Sei sulla pagina 1di 7

Surface and Coatings Technology 176 (2003) 124130

A nanoindentation study of copper films on oxidised silicon substrates


D. Beegana,b, S. Chowdhurya,b, M.T. Laugiera,b,*
b a Department of Physics, University of Limerick, Limerick, Ireland Materials and Surface Science Institute, University of Limerick, Limerick, Ireland

Received 26 February 2003; accepted in revised form 7 May 2003

Abstract The indentation hardness of copper films deposited on oxidised silicon substrates is investigated by two methods. The first uses the Oliver and Pharr method for analysing the loaddisplacement curves and the second involves the measurement of the residual indent impression with AFM. Material pile-up was found at the indent edges, and this area needs to be accounted for in hardness measurement. The effect of the pile-up area is investigated and the true film hardness is determined. 2003 Elsevier B.V. All rights reserved.
Keywords: B nanoindentation; B atomic force microscopy (AFM); D copper

1. Introduction Nanoindentation is an instrumented indentation method, which is widely used to determine the mechanical properties of both bulk solids and thin films w15x. In nanoindentation, an indenter tip with a known geometry is driven into a specific site of the material to be tested by applying an increased normal load. Upon reaching the preset maximum value, the normal load is reduced until partial or complete relaxation. At each stage of the experiment, the position of the indenter relative to the sample surface is precisely monitored and a load displacement curve is obtained. Material hardness is generally obtained by dividing the maximum applied load by the residual contact area. Classical macro- or micro-hardness experiments required the imaging of the indents in order to measure the contact area. There have been various methods developed to calculate the contact area directly from the loaddisplacement curves. The most commonly used method, developed by Oliver and Pharr w2x, estimates the contact area from the contact depth, determined from the loaddisplacement curves (Fig. 1). This makes the imaging of the indent impression unnecessary. However, during the indentation of a soft strain hardenable mate*Corresponding author. Tel.: q353-61-202257; fax: q353-61202423. E-mail address: michael.laugier@ul.ie (M.T. Laugier).

rial, particularly a soft thin film on a hard substrate, pile-up of the material on the faces of the indenter is often found w311x. If it is assumed that this piled-up material supports the load during the indentation process, it must be accounted for in the calculation of the hardness value. The hardness measured by the Oliver and Pharr method does not take this pile-up into account, leading to an overestimate in the hardness value w3,4,12x. In this paper, the hardness of copper films on oxidised Si substrates is investigated. These films are sputter deposited to thicknesses of 950 and 1400 nm. The hardness of these films is initially determined using the Oliver and Pharr method. In order to determine the actual load-bearing area, the indentations are directly measured via AFM. The effect of pile-up on the area determination and the hardness values is investigated. 2. Experimental 2.1. Film deposition The P-type silicon (100) wafers (Wacker, rs230 mV cm) were used as substrates in this work. Specimen cleaning was carried out in clean room conditions. A thin layer of SiO2 was then thermally grown on the substrates. The SiO2 film thickness was measured by ellipsometry.

0257-8972/03/$ - see front matter 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0257-8972(03)00774-6

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130

125

A minimum of five indents was performed per load. A Berkovich diamond indenter was used for all indents and the Oliver and Pharr method w2x was used to determine the reduced elastic modulus Er and the hardness H. The Oliver and Pharr analysis is based upon relationships developed by Sneddon w14x for the penetrations of a flat elastic half space by different probes with particular axisymmetric shapes (e.g. a flat-ended cylindrical punch, a paraboloid of revolution and a cone). In general, the relationships between penetration depth h and load P for such indenter geometries can be represented in the form Ps ah y hf.m (1)

Fig. 1. (a) Loaddisplacement curve, showing the values used in the Oliver and Pharr method and (b) cross-section of an indentation.

The Copper films were deposited in a Leybold Lab 500 RF magnetron sputtering system (RF 13.56 MHz). The chamber base pressure was below 2.3=10y5 mbar and the process pressure was 4=10y3 mbar. A 99.9999% pure copper target was used for film deposition with an applied RF power of 40 W. The copper deposition rate was 9.6 nmymin and film thicknesses were varied by varying the deposition time. In this paper, we look at the copper films of thicknesses 950 and 1400 nm. An in-situ quartz crystal oscillator was used to monitor the film thickness. The Film thickness was measured subsequently by measuring a step height with a Zygo white light interferometer. 2.2. Determination of film mechanical properties The mechanical properties of the films were characterised using a nano hardness tester (NHT) developed by CSEM instruments, Switzerland. The system has load and displacement resolutions of 1 mN and 0.3 nm, respectively. Indents were performed over a range of loads from 0.5 to 100 mN. We used a constant strain rate method w13x, by setting 1yP(dPydt) to be 2 miny1 and a pause of 10 s was set at maximum load.

where a contains geometric constants, the sample elastic modulus, the sample Poissons ratio, the indenter elastic modulus and the indenter Poissons ratio, hf is the final unloading depth and m is a power law exponent that is related to the geometry of the indenter; for a flat-ended cylindrical punch, ms1, for a paraboloid of revolution, ms1.5 and for a cone, ms2. From Eq. (1), two observations were made: firstly, the slope of the unloading curve changes constantly due to a constantly changing contact area; and secondly, if the unloading curve can be fitted by a power law expression, then a derivative dPydh applied at the maximum loading point (hmax, Pmax) should yield information about the state of contact at that point. This derivative is termed the contact stiffness S, and is given by Ssb 2 EryA (2)

yp

where A is the projected area of tip-sample contact and b is used to account for the triangular and square cross sections of many indenters used in nanoindentation studies. The reduced modulus accounts for deformation of both the indenter and the sample, and is given by 1 1ynS. 1ynIi. s q Er ES EI
2 2

(3)

where E and n are the Youngs modulus and Poissons ratio with the subscripts S and I meaning sample and indenter, respectively. The hardness is thus calculated from Hcs Pmax A c fhc. (4)

The area of contact at peak load is determined from

126

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130

nm in the z direction. The images obtained were processed by TopoMetrix SPM Lab NT Version 5.0 software supplied with the microscope. The image processing consists of a third order two-dimensional levelling of the surface. The line scan tools on the SPM software were used for the measurement of the indents. 3. Results and discussion 3.1. Loaddisplacement curves Fig. 2 shows typical loaddisplacement for the 1400 nm copper film. We can see that for all loads, the penetration depth is less than the film thickness. For the 950 nm film only at loads greater than 80 mN, the maximum penetration depth (hm) is greater than the film thickness. The loaddisplacement behaviour is typical of that for a soft metal and it is seen that there is very little elastic recovery indicating that the deformation occurs mainly by plastic processes. During the pause at maximum load, the displacement is seen to increase due to creep. The films are constrained by substrates and thus their properties depend on the penetration depth even in a depth shallower than the film thickness. In order to measure the film-only properties, a commonly used rule of thumb is to limit the indentation depth to less than 10% of the film thickness w8,15x. It is found from the loaddisplacement curves that in order to penetrate to 10% or less of the film thickness, the maximum load is 1 mN for the 1400 nm film. The corresponding load for the 950 nm film is found to be 0.5 mN. 3.2. Hardness behaviour The hardness values for all films have been plotted as a function of the plastic depth (Fig. 3). The hardness of the films has been calculated using the Oliver and Pharr method. The results shown are an average of at

Fig. 2. Series of loaddisplacement curves for 1400 nm Cu film on oxidised Si substrate.

the geometry of the indenter and the depth of contact. It is assumed that the indenter geometry can be described by an area function which relates the cross-sectional area of the indenter to the distance from its tip h. Given that the indenter does not itself deform significantly, the projected contact area at peak load can then be computed from the relation Asfhc. (5)

where hc is related to the deformation behaviour of the material and the shape of the indenter. In fact, hcs hmaxyhs where hs is defined as the elastic displacement of the surface at the contact perimeter and can be calculated for specific geometries using displacement equations from Sneddons analysis. For each three specific tip shapes (flat-ended punch, paraboloid of revolution and cone) hssPmax yS, where is a function of the particular tip geometry. Thus, hc is given by hcshmaxy Pmax S (6)

where the geometric constant for the conical indenter is 0.72, for the flat punch, s1 and for the paraboloid of revolution, s0.75. 2.3. AFM analysis The AFM measurements were carried out with a Surface Probe Microscope Topometrix Explorer . The images were collected in non-contact mode. The cantilevers used were high resonance frequency (HRF) silicon I shaped cantilevers, frequency range 279318 kHz. The dimensions of the arm of the cantilever were: length 132 mm; width 2628 mm; thickness 3.84.5 mm. These cantilevers end with a pyramidal tip: 36 mm base, 1015 mm high, with tip radius -20 nm and force constant: 2546 Nmy1. The scanner used has 10 mm z and 100 mm xy with background noise of 310

Fig. 3. Variation in the hardness as determined by the Oliver and Pharr method with the plastic depth for both Copper films.

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130

127

from these AFM images. Fig. 5 shows the area taken as the projected contact area. The contact point (c.p.) is taken to be the maximum point of the pile-up, shown as the dotted line in Fig. 5a. The area is then calculated by assuming that the load supporting area of the pile-up forms an arc of radius R, at the indent edges. By measuring the indent edge a and the radius R, it is thus possible to calculate the total piled-up area. The total pile-up area is given by: Apile-ups3C
B

pR2 a y D 6 2

R2y

a2 E F 4G

(7)

The triangular indent area is then calculated from Atriangles0.433 a2 (8)

We can now calculate the total projected contact area, which is used in Eq. (4) to determine the film hardness. From Fig. 6 we can see that the pile-up measured by the AFM is found to increase with increasing penetration depth as has been previously found by Chen and Vlassak and Moody et al. w17,18x. The 1400 nm film exhibits
Fig. 4. (a) 60 mN Indent on 950 nm Copper film and (b) line scan of indent showing pile-up around indent edges.

least five indentations with the error bars representing "1 S.D. For the films, the hardness remains constant at penetration depths greater than 200 nm, with values between 3 and 3.5 GPa for both films. At the lower depths, however, there is an increase in hardness, up to 4.5 GPa and 5.2 GPa for the 950 nm and 1400 nm films, respectively. This is due to an indentation size effect, which is commonly found in soft metal films and has been related to strain gradient plasticity w16x. At penetration depths approaching the film thickness, the hardness does not seem to be affected by the substrate which has a much higher hardness of ;12.5 GPa. In this case, the 10% rule is too restrictive. The film is soft in comparison with the substrate and it seems that all the plastic deformation takes place in the film and the substrate does not yield until the indenter penetrates the substrate. 3.3. AFM measurement In materials with low strain hardening capacity, such as copper, pile-up is commonly found. The AFM imaging of the indents is used to determine the extent and nature of the pile-up found around the indents. In Fig. 4a, we can see the pile-up evident approximately 60 mN indent in 950 nm copper film. Fig. 4b shows a line scan of this indent. The load-bearing area is determined

Fig. 5. (a) Projected area determination and (b) cross section showing pile-up contact point determination.

128

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130

Fig. 6. The variation in pile-up height with penetration depth for the 950 nm Copper film.

pile-up down at loads greater than 20 mN and the 950 nm film has pile-up down at loads greater than 5 mN. For the indents where no pile up is evident (see Fig. 7a and b), we calculate the hardness using the projected area of the triangle only. There are two main factors to be considered when comparing the hardness as measured from the AFM and that obtained using the Oliver and Pharr method. Firstly, the Oliver and Pharr method is based on equations obtained from a solution for a purely elastic contact developed by Sneddon. This may not give accurate results for elasticyplastic indentation. Secondly, the hardness measured by the AFM is the traditional hardness as calculated from the residual area of the indent, whereas the Oliver and Pharr method calculates a hardness value based on an area of contact at maximum load. In the case of soft materials where no pile-up is observed, it is found that these two values are very similar. The residual area as measured by AFM may be taken to be the true value, and any deviations between the values measured and those obtained by the nanoindentation analysis are attributed to inaccuracies in the procedure by which the contact areas are obtained from the experimental loaddisplacement curve. There are a number of different parameters measured from these AFM images that we can compare with those obtained by the NHT and predicted by the Oliver and Pharr method. Firstly, we can compare the AFM measured residual depth with the final depth value obtained with the NHT. It is found that for both films, at nearly all loads, the AFM measured values of residual depth are slightly less than those measured by the NHT (Fig. 8). This may be due to relaxation in the film in the time between indentation and imaging. The contact edge a, shown in Fig. 5a is equal to 7.53=hc from the shape of the tip. Thus, it is possible to compare the values of a

Fig. 7. (a) 1 mN Indent in 950 nm Copper film and (b) cross-sectional line scan of the indent showing no pile-up.

measured by the AFM with those calculated from the hc values given by the loaddisplacement curves. Fig. 9 shows the values obtained by both methods. The contact edge measured by the AFM is always slightly larger than that predicted by the Oliver and Pharr method from the loaddisplacement curve. The difference in contact edge will of course lead to a difference in the contact area. If we measure the area of the indents at all loads and excluding the pile-up area, compare it with the area calculated by the Oliver and Pharr method, we see that the AFM values are always larger (Fig. 10). Fig. 11 shows the hardness calculated using the area measured with the AFM and compares them with those

Fig. 8. Residual depth as measured with the AFM compared with the value obtained from the NHT.

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130

129

Fig. 9. Contact edge a, as measured by AFM compared with that predicted by the Oliver and Pharr method.

Fig. 11. Variation in hardness with applied maximum load. The values shown are for both Copper films as measured by the NHT and calculated from the AFM measurements.

obtained from the Oliver and Pharr method. The AFM values are found to be much lower than those found using the Oliver and Pharr method for all loads. The AFM values are constant, between 1.7 and 2.2 GPa at the loads greater than 20 mN. The increase in hardness at lower loads is still evidenced with values reaching 2.8 GPa at 0.5 mN. 4. Conclusions It has been seen that the Oliver and Pharr method of evaluating the load displacement curves for these copper films leads to inaccurately high values of hardness. This method does not include the pile-up of material in the estimation of the contact area. By imaging the indents with AFM, it is possible to determine the total projected contact area and thus determine a more accurate value for the film hardness.

The values measured with the AFM for the residual depth h0 and the contact edge a, are found to be different to those estimated from the loaddisplacement curve. Due to these differences, the areas calculated by the two methods do not agree, even when no pile-up is found in the hardness values. The values measured via AFM are always less than those measured with the NHT. This leads to a lower value of hardness even when no pileup occurs around the indent edges. Although, there may be errors associated with the hardness values determined from the AFM residual area, these may be considered to be the more accurate values especially in the case of a soft film on a hard substrate, where the pile-up takes place. Acknowledgments Support for this work by the Higher Education Authority (HEA), through the Materials and Surface Science Institute (MSSI), University of Limerick is gratefully acknowledged. References
w1x M.F. Doerner, W.D. Nix, J. Mater. Res. 1 (1986) 601. w2x W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (1992) 1564. w3x W.J. Poole, M.F. Ashby, N.A. Fleck, Scr. Mater. 34 (1996) 559. w4x G.M. Pharr, Mater. Sci. Eng. A253 (1998) 151. w5x R. Saha, W.D. Nix, Mater. Sci. Eng. A319321 (2001) 898. w6x R. Saha, W.D. Nix, Acta Mater. 50 (2002) 23. w7x Z.-H. Zu, D. Rowcliffe, Surf. Coat. Technol. 157 (2001) 231. w8x D.E. Kramer, A.A. Volinsky, N.R. Moody, W.W. Gerberich, J. Mater. Res. 16 (2001) 3150. w9x T.Y. Tsui, J. Vlassak, W.D. Nix, J. Mater. Res. 14 (1999) 2196. w10x T.Y. Tsui, W.C. Oliver, G.M. Pharr, Mater. Res. Soc. Symp. Proc. 436 (1997) 207.

Fig. 10. Area of indents as determined by AFM (not including pileup) and by the NHT using the Oliver and Pharr method.

130

D. Beegan et al. / Surface and Coatings Technology 176 (2003) 124130 w14x w15x w16x w17x w18x I.N. Sneddon, Int. J. Eng. Sci. 3 (1965) 47. A.C. Fischer-Cripps, Vacuum 58 (2000) 569. W.D. Nix, H. Gao, J. Mech. Phys. Solids 46 (1998) 411. X. Chen, J.J. Vlassak, J. Mater. Res. 16 (2001) 2974. N.R. Moody, A. Strojny, D. Medlin, S. Guthrie, W.W. Gerberich, Mater. Res. Soc. Symp. Proc. 522 (1998) 281.

w11x A. Bolshakov, W.C. Oliver, G.M. Pharr, Mater. Res. Soc. Symp. Proc. 436 (1997) 141. w12x R. Saha, Z. Xue, Y. Huang, W.D. Nix, J. Mech. Phys. Solids 49 (2001) 1997. w13x B.N. Lucas, W.C. Oliver, G.M. Pharr, J.-L. Loubet, Mater. Res. Soc. Symp. Proc. 436 (1997) 233.

Potrebbero piacerti anche