Characterization of the mechanical and thermal interface of copper films on carbon substrates modified by boron based interlayers.

The manipulation of mechanical and thermal interfaces is essential for the design of modern composites. Amongst these are copper carbon composites which can exhibit excellent heat conductivities if the Cu/C interface is affected by a suitable interlayer to minimize the Thermal Contact Resistance (TCR) and to maximize the adhesion strength between Cu and C.In this paper we report on the effect of boron based interlayers on wetting, mechanical adhesion and on the TCR of Cu coatings deposited on glassy carbon substrates by magnetron sputtering. The interlayers were 5 nm thick and consisted of pure B and B with additions of the carbide forming metals Mo, Ti and Cr in the range of 5 at.% relative to B. The interlayers were deposited by RF magnetron sputtering from either a pure B target or from a composite target. The interlayer composition was checked by Auger Electron Spectroscopy and found to be homogenous within the whole film.The system C-substrate/interlayer/Cu coating was characterized in as deposited samples and samples heat treated for 30 min at 800 °C under High Vacuum (HV), which mimics typical hot pressing parameters during composite formation. Material transport during heat treatment was investigated by Secondary Ion Mass Spectroscopy (SIMS). The de-wetting and hole formation in the Cu coating upon heat treatment were studied by Scanning Electron Microscopy (SEM) and Atomic Force Microscopy (AFM). The adhesion of the Cu coating was evaluated by mechanical pull-off testing. The TCR was assessed by infrared photothermal radiometry (PTR). A correlation between the adhesion strength and the value of the TCR which was measured by PTR was determined for as deposited as well as for heat treated samples.

Introduction and motivation

Interfaces in modern high performance materials play an ever increasing role. Their properties influence mechanical, electrical and thermal characteristics to a high degree since they may constitute important adhesive zones or scattering centers for conduction carriers or phonons. In novel materials as e. g. superhard nanocomposites [1-3] they may even be the main features responsible for novel physical properties such as hardness values which exceed those of all involved constituents. Also Metal-Matrix-Composites (MMCs) represent a material class with a high interface fraction. By changing the compositional ratio of the constituents, mass density or electrical and thermal conductivity can be changed. In a first approximation simple rules of mixture may be employed for a prediction of the changes in properties, but these do not take into account interface effects.

MMCs may consist of materials with distinctly different physical and chemical properties. copper–carbon composites [4-6] are a good example for this material class. They have a high potential for an application as heat sinks for electronic components. By changing the carbon content in the Cu-Matrix it is possible to adjust the Coefficient of Thermal Expansion (CTE) to match the CTE of standard electronic materials such as silicon while retaining the high thermal conductivity of copper for heat removal. Of course this application requires a reliable, stable joining of the two components, Cu and C. C may be included into the Cu Matrix in the form of fibers [7-10] with an amorphous or a graphitic structure [11], as granulates or flakes [12,13], or by reactive deposition processes [14,15]. Also including C in the form of diamond granulate is under consideration since diamond would add its excellent thermal conductivity to the high thermal conductivity of Cu [16]. Even first industrial products based on copper diamond composites are available [17,18]. Nonetheless, to make use of the previously mentioned facts not only the mechanical interface between Cu and C has to be strengthened but also the thermal interface has to be modified in a way that the electronic heat conduction mechanism of Cu can be matched to the phononic heat transport mechanism of diamond [19-21]. A promising approach to achieve this is the application of thin interlayers which may serve as wetting and adhesion promoters for the stabilization of the mechanical interface between Cu and C [10,22-24] as well as promoters for the heat transfer across that interface. Boron and metal doped boron was considered in the present work since (i) B may be beneficial to match the phononic heat conduction of C because of its low mass and (ii) the metal may capture the electronic heat conduction mechanism. In addition, the carbide forming metals Mo, Ti and Cr were chosen to enhance the mechanical interface by carbide formation.

In a first approach the mechanical and thermal interface was characterized on plane substrates. In this work glassy carbon (Sigradur G) was chosen as substrate because larger samples are available, which is paramount for the characterization of the thermal interface.

The paper is structured as follows: Section 2 describes the experimental procedures which were used to prepare and characterize the samples. The experimental results are also presented there. In Section 3 the experimental results relating to the characterization of the Cu/C interface in respect to mechanical, chemical and thermal properties are presented. The final Section 4 gives an overview of yet unresolved points as well as an outlook on future work.

Experimental

The interlayer which was used for the modification of the C interface is based on boron which was either used as pure material or doped with Mo, Ti or Cr additions. It was intended to add an amount of 2–5 at.% metal to the boron to achieve a significant influence of the metallic component in the interlayer on the one hand without changing the role of B as the majority component of the interlayer on the other hand.

Pure boron was deposited by RF magnetron sputtering from a planar magnetron source (AJA ST20 in as delivered magnet configuration). The target diameter was 5 cm and depositions were carried out at room temperature (RT) at an Ar pressure of 0.5 Pa and a RF sputtering power of 100 W. The distance target/substrate was 6 cm. These deposition parameters yielded a deposition rate of 0.18 nm/s. The metals were added to the B films by using a composite target which basically consisted of small pieces of Mo, Ti and Cr placed onto the B target. The size of the metal pieces was calculated in a way that, considering the sputtering yields of the metals and of B taken from [25], a composition of the film in the range mentioned above should be achieved. The deposition parameters were the same as for pure B and the deposition rate of the composite film remained essentially unchanged. No signs of arcing or other process instabilities caused by the metal pieces were detected. To check the film composition layers of approx. 30 nm thickness were first deposited on natively oxidized Si wafers. The composition was investigated by dynamic Auger Electron Spectroscopy (AES) in a VG Microlab 310F system. The evaluation of the AES data yielded metal contents of about 3 at.% for all considered metal additions. The distribution of the metal was very homogenous within the layers as depth resolved AES showed. Therefore the use of the described composite target was deemed feasible for the production of the metal doped layers.

All further samples consisted of glassy carbon substrates (Sigradur G, [26]) with an area of 10 × 20 mm2 and a thickness of 2 mm. Onto these B and metal doped B interlayers were deposited. The interlayer thickness was chosen to be 5 nm to keep the total amount of interlayer material within the Cu/C system as low as possible. The deposition parameters for the interlayers were identical to those mentioned before. Finally, the interlayers were covered with Cu-coatings of 300–1500 nm thickness which were manufactured by DC magnetron sputtering at a Power of 200 W at RT at a rate of 20 nm/s at an Ar pressure of 0.5 Pa.

Substrate pre-treatment involved degreasing by ultrasonic cleaning in acetone and methanol followed by vapor phase cleaning in methanol vapor. The samples were stored at 100 °C until being inserted in the deposition chamber to guarantee a dry surface. Two samples could be mounted on a substrate holder and inserted in the deposition chamber via a load-lock system. In all cases the sputtering process started after a base pressure of 10−4 Pa was reached. The data of the deposition equipment and a summary of the deposition parameters are given in Table 1.

Table 1

Basic deposition parameters.

Sputter plant	ALCATEL SCM450, turbomolecular pumped
Base pressure	10−5 Pa
Working gas/pressure/measurement	Ar/0.4 Pa/Baratron gage
Substrate material	Glassy carbon, Sigradur G
RMS-roughness of C substrate	< 3 nm
Distance target/substrate interlayer	60 mm
Target diameter interlayer	50 mm
Deposition rate at the substrate interlayer	0.18 nm s−1
Layer thickness interlayer	5 nm
Deposition temperature interlayer	Room temperature
Distance target/substrate Cu	100 mm
Target diameter Cu	100 mm
Deposition rate at the substrate: Cu	1.5 nm s−1
Layer thickness Cu	300–1500 nm
Deposition temperature Cu	Room temperature

The samples were characterized in respect to their de-wetting behavior upon heat treatment by Scanning Electron Microscopy (SEM) with a FEI XL30 ESEM and contact mode Atomic Force Microscopy (AFM) performed by a TOPOMETRIX EXPLORER with Si3N4 tips with an opening angle of 50°. Mechanical adhesion of the Cu coatings on the various interlayers was evaluated by a custom built mechanical pull off tester. The effects of heat treatment on the depth resolved chemical composition were investigated by Time Of Flight Secondary Ion Mass Spectroscopy (TOF SIMS) using a TOF SIMS V instrument from ION TOF GmbH, Germany. Finally, the Thermal Contact Resistance (TCR) of the samples was determined by infrared photothermal radiometry (PTR) in modulated reflection mode at the University of Reims. The basics of this technique will briefly be described in Section 3.

Results and discussion

Wetting and interdiffusion

The first set of samples consisted of 300 nm Cu deposited on 5 nm B and metal doped B interlayers. Before subjection to thermal treatment the Cu coating was smooth, dense and featureless with a RMS roughness below 2 nm and grain sizes in the 20–50 nm region [27]. After heat treatment at 800 °C for 30 min under High Vacuum (HV), which mimics typical hot pressing temperatures and durations which may be involved in the formation of a metal matrix composite, the Cu coating recrystallized [27] and holes were formed in the coating which are an indication of de-wetting [28,29]. The number and size of the holes, however, depends on the kind of the interlayer used, as SEM micrographs in Fig. 1 show for a sample without interlayer (Fig. 1a) and a sample with a Ti doped B interlayer (Fig. 1b). The dark spots in the SEM micrographs represent holes in the Cu film. The degree of de-wetting can be quantified by determining the total area of the holes in relation to the whole image area. This can be achieved by thresholding the SEM images since the contrast between the Cu covered regions and the holes is reasonably high (see Fig. 1). After thresholding the ratio of the number of black pixels (value below threshold; holes) and of the total number of pixels is formed. This ratio is the total area of the holes in relation to the total image area, given in %. These data are displayed in Fig. 2. It is visible in Fig. 2 that the total hole area is significantly higher for the sample without an interlayer (around 3%). All samples containing an interlayer exhibit values around or below 1%. Regardless whether a doping metal is present in the boron layer or not, the hole density is reduced by a factor of 0.3 ± 0.1 when the hole density of the sample without interlayer is taken as a reference. The dataset presented here represents the samples exhibiting the lowest hole areas for different process parameters. It is extracted from [30] where data for other interlayer thicknesses and working gas compositions can be found.

Fig. 1

SEM micrographs of 300 nm Cu deposited on (a) Sigradur G without interlayer and (b) on a 5 nm Ti doped B interlayer. Both samples were heat treated at 800 °C under High Vacuum for 30 min. The formation of large holes (dark areas) in the Cu layer can clearly be observed in (a).

Fig. 2

Total hole area as determined by image analysis (see text) for the samples containing various types of interlayers. The sample without interlayer shows a significantly higher total hole area than the other ones.

Hole formation has also been verified by Atomic Force Microscopy as Fig. 3 shows. Fig. 3a shows the surface of a sample without interlayer, Fig. 3b a sample with a 5 nm Ti-doped B interlayer. It is visible from the topographic data that (i) holes form between the crystallites of the film and (ii) that the Cu film without interlayer recrystallizes into 3d crystallites while the Cu film on the B/Ti-interlayer forms much flatter, 2d like grains. This 2d grain growth is also typical for all other samples with an interlayer, regardless of the interlayer composition. If one determines the average RMS roughness in regions which contain no large holes or other significant irregularities (see rectangles indicated in Fig. 3a and b) then the average RMS value for Cu coatings directly deposited onto C is about 15 ± 2 nm while the one for the systems with interlayers amounts to 7 ± 2 nm. This is another indication for the 3d like recrystallization mode for samples without interlayer. The SEM data in combination with AFM data and the data on the total hole area therefore clearly suggest a wetting promoting property of the interlayers.

Fig. 3

Contact mode topographic AFM scans of 300 nm thick Cu-coatings on heat treated substrates (30 min, 800 °C, HV). (a) Sample without interlayer and (b) sample with 5 nm Ti doped B interlayer. Bright regions correspond to elevated sample positions. Three dimensional grain growth is clearly observable in (a) while crystallites in (b) essentially remain flat. Rectangles indicate typical areas where RMS values for all samples were determined. The main criterion of area selection is the absence of holes or other significant irregularities.

The chemical processes which are triggered by thermal treatment were investigated by TOF SIMS. One representative TOF SIMS depth profile is shown in Fig. 4 for an untempered sample (Fig. 4a) and a tempered sample (Fig. 4b) containing a 5 nm B/Ti interlayer. The apparent broadening of the interlayer in the untempered case is caused by the roughness evolution during the dynamic SIMS measurements. Despite this effect, after the thermal treatment it can be observed that B, Ti and C are diffusing into the Cu layer where the diffusion profile of Ti basically follows the one of B. Within the C substrate the Ti-profile does not change significantly while the B signal increases (see Fig. 4b with respective material redistribution indicated by arrows). A similar behavior was also observed for all other metal doped interlayers except the Cr doped ones where a significant enrichment of Cr was detected at the sample surface. The mechanism of this enrichment is related to the immobilization of Cr by oxidation at the sample surface and has been discussed elsewhere [31]. The pure B interlayer showed similar diffusion profiles as the one depicted in Fig. 4b, i. e. diffusion of B and C into the Cu coating as well as an increasing B-signal in the C substrate. This may be an indication of the formation of B4C, because in the vicinity of the interface the B signal is significantly higher than the C signal. The formation of Cu rich intermetallic compounds as e. g. Cu4Ti, however, seems unlikely because in no sample the Cu profile was correlated to the profile of the metallic dopant within the Cu rich region. In addition only Ti forms compounds with Cu while Mo and Cr are immiscible with Cu [32]. For the Cu poor region located within the C-substrate the Cu and Ti signals seem to be correlated after heat treatment, but the Ti-signal is significantly higher than the Cu signal. This might be an indication for the formation of Ti2Cu [32] within this region. For Mo and Cr no such behavior was observed. Also the formation of carbides with the metal dopants could not be confirmed by the SIMS data. For Cu coatings on C without any interlayer essentially no material transport after heat treatment was observed. Generally the SIMS investigations have shown that that boron can penetrate the C substrate, while the metal dopants preferably diffuse into the Cu layer. The diffusion range for each material can be estimated to be 100 nm, except for Cr which diffuses through the whole Cu film until the surface is reached.

Fig. 4

SIMS profiles through a sample containing a 5 nm Ti doped B interlayer (a) before heat treatment and (b) after heat treatment. Arrows indicate into which sample regions different materials diffuse preferably.

Mechanical properties

After establishing the positive effect of the interlayer in respect to the de-wetting behavior a second set of samples with a Cu-layer of 1500 nm thickness was produced to investigate the effect of the interlayer on mechanical adhesion by pull-off testing. Within one deposition run two samples could be manufactured, one of which was again thermally treated at 800 °C for 30 min under HV. Four identical sample sets were produced to gain reasonable statistics for pull-off testing. A cylindrical stainless steel stud was glued to the copper surface by 3 M Scotch Weld™ Brand adhesive tape. The adhesive joint was thermally activated at 100 °C for 60 min. The pull off forces for removing the Cu coating from the C surface bearing the interlayer are given in Fig. 5 for both, as deposited (light bars) and thermally treated (dark bars) films. The common feature in Fig. 5 is that thermal treatment decreases adhesion values in all cases which can be attributed to the observed recrystallization and de-wetting processes. In the case of thin Cu coatings hole formation is present for all samples, although to a different extent (Fig. 2). For the thick Cu coatings used for the adhesion measurements void formation at the Cu–C interface is to be expected in association with recrystallization as it was observed in [10]. These voids reduce the total contact area of the coating with the substrate thus leading to adhesion loss. Also the interdiffusion of the different elements may be associated with vacancy agglomeration due to the Kirkendall effect thus also contributing to the formation of voids at the coating–substrate interface. Nonetheless, after heat treatment in almost all cases the adhesion values of the samples containing interlayers are significantly higher (approximately a factor of 1.6) than for Cu directly deposited onto C. The only exception is the sample with the Ti doped interlayer which exhibits slightly lower adhesion values compared to the one without interlayer.

Fig. 5

Adhesion values for samples containing various types of interlayers before (light bars) and after (dark bars) heat treatment. Dark regions within the bars represent the error region. Heat treatment reduces adhesion strength in all cases.

Thermal properties

The thermal properties of the samples were investigated by PTR. The system uses a modulated laser beam to trigger thermal waves within a medium. The PTR system at the University of Reims is schematically displayed in Fig. 6 and consists of three main components: (i) a diode pumped solid state laser (SDL-532-300T, 300 mW, 1.7 mm beam diameter at 532 nm wavelength) modulated by an acousto-optical modulator (AASA MT80-A1, driver MODA80-D4-30) for the excitation of thermal waves, (ii) a liquid-nitrogen cooled HgCdTe infrared detector (Graseby Infrared HCT-100-B) and the two off-axis, gold coated paraboloidal mirrors to measure the surface temperature of the Cu-layer, and (iii) a lock-in amplifier (SRS 850 DSP) which is necessary to filter the small periodic signal amplifications from the large radiative background of the stationary sample temperature. The system is capable of generating thermal waves with frequencies from 0.1 Hz to 100 kHz. More details on the PTR method and set ups for frequency domain measurements may be found in [33-36].

Fig. 6

Schematic of the infrared radiometry setup used to determine the TCR.

For each sample the amplitude and phase of the PTR-signal were measured as a function of the modulation frequency f of the laser intensity and subsequently were normalized to the signal measured for a thick, polished Tantalum reference sample with known thermal parameters. Fig. 7 shows normalized phases of samples with and without bonding layers, before heat treatment. As can be seen the bonding layer has a significant influence on the dependence of the normalized phase on the frequency of the thermal wave. The measured data were analyzed in the frame of a two-layer model consisting of the thin Cu-layer and the Sigradur substrate suspended in air, with the TCR or its reciprocal, the thermal contact conductance G = 1/TCR as fit parameters. In this study, the effect of the bonding layer on the thermal wave propagation was included in the TCR.

Fig. 7

Dependence of the normalized phase on the frequency of the thermal wave. Symbols represent measured values, lines result from fitting the data by a 1-D model of thermal wave propagation in layered media (for details see text).

Calculations were made for one-dimensional (1-D) and 3-D heat propagation configurations [34-37]. As can be seen in Fig. 7, the 1-D model is accurate only for frequencies larger than 3–5 kHz. For lower frequencies, 3-D effects become dominant and they obscure the effect of TCR. Below 3 kHz the deviations can be fitted by a 3-D model, but this requires the knowledge of the sizes of the heating and detection spots and of the intensity distribution of the laser beam, without providing additional information on the TCR. However, we found that within the same sample series, the ratios of the complex PTR signals to the signal of one member in the series cancels the influence of 3-D effects. Thus the amplitudes and/or phases normalized in such a way can still be analyzed in the frame of the 1-D model only. Details of this procedure will be published elsewhere.

All theoretical curves were calculated using thermophysical data of Cu, Sigradur and air from literature. The thickness of the Sigradur substrate and of the Cu-layer was determined from experimental data regarding the sample preparation. Whereas the thermal parameters do not seriously affect the TCR, there is a strong influence of the Cu-layer thickness on the TCR value. Therefore, when fitting theoretical curves to the experimental data, the known Cu-layer thickness was allowed to vary within its margins of error.

For the determination of the TCR all interlayer types have been considered except the one containing Cr because of the massive oxidation of Cr which was transported to the surface of the Cu layer upon heat treatment [31]. In general the absolute values of the TCR were found to be very low for both sample types, thermally untreated and thermally treated. They are in the range from 10−6 to 10−7 Wm2 K−1. In combination with the low thermal conductivity of glassy carbon (thermal conductivity of Sigradur G at 300 K: 6.3 Wm−1 K−1) this makes it very hard to resolve absolute differences in TCR below 10−7 Wm2 K−1. Therefore only relative differences will be given in the following. The lowest TCR was found in the thermally untreated sample with the B/Mo interlayer. This sample was taken as a reference. Subsequently the phase differences for all samples in relation to the reference sample were determined and the relative TCR was calculated by fitting these data to a 1-D model of thermal waves in layered media. These relative values are displayed in Fig. 8 for thermally untreated and thermally treated samples. It is visible from Fig. 8 that thermal treatment increases the TCR except in the case of the B interlayer, where it is reduced in relation to the untreated sample. This increase in TCR can be interpreted in terms of void formation at the interface between the Cu coating and the C substrate. By this mechanism new interfaces and regions of low thermal conductivity (air filled cavities) are generated which increase the TCR because of a higher reflection or decreased transmission of thermal waves through the voids and vacancy agglomerations present at the interface. As can be seen from Fig. 8 after heat treatment the TCR of the samples with an interlayer is lowered by a factor of 0.4 ± 0.1 in relation to the sample without interlayer. This is in good agreement with the reduction in total hole area which was given by a factor of 0.3 ± 0.1 (see Fig. 2). This quantitative agreement is most probably fortuitous but shows the importance of the formation of pores, voids and other defects at interfaces for tuning the thermal properties of a composite material. In the case of the B interlayer the TCR of the heat treated sample is even reduced in comparison to its value before tempering. Here the effect of void formation may be overcome by the formation of a graded interface due to the diffusion of B into the C substrate which leads to a gradual shift in thermal properties.

Fig. 8

TCR values before (dark bars) and after (light bars) heat treatment. All values are normalized to the TCR of the sample with the Mo interlayer which was not heat treated.

Regarding the comparison of adhesion and the interface thermal barrier, we found that adhesion correlates with the reciprocal of TCR. Therefore, for the sake of clarity, it is the parameter thermal contact conductance, G = 1/TCR, that is shown in Fig. 9a and b. In Fig. 9a both, adhesion values as determined from the pull-off test and G values are normalized to the respective adhesion values of the sample with the B/Mo interlayer. It can be seen that samples with low adhesion also show a low G and vice versa. The situation after heat treatment is given in Fig. 9b where the adhesion and G are given in relation to their values before tempering. All samples which experience a significant loss in adhesion also show a decrease in the G relative to their values before heat treatment. In particular, G for the sample without interlayer decreases by a factor of 3.3. Only for pure B interlayers G is moderately increased by about 20% and adhesion still retains about 80% of its value before heat treatment.

Fig. 9

Comparison of adhesion values and G = 1/TCR (a) before heat treatment, G values normalized to the sample with the B/Mo interlayer and (b) after heat treatment, G values in relation to their respective quantity before heat treatment.

In connection with the data from SIMS, SEM and of the pull-off test the modification of the TCR (or G) by interlayers and temperature treatment may be caused by the following mechanisms:

in the case of the thermally untreated samples the interlayer may act as a matching element which couples the electronic heat conduction within Cu to the phononic heat conduction within C. In addition the promotion of adhesion by the interlayer also has a positive effect on the TCR as can be seen in Fig. 9a, where the highest relative adhesion value corresponds to the lowest TCR (highest G) in the case of B/Mo. Low adhesion values result in a high relative TCR (low G), as observed for samples without interlayer and with a pure B interlayer; and

in the case of the tempered samples the mechanisms of elemental interdiffusion (see SIMS data) and carbide formation due to the presence of boron also lower the TCR. De-wetting and void formation on the surface of the Cu coating or at the interface between Cu and C [10,29] will counteract these effects and lead to an increase in TCR. If the adhesion after heat treatment decreases significantly, the TCR increases (G decreases, Fig. 9a). This effect may also be supported by the metal doping of the interlayer because metal diffusing into the Cu layer could act as an impurity in Cu thus negatively influencing its thermal properties. Only in the case of the boron interlayer adhesion after tempering remains close to its initial value and the TCR after heat treatment is slightly lower (G slightly higher) than before heat treatment. As boron diffuses into the copper layer as well as into the carbon substrate a graded interface is formed which gradually matches the thermal properties of Cu and C to each other. A possible consequence for thermal treatment is therefore that it might be reasonable to reduce the treatment time to minimize the adhesion loss which is correlated to the reduction in G. As even the presence of an only 5 nm thick interlayer has a measurable effect on G (Fig. 9) a further reduction of the interlayer thickness might be worth considering to reduce the contamination of Cu by interlayer material diffusing into it.

Conclusion and outlook

The present work has shown the positive influence of thin interlayers on the mechanical and thermal properties of Cu films deposited on glassy carbon. Wetting, mechanical adhesion and TCR have been shown to be connected to each other and can be influenced by the choice of the interlayer composition.

A major conclusion from the present work is that even very thin interlayers can significantly influence the thermomechanical properties of Cu–C systems. This is of special importance in respect to carbon containing metal matrix composites where the addition of an interlayer which promotes the connection between the Cu matrix and C-fibers or diamond granulate can also be considered as the addition of an impurity to the system. As it is well known, impurities may have significant influence especially on the thermal and electrical properties of a given material [38], and so it is of great benefit that the interlayer thickness can be kept in the low nm range as the above data have shown.

It was also shown that, although the addition of carbide forming metals has beneficial effects on the thermomechanical properties of the present samples, the addition of the pure B interlayer yields the best results in respect to TCR and adhesion after temperature treatment (see Fig. 9b, where the sample containing the B interlayer shows a higher G value after temperature treatment when compared to the respective value before). This is of great importance in respect to application since it makes B a candidate material for thin interlayers in Cu–C composites. The addition of a carbide forming metal does not seem prerequisite for a significant improvement of the thermomechanical properties. Investigations on pure metallic interlayers which were previously done [39,40] show that a positive influence on the thermomechanical and wetting properties in similar systems often emerges only at considerably higher interlayer thicknesses.

The focus of future work will be the study of similar layer systems on diamond samples. Special attention will be paid to the fact whether the same trends in thermal properties can be transferred from glassy C to diamond surfaces. This is experimentally demanding since the thermal characterization of diamond is less straight forward due to (i) the transparency of the substrate both, in the VIS and IR, and (ii) the small sample sizes. From these data, finally, an optimum interlayer material can be chosen which allows the production of Cu–diamond composites with optimized thermal interfaces.