The inventions described below relate the field of wafer carriers used to hold wafers during chemical mechanical planarization.
Integrated circuits, including computer chips, are manufactured by building up layers of circuits on the front side of silicon wafers. An extremely high degree of wafer flatness and layer flatness is required during the manufacturing process. Chemical mechanical planarization (CMP) is a process used during device manufacturing to flatten wafers and the layers built-up on wafers to the necessary degree of flatness.
Chemical mechanical planarization is a process involving polishing of a wafer with a polishing pad combined with the chemical and physical action of a slurry pumped onto the pad. The wafer is held by a wafer carrier, with the backside of the wafer facing the wafer carrier and the front side of the wafer facing a polishing pad. The polishing pad is held on a platen, which is usually disposed beneath the wafer carrier. Both the wafer carrier and the platen are rotated so that the polishing pad polishes the front side of the wafer. A slurry of selected chemicals and abrasives is pumped onto the pad to affect the desired type and amount of polishing. Using this process a thin layer of material is removed from the front side of the wafer or wafer layer. The layer may be a layer of oxide grown or deposited on the wafer or a layer of metal deposited on the wafer. The removal of the thin layer of material is accomplished so as to reduce surface variations on the wafer. Thus, the wafer and layers built-up on the wafer are very flat and/or uniform after the process is complete. Typically, more layers are added and the chemical mechanical planarization process repeated to build complete integrated circuit chips on the wafer surface.
As integrated circuit chips have become more complex, the standards of flatness and planarization necessary to create the integrated circuits and to achieve high yield in the process have become more and more stringent. At the same time, zones of differential response to the CMP process (due to the topography and underlying architecture of the particular integrated circuits built up on the wafer creates) and prime wafer height variations have become significant relative to the degree of flatness required for many processes in the manufacturing process. In prime wafers, high spots and low spots spanning millimeter-scale zones may be randomly located over the surface of wafer, and the height differential may be on the order of 100 nm. In processed wafers, high spots and low spots spanning millimeter-scale zones tend to be predictably and uniformly arrayed across the surface of the in-process wafer. The arrangement of high spots and/or low spots depends on the particular architecture of the built-up wafer, but is generally predictable from run to run, and it appears that the underlying architecture results in soft spots subject to an increase removal rate under CMP, or in hard spots that are resistant to removal of surface material. The excess wear leading to the low spots is referred to a dishing, and it is problematic because, among other reasons, it limits the resolution of lithography and creates high spots subject to thinning. (It may be appreciated that the terms soft-spot or hard-spot do not necessarily refer to the measured hardness of the wafer surface, and refer more generally to resistance or susceptibility to polishing).
The differential polishing described above differs from typical cross-wafer or center-to-edge differential polishing. Center-to-edge differential polishing describes the uniform over-polishing or under-polishing of the edge of the wafer compared to the center of the wafer. CMP processes have known inherent challenges in controlling center-to-edge uniformity. For example, oxide polishing is typically center-slow (the wafer's edge is polished faster than the wafer's center), while metal and prime wafer polishing are typically center-fast. Thus, annular zones of the wafer are polished differently.
The effects of uneven polishing have previously been addressed by adjusting the backpressure applied to different annular zones of wafer during polishing. The wafer is held against the polishing pad by a wafer carrier. The wafer carrier includes a pressure plate which is used to apply pressure on the back side of the wafer. For processes known to result in differential polishing rates, wafer carriers adapted to apply different backpressure to different zones of the wafer have been used. Pressure plates have been modified to provide for the application of different backpressure on the edge of the wafer and the center of the wafer, and in some cases a third concentric annular zone between the center and the edge. The current systems are useful in many processes. However, the soft-spot and hard-spot differential polishing identified above is not addressed by annular zoned backpressure systems, as the differential wear occurs over the entire surface of the wafer, and results in inconsistent polishing in all the annular zones.
The methods and devices shown below provide for precisely controlled, millimeter scale (or smaller) adjustments to wafer backpressure applied to a wafer during chemical mechanical polishing. A wafer carrier is provided with numerous distensible elements arrayed on or near the bottom surface of the pressure plate. The distensible elements comprise expandable chambers disposed on a disk fitted between the pressure plate and the wafer. The distensible elements are controlled with necessary valves and pressure sources and computer control systems to exert excess backpressure at select small zones of the wafer. For wafers with known patterns of hard areas which experience relatively low local removal rates, the disk is placed in registration with the wafer, and polishing is performed while distensible elements corresponding to the hard areas are operated to provide excess backpressure corresponding zones of the wafer.
Various embodiments of the distensible elements are described, including silicon bladders, shape memory elements, and electrostatic plates.
The backpressure applicator is preferably constructed as an array of MEMS devices.
A valve layer 37 contains numerous pressure control valves 38, 39, 40, 41, 42 and 43 with outlets 44 aligned to corresponding holes in the pipe layer. Inlet ports 45 provide for fluid communication from a pressurized fluid source, relief ports 46 provide for exhaust of fluid from the chamber, and outlet ports 47 provide for fluid communication from the valves to the fluid pathways 34 of the pipe layer. The valves are operated as necessary to maintain desired pressure in the chambers. The valves may be individually addressed, like pixels on a display, with appropriate chip addressing technology.
The layers may be fabricated from silicon wafers and other materials using known etching and deposition, and layering techniques. The wafer-contacting layer may be provided in fairly thick layer of silicon, then heat pressed onto the pipe layer with the glass layer sandwiched between them, so that the glass layer melts to fuse the other two layers together. The wafer-contacting layer may also be made of material of different flexibility, such as silicon nitride, parylene, bisbenzo-cyclobutene (BCB), resins such as cyclohexanone (SiLK®) and other polymers that can be evenly deposited with appropriate assembly techniques. Even elastomeric materials such as silicone rubber may be used for the wafer contacting layer. The glass layer may be deposited on the pipes layer, and voids may be etched from the deposited glass to leave a grid of glass beads about 100 nm deep and 100 microns wide. The channels of the pipes layer may be formed by several etchings of a silicon wafer, with a first bore of about 500 micron diameter, a second bore of about 50 micron diameter, and a final bore of about 200 nm. The wafer 31 and pipe layer 33 are pressed together and heated in order to melt the glass layer and thereby fuse the glass beads to the wafer contacting layer. The wafer-contacting layer may then be ground to a thickness that permits sufficient flexure of the layer to distend under pressure applied to it through the pipes layer (20 to 50 microns).
The valves layer may be constructed with known techniques for manufacture of MEMs valves on silicon and glass substrates. MEMS micro-arrays described in Vandelli, Development of a MEMS Microvalve Array for Fluid Flow Control, the silicon micro-fabricated valve described in Henning, et al., Evaluating the use of MEMS-based gas and fluid delivery systems, silicone-on-silicon valves and many other valves may be used.
In use, the valves are controlled to pressurize chambers corresponding to known hard spots on wafers. The location of hard spots depends on many complex factors, but for given IC architecture and CMP process variables, hard spots and soft spots occur in predictable areas of the wafer surface. On a wafer with many IC build up on the surface, an ordered matrix of hard spots is typical, so that, relative to the wafer, areas requiring more or less back pressure may be empirically determined. During polishing corresponding chambers of the zoned pressure applicator may be pressurized to increase the removal rate on these hard spots to match the removal rate of the surrounding wafer surface. This requires registration between the wafer and the zoned pressure applicator, and this is easily achieved by first, determining empirically the location of hard spots of a model wafer, aligning the wafer to be polished with the zoned pressure applicator with any suitable registration feature. Because the zoned pressure applicator and the wafer to be polished are fabricated from similarly substrates, the notch or flat of each may be aligned during polishing to keep the array of chambers aligned with the matrix of hard spots. During polishing, the chambers associated with hard spots may be pressurized to increase the back pressure applied to the wafer at each hard spot. Should the wafer be subject to hard spots with differing degrees of resistance to polishing, the back pressure may be varied accordingly to achieve a simultaneous endpoint for all regions of the wafer. Where pressure regulating valves are used, the pressures in various chambers may be controlled on a chamber-by-chamber basis, so that each chamber may be pressurized to a different degree. In more rudimentary operation, the valves may be operated without a pressure regulating function, and the pressurized fluid source may be provided at a predetermined, desirable pressure corresponding to the desired overpressure, and the valves may be opened to subject the associated chambers to the desired overpressure.
As illustrated in
A computer control system with appropriate software and memory, along with electronics for addressing the valves, is used to control the valves as desired. The location of hard spots is entered into the computer and stored in memory, and the computer is programmed to control valves corresponding to the hard spots to increase the backpressure on the wafer in regions corresponding to the hard spots.
The backpressure applicator disk may be provided with a complete array of valves, as illustrated in the previous figures. However, for high volume production of process wafers, custom-made backpressure applicator disks with low-density arrays of valves, arranged on the disk to correspond to the known hard spots, may be most economical. A custom backpressure applicator disk is illustrated in
Though the distensible elements are conveniently provided as expandable chambers constructed as indicated in reference to
Many MEMs type devices may be formed in the back pressure applicator and operated to provide high density, addressable array of back pressure zones. As illustrated in
The high-resolution backpressure applicator can be used in other process. For example, prime wafers exhibit randomly dispersed local millimeter scale non-planarities of 100 nm or more. While insubstantial in current processes, these non-planarities may be considered substantial relative to increasingly dense and small-scale architectures. These prime wafers may be polished to eliminate the local plateaus with the backpressure applicator by first imaging a particular disk to determine the location of high spots, feeding information regarding the location of high spots to the control system, and then polishing the particular disk while distending elements disposed in proximity to the high spots. Also, though it is overkill, the high-resolution backpressure applicator can be used in place of current annular zone backpressure devices to address cross-wafer variations in removal rates. For center-slow processes, a circular grouping of elements may apply over pressure to the center region of the wafer. For edge-slow (center-fast) processes, and annular grouping of distensible elements may be pressurized. Using this procedure, a large number of distinct annular zones can be differentially polished, and the annular zones may be adjusted during the process, or between polishes, without replacing the carrier or carrier components.
While the inventions have been described in the context of chemical mechanical polishing, they may be employed in any polishing or grinding system. The various materials and methods may be modified as needed for particular applications. For example, the chambers may be provided in any shape, and the material for the wafer contacting layer may be varied to provide more or less flexibility than the silicon membrane described in relation to the figures. The density and arrangement of the distensible elements may also be varied to suit the application, as described in relation to
This application is a continuation of U.S. patent application Ser. No. 10/452,411, filed May 30, 2003, now U.S. Pat. No. 7,008,309.
Number | Date | Country | |
---|---|---|---|
Parent | 10452411 | May 2003 | US |
Child | 11369700 | Mar 2006 | US |