As I understand it the point of discussion, or at least the question that was posed to me, was whether an SSMB based system can perform the equivalent to what ASML’s high NA system could do. As I see it, assuming that an SSMB based system can deliver on a 1 kW+ light source, which means that an SSMB based system wouldn’t be hitting the same photon quantity constraints, there’s no reason for optical complexity to be an inhibiting factor. Technically, if your light source is powerful enough, you don’t even need to start with a lower NA optical design. I am skeptical of SSMB materializing as a solution for industry for a lot of other reasons, but high NA optics is not one of them for me.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
You’re the one who brought up the anamorphic mirrors and started going into the reasons behind them. I was just highlighting why the rationale latent in that point actually strengthened my argument. I don’t want to be petty here but if we’re going to lob charges of being too theoretical there was no reason for you to bring up 4x8 projection either.
Agree on the skepticism on SSMB.
I’m not familiar with the reliability of SSMB. I don’t think SSMB could be run 24/7 at >90%availability. There’s also the MTBI, downtime issue that I would question. Could you comment on these?
Dude, relax. I couldn’t have inferred what you were talking about from the short comment of yours.
Not lobbing anything towards you, bro. It’s my bad I didn’t understand where you were coming from. But I’m glad I know what you meant to say now.
And I hope you also understand my statement that source is not as big a issue for ASML to deliver the HiNA EUVL (under their current design decisions).
Synchrotrons are actually more mechanically reliable and less mechanically complex than plasma generated light sources imo. I think they’d be advantageous compared to plasma generated light sources on factors like reliability, maintenance, and downtime.
Where I’m skeptical is operational infrastructure and production workflow and their related economics. I think the idea of using a central SSMB facility to power several scanners is a feasible concept. But then you would have to design a bunch of fabs around each light terminal. Is this going to be a cost efficient arrangement? Even if it is, how long would it take to figure out all the details to make this kind of production facility operationally efficient? The primary advantage of a more traditional lithograph machine is that it’s literally plug and play into current fab designs. And designing a new fab process and workflow to incorporate this different infrastructure is imo where the real points of friction are.
Ultimately, I think in theory at least SSMB is probably fine as an alternate technology to provide for future production of advanced nodes, but because of all the details that need to be worked out for adoption into production I don’t think it’s a viable candidate for a fast timeline to catch-up. That said, if you can get SSMB working and integrated into production, it might have some advantages for upgrading production processes, because everything downstream of your light source is now modularized, so the cadence of your upgrades for those components won’t be restricted by cadence of upgrade for a whole integrated machine, and you can tune your synchrotron to generate even shorter wavelengths if that’s a viable future path to squeezing out anymore progress in the shrinking game.
Maybe you should relax, since my “short comments” weren’t initially directed at you, and since I wasn’t the one that seemed to take issue with being “too theoretical”.
By the way, I don’t think even 1KW is enough either
Why?
The incident angle of 0.55 HiNA is ~18deg in the scan direction. It would seems you’ll need 10x the original source light to be at the level of current EUVL which I think is ~10deg.
The current EUVL is at 250W So I think you’d need at least 2500W for 0.55NA (if want to keep changes from current EUVL to a minimal). This is what I inferred from the chart on the right.
Last edited:
You might be able to get away without needing that much increase in your light source power if your light loss along the optical path or at the point of generation is also much less. One of the headaches with LPP is that you get a lot of light loss even at point of generation just because of factors like errant scatter from metal vapor in the vacuum chamber, loss of reflectivity in the light accumulator due to metal deposition build up, and a less coherent beam profile from the emission source. Arguably without those issues your LPP light sources wouldn’t need to be nearly as powerful as they are. But it’s really hard to know for sure without knowing what illumination power at the scanning point is. All we have are the specs for illumination power at thebpoint of generation.
That said, it’s worth keeping in mind that that 10x light loss is not just going to impact an SSMB light source but an LPP one too, so presumably ASML must’ve figured out something out to get around needing a 10x increase in light source power, and I highly doubt the anamorphic projection is going to be responsible for the vast majority of any 10x difference in illumination at the point of scanning. My best guess is that they’ve both improved the power of their LPP light source while making significant improvements to the coherence of the beam profile and light loss in the light accumulator. These complications with plasma produced sources is probably one of those other engineering parameters where SSMB may offer a notable advantage.
Last edited:
- Status