Kelsey Wooley | Eulitha: Can silicon carbide AR waveguides be etched with production-grade uniformity using DUV displacement Talbot lithography?
16:19 - 17:12
Other snippets from this talk
Summary of the clip:
Can silicon carbide AR waveguides be etched with production-grade uniformity using DUV displacement Talbot lithography?
The data was shared earlier this month at SPIE Photomasks and EUV by TEL, which showcased silicon carbide etching using the photolithography and TEL etch tools. The study's purpose was to demonstrate a highly manufacturable, repeatable path to silicon carbide waveguides that could be implemented right away with mature technologies. Rather than proposing new process modules, the work combined existing photolithography and etch platforms to target immediate product adoption.
TEL reported excellent results. The etch depth critical dimension variation was sub 2%. The extremely vertical sidewalls transferred well, and there was a well-controlled systematic CD bias. Those are exactly the metrics needed for product adoption in AR waveguide manufacturing, because they address uniformity, profile control and run-to-run repeatability. The combination indicates the photolithography and etch sequence can deliver the dimensional control that waveguide products require.
For AR waveguide manufacturing, the key question is whether the lithography and etch can deliver repeatable critical dimensions without exotic process steps. The TEL results show sub-2% etch depth CD variation, vertical sidewalls and a controlled systematic CD bias from the photolithography plus TEL etch flow. That combination points to a manufacturable, repeatable route for silicon carbide waveguides using mature tools, which is what product adoption requires.
In this short video, you can learn:
* TEL presented at SPIE Photomasks and EUV silicon carbide etching that used the photolithography and TEL etch tools.
* The study demonstrated sub-2% etch depth CD variation, extremely vertical sidewalls, and a well-controlled systematic CD bias.
* The results were positioned as a highly manufacturable, repeatable path to silicon carbide waveguides that can be implemented right away with mature technologies.
📋 **Clip Abstract** TEL shared results at SPIE Photomasks and EUV showing silicon carbide etching with the photolithography and TEL etch tools. The study reported sub-2% etch depth CD variation, vertical sidewalls and controlled systematic CD bias, supporting a highly manufacturable, repeatable route to silicon carbide waveguides using mature technologies.
About the speaker:
* Speaker: Kelsey Wooley
* Company: Eulitha
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#SiliconCarbideEtch, #SubTwoPercentCD, #VerticalSidewalls, #SystematicCDBias, #ARWaveguides, #SemiconductorLithography
This is a highlight of the presentation:
Displacement Talbot Lithography for Scalable AR Waveguide Manufacturing
More Highlights from the same talk.
03:02 - 04:28
How does displacement Talbot lithography deliver millimetre-scale depth of focus, and what does it give up in return?
How does displacement Talbot lithography deliver millimetre-scale depth of focus, and what does it give up in return?
DTL leverages all the mature auxiliary technologies already present in the semiconductor industry. Its difference from projection lithography lies in how the aerial image is created: because it uses an alternate method, it achieves an effectively much larger depth of focus. Projection lithography relies on reduction optics that form a focal plane where the photoresist must sit. With DTL, you get millimeters of effective depth of focus, so curved surfaces, warped wafers and high topography can all be patterned.
A second advantage is that, because there are no reduction optics, the exposure field is very large. It is one-to-one with the size of the photomask. A typical industry-standard photomask is six inches, and DTL can transfer all of that as the exposure field, ending up at about 140 millimeters squared. That large field is what makes the approach interesting for this audience.
The no-free-lunch point, however, is that DTL needs periodic structures. It cannot print any arbitrary image; the pattern does need to be periodic. That constraint defines what is given up in return for the millimetre-scale effective depth of focus and the large one-to-one exposure field. So the advantage for periodic structures is clear, but the method is not a general-purpose arbitrary-image lithography.
In this short video, you can learn:
* DTL achieves millimetres of effective depth of focus because it creates the aerial image differently from projection lithography, allowing patterning over curved surfaces, warped wafers and high topography.
* With no reduction optics, DTL's exposure field is one-to-one with the photomask; a six-inch industry-standard mask yields about 140 millimeters squared.
* The trade-off is that DTL requires periodic structures and cannot print arbitrary images.
📋 **Clip Abstract** Displacement Talbot lithography uses mature auxiliary semiconductor technologies but differs from projection lithography by creating an aerial image with an alternate method, giving millimetres of effective depth of focus. It can therefore pattern curved surfaces, warped wafers and high topography, and because it has no reduction optics its exposure field is one-to-one with a six-inch photomask, about 140 millimeters squared, though it requires periodic structures and cannot print arbitrary images.
About the speaker:
* Speaker: Kelsey Wooley
* Company: Eulitha
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#DisplacementTalbotLithography, #ReductionOptics, #ExposureField, #PeriodicStructures, #ARWaveguides, #SemiconductorLithography
09:10 - 10:31
Why does reticle stitching erase the throughput advantage of traditional projection lithography for AR waveguides?
Why does reticle stitching erase the throughput advantage of traditional projection lithography for AR waveguides?
Compared with traditional projection lithography, DTL's tool capital is significantly less—on the order of one-tenth the price. Because DTL has no stitching within the device, you need only two photomasks per product: a front side and a back side. Traditional projection lithography uses reduction, so large AR devices are often three or four reticles stitched together; most designs seen are three or four, especially high FOV.
That stitching forces six to eight masks per product, meaning significantly more mask management than the DTL route. Consumables across the board are the same—all off-the-shelf, industry standard. DTL throughput is currently about 40 wafers per hour, with 300 millimeter full wafer exposure. Traditional projection systems are much faster, significantly faster. But when they stitch multiple reticles together, they take a huge productivity hit.
So the traditional speed advantage is not realized for large AR devices: stitching multiple reticles causes a huge productivity hit, erasing the throughput lead. DTL avoids that by exposing full 300 millimeter wafers with only two masks and no in-device stitching, while consuming the same off-the-shelf industry-standard consumables. Tool capital is one-tenth, mask count is two versus six to eight, and throughput holds at 40 wafers per hour.
In this short video, you can learn:
* DTL tool capital is roughly one-tenth that of traditional projection lithography.
* DTL needs only two photomasks per product (front and back) because there is no stitching within the device, versus six to eight masks for traditional projection lithography stitching three or four reticles.
* DTL exposes 300 mm full wafers at about 40 wafers per hour, while traditional projection systems are significantly faster but take a huge productivity hit when stitching multiple reticles.
📋 **Clip Abstract** DTL reduces tool capital to about one-tenth of traditional projection lithography and needs only two photomasks per product—front and back—because it avoids in-device stitching. Traditional projection systems are significantly faster in raw throughput, but stitching three or four reticles into large AR devices requires six to eight masks and imposes a huge productivity hit, leaving DTL's 300 mm, 40 wafer-per-hour exposure as a scalable alternative.
About the speaker:
* Speaker: Kelsey Wooley
* Company: Eulitha
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#DTLLithography, #ReticleStitching, #PhotomaskManagement, #300mmFullWafer, #ARWaveguides, #ProjectionLithography
06:29 - 08:04
After 18 months of biweekly 150 nm pitch runs, how stable were line-edge roughness and CD uniformity?
After 18 months of biweekly 150 nm pitch runs, how stable were line-edge roughness and CD uniformity?
Over a year and a half, wafers ran every two weeks through an in-house Fablex DUV tool at 150 nm pitch for a reliability and repeatability study. The work tracked LER and CD uniformity at that cadence, deliberately choosing a small pitch. LER was measured at 16 spots per wafer, so reported averages reflect within-wafer sampling.
Average LER came out just over one nanometer, and even with 16 spots measured per wafer the result held up very well. The displacement Talbot process forms the image with an averaging effect, which makes such good structures understandable. It is nice to see that this kind of long time study upholds that.
For CD uniformity, inline wafer variation stayed continuously below two nanometers, and the average was just a little over one nanometer. The results were presented as an overall PDK of what the technology is capable of doing, covering the same 150 nm pitch. Together, the LER and CD figures define the repeatability demonstrated by the long-run data collection.
In this short video, you can learn:
* Biweekly 150 nm pitch runs through an in-house Fablex DUV tool over 18 months produced an average LER of just over one nanometer from 16 spots per wafer.
* Inline wafer CD uniformity remained continuously below two nanometers at 150 nm pitch, with an average just a little over one nanometer.
* Displacement Talbot imaging's averaging effect is credited for the strong structures, and the year-and-a-half time study upheld them.
📋 **Clip Abstract** A year-and-a-half study ran wafers every two weeks through an in-house Fablex DUV tool at 150 nm pitch to test reliability and repeatability. Average LER was just over one nanometer measured at 16 spots per wafer, while inline CD variation stayed below two nanometers and averaged just over one nanometer, presented as an overall PDK of capability.
About the speaker:
* Speaker: Kelsey Wooley
* Company: Eulitha
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#DisplacementTalbotLithography, #LineEdgeRoughness, #CDUniformity, #150nmPitch, #DUVLithography, #ARWaveguides