DTIC ADA1049431: Low-Loss Fiber Waveguides.

ADA104943

LOW-LOSS FIBER WAVEGUIDES

J. A. Harrington

Hughes Research Laboratories
3011 Malibu Canyon Road
Malibu, CA 90265

September 1981

N00014—79—C—0691
Annual Technical Report
1 August 1980 to 30 June 1981

Approved for public release; distribution unlimited.

DTIC

E'. r " '"P?

OCT 2 1981 ’

A

H

Prepared for

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

Si 10 2 097

UNCLASSIFIED _

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

I REPORT DOCUMENTATION PAGE

For r-* umber

6 1

2 GOVT ACCf

ADr Ao 1

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

LOW-LOSS FIBER '.w

I < *

AVECUIDES -

[7, AUTHO«C*J

)j J.Ay Harrington ^

9. PERFORMING ORGAN! 2 ATION NAME AND AOORESS

Hughes Research Laboratories

3011 Malibu Canyon Road

Malibu, California 90265 _

H. CONTROLLING OFFICE NAME AnO AOORESS

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,

/£

READ INSTRUCTIONS
BEFORE COMPLETING FORM

3. RECIPIENT'S CATALOG NUMBER

%. TYPE OF Rga^JM-■# ^ RTOtT c6v|WO

: Annual Technical .

/ 1 Aug »8j/~ 3/ Jun' 181,

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

T CONTRACT OR^TranT HUMBER/*; ~

jl NOOO14-79-C-0691

to PROGRAM EL EMEN T PROJECT. TASK
AREA 6 WORK UNIT NUMBERS

12. r O pbrt DATE _

13 ”

57

IS SECURITY CLASS tol this report)

/cKj

UNCLASSIFIED

i S« DECLASSIFICATION DOWNGRADING
SCHEDULE

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

16. SUPPLEMENTARY NOTES

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00 , jaCts 1473 Moition or i nov ss is obsolete

UNCLASSIFIED

SECURITY CL A SSI FIC AT ION OF THIS PAGE r Wh*n Ce a Entered)

UNCLASSIFIED _

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.

UNCLASSIFIED

5ECj PiTv Cl AS'lfiC ATtON Of T

S P*GE'»***n Omtm

PRECEDING PA01 BLANK “WOT

FIGURE

1

2

3

5

6

7

8

9

10

11

12

13

14

15

16

17

18
19

LIST OF ILLUSTRATIONS

UL

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

5

LIST OF ILLUSTRATION'S (.Conti inued *

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

SECTION 1

INTRODUCTION AND SUMMARY

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

_3

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

WAVELENGTH, (im

30.0

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,

_3

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

_2

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.

9

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.

10

SECTION I

FIBER MATERIALS PREPARATION

A. INTRODUCTION

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

B. APPROACHES TO BE TAKEN

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

11

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.

C. EXPERIMENTS AND RESULTS IN MATERIALS SELECTION AND PREPARATION

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.

2TlOH(s>

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)

and

*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.

13

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
congruency.

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.

14

r

i

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).

15

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.

Jr RECEDING PAGE BLANK-MOT Fli>JQ>

SECTION 3
FIBER FABRICATION

A. INTRODUCTION

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.

B. CZOCHRALSKI GROWTH OF SC FIBERS

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 ’
c

( 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

Trr

(2)

where r is the radius of the fiber.

19

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

R~

(3)

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

(4)

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;
i.e.,

(5)

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,

20

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.

C. PRESSURE-FED SC FIBER GROWTH

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

|

k

21

RAP GAS OUT

A

I

i u-i Jb i

Figure 4 .

SC fiber crucible for growth and
RAP purification of crystalline
fibers.

10304-7

RAP GAS OUT

GROWTH

ORIFICE

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.

25

i-nECEDlWG PAGE BUNK -NOT F1U»22

SECTION’ 4
FIBER EVALUATION

A. INTRODUCTION

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
dependence:

(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

27

TABLE L

COMMON METHODS USED TO MEASURE ATTENUATION
IN’ OPTICAL FIBERS

Attenuation

Experimental Technique

* T

Insertion-loss measurement with

laser IR spectroscopy

"s

Scattering sphere

Rayleigh-Brillouin scattering

a 4

Laser calorimetrv

A

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).

B. EXPERIMENTAL METHODS

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.

28

I U‘J4‘J 1 b

LIGHT SHIELD

0949-S

1930°K NERNST
GLOWER

HgCdTe

DETECTOR

Ge LENS
f • 3.8 cm
DIAM = 2.5 cm

KRS-5 FIBER -

' OCLI FILTER WHEEL
2.5 - 14.5 MICRONS
FWHM PASSBAND 3 A/ 100

6° FOV APERTURE
X CHOPPER AT 270 Hz

BIAS AND PREAMP PREAMP

.UN6

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

30

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.

C. TRANSMISSION SPECTRA OF KRS-5 FIBER

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
measurements.

.

31

Figure 9. Scattering measurements at
IR Laser wavelength.

32

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

D.

SCATTERING SPHERE MEASUREMENTS

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

, 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

34

Figure 11. Calculation of

35

SCATTERED LIGHT FLUX, mW SCATTERED LIGHT FLUX, mW

10949-9

POSITION ALONG FIBER, CM

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

36

10949-10

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
b

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.

38

. dB/m

10 20 30 40 50 60 70 80 90 100 110

POSITION ALONG FI8ER, CM

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

1000X

1000X

AVERAGE GRAIN SIZE:

3 um

34 fim

a T (dB/m):

0.79

1.14

a s (dB/m):

0.37

0.51

a A (dB/m):

0.41

0.63

a s /a r :

47%

45%

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

40

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.

E. INTERPRETATION OF SCATTERING SPHERE RESULTS

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

-2

• 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

_9

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

F. STRESS-INDUCED BIREFRINGENCE IN FIBERS

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.

44

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

45

r

MECHANICAL PROPERTIES

9000

8000

7000

6000

5000

4000

3000

2000

1000

Figure 18. Mechanical strength of KRS-5 fiber.

Ub

STRESS, PSI

'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 nsi.cr t.'.e

s

effects of voids introduced along grain boundaries ana disiocations.

H. INTERPRETATION OF ATTENUATION IN STRESSED FIBERS

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.

49

trtECEDIWG PADS BLAMK-NOT FILM®

SECTION' 5

FUTURE PLANS .AND RECOMMENDATIONS

A. FIBER FABRICATION

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.

51

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.

B. OPTICAL EVALUATION AND ANALYSIS

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

53

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
..in 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.

C. MATERIAL CONSIDERATIONS

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
4

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.

54

PRESENTATION’S AND PAPERS

,1

'i

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.' !

PRECEDING PAGE BLANK -WOT FlwE)

REFERENCES

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.

57