Skip to main content

Full text of "DTIC ADA1049433: Low-Loss Fiber Waveguides."

See other formats



J. A. Harrington 

Hughes Research Laboratories 
3011 Malibu Canyon Road 
Malibu, CA 90265 

September 1981 

Annual Technical Report 
1 August 1980 to 30 June 1981 

Approved for public release; distribution unlimited. 


E'. r " '"P? 

OCT 2 1981 ’ 



Prepared for 

Office of Naval Research 
800 North Quincy Street 
Arlington, VA 22217 

Si 10 2 097 


SECURITY CLASSIFIC ATiqn OF This PAGE fW»»«n Data Entered) 


For r-* umber 

6 1 


ADr Ao 1 

I 4.\ TITLE cM*d Subtil!*) 


I < * 


[7, AUTHO«C*J 

)j J.Ay Harrington ^ 


Hughes Research Laboratories 

3011 Malibu Canyon Road 

Malibu, California 90265 _ 


Material Sciences Division 

Office of Naval Research 

800 No. Quincy St., Arlington, VA 22217 

14. MONITORING AGENCY NAME ft A O C R E S Si fit^jUIS^en t from Contmlh, 




%. TYPE OF Rga^JM-■# ^ RTOtT c6v|WO 

: Annual Technical . 

/ 1 Aug »8j/~ 3/ Jun' 181, 

j'V PERFORMING r f - r-'- NUMBER^ 


jl NOOO14-79-C-0691 


12. r O pbrt DATE _ 

13 ” 


IS SECURITY CLASS tol this report) 




16. DISTRIBUTION STATEMENT (ol this Report, 

Approved for public release; distribution unlimited. 

IT. DISTRIBUTION STATEMENT (ot the abstract entered m flhuk .*<_>. it different from Report) 

\G t 


I 19. KEY WORDS (Continue on reverse aide it necessary onrl identify e»%* block number) 

Infrared fibers, optical materials, fiber optics, light scattering, 
crystal growth. 

12$ ABSTRACT (Continue on reverie aide ft neceatery and Identity bv block numb .r) 

This annual report summarizes our second year of effort to fabricate 
low-loss fiber waveguides with potential loss near 10“^ dB/km. To 
develop such low-loss waveguides, we have used alkali and thallium hal¬ 
ides to form poly- and single-crystal fibers. To date our best poly¬ 
crystalline KRS-5 (TIBrI) fiber remains 5 orders of magnitude above the 
goal of 10 - 3 dB/km. 

00 , jaCts 1473 Moition or i nov ss is obsolete 




TV CL. A$Sl P lC * T ION OP T mS P*G£ 0»i* /Tn/«redj 

Our approach to developing low-loss waveguides involves three interrelates 
tasks: (L) fiber materials preparation, (2) fiber fabrication, and 

(3) optical evaluation and analysis. In the first task, we have combined 
our reactive atmosphere process (.RAP) ehemistrv techniques, which save 
proven so successful for alkali halides and zone refininv; to prepare 
j thallium halide crvstais for use in fiber fabrication. '.-.'bile we .-.are yet j 
( to evaluate these crvstais optically, we expect substantial reduction in j 
Che anion impurity concentration. 

In the second task, we devoted the majority of our efforts to the fabri¬ 
cation of SC fibers. The fabrication of SC fibers holds, we feel, the 
greatest promise for achieving the program goal of 10 - ^ dB/km. By going 
from poly- to single-crystal waveguides, we will minimize scattering 
losses associated with grain boundaries (impurities decorating grains, 
inherent surface roughness, and grain boundary separation), and with the 
fabrication process (for example, stress-induced birefringence during 
extrusion). Our first attempts at SC fiber growth involved an inverted 
Czochralski method, combined with RAP chemistry to grow SC KC1 fiber. In 
this method, molten KC1 in the presence of a RAP environment is directed 
vertically downward through a quartz, fiber-forming orifice. To date, we 
have not been able to prepare any SC fiber using this apparatus because 
of the inability to precisely control the temperature at the liquid- 
solid interface. Another SC fiber growth method we have developed uses 
our extrusion apparatus. In this pressure-fed SC fiber growth, only a 
small portion of the solid is melted near the exit of the extrusion die. 
This forms the growth interface for fabrication of SC fibers. 

The third task, optical evaluation and analysis, has been directed at 
understanding the nature of scattering and absorptive losses in our IR 
fibers. Using a scattering sphere we measured the attenuation due to 
scattering, i g , and the total attenuation,a^, as a function of IR laser 
wavelength. We found that for most fibers the scattered light varies 
between 40 and 60%’of the total light. Further, we found that 
and that there is no simple wavelength dependence for a s . This, we feel, 
is a result of scattered light being re-absorbed at the surface of the 
fiber. In general, our scattering measurements reveal that the surface 
quality of our fibers is a major source of fiber attenuation. Additional 
measurements of the attenuation in KRS-5 fiber while the fiber is under a 
constant tensile load have indicated the possibility that stress-induced 
birefringence may also be a major contributor to the fiber's losses. In 
particular, KRS-5, and very likely other ductile fiber materials (the 
thallium and silver halides), exhibits residual strain induced during the 
extrusion process. This strain and resultant variation in the index of 
refraction contributes to the attenuation coefficient due to scattering. 


5ECj PiTv Cl AS'lfiC ATtON Of T 

S P*GE'»***n Omtm 






















Projected transmission in IR fibers . 

Comparative composite thermogram of (a) ALFA ultrapure 
thallous iodide and (b) its RAP zone refined 

counterpart. 15 

Comparative composite thermogram of (a) ALFA ultrapure 

thallous bromide and (b) its RAP-zone-refined 

counterpart. 16 

SC fiber crucible for growth and RAP purification 

of crystalline fibers . 22 

Our first apparatus for inverted Czochralski growth 

of SC fibers. 23 

Special die for pressure-fed, SC fiber growth . 24 

C0 2 laser insertion loss apparatus . 29 

IR fiber monochromator for 2.5 to 14.5 um spectral 

information. 30 

Scattering measurements at IR laser wavelength . 32 

Transmission of two KRS-5 fibers from 2.5 to 14 um. 33 

Calculation o f a s. 35 

Scattering losses in KRS-5 fiber at 10.6 um. 36 

Effect of output end finish on . 37 

Scattering losses in KRS-5 fiber at two IR laser 

wavelengths. 39 

Attenuation at 10.6 um in KRS-5 fiber with different 

grain size . 40 

Percentage of scattering loss at 10.6 um. 42 

Attenuation at IR laser wavelengths . 43 

Mechanical strength of KRs-5 fiber. -»6 

Attenuation in KRS-5 fiber under applied tensile load .... 47 



Ab-> Tpt ion In KRS-5 liber .it LO.o ...m while -:e 
is under a constant tensile load. 

Laser heating tor crucible-less SC fiber growth 



For the past several years, Hughes Research Laboratories has been actively 
engaged in the development of ultra-low-loss fiber waveguides for future long¬ 
distance communications systems. With support from contract (ONR, DARPA/NRL, 
and RADC) and Hughes IR&D funds, we have fabricated fibers from highly trans¬ 
missive, infrared (IR) crystals. The projected losses for the alkali and 
thallium halides we have studied are near 10""^ dB/km. Unfortunately, the best 
crystalline IR fiber made from KRS-5 (TIBrI) has a measured loss of 300 dB/km 
at 10.6 am — three orders of magnitude above the intrinsic loss for KRS-5 at 


this wavelength, and far above the program goal of 10 dB/km. 

Our approach to achieving the ultimate low-loss fiber has concentrated 
on developing waveguides from crystalline materials. This approach is based 
on theoretical predictions which show that a variety of IR crystals should 
have losses no greater than LQ -2 dB/km near 5 um.^» 2 For review, we show 
projected losses for KRS-5 and KC1 in Figure 1, along with those of fused 
Si02 for comparison. From these familiar V-shaped loss curves, we see that 
the advantage of crystalline materials derives from the shift of the multi¬ 
phonon edge (Reststrahl peak) to longer wavelengths. For fused Si02 or other 
oxide glasses, the multiphonon edge occurs at shorter wavelengths because of 
the high vibrational frequencies of the Si-0 bond. 

From Figure 1 we see that this limits the ultimate loss in these materials 
to 0.2 dB/km at 1.55 um, a value only recently achieved in kilometer-long 
lengths of silica-based fiber 2 . The crystalline materials, therefore, are in 
principle, excellent candidates for next generation waveguides if we can reduce 
the dominant extrinsic attenuation losses that now severely limit our crystal¬ 
line fibers. 

During the second year of our ONR program we have studied the purification 
of thallous halides and developed new methods to fabricate single crystal (SC) 
fibers. Combining reactive atmosphere process (RAP) chemistry techniques, 
which have proven so successful for alkali halides and zone refining, we have 
prepared thallous halide crystals for use in fiber fabrication. While we have 

1.0 2.0 

3.0 5.0 10.0 20.0 



Figure 1. Projected transmission in IR fibers 

yet to evaluate these crystals optically, we expect substantial reduction in 
the anion impurity concentration. The fabrication of SC fibers holds, we feel, 


the greatest promise for achieving the program goal of 10 dS/km. 3v going 
from poly to single crystal waveguides, we will minimize scattering Losses 
associated with grain boundaries (impurities decorating grains, inherent sur¬ 
face roughness, and grain boundary separation), and with the fabrication pro¬ 
cess (for example, stress-induced birefringence during extrusion). 

Our first attemps at SC fiber growth involved an inverted Czochralski 
method, combined with RAP chemistry to grow SC KC1 fiber. In this method, 
molten KCL is directed vertically downward through a quartz, fiber-forming 
orifice. The melt and fiber are enclosed in an RAP atmosphere, both to purify 
the KCL and to reduce the surface tension between the quartz orifice and KC1 
fiber. To date, we have not been able to prepare any SC fiber using this 
apparatus because of the inability to precisely control the temperature at the 
liquid-solid interface. For better temperature control we have added an 
auxiliary heater to the apparatus. 

Another SC fiber growth method we have developed used our extrusion 
apparatus. In this pressure-fed SC fiber growth, only a small portion of the 
solid is melted near the exit of the extrusion die. This forms the growth 
interface for fabrication of SC fibers. 

In addition to fabricating IR fibers, we have been evaluating the optical 

and mechanical properties of the waveguides. Our optical studies have 

endeavored to separate scattering and absorptive losses in polycrystalline 

KRS-5 fibers. Using a scattering sphere we measured the attenuation due to 

scattering, a , and the total attentuation, as a function of IR laser 

wavelength. We found that for most fibers the scattered light varies between 


40 and 60% of the total light. Further, we found that \ , and that 

there is no simple wavelength dependence for jig. This, we feel, is a result 
of scattered light being re-absorbed at the surface of the fiber. In general, 
our scattering measurements reveal that the surface quality of our fibers is 
a major source of fiber attenuation. 

Additional measurements of the attenuation in KRS-5 fiber while the fiber 
is under a constant tensile load have indicated the possibility that stress- 
induced birefringence may also be a major contributor to the fiber's losses. 


In particular, KRS-5, and very likely ocher ductile fiber materials (.the 
thallium and silver halides), exhibits residual strain induced during the 
extrusion process. This strain and resulting variation in the index of refrac¬ 
tion contributes to the attenuation coefficient due to scattering. 

In Sections 2 through 4 we discuss the technical progress in three areas: 
Section 2, Fiber material preparation; Section 3, Fiber fabrication; and 
Section 4, Fiber evaluation. In Section 5, our recommendations for future 
work are given. 





The fabrication of Low-loss crystalline fibers requires ultra pure 
starting materials and new fiber fabrication techniques. In this section we 
describe the chemistry involved in purifying thallous halides and our attempts 
to prepare this material in ultra-pure form. Because we have been so success¬ 
ful in the extrusion of polycrystalline thallous halides, we have devoted our 
efforts toward preparing very pure thallous halides for use in both extrusion 
and single-crystal (SC) fiber growth. The primary emphasis has been on the 
binary, medium-melting solid solution of thallous bromide and thallous iodide, 

or KRS-5. The composition of this material is TlBr. ..I. ... 

0.46 O.o4 


In order to obtain uniformity and homogeneity in the SC fiber, crystal 
growth must be carried out as a steady-state operation, with the crystal 
growth direction along the fiber axis. Because of the relatively long crystal 
dimension in this direction, a fast crystal growth rate is essential; and, 
therefore, a melt growth method, such as the Czochralski method, the Bridgman 
method, or some method that is derivative of these, is much preferred over 
solution or flux growth methods. 

To obtain fast, steady-state crystal growth from the melt, the charge 
material must be a congruent melter. This means that the charge material 
must have identical chemical compositions in its crystalline and molten phases 
(In that case, no segregation or partition takes place across the crystal-to- 
melt interface; therefore, the compositions of the crystal and the melt are 
invariant during crystal growth - a necessary condition for steady-state crys¬ 
tal growth.) In the ideal sense, melting congruency may be interpreted to 
mean that the material is a pure chemical compound which is thermochemicaily 
stable in the vicinity of its melting point. In less strict and more realisti 
terms, it is sufficient that the concentrations of any impurities present in 


the material be below particionable levels. We have employed the technique 
of cone refining under a stream of reactive gases * RAP cone refining,* to raise 
the purity levels of commercially available challous halides to the level of 
melting congruency of these materials. 

There is already some evidence derived from extruded polycrystalline 
KRS-5 fibers that suggests that the mechanical properties of this material, 
namely, its yield strength, could be improved. Such improvement may be 
brought about by the addition of controlled quantities of specific impurities. 
For example, yield strength may be increased by blocking slip, which in turn 
may be accomplished by rendering slip planes less "smooth" through the sub¬ 
stitution of size- or coordination-mismatched impurity species for some host 
ions. Dopants that may strengthen the thallous halides include Sr" 1 "" 1 ", Rb + , 

|. j |- 

Ag , and Eu . In the past, Rb and Sr have greatly strengthened the 
alkali halides. 


The components in KRS-5 are thallous bromide and thallous iodide; those 
of KRS-6 are thallous bromide and thallous chloride. The unary thallous 
halides were obtained by us in Che form of ultrapure-grade powders from Alfa 
Products, Thiokol/Ventron Division. These compounds are susceptible to hydrol¬ 
ysis by moisture from the air, and therefore their anion purity is subject to 
progressive degradation even with intermittent exposure to the atmosphere. 

Hydrolysis will convert the thallous halide to thallous hydroxide, as in 
the reaction, 

TlX(s) + H 2 0(g) - T10H(s; + HX(g), (a) 

where X=I , Br , or Cl . Although this reaction may be incident only on the 
surface of the solid, it may occur extensively in the raw materials since 
these are in powder form. 

Thallous hydroxide readily decompoases upon hearing 

:ia me reaction. 


139 “ — 11,0(3) -r H0 1 g ) . 

Note that the water molecule that is a product of reaction (b) may serve as a 
reactant in reaction (a). If the water is completely recycled, the overall 
reaction will be 

2TLX(s) + H 2 0(g) J 3 ^- T1 2 0(s) + 2HX(g), (c) 

in which case each molt of water will displace two moles of halide ion. How¬ 
ever, part of the water produced in reaction (b) is lost during calcination, 
and the actual extent of hydrolytic degradation will result in an [X ] to 
[H o 0] ratio between one and two.* 

Thallous oxide (T1 2 0) is a basic oxide, and has a melting point of 300°C, 
which is far below the melting points of thallous iodide, thallous bromide, or 
thallous chloride. (These are, respectively, 442.5 ±1.0, 461.5 ±1.0, and 
431.5 ±1.0°C by our own DTA measurements.) In molten form, thallous oxide 
will react with silica, and consequently cannot be contained by vitreous-silica 
ware. Indeed, our initial experiences with the "ultrapure" thallous halide 
powders in glass or fused-quartz containers lead us to the inevitable conclu¬ 
sion that these materials would have to be stripped of their hydroxide and/or 
oxide content before they could be melted in such containers. 

The chemical stripping process that we adopted was the reversal of 
reaction (a) or (c), viz., 

T10H(s) + HX(g) ^~ 39 °c TlX(s) + H,,0(g), <d) 


*These anion impurities, OH and 0 , derived from the action of water, degrade 
the optical transparency of the host in the infrared. The oxvgen-to-hydrogen 
stretch is active at •'-3 urn, and the oxygen-to-metal stretch is active at '40 urn. 


n 2 0(s) + 2HX(g) - }Q0 - — -T1X I's) + H.,0 (g) ; e> 

i.e., heating the contaminated thalLous halide powder under a stream of the 
corresponding hydrogen halide gas without melting its thallous oxide content. 
Helium was used as the carrier gas so that the HX gas flow rate could be main¬ 
tained at controllable levels. After approximately five hours of this treat¬ 
ment the thallous halide powder could be melted in vitreous-silica ware with¬ 
out corrosion of the latter. 

Further purification of the thallous halide was carried out by RAP-zone 
refining with the same gas mixture. The iterative crystallization process, 
which is the basis of zone refining, serves to clean up the charge material 
via the mechanism of segregation; i.e., the expulsion from the crystal lattice 
of species that are alien to it. However, there are compensation reactions, 
either between two different alien species or between alien species and native 
point defects that may inhibit the segregation mechanism. In RAP-zone refin¬ 
ing, in which a second mechanism is simultaneously provided for anion purifica 
tion, many of these compensation reactions become inoperative, and the segrega 
tion mechanism is more potent. 

Figures 2 and 3 show comparative DTA thermograms of pre- and post-RAP- 
zone-refined thallous iodide and thallous bromide, respectively. More sig¬ 
nificant than the raising of the melting point of the material by the RAP- 
zone-refining in both cases (2.5°C for Til and 2.0°C for TIBr) is the sharpen¬ 
ing of the knee of the melting endotherm, which is characteristic of melting 

RAP-zone-refined thallous bromide and thallous iodide were used to grow 

an SC ingot of KRS-5. Its melting point was 414.0 ±1.0°C by DTA measurement. 

Another SC ingot of KRS-5 doped with strontium iodide to a starting concentra- 
2 + 

tion of LOO Sr ions per million cations was similarly grown. Its melting 
point did not deviate from that of its undoped counterpart within the limits 
of accuracy of the measuring procedure. There was also evidence of incomplete 
dissolution of the dopant. 




Figure 2. Comparative composite thermogram of (a) ALFA ultrapure thallous 
iodide and (b) its RAP zone refined counterpart. The melting 
point read from (a) is 440°C, and that from (b) is 442.5°C. 

The solid-solid transition occurs at 197.5°C in (a) and at 
182.5° in (b). The shallow endotherm at around 90°C in ia), 
presumably the evaporation of moisture, does not appear in (b). 


It is expected that our zone-refining work on the thallous halides 

will Lead to purer starting materials, and therefore lover-loss fibers. 

Recentlv, .\rtvushenko, et al. 4 at the Lebedev Russian Institute reported hulk 
.... . , - , -5 -i 

absorption coeirieients between j-o x Lh cm , or m-- :o u-. purt: t=t 

bulk KRS-5, T1C1, and KRS-b (TIBrCl). Fley have indicated tnat H, 1 and ot/.er 
anion impurities such as C0~ , and HCO^, and CIO.^ limited the bulk transmission 
and that their "repeated crystallization" (probably zone-refining) improved 
the material. So far, we have not optically evaluated our purified materials, 
but we plan to do IR laser calorimetry and evaluate fibers made from these 
zone-refined materials during the upcoming year. 




The major emphasis in fiber fabrication has been on single-crystal (SC) 
fiber growth. SC fibers offer significant advantages over polycrystalline 
fibers. Some advantages are smaller scattering losses through reduction in 
strain birefringence, elimination of grain boundaries, and >'etter surface 
quality. During the past year we have examined the chemistry and kinetics of 
SC fiber growth using an inverted Czochralski growth method, and have attempted 
to prepare SC fibers using a modified extrusion technique. 


We began our study of SC fiber growth by examining the kinetics of crystal 
growth for a system (fiber) which must be grown at a very uniform rate for good 
surface quality. The key factor in producing a uniform fiber is the surface 
tension of the melt. This factor will determine the kinetics of heat and mass 
transport at the growth interface. 

The crystal growth rate, R, based exclusively on a heat balance across 
the growth interface, is given by 

• _ kM dT 

oH dz ’ 

( 1 ) 

where k is the thermal conductivity of the crystal in the growth direction, M 
is its molecular weight, p is its density just below the melting point, H^ is 
its molar latent heat at crystallization, and dT/dz is the longitudinal or 

axial thermal gradient in the crystal just inside the growth interface. The 

2 • • 

rate of momentum transport from melt to crystal is ir . cR’R. or 



where r is the radius of the fiber. 


The inverted Czochralski apparatus that we proposed using is shown in 
Figure 4. The molten halide salt is formed into fiber in the precision bore 
capillary. An auxiliary heater is used to obtain precise control of the growth 
interface. The entire system is made of quartz. in our considerations of 
fiber growth kinetics, the friction between the canillary may be important. 

Our experiments will give us empirical values for this force. 

The virtual force in Equation (2) may now be included in the force- 
balance equation for the inverted Czochralski system. We have, for the force- 
balance equation. 

2 i 2 

wr hpg = 2r ry + r : 

. 9 



where h is the hydrostatic head of melt over the growth interface, o' is its 
density, and y its surface tension. (The error in using the same r for both 
the melt and the crystal is negligible.) If these are indeed the only forces 
that come into play at the growth interface, then a comparison of the two 
terms on the right-hand side of the equation will yield (assuming r * 0.015 cm, 
and R i 10 cm/hr) 

l ° 9 • 

2 z2 
rr pR 


However, this equation is really incomplete in that it does not take into 
account the resistance to fluid flow through the capillary duct. At R= 10 cm/hr, 
it is safe to assume viscous flow, and the Hagen-Poiseuille equation applies; 


where Ap is the pressure drop across a length, L, of duct due to resistance to 
flow by a fluid of viscosity, u (which has been estimated as equal to 1.1 cp 
for molten KC1 from parameters tabulated in Janz' Molten Salts Handbook , 
Academic Press, 1967). Setting L/h = 1/2, and comparing this resistive force 
to Che driving force (i.e., the hydrostatic force), we will obtain the r„tio, 


Our initij 

% • '• 0.0003. 

~r“ha g 

1 attempts at growing SC potassium chloride fiber were carried 
out in a :used-quartz crucible, shown schematically in Figure 5. We used 
h = 12 cm and L = 3 cm. It eventually became apparent that control of the 
growth would require a closer control of the temperature of the capillary duct. 
To achieve this we used an auxiliary heater for this duct. This current 
apparatus is shown in Figure 4. Using the first inverted Czochralski apparatus 
for KC1 fiber growth, we were unsuccessful in producing any fiber. With the 
better temperature control afforded by the second growth apparatus (Fig¬ 
ure 4), we will be able to grow alkali and thallous halide fibers in next 
year's program. 


We have developed a new method of growing SC fibers using the extrusion 
apparatus normally used to fabricate polycrystalline fibers. In this method 
only a small portion of the crystal is melted. Figure 6 shows the special die 
we needed for forming SC fiber. The solid KRS-5 or TIBr (materials we have 
tried to date) are placed in the extrusion die body as normally done in the 
extrusion of polycrystalline fiber and the material is forced through the 
special die. In the narrow fiber-forming region of the die the crystal is 
melted and an SC fiber results. This method, which we call pressure-fed 
crystal growth, has an advantage over the first approach by having a very small 
melt zone, with minimal exposure to Che contaminating outside environment. 
Furthermore, since the method used our fiber extruder, long fiber lengths are 
possible from che start. Finally, there is no pulling on the finished fiber. 

In other SC fiber techniques, the fiber is pulled by a take-up reel. This may 
incduce strain in the fiber which can lead to scattering losses. 

We have attempted a number of pressure-fed SC fiber growths using both 
quartz and stainless steel dies in the shape shown in . .ure 6. All quartz 
dies cracked. We now believe this is due primarily to attack of the quartz by 







i u-i Jb i 

Figure 4 . 

SC fiber crucible for growth and 
RAP purification of crystalline 





Figure 5. Our first apparatus for inverted 
Czochralski growth of SC fibers. 

TIOH, and not che pressure placed on che die in the quasi-extrusion process. 
The metal die also failed during our first attempt because the melt became too 
hot and adhered to the metal. Future work will use other die materials, such 
as sapphire, to prevent interaction between the halide and the die capillary. 





The optical evaluation of both fiber and bulk materials provides the 

foundation of our understanding of loss mechanisms in transparent materials. 

Our analysis of the optical attenuation in crystalline IR fibers has involved 

determining the total absorption coefficient, a^, as well as the attenuation 

from scattering, , and absorption,a , processes. In general, 

^ A 

\ = a s + a A ’ (6) 

so that evaluation of , and provides a thorough understanding of the 

loss mechanisms in fibers. 

Determination of the various attenuation coefficients can be accomplished 
in several different ways. In Table 1 we summarize the various experimental 
methods which are most commonly used. In our measurements we determined 
using both IR laser sources (insertion loss) and a small IR spectrometer of 
our own construction. The scattering losses were obtained using a scattering 
sphere. Finally, a was obtained from a and a using Equation (6). 

A i J 

Even more information is obtained about the individual loss mechanisms 
if we study a as a function of wavelength, A. In particular, we expect a 
priori that a^ will be a strong function of A (Rayleigh scattering). There¬ 
fore, we rewrite Equation (6) to incorporate an explicit wavelength 

(A) = (r^ + B) + a^A). (7) 

A represents Rayleigh scattering mechanisms, while B accounts for the 
frequency-independent scattering from particles whose size is large with 
respect to the wavelength. Equation (7) has been used^ quite successfully 
to account for the observed a^(A) behavior in silica-based fibers. Thus, by 





Experimental Technique 

* T 

Insertion-loss measurement with 

laser IR spectroscopy 


Scattering sphere 

Rayleigh-Brillouin scattering 

a 4 

Laser calorimetrv 


analogy, we anticipated that it would apply to our fibers at IR wavelengths. 
However, our experiments demonstrate that a,, (A) does not have as simple a wave¬ 
length dependence as that expressed in Equation (7). 


The insertion loss measurements were made at 10.6, 5.2, 3.8, and 2.8 _m 
using CO^. CO and DF/HF lasers. A diagram of the apparatus used at 10.6 urn 
is shown in Figure 7. A unique feature of this set-up is the blockage of the 
signal reflected from the input end of the fiber. Because of the high refrac¬ 
tive indices of our fibers, a large fraction (17% for KRS-5) is reflected at 
the front surface. This reflected energy can alter the laser output, creating 
a different incident intensity than that measured without the fiber in place. 

We blocked the reflected beam by using the quarter-wave plate and wire-grid 
polarizer, thus eliminating any enhancement of the incident energy. 

Spectral information is obtained by using a broadband source (Nernst 
glower) and a simple monochromator. In Figure 8 we have diagrammed the 
essential features of our IR fiber monochromator. The wavelength selection 
is performed by rotating an 0CL1 filter wheel (2.5 to 14.5 i.m) at the output 
end of the fiber. Data are first recorded with the fiber in place, then a 
background or reference spectrum is recorded. The sample and reference data 
are established and the transmission is arbitrarily set equal to 1.0 at 10.6 ..m. 


I U‘J4‘J 1 b 



1930°K NERNST 



f • 3.8 cm 
DIAM = 2.5 cm 


2.5 - 14.5 MICRONS 




Figure 8. IR fiber monochromator for 2.5 to 1^.5 „m 
spectral information. 


Suae coring measurements were maue usinc i small, _1.5 err. Ji^meter 

scattering sphere. Scatterin**; spheres have jeen ised f ?r fiber measurements 

o . , 5 

on silica-glass fibers and have been uuuptcu :jr us.- on LR Liners. Ihe 

rad Lat ion sear re red within the diffuse 40 Id—. - area -.piiere is I Icoted by a 
pryoeleetric detector t.see Figure 9). The laser beam is chopped and phase 
sensitive detection is used to eliminate background radiation. The measure¬ 
ment is made by carefully sliding the sphere along the fiber and recording 
the differential scattered light as a function of position of the sphere 
along the fiber. The sphere and detector are calibrated at the end of the 
measurement by drawing the output end of the fiber into the sphere and mea¬ 
suring the detector signal for a known-power output of the fiber. The fibers 
were studied at various IR laser wavelengths (.see Figure 9) to obtain our 
frequency dependent data. 


The relative transmission of two L-m long KRS-5 fibers is given in Fig¬ 
ure 10. The solid line shows a fiber with some impurity absorption, while 
the dashed line represents a fiber free of observable impurity effects. The 

absorption of 6 -L urn is due to H n 0, and the 9.6 am band is probably due to 

“ 9 

a C-0 bonded impurity. The latter band has been seen in alkali halides , 

but its exact source has never been identified. In the case of the fibers we 
feel that the impurity absorption is on the surface of the fiber, and the fiber 
is acting like an ATR (attenuated total reflection) plate to reveal surface con¬ 
taminants. The most striking feature of the data, however, is the decrease in 
transmission at the shorter wavelengths. Later, we will show that - • ”, 

This behavior is characteristic of the fiber; bulk KRS-5 exhibits no decrease in 
transmission over this wavelength range. The most reasonable explanation for the 
decrease in transmission is scattering. In the next section, we will discuss the 
scattering measurements in greater detail and offer an explanation for the 
observed spectral characteristics. The transmission of the fiber (solid line 
in Figure 10) at IR laser wavelengths, made by insertion loss techniques, is 
detailed in the tabie in Figure 10. When these points are plotted, along 
with the spectral data, we see that there is good agreement between the two 



Figure 9. Scattering measurements at 
IR Laser wavelength. 


Figure 10. Transmission of two KRS-5 fibers from 2.5 to L-+ urn 
Discrete joints are values of a^, at IR laser 



The calculation of a, has been discussed in Last '.'ear's ONR contract 

, o 

tenure. Ve summarise the basic relat ion.snins needed to set err. me .. in 
* *\ 

Figure LI. From the derining equation for the amount or iian.t, a L , scatterec 
from a length, dx, of fiber at a distance ,x, from the end of the fiber, 

dig = -I (x)cigdx, ( 3) 

from which we obtain a^. The actual value of depends on whether is 
obtained by integrating Equation (8) from x=0 to x= •' or whether a is obtained 
at a few selected points using Equation (8) directly. In the first case, an 
average value, , is obtained which includes the strong scattering contri¬ 
bution from the output end (x=L) of the fiber (see Figure L2). 

Differential light scattering intensity for two KRS-5 fibers is shown 
as a function of position along the fiber in Figure 12. In the top curve, 
we note the presence of a hot spot in the fiber 28 cm from the input end. 

The hot spot occurs where the fiber hung in the extruder overnight. At this 
point, the grains are large. Generally, we observe few hot spots in our 
present fibers, and the differential light scattering data appear as shown 
in the lower curve of Figure L2. Both sets of data, however, reveal exces¬ 
sive scattering from the output end of the fiber. This is characteristic of 
all our fibers and results from strong end reflections (17T) which couple 
light into higher order lossy modes. It is not possible to eliminate these 
end effects by terminating Che output in an index matching fluid because 
there is no suitable matching liquid with a refractive index as high as that 
of KRS-5 (n=2.37). Instead, we mathematically eliminate the end effect by 
calculating cig from Equation (8) at several points near the middle of the 

The differential light scattering data of Figure 12 is used to calculate 
a . For the lower curve in Figure 12 we have plotted the values, a (ealeu- 
lated using Equation (8))in Figure 13. As expected, the data show that a^ 
is fairly constant over most of the fiber's length (note the break in abcissa 
in Figure 13). The solid circles are data taken on a fiber with ends polished 
using 0.3 urn grit paper. This is our standard end finish for most fiber 


Figure 11. Calculation of 





Figure 12. Scattering Losses in KRS-5 fiber at 10.6 uni. 



Figure 13. 

Effect of output end finish on a,. 

measurements. From Figure L3 we see that the increase in at x=■ aepends on 

the end finish, and that the coarse grit, 9 urn polish leads to the greatest 

increase in a, at the output end. The coarse end finish leads to more 

scattering at the end surface and thus greater losses, as expected. The best 
polish reduces end scattering losses to the point where there is little change 
in ig over the length of the fiber. We also note from Equation 13 that the 
average value of 0.407 dB/m (includes output end scattering) for the 
solid-circle data is greater than the value of = 0.3 dB/m obtained from 
Equation (8) and data taken near the beginning of the. fiber. Therefore, we 
can regard a = 0.3 dB/m as more accurately representing the true scattering 
loss of the bulk fiber. 

The effect of end scattering is even more pronounced at shorter wave¬ 
lengths. Figure 14 shows data on the same fiber shown in Figure 13 with a 
standard end finish at two laser wavelengths. At 10.6 pm (solid line in 
Figure 14), is observed to be relatively constant (note expanded ordinate 
scale as compared to Figure 13). At 5.3 pm, however, the end scattering 
losses are severe, increasing almost an order of magnitude over the constant 
value of We expected this effect at shorter wavelengths because scat¬ 

tering increases as the wavelength decreases. 

It seems reasonable to attribute scattering losses to the polycrystalline 
nature of the waveguide. Specifically, the grain size would appear influen¬ 
tial in determining the optical properties of the fiber. To check this 
hypothesis, we extruded two KRS-5 fibers from the same starting billet with 
different grain sizes. Figure 15 shows the surfaces of the 3 pm and 34 pm 
average grain size fibers, along with the measured attenuation coefficients 
at 10.6 pm. We observed that even though varied slightly for each fiber, 
the ratio, was nearly identical for the two fibers. This means that 

the scattering contribution does not depend on grain size, at least within 
these limited test data. This is reasonable because these are cubic mate¬ 
rials, and therefore, the random orientation of the grains by themselves 
should not lead to scattering losses. It is possible, however, that impuri¬ 
ties decorating grain boundaries and inhomogeneous strain fields could lead 
to apparent grain size effects. At this point, we have not seen any evidence 
of these effects. 


. dB/m 

10 20 30 40 50 60 70 80 90 100 110 


Figure 14. Scattering Losses in KRS-5 fiber 
at two IR laser wavelengths. 




3 um 

34 fim 

a T (dB/m): 



a s (dB/m): 



a A (dB/m): 



a s /a r : 



Figure 15. Attenuation at 10.6 ym in KRS-5 
fiber with different grain size. 


We have collected all scattering measurements at 10. o -m to .study the 
fraction of the total light scattered out of the fiber. The histogram in 
Figure !o gives the number of fibers within a small . range. It is 

’ 3 r 

evident that most fibers studied scatter between -a) to o0' of the >t il light. 

Some fibers, however, scatter only a few percent. In these cases, the total 

fiber loss is not reduced through the reduction of a_, as might be expected. 

Instead, is as high or higher than fibers scattering 50% of the light. 

Finally, the fiber attenuation data at four laser wavelengths is given 

in Figure L7. In the two curves, a is plotted versus X * to help reveal 

the inverse power dependence expected for from Equation 7. We observe from 

the data that for the two fibers studied, a varies, to a good approximation, 

as \ “. However, a does not exhibit any simple wavelength dependence. The 
S -2 -4 

solid lines in the graph show that \ or \ (Rayleigh scattering) do not 
fit the data. We must conclude that the simple frequency dependence of 
Equation 7 for ot,, does not hold for our fibers. 


The following two observations summarize the primary results of our 
attenuation measurements on K.RS-5 fibers: 


• The total attenuation coefficient, a , varies as \ from 2.8 to 
10.6 urn. 

• The scattering attenuation coefficient, a,, is not given bv (—r + B) . 

b \ 4 

At the outset, we expected that scattering and absorptive losses could 

be treated independently, or that a = a + a,. Further, we assumed a, could 

T A S S 

be broken down into a frequency dependent term which describes Rayleigh-type 
scattering losses, and a frequency independent term representing scattering 
from defects much larger than the wavelength of light. This general form of 
analysis has been applied to silica fibers with excellent success. Our results 
indicate that the situation is not so simple for our fibers. This, we feel, 
is due to an interaction between scattered and absorbed light which prevents 
us from treating each contribution separately, as described by Equation 6 . 

One of the major sources of attenuation is the surface quality of the 
fiber. Because these fibers are extruded, the surface is not as smooth as the 

Figure 15. 

a s 

Attenuation l fl.tOfin in KRS-5 
fiber with difTerent grain size. 

Figure 16. 

Percentage of scattering 
loss at 10.6 urn. 

rrue-r.irmed surface or drawn glass fibers. The IR fiber surface is also 
susceptible to chemical contamination from moisture, dust, and hydrocarbons. 

We see the effects of such contamination in the spectral data of Figure 10. 
shun Iight is scattered in the fiber from chemical defects or residual strain 
:ields, some •>j the light will be scattered at angles near the critical angle 
i25 3 for KXS-5). This light will escape along the surface of the fiber and 
some will be absorbed by the surface. Thus, we will not collect this scattered 
Light in our scattering sphere, and hence, this light will not properly con¬ 
tribute to -t . This interaction between scattering and absorption leads to 

an effectively smaller a,, owing to the reabsorption of scattered light. We 

b -4 

notice chis trend in the curve in Figure 17. We expect a strong •. 

scattering dependence, but in fact observe a much weaker dependence of a, on 

X, especially at the shorter wavelengths. The data show that a does not 

_ ) 7 * 

even vary as \ 

The best correlation of attenuation and wavelength is observed (see 

-2 -? 

Figure 17) in the a data. Here, a-j varies as X . A X -dependence can be 
associated with surface absorption^; and -because scattering and absorption 

losses are interrelated, it may seem reasonable that a could be described 

-2 T 
by a composite X -type term. This is, however, speculative, and further 


analysis is necessary to properly explain the observed v \ “ behavior. 


The effect of strain on the optical properties of IR fibers is particu¬ 
larly important in the ductile fiber materials where the stress-induced bire¬ 
fringence can be significant. To determine the role of strain and its rela¬ 
tion to the attenuation in polycrystalline KRS-5 fibers, we have undertaken a 

series of experiments to measure the attenuation coefficients a , a , and a 

T S A 

as a function of applied uniaxial stress on a KRS-5 fiber. Our results 
indicate a strong increase in the attenuation as the fiber is loaded up to 
its ultimate strength. 


experimental set-up and results 

The measurements were carried ouc at 10. o „;n using the optical apparatus 

described in Seccion 4B. The Tiber was suspended vertically and leaded with 

known amouncs or water added to a vessel hanging t rom 1 hot tea clump. Ihe 

gauge length was approximately 75 cm or 3/4 of the fiber's length. As each 

load was added, x and a, were recorded. The measurements continued to the 
T S 

fracture point of the fiber. 

A typical stress-strain curve of a KRS-5 fiber is shown in Figure IS. 
This curve can be used to determine the strain in our measurements, although 
this was not done. We note from Figure L 8 the large region of plastic flow 
beyond the yield point. This is characteristic of these ductile polycrvstal- 
line fibers. In practice, the strain beyond yield is a function of the gauge 
length and loading time because this elongation is due partly to creep. 

The data in Figure 19 shows the strong increase in both and as 
the applied load is increased. Below the yield point, the fiber loss always 
returns to the zero load value when the load is removed (no hysteresis). 

Above the yield point, the plastically deformed fiber does not return to its 
original loss value; instead, an extra absorption proportional to the amount 
of plastic deformation is added. In Figure 20 we have replotted the data in 
Figure 19 after subtracting the attenuation at zero load. The excess loss in 
Figure 20 is seen to obey a simple power law relationship with respect to 
applied stress. 

The most likely source of the excess attenuation is stress-induced bire¬ 
fringence. Photomicrographs of KRS-5 fiber placed between crossed polarizers 

in both a relaxed and stressed state indicate considerable birefringence. 

This birefringence, of course, leads to scattering as a result of variations 
in the refractive index. M. Sparks^ has recently shown that within the pro¬ 

portional limit fiber losses resulting from inhomogeneous strains should vary 
as the square of the change in refractive index. An, or 













Figure 18. Mechanical strength of KRS-5 fiber. 



'U !' 

where * is the app Lied stress, and q is the stress-optic coef:icienr. 

1 .0 ’. 5 

results from Figure 20 show that i • and a : . This is in ;o-'j 

agreement with Spark's ca 1culations. Furture experiments will further verify 

tiiese results, especial lv v. , and a more complete taeorv will c t.'.e 


effects of voids introduced along grain boundaries ana disiocations. 


Our measurements of attenuation in stressed fibers have shown that strain 
birefringence is a significant Loss mechanism in our fibers. The very forma¬ 
tion of a poLycrystalline fiber by extrusion induces enormous strain in the 
fiber. This fact, coupled with the ready deformation of these ductile mate¬ 
rials, means that large strains will always be present in the as-extruded 
fiber. In tact, we now believe that, aLong with the surface quality, stress- 
induced variations in the refractive index (birefringence) may be a major 
source of fiber loss. In the future we will index profile the fibers to 
measure An directly. 

The removal of strain will be difficult in our polycrystalline fibers. 
Annealing above LOO^C will induce grain growth and weaken the waveguide. 

The solution seems to be single crystal fibers. SC fibers should be inher¬ 
ently less strained because the fabrication methods are less mechanically 
severe than extrusion. Further, SC fibers can be annealed to remove process- 
relation strain. 






The first two years of ONR sponsored research has indicated several new 
directions for future work. The most important is the method of fabrication. 
Since the primary objective remains to develop low-loss fibers, we have 
decided to devote the majority of our fiber fabrication to SC fiber growth. 

In addition to the two approaches outlined in Section 3 to prepare SC fibers, 
namely, pressure-fed growth using the extruder, and the inverted Czochralski 
technique, we propose to add a new and novel growth method. This method is a 
laser initiated, crucibleless growth technique. 

In this technique, laser beams are directed at the surface of the crystal 
to melt only a small portion of the surface of the fiber material. A seed rod 
(for example, a platinum wire) is dipped in the melt and a fiber is pulled 
vertically upward as in Czochralski growth. Figure 21 is a diagram of the 
concept. The advantages of this method over other SC techniques are: 

• The fiber surface is not formed by any guide (such as a quartz 
capillary). Instead, Che surface is free-formed. By analogy with 
glass drawing, we can expect a very smooth surface, and thus greatly 
reduced losses from surface roughness. 

• There is less contamination of the melt by the surrounding envrion- 
ment because the melt is contained by its own crucible (the sur¬ 
rounding solid fiber material) rather than by a conventional 
crucible. As the melt is depleted during fiber growth, new mate¬ 
rial is melted. 

• Better control of the fib'er diameter is expected as opposed to 
conventional Czochralski growth. 

Using laser heating, we may finely control the laser power to control 
the thermal profile. In our past experiments to seed a fiber from a KC1 
melt contained in a standard resistance-heated cruiCible we were unable to 
start the fiber because we could not control the temperature at the seed- 
liquid interface. With laser heating we have a much smaller melt region, and 
we expect not only easier seeding, but also good control of the fiber diameter 
by using an optical monitor to vary the laser power. 


Our first laser heating experiments will be carried out on the thallium 
halides. KRS-5 and TIBr absorb strongly in the blue-green region; thus, ve 
will use an Ar ion laser tor melting. In a simple test, we required only 1 ’■< 
of unfocused multiline Ar ion laser power to melt KRS-5. In the actual fiber 
growth experiments, we anticipate using an auxiliary resistance heater to bring 
the material close to the melting point (MP for KRS-5 is 410°C). The laser 
beams (minimum of two) would then be used to melt a small portion of the solid. 
We also have available a CO 2 laser for heating. A simple vertical pulling 
apparatus will be used to make the first SC fiber. 

A small effort will continue in the extrusion of thallium halide, fibers. 
These fibers allow us to study the basic loss mechanisms in IR fibers and 
eventually will serve as a useful comparison to the SC fibers made from the 
same materials. We also will be extruding some of our new purified RAP 
thallium halides. 


The optical evaluation of our fibers will center on measurement of a , 
a , and a at infrared wavelengths. In particular, we want to study the fre- 

^ A 

quency dependence of the absorption to learn more about the absorptive and 
scattering losses. 

We will measure the index of refraction variations in our fibers by index 
profiling, and by evaluating the waveguides' birefringence. In the latter 
case, we will use a laser polariscope to obtain the average value of the bire¬ 
fringence across the fiber. As discussed in Section 2, we believe An may be 
quite large for our fibers and for most of the ductile crystalline materials 
in general. 

We feel that a major portion of our present fiber losses are a result of 
surface roughness. Polycrystalline fibers are prone to a certain amount of 
surface roughness because of the finite grain size. We will study the effect 
of grain size on the fiber's attenuation by measuring, a_, a„, and a. in 
fibers with different grain sizes. Grain size variation in fibers will be 
accomplished by thermally inducing grain growth in fibers. The surface 
roughness of SC fibers is expected to be substantially less than in polvcrys- 
talline fibers. However, surface irregularities may result from the SC growth 


pr-'cess because .it the difficulty in controlling the fiber diameter. Similar 
ilV'.'rpt ion measurements will be made on the SC fibers, and the results will be 
'm:\ired to those for polycrystalline fibers. 

.'bully, Lurther limited studies will be made using our 7sbry-?er.'t 
Lnterferometer. Measurements of Rayieigh-Briliouin scattering will continue doped and pure KC1, and scattering in the thallium halides will be initiated. 
These light scattering studies have helped us understand the nature of the 
Ray Leigh scattering in bulk fiber materials. By making polarization-’ 
dependent light scattering measurements, and by evaluating the Landau-Placzek 
ratio, R , as a function of temperature, we are able, in principle, to 
differentiate between the various elastic light scattering mechanisms. The 
thrust into the thallium halides will supplement our integrating sphere mea¬ 
surements on K.RS-5 fibers, and thus help us understand the nature of our 
scattering Losses in fibers. 


Throughout our development of low-loss IR fibers, we have emphasized the 

metal halides because of their potentially small residual absorption. Lately, 

we have focused our attention on the thallium halides, even though these 

crystals have not exhibited the low-loss already demonstrated for bulk alkali 

halides, in particular,K.C1. We feel strongly that the thallium halides afford 
the best near-term solution for obtaining SC fibers with losses near 10 1 dB/km 
(year-end goal). One reason for this is the ease with which fibers can be 
made from these low melting point, ductile materials. Another is that these 
materials do not cleave, so that microcleavage cracks, which plagued extruded 
KC1 fibers, will not occur in SC thallium fibers. The problem with the thal¬ 
lium halides is that they have not been as thoroughly studied as the alkali 
halides, and thus the impurity levels are greater in these materials than, for 
example, KCl. 





l. J. A. Harrington, >1. Braunstein, B. Bobbs, and R. Braunstein, "Scattering i 

Losses in Single and PoLycrystalline Materials for Infrared Fiber Applies- : 

tion," presented at the Physics of Fiber Optics Meeting held at tile \ 

annual American Ceramic Society Meeting, 28-10 April, 1980, Chicago, • 

Illinois. Published in Adv. in Ceramics, 2, Phvsics of Fiber Optics , 

ed. by B. Bendow and S. Mitra, p. 94, Amer. Cer. Soc., Columbus, Ohio, i 

1981.' ! 



L. M. Sparks and L. DeShazer, "Theoretical Overview or Losses in Infrared 
Fibers," Infrared Fibers (0.8 - L2 _in), Proc. Soc. Photo-Opt. Instr. 
Eng. 26b , 3-9 (1981). 

2. J. A. Harrington, "Crystalline Infrared Fibers," ibid, 10—15. 

3. T. Miya, Y. Terunuma, T. Hosaka, and T. Miyashita, "Ultimate Low-Loss 
Single-Mode Fiber at 1.55 am" Elect. Lett., _L5_, No. 4, L06-L08 (1979). 

4. V. G. Artyushenko, et al., "Thallium Halide Crystals with Low Optical 
Losses," Sov. J. Quantum Electron., _10, 1181-1182 (1981). 

5. D. L. Philen and F. T. Stone, "Direct Measurement of Scattering Losses 
in Single-Mode and Multimode Optical Fibers," in Advances in Ceramics, 
Physics of Fiber Optics, 2_, ed. by B. Bendow and S. Mitra, Amer. Cer. 
Soc., Columbus, Ohio, 237-245 (1981). 

6. A. R. Tynes, et al., "Loss Mechanisms and Measurements in Clad Glass 
Fibers and Bulk Glass," J. Opt. Soc. Amer., 6_U 143-153 (1971). 

7. J. A. Harrington, "Low-Loss Fiber Waveguides," ONR Annual Report, 

1 August 1979 - 31 July 1980, Contract No. N00014-79-C-0691, 

Occ. 1980. 

8. J. A. Harrington, "infrared Fiber Optics for CO 2 Laser Applications," 
CO 2 Laser Devices and Applications Proc. of Soc. Photo-Opt. Instr. 
Eng., _227, 133-137 (1980). 

9. J. M. Rowe and J. A. Harrington, "Extrinsic Absorption in KC1 and KBr 
at CO 2 Laser Frequencies," J. Appl. Phys. 4_7, 4926-4928 (1976). 

10. H. E. Bennett, "Scattering Characteristics of Optical Materials,” 

Opt. Eng. _1_7, 480-488 (1978). 

11. M. Sparks, "Theoretical Studies of Low-Loss Optical Fibers," Final 
Report, Contract No. N00173-79-C-036I, August, 1980.