Int J Adv Manuf Technol (1999) 15:281–286 1999 Springer-Verlag London Limited
Characterisation of Epoxy Resins for Microstereolithographic Rapid Prototyping C. R. Chatwin, M. Farsari, S. Huang, M. I. Heywood, R. C. D. Young, P. M. Birch, F. Claret-Tournier and J. D. Richardson School of Engineering, University of Sussex, Falmer, Brighton, UK
Two epoxy resins are investigated using non-degenerate fourwave mixing. The materials assessed are optimised for use with a UV argon-ion laser. The holographic gratings were written at a wavelength of = 351.1 nm for an irradiance range 0.5–3.0 W/cm2 and read at = 632.8 nm in order to compare the reactivity, curing speed, shrinkage and resolution of the resins. These experiments were carried out to prove the suitability of the photopolymerisation systems for use with a new microstereolithographic technique. This new technique exploits a spatial light modulator to create 3D components using a completely planer, layer-by-layer, process of exposure which is described herein. With this procedure it is possible to build components with dimensions in the range of 50 mm to 50 m, and feature sizes as small as 5 m with a resolution of 1 m. Keywords: Diffractive optics; Four-wave mixing; Microstereolithography; Optoelectronics; Photochemistry
1.
Introduction
Component prototyping using stereo lithographic techniques were first demonstrated in 1986 with the first commercial example of a fast free form fabrication technology appearing in 1987; this system was developed by 3D Systems Inc. Since then the market has expanded to approximately ten suppliers using systems based on several different technologies to achieve component build. The dominant processes are: stereolithography, laminated object manufacture, fused deposition modelling, selective layer sintering and solid ground curing. With the exception of the latter process, all of these methods rely on scanning either a laser beam or a material deposition head to facilitate the curing of slices representing planer cross-sections of the part to be constructed. The completed part is formed Correspondence and offprint requests to: Professor C. R. Chatwin, School of Engineering, University of Sussex, Falmer, Brighton, BN1 9RQ, UK. E-mail:
[email protected]
by the repeated application of layers of the fresh material, which is cured layer-by-layer. The system described in this paper is a new rapid prototyping process capable of producing components with small to micro dimensions with applications in industrial, scientific and medical products. This new stereophotolithography technique uses a spatial light modulator to create 3D components using a completely planer (layer-by-layer) process of exposure [1, 2]. This is a major advance over alternative methods because layers are now concurrently cured over the entire surface as opposed to incrementally building the layer itself. It is now possible to build components with dimensions in the range of 50 mm to 50 m (micro to small form factors), with feature sizes as small as 5 m with a resolution of 1 m; this is two orders of magnitude smaller than the technologies currently available. In order for a photopolymer to be suitable for microstereolithography, several requirements must be fulfilled. The material should have good spreading characteristics, low viscosity and low shrinkage from liquid to solid [3, 4]. In addition, it must have high resolution to permit good build accuracy, as single parts must be built with small dimensions [5]. In this work we investigate two commercially available photopolymers, Ciba-Geigy Cibatool SL 5180 and DuPont Somos 6100. These are special epoxy resins, which polymerise with the help of a cationic photo-initiator. Their composition is proprietary, but they are optimised for use with an argon ion laser at 0 = 351.1 nm. We perform holographic measurements on them in order to compare their: reactivity, polymerisation reaction rate, and the extent of volume shrinkage during polymerisation at this wavelength [6].
2. Microstereolithography Set-Up The overall optoelectronic system employed for this purpose is shown in Fig. 1. This optical system operates in the UV using the 351.1 nm line of an argon ion laser. Such a source reliably produces a Gaussian beam and provides flexible operation at these wavelengths. The principal physical components
282
C. R. Chatwin et al.
Fig. 1. Overview of control system architecture.
of the system consist of: an ultraviolet laser light source; an optical shutter; beam conditioning optics; a spatial light modulator; a multi-element lithographic lens system; a high-resolution translation stage. The general mode of operation may be summarised in six steps: 1. Describe the part using a suitable computer-aided design environment supporting solid or surface modelling. 2. Orientate and create planar slices of the target part and initialise the photolithographic subsystems. 3. Retrieve a slice of the target part as a bitmap image and display it on the optoelectronic mask. 4. Open the shutter and initiate a timer countdown for the photo-polymer cure time. 5. Close the shutter and initiate resin re-layering process. 6. Loop to step 3 if further slices exist. The spatial light modulator is the critical interface between the electronic and physical embodiment of the part. Twodimensional computer aided design images from a 3D solid or surface modeller are orientated and sliced at uniform increments along the chosen plane. Each slice is then converted to a bitmap format and used to drive the spatial light modulator, the UV laser beam thus being modulated with the level slice image. The subsequent optics are responsible for re-imaging this to the plane of the component build. Formation of the component takes place in a resin bath. Specifically, the photopolymer resin and incident light are selected such that incident light produces cationic polymerisation, hence initiating the curing of selective regions of the resin. The period of exposure is controlled by a shutter placed in the beam path. Once curing
of a layer is complete, the component is “dipped” further into the bath such that a new layer of resin is formed over the surface that was last cured. The process then repeats over the same sequence of operations for the next layer until the component is completed. Further details of the general procedure associated with the rapid prototyping process may be found in standard texts such as [3]. Patents describe the specific details of the optical [2] and resin recoating systems [7]. The experimental optical system is divided into four sections: beam conditioning, spatial light modulator, beam imaging optics and reaction kinetics, see Fig. 2. 2.1 Beam Conditioning Optics
The light source for the miniature rapid prototyping experiment is an argon ion laser, which provides a beam with a nominally Gaussian irradiance distribution with linear polarisation. In order to obtain a uniform beam energy distribution within the subsequent optical components and maximise the contrast ratio of the SLM, the beam is reshaped into a rectangular uniform irradiance to match the dimensions of the SLM. The diffractive optical element used is a Digital Optics Corporation Gaussian to Square Beam Converter. This is an 8 level device designed for a wavelength of 351.1 nm. In practice, the argon ion laser gives a poor approximation to a Gaussian, which results in the diffractive optical element not giving a flat irradiance pattern. The method is, however, considerably more energy efficient than sampling the central section of a Gaussian beam with an aperture. For example, if the central section of a Gaussian beam is sampled such that the intensity variation is 5% across the aperture, only 5% of the beam power can used.
Microstereolithographic Rapid Prototyping
283
Fig. 2. Experimental schematic.
The low quality of the beam produced by the diffractive optic can be redressed by the incorporation of a rotating ground glass screen diffuser. This also destroys spatial coherence and reduces speckle in the image plane of the reducing lens. The disadvantage of this arrangement is that the beam is no longer collimated and it becomes depolarised. The polarisation is linearised before the beam enters the SLM by the addition of a linear polariser sheet after the ground glass screen. 2.2 Spatial Light Modulator
A polysilicon thin film twisted nematic spatial light modulator (SLM) is used as a dynamic photolithographic mask. The SVGA SLM was supplied by CRL and has 800 × 600 pixels, pixel size 26 m × 24 m, 50% fill factor and a contrast ratio of 200:1. A cube UV polariser converts the polarisation modulation to a complex amplitude modulation. The CRL SLM is not damaged at wavelengths longer than 351 nm. However, at wavelengths much shorter than 350 nm the SLM indium tin oxide electrodes, and the liquid crystal itself, would start to suffer damage. This rules out helium–cadmium lasers, which operate at 325 nm. Beam Imaging Optics
In order to reduce the size of the image displayed on the SLM to that required for the component build, a high-quality, multielement, fused silica lens system with an NA of 0.2 was designed and fabricated. It has been optimised to operate at both 351.1 nm and 364 nm, and for image reductions of 10 or 20 times.
3.
Performance of Resins
The materials were characterised using non-degenerate fourwave mixing. The method and the results are described in the following paragraphs.
3.1 Experimental Set-Up
The principles of non-degenerate four-wave mixing are described in the literature [8, 9]. Our experimental set-up is shown in Fig. 2. An Ar-ion laser with a beam diameter R = 1.57 mm operating on the 351.1 nm line and with polarisation parallel to the plane of incidence was used to polymerise the resin. A pinhole was used to filter any noise in the beam. The laser beam was split into two equal power beams Ir and Is using a quartz cube beam splitter. The two beams overlap on the sample, generating a light pattern I(x) = I0 (1 + mcos(kx))
(1)
where I0 = Is + Ir, m = 2(IrIs) (Ir + Is) is the modulation index and k is the spatial frequency of the grating. The spatial wavelength ⌳G is given by the relation 1/2
⌳G =
0 2n sin(0)
−1
(2)
where 20 is the angle between the two beams and n is the refractive index of the material. The beams are controlled by an electronic shutter with a closure response time of 6 ms. The grating is read using a 5 mW He-Ne laser operating at = 632.8 nm, with the plane of polarisation perpendicular to the plane of incidence. The resins are not reactive at this wavelength and as a result this beam does not modify the grating by causing further curing. The diffracted beam is detected using a silicon photodiode, connected to a LeCroy 9361 digital oscilloscope. If the response of the material is a linear function of the beam intensity, the light pattern produces a refractive index grating n(x) = n0 + . ⌬n(1 + cos(kx))
(3)
where ⌬n is the amplitude change in the refractive index. The grating diffraction efficiency is defined as the ratio of the intensity of the diffracted beam over the intensity of
284
C. R. Chatwin et al.
the incident beam. If ⌬n is small and there is no change in the absorption constant, ␣, of the material during polymerisation, then can be approximated by the Kogelnik equation [10], thus = exp(−ad/cos )sin2
冉
⌬n d cos
冊
(4)
where d = 100 m is the thickness of the sample and is the angle of incidence of the reading beam. and satisfy the Bragg condition 2⌳ sin =
(5)
The change in the refractive index is due to the change of the density of the material (shrinkage). There has been no proof of a direct relation between volume shrinkage and curl occurring in the cured material, mainly because it is almost impossible to find material formulations that differ only in curl shrinkage, while leaving other parameters unchanged [3]. It is, however, safe to assume that a high diffraction efficiency would indicate high curl during building. The induced refractive index change in UV curable photopolymers is, in general, nonlinear with respect to the illuminating irradiance. The refractive index grating that results from the sinusoidal fringe pattern generated by the interference of the two writing beams can be more closely represented as a square wave grating which leads to higher-order terms in the wave diffracted from the grating [9]. However, it has been shown [11] that for small grating amplitudes, the Kogelnik formula is a good approximation for the relation between the first-order diffraction efficiency 1 and the amplitude of the first harmonic grating ⌬n1.
3.2
Sample Preparation
The samples were prepared and all measurements taken in a controlled environment with a temperature of T = 19°C and a humidity of approximately 50%. The preparation of the samples involved placing a drop of resin between two glass plates separated by 0.1 mm thick Mylar spacers. All the measurements were taken within four hours of the manufacture of the samples. The DuPont Somos 6100 was quite viscous (see Table 1) and demonstrated scattering when illuminated by laser light. There were no imperfections visible to the naked eye but when examined under the microscope, very small bubbles were visible. DuPont recommend that the resin is used at a temperature of 26°C, so its performance would be improved at this temperature. It should be pointed out, however, that after these experiments were completed, DuPont released a new photopolymer, the DuPont Somos 7100, which is promoted as possessing a “complete absence of bubbles”. In comparison, the Ciba-Geigy Cibatool SL 5180 was a smoother, lower viscosity resin (Table 1), which exhibited hardly any scattering. For the measurements, only the samples that transmitted a uniform beam (hence bubble free) were used. After the experiment, these were re-examined for “bubble trapping”.
Table 1. Viscosity and refractive index of the investigated materials. Material
Viscosity (cps, 25°C)*
Ciba-Geigy Cibatool 265 SL 5180 DuPont Somos 6100 390
Refractive index 1.49 ± 0.05 1.52 ± 0.05
*Information provided by the manufacturers’ data sheets.
3.3 Results
In order to calculate the grating spacing ⌳ and Bragg angle from Eqs (2) and (5), respectively, the refractive indices of the resins must be known. They were measured using an Abbe´ refractometer and are reported in Table 1. Figure 3 shows the time history monitoring diagrams of the two photopolymers where the diffraction efficiency is plotted versus irradiation time. The different response of each material is presented as a normalised logarithmic plot of the evolution of the diffraction efficiency over a time of 45 s. The power of each of the writing beams is 10 mW (0.52 W/cm2) and the angle between the two writing beams is 20 = 12°. It is evident that the DuPont Somos 6100 has a completely different response to that of the other material. That is to say, the Ciba-Geigy Cibatool SL 5180 resins rise to a maximum diffraction efficiency, which then decreases. This type of response is similar to previous reports which have used fourwave mixing to assess UV-curable resins [6, 11]. The DuPont Somos 6100, however, demonstrates an initial peak in the diffraction efficiency, then a decrease followed by a slow increase to the final value. In order to further investigate this behaviour, exposure times shorter than the initial rise time of the diffraction efficiency were used. The result of this was that the shape of the slope did not change; however, the final diffraction efficiency was lower. If the writing beams were blocked after the initial peak, then the diffraction efficiency would evolve as in Fig. 3. This leads us to believe that the first peak is due to the creation of the cations in the material, while the second rise in the diffraction efficiency is due to the polymerisation which, in
Fig. 3. Time evolution of the diffraction efficiency. Chiba-Geigy Cibatool SL 5180; —, DuPont Somos 6100.
Microstereolithographic Rapid Prototyping
285
the case of DuPont Somos 6100, is much slower than the initial cation creation. As the polymerisation of the material occurs regardless of the continuation of UV irradiation, curing time should be regarded as the time to the initial peak in the diffraction efficiency. Plotting diffraction efficiency versus time provides information about the delay time until the onset of diffraction, the final value of the diffraction efficiency and the rate of increase of diffraction [3, 6]. These parameters are discussed in the following subsections. 3.3.1 Delay
The delay time t0 characterises an epoxy resin system in terms of its photosensitivity. The rise time to 10% of the maximum diffraction efficiency versus the beam irradiance for the two photopolymers under investigation is plotted in Fig. 4. It can be seen that, under the experimental conditions employed, the DuPont Somos 6100 photopolymer has a faster response, particularly at lower irradiances. 3.3.2 Diffraction Efficiency
The diffraction efficiency of the two photopolymers was measured for different beam powers and grating spacing, 30 s after opening the beam shutter, when the diffraction had reached a constant value. It was observed that the irradiance did not affect the diffraction efficiency for the range of beam powers, 0.5–3.0 W/cm2, investigated. The variation of the diffraction efficiency with grating spacing is shown in Fig. 5. As indicated previously, a high diffraction efficiency is indicative of a high curl distortion. The DuPont Somos 6100 exhibited lower diffraction efficiency, it is therefore expected to have a lower curl distortion during build. From Fig. 5 it is observed that the diffraction efficiency decreases when the spatial wavelength decreases. This is due to the reduction of the modulation index, caused by the limited resolution of the material. It is noted that there is a slightly larger drop in the case of the Ciba-Geigy Cibatool SL 5180 than in the case of DuPont Somos 6100. Overall, however, the resolution of the two materials investigated was exceptionally high. Specifically, it was possible to write (but not measure, owing to scattering), gratings with a fringe period of 0.28 m,
Fig. 4. Dependence of the diffraction delay on the writing beam irradiance. 䊐, Ciba-Geigy Cibatool SL 5180; 왕, DuPont Somos 6100.
Fig. 5. Dependence of diffraction efficiency on grating spacing in the resin. 䊐, Ciba-Geigy Cibatool SL 5180; 왕, DuPont Somos 6100.
which corresponds to a feature size of 0.14 m. Since most stereolithographic machines require an accuracy in the order of fractions of a millimetre, the resolution of the investigated resins is more than adequate for their intended purpose. In our microstereolithography project we require 1 m resolution, hence, as far as resolution is concerned, both photopolymers were found to be suitable. 3.3.3 Rate
The rate at which diffraction increases is an indication of the reactivity of the resin. This gives information about the reaction rate of the double bonds of the monomers and oligomers that constitute the resin. Unfortunately, since the composition of the materials is confidential, it is not possible to calculate how many reactions correspond to each photon of UV radiation. Hence, the following comments are limited to a comparison between the resins. Their diffraction rate as a function of the irradiance of the writing beams is shown in Fig. 6. It can be seen that, under the experimental conditions employed, the DuPont Somos 6100 photopolymer has a high diffraction build-up rate, which increases dramatically with irradiance. The Ciba-Geigy Cibatool SL 5180 has a much lower buildup rate.
Fig. 6. Dependence of the diffracted beam build-up rate on the writing beam irradiance. (This variable is a function of the reaction rate). 䊐, Ciba-Geigy Cibatool SL 5180; 왕, DuPont Somos 6100.
286
3.4
C. R. Chatwin et al.
Discolouration
A common problem with UV curable resins is the change of the absorption of the material whilst curing. This is particularly significant within the context of rapid prototyping, as a decrease in the absorption (bleaching) would allow light to pass into masked areas of previously exposed layers. In order to investigate this effect, a simple experiment was conducted; the transmission of the beam of a He-Ne laser through a resin sample was monitored as the sample was exposed to UV radiation. The samples used were prepared in a similar way to those used for the non-degenerate four-wave mixing. In the case of DuPont Somos 6100, the transmission of the He-Ne beam decreased, which indicates an increase in the absorption of the resin as it was cured. This decrease was no more than 1%, and is therefore not expected to significantly influence the build of rapid prototyping components. There was no measurable change in the transmission of the He-Ne beam in the case of the Ciba-Geigy Cibatool SL 5180 photopolymer.
4.
Conclusions
From the four-wave mixing studies it is clear that photopolymers designed for conventional stereolithography will permit the manufacture of 3D microcomponents with feature sizes as small as 5 m and resolutions down to 1 m. The operational envelope of the machine encompasses components that will fit into a cubic volume which ranges from 50 m × 50 m × 50 m to 50 mm × 50 mm × 50 mm. The gross component dimensions constrain feature resolution as the SLM has a constant spatial bandwidth product. Hence, as the size of the prototype component is increased, the resolution of the component is decreased. Thus, using a 1000 × 1000 pixel SLM to produce a 1 mm diameter component by imaging through a photolithographic imaging lens will give a component resolution of 1 m. Using the same SLM to produce a 50 mm diameter component will give a resolution of 0.05 mm. Even this lower resolution is twice as good as the best laser scanning system, which can only achieve an accuracy of 0.1 mm owing to the size, off-axis aberrations and the irradiance of the focused beam. IBM – T. J. Watson Research Center [12] have recently reported a 2048 × 2048 pixel SLM, this will permit the fabrication of a 50 mm component to better than 25 m accuracy, i.e. 1 thousandth of an inch, the benchmark for standard engineering tolerance. This will significantly extend the operational envelope of the SLM based process. Hence, performance is set to improve greatly, which is not the case for scanning laser beams. Apart from being able to operate in size and accuracy regimes impossible with other techniques, this new manufacturing technology delivers significant advantages over existing solutions, which are: 1. The process does not use a focused beam, it is therefore far more controllable than conventional processes and avoids secondary parasitic polymerisation.
2. It polymerises the component layer-by-layer, hence the time required to solidify a layer is independent of its complexity. 3. The resin forming the part is fully cured and does not require post-curing, which is a major source of component distortion with other processes. This also eliminates a complete manufacturing process, reducing component cycle-time and removing the need for post-curing capital equipment and its associated running cost. 4. Because the component is cured layer-by-layer the internal residual stresses within it will be reduced, resulting in less distortion and greater accuracy. Acknowledgements
The authors wish to thank CRL Smectic Technologies Ltd for supplying the SLM devices. We gratefully acknowledge the EPSRC, Design and Integrated Production Program, for supporting this research on grant number GR/ L31814 [1].
References 1. C. R. Chatwin, R. C. D. Young, M. I. Heywood, S. Huang and M. Farsari, “Manufacture of fully three dimensional microcomponents”, Technical Report, pp. 1–27, RAP/PR/SUS/ EPSRC971102, EPSRC grant GR/L31814, University of Sussex, November 1997. 2. R. C. D. Young, C. R. Chatwin, M. I. Heywood, “Use of dynamic masks for object manufacture”, Patent application no. GB9615840.7; Patent reference PAT/RYCC960725, July 1996. 3. P. F. Jacobs, “Rapid prototyping and manufacturing”, CASASME, 1992. 4. A. Schulthess, B. Steinmann and M. Hofmann, “New applications for Cibatool SL epoxy resins”, Proceedings of the 28th International Symposium on Automotive Technology and AutomationRapid Prototyping in the Automotive Industries, pp. 41–50, 1995. 5. A. Bertsch, J. Y. Je´ze´quel and J. C. Andre´, “Study of the spatial resolution of a new 3D microfabrication process: the microstereophotolithography using a dynamic mask-generator technique”, Journal of Photochemistry and Photobiology A:Chemistry, 107(1–3), pp. 275–281, 1997. 6. M. Hunziker and P. Bernhard, “Development of resin systems for stereolithography: holographic cure monitoring”, Proceedings of the 1st National Conference on Rapid Prototyping, Dayton, OH, pp. 79–85, 1990. 7. M. I. Heywood, R. C. D. Young and C. R. Chatwin, “Reapplication of materials for object fabrication”; Patent reference RP/PAT/MIH960725; Patent application no. GB9615840.7, July 1996. 8. C. Brau¨chle and D. M. Burland, “Holographic methods for the investigation of photochemical and photophysical properties of molecules”, Angewandte Chemie, International Edition in English, 22(8), pp. 582–598, 1983. 9. C. Carre´, D. J. Lougnot and J. P. Fouassier, “Holography as a tool for mechanistic and kinetic studies of photopolymerization reactions – a theoretical and experimental approach”, Macromolecules, 22(2), pp. 791–799, 1989. 10. H. Kogelnik, “Coupled wave theory for thick hologram gratings”, Bell Systems Technical Journal, 48(9), pp. 2909–2947, 1969. 11. J. Marotz, “Holographic storage in sensitized polymethyl methacrylate blocks”, Applied Physics B, 37(4), pp. 181–187, 1985. 12. Sunny Bains, News article in Laser Focus World, 33(12), pp. 19– 20, 1997.
