1. Introduction

The next generation of ground-based optical telescopes are currently under development. These telescopes will have primary mirrors of 30–50m in diameter and are termed Extremely Large Telescopes (ELTs). Most design studies for ELTs have identified the need for an integrated large adaptive mirror ranging in size from 2–4 metres and either flat, convex or concave in profile. Currently there is a move towards large, ultra-thin glass mirrors, however these are fragile, costly to produce and unlikely to be made to the sizes required for ELTs, needing a less desirable segmented arrangement of smaller mirrors to obtain the required diameter. Telescopes currently utilising or planning to implement a thin-glass adaptive secondary are the Mutiple Mirror Telescope (MMT) [1], Very Large Telescope (VLT) [2] and Large Binocular Telescope (LBT) [3], with glass-shell thicknesses of 1.6–2mm and diameters of 600–900mm. It is generally known that costly breakages have occurred on some projects utilising glass-shell mirrors.

An alternative solution could be to use carbon-fibre composite (CFC) substrates - these are very robust even at high length to thickness aspect ratios and are scalable to the maximum sizes proposed. Some of the benefits in using CFC material are its low density, high stiffness and good thermal stability - these properties and others can be optimised for the project in question by careful choice of the fibre/resin ply matrix and design of the laminate lay-up sequence. The main concerns over using CFC that are perceived within the community are material inhomogeneity - does the Young’s modulus (E), coefficient of thermal expansion (CTE) and other bulk properties change significantly over a large mirror - that could adversely affect performance. In addition there are several questions that need addressing over any lifetime problems associated with repeated stressing (actuation) of the material and general environment (water absorption etc.) exposure. The Japan ELT working group [4] in particular performed studies into the moisture absorption of CFCs and observed a shape change due to water uptake at a level higher than that for thermal deformations, highlighting the requirement for an environmental encapsulation.

The main aims of this project were: to build a demonstrator, thin carbon-fibre mirror of optical quality (visible to infra-red wavelength suitable) and low form errors; provide the mirror with a stress-free coating suitable for grinding, polishing and providing environmental resistance; mount this mirror on a (preferably lightweight) partially active support system to allow testing of influence functions and address lifetime issues; completion of detailed finite element models to investigate design and performance parameters of the mirror and any support systems and compare the predicted performance with results from laboratory testing. This paper discusses the production methods and key properties of the metal coated CFC mirror, the system tests and corresponding simulations will be presented in a follow-up paper.

2. CFC mirror blank production

The creation of CFC mirrors is via a replication process - a master mould of the inverse form is required. We used a mould from a previous project which set the size of our finished demonstrator mirror at 27cm diameter. This mould was formed from stainless steel and is convex in form to allow the production of a spherical, concave mirror of radius of curvature 2950mm. The form accuracy of this mould is better than 1µm and the surface is polished to an Ra (surface roughness) of approximately 4nm. In addition a back-press mould was manufactured; this is required to ensure that the back surface of the mirror is smooth and uniform for the subsequent electroplating process.

The core material for the mirrors comprises sheets of a unidirectional (UD) carbon-fibre mat, pre-pregnated with a toughened cyanate-ester resin. The type used here is M55J/LTM123 (fibre/resin); this material is space grade approved for low outgassing, the toughened cyanateester matrix is highly resistant to microcracking and has very low (0.22%) moisture absorption properties [5] compared to other polymer matrices. The Toray M55J fibre is a high tensile modulus fibre (E=540GPa [6]) and as discussed by Cheng[7], may be more prone to inconsistencies than lower modulus fibre types. The fibre/matrix selection was partially based on the ready availability of the material at the time - there are other combinations which may prove to be more advantageous with further investigation.

 

Fig. 1. Cross section through the thickness of a completed mirror. Note that the nickel layer totally encapsulates the CFC core.

Download Full Size | PDF

The UD ply sheets are placed over the mould in a specific lay-up pattern, given by

the S indicates a full symmetric lay-up sequence about the centre plane of the composite, the numbers denoted are angles in degrees of the fibre direction with reference to the first ply fibre direction. For the 27cm diameter demonstrator mirror, 24 plies were used to give a composite thickness of 2.98mm, the stack sequence is illustrated in Fig. 1. Once the plies have been placed over the mould a breather membrane is then placed between the last ply and the back-press mould to prevent air bubbles forming during the curing process. This whole assembly is then placed in an autoclave at high temperture to achieve a full cure of the cyanate ester resin (and so ensure its properties of minimum water absorption and no shape change after removal from the mould). Shrinkage of the mirror does occur due to the curing process which results in a mirror of reduced radius of curvature than the mould - this amounts to an approximate 5% change. A chemical release agent is not required for this process. It is also found that the surface roughness increases tenfold from that of the mould, the Ra of the stainless steel mould being 4nm and that of the CFC mirror surface being around 35–40nm; this is likely due to fibre print-through on the surface resin layer.

2.1. Form errors

Although the mould is spherical it is generally found that the mirror deforms with an astigmatic form error on removal from the mould. This is most likely due to angular errors in placement of the plies during the manual lay-up procedure, resulting in an asymmetric stack sequence about the centre plane and hence an imbalance of forces either side post-cure.

Finite-element models have been created to verify the key factors involved in the process that leads to the undesired deformations. The models consist of a stack of 24 or 32 discs, 20cm in diameter and 100µm thick, an aspect ratio similar to that in reality, but consisting of a flat, rather than concave profile to allow the use of a specific type of mesh in the Comsol Multiphysics software. A diagram highlighting the model construct is shown in Fig. 2(a), an 8-ply version is illustrated for clarity. The premise for this model is that the deformations come about due to shrinkage of each ply, primarily in one axis, during the curing process. This was simulated in the model by creating an orthotropic material with negative CTE in 2 axes (to represent the resin shrinkage) and zero CTE in the axis corresponding to the fibre axis direction (since it is assumed that very little shrinkage occurs in this direction). The mechanical properties were set to be those of M55J/LTM123 UD fibre sheet (60% Vf), as given by Advanced Composites Group who manufacture the material. The linear shrinkage perpendicular to the fibre direction was estimated to be 0.1%, and so a suitable thermal model was constructed to represent this. The pertinent outcome of this model is that an asymmetry in the stack sequence due to an angular misalignment that is not mirrored across the symmetry plane does indeed cause an astigmatic form error (as shown in Fig. 2(b) and (c)) similar to those observed in the actual CFC mirror bases. Also, as might be expected, the further from the central symmetry plane that the error occurs the greater the amount of deformation observed as the moment of force increases. Other important trends that were noted are that increased isotropy gives improved resistance to deformations on ply misalignment but only if there is at least one sequence repeat. Also, ensuring that the plies are arranged in perpendicular pairs, rather than laying up in clockwise fashion, provides additional resistance to bending moments caused by ply misalignments. The results of these trends are shown in Fig. 3. For the perpendicular-paired lay-up the same trend is observed, but for the clockwise lay-up there is a different trend depending on whether the θ increment is an even fraction of 45° or a fraction of 60°.

 

Fig. 2. (a) Basic model construct to investigate effect of lay-up errors on mirror form. (b) 5° error for a 45° lay-up pattern, the P-V=2.25×10^-6m. (c) 5° error for a 90° lay-up pattern, the P-V=19.68×10^-6m.

Download Full Size | PDF

 

Fig. 3. Graphs showing maximum deformation of a flat mirror on curing when the ply immediately below the surface ply is misaligned by 5°. The number of pattern repeats is increasing as θ increases. Left: 32 ply lay-up, with θ increments as even fractions of 45°; right: 24 ply lay-up, θ increments mostly fractions of 60° and a 45° set for comparison.

Download Full Size | PDF

In reality the cause of the deformations is more complex; random errors in ply orientation can occur on multiple layers either side of the symmetry plane and the resin flow characteristics during curing can introduce additional strains. However, the case of one larger error, as indicated in these results, is representative of how the structure as a whole is able to resist deformations due to angular errors in ply placement. At best a 0.5° placement accuracy could be achieved by the manual layup procedure, although within 1° is more realistic.

We experimented with 45° increment lay-ups and 15° increment lay-ups. Although the 15° lay-up does not allow for any repeats when using only 24 plies, this still provided a much improved form error compared to the 45° one, reducing it from approximately 350µm to 30µm on the best piece. In choosing which lay-up is best, consideration must also be given to the improved characteristics in terms of mechanical properties (increased isotropy) that a smaller angular increment lay-up gives.

3. Coating and polishing

The Optical Science Laboratory at UCL has been involved in CFC mirror development since 2002. Initial prototypes by Kendrew et. al [8, 9] used honeycomb structures sandwiched between CFC facesheets with a layer of electroformed Ni bonded on top. The thin electroformed Ni plate was used as the first layer in the composite layup and since this whole structure was autoclaved to cure the resin, the differences in CTE of the different materials used (carbon fibre, epoxy, aluminium, nickel) resulted in the piece having very high internal stresses. These stresses are released during the grinding process causing the mirror form to undergo unpredictable change. The alternative mirror constructs described here are devised to avoid this problem.

The procedure described here uses a carbon-fibre/cyanate ester mirror blank and the nickel layer is applied post-production. The method used to apply this layer is a room temperature, electrochemical deposition performed at the Optical Science Laboratory. The all-over coating of nickel, the cross-sectional profile as illustrated in Fig. 1, is suitable to be ground and polished (i.e. has the required hardness needed to enable an optical quality finish) and yet has extremely low internal stresses to ensure the substrate does not deform during processing.

3.1. Electroplating method

A number of different nickel coating processes have been investigated. We have opted for a nickel sulphamate electrochemical deposition process and have set up our own small plant in-house (see Fig. 4(a)). The vast majority of industrial plants use high temperature baths, usually with chemical additions, which for our application create unwanted stresses in the nickel deposit. We use a room temperature (22°C), low current density procedure to deposit a zero-stress nickel layer which totally encapsulates the CFC mirror core.

The first stage in the coating process is surface preparation; this is a critical stage and if it is not performed properly results in fibre print-through, delamination of the coating and generally poor electroplating. Since the M55J/LTM123 system used here has a very high fibre content the substrate gains a low electrical conductivity once the insulating layer of surface resin has been removed. The surface layer of resin is removed using a course grinding compound front and back of the CFC mirror core and up to a third of the fibre thickness is also removed to provide an evenly smooth surface front and back with an average Ra of approximately 5µm. The mirror is then meticulously cleaned with a degreasing compound (we use SD-Klene S416) and thoroughly rinsed in distilled water. Other systems we have tested that have a higher percentage of resin are generally non-conductive even after the surface layer is removed. We have identified a treatment for surfaces such as these using a fine conductive, silver particle spray paint/adhesive. We have tested the procedure and subsequent nickel coating and initial tests do not show any difference to those pieces that do not have this intermediate coating, although more tests would be required and a suitable delivery system would need to be implemented to ensure a coating of equal thickness all over (we applied the spray manually using an airbrush system).

 

Fig. 4. (a) OSL’s nickel sulphamate plating bath for coating the CFC mirror samples. The tank is a 60cm cube with two titanium anode baskets containing the nickel, and the cathode (mirror sample) is suspended in-between them. (b) The electrode rig - the mirror is supported at 3 points, the bottom support is a spinning titanium conductive contact which supplies the plating current to the mirror, the mirror is rotated in the vertical plane via a motor driven gear system attached to a spinning silicone rubber contact.

Download Full Size | PDF

On completion of the surface grinding and cleaning the mirror is placed into the plating rig (Fig. 4(b)), the surface is wetted down with distilled water and then the entire assembly sub-merged in the tank. The plating rig provides a moving electrode contact by means of spinning the mirror and so allows the mirror to be completely coated. The required current is calculated based on a current density of 0.01Acm^-2 over the entire substrate surface area to be coated; the deposition rate at this current density is approximately 12.5µm of nickel per hour of plating time. We use a well regulated power supply which maintains the current at the correct level by automatically adjusting the voltage as the resistance of the circuit changes during initial plating. The encapsulation is achieved by using two anodes, allowing an even coating of the mirror substrate on all sides. The encapsulation also improves environmental resistance properties by sealing the CFC edges to moisture ingress. A constant mass transport of the electrolyte across the mirror surface is crucial to ensure an even thickness/rate of deposit of the nickel. Currently we have a reduced efficiency of mass transport towards the centre of the mirror which is resulting in an overplating at the edges of the mirror which need to be ground away afterwards. This issue is being addressed and we expect to arrive at a satisfactory solution over the next few months.

 

Fig. 5. The setup used to test the optic during grinding and polishing. The 633nm laser beam from the interferometer is brought down onto the optic via a 700mm diameter flat mirror to allow the flexible test optics to be supported in a horizontal, gravitationally symmetric position.

Download Full Size | PDF

 

Fig. 6. Left: original surface Ra (note fibres are visible) prior to plating surface preparation. Right: surface post plating and polishing (note, the finest grade polish was not used here); fibres are not visible and small, random polishing scratches are present.

Download Full Size | PDF

3.2. Grinding and polishing

Since the nickel sulphamate process produces a deposit which is matte-grey in colour, the surface must be ground and polished to provide a reflective, optical quality finish. If neccessary this can be further coated with an e.g. aluminium or gold vapour deposit for improved reflectivity in the required wavelength. The methods for polishing a flexible mirror must be modified from the traditional methods otherwise an incorrect form can be produced. Figure 5 shows the optical set-up used to measure the mirror surface during the polishing process. An f11 test lens is used in theWyko 6000 interferometer and the beam is brought down on to the mirror surface via a large (700mm) folding mirror flat. During measurement the flexible mirror is supported horizontally in a matched curvature moulded base with a compliant layer separating them.

If the nickel has been deposited at the correct rate then the coating can be ground and polished using increasing grades of silicon carbide grits. If the nickel has been deposited too quickly (and is hence very hard due to the decreased grain size) diamond would be required. Figure 6 shows the surface quality of the unprocessed CFC mirror core and the result after coating and polishing. The surface Ra can be improved further by using a pitch polish, but this was not done here due to time constraints, however it is clear there is no evidence of high frequency fibre print-through visible on the finished surface. The completed mirror is shown in Fig. 7 with an RMS form error of approximately half a wave (λ=633nm), this being mainly due to astigmatism.

4. Material properties

The first 27cm Ni-CFC mirror we have produced has a finished mass of 347g and a thickness of 3.1mm. This corresponds to an average density of 1949kgm^-3 which according to the theoretical values tabulated in Table 1 puts the Ni layer thickness somewhere between 50–75µm; experimentally we estimated a thick Ni layer of 80–100µm, this discrepancy needs to be investigated. The average density obtained is still lower than the density of glass (ρ=2200 kgm^-3) whilst being significantly stiffer having a Youngs modulus of 115GPa (theoretical) compared to 70Gpa. The final radius of curvature of the reflective surface is 2700mm which is within the testing range for our large optical table set-up.

It has been difficult to test the thermal properties of the mirror accurately in-house. Some basic tests have been performed to investigate the thermal properties of a smaller 12cm mirror sample constructed of the same materials as the current larger sample, although with a θ=[0/90/45/-45]_S layup and a non-equal nickel layer thickness either side of 20µm difference (the thicker bottom layer of 60µm was not ground compared to the top layer which was ground and polished). The 12cm mirror was placed in thermal contact with a temperature controllable hotplate and the change in radius of curvature (R) was measured using a contact stylus profilometer (Form Tallysurf); a thermocouple was attached to the upper surface of the mirror. The initial R of this sample was 2625mm and for a ΔT=10K, the ΔR=5mm, corresponding to a 0.2% change. In addition there is no evidence of fibre print-through appearing with temperature change. The finished 27cm prototype mirror undergoes the same amount of grinding top and bottom to ensure equal nickel layer thickness, this balance improves thermal performance.

The CTE is strongly dependent on the Ni layer thickness. Table 1 shows the calculated change in properties for a CFC mirror with the layup specified in Section 2 as the thickness of the nickel layer increases; it is possible to produce a mirror with a coefficient of thermal

 

Fig. 7. The 27cm diameter polished Ni-CFC mirror prior to mounting on an actuated base-plate.

Download Full Size | PDF

Table 1. Showing how the theoretical elastic and thermal properties of a composite mirror of base CFC thickness of 3mm (fibre type and layup as specified in §2) changes with the thickness of nickel deposited. The nickel is deposited on the front and back of the mirror.

View Table

expansion (in plane) of zero with a nickel layer thickness of approximately 33µm.

The sample has also been immersed in water at 90°C for 1 hour and for a month in room temperature water; the surface coating did not suffer any degradation that was apparent on visual inspection or on surface interferometry using an RST500 microscope interferometer.

Thermal and structural properties will be verified by a detailed materials analysis (destructive testing) once all non-destructive experiments are completed.

5. Ongoing and future work

The Ni-CFC mirror has been fixed to a lightweight backing structure comprised of aluminium honeycomb sandwiched with CFC sheets on which 19 stainless steel column flexures are mounted. Three of these columns are fully bi-polar piezo stack actuators. Comparisons of the influence functions of the system with computer modelled predictions will be made and then a lifetime actuation test performed to see if there have been any unwanted surface changes in the mirror before and after the test.

In addition further experiments on the electroplating procedure will be conducted, primarily into improved mass transport in the tank or suitable shielding to reduce the excessive edge plating currently observed. This should create a nickel layer with uniform properties (hardness, grain structure, thickness) and reduce post-processing time.

Devices and methods to control placement of the fibre mats during layup are also being investigated with our manufacturer B3 Technologies Ltd; with these in place consistency of form between repeat sample mirrors and a large reduction in the form replication error will be looked for. Future work will also be testing an alternative fibre/resin pre-preg for comparison.

6. Conclusion

The Ni-CFC mirror is looking to be a promising future technology, offering an alternative to thin, lightweight adaptive mirrors. CFC mirrors can be made thin and lightweight more easily than glass and can be made in large continous pieces without breaking, allowing the formation of mirrors up to the size of the autoclave (commercial airline wings are made this way for a size comparison). This toughness also means that CFC mirrors can be handled and cleaned safely and with minimal risk of damage. CFC mirrors can offer considerable cost savings both in basic material procurement but additionally in future handling and shipping costs due to their robust nature. A favourable estimate places the cost at about a tenth the cost of a similar glass shell mirror. A non-segmented deformable mirror would also simplify the set-up and running of the AO system and offers the potential for better image quality due to reduced scattered light - both from the lack of segment edges and the ability to clean the mirror to remove the build up of dust - and fewer diffraction effects.

The results from a mounted and partially actuated 27cm diameter demonstrator system using this new mirror substrate and finite element analysis system simulations will be presented in a follow-up paper.