHR USA) temperature) objective were used polewas piece870 rather K. and Resistance than EDS thermocouples from temperature Oxford for measuringdetectorsInstruments temperatures. (RTDs) Max (Omega 80 SDD SEM Engineering (Oxfordshire, of the MLs Inc., UK was Norwalk,). The done crystal with CA, structure aUSA) JEOL-JSM were of the used 6500Fdeposited rather (JEOL thanMLs Ltd., thermocoupleson the Tokyo, silicon Japan) in secondaryforsubstrate measuring electron was temperatures. analyzed mode. using TEM SEM XRD of o f thewith the MLs MLsa Bruker waswas D8done Discover with with aa (Billerica,JEOL JEOL-JSM JEM-2100F MA, 6500F US) (JEOL with (JEOL Ltd.a Cu Ltd.,, Tokyo(Kα) Tokyo, , Japan)Japansourc with) ine scanning the secondary ultra-high at 1electron s per resolution step mode. with a (UHR)TEM step ofsize objectivethe of MLs0.02°. was The pole doneenergy piece with of and the a EDSJEOLbeam from JEMwas -402100F Oxford keV at(JEOL 40 Instruments mA. Ltd., MaxTokyo, 80The SDD mobility Japan) (Oxfordshire, ofwith the MLthe films UK).ultra was The-high measured crystal resolution structure using (UHR the offour) theobjective-point deposited Ecopia pole HMS MLs piece- on3000 and the Hall EDS silicon measurement from substrate Oxford was analyzedInstrumentssystem using (Bridge Max XRD Technology, 80 with SDD a ( BrukerOxfordshire, Chandler D8 DiscoverHeights, UK). The AZ, (Billerica,crystal USA structure). MA, US)of the with deposited a Cu (K MLsα) sourceon the silicon scanning substrate was analyzed using XRD with a Bruker D8 Discover (Billerica, MA, US) with a Cu (Kα) at 1 s per step with a step size of 0.02◦. The energy of the beam was 40 keV at 40 mA. The mobility source scanning at 1 s per step with a step size of 0.02°. The energy of the beam was 40 keV at 40 mA. of the ML films was measured using the four-point Ecopia HMS-3000 Hall measurement system The mobility of the ML films was measured using the four-point Ecopia HMS-3000 Hall measurement (Bridgesystem Technology, (Bridge Technology, Chandler Heights,Chandler AZ,Heights, USA). AZ, USA).

Figure 2. Thermoelectric property measurement setup. Two samples could be tested at the same time. RTD, resistance temperature detector.

Figure 2. Thermoelectric property measurement setup. Two samples could be tested at the same time. FigureRTD, 2. resistanceThermoelectric temperature property detector. measurement setup. Two samples could be tested at the same time. RTD, resistance temperature detector.

Coatings 2018, 8, 109 4 of 10

Coatings 2018, 8s 2018, 8, x 3. Results Coatings3. Results 2018 , 8s 2018, 8, x The crystal structure of the thin films on silicon was investigated with XRD where all peaks are The crystal structure of the thin films on silicon was investigated with XRD where all peaks are identified3. Results with respective planes except for two. Figure3 shows some small crystalline peaks at 37, 40, identified with respective planes except for two. Figure 3 shows some small crystalline peaks at 37, 60 and 76 degrees from SiC. The peaks around 27, 45, 57, 65 and 70 degrees are from Si. The recorded 40, 60The and crystal 76 degrees structure from of theSiC thin. The films peaks on aroundsilicon was 27, 45,investigated 57, 65 and with 70 degreesXRD where are allfrom peaks Si. Theare siliconidentifiedrecorded peak at siliconwith 27 degreesrespective peak at is 27 planes similar degrees except to is the similar for one two. foundto Figurethe byone 3 Surana foundshows by etsome al.Surana [small25]. et Thecrystalline al. recorded[25]. The peaks recorded silicon at 37 peak, at 6540,silicon degrees 60 and peak was76 atdegrees identified65 degrees from from wasSiC . YangidentifiedThe peaks et al. from [around26]. Yang In 27,general, et 45, al .57, [26]. the 65 films Inand general, 70 show degrees the some filmsare crystallinity from show Si. some The with the Sirecordedcrystallinity (101) peak silicon with coming peak the at Sifrom 27 (101) degrees the peak substrate, is coming similar and from to later, the the one TEM substrate, found further by and Surana shows later, et theTEM al. distributions [25]. further The shows recorded revealing the thatsilicondistributions the films peak have at revealing 65 some degrees pockets that was the of identified films grains have among from some Yangotherwise pockets et al . of amorphous[26]. grains In general, among material. othe the filmsrwise The small show amorphous peaks some are likelycrystallinitymaterial. from small The with small amount the peaks Si of (101) Siare and likely peak SiC comingfrom that aresmall from crystalline, amount the substrate, of as Si willand and beSiC shown laterthat, areTEM in crystalline TEM further later. , shows as The will broader the be peaksdistributionsshown at 27, in 57 TEM and revealing later. 65 degrees The that broader arethe amorphousfilms peaks have at 27, some peaks,57 and pockets which65 degrees of is grains the are reason amorphous among to othebelieve peaks,rwise that which amorphous the is films the are not entirelymaterial.reason to polycrystalline,The believe small that peaks the butarefilms likely have are somenotfrom entirely small areas amount ofpolycrystalline grains of andSi and some, but SiC have areasthat someare of crystalline amorphous areas of ,grains as material will and be for shown in TEM later. The broader peaks at 27, 57 and 65 degrees are amorphous peaks, which is the bothsome materials areas inof amorphous the ML. material for both materials in the ML. reason to believe that the films are not entirely polycrystalline, but have some areas of grains and some areas of amorphous material for both materials in the ML.

Figure 3. X-ray diffraction spectra for a Si/SiC multilayer on silicon substrate where 2θ is the Figure 3. θ diffractionX-ray angle diffraction and the spectra intensit fory is a shown Si/SiC in multilayer counts per on second silicon (cps) substrate. where 2 is the diffraction angleFigure and the3. X intensity-ray diffraction is shown spectra in counts for a per Si/SiC second multilayer (cps). on silicon substrate where 2θ is the diffractionFigure 4 shows angle anda cross the -intensitsectiony ofis shownthe ML in thin counts film per that second was (cps) investigated. using scanning electron microscopyFigure4 shows (SEM). a cross-section The cross-section of the of MLthe ML thin thin film film that on was the investigatedsilicon and the using quartz scanning was used electron for microscopythe SEMFigure (SEM).and 4 shows the TheTEM a cross cross-section analysis.-section In of both of the the cases,ML ML thin the thin film films film that were on was the depositedinvestigated silicon andat using773 the K scanning quartzand tested was electron in used the for microscopy (SEM). The cross-section of the ML thin film on the silicon and the quartz was used for the SEMthermoelectric and the TEMproperty analysis. measurement In both system cases, from the 323 films K to were 903 depositedK. The SEM at image 773 Kof andthe multilayer tested in the the SEM and the TEM analysis. In both cases, the films were deposited at 773 K and tested in the thermoelectricthin film structure property on measurementthe silicon substrate system shown from in 323 Figure K to 4 903 is used K. The to measure SEM image and ofverify the the multilayer ML thermoelectricstructure, layer property thickness measurement and uniformity system of the from thin 323 film; K to however, 903 K. The it SEMis difficult image toof tellthe howmultilayer much thin film structure on the silicon substrate shown in Figure4 is used to measure and verify the ML thinintermixing film structure is occurring on the at silicon the interfaces substrate between shown thein Figurelayers and4 is difficultused to measureto get good an ddata verify for the Si/CML structure, layer thickness and uniformity of the thin film; however, it is difficult to tell how much structure,ratio for the layer SiC thickness layer. and uniformity of the thin film; however, it is difficult to tell how much intermixingintermixing is occurring is occurring at at the the interfaces interfaces between between thethe layers and and difficult difficult to to get get good good data data for forthe theSi/C Si/C ratioratio for thefor the SiC SiC layer. layer.

Figure 4. SEM image of cross-section of a 10-nm layer multilayer film on silicon.

Figure 4. SEM image of cross-section of a 10-nm layer multilayer film on silicon. Figure 4. SEM image of cross-section ofa 10-nm layer multilayer ﬁlm on silicon.

Coatings 2018, 8, 109 5 of 10 Coatings 2018, 8s 2018, 8, x Coatings 2018, 8s 2018, 8, x

Figure5 5 shows shows aa transmissiontransmission electronelectron microscopymicroscopy (TEM)(TEM) imageimage ofof thethe substratesubstrate andand MLML thatthat Figure 5 shows a transmission electron microscopy (TEM) image of the substrate and ML that waswas milledmilled out with a focused ion beam (FIB). The substrate used for this ML thin f filmilm was quartz. was milled out with a focused ion beam (FIB). The substrate used for this ML thin film was quartz. TheThe substrate andand filmfilm cancan bebe distinguished, andand thethe ML filmfilm layerlayer size is shown. Figure6 6 shows shows thatthat The substrate and film can be distinguished, and the ML film layer size is shown. Figure 6 shows that thethe filmfilm hashas pocketspockets of different material behavior.behavior. TheThe fast Fourier transforms (FFTs) showshow that the the film has pockets of different material behavior. The fast Fourier transforms (FFTs) show that the pocketspockets in thethe MLML filmsfilms mostlymostly crystallinecrystalline andand areare likelylikely grains.grains. TheThe materialsmaterials mapping data from pockets in the ML films mostly crystalline and are likely grains. The materials mapping data from EDX with the TEM measurements were inconclusive likely due to the FIB mixing of the materials. EDX with the TEM measurements were inconclusive likely due to the FIB mixing of the materials. TheThe crystallinitycrystallinity isis mostmost likelylikely more more from from the the Si Si than than from from the the SiC SiC because because deposition deposition is doneis done closer clos toer The crystallinity is most likely more from the Si than from the SiC because deposition is done closer theto the recrystallization recrystallization temperature temperature of Siof andSi and not not SiC. SiC. There There is stillis still some some crystallinity crystallinity in thein the SiC SiC layer layer as to the recrystallization temperature of Si and not SiC. There is still some crystallinity in the SiC layer theas the (200) (200) plane plane shows shows up up very very well. well. as the (200) plane shows up very well.

(a) (b) (a) (b) Figure 5. TEM of substrate and multilayer (ML) film (a) and FFTs of the same film (b). Figure 5.Figure TEM 5.ofTEM substrate of substrate and multilayer and multilayer (ML) film (ML) (a) ﬁlmand (FFTsa) and of FFTs the same of the film same (b). ﬁlm (b).

Figure 6. TEM bright field emission image of the grains of the silicon and the silicon carbide layers Figure 6. TEM bright field emission image of the grains of the silicon and the silicon carbide layers Figureshowing 6. someTEM crystallinity bright ﬁeld emissionwithin the image white of circles. the grains of the silicon and the silicon carbide layers showingshowing somesome crystallinitycrystallinity withinwithin thethe whitewhite circles.circles. Figure 7 shows the plot of resistivity versus temperature for MLs on quartz and mullite. In the Figure 7 shows the plot of resistivity versus temperature for MLs on quartz and mullite. In the curreFigurent research,7 shows quartz the plot and of mullite resistivity substrates versus temperatureare used for forthe MLsML thermoelectric on quartz and mullite.measurements In the current research, quartz and mullite substrates are used for the ML thermoelectric measurements currentsince the research, Si substrate quartz interferes and mullite with substratesthe measurements. are used forThe the MLs ML on thermoelectric the quartz and measurements mullite substrates since since the Si substrate interferes with the measurements. The MLs on the quartz and mullite substrates theshow Si substratesimilar resistivities; interferes withhowever, the measurements. the resistivity Theof the MLs thi onn film the on quartz the quartz and mullite is lower substrates compared show to show similar resistivities; however, the resistivity of the thin film on the quartz is lower compared to the mullite substrate. Since quartz and mullite are both insulators, the decrease in resistivity on the the mullite substrate. Since quartz and mullite are both insulators, the decrease in resistivity on the quartz substrate may be from the quality of film and interfaces made. The resistivity data gathered quartz substrate may be from the quality of film and interfaces made. The resistivity data gathered

Coatings 2018, 8, 109 6 of 10 similar resistivities; however, the resistivity of the thin ﬁlm on the quartz is lower compared to the mullite substrate. Since quartz and mullite are both insulators, the decrease in resistivity on the quartz

Coatingssubstrate 2018, may 8s 2018, be 8, from x the quality of ﬁlm and interfaces made. The resistivity data gathered in this study are signiﬁcantly higher than data previously shown in the literature on the silicon substrate, inlikely this due study to the are conductivity significantly of higher the silicon than substrate, data previously but follow shown the same in the semiconductor literature on trend the silicon for the substrate,temperature likely range due tested to the [ 19conductivity,21,22]. of the silicon substrate, but follow the same semiconductor trend for the temperature range tested [19,21,22]. ·cm) ·cm) Ω (

Figure 7. Resistivity of the two candidate substrates as a function of temperature. Figure 7. Resistivity of the two candidate substrates as a function of temperature. Figure8 shows the Seebeck coefficient measurements of the ML thin films on quartz and mullite. TheFigure multilayer 8 shows thin the films Seebeck on the coefficient quartz and measurements the mullite samplesof the ML have thin veryfilms low on quartz Seebeck and coefficient mullite. Thevalues multilayer until the thin temperature films on the reaches quartz the and value the of mullite 800 K. samples The multilayer have very thin low films Seebeck reach a coefficient stunningly valueshigh Seebeck until the coefficient, temperature producing reaches overthe value 100 mV. of 800 The K. films The aremultilayer n-type as thin the films two multilayerreach a stunningly materials highare intrinsicallySeebeck coefficient n-type., Theproducing giant Seebeck over 100 coefficient mV. The valuesfilms are of then-type ML as thin the films two multilayer on the quartz materials and the aremullite intrinsically were repeatable n-type. The and giant were Seeb observedeck coefficient at the multiple values temperatureof the ML thin ramps films on on the the multiple quartz and thin thefilm mullite samples. were Table repeatable1 shows thatand thewere Seebeck observed coefficient at the multiple decreases temperature by 5.8% and ramps 6.5% on on the the MLs multiple on the thinquartz film and samples. the mullite Table substrates, 1 shows that respectively, the Seebeck after coefficient 3 cycles. decreases The metal by contacts 5.8% and started 6.5% to on delaminate the MLs onslightly the quartz from and the MLthe thinmullite films substrates, on the quartz respectively substrate, after after 3 justcycles. one The cycle. metal The contacts best thermoelectric started to delaminateproperties wereslightly obtained from the from ML the thin ML films thin onfilm the on quartz the quartz substrate substrate, after but just the one ML cycle. thin The film best had thermoelectricsome cracks near properties the metal were contact obtained after onefrom thermal the ML cycle thin duefilm toon the the mismatch quartz substrate, in thermal but expansion the ML thincoefficient film had between some cracks the quartz near substrate, the metal metal contact contact after one and thermal ML thin cycle film. due However, to the mismatch data are still in thermalcollected expansion for this sample coefficient for two between more the cycles, quartz so the substrate, mechanical metal cracking contact isand not ML enough thin film. to compromise However, datathe wholeare still film collected for studies. for this Ideal sample contacts for two for thismore system cycles, could so the not mechanical be obtained cracking [27]. This is not is an enough issue if tothese compromise MLs are tothe be whole used film in a module.for studies. This is an issue if these MLs are to be used in a module.

FigureFigure 8 8.. SeebeckSeebeck coefficients coefﬁcients of of ML ML thin thin films ﬁlms on on the the mullite mullite and and the the quartz quartz substrates. substrates.

Table 1. Degradation of Seebeck coefficients of ML thin films on different substrates tested at 870 K.

Test # Seebeck Coefficient on Quartz (μV/K) Seebeck Coefficient on Mullite (μV/K) 1 −2600 −2300 2 −2500 −2250 3 −2450 −2150

Coatings 2018, 8, 109 7 of 10

Table 1. Degradation of Seebeck coefﬁcients of ML thin ﬁlms on different substrates tested at 870 K.

Test # Seebeck Coefﬁcient on Quartz (µV/K) Seebeck Coefﬁcient on Mullite (µV/K) 1 −2600 −2300 2 −2500 −2250 3 −2450 −2150

Finally, the mobility of the ML thin films was measured at room temperature using a the 4-point probe Hall mobility system to separate the carrier density component of the electrical conductivity. The mobility is low and looks very different for the multilayer thin films on the two substrates. It is also noted that the room temperature carrier concentration is low for the multilayer thin films on the two substrates. Lower values of the mobility and the carrier density are indicative of amorphous materials. Table2 shows the mobility and carrier concentration of the films on quartz and mullite substrates. The expected mobility of amorphous silicon and the silicon carbide is in the range of 0.1 to 10 cm2/Vs [28,29]. The mobility values of the multilayer thin films in this study are in this range. This verifies that the multilayer thin films in this study have some amorphous nature.

Table 2. The mobility and the carrier concentration values of Si/SiC ML thin ﬁlm samples on the mullite and the quartz substrates at room temperature.

Substrate Carrier Mobility (cm2/Vs) Carrier Concentration (cm−3) Quartz 3.9 × 10−2 −3.2 × 1017 Mullite 1.0 × 100 −3.3 × 1015

4. Discussion A limited number of TE property measurement tests were repeated three times to determine the reproducibility of the thermoelectric measurements. The experiments show good repeatability for the Seebeck coefficient from ML thin films deposited on the mullite substrates where the ML thin films, and the metal contacts did not crack. One interesting phenomenon in our ML thin films is the appearance of the high Seebeck coefficient effect at higher temperature on the quartz and the mullite substrates. The turning on of the giant Seebeck coefficient is not well understood, but it is hypothesized that it is a strain-induced effect due to the mismatch of the interfaces since the electrical resistivity behaves as a regular semiconductor would. The strain-induced Seebeck coefficient seen at higher temperatures has been suggested and modeled in several studies with the band convergence and the enhanced effective mass [30–33]. The layers with some limited crystallinity likely line up better and allow the emergence of phenomena at the interfaces when the thin film samples are given enough thermal energy to reduce the strain in the films. This is a possible mechanism because the thermal expansions of the two multilayer materials is different enough to cause interface shifting at higher temperatures, and the carriers in one material could activate and transport in the other at higher temperatures. Our testing apparatus could not be used to drive the temperature above ~900 K, and thus, there are no data available above 900 K. In addition, as discussed before, our metal contacts began to degrade when they were tested to 900 K, but unlike the Seebeck coefficient, the resistivity had significantly increased after each test likely due to cracking of the metal contacts and annealing. That is why we need both the better electrical contacts and the higher temperature measuring system to further study the emergent phenomena that have been observed. The existence of Giant Seebeck in this system is very significant because the Seebeck is almost 4 times higher than a normal Si system, which means the zT would be a factor of 16 better if the resistivity was lower. Table3 shows the figure of merit, zT, values for 10-nm layers of Si/SiC ML structure. There is an observed tradeoff in the thermoelectric properties that yielded a relatively low figure of merit, Coatings 2018, 8, 109 8 of 10 zT, where zT =α2T/$; α is the Seebeck coefficient, which is a measure of the amount of voltage per degree temperature difference; T is the temperature; $ is the electrical resistivity, and k is the thermal conductivity. If the MLs possessed better resistivity with a simultaneous increase in Seebeck coefficient, then the MLs figure of merit would be increased significantly and surpass many known materials.

Table 3. Si/SiC ML on different substrates.

Substrate zT at 870 K Quartz 0.08 Mullite 0.07

The higher Seebeck coefficients at high temperature suggest that the giant Seebeck coefficient effect has been observed in the Si/SiC ML thin film system. This is a very high Seebeck coefficient for films made with ion beam deposition. The giant Seebeck coefficient is a term reserved for systems enhancing the Seebeck coefficient beyond the bulk values with the quantum confinement. Most of the studies reporting giant Seebeck coefficient were observed in the systems utilizing oxides and precise doping. Ohta et al. [34] and Lee et al. [9] have both observed the giant Seebeck coefficient in SrTiO3 systems with values near −1000 µV/K, and Song et al. [35] have observed the giant Seebeck coefficient for MnO2. More recently, the giant Seebeck coefficient seen in the silicon nanoparticles is in the range of our ML thin films [36]. The most recent giant Seebeck system is the SnSe system with ~1000 µV/K [37]. These studies provide additional evidence that Hicks and Dresselhaus’s [6] theoretical work is of current significance. Full decoupling of the carrier concentration and Seebeck coefficient may never be achieved, but the recent evidence shows that the giant Seebeck coefficient could be produced in a high temperature material that has some amorphous nature. Systems designed for a high Seebeck coefficient like superlattices and multilayers have the potential for performing carrier concentration tuning, thus shifting the maximum of the zT versus carrier density plot to the left and upward for a higher, more efficient material with relatively lower carrier concentration, which is the reverse of the approach followed recently by the materials science community in skutterudites, clathrates and chalcogenides [38]. The turning on of the giant Seebeck coefficient does not happen until high temperature, where carriers are activated. Once carriers are activated in the ML, then they can transport. The giant Seebeck is then observed because of the layered structure and enhanced density of states. This gives more energy per electron, which increases the Seebeck coefficient. The resistivity is not affected significantly from our data because the resistivity is not enhanced and acts like a regular semiconductor, and the level of crystallinity in our films allows for the quantum confinement to enhance the Seebeck coefficient.

5. Conclusions The ion-beam-based deposition system was configured to grow Si/SiC multilayer thin films on mullite and quartz substrates. Typical ML thin films consisted of 31 bi-layers of a 10-nm layer thickness. The ML thin films were found to be have some amorphous nature with some regions of crystallinity. An impressive Seebeck coefficient of ~2600 µV/K was measured at this temperature along with a higher, less than desired, electrical resistivity of 3.4 Ω·cm. zT of 0.08 was calculated when these values were combined with a previously measured thermal conductivity of 2 W/mK. We have measured the highest Seebeck coefficient to date with Si/SiC multilayers on insulating substrates with values greater than −2000 µV/K. We showed evidence that there are enhancements in the Seebeck coefficient to the quantum confinement. Quantum confinement can enhance the Seebeck coefficient because of an enhanced density of states and effective mass, which increases the energy per electron, but this does not happen until electrons have promoted the conduction band, so that is why the giant Seebeck coefficient is not seen until high temperatures. With superlattices or multilayers made of semiconductor materials, Coatings 2018, 8, 109 9 of 10 the electrical conductivity is poor compared to traditional bulk TE materials that are in the semi-metal materials class. It is believed that the density of states of the effective mass was increased in our ML thin films on the mullite and the quartz substrates, which resulted in an increase of the Seebeck coefficient at higher temperatures. The turning on of the giant Seebeck coefficient at high temperature is not completely understood, but it is thought to be due to a strain-related effect.

Acknowledgments: This work is supported by the DOE under grant number DE-SC0004278. Special thanks to Jack Clark, Surface Analytics, and Colorado School of Mines for help preparing and analyzing surfaces. Author Contributions: John D. Williams, Casey C. Farnell, and Cody C. Farnell conceived and designed the experiments; Corson L. Cramer performed the experiments; Corson L. Cramer and Roy H. Geiss analyzed the data; Corson L. Cramer wrote the paper. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Larbi, F.; Abdelkader, B.; Sib, J.D.; Bouizem, Y.; Chahed, L.; Amaral, A. Density of states in intrinsic and n/p-doped hydrogenated amorphous and microcrystalline silicon. J. Mod. Phys. 2011, 2, 1030–1036. 2. Park, N.-M.; Choi, C.-J.; Seong, T.-Y.; Park, S.-J. Quantum conﬁnement in amorphous silicon quantum dots embedded in silicon nitride. Phys. Rev. Lett. 2001, 86, 1355. [CrossRef][PubMed] 3. Car, R.; Parrinello, M. Structural, dymanical, and electronic properties of amorphous silicon: An ab initio molecular-dynamics study. Phys. Rev. Lett. 1988, 60, 204. [CrossRef][PubMed] 4. Zacharias, M.; Blasing, J.; Veit, P.; Tsybeskov, L.; Hirschman, K.; Fauchet, P.M. Thermal crystallization of

amorphous Si/SiO2 superlattices. Appl. Phys. Lett. 1999, 74, 2614–2616. [CrossRef] 5. Ley, L.; Kowalczyk, S.; Pollak, R.; Shirley, D.A. X-ray photoemission spectra of crystalline and amorphous Si and Ge valence bands. Phys. Rev. Lett. 1972, 29, 1088. [CrossRef] 6. Hicks, L.D.; Dresselhaus, M.S. Effect of quantum-well structures on the thermoelectric figure of merit. Phys. Rev. B 1993, 47, 12727. [CrossRef] 7. Chen, G. Nanoscale heat transfer and nanostructured thermoelectrics. IEEE Trans. Compon. Packag. Technol. 2006, 29, 238–246. [CrossRef] 8. Ezzahri, Y.; Zeng, G.; Fukutani, K.; Bian, Z.; Shakouri, A. A comparison of thin film microrefrigerators based on Si/SiGe superlattice and bulk SiGe. Microelectron. J. 2008, 39, 981–991. [CrossRef] 9. Lee, K.; Mune, Y.; Ohta, H.; Koumoto, K. Thermal stability of giant thermoelectric Seebeck coefficient for

SrTiO3/SrTi0.8Nb0.2O3 superlattices at 900 K. Appl. Phys. Express 2008, 1, 015007. [CrossRef] 10. Liu, W.L.; Borca-Tasciuc, T.; Liu, J.L.; Taka, K.; Wang, K.L.; Dresselhaus, M.S.; Chen, G. In-plane thermoelectric properties of Si/Ge superlattice. In Proceedings of the ICT 2001, XX International Conference on Thermoelectrics, Beijing, China, 8–11 June 2001; pp. 340–343. 11. Sztein, A.; Bowers, J.E.; DenBaars, S.P.; Nakamura, S. Polarization ﬁeld engineering of GaN/AlN/AlGaN superlattices for enhanced thermoelectric properties. Appl. Phys. Lett. 2014, 104, 042106. [CrossRef] 12. Ghamaty, S.; Elsner, N.B. Si/SiGe quantum well thermoelectric materials and devices for waste heat recovery from vehicles and industrial plants. In Proceedings of the International Symposium on Nano-Thermoelectrics, Toyonaka, Japan, 11–12 June 2007. 13. Bulusu, A.; Walker, D.G. Modeling of thermoelectric properties of semi-conductor thin ﬁlms with quantum and scattering effects. J. Heat Transf. 2007, 129, 492–499. [CrossRef] 14. Dresselhaus, M.S.; Chen, G.; Tang, M.Y.; Yang, R.G.; Lee, H.; Wang, D.Z.; Ren, Z.F.; Fleurial, J.; Gogna, P. New directions for low-dimensional thermoelectric materials. Adv. Mater. 2007, 19, 1043–1053. [CrossRef] 15. Elsner, N.B.; Ghamaty, S.; Norman, J.H.; Farmer, J.C.; Foreman, R.J.; Summers, L.J.; Olsen, M.L.;

Thompson, P.E.; Wang, K. Thermoelectric performance of Si0.8Ge0.2/Si heterostructures synthesized by MBE and sputtering. AIP Conf. Proc. 1994, 316, 328–333. 16. Yu, T.; Efstathiadis, H.; Matyi, R.; Haldar, P.; Ghamaty, S.; Elsner, N. Large area depositions of Si/SiC quantum well ﬁlms for thermoelectric generator applications. MRS Proc. 2005, 886, 0886-F04-08. [CrossRef] 17. Zheng, Z.H.; Fan, P.; Chen, T.B.; Cai, Z.K.; Liu, P.J.; Liang, G.X.; Zhang, D.P.; Cai, X.M. Optimization in fabricating bismuth telluride thin ﬁlms by ion beam sputtering deposition. Thin Solid Films 2012, 520, 5245–5248. [CrossRef] Coatings 2018, 8, 109 10 of 10

18. Hendricks, T.J. Comparison of skutterudites and advanced thin-ﬁlm B4C/B9C and Si/SiGe materials in advanced thermoelectric energy recovery systems. In Proceedings of the ICT 2005, 24th International Conference on Thermoelectrics, Clemson, SC, USA, 19–23 June 2005; pp. 369–375. 19. Ghamaty, S. Quantum Well Thermoelectrics for Converting Waste Heat to Electricity; Hi-Z Technology Inc.: San Diego, CA, USA, 2005.

20. Martin, P.; Olsen, L. Scale up of Si/Si0.8Ge0.2 and B4C/B9C Superlattices for Harvesting of Waste Heat in Diesel Engines; Pacific Northwest National Lab.: Richland, WA, USA, 2003. 21. Ghamaty, S.; Elsner, N.B. Thermoelectric Device with Si/SiC Superlattice n-legs. U.S. Patent 6,828,579, 7 December 2004. 22. Jurgensmeyer, A.L. High Efficiency Thermoelectric Devices Fabricated Using Quantum Well Confinement Techniques; M.S. Colorado State University: Fort Collins, CO, USA, 2011; pp. 1–44. 23. Aubain, M.S.; Bandaru, P.R. Determination of diminished thermal conductivity in silicon thin films using scanning thermoreflectance thermometry. Appl. Phys. Lett. 2010, 97, 253102. 24. Mazumder, M.; Borca-Tasciuc, T.; Teehan, S.C.; Stinzianni, E.; Efstathiadis, H.; Solovyov, S. Temperature dependent thermal conductivity of Si/SiC amorphous multilayer films. Appl. Phys. Lett. 2010, 96, 093103. [CrossRef] 25. Surana, K.; Lepage, H.; Lebrun, J.M.; Doisneau, B.; Bellet, D.; Vandroux, L.; Le Carval, G.; Baudrit, M.; Thony, P.; Mur, P. Film-thickness-dependent conduction in ordered Si quantum dot arrays. Nanotechnology 2012, 23, 105401. [CrossRef][PubMed] 26. Yang, S.; Cai, W.; Zeng, H.; Xu, X. Ultra-fine β-SiC quantum dots fabricated by laser ablation in reactive liquid at room temperature and their violet emission. J. Mater. Chem. 2009, 19, 7119–7123. [CrossRef] 27. Cho, N.-I.; Jung, K.; Choi, Y. Improved ohmic contact to the n-type 4H-SiC semiconductor using cobalt silicides. Semicond. Sci. Technol. 2004, 19, 306. [CrossRef] 28. Carlson, D.E.; Wronski, C.R. Amorphous silicon solar cell. Appl. Phys. Lett. 1976, 28, 671–673. [CrossRef] 29. Le Comber, P.G.; Spear, W.E. Electronic transport in amorphous silicon films. Phys. Rev. Lett. 1970, 25, 509. [CrossRef] 30. Hinsche, N.F.; Mertig, I.; Zahn, P. Effect of strain on the thermoelectric properties of silicon: An ab initio study. J. Phys. Condens. Matter 2011, 23, 295502. [CrossRef][PubMed]

31. Zou, D.; Liu, Y.; Xie, S.; Lin, J.; Li, J. Effect of strain on thermoelectric properties of SrTiO3: First-principles calculations. Chem. Phys. Lett. 2013, 586, 159–163. [CrossRef] 32. Lv, H.Y.; Lu, W.J.; Shao, D.F.; Sun, Y.P. Enhanced thermoelectric performance of phosphorene by strain-induced band convergence. Phys. Rev. B 2014, 90, 085433. [CrossRef] 33. Fei, R.; Yang, L. Strain-engineering the anisotropic electrical conductance of few-layer black phosphorus. Nano Lett. 2014, 14, 2884–2889. [CrossRef][PubMed] 34. Ohta, H.; Kim, S.; Mune, Y.; Mizoguchi, T.; Nomura, K.; Ohta, S.; Nomura, T.; Nakanishi, Y.; Ikuhara, Y.;

Hirano, M.; et al. Giant thermoelectric seebeck coefﬁcient of a two-dimensional electron gas in SrTiO3. Nat. Mater. 2007, 6, 129–134. [CrossRef][PubMed]

35. Song, F.F.; Wu, L.; Liang, S. Giant seebeck coefﬁcient thermoelectric device of MnO2 powder. Nanotechnology 2012, 23, 085401. [CrossRef][PubMed] 36. Ashby, S.P.; Thomas, J.A.; García-Cañadas, J.; Min, G.; Corps, J.; Powell, A.V.; Xu, H.; Shen, W.; Chao, Y. Bridging silicon nanoparticles and thermoelectrics: Phenylacetylene functionalization. Faraday Discuss. 2014, 176, 349–361. [CrossRef][PubMed] 37. Gharsallah, M.; Serrano-Sánchez, F.; Nemes, N.M.; Mompeán, F.J.; Martínez, J.L.; Fernández-Díaz, M.T.; Alonso, J.A. Giant seebeck effect in Ge-doped SnSe. Sci. Rep. 2016, 6, 26774. [CrossRef][PubMed] 38. Snyder, G. J.; Toberer, E.S. Complex thermoelectric materials. Nat. Mater. 2008, 7, 105–114. [CrossRef] [PubMed]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).