ABOUT THE EFFECT OF CARBON NANOTUBES ON THE OVERALL THERMAL CONDUCTIVITY IN Cu Ti/CNT COMPOSITES

Transcription

1 Powder Metallurgy Progress, Vol.10 (2010), No ABOUT THE EFFECT OF CARBON NANOTUBES ON THE OVERALL THERMAL CONDUCTIVITY IN Cu Ti/CNT COMPOSITES Ch. Edtaier, T. Keppert Abstract Copper atrix/carbon Nanotubes (CNTs) coposites and titaniu as active eleent were considered and their roo teperature theral conductivities easured using a steady state heat flow ethod. The evolutions of the theral conductivity as a function of CNT and titaniu concentrations in Cu Ti/CNT coposites are presented. As opposed to expectations, the theral conductivity of the coposites was below pure copper. However, a positive effect of the CNT additions was clearly detectable as the theral conductivity is higher than unreinforced Cu Ti alloys. Keywords: carbon nanotubes, MMCs (Metal Matrix Coposites), fibre/atrix bond, theral properties INTRODUCTION In recent past decades the proposed outstanding intrinsic properties of Carbon Nanotubes CNTs (ainly ulti-walled Carbon Nanotubes MWNTs) have inspired researchers and fired speculations (and stiulated several project proposals as well), as to whether CNTs could be used in etal atrix coposites MMCs to raise the overall theral conductivity of the coposite bodies [1,2]. Highly conductive filler aterials like CNTs as well as diaonds cae into the focus of MMC research, as heat producing coponents in odern icro-, and power-electronics deand iproved cooling efficiencies [3]. While iproveents in theral properties (theral conductivity and CTE) in diaond MMCs see to be rather easily attainable and technologically advanced, scientific results of CNTs in MMCs are rather scace and ostly disappointing. Although the outstanding intrinsic (therophysical) properties of CNTs are undoubtable, the scientific question is still not answered, if CNTs can positively contribute to an enhanceent of coposite theral conductivity. Few works using Carbon Nanofibres (CNFs) were perfored and show no enhanceent, respectively, a pronounced degradation in theral conductivity [4,5]. This unexpected behaviour can be attributed to a nuber of challenging probles one has to face during ipleentation of the extreely tiny fibres in a etallic atrix: difficulties arising by the high aspect ratio of CNT fibres copared to the relatively large scale diaond crystals, thus leading to probles in unifor distribution of the phases (also aqueous based ones), bundling of the naturally entangled and tortuous CNTs in their asproduced condition, surface and interface probles, alignent of the fibres and equal containations with catalyst particles and carbon/graphitic by-products as well. Although any approaches exist to overcoe those probles, any seeed to be unsolved. The foration of hoogeneous dispersions ight be supported by coating the nanofibres with the atrix etal followed by hot pressing copaction. Due to the lack of cheical reactivity and wetting between carbon and the atrix etal, the investigation of Christian Edtaier, Tiothy Keppert, Vienna University of Technology, Vienna, Austria

2 Powder Metallurgy Progress, Vol.10 (2010), No ethods to adjust the interface between the etal and the carbonaceous inclusion towards excellent theral and echanical behaviour plays an essential role. Several potentially interesting reactive atrices degrade the fibres at high teperatures by severe reactions and therefore developent has concentrated on atrices cheically inert towards carbon, such as copper. In order to address the theral conductance between etal/inclusion conductors interfaces have to be iproved, if a layer of electron-conducting aterial is introduced at the interface between the two phases. This interediate layer ust have high bond strength to the fillers and the atrix. Titaniu was identified as an appropriate candidate eleent to address the interfacial bonding between copper and carbon nanotubes. Titaniu can either react with the carbon to give titaniu-carbide or be dissolved within the copper atrix. It has to be considered, that any dissolved Ti in Cu significantly reduces the theral conductivity of the atrix. Here we present results on the influence of titaniu additions in copper/ultiwalled carbon nanotubes coposites on the evolution of theral conductivity. We further copare the aounts of CNTs and titaniu needed to confer good bonding within the Cu Ti/MWNT couple. EXPERIMENTAL PROCEDURE Multi-walled carbon nanotubes fabricated by a cheical vapour deposition process were used in stable aqueous dispersions (FutureCarbon GbH, Bayreuth, Gerany). Titaniu etallization was perfored by the addition of titaniu hydride TiH 2 powders to the CNT suspensions, drying and subsequent heat treatent under a constant flow of hydrogen gas (1 L in -1 ) at teperatures up to 1273 K. Those as-etallized CNTs were redispersed in an aqueous solution of 1M copper nitrate. An aqueous solution of 1M oxalic acid was slowly added to the copper nitrate Ti-etallized CNT suspension under vigorous stirring. A constant teperature of 353 K was adjusted during unification. Copper oxalate with incorporated CNTs is fored instantaneously. The precipitate was subsequently filtered off, dried and decoposed to etallic copper Ti/CNTs coposite powders under a constant flow of hydrogen gas (1.5 L in -1 ) at teperatures up to 773 K. An overall volue fraction of CNTs was adjusted between roughly 0.7 and 2.5 vol.%, titaniu content was varied between 0.2 and approxiately 3 vol.%. However, the coposition of the coposite powders and the consolidated coposite saples were controlled by C-analysis (induction furnace cobustion) and ICP-OES easureents respectively. The coposite powders were consolidated by die pressing, sintering under a constant hydrogen gas strea (1 L in -1 ) at teperatures up to 1173 K, followed by hot pressing in graphite dies at a axiu teperature of 1173 K and a axiu pressure of 42 MPa. The isotheral tie was set to 5 hours. Pure copper (99.99 wt.%), Cu Ti alloy and Cu TiC saples were prepared as reference aterials. Cu Ti alloys were produced by two different ethods: first by vacuu induction elting of the pure substances, second by the sae ethod described above for the preparation of Cu Ti/MWNT saples by decoposing Cu-oxalate/TiH 2 coposite precursors, and consolidated by hot pressing. Cu TiC reference aterials were produced by decoposing Cu-oxalate TiC ixtures under H 2 and hot pressed as well. The TiC powders used had a ean particle size of 2 µ. Consolidated saples were characterized by LOM, SEM and the condition of the MWNTs ebedded in the atrix was followed by FEG-SEM analysis.

3 Powder Metallurgy Progress, Vol.10 (2010), No Thero-physical properties were deterined on cuboid-shaped saples. In the case of CNT containing saples the direction of easureent was perpendicular to the direction of hot pressing (cast and hot pressed reference saples are assued to behave isotropically). The theral conductivity κ c/ was easured by a steady state, equal heat-flow coparative ethod against a brass reference. To this end, the teperature gradient in the saples was copared to that in the reference by eans of two Pt100 eleents in both saple and reference. The syste was calibrated with easureents on 5 N pure silver, aluiniu or copper reference saples. We evaluate the overall uncertainty of the easured theral conductivities to within ± 2% of the indicated values. RESULTS Two different series of Cu CNT coposites were fabricated: the first attributed the influence of the CNT concentration in Cu and Cu Ti atrix saples, the later realizing a constant Ti concentration of 0.25 vol.% Ti. The second series intended to study the influence of the Ti concentration at a constant CNT concentration (0.7 vol.% CNTs). This series was copared to atrix reference saples of Cu Ti and Cu TiC with increasing Ti and TiC content. Carbon analysis revealed soe divergences between desired and effective aounts of CNTs in the range of ± 5% (rel.) and ight be attributed to the appearance of aggloerates in the delivered CNT dispersions. However, the easured concentrations by C-analysis were subsequently used as effective CNT concentrations in the coposites and displayed in the graphs. The titaniu concentration in the saples was varied roughly between 0.2 and 3 vol.%. The density of the consolidated coposites is close to 100% for low CNT loaded saples. High volue fraction of CNTs in the coposites can cause porosities up to 5%. We generally note that the carbon nanotubes used in the experients are ultiwalled carbon nanotube MWNTs with outer diaeters up to about 50 n and have large but finite aspect ratios and are not straight. A theoretical density of 1.65 g c -3 was used for calculating the volue fraction fro the used weight fractions in experiental investigations. The FEG-SEM analysis of the atrix etched surface in a Cu Ti /CNT coposite is given in Fig.1. The evolution of the theral conductivity for Cu/CNT and Cu Ti/CNT coposites at a constant Ti concentration of 0.25 vol.%. vs. CNT-concentration and Cu TiC saples in Fig.2. Note that the Cu TiC graph is displayed against its C content and the CNT concentration is given as a carbon concentration. Figure 3 shows the evolution of the theral conductivity for Cu/Ti, and Cu TiC reference saples respectively, and Cu Ti/CNT coposites; the CNT concentration is constant at 0.7 vol.%; the shown Cu TiC line is displayed by its Ti content. Figure 4 displays the nueric evolution of the coposite theral conductivity following the DEM schee for a Cu TiC coposite, assuing h b = 4.2E 7 W -2 K -1, κ Cu = 405 W -1 K -1, κ TiC = 30 W -1 K -1. The 2D and 3D-odels of Hatta- Taya were applied to calculate the evolution of Cu/CNT coposites, assuing a atrix theral conductivity κ = 360 W -1 K -1, a longitudinal fibre conductivity κ i 33 = 2000 W -1 K - 1 and a transversal conductivity κ i 11 = 2.4 W -1 K -1. The theral conductivity data easured were noralized using a κ value of 360 W -1 K -1 for Cu Ti/CNT coposites and 405 W - 1 K -1 for Cu CNT coposites.

4 Powder Metallurgy Progress, Vol.10 (2010), No Fig.1. FEG-SEM iage of Cu Ti/CNT coposite (2 vol.% CNT, 0.25 vol.% Ti), atrix etched by hot nitric acid. Fig.2. Evolution of the theral conductivity for Cu/CNT and Cu Ti/CNT coposites; the Ti concentration is constant at 0.25 vol.%; the shown Cu TiC line is displayed by its calculated theoretical C content, the CNT concentration is given as carbon concentration. DISCUSSION The consolidated saples (surface etched by nitric acid) were analyzed by FEG- SEM (Fig.1) showing that the CNTs are still present and not totally consued during etallization processing, although bundles of CNTs are still visible in cases where no titaniu was added (not displayed). Sporadic individualized MWNTs can be observed sticking out of the copper atrix, whereas areas of uncovered MWNTs bundles are visible in other areas. Obviously the presence of an active eleent like titaniu enhances the deposition of copper on the priarily etallized CNTs and incorporates the nanotubes ore easily. The active eleents ight act as nuclei resulting in a preferred deposition of the atrix on those sites. Recent reports of Cho [6], Chu [7], Ki [8] (and ore recent of Yaanaka [9] on Ni CNT coposites), on the successful fabrication of CNT-reinforced Cu atrix coposites by eans of a particle-copositing ethod followed by spark plasa sintering (SPS), fed speculations about the contribution of CNTs to an increase in theral conductivity at low volue fractions. Their theral conductivities, however, showed either no enhanceent or quite liited contributions at volue fractions of up to 1.5 vol.% (Cho et al). Although they stated a axiu increase of roughly 2.5%, the enhanceents could be concealed by the uncertainty of the easured theral conductivities to within ± 5 10 W -1 K -1 of the indicated values, which is evident in laser flash easureents. At higher volue fractions, Cho et al. observed a decrease in theral conductivity which they attribute to the increasing difficulties in dispersing bundled CNTs with an increase in the CNT content. Other researchers reported siilar disappointing results for ceraic coposites (CMCs) or for MMCs with CNFs, see references in [6].

5 Powder Metallurgy Progress, Vol.10 (2010), No In addition to probles of dispersion, the interface between CNTs and etal atrix plays a crucial role for κ c, the theral conductivity in coposites, even when CNTs are well dispersed. It is well known that the interfacial theral resistance causes a size effect in the overall coposite theral conductivity, which akes it difficult to increase κ c when the size of the inclusions is below a certain liit, even if highly conductive fillers are used [10]. CNTs are clearly sall enough to create such an effect [6]. However, fro all above-entioned reports, the question whether CNTs can contribute to the overall theral conductivity in a coposite is still not fully answered, as they do not address the role of interfacial eleents. They are essential, as pure Cu and carbonaceous allotropes show no utual reactivity. With this contribution we wanted to clarify if the cobination of carbide foring eleents and CNTs could ever contribute to κ c the overall theral conductivity of coposites. On a first sight, Fig.2 indicates siilar results to other research groups: no contribution of the CNTs to κ c with increasing volue fractions; first the conductivity reains constant up to 0.5 vol.% CNTs and then slowly decreases with increased CNT content. Saples with titaniu additions show a siilar conductivity evolution up to roughly 1.7 vol.% CNTs, whereas increasing volue fractions of Ti etallized CNTs give a pronounced decrease in κ c. We first assue that this is a result of the atrix solid solution foration between copper and titaniu, responsible for a sharp decrease in κ, the atrix theral conductivity. A very sall increase (copared to a pure copper atrix) in κ c ight be interpreted at a very low volue fraction (up to 0.5 vol.%), but is ore or less concealed by the uncertainty of the easured theral conductivity values. Special ephasis was put on the preparation of reference saples, as Cu Ti/CNT saples should not be stringently rated against pure copper, as is often done. In this context it was also of interest to evaluate the possible foration of TiC in Cu Ti/CNT coposites and deterine their possible contribution to theral conductivity degradation in copper coposites. However, the atrix reference saples in Fig.3 show the expected TC evolution of copper with increasing titaniu content, in line with data reported elsewhere or siilar reported data [4], whereas a different behaviour is notable between olten saples and hot pressed ones. This can be attributed to a better dissolution of titaniu in the case of liquid processing, whereas by powder etallurgy the condition of the titaniu is iportant, as natural oxidation layers of powders used and oxidation during handling respectively ust be considered. Therefore the Ti content in solid solution is lower, thus resulting in an increase of theral conductivity copared to a fully dissolved Ti in Cu by liquid etal processing of bulk substances. Cu TiC saples generally exhibit a uch higher theral conductivity copared to all reference saples and the Cu Ti/CNT coposites. As TiC is not soluble in the Cu lattice the theral conductivity degradation is uch lower copared to the reference saples and thus ainly driven by the intrinsic theral conductivities of TiC (around 30 W -1 K -1 ) and Cu and their proportion. If we assue utual dissolubility between the TiC phase and Cu, the theral conductivity of such coposites can be calculated in the siplest for by the rule of ixture (ROM) or ore sophisticated by the DEM schee (see details and forulation in [10]). However, in both cases the theoretical theral conductivity of Cu TiC coposites should be uch higher than the experiental values reflect (see Fig.4).

6 Powder Metallurgy Progress, Vol.10 (2010), No Fig.3. Evolution of the theral conductivity for Cu Ti, and Cu TiC reference saples respectively, and Cu Ti/CNT coposites; the CNT concentration is constant at 0.7 vol.%; the shown Cu TiC line is displayed against its Ti content. Fig.4. Nueric evolution of the coposite theral conductivity for CNT coposites following the 2D and 3D-odels of Hatta-Taya according eq. (1) and (2), and the ROM and irom odels, assuing κ = 360 W -1 K -1, κ i 33 = 2000 W -1 K -1, κ i 11 = 2.4 W -1 K -1. A second graph is displayed for the ROM odel using κ i 33 = 6600 W -1 K -1. The DEM schee was applied for a Cu TiC coposite, assuing h bd = 4.2E 7 W -2 K -1, κ Cu = 405 W -1 K -1, κ TiC = 30 W -1 K -1 according forulations given in [10]. The easured theral conductivity data were noralized using a κ value of 360 W -1 K -1 for Cu Ti/CNT coposites and 405 W -1 K -1 for Cu CNT coposites. However, the rearkable feature of Fig.3 is that with increasing Ti concentration all Cu Ti/CNT coposites` saples (constant CNT concentration of 0.7 vol.%) possess a significant higher theral conductivity copared to atrix reference saples. Hence it is verified, that CNTs can positively contribute to the overall coposite theral conductivity, with the constraint that an active eleent like titaniu ust be present. Additionally, the

7 Powder Metallurgy Progress, Vol.10 (2010), No coparison of Cu Ti/CNT and Cu TiC coposites verifies that the CNTs are not fully consued and converted to TiC, as the conductivity of the CNT coposites is lower. Theoretical considerations Generally predictive schees for physical properties of two- or even ulti-phase aterials are an iportant tool for assessing the potential of, and then tailoring of coposite aterials for specific applications. There are several predictive schees described in literature, aong the ost powerful for predicting the theral conductivity of fibre coposites is the odel of Hatta-Taya [11] and which is based on the Eshelby ethod: here the elastic proble with stress (σ ij ), strain (ε ij ) and stiffness (C ijkl ) tensor correspond to the theral proble with heat flux (q i ), teperature gradient (T i ) and the theral conductivity (κ ij ) respectively. Following the Hatta-Taya odel, the coposite theral conductivity for a copletely rando distribution and a 2D (circular cylindrical) inclusion isorientation can be calculated by: fi( κi κ) 0.5( κi κ) + 2κ κc = κ 1+ (1) κ( κi κ)( 2 fi) + 2κ For the 3D orientation case the odel is defined by: fi( κ κi ) 0.5( κi κ) + 3κ κc = κ 1 (2) 33 2 Rκ( κi κ)( 2 fi) + 3κ R = 1.5 f (3) i Note that the shape tensor S ij - given in the original paper of Hatta & Taya - for a cylindrical fibre derives to S 11 = 0.5 and S 33 = 0 siplifying the equations in the way given above. The odel is also based on the assuption that the inclusions are copletely surrounded by the isotropic atrix; therefore it is only valid up to an interediate range of fibre volue fraction, f i 0.5. Other effects such as fibre non-straightness, theral interface conductance and inclusion size can be incorporated by odifying the effective inclusion theral conductivity. The liited heat transfer capability at the interface can be considered by the presence of a very thin interfacial barrier layer, described by the theral interface conductance h bd. This barrier layer reduces the equivalent theral conductivity of the fibres. The effects of the theral interface conductance h bd give rise to the well-known forulation by Hasselan and Johnson, which is described in detail elsewhere and not considered here. Calculations were perfored assuing a copper titaniu atrix coposite with 0.25 vol% Ti (κ = 360 W -1 K -1, an approxiated value, estiated by the experiental results in Fig.3) and a theoretical intrinsic CNT theral conductivity in longitudinal fibre direction of κ 33 i = 2000 W -1 K -1. κ 11 i, the radial fibre (CNT) theral conductivity was assued to be in the typical order of graphite fibres (2.4 W -1 K -1 ). An additional graph for the ROM odel is displayed using κ 33 i = 6600 W -1 K -1, a value often cited for single wall nanotube SWNTs. The DEM schee was applied to calculate the theoretical conductivity for a Cu TiC coposite; details of forulations are given in [10]. The calculations were perfored assuing h bd = 4.2 E 7 W -2 K -1, κ Cu = 405 W -1 K -1, κ TiC = 30 W -1 K -1. Note that the easured theral conductivity data displayed in Fig.4 were noralized using a κ value of 360 W -1 K -1 (Cu Ti/CNT coposites) and 405 W -1 K -1, respectively (Cu CNT coposites). Results presented in Fig.4 consider the different odels described by equations (1) (2) and by the ROM and irom odels to illustrate the influence

8 Powder Metallurgy Progress, Vol.10 (2010), No of MWNT additions on the overall coposite s conductivity and the odels. Both are copared with experiental results. In general the highest increase in conductivity with increasing nanotube content is predicted by the ROM odel, the least by irom, thus both represent the extree boundaries. Considering the steric orientation of the CNTs within the atrix, represented by the 2D and the 3D odels, results in a degradation of theral conductivity. The corresponding graphs are, as expected, between the ROM and irom odel. Coparing the theoretical calculations and the experiental results shows that the few coposites with very low CNT loadings fit the odels, especially the one with the highest conductivity, which perfectly narrows the ROM odel using κ i 33 = 6600 W -1 K -1. In this case we can iagine being quite close to the upperost boundary theoretically possible in such systes. Hence, we can describe the heat transfer capability at the atrix/fibre interfaces being optial, if low inclusion loadings and concoitantly additions of titaniu are fulfilled. Nevertheless, with increasing CNT content the conductivity contribution of the CNTs diinishes rapidly and all odel predictions fail. At high MWNT content the ipact of active eleents diinishes and results in the observed degradation. This can be attributed to the non-straight, entangled and tortuous nature of CVD as-produced MWNT fibres. Hence the necessary contact between fibres and atrix for heat and load transfer cannot be established, as the atrix cannot penetrate the fibre bundles thoroughly. This is also reflected by an increasing porosity of the saples with increasing MWNT concentration thus contributing to the contact probles between fibres and atrix. To rate the influence of larger titaniu additions, we have to face the possibility of severe reactions between the CNTs and the Ti, possibly fully consuing the fibres or the foration of undesirable thick TiC layers, acting as physical theral barrier to the heat flow across the interface due to the low intrinsic theral conductivity of TiC. In order to reduce and exclude excessive TiC foration, the aount of titaniu should be strictly liited. CONCLUSION Investigations presented here indicate that the presence of MWNTs in Cu Ti/CNT coposites can contribute to an increase in coposite theral conductivity. This finding is restricted to the additional presence of an active eleent like titaniu in the atrix. If no carbon affine eleents are added, the theral conductivity is not positively influenced. It ust be also clear, that the theral conductivity of a pure copper reference will be not (under any circustance) be exceeded. At high CNT contents the ipact of the presence of an active eleent diinishes and results in a pronounced degradation of thero-physical properties. This can be attributed to the non-straight, entangled and tortuous nature of CVD as-produced MWNT fibres. Hence the necessary contact between fibres and atrix for heat and load transfer cannot be established, as the atrix cannot penetrate the fibre bundles thoroughly. The sae effect is visible for increasing Ti content, as the coposite conductivity rapidly decreases due to the low intrinsic conductivities of Ti and TiC. FEG-SEM investigations were perfored at the university service institution for transission electron icroscopy USTEM within the Vienna University of Technology, Austria. REFERENCES [1] Bakshi, SR., Lahiri, D., Agarwal, A.: Int. Mater. Reviews, vol. 55, 2010, p. 41 [2] Baughan, RH., Zakhidov, AA., de Heer, WA.: Science, vol. 297, 2002, p. 787

9 Powder Metallurgy Progress, Vol.10 (2010), No [3] Zweben, C. In: Power Electronics Technology Conference and Exhibition. Baltiore, Maryland, USA, 2005, p. 10 [4] Lloyd, JC., Neubauer, E., Barcena, J., Clegg, WJ.: Cop. Sc. and Techn., vol. 70, 2010, p [5] Lloyd, JC.: PhD thesis. Cabridge : Univ. of Cabridge, p. [6] Cho, S., Kikuchi, K., Miyazaki, T., Takagi, K., Kawasaki, A., Tsukada, T.: Scripta Mater., vol. 63, 2010, p. 375 [7] Chu, K., Wu, Q., Jia, C., Liang, X., Nie, J., Tian, W., Gai, G., Guo, H.: Cop. Sc. and Techn., vol. 70, 2010, p. 298 [8] Ki, KT., Eckert, J., Liu, G., Park, JM., Li, BK., Hong, SH.: Scripta Mater., vol. 64, 2011, p. 181 [9] Yaanaka, S., Gonda, R., Kawasaki, A., Sakaoto, H., Mekuchi, Y., Kuno, M., Tsukada, T.: Mater. Trans., vol. 48, 2007, p [10] Tavangar, R., Molina, JM., Weber, L.: Scripta Mater., vol. 56, 2007, p. 357 [11] Hatta, H., Taya, M.: J of Applied Physics, vol. 58, 1985, p. 2478