Tutorial #34
K. O’Donnell, J. Kostetsky, and R. Devito, NEXX Systems,Inc.
V. Bellido-Gonzalez, S. Powell, and D. Monaghan, Gencoa, Ltd.
October 2003

Based upon a paper presented at the IMAPS Flip Chip Technology Workshop, Austin, Texas, June 15 – 18, 2003

Introduction

The transition to lead free solder requires a change in the under bump metallurgy (UBM) solder wettable layer from copper to nickel due to the lower consumption rate of Ni by high-Sn solders. Nickel deposited by evaporation has high film stress. Nickel deposited by electroless plating using NiP is amorphous, which facilitates rapid atomic diffusion of the solder through the UBM. The failure mechanism in electroless nickel with lead-free SnAgCu solder is similar to that in eutectic SnPb solder.

Nickel with 7wt.% vanadium, which is non-magnetic and therefore more easily deposited by magnetron sputtering, has been used successfully with eutectic SnPb solder. However, there are serious issues with NiV as a UBM for eutectic SnAgCu solder due to the much higher solubility of Cu in SnAgCu [1].

In this work, pure nickel is deposited by magnetron sputtering, using an innovative new magnetron design [2], which overcomes the problems of conventional sputtering techniques for magnetic target materials. Conventional magnetron sputtering of magnetic targets requires the use of thin (< 4 mm) targets, which need to be replaced frequently due to the low target utilization of 10 – 20%. The Loop magnetron used to sputter nickel in this work can sputter 10 – 20 mm thick targets of magnetic nickel with a target utilization of 50 – 60%.

Stress control in thin films is achieved by varying the degree of energetic particle bombardment during sputtering [3]. Compressive stress is usually attributed to an ‘atomic peening’ mechanism in which reflected neutral atoms bombard the growing film at low sputtering pressures. An increase in sputtering pressure increases the frequency of gas phase collisions, reducing the kinetic energy of sputtered neutral atoms and reflected neutrals bombarding the growing film. This reduction in ‘atomic peening’ reduces compressive stress. The sputtering pressure at which a stress reversal from compressive to tensile occurs increases with the atomic mass of the metal being sputtered.

The majority of metal films deposited by sputtering are in tensile stress. RF Bias of substrates during deposition increases bombardment of the film with energetic ions and causes a stress reversal from tensile to compressive.

Data are presented for stress and resistivity in Ni and NiV (7 wt.% V) deposited at different sputtering pressures and with different bias voltages applied to the substrate during deposition. Microstructural analysis of the films was performed using Field Emission Scanning Electron Microscopy (FESEM), Atomic Force Microscopy (AFM) and Focused Ion Beam (FIB) to determine the correlation between the microstructure and the macroscopic film properties (stress and resistivity).

Experimental Methods

All films were DC sputter deposited using the NEXX Systems Nimbus 310 on 200 mm silicon wafers. Substrate bias in the range 0 to 200 voltswas used during deposition of some films. RF substrate bias is applied to a stationary electrode in the Nimbus to which the wafer tray is capacitively coupled. The sputtering pressure was varied from 5 to 2 mTorr to reduce the tensile film stress. Nickel films were sputtered using a Loop magnetron. NiV, which is non-magnetic was sputtered using a conventional magnetron.

Film stress was calculated from substrate curvature measured before and after deposition using a Tencor Flexus laser deflection technique. The sheet resistance of films was measured using the Prometrix Omnimap four-point probe automated wafer-mapping tool. Film thickness was measured using a Tencor Alpha-Step profilometer. Field Emission Scanning Electron Microscopy (FESEM), Atomic Force Microscopy (AFM) and Focused Ion Beam (FIB) analysis of films was performed by Analytical Answers, Inc.

Results and Discussion

Figure 1 shows film stress versus thickness for Ni and NiV films. RF Bias is used to reduce the large tensile stress in NiV. Nickel films deposited without RF Bias have very low stress.

Figure 1: Stress Vs Thickness for Ni and NiV

The increase in film stress with film thickness in Figure 1 is due to the increase in thermal stress with deposition temperature. Figure 2 shows the deposition temperature and stress versus film thickness.

Figure 2: Deposition Temperature and Stress Vs. Thickness for Ni Films

The intrinsic film stress was calculated by subtracting the thermal stress from the total measured film stress. Figure 3 shows the low stress in nickel (3000 Å) is reduced with increased substrate bias.

Figure 3: Stress Vs bias for Ni (3000 Å)Films

Figure 4 shows how sputtering at lower pressure reduces the stress in both Ni and NiV. The effect is not large due to the low atomic mass of Ni and NiV.

Figure 4: Stress Vs Pressure for 3000 Å) thick films of Ni and NiV.

Tables I and II show resistance data for the nickel and NiV films with stress values plotted in figures 3 and 4. Ni with 7 wt.% vanadium has a resistivity approximately five times higher than nickel.

Table I: Resistance Vs Bias Voltage for 3000 Å thick films of Ni and NiV.

View Ni Graph     View NiV Graph

Resistance increases in both Ni and NiV with increasing pressure and substrate bias. The increase is less than 5% in both cases.

Table II: Resistance Vs Pressure for 3000 Å thick films of Ni and NiV.

View Ni Graph     View NiV Graph

RMS roughness measured using AFM is shown in tables III – V as a function of Bias voltage, Pressure and film Thickness.

Table III: RMS roughness Vs Bias Voltage for Ni (3000 Å) deposited at 3.7 mTorr.

View Graph

Table IV: RMS roughness Vs Pressure for Ni (3000 Å)

View Graph

Table V: RMS roughness Vs Thickness for Ni with and without substrate bias.

View Graph

Increased substrate bias and lower sputtering pressure, both reduce surface roughness and stress. The reduction in tensile stress is due to the increase in energetic particle bombardment of the films during deposition, resulting in a more dense microstructure.

In the nickel film with substrate bias, the grain size is 30 – 50 nm over the first few thousand angstroms and increases with increasing film thickness to a few hundred nm at the surface of the film. This microstructure is similar to that of NiV films sputter deposited with and without substrate bias [4]. In the nickel film without substrate bias, the microstructure is dramatically different. Single grains of 0.5 – 1 mm are visible, with no nanocrystalline grains.

Conclusions

Nickel films 0.25 – 2 mm thick, sputter deposited using an innovative new magnetron designed for sputtering of magnetic materials have low (50 – 450 MPa) film stess without the use of ion bombardment during deposition. The absence of vanadium and substrate bias result in grain sizes of 0.5 – 1 mm, making sputtered nickel a good candidate as a replacement for NiV, as a UBM layer for use with lead-free solders.

Acknowledgements

SEM, AFM and FIB analysis were performed by Analytical Answers, Inc. in Woburn, Masachusetts

References

[1] K. Zeng and K.N. Tu, Materials Science and Engineering, R 38, No.2, 14 June 2002.

[2] Gencoa [V.Bellido-Gonzalez, D.P.Monaghan, Intl patent application 0126721.0, 7 Nov 01.

[3] J. A. Thornton and D. W. Hoffman, J. Vac. Sci. Technol. 14, 164 (1977).

[4] K. O’Donnell, J. Kostetsky and R.S. Post, Flip Chip Technology Conference, Austin Texas 2002.

For More Information

A full version of the original paper, including 24 photographs not included here, is in the proceedings of FlipChip Technology Workshop 2003, June 15 – 18, Austin, Texas.