process dependencies in nanoimprint

Transcription

1 Modeling and mitigating pattern and process dependencies in nanoimprint lithography 23 June 2011 Hayden Taylor Singapore-MIT Alliance for Research and Technology formerly based at: Microsystems Technology Laboratories, MIT

2 Collaborators and acknowledgements Funding MIT Singapore-MIT Alliance Duane Boning Danish National Advanced Cai Gogwilt Technology Foundation Matt Dirckx Eehern Wong NIL Technology Melinda Hale Kristian Smistrup Theodor Nielsen Helpful discussions Brian Bilenberg Hella Scheer Yoshihiko Hirai University of California, San Diego Dave White Andrew Kahng Yen-Kuan Wu 2

3 Spun-on vs droplet-dispensed resist in NIL Spun-on resist residual layer Droplet-dispensed resist time Resist viscosity 10 3 Pa.s Applied pressures ~ 5 MPa Thermoplastic or UV-curing Viscous resist squeezing Elastic stamp deflections S.Y. Chou et al., Appl. Phys. Lett. vol. 67 pp , 1995 S. Fujimori, Jpn. J. Appl. Phys. vol. 48 p. 06FH01, 2009 Resist viscosity < 0.1 Pa.s Applied pressures ~ 5 kpa Droplets tailored to pattern Key figure of merit: filling time Gas trapping and dissolution M. Colburn et al., SPIE 3676, pt.1-2, ,

4 NIL pattern and process dependencies have systematic and random components 4

5 Nanoimprinting of spun-on layers exhibits pattern dependencies Two relevant timescales for pattern formation: Local cavity filling Residual layer thickness (RLT) homogenization 5

6 We need a unified simulation approach for micro- and nano-embossing/imprinting Initial polymer thickness, r 0 10 mm 1mm Biological micro-/nano-devices stamp w r 0 polymer substrate 100 µm Tissue engineering Diffractive optics 10 µm 1 µm 100 nm Semiconductors Hard-disk drives Photovoltaics Photonics Flat-panel displays Metamaterials Planarization 1 nm 10 nm 100 nm 1 µm 10 µm 100 µm Cavity width, w Taylor, NNT

7 We need a unified simulation approach for micro- and nano-embossing/imprinting Initial polymer thickness, r 0 10 mm 1mm 100 µm 10 µm 1 µm 100 nm Biological micro-/nano-devices Tissue engineering Diffractive optics Photovoltaics Flat-panel displays Planarization Semiconductors Photonics Metamaterials Hard-disk drives Cavity 1 nm 10 nm 100 nm 1 µm 10 µm 100 µm width, w Taylor, NNT

8 Key: model impulse response g(x,y,t) of resist layer Resist Model in space: x g Mechanical impulse applied uniformly over small region at time t = 0 Resist Substrate Model in time: Newtonian: impulse response constant in time for t > 0 Viscoelastic: impulse response is function of time. Taylor, NNT After Nogi et al., Trans ASME: J Tribology, (1997) 8

9 Change in topography is given by convolution of impulse response with pressure distribution Small, unit disp. p(x,y,t) )? Stamp Resist Substrate Time increment p ( x, y, t) g( x, y, t) t 1 Pressure Impulse Unit displacement response in contact region? Taylor, NNT

10 Contact pressure distributions can be found for arbitrary a stamp geometrieset es 2.3 µm-thick polysulfone film embossed at 205 C under 30 MPa for 2 mins Stamp design Simulated pressure Optical micrograph Cavity 200 µm MPa Taylor et al., SPIE 7269 (2009).

11 Successful modeling of polysulfone imprint 2.3 µm-thick polysulfone film embossed at 205 C under 30 MPa for 2 mins Taylor et al., SPIE 7269 (2009). 11

12 Representing layer-thickness reductions 12

13 Modeling stamp and substrate deflections Indentation Indentation and bending t stamp λ λ Magnitude of stamp deflection (log scales) ~4 λ/t stamp 13

14 Simulation method: step-up resist compliance PMMA 495K, c. 165 C, 40 MPa, 1 min Taylor, NNT

15 Abstracting a complex pattern Local relationships between pressure history and RLT: Abstractions: Stamp: pointload response Resist: impulse response HK Taylor and DS Boning, NNT 2009; SPIE 7641 (2010) Wafer: point-load response 15

16 Our NIL simulation technique has been experimentally validated PMMA 495K (200 nm), 180 C, 10 min, 16 MPa, 10 replicates Si stamp Residu ual layer thickn ness (mi icron) cavity 0.1 protrusion 1 mm Cavity proportions filled A B C D E F G H 550 nm-deep cavities: Exp t Simulation A C E G B D F Lateral position (mm) H

17 Simulation time Simulation time (s) N 10 4 Expected: time ~ O(N 2 logn) Stamp 1 Feature-scale Stamp 2 Simulation size, N Stamp 2 Abstracted 17

18 Strengths of the simulation method A unified simulation approach Can cope with any layer thickness Can integrate feature sizes ranging over many orders of magnitude Can model any linear viscoelastic material Speed At least 1000 times faster than feature-level FEM Implicit periodic boundary conditions are useful Realistic representation of whole-wafer wafer imprint of many chips Can use edge-padding for non-periodic modeling Suited to quick adaptation for new NIL configurations Use to explore the use of flexible stamps and substrates Explore the imprinting of non-flat substrates Micro-contact t printing; roll-to-roll 18

19 Varying stamp s bending stiffness: simulations Stamp thickness: 5 mm 0.5 mm 0.12 mm Features 200 nm 4 mm Residual layer thickness 19

20 Long-range compliance and short-range rigidity are both desirable in a TNIL stamp Long-range compliance to allow the stamp to conform to random wafer nanotopography Short-range rigidity to limit systematic pattern dependenciesd Making the stamp soft (i.e. polymeric) or thin satisfies the first aim but not the second Structuring the stamp can meet both needs 20

21 Structured stamps provide long-range compliance and short-rangerange rigidity A mechanical model of a structured stamp is needed: To ensure adequate long-range compliance while keeping fabrication affordable and maximizing the stamp area available for product features. T Nielsen, et al., Proc. 18th IEEE Conf. MEMS 2005, pp HK Taylor, K Smistrup, and DS Boning, MNE

22 Even a small flexure-gap increases wafer-scale stamp compliance several-fold HK Taylor, K Smistrup, and DS Boning, MNE

23 Simulations using a measured wafer topography illustrate long-range compliance Roughness spectra of three virgin silicon wafers HK Taylor, K Smistrup, and DS Boning, NNT

24 Simulations using a measured wafer topography illustrate long-range compliance HK Taylor, K Smistrup, and DS Boning, NNT

25 Simulations using a measured wafer topography illustrate long-range compliance Mean within- Mesa-to- mesa std. dev. (nm) mesa std. dev. (nm) Undeformed stamp topography Simulated RLTs t g = 100 µm t g = 150 µm no grooves HK Taylor, K Smistrup, and DS Boning, NNT

26 Die-scale simulations show that structuring the stamp reduces local pattern dependencies RH Pedersen, et al., J. Micromech. Microeng., vol. 18, p , HK Taylor, K Smistrup, and DS Boning, MNE

27 Structured stamps also allow for decoupling of differently patterned adjacent mesas HK Taylor, K Smistrup, and DS Boning, MNE

28 Cavity-filling time depends on length-scale of pattern-density variation, and stamp stiffness Lower-density region fills by: Lateral flow Lateral flow and stamp deflection HK Taylor, NNT

29 Cavity-filling time depends on length-scale of pattern-density variation, and stamp stiffness HK Taylor, NNT

30 If imprinted layer is an etch-mask, RLT specifications depend on resist properties (h + r max )/r max must be large enough for mask to remain intact throughout etch process Largest allowable r max r min is likely determined by lateral etch rate and critical dimension specification HK Taylor, NNT

31 Time to satisfy target for RLT uniformity scales as ~W 2 for ρ above a threshold W (µm) HK Taylor, NNT

32 We postulate a cost function to drive the insertion of dummy fill into rich designs tˆ fill N 2 Wi p0 r0 h 1 0 i r0 h i 2 W i Abutting windows of size W i swept over design ρρ i is maximal density contrast between abutting windows in any location Objective is to minimize sum of contributions from N+1 window sizes h: protrusion height ht on stamp r 0 : initial resist thickness 32

33 We postulate a cost function to drive the insertion of dummy fill into rich designs insertion of dummy fill into rich designs N W N i i i fill h r h r p W t ˆ W i 2 33

34 A simple density-homogenization scheme offers faster filling and more uniform RLT Metal 1 of example integrated circuit: min. feature size 45 nm Stamp protrusion pattern density: without dummy fill Characteristic feature pitch (nm) Predominant feature orientation µm HK Taylor, NNT

35 A simple density-homogenization scheme offers faster filling and more uniform RLT Density: without fill Density: with fill Designed protrusion 1 µm Available for dummy µm 0 35

36 If stamp cavities do not fill, smaller RLTs are possible but RLT may be less uniform HK Taylor, NNT

37 Increasing keep-off distance may reduce IC parasitics, but degrades RLT performance HK Taylor, NNT

38 Summary: modeling and mitigation of process and pattern dependencies in NIL Thermal NIL Modeling Mitigating Stamp s Resist s elastic plastic deflection deformation Pattern abstraction Structured stamps: long-range compliance, short range rigidity Design rules for pattern density uniformity; dummy fill insertion Ongoing: extend to UV-NIL: Capillary pressures Gas bubble trapping Droplet spreading Mechanical proximity correction