Quality control of the CFRP mirror manufacturing process at NPF
N. Soto, C. Lobos, P. Mardones, A. Bayo, C. Rozas, S. Castillo, G. Hamilton, L. Pedrero, S. Zúñiga-Fernández, K. Maucó, H. Hakobyan, C. García, M. R.Schreiber, W. Brooks
QQuality control of the CFRP mirror manufacturing process atNPF
N. Soto a,b , C. Lobos a,b , P. Mardones b , A. Bayo a,b,d , C. Rozas b,c , S. Castillo a,b , G. Hamilton c,d ,L. Pedrero b,c , S. Z´u˜niga-Fern´andez a,b,e , K. Mauc´o a,b , H. Hakobyan b,c,d , C. Garc´ıa b,c,d , M. R.Schreiber b,c , and W. Brooks b,c,da Instituto de F´ısica y Astronom´ıa, Facultad de Ciencias, Universidad de Valpara´ıso b N´ucleo Milenio de Formaci´on Planetaria - NPF, Valpara´ıso, Chile c Universidad T´ecnica Federico Santa Mar´ıa, Av. Espa˜na 1680, Valpara´ıso, Chile d Centro Cient´ıfico Tecnol´ogico de Valpara´ıso - CCTVal, Valpara´ıso, Chile e European Southern Observatory, Santiago, Chile
ABSTRACT
The surface quality of replicated CFRP mirrors is ideally expected to be as good as the mandrel from which theyare manufactured. In practice, a number of factors produce surface imperfections in the final mirrors at differentscales. To understand where this errors come from, and develop improvements to the manufacturing processaccordingly, a wide range of metrology techniques and quality control methods must be adopted. Mechanical andoptical instruments are employed to characterise glass mandrels and CFRP replicas at different spatial frequencyranges. Modal analysis is used to identify large scale aberrations, complemented with a spectral analysis atmedium and small scales. It is seen that astigmatism is the dominant aberration in the CFRP replicas. Onthe medium and small scales, we have observed that fiber print-through and surface roughness can be improvedsignificantly by an extra resin layer over the replica’s surface, but still some residual irregularities are present.
Keywords:
Surface metrology, CFRP mirror, Mirror production, Optical testing
1. INTRODUCTION
Production of low-cost segmented primaries for four to eight meter class infrared telescopes is one of the maintechnology requirements of the future ground-based infrared interferometers such as the Planet Formation Imager(PFI). Developments in the manufacturing process of replicated carbon fiber reinforced polymers (CFRP)mirrors have been made by the collaboration between engineers, astronomers and experimental physicist atN´ucleo Milenio de Formaci´on Planetaria (NPF) and Centro Cient´ıfico y Tecnol´ogico de Valapra´ıso (CCTVal).A key part of the development of the production process of CFRP mirrors involves the design and implemen-tation of quality control procedures that allow to guide the manufacturing in a cost-and-time effective manner.These procedures have to be designed to quantify the main known quality issues of CFRP mirrors, such as fiberprint-through and large shape deformation, which are specific to CFRP-based mirrors. Also, the metrology ofthe convex mandrels from which the replicas are made must be considered. Although optical methods for themeasurement of convex optics are well known, their implementation is not cheap and poorly scalable if a slightlylarger optical element is not readily available. In this work, we present the current status of the quality control process of lightweight mirror productionat NPF. A summary of the manufacturing process is first introduced and the main surface errors of replicatedCFRP mirrors are then described. The used metrology procedures and some example results are finally presentedwith a discussion of future improvements.
Further author information: (Send correspondence to N. Soto)N. Soto: E-mail: [email protected] a r X i v : . [ a s t r o - ph . I M ] D ec . PROCESS The manufacturing process of replicated CFRP optical surfaces consists of five main steps: First a convex mold,known as mandrel, is polished until a spherical high quality surface is achieved. Then “prepreg” laminates ofCFRP are layered over the mandrel surface with specific orientations in order to achieve a quasi-isotropic layup,that aims to reduce curing-induced deformations. The layup is then cured in an oven or autoclave using vacuumbagging.
Figure 1. The CFRP Mirror production process. Starting from an optically polished convex spherical mandrel, CFRPlayers are stacked in a quasi-isotropic layup. Using vacuum bagging the replica is cured. The mirror is obtained aftersmoothing the surface and a thin layer of metal coating
As out-of-the-oven replicas present high fiber print-through, it is necessary to apply an extra layer of resinto achieve a smoother surface. The resin is sandwiched between the mandrel and the rough replica to be curedat room temperature to minimise distortions. A thin layer of release agent must be applied over the mandrel toprevent the resin from bonding with it.
Figure 2. Extra resin layer application. The first step shows the rough replica obtained after curing in the oven. Thena liquid resin is sandwiched between the mandrel and the replica. After curing and release from the mandrel, a smoothreplica is obtained
Each step introduces a potential error factor in the replication process that can degrade the optical perfor-mance of the final CFRP mirror. Controlling the quality of both mandrel and replicas at every step is thenessential to understand the critical factors that induce the predominant errors in the replication process. .1 TYPES AND SOURCES OF SURFACE ERRORS
During the replication process different factors influence the behaviour of the composite, therefore, the sources ofimperfections are varied in nature and “range of action”, adding to a final surface that has artifacts distributedthrough different scales depending on the underlying physical processes causing them. The main imperfectionsdetectable on a CFRP mirror are: shape distortion, fiber print-through, micro-bubbles and micro-cracking ofthe polymer matrix.
Figure 3. One of the first CFRP prototypes showing the typical surface errors: clear fiber print-through, shape distortionsand micro-bubbles.
To identify the causes and quantify the effects of the dominant imperfections it is necessary to measure andanalyse the surfaces involved in the process.The first possible source of imperfections is the mandrel that is being replicated, as it is expected that the finalsurface quality of the replicas is related directly to that of the mould, at best in a one to one fashion. A carefulmeasurement of the mandrel quality is then needed to assess the expectations of the final result. Nevertheless,optical metrology of convex surfaces is not trivial due to the lack of a real optical focus.Mechanical stress during the curing cycle due to layup misalignment and different thermal behaviour ofcarbon fibers and epoxy resin is the main source of shape errors in the replicas. Careful alignment of the CFRPlayers is necessary to achieve a quasi-isotropic layup that present minimal shape distortion. Even with perfectlyaligned layers, fabrication errors of the laminates may induce an-isotropic behaviour. The temperature ramps ofthe curing process also influence the amount of distortion for a given layup.Fibers print-through is a known quality killer of CFRP mirrors. The structure of the underlying laminatesis easily transferred to the surface due to the different thermal behaviour of the polymer matrix and the carbonfibers. Even when applying an extra resin layer to mitigate this effects, residual print-through may still bepresent in the corrected surface. Its predominance is linked with the thickness of the fibers that are being used,the chemical shrinkage of the epoxy and is also dependent on the laminate type: woven patterns clearly showmore print-through when compared to the unidirectional ones. Print-through dominates mid to high spatialfrequencies of the surface.The “degradation” level of the polymer, known as matrix state, can also deteriorate the surface quality ofthe replicas for high precision replication. Partially cured epoxy will not react as expected in the curing cycleand will produce a range of surface artifacts such as bubbles, voids and micro-crackings. This defects, as far asthey are not predominant, will deteriorate the surface in the middle and small scales, increasing the scatteringof the reflected light.Analysing the surface errors by their spatial frequency helps to pinpoint the sources and quantify the effectsof the different types of surface imperfections. Form errors present low spatial frequency deviations from theesired shape. Fiber print-through waviness and roughness can be seen as clear mid and high frequency surfaceerrors in the direction orthogonal to the fiber’s orientation.
3. METROLOGY PROCEDURES
In our lab, metrology procedures have been introduced at different stages of the manufacturing process to quantifythe impact of new techniques in the final quality of replicated mirrors. The main parameters to measure are thesurfaces shape and their roughness. We are interested in determining the degradation of surface quality fromthe mandrel to the replicas.The final optical quality of the mandrel is difficult to quantify due to its convex shape, so we rely on mechanicalinstruments for its metrology. Sphericity is controlled locally using a ring spherometer. At least 30 points aremeasured across the surface. Gauge precision of 1 µm ensures constant radius of curvature (RoC) within 30mm accuracy over all the surface for mandrels with RoC = 2400 mm when using a 40 mm ring, as given by theexpression: ∆ R = − r s ∆ s (1)Where r is the radius of the spherometer ring, s the measured sagitta and ∆ R is the uncertainty in the RoCmeasurement for a given sagitta uncertainty ∆ s . Figure 4. Left: Sphericity measurement of the mandrel. If sag measurements are constant within 1 µm over the wholesurface the mandrel is considered acceptable. Right: CFRP replica surface profiling. Profiles must be perpendicular tothe fibers orientation of the fisrt CFRP layer. Residual scratches and dents from the polishing process spotted by visual inspection are periodically controlledvia microscopy to evaluate surface degradation. Polarised light is also used to evaluate the material stress oftransparent mandrels.Efforts are being made to develop a custom optical characterisation method for large convex spheres. Inthe meantime, due to the homogeneity of the glass mandrel surface produced by traditional polishing, mechan-ical measurements seem sufficient to provide a characterisation of its quality. Nevertheless, in a productionenvironment, quicker methods will be necessary.Improving the surface smoothness is a key challenge in the development of CFRP mirrors, as surface irregu-larities have a negative impact in the performance of optical elements due to the scattering of the incident light.A good parameter to establish the surface quality is its roughness, as it determines the ratio of specular to totalreflected radiation by the relation: R s R t = e − (4 π σλ ) (2)here R s is the specular reflectance, R t is the total reflectance of a given material, σ is the surface roughnessand λ is the radiation wavelength. For example, to achieve a 95 % of specular reflection in the visible range( λ = 532 nm), it is necessary to achieve a surface roughness σ <
10 nm.It is important to recall that surface irregularities are distributed at different scales. As mentioned in the pre-vious chapter, CFRP replicas are dominated by the wavy pattern of the fibers rather than random homogeneoussurface roughness. Measurements of surface quality have to consider scales in the order of the material’s patternsdimensions, i.e. in the 0.01 - 10 mm range for CFRP to cover features arising from single fibers to packs ofa few thousands of them (which is the typical manufacturing tow size). We therefore divide our measurementsin short and mid spatial frequencies to distinguish between the random roughness of the polymer and the wavypattern of the fibers.Both mid and short spatial frequencies are measured using a mechanical stylus profilometer, the MitutoyoSJ-410, first during the mandrel fabrication and then for each replica before and after applying the extra resinlayer. Environmental vibrations make precise mechanical measurements difficult, so the elements must be placedover a vibration isolation support. We use an optical table with passive vibration dampeners. Twelve 0.8 mm“short” profiles are measured at random over the surface to measure its roughness in the 0.002-0.08 mm spatialperiod range. Four additional 40 mm “long” profiles are taken to characterise the waviness in the 0.08-8 mmrange. The mandrel is considered acceptable when its roughness is less than 10 nm RMS, as its quality determinesthe base level to assess the improvement margin for the CFRP replicas.Replicas are also characterised optically if their quality is sufficient. As the epoxy resin has ∼
4. RESULTS AND DISCUSSION
So far our tests have shown that, although the application of an extra resin layer improves considerably theCFRP surface, residual irregularities are still an issue in the production of replicated mirrors. Figure 6 showsthe comparison of the power spectral densities for high and mid spatial frequencies measurements of the mandreland a CFRP replica after the application of the extra resin layer.It can be seen that the high spatial frequency content of the roughness profiles is almost equivalent, showingthat the cured epoxy resin can achieve very smooth surfaces and is limited only by the mandrel quality. Mean-while, the CFRP replica shows a larger mid spatial frequency content up to 60 cycles/mm, which is equivalentto larger irregularities with periods down to 0.016 mm, i.e, the width of two fibers.Given the thickness of the additional epoxy layer ( ∼ µm ) it is possible to polish the solid resin to improveits surface quality, at the expense of adding steps to the manufacturing process. A new layer orientation layupis also being investigated to smooth mid spatial frequency components from fiber print-through (see Castilloet al. in this same SPIE volume), and a different room temperature curing resin has shown promising resultsincreasing the replication fidelity.Shape deformations are also difficult to eliminate given the sensitivity of the process to many variables thatinfluence the behaviour of the CFRP layups. Astigmatism is the principal optical aberration for 20cm CFRPmirrors prototypes, as shown in figure 7, caused by the mechanical stress in the curing process. A better controlof the temperature ramps and the layer’s alignment is then of critical importance to improve the final mirrorsurface quality. Nevertheless, due to the high flexibility of the thin CFRP replicas, it is possible to correctcuring-induced deformations using adjustable supporting cells. So far, 3D-printed prototype cells have beenproven useful to reduce large scale aberrations (see, Bayo et al. in this SPIE volume)Another aspect to consider is the degradation over time due to environmental causes. Systematic assessmentof mirror degradation in working conditions is critical to establish the potential use of CFRP mirrors in futureastronomical telescopes. We have transported a 20cm replica to Paranal observatory and left it exposed to the andrel lappingSag whithin 1 ?Mandrel polishingRq 10 nm?Sphericity checkSurface pro ling
Readyfor surfacereplication
CFRP cureSurface pro lingExtra resin cureSurface pro lingRq 10 nm?Metal coatingOpticalcharacterisation Analise prototype
Glass mandrel CFRP replica
Figure 5. Manufacturing process for the development of CFRP mirrors including metrology procedures. elements to evaluate its quality degradation over a long period of time. The exposed replica at Paranal was notsignificantly affected by environmental factors in a one month period, and it will be evaluated periodically toaccount for long term effects such as extreme humidity variations.
5. CONCLUSIONS
We have introduced a detailed description of our current metrology procedures for the development of replicatedCFRP mirrors. Both mandrel and replicas are measured systematically across the different stages of the manu-facturing process. We rely on mostly manual measurements for the development process using well establishedmetrology instruments. The mandrel shape is controlled with an spherometer to ensure a constant sag acrossits surface. Stylus profilometry is used to characterise the surface quality of both mandrel and mirrors. Finalmirrors are measured using traditional optical methods.So far, we have been able to improve the quality of the mirror prototypes considerably, based on a systematicmethod of research and process development. Still, further improvements are necessary to meet the requirementsfor telescope mirrors in the visible to mid infrared range. Also, automation of the different sub-processes,including the metrology, is crucial to reduce the time and cost of producing CFRP mirrors. Implementation ofautomated procedures, based in the experience gained so far, will be the next challenge in the road to establisha scalable production process.
200 400 600 800 1000Spatial frequency [cycle/mm]1009080706050 P o w e r S p e c t r a l D e n s i t y [ d B / c y c l e s / mm ] Roughness PSD
MandrelReplica 0 20 40 60 80 100Spatial frequency [cycle/mm]10080604020 P o w e r S p e c t r a l D e n s i t y [ d B / c y c l e s / mm ] Waviness PSD
MandrelReplica
Figure 6. Comparison of mandrel and replica spatial frequencies. High spatial frequency distribution is unaltered inthe replication process as shown in the left panel. Mid spatial frequency content of the CFRP replicas is still significantcompared with the mandrel as shown in the right panel x cm y c m CFRP mirror deformation m Figure 7. Coated mirror modal surface reconstruction from Shack-Hartmann wavefront sensor: tip, tilt and defocusremoved. Astigmatism is the predominant aberration
ACKNOWLEDGMENTS
All the authors acknowledge financial support from Iniciativa Cient´ıfica Milenio v´ıa N´ucleo Milenio de Formaci´onPlanetaria. N.S and G.H acknowledge financial support from the PIIC program at USM. A.B acknowledgessupport from FONDECYT grant 1190748, A.B. and N.S. acknowledges support from ESO Comit´e-Mixto andA.B. from QUIMAL funding agencies. M.S., S.C and C.L acknowledge support from the ALMA-CONICYT fund.S.Z-F acknowledges financial support from the European Southern Observatory via its studentship program andANID via PFCHA/Doctorado Nacional/2018-21181044. K.M. acknowledges financial support from CONICYT-FONDECYT project no. 3190859
REFERENCES [1] Monnier, J. D., Ireland, M. J., Kraus, S., Baron, F., Creech-Eakman, M., Dong, R., Isella, A., Merand, A.,Michael, E., Minardi, S., et al., “Architecture design study and technology road map for the planet formationmager (pfi),” in [
Optical and Infrared Interferometry and Imaging V ], , 99071O, International Societyfor Optics and Photonics (2016).[2] Burge, J. H., “Measurement of large convex aspheres,” in [ Optical telescopes of Today and Tomorrow ], ,362–373, International Society for Optics and Photonics (1997).[3] Romeo, R. C. and Martin, R. N., “Progress in 1m-class lightweight cfrp composite mirrors for the ultratelescope,” in [ Optomechanical Technologies for Astronomy ], , 62730S, International Society for Opticsand Photonics (2006).[4] Thompson, S. J., Bichon, S., and Grant, R. J., “Influence of ply misalignment on form error in the manufac-turing of cfrp mirrors,” Optical Materials Express (1), 79–91 (2014).[5] Massarello, J. J., Welsh, J. S., Hochhalter, J. D., Maji, A. K., and Fuierer, P. A., “Fiber print-throughmitigation technique for composite mirror replication,” Optical Engineering (12), 123401 (2006).[6] Bennett, H. and Porteus, J., “Relation between surface roughness and specular reflectance at normal inci-dence,” JOSA (2), 123–129 (1961).[7] Aikens, D. M., DeGroote, J. E., and Youngworth, R. N., “Specification and control of mid-spatial frequencywavefront errors in optical systems,” in [ Optical Fabrication and Testing ], OTuA1, Optical Society of America(2008).[8] Steeves, J. and Pellegrino, S., “Ultra-thin highly deformable composite mirrors,” in [ ], 1523 (2013).[9] Zaldivar, R. J., Kim, H. I., and Ferrelli, G. L., “Hygroscopic and thermal stability of high precision replicatedepoxy composite mirrors,”
Optical Engineering56