J. Am. Ceram. Soc., 90  533–540 (2007) DOI: 10.1111/j.1551-2916.2006.01410.x r 2006 The American Ceramic SocietyJournal Oxide Materials with Low Thermal Conductivity Michael R. Winter and David R. Clarkew Materials Department, College of Engineering, University of California, Santa Barbara, California 93160-5050A heuristic approach to identifying candidate materials with materials so that we can compare the intrinsic thermal conduct-low, temperature-independent thermal conductivity above room ivity of materials as it is well known that porosity can signiﬁ-temperature is described. On the basis of this approach, a num- cantly decrease the measured thermal conductivity.ber of compounds with thermal conductivities lower than that of8 mol% yttria-stabilized zirconia and fused silica have beenfound. Three compounds, in particular, the Zr3Y4O12 delta II. Criteria for Low Thermal Conductivityphase, the tungsten bronzes, and the La2Mo2O9 phase, exhibit Although no theory exists for predicting which compounds willpotential for low thermal conductivity applications. As each can have a low thermal conductivity at high temperatures, there areexhibit extensive substitutional solid solution with other, high a number of guidelines that we have used to identify candidateatomic mass ions, there is the prospect that many more com- materials. The ﬁrst are expressions for what has become knownpounds with low thermal conductivity will be discovered. as the minimum thermal conductivity, kmin.8–10 This is the low- est thermal conductivity that a non-metallic solid can exhibit based on its atomic number density, its phonon spectrum, and I. Introduction other parameters.8 Of particular interest to our work is the high- temperature limit to the minimum thermal conductivity, whichW ITHsome exceptions, the majority of non-metallic com- pounds exhibit a thermal conductivity that decreases withincreasing temperature above room temperature.1–4 Two not- is based on the assumption that the phonon mean free path ap- proaches the mean inter-atomic distance. A number of similar expressions have been derived for this high-temperatureable exceptions are fused silica2 and stabilized zirconia4–6 (and limit.8–10 For instance, Cahill and Pohl9 have expressed this asthe structurally related defect ﬂuorite compounds ceria, thoria,and urania3). These two materials also have the lowest thermal kB 2=3conductivity of any fully dense oxide (Fig. 1). kmin ¼ n ð2vt þ vl Þ (1) 2:48 One of the applications of yttria-stabilized zirconia (YSZ)that takes advantage of its unusually low thermal conductivity is where n is the number of atoms per unit volume and vt and vl areas a thermal barrier coating for combustors and turbine blades the transverse and longitudinal sound velocities. An alternativein gas turbine engines, enabling them to operate at higher gas expression, based on the same limiting conditions for the pho-temperatures than without the coating. To further increase en- non mean free path and the effective atomic masses, that alsogine performance, there has been interest in ﬁnding an alterna- gives almost the same values for simple compounds but is ex-tive material with still lower, temperature-independent thermal pressed in more easily obtained physical parameters is:10,11conductivity up to high temperatures. The discovery that certainrare-earth zirconates4,7 with the pyrochlore crystal structure À2=3have lower thermal conductivity than YSZ and also exhibit an kmin ¼ 0:87kB O ðE=rÞ1=2 (2)almost ﬂat temperature dependence has demonstrated that sta-bilized zirconia is not unique among crystalline oxides in this where Oa is an effective atomic volumerespect and has spurred interest in the search for other oxides Mthat also have low thermal conductivity. Although the pyro- Oa ¼ (3)chlore zirconates have a crystal structure closely related to the mrNAbasic ﬂuorite structure of stabilized zirconia, the fact that theirconductivity is lower than YSZ suggests that there is potential where M is the mean atomic mass of the ions in the unit cell, m isfor many other materials with low thermal conductivity. the number of ions in the unit cell, r is the density, and E is In this contribution, we describe some of the results of our Young’s modulus. kB and NA are Boltzmann’s constant andsearch for oxide materials having low thermal conductivity. This Avagadro’s number.paper is organized as follows: after a description of our ap- These expressions indicate that a large mean atomic mass andproach to identifying candidate materials with low thermal con- a low elastic modulus favor low thermal conductivity. Althoughductivity and a brief description of experimental details, we helpful as a ﬁrst step in selecting materials, one would not iden-present our results in sections devoted to different classes of tify silica glass as having an especially low thermal conductivitymaterial. For the purposes of this work, we include only those on this basis nor do they provide any additional insight into thematerials that proved to have thermal conductivities lower than types of crystal structure that may exhibit low thermal conduct-that of dense YSZ. Although arbitrary, it serves to identify those ivity.materials that might, for instance, be of interest in future ther- To proceed, we have been guided by two additional concepts.mal barrier coating systems. Our focus has been on fully dense One is to introduce randomly distributed point defects into a structure at a sufﬁciently high density that they will cause in- elastic phonon scattering, thereby decreasing the phonon mean N. Padture—contributing editor free path and decreasing the attainable thermal conductivity. There are a number of classic examples of this effect. These in- clude the decrease in thermal conductivity of AlN with the con- Manuscript No. 21956. Received June 29, 2006; approved October 4, 2006. centration of oxygen impurities12 and the increase in thermal This work was supported by the Ofﬁce of Naval Research as part of the TBC MURI conductivity of diamond when it is made of isotopically puregrant through contract N014-02-1-0801. w Author to whom correspondence should be addressed. e-mail: clarke@engineer- carbon.13 Similarly, both the alloying of germanium with sili-ing.ucsb.edu con14 and the alloying of indium with GaAs15 decrease the ther- 533
534 Journal of the American Ceramic Society—Winter and Clarke Vol. 90, No. 2 5 amorphous structure precluding the existence of long wave- length phonons so that the phonon mean free path is limited by m-ZrO2  the size of the SiO4 tetrahedral structure. Using the simple De- Thermal Conductivity, W/mK 4 bye model for thermal conductivity and using the measured thermal conductivity, ﬁrst Kittel19 and then Kingery1 calculated Dense Nano-YSZ Fused SiO2 that the phonon mean free path corresponds closely to the size 3 YSZ  (excluding of the tetrahedral unit. Although no non-silicate glass is known (dense)  radiation)  to be stable to high temperatures, it does suggest that crystal structures with various forms of short-range disorder and con- 2 sisting of structural units that are loosely connected may also Gd 2 Zr 2O 7  exhibit low thermal conductivity. Taken together, these guiding concepts suggest that materials 1 having low thermal conductivity at high temperature will be those that have high mean atomic mass, weak, non-directional bonding as well as extensive disorder at a variety of length scales 0 0 500 1000 1500 down to that corresponding to inter-atomic spacings. Not all of Temperature (°C) these structures will be stable to the very highest temperatures required for thermal barrier application but may help identifyFig. 1. Thermal conductivity of a number of oxide materials including other compounds with higher melting temperatures. This is anmonoclinic zirconia,1 dense yttria-stabilized zirconia (YSZ),4,6 the alternative to the search approach based on ﬁrst identifyingGd2Zr2O7 pyrochlore,6 and, for comparison, fused silica.2 There are compounds that have the highest melting temperatures, such asseveral reports of the thermal conductivity of the pyrochlores but only appears to be the basis for a number of other searches.7,20,21Wu et al. 6 is for fully dense material.mal conductivity of both these semiconductors. Kingery1 also III. Experimental Procedureshowed that solid solution alloying in the NiO–MgO system Unless otherwise described, all the materials studied in this workdecreased the thermal conductivity. However, in each of these were performed by what will be referred to as standard powdermaterials, as well as alloying in a variety of minerals16 and mixed processing routes in which commercially available oxide pow-halides, cyanides, and ﬂuorides,17 the effect is to decrease the ders were ﬁrst mixed together, milled, ground, and then sinteredthermal conductivity in the 1/T regime of thermal conductivity. to full density. Deviations from this method were necessary forNo effect has been reported in the high-temperature regime but compositions that did not densify easily and these are describedthere is no reason to suppose that alloying, or mass disorder, will where appropriate in the sections devoted to specific materials.also not be effective in the high-temperature regime. As we will In the standard process, the powders were mixed together andshow, this can, in fact, be quite effective in reducing thermal attrition milled for 3 h in deionized water with zirconia grindingconductivity at high temperatures, especially when alloying is media. The milled powder was passed through a o325 meshused to create disorder on more than one of the sub-lattices in a sieve, dried, and ground to break up large agglomerates. Then,crystal structure. the powders were formed into disks by uniaxial cold pressing (80 The second concept is to seek structures that have the op- MPa) without a binder and sintered in air, typically at 16001C.posite characteristics of materials with very high thermal con- Once dense samples were created, they were machined to size forductivity.18 Such materials, for instance diamond, beryllium thermal diffusivity measurements using a surface grinder. Someoxide, silicon carbide, aluminum nitride, and gallium nitride, materials had to be hot pressed. These were also ground to sizehave low atomic mass; highly directional, usually covalent, after removing any graphite that had adhered to the surfaces asbonding; high lattice stiffness; high purity; low defect concen- well as any reaction zone. Finally, all the samples were annealedtration; and minimum disorder in addition to being electrical in air at 10001C for 30 min to equilibrate them before testing.insulators. These common characteristics lead to high phonon The density was measured using the Archimedes technique. Thevelocities and minimization of phonon scattering. Silica glasses values are listed in Table I, together with the method used toare the classic example of materials that exemplify this set of make the samples.opposite characteristics. Although silica glass has a low density X-ray diffraction analysis and Raman spectroscopy wereand is primarily covalent, it has a ﬂexible structure consisting of used to conﬁrm that the materials had the desired crystal struc-corner-shared SiO4 tetrahedra that can rotate and ﬂex about one ture and were single phase. The thermal conductivity, k, of eachanother. The contrary behavior of silica glasses was attributed, of the materials was determined from separate measurements ofas long ago as 1949 by Kittel,19 to being a consequence of the the thermal diffusivity, a, heat capacity, C, and density, r, using Table I. Sample Density and ProcessingCompound Measured density (kg/m3) X-ray density (kg/m3) Density (%) Processing7YSZ 6.04 6.134 98.5 Hot pressed/16001C/60 minGdPO4 5.94 6.061 98.0 Hot pressed/12001C/30 minLaPO4 4.60 5.067 90.8 Hot pressed/11501C/45 minYPO4 3.63 4.262 85.2 Hot pressed/15501C/60 minGd0.5La0.5PO4 5.48 5.564 98.5 Hot pressed/12001C/30 minZr3Y4O12 5.45 5.49 99.3 Sintered in air/16001C/2 h0.1WO3–0.9Nb2O5 4.70 4.74 99.2 Hot pressed/12751C/30 minWNb12O33 4.70 4.76 98.7 Hot pressed/12751C/30 minW4Nb26O77 4.78 4.93 97.0 Hot pressed/12751C/30 minW3Nb14O44 5.02 5.02 100 Hot pressed/12751C/30 min(3.5Eu–3.5Tm–7Y) SZ 6.25 6.24 100 Hot pressed/15501C/60 min(3.5Eu–3.5Yb–7Y) SZ 6.17 6.25 98.7 Hot pressed/15501C/60 minLa2Mo2O9 5.02 5.564 90.2 Sintered in air/10501C/10 h
February 2007 Oxide Materials with Low Thermal Conductivity 535 Table II. Atomic Mass and Ionic Radii (Ionic Radii after Shannon43)Ion Zr41 Hf41 Y31 Tm31 Eu31 Yb31 La31 Gd31Atomic mass (amu) 91.2 178.58 88.91 168.93 151.96 173.04 138.91 157.25Ionic radius (nm) 0.084 0.083 0.1019 0.0994 0.1066 0.0985 0.1160 0.1053the relationship In a very important practical development, Zhu and Mill- er25,26 at NASA Glenn have demonstrated that coatings stabil- k ¼ rCP k (4) ized with a combination of trivalent ions have exceptionally low thermal conductivity. These coatings have been referred to as Thermal diffusivity was measured with an Anter Flashline ‘‘cluster-doped’’ coatings and are based on the substitution of3000 thermal ﬂash system (Pittsburgh, PA) based on the tech- three different trivalent cations, typically Y31 ion plus one ionnique pioneered by Parker et al.22 This instrument uses a high- larger and one smaller. Unlike other coating materials, the ther-speed xenon lamp and measures the corresponding increase in mal conductivity of these coatings did not change substantiallybackside temperature with an InSb infrared detector. All the over extended periods of time at high temperatures.27samples were ground coplanar to a thickness of approximately 1 An alternative approach to reducing thermal conductivity ismm and a diameter of approximately 12.4 mm. To minimize to replace some of the Zr41 ions with Hf41 ions. In contrast toradiative heat transport through the sample, which can lead to the rare-earth substitutions used in the ‘‘cluster’’-doped zirconia,an apparently higher diffusivity, both sides were coated with a Hf41 ions do not introduce vacancies into the oxygen sub-lattice10 nm Ti adhesion layer and a 750 nm of Pt layer via electron but produce random mass disorder into the cation sub-lattice. Abeam physical vapor deposition. Both sides were also spray peculiar feature of Hf41 ions is that they are chemically similarcoated with a colloidal graphite approximately 10 mm thick to to Zr41 ions, have a similar ionic radius but are almost twice asensure full absorption of the ﬂash at the front surface and max- massive.imum emissivity from the backside. The samples were thenheated to drive off any remaining solvents. The mass of the (1) Multiple Dopant Substitutionsample was measured before and after the test to ensure that the To evaluate the possible effect of multiple stabilizer dopants ingraphite coating remained intact. Thermal diffusivity was meas- fully dense material as distinct from coatings, we produced twoured at 1001C intervals from 1001 to 10001C and then allowed to sets of dense zirconia material stabilized with three different tri-cool to 1001C before a ﬁnal measurement was made. A com- valent ions, Y31 and Eu31, and either Yb31 or Tm31. The ionicparison of the diffusivities measured at 1001C at the beginning radii are listed in Table II. The total concentration of the sta-and end of the test series was made to ensure that there were no bilizer ions was 14 m/o, so that the material was tetragonal, anchanges in the sample or in the coating integrity. All the meas- expectation conﬁrmed by X-ray diffraction and Raman spec-urements were made in nitrogen. The thermal diffusivity was troscopy. The stabilizer concentration was chosen as the work ofdetermined using the correction described by Clarke and Tay- Zhu and Miller25 indicates that coatings made with this totallor,23 which accounts for radiative losses that occur due to non- stabilizer concentration had the lowest thermal conductivity inadiabatic conditions. Calibration of the thermal ﬂash system their thermal gradient testing. The powders of these materialswas performed with a standard sample of pyroceram 9606 glass– were prepared in the standard way but were then densiﬁed byceramic. The pyroceram sample was purchased from Anter Cor- hot pressing at 20 MPa for 1 h at 15501C as they could not beporation (Pittsburgh, PA). The day-to-day variation in the ther- densiﬁed by sintering. Even after sintering for 2 h at 16001C,mal diffusivity of the pyroceram sample was 2%–3%. Together they only attained a density of B70%.with variations in the heat capacity measurements, the uncer- The thermal conductivity of two of the fully dense materials istainty in the reported values here of thermal conductivity is es- shown in Fig. 2. One, the composition stabilized withtimated to be 75%. 7YO1.513.5 EuO1.5 and 3.5 TmO1.5, had a temperature-inde- Heat capacity was measured with a Netzsch Pegasus 404C pendent conductivity of B2.0 W/mK, independent of tempera-differential scanning calorimeter (Estes, CO). Samples were ture. This value is almost the same as has been reported forground coplanar to approximately 1 mm thickness and approxi- zirconia stabilized with 14 m/o YO1.5 28 but much higher thanmately 5.5 mm diameter. Heat capacity was determined accord- that reported for the NASA coatings. The other, the one sta-ing to the ASTM standard E1269-95 in a platinum crucible from bilized with 7YO1.513.5 EuO1.5 and 3.5 YbO1.5, exhibited veryroom temperature to 10001C in an argon atmosphere.24 3.0 7YSZ+ 3.5 EuO + 3.5 TmO IV. Decreasing Thermal Conductivity of Zirconia by 7YSZ+ 3.5 EuO + 3.5 YbO Thermal Conductivity (W/mK) Substitution on the Cation Sub-Lattice 2.5YSZ has several of the characteristics of low conductivity de-scribed in Section II. Without stabilization by yttria (or anotherstabilizer ion, such as the trivalent rare-earth ion, divalent cat-ions, or mixed-valence ions), zirconia is monoclinic and exhibits 2.0the classic 1/T temperature dependence.1 The substitution ofY31 for Zr41 in the ﬂuorite structure of ZrO2 stabilizes the te-tragonal phase by introducing vacancies into the oxygen sub-lattice. At typical concentrations of 8 m/o YO1.5, the spacing of 1.5the oxygen vacancies is B1.25 nm. (At higher yttria concentra- (Zr,Hf) Y Otions, the concentration of vacancies increases but clusteringbegins to occur.) The unusually low thermal conductivity of 1.0YSZ is generally attributed to enhanced phonon scattering from 0 200 400 600 800 1000 1200oxygen vacancies. Although the Y31 ions substitute for Zr41ions in the cation sub-lattice, their atomic mass is almost iden- Temperature (°C)tical to that of the Zr41 ions and hence are unlikely to cause any Fig. 2. Comparison of the thermal conductivity of two tri-dopedmass disorder on the cation sub-lattice. zirconias and the equi-molar hafnia-zirconia material.
536 Journal of the American Ceramic Society—Winter and Clarke Vol. 90, No. 2pronounced radiative transport effects from almost 3001C on- 3.0wards. Why radiative effects are so pronounced in this compos-ition is not known but was duplicated with a second sample. All Thermal Conductivity (W/m K)the samples were white indicating extensive internal optical scat- 8YSZtering. Nevertheless, at 1001C, the thermal conductivity was 2.5somewhat lower at B1.8 W/mK. Two conclusions can be drawnfrom these results. The ﬁrst is that it is the mean value of the Zr3Y4O12ionic mass of the stabilizer ions that is important rather than thevariation in ionic mass or size. The second, and of crucial im- 2.0portance for the design of coatings, is that the low thermal con-ductivity of the ‘‘cluster’’-doped coatings is not intrinsic to thecomposition per se but most probably due to the presence of (Zr,Hf)3Y4O12porosity incorporated during the deposition of the coating and 1.5its resistance to densiﬁcation during deposition and subsequenthigh-temperature exposure. 1.0(2) Hafnia Substitution 100 200 300 400 500 600 700 800 900 1000As described in detail elsewhere, hafnia substitution de- Temperature (°C)creases the thermal conductivity of YSZ across the full solid–solution series.29 The most notable reduction, however, Fig. 3. Comparison of the thermal conductivity of the Zr3Y4O12 deltaoccurs close to the equi-molar mixtures of hafnia and zir- phase and the hafnia–zirconia counterpart with that of dense yttria-conia, such as ðZr0:5 Hf 0:5 Þ0:87 Y0:13 O1:94 tetragonal and the stabilized zirconia (8YSZ).ðZr0:4 Hf 0:6 Þ0:76 Y0:24 O1:88 cubic phase. The data for the formercomposition are shown in Fig. 2 and, for comparison, with the arranged in a tetragonal structure. According to the ﬁrst crystalthermal conductivity of the multiple-stabilizer zirconia, illus- structure determination, the W and Nb ions were randomlytrates the effectiveness of mass disorder on the cation sub-lattice. distributed among all the octahedral sites. Further structuralIn contrast to the cluster-doped materials, these could be readily reﬁnement by neutron diffraction indicates that, at room tem-densiﬁed to better than 96% by sintering in air at 16001C for 2 h. perature at least, there is a preference for the W to be in anThe measured lattice parameters and molecular weight were octahedral co-ordination and to be distributed among the octa-used to calculate the X-ray density of the individual compos- hedra in the vicinity of the block-sharing W ions. In selectingitions. these compounds for study, it was also hypothesized that the corner sharing of the blocks and the disorder along the c-axis associated with the W ions might provide the vibrational free- V. Thermal Conductivity of the Delta Phase Zr3Y4O12 dom that would lead to low thermal conductivity. In addition,Previous studies have shown that the thermal conductivity of several of the tungsten bronzes have some of the highest meanzirconia decreases with increasing yttria concentration up to atomic masses and densities of the oxides and several are stableabout 20 m/o YO1.5 but then is approximately constant.28 This to well over 14001C. Also, the size of the blocks varies with thesaturation behavior has been attributed to ordering of the sub- W/Nb ratio, leading to a series of line compounds in the phasestitutional Y31 ions and the oxygen vacancies, limiting the ran- diagram, as shown in Fig. 4.dom disordering on the oxygen sub-lattice.30,31 The maximum Four compounds were investigated and are indicated in theconcentration that seems to have been studied is 30% Y2O3, well diagram of Fig. 4. Three were single-phase line compounds hav-within the cubic, ﬂuorite phase. However, the ZrO2–YO1.5 phase ing different W/Nb ratios, W3 Nb14 O44 , (D), WNb12 O33 (B), anddiagram indicates that another crystalline phase, the delta phase, W4 Nb26 O77 (C). The fourth was a deliberately made two phaseZr3Y4O12, exists.32 mixture of Nb12 WO39 and Nb60 WO183 marked A on Fig. 4. All The experiment indicates that this compound does have a low the compounds were made from oxides and were calcined atand temperature-independent thermal conductivity as shown by 11001C for 2 h in air. They then had to be hot pressed and thenthe data in Fig. 3. Its thermal conductivity parallels the thermal annealed in air at 10001C for 1 h.conductivity of YSZ but has a value almost 30% lower up to at The measured thermal conductivity (Fig. 5) indicates that theleast 10001C. The delta-phase composition was readily sintered tungsten bronzes do indeed have low thermal conductivity andin air at 15501C for 2 h but could not be fully ordered even after are temperature independent. The actual values are similar to3 weeks at 12501C. those of both YSZ and fused silica. Notably, both the single- phase materials and the two-phase material exhibited similar(1) Hafnia Substitution in the Delta Phase conductivities.Based on the success in lowering the thermal conductivity bysubstituting Zr41 ions with Hf41 ions,29 the equi-molar com-position ðZr0:5 Hf 0:5 Þ3 Y4 O12 was prepared in a fully dense form. VII. Lanthanide OrthophosphatesIts thermal conductivity as a function of temperature is shown in Orthophosphates have crystal structures that are analogous toFig. 3 and demonstrates clearly that introducing mass disorder the network silicates with various arrangements of corner- andinto the cation sub-lattice can signiﬁcantly reduce thermal edge-sharing PO4 tetrahedra rather than the SiO4 tetrahedra.35conductivity from room temperature up to at least 10001C. Based on the guidelines described in ‘Section II,’ they are a nat- ural choice of compounds for searching for low thermal con- ductivity. Perhaps the best known example is the aluminum VI. Tungsten Niobates phosphate (AlPO4) but an extensive series of lanthanide-basedIn contrast to the zirconia-based oxides, the basis for selecting orthophosphates (LnPO4), also exist, which have a signiﬁcantlythe tungsten niobates, sometimes referred to as the ‘‘tungsten higher density because of the higher atomic mass of the lantha-bronzes,’’ for study is that their crystal structures have some nide ion. They form two distinct, but closely related, crystalfeatures common to silica. They consist of blocks of corner- structures, the monazite and the xenotime structures, dependingshared NbO octahedra linked together with W ions.33,34 This is on the ionic size of the lanthanide ions.36 The monazite structureillustrated for the W3 Nb14 O44 compound in Fig. 4 showing the is typically formed with the larger lanthanide ions and the xeno-blocks of 4 Â 4 corner-sharing octahedra joined by W ions and time structure with the smaller lanthanide ions. Both structures
February 2007 Oxide Materials with Low Thermal Conductivity 537 (a) (b) 1500 1476 1470 1464 1440 Liquid 1435 1435 1400 1385 1380 1357 1358 1365 1335 B C 1340 1356 Temperature (°C) 1300 A D ~1265 ~1275 9:8 ~1270 1:11 1:15 2:7 ~1245 ~1210 9:16 1200 9:17 ~1115 1100 ~1090 4:9 13:4 8:5 "3.8" 1000 6:1 20 7:3 40 1:1 60 80 Nb2O5 WO3 Mol.%Fig. 4. (a) Schematic of the crystal structure of W3Nb14O44 looking down the tetragonal axis. The structure consists of blocks of 4 Â 4 corner-sharingNbO octahedra joined by the tungsten ions shown as circles in the square unit cell. (b) Section of the Nb2O3–WO3 phase diagram illustrating thecompositions of the compounds, A, B. C and D measured.are based on chains of alternating phosphate tetrahedral and RE rationale for the last was that although it has a different crystalpolyhedra. Because of the well-known lanathanide ion contrac- structure, it has the same unit cell volume as does YPO4 and sotion, these compounds are formed with the lighter and heavier one might expect that the average inter-atomic spacing would beions, respectively. The boundary between the monazite and the same. The materials were made from chemically precipitatedxenotime structures occurs between the GdPO4 and the precursor powders prepared by adding phosphoric acid, drop byTbPO4 compounds, the monazite being formed with the Gd31 drop, to the appropriate nitrate solution in deionized water.ions and the xenotime being formed with the Tb31 ions. The powders were washed several times with deionized water It has been reported that the monazite compound CePO4 has and centrifuged between washings to avoid powder loss. Then,a thermal conductivity of B3.1 W/mK at room temperature and they were calcined at 7001C for 1 h in air. Before hot pressing,a value of B1.8 W/mK at 5001C37, suggesting a weak 1/T de- the powders were heated above 5001C to remove any water aspendence. Similarly, the xenotime compounds YPO4, LuPO4, they are hygroscopic. The YPO4 was hot pressed at 15501C,YbPO4, and ErPO4 all have a thermal conductivity of B12 W/ whereas the monazite compounds were hot pressed at 12001C.mK at room temperature.38 The thermal conductivity of LaPO4 Even under these conditions, not all the materials were fullyat room temperature has previously been reported.39 dense; the YPO4 samples were B85% dense and the LaPO4 To compare the thermal conductivity of a number of the samples were 90% dense. Both the GdPO4 and ðLa0:5 Gd0:5 ÞPO4monazite and xenotime structures, we measured the conductiv- samples were 98% dense. Two samples of each compositionity of YPO4 (a xenotime compound) and GdPO4 and LaPO4, were made.both having the monazite structure. We also investigated the The measured thermal conductivities are shown in Fig. 6. Theco-doped ðLa0:5 Gd0:5 ÞPO4 , another monazite compound. The most striking feature of the data is that the orthophosphate with
538 Journal of the American Ceramic Society—Winter and Clarke Vol. 90, No. 2 3.0 3.0 W Nb 0 WNb O 8YSZ Thermal Conductivity (W/mK) Thermal Conductivity (W/mK) W Nb O 2.5 Two phase Tri-doped YSZ Zr Y O 2.0 W Nb O (Zr,Hf) Y O 2.0 (Zr,Hf) Y O LaPO 1.0 1.5 La Mo O 1.0 0 200 400 600 800 1000 1200 0.0 0 200 400 600 800 1000 1200 Temperature (°C) Temperature (°C)Fig. 5. Measured thermal conductivity of the tungsten–bronzes indicated. Fig. 7. Thermal conductivity of La2Mo2O9 compared with the other materials investigated in this work. The density of the La2Mo2O9the xenotime structure has the classic 1/T dependence, whereas samples was 90%.the three with the monazite structure are almost indepen-dent of temperature. Furthermore, the co-doped monazite,ðLa0:5 Gd0:5 ÞPO4 , clearly has quite different conductivity char- time structure, some of the tetrahedra are linked to the REO8acteristics from the xenotime YPO4 even though the average polyhedra by corner sharing and others are edge shared. Allinter-atomic distance is the same. Although the atomic mass of these facts suggest that the monazite structure is more disor-the cations is rather similar, the difference in temperature de- dered at the inter-atomic length scale than the xenotimependence appears to be directly related to the difference in the structure, consistent with the signiﬁcantly lower thermal con-way in which the basic building blocks are connected. We also ductivity.note that although the three monazites have different unit cell A more extensive survey of the orthophosphates is clearlyvolumes, they have rather similar thermal conductivities. The warranted but we had difﬁculty in making large enough, fullymost obvious difference between the xenotime and monazite dense pellets for thermal diffusivity measurements to explorestructures is that the xenotime structure, which has a higher other ionic substitutions.crystal symmetry, consists of REO8 polyhedra whereas the poly-hedra in the monazites are REO9.36 This would suggest that thebonds are weaker in the monazite than in the xenotime struc- VIII. Lanthanum Molybdatetures. The other distinct difference is that there is much greater Lanthanum molybdate, La2Mo2O9, is a recently discoveredvariation in the RE–O bond length in the monazite, with each of compound that has attracted considerable attention because itthe nine bonds being different than in the xenotime structure exhibits a two to three order of magnitude increase in ionicwhere there are only RE–O bond lengths.36 Similarly, the RE–P conductivity to a value comparable with that of YSZ, when thebond length in the xenotime is shorter than either of the two crystal structure changes on heating at B5851C.40 Our originalRE–P bond lengths in the monazites. In addition, in looking motivation for studying the thermal conductivity of this com-along the  axis, it is evident that there is some rotation of pound was simply to ascertain whether the change in crystalthe PO4 tetrahedra about the  direction in the monazite, structure would affect the thermal conductivity, the rationalewhereas the tetrahedra are all parallel in the xenotime structure. being that a compound with a very high oxygen ionic conduct-Finally, in the monazite structure, the PO4 tetrahedra are linked ivity would have considerable vibrational entropy of the oxygento the REO9 polyhedra by corner sharing, whereas in the xeno- sub-lattice and this might be reﬂected by strong phonon scat- tering and hence low thermal conductivity. However, as our 8 measurements show in Fig. 7, there was no discernable change GdPO4 in thermal conductivity as the crystal structure changed and al- 7 LaPO4 most no variation with temperature. The samples were made by Thermal Conductivity (W/m K) (Gd,La)PO4 attrition milling of the oxides for 3 h in deionized water, fol- 6 lowed by grinding and then calcination in air for 16 h at 8001C. YPO4 They were then cold pressed and sintered in air at 10001C for 5 48 h. They could not be prepared by hot pressing, presumably because of the reaction to form molybdenum carbides. Four 4 samples were made on separate occasions. Each had a density of B90% and all showed almost identical thermal conductivity. 3 Using Maxwell’s relation1,41 to account for the porosity, f, 2 1 kdense ¼ kmeasured (5) ð1 À 1:5fÞ 1 the thermal conductivity of dense lanthanum molybdate would 0 100 200 300 400 500 600 700 800 900 1000 be 17% higher than shown in Fig. 7, namely 0.82 W/mK at 1001C and 0.95 W/mK at 10001C. Temperature (°C) Remarkably, our results show that the thermal conductivity isFig. 6. Comparison of the thermal conductivity of the xenotime-struc- lower than that of any other crystalline compound we havetured YPO4 and three of the monazite compounds. The YPO4 material measured or found in the literature. Part of the reason for its lowtested was 85% dense and the monazites were all 90% of full density. thermal conductivity is undoubtedly its large mean atomic mass.
February 2007 Oxide Materials with Low Thermal Conductivity 539However, this alone does not explain its exceptionally low con- Table III. Calculated Average Atomic Volumeductivity. It is noted, though, that in common with the zirconatepyrochlores, it also has a high concentration of ‘‘structural Compound Average atomic volume O (nm3)vacancies.’’ 7YSZ 0.1158 YPO4 0.1194 LaPO4 0.1278 IX. Discussion and Conclusions Zr3Y4O12 0.1308As mentioned in Section I, it was once considered that YSZ was Gd2Zr2O7 0.1330exceptional for a crystalline material because it exhibits both a W3Nb14O44 0.1385low thermal conductivity at room temperatures and a conduct- La2Mo2O9 0.1410ivity that is constant to high temperatures. However, as the re-sults presented here indicate, several compounds with lowerthermal conductivity than YSZ have been identiﬁed. While the cancies can be expected to limit the allowable lattice phonons asapproach we have used to identify these compounds is, by its well as produce phonon scattering when they are occupied.very nature, qualitative it does suggest that there will be many However, as they are not randomly distributed, by analogyother compounds that will also have a lower thermal conduct- with the oxygen vacancy ordering in YSZ at high yttria con-ivity than YSZ. centrations, it is unlikely that they can be more effective in pho- On the basis of the materials identiﬁed, three strategies ap- non scattering than they are in that material. Several of thepear to be most promising in identifying low thermal conduct- compounds also appear to have unusually large thermal ellips-ivity materials. One is to search for silicate-like structures with a oids for the oxygen ions as measured by X-ray diffraction atlarge effective mass that incorporate structural units that are room temperature, suggesting that they may be particularly ef-joined together by rather ﬂexible linkages. A second is to seek fective in phonon scattering. Another characteristic, though, ofout compounds with low crystal symmetry. The third is to the oxygen sub-lattice in the delta phase as well as both the cat-introduce mass disorder on the cation sub-lattices of a crystal ion and anions sub-lattices in the pyrochlore compounds is thatstructure that is known to have rather low thermal conductivity. they have a Kagome arrangement.44 This is an unusual featureThe concept of introducing mass disorder to lower thermal con- for the anion sub-lattice in solids but is likely to support unusualductivity is not, of course, new. However, hitherto it has only phonon modes.been shown to be effective in the temperature regime where the One of the correlations that we have noted is that all thesethermal conductivity is dependent on temperature, the 1/T re- compounds have rather broad Raman lines, and in some cases,gime. This work demonstrates that the mass disorder strategy is rather poorly deﬁned Raman spectra. An example is shown inalso a viable one in the minimum thermal conductivity regime, Fig. 8. Although we are not aware of any direct correlation be-the regime where the thermal conductivity is temperature inde- tween thermal conductivity and the width of the Raman lines,pendent. the phonon lifetime, t, is inversely proportional to the FWHM At this stage, very little is known about the phase equilibria of of the Raman lines when the lines are not instrumentally limit-the La2Mo2O9 and related compounds to ascertain whether they ed.45 As the phonon mean free path can be expected to be pro-might have applications as thermal insulators or in other appli- portional to the phonon lifetime, then if the mean phononcations beyond fast-ion conductors. It is also unknown how velocity remains unchanged, there will be a correlation betweentheir melting temperatures vary with the rare-earth ion sub- the thermal conductivity and the width of the Raman lines.stitution for the lanthanum ion. Similarly, although the Lastly, it is appropriate to mention that although all the sam-orthophosphates have attractively low thermal conductivity, ples were the same thickness and were either white or darkcontrolling their stoichiometry during deposition has proven colored before metal coating, some nevertheless showedproblematic and any free phosphate is likely to corrode existing appreciable ‘‘transparency’’ in the thermal diffusivity measure-superalloys. Of the compounds investigated in the course of our ments even at relatively low temperatures. Interestingly, increas-studies, the zirconia–yttria delta phase therefore appears to be ing the thickness of the platinum layer to 2.25 mm on both sidesthe most promising for possible application in thermal barrier did not affect the observation. Also, as indicated in Fig. 2, therecoatings. The delta phase is reported to be stable to B13461C could be a significant difference with only a minor change in thebefore transforming to cubic zirconia and the composition itself stabilizer ion. The transparency is not the penetration of thedoes not melt until 24611C. In addition, the constituent elements ﬂash of light through the absorbing layer and the sample that isare known to have similar vapor pressures and hence it is likelythat the compound can be deposited by standard, one-sourceelectron beam deposition. As it is not phase compatible withalumina, the oxide formed by high-temperature oxidation of the Zr3Y4O12majority of bond-coat alloys, it would need to be used in a two-layer conﬁguration with standard 8YSZ as an inner layer pro-viding the thermal cycle resistance. Finally, it is noted that the Relative Intensityzirconia–yttria delta phase is just one of a family of compoundswith the same crystal structure and that cation substitution (Zr,Hf)3Y4O12can be made on both the Zr site and on the Y site, so there apossibility of systematically varying the cations to investigate theeffect on thermal conductivity as has been done for the pyro- (Zr0.75Y0.25) O1.875chlore compounds through modeling.42 It is interesting to speculate as to why the pyrochlore zirco- (Zr0.75 Hf0.25) O1.875nates, the delta phase, the tungsten bronzes, and the lanthanummolybdate all exhibit such low thermal conductivity. One com-mon characteristic is that they have more than one cation sub-lattice and hence the possibility of different phonon modes as 100 200 300 400 500 600 700well as disorder on each. They all have a large average atomicvolume, O, as shown in Table III and calculated from published Wavenumber (cm −1)data on densities and formula units per unit cell. Another com- Fig. 8. Raman spectra of the indicated compounds recorded with a 514mon characteristic of the unit cells of these compounds is that nm argon ion laser. The spectra are characterized by broad Ramanthey all incorporate ‘‘structural vacancies.’’ These oxygen va- bands much broader than that of most oxides.
540 Journal of the American Ceramic Society—Winter and Clarke Vol. 90, No. 2 13usually considered in papers on measuring diffusivity by the L. Wei, P. K. Kuo, R. L. Thomas, T. R. Anthony, and W. F. Banholzer, ‘‘Thermal Conductivity of Isotopically Modiﬁed Single Crystal Diamond,’’ﬂash technique. As the center wavelength of the xenon ﬂash is at Phys. Rev. Lett., 70, 3762–7 (1993).B1.8 mm, it is unlikely that any light penetrates as it is a wave- 14 B. Abeles, D. S. Beers, G. D. Cody, and J. P. Dismukes, ‘‘Thermal Conduct-length to which the carbon over coat is strongly absorbing and ivity of Ge-Si Alloys at High Temperatures,’’ Phys. Rev., 125, 44 (1962). 15the platinum layer is highly reﬂective. Instead, it appears to be M. S. Abrahams, R. Braunstein, and F. D. Rosi, ‘‘Thermal, Electrical and Optical Properties of (In,Ga)As Alloys,’’ J. Phys. Chem. Solids, 10, 204–10 (1959).associated with radiation from the heated side through the sam- 16 F. Birch and H. Clark, ‘‘The Thermal Conductivity of Rocks and Itsple to the other side that the detector senses. As the majority of Dependence Upon Temperature and Composition,’’ Am. J. Sci., 238, 529–58,the zirconia samples are intrinsically transparent into the mid- 613 (1940). 17infra-red, despite the presence of optical scattering, the radiative D. G. Cahill, S. K. Watson, and R. O. Pohl, ‘‘Lower Limit to the Thermal Conductivity of Disordered Crystals,’’ Phys. Rev., 46  6131–40 (1992).contribution may be especially sensitive to the arrangement of 18 G. A. Slack, ‘‘Nonmetallic Crystals with High Thermal Conductivity,’’pores in the samples that are close to being fully dense and hence J. Phys. Chem. Solids, 34, 321–35 (1973).containing a relatively few number of pores. Under such cir- 19 C. Kittel, ‘‘Interpretation of the Thermal Conductivity of Glasses,’’ Phys.cumstances, there can be direct radiative paths through regions Rev., 75, 972 (1949). 20 X. Q. Cao, R. Vassen, and D. Stoever, ‘‘Ceramic Materials for Thermalof the sample between pores. This effect is likely to be dependent Barrier Coatings,’’ J. Eur. Ceram. Soc., 24, 1–10 (2004).on the thickness of the samples tested, being particularly pro- 21 R. Vassen, F. Tietz, G. Kerkhoff, and D. Stoever, ‘‘New Materials fornounced for thin samples. Advanced Thermal Barrier Coatings’’; pp. 1627–35 in Proceedings of the 6th Liege Conference. Materials for Advanced Power Engineering, Edited by J. Lecomte-Beckers. Forschungszentrum Julich, Julich, Germany, 1998. ¨ ¨ 22 W. J. Parker, R. J. Jenkins, C. P. Butler, and G. L. Abbott, ‘‘Flash Method of X. Summary Determining Thermal Diffusivity, Heat Capacity and Thermal Conductivity,’’Heuristic arguments have been used to identify possible mate- J. Appl. Phys., 32  1679–84 (1961). 23 L. I. Clarke and R. Taylor, ‘‘Radiation Loss in the Thermal Flash Method forrials with low thermal conductivity at high temperatures. These Thermal Diffusivity,’’ J. Appl. Phys., 46  714–9 (1975).arguments suggest that candidate materials can be identiﬁed by 24 ASTM E1269-95. Standard Test Method for Determining Specific Heatsearching for compounds with a combination of high atomic Capacity by Differential Scanning Calorimetry. ASTM International, West Con-mass, ﬂexible ionic structures, non-directional bonding, and ex- shohocken, PA. 25 D. M. Zhu and R. A. Miller, ‘‘Low Conductivity and Sintering Resistanttensive site disorder. We also ﬁnd that substitutional alloying on Thermal Barrier Coatings’’; U.S. Patent 7001859 B2, 2006.the cation sub-lattices is a viable approach to reducing thermal 26 D. M. Zhu and R. A. Miller, ‘‘Thermal Conductivity and Sintering Behaviorconductivity in the kmin regime. Materials that exhibit low ther- of Advanced Thermal Barrier Coatings,’’ Ceram. Eng. Sci. Proc., 23, 457–68mal conductivity generally also exhibit Raman spectra with very (2002). 27 D. M. Zhu, Y. L. Chen, and R. A. Miller, ‘‘Defect Clustering and Nanophasebroad peaks. It is also concluded that compounds with low Structure Characterization of Multi-Component Rare Earth Doped Zirconia–crystal symmetry and with a Kagome arrangement of ions might Yttria Thermal barrier Coatings,’’ Ceram. Eng. Sci. Proc., 24, 525–34 (2003). 28also exhibit unusually low thermal conductivity. J-F. Bisson, D. Fournier, M. Poulain, O. Lavigne, and R. Mevrel, ‘‘Thermal Conductivity of Yttria–Zirconia Single Crystals, Determined with Spatially Resolved Infrared Thermography,’’ J. Am. Ceram. Soc., 83  1993–8 (2000). 29 M. R. Winter and D. R. Clarke, ‘‘The Thermal Conductivity of Yttria- Acknowledgments Stabilized Zirconia–Hafnia Solid Solutions,’’ Acta Mater., 54, 5051–9 (2006). 30 The authors are grateful to Dr. Kurt Sickafus of Los Alamos National J. P. Goff, W. Hayes, S. Hull, M. T. Hutchings, and K. N. Clausen, ‘‘DefectLaboratory for discussions on pyrochlore and delta phases and for providing an Structure of Yttria-Stabilized Zirconia and Its Inﬂuence on the Ionic Conductivityunpublished copy of his work on characterizing crystal structures in terms of layers at Elevated Temperatures,’’ Phys. Rev., B59  14202–19 (1999). 31of cations and anions. T. R. Welberry and B. D. Butler, ‘‘Diffuse X-Ray Scattering from Disordered Crystals,’’ Chem. Rev., 95, 2369–403 (1995). 32 H. G. Scott, ‘‘The Yttria–Zirconia Delta Phase,’’ Acta Crystallogr., B33, 281–2 (1977). 33 References R. S. Roth and A. D. Wadsley, ‘‘Multiple Phase Formation in the Binary System Nb2O5–WO3: III. The Structure of the Tetragonal Phases,’’ Acta Crystal- 1 W. D. Kingery, Introduction to Ceramics, 2nd Edition. Wiley Interscience, New logr., 19, 38–42 (1965). 34York, 1976. R. S. Roth and A. D. Wadsley, ‘‘Multiple Phase Formation in the Binary 2 K. L. Wray and T. J. Connolly, ‘‘Thermal Conductivity of Clear Fused Silica at System Nb2O5–WO3: IV. The Block Principle,’’ Acta Crystallogr., 19, 42–7 (1965). 35High Temperatures,’’ J. Appl. Phys., 30  1702–6 (1959). R. C. Buchanan, Materials Crystal Chemistry. CRC Press, Boca Raton, FL, 3 W. D. Kingery, ‘‘Thermal Conductivity: XII. Temperature Dependence of 1997. 36Conductivity for Single Phase Ceramics,’’ J. Am. Ceram. Soc., 38  251 Y. Ni, J. M. Hughes, and A. N. Mariano, ‘‘Crystal Chemistry of the Monazite(1955). and Xenotime Structures,’’ Am. Mineral., 80  21–6 (1995). 4 37 M. J. Maloney, ‘‘Thermal Barrier Coatings Systems and Materials’’; U.S. Y. Hikichi, T. Nomura, Y. Tanimura, S. Suzuki, and M. Miyamoto, ‘‘Sin-Patent 6117560 (2000) ﬁled 1996. tering and Properties of Monazite-Type CePO4,’’ J. Am. Ceram. Soc., 73  5 S. Raghavan, H. Wang, R. B. Dinwiddie, W. D. Porter, and M. J. Mayo, 3594–6 (1990). 38‘‘The Effect of Grain Size, Porosity and Yttria Content on the Thermal Conduct- Y. Hikichi, T. Ota, K. Dainmon, T. Hattori, and M. Mizuno, ‘‘Thermal,ivity of Nanocrystalline Zirconia,’’ Scr. Mater., 39  1119–25 (1998). Mechanical and Chemical Properties of Sintered Xenotime-Type RPO4 6 J. Wu, X. Wei, N. P. Padture, P. G. Klemens, M. Gell, E. Garcia, P. Miranzo, (R 5 Y,Er,Yb or Lu),’’ J. Am. Ceram. Soc., 81  2216–8 (1998). 39and M. I. Osendi, ‘‘Low Thermal Conductivity Rare-Earth Zirconates for Poten- O. Sudre, J. Cheung, D. Marshall, P. E. D. Morgan, and C. G. Levi, ‘‘Thermaltial Thermal Barrier Coating Applications,’’ J. Am. Ceram. Soc., 85  3031–5 Insulation Coatings of LaPO4,’’ Ceram. Eng. Sci. Proc., 4, 367–74 (2001). 40(2002). F. Goutenoire, O. Isnard, R. Retoux, and P. Lacorre, ‘‘Crystal Structure of 7 R. Vassen, X. Cao, F. Tietz, D. Basu, and D. Stover, ‘‘Zirconates as New La2Mo2O9, A New Fast Oxide-Ion Conductor,’’ Chem. Mater., 12, 2575–80Materials for Thermal Barrier Coatings,’’ J. Am. Ceram. Soc., 83  2023–8 (2000). 41(2000). K. W. Schlichting, N. P. Padture, and P. G. Klemens, ‘‘Thermal Conductivity 8 G. A. Slack, ‘‘The Thermal Conductivity of Nonmetallic Crystals’’; pp. 1–71 in of Dense and Porous Yttria-Stabilized Zirconia,’’ J. Mater. Sci., 36, 3003 (2001). 42Solid State Physics, Vol 34, Edited by F. Seitz and D. Turnbull, Academic Press, P. K. Schelling, S. R. Phillpot, and R. W. Grimes, ‘‘Optimum PyrochloreLondon, 1979. Compositions for Low Thermal Conductivity By Simulation,’’ Philos. Mag. Lett., 9 D. G. Cahill and R. O. Pohl, ‘‘Lattice Vibrations and Heat Transport in 84, 127 (2004). 43Crystals and Glasses,’’ Annu. Rev. Phys. Chem., 39, 93–121 (1988). R. D. Shannon, ‘‘Revised Effective Ionic Radii and Systematic Studies of 10 D. R. Clarke, ‘‘Materials Selection Guidelines for Low Thermal Conductivity Interatomic Distances in Halides and Chalcogenides,’’ Acta Crystallogr., A32, 751Thermal Barrier Coatings,’’ Surf. Coat. Technol., 163–164, 67–74 (2003). (1976). 11 44 C. G. Levi, ‘‘Emerging Materials and Processes for Thermal Barrier Systems,’’ K. Sickafus, ‘‘Layered Atom Arrangements in Complex Materials’’; LANLCurr. Opin. Solid State Mater. Sci., 8, 77–91 (2004). report LA-14205, to be published. 12 45 G. A. Slack, R. A. Tanzilli, R. O. Pohl, and J. W. Vandersande, ‘‘The Intrinsic J. M. Ziman, Electrons and Phonons. Oxford University Press, Oxford,Thermal Conductivity of AlN,’’ J. Phys. Chem. Solids, 48  641–7 (1987). 2001. &