High-throughput W/ Ti barrier sequential deposition
06/01/1998
High-throughput W/Ti barrier sequential deposition
Thierry Castan, Motorola Semiconductor, Toulouse, France
Jean-Luc Peyre, Klaus Beschorner, Tokyo Electron Limited, Europe
A newly developed in situ W/Ti deposition process, using a pulse of dry air between sequential deposition steps, achieves the same barrier properties as those from conventional processing with two separate systems. Tests show that such sequential depositions can be done using currently fielded process equipment.
As device dimensions shrink, junction depths must also be reduced. Such shallow junctions require metal contact barrier layers that prevent "spiking" of aluminum through the junction. Spiking can occur during thermal cycles that commonly follow metal deposition.
Use of a self-aligned silicide (or "salicide"), such as platinum silicide (PtSi), guarantees extremely good metallurgical contact to the silicon substrate in the junction region. Salicides serve either as high barrier Schottky or ohmic contacts to shallow junctions in fabricated devices. However, a silicide layer alone is an insufficient diffusion barrier to aluminum spiking. When the temperature exceeds 350?C in the course of final device fabrication, the silicide requires an additional barrier layer to protect it from the aluminum.
 |
Figure 1. IC contact structure using TiW as a diffusion barrier to prevent reaction between PtSi and AlCu.
Titanium tungsten (TiW) is often used as a diffusion barrier between a PtSi contact to silicon and an aluminum alloy interconnect. A typical ohmic contact structure (Fig. 1) incorporates a PtSi layer (50-100 nm thick) in direct contact with heavily doped Si regions, covered with a sputter-deposited TiW layer (100-200 nm thick), and a final layer of aluminum copper alloy (AlCu) on the TiW. When the metal is patterned, the TiW remains under the Al layer and becomes part of the final interconnect structure. Most semiconductor manufacturers use two separate pieces of equipment for depositions of TiW and AlCu. Exposure to air between the two layers (termed an "air-break") provides the oxygen and nitrogen (O2 + N2) necessary for grain boundary stuffing to block rapid diffusion paths (see, "Grain boundary stuffing" on p. 128). Our work investigated barrier layer integrity and efficiency by varying process conditions during and between depositions of the two layers.
Unstuffed TiW barrier failure and subsequent junction spiking can occur during sintering of the contact, destroying the device. Silicon dissolves, reacts with TiW to form WSi2 and TixW1-xSi2 compounds, and migrates into the AlCu line. Silicon that is consumed in this reaction series is replaced with aluminum that penetrates the junction in a spike (interdiffusion mechanisms).
Experimental procedure
All experiments used a standard MRC Eclipse system with four isolated processing chambers, including pre-etch, TiW and AlCu sputtering chambers. This system achieves wafer heating via the injection of argon through the backplane. Our TiW deposition process was done at 150?C to ensure stable compressive film stress. AlCu sputtering was done at 200?C.
An in situ RF etch removed any residual contamination or oxide on the wafer surface prior to TiW deposition. TiW (1500 ?) and AlCu (4500 ?) films were sputter deposited on <100> and <111> Si wafers with different process conditions between the two layers, as outlined in the table. All samples were annealed for 2 hrs at 450?C in a conventional furnace, and then the two metal layers were wet etched. Final silicon surface inspection identified any etch pits due to aluminum penetration through the TiW barrier.
 |
Figure 2. SEMs of a typical defect on an a) <100> wafer and b) <111> wafer.
The shape of these pit defects corresponds to the crystallographic axis of the wafers: rectangular for <100> wafers, triangular for <111> (Fig. 2). To confirm that the detected defects represent aluminum spiking through the boundary layers, we used x-ray analysis to detect aluminum in the silicon at defect points.
 |
Results and discussion
First, the experiments confirmed the barrier quality of the standard air-break process (see table): an air exposure as short as 5 min is sufficient to stuff the grain boundaries. The barrier integrity is fundamentally improved by blocking rapid diffusion paths.
We performed secondary ion mass spectrometry (SIMS) to study the TiW/Al interface. SIMS detects the presence of oxygen and nitrogen atoms incorporated in the film by progressively etching layers and detecting their elementary make up. We clearly observed nitrogen and oxygen peaks at a depth of 0.45 ?m on the recorded profiles of reference wafers (Fig. 3); this depth corresponds to the interface between the aluminum and the underlying TiW metal.
 |
Figure 3. SIMS graph produced with the standard air-break between TiW and AlCu depositions.
These peaks are rather large because of the roughness of the surface when the ion beam reaches the interface. The quantity of atoms detected can be determined by calculating the area under each peak. A typical dose of 1.6 ? 1017 oxygen atoms/cm2 was thus measured for the reference wafers in the air-break process sequence.
As expected, and generally admitted [1-4], it is not possible to form a proper barrier layer using a direct sequential deposition of barrier and AlCu films (see table, group 1). Indeed, the unstuffed barrier produced 150,000 defects/cm2.
 |
Figure 4. SIMS graph for direct sequential deposition without in situbarrier stuffing. The nitrogen and oxygen peaks are very low, coming from residual background gases.
During the subsequent annealing step at 450-500?C, aluminum can react with pure TiW to form various compounds, such as TiAl3. Aluminum can then diffuse through grain boundaries toward the silicon substrate.
No barrier stuffing is possible with residual gas contaminants inside a PVD system; SIMS signals are very weak for oxygen and nitrogen species (Fig. 6). We noted that the oxygen dose was about three times lower than was the case with the reference wafers that had been exposed to air. These results are also correlated with electrical measurements on n-si/PtSi/TiW/AlCu Schottky contacts for which shifted characteristics were observed after high temperature stabilization.
 |
Figure 5. Sequential processing without an air-break in the PVD system. The gas-flash chamber provides sufficient material for TiW barrier stuffing prior to AlCu deposition.
A nitrogen gas flash just before deposition of the aluminum layer in the second chamber of the PVD system (Fig. 7) was clearly beneficial. The defect densities (see table, groups 2 & 3) were reduced by 10? at 200?C, and by 100? at 400?C, indicative of the thermal activation of the reaction due to the backplane gas heating. However, when aluminum was deposited in the next chamber after the 400?C N2 gas flash (see table, group 4) the defect counts rose back to the levels seen in the 200?C N2 gas flash (group 2). This increase in defect levels could be explained by nitridation of the aluminum target, which probably leads to the deposition of a preliminary nitrided aluminum layer.
In the final experiment, a similar gas pulse of dry air (20% oxygen in nitrogen) injected at 400?C replaced pure nitrogen to enhance the oxidation of the TiW (group 5). The dry air pulse produced excellent barrier characteristics, similar to the reference air-break process; no defects were identified on the silicon surface.
The excellent results obtained with the dry air pulse process show that an oxygen-containing ambient is clearly needed for in situ grain-boundary stuffing; nitrogen alone is insufficient. The duration of the gas flash does not appear to influence the result strongly, and was finally fixed at 45 sec. SIMS analysis confirmed that the oxygen and nitrogen peaks are close to those obtained in reference wafers (oxygen dose of 1.3 ? 1017 atoms/cm2). The processing sequence used for groups 4 and 5 allows us to maintain high machine throughput, as long as the duration of the gas flash does not exceed 60 sec. Complementary characterizations show that film resistivity, reflectivity, and stress are not significantly affected by the gas flash.
Conclusion
This work evaluated a sequential PVD process for deposition of interconnect materials required for advanced bipolar technology. We identified a promising integrated process for sequential deposition of barrier and aluminum layers without the need of reactive sputtering. Specifically, an in situ N2/O2 gas flash in an intermediate chamber between the two deposition steps eliminates the usual air exposure required for barrier stuffing. The system productivity is clearly improved, reaching, in the case of this work, a factor of 2?, while obtaining high uniformity, controlled stress, and low particle levels.
Acknowledgments
The authors thank Murielle Delage of Motorola for the extensive work on this study, and Marc Nougaro and Jean-Luc Hamon of Tokyo Electron Limited for the maintenance of the system.
References
1. V. Hoffman, "Tungsten/titanium diffusion barrier metallization," Varian Semiconductor equipment operations, Report No. 33.
2. B.S. Chiou, H.S. Lo, P.H. Chang, "Barrier effect of e-beam evaporated tungsten interlayer in Al/W/PtSi metallization layer," Journal of Electronics Materials, Vol.17, No. 5, 1988.
3. C. Canali, F. Fantini, E. Zanoni, "Electrical degradation of n-Si/PtSi/(Ti-W)/Al Schottky contacts induced by thermal treatments," Thin Solid Film, Vol. 97, No. 4, Nov. 1982.
4. M.A. Nicolet, M. Bartur, "Diffusion barriers in layered contact structures," J. Vac. Sci. Technol., Vol. 19, No. 3, Sept./Oct. 1981.
Thierry Castan received his engineer diploma from Conservatoire National des Arts et M?tiers de Toulouse, France. He joined Motorola in 1981 and has worked in epitaxy, ion implantation, and thin films.
Jean-Luc Peyre received his engineer diploma from the Institut National Polytechnique de Grenoble, France, and a DEA from Scientific University in 1985, all in physics of components. Over the last 12 years, he has been involved in excimer laser stimulated processes, photonic integrated circuits development, plasma etching, and deposition. He joined the semiconductor equipment division of MRC Europe (now Tokyo Electron) in 1996 as a process engineer in metallization technology. Tokyo Electron Ltd., fax 39/39-605-8004, jeanluc_peyre@
compuserve.com.
Klaus Beschorner received his MS degree in physics from Stuttgart University in 1988. He joined IBM Germany as a process engineer in 1990 and developed etch, laser ablation, and sputtering manufacturing processes and equipment. In 1995, he joined MRC`s Semiconductor Equipment Division (now Tokyo Electron Limited) to head their PVD field process support in Europe.