Characterisation of sputter deposited niobium and boron interlayer in the copper–diamond system

Abstract

In most metal matrix composites (MMCs) interfaces are decisive but hard to manipulate. Especially copper–carbon composites can exhibit excellent mechanical and thermal properties only if the Cu/C interface is modified by an optimised interlayer. Due to the excellent thermal conductivity and mechanical stability of diamond this form of carbon is preferred as reinforcement in heat sink materials (copper–diamond composite) which are often subjected to severe thermal and mechanical loads.

In the present case niobium and boron interlayers of various thicknesses were deposited on diamond and vitreous carbon substrates by magnetron sputter deposition. After the coverage of all samples by a copper film, a part of the samples was subjected to heat treatment for 30 min at 800 °C under high vacuum (HV) to simulate the thermal conditions during the production of the composite material by uniaxial hot pressing.

De-wetting during heat treatment leads to the formation of holes or humps in the Cu coating. This effect was investigated by scanning electron microscopy (SEM) and atomic force microscopy (AFM). A comparison of time-of-flight secondary ion mass spectroscopy (TOF SIMS) profiles of heat treated samples with those of as deposited ones showed the influence of interdiffusion during the heating process. Diffusion behaviour and chemical composition of the interface were also studied by cross sectional transmission electron microscopy (X-TEM) investigations using focused ion beam (FIB) cut samples. The thermal contact resistance (TCR) of the interface was calculated from results obtained from modulated infrared radiometry (IR). Thin interlayers suppressed de-wetting most effectively and consequently the TCR at the Cu–diamond interface was found to decrease. Therefore they are promising candidates for optimising the Cu–diamond interface.

Highlights

► The interface in the Cu–C system was modified by boron and niobium interlayers. ► Thermal treatment of the samples simulated the production process of Cu–diamond MMCs. ► The TCR of the interface was investigated by modulated infrared radiometry. ► Chemical- and diffusion processes were investigated by TEM of interlayer cross sections. ► Very thin interlayers improve the mechanical and thermal interfaces.

1. Introduction and motivation

The ever increasing requirements posed on industrial materials, such as tuneable electrical and mechanical properties, high wear resistivity and application in extreme environments combined with reliable operation led to the development of new composite materials in the past decades. One composite material class are the so called metal matrix composites (MMCs), which combine the properties of the reinforcement with those of the matrix material [1,2]. In the case of MMCs the material requirement of tuneable properties can be realised by changing the volume or mass fractions of the constituents.

The relevance of the interface between matrix and reinforcements increases, as the volume fraction of the reinforcement increases and its lateral dimensions decrease. In many cases the interface is an essential factor for increased performance. Mechanical properties as e.g. wear resistivity, may well surpass those of the individual components [3–5]. The composition of the interface is decisive for the adhesion between matrix and reinforcement which, in turn, influences the thermal conductivity, the scattering of phonons and electrons or the intrinsic forces within the composite material.

Due to the extreme characteristics of diamond regarding e. g. hardness or high thermal conductivity combined with a low coefficient of thermal expansion (CTE) [6–8], it is widely used as reinforcement in MMCs. The size and arrangement of diamonds depend on the required application as e.g. cutting or grinding tools. While in this case titanium, tungsten, stainless steel or Cu alloys are used as matrix material, the matrix in heat sink materials mostly consists of silver, aluminium or copper because of the higher thermal conductivity of the latter metals when compared to the former. An additional difference is the significantly higher volume fraction of diamond reinforcements in heat sink materials.

The high thermal conductivity of diamond and copper makes this combination very attractive for heat sink applications. In addition it is possible to tailor the CTE by changing the diamond content in the composite to match the CTE of the device to cool. There is a wide field of cooling applications ranging from CPUs, laser diodes, high brightness LEDs or high current electronics to cooling elements for the inner wall of beam lines as e.g. in the LHC at CERN or nuclear fusion plants like ITER.

The incorporation of diamond particles into a copper matrix faces the issue of the total de-wetting of Cu and C [9–11] as well as the complete thermodynamic immiscibility of these two materials [9]. For the application as effective heat sink material not only the mechanical adhesion has to be improved, but the thermal interface is even more decisive. The most promising way to match the electronic heat conduction mechanism of copper to the phononic heat transport mechanism of diamond and to improve the mechanical interface is the application of thin interlayers [12]. In the present case boron and niobium were investigated as interlayer materials. B is expected to match the phononic heat transport mechanism of diamond, as its mass is only slightly lower than the mass of carbon, and the electronic one of Cu because of its semiconducting properties. Nb as carbide forming metal was considered as a promising interface material to enhance the wetting properties, strengthen the mechanical interface and capture the electronic heat transport mechanism of Cu.

To model the surface of reinforcements in MMCs, diamond- and vitreous carbon substrates of plain geometry were coated by sputter deposition. Some previous experiments, which identified a thin interlayer as most effective in the case of boron were partially repeated to check reproducibility and to offer the possibility of comparison with previous works [13,14].

2. Experimental

In order to modify the Cu–diamond interface pure boron was deposited by RF magnetron sputtering and pure niobium by DC-magnetron sputtering from a planar magnetron source (AJA ST20). Synthetic single crystal diamonds, Sumicrystal Type ECO Ib < 100 > with a nitrogen content of 10–100 ppm and the dimensions 3.5 mm × 3.5 mm × 1.2 mm were used as substrate material [15]. To investigate if the microstructure of the carbon substrate plays a significant role, vitreous carbon substrates (Sigradur G [16,17]) were coated with identical interlayers. The lateral dimensions of the Sigradur substrates are 10 mm × 20 mm × 2 mm. In all cases the interlayer was covered by a copper top layer with a thickness of either 300 nm or 1300 nm, deposited by DC-magnetron sputtering as well. All deposition parameters are given in Table 1.

Table 1.

Sputter deposition parameters.

Parameter/material	B	Nb	Cu
Sputter mode	RF	DC	DC
Working gas/pressure	Ar/0.4 Pa
Base pressure	10− 4 Pa
Target diameter	50 mm	50 mm	100 mm
Sputtering power	100 W fw	100 W	200 W
Distance target-substrate	65 mm
Temperature	Room temperature

Interlayers with thicknesses ranging from 5 nm to 100 nm were investigated to determine the optimal amount of interlayer material to mechanically strengthen the interface as well as to match the different heat conduction mechanisms.

All synthetic diamond samples were coated without any prior surface treatment. Only short wiping with acetone removed residua of the production casing. Vitreous carbon samples were subjected to ultrasonic surface cleaning in acetone and ethanol before the deposition process. Additional vapour phase cleaning and storage at 100 °C before inserting the substrates into the vacuum vessel removed further adsorbates and guaranteed a dry, fat-free surface. After the deposition processes a part of the samples was exposed to a temperature of 800 °C under high vacuum (HV) conditions for 30 min to simulate the thermal conditions during the production of the MMC.

Two samples combining the same substrate material with the same interlayer material and thickness were coated in the same deposition run. In this way two completely identical sets of samples, each set consisting of two samples (one thermally treated and one untreated) could be produced. The first set was used for SEM-, AFM-, and TOF-SIMS investigations. The second set of samples was analysed by infrared radiometry (IR). For cross sectional TEM investigations another set was prepared under exactly the same deposition conditions (parameters given in Table 1). IR analysis of nominally equal samples (identical deposition conditions and thermal treatment) yielded comparable thermal contact resistances.

As a result of heat treatment de-wetting processes led to the formation of holes or humps, in the Cu coating, depending on the thickness of the Cu layer. This effect was characterised by SEM (FEI XL30) and AFM (Topometrix Explorer in contact mode with Si₃N₄ tips with an opening angle of 50°). Interdiffusion processes during heat treatment were analysed by comparing depth resolved TOF SIMS (TOF SIMS V equipment by ION TOF GmbH, Germany) profiles of heat treated samples with those of as deposited ones. The core characteristics of an adaptive heat sink MMC, the thermal interface properties, were characterised by calculating the TCR from data gained by modulated infrared radiometry (IR) operated in reflexion mode.

For TEM-studies either B or Nb interlayers with a thickness of 20 nm were deposited on diamond substrates and again covered by a 1 μm thick Cu coating. A thickness of 20 nm was chosen to be able to clearly distinguish between the two interfaces, C/interlayer and interlayer/Cu. A part of the samples was subjected to thermal treatment at 800 °C for 30 min and another part for 60 min at the same temperature. Inside the FIB chamber another protective Pt layer was deposited onto the 1 μm Cu top layer by electron beam- and ion beam induced deposition. Thin lamellae of the Cu layer, the interlayer and a part of the diamond substrate were prepared by a FEI Quanta 200 3D FIB equipment operated with Ga ions. After further thinning of the lamellae in an Ar ion mill and He-plasma cleaning the TEM investigation of these cross sections was done by a field emission gun (FEG)-TEM TECNAI F20 with a LaB₆ electrode operated at an extraction voltage of 200 kV in scanning TEM mode (STEM). To analyse the chemical composition and the diffusion characteristics electron energy loss spectroscopy (EELS) and energy dispersive X-ray spectroscopy (EDX) were used.

Spectral images were obtained by EELS performing line scans across the interlayer cross sections and the surrounding diamond and cu. Complete spectra are recorded at certain points along this line. At certain lateral positions corresponding to the diamond substrate, the interlayer and the two interface spectra were extracted from the three dimensional spectral images and compared to each other.

3. Results and discussion

3.1. De-wetting and void formation

Temperature treatment has a significant influence on the formation of holes, voids and bulges on both, the substrate coating and the coating ambient interface. As shown in previous works [13,14] heat treatment at 800 °C for 30 min under HV conditions results in recrystallisation and movement of grain boundaries [18,19]. The measurements on vitreous carbon substrates were repeated (new samples, same conditions) to check reproducibility. Holes are formed in the Cu coating in the case of 300 nm film thickness. In the case of thicker Cu films (approximately 1 μm) this de-wetting process does not primarily result in the formation of holes, but in the formation of bulges due to the recrystallisation of Cu [11,12]. Furthermore compressive stresses are generated by the different CTEs of Cu (16 ppm/K) and carbon in its different modifications (1–1.2 ppm/K) [6–8,11,12]. They lead to buckling instabilities which are a source of the formation of bulges within a coating. Bulge formation is associated with the formation of voids at the interface between substrate and coating as it was illustrated in [11,18] for vitreous C and Cu. A direct consequence of these temperature induced effects is a modification of the thermal contact resistance (TCR) between substrate and coating. Interlayers, in general, have an influence on the above phenomena and can, in the best case, significantly reduce them, thus reducing also the TCR.

To quantify the influence of B and Nb interlayers on de-wetting the method described in [13,14] for the quantification of the total hole area was employed. Vitreous C substrates with B and Nb interlayers of different thicknesses were coated with 300 nm thick Cu and subjected to heat treatment for 30 min at 800 °C for 30 min. The hole area in % of the total sample area for this series of samples is displayed in Fig. 1. This relative value is the most significant describing the effect of hole formation. It is clearly visible in this diagram that all interlayers reduce the total area of holes compared to the case without any interlayer by about 1.2–2.2%. This is a relative change of about a factor of 0.1–0.3 when compared to the sample without interlayer. A thin Nb interlayer (5 nm) reduces the area of holes by up to 1% compared with a thick interlayer (100 nm). Thick B interlayers reduce the number and area of holes more efficiently than thin ones do, but all results for B lie in a range of about 0.4%. Obviously the thickness of the interlayer is a more decisive parameter in the case of Nb than it is in the case of B. Due to that fact, further investigations concentrated on thin interlayers of both elements. In addition there are two technological arguments favouring thin interlayers, (i) being the low amount of deposited material so that the interlayer may be considered as a single interface rather than a layer bounded by two interfaces, and (ii) the significant lower usage of material and deposition time, which reduces the production costs. The former argument will be advantageous for the interpretation of thermal data obtained by infrared radiometry.

Fig. 1.

Hole area ratio of the sample surface calculated from SEM micrographs; 300 nm Cu on vitreous carbon substrates with B and Nb interlayers of different thicknesses; all samples heat treated for 30 min at 800 °C.

To check whether the above results can qualitatively be transferred to single crystalline diamond the topographical situation is illustrated in Fig. 2. Fig. 2(a) shows a SEM-micrograph of a synthetic diamond substrate coated with 1300 nm of Cu without any interlayer, while Fig. 2(b) shows an analogous sample with a 5 nm thick Nb interlayer. Although more distinct grain boundaries can be observed in the latter case the grains appear to be flatter than in the sample without interlayer, where 3D grain growth and bulging can clearly be observed. This is consistent with the results presented for Cu without interlayer on vitreous carbon substrates.

Fig. 2.

SEM micrographs with 1.3 μm Cu deposited on synthetic diamond substrates with (a) no interlayer and (b) a 5 nm Nb interlayer; both samples thermally treated for 30 min at 800 °C.

To quantify the topography of the 1300 nm thick Cu top layer on synthetic diamond substrate samples with different B and Nb interlayers were investigated by atomic force microscopy (AFM). Fig. 3 displays the R_a roughness values of the nsample surfaces. The most obvious fact is that all samples without thermal treatment are much smoother than those subjected to heat treatment at 800 °C for 30 min. This verifies that void and bulge formations are a direct consequence of grain boundary formation and movement during the heating process. In addition to that the AFM measurements show that all interlayers enhance the wetting properties of the Cu film, especially the 5 nm Nb interlayer, as all samples with interlayers exhibit a smoother Cu surface.

Fig. 3.

R_a roughness values calculated from contact mode topographic AFM scans of 1.3 μm Cu films on synthetic diamond substrates with various interlayers; comparison of heat treated samples (30 min @800 °C) with not heat treated ones.

3.2. Interdiffusion characteristics

Interdiffusion processes induced by the thermal treatment of the samples for 30 min at 800 °C were investigated by TOF SIMS measurements. One general result of these investigations is that B and Nb diffuse into the Cu coating in all cases, irrespective of the interlayer thickness. Fig. 4 shows a TOF SIMS depth profile of a 1300 nm Cu coating on a synthetic diamond substrate with a 5 nm B interlayer. Results for the sample without heat treatment are displayed in Fig. 4a and for the heat treated sample in Fig. 4b. The comparison of these two diagrams illustrates the diffusion of B into the Cu coating. Moreover the C signal in Fig. 4b follows the profile of the B signal, which might indicate a chemical reaction at the interface, i.e. the formation of boron carbide. The same effect was observed in [14] for vitreous carbon substrate, thus it indicates that the nature of the C-substrate is not of primary importance insofar as the chemical effects of heat treatment are concerned.

Fig. 4.

Comparison of TOF SIMS profiles of thermally untreated (a) and heat treated samples (b) (30 min at 800 °C) for 5 nm B interlayer on synthetic diamond substrates covered by 1.3 μm Cu, relevant regions are arrow-marked in both diagrams.

For Nb interlayers the interdiffusion characteristics are similar. In Fig. 5 the untreated (a) and the thermally treated (b) cases are compared for a 1300 nm Cu top layer on a synthetic diamond substrate with a 5 nm Nb interlayer. As it is visible in Fig. 5b Nb does not penetrate as far into the Cu layer as B does for the same heat treatment parameters (see Fig. 4b).

Fig. 5.

Comparison of TOF SIMS profiles of thermally untreated (a) and heat treated samples (b) (30 min at 800 °C) for 5 nm Nb interlayer on synthetic diamond substrates covered by 1.3 μm Cu, relevant regions are arrow-marked in both diagrams.

The dependence of the diffusion behaviour on the interlayer thickness is illustrated in Fig. 6 for the case of the Nb interlayer. Fig. 6 displays the comparison between TOF SIMS profiles of a 5 nm (a) and a 30 nm (b) Nb interlayer on diamond substrates covered by a 1300 nm Cu top layer. In both cases the samples were subjected to thermal treatment for 30 min at 800 °C. In the case of the 30 nm interlayer much more Nb diffuses into the Cu coating and consequently to the sample surface than in the case of the 5 nm Nb interlayer. This fact suggests that the higher surface roughness measured by AFM for higher Nb interlayer thicknesses (see Fig. 3) is caused by the diffusion of interlayer material to the sample surface.

Fig. 6.

TOF SIMS profiles of synthetic diamond substrates covered with 1.3 μm Cu; comparison of 5 nm Nb (a) with 30 nm Nb (b) interlayer; both samples thermally treated for 30 min at 800 °C, relevant regions are arrow-marked in both diagrams; panel a displayed only for comparability (identical to Fig. 5b).

The analysis of the difference in the element concentration before and after thermal treatment of the samples has shown that the main amount of the reallocated interlayer material diffuses towards the Cu surface and a smaller part diffuses into the substrate material. As the main amount of the interlayer material remains near the substrate/coating interface, this effect is another indicator for the carbidisation of B and Nb at the diamond–interlayer interface.

3.3. Thermal properties and infrared radiometry (IR)

One of the most important properties of the Cu–C system regarding the manufacturing of MMCs and consequently effective heat sink materials, the thermal interface, was characterised by infrared radiometry (IR). In the used IR measurement systems thermal waves are excited in the sample by a modulated laser beam. The thermal waves are partially reflected and transmitted at every interface. By varying the modulation frequency the depth of penetration can be varied to obtain data from different sample regions. The frequency dependent characteristics of the phase shift and the amplitude of the response signal (infrared radiation) can be fitted by theoretical models [20,21]. The thermal conductivity equation in combination with the relation of Wiedemann and Franz can only be solved for certain boundary conditions and parameters for planar waves (two dimensional model). If the interlayer thickness is neglected, which can be assumed for thin interlayers, the interlayer can be considered as interface modification and its properties can be interpreted as an effective TCR. The theoretical characteristics can be fit to the one of the measured amplitude to determine the TCR. A more detailed description of such systems can be found in [14,22–27].

A general result of the present measurements is that all values for the TCR are in the range of 10^− 6–10^− 8 m²K/W. In the case of diamond substrates the TCR values could be determined as an absolute value, whereas only relative values could be calculated for vitreous carbon substrates [14] due to systematic restrictions.

In Fig. 7 TCR values for synthetic diamond substrates coated with various interlayers and a 1300 nm Cu top layer are displayed. In all cases the TCR of the untreated samples is higher than that of the thermally treated ones, except in the case of the 5 nm B interlayer. In this case the values for the untreated and the thermally treated sample are identical. A similar result has been obtained for the 5 nm B interlayer on vitreous carbon substrates [14]. In addition to that this value is the same as for the heat treated 5 nm Nb interlayer (2 · 10^− 8 m²K/W). Moreover these cases are the only ones for which the value of the TCR falls below the result for the thermally treated sample with no interlayer. For the 5 nm Nb interlayer another interesting fact can be observed. The difference between the untreated and the thermally treated case is the highest of all samples. The TCR value of the thermally treated sample is only 5% of that of the untreated sample. In addition the TOF-SIMS measurements identified the 5 nm Nb interlayer as the one with the lowest amount of interlayer material reallocated after thermal treatment. As nearly the whole interlayer material stays at the interface one possible explanation for the high difference in the TCR is that nearly all interlayer material is fixed at the interface as niobium carbide forms by thermal treatment. Also the low contamination of Cu and C by Nb leaves the thermal characteristics of the constituents unchanged.

Fig. 7.

Thermal contact resistance of various interlayers on diamond substrates coated with 1.3 μm Cu for thermally treated and untreated samples.

These results verify the matching properties between the phononic (C) and electronic (Cu) heat conduction mechanism of thin B and Nb interlayers. For higher interlayer thicknesses material transport processes at the interface between Cu and C during thermal treatment, as e.g. diffusion of the interlayer material into Cu leads to an increase of the TCR.

3.4. STEM-study of interfacial properties

To investigate the chemical composition of the Cu–interlayer and the interlayer–diamond interface in more detail and to learn about the diffusion behaviour at these decisive positions STEM analyses of sample cross sections were performed. The samples consisted of diamond substrates, a 20 nm interlayer and a copper top layer. Two different thermal treatment times were chosen, 30 min and 60 min, both at 800 °C under high vacuum.

Fig. 8 shows images of cross sections of two thermally treated samples with B interlayer. In Fig. 8a diffusion zones are clearly visible at both interfaces (marked by red bars), formed during thermal treatment for 30 min at 800 °C. At the diamond–B interface the diffusion zone was found to have a thickness of 4.5 nm in both thermally treated cases for 30 and 60 min as well as for the untreated ones. EELS identified these diffusion zones to be boron carbide layers. At the B–Cu interface there is no interdiffusion in thermally untreated samples, whereas in both thermally treated cases the formation of a diffusion zone with a thickness of 4 nm could be observed. The difference between the thermally treated samples for 30 min and 60 min is the diffusion of interfacial material through the whole B interlayer in the latter case (Fig. 8b). This effect results in the formation of dark areas in the interlayer. EELS analysis identified these areas to be boron carbide diffusing from the diamond–B interface through the interlayer even into the B–Cu diffusion zone.

Fig. 8.

TEM micrographs of cross sections of B interlayer; (a): diffusion zones (red bars) visible after thermal treatment for 30 min at 800 °C; (b): interdiffusion of interface material through hole interlayer (dark areas) after heat treatment for 60 min at 800 °C.

The results of the EELS analysis for the B interlayer are illustrated in Fig. 9a,b,c. The comparison of the diamond peaks obtained from the diamond substrate with the ones recorded at the diamond–B interface indicates the formation of boron carbide due to the carbidisation of diamond at the interface in all three cases (thermally untreated (a), treated for 30 min (b), treated for 60 min (c)). The carbidisation is indicated by the transformation from a clear diamond peak to a carbon peak in general and especially by the appearance of the typical carbon pre-peak at the interface. The difference in fine structure when comparing the B peak at the diamond/B interface with the B peak obtained from the central part of the interlayer indicates the formation of boron carbide at the interface. In the case of thermal treatment for 60 min (Fig. 9c) the B peak at the central part of the interlayer and at the B–Cu interface shows the same modification in fine structure. This is due to the interdiffusion of boron carbide through the whole B layer.

Fig. 9.

EELS spectra recorded from B interlayer cross sections at the diamond substrate, the diamond-B interface, the B interlayer and the B–Cu interface of samples not subjected to thermal treatment (a), heat treated samples at 800 °C for 30 min (b) and thermally treated samples at 800 °C for 60 min (c).

For Nb interlayers, Fig. 10a displays the cross section of a sample subjected to thermal treatment for 30 min at 800 °C. The diffusion of Cu into the Nb interlayer is clearly visible in this image as the brighter part of the interlayer. No interdiffusion was found at the diamond–Nb interface for all samples. Also no clear EELS results could be obtained concerning the chemical composition of the interfaces because of (i) the poor EELS-yield of niobium and (ii) the delayed edge of Nb in the EELS-spectrum (see Fig. 10b). This delayed edge coincides with the carbon peak to a certain extent, as it is illustrated in Fig. 10b, in which the EELS spectra at the diamond–Nb interfaces for the three cases of thermal treatment are compared. There is a small chemical shift in the carbon peak and a difference in fine structure too, but as there are no observable differences in the delayed Nb edge, this is too few information to clearly verify the existence of niobium carbide.

Fig. 10.

TEM micrograph of the cross section of Nb interlayer (a) and resulting EELS spectra for the three different cases of thermal treatment recorded at the diamond–Nb interface (b). No significant difference in the spectra is visible.

Due to the poor EELS yield of Nb the diffusion behaviour at the Nb–Cu interface was characterised by EDX. In samples without any thermal treatment no interdiffusion was found as it is illustrated in Fig. 11a. Thermal treatment for 30 min at 800 °C results in the diffusion of Cu into the Nb interlayer and Nb into the Cu top layer, each about 5 nm deep, which leads to a diffusion zone with a width of about 10 nm (Fig. 11b). Fig. 11c illustrates the situation after 60 min thermal treatment at 800 °C. While Cu diffuses about 5 nm into the Nb interlayer, Nb diffuses 17 nm into the Cu top layer, which results in an interdiffusion zone with a width of about 22 nm which is the same thickness as the initial interlayer. Therefore prolonged thermal treatment might result in a contamination of the Cu matrix which may deteriorate the thermal conductivity of a potential Cu–diamond MMC.

Fig. 11.

EDX results of Nb interlayer cross sections recorded at the Nb–Cu interface for the three different cases of thermal treatment ((a): untreated, (b): 30 min at 800 °C, and (c): 60 min at 800 °C). Increase of diffusion lengths with the duration of heat treatment is clearly visible.

4. Conclusion and outlook

The present paper has shown that B and Nb interlayers can positively influence the thermal properties of interfaces in Cu–diamond systems. The surface of diamond particles, which are used as reinforcements in Cu matrix composites, was modelled by coating diamond substrates of plain geometry. After deposition of the interlayer and the Cu top layer the identification and characterisation of the interface is more effective using this model system as if diamond particles would have been used as substrate material.

In the case of thin interlayers it was possible to replace the interlayer by an effective thermal contact resistance (TCR) which makes the modelling of the system easier. The TCR could be quantitatively determined by IR radiometry and it was shown that thin interlayers can reduce the TCR by about one order of magnitude after heat treatment of the samples, when compared to pure Cu–diamond samples.

Chemical effects at the interface include the formation of carbides and the diffusion of the interlayer material into both, the Cu top layer as well as into the diamond substrate. Heat treatment also changes the morphology of the Cu top layer due to recrystallisation and de-wetting. For both, chemical and morphological modifications, trends observed for vitreous carbon substrates [14] could be transferred to diamond substrates.

In the case of B interlayers the formation of boron carbide even in samples which were not subjected to thermal treatment, found in samples with 20 nm B interlayer by EELS, is in perfect agreement with the results of the infrared radiometry measurements. This fact explains the exact equal values of the TCR for samples with 5 nm B interlayer before and after thermal treatment for 30 min at 800 °C. Furthermore the thickness of the boron carbide layer observed by STEM suggests the complete transformation of the 5 nm B interlayer to boron carbide in samples investigated by IR. Regarding the interdiffusion behaviour at both interfaces the results of the TOF SIMS measurements are consistent with the STEM results.

For Nb interlayers only the high TCR values of samples not subjected to thermal treatment could be explained satisfactorily by EELS as no niobium carbide could be found in these samples. The lower TCR after thermal treatment is obviously not caused by interdiffusion of Nb and C or the formation of niobium carbide. On the other hand the fact that no niobium carbide was observed does not exclude the possibility of formation of niobium carbide in the investigated samples due to the mentioned problems, as there are the poor EELS yield and the delayed edge of Nb in the EELS spectrum. The very large difference in the TCR between thermally treated and untreated samples indicates a transformation which cannot be explained just by the interdiffusion of Nb and Cu alone, which is observed by TOF SIMS. Analysing the difference in element concentrations before and after thermal treatment by TOF SIMS shows an interdiffusion of C and Nb. This indicates a possible chemical transformation at the diamond–Nb interface or, at least, the formation of a graded interface between Nb and diamond which can influence the transmission of phonons.

A very general conclusion of the present work is that very thin interlayers improve the thermal and mechanical interface properties of the Cu–C system even more significantly than interlayers with a higher thickness. This is especially relevant if the manufacturing process of diamond reinforced metal matrix composites is taken into consideration. The requirement of thin interlayers results in low deposition time and material consumption which consequently might reduce the production costs.

Future work will focus on the deposition and analysis of Nb and B interlayers on granular synthetic diamond. The main issue will be, whether experimental results from two-dimensional plane samples can be transferred to three-dimensional objects or not. Moreover the manufacturing process of copper–diamond MMCs offers many degrees of freedom (e.g. interlayer uniformity, determination of the composite density by variation of hot pressing parameters), which are influencing the mechanical and thermal properties for the final material. The optimum parameters combined with optimised interlayers have to be found to produce an effective high performance heat sink material.