- Email: [email protected]
Characterisation of the interface region in stepwise bias-graded layers of DLC films by a high-resolution depth profiling method
Thin Solid Films 482 (2005) 63 – 68 www.elsevier.com/locate/tsf
Characterisation of the interface region in stepwise bias-graded layers of DLC films by a high-resolution depth profiling method C. Ziebert*, C. Bauer, M. Stqber, S. Ulrich, H. Holleck Forschungszentrum Karlsruhe, Institut fu¨r Materialforschung I, Postfach 3640, 76021 Karlsruhe, Germany Available online 1 January 2005
Abstract The stepwise graded layer concept solves the problem of high internal compressive stresses in diamond-like carbon (DLC) films by adjusting a graded constitution by stepwise increasing the ion energy, i.e., the bias voltage, during sputter deposition. In order to optimize this concept, the detailed characterisation of the interface zones between the layers sputtered with different bias voltage plays a key role. The small-angle cross-section method (SACS) has been developed as a special nanoindentation (NI) technique to perform a high-resolution depth-profiling of the mechanical properties on the nanometer scale in multilayer or nanolaminated composite systems and especially to characterise their interfacial regions. Using improved area correction functions, by varying the maximum load, and by separating the instrumental broadening from the measured hardness profiles, it was possible to significantly improve the sensitivity and the resolution of the SACS. This allowed for the first time to investigate the dependence of the expansions of the interface zones between the graded layers on measuring and processing parameters. By using SACS, depth profiles of hardness and elastic modulus in dependence of applied load and layer thickness ratio have been measured. D 2004 Elsevier B.V. All rights reserved. Keywords: Diamond-like carbon (DLC); Stepwise bias-graded layer design; Interface characterisation; Depth-profiling; Mechanical properties
1. Introduction Due to their high hardness, chemical inertness and excellent tribological properties, amorphous carbon coatings with significant fractions of sp3 bonds, often called diamond-like carbon (DLC) coatings, are of great interest for technological applications such as wear-resistant coatings in magnetic storage disks [1,2] or protective coatings in the food industry [3], friction-reducing coatings in microelectromechanical systems [4], or biocompatible coatings [5,6]. In sputter deposition, a strong ion bombardment of the growing carbon film is applied in order to provide a strong densification of the amorphous carbon network [7,8], which is necessary to increase the fraction of diamond-like sp3 bonds. The sputter deposition mechanism differs from that, when the films are grown from energetic C+ ions using * Corresponding author. Tel.: +49 7247 82 2919; fax: +49 7247 82 4567. E-mail address: [email protected] (C. Ziebert). 0040-6090/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2004.11.117
filtered cathodic vacuum arc (FCVA) [9] or mass-selected ion-beam (MSIB) deposition method [10]. During sputtering, the Ar+ ions act only to bombard the growing DLC film and are not incorporated into the film in contrast to the C+ ions in FCVA and MSIB which are energetic enough to enter subsurface sites. The Ar+ ions displace C atoms into subsurface positions (knock-on subplantation), densify the surface, and provide conversion from sp2 to sp3 bonds by a high local compressive stress. Because the three different mechanisms have different process parameters, the sp3 content does not vary linearly with stress as in FCVA [9] but remains constant until a certain threshold is reached, and then increases rapidly [7]. Thus, the high internal compressive stresses in DLC films limits either their maximum thickness or their maximum sp3 bond fraction. The stepwise bias-graded layer design was shown to be successful for solving this problem [11,12]: in order to improve the adhesion and to manage the internal stresses, a graded constitution of the growing DLC films is adjusted by a
64
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
stepwise increase of the ion energy, i.e., the bias voltage, during deposition. While the ion bombardment provides a high density and a high hardness, the modification of the interface regions between the substrate and the graded layers ensures a good adhesion. By using the stepwise bias-graded layer design, it has been possible to deposit stable adherent DLC films with a maximum thickness of 10 Am and a maximum hardness of 5300 HV0.05. However, to optimize this concept, a depth profiling of the mechanical properties particularly at the interfaces between the layers deposited with different bias voltage is of great interest. Combining experimental investigations of the mechanical properties at the nanoscale with theoretical work on this field by MonteCarlo [13] or molecular dynamics simulation [14,15] will result in a deeper understanding and a better optimization of amorphous layer systems. Only recently, three methods for a depth-resolved measurement of hardness and Young’s modulus in the nanometer range were developed: the load variation method, the constant load method (CLM) and the cross section method. The basis of all methods is depth-sensitive nanoindentation (NI). Here, a diamond tip is pressed into the sample as an indenter and removed again after reaching a maximum load or depth, while the load F and the displacement or penetration depth h t of the indenter are recorded. The simplest method to obtain depth profiles is a stepwise increase in load, and thereby increasing the probed depth, during indentation in vertical direction to the surface. Therefore, the depth resolution available with this load variation method (LVM) decreases with increasing load [16]. In the constant load method (CLM), the layers are removed in a stepwise manner by ion sputter etching. Indentation takes place vertically to the surface at constant load. Resolution is somewhat better than for the LVM, as the measurements can always be performed at the same small load. On the other hand, this method is destructive and modifications of the specimen properties by ion bombardment have to be taken into account. In case of the cross-section method (CSM) developed by Kunert et al. for studying carbon implantation into Ti–6Al–4V steel [16], indentations take place at constant load along an appropriately prepared specimen cross section. In principle, this method provides the best depth resolution. However, its use for studying thin nanoscale films and in particular interface regions in stepwise bias-graded layers of DLC films is limited, as the minimum possible distance between two indentations required is 2 Am in a CSIRO UMIS2000 system and 100 nm in a HYSITRON TriboScopeR system. Therefore, the small-angle cross-section method (SACS) has been developed as a special nanoindentation technique, which allows to significantly increase the number of measurement points in the single layers and to use the highresolution capacity of scanning probe methods [17]. The SACS has been optimised by varying the maximum load, using improved area correction functions and subtracting
the instrumental broadening from the measured hardness profiles. Using the SACS on a UMIS2000 system, the mechanical properties of two different bias-graded DLC films were examined at the nanoscale. This allowed for the first time to investigate the dependence of the expansions of the interface zones between the graded layers on measuring and processing parameters. By using SACS, depth profiles of hardness and elastic modulus in dependence of applied load and layer thickness ratio have been measured.
2. Experimental 2.1. Deposition and characterisation of stepwise biasgraded DLC films The DLC films were deposited onto polished commercial M15 hard metal cutting inserts (88.5 wt.% WC, 11 wt.% Co, 0.5 wt.% Ta(Nb)C; 12124.5 mm) by nonreactive d.c. magnetron sputtering of a pure C target in an argon atmosphere. Prior to the deposition the recipient was evacuated to a residual pressure of less than 5104 Pa and the substrates were cleaned by ion etching at a r.f. substrate bias of 800 V for 15 min in argon. The constant deposition parameters applied were a d.c. sputtering power of 500 W, which corresponds to a power density of 11.32 W/cm2, an argon pressure of 0.6 Pa and a target–substrate distance of 5 cm. Following the graded layer concept developed by Stqber et al. [18], stepwise bias-graded DLC films composed of three layers with different properties were realized by increasing the ion energy, i.e., the substrate bias voltage U B in three steps as shown in the schematic presentation in Fig. 1. The deposition started with a bias voltage of 0 V to initiate a high adhesion of the growing film. Then, the bias voltage was increased stepwise to 150 V and finally to 300 V to produce a hard film surface. To investigate the influence of the thicknesses D i of the three graded layers on the expansion of the interface regions d 12 between 1. and 2. layer and d 23 between 2. and 3. layer, two different layer deposition time characteristics and thus thickness ratios D 1:D 2:D 3 of the three graded layers were adjusted. The two samples with reversed ratio of layer
Fig. 1. Bias-graded layer design of DLC films with the expansions of the interface regions d 12 between 1. and 2. layer and d 23 between 2. and 3. layer.
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
thicknesses will be called DLC I (D 1:D 2:D 3=1:2:3) and DLC II (D 1:D 2:D 3=3:2:1) in the following. The constitution and the microstructure of the films were analysed by X-ray diffraction, scanning electron microscopy and transmission electron spectroscopy, which showed that all films were amorphous. The total film thickness was determined by a Calotest to be 4.6 Am for DLC I and 4.9 Am for DLC II. The Vickers hardness, which was measured by a Fischerscope H-100, was 4100 HV0.05 for DLC I and 2700 HV0.05 for DLC II and the critical loads of failure were about 22 N for both samples. 2.2. Small-angle cross-section method In the small-angle cross section method (SACS) described in Ref. [17] for TiN/ZrN nanolaminated composite coatings, the area to be investigated by nanoindentation on the different layers is drastically enlarged by preparing a cross-section under a very small angle of about 0.028 to 0.18 as shown schematically for a much larger angle in Fig. 2(a) for a three-layered DLC film system. To prepare the specimens, the nanogrinding method developed by the Institute of Microtechnology (IMT) in Hanover is applied, which allows for roughness of few nanometers only [19]. Thus, the accuracy of the nanoindentation results is almost not affected by the remaining surface roughness. Due to the very small angle of the cross-sections, a sufficient number of indentations can be made in each single layer and even in the interface regions of the DLC films, which considerably increases resolution. Fig. 2(b) explains the geometrical parameters in SACS: D i with i=1, 2, 3 is the thickness of the three graded layers and X i with i=1, 2, 3 is the related distance on the sample surface, which is determined by the small cross section angle a. Using the simple geometric formula h j =x j sina, the distance x j passed by the nano-
Fig. 2. (a) Principle of small-angle cross-section method (SACS). (b) Geometrical parameters in SACS.
65
indenter, while it performs a nanoindentation linescan over the different layers, can be transformed into the depth information h j . However, as the real indenter tip geometry never follows the ideal relationship, calibration first had to be made on quartz glass and sapphire, especially for the small penetration depths needed to resolve the interface regions. In the range between 1 and 30 mN, 10 load–displacement curves were recorded for every load in 2 mN steps. They were averaged, the instrumental stiffness was determined and the area function for the Berkovich tip (three-sided pyramid with an enclosed angle of 142.38, tip radius of 150 nm) was fitted with an 8th order polynomial. To evaluate the load– displacement curves, a method similar to the Oliver–Pharr method [20] was employed. The use of the improved software Indentanalyser 1.5 (ASMEC), which considered the tip rounding, reduced the scatter of the data substantially and provided the values of hardness and Young’s modulus in accordance to ISO 14577 [21].
3. Results and discussion Fig. 3(a) shows the load–displacement curves of the hard metal substrate and the three bias-graded layers of a DLC film (DLC I: 0 V/150 V/300 V; thickness ratio D 1:D 2:D 3=1:2:3; a=0.098) recorded on a small-angle cross-section in the centre of the layers at 20 mN maximum load. None of the curves reveals any pop-in events or other features, e.g., due to the influence of the underlying layers. The smallest value of the maximum penetration depth h max was determined for the layer deposited with U B=300 V (1) to be 190 nm and the maximum value of 230 nm was found for the layer deposited without a bias voltage (4). Another disturbing factor might be the pile-up of material during indentation, which leads to an incorrect determination of the contact area. According to Bolshakov and Pharr [22], however, this pile-up effect only plays a major role, if the quotient of the maximum displacement h max and the plastic residual depth upon unload h p is larger than 0.7. Here, this quotient reached a maximum value of 0.6 in the softest layer; thus, the above condition was fulfilled. From a set of similar load–displacement curves recorded across a line over the different layers and the hard metal substrate (see Fig. 2(a)), the hardness depth profile of DLC I presented in Fig. 3(b) could be determined by the SACS. Starting from the left side (sample surface), the stepwise decrease of the hardness with increasing depth for the graded layers deposited with bias voltages of 300 (3. layer), 150 (2. layer), and 0 (1. layer) is clearly visible. Whereas the 3. layer sputtered with the highest bias voltage of 300 V has the highest hardness of 34.2 GPa, because the highest ion bombardment provided the highest densification of the carbon network, the 1. layer deposited without bias voltage shows a hardness value of 17.1 GPa, which is lower than the value of the hard metal substrate of 18.6 GPa.
66
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
up which might have been present at 20 mN maximum load, but can be completely neglected at 2 mN maximum load. However, even at a maximum load of 2 mN, another issue of the SACS has to be taken into account. When looking back to Fig. 2(b), at a distance s j from the interface between two graded layers on a small-angle cross-section, the effective layer thickness d j can be calculated as d j =s j sina. For example, for a=0.098 and s 1=20 Am, an effective layer thickness of d 1=31 nm only is obtained. In Fig. 4(b), the load–displacement curves of the 2. layer (A), the 1. layer (D) and the interface region (B and C) are shown, which indicate that at a maximum load of 2 mN, the maximum penetration depth in the layers and the interface region was between 40 and 50 nm, which is larger than the effective layer thickness. Thus, since the effective layer thickness decreases strongly due to geometrical reasons with decreasing distance to the interface, the indenter finally penetrates into the softer layer which is situated under the layer under investigation, when passing a certain distance. Consecutively, the calculated hardness value becomes too small, which leads to a broadening of the measured hardness profile. However, this broadening can be theoretically approximated by a weighting of the effective area fractions
Fig. 3. (a) Typical load–displacement curves of hard metal substrate and the three bias-graded layers of a DLC film (DLC I: 0 V/150 V/300 V; thickness ratio D 1:D 2:D 3=1:2:3; a=0.098) at a maximum load of 20 mN. (b) Depth profile of hardness of DLC I determined by SACS.
These values are in good agreement with those determined by microindentation or nanoindentation on three singlelayer DLC samples, which have been prepared by applying the same bias voltages of 0, 150, and 300 V. In Fig. 3(b), there are two large interfacial regions of gradual hardness decrease between the graded layers of 375 nm between 3. and 2. layer and of 485 nm between 2. and 1. layer. To investigate the influence of the applied load in the nanoindentation experiment on the apparent expansion of the interface regions, the maximum load has been reduced by a factor of 10 to 2 mN. Fig. 4(a) compares the hardness depth profiles at the interface region between the 1. and the 2. graded layer of DLC I recorded at 20 (-n-) and 2 mN (-w-). The large influence of the maximum load is obvious because reducing the load from 20 to 2 mN leads to a large subsequent decrease of the measured interface expansion from 485 to 38 nm, which is indicated by the shaded region. Both layers can still be discriminated very well by their average hardness values of 20.0 (1. layer) and 27.2 GPa (2. layer). The little increase in the measured hardness of the 1. layer in comparison to Fig. 3(b) might be due to a little pile-
Fig. 4. (a) Depth profile of hardness at the interface region between 2. layer (U B=150 V) and 1. layer (U B=0 V) of DLC I at a maximum load of 20 (-n-) and 2 mN (-w-), together with theoretical profile broadening (- -). (b) Load–displacement curves of 2. layer (A), 1. layer (D) and interface region (B and C) of DLC I.
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
67
of the indenter in both layers A 1 and A 2 and their averaged hardness values H 1 and H 2: Htheo ¼
A2 H2 þ A1 H1 A2 þ A1
ð1Þ
and it can be separated from the measured hardness profile. The calculated theoretical broadened profile indicated in Fig. 4(a) by the bold dashed line, gives a value for the theoretical broadening of 24 nm, which can be subtracted from the measured expansion of 38 nm giving a residual expansion of the interface between the 1. and the 2. graded layer of DLC I of V14 nm. A further reduce of the maximum load to 1 mN did not give any improvements because with the corresponding shallow penetration depth it was not possible to clearly discriminate between the graded layers due to a large scatter in the measured data. By using the SACS, not only the hardness depth profile but also depth profiles of other mechanical properties can be determined. As an example, the Young’s modulus depth profile at the interface region between the 3. layer (U B=300 V) and the 2. layer (U B=150 V) of DLC I at a maximum load of 2 mN is shown in Fig. 5. The profile reveals that the interface expansion value of 98 nm deduced from the Young’s modulus data without correction of the theoretical broadening is much larger than the value of 24 nm taken from the hardness data. A possible reason could be that the elastic stress fields governing the Young’s modulus have a far longer interaction range than the plastic deformation which determines the hardness value. Thus, the elastic stress fields under the indenter tip reach far into the underlying layer when coming close to the interface between two graded layers. However, because the theoretical broadening is necessary in order to accomplish a good comparison, new models will have to be developed to determine the corrected interface expansion for the Young’s modulus.
Fig. 5. Depth profile of Young’s modulus at the interface region between 3. layer (U B=300 V) and 2. layer (U B=150 V) of DLC I at a maximum load of 2 mN.
Fig. 6. Comparison of depth profiles of hardness of two DLC films with different thickness ratios at a maximum load of 5 mN: (a) DLC I and (b) DLC II.
To investigate the influence of film processing parameters the hardness depth profile of another sample DLC II (D 1:D 2:D 3=3:2:1; a=0.048) with a different ratio of the layer thicknesses, was recorded by SACS. In Fig. 6, the hardness depth profiles at a maximum load of 5 mN of the two DLC films with different thickness ratio are compared. A load of 5 mN was chosen because the differences between both films are visible better across the complete profile than at 2 mN load. The macroscopic hardness values were largely different because of the large penetration depth of more than 1 Am. Because in sample DLC II the hardest layer was only 0.82 Am thick, the microindenter already penetrated into the softer 2. layer leading to a measured Vickers hardness of 2700 HV0.05 in comparison to 4100 HV0.05 for the sample DLC I with a three-times thicker upper layer. However, the comparison of their hardness depth profiles in Fig. 6 clearly proves that the hardness of the three individual graded layers is the same in both films. The film DLC I with a thickness ratio of D 1:D 2:D 3=1:2:3 showed interface
68
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
the possibility to form sp2 or sp3 bonds makes the situation even more complex in amorphous carbon layers. Therefore, both other systems should also be investigated in detail before future studies will focus on the stress-state and the deformation and densification of amorphous carbon networks. A combination of experiments and theoretical simulations will be needed to get new insights into the subplantation model, to develop new models for the behaviour of mechanical layer properties across an interface and to find the optimal expansion of the interface regions to achieve optimum adhesion and maximum thickness of DLC films.
Acknowledgement Fig. 7. Comparison of thickness of interface regions d 12 and d 23 of two DLC films with different thickness ratio determined for a 2 mN load: (a) DLC I and (b) DLC II.
expansions d 12=180 nm and d 23=140 nm, which are indicated by the shaded regions, while the film DLC II with a reversed ratio of D 1:D 2:D 3=3:2:1 showed a larger value d 12=290 nm and a smaller value d 23=110 nm. Fig. 7 compares the corresponding expansions of interface regions for the two films at 2 mN load. On the one hand, there is a strong influence of the ratio D 1:D 2, which leads to d 12=14 nm when the thickness of the 2. layer is twice the thickness of the 1. layer, and to an eight-times larger interface expansion of 111 nm, when the thickness of the 2. layer is 3/4 of the thickness of the 1. layer. This shows that the thicker soft 1. layer forms a larger interface to the 2. layer, which might improve the adhesion of the system but leads to a macroscopic hardness decrease. On the other hand, almost no difference was found in d 23 with 24 nm or 22 nm when the thickness of the 3. layer is 3/2 of the thickness of the 2. layer or 1/2 of the thickness of the 2. layer, respectively.
4. Conclusions For amorphous stepwise bias-graded DLC films, it has been shown that the small-angle cross-section nanoindentation method (SACS) is a versatile tool for high-resolution depth-profiling of mechanical properties. It was revealed that the hardness values of the individual graded layers are independent on the thickness ratio of the graded layers, while the expansion of the interface regions between them showed a strong influence on this ratio and on the maximum applied load. However, it should be stressed that the in the case of amorphous carbon layers many different issues have to be taken into account. In comparison to crystalline nanolaminated coatings, such as TiN/ZrN multilayers, it is not possible to investigate the lattice parameters at the interfaces, e.g., by TEM, and in comparison to simpler amorphous materials such as B4C
The authors would like to thank C. Kourouklis of IMT Hanover for the preparation of nanogrinded small-angle cross-sections.
References [1] B. Bhushan, Diamond Relat. Mater. 8 (1999) 1985. [2] J. Robertson, Thin Solid Films 383 (2000) 81. [3] H. Schulz, H.J. Scheibe, P. Siemroth, B. Schultrich, Appl. Phys., A 78 (2004) 675. [4] D.F. Wang, K. Kato, N. Umehara, Surf. Coat. Technol. 123 (2000) 177. [5] R. Lappalainen, H. Heinonen, A. Antla, S. Santavirta, Diamond Relat. Mater. 7 (1998) 482. [6] M.I. Jones, I.R. McColl, D.M. Grant, K.G. Parker, T.L. Parker, Diamond Relat. Mater. 8 (1999) 457. [7] J. Schwan, S. Ulrich, H. Roth, H. Ehrhardt, S.R.P. Silva, J. Robertson, R. Samienski, R. Brenn, J. Appl. Phys. 79 (1996) 1416. [8] S. Logothetidis, M. Gioti, P. Patsalas, C. Charitidis, Carbon 37 (1999) 765. [9] D.R. McKenzie, D. Muller, B.A. Pailthorpe, Phys. Rev. Lett. 67 (1991) 773. [10] Y. Lifshitz, S.R. Kasi, J.W. Rabelais, W. Eckstein, Phys. Rev., B 41 (1990) 10486. [11] H. Holleck, M. Stqber, Method of manufacturing a composite material structure, US Patent No. 6, 110, 329 (2000); EU Patent No. EP 0912774B1 (2002). [12] C. Bauer, H. Leiste, M. Stqber, S. Ulrich, H. Holleck, Diamond Relat. Mater. 11 (2002) 1139. [13] C. Mathioudakis, P.C. Kelires, Y. Panagiotatos, P. Patsalas, C. Charitidis, S. Logothe- tidis, Phys. Rev., B 65 (2002) 205203. [14] A.Yu. Belov, H.U. J7ger, Comput. Mater. Sci. 24 (2002) 154. [15] S.B. Sinnott, R.J. Colton, C.T. White, O.A. Shenderova, D.W. Brenner, J.A. Harrison, J. Vac. Sci. Technol., A 15 (1997) 936 (Part 1). [16] M. Kunert, B. Baretzky, S.P. Baker, E.J. Mittemeijer, Metall. Mater. Trans., A 32 (2001) 1201. [17] S. Ulrich, C. Ziebert, M. Stqber, P. Schlogmacher, H. Holleck, E. Schweitzer, M. Gfken, Surf. Coat. Technol. 188 – 189 (2004) 331. [18] M. Stqber, S. Ulrich, H. Leiste, A. Kratzsch, H. Holleck, Surf. Coat. Technol. 116-119 (1999) 591. [19] H.H. Gatzen, J.C. Maetzig, Precis. Eng. 21 (1997) 134. [20] W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (1992) 1564. [21] ISO/FDIS 14577-1:2002; Metallic materials—Instrumented indentation test for hardness and materials parameters. [22] A. Bolshakov, G.M. Pharr, J. Mater. Res. 13 (1998) 1049.
Characterisation of the interface region in stepwise bias-graded layers of DLC films by a high-resolution depth profiling method C. Ziebert*, C. Bauer, M. Stqber, S. Ulrich, H. Holleck Forschungszentrum Karlsruhe, Institut fu¨r Materialforschung I, Postfach 3640, 76021 Karlsruhe, Germany Available online 1 January 2005
Abstract The stepwise graded layer concept solves the problem of high internal compressive stresses in diamond-like carbon (DLC) films by adjusting a graded constitution by stepwise increasing the ion energy, i.e., the bias voltage, during sputter deposition. In order to optimize this concept, the detailed characterisation of the interface zones between the layers sputtered with different bias voltage plays a key role. The small-angle cross-section method (SACS) has been developed as a special nanoindentation (NI) technique to perform a high-resolution depth-profiling of the mechanical properties on the nanometer scale in multilayer or nanolaminated composite systems and especially to characterise their interfacial regions. Using improved area correction functions, by varying the maximum load, and by separating the instrumental broadening from the measured hardness profiles, it was possible to significantly improve the sensitivity and the resolution of the SACS. This allowed for the first time to investigate the dependence of the expansions of the interface zones between the graded layers on measuring and processing parameters. By using SACS, depth profiles of hardness and elastic modulus in dependence of applied load and layer thickness ratio have been measured. D 2004 Elsevier B.V. All rights reserved. Keywords: Diamond-like carbon (DLC); Stepwise bias-graded layer design; Interface characterisation; Depth-profiling; Mechanical properties
1. Introduction Due to their high hardness, chemical inertness and excellent tribological properties, amorphous carbon coatings with significant fractions of sp3 bonds, often called diamond-like carbon (DLC) coatings, are of great interest for technological applications such as wear-resistant coatings in magnetic storage disks [1,2] or protective coatings in the food industry [3], friction-reducing coatings in microelectromechanical systems [4], or biocompatible coatings [5,6]. In sputter deposition, a strong ion bombardment of the growing carbon film is applied in order to provide a strong densification of the amorphous carbon network [7,8], which is necessary to increase the fraction of diamond-like sp3 bonds. The sputter deposition mechanism differs from that, when the films are grown from energetic C+ ions using * Corresponding author. Tel.: +49 7247 82 2919; fax: +49 7247 82 4567. E-mail address: [email protected] (C. Ziebert). 0040-6090/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2004.11.117
filtered cathodic vacuum arc (FCVA) [9] or mass-selected ion-beam (MSIB) deposition method [10]. During sputtering, the Ar+ ions act only to bombard the growing DLC film and are not incorporated into the film in contrast to the C+ ions in FCVA and MSIB which are energetic enough to enter subsurface sites. The Ar+ ions displace C atoms into subsurface positions (knock-on subplantation), densify the surface, and provide conversion from sp2 to sp3 bonds by a high local compressive stress. Because the three different mechanisms have different process parameters, the sp3 content does not vary linearly with stress as in FCVA [9] but remains constant until a certain threshold is reached, and then increases rapidly [7]. Thus, the high internal compressive stresses in DLC films limits either their maximum thickness or their maximum sp3 bond fraction. The stepwise bias-graded layer design was shown to be successful for solving this problem [11,12]: in order to improve the adhesion and to manage the internal stresses, a graded constitution of the growing DLC films is adjusted by a
64
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
stepwise increase of the ion energy, i.e., the bias voltage, during deposition. While the ion bombardment provides a high density and a high hardness, the modification of the interface regions between the substrate and the graded layers ensures a good adhesion. By using the stepwise bias-graded layer design, it has been possible to deposit stable adherent DLC films with a maximum thickness of 10 Am and a maximum hardness of 5300 HV0.05. However, to optimize this concept, a depth profiling of the mechanical properties particularly at the interfaces between the layers deposited with different bias voltage is of great interest. Combining experimental investigations of the mechanical properties at the nanoscale with theoretical work on this field by MonteCarlo [13] or molecular dynamics simulation [14,15] will result in a deeper understanding and a better optimization of amorphous layer systems. Only recently, three methods for a depth-resolved measurement of hardness and Young’s modulus in the nanometer range were developed: the load variation method, the constant load method (CLM) and the cross section method. The basis of all methods is depth-sensitive nanoindentation (NI). Here, a diamond tip is pressed into the sample as an indenter and removed again after reaching a maximum load or depth, while the load F and the displacement or penetration depth h t of the indenter are recorded. The simplest method to obtain depth profiles is a stepwise increase in load, and thereby increasing the probed depth, during indentation in vertical direction to the surface. Therefore, the depth resolution available with this load variation method (LVM) decreases with increasing load [16]. In the constant load method (CLM), the layers are removed in a stepwise manner by ion sputter etching. Indentation takes place vertically to the surface at constant load. Resolution is somewhat better than for the LVM, as the measurements can always be performed at the same small load. On the other hand, this method is destructive and modifications of the specimen properties by ion bombardment have to be taken into account. In case of the cross-section method (CSM) developed by Kunert et al. for studying carbon implantation into Ti–6Al–4V steel [16], indentations take place at constant load along an appropriately prepared specimen cross section. In principle, this method provides the best depth resolution. However, its use for studying thin nanoscale films and in particular interface regions in stepwise bias-graded layers of DLC films is limited, as the minimum possible distance between two indentations required is 2 Am in a CSIRO UMIS2000 system and 100 nm in a HYSITRON TriboScopeR system. Therefore, the small-angle cross-section method (SACS) has been developed as a special nanoindentation technique, which allows to significantly increase the number of measurement points in the single layers and to use the highresolution capacity of scanning probe methods [17]. The SACS has been optimised by varying the maximum load, using improved area correction functions and subtracting
the instrumental broadening from the measured hardness profiles. Using the SACS on a UMIS2000 system, the mechanical properties of two different bias-graded DLC films were examined at the nanoscale. This allowed for the first time to investigate the dependence of the expansions of the interface zones between the graded layers on measuring and processing parameters. By using SACS, depth profiles of hardness and elastic modulus in dependence of applied load and layer thickness ratio have been measured.
2. Experimental 2.1. Deposition and characterisation of stepwise biasgraded DLC films The DLC films were deposited onto polished commercial M15 hard metal cutting inserts (88.5 wt.% WC, 11 wt.% Co, 0.5 wt.% Ta(Nb)C; 12124.5 mm) by nonreactive d.c. magnetron sputtering of a pure C target in an argon atmosphere. Prior to the deposition the recipient was evacuated to a residual pressure of less than 5104 Pa and the substrates were cleaned by ion etching at a r.f. substrate bias of 800 V for 15 min in argon. The constant deposition parameters applied were a d.c. sputtering power of 500 W, which corresponds to a power density of 11.32 W/cm2, an argon pressure of 0.6 Pa and a target–substrate distance of 5 cm. Following the graded layer concept developed by Stqber et al. [18], stepwise bias-graded DLC films composed of three layers with different properties were realized by increasing the ion energy, i.e., the substrate bias voltage U B in three steps as shown in the schematic presentation in Fig. 1. The deposition started with a bias voltage of 0 V to initiate a high adhesion of the growing film. Then, the bias voltage was increased stepwise to 150 V and finally to 300 V to produce a hard film surface. To investigate the influence of the thicknesses D i of the three graded layers on the expansion of the interface regions d 12 between 1. and 2. layer and d 23 between 2. and 3. layer, two different layer deposition time characteristics and thus thickness ratios D 1:D 2:D 3 of the three graded layers were adjusted. The two samples with reversed ratio of layer
Fig. 1. Bias-graded layer design of DLC films with the expansions of the interface regions d 12 between 1. and 2. layer and d 23 between 2. and 3. layer.
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
thicknesses will be called DLC I (D 1:D 2:D 3=1:2:3) and DLC II (D 1:D 2:D 3=3:2:1) in the following. The constitution and the microstructure of the films were analysed by X-ray diffraction, scanning electron microscopy and transmission electron spectroscopy, which showed that all films were amorphous. The total film thickness was determined by a Calotest to be 4.6 Am for DLC I and 4.9 Am for DLC II. The Vickers hardness, which was measured by a Fischerscope H-100, was 4100 HV0.05 for DLC I and 2700 HV0.05 for DLC II and the critical loads of failure were about 22 N for both samples. 2.2. Small-angle cross-section method In the small-angle cross section method (SACS) described in Ref. [17] for TiN/ZrN nanolaminated composite coatings, the area to be investigated by nanoindentation on the different layers is drastically enlarged by preparing a cross-section under a very small angle of about 0.028 to 0.18 as shown schematically for a much larger angle in Fig. 2(a) for a three-layered DLC film system. To prepare the specimens, the nanogrinding method developed by the Institute of Microtechnology (IMT) in Hanover is applied, which allows for roughness of few nanometers only [19]. Thus, the accuracy of the nanoindentation results is almost not affected by the remaining surface roughness. Due to the very small angle of the cross-sections, a sufficient number of indentations can be made in each single layer and even in the interface regions of the DLC films, which considerably increases resolution. Fig. 2(b) explains the geometrical parameters in SACS: D i with i=1, 2, 3 is the thickness of the three graded layers and X i with i=1, 2, 3 is the related distance on the sample surface, which is determined by the small cross section angle a. Using the simple geometric formula h j =x j sina, the distance x j passed by the nano-
Fig. 2. (a) Principle of small-angle cross-section method (SACS). (b) Geometrical parameters in SACS.
65
indenter, while it performs a nanoindentation linescan over the different layers, can be transformed into the depth information h j . However, as the real indenter tip geometry never follows the ideal relationship, calibration first had to be made on quartz glass and sapphire, especially for the small penetration depths needed to resolve the interface regions. In the range between 1 and 30 mN, 10 load–displacement curves were recorded for every load in 2 mN steps. They were averaged, the instrumental stiffness was determined and the area function for the Berkovich tip (three-sided pyramid with an enclosed angle of 142.38, tip radius of 150 nm) was fitted with an 8th order polynomial. To evaluate the load– displacement curves, a method similar to the Oliver–Pharr method [20] was employed. The use of the improved software Indentanalyser 1.5 (ASMEC), which considered the tip rounding, reduced the scatter of the data substantially and provided the values of hardness and Young’s modulus in accordance to ISO 14577 [21].
3. Results and discussion Fig. 3(a) shows the load–displacement curves of the hard metal substrate and the three bias-graded layers of a DLC film (DLC I: 0 V/150 V/300 V; thickness ratio D 1:D 2:D 3=1:2:3; a=0.098) recorded on a small-angle cross-section in the centre of the layers at 20 mN maximum load. None of the curves reveals any pop-in events or other features, e.g., due to the influence of the underlying layers. The smallest value of the maximum penetration depth h max was determined for the layer deposited with U B=300 V (1) to be 190 nm and the maximum value of 230 nm was found for the layer deposited without a bias voltage (4). Another disturbing factor might be the pile-up of material during indentation, which leads to an incorrect determination of the contact area. According to Bolshakov and Pharr [22], however, this pile-up effect only plays a major role, if the quotient of the maximum displacement h max and the plastic residual depth upon unload h p is larger than 0.7. Here, this quotient reached a maximum value of 0.6 in the softest layer; thus, the above condition was fulfilled. From a set of similar load–displacement curves recorded across a line over the different layers and the hard metal substrate (see Fig. 2(a)), the hardness depth profile of DLC I presented in Fig. 3(b) could be determined by the SACS. Starting from the left side (sample surface), the stepwise decrease of the hardness with increasing depth for the graded layers deposited with bias voltages of 300 (3. layer), 150 (2. layer), and 0 (1. layer) is clearly visible. Whereas the 3. layer sputtered with the highest bias voltage of 300 V has the highest hardness of 34.2 GPa, because the highest ion bombardment provided the highest densification of the carbon network, the 1. layer deposited without bias voltage shows a hardness value of 17.1 GPa, which is lower than the value of the hard metal substrate of 18.6 GPa.
66
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
up which might have been present at 20 mN maximum load, but can be completely neglected at 2 mN maximum load. However, even at a maximum load of 2 mN, another issue of the SACS has to be taken into account. When looking back to Fig. 2(b), at a distance s j from the interface between two graded layers on a small-angle cross-section, the effective layer thickness d j can be calculated as d j =s j sina. For example, for a=0.098 and s 1=20 Am, an effective layer thickness of d 1=31 nm only is obtained. In Fig. 4(b), the load–displacement curves of the 2. layer (A), the 1. layer (D) and the interface region (B and C) are shown, which indicate that at a maximum load of 2 mN, the maximum penetration depth in the layers and the interface region was between 40 and 50 nm, which is larger than the effective layer thickness. Thus, since the effective layer thickness decreases strongly due to geometrical reasons with decreasing distance to the interface, the indenter finally penetrates into the softer layer which is situated under the layer under investigation, when passing a certain distance. Consecutively, the calculated hardness value becomes too small, which leads to a broadening of the measured hardness profile. However, this broadening can be theoretically approximated by a weighting of the effective area fractions
Fig. 3. (a) Typical load–displacement curves of hard metal substrate and the three bias-graded layers of a DLC film (DLC I: 0 V/150 V/300 V; thickness ratio D 1:D 2:D 3=1:2:3; a=0.098) at a maximum load of 20 mN. (b) Depth profile of hardness of DLC I determined by SACS.
These values are in good agreement with those determined by microindentation or nanoindentation on three singlelayer DLC samples, which have been prepared by applying the same bias voltages of 0, 150, and 300 V. In Fig. 3(b), there are two large interfacial regions of gradual hardness decrease between the graded layers of 375 nm between 3. and 2. layer and of 485 nm between 2. and 1. layer. To investigate the influence of the applied load in the nanoindentation experiment on the apparent expansion of the interface regions, the maximum load has been reduced by a factor of 10 to 2 mN. Fig. 4(a) compares the hardness depth profiles at the interface region between the 1. and the 2. graded layer of DLC I recorded at 20 (-n-) and 2 mN (-w-). The large influence of the maximum load is obvious because reducing the load from 20 to 2 mN leads to a large subsequent decrease of the measured interface expansion from 485 to 38 nm, which is indicated by the shaded region. Both layers can still be discriminated very well by their average hardness values of 20.0 (1. layer) and 27.2 GPa (2. layer). The little increase in the measured hardness of the 1. layer in comparison to Fig. 3(b) might be due to a little pile-
Fig. 4. (a) Depth profile of hardness at the interface region between 2. layer (U B=150 V) and 1. layer (U B=0 V) of DLC I at a maximum load of 20 (-n-) and 2 mN (-w-), together with theoretical profile broadening (- -). (b) Load–displacement curves of 2. layer (A), 1. layer (D) and interface region (B and C) of DLC I.
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
67
of the indenter in both layers A 1 and A 2 and their averaged hardness values H 1 and H 2: Htheo ¼
A2 H2 þ A1 H1 A2 þ A1
ð1Þ
and it can be separated from the measured hardness profile. The calculated theoretical broadened profile indicated in Fig. 4(a) by the bold dashed line, gives a value for the theoretical broadening of 24 nm, which can be subtracted from the measured expansion of 38 nm giving a residual expansion of the interface between the 1. and the 2. graded layer of DLC I of V14 nm. A further reduce of the maximum load to 1 mN did not give any improvements because with the corresponding shallow penetration depth it was not possible to clearly discriminate between the graded layers due to a large scatter in the measured data. By using the SACS, not only the hardness depth profile but also depth profiles of other mechanical properties can be determined. As an example, the Young’s modulus depth profile at the interface region between the 3. layer (U B=300 V) and the 2. layer (U B=150 V) of DLC I at a maximum load of 2 mN is shown in Fig. 5. The profile reveals that the interface expansion value of 98 nm deduced from the Young’s modulus data without correction of the theoretical broadening is much larger than the value of 24 nm taken from the hardness data. A possible reason could be that the elastic stress fields governing the Young’s modulus have a far longer interaction range than the plastic deformation which determines the hardness value. Thus, the elastic stress fields under the indenter tip reach far into the underlying layer when coming close to the interface between two graded layers. However, because the theoretical broadening is necessary in order to accomplish a good comparison, new models will have to be developed to determine the corrected interface expansion for the Young’s modulus.
Fig. 5. Depth profile of Young’s modulus at the interface region between 3. layer (U B=300 V) and 2. layer (U B=150 V) of DLC I at a maximum load of 2 mN.
Fig. 6. Comparison of depth profiles of hardness of two DLC films with different thickness ratios at a maximum load of 5 mN: (a) DLC I and (b) DLC II.
To investigate the influence of film processing parameters the hardness depth profile of another sample DLC II (D 1:D 2:D 3=3:2:1; a=0.048) with a different ratio of the layer thicknesses, was recorded by SACS. In Fig. 6, the hardness depth profiles at a maximum load of 5 mN of the two DLC films with different thickness ratio are compared. A load of 5 mN was chosen because the differences between both films are visible better across the complete profile than at 2 mN load. The macroscopic hardness values were largely different because of the large penetration depth of more than 1 Am. Because in sample DLC II the hardest layer was only 0.82 Am thick, the microindenter already penetrated into the softer 2. layer leading to a measured Vickers hardness of 2700 HV0.05 in comparison to 4100 HV0.05 for the sample DLC I with a three-times thicker upper layer. However, the comparison of their hardness depth profiles in Fig. 6 clearly proves that the hardness of the three individual graded layers is the same in both films. The film DLC I with a thickness ratio of D 1:D 2:D 3=1:2:3 showed interface
68
C. Ziebert et al. / Thin Solid Films 482 (2005) 63–68
the possibility to form sp2 or sp3 bonds makes the situation even more complex in amorphous carbon layers. Therefore, both other systems should also be investigated in detail before future studies will focus on the stress-state and the deformation and densification of amorphous carbon networks. A combination of experiments and theoretical simulations will be needed to get new insights into the subplantation model, to develop new models for the behaviour of mechanical layer properties across an interface and to find the optimal expansion of the interface regions to achieve optimum adhesion and maximum thickness of DLC films.
Acknowledgement Fig. 7. Comparison of thickness of interface regions d 12 and d 23 of two DLC films with different thickness ratio determined for a 2 mN load: (a) DLC I and (b) DLC II.
expansions d 12=180 nm and d 23=140 nm, which are indicated by the shaded regions, while the film DLC II with a reversed ratio of D 1:D 2:D 3=3:2:1 showed a larger value d 12=290 nm and a smaller value d 23=110 nm. Fig. 7 compares the corresponding expansions of interface regions for the two films at 2 mN load. On the one hand, there is a strong influence of the ratio D 1:D 2, which leads to d 12=14 nm when the thickness of the 2. layer is twice the thickness of the 1. layer, and to an eight-times larger interface expansion of 111 nm, when the thickness of the 2. layer is 3/4 of the thickness of the 1. layer. This shows that the thicker soft 1. layer forms a larger interface to the 2. layer, which might improve the adhesion of the system but leads to a macroscopic hardness decrease. On the other hand, almost no difference was found in d 23 with 24 nm or 22 nm when the thickness of the 3. layer is 3/2 of the thickness of the 2. layer or 1/2 of the thickness of the 2. layer, respectively.
4. Conclusions For amorphous stepwise bias-graded DLC films, it has been shown that the small-angle cross-section nanoindentation method (SACS) is a versatile tool for high-resolution depth-profiling of mechanical properties. It was revealed that the hardness values of the individual graded layers are independent on the thickness ratio of the graded layers, while the expansion of the interface regions between them showed a strong influence on this ratio and on the maximum applied load. However, it should be stressed that the in the case of amorphous carbon layers many different issues have to be taken into account. In comparison to crystalline nanolaminated coatings, such as TiN/ZrN multilayers, it is not possible to investigate the lattice parameters at the interfaces, e.g., by TEM, and in comparison to simpler amorphous materials such as B4C
The authors would like to thank C. Kourouklis of IMT Hanover for the preparation of nanogrinded small-angle cross-sections.
References [1] B. Bhushan, Diamond Relat. Mater. 8 (1999) 1985. [2] J. Robertson, Thin Solid Films 383 (2000) 81. [3] H. Schulz, H.J. Scheibe, P. Siemroth, B. Schultrich, Appl. Phys., A 78 (2004) 675. [4] D.F. Wang, K. Kato, N. Umehara, Surf. Coat. Technol. 123 (2000) 177. [5] R. Lappalainen, H. Heinonen, A. Antla, S. Santavirta, Diamond Relat. Mater. 7 (1998) 482. [6] M.I. Jones, I.R. McColl, D.M. Grant, K.G. Parker, T.L. Parker, Diamond Relat. Mater. 8 (1999) 457. [7] J. Schwan, S. Ulrich, H. Roth, H. Ehrhardt, S.R.P. Silva, J. Robertson, R. Samienski, R. Brenn, J. Appl. Phys. 79 (1996) 1416. [8] S. Logothetidis, M. Gioti, P. Patsalas, C. Charitidis, Carbon 37 (1999) 765. [9] D.R. McKenzie, D. Muller, B.A. Pailthorpe, Phys. Rev. Lett. 67 (1991) 773. [10] Y. Lifshitz, S.R. Kasi, J.W. Rabelais, W. Eckstein, Phys. Rev., B 41 (1990) 10486. [11] H. Holleck, M. Stqber, Method of manufacturing a composite material structure, US Patent No. 6, 110, 329 (2000); EU Patent No. EP 0912774B1 (2002). [12] C. Bauer, H. Leiste, M. Stqber, S. Ulrich, H. Holleck, Diamond Relat. Mater. 11 (2002) 1139. [13] C. Mathioudakis, P.C. Kelires, Y. Panagiotatos, P. Patsalas, C. Charitidis, S. Logothe- tidis, Phys. Rev., B 65 (2002) 205203. [14] A.Yu. Belov, H.U. J7ger, Comput. Mater. Sci. 24 (2002) 154. [15] S.B. Sinnott, R.J. Colton, C.T. White, O.A. Shenderova, D.W. Brenner, J.A. Harrison, J. Vac. Sci. Technol., A 15 (1997) 936 (Part 1). [16] M. Kunert, B. Baretzky, S.P. Baker, E.J. Mittemeijer, Metall. Mater. Trans., A 32 (2001) 1201. [17] S. Ulrich, C. Ziebert, M. Stqber, P. Schlogmacher, H. Holleck, E. Schweitzer, M. Gfken, Surf. Coat. Technol. 188 – 189 (2004) 331. [18] M. Stqber, S. Ulrich, H. Leiste, A. Kratzsch, H. Holleck, Surf. Coat. Technol. 116-119 (1999) 591. [19] H.H. Gatzen, J.C. Maetzig, Precis. Eng. 21 (1997) 134. [20] W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (1992) 1564. [21] ISO/FDIS 14577-1:2002; Metallic materials—Instrumented indentation test for hardness and materials parameters. [22] A. Bolshakov, G.M. Pharr, J. Mater. Res. 13 (1998) 1049.