Double patterning grabbing more litho headlines these days. Interesting development. The current tools may not be good enough though (>5 nm overlay spec).

Yes it's interesting. It suggests several things and consequences.

First, it suggests options are running out for 32 nm and beyond. Going to another wavelength is not easy, especially one that is more absorbing.

Second, double patterning is an "easy" way to reduce pitch, but it does not help how to reduce the feature size. For that, you have to trim the line, or shrink the space. But you still have to improve the quality of the starting (wider) feature to the next node, even if it is nominally the same size as one or two generations ago. It's still easier, much easier to go to immersion single exposure.

Third, some low-end foundries, especially in China, still do not have their 193 nm process set up, since they are still stuck at 0.18 or 0.13 micron, using the older 248 nm wavelength. The double patterning philosophy (easy to learn and pick up) will immediately enable them to do 90-65 nm level technology today, without making a painful transition to 193 nm processing. At least one foundry has 90/65 nm capability already, so this means they can go to 45/32 immediately. This should be more than enough for their needs today. It's a great catch-up and/or overtake opportunity for them.

The evidence builds...

http://spiedl.aip.org/getabs/servlet/Getab...=cvips&gifs=yes

Proceedings of SPIE -- Volume 6517

Carbon accumulation and mitigation processes, and secondary electron yields of ruthenium surfaces

B. V. Yakshinskiy, R. Wasielewski, E. Loginova, and Theodore E. Madey

(published online Mar. 21, 2007)

Metallic ruthenium capping layers ~2 nm thick protect and extend the lifetimes of Mo/Si multilayer mirrors used in extreme ultraviolet lithography (EUVL) applications. However, Ru-capped mirrors experience a loss of reflectivity after prolonged exposure to EUV radiation. In the present work, we use ultrahigh vacuum surface science methods to address several aspects of Ru surface chemistry that may impact on Ru capping layer stability and mitigation processes. (1) We characterize the composition and stability of Ru surfaces that simulate surfaces of Ru-capped multilayer mirrors, under exposure to different background gases (water, methyl methacrylate (MMA)) and to electron irradiation. Evidence for some mitigation of carbon accumulation during electron bombardment in MMA + water vapor is found. (2) We report the photon-energy dependence of secondary electron yield (SEY) measurements for clean Ru, O-dosed and C-dosed Ru, and Ru-capped multilayer mirrors using synchrotron radiation near 13.5 nm at Brookhaven National Synchrotron Light Source (NSLS). Much of the radiation-induced chemistry on the surfaces of capping layers is induced by low-energy secondary electrons rather than direct photoexcitation, so the SEY is an important parameter affecting mirror lifetimes in EUVL.

©2007 COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

From: US patent application 20060088787

"...However, unlike ion beam lithography, EBL suffers from a serious intrinsic problem, namely electron proximity effects which stem from polymer/electron interactions at the molecular level. These effects arise from secondary electrons that induce severe degradation of the pattern definition, as the uniform exposure of the resist by the incident electron beam gives rise to a non-uniform distribution of actual exposure in the pattern writing area. During the past two decades, attempts have been made to decrease the electron proximity effects (see, for example: A. N. Broers, IBM J Res. Develop. 32, 502 (1988)) but no significant breakthrough has been achieved.

[0008] A similar intrinsic problem arises using extreme ultraviolet lithography (EUVL at 13.4 nm, or "13 nm" Lithography)), viz. the generation of photoelectrons in the polymeric resist. These photoelectrons, like the secondary electrons in electron beam resists, induce resolution degradation from line broadening due to the off-axis pathways of photoelectrons..."

Hi everybody!
It's been such a long time, I stopped in 1990 (remember, 2µm)...
Electron beams probably have the same weakness as EUVL or worse.

But what would happen (happens?) with a proton beam?
The displaced electrons would have a moderate energy.
The protons deviate within the resist but should concentrate their energy better than UV.
One can expose at several energies to put a uniform dose along the depth.
Hydrogen atoms are (were!) not very polluting and can be outgassed.

Deuterium would be slightly better for the energy of secondary electrons.

Right, I guess the irradiation weakens the resist chemically. But what if we use this weakening and etch the "resist" away at the irradiated zones?

If one fears to damage the semiconductor, maybe an additional film under the resist could be etched away after etching the resist.

Already tried and abandoned?

Hello,

That is an interesting suggestion. With protons or ions, their range is smaller than that of electrons, but since they stick around, they 'stuff' the resist or wherever they land. Also since they are slower they have more chance to interact Coulombically.

Beam etching is very promising. I think the main reason it didn't take off was the beam got blocked by the reactant. But maybe you can still use the beam itself to etch if it's not pure implanting/sputtering.

Departing from classical photoresist:

PTFE, FEP, maybe ETFE (would adhere better) etc are chemically stable but very sensitive to ionizing radiations. Something like 100krad makes them weak and supposedly etchable. My calculation gives a very reasonable proton dosis.

An underlying PI would protect the semiconductor. Very resistent to ionizing radiations, but chemically less stable, so it could be etched through the locally removed Ptfe and thus expose the semiconductor.

Hi Enthalpy,

Can you define 'etchable' for us? (like etch using what...)

Thanks...

Data is scarce for such low energies. To stop within 50nm of PTFE, protons would have roughly 1keV energy, alphas maybe 4keV, fluorine 35keV, silicon a bit more.

The nice news is that "secondary electrons" have no noticeable energy.
Also, the wavelength for protons slowed down to 20eV is 0.2pm.

Ptfe needs a small dose, maybe 1Mrad = 10kJ/kg (check with protons, not gammas!), in which case 1J/m2 or 1mC/m2 would already suffice, charging the film to <2.5V so breakdown is doubtful.

How to etch? How should I know? Hydrogen plasma maybe, with PTFE? You forget photoresists had 2µm thickness when I changed my job!

The secondary electrons come from ionization by the incident particles along the way. It is the secondary electrons, regardless of the primary radiation (electrons, protons, EUV, X-rays, etc.) that break the bonds in the resist. They do so only after they have slowed to a stop (or near-stop). So each secondary needs a varying distance to do so. It could be sub-nanometer, it can also be tens of nanometers. This variability is the fundamental limit to width control.

These high-energy beams should be put to other more significant use than lithography. They are no match for the improvement of the patterning tricks of today. Fortunately or unfortunately.

Dear Enthalpy, thanks for the valuable calculations and suggestions. Maybe response to oxygen and fluorine plasma is also affected by the implantation!

Dear guiding_light, where is the data for such low energy electron stopping distance?

There are lots of references, including this one.

1123, for stopping power, you basically have to add or integrate the mean free path distances from the initial energy down to zero. But you also have to recognize that this distance won't be a straight line but a total travel distance which would include a lot of direction changes. But clearly the path could be quite long for low energy electrons below 5 eV or so.

Whether you use protons at 1keV or silicon at 40keV, the secondary electrons (if any!) will have <1eV energy, so they will have no chemical effect.

I even doubt protons of 1keV are ionizing enough to weaken PTFE chemically, that's why fluor or silicon might be better: At least, they will displace atoms, thus breaking chemical bonds.

Is there any data that allows useful predictions? Or should it rather be experimented?

Dear Enthalpy,

After some searching, found this article which you may find answers your questions mostly to your satisfaction.

http://www.materialstoday.com/pdfs/june_20...006_review2.pdf

While the imaging is still based on secondary electrons with the same range at low energy, at least the origin points of these electrons appears to be more confined than with primary electrons.

Thanks for the link, guiding_light!

The penetration depth of 1-2MeV protons in Si is muuuuuch more than what I saw for water - unusable for <45nm lithography. One explanation is that hydrogen in water brakes protons better than Si does, so one should replace PTFE by ETFE or PVDF or another radiation-sensitive polymer containing hydrogen, or use heavier ions like F or Si.

The very deep penetration of 2MeV protons in Si proves that at such a low energy, protons don't ionize matter and won't have any interesting chemical effect, so pushing the atoms instead should be the right method.

It's no problem.

Here is another by the same group I think: http://www.ciba.nus.edu.sg/publications/PBW/pbw2004_5.pdf

A proton with MeV energy stopping in ~ 100 microns, comes out to about eV/angstrom energy loss. The paper mentions energy transfers peaked around 100 eV, which means they would be very rare even in thin resist. It suggests probably a high dose is needed to make up for relatively low interaction.

Interesting papers!

But why do they all use protons of MeV energy that go through the resist and stop deep in the silicon - where I suppose they have unwanted effects, and in the gate insulators as well - instead of 1keV that would stop within the masking polymer? With the two-layers polymer (ETFE+PI), preserving the silicon but etching the polymer down to the silicon looks feasible.

Will this be the cutting etch of technology?

I don't know why 1 MeV is preferentially used either. Near the end of range, there is some more forward lateral scattering though.

Digging through some old pdfs, found this piece of history. Back in 2000, they only expected EUV to last to 35 nm node (and that's with the most powerful optics available). Yet now we are close to 32 nm node, and it is already expected that immersion and/or double patterning will be the only game in town at that point. You can talk about EUV at 22 nm, but that puts it in the "game over-time" realm that the other optical techniques are already at.

From Intel's patent 7235344:

"Electron absorption is recognized as one recognized mechanism by which EUV photoresist films receive patterning signals to form lithographic features.

These electrons are the product of the original aerial image from EUV photons impinging upon and ionizing atoms in the photoresist. The distance an electron travels from the point of ionization within the film or at the uppermost part of the substrate (e.g. the substrate surface) to a PAG is a propagation length. Typically too, PAGs such as triphenyl sulfoniums are relatively small species (e.g., volume on the order of about one cubic nanometer) and include relatively electron transparent moieties (e.g. hydrocarbons and sulfur) and therefore have a relatively limited electron capture cross-section. Thus, the uncontrolled propagation length of electrons within the photoresist blurs the original aerial image by a finite amount, limiting the resolution of the film and contributing to feature roughness. "

While it seems they finally get it, some clarification may still be necessary.

Rather than electron "absorption", "inelastic scattering" may be more accepted by the electron community.

Also it is not just one mechanism, it is the dominant one, as the cross-section for electron inelastic scattering causing chemical excitation is much higher than for the original photon absorption.

Intel disclosed their 32 nm process, it will be using immersion lithography. They have to use immersion lithography forever (effectively), just to improve their cost of ownership.



Samsung disclosed their 30 nm NAND process, it will be using a self-aligned double patterning technology. Again, this powerful halving-in-one step technique asssures it will be used forever (effectively).

Some graphic depiction of EUV damage to semiconductor devices (CCDs for EUV solar imaging telescope).

Some of the damage to the EIT is a coating of foreign material, not EUV damage.

Yes, the contamination was believed to be H2O (ice). Periodically they would bake it off. But the dose-dependent reductions of CCE are related to the EUV exposure.