Patent Application
20080030877
Features, aspects, and embodiments of the inventions are described in conjunction with the attached drawings, in which:
Thin-beam Directional Crystallization, or Thin-beam Directional ‘Xtallization (TDX), fabrication methods can combine poly-silicon's inherent advantages with efficient volume-oriented production capabilities. The end result can be superior electron mobility, flat surface topology, a large process window, and greater throughput. Different types of lasers can be used in thin-beam directional crystallization, for example, in one embodiment a solid state laser can be used. In another embodiment a high power Excimer laser can be used in the TDX process. A master oscillator power amplifier (MOPA) configuration that was originally developed for semiconductor of microlithography applications can also be used. The laser can operate at 351 nanometers and provide over 900 watts of power with exceptional pulse-to-pulse stability and high reliability. Other wavelengths can also be used, for example, 308 nanometers. Generally, any wavelength that is strongly absorbed by the material to be melted, e.g., silicon, can be used. A TDX system is described in co-pending U.S. patent application Ser. No. 10/781,251 entitled “Very High Energy, High Stability Gas Discharge Laser Surface Treatment System,” filed Feb. 18, 2004; U.S. patent application Ser. No. 10/884,101 entitled “Laser Thin Film Poly-Silicon Annealing Optical System,” filed Jul. 1, 2004; U.S. patent application Ser. No. 10/884,547 entitled “Laser Thin Film Poly-Silicon Annealing System,” filed Jul. 1, 2004; and U.S. patent application Ser. No. 11/201,877 entitled “Laser Thin Film Poly-Silicon Annealing Optical System,” filed Aug. 11, 2005, which are incorporated herein by reference as if set forth in full.
The TDX optical system used in conjunction with the systems and methods described herein can convert laser light into a very long thin uniform beam and deliver it onto the silicon film. In addition, it can be configured to stabilize the beam's energy, density and pointing; all of which can improve the consistency of the TDX process. In one embodiment, each pulse can expose an area of approximately 5 microns wide and 730 millimeters long. The length of the beam can be matched to the substrate width so that the glass is processed in a single pass. This can help to ensure a high degree of uniformity and rapid throughput. During exposure, the panel can be scanned at a constant velocity and the laser can be triggered to fire at a pitch, or step size of, e.g., 2 microns. The pitch can be chosen so that the melt region always seeds from the high quality crystals of the previous pulse, producing long directional poly-silicon crystals. Each pulse also melts the large ridge or protrusion at the center of the previous melt region, resulting in a more planar surface.
The TDX process is based on a form of controlled super lateral growth where the melt region re-solidifies laterally from the edges and towards the center. In contrast to ELA where crystal growth proceeds vertically from within the silicon layer, lateral growth produces large directional poly-silicon grains with high electron mobility. The TDX process has a much larger process window than ELA because it relies on spatially controlled complete melting of the silicon film and avoids energy sensitive partial film melting.
The use of a System on Glass (SOG) design approach is another evolving arena that is only made possible with poly-silicon. The higher electron mobility and smaller size transistors that are achievable with TDX processes described herein allow the drive electronics to be fabricated directly into the thin Si coating. This provides a powerful method to reducing panel cost and also improves panel robustness by decreasing the need for tab bond connections. Poly-silicon's much higher electron mobility allows for additional integration of drive electronics such as integrating digital-to-analog converter (DAC) on the substrate and reducing the number of drivers, e.g., by using faster drivers to control more TFT switches.
The overall cost savings with SOG can be very dramatic, especially for processing large panels that consist of many small LCD screens. Using a conventional a-Si approach with separate tab bonded drive electronics for each screen; the drive chips can comprise a significant percentage of the cost per screen as well as an expensive additional assembly step. In comparison, SOG with poly-silicon allows the drive electronics to be efficiently fabricated during the backplane manufacturing process.
With this in mind,
Protrusions 108 can exist at the last point of freezing generally at or near the center of the irradiated surface. Protrusion 108 can be caused when the two edges grow into each other. At or near the center where the two edges grow together the crystal structures generally will not match because each edge is “seeded” from opposite sides of the melted region and these sides originate from different randomly generated seeds. Where the two growth regions meet the crystals will push into each other and push up from the surface. The height of these protrusions 108 can be on the order of a film thickness. The film thickness is commonly about 50-100 nm, however, other film thicknesses are possible.
Protrusion 108 breaks up the uniformed crystallized structure of the surface. Further, as discussed above, the pattern of protrusions 108 that appear after annealing can also make it difficult to deposit a uniform gate dielectric layer, leading to non-uniformity in the TFT performance across the panel. In order to remove protrusion 108 it can be re-melted in the next laser shot.
For example, film surface 102 can be moved a certain step size under the laser for the next shot. The step size must be set, however, to ensure that sufficient laser energy is imported to protrusion 108 so as to ensure protrusion 108 melts. Accordingly, the need to re-melt each protrusion 108, limits the maximum step size that can be achieved. In the example of
Generally, however, the step size must be kept much less than the theoretical maximum, e.g., by several hundred nanometers where the laser pulse width is about 5 μm. The actual step size will be less than the maximum theoretical step size because greater energy is required to re-melt the protrusions 108, due to protrusion 108 being thicker than the rest of film surface 102. Additionally, protrusions 108 can scatter the laser light. So, not only will it take more energy to re-melt protrusion 108 due to its thickness, more energy will also be needed to make up for laser energy scattered by protrusion 108.
It should also be noted that the laser beam width must be controlled to avoid the formation of nucleated grains 204 as illustrated in
If the beam is too wide, then as lateral solidified regions 206 and 208 grow to the center nucleated region 204 and two protrusions 210 and 212 can occur. Protrusions 210 and 212 can be caused when the edges grow into nucleated region 204. The crystallized structures of each lateral solidified region 206 and 208 generally will not match nucleated region 204 because each edge is “seeded” from opposite sides of the melted region. Where the mismatched structures meet the crystals will push into each other and push up from the surface. As discussed above, it is generally preferable that the crystallized structure of an LCD formed when the film surface 202 solidifies be uniform. Protrusions 210 and 212 break up the uniform crystallized structure of the surface. Therefore, it can be advantageous to limit the beam width such that nucleated region 204 does not occur. For example, in one embodiment, the beam width is approximately 5 μm; however, it will be understood that the beam width will depend on a particular embodiment. As long as each side can grow together before nucleation occurs, the fine-grain nucleated region 204 will not occur.
As discussed above, film surface 102 can be moved, or stepped underneath the beam to melt protrusion 108. Surface 102 can, for example, be moved to the left a little less than one half the pulse width. Protrusion 108 can then be re-melted, along with a small portion of lateral solidified region 104, all of lateral solidified region 106 and a portion of un-irradiated amorphous-Si 114. As a lateral solidified region grows from the left to the right it will seed from lateral solidified region 104, continuing the crystalline structure of lateral solidified region 104 until meeting in the middle to form a new protrusion. This can be seen with respect to
Referring to
A laser beam short-axis profile that roughly correlates the intensity profile of the beam to the required melting temperature of the film can be preferred. The profile can be tailored to enable the maximum per-pulse step distance without exceeding the damage threshold.
In other words, by using a short axis, spatial intensity profile, such as illustrated in
Using a specially designed laser 802 and custom beam forming optics 804, substrate 809 can be exposed with a long thin beam 808. A beam forming optical system 804 can produce a short-axis spatial laser beam profile, e.g., as discussed above with respect to
Thin-beam Directional Crystallization with a short-axis spatial intensity profile as described above can be much more efficient than ELA, with much fewer pulses used to expose each area, e.g., less than the 20-40 pulses used in ELA. This can provide much higher panel throughput. In addition, the process window can be much larger than ELA because it does not rely on partial melting, which also helps to improve yield. Since the entire panel can be exposed in a single pass, the Thin-beam Directional Crystallization with a short-axis spatial intensity profile process as described above can also avoid the non-uniformity caused by the overlapping regions that are seen in multi-pass exposure techniques such as SLS and ELA.
The practical realization of the Thin-beam Directional Crystallization can include, for example, three major components in the system: the laser 802, the beam forming optics 804 and the stage 810. In one embodiment a specially designed high power laser 802 with a carefully chosen combination of power, pulse frequency and pulse energy to support the long beam and high scan rates can be used. This laser 802 can, for example, provide 900W of power, which is almost three times the current ELA laser power, to ensure the highest throughput. In one embodiment a laser 802 originally designed for the demanding semiconductor lithography application can be used to ensure good uniformity of the poly-Si and the TFT performance throughout the substrate.
In one embodiment stage 810 can be moved under long thin beam 808 using a stepper or translator. In this way the portion of panel 809 that is under beam 808 can be controlled so that various parts of panel 809 can be processed. In one embodiment, panel 809 can be an amorphous silicon coated glass panel. Thus, beam 808 can be used to melt a silicon film surface on panel 809.
An optical system was developed to create the optimal beam shape. In one embodiment the optimal beam shape can be long enough to cover the entire width of a substrate and narrow enough to optimize the crystallization process. Particular care can be taken with the design of the projection optics contained in the optical system to ensure thermal stability and controlled Depth of Focus (DOF) under high power loads, and to maximize the optics lifetime.
In one embodiment, to ensure rapid motion in the scan direction, the laser must operate at a high repetition rate, for example, at 6 kHz, and the stage speed can, e.g., be 12 mm/sec for an approximately 2 micro pitch. The substrate can be exposed in a single pass, which requires approximately a 150 mJ/pulse to expose a Gen4 substrate. In one embodiment a thin beam crystallization system with a 6 kHz, 900W laser can process an entire Gen4 panel in as little as 75 seconds.
A more detailed explanation of example embodiments of a surface treatment system 800 that can be used in accordance with the systems and methods described herein are described in U.S. application Ser. Nos. 10/781,251; 10/884,101; 10/884,547; and 11/201,877.
At the onset of crystal growth in, e.g., TDX processing of silicon films, crystallographic orientation of the film is generally random due to random formation of crystalline seeds within an amorphous film. On each pulse of the process, one side of the melted silicon formed by beam irradiation re-solidifies laterally and epitaxially from grains grown on previous iterations. The other side grows laterally from newly formed seeds from the initially amorphous portion of the film under the beam. As the beam and/or substrate are moved relative to each other.
As a beam scans across an, e.g., amorphous silicon coated glass panel the TDX process can induce crystallographic texture in the scan direction, normal to the film, or both. The crystallographic texture is formed because as the panel moved under the beam or the beam moves across the panel one side of the molten silicon zone formed by the beam irradiation re-solidifies laterally and epitaxially from grains grown on previous iterations. Thus, as the beam scans across the amorphous silicon coated glass panel a crystallographic texture can form because with each shot a portion of silicon seeds from the portion before it.
In other words, because the process described above produces directionally solidified material, e.g., polycrystalline silicon, the material produced can include a “texture,” since texture often evolves in directionally solidified material. Such texture can occur in the scan direction, normal to the scan direction, or both. The texture produced can depend on the material, the film thickness, process variables, and phase transformation. For example, in a TDX process, texture development can be effected by the step size, the incident energy density, the shape of the laser beam intensity profile, the user wavelength, and the laser pulse duration.
As illustrated in
But as the long grains are formed, the texture can vary across portion 1018. This variation in texture can produce non-uniformity in the performance of transistors 1008 formed on treated film 1000. In other words, variations in texture can result in variations in mobility and other parameters that effect transistor 1008 performance. This decreases the uniformity of transistor 1008 performance, which can have a negative impact on display performance.
In certain embodiments crystallographic texture formation can be stopped by disrupting the epitaxial lateral growth at predetermined locations. By disrupting the epitaxial lateral growth, subsequent epitaxial lateral growth in each new section is re-initiated from new seeds, thereby randomizing the crystallographic orientation of the growing grains.
As shown in
In one embodiment epitaxial lateral growth can be stopped and restarted approximately every 10-20 micrometers, or with a pitch that is matched to the layout of transistor 1008. Referring to
Crystallized film 1000 of
The performance of TFT's 1008 formed on film 1002 will not be as good as transistors formed on film 1000, due to the lower quality of film 1002; however, it has been shown that uniformity is more important for the display area, whereas quality is more important for transistors formed in the circuit area. Thus, by selectively including both types of films, performance for both regions can be better optimized.
Accordingly, when processing a panel, variations in the process can be used for different areas to optimize overall performance by trading off quality versus uniformity. For example, high quality crystallized film for display circuit areas and more uniform crystal film for display area.
Panel 1100 can be processed, e.g., from bottom to top by moving panel 1100 under laser beam 1102 in the direction of the arrows shown at the bottom. The step size for each shot of beam 1102 can be varied as required to produce regions 1104 and 1106. This can be done by varying the rate of translation of panel 1100. In other embodiments, panel 1100 can move at a constant rate while the firing rate of laser 1102 is varied, i.e., to produce intentional overshoots 910 in region 1106.
For example, the placement of circuit area 1108 and display area 1110 can be based on a predetermined layout or mapping of panel 1100. This layout or mapping can be pre-loaded or continuously fed to a controller such that step distances between laser pulses can be varied on a shot by shot basis. One or more panels 1100 can then be processed using the predetermined layout of panel 1100 to guide what process is used in areas 1108 and 1110 of panel 1100.
For example, OLED displays can require a high degree of uniformity from the pixel addressing TFTs, while high performance is generally not necessary. Thus, in one embodiment, a step size larger than the lateral growth length can be used to process a display area 1110. Generally while the step size can be larger than the lateral growth length, it can also be less than twice the lateral growth length. For example, display area 1110 can have a step size to optimize uniformity, e.g., between approximately 2.5 and 3.5 μm for a beam width of 5 μm. Conversely, the digital circuit area 1108 is generally not going to be seen and therefore, visual artifacts are generally not important. Performance can, however, be important in the digital circuit area 1108 because, for example, high performance can lead to higher speed digital circuits. Thus, a step size that is less than the lateral growth length can be used. For example digital circuit area 1108 can use a step size of less than 1 μm. Crystallization can still take place in a single pass.
In one embodiment no reprogramming is necessary during the scan because triggering the laser pulse 1102 to fire can occur as panel 1100 moves under the beam and the different step sizes can be accomplished by, e.g., changing the timing of laser pulses 1102 and/or the rate at which the panel moves relative to the laser and/or the laser moves relative to the panel. Additionally, areas that do not require crystalline material can be left un-irradiated.
While examples that use a thin-beam directional crystallization process to process amorphous silicon glass panels it will be understood that any directional solidification process where step-size influences the resulting polycrystalline material's uniformity and quality, i.e., grain size, crystallographic orientation, etc. can benefit from the systems and methods described herein.
In other embodiments, the ability to control the step size can be used to improve the quality of a display. For example, when a uniform step size is used, a periodic stripped pattern can be produced that can be visible to the viewer in the display area. The stripped pattern is produced by the overlapping application of the laser. As seen in
A TDX scan with a constant step size 1202 will generally be repetitive. If a display surface is too repetitive the eye can pick up small flaws in the surface. Further, flaws in the surface may be repeated due to the uniformity of the scan. To make any flaws in the display surface more difficult for the eye to pick up a TDX scan with intentional non-uniform step distance can be used. The non-uniform step distance can help to disrupt any visual effect perceived in, e.g., an LCD or OLED display due to an otherwise constant and thus periodically appearing step distance. In one embodiment the step size can be varied within a certain range, e.g., 1 to 2 μm. In another embodiment the step size can by a certain range, e.g., 1 to 2 μm.
While certain embodiments of the inventions have been described above, it will be understood that the embodiments described are by way of example only. Accordingly, the inventions should not be limited based on the described embodiments. Rather, the scope of the inventions described herein should only be limited in light of the claims that follow when taken in conjunction with the above description and accompanying drawings.
