ELSEVIER Microelectronic Engineering 41/42 (1998) 575-578

MICROELECTRONIC ENGINEERING

H o t e m b o s s i n g in p o l y m e r s as a d irec t w a y to p a t t e r n res is t

R.W. JaszewskP, H. Schifl", J. Gobrecht ~ and P. Smith ~

:' Laboratory for Micro- and Nanostructures, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland ~' Institute for Polymers, Swiss Federal Institute of Technology ETHZ, CH-8092 Ztirich, Switzerland

In this work, we investigated the possibilities of Hot Embossing Lithography as a new nanoreplication technique. Different structures with feature sizes down to 50 nm were successfully replicated into a resist over an area of up to 10 cm=. These polymer structures were then further transferred into a hard material by means of two different pattern transfer techniques. The aspect ratio and the shape of the various replicated nanostructures were monitored throughout the process by means of scanning electron and Ibrce microscopies.

! . INTRODUCTION

Current fabrication methods for nanostructures below 100 nm are either based on very slow serial processing (e-beam and SPM lithographies), require very costly technologies (X-ray lithography), or apply processes which result in a random patterning of the surface [1]. Hot Embossing Lithography (HEL) is a recently developed high-throughput low- cost replication method, which has demonstrated 10- nanometer resolution 12] . This manufacturing technique exploits the advantages of polymer molding, which makes it possible to pattern large surlaces with nanostructures in a parallel manner, and avoids the limitations of lithography by conventional exposure and development. During the embossing step, the original pattern is directly translErred into a thermoplast, which acts as a resist. When heated above its glass transition temperature, the polymer becomes viscous and conforms exactly to the embossing shim by filling the cavities of the surface relief. Alter it has cooled down, the replica is demolded l¥om the master. Our experiments have shown that the structures on the master are left undamaged by the replication process and can be re- used directly, thus permitting a large number of replicas to be obtained from a single master. The results of replication of diftErent structures ,arc presented in this paper. Wc also show that it is possible to achieve high enough aspect ratios for a variety of subsequent pattern transfer processes and present results obtained with two different techniques.

2. EXPERIMENT HEL consists basically of two different steps: the

first is imprint replication, where a polymer is structured by means of a hard master in a molding process (fig. l). In the second step, this surface relief is further transferred into a hard material.

Several pattern transfer techniques are available such as lift-off, Reactive Ion Etching (RIE), wet chemical etching or electroplating.

7

/ ] = th in fihzn OI pc*l } 111 (2 I

'~ '~ 1 1 pre~,stiie & heat }~mbogs;in g Rill I n a n o s tr t lct t l t ed

shim

I I i I i I ] Den'lolding a , , ~ , ~ , . . ~ - - r e m a i n i n g thin i"

Fig 1. First step: imprint replication in polymers. The thickness c~f the remaining polymer layer has been ex, ggerated.

2.1 Polymer films Polymethyhnethacrylate (PMMA) posscsscs a

number of crucial properties that are necessary for successful nanoreplication and was therefore chosen as the resist material lbr this technique. It is widely available in a broad range of molecular weights, is easily dissolved, and has small temperature and pressure shrinkage coefficients [31. On the other hand, it is the standard resist for e-beam lithography, and therefo,e interesting cross-links between imprint and e-beam lithographies could arise. PMMA, along with other polymers, has been widely used in Transmission Electron Microscopy sample preparation fin surface relief transfer. These investigations situated the resolution limit around 10 nm [4], corresponding to what HEL has achieved [2 I. Other polymers can also be structured and good results have been obtained with Polycarbonate [5 [. Granulates and powders of PMMA were dissolved in a number of organic solvents, of which chlorobenzene yielded the best results. The use of

0167-9317/98/$19.00 © Elsevier Science B.'~ All rights reserved. Pl I: SO 167-9317(98)00135-X

576 R. W. Jaszewski et al. / Microelectronic Engineering 41/42 (1998) 575-578

ethylmethylketone or ethylacetate produced films with significant surface nonuniformities in the form of radial waves. This is most probably the consequence of local density fluctuations of the solution and of the high volatility of these solvents [6]. Polished and cleaned silicon wafers were spincoated using solutions of PMMA with molecular weights ranging from 500k down to 25k. The coated wafers were then baked for 10 minutes at 170°C. The film thickness was chosen depending on the master characteristics (see section 3), and was between 50 and 300 nm.

2 . 2 Embossing: the parameters When heated above its glass transition temperature

(T~), the polymer reaches a viscoelastic state and can be shaped mechanically without damage to the shim. After conforming to the master, the polymer film is cooled to below the T~ and demolded. If not done carefully, the demolding step can lead to extended damage of the master, particularly for high aspect ratio structures. The Tg of PMMA is around 105°C, the exact value being dependent upon its molecular weight and other film properties [7]. We used embossing temperatures in the range of 130-160°C and pressures from 100 down to 20 bar, while the embossing time was typically a few minutes. The temperature, pressure, as well as embossing and cooling times on our press are computer controlled. Although it is possible to carry out the embossing under vacuum conditions, it was not necessary to do so in our experiments.

2 . 3 Pattern transfer by lift-off Hot embossing produces a surface relief in the

polymer resist, but in order to uncover the surface of the underlying silicon substrate, one has to remove a thin layer of polymer. Its thickness was between 20 and 50 nm and depended on the polymer used, the embossing parameters and the master characteristics. Removal was done by anisotropically etching the sample in an oxygen plasma. Next, a thin layer of titanium was evaporated onto the sample, and the metal was lifted off from the unstructured regions by dissolving the PMMA in chlorobenzene (fig.2).

Polymer ~ closed window replica

Dr:, etching ~ ~ ~' ~¢ . . . . . . ~ ..L .... ,O2 RIE ) ~ ~ ! ~ open wi,ldow

Deposition . . • metal of metal ~ ~ ] . . . . . polymer mask

I ,ift-ofl" of the ~ ? ~ ~ .~ : ~ nanostructured polymer mask ~ ~ ] metal layer

Fig 2. Schematic of the lift-off process.

2 . 4 Pattern transfer by direct RIE For pattern transfer by dry etching, a 200 nm-thick

SiO 2 layer was thermally grown upon the silicon substrate before spincoating PMMA. After embossing, the remaining thin polymer layer was removed by 02 RIE etching, as in the lift-off process. Next, the underlying SiO 2 layer was anisotropically etched in a CHF3/O 2 gas mixture using the polymer as a mask (fig.3).

Polymer closed window replica SiO 2 layer

Dry etching ~ ~ ~ ~ ~

(CHF 3 RIE) ~ . . . k . . . . . , ' - - etched Si©-~

Fig 3. Schematic of the direct etching process.

3 . R E S U L T S

Based on our experiments, the critical parameter in the embossing process is the temperature, with pressure and embossing time playing only a secondary role. In fact, the pressure and the embossing time can be optimized so as to make the process cycle time as short as possible. On the other hand, our experiments have shown that the right choice of embossing temperature as well as polymer thickness and type depends on the kind of structures which are to be replicated. Therefore, these parameters had to be adapted to the characteristics of each master. The attributes to he considered are the density, size, aspect ratio, profile type (see 3.2) and uniformity of the structures on the master. Features with different heights or alternating positive and negative structures are particularly difficult to replicate. Special attention must be theretbre given to the flow properties of the embossed polymer [8].

3 . 1 Resolution limits & structures Different structures, including lines, squares, dots,

curves and checkers, were successfully replicated (figs.4&5) and the conformation to the master was, as far as we could investigate, almost perfect. The minimum replicated feature size was 50 nm. Due to the high replication fidelity, the aspect ratios on the replica were equal to those on the master, which varied between 3:2 and 2:1.

3 . 2 Positive / negative profiles Most common structures can be divided in positive

and negative types, determined by the relative amount of polymer that has to be displaced during the embossing (fig.6).

R. W. Jaszewski et al./Microelectronic Engineering 41/42 (1998) 575-578 577

Figs 4&5. SEM images at a 85 ° tilt angle of structured PMMA. The "tombstones" and the boxes are 125 nm wide.

[ Positive profile ] [ Negative profile J

i l l a s t e l - I

Polymer 1 , replica

Fig 6. DifJbrent profile O'pes o f embossing masters.

In our replication experiments, the negative profiles proved to be less difficult to emboss. The investigation of the replica with SEM and SFM techniques was also lacilitated compared to the positive profiles.

3 . 3 Master durabi l i ty and large area e m b o s s i n g

Replication fidelity and master durability in our experiments depended on the adhesion between the shim and the polymer. The surface energy of the shim can be significantly lowered by the deposition of a thin anti-adhesive layer. We have previously shown that masters covered with ultra-thin teflon- like fihns can be used to emboss for at least 50 times without requiring an intermediate cleaning process [9]. In fact, most of the degradation of our samples was due to dust particles which were occasionally caught between the master and the

polymer, and which damaged both of them during the embossing step. The largest master which was replicated was a structured 10 cm= piece of silicon wafer, featuring grooves 200 nm-wide with a period of 660 nm. 11 was made by laser interference lithography trod anisotropic etching.

3 . 4 Pattern transfer: removal of the remaining polymer layer

Removal of the thin residual polymer layer results in a significant loss of aspect ratio, which drops to about 1:1 after RIE etching (fig.7). The etch rate of the polymer was as low as 40 nm/min, thus permitting a good control of the process.

r {nm] [ Po lymer repl ica

.fll /' ( I', ! 1 5 0 i j , l/L__. :! j L . i

0

0 2 4 6 I!,m] Aspect ratio: 3:2

[nmlf i~°lymorreplicaat-tel~" 1 anisotropic etctain~ ]

120 ~ l

[ i ' llli :ql Or- . . . . . . . . . . .

[ i i i i i i i

0 1 2 3 [ !~,in]

Aspect ratio: I: 1

Fi<~ 7. h!fluence qf O, RIE etching (m structure shape and aspect ratio, as monitored by SFM.

3 . 5 Pattern transfer: lift-off A 40 nm-thick layer of titanium was evalxmited

over the entire sample, and the polymer mask lifted- off in chlorobenzene. Large areas covered by 80 nm structures were successfully transfened with an aspect ratio of 1:2 (fig.8). For these structures, the deposition of a thicker metal layer resulted in a continuous titanium coverage and the metal could not be lifted off. Therefore, the maxinmm aspect ratio achievable with this technique was about 35~7~ of that of the master.

3 . 6 Pattern transfer: direet RIE etching In a second approach, we directly translcrred Ihe

pattern into the SiO, substrate (fig.3). TIno etch rLltOS

578 R.W. Jaszewski et al./Microelectronic Engineering 41/42 (1998) 575-578

of the polymer and the oxide were comparable; therefore, the final aspect ratio was close to 1: 1. The mask is chemically modified during etching and is difficult to remove afterwards. The etching also leaves tiny pieces of residual matter behind (fig.9), which stick to the substrate even alter an acetone rinse. By this technique, the aspect ratio of the hard replica is about 70% of that of the master.

Fig 8. SEM intages of 80 nm-diameter titanium dots, 40 nm in height.

r

5 0 0 n m

Fig 9. SEM image (50°tilt) of structures etched in SiOs. The structures are 15Ohm wide.

4 . C O N C L U S I O N S

Replication seems to be the solution to the throughput and cost issues in sub-100 nanometer lithography. Several novel techniques address this topic such as Microcontact Printing, where a monolayer of resist is deposited onto a surface by means of an elastic stamp [10]. In Hot Embossing Lithography, the resist is directly patterned and an aspect ratio comparable to that of the master is obtained. Limitations due to the use of energetic beams and chemical development are avoided. In this work, the best results were obtained with uniform reliefs featuring negative profiles. The structured polymer was used as a mask to transfer the pattern into a hmd material. Both lift-off and direct RIE etching yielded high quality hard replicas. The aspect ratio and resolution seem to have great potential for improvement, and, for the moment, they are only limited by the imperfections of the embossing shim, These preliminary results lead us to think that HEL has the capabilities to provide a powerful nanomanufacturing technique for the future.

5 . A C K N O W L E D G E M E N T

We would like to thank D.Btichle, Y.Bonetti, T.A.Jung, T.Mezzacasa, C.R.Musil trod H.A.Biebuyck (IBM Zfirich) for their valuable collaboration and contributions. This work is supported by the Swiss National Science Foundation (Priority programme NFP36).

R E F E R E N C E S

[11 L.Piraux et al., AppI. Phys. Lett., vol.65. 19, 2484-2486, 1994

[2] S.Y.Chou and P.R.Krauss, Microelectronic Engineering, 35. 237-240, 1997

[3] S.Y.Chou, P.R.Krauss and P.Renstrom, Science, 272, 85-87, 1996

[4] L.Reimer and C.Schulte, Die Naturwissensch~lfien, 53,489-497, 1966

[5] M.T.Gale, L.G.Bm'aldi and R.E.Kunz, Proc. SPIE (The htternational Society for Optical Engineering), vol.2213, 2- I 0, 1994

[6] L.L.Spangler, J.M+Torkelson and J.S.Royal, Polymer EngiHeering and Science, vol.30, l I, 644-653, 1990

[7] G.Beaucage, R.Composto and R.S.Stein, J. ~f Pol. Science B+ 3 I, 3, 319-326, 1993

[8] L.G.Baraldi, Dissertation ETHZ no. 10762+ 44- 61, 1994

[9] R.W.Jaszewski, H.Schift, P.Gr6ning ~md G.Margaritondo, Microelectronic Engineering, 35, 381-384, 1997

[10I H.A.Biebuyck, N.B.Larsen, E.Delamarche mid B.Michel, IBM Journal of Research (otd Development, vol.4 I, no. I-2, 159-170, 1997