Microlens array produced using hot embossing process

Microelectronic Engineering 60 (2002) 365–379www.elsevier.com/ locate /mee

Microlens array produced using hot embossing process*N.S. Ong , Y.H. Koh, Y.Q. Fu

School of Mechanical and Production Engineering, Nanyang Technological University, Nanyang Avenue, Singapore639798, Singapore

Received 1 May 2001; accepted 30 August 2001

Abstract

In this paper, the fabrication of molds that are suitable for the production of microlens arrays using thereplication technique is discussed. Variation of parameters in the replication process were investigated. Afocused ion beam was used to fabricate the microlens cavities on three materials, with silicon showing the bestresult. Hot embossing was used to produce replicated polycarbonate lens array. The temperature of the mold andthe embossing force were the two parameters varied. The microlens array produced using the embossingreplication process demonstrates the possibility of nanometre fabrication. 2002 Elsevier Science B.V. Allrights reserved.

Keywords: Microlens array; Focused ion beam; Hot embossing

1. Introduction

The increase in research into control-by-light systems has widened the market for the use ofmicrolens array. Indeed, microlens array has a large field of application; for example high-speedphotography, telecommunication industry that couple light in and out optical fiber waveguides andoptical communication [1]. Also the use of optical control systems over their electrical counterpartsoffers a large number of advantages such as immunity to electromagnetic interference, safety inflammable areas, weight and cost savings, etc. Many methods of fabricating microlens were presented.Some examples are contactless embossing molding [2], the melted photoresist method [3] andmicrojet fabrication [4]. In this research, a focused ion beam is used to produce lens pattern on a moldmaterial, which will then be used for embossing to produce microlens array.

Focused ion beam (FIB) technology is well known and widely used in semiconductor manufactur-ing. The FIB system uses liquid metal gallium as the ion source. In the ionization process, gallium

*Corresponding author. Tel.: 165-799-55-37; fax: 165-791-18-59.E-mail address: [email protected] (N.S. Ong).

0167-9317/02/$ – see front matter 2002 Elsevier Science B.V. All rights reserved.PI I : S0167-9317( 01 )00695-5

366 N.S. Ong et al. / Microelectronic Engineering 60 (2002) 365 –379

Fig. 1. Steps involved in hot embossing process.

atoms tend to lose one electron, thus becoming singly-charged positive ions. Being charged particles,ions can be accelerated, focused and controlled by electrostatic fields. Their relatively high mass(compared with that of subatomic particles) allows them to be used to induce the milling anddeposition effects. Factors affecting FIB milling such as beam limiting aperture size, neighbouringspace of beam spot, dwell time and milling sequence were reported [5–7].

Fig. 1 shows the steps involved in hot embossing. A sheet of plastic foil /material is sandwichedbetween a mold (embossing tool) and an optically smooth backing plate was heated under pressure toa temperature (typically . 508C) above the softening temperature (T ) of the plastic. Higherg

temperatures are favourable as the lower viscosity of the polymer facilitates the molding process.After molding, the polymer is cooled down to below the glass transition temperature. The moldingforce is maintained during cooling in order to preserve the polymer microstructures from distortion.Once the polymer is cooled to below T and the pressure is released, the plastic can be separated fromg

the mold to give a high quality copy of the planar microstructure. No material shrinkage was foundduring the hot embossing process [8]. A simple filling mechanism governing the flow of polymer inhot embossing was described [9]. The factors governing hot embossing were reported [10];temperature, embossing force and time were the three main factors. This technique works extremelywell for shallow microstructures (relief depths less than 1 mm). Deeper structures, where aspect ratiowas as high as seven, were also reported [11]. Such an embossing technique can be carried out in thelaboratory using a relatively unsophisticated hot press.

2. Fabrication of micro-molds

The manufacturing of the microlens array involved 2 steps. Firstly, FIB milling is done on a moldmaterial. Once the pattern obtained on the mold is deemed satisfactory, it is then used as a mold toemboss the plastic lens array.

2.1. Focused ion beam milling

The material used for the mold must be polished to surface finish of below 10 nm Ra. This isbecause when the ion hits the surface of the material, the depth of material being removed will beaffected if there are large surface irregularities. The material was ground by abrasives paper of gritsize 180, 400, 800, 1000 progressively. The surface was then polished with 3- and 1-mm sized

N.S. Ong et al. / Microelectronic Engineering 60 (2002) 365 –379 367

Fig. 2. Fabrication of microlens profile using FIB.

diamond paste at a rotating speed of 150 rpm. The polished material was then cleaned in an acetoneultrasonic bath.

Fig. 2 shows the lens profile generation. The inner circle path will be programmed with a higher iondose while the ion dose in the outer circle path will be subsequently reduced. It is this varation in theion dose that generates the depth of the lens profile. At a constant aperture, the depth increases as theion dose increases. A higher ion dose indicates that there are a greater number of incident ions perunit area. Each of them removes particles from the material and thus generate a deeper feature. Thesequence used in the fabrication of lens pattern was from periphery to the center of the profile. A250-mm aperture size was used for the FIB system. The dwell time was set at 5 ms. Fig. 3 shows thedesign lens profile. The diameter is 70 mm with a depth of 3.45 mm.

Three different materials were used in the fabrication of the mold using FIB. As can be seen inFigs. 4 and 5, there are numerous small pitting holes on the mold surface. It is believed that one of thereasons for this problem is that the material is not homogenous. Another reason is because the grain

Fig. 3. Microlens profile.


Fig. 4. SEM picture of a lens profile on pure nickel.

structure for pure nickel and stainless steel materials were too big. This causes an uneven ‘tearing’effect due to the bombardment of the ions during the ion milling process. A lens profile was alsomilled on a silicon wafer using FIB (see Fig. 6). A surface finish of around 7 nm Ra was obtained.

Fig. 5. SEM picture of a microlens profile on stainless steel.


Fig. 6. SEM picture of a microlens profile on silicon wafer.

Silicon was therefore selected for the lens array mold fabrication. Furthermore, it was noted thatsilicon has several advantages [12]. The silicon must be thick enough to withstand the force appliedduring the embossing process. A 7 3 5 lens array was milled on a 9-mm thick silicon (see Fig. 7).

2.2. Hot embossing

In the embossing process, polycarbonate material was sandwiched between a flat nickel plate andthe mold. Heat was applied to above the T temperature (1488C) of the polycarbonate material.g

Embossing force was applied and held for 20 min. The mold was subsequently cooled to 268C withthe force maintained to preserve the microstructures of the lens. Demolding was then performed.Demolding is the separation of the mold from the embossed polymer structure by a vertical movementof the mold.

3. Results and discussion

Fig. 8 shows a lens profile milled on the 9-mm thick silicon material. It can be seen that the surfaceprofile was very well defined and the surface roughness was measured to be around 4 nm Ra.


Fig. 7. Microlens array for mold 4.

3.1. Embossing with varying temperatures (temperature test)

The silicon mold was used in this experiment. The temperature of the plastic (in this research, theplastic used was polycarbonate) was varied while the embossing force was kept constant at 2.22 kN. Itwas found that the surface finish of the mold deteriorates after each embossing. The following couldbe the reasons. Firstly, the polycarbonate (PC) was demolded at room temperature. There was friction

Fig. 8. 2-D profile of mold 2 measured using WYKO NT 2000.


generated between the mold surface and the cooled polycarbonate material during demolding.Secondly, frictional force introduced during the flow of plastic into the lens cavity may have causedthe Ra of the mold to deteriorate. However, the dimensions of the mold cavity were found to beunchanged. Therefore, new molds were milled and used in each embossing process so that a fair testis employed when comparing the surface finish of the mold cavity and the PC lens.

In order to reduce thermally induced stresses in the material as well as replication errors due to thedifferent coefficient of thermal expansion between the embossing tool and polymer, the change oftemperature during the cooling should be as small as possible [13]. Therefore, a test was done toascertain whether the demolding temperature could help reduce the surface roughness of the embossedlens and the mold cavity surface. A 4 3 4 lens pattern array (mold 3), of diameter 65.2 mm and depth4.6 mm, was fabricated using the FIB system. The demolding temperature was raised to 608C. Theembossing temperature and force were 1988C and 2.22 kN, respectively. The replicated lens diameterwas 63.5 mm with a depth of 4.9 mm.

Fig. 9b shows the 3D profile of replicated lens molded by mold 3 with a demolding temperature of608C. It can be seen that a part of the replicated lens was damaged. It was discovered that at a higherdemolding temperature, the PC lens tends to be sticky and was stuck inside the silicon mold. Thisbehaviour makes demolding very difficult as it damages the replicated lens. The height of thereplicated lens profile was large in comparison to the mold cavity. During demolding, the PC lens wasstuck inside the cavity and became distorted when taken out. The PC lens, therefore, experienceplastic deformation during the demolding process.

Fig. 10 shows the comparison of the profiles of mold 3. The surface of the mold also deteriorates ata higher demolding temperature. But more importantly, plastic deformation and the damageintroduced to the plastic lens at a higher demolding temperature renders the use of higher demoldingtemperature unfavourable in this work.

Fig. 11(b) shows the lens produced by mold 3 using the same temperature and embossing force of1988C and 2.22 kN, respectively. The only difference was that the lens was demolded at roomtemperature. The profile was not damaged or distorted. The height measured was 4.6 mm. Themechanical behaviour of plastics is very dependent on the service temperature. At low temperatures,plastic behaves as a hard solid with a high modulus and a low extensibility. However, as the

Fig. 9. Profile of replicated lens of mold 3 (demold at 608C).


Fig. 10. Profile of mold 3; (a) after molding, (b) before molding.

temperature increases, the modulus decreases and extensibility increases slightly. Therefore, it can besaid that distortion and extension of lens profile occurs more extensively at a higher demoldingtemperature.

Fig. 12 shows SEM pictures of the replicated lens array molded using mold 4 where the embossingtemperature and force were 1988C and 2.22 kN, respectively. Table 1 gives a summary of data for thevarious molds that were used in this test. The dimensions of the molds vary slightly as it is notpossible to obtain the same dimensions using FIB. However, the purpose of this research work is toobtain the optimal temperature setting for hot embossing of PC material on a silicon mold.

Four points on the replicated plastic lens were measured (see Fig. 13). The average Ra values wereplotted (see Fig. 14). The results obtained from the second molding of the same mold were not used inthis analysis as the mold surface deteriorated as mentioned earlier. From Fig. 14, it can be seen thatthe Ra value of the replicated plastic lens was higher than the mold. The surface roughness of themold cavities before molding showed good consistency. Ra values of below 8 nm are obtainable.Therefore, the surface roughness (Ra) of the mold cavities before molding was considered to beconstant. The Ra values of the replicated lens above 1768C shows an increasing trend that was similarto the Ra values of the mold cavities. But the result from the embossing done at 1688C shows acontradiction. It was thought that due to the low temperature, the viscosity of the polycarbonate

Fig. 11. Profile of replicated lens of mold 3; (demold at room temperature, 268C).


Fig. 12. SEM picture of replicated 735 microlens pattern.

material increases. This viscosity limits the smooth flow of the polycarbonate into the mold cavity asit seems unable to provide enough energy/work to form the curvature on the material surface [14].This could be the reason for the poor surface finish obtained at low embossing temperature. Asreported [8], a higher temperature is favourable because of the lower viscosity that facilitates themolding process. However, an embossing temperature that is too high is also undesirable, as a largechange in temperature during the cooling phase, will lead to poor surface finish of the replicated lens.In addition, a high temperature tends to ruin the surface profile of the microlens [13]. Therefore, toobtain a good surface finish, a balance between plastic viscosity and the change of temperature duringthe cooling phase had to be obtained. Judging from Fig. 14, the Ra value is lowest at around 1828C.

Table 1Summary of mold data

Mold no. 1 2 4 3 52Ion dose (nC/mm ) 10 5.3 3.5 3 3

per mill depth (mm)Extraction current 5.5 4 3 3.5 2.2(mA)

Diameter 70.5 67.8 65.4 65.2 64.6Height 6.3 5.7 5.5 4.6 3.45

Number of 1st trial 2nd trial 1st trial 2nd trial 1st trial 1st trial 2nd trial 1st trialembossing

Embossing 187 176 187 176 198 198 198 176temperature (8C)

Demolding 26 26 26 26 26 60 26 26temperature 8C


Fig. 13. Four points were measured to obtain the average surface roughness of the replicated lens.

3.2. Embossing with varying force ( force test)

Mold 5 (refer to Table 1 for mold dimensions) was used for this experiment with varying force. Thetemperature of the mold was set constant at 1828C and demolded at 268C. Fig. 15 shows the variationof the profiles of the lens with increasing force. In this work, zero applied force means the moving

Fig. 14. Comparisons of the surface finish of the mold (before and after molding) and the surface finish of replication plasticlens.


Fig. 15. Embossing force applied, (a) Zero force (b) 4.44 kN (c) 6.66 kN (d) 11.1 kN.

half of the hot press was closed, with the force sensor indicating a zero value. It was noted that withincreasing force, the profile of the lens was more defined. Embossing force above 6.66 kN yields goodresults. However, too high a force may cause damage to the silicon mold. Crack was observed in thesilicon mold when embossing was done at 11.1 kN. Fig. 16 shows the replicated lens array embossedusing a force of 11.1 kN and a temperature of 1828C.

Similar to the previous experiments, the surface roughness were measured at four locations on thereplicated plastic lens. The difference in diameter and height were calculated as shown by:

Mold dimension 2 replicated plastic lens dimension]]]]]]]]]]]]]]]]

mold dimension

Fig. 17 shows the difference in diameter and height versus embossing force. It can be seen that thedifference in diameter decreases as the embossing force increases. The lens profile embossed withzero applied force was believed to be under-filled as it was significantly smaller than the originalmold. The difference in diameter and height were 3.4 and 0.43%, respectively. It was thought thatsmall fillet of the replicated plastic lens was formed at the edge of the mold (see Fig. 18) when a smallembossing force was used. Thus, a smaller lens diameter was expected when a lower embossing forcewas used. However, when a higher embossing force was used, the polycarbonate was forced to formthe shape of the mold cavity. Thus, the difference in diameter reduces.

An interesting phenomenon was observed in that the height of the replicated lens was larger thanthe depth of the mold. Stretching of the total structure occurs during demolding due to two mainreasons [9]: adhesion at the mold surface and friction due to surface roughness. It was thereforeconcluded that the polycarbonate undergoes plastic deformation during demolding. Stretching of thepolymer should generally be avoided by ensuring good surface quality of the mold. This could be


Fig. 16. (a) Microscopic picture of replicated 735 lens array embossed using a force of 11.1 kN and temperature of 1828C;(b) zoom-in SEM picture of one of the replicated lens profile.

done by adding demolding agent to the polymer or use of anti-adhesive layer on the mold cavitysurface.

Fig. 19 shows that there is an increase in mold surface roughness as the number of moldingincreases for the same mold. The Ra value of the mold cavity increases sharply for the first fewmoldings. After the fifth molding, the Ra values tend to stabilise at around 55 nm.

Fig. 20 shows the average surface roughness of the replicated lens and as a comparison, the averagesurface roughness of the mold before molding was also plotted. It can be seen that at zero appliedforce, the surface finish is good (16.2 nm Ra) as some part of the plastic may not come in contact withthe mold cavity surfaces. Thus, is the inherent advantage of contactless embossing molding, as thereplicated plastic does not depend on the quality of the mold. But due to the low embossing forceused, the profile and dimension of the lens are unpredictable and furthermore, the profile was not well

Fig. 17. Molding force versus difference in replicated plastic lens dimensions.


Fig. 18. Small fillet of the replicated plastic lens formed with low embossing force.

defined. The surface finish of the lens deteriorates when the embossing force was at 2.22 kN. At a lowembossing force (2.22 kN), the polycarbonate material fills the cavity but did not stretch in the mold.As the embossing force increases, the polycarbonate was squeezed and stretched inside the mold.Therefore, this could be the reason for both the reduction of Ra values and better profiles (refer to Fig.15) for the replicated lens.

Despite the poor surface finish of the mold as it deteriorates with the number of moldings, thereplicated lens surface roughness remain constant at higher embossing force (greater than 4.4 kN).

Fig. 19. Average mold roughness versus number of molding.


Fig. 20. Average surface roughness of replicated lens vs. embossing force. Average surface roughness of mold beforemolding included for comparative purpose.

Therefore, using an embossing force of between 6.66 and 11.12 kN is favourable in this work. Thebest surface roughness of around 11 nm Ra was obtained.

4. Conclusion

The study was undertaken with the aim of using focused ion beam to mill microlens array on amold material, which can be later used for replication using a hot embossing process. Conventionalmethods, such as photoresist-based techniques, cannot be used to produce patterns on the substratematerial directly. Hot embossing was used primarily due to its low cost, short setup time andreplication with close proximity.

Materials such as pure nickel and stainless steel were milled using FIB, with very poor results.Silicon, which has several advantages [11], proved to be a suitable material for FIB. Embossingparameters investigated were temperature and embossing force. It was found that using highdemolding temperature is not favourable, due to the increase in extensibility of plastic, which leads todistortion and extension of the lens profile. The optimal temperature for embossing polycarbonate intoFIB machined silicon mold was found to be 1828C. It was thought that for a good replicated surfacefinish, a balance between the plastic viscosity and the change of temperature during the cooling phasehad to be obtained.

The diameter of the lens was found to be replicated with close proximity (0–3%) and the differencereduces when higher forces were used. It was thought that the formation of small fillet reduces thereplicated diameter. Heights of the replicated lens were found to be larger than the mold. Theadhesion at the surface and friction due to surface roughness of the mold could be a reason for thisphenomenon.

The embossing force test showed that the profile of replicated lens and the surface finish weregenerally better with a higher embossing force, despite the deterioration of the surface finish of themold. A surface finish of around 11 nm Ra was obtainable for the lens.


Acknowledgements

The authors would like to express their thanks to Cecila Chee from Inco Alloys Pte Ltd., and DanielLoh and Gordon Brinser from Wacker Siltronic Pte Ltd. for sponsoring the mold materials for thisresearch work.

References

[1] N.F. Borelli, Microoptics Technology Fabrication and Applications of Lens Arrays and Devices, Marcel Dekker Inc,NY, 1999.

[2] J. Schulze, W. Ehrfeld, H. Muller, A. Picard, Compact self-aligning assemblies with refractive microlens arrays madeby contactless embossing, in, Proc. of SPIB 3289 (1998) 22–32.

[3] M. Hutley, R. Stevens, D. Daly, Microlens arrays, Proc. SPIIE 1573 (1991) 110–121.[4] D.L. MacFarlane, V. Narayan, J.A. Tatum, W.R. Cox, T. Chen, D.J. Hayes, Microjet fabrication of microlens arrays,

IIEEB Photonics Technol. Lett. 6 (9) (1994) 1112–1114.[5] Y.Q. Fu, K.A. Ngoi, N.S. Ong, N.P. Hung, Influence analysis of dwell time on focused ion beam micromachining in

silicon, Sensors Actuators, A: Physical 79 (3) (2000) 230–234.[6] Y.Q. Fu, K.A. Ngoi, Microfabrication of microlens array by focused ion beam technology, Microelectron. Eng. 54

(2000) 211–221.[7] Y.Q. Fu, K.A. Ngoi, N.S. Ong, Microfabrication of diffractive optical element with continuous relief by focused ion

beam, Microelectron. Eng. 54 (2000) 287–293.[8] L.W. Lin, J.C. Chun, B. Walter, M. Heckele, Microfabrication using silicon mold inserts and hot embossing, Micro

Machine and Human Science, in: Proc. of the seventh International Symposium (1996) 67–71.[9] L.J. Heyderman, H. Schift, C. David, J. Gobrecht, T. Schweizer, Flow behaviour of thin polymer films used for hot

embossing lithography, Microelectron. Eng. 54 (2000) 229–245.[10] R.W. Jaszewski, H. Schift, J. Gobrecht, P. Smith, Hot embossing in polymers as a direct way to pattern resist,

Microelectron. Eng. 41 (1998) 575–578.[11] H. Becker, U. Heim, Silicon as tool material for polymer hot embossing, Twelfth IEEE International Conference on

MEMS ’99 (1999) 228–231.[12] L.W. Pan, L.W. Lin, N. Jun, Cylindrical plastic lens array fabricated by micro intrusion process, Twelfth IEEE

International Conference on MEMS ’99 (1999) 217–221.[13] H. Becker, U. Heim, Hot embossing as a method for fabrication of polymer high aspect ratio structures, Sensors

Actuators 83 (2000) 130–135.[14] H.H. Yang, M.C. Chou, A. Yang, C.K. Mu, R.F. Shyu, Realization of fabrication microlens array in mass production,

Proc. SPIE 3739 (1999) 178–185.

Documents

Microlens array produced using hot embossing process