Liners for tungsten plug applications with long throw and ionized PVD
07/01/2000
D.C. Butler, Keith Buchanan, Stephen R. Burgess, Trikon Technologies Ltd., Newport, UK
N. Urbansky, S. Schmidbauer, Infineon Technologies, Dresden, Germany
overview
The properties of Ti contact and TiN barrier layers deposited by PVD are shown to depend on the target-to-substrate distance. A long throw process, combined with low chamber pressure, extends PVD processes below 0.17mm design rules. Ionized PVD processing extends PVD capability even further.
As integrated circuits are continuously scaled down, the requirements of multilevel interconnect technology become more stringent. At sub-0.5mm device nodes, tungsten plugs combined with aluminum wiring form the main scheme used in interconnect fabrication. In W plug processing, Ti contact and TiN barrier layers are essential prerequisites for successful device manufacture. For Ti, bottom coverage is important to minimize contact and via resistance, and for TiN, sidewall coverage is needed to maintain the integrity of the W fill process.
 Trikon Technologies' single-shot system. |
This work discusses the application of long throw (a coherent non-ionized approach) and ionized physical vapor deposition (PVD) processing to contact level manufacturing in a high volume DRAM production environment. The studies show how these complementary processes extend the application of PVD-only liners to below 0.1mm.
The application of a new in situ ionized PVD (i-PVD) Ti/TiN liner for via level interconnects is also reviewed. Early data are discussed showing how an in situ approach reduces via resistance compared to non-ionized/ionized liner combinations.
For these examples, one major benefit of such approaches exists at the chemical mechanical polishing (CMP) step. When using CMP to remove the W and liner blanket metals, selectivity to the Ti layer is typically high, and the time to remove can be long, even for such relatively thin films. An added complication is that any overpolish on the Ti film can lead to oxide erosion and potential yield loss. Any technique that reduces the necessary Ti thickness can therefore have a direct benefit on yield.
Long throw processing
In long throw processing, the distance between the target and the substrate is increased so that the proportion of near-normal sputter species incidents on the wafer is increased. A first-generation long throw chamber (Trikon's Hi-Fill module) uses a spacing of 245mm for both Ti and TiN processes [1]. This module has been used successfully in volume production, offering similar step coverage to physically collimated chambers, but with superior particle performance and lower cost of ownership. The ability to run clean Ti and TiN sequences in the same module is a particular feature of this type of approach.
 Figure 2. 430mm throw Ti base coverage at a) 0.4mtorr and b) 0.15mtorr. |
When used in advanced W plug applications, liners deposited at the 245mm spacing can show limitations. To minimize contact resistance, Ti field thickness must increase to compensate for the natural decline in base coverage as aspect ratios rise. In response, the extended long throw chamber (called Ultra) is an evolution of the earlier chamber, with target-to-substrate spacing increased to 430mm [2]. Modeling and experiment have shown that base coverage can be significantly improved at longer throws when running at process pressures low enough to minimize gas scattering events. Consequently, the extended spacing is best suited to nonreactive processing, because reactive processes running in Ar/N2 atmospheres typically use pressures in excess of 1mtorr. Consequently, the final appearance of the module uses a long stroke platen to allow 430mm Ti operation, and 245mm TiN operation. Ti and TiN are run sequentially in the same wafer visit, and the target is de-nitrided onto a shutter between wafers.
Effect of throw distance on contact parameters
For all the work discussed here, the contact process consists of Ti-only followed by a rapid thermal processing step to form the TiN.
The 430mm and 245mm Ti processes were evaluated for DRAM and embedded DRAM production using 0.25mm and 0.20mm technology at the silicon contact level. The tests were done on a 200mm Sigma fxP platform and included wafer heating to remove moisture prior to deposition.
 Figure 3. Contact resistance results for 0.17mm contacts using standard and low-pressure processes. |
The contact resistance distribution for 245mm Ti and 430mm Ti deposition is shown for 0.25mm and 0.20mm contacts in Fig. 1. For the 0.25mm example, reducing the Ti thickness at the 245mm mode causes the contact resistance to rise out of control as the thickness of Ti at the bottom of the contact hole falls below the threshold value needed for minimum Rc. At the extended 430mm distance, Rc is low and in control at a field thickness of 15nm, 40% thinner than the 25nm process of record run at the medium distance. Similarly at 0.20mm contact size, running at the longer distance gives controllable characteristics at lower field thicknesses.
Low pressure, long throw processing
The extendibility of long throw processing in terms of base coverage is strongly dependent on the minimum process pressure that can be reliably maintained. New work based on magnetron field design has enabled the process to be run at pressures down to 0.15mtorr, offering near-zero scatter processing and further increasing the proportion of near-normal species arriving at the wafer. This 63% reduction in process pressure results in a 30% increase in absolute Ti base coverage for holes of aspect ratio 2.6:1. Figure 2 shows SEMs of Ti base coverage for 430mm mode processing at 0.4mtorr and 0.15mtorr, taken at the edge of a 200mm wafer. A sloped film profile at the hole base is typical with a long throw approach, caused by shadowing of the hole base by the sidewall out of line of sight of the target. This effect is ultimately expected to determine the extendibility of these techniques, triggering a switch to i-PVD.
For a 0.17mm contact design, compared to 25nm Ti run at the same 430mm distance but at higher pressure, the benefits of low-pressure operation are immediately apparent, as shown in Fig. 3. Running at very low pressures allows a further 5nm reduction to be made in Ti field thickness, compared to the 25nm run at 430mm and 0.4mtorr.
 Figure 4. Schematic diagram of an ionized PVD (i-PVD) module. |
As would be expected, deposition rate falls as the target-to-substrate distance is increased. Properly scheduled, this would have minimal effect on system throughput since the required film is thinner and target power can be increased to compensate. In addition, in the low pressure, long throw configuration, the modifications made to the magnetron field improve electron capture and allow use of target powers in advance of what might be expected at such low pressures. For the liner thicknesses used in this work, the throughput for all module configurations exceeds 38wph. Another advantage of moving to longer distances and lower pressures is the reduction in field nonuniformity, which will further benefit the CMP step.
Ionized PVD for contact level processing
Shrinking beyond 0.17µm design rules eventually compromises the performance of even the most advanced, nonionized PVD technique. Ionized sputter modules are the last word in normalized deposition, offering degrees of base coverage far in advance of any other PVD technique.
 Figure 5. Contact resistance results for 0.19mm contacts using i-PVD and the Ultra system. |
To follow the long throw modules, Trikon has developed an i-PVD module based on the internal coil approach [3]. In this scheme, a high-density plasma is generated between the target and the wafer. At the optimized process conditions, sputtered material is ionized and accelerates towards the biased wafer, such that metal ions arrive at normal or near-normal incidence. The wafer bias can be generated by self-biasing, or by applying a separate biasing supply. With this approach, film coverage at the bottom of deep holes can be significantly increased.
This module features some novel hardware designs that increase process flexibility. In the approach reported here, the ionizing coil is operated in a net deposition mode. That is, it receives deposition from the target, but does not re-sputter it. Therefore, the coil can be made out of material selected on the basis of cost and manufacturability, and can be easily temperature controlled. If the coil is operated in a net etch mode, it has to be made from the same high purity material as the target itself, which can add cost and make attachment of temperature cooling channels to the coil difficult. Figure 4 shows a simple representation of this module.
To test the suitability of this design as the next-generation replacement for the long throw approach, lots with 0.19mm contacts to Si were processed through the standard pressure Ultra and i-PVD modules. Aspect ratios are still moderate at 2.6:1, so they were still expected to be within the capability of the nonionized approach. As shown in Fig. 5, 15nm Ti deposited in the i-PVD module shows promise, but 20nm gives the lowest, tightest response and compares well to 25nm deposited at the 430mm distance.
 Figure 6. Ionized PVD base coverage in 2.6:1 aspect ratio holes, with a) Ti and b) TiN. |
The conclusion is that at this generation, the i-PVD approach does not offer any significant technology advantages over the long throw modules. When compared to the low pressure Ultra, base coverage is similar at this aspect ratio, as demonstrated by comparable contact resistance at the same field thickness. A more important conclusion, however, is that the i-PVD process window looks to be manageable, i.e., that a change in Ti thickness produces a predictable response in Rc characteristics. While a departure from the long throw may be unnecessary here at the 0.19mm node, the i-PVD technique is expected to be needed and capable at some point after 150nm.
i-PVD extendibility: dual Ti/TiN operation for via applications
null
Although not relevant to the contact process described here, a significant proportion of the i-PVD development work was devoted to developing a module capable of running both Ti and TiN. If a single i-PVD module could be used for both layers, significant reductions in cost of ownership are to be expected.
The novel module design reported earlier allows dual Ti/TiN in the single chamber. The issue lies with minimizing N2 content in the atmosphere at the time of pure Ti deposition. Following TiN deposition, the wafer is removed and a Ti-only paste step is performed onto a shutter to remove N2 from the ambient. Relatively short paste times are required, in accordance with high productivity manufacturing. During the next wafer Ti step, since the coil is run in a net-deposition mode, TiN build-up on the coil is not removed and does not contribute any free N2 to the process atmosphere.
Figure 6 shows base and lower sidewall coverage into 2.6:1 aspect ratio holes for Ti and TiN deposited in i-PVD mode. Base coverage is measured at 71% for Ti and 54% for TiN. In addition, the sloped film profile typical of long throw modules is not observed in the i-PVD mode.
 |
Electrical data taken in later work showed an unexpected benefit of the dual i-PVD Ti/TiN technique. Following W deposition, CMP, plug formation, and Al interconnect, a reduction in via resistance of ~20% was measured in 3.5:1 aspect ratio vias when compared to i-PVD Ti/non i-PVD TiN liners or vice versa. Further, a similar 20% reduction was noted compared to the non i-PVD Ti/TiN liner process of record, using film stacks 60% thinner. When transferred to full production, such a reduction in field thickness can be expected to significantly benefit fab output by cutting CMP time and further improving device yield.
Comparison of long throw and i-PVD data
Moving to very long throw distances will affect basic film parameters in addition to actual base coverage. The table compares basic parameters for the three configurations; at the 245mm spacing and 0.8mtorr, 430mm spacing and 0.4mtorr, and 430mm spacing and 0.15mtorr. Representative i-PVD data is added for comparison.
Conclusion
Working at a very long throw distance in PVD processing enables significant reductions in Ti field thickness, resulting in reduced CMP times, and corresponding benefits in the degree of interconnect dishing and oxide erosion. Moving to very low process pressures extends the application of such techniques to <0.17mm design rules.
To follow the long throw modules, an i-PVD Ti process was shown to have capability at contact level. First, data was discussed for the application of an in situ i-PVD Ti/TiN module to via level interconnects. Up to 20% reduction in via resistance has been observed for very thin Ti/TiN layers compared to liners deposited up to 2.5 times thicker by conventional means.
Long throw and ionized PVD modules are extendible beyond the nodes previously believed to be the limit of all-PVD liner schemes. Dual function, all-PVD Ti/TiN approaches such as those presented here offer familiar, low maintenance, cost effective alternatives to CVD-based liner processes.
Acknowledgments
Hi-Fill is a registered trademark of Trikon.
References
- D.C. Butler, P.J. Holverson, P. Rich, J. Hems, G.A. Dixit, S. Poarch, R.H. Havemann, "Enhanced Bottom Coverage of Sub-Micron Contact Holes using a Novel Hi-Fill Ti/TiN Sputter Process," Advanced Metallization for ULSI Applications in 1994, pp. 277-284, 1994.
- S.M. Rossnagel, J. Hopwood, " Metal ion deposition from ionized magnetron sputtering discharge," J. Vac. Sci. Technol. B 12(1), pp. 449-453, Jan/Feb 1994.
- N. Urbansky, M. Harris, D.C. Butler, P. Rich, K. Buchanan, C. Goergens, "Advanced Long Throw PVD for Contact to Silicon and Via Applications," Advanced Metals Conference 1999, proceedings yet to be published.
David Butler is the technical marketing manager for PVD at Trikon Technologies. After completing his BSc in physics at Loughborough University (UK) in 1983, he worked at Pilkington plc, depositing coatings onto glass and plastic substrates. He joined Trikon as a process engineer in 1989 before taking up his current role in 1995. He has written and presented widely on Trikon's PVD technologies, including liner deposition and Forcefill high-pressure Al plug fill. Trikon Technologies Ltd., Ringland Way, Newport NP18 2TA UK; ph 44/-1633-414058, fax 44/-1633-414180, email [email protected].
Keith Buchanan graduated from The University of Glasgow with a degree in electronics and electrical engineering. He joined Electrotech's PVD group in 1994 and is currently process integration manager.
Stephen R. Burgess received his MA in solid state and atomic physics from Worcester College Oxford University in 1993. He received his PhD for research into semiconductor heterojunctions in 1996. He is currently metallisation process engineer at Trikon Technoloigers.
Norbert Urbansky studied electrical engineering at the Dresden University of Technology where he received his engineer's certification in 1978 and defended his dissertation in 1984. From 1983 to 1998, he worked at the Institute for Semiconductor and Microsystems Technology at the Dresden University of Technology in the area of bumping technologies for flip chip applications and advanced interconnection technologies. In 1998, he transferred to the Metal Group at Infineon in Dresden, where he works on liner metallizations.
Sven Schmidbauer received his MS degree in physics from the University of Chemnitz, Germany, in 1991. He then joined a research group in solid state physics at the University of Chemnitz working on vacuum arc evaporation for hard-coating applications. Since 1994, he has been working on liner and AlCu-based metallizations and is now heading the metal group at Infineon in Dresden.