Adhesive strength of electroplated copper films

Adhesive strength of electroplated copper films

Journal of Materials Processing Technology 114 (2001) 252–256 Adhesive strength of electroplated copper films$ C.H. Seaha,*, S. Mridhab, L.H. Chanc a...

479KB Sizes 2 Downloads 31 Views

Journal of Materials Processing Technology 114 (2001) 252–256

Adhesive strength of electroplated copper films$ C.H. Seaha,*, S. Mridhab, L.H. Chanc a

Thin Film Department, Chartered Silicon Partners Pvt. Ltd., 60 Woodlands Industrial Park D, Street 2, Singapore 738406, Singapore b School of Materials Engineering, Nanyang Technological University, Singapore, Singapore c Research and Development Department, Chartered Semiconductor Manufacturing Ltd., Singapore, Singapore

Abstract A study has been conducted to characterise the scratch resistance and adhesive strength of copper films electroplated on to thin layers of W/TiN/Si(1 0 0) and Cu/Ta/Si(1 0 0) substrates. The properties were characterised with a microscratch tester using diamond indenter of different sizes. For the electroplated copper film on W seed, a 200 mm tip with an increased loading produced initially a minor delamination corresponding to damage along the sides of the scratch path followed by cracking at the centre of the scratch. A complete delamination of the film, causing tearing of the film from the underlying substrate, occurred at a critical load of 16 N. In the case of the film on Cu seed, no delamination was observed within the loading range up to 22 N under a 200 mm tip. When a 50 mm indenter was used a minor delamination was seen initially at a force of 4 N, and this delamination subsequently increased with increasing load, but no full delamination was found within the loading range of up to 22 N. The full delamination of the film on Cu seed occurred at 3 N when the scratch testing was performed using a 20 mm diamond tip. The failure phenomenon for the film on Cu seed is different from that on W seed. The initial failure was caused by radial cracking in the film, which turned into chipping of the film as the loading force increased. The film then completely delaminated, causing substantial debris to be strewn along the scratch path. # 2001 Elsevier Science B.V. All rights reserved. Keywords: Scratch resistance; Indenter; Delamination; Radial cracking

1. Introduction As the semiconductor drives towards faster circuits, the RC delay due to metallisation layers needs to be reduced. Thanks to its lower resistivity and higher electromigration and stress migration resistance, copper has established itself to be a very promising substitute for aluminium in interconnections. However, copper is very difficult to pattern. It has been then shown that alternating interconnect architectures such as damascene and dual damascene, coupled with chemical mechanical planarisation (CMP) could resolve this patterning problem of copper layers [1,2]. In this process, the dielectric is patterned with a conventional dry-etch process to form trenches and vias for horizontal and vertical interconnects. An adhesive layer (usually Ta/TaN) and a copper layer are then deposited to fill (and overfill) these holes. Finally, using CMP, the excess of copper is removed in the field areas, to leave metal only in vias and lines. Because CMP leaves the wafer surface planar on a global scale, this sequence may be easily repeated to add multiple layers of metal. $

Accepted for publication as part of the APCMP-S conference. Corresponding author. E-mail address: [email protected] (C.H. Seah). *

To ensure that there is no residual copper and barrier material in the region between the trenches, and hence no shorting of any two copper lines, requires that one clears excess copper and barrier material everywhere on the die and wafer. This requirement implies overpolish in some regions of the die and wafer, leading to dishing of copper and erosion of oxide. Copper dishing and oxide erosion lead to considerable surface non-planarity and cause various process integration problems. They also reduce the crosssections of the copper lines and dielectric spacings, leading to an increase in interconnect resistance (signal relay times) and related deterioration of device performance. Therefore, a high selectivity between Cu and SiO2 is required and it has been reported in copper CMP to be between 130 and 150% [3]. As a result the adhesion of the electroplated copper film with the underlying seed and barrier layers onto the oxide etched trenches and vias has become of paramount importance. This study focused on the characterisation of such films in term of scratch resistance and adhesive strength. Copper films were electroplated on thin layers of W/TiN/ Si(1 0 0) and Cu/Ta/Si(1 0 0) substrates and the properties of these films were then characterised with a microscratch tester using a diamond indenter of different sizes.

0924-0136/01/$ – see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 4 - 0 1 3 6 ( 0 1 ) 0 0 6 2 4 - 0

C.H. Seah et al. / Journal of Materials Processing Technology 114 (2001) 252–256

2. Procedure The substrate used was (1 0 0)-oriented p-type Si, on which ˚ TiN or 350 A ˚ Ta was deposited as the diffusion either 1200 A ˚ W, acting as the seed material barrier layer. A layer of 1000 A for the copper electroplating process and also the diffusion barrier, was deposited onto the TiN barrier-coated Si wafer, ˚ Cu seed layer was deposited onto the Ta while a 1200 A barrier-coated Si wafer. The electrolyte was composed of CuSO45H2O and H2SO4 solution. The electrochemical deposition on both seeded substrates was done at 308C with applied voltages of 5 and 10 V, current densities of 0.10, 0.15 and 0.20 A/cm2 and a deposition time of 30 s. The anode used was a piece of electronic grade copper sheet, while the cathode was the substrate to be plated. A CSEM Instruments Micro Scratch Tester was used to measure the adhesive strength of the copper film by applying a load ranging from 0 to 20 N which generated scratches with a spherical Rockwell C diamond stylus. Tip radii of 20 and 200 mm were used for the Cu seeded and W seeded substrates, respectively; the tip was drawn across the filmsubstrate system to be tested, under progressive loading at a fixed rate of 10 mm/min. The critical load (Lc) was defined as the smallest load at which a recognisable failure occurred. The acoustic emission (AE) detection and the tangential force (Ft) were also recorded. The run is performed for three times to complete the test. Field emission scanning electron microscopy (FESEM) and atomic force microscopy (AFM) were then employed to characterise the microstructure of the scratch marks on the as-deposited copper films. 3. Results and discussion 3.1. Critical loads Fig. 1 shows a typical set of results obtained with the microscratch tester for the electroplated copper films on both

253

W and Cu seeded silicon substrate. In the graph, both the acoustic emission (AE) signal and the tangential force (Ft) are plotted as a function of applied load. For both specimens, there are distinct critical loads at which different failure modes occurred and could be recognised by viewing the specimens under the SEM. These critical loads are also seen to correspond directly with sudden changes in both the acoustic signal and the Ft. 3.1.1. Critical load for W seed Copper films were plated onto W seeded substrates using 0.10, 0.15 and 0.20 A/cm2 and at 5 and 10 V. The films produced using 5 V and those produced with 10 V and 0.10 A/cm2 gave a wide variation of critical loads across the surface during the scratch test, which could be attributed to inhomogeneity in the film. It was difficult to determine whether failures in the film of these specimens were due to the variation in film composition/thickness or to the actual scratch test. As such the results obtained were not used for analysis. However, films produced with 10 V and 0.15 A/ cm2 and 10 V and 0.20 A/cm2 gave reproducible results, as are tabulated in Table 1. These specimens were tested with a 200 mm tip. 3.1.2. Critical load for Cu seed A full set of measurements was obtained for copper film specimens on Cu seed, plated using 0.10, 0.15 and 0.20 A/ cm2 at 5 and 10 V. The specimens were tested using a 20 mm diamond tip except for one plated at 5 V and 0.10 A/cm2. This specimen was used to find the optimum conditions for the test, which involves performing several scratches. The initial scratching was performed using a 200 mm tip and no delamination was observed up to 21.6 N, only radial cracking being found along the scratch path. The 50 mm tip was then used to perform the scratching. Only minor delamination was observed at a force of 4.5 N, after which increased delamination was observed at 5.2 N and above. However, no

Fig. 1. Acoustic emission signal and the tangential force obtained for a scratch performed on electroplated copper film on: (a) W seed layer (10 Vand 0.2 A/cm2) with a 200 mm diamond tip and (b) Cu seed layer (10 V and 0.2 A/cm2) with a 20 mm diamond tip, using a loading rate of 5–20 N/min, and a speed of 10 mm/min.

254

C.H. Seah et al. / Journal of Materials Processing Technology 114 (2001) 252–256

Table 1 Adhesive strength of electroplated copper film on W seed layer using a 200 mm diamond tip, a loading rate of 20 N/min, and a speed of 10 mm/min Specimen 2

10 V and 0.15 A/cm 10 V and 0.20 A/cm2

Lc1 (minor delamination) (N)

Lc3 (cracks at centre of scratch) (N)

Lc5 (delamination) (N)

12.3 11.5

13.0 13.8

16.1 15.1

full delamination was found within the loading range. Based on these observations, a diamond tip of 20 mm was used to perform the test on the other specimens, the results obtained being tabulated in Table 2. Using a 200 mm diamond tip, delamination occurred at about 16 N for the film on W seed (Table 1), while it was not observed in the film on Cu seed within the loading range of 0–21.6 N. This observation indicates that the adhesive strength of the electroplated copper film on Cu seed layer is much better than that on W seed. When the 50 mm indenter tip was used on the same Cu seed specimen, minor delamination was initially observed at a force of 4.5 N, after which increased delamination was seen from 5.2 N onwards. However, no full delamination was found within this loading range. Full delamination of the film on Cu seed only occurred at 2.6–2.8 N when a 20 mm diamond tip was used (Table 2). These observations indicate that the electroplated film on Cu seed has a higher adhesive strength than that deposited on W seed. However, the force required for the earlier failure in the film on Cu seed layer, which is the chipping of the film, was found to be in the range between 0.7 and 2.1 N (Table 2). No definite trend was noticed in terms of critical load for failure in the films produced with different applied voltages and current densities. Further investigation would be needed in order to explain why this wide range of loads was obtained. Randall [4] encountered four distinct failure points when ˚ SiN/SiC passivation layer was scratched with a a 4000 A progressive load range of 0–15 N and a diamond tip of radius 20 mm. The first critical force value of 3.26 N corresponded to damage along the sides of the scratch path. At 9.77 N first cracking appeared, after which continuous flaking occurred at 10.60 N. Shortly afterwards, the film was seen to completely delaminate at a critical force of 11.74 N. This result shows that the SiN/SiC passivation layer has better scratch resistance than the electroplated copper film.

Table 2 Adhesive strength of electroplated copper film on Cu seed layer using a 20 mm diamond tip, a loading rate of 5 N/min, and a speed of 10 mm/min Specimen

Lc1 (chipping of film) (N)

Lc5 (delamination) (N)

5 V and 0.15 A/cm2 5 V and 0.20 A/cm2 10 V and 0.10 A/cm2 10 V and 0.15 A/cm2 10 V and 0.20 A/cm2

– 0.7 2.1 1.0 1.2

2.4 2.6 2.8 2.8 2.8

3.2. Morphology of critical failure points Figs. 2 and 3 show the SEM micrographs of the two specimens corresponding to the distinct critical failure points in Fig. 1. For an electroplated copper film on W seed layer, Fig. 2 exhibits three distinct critical failure points, namely (a) minor delamination, (b) cracks at the centre of the scratch and (c) full delamination. Fig. 3 shows two critical failure points, namely (a) chipping of the film and (b) full delamination, for the copper film on Cu seed layer. The first critical failure point, Lc1, is characterised by a sharp drop in AE in Fig. 1(a) and represents the first signal of delamination from the substrate, which is seen as cracking of the film at the sides of the scratch path in Fig. 2(a). The second critical failure point, Lc3, shows a smaller jump in AE and corresponds to cracking in centre of scratch. Interestingly enough, the SEM micrograph (Fig. 2(b)) also shows the greater propagation of scratches away from the scratch path and some buckling to the right of the image. As this is a multilayered film it is probable that the buckling effect is caused by mismatch between the uppermost layer(s) and those beneath, resulting in delamination at a certain distance from the track. Final failure, Lc5, is characterised by total delamination from the substrate, an event which is symbolised by another small drop in AE and a sharp increase in Ft. This can also be observed accurately from the micrograph in Fig. 2(c). In the case of the copper film deposited onto Cu seed, only two distinct critical loads corresponding directly with sudden changes both in the acoustic signal and the frictional force could be seen in Fig. 1(b). A sharp drop in AE characterises the first critical failure point, Lc1, and it corresponds to chipping of the film at the sides of the scratch path as revealed in Fig. 3(a). The second critical failure point, Lc5, shows a smaller jump in AE and a sharp increase in Ft, representing a sign of full delamination from the substrate, which is seen as removal of a chip or flake from the surface, see Fig. 3(b). While the SEM shows the damage in the film, it cannot measure it quantitatively, so the AFM has been used to measure selected sections of a scratch and also to determine the extent of pile-up along its edges, and other plastic or brittle deformation behaviour, see Fig. 4. Two-dimensional AFM images of scratches made on the copper film specimens deposited on Cu seed with 0.10 and 0.20 A/cm2 at 10 V using a load of 20 N and a diamond tip of 20 mm are shown in Fig. 4. The film deposited using high cathode

C.H. Seah et al. / Journal of Materials Processing Technology 114 (2001) 252–256

255

Fig. 2. Progressive load scratch data for an electroplated copper film on W seed layer which exhibits three distinct critical failure points.

current density (Fig. 4(b)) produced more plastic deformation than that formed using lower current density (Fig. 4(a)). For film plated using a low cathode current density, plastic deformation is restricted, whereas for that plated using a higher current density, the deformation of pile-up material

on either side of the advancing indenter becomes significant. A previous study [5] has shown that plating at a higher current density formed small copper grains. This could explain why more plastic deformation was found on the film plated with higher current density.

Fig. 3. Progressive load scratch data for an electroplated copper film on Cu seed layer which exhibits two distinct critical failure points.

256

C.H. Seah et al. / Journal of Materials Processing Technology 114 (2001) 252–256

Fig. 4. Two-dimensional AFM images of scratches made on the copper film specimens deposited on Cu seed using: (a) 0.10 and (b) 0.20 A/cm2 at 10 V.

3.3. Adhesion strength of copper films The electroplated film on Cu seed has better adhesive strength than that deposited on W seed. Using a 200 mm diamond tip, delamination occurred at about 16 N for the film on W seed (Table 1) while it was absent in the film on Cu seed within the loading range of 0–21.6 N. This indicates that the adhesive strength of the electroplated copper film on Cu seed layer is much higher than that on W seed. Full lamination of the film on Cu seed occurred only at 2.6–2.8 N when a 20 mm diamond tip was used (Table 2). The SEM micrographs of the scratch test surface for the electroplated copper film on W seed layer, Fig. 2, exhibits three distinct critical failure points, namely (a) minor delamination, (b) cracks at the centre of the scratch and (c) full delamination, while for the film on Cu seed, Fig. 3 shows only two distinct critical failure points, namely (a) chipping of the film and (b) full delamination. There is no crack at centre of the scratch observed for the latter, suggesting that the buckling effect is not occurring between the electroplated copper film and the underlying layers, which has been observed for that on W seed (Fig. 2(b)). This further supports the observation that electroplated film on Cu seed has better adhesion.

4. Conclusion 1. For the electroplated copper film on W seed, a 200 mm tip with an increased loading gave a minor delamination

corresponding to damage along the sides of the scratch path followed by cracking at the centre of the scratch. A complete delamination causing tearing of the film from the underlying substrate occurred at a critical load of 16 N. In the case of the film on Cu seed, no delamination was observed within the loading range up to 22 N under the 200 mm tip. 2. When a 50 mm indenter was used, a minor delamination was seen initially at a force of 4 N, and this delamination subsequently increased with increasing load, but no full delamination was found within the loading range up to 22 N. The full delamination of the film on Cu seed occurred at 3 N when the scratch testing was performed using a 20 mm diamond tip. 3. The failure phenomenon for the film on Cu seed is different from that on W seed. The initial failure was caused by radial cracking in the film, which turned into chipping of the film as the loading force increased. The film then completely delaminated, causing substantial debris to be strewn along the scratch path. References [1] C.W. Kaanta, et al., VMIC Conf. Proc. 8 (1991) 144. [2] B. Luther, et al., VMIC Conf. Proc. 10 (1993) 15. [3] Z. Stavreva, D. Zeidler, M. Plo¨ tner, G. Grasshoff, K. Drescher, Microelectron. Eng. 33 (1997) 249. [4] N. Randall, Mater. World 6 (8) (1998) 476. [5] C.H. Seah, S. Mridha, L.H. Chan, IITC 1 (1998) 157.