DATA ANALYSIS: SURFACE COATINGS
Beauty is skin deep
Felix Grant assesses the value of data analysis software in the development and use of surface coatings
‘These lenses have a surface coating,’ said my partner, waving her expensive light-reactive spectacles at me, ‘and it’s a nightmare because it comes off. Stick that in your article!’
So I have (stuck it in my article, that is), because it served to bring together and focus several different skeins of thought. Surface coating is one of the oldest human technologies, but still at the cutting edge and a prime consumer of scientific computing. Biomimetic or nanoscale, from reduction of optical reflections in lens systems to protection of components in orbit, taking in paper manufacture and dental treatment in between it is crucial to almost every area of endeavour. How might we make an optical surface coating less likely to come off? The obvious answer is improved adhesion, but to strengthen it or make it self-healing would be alternative approaches. After a detour through various mesoscale approaches it seems fairly certain that the answer, whatever it is, will lie in nanotechnology.
The archetypal adhesive surface coating is a protective layer of paint, pitch or varnish, first recorded in Genesis 6:14. Noah’s account lacks computational detail, but requirements for precision have certainly tightened since Cennini  specified, some time around the end of the 14th century CE: ‘take one pound, or two, or three, or four, of linseed oil, and put it into a new casserole...’. Even the most mundane oil or alkyd paint descendent of Cennini’s casserole is the product of a highly computerised and data-intensive process; a series of recent patents [2-4] by Friel et al specify a multiple precursor site manufacturing system completely dependent on a computing extranet. Not only does modern manufacture have to provide improving and usage qualities of increasing consistency, it must consider issues such as byproducts, post-application maintenance, and disposal.
One large corporation with extended surface coating interests includes a considerable (and surprisingly eclectic) data analytic software library within its paints manufacturing inventory. Mathematica is used by the future products research division. JMP, Maple, Matlab S-Plus and SigmaPlot all feature in science roles under product development. Production is monitored by Statistica and improvements informed by Statistica Data Miner.
Paints themselves have moved a long way beyond simple protection and visual appearance enhancement. Solar-power generating paints, though not yet as with us as some would like to suggest, are on the horizon. Steel producer Corus, for instance, has invested in a collaborative project with Australian dye-sensitised solar-cell systems specialist Dyesol at its Shotton coating plant. A variety of patents (for example, Breen et al , 2009) relate to means of applying such cells as surface coatings in paint or ink forms.
Piezoelectric paints and inks have a ready scientific application in the direct real-time generation of deformation data from the surfaces to which they are applied; I have just been watching a study of a deep submersion research hull using this method. Piezoelectric paint applied as a matrix of grid squares to the inner surface and linked to data logger terminals provides a continuous and highly detailed flow of information on performance and stress as pressure increases. Output is monitored through graphic digital dashboards and stored for more detailed future analysis.
Self-healing surface coatings are not a new idea, but a rapidly developing one. Physical systems involve biomimetic delivery of epoxy healing agents through three-dimensional capillary networks (the design of which, to get a balance between effectiveness and structural weakening, is a computational analytic issue in itself – see, for example, the 2007 Nature letter  from Toohey et al) printed using fugitive inks.
An alternative to physical self-healing involves nanoscale ingredients, such as the work led by Dr Stefan Bon at Warwick University . In this example, silica-based particles of around 25nm in diameter are used in a short, low-cost, high-efficiency production process that allows film coatings with a number of useful properties including self-healing of surface scratches, designed-in degrees of semipermeability to gasses and fluids, and improved interfacing to other materials. Between the two comes use of mechanophores – embedded nanoparticles that change under stress to act as analogues of blood platelets in healing microscopic cracks. All such programs involve intensive in-silico model design, generation, exploration and analysis before any physical manifestation is considered.
My shadier acquaintances talk of ‘actively intelligent’ paints containing new generation ecosystems of nanotech devices. Some harvest ambient energy, others utilise that energy to gather information or to transmit the gathered information onward. The devices don’t work individually; numbers of them across a surface collectively form redundant holographic systems. The best developed examples, I am told, are simple transducers; a painted wall can be a very effective radio microphone feeding low-power (and therefore difficult to detect) signals to a storage device hidden in a safer nearby location. More complex capability matrices are confidently expected to follow fairly quickly, with ranges of intermixed biosensors high on the list. There is also talk of simple preliminary data analysis being done before transmission, by distributed devices within the film. Applying such films covertly to large surfaces might be difficult, but various industries already have the need for closely matching patches to existing surface coats. Accelrys, whose data-analysis products have a track record in the coatings field, offer the example of automotive refinishing where a hand-held scanner gives the visible spectrum of a particular car. Data-analytic software (in Accelrys’ case, Pipeline Pilot Enterprise Server or PPE) bypasses the multiple trial and error reformulations involved in consensus visual matching to yield an optimum mix for the individual job in hand.
Beyond their genesis in politicomilitary paranoias, ideas of this sort would have a wide range of applications in scientific data capture and processing. These days, biology is no longer limited to living things, and can permeate a coating as readily as be sensed by it. In particular, DNA offers a whole new world of data encoding possibilities. Coatings containing nanoscale information bearing DNA constructs (see box, ‘DNA origami’) offer a vision of informational carpets knitted by self-assembly and applied to surfaces. It’s eerily reminiscent of Quatermass and the Pit , four decades ago, with its small Martian spacecraft whose artificial intelligence was built into their structural materials to allow maximum carrying capacity. Perhaps the day will come when my DSV hull (or any other research vehicle) can be given a layer of computing hardware primer, a lick of Linux undercoat, and a quick slap of R-analytic paint?
Analytic stages in the design of a four-layer thin-film lens coating using TFcalc.
Surface properties, structural behaviour, potential for degradation, qualities of nanoparticulate fillers and additives, all occur at or across different levels from which data must be collected and analysed for total combined interactive effect. In discussion of its products’ (Materials Studio and PPE) analytic role, Accelrys points out that modern coatings require integrated consideration and understanding of factors across all of these different measurement scales – from the performance of a complete DSV hull, in my example, down through formulation of the coating itself and its components, then beyond into the ‘submolecular quantum mechanical scale’ at which bonding and other interactions occur between coating and substrate.
Coming back to optics, where I opened, the problems are different. Paint films tend to have thicknesses around a tenth of a millimetre and are not normally required to transmit light with great fidelity. Optical lens coatings, designed to minimise reflections, are far thinner – typically between 100 and 175nm, for visible spectrum applications. At that thickness (spider silk extrusion ducts, by way of comparison, are three orders of magnitude coarser), biomimetic microvascular systems are impractical and nanotechnology would be the only option. Even then, the size of the nanoparticle itself would, at around 20 per cent of the coating thickness, be a significant factor in optical performance of the film.
Designers at a large Japanese lens manufacturing company (serving medical, scientific and photographic markets, as well as spectacles) tell me, nevertheless, that investigation of such possibilities is actively under way. Improved adhesion, like initial scratch resistance and a degree of surface self-healing, are fruits already brought to market. More dramatic self-healing is further away. In addition to generic products and in-house specialist examples, optics has its own market for design and analysis software, including the likes of TFcalc dedicated to thin-film coating calculations, but inputs need to be established first. The relative importance of antireflective efficiency, self-healing convenience, and reduced optical quality, further complicated by varying prescription curvature, is a very subjective judgement in spectacle design. In scientific and medical work, things are somewhat more objectively predictable; in photographic work for visual reference, even less so. On top of this, the actual layer structure of a coating can be complex in itself with (depending on application) anything from one or two up to several thousand layers. For any possible improvement, therefore, a great deal of psychoperceptual data analysis needs to be done. Deriving a possible improvement in the coating, generating physical prototypes across a range of realistic criteria, then gathering response data for analysis, then to start again on the basis of findings, is a costly and time-consuming loop.
Not all optical applications are so demanding. In particular, coating thickness and clarity become less significant as component size grows. For example, experimental prototype components are being designed in both Europe and the US for large astronomical reflector telescope dishes that would drift above the ecliptic and gaze into deep space without human maintenance. Among the strategies for prolonging their lonely operational lives are massively redundant distributed systems with both nanoparticle and fluid flow self-healing protective coatings similar to those for UV protection of orbital components  – since each individual component will be an element in a larger array, optical quality at its surface will be relatively unimportant.
Coming back from deep space to the human eye, contact lenses make a very different set of demands upon their surface coatings than spectacle lenses. The usual optical considerations are largely irrelevant when the lens is awash with tears, while close contact with the surface of the eye introduces new ones.
A few years ago , I mentioned my colleague Hafeez Jeraj’s investigation of factors affecting adhesion in a naturally occurring and undesirable surface coating: colonies of bacteria such as Pseudomonas aeruginosa. The problem hasn’t gone away, but progress is being made in seeking ways to combat it, including application of an antimicrobial counter coating to the lens. Such a coating has as its priorities a broad activity against as many microbial colonisations as possible (P. aeruginosa is not the only culprit) without triggering rejection or other problematic responses. It must, of course, also be transparent and retain its activity when attached to the lens surface. To an even greater extent than with conventional optical coating, finding suitable candidates for use in close contact with the cornea involves exhaustive data analysis – and a requirement for extracting maximum information from a minimum number of trials.
A promising line of development is the use of cationic peptide layers. Having developed a suitable material, means of attachment need to be developed and evaluated. A study  supported by the Australian Research Council, for example, explored photochemical attachment for two peptide coatings involving modification of the lens surface using 4-azidobenzoic acid and 4-fluoro-3-nitrophenyl azide with, on analysis of the data, significantly different balances of effectiveness, comfort, and bioacceptance.
Other biofilms present related problems across a diverse range of engineering technologies, from medical to petroleum extraction, contact lenses are close enough to spectacles for an aesthetically satisfactory closing of the narrative loop. My correspondents in the Japanese optical industry tell me that their future thinking is turning away from coatings as discrete entities, and towards multipurpose films whose function changes with depth and interpenetrates with the substrate. Glass is not, in the long run, the only high-quality option for optical design and has in many cases already been replaced by various organic alternatives. Antireflective coatings may well become functions of, rather than adjuncts to, complex designed surfaces with no clearly defined boundary on lenses with their own self-healing properties. Properties will vary with function, reflections dealt with through nanotexturing and disappearance of clear cut refractive index variation boundaries. All of which means that the ‘nightmare’ of peeled coatings should disappear.
Accelrys - Pipeline Pilot, Materials Studio, http://accelrys.com/company/contact/
Corus - Steel; solar paint coatings, http://www.corusgroup.com/en/contact/general_enquiries/
Dyesol - Dye solar cells, http://www.dyesol.com/index.php?page=Contact
Ilford Photographic - Papers coatings, http://www.ilfordphoto.com/contact.asp
Insightful - S-Plus, email@example.com
Maplesoft - Maple, firstname.lastname@example.org
MathWorks - Matlab, email@example.com
SAS - JMP, http://www.jmp.com/forms/jmp_contact_nonus.shtml
Software Spectra Inc - TFcalc, firstname.lastname@example.org, http://www.sspectra.com
Statsoft - Statistica, Statistica Data Miner, email@example.com
Systat Software - SigmaPlot, SigmaStat, firstname.lastname@example.org
Wolfram Research - Mathematica, email@example.com
The references cited in this article can be accessed through the Scientific Computing World website. Please go to www.scientific-computing.com/features/referencesapr09.php