Nanoimprint Lithography Tool

1. Theory

• Definition: Nanoimprint lithography forms patterns by pressing a patterned template into a liquid or softened resist and hardening it by cooling (thermal NIL) or UV curing (UV-NIL, J-FIL) (Cornell; Resnick and Choi).
• Flow physics: Filling is a squeeze flow in which velocity scales as pressure times the cube of film thickness over viscosity and the square of feature width (Ryu et al.; Cornell).
• Residual layer: A thin layer remains under the features and must be etched away; its thickness should be between 20 $\mathrm{nm}$ and a quarter of the feature depth (Cornell).
• Resolution: Set by the template, better than 10 $\mathrm{nm}$, because no optics are involved (Resnick and Choi; Cornell).

2. Working Principle

• Thermal NIL: Heat above $T_g$, press at hundreds of psi, cool and separate (Cornell).
• J-FIL: Inkjet sub-picolitre drops of low-viscosity resist on each field, lower the template, let capillary filling complete in about 1 s, UV cure and separate (Resnick and Choi; Ito et al.).
• Alignment: Piezo magnification and DMD heating correct field distortion for overlay (Resnick and Choi).
• Templates: Masters written by e-beam are replicated, with anti-sticking coatings for release (Cornell; Patsnap; Electron-Beam Lithography Writer page).

3. Applications

• NAND flash and memory: Overlay of 4 $\mathrm{nm}$ already meets NAND needs (Resnick and Choi).
• Low-cost fine patterning: Tools costing far less than EUV (Patsnap).
• Research and photonics: Thermal NIL of polymers for nanostructures (Cornell; Ryu et al.).

Filling and Residual Layer

Name	Formula	Related Formulae	Variables/Constants
Squeeze-flow fill rate	$V\left(t\right)=\frac{P\left(t\right)H{\left(t\right)}^3}{3\eta {\left(2S\right)}^2}$	$V\sim \frac{p_{eff}{h}^3}{\eta R_{eff}^2}$	$V:imprintvelocity$ $P:pressure$ $H:polymerthickness$ $\begin{array}{r}\eta :viscosity,2S:indenterwidth;Ryuet\\ al.,Eq.4,validforwidthoverthickness\\ above1;Cornell:thesamescalingwith\\ effectivepressureandstampsize\end{array}$ $\begin{array}{r}(assumed:P=1MPa,H=300\mathrm{nm},\eta =1\times 10^4\\ Pas,2S=2.2\mu \mathrm{m};computed:V=0.19\mu \mathrm{m}/s,\\ about0.5stomove100\mathrm{nm})\end{array}$
Residual layer	$d=t-Fz$	$20\mathrm{nm}<d<\frac{z}{4}\Rightarrow 20+Fz<t<z\left(F+0.25\right)$	$t:initialfilmthickness$ $F:fillfactor$ $\begin{array}{r}z:featuredepth;Cornell:theserules,\\ residualremovedbyO2RIEatabout100to\\ 110\mathrm{nm}/min;ResnickandChoi:inkjet\\ dispensingavoidsnon-uniformresidual\\ layers\end{array}$ $\begin{array}{r}(assumed:z=100\mathrm{nm},F=0.5;computed:\\ 70\mathrm{nm}<t<75\mathrm{nm},anarrowwindow)\end{array}$
Thermal imprint conditions	$\log \frac{t}{t_0}=\frac{-C_1\left(T-T_g\right)}{C_2+T-T_g}$	$\mathrm{PMMA}:180\text{to}200{}^{\circ }C,600\mathrm{psi}$	$T_g:glasstransition$ $\begin{array}{r}{C}_1,C_2:WLFconstants;Cornell:WLF\\ relationandresistconditions;Ryuet\\ al.:PMMAat438K,10to12.5\mathrm{bar}for1\\ s,fillingrising2.4to6.5timesfor\\ thickerfilms\end{array}$ $\begin{array}{r}(computed:600\mathrm{psi}is4.1MPa,about40\\ \mathrm{bar})\end{array}$

High-Volume J-FIL Performance

Name	Formula	Related Formulae	Variables/Constants
J-FIL throughput	${\text{wph}}_{station}=20\text{to}22.5$	${\text{wph}}_{cluster}=4\times {\text{wph}}_{station}$	$\begin{array}{r}ResnickandChoi:20wphperstationwith\\ 1.1sfill,80wphforfourstations;\\ Patsnap:22.5wphperstationand90wph\\ fortheFPA-1200NZ2Ccluster(seetheQ/A)\end{array}$ $\begin{array}{r}(computed:82full26x33\mathrm{mm}fieldsona\\ 300\mathrm{mm}wafertakeabout91soffillingat\\ 1.1s,halfofthe180sperwaferat20\\ wph)\end{array}$
Overlay	$O{L}_{SMO}<2.2\mathrm{nm},O{L}_{MMO}=4.0\mathrm{nm}(3\sigma )$	$\text{piezo magnification}+\text{DMD heating}$	$\begin{array}{r}SMO,MMO:single-machineandmix-and-match\\ overlay;ResnickandChoi:thesevalues,\\ 4.0\mathrm{nm}sufficientforNAND,distortion\\ residualsbelow1.3\mathrm{nm};Patsnap:3.4\mathrm{nm}\\ MMOand2.5\mathrm{nm}SMOin2018\end{array}$ $\begin{array}{r}(myreading:in1xcontactprinting,field\\ distortionmustbecorrectedthermallyand\\ mechanicallybecausenoprojectionlens\\ canscaleit)\end{array}$
Defectivity and mask life	${\rho }_{def}\approx 1.1\mathrm{c}{\mathrm{m}}^{-2}\text{at 325 wafers}$	$N_{particle}=8\times 10^{-4}\text{per wafer}$	$\begin{array}{r}ResnickandChoi:about3\mathrm{c}{\mathrm{m}}^{-2}initially,\\ 1.1\mathrm{c}{\mathrm{m}}^{-2}after325wafers,particleadders\\ 0.0008perwafer,replicamasklifetarget\\ above1000wafers;Patsnap:non-fillbelow\\ 1\mathrm{c}{\mathrm{m}}^{-2},earlytemplatelifeunder100\\ wafers\end{array}$ $\begin{array}{r}(computed:0.0008perwaferisone\\ particleevery1250wafers)\end{array}$

Resolution, Release and Cost

Name	Formula	Related Formulae	Variables/Constants
Resolution	$\text{better than }10\mathrm{nm}(\text{J-FIL})$	$14\mathrm{nm}\text{line width (FPA-1200NZ2C)}$	$\begin{array}{r}ResnickandChoi:betterthan10\mathrm{nm}with\\ LERandCDUmeeting2\mathrm{nm};Patsnap:14\mathrm{nm}\\ minimumlinewidth,10\mathrm{nm}target;Cornell:\\ about10\mathrm{nm}featuresfrome-beammasters\end{array}$ $\begin{array}{r}(myreading:resolutionissetbythe\\ template,writtenonane-beamtool,not\\ byoptics)\end{array}$
Template release	${\theta }_{water}\approx 110^{\circ }(\text{FDTS})$	$\text{CTE mismatch}\Rightarrow \text{separation stress}$	$\begin{array}{r}Cornell:FDTSanti-stickingmonolayerby\\ vapourdeposition,siliconorquartz\\ masters,separationtroublewithCTE\\ mismatch;Itoetal.:separationasa\\ resistrequirement\end{array}$ $\begin{array}{r}(myreading:alow-energysurfaceletsthe\\ curedresistreleasewithouttearing\\ features)\end{array}$
Cost	$C_{NIL}<50\text{M USD},C_{EUV}\approx 150\text{to}200\text{M USD}$	$\text{no multiple patterning or immersion steps}$	$\begin{array}{r}Patsnap:toolcostestimates;Resnickand\\ Choi:costofownershipadvantagefrom\\ avoidingmultiplepatterningandimmersion\\ andfromlongermasklife\end{array}$ $\begin{array}{r}(computed:theNILtoolisatmostathird\\ toaquarterofanEUVscanner)\end{array}$

Important Questions

Ques: What controls how fast a nanoimprint resist fills the template?

Pressure, film thickness, viscosity and feature width. Ryu et al. use the Stefan squeeze equation, $V=P{H}^3/(3\eta (2S{)}^2)$, valid when feature width exceeds film thickness, and show it overestimates filling near unity, so they add a correction that matches experiments within 10 per cent. Their PMMA filling rose 2.4 to 6.5 times for 1.5 to 2.5 times thicker films. With an assumed 1 MPa, 300 $\mathrm{nm}$ film, $10^4$ Pa s and 2.2 $\mu \mathrm{m}$ features, $V$ is about 0.19 $\mu \mathrm{m}$/s, so displacing 100 $\mathrm{nm}$ takes about 0.5 s (computed). Cornell gives the same scaling and the WLF temperature dependence, which is why thermal NIL runs PMMA at 180 to 200 ${}^{\circ }\mathrm{C}$ and 600 $\mathrm{psi}$, about 4.1 MPa (computed). Reference: J. Ryu, H. Lee, S.-H. Lee, H. Lim and J. Lee, ‘A modified squeeze equation for predicting the filling ratio of nanoimprint lithography’, Nano Convergence 4, 14 (2017): Stefan squeeze V(t) = P H^3/(3 eta (2S)^2) valid for indenter width over thickness above 1 and overestimating near unity, modified equation within 10 per cent, filling ratios 86 to 64 per cent as pressure rate rose from 5.5 to 50 bar/s, filling up 2.4 to 6.5 times for 1.5 to 2.5 times thicker films, PMMA 300 nm at 438 K, 10 to 12.5 bar held 1 s (fetch-tool summary); Cornell NanoScale Facility, ‘Thermal Nanoimprinting Basics’ (author not stated): imprint above Tg with pressure, Nanonex resists 140 C and 300 psi, mrI-2000 150 C and 350 psi, PMMA 180 to 200 C and 600 psi, imprint velocity V ~ p_eff h^3/(v R_eff^2), WLF log(t/t0) = -C1 (T - Tg)/(C2 + T - Tg), residual layer d = t - F z with 20 nm < d < z/4, FDTS anti-sticking monolayer with about 110 degree contact angle, silicon or quartz masters, about 10 nm features from e-beam masters, Nanonex NX-2000 with pressurized membrane, O2 RIE of residual layer at about 100 to 110 nm/min, non-fill, uneven residual layers and CTE separation problems (fetch-tool summary).

Ques: Why must the residual layer be controlled?

Because it must be etched through without eroding the features. Cornell defines the residual thickness as $d=t-Fz$, with initial film t, fill factor F and feature depth z, and recommends 20 $\mathrm{nm}$ $<d<z/4$; thinner layers are not recommended, and the residual is removed by oxygen RIE at about 100 to 110 $\mathrm{nm}$/min. For an assumed 100 $\mathrm{nm}$ depth at 50 per cent fill, the initial film must lie between 70 and 75 $\mathrm{nm}$ (computed), a narrow window. Cornell also warns that mixed-density patterns give uneven residual layers because the polymer cannot flow far, and Resnick and Choi note that dispensing resist drops locally, as in J-FIL, avoids the non-uniform residual layers of spin-coated films. Reference: Cornell NanoScale Facility, ‘Thermal Nanoimprinting Basics’ (author not stated): imprint above Tg with pressure, Nanonex resists 140 C and 300 psi, mrI-2000 150 C and 350 psi, PMMA 180 to 200 C and 600 psi, imprint velocity V ~ p_eff h^3/(v R_eff^2), WLF log(t/t0) = -C1 (T - Tg)/(C2 + T - Tg), residual layer d = t - F z with 20 nm < d < z/4, FDTS anti-sticking monolayer with about 110 degree contact angle, silicon or quartz masters, about 10 nm features from e-beam masters, Nanonex NX-2000 with pressurized membrane, O2 RIE of residual layer at about 100 to 110 nm/min, non-fill, uneven residual layers and CTE separation problems (fetch-tool summary); D. J. Resnick and J. Choi, ‘A review of nanoimprint lithography for high-volume semiconductor device manufacturing’, Adv. Opt. Technol. 6, 229 (2017): J-FIL resolution better than 10 nm, LER and CDU meeting 2 nm, single-machine overlay below 1.2 nm in field and 2.2 nm across wafer, mix-and-match 4.0 nm to ArF immersion and sufficient for NAND, 20 wph per station (80 wph four-station) needing 1.1 s fill, defectivity about 3 per cm2 initially and 1.1 per cm2 at 325 wafers mask life, particles 0.0008 per wafer, replica mask life target above 1000 wafers, sub-1.0 pL drops, inkjet-dispensed resist avoiding non-uniform residual layers, piezo magnification and DMD heating for distortion with residuals below 1.3 nm (fetch-tool summary).

Ques: Is J-FIL throughput 80 or 90 wafers per hour?

Both, at different times. Resnick and Choi report 20 wafers per hour per station, 80 for a four-station tool, and tie it to a 1.1 s fill time; earlier results were 15 wph at 1.5 s. The Patsnap roadmap gives 22.5 wph per station in 2018 and 90 wph for the FPA-1200NZ2C cluster, consistent with $4\times 22.5$ (computed). The 90 wph figure is the more recent and more likely current value, and 80 the 2017 status (my reading). At 20 wph, 180 s per wafer per station must cover about 82 full fields, whose filling alone takes about 91 s at 1.1 s each (computed), so fill time remains the main lever. Ito et al. list overlay, throughput and defectivity as the HVM criteria. Reference: D. J. Resnick and J. Choi, ‘A review of nanoimprint lithography for high-volume semiconductor device manufacturing’, Adv. Opt. Technol. 6, 229 (2017): J-FIL resolution better than 10 nm, LER and CDU meeting 2 nm, single-machine overlay below 1.2 nm in field and 2.2 nm across wafer, mix-and-match 4.0 nm to ArF immersion and sufficient for NAND, 20 wph per station (80 wph four-station) needing 1.1 s fill, defectivity about 3 per cm2 initially and 1.1 per cm2 at 325 wafers mask life, particles 0.0008 per wafer, replica mask life target above 1000 wafers, sub-1.0 pL drops, inkjet-dispensed resist avoiding non-uniform residual layers, piezo magnification and DMD heating for distortion with residuals below 1.3 nm (fetch-tool summary); Patsnap, ‘Canon nanoimprint lithography roadmap 2015-2026’: FPA-1200NZ2C four-station cluster, first delivered outside Canon in October 2024, 14 nm minimum linewidth (5 nm node equivalent) with a 10 nm target, overlay 5 to 7 nm (2017), 3.4 nm mix-and-match and 2.5 nm single machine (2018), sub-3 nm from 2020, throughput 15 then 22.5 wph per station and 90 wph cluster, non-fill defects below 1 per cm2, template life under 100 wafers in 2015 to 2016 with a 1000-wafer target, tool cost below 50 million dollars against about 150 to 200 million for EUV, fill 1.2 to 1.5 s (2016) to 1.1 s (2018) (fetch-tool summary); T. Ito, K. Emoto, T. Takashima, K. Sakai, W. Liu, J. DeYoung, Z. Ye and D. LaBrake (Canon), ‘Nanoimprint System for High Volume Semiconductor Manufacturing; Requirement for Resist Materials’, J. Photopolym. Sci. Technol. 29, 159 (2016), abstract: overlay, throughput and defectivity as criteria, FPA-1200NZ2C cluster, J-FIL photo-curable resist requirements for jetting, spread and filling, UV sensitivity, separation and etch durability (fetch-tool summary).

Ques: How good are nanoimprint overlay and defectivity?

Good enough for memory. Resnick and Choi report single-machine overlay below 1.2 $\mathrm{nm}$ within a field and 2.2 $\mathrm{nm}$ across the wafer, and 4.0 $\mathrm{nm}$ mix-and-match to an ArF scanner, which they call sufficient for NAND; piezo magnification and DMD heating correct distortion with residuals below 1.3 $\mathrm{nm}$. Patsnap lists 3.4 $\mathrm{nm}$ mix-and-match and 2.5 $\mathrm{nm}$ single-machine overlay in 2018. For defects, Resnick and Choi start near 3 per cm${}^2$, reach 1.1 per cm${}^2$ after 325 wafers, and cut particle adders to 0.0008 per wafer, one in 1250 wafers (computed), against a replica-mask life target above 1000 wafers. Because the template touches the resist, each particle can damage the mask (my reading). Reference: D. J. Resnick and J. Choi, ‘A review of nanoimprint lithography for high-volume semiconductor device manufacturing’, Adv. Opt. Technol. 6, 229 (2017): J-FIL resolution better than 10 nm, LER and CDU meeting 2 nm, single-machine overlay below 1.2 nm in field and 2.2 nm across wafer, mix-and-match 4.0 nm to ArF immersion and sufficient for NAND, 20 wph per station (80 wph four-station) needing 1.1 s fill, defectivity about 3 per cm2 initially and 1.1 per cm2 at 325 wafers mask life, particles 0.0008 per wafer, replica mask life target above 1000 wafers, sub-1.0 pL drops, inkjet-dispensed resist avoiding non-uniform residual layers, piezo magnification and DMD heating for distortion with residuals below 1.3 nm (fetch-tool summary); Patsnap, ‘Canon nanoimprint lithography roadmap 2015-2026’: FPA-1200NZ2C four-station cluster, first delivered outside Canon in October 2024, 14 nm minimum linewidth (5 nm node equivalent) with a 10 nm target, overlay 5 to 7 nm (2017), 3.4 nm mix-and-match and 2.5 nm single machine (2018), sub-3 nm from 2020, throughput 15 then 22.5 wph per station and 90 wph cluster, non-fill defects below 1 per cm2, template life under 100 wafers in 2015 to 2016 with a 1000-wafer target, tool cost below 50 million dollars against about 150 to 200 million for EUV, fill 1.2 to 1.5 s (2016) to 1.1 s (2018) (fetch-tool summary).

Ques: How do nanoimprint resolution and cost compare with EUV?

Comparable resolution at lower tool cost. Resnick and Choi state J-FIL resolution better than 10 $\mathrm{nm}$ with LER and CDU meeting 2 $\mathrm{nm}$, and Patsnap gives a 14 $\mathrm{nm}$ minimum linewidth for Canon’s cluster with a 10 $\mathrm{nm}$ target; the template, written by e-beam, sets this resolution (my reading; Electron-Beam Lithography Writer page). Patsnap estimates the tool at below 50 million dollars against about 150 to 200 million for an EUV scanner, a third to a quarter of the cost (computed), and Resnick and Choi attribute a cost-of-ownership advantage to avoiding multiple patterning and immersion steps. EUV still leads in throughput and logic overlay (EUV Scanner page). Reference: D. J. Resnick and J. Choi, ‘A review of nanoimprint lithography for high-volume semiconductor device manufacturing’, Adv. Opt. Technol. 6, 229 (2017): J-FIL resolution better than 10 nm, LER and CDU meeting 2 nm, single-machine overlay below 1.2 nm in field and 2.2 nm across wafer, mix-and-match 4.0 nm to ArF immersion and sufficient for NAND, 20 wph per station (80 wph four-station) needing 1.1 s fill, defectivity about 3 per cm2 initially and 1.1 per cm2 at 325 wafers mask life, particles 0.0008 per wafer, replica mask life target above 1000 wafers, sub-1.0 pL drops, inkjet-dispensed resist avoiding non-uniform residual layers, piezo magnification and DMD heating for distortion with residuals below 1.3 nm (fetch-tool summary); Patsnap, ‘Canon nanoimprint lithography roadmap 2015-2026’: FPA-1200NZ2C four-station cluster, first delivered outside Canon in October 2024, 14 nm minimum linewidth (5 nm node equivalent) with a 10 nm target, overlay 5 to 7 nm (2017), 3.4 nm mix-and-match and 2.5 nm single machine (2018), sub-3 nm from 2020, throughput 15 then 22.5 wph per station and 90 wph cluster, non-fill defects below 1 per cm2, template life under 100 wafers in 2015 to 2016 with a 1000-wafer target, tool cost below 50 million dollars against about 150 to 200 million for EUV, fill 1.2 to 1.5 s (2016) to 1.1 s (2018) (fetch-tool summary); Electron-Beam Lithography Writer page of this vault; EUV Scanner page of this vault.

References

1. Cornell NanoScale Science and Technology Facility, “Thermal nanoimprinting basics.” Available: https://www.cnfusers.cornell.edu/sites/default/files/Equipment-Resources/NanoImprint%20Basics_0.pdf
2. J. Ryu, H. Lee, S.-H. Lee, H. Lim and J. Lee, “A modified squeeze equation for predicting the filling ratio of nanoimprint lithography,” Nano Convergence, vol. 4, 14, 2017. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC5487822/
3. D. J. Resnick and J. Choi, “A review of nanoimprint lithography for high-volume semiconductor device manufacturing,” Advanced Optical Technologies, vol. 6, no. 3-4, pp. 229-241, 2017. Available: https://www.frontiersin.org/articles/10.1515/aot-2017-0020/pdf
4. Patsnap, “Canon nanoimprint lithography roadmap 2015-2026.” Available: https://www.patsnap.com/resources/blog/articles/canon-nanoimprint-lithography-roadmap-2015-2026/
5. T. Ito, K. Emoto, T. Takashima, K. Sakai, W. Liu, J. DeYoung, Z. Ye and D. LaBrake, “Nanoimprint system for high volume semiconductor manufacturing; requirement for resist materials,” Journal of Photopolymer Science and Technology, vol. 29, no. 2, pp. 159-168, 2016 (abstract read). Available: https://www.jstage.jst.go.jp/article/photopolymer/29/2/29_159/_article
6. Electron-Beam Lithography Writer page of this vault.
7. EUV Scanner page of this vault.

Related pages in this vault: Electron-Beam Lithography Writer, EUV Scanner, High-NA DUV Immersion Scanner, Overlay Metrology Tool, Defect Detection Fundamentals, Laser Direct-Write and Interference Lithography System.