Sputtering Target With Improved Ring Contact Interaction

Monolithic aluminum targets and targets bonded to aluminum alloy backing plates have several advantages over copper alloy backing plates. The most important benefits are improved cost of ownership, potential to reduce grain size, ease of handling, and capability for long life.
 
With further development of 300 mm and system-in-package (SiP) process working at high power, the requirements for target assembly mechanical performance are constantly escalating. With increasing target diameter and applied sputtering power, mechanical stability becomes a very important performance factor. The target assembly is subjected to high temperature on the target face and cooled by water flow on the backing plate, which creates a temperature gradient. The target assembly separates the cooling water from the vacuum chamber, thus generating a pressure differential. The pressure and temperature gradients cause the assembly to deflect away from the magnet. The target deflection should be minimized to maintain good uniformity and overall performance.

The monolithic target design uses high-purity metal without bonding to a backing plate so there is no risk of target separation from a backing plate. But because of low tensile strength of pure metals, it may deflect significantly during sputtering. Another approach is to join a metal target to a backing plate made of material with increased yield and ultimate strength.

In this article, we will present improved target design and target performance results for aluminum, and copper target assemblies with improved interaction with the ceramic ring (T-Mod). One of the existing problems of all known target assemblies, especially those used in 300 mm systems, is undesirable interaction with the ceramic ring resulting in scratching and damaging the ceramic ring. Because of the scratches shown in Figure 1a , the ceramic ring should be cleaned after target use and replaced more frequently, which causes production downtime and increases the cost of ownership.

1. Target has black marks on the flange after sputtering (a). Target with no black marks on the flange used T-Mod pins (b).

Experimental

Metal targets, 300 mm Al-0.5%Cu, and 5N copper used in this experiment were production targets with modified backing plate flanges to accommodate plastic pins. Targets with solid T-Mod pins were machined with the usual tolerances. Under experimental conditions, it was found that the optimal height of the pins was ~0.01-0.015 in. above the flange. In this work, solid low-friction polymer was used to make pressure pads or pins; however, any other suitable material can be used.

Pins were positioned into the machined grooves under cleanroom conditions so that the targets could be packed and shipped to customers, and were ready for use. Metal deposition was conducted using 300 mm Applied Materials' sputtering systems at Infineon and 200 mm Endura 5500 system at Tosoh Application Laboratory. Film quality throughout the target lifetime was checked for sheet resistance uniformity and particles.

The objective of this design was to create reliable technology, providing high-temperature stability of the assembly during sputtering without scratching the ceramic ring. The proposed method was based on positioning the T-Mod inserts at the locations where most deformation of the backing plate happens during sputtering. The proposed assembly is shown in Figure 2 .

2. The geometry of simulation, target, backing plate and ceramic ring are shown.

FEA analysis of target deformation

Finite element analysis (FEA) was performed to evaluate the thermal and structural performance of 300 mm Al-0.5%Cu targets for Applied Materials' 300 mm systems. The model assumed that the target and backing plate were in mechanical contact by a proprietary Tosoh bonding process (Forte), and can separate at the grooves if sufficient parting force is encountered. Also, the heat transfer was allowed only at the locations where the target and backing plate were in contact. The cooling water was at 21°C and 35 psi pressure, with a convective heat-transfer coefficient of 0.262 W/cm²•°C.

The geometry for the simulation was generated from an Inventor CAD model (Fig. 2 ). The assembly includes an Al-0.5%Cu target, Al6061 or Al7075 alloy backing plate, alumina ceramic ring, and five pins (T-Mod).

Model geometry

Simplifications were made to the model to reduce the amount of computational power needed to get an overall idea of how the assembly interacts during sputtering. The first simplification was to take advantage of the part's cylindrical symmetry and only model one quarter of the entire part. A second simplification removed the O-rings that are on either side of the ceramic ring and replaced them with a given displacement of the ceramic ring. This simplification greatly reduced the computational load, although it is by no means an insignificant interaction during actual sputtering.

The target's sputtering surface is based on the erosion profile of a spent target, and assumes the target is at half-life during this analysis.

Contact — There were two types of contacts used in this model: bonded and frictionless. A bonded contact assumes that the contacting faces of the two parts are rigidly connected to each other; the examples of “bonded contacts” are T-Mod pins in slots (where T-Mod pins are positioned) and the target backing plate. The “frictionless” contact slides freely on each surface and only transfers forces normal to those surfaces. The examples of frictionless contacts are between the T-Mod pins and ceramic ring and between the ceramic ring and O-ring groove.

Mesh — The mesh contains 45,949 nodes and 24,635 elements. The smaller the mesh, the higher level of accuracy of the simulation and the larger the calculation becomes. Mesh refinement was only used for the contacts between the T-Mod pins and ceramic ring.

Environment

The model used a static thermal stress analysis. A steady-state thermal solution was found, and those results were reentered into a static structural model in which additional supports and loads were added to determine the overall solution.

Thermal loading — The thermal loading was a steady-state solution consisting of heat generated from the sputtering power (15 kW) on the target face and the cooling water (21°C) on the backside of the backing plate. A transient thermal solution, which is the case in the actual sputtering process, could be performed if the sputtering power on and off cycles were known. To further simplify the model, the water was assumed to be stagnant on the backside of the backing plate and all contacts transferred heat with 100% efficiency.

Structural loading — The structural loads on the model were the water pressure on the backing plate, the pressure difference of 1 atm caused by a vacuum on one side, and the compression of the O-ring (displacement of the ceramic ring). The 1 atm (0.101 MPa) of pressure was applied to the surface that was exposed to vacuum, and the circulating cooling water applies a 0.21 MPa pressure on the backside of the backing plate.

Compression of the O-ring was not modeled; instead, the ceramic ring was given an initial displacement to compress the T-Mod pins to within 0.027 mm of the backing plate flange. This assumption was made to get a baseline comparison between the model with and without T-Mod pins, not knowing how close the actual ceramic ring comes to contacting the target flange.

Structural supports — The structural supports needed for this model were used to simulate the symmetry boundary condition, fix bolt holes, and provide additional support for the ceramic ring. The symmetry boundary condition constrains the model to zero displacement normal to the surfaces. This was applied to the target, backing plate, and ceramic ring. There were five bolt holes that had a zero displacement constraint applied to simulate the mounting of the target in the sputtering system. The ceramic ring was supported on the inside diameter (ID) and outside diameter (OD) surfaces to allow translation only in the Y axis, which is normal to the sputtering surface. This constraint was added to help with the convergence of the static model.

Solution

“Solution” contains the calculated response for the geometry of simulation described above and the given loading conditions defined in “Environment.” Thermal expansion calculations use a constant reference temperature of 22.0°C for all bodies in the model. Theoretically, at a uniform temperature of 22.0°C, no strain results from thermal expansion or contraction.

Structural results — The results of the analysis showed that the addition of the T-Mod pins reduced the contact pressure of the ceramic ring on the OD of the O-ring groove. This is better explained in the context of how the entire system responded to the loads that were applied. The maximum target deflection was 2.27 mm and occurred at the center of the assembly, which was expected. Figure 3 shows the stress on the five fixed bolt holes, which is where the maximum occurs. This stress contour was not obvious before the simulation, and we used the advantage of the three-dimensional analysis to show the material's response between the mounting holes. Isolating the O-ring groove and increasing the scaling factor of the deformation showed that, under the current loads, the OD of the groove deflects between 0.077 and 0.171 mm, which is greater than the initial gap of 0.027 mm. Figure 4 is a plot of total deformation at an elevated scale factor to better visualize the deflection. The underformed model is outlined by the solid black lines for reference. Once this deformation pattern is observed, it is understandable how the OD of the O-ring groove contacts the ceramic ring and produces scratch marks observed after target use (Fig. 1 ).

3. The equivalent stress contours on the five fixed bolt holes.

4. A plot of total deformation at an elevated scale to better visualize the deflection. The under-formed model is outlined by the solid black lines for reference.

The contact pressure between the target flange and ceramic ring was reduced when the T-Mod pins were included in the assembly. For comparison, the contact pressure between the target flange and ceramic ring was plotted with and without T-Mod pins in Figure 5 . The plots show the wire frame of the geometry and flange surface that have a contact pressure >11 MPa (colored region). The figures show that there is less contact pressure by the mounting holes in both cases. In addition, the T-Mod pins reduced the value of contact pressure between the mounting holes.

5. The contact pressure between the target flange and ceramic ring with (top) and without (bottom) Teflon pins.

Thermal results — The steady-state solution between the sputtering power of 15 kW and cooling water of 21°C found a maximum temperature of 76.68°C. Further understanding of how the target flange and ceramic ring contact each other could be observed with the following model improvements. Incorporation of the O-rings on either side of the ceramic ring would be a more realistic interpretation of how the T-Mod pins interact with the entire assembly. Modeling the contact area between the target flange and ceramic ring as a frictional surface instead of frictionless would give more realistic surface stress than the current model. A simulation investigating vibration during the sputtering process would also add to the understanding. The refinement of the mesh would be helpful in increasing the overall accuracy of the model.

Sputtering results of 200 mm aluminum and copper targets

A sputtering test at Tosoh included cycling (up to 20 kW power). The flange on the O-ring groove side had 16 holes milled ~0.218 in. in diameter and 0.118 in. deep. The holes were equally spaced between the bolt holes at the same radius.

T-Mod pads ~0.218 in. diameter × 0.133 in. tall were placed into each hole so that T-Mod pins were ~0.015 in. above the flange.

After we placed an O-ring into the target groove to prevent contaminating the groove, the flange was coated from the target OD to the O-ring with red layout ink. Ink was also applied on the Teflon pins (note that this is the part of the target flange that is not in the vacuum). The ceramic ring for the Endura chamber was cleaned and checked for markings on the ceramic ring.

In the next test, we installed the target into the chamber and pumped down to 4 × 10^-7 Torr, then sputtered the target for 10 kW/hr at 13 kW. After sputtering, the system was cooled, then we vented the chamber and took photos to show where the target hit the ceramic ring. In this experiment, we found no red ink on the flange except where the pads were located and contacted ceramic ring.

The T-Mod pins that were 0.015 in. above the target flange did a good job preventing the target from rubbing the ceramic ring. The red layout ink appeared on the ceramic ring only in the area where T-Mod pads were in contact with the ring, which implied that there was no rubbing on the flange. The assembly demonstrated very good reliability during spatter tests, and was recommended for manufacturing.

A 200 mm copper sputtering target was manufactured and sputtered at a customer location. Figure 6 shows the clean flange from the target with T-Mod pins, and the formation of black marks on the flange without T-Mod pins.

6. Copper target after sputtering without (left) and with (right) T-Mod pins.

Sputtering results of 300 mm aluminum targets

Qualification of the T-Mod pin 300 mm concept target was done at IFX300 Dresden on an Endura Classic (Endura1) Cluster Tool. Two Tosoh SMD Al-0.5%Cu Forte bonded to Al6061 targets were mounted into hot AlCu ESC chambers and sputtered to ~1200 kW/hr. No rubbing marks on either the backing plate or the ceramic ring were observed during the qualification runs (Fig. 1b). The pins were still firmly in place, and the backing plate deflection was im-measurable after sputtering. Particle data was generated during tool check twice a week (Anko) by depositing a 240 nm thick AlCu layer onto bare silicon wafers. Particle sizes between 0.16 and 5.0 µm were measured at SP1. The qualification targets both showed excellent particle performance (Fig. 7 ). The mean count of both targets was six particles per wafer, with a median value of three counts in a 26-day timeframe. There were no spec violations (spec <30 counts) observed. Large particle count (>5.0 µm) was one or below for all Anko wafers processed with Tosoh qualification targets. For comparison, four chambers at a different Endura1 mainframe equipped with AlCu targets bonded to a CuCr backing plate and no anti-rubbing interlayer showed an overall mean of 11 particle counts, with a median of nine counts.

7. Particle values of Anko check twice a week. Data was generated by deposition of 240 nm AlCu on bare silicon wafers in an Endura300 classic hot ESC chamber (Endura1) at IFX300 Dresden. The installed target was a Tosoh AlCu, bonded to an Al6061 backing plate with T-Mod pins.