esolution, LE, and Sensitivity Limitations of Photoresist Gregg M. Gallatin 1, Patrick Naulleau,3, Dimitra Niakoula, obert Brainard 3, Elsayed Hassanein 3, ichard Matyi 4, Jim Thackeray 4, Kathleen Spear 4, Kim Dean 5 1., Newtown, CT. Lawrence Berkeley National Laboratory, Berkeley, CA 3. University of Albany, College of Nanoscale Science and Engineering, Albany, NY 4. ohm and Haas Microelectronics, Marlborough, MA 5. SEMATECH, Albany, NY Advanced Materials esearch Center, AMC, International SEMATECH Manufacturing Initiative, and ISMI are servicemarks of SEMATECH, Inc. SEMATECH, the SEMATECH logo, Advanced Technology Development Facility, ATDF, and the ATDF logo are registered servicemarks of SEMATECH, Inc. All other servicemarks and trademarks are the property of their respective owners
This work is part of a SEMATECH-funded program designed to 1. Generate, analyze, and compare multiple resists to an LE model. Use the experimental results to verify, expand, and finetune the LE model 3. Use the verified model to determine approaches for breaking the esolution, LE, and Sensitivity (LS) tradeoff At the end of the talk, two ideas for breaking the LS tradeoff will be discussed: 1. Anisotropic acid diffusion Anisotropic deprotection blur NOT EUV Specific. Increased quantum yield ~ EUV Specific For a detailed discussion of these ideas see Gallatin, et. al., SPIE 007.
Agenda 1. Describe the LS tradeoff. Show how it relates to an LE model 3. Describe the experiments done so far 4. Present comparisons of the experimental results to the LE model 5. Discuss the (two) approaches for breaking the LS tradeoff See See Poster Poster E-P0: E-P0: obert obert Brainard, Brainard, et et al., al., for for a detailed detailed description description of of the theresists studied studied See See Poster Poster E-P04: E-P04: Patrick Patrick Naulleau, Naulleau, et et al., al., for for a a detailed detailed evaluation evaluation of of resist resist resolution resolution metrics metrics 3
The esolution, LE, Sensitivity (LS) Tradeoff esist esolution: PEB Diffusion or esist Blur. Smaller is better LE: Line Edge oughness... Smaller is better Sensitivity: Dose-to-Size. Smaller is better BUT data and modeling indicate the following constraint : 3 Blur LE Dose ~ Constant LS Tradeoff For a standard chemically amplified resist and process cannot have blur, LE and dose all small at the same time. You can t always get what you want Mick Jagger This type of behavior has been found by many researchers: Lammers, et al., SPIE 007, Bristol, et al., SPIE 007, Brainard, et al., SPIE 004, Gallatin SPIE 005,... 4
How Does the LE Model LS Tradeoff LE MODEL contains 3 fundamental processes Gallatin SPIE 005 1. Exposure: Image intensity ~ Probability distribution for acid release. Acid release positions are statistical Sensitivity. PEB: Acids diffuse and deprotect the resist PEB diffusion range esolution (resist blur ) 3. Development: Spatial distribution of deprotection determines final resist profile Line edge statistics LE do the math σ LE C I I edge T α Q E size 3 I = Image intensity T = esist Thickness α = Absorptivity Q = Quantum Efficiency LE Constant Dose Blur = PEB Diffusion ange LS Tradeoff BONUS: Get the explicit analytic form for the frequency content of the LE Compare predicted content to experiment 5
EUV LE DATA Combined effort of LBL, CNSE, and ohm and Haas with SEMATECH funding Exposures were done on the 0.3 NA MET at Berkeley Three different resists with 4 or 5 different base loadings each esist Names: 5435 571 5496 Features imaged: 50 nm and 60 nm 1-to-1 lines/spaces CD and LE data through dose and focus at each base loading LE and PSDs computed from average of left and right edge data at each focus, dose, and base loading condition LE computed from both filtered and unfiltered PSD data Best Dose is as indicated in the graphs and tables Best Focus is set to 0, by definition esist blur values,, are fit using the analytical PSD formula: esist blur is the only fitting parameter 6
Example esult: esist 5435, Base Loading G, 50 nm 1-1 lines/spaces Best Dose LE(unfiltered) Best Focus Dose\Focus -150-100 -50 0 50 100 150 1.8 5.5 8.8 13.44 8.4 6. 5. 5.5 14.11 11.7 4.4 5. 5.1 7.4 14.8 9.9 5. 4.6 4.6 5.3 8. 15.56 6.8 6.3 5. 4. 5.1 5.7 15.8 16.34 8.4 4.9 4.3 5. 5.1 14. 17.15 10.7 5. 4.7 4.9 8. 18.01 7.3 6.5 6.8 10.7 16.5 3. 18.91 14.7 8.6 8.8 11.5 5. Best Dose ( blur ) Best Focus Dose\Focus -150-100 -50 0 50 100 150 1.8 18 16 13.44 5 0 17 14.11 19 0 5 18 17 14.8 0 0 18 0 1 15.56 1 4 1 16 17 19 5 16.34 17 16 18 30 17.16 3 5 19 19 38 18.01 0 18 9 1 1 1 18.91 18 19 4 0 3 Average = 1. nm D = 4. nm rms Best Dose Best Dose 7
LE versus Dose Solid Lines = Filtered LE (High frequency noise removed from PSD before computing LE) esist 5435 LE 3σ 10 8 6 4 50 nm 1-1 60 nm 1-1 Dashed Lines = Unfiltered LE (aw PSD used to compute LE) Comments Difference between filtered and unfiltered LE values ~ 0.5 nm Saturation at high dose is clearly evident Most of the difference between 50 nm and 60 nm LE comes from the slight difference in image log slope( I / I ) edge esist 571 esist 5496 LE 3σ LE 3σ 0 0 5 10 15 0 5 30 35 10 8 6 4 0 0 10 0 30 40 50 60 10 8 6 4 60 nm 1-1 60 nm 1-1 50 nm 1-1 50 nm 1-1 0 0 5 10 15 0 5 30 35 Dose-to- Size (mj/cm ) Dose-to- Size (mj/cm ) Dose-to- Size (mj/cm ) 8
LE versus Dose Add saturation to model Dots = LE Data Lines = Model fit to the data after incorporating saturation 1 1 E ( size E 1 e sat ) / E size Esat esist 5435 esist 571 LE 3σ LE 3σ 10 8 6 4 0 0 5 10 15 0 5 30 35 10 8 6 50 nm 1-1 60 nm 1-1 50 nm 1-1 Dose-to- Size (mj/cm ) Comments 4 60 nm 1-1 esults aren t bad, but more work needs to be done. Specifically: esist 5496 LE 3σ 0 0 10 0 30 40 50 60 10 8 6 50 nm 1-1 Dose-to- Size (mj/cm ) Match fitting parameter E sat to actual resist parameters, e.g., PAG loading, base loading, absorption, quantum efficiency, etc. 4 60 nm 1-1 0 0 5 10 15 0 5 30 35 Dose-to- Size (mj/cm ) 9
esist Blur Value,, versus Dose esist 5435 Blur 8 6 4 50 nm 1-1 Solid Lines = at best dose and focus Dashed Lines = Average over all dose and focus values Comments esist 571 Blur 0 18 16 14 0 5 10 15 0 5 30 35 18 16 60 nm 1-1 60 nm 1-1 Dose-to- Size (mj/cm ) 50 and 60 nm blur values approximately the same. Possibly some systematic variation with dose. 1σ variation in ~ 1 to 4 nm Difficult to distinguish systematic from random behavior. esist 5496 Blur 14 1 0 10 0 30 40 50 60 0 18 16 14 60 nm 1-1 50 nm 1-1 50 nm 1-1 1 0 5 10 15 0 5 30 35 Dose-to- Size (mj/cm ) Dose-to- Size (mj/cm ) 10
How well does the model PSD shape fit the data PSD shape? Next 5 slides show an example comparison of model PSD to data PSD esist 5435, 50 nm 1-to-1 lines/spaces Plots show model PSD and data PSD at all dose and focus and base loading (dose-to-size) values ed curves = Model PSD fit to one parameter, the value of Blue curves = Data PSD esults indicate that the PSD shape is remarkably stable through dose, focus, base loading, and resist type Will go quickly through the next 5 slides. Just look at the general comparison of the model and data PSDs 11
esist 5435 Dose-to-Size = 3.09mJ/cm =5 nm = nm =3 nm =5 nm =6 nm =18 nm =3 nm =18 nm =3 nm =17 nm =18 nm 1
esist 5435 Dose-to-Size = 4.73mJ/cm =3 nm =1 nm =18 nm =3 nm =1 nm =16 nm =14 nm =3 nm =6 nm =3 nm =0 nm =5 nm =6 nm = nm =0 nm =6 nm =0 nm = nm =0 nm =19 nm =16 nm =15 nm =15 nm 13
esist 5435 Dose-to-Size = 7.50mJ/cm =0 nm =16 nm =6 nm =1 nm =8 nm =8 nm =8 nm =1 nm =30 nm =8 nm =9 nm =1 nm =6 nm =1 nm =4 nm =6 nm =5 nm =1 nm =4 nm =18 nm =1 nm =8 nm =5 nm =8 nm =3 nm =17 nm =18 nm =8 nm =19 nm =18 nm =18 nm =7 nm =1 nm =6 nm =4 nm =30 nm 14
esist 5435 Dose-to-Size = 14.8mJ/cm =18 nm =16 nm =5 nm = nm =0 nm =17 nm =19 nm =0 nm =5 nm =18 nm =17 nm =0 nm =0 nm =18 nm =0 nm =1 nm = nm =1 nm =4 nm =1 nm =16 nm =17 nm =19 nm =5 nm = nm =17 nm =16 nm = nm =18 nm =30 nm =3 nm =5 nm =19 nm =19 nm =38 nm =0 nm =18 nm =9 nm =1 nm =1 nm =1 nm =18 nm =19 nm =4 nm =0 nm =3 nm 15
esist 5435 Dose-to-Size = 30.87mJ/cm =18 nm =19 nm =18 nm =16 nm =16 nm =1 nm =16 nm =17 nm =18 nm =1 nm =15 nm =14 nm =17 nm =1 nm =19 nm =17 nm = nm =0 nm =18 nm =3 nm =1 nm =16 nm =14 nm =17 nm =1 nm =18 nm =19 nm =9 nm =18 nm = nm =18 nm = nm =8 nm = nm =17 nm =1 nm =3 nm =16 nm =18 nm =38 nm =4 nm =3 nm =16 nm =31 nm = nm =18 nm 16
1. Anisotropic esist Blur Spherical Deprotection Blur: = = x y z z y Volume ~ x y z x Anisotropic Deprotection Blur: Shrink in x and y Expand in z x x s Volume ~ s x s y y s z s y = x y z z s z s = Scaling Factor s 1 do the math LE ~ 1/ s Horizontal Blur ~ 1/s Dose = Fixed Improved LE and resolution Fixed dose Anisotropic diffusion explicitly breaks the LS tradeoff. 17
. Quantum Yield = Q 48 nm and 193 nm photons release acids Energy required to release an acid is at most ~ 5 to 6 ev. EUV photon has 9 ev of energy Each EUV photon has enough energy to release at least 9 ev/5 ev~18 acids DATA: Q ~ 1/ 3 DUV Brainard, et al., SPIE 004 QEUV ~ acid bottleneck Neureuther, et al., JVST B 006 1 in 3 absorptions results in acid release EUV is not close to using its energy efficiently BUT Not all acids can be released in the same position adds blur Assume acids are released along random walk path with steps spaced by ρ 1/3 Q released acids Extra exposure blur r ~ (Q 1) 1/ ρ 1/3 18
. Quantum Yield = Q Combine deprotection blur from each acid with the random walk distribution of released acids Net Blur adius Acid Photon Acid Net Blur adius ~ / 3 ( Q 1) + r = + ρ Acid Acid Added Blur 3 1 0 = Deprotection blur radius = FWHM/ ~ 15nm Net Blur adius Q = 1 0 19 18 17 16 15 Q = Q = 8 Q = 4 Q =16 Spatial distribution of Q released acids 0.05 0.10 0.15 0.0 Clearly want high PAG density to avoid significantly increasing blur ρ (PAG s/nm 3 ) 19
. Quantum Yield = Q LE 1 α Q E 3 Increase Q and Decrease E with Q E = constant Explicit gain in sensitivity If net blur = net ~ no change in resolution Increased quantum yield can break the LS tradeoff. No change in LE NOTE: Some recent experiments show decreasing E and decreasing LE with increasing PAG loading Choi, et al., SPIE 007 Leunissen, et al., MNE 005. 0
Conclusions and Future Tasks: LS tradeoff predicts that for a standard chemically amplified resist and process you cannot get high resolution, low LE and low sensitivity all at the same time. Both anisotropic dissolution and increased quantum yield are good candidates for breaking the LS tradeoff If you try sometimes you just might find you get what you need. Mick Jagger Future: 1. Continue verification, expansion, and finetuning the LS model based on experimental results.. Need to consider non-mean-field behavior. Smith, Biafore, obertson, SPIE 007 3. Determine feasibility of implementation. This work is funded by SEMATECH 1