Window Materials
Durable 3?x2013;5 m transmitting infrared
window materials
Properties of 3?x2013;5 m infrared-transmitting window materials are reviewed, with an emphasis on durable materials for applications in environments involving moisture, impact by solid and liquid particles, high temperatures and rapid heating rates. Infrared, visible and ultraviolet transmission windows are compared for MgF2, aluminum oxynitride, sapphire, spinel, MgO, Y2O3, calcium aluminate, SiO2, CaF2, LiF, ZnS, ZnSe, GaAs, GaP, Si and Ge. Emission at elevated temperature, reflection and optical scatter are also discussed. A comparison of mechanical and thermal properties is given, as is a brief discussion of rain and particle erosion resistance.
Author Keywords: Infrared; Ultraviolet; Visible; MgF2
1. Introduction
Excellent optical quality single-crystal materials such as sodium chloride and potassium bromide have been available for many decades as infrared-transparent windows for laboratory use. However, poor mechanical properties and susceptibility to attack by water prohibit the use of these materials in all but the most benign environments. For this reason, there has been a continuing search for more robust materials that can be used in environments involving moisture, impact by solid and liquid particles, high temperatures and rapid heating rates. This review emphasizes durable materials available for the 3?x2013;5 m (midwave) atmospheric infrared transmission window. Some `2-color’ materials that have good transmission in both the midwave and 8?x2013;14 m (long wave) atmospheric windows will also be discussed.
Table 1 lists many infrared window materials and indicates whether common forms are available as single crystals, polycrystalline materials, or glasses. Most window materials are electrical insulators, but Ge, Si, GaAs and GaP have sufficiently small bandgaps (0.7, 1.1, 1.4 and 2.2 eV) that their optical properties are degraded at temperatures well below their chemical stability limits. Major sources of information on infrared window materials are listed in Refs. [1, 2, 3, 4, 5, 6, 7, 8].
Table 1. Three classes of infrared window materials
(1K)
2. Transmission window
Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5 compare the infrared transmission of window materials on wavenumber and wavelength scales. In Fig. 1, sapphire, ALON (aluminum oxynitride), spinel and yttria are representative of the best optical quality available for each material. Magnesium fluoride in Fig. 1 is hot pressed, polycrystalline Irtran-1®, a material formerly sold by Kodak. The spectrum of this material, which is representative of currently available hot pressed MgF2, exhibits a sharp impurity band at 3565 cm-1 and a weak, broad impurity band centered near 3500 cm-1. Both bands probably arise from OH impurities.
Magnesium oxide in Fig. 1 (Kodak Irtran-5) is a low quality hot-pressed material with a broad impurity band at 3000?x2013;3500 cm-1 and sharper impurity bands at 1550 and 1320 cm-1. These bands are not seen in single-crystal MgO [9]. The most important comparison in Fig. 1 is the long wave cutoff, which increases in the order ALON<sapphire<spinel<MgOY2O3<MgF2. The shorter the cutoff wavelength, the less useful the window is at elevated temperature. The loss of transmittance with decreasing wavelength in the spectra of MgF2 and Y2O3 is attributed to Rayleigh optical scattering by centers that are smaller than the wavelength of light.
Fig. 2 shows that calcium aluminate glass, quartz (crystalline SiO2) and fused silica (SiO2 glass) all have shorter wavelength cutoffs than sapphire. Calcium aluminate also has a strong band near 3000?x2013;3500 cm-1 that is probably due to OH groups. The infrared spectrum of lanthana-doped yttria (polycrystalline material with the composition 0.09La2O3•0.91Y2O3) is
nearly identical to that of yttria.
Fig. 3 shows that single-crystal MgF2 has the same long-wave cutoff as polycrystalline MgF2, but lacks the impurity bands seen
in Irtran-1. The cutoff for LiF is at shorter wavelength and that of CaF2 is at longer wavelength, relative to MgF2. These
materials illustrate the general trend that the long-wave cutoff tends toward longer wavelength as the atoms in a structure become heavier (Li<Mg<Ca).
Fig. 4 compares ZnSe with two varieties of ZnS. These materials are available today as chemical-vapor-deposited, polycrystalline materials. (The historically important hot pressed ZnS, Irtran-2, is no longer available.) ZnSe, an excellent
`2-color’ (3?x2013;5 and 8?x2013;14 m) window material with extremely low absorption, is not as durable as ZnS. Standard
grade (yellow) ZnS is useful in part of the 8?x2013;14 m region, but significant optical scatter reduces its utility in the
3?x2013;5 m region. However, Multispectral® (colorless) ZnS (also called Cleartran® and Waterclear® ZnS), prepared by
heat treatment of standard ZnS, has low optical scatter in the 3?x2013;5 m window. Standard ZnS has a band near 1700
cm-1 attributed to stretching of Zn?x2013;H impurities [10].
The semiconductor materials Ge and GaAs in Fig. 5 are excellent optical windows for both the 3?x2013;5 and 8?x2013;14
m regions at 20°C. However, free carrier absorption becomes significant in Ge above 100°C [11] and in GaAs above 400°C
[12]. GaP is an excellent 3?x2013;5 m window up to 600°C, but has significant intrinsic absorptions in the 8?x2013;14 m
region [13]. Si is also a good 3?x2013;5 m window up to about 250°C [11], but has numerous absorptions in the
8?x2013;14 m region. Many of these absorptions are attributed to oxygen impurities. It is reported that float-zone growth of
single crystal Si, a technique that eliminates quartz crucibles as a source of oxygen, produces Si that is free of impurity bands in
the 8?x2013;14 m range [14].
The ultraviolet and visible optical windows of the materials in Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5 are shown in Fig. 6, Fig. 7
and Fig. 8. Sapphire in Fig. 6 has excellent ultraviolet optical quality. Lower grades of sapphire have an absorption band
centered at 205 nm [15]. Fig. 7 compares the performance of an excellent optical quality specimen of polycrystalline yttria to
tabulated data for single-crystal yttria. It is typical of polycrystalline materials to cut off in the ultraviolet region at longer
wavelength than a single crystal. This is likely due to optical scatter. Fig. 8 shows that silica, quartz, CaF2, calcium aluminate
and multispectral zinc sulfide transmit throughout the visible region. ZnSe and GaP provide just part of the visible window, while
GaAs, Si and Ge cut off in the near infrared.
(6K)
Fig. 6. Ultraviolet/visible transmittance of single-crystal MgF2, (John H. Ransom Laboratories, 2.7 mm thick), polycrystalline
MgF2 (Kodak hot pressed Irtran-1®, 2.0 mm thick), single-crystal LiF (5.4 mm thick), sapphire (Union Carbide single crystal, 60°
cut, 2.6 mm thick), spinel (Coors polycrystalline MgAl2O4, 1.7 mm thick) and polycrystalline ALON (Raytheon, 1.9 mm thick).
(6K)
Fig. 7. Ultraviolet/visible transmittance of polycrystalline yttria (Raytheon Y2O3, 2.0 mm thick), polycrystalline lanthana-doped
yttria (GTE 0.09La2O3•0.91Y2O3, 2.0 mm thick), single crystal MgO and single crystal Y2O3. Spectra of 2.0 mm thick single crystal
materials were computed from the optical constants n and k given by Palik [8] using Eq. 1, Eq. 2 and Eq. 3.
(10K)
Fig. 8. Ultraviolet/visible/near infrared transmittance of quartz (4.7 mm thick), fused silica (5.3 mm thick), CaF2 (Kodak Irtran-3®,
1.1 mm thick), calcium aluminate glass, (2.6 mm thick) multispectral ZnS (Raytheon, 5.2 mm thick), ZnSe (Raytheon, 7.1 mm thick),
GaP (Texas Instruments, 3.0 mm thick), standard ZnS (Raytheon, 6.0 mm thick), GaAs (single crystal, 0.65 mm thick), Si (2.8 mm
thick) and Ge (2.8 mm thick).
Tabulated [7, 8] optical constants used to reconstruct the transmittance of single crystal MgO and Y2O3 in Fig. 7 are n and k,
the real and imaginary parts of the refractive index. To convert n and k to transmittance, the following formulas were used:
(1)
(2)
(3)
Eq. 3 presumes that scatter is negligible, which may not be true at ultraviolet wavelengths. The absorption coefficient () and
sample thickness (b) are customarily expressed in units of cm-1 and cm, respectively. Other parameters in Eq. 1, Eq. 2 and Eq.
3, except wavelength, are dimensionless.
In high quality windows, bulk absorption is low enough that a significant fraction of absorption occurs at the surfaces. For
example [16, 17], at a wavelength of 5.25 m, 1-cm thick polycrystalline, cast CaF2 absorbs 0.042% of the light in the bulk
and absorbs 0.0028% at the two surfaces. In the case of polycrystalline `FLIR-grade’ ZnSe at 8 m wavelength, a 1-cm thick
window absorbs 0.06% of the light in the bulk and 0.3% of the light at the two surfaces [18].
3. Emission at elevated temperature
For use at high temperature, the most important limitation of an infrared window is emission of light from the window itself.
Emission can exceed the signal being observed, especially since the window is much closer to the detector than the object being
viewed. Emission has been reported for Irtran-1, Irtran-2 and Irtran-3 hot pressed materials [19], and for sapphire, yttria,
spinel, ALON and fused silica [20, 21]. Emission is related to absorption: the greater the absorption at a given wavelength, the
greater the emission.
In general, the long wave absorption edge of an infrared window material shifts to shorter wavelength as the temperature rises.
Fig. 9 illustrates this point for spinel [22]. For use in the 3?x2013;5 m atmospheric window, an important comparison in Fig. 1
is the long wave cutoff, which increases in the order ALON<sapphire<spinel<MgOY2O3<MgF2. At room temperature,
each of these windows can be used at wavelengths close to the cutoff. At elevated temperature, there is significant emission
near the cutoff and the cutoff shifts to shorter wavelength.
(6K)
Fig. 9. Temperature dependence of long wave absorption edge of Coors spinel [22]. In general, the absorption edge for most
materials shifts to shorter wavelength as temperature increases.
For a window with absorption coefficient , thickness b, and single-surface normal reflectance R (Eq. 1), the normal emittance
is given by
(4)
A blackbody has an emittance of 1, whereas a perfectly transparent window has an emittance of 0. If we know the absorption
coefficient of a material as a function of temperature, we can use Eq. 4 to compute the emittance as a function of temperature.
A multiphonon model for the absorption of window materials near the long wave cutoff has been developed [23, 24] and
implemented in a computer code [25] sold under the name Optimatr® [26]. Absorption coefficients computed by this code
closely match those measured for oxide and halide window materials over a wide range of temperature. From calculated
absorption coefficients, Eq. 4 was used to compute the emittance of sapphire, ALON, spinel and yttria in Fig. 10. Just as the
absorption edge shifts to shorter wavelength with increasing temperature in Fig. 9, the emission edge also shifts to shorter
wavelength with increasing temperature in Fig. 10. Fig. 11 compares the emission edges of the four materials of Fig. 11 at a
single temperature. The important point is that the shorter the wavelength of the cutoff in the absorption spectrum, the shorter
the wavelength of the emission edge.
(30K)
Fig. 10. Emittance of window materials computed with Eq. 4 using absorption coefficients computed by the program Optimatr
[26].
(7K)
Fig. 11. Comparison of calculated emittance of ALON, sapphire, spinel and yttria at 700 K using absorption coefficients
computed with Optimatr [26]. Note the correspondence between emittance and the long wave cutoff behavior in Fig. 1.
There is an important limitation on the use of the multiphonon model for absorption and emission. The model is based on the
behavior of the absorption edge where the intrinsic absorption of the material dominates over effects of impurities and crystal
defects. In the window region where the material of interest has no significant absorption, impurities and defects dominate the
observed optical behavior. For example, Fig. 12 compares the measured [27] emittance of yttria at 1000°C with that
calculated by Optimatr. The calculated emittance is negligible at wavelengths shorter than 4.5 m. The observed emittance has
a plateau with 0.08 stretching from 3.5 to 5 m. Although OH impurities in Y2O3 could be expected to give rise to
absorption and emission near 3 m, it is not clear how to account for emission in the 4?x2013;5 m region. The lesson from
Fig. 12 is that the emittance of a window material in its window region must be measured and cannot be accurately modeled. If
the measured and modeled behavior are in agreement over some range of temperature, it is probably safe to extrapolate the
model to other temperatures to predict window performance.
(6K)
Fig. 12. Comparison of measured [27] emittance of yttria at 1000°C with that calculated by Optimatr [26] using Eq. 4.
The example in Fig. 12 was chosen to illustrate poor agreement between measured and calculated emittance. Spinel, ALON
and ultraviolet-grade sapphire show excellent agreement between theoretical and measured emittance at wavelengths as short
as 3 m, with no obvious impurity emission [20, 21]. By contrast, a lower grade of sapphire has an emission band centered
near 3 m that is not predicted by the model [20, 21].
4. Refractive index
Good sources of information on refractive index of optical materials are available [26, 28, 29]. Table 2 lists the refractive index
(n) of many midwave infrared materials near 20°C, and indicates the temperature dependence (dn/dT). Single crystals of
sapphire, MgF2 and hexagonal SiC have 3- or 6-fold axes of symmetry called the optical axis or the ordinary direction. The
refractive index in the direction of the 3- or 6-fold axis is designated no. The refractive index on the 2-fold symmetry axis
perpendicular to the optical axis is designated ne. Other materials in Table 2 have isotropic refractive index.
Table 2. Refractive index and absorption coefficient near 300 K
(1K)
In Table 2, fluorides have the lowest refractive index, followed closely by oxides. Materials with heavier atoms have higher
refractive index. Oxides and fluorides have a relatively small temperature dependence dn/dT. The chalcogenides (ZnS, ZnSe,
AMTIR) have greater values of dn/dT and the semiconductors (Ge, Si, GaAs, GaP) have the greatest temperature
dependence.
The higher the refractive index, the greater the reflection loss. The theoretical transmittance in air of a flat plate with no
absorption or scatter loss is given by
(5)
Table 2 lists the transmittance calculated with Eq. 5. This reflection-limited transmittance accounts for the different values of
baseline transmission observed in Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5.
5. Optical scatter
Light striking an ideal optical window at normal incidence is transmitted through the window with no change in direction. In all
real materials, some light is scattered in all directions from both surfaces and from the bulk of the optical window.
Total integrated scatter (TIS) [48] in Table 3 is measured with an integrating sphere by collecting all light scattered between
2.5° and 70° from the normal direction in either the forward hemisphere or the back hemisphere. Light scattered <2.5° is
considered to be specular for the sake of this measurement. Light scattered by more than 70° is generally negligible. TIS is a
dimensionless ratio of the power (W) of scattered light divided by the power of incident light (measured with no sample in the
beam). TIS can be measured in either the forward (transmitted) or backward (reflected) hemisphere.
Table 3. Optical scatter
(39K)
An alternate measurement called bidirectional transmittance distribution function (BTDF) [53] measures scattered light as a
function of angle. BTDF is defined as (Ps/Pi)/( cos ), where Ps is the power of scattered light, Pi is the power of incident
light, is the scattering angle measured from the incident (normal) direction, and is the solid angle of scatter viewed by the
detector. When is very small, the detector collects specularly transmitted light. As increases from zero, the detector collects
scattered light whose intensity decreases rapidly with angle. By integrating the scattered light from 2.5° to 70°, the BTDF
measurement gives results similar to the TIS measurement, as verified with lanthana-doped yttria at 0.63 and 3.39 m [52] The
bidirectional reflectance distribution function (BRDF) measured scatter in the backward hemisphere.
Polycrystalline materials generally exhibit more bulk scatter than single crystals and glasses. Polycrystalline materials with
noncubic crystal structures tend to have the greatest scatter because the refractive index changes from grain to grain (because
grains are oriented randomly). This causes a light ray to shift its course as it passes from grain to grain. If the grain size is much
smaller (1/10) than the wavelength of light, the scatter by individual grains is not significant. Grain boundaries in polycrystalline
materials give rise to scatter because the boundaries generally have a different composition (and different refractive index) from
the bulk grain.
6. Mechanical and thermal properties
Table 4 lists mechanical and thermal properties of interest in predicting the performance of optical windows. Unless otherwise
indicated, properties apply near 300 K. Although many properties are intrinsic to the material, strength, hardness and thermal
conductivity are extrinsic properties that depend on microstructure and impurities.
Table 4. Physical properties
(25K)
The most variable property in Table 4 is strength, which measures the force per unit area required to fracture the material in
flexure or tension. A brittle material fractures when the stress exceeds the strength of the weakest flaw that is present, usually at
the surface. In general, strength decreases with increasing area under load, because the probability of a weak flaw being
present increases with area. Table 5 illustrates the general observation that strength increases if the surface finish is improved
[16, 17]. In Table 5, the strength of CaF2 is more than doubled by improved polishing. The strength also doubles when CaF2 is
annealed, which presumably `heals’ some crystal defects.
Table 5. Effects of surface and bulk treatments on strength of polycrystalline cast CaF2 [16, 17]
(60K)
In another example of the dependence of strength of surface and bulk treatments, the median strength of sapphire samples
increased from 660 MPa to 850 MPa with annealing, to 1120 MPa with noncontact `superpolishing’, and to 1570 MPa with
surface ion implantation [79]. The orientation of polishing scratches relative to the tensile axis of a flexure bar also affects
strength. Specimens of several different optical materials polished perpendicular to the tensile direction were 20?x2013;40%
weaker than specimens polished parallel to the tensile direction [80].
Strength can also depend on the atmosphere in which testing is performed, with water and oxygen playing critical roles in the
surface chemistry of a developing crack tip as a material fractures. In one example, the strength of ZnS was independent of
temperature over the range -100°C to +600°C when tested in dry nitrogen atmosphere [81]. When tested in air, the strength of
ZnS increases by 50 [82]?x2013;100% [78] when the temperature is raised from 25°C to 500°C.
The strength of a material can depend on its microstructure. Cases are known in which a manufacturer quotes the strength of a
material whose microstructure is not optimized for optical performance. A better optical grade of the material might have less
strength than that quoted. If strength is critical to a particular design, there is no substitute for measuring the strength of the
particular grade of material that will be used with the particular finish that will be used.
7. Thermal stress resistance
When a brittle optical window is heated rapidly, part of the window becomes warm and expanded, while other parts are still
cool and not expanded. If heating is rapid enough, the stress between the warm and cool regions is sufficient to fracture the
window. A figure of merit that describes relative resistance to failure by thermal stress is designated R’ [83]:
(6)
where is the strength of the material, is Poisson’s ratio, k is the thermal conductivity, is the thermal expansion coefficient,
and E is Young’s modulus.
Eq. 6 compares the maximum allowable temperature difference (Tinside-Toutside) to which internally heated cylinders made of
different materials can be subjected under conditions of convective heat transfer. The maximum allowable temperature
difference is proportional to R’. Eq. 6 applies under conditions of `mild’ heating, defined by the condition bh<k, where b is the
thickness of the material, h is the surface heat transfer coefficient, and k is the thermal conductivity of the material (W/m K).
The heat transfer coefficient is the heat entering the material per unit area per unit temperature difference between the
atmosphere and the surface (W/m2 K).
Table 6 gives the thermal shock figure of merit for materials in Table 5 computed with physical properties that apply near 300
K. Since the physical properties generally depend on temperature, and since thermal stress failure generally occurs above 300
K, the R’ values in Table 6 should only be considered for their qualitative significance.
Table 6. Thermal shock figure of merit (R’) near 300 Ka
(11K)
Table 6 tells us that the order of thermal stress resistance is SiC>Si>(GaP, sapphire, GaAs and Ge)>(fused silica, MgO,
ZnS)>(ZnSe, ALON, Y2O3, spinel, MgF2)>(CaF2, calcium aluminate)>AMTIR-1. SiC offers promise as a
thermal-stress-resistant optical window, but it is still developmental and has not yet reached its potential for optical
transmission. It is not known whether pure silicon carbide will have a useful midwave infrared transmission window at elevated
temperature. Si has very high thermal stress resistance, but its useful upper operating temperature is only 250?x2013;300°C
because of free carrier absorption. So although Si can be heated rapidly, it cannot be taken to high temperature.
In the next category of materials, GaP has the highest thermal stress figure of merit Its upper use temperature is probably near
600°C, as dictated by free carrier generation. Sapphire is also excellent, but not nearly so resistant as the 300 K value of R’
suggests. The strength and thermal conductivity of sapphire fall rapidly above 300 K, so the value of R’ for sapphire at a more
realistic thermal stress temperature (such as 200?x2013;500°C) is much lower than the value in Table 6. Nonetheless, sapphire
has the greatest demonstrated thermal stress resistance of commercially available, durable, midwave infrared materials [85].
GaAs and Ge also have excellent thermal stress resistance, but their upper operating temperatures are approximately 400°C
and 100°C, respectively, due to free carrier generation. GaP and sapphire are therefore the top two candidates for a
high-thermal-stress environment.
In the next tier, ZnS is indeed very resistant to thermal stress. Fused silica does not have sufficient bandpass for midwave use
and MgO is not a commercially available optical material at this time (although it was formerly produced as hot pressed
Irtran-5®). The remainder of the order of thermal stress resistance derived from Table 6 is qualitatively correct. Table 7 gives
the observed thermal stress resistance of optical dome materials tested under aerothermal heating conditions in a wind tunnel.
Results are consistent with the qualitative predictions of Table 6.
Table 7. Observed thermal stress capability of infrared-transmitting domes [85]
(5K)
8. Erosion resistance
Optical windows for external use are eroded by high speed impact with rain, and low or high speed impact with solid particles
such as blowing sand. Research effort is currently focused on creating durable protective coatings for infrared optical windows.
One way to measure relative rain impact resistance is with a laboratory waterjet. Fig. 13 shows the damage threshold velocity
for 300 impacts on the same spot by a waterjet delivered from a 0.8-mm diameter nozzle [86]. The damage threshold velocity
for a material is the waterjet velocity below which no damage is observed after 300 impacts and above which damage is
observed. ALON is expected to come between spinel and sapphire on this graph. Yttria would probably be near MgF2.
Multispectral ZnS and ZnSe have lower damage threshold velocities than standard ZnS, which is the data point shown in Fig.
13.
(6K)
Fig. 13. Waterjet damage threshold velocity (m/s for 300 impacts of a 0.8-mm diameter jet) is correlated with the logarithm of
fracture toughness (MPa) of optical materials [86].
Sand erosion is less well understood than rain erosion. As an example of recent work, the behaviors of ZnS and Ge were
compared [87]. For perpendicular incidence of 0.3- to 0.6-mm diameter sand at a velocity of 34 m/s, the erosion rate of Ge
(expressed as mg of mass lost from the window per kg of sand) was 75% greater than that of ZnS. When the sand velocity
was increased to 59 m/s, the erosion rate of Ge was 8% less than that of ZnS.
Ge and ZnS are relatively easily eroded. The oxides sapphire, ALON and spinel are much more durable and have generally
acceptable rain and sand erosion resistance for subsonic impact speeds. No material has adequate resistance for supersonic
rain impacts.
9. Future directions
There remains a need for more thermal-stress-resistant midwave infrared materials. There is also a need for greater erosion and
thermal stress resistance in materials that cover both the midwave and long wave regions. The most durable midwave infrared
materials (sapphire, ALON and spinel) all have more optical emission than desired at elevated temperature.
Chemical-vapor-deposited diamond holds promise for providing thermal stress and erosion resistance in the long wave region,
but its midwave absorption and emission rule it out for midwave applications (except as a thin coating). Research in erosion
resistance is evaluating both hard and soft coatings. Thermal stress resistance must focus on the bulk window material, because
there is no coating that can carry heat away fast enough. SiC, AlN and Si3N4 each promise improved thermal stress resistance
(and outstanding erosion resistance), but none is available in a clear, optical form. All three will restrict the available window
because of high optical emission at elevated temperature. It is conceivable that the range of material properties could be
expanded with nanometer-scale starting materials. In the absence of new materials, improved performance of infrared systems
requires engineering innovations to protect the window from the harshest conditions.
Acknowledgements
We are indebted to Mel Nadler and Andy Wright for obtaining the spectra used in the article and to Sally Harris for preparing
most of the graphics. This work was supported by the Office of Naval Research.
References
1. W.L. Wolfe, Optical materials, in: W.D. Rogatto (Ed.), The Infrared and Electro-Optical Systems Handbook, Vol. 3, SPIE
Optical Engineering Press, Bellingham, Washington, 1993, pp. 1?x2013;78. ISBN 0-8194-1072-1.
2. W.L. Wolfe, Optical materials, in: W.L. Wolfe, G.J. Zissis (Eds.), The Infrared Handbook, Infrared Information and
Analysis Center, Environmental Research Institute of Michigan, Ann Arbor, MI (1985 edn.). ISBN 0-9603590-1-X.
3. J.A. Savage, Infrared Optical Materials and Their Antireflection Coatings, Adam Hilger, Bristol, 1985. ISBN
0-85274-790-X.
4. P. Klocek (Ed.), Handbook of Infrared Optical Materials, Marcel Dekker, New York, 1991. ISBN 0-8247-8468-5.
5. S. Musikant, Optical Materials, Marcel Dekker, New York, 1985. ISBN 0-8247-7309-8.
6. D.C. Harris, Infrared Window and Dome Materials, SPIE Optical Engineering Press, Bellingham, Washington, 1993. ISBN
0-8194-0998-7.
7. E.D. Palik (Ed.), Handbook of Optical Constants of Solids, Academic Press, New York, 1985. ISBN 0-12-544420-6.
8. E.D. Palik (Ed.), Handbook of Optical Constants of Solids, Vol. II, Academic Press, New York, 1991. ISBN
0-12-544422-2.
9. J.T. Gourley and W.A. Runciman. J. Phys. C 6 (1973), p. 583. Abstract-INSPEC
10. K.L. Lewis, G.S. Arthur and S.A. Banyard. J. Cryst. Growth 66 (1984), p. 125. Abstract
11. D.T. Gillespie, A.L. Olsen and L.W. Nichols. Appl. Opt. 4 (1965), p. 1488.
12. P. Klocek, T. McKenna and J. Trombetta. Proc. SPIE 2286 (1994), p. 70. Abstract-Compendex
13. P. Klocek, J.T. Hoggins and M. Wilson. Proc. SPIE 1760 (1992), p. 210. Abstract-Compendex
14. K.V. Ravi. J. Spacecraft Rockets 30 (1993), p. 791.
15. F. Schmid, C.P. Khattak and D.M. Felt. Bull. Am. Ceram. Soc. 73 (1994), p. 39. Abstract-Compendex
16. R.T. Newberg, J. Pappis, Casting of Halide and Fluoride Alloys for Laser Windows, AFCRL-TR-76-0011. Air Force
Cambridge Research Laboratories, Hanscom Air Force Base, MA, February 1976.
17. P. Miles. Opt. Eng. 15 (1976), p. 451. Abstract-INSPEC | Abstract-Compendex
18. C.A. Klein, R.P. Miller and D.L. Stierwalt. Appl. Opt. 33 (1994), p. 4304. Abstract-Compendex | Abstract-INSPEC
19. S.E. Hatch. Appl. Opt. 1 (1962), p. 595.
20. R.M. Sova, M.J. Linevsky, M.E. Thomas and F.F. Mark. Proc. SPIE 1760 (1992), p. 27. Abstract-Compendex
21. R.M. Sova, M.J. Linevsky, M.E. Thomas and F.F. Mark. Johns Hopkins APL Technical Digest 13 (1992), p. 368.
Abstract-INSPEC | Abstract-Compendex
22. M.E. Thomas, R.I. Joseph and W.J. Tropf. Appl. Opt. 27 (1988), p. 239. Abstract-INSPEC
23. M.E. Thomas and R.I. Joseph. Proc. SPIE 929 (1988), p. 87.
24. M.E. Thomas and R.I. Joseph. Johns Hopkins APL Technical Digest 9 (1988), p. 328. Abstract-INSPEC |
Abstract-Compendex