Significant improvements in optical materials have been made in the past 15 years: infrared transmitting windows such as KC1 or ZnSe having absorption coefficients two to three orders of magnitude lower than that previously achievable, infrared transmitting fibers with losses as low as 10 dB/km, inexpensive wide-band antireflective surfaces with reflectance coefficients as low as 0.001, sapphire-like mechanical and thermal properties in a cubic and, hence, nonbirefringent crystal structure called ALON, large hot-pressed spinel structures with good transparency, the list goes on and on. Why then do we still insist on better optical materials? The reasons are as varied as is modern technology, and many of the applications are in quite unrelated fields. Let me outline here just a few of the reasons which I am particularly acquainted with. I speak from a Department of Defense background. Many of the drivers for technology development and a significant amount of the development money (although, unfortunately, usually not enough to cover all that needs to be done) comes from defense-related efforts. What are some of these drivers, and to what directions do they point for additional development?

The Heat Exchanger Method (HEM), a new crystal growth process, is in commercial production for 20 cm diameter sapphire crystals and 40 cm diameter silicon ingots for optical applications. The simplicity of the HEM combined with a very high degree of control of the submerged, solid-liquid interface allows growth of high-quality crystals. The HEM is also being adapted for the growth of Co:MgF₂, Ti:A1₂0₃ and Cr:A1₂0₃ crystals for laser applications.

Yttria is of interest for infrared applications because of its high wavelength cutoff (9 μm) which results in low emissivity, combined with high refractoriness (M.P. 246°C). Pore-free, single phase polycrystalline yttria is transparent since the structure is C-type cubic. An addition of 6 to 14 mole percent lanthana yields a composition which is two-phase (cubic plus hexagonal) during a high temperature sintering and transforms to a single cubic phase during a lower temperature anneal. This permits the attainment of low porosity levels by a unique transient second solid phase sintering mechanism. Sintering of discs and domes to near-net-shape has proven feasible. Room-temperature transmittance, absorption, and scattering measurements have been determined as a function of process parameters. Absorption coefficients have been lowered by a factor of five over the last year in an ONR sponsored program to 0.02 cm^-1. A significant absorption peak at 3246 cm^-1 was removed by controlled oxygen pressure annealing. An apparatus was developed to measure transmittance as a function of temperature and atmosphere. Increased absorption was observed above 1000°C in 0₂ and 1300°C in H₂. Curves of transmittance as a function of oxygen partial pressure were determined. These were interpreted in terms of stoichiometry and the role of point defects on optical transmittance. An interrelationship between high temperature optical and electrical properties is discussed.

Optical and mechanical properties of aluminum oxynitride (ALON) and magnesium aluminate spinel (MgA1₂0₄) are presented as well as some optical properties of spinel and ALON hemispherical domes. These materials are transparent in the visible and mid IR and are durable polycrystalline ceramics.

This paper will provide an overview of AMMRC's current optical material programs. Projects include the development of reinforced radome materials, high strength multi-mode transparent windows, tunable laser crystals, crystal growth of Nd:YAG by VSOM, hard optical coating materials, epitaxially grown thin film crystals, glasses for laser absorption, low-cost hermetically sealed optical fibers, and oxynitride glasses. Ongoing work in the laser crystals, multi-mode windows (AlON), and oxynitride glasses areas will be reviewed in some detail.

Very little is known regarding the optical, mechanical and thermal characteristics of calcium lanthanum sulfide (CaLa₂S₄). Property data including absorption coefficient, Knoop hardness, flexural strength, Young's modulus, Poisson's ratio, thermal expansion, thermal conductivity and calculated thermal shock resistance R' are presented for CaLa₂S₄ at its present state of development.

Binary and ternary phosphides have recently been identified in our laboratory as promising classes of materials which can be used as potential novel infrared (IR) transmitting ceramics. These materials might be employed in the 8-12 μm region and in environments involving severe thermal and mechnical stresses. In this presentation the current status of research and developing technologies involving these materials will be summarized. Initial target materials selected for study include ZnP₂, ZnSiP₂, ZnZrP₂, ZnGeP₂ and mixed systems of the latter three compounds. Preliminary synthesis and characterization efforts will be described. Characterization techniques used to date include: Fourier Transform Infrared (FTIR) spectroscopy, X-ray powder diffraction, elemental analyses, thermogravimetric analyses, and microhardness determination. Data from these measurements will be presented. Problems involved in sample preparation and impurity content will be reviewed. Potential problems in processing these materials and in their use as IR trans-mitting ceramics will also be discussed.

Calcium lanthanum sulfide (nominally CaLa₂S₄) has been under development for almost 4 years as an 8-12 μm transmitting ceramic. During that time, by concentrating the effort on optimizing the optical properties of this material, the absorption coefficient at 10.6 um has been decreased from 40 cm^-1 to 0.6 cm^-1. Most of the present absorption is thought to be extrinsic. Studies this year have dealt with a number of areas. One of these studies was an attempt to determine the cause of the remaining extrinsic absorption. During the course of this investigation, the presence of additional phases, including calcium sulfide (CaS), has been noted. The presence of CaS is somewhat surprising as all material investigated was 12-30 mole percent calcium deficient. Point defects and "trace" impurities have also been observed. These have been characterized using electron paramagnetic resonance and excitation/emission spectroscopies. Some of the defects and impurities have been identified and their location in either calcium lanthanum sulfide or second phase CaS determined. The possible effects of these defects and impurities on the observed infrared absorption of calcium lanthanum sulfide ceramics will be discussed.

Polycrystalline ZnS for infrared optical applications can be grown by chemical vapour deposition (CVD) over a wide range of process parameters such as pressure, temperature, Zn/H₂S ratio, growth rate and purity. In order to optimise the physical properties of this material to meet specific applications such as 8-12 μm window or multispectral visible and infrared window applications it is necessary to know how these growth parameters affect the optical and mechanical properties of the material, some of this data has been obtained by utilising a small scale low pressure research reactor in two configurations. In the first, a simple but effective vertical laboratory furnace was used to grow mainly 8-12 μm quality material for the determination of hardness and fracture toughness data, and in the second a more complex enclosed carbon reactor within a vertical laboratory furnace was used to grow mainly material for subsequent hot isostatic pressing. Four pLates of material of the order of 50 mm x 250 mm x 3 to 4 mm thick available from each growth experiment were characterised in terms of their optical, mechanical and structural properties. Thus material from a research reactor has provided data which goes much of the way towards enabling the best type of product to be defined for a particular requirement.

Sapphire, spinel, and ALON (aluminum oxynitride) have been identified as candidate dome materials for ultraviolet through 5 μm wavelength applications. They possess optical, mechanical, and thermal properties that are superior to those of currently used Irtran-1 domes. Optical performance of these materials in the visible wavelength region far exceeds that of Irtran-1, while infrared properties reported here vary from worse than to better than Irtran-1 domes. Reported in this paper are measurements of optical scatter and transmittance at 0.4762, 0.6471, and 3.39 μm, which represent a large range of values obtained on these materials in dome form. Processing changes over the last few years have produced improvements in both scatter and transmittance, provided that a good surface finish is maintained. Higher index of refraction will, of course, limit the ultimate transmittance for uncoated domes of these materials to slightly less than that of Irtran-1, which has also improved in the same time period. Calculations indicate maximum transmittance at 3.39 pm to be 0.95 to 0.96 for Irtran-1 and 0.87 to 0.88 for spinel, a difference of 0.08. Current measurements at the Naval Weapons Center confirm values of 0.88 for spinel, while the best Irtran-1 dome gave a value of less than 0.92.

Clay Engineering Inc. currently has a proposal before DARPA to manufacture large optical quality diamond for use as optical material. The manufactured diamond will be approximately 100 mm in diameter by 100 mm long. The cost of producing the diamond is expected to be three dollars per carat. It is expected that total impurities of a few parts per billion can readily be obtained. A study of diamond is a study of the effects of impurities. The elements boron and nitrogen can replace carbon atoms in the lattice structure, making diamond a "P" or "N" type semiconductor. Diamonds which are not semiconductors are classified as type IIa. The presence of B or N in the lattice causes diamond to photoconduct in ultraviolet light. All type I and III) and most type IIa diamonds photoconduct. The manufactured diamond will not photoconduct and will have an electrical resistivity greater than 1018 ohm*m. All non-lattice impurities are in the form of inclusions which dramatically affect the mechanical properties of diamond. High purity diamond has a coefficient of absorption of order 10^-3 cm^-1 at wavelengths of 8 to 12 micro metres, which makes it useful for infrared applications. It also has a low coefficient of absorption at wavelengths greater than 12 micro metres. For missile and aircraft applications, diamond is relatively immune to erosion or pitting damage by sand and rain. Diamond will readily withstand the stagnation temperature of Mach 3 flight and will go to Mach 4.8 with an anti-reflective coating to protect it from oxygen attack. Diamond is highly resistant to thermal shock, which makes it valuable for high energy laser applications. Using R = St (1-)) k/Ea as a measure of thermal shock resistance, diamond is 10⁷ w/m vs "sapphire" and Zerodur at 10⁴ and fused quartz at 1.45x10³. Diamond does not perform well in the 2.5-7.5 micro metres and less than 0.4 micro metres wavelengths. Intense beams of less than 0.4 micro metres energy can create color centers in diamond. For laser pulses of such short duration that thermal shock is not a problem, diamond will take less peak power than some competing materials, such as quartz. One could take advantage of the superior strength of diamond and use a thinner slice to obtain equal peak power capacity.

On a chemical basis, future directions in glass research for optics will emphasize low temperature processing, new materials, and composites. The key contributions from chemistry will come from organometallic synthesis, the application of theoretical chemistry, polymer reaction chemistry, chemical physics, chemical kinetics, polymer glass science and fluorine chemistry. These conclusions were derived from the indicators generated through the AFOSR fundamental glass research program and reflects the new approaches arising through the planning process.

The sol-gel process is a chemical approach to making optical materials at low temperature. Through hydrolysis and condensation reactions, a metal alkoxide such as tetraethyl-orthosilicate (TEOS) is converted largely to high surface area silica gel at room tempera-ture. After drying, the result is a rigid monolithic shape of bulk density about half that of conventional fused silica. The reduced weight of the shape is due to interconnected microporosity. The average pore size is generally smaller than 10 nm, and the material is transparent to visible light.

In the past, a great deal of theoretical and experimental efforts have been devoted to the development of glasses with low nonlinear refractive index mainly because of laser requirements. Recently, however, interests in rapid self-induced refractive index changes in materials have focussed the importance of high nonlinear refractive indices. This report is concerned with the theoretical and experimental development in the preparation of glasses with very high nonlinear refractive index. In glasses, the average observed value of nonlinear refractive index, n₂, is about 0.01x 10^-11 esu. Changes in n₂ are dependent on the third order electronic polarization which in turn is governed by (a) the linear electronic polarization (b) the absorption frequency and (c) the product of the density of polarizable atoms or molecules and oscillator strength. With these consideration, theoretical calculation of n₂ have been made for glass samples, and discussed with their IR and visible spectra.

Compositionally tailored glasses offer potential for ultra low thermal distortion (less than 10^-6/°C) optical components at a variety of wavelengths of current interest. Based on existing data, and calculations performed in this work, certain families of phosphate, fluorophosphate, silicate and heavy metal fluoride glasses (HMFG) appear to be promising candidates for ultralow distortion in the near-IR to near-UV range, while HMFG are also promising for mid-IR use. Although some general correlations between properties and composition can be established using existing data, the identification of the best candidates for low distortion will require more detailed studies of the dependence of dn/dT, CTE and stress-optic coefficient on composition.

Glasses are usually formed by solidification from the melt. In the last few years, sol-gel methods have received considerable scientific attention among ceramic and glass scientists as an alternative method of glass formation. The apparent advantages of sol-gel processing over the conventional melting method for glass preparation are: (1) lower temperature of preparation, (2) higher purity, (3) non-crystalline solids outside the range of normal glass formation, (4) increased homogeneity, (5) new crystalline phases from new non-crystalline solids, (6) better glass products from special properties of gels, (7) special optical glass. We suggest that other potential advantages relevant to manufacture of optical components include: 1) rapid production, 2) large shapes, 3) as-formed near net shape, 4) as-formed optically smooth surface, 5) unique indices of refraction, 6) density derived index gradients, 7) unique absorption bands, 8) ease of sealing and joining of components, 9) reproducibility, and 10) computer aided processing.

The glass-forming areas of two new systems based on the oxides of lead, gallium, and bismuth and cadmium, gallium, and bismuth have been mapped out. These glasses can be melted at 1000°C for 20 minutes and cast into homogeneous pieces at least as large as 30 mm X 30 mm X 15 mm. Their expansion coefficients (25°C-200°C) range from 83 X 10^-7/°C to 112 X 10^-7/°C, and they are relatively soft with Knoop hardness (100g) in the 225 region. The refractive index (ND) is in the vicinity of 2.4. The lower absorption edge is about 470 nm (yellow color) but they transmit out to 81m in 2 mm thickness. This is the best infrared transmission for any oxide glass which is sufficiently stable to devitrificationon cooling from the melt to enable sizeable pieces to be formed.

Most solid state laser materials exhibit pronounced concentration quenching; at Nd³⁺ ion concentrations above ≈2 x 10²⁰ cm^-3 non-radiative transfer of excited state energy between neodymium ions inhibits fluorescence and sharply reduces the quantum efficiency. Although other ions such as hydroxyl and many transition metals produce similar quenching effects, they are extrinsic contaminants and Nd self-quenching represents the material baseline. Reducing concentration quenching behavior is important for applications such as miniaturized lasers for electro-optic applications where high dopant concentrations and high induced emission cross-sections are desirable.

The recent move to shorter wavelengths in laser fusion research via harmonic conversion of the Nd: glass fundamental has resulted in the need for special filter glasses of sizes, optical homogeneity, and transmission properties not available in conventional filters. Typically, filter glasses are produced in diameters less than 25 cm and thicknesses less than 1 cm. In contrast, beam filters up to 90 cm diameter are required for Lawrence Livermore National Laboratory's NOVA laser, with homogeneity equivalent to that of the highest quality optical glass. Additionally, laser damage and thermal shock considerations require volume absorption, and thus greater thicknesses than are normally produced. Fabrication of such filters necessitates utilizing new glass types along with dopant combinations chosen specifically for the application. In the present paper we review the design and fabrication of several such filter glasses for large aperture frequency-converted laser systems.

The refractive index of an infrared transmitting glass, AMTIR-1, has been measured at infrared wavelengths and cryogenic temperatures. Experimental data are presented for wave-lengths of 7-12 micrometers (pm) and for temperatures of 170-350K. The change in index with temperature, dn/dT, was determined as well. An overall accuracy in index of five parts in the fourth decimal place was achieved. In addition to results, the equipment, measurement procedures, and error estimates are also described.

The fabrication of halide glass fibers is much more difficult when compared with oxide glass fibers. This is because of the high fluidity of most molten halides, their unusual viscosity-temperature characteristics and the poor chemical durability of the glass. Protective coatings are necessary and because of the high expansion coefficients of halide glasses and their relatively low softening temperatures the choice of an ideal coating is difficult. Some of the critical problems associated with fiber formation and the choice of proper coatings are discussed. The effects of the penetration of water through Teflon coatings and that of the presence of crystallites on tensile strengths of fluoride fibers are presented.

A quantitative assessment of bulk chalcogenide glass material properties was conducted for the As-S, As-Se, As-Se-Te, Ge-As-Se, and Ge-As-Te glass systems. The promising glass candidates were subsequently drawn into optical fibers by the crucible technique, and fiber attenuation losses were measured at wavelengths of He:Ne and CO₂ laser sources.

Long wavelength (X > 22 μm) lead-salt diode lasers are useful for spectroscopy studies and also for long distance fiber-optical communications. Double heterojunction diode lasers have now been fabricated using a new material system, Pb_1-x Eu_xSe_yTe_1-y. These structures were grown lattice-matched to (100) oriented PbTe substrates by molecular beam epitaxy. Laser operation up to 190K pulsed, 147K CW, has been attained with up to 1 mW single mode output power. The growth of single quantum well lead-chalcogenide diode lasers will be described. The threshold current of these quantum well lasers increases relatively slowly with temperature, yielding CW operation up to 174K (at 4.41 μm wavelength), and pulsed operation up to 260K (at 3.97 μm). To achieve single mode operation, a simple technique has been developed for the fabrication of lead-salt C³ (cleaved-coupled-cavity) diode lasers. The improvement in spectral purity and the reduction in threshold current of these coupled cavity lasers will be discussed.

An overview is presented of IR fibers, including prospective crystalline and vitreous materials; fiber fabrication methods such as drawing from preforms, crucible draw, pedestal growth and extrusion; and applications such as ultralong repeaterless links, IR imaging, and IR power transmission.

The optical absorption of tetrahedrical anions such as phosphates and sulfates has been studied in fluorozirconate glasses. A ternary fluoride glass in the system ZrF₄-BaF₂-LaF₃, was doped with calculated amounts of KHSO₄ or NaPO₃. A significant reduction of the OH-absorption has been achieved using a low temperature fluorination treatment and a high temperature refining under a very dry atmosphere. A minimum transmission loss of 58 dB/km at 2.9 μm was observed on a fiber 60 meters in length. Increasing heating time or temperature for the refining stage however results in higher scattering losses.

There is increasing interest in IR window materials with transmissionn limits of > 5 μm, increased mechanical toughness, and higher temperature capability. ZrO₂, which transmits to 7-8 μm is a possible candidate material for such needs. While its IR cut-off is at somewhat shorter wavelengths than other candidates (e.g. Y₂0₃ with a cut-off of - 9 μm), ZrO₂ offers the potential of significant toughening via a two phase, i.e. partially stabilized, structure. Such structures, and the relevant toughening mechanisms are reviewed, and the resultant mechanical properties discussed, particularly for the Zr0₂-Y20₃ system. For example, some materials in this system can give strengths of about 1.5 GPa (200,000 psi) at 22°C and nearly 0.7 GPa (100,000 psi) at 1500°C for laboratory test specimens. Preliminary optical measurements, including some outlining optical scattering effects of the second (precipitate) phase, required for mechanical toughening are presented. These suggest that useful transmission in the range of interest is achievable with these toughened ZrO₂ materials.

Recent improvements in preform processing and fiber drawing techniques have resulted in glass-cladded fluoride glass fibers having losses under 10 dB/km. Multicomponent zirconium fluoride glass was used, and care was taken to reduce impurities such as transition metals and water. The reduction of scattering centers was also a major concern. Preforms were made using the rotational casting approach, which resulted in glass-cladded preforms having no observable core-clad defects. The preforms were coated with teflon, and drawn into fibers using an r.f. induction furnace. The optical attenuation of the fibers was measured in the infrared region. The minimum loss occurred around 2.5 microns. The fiber scattering loss was also measured. A variety of lasers were used for this measurement, including an infrared color-center laser to obtain scattering data directly in the infrared region.

Ordinary glass-metal seals are reported here to give rise to currents as a result of light illumination. Responsivity observed here is of the order of 10 mA.W . Signal-polarity results suggest that the seal acts as a Schottky barrier. Photoionization of impurities, particularly iron oxide, in the glass energy gap is believed to be the detection mechanism. A method is suggested for increasing responsivity. Development of glass optical detectors may be advantageous for many applications involving optics and glass, including integrated optics, particularly in view of the low cost, the negligible dark current, and the guiding capabilities of glass fibers. The detection properties of glass-metal seals should be considered when evaluating experimental results, such as the optogalvanic effect, involving discharges as detectors of light.

Through the reinforcement of glass matrices with high modulus graphite fibers a family of highly stable composites has been developed. The fabrication and properties of these composites will be described. It will be shown that tensile strength and fracture toughness far above that of the parent glass matrix can be achieved. Composite thermal expansion behavior will be described in detail with CTE values of less than 1 X 10-''C^-1 achievable in specific instances. This combination of mechanical strength and to CTE recommends these materials for applications reauiring extreme environmental and dimensional stability which have traditionally utilized unreinforced glasses and graphite-epoxy composites.

A novel family of "germania-modified cordierite" ceramics has been developed for fabrication of precision optical components. Compositions in this family possess zero average coefficient of thermal expansion (CTE) over selected temperature regions. Moreover, the zero CTE can be "tailored" to a desired temperature range of operation by varying the relative proportions of Ge0₂ and Si0₂ in the modified cordierite structure. Thermal diffusivity of these ceramics is more than twice that of ULETM glass, a current state-of-the-art material for precision mirror substrates. In addition, germanium-cordierites possess high specific stiffness and low thermal moment/rigidity ratio. Controllable fine-grained structure enables good polishability and offers the potential for an optical quality surface. These thermal and mechanical properties make Ge-cordierite an excellent material for optical and structural applications. The theoretical basis for germania substitution for silica in the pure cordierite struc-ture has been examined. Material characteristics including solid solubility and sinterability have been considered. All the germanium-cordierite compositions presented in this study are morphologically single phase and microstructurally homogeneous materials. These properties along with "tailorable" zero CTE characteristics offer the potential of employing cordierites for high precision optics.

Heat pipes offer the potential of vibrationless cooling of optical surfaces while maintaining a high degree of temperature uniformity on the cooled surface. The objective of the present program is to develop and demonstrate prototype heat pipes for this application. The material of construction is silicon; the pqwer density range is 5 to 50 Watts/per square centimeter with a nominal objective of 30 W/cm². This paper describes the first eighteen months of work, during which the contract goals were met. The program was carried out by Thermacore on Contract F33615-82-C-5127 for the Department of the Air Force, Aeronautical Systems Division, Wright-Patterson Air Force Base, Ohio. Dr. Alan K. Hopkins of the Materials Laboratory supplied technical supervision of the program for the Air Force.

The low energy laser damage resistance of plastics, especially low energy multiple-shot damage threshold, has been a significant barrier to its applications for pulsed lasers. Applications of plastics include bulk material for optics (e.g., lenses and windows), hosts for Q-switches, and blast shields in laser fusion experiments. We review recent work evaluating the damage threshold of commercial and specially fabricated plastics. Our former work has concentrated on PMMA. We also review some of the other literature reporting damage resistance better than glass! The commercial grade PMMA demonstrated a damage threshold of 41 J/sq. cm. for a 1.06 micrometer wavelength pulse of 8 nsec duration. Improvement in PMMA has been small for single-shot damage but significant for multi-shot damage. For example, the 1,000 shot at 1 pulse per second rate threshold was reached at 75% of single shot threshold for our fabricated PMMA. We never reached a 1,000 shot threshold for commercial PMMA even though we illuminated at 6% of single-shot damage threshold. We also report our recent results of the benefits of distillation (both vacuum and atmospheric) and/or filtration of the starting monomers and uv and thermal polymerization methods of fabricating PMMA. Copolymer and plasticized polymer considerations will also be discussed. In addition, polymers other than PMMA are being investigated.

New applications of advanced semiconductor materials, as for example in two-color missile seekers or focal plane arrays generate a need for faster, more detailed material characterization at the wafer level. Conventional contact bonding methods are severely limited in this respect, especially in high production rate situations. A cost effective technique that seeks to satisfy this need is presented here together with results of a feasibility investigation using wafers of very high quality single crystal CdS, an advanced detector material for the STINGER-POST air defense missile. These results indicate that Faraday Rotation (FR) could be used initially as the first step in screening wafers, and could potentially become the preferred method of characterization.

Birefringence in optical thin films due to structure on a scale large compared to atoms but small compared to optical wavelengths, known as form birefringence (FB), was observed almost a century ago. More recently, studies of obliquely deposited metal films stimulated new interest in birefringent films. The link between structure, which is predominantly columnar in evaporated thin films, and birefringence has been conclusively demonstrated through ellipsometric measurement and modeling. Direct measurements of form birefringence are especially tedious in tilted films, since essentially four quantities must be derived: three indices of refraction and the film thickness. Clearly, four measurements are required; Horowitz' used an ellipsometric method to perform such measurements on a zirconium oxide (Zr0₂) film. Later, a 4.6-μm-thick film of Zr0₂ was obliquely deposited; spectrophotometric measurements revealed its utility as a half-wave plate. A parallel effort directed at understanding FB films through computer simulations has been undertaken by Sikkens and Liao. These simulations can be specialized to include defects, epitaxy, and anisotropic surface mobility. Applications of obliquely deposited FB films of familiar thin film materials can be anticipated if their structure and performance can be more thoroughly understood.

Computer simulation of thin film growth has been used extensively to gain insight into the origin and nature of the microstructure of vapor deposited thin films. Usually, however, no attempts are made to predict film properties other than column angle and film density from such simulations. The aim of our work is to derive quantitative data from computer simulations in order to be able to predict relevant properties of optical coatings. The deposition of 2,500 - 25,000 particles has been simulated on different computers by random deposition of two-dimensional hard disks, using a simple relaxation scheme. Statistical analysis of the results yields quantitative data for the density, column angle and column period. On the basis of these results, a simple model has been developed for the microstruc-ture of a three-dimensional film. The birefringence and the shape of water penetration fronts in evaporated optical coatings, predicted from this model, are confirmed by experiment.