(1) Field of the Invention
This invention relates to a versatile maskless patterning system with capability for: selecting rapidly among a plurality of projection lenses, providing multiple resolutions, enabling optimization of the combination of the imaging resolution and exposure throughput, and making possible cost-effective fabrication of microelectronics packaging products.
(2) Description of Related Art
Microelectronic packaging technology is a key driver leading to advancements in the performance of electronic systems that are integrated on circuit boards. Especially for high-processing-power systems, the packages may be the most crucial factor since they gate the rate of communication between different chips, and in fact, a high-end processor may be only as good as the packages that are used to interface it with the other components on the board. Additionally, given the importance of portability for commercial, but especially for all types of military systems, the size of the package is important as well. Recent trends in packaging technology have resulted in small, high-performance packages by means of designs having fine linewidths, multi-level circuitry, and solder-bump interfaces to the circuit board. Fabrication of packages having these advanced design features can be performed only with high-resolution patterning systems that have fine-alignment capability—resolutions down to 10 μm and alignment capability better than 3 μm are required for next-generation microelectronic packages.
While microelectronics packages have traditionally been fabricated by means of conventional lithography tools such as contact printers, the requirements for higher resolution and finer alignment for the multiple layers of the package have led to the need for a mix-and-match approach, based on contact printing, proximity printing, direct-write, as well as projection lithography. The large-area seamless-scanning projection lithography systems developed by Anvik Corporation (U.S. Pat. Nos. 4,924,257 and 5,285,236) meet both the resolution and alignment requirements for next-generation microelectronic packages—and deliver high throughputs—and are therefore highly attractive for large-volume production of high-end packages.
A massively parallel maskless lithography technology newly developed by Anvik (U.S. Pat. No. 6,312,134) has promising applications for the production of a wide variety of electronic systems, that require only small-to-moderate-volume production of many specialized components, rather than a large volume of a single component; for example, for military missile defense applications, which require large networks of sensor, communication, and networking equipment, but in limited quantities.
Anvik's prior-art maskless lithography system uses a spatial light modulator (SLM) array instead of a Mylar or glass mask. The SLM is a megapixel display device which can be updated at multi-kHz frame rates. The SLM fully removes the need for hard glass masks, eliminating many of the problems associated with the use of masks, including the cost of purchasing and maintaining masks; the cost of lost yield resulting from masks having defects; and the production time that is lost while waiting for delivery of masks from mask houses. These factors can be particularly significant for electronic components manufactured in low volumes, such as many military electronics, where mask-related issues may be the primary influences affecting the price and delivery of the component.
In spite of the above cited advantages of the prior art maskless lithography technology, there is still a need for a more versatile maskless patterning system where the demagnification of the SLM image onto the substrate can be rapidly and easily varied in order to optimize the trade-off between pattern definition and throughput. Such a versatile system would be capable of patterning a large range of feature sizes on a wide variety of substrate materials, over very large areas, the substrate size limited only by the travel range of the scanning stage. In addition, it would be capable of patterning different regions on the same substrate with different resolution and feature sizes.
This invention provides a capability for selecting rapidly among a plurality of projection lenses in a maskless patterning system, providing multiple choices for resolution and feature size, and enabling optimization of the combination of the imaging resolution and exposure throughput.
An object of this invention is to enable the cost-effective fabrication of microelectronic packaging products.
Another object of this invention is to provide versatility to a single maskless patterning system.
A feature of this invention is the use of a plurality of projection lenses of different demagnifications mounted on a turret in such a way that they can be rapidly and easily interchanged while maintaining alignment and focus.
Another feature of this invention is the use of multiple spatial light modulators to provide a composite image with a greatly enhance pixel count for demagnification by the selected projection lens.
Another feature of this invention the use of software that recognizes those high detail areas of the virtual mask that require small feature size and that controls the projection lens turret and scanning stage in such a way that these high detail areas are patterned with the necessary high resolution while allowing the other areas to be patterned more rapidly at lower resolution.
An advantage of this invention is that a single versatile system is capable of patterning a large range of feature sizes on a wide variety of substrate materials, over very large areas, with the substrate size limited only by the travel range of the scanning stage.
Another advantage of this invention is that selection of the projection lens and illumination source can be accomplished easily and rapidly so that system down time between production runs or steps in production runs is minimized.
Another advantage of this invention is that a single versatile system is capable of patterning the different layers involved for each of the process steps in the multi-step fabrication of microelectronic packaging products, with the optimum combination of resolution and exposure throughput for each step.
Another advantage of this invention is that patterning different regions on the same substrate with different resolution and feature size can be accomplished, making it possible to increase overall throughput by using the highest resolutions only on those regions of the substrate where the smallest feature sizes are needed.
Another advantage of this invention is that the patterning of the various areas on the substrate with different feature size can be accomplished either sequentially or simultaneously or any combination thereof, making it possible to further improve throughput.
Another advantage of this invention is that the virtual mask for each layer of the microelectronic packaging product to be manufactured can be designed in such a way that the features requiring the highest resolution are deliberately located on specific limited regions of the substrate, making it possible to further improve throughput by taking fuller advantage of the ability to use the highest resolutions only on those regions of the substrate where the smallest feature sizes are located.
Other objects, features, and advantages of the invention will be apparent from the following written description, claims, abstract, and the annexed drawings.
Operation of the maskless patterning system is achieved by scanning stage 12 while each of the SLMs 5 displays a sequence of frames—each frame representing a portion of an entire printed pattern—mimicking a prior art hard mask scanning synchronously with the substrate. The illumination of each SLM 5 is provided by pulsed laser sources 1 and the illumination pulses are synchronized to the data stream that configures the SLMs. Pixels which are “on” in each SLM reflect the illumination into the aperture of the corresponding projection lens and are therefore imaged onto the substrate; “off” pixels reflect illumination away from the lens and therefore are not imaged.
An important feature of the versatile maskless patterning system is the use of a plurality of projection lenses 7 mounted in lens mount 8, which can be fixed or movable. Each projection lens has a different demagnification ratio and is mounted in lens mount 8 in such a way that it is precisely in focus with the plane of the substrate 11 and in accurate registration with the patterning location. Each lens is of a different magnification, the function of which is to provide multiple resolutions and enable optimization of the combination of the imaging resolution and exposure throughput. In one preferred embodiment, there are three lenses 7 in lens mount 8 covering a range from 1 to 14× demagnification. Sequential, simultaneous or any combination thereof exposures of the substrate to patterning projected beams with different selected resolutions is achieved by controlling selected SLMs 5 and laser sources 1. Any of the projected beams 9 can be turned off entirely by adjusted all of the pixels in the corresponding SLM 5 to the “off” position.
In simplified embodiments of the versatile maskless patterning system, a single illumination source 7 is used with a single SLM 5. A plurality of projection lenses 7-1 . . . 7-n is mounted in movable lens mount 8. The projection lens 7 can be arranged in a linear array, in a 2-D array, or along the circumference of a circle. Movable lens mount 8 which can be translated or rotated. Each projection lens 7 has a different demagnification ratio and is mounted in lens mount 8 in such a way that translation or rotation of the mount selects a particular projection lens and demagnification ratio. The lenses are mounted and pre-aligned in mount 8 so that when any lens is brought into imaging position it is precisely in focus with the plane of the substrate 11 and in accurate registration with the patterning location. Each lens is of a different demagnification, the function of which is to provide multiple resolutions and enable optimization of the combination of the imaging resolution and exposure throughput. In one preferred embodiment, there are three lenses 7 in lens mount 8 that can be rotated to cover a range from 1 to 14× demagnification.
In one preferred embodiment, a DMD array with 1024×768 micro-pixel array is used having a pixel pitch of 14 μm. A DMD of this type is compatible with reasonable levels of fluence in the visible through the ultraviolet wavelength range, with a damage threshold of 100 mJ/cm2 (comparable to the damage thresholds for chrome masks in prior art lithography), and an average power rating of 10 W/cm2.
Since the substrate rests on a stage which moves at a constant velocity, each pulse that illuminates the DMD spatial light modulator, exposing the substrate, illuminates a different region of the substrate. The system therefore operates such that entire DMD array is reconfigured for each pulse, mimicking a scanning mask. Since the substrate scans at a constant velocity, while the DMD spatial light modulator remains stationary, image blurring results due to the finite duration of the illumination pulse, and therefore the pulse width must be short enough to maintain the blurring below a maximum-acceptable level, which we typically set at less than ⅕th the minimum feature size. In order to optimize throughput, the system should operate at the fastest possible pattern transfer rate, limited either by the frame rate of the DMD or the maximum repetition rate of the laser.
In a preferred embodiment, the translation speed of the stage is set to a value where, between consecutive laser pulses, the substrate moves a distance that is only a fraction of the width of the hexagonal illumination beam. Each point on the substrate is thus illuminated by a number of different pulses as the stage carries the substrate through the illuminated region, thereby integrating the energy along the scan direction and improving the uniformity of the delivered dose along that axis.
The ‘address size’, is an important parameter for both mask-based and maskless systems. For example, with a mask-based system, a mask is written either with a laser or an e-beam tool, using pattern data produced on a CAD program. Curved features, such as circles, are continuous, smooth patterns, when created on a CAD program, but when they are digitized for mask writing, the smooth curves must be approximated by a step-wise pattern constrained to locations on an x-y grid. The ‘address size’ is the smallest x-y separation of points on the grid. As shown in
The maskless lithography system also has an address size, in this case the pitch of the x-y pixel array in the image plane. For a preferred embodiment of the versatile maskless system, designed for 10 μm resolution, the address sizes is as small as 1 μm—which is 1/10th the size of the smallest feature—so that feature definition will compare favorably with that obtained with hard masks.
Since the size of the pixels on the spatial light modulator is fixed, the address size can only be changed by varying the magnification of the projection lens. For a preferred embodiment of the versatile maskless patterning system, a DMD type spatial light modulator is used with a fixed pixel size of approximately 14 μm. Thus when patterning features several hundred microns in dimension, 1:1 imaging is suitable, since the pixel size is substantially smaller than 1/10th the feature size. However, for the finest curved and off-grid features of sizes down to 10 μm, as required for high-performance electronic packages, a reduction ratio of approximately 14:1 would be needed in order to have an address size 1/10th the size of the features. For features of the order of many tens of microns to >100 μm, intermediate address sizes would be needed. For this preferred embodiment of the versatile maskless patterning system, three different projection lenses are used, each covering a different magnification range.
General Considerations Regarding Throughput
For microelectronic packaging products, conductor traces, layer-to-layer vias, and other features may range in size from as small as 10 μm and as large as several hundred microns. Vias for ball grid arrays, which connect the chip package to the circuit board, are generally several hundred microns in size. Thus the versatile maskless system for fabricating microelectronic packages should be capable of efficiently patterning features over a wide range of sizes, from approximately 10 microns, up to several-hundred microns, while delivering optimum throughput for the full range of feature sizes.
While a 14:1 reduction system would satisfy the address size requirements over the full range of resolutions, from 10 μm up to several hundred microns, the large reduction ratio has the disadvantage that, from optical design considerations, it leads to a smaller field size compared with a 1:1 system. For example, the image area for the larger reduction ratio is approximately 200 times smaller, a factor large enough to have a significant effect on the throughput of the maskless tool. Therefore, to optimize throughput, it is desirable to operate at the largest acceptable address size. In order to meet this requirement, the preferred embodiment of the versatile maskless patterning system uses three different lenses, each producing different address sizes by covering different demagnification ranges.
For the coarsest features, of the order of hundreds of microns, a reduction lens operating at demagnification ratios ranging from 1:1-2:1 (providing a pixel image size of 14-7 μm) is required. For intermediate resolutions, a lens covering a range from 2:1-4:1 (providing a pixel image size of 7-3.5 μm) is required; and for the finest features, of the order of tens of microns, a lens ranging from 4:1-14:1 (providing a pixel image size of 3.5-1 μm) is required. Having three projection lenses covering this full range, rather than a single lens, is preferable because it would reduce the overall complexity of the imaging system given that the design and construction of three individual lenses, each covering a small magnification range, is much simpler than the design and construction of a single lens covering the full magnification range. For example, the highest-reduction lens, operating for feature sizes from 1-3.5 μm, would need a field size of only 4 mm for 3.5 μm resolution and 1.5 mm for 1 μm resolution, comparable to the specifications of a standard microscope objective. The low reduction-ratio lens would operate at a field size of approximately 1 cm for 14 μm resolution, and 0.5 cm at 7 μm resolution, performance specifications that are only slightly more demanding than those of a high-quality camera lens.
In order to optimize the throughput at various reduction ratios, it is sometimes advantageous to use several different illumination sources, each one for a particular reduction ratio range. This is because a change in the projection lens reduction ratio causes a quadratic change in the exposure pulse intensity on the substrate. In a preferred embodiment of the versatile maskless patterning system, two different laser sources are utilized.
For patterning the coarser features by using the projection lens operating in the range of 1:1-2:1 reduction, a XeF excimer laser source is used. This laser has an output wavelength of 351 nm, compatible with i-line resists and operates at 1000 Hz, producing around 25 mJ/pulse. In this operating regime the image field size ranges from approximately 0.25 to 1 cm2 in the image plane, and, therefore, assuming a typical 50% optical system throughput, the exposure intensity ranges from 12 to 50 mJ/cm2 per pulse in the image plane. While the repetition rate of the laser source is significantly lower than the 10 kHz frame rate of the DMD, the full utilization of the output power from the laser results in a high-throughput system.
However, if operating at 14:1 demagnification with the same excimer laser, the energy density in the image plane would range from 2 to 8 J/cm2 per pulse, which is far more than needed, considering that typical resists have a dosage requirement of, at most, several-hundred mJ/cm2. Consequently the laser would have to be operated at much lower power than its peak capability, significantly reducing the throughput. A better choice is a high-power solid-state laser system, operating at lower power per pulse, but much higher repetition rate, such as a frequency-quadrupled Nd:YLF laser with output at 351 nm, producing 5 W of power at a repetition rate of 10 kHz. At 0.5 mJ/pulse output from such a laser, with 50% optical system efficiency, operating with 14× reduction, the energy density in the image plane is approximately 50 mJ/cm2, comparable to that obtained at 1:1 magnification with an excimer laser source, and suitable for exposing a variety of resists. For magnifications in the middle of the range, around 4-10×, the system offers the user the flexibility to choose the light source that optimizes the throughput.
Thus the preferred embodiment of the versatile maskless patterning system is designed to optimize performance by allowing the user to obtain the best feature definition without sacrificing patterning speed. Following are some specific calculations of throughputs.
The throughput estimates are for patterning resists having a relatively high dosage requirement of 200 mJ/cm2. For purposes of achieving uniform exposure at all pixel sites on the substrate, each pixel with is assumed to be exposed to twenty pulses. Therefore, the laser intensity would be decreased to levels below the maximum possible output in order to achieve an energy density of 10 mJ/cm2 in the image plane.
It is assumed that the scan speed of the stage is set such that twenty pulses expose each pixel for the set repetition rate of the laser. For example, with 1:1 magnification, the XeF excimer laser would be used, with the energy density per pulse at the substrate set to be approximately 10 mJ/cm2, and the image area approximately 1.5×1.5 cm. Thus, the DMD must scan the image across its width after 20 pulses (see
For reduction ratios in between 1:1 and 14:1, the throughput would be somewhere between the two extremes discussed above. Additionally, we emphasize that with this system, using multiple magnification ratios and two laser illumination sources, the operator has the flexibility to select the best trade-off between feature definition and system throughput.
This throughput can be increased by a factor of 2 or more by using multiple DMDs. (see
Under these conditions, one additional DMD would nearly double the throughput, two additional DMDs would triple it, and so on. As an example, by multiplexing 4 DMDs, a system with a resolution of 14 μm will deliver a throughput of 160 18×24 in. panels/hr.
While additional DMDs would call for larger field sizes for the projection lenses as well as additional illumination system optics, thereby increasing the complexity of the optical design and adding cost to the system, the increase would be less than linear in proportion to the throughput enhancement due to efficient DMD multiplexing. Further, each system could be designed on an individual basis to achieve the best trade-off between throughput and system cost. It is likely that the marginal cost for doubling or even tripling the throughput by adding one or two additional DMDs would be worth the additional system cost for most high-volume production environments.
General Considerations Regarding Processing Methods
The versatile maskless patterning system can be used to perform three different processing methods. In method #1, the system is used to pattern a wide variety of feature sizes on a wide variety of substrate materials. For method #2, the system is used to sequentially pattern the different layers comprising a single multi-layer package. Finally, for method #3, the system is used to pattern different regions on a single layer with different resolution and feature size. A preferred manufacturing process might use a single one of these methods or several of these methods in combination.
Method #1 is described as follows. Here the versatile maskless system is being used to pattern several different types of substrate materials in consecutive low-volume production runs. Each substrate material corresponds to a different type of microelectronics packaging product and has different desired feature size and exposure throughput. For each low-volume run, the appropriate projection lens and illumination source is chosen to provide optimization of the combination of the imaging resolution and exposure throughput for each substrate material. The turret mounting and focus pre-alignment of the projection lenses makes rapid changing of the lens possible. For each production run, the substrates can be either coated with photoresist for a stand photolithographic pattering process or be uncoated and directly patterned by photoablation. Each substrate is then mounted on the planar scanning stage and aligned with regard to the projected image of the spatial light modulator. The stage is then scanned while the spatial light modulator displays a sequence of frames, each frame representing a portion of the desired entire printed pattern on the substrate. Once the entire substrate is patterned, it is removed from the scanning stage and sent on to the next step in the manufacturing process. The next substrate is then mounted and the process repeated until the production run is completed.
Advantages of the versatile maskless system for method #1 are that a single system is capable of patterning a large variety of feature sizes in a wide variety of substrate materials, that the substrate size is limited only by the travel range of the scanning stage, and that selection of the projection lens and illumination source can be accomplished easily and rapidly so that system down time between production runs is minimized.
Method #2 is described as follows. Here the versatile maskless system is being used to sequentially pattern the different layers comprising a single multi-layer microelectronic packaging product in a low-volume production run. Each layer has a different desired feature size and exposure throughput. Some layers in the same package might be patterned by photolithograpy and some by photoablation. For each layer, the appropriate projection lens and illumination source is chosen to provide optimization of the combination of the imaging resolution and exposure throughput. Again, the turret mounting and focus pre-alignment of the projection lenses makes rapid changing of the lens possible. Each substrate containing the layer of interest is then mounted on the planar scanning stage, aligned, and the stage is scanned while the spatial light modulator displays a sequence of frames. Once the layer is patterned, the corresponding substrate is removed from the scanning stage and sent on the next steps in the manufacturing process that will add new layers to be patterned. For photolithographic patterning, the patterned photoresist would be developed, the underlying material etched, the remaining photoresist stripped, a planarizing layer deposited, a new layer of material to be patterned deposited on top of that, followed by a new layer of photoresist. For photoablative patterning, a planarizing layer would be directly deposited and a new layer of material to be patterned deposited on top of that. The substrate would then be returned to the versatile maskless system for patterning of the new layer. Once all of the layers in all of the substrates are patterned, the production run is completed.
Advantages of the versatile maskless system for method #2 are that a single system is capable of patterning a large variety of feature sizes that are necessary for each layer of a multi-layer package, that the substrate size is limited only by the travel range of the scanning stage, and that selection of the projection lens and illumination source can be accomplished easily and rapidly so that system down time between processing each layer is minimized.
Method #3 is described as follows. Here the virtual maskless system is being used to pattern different regions on a single layer of a microelectronic packaging product. Each region has a different optimum combination of feature size and exposure throughput. The software that controls the system and provides the frame data to the spatial light modulator is given the additional capability of recognizing those regions of the virtual mask that require different feature sizes. The appropriate projection lens and illumination source are selected for a particular feature size and the planar scanning stage is scanned to cover only those regions on the substrate that correspond to that feature size. This is repeated for all of the different regions that make up the layer. The substrate remains fixed on the scanning stage throughout this process. In this way the high detail areas are patterned slowly with the necessary high resolution while the other areas on the substrate can be patterned more rapidly at lower resolution. The slow high resolution scan skips over those areas of the substrate where there are no high resolution features, greatly reducing the time necessary to do the entire high resolution scan.
Advantages of the versatile maskless system for method #3 are that a single system is capable of patterning a large variety of feature sizes that are necessary for each region of a layer of a microelectronic packaging product, that the substrate size is limited only by the travel range of the scanning stage, and that selection of the projection lens and illumination source can be accomplished easily and rapidly so that system down time between processing each region is minimized.
General Considerations Regarding Product Design
In order to take fuller advantage of the ability to pattern different regions of the substrate with different feature sizes, it is possible to design the virtual mask for each layer in such a manner that the small features are deliberately clustered in specific regions of the substrate. In this way the fractional area of the substrate with high detail that requires slow, high resolution patterning is minimized and the patterning system throughput is enhanced as the high resolution scan can skip over a large majority of the area of the substrate.