Target Solves Step Coverage Issues

Metals and alloys have been used for sputtering target materials in microelectronic applications including contacts, vias, interconnections, barrier layers for copper diffusion, and copper seed layers for copper electroplating.¹ One of the most challenging problems is achieving conformal step coverage for small vias and interconnect lines.
 
Various methods have been developed to obtain collimated beams, including placing a physical collimator between the target and substrate to screen off the off-normal beams,^2,3 extending the target-to-substrate spacing to deposit only the near-normal beams, and using an inductively coupled plasma for enhanced ionization or an ion metal plasma (IMP) to direct ionized atoms straight on to a biased substrate.^4-6 Recently, self-ionized plasmas (SIPs) have been used in conjunction with extended target-to-substrate spacing to enhance collimation.^7,8 All these methods, however, require extensive system modification at high cost. The complicated systems also require more control parameters in processing and more frequent system maintenance. Therefore, methods for providing the desired conformal coverage have been sought with associated simplicity and low cost of ownership.

Good step coverage is associated with well-collimated sputtered beams, not perfect collimation, which may result in poor step coverage. In this work, desired beam collimation is achieved by tailoring the alloy microstructure, specifically the grain size of a target. The physical principle is based on the fact that the atoms sputtered off from recessed grain boundaries are more focused than those sputtered from a flat grain surface, akin to a focused light beam emerging from a concave cup in a flashlight despite the multiple scattering of light.

The fraction of collimated beams increases by introducing more grain boundary grooves or by refining grain size. One way to check the magnitude of collimation is to measure the deposition yield, namely deposited mass per energy (kW-h) on the wafer. The higher the fraction of off-normal beams, the less material that will deposit on the wafer. The other method is, of course, to check the step coverage. This work presents experimental data that demonstrates the effect of grain size on beam collimation by employing both methods.

We also performed atomic force microscopy (AFM) examination, which indicated that beam collimation is largely caused by the focusing power of the grain boundary grooves that develop after target burn-in. Optimum focusing power is determined by the grain size that delivers a maximum well-defined groove line density. We address various methods of making such targets.

We fabricated sputtering targets with various grain sizes to investigate the effects of grain size on deposition yield and step coverage. The target materials were Honeywell 6N grade copper and three copper alloys, each containing 0.5 atomic percent of aluminum (at.%Al), silver and titanium, respectively. Grain size was measured by a lineal intercept technique described in ASTM E112 standard test method. Among the four copper alloys, 6N copper showed the coarsest grains (30-120 µm), with grains larger than 100 µm occupying a significant portion of target surface area. Copious twins characterized the microstructure of 6N copper and Cu-0.5at.%Al alloys. Average grain size of CuAl and CuAg was in the range of 30-40 µm, and that of CuTi was 60 µm (Fig. 1 ).

1. 6N grade copper and then copper alloy targets (each with 0.5 atomic percentage aluminum, silver and titanium) were fabricated to investigate effects of grain size on deposition yield and step coverage.

The other set of copper targets include 6N grade APX copper and ECAE copper having average grain size of 50 and 17 µm, respectively. APX and ECAE are process identifications. The fine grain size of the latter target was produced by employing a Honeywell-patented Equal Channel Angular Extrusion (ECAE) method. These two post-mortem targets were used in evaluating the effect of grain size on target erosion.

2. Two base alloys, 3N5 grade tantalum and 5N5 grade Ti-5at.%Zr with grain sizes of 50 and 10 µm, respectively, were sputtered in an Endura IMP chamber, but without activating the RF coil or pedestal bias power.

A third set of targets includes two different base alloys, 3N5 grade tantalum and 5N5 grade Ti-5at.%Zr having an average grain size of 50 and 10 µm, respectively (Fig. 2). Table 1 summarizes the alloys, their grain sizes, process conditions and the items examined.

Copper, tantalum and TiZr targets were made in Applied Materials' Endura IMP configuration (13.578 in. surface diameter). Copper films were deposited in a P5500 Endura IMP chamber without activating either the RF coil or pedestal bias power. This particular arrangement was employed to capture the true distribution of sputtered atoms without the effect of ionization by the coil and bias electric field of the pedestal. A nominal target-to-substrate spacing was 140 mm. The process chamber pressure was in the range of 20 mTorr at a 58 sccm argon flow rate. Typical base chamber pressure was in the range of low 10^-6 Pa (>10^-8 Torr). All depositions were carried out at ambient temperature on 200 mm diameter wafers. We measured the weight of the wafers before and after deposition to determine the weight of the deposited films. Deposition yield was calculated by the film weight per unit input energy (g/kW-h). All comparison depositions were made by using the targets with the same magnitude of target burn-in histories (15 kW-h) to equalize effects from target erosion.

For the APX and ECAE copper targets, the eroded surfaces were examined using AFM to evaluate the nature of erosion grooves that developed by sputtering.

Grain size effect on step coverage for the tantalum and TiZr targets was examined by first depositing nitride films by reactive PVD in a conventional Widebody Endura chamber at a nominal 55 mm target-to-substrate spacing. The deposition was carried out at ambient temperature with 4 kW power and at 4 mTorr chamber pressure with 55 sccm process argon and 75 sccm N₂ under separate mass flow controllers. In addition, a tantalum film was deposited in the IMP chamber at 140 mm target-to-substrate spacing. In this run, deposition was done at 4 kW DC power with 2.5 kW RF and 400 W bias power to enhance ionization and step coverage. The nitride film was prepared by reactive sputtering at 200°C under 20 mTorr argon (28 sccm) and N₂ (28 sccm) mixture. Step coverage was examined by SEM for vias with various sizes.

Deposition yield for copper alloys was determined by measuring the film weight on a wafer that was deposited at 2-6 kW power with 1 kW increment. By reducing the deposition time progressively from 1800 to 600 seconds with increasing power from 2-6 kW, each wafer gained ~1.2 g of copper film. Fairly thick films were deposited to warrant the accuracy in determining the deposition yield. Figure 3 summarizes the deposition yield for four copper alloys. Deposition yield increased initially with increasing power and tended to stabilize after 4 kW. A notable finding was that the deposition yield was 4-8% higher in the entire range of power for the finer grain size Cu-Ti, Cu-Ag and Cu-Al targets compared with that of the large grain size 6N copper target, showing a clear correlation between grain size and deposition yield.

3. Deposition yield tends to stabilize after 4 kW. Yield is 4-8% higher for the finer-grained copper alloys.

The surface profile of eroded copper target was examined by AFM to see the nature of grain boundary grooves that developed during target usage. We obtained AFM micrographs from the cut pieces of spent APX and ECAE copper targets. Average width of the erosion groove for the APX target was ~7 µm wide for the 50 µm grain size target. This suggests that erosion grooves would overlap if grain size became smaller than this width, namely 7 µm. In a 17 µm grain ECAE copper target, ~10 µm diameter craters developed instead of erosion grooves. This crater diameter is about half of the grain size, hinting that the craters might have formed at triple junctions of grain boundaries where the atomic packing density is loosest and thus erosion is expected to be highest. Of significant implication is that, if grain size became smaller than 10 µm, it would be difficult to form well-defined grooves or craters.

5. Looking at sidewall coverage, the smaller grain size TiZr target delivered visibly better coverage in regard to the total thickness of deposited film.

Finally, in the third set, the performance of two dissimilar alloys with two different grain sizes were compared — tantalum and Ti-5at.%Zr. Since sputtering yield varies with the atomic mass of target element, only step coverage was examined for these two alloys. Both tantalum and TiZr nitride films were deposited at 55 mm target-to-substrate spacing in a Widebody chamber. In addition, a TaN film was deposited at 140 mm target-to-substrate spacing in an IMP chamber with both RF and bias power on to see the effect of ionization and bias field on deposition.

Figure 5 compares the step coverage for TaN and TiZrN films. The smaller grain size TiZr target delivered visibly better step coverage when comparing the sidewall coverage in regard to the total thickness of deposited film. In Figure 6 , the step coverage of TiZrN film deposited by a conventional DC sputtering in a Widebody chamber is compared with that of TaN film deposited by ionized metal plasma in an IMP chamber. It is apparent that the smaller grain size TiZr target rendered better step coverage than IMP tantalum, hinting that better step coverage can be achieved via grain size control alone without the aid of ionized metal plasma and bias field.

Discussion

Thus far, we have given several examples that demonstrate the effect of grain size on deposition yield and step coverage. A general consensus is that there is an optimum grain size that delivers higher deposition yield and better step coverage. AFM study revealed that the focusing power of the grain boundary grooves is largely responsible for the improved deposition yield. Since the atoms sputtered off from the recessed grain boundaries are more focused, the depth and width of the grooves are considered to be important parameters in determining the focusing power. AFM revealed that the groove width in the conditioned copper target was in the range of 7 µm (Fig. 4 ). Thus, well-defined grooves may not develop in an alloy with grain size smaller than this width.

6. Even with a thicker film and the aid of IMP and bias field (left), the TiZrN film exhibited better step coverage because of the grain size of the target.

According to SRIM Monte Carlo simulation based on ZBL stopping model,⁹ the sputtering yield for 500 eV Ar⁺, the number of sputtered atoms per bombarding argon ion, is 2.5 for copper, 0.75 for aluminum, 0.65 for titanium, and 2.4 for silver. Thus, the improved deposition yield for the alloyed copper targets cannot be caused by the minor alloying elements. If it were, the sputtering yield might have been reduced.

4. Average width of the APX erosion groove is 7 µm for the 50 µm target. Rather than grooves, craters formed on the ECAE target surface about half the size of the grains (17 µm), possibly at the triple junction of the grain boundaries.

The second important factor is the groove line density that is in turn determined by the grain size. The larger the grain size, the lower the groove line density and thus the lower the effective focusing power. On the other hand, well-defined grooves do not develop in a target with too small grain sizes as the erosion grooves begin to overlap. Therefore, there is an optimum grain size that delivers the highest deposition yield. Based on the experimental results, the optimum grain size appears to be in the range of 10-50 µm.

One notable finding is that targets can be made to have self-collimating power by simply optimizing the grain size without the aid of a physical collimator nor installing costly devices such as IMP and SIP systems. In fact, superior step coverage was achieved by using a small grain size Ti-5at.%Zr target in a conventional sputtering chamber without the aid of ionized metal plasma nor biased substrate (Fig. 6 ).

Self-collimating targets not only improve the deposition yield and step coverage, but also extend the target and kit life because of efficient usage of target materials by reducing the loss to the sidewalls. In a fine-tuned grain target, beam collimation is assisted by the micro grain boundary grooves and thus macro erosion groove has less impact on beam collimation. Thus, less frequent target-to-substrate spacing adjustment is needed to correct the focusing effect of the erosion grooves that develop along the tracks of rotating magnets in a DC magnetron sputtering system. Besides, fine-tuned grain size target allows the usage of existing systems and thus lends tremendous savings in ownership by avoiding a costly upgrade to an IMP or SIP system. Simple process parameters are another advantage in a conventional PVD system that does not require additional control for coil and substrate bias power as in the IMP process.

Finally, we address potential methods for producing fine grain size targets. Grain size control is done traditionally by a judicious choice of thermo-mechanical treatments (TMTs), namely cold and/or hot rolling followed by thermally activated recrystallization, and recently by dynamic recrystallization that occurs during shear deformation using the ECAE method. Other methods include solute-induced grain refinement in combination with proper TMT as in TiZr alloy, and arc-spray or continuous splat quenching on a chilled mold. The latter method can be used to produce a target in a final shape with fine columnar grains and a high number of grain boundaries. Rapid deposition by high-power PVD or high-current electrochemical plating can be employed to achieve desired grain size and orientation for the copper target, for example.

The need for collimated atom beams has been increasing for microelectronic applications as the critical dimensions are decreasing. In response to this demand, sputtering systems with a physical collimator, ionized IMP and self-ionizing devices have been developed. In this work, we showed that focused atom beams can be obtained simply by tailoring the grain size of sputtering target. This new approach is based on the fact that atoms sputtered off from recessed grain boundaries are more focused and impinge directly onto the substrate. The beam focusing power was evaluated for targets with various grain sizes by measuring deposition yield and examining the step coverage for various via features. AFM examination showed that the average width of grain boundary grooves was ~7 µm for copper target, and well-defined grooves did not develop when the grain size became smaller than the groove width. If grain size becomes too large, groove line density is low and thus, low focusing power. In consideration of these, optimum grain size is thought be in the range of 10-50 µm. However, further systematic deposition study and theoretical modeling is needed to determine the exact criteria for optimum grain size. It should be emphasized that all deposition yield study must be conducted with targets with the same magnitude of burn-in history because deposition yield is very sensitive to the depth of erosion groove, mainly because of the increased redeposition within the groove as targets erode.

Since focused atom beams have less off-normal beams, deposition yield and step coverage were improved, and target and shield life would be extended. It is demonstrated that step coverage performance superior to that of IMP can be achieved by simply adopting a target with optimum grain size. This would allow a considerable cost saving by avoiding system upgrades. Finally, this paper proposed various potential methods of making targets with controlled grain sizes, including arc-spray or continuous splat quenching on a chilled mold. This is certainly a challenge, but would open new ways of making targets.