論文

Space observation is inevitable to know about the formation of the universe and galaxies, and for imaging far objects. Therefore, astronomers always strive to build better telescopes with high resolution and stability. Methods to fabricate low-weight large-size precision telescope mirrors are being continuously developed around the globe. Material selection and processing method are two important aspects in fabricating high performance mirrors. ZerodurTM, a nano-crystalline glass ceramic, has found wide ranging application in earth-bound and space-borne astronomical telescopes due to its near-zero thermal expansion coefficient. To finish this kind of material, Shape Adaptive Grinding (SAG) with contact type elastic abrasive tools is a viable process that achieves excellent surface roughness (Ra < 1 nm) and form error (P-V ~ 20 nm). However, the relatively high hardness of ZerodurTM causes material removal by mechanical means to be rather low. In this paper, a new SAG+ process is proposed where cerium oxide slurry is used together with diamond based SAG tools to leverage chemo-mechanical interaction and improve the material removal rate. Cerium oxide acts as a reducing agent that destabilizes silica crystals. This allows the Si-O bond to weaken and enhances material removal during polishing. Various grades of resin bonded diamond SAG tools were used in combination with cerium oxide slurry to process lapped ZerodurTM samples on an industrial robot. SAG+ could achieve at least 50% reduction in the polishing time while retaining similar material removal depth and surface finish. The capabilities of SAG+ were also demonstrated in pre-polishing of silicon X-ray mirror segments, whereby the surface roughness reduced by 47-62% for different grades of polishing tools. These results on robotic polishing of large mirror substrates employing proposed SAG+ technology seems a promising advancement towards sustainable manufacturing, where energy footprint can be significantly reduced by shortening the overall polishing cycle time.

Complex surfaces are increasingly used in optics, biomedical devices, and aerospace industries. Additive manufacturing and five-axis milling can produce complex shapes, but additional finishing processes are needed to meet the surface quality demands. This keynote paper reviews shape-adaptive finishing processes that enable uniform surface quality improvement without altering the shape. Key factors affecting shape adaptability, processing efficiency, and material applicability are analyzed and optimal process selection strategies for finishing complex surfaces are discussed. Future possibilities and research topics in this area, including process modeling/simulation, digital twin, surface/subsurface metrology, functionality evaluation, and applications of advanced robotics and artificial intelligence, are outlined.

In this paper, the material removal characteristics and mechanism of submerged pulsating air jet polishing (SPAJP) process are investigated by polishing experiment on K9 glass and computational fluid dynamics simulation. The experimental results show SPAJP can etch almost perfect Gaussian footprints with high stability. Under comparable condition, SPAJP achieves nearly 300 times higher material removal rate (3.0 × 10−3 mm3/min) than that of traditional fluid jet polishing (FJP) (9.9 × 10−6 mm3/min) at the cost of a moderate degradation in processed surface roughness from Sa 3.97 nm by FJP to Sa 20.49 nm by SPAJP. Jet evolution, abrasive behavior and material removal mechanism in SPAJP are clarified by numerical modeling and experimental jet observation. It is found that SPAJP relies on jet instability of the submerged airflow beam to entrain and accelerate abrasive particles in the slurry, achieving stable and efficient polishing removal on the workpiece surface. To explore the potential of the process for deterministic polishing, an array structure surface is then designed and polished by SPAJP. The processing results are consistent with numerical prediction results, indicating that the new process has good material removal controllability which makes it applicable to deterministic polishing. This study highlights the possibility of using SPAJP to produce ultra-precise surfaces in terms of form control or fabricate high precision optical components in a more efficient and convenient manner.

Nanoscale roughness with ultra-precise form control can be readily achieved using compliant finishing methods such as bonnet polishing. However, their weak point lies in the difficulty of removing mid-spatial-frequency (MSF) waviness in the typical range from 0.1 to 5.0 mm wavelength. To overcome this shortcoming, a bonnet tool filled with viscoelastic fluid is developed and a comprehensive model is established to disclose its distinct removal behavior in the MSF range. The model considers tool viscoelasticity, stress distribution and workpiece topography. Experiments show high consistency with theoretical predictions, and show that MSF waviness can be effectively reduced using the proposed method.

Wavelength dispersion is a universal phenomenon in refractive optics that degrades optical performances. Current mitigation methods are being greatly challenged by the demanding requirements of miniaturization, low aberrations, high throughput, and accurate manufacture. Here, we propose an unconventional monolithic honeycomb-like lens concept with topological degrees of design freedom, capable of near diffraction-limited achromatic performance. To automate the design process, a methodological framework is presented that directly evolves the honeycomb lens from a flat window, instead of a knowledge-based starting geometry. The honeycomb lens performance at the full field of view is remarkably better than either traditional aspheric lenses or commercial cascaded lenses. It also overcomes limitations of flat metalenses, which have extremely small aperture sizes and suffer from multi-order aberrations. This topological honeycomb lens concept may pave the way to inexpensive and compact achromatic optics for focusing and imaging applications in consumer cameras, wearable optics, virtual reality, etc.

We present a random unicursal tool path for sub-aperture polishing and compare polishing with the random and raster tool paths. This new complex tool path is useful for reducing mid-spatial frequencies in polished surfaces. © 2008 Optical Society of America.

Compliant machining processes, such as bonnet polishing, can be used on hard and brittle ceramic materials such as alumina and silicon carbide, to produce ultra-precise surfaces with sub-micron form accuracy and nanometric surface roughness. However, a comprehensive understanding of the removal mechanism in such process is lacking. In this paper, an analytical model is proposed that is based on the existence of “three zones” in compliant machining process, namely elastic recovery, plastic removal and brittle fracture. The inherent relationships of the three critical pressures with actual pressure, due to compression of the elastic bonnet tool and asperity effect, are established and analyzed in association with different material removal behaviors. Analysis indicates that pad asperity plays an important role in material removal and that lower material hardness combined with higher tensile strength contributes to enlarged plastic removal zone, and thus higher manufacturability. Removal footprints and polishing tests were then generated to verify accurate prediction of the material removal rate under different conditions and demonstrate effectiveness of the proposed model.

Producing X-ray imaging space telescopes is a very expensive endeavor, due in great part to the difficulty of fabricating thin mirrors for Wolter type-I optical assemblies. To meet this challenge, replication from optical molding dies (also called mandrels) has become the preferred method, as it is reliable and economical. Several replication methods exist: in the case of the ASTRO-H mission, DC magnetron sputtering was used to deposit Pt/C multilayer coating on glass molding dies. The multilayer coating was then bonded with epoxy to aluminum shells and then separated from the die. Another mirror replication method consists of slumping thin glass sheets over a full (or a section of) revolution molding die under high temperature. This method was demonstrated in the case of the NuSTAR mission. But the challenge of fabricating truly aspheric Wolter type molding dies, which are capable of highly accurate angular resolution (below 5 arcs), remains very expensive and time consuming. In this paper, three methods for producing X-ray optic molding dies are presented. Each method uses a different substrate material and process chain, as follows: electroless nickel plated aluminum (first diamond turned then correctively polished), fused silica (first precision ground then correctively polished), and CVD silicon carbide (which can be finished entirely with a newly developed Shape Adaptive Grinding process). The process chains employed for each method are explained in details, and their relative merits discussed. A way forward for the next generation of X-ray telescopes after ASTRO-H is then drawn out.

Vibratory finishing is widely used in mass-finishing of components. However, it is difficult to guarantee shape accuracy and predict surface roughness evolution. Here, a material removal model is proposed that is based on Preston's law and statistical overlapping of particle trajectories computed by Discrete Element Method (DEM). After validating DEM motion by particle image velocimetry and rope experiment, the occupancy ratio is studied as function of media type and mixing. Predictions of removal volume and surface roughness agree with experimental results within 12% on average, making this simulation method a very practical tool that reduces the need for extensive trial runs and barrel reloading.

In the context of Industrie 4.0, we have previously described the roles of robots in optical processing, and their complementarity with classical CNC machines, providing both processing and automation functions. After having demonstrated robotic moving of parts between a CNC polisher and metrology station, and auto-fringe-Acquisition, we have moved on to automate the wash-down operation. This is part of a wider strategy we describe in this paper, leading towards automating the decision-making operations required before and throughout an optical manufacturing cycle.

The recent upsurge in the demand for off-axis and complex "freeform" optical surfaces is driving the development of novel processes for their fabrication. This paper focuses on recent developments of the Precessions CNC polishing process for freeform surfaces, including off-axis as a special case. First, the surface-prescription and metrology-data, and their relation to the data-input for the polishing machines, are considered. The relevance of consistent coordinate frames is emphasised. An outline of how the process can 'polish' a ground freeform part (improve the texture), and then 'figure' the part (reduce the form errors) is given. Specific experimental case-studies are then presented, illustrating the versatility of the process on different materials and forms. Recent work is included in which the process-speed has been moderated in order to remove tens of nanometres of stock material, rather then the more usual hundreds of nanometres to tens of microns as in the standard Precessions process. The relevance of this to improving the ultimate surface-precision that should be achievable by this method is described. As a final illustration, the potential of the process to the rapid fabrication of the hundreds to thousands of 1-2 metre class mirror segments required for extremely large telescopes is considered.

The Precessions™ process, and the 200mm capacity 'All' machine which embodies it, are outlined in the accompanying paper 'New method to control form and texture on industrially-sized lenses'. In this paper, we summarise the development of the process to address larger optical components. The first machines constructed for this application are of 600mm capacity. A conceptual design of a 1-2m class machine of bridge-configuration has also been produced. Unlike the All machine, the work-piece on these larger machines lies in a horizontal plane i.e. with the rotation axis of the work-piece spindle vertical. This configuration is preferred because it allows the use of a hydrostatic support-system for the work-piece; important for large optics in general, and particularly so for light-weight optics. The machine can also provide a clear vertical path above the work-piece for optical testing, or for access by lifting gear. It is interesting to consider how the Precessions™ process can be scaled. As an example, consider first that the type of surface-form remains the same; it might, for example, be an f/2 parabola. Consider then that the work-piece diameter, the bonnet-diameter, the polishing spot-sizes used, and the tool-path on the surface, are all stretched by a factor of two. The area of the polishing spot and that of the work-piece have both increased by a factor of 4, but the number of convolutions of the spiral tool-path remains the same. Since the spot instantaneously addresses the same fraction of the overall work-piece surface-area, the terms all cancels in relation to cycle-time. However, for the same tool-rpm, the larger bonnet delivers a proportionally higher surface-speed, because the polishing action is further from the axis of rotation. Therefore, the volumetric removal rate (proportional to contact-area times surface-speed) tends to vary as the cube rather than the square of the scale-factor. As a result, the larger bonnet working over the larger work-piece with the same polishing pressure, will remove twice the depth of material in the same time. The above scaling could, in principle, reduce cycle times on larger parts. In practice, we have found that excessive removal-rates lead to impractically short dwell-times, especially near the centre of a spiral tool-path. Such scaling also demands increased traverse-speeds and accelerations from the machine, and this rapidly becomes a limitation. Moreover, as the process converges on final form, smaller depths of material need to be removed. In practice, we have found it necessary to moderate the process by diluting the slurry and by moving from a polyurethane tool-surface to a material that delivers a lower-removal rate, such as Multitex. We are also developing bonnets that are compatible with operation at a reduced polishing pressure, in order to give additional control for fine-removal. The current status of the process development is as follows. A form error of ∼80nm peak-to-valley has been achieved aspherising a 100mm diameter part. This work revealed some subtle but significant limitations in the optimisation code, mostly concerned with convergence and residual ripples. Recent work has focussed on this aspect, and the revised code is now ready for polishing trials. A 150mm diameter part is currently in-process, and work is about to commence on a 300mm fused silica mirror for a space application.