Content uploaded by John Ballato
Author content
All content in this area was uploaded by John Ballato
Content may be subject to copyright.
On the fabrication of all-glass optical fibers from crystals
J. Ballato,1,a兲T. Hawkins,1P. Foy,1B. Kokuoz,1R. Stolen,1C. McMillen,1,2
M. Daw,1,3 Z. Su,1,3 T. M. Tritt,1,3 M. Dubinskii,4J. Zhang,4T. Sanamyan,4and
M. J. Matthewson5
1Center for Optical Materials Science and Engineering Technologies (COMSET), School of Materials
Science and Engineering, Clemson University, Clemson, South Carolina 29634, USA
2Department of Chemistry, Clemson University, Clemson, South Carolina 29634, USA
3Department of Physics and Astronomy, Clemson University, Clemson, South Carolina 29634, USA
4US Army Research Laboratory, AMSRD-ARL-SE-EO 2800 Powder Mill Road, Adelphi, Maryland 20783,
USA
5Department of Materials Science and Engineering, Rutgers, The State University of New Jersey,
607 Taylor Road, Piscataway, New Jersey 08854, USA
共Received 10 August 2008; accepted 9 January 2009; published online 13 March 2009兲
The highly nonequilibrium conditions under which optical fibers conventionally are drawn afford
considerable, yet underappreciated, opportunities to realize fibers comprised of novel materials or
materials that themselves cannot be directly fabricated into fiber form using commercial scalable
methods. Presented here is an in-depth analysis of the physical, compositional, and selected optical
properties of silica-clad erbium-doped yttrium aluminosilicate glass optical fibers derived from
undoped, 0.25, and 50 wt %Er3+-doped yttrium aluminum garnet 共YA G兲crystals. The
YAG-derived fibers were found to be noncrystalline as evidenced by x-ray diffraction and
corroborated by spectroscopic measurements. Elemental analysis across the core/clad interface
strongly suggests that diffusion plays a large role in this amorphization. Despite the noncrystalline
nature of the fibers, they do exhibit acceptable low losses 共⬃0.15–0.2 dB/m兲for many
applications, broad-band emissions in the near-infrared, and enhanced thermal conductivity along
their length while maintaining equivalent mechanical strength with respect to conventional silica
optical fibers. Further, considerably higher rare-earth doping levels are realized than can be achieved
by conventional solution or vapor-phase doping schemes. A discussion of opportunities for such
approaches to nontraditional fiber materials is presented. © 2009 American Institute of Physics.
关DOI: 10.1063/1.3080135兴
I. INTRODUCTION
The vast majority of optical fiber is made from combi-
nations of core and clad compositions that are variations on
the same theme. In other words, the cladding and core are
essentially the same material but with a minor compositional
perturbation to adjust the refractive indices to permit light
confinement and guidance. Such similarities between core
and clad materials yield additional benefits such as amena-
bility to codrawing and well-matched thermal expansion co-
efficients. However, using the same family of core and clad
materials severely restricts the choice of materials from
which fibers can be made. Over the years, there have been
occasional successes in fabricating optical fibers comprised
of considerably dissimilar materials including all-glass
fibers,1–3crystalline fibers,4,5and glass-clad crystal core
fibers.6,7However, the trend within the optical fiber commu-
nity has been more focused on developing fiber geometries
of ever-greater complexity than on new combinations of ma-
terials that could enable additional and value-added optical
and optoelectronic functions.
This paper provides a thorough material characterization
of optical fibers prepared from a crystalline starting material
for the core; specifically Er3+-doped yttrium aluminum gar-
net 共YAG兲. The choice of Er:YAG was due to recent work
using conventional fiber draw techniques to make glass-clad
fibers with cores purportedly containing 共chromium doped兲
YAG.8,9In those reports, a single crystal of 共Cr doped兲YAG
was sleeved inside a silica glass tube and drawn into fiber
using a conventional fiber draw approach. The core material
is molten at the temperature where the glass capillary tube
共and optical cladding兲softens.
In the present work, silica glass-clad fibers of varying
diameters were drawn from preforms containing undoped
and erbium-doped YAG single crystals in the core. These
preforms were drawn into fiber at temperatures above the
melting point of the YAG core crystal. X-ray diffraction was
used as the primary tool to examine crystallinity in the re-
sultant fiber core. Optical absorption and emission measure-
ments of rare-earth-doped fiber samples were made to pro-
vide a secondary indication of crystallinity since spectral
linewidths broaden in amorphous hosts. Diffusion from clad-
ding to core, which is likely at the high draw temperatures,
was investigated using energy dispersive x ray 共EDX兲for
elemental analysis. Lastly, since the ultimate goal is the prac-
tical use of the fiber, both the strength and the thermal con-
ductivity of the drawn fibers—properties important for high-
power amplifier applications—were studied.
Although our initial hope was that the fibers would pos-
sess some degree of YAG crystallinity, the resultant cores
a兲Electronic mai: jballat@clemson.edu.
JOURNAL OF APPLIED PHYSICS 105, 053110 共2009兲
0021-8979/2009/105共5兲/053110/9/$25.00 © 2009 American Institute of Physics105, 053110-1
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
were found to be a fully amorphous yttrium-aluminosilicate
glass. However, the fibers did exhibit beneficial physical and
optical properties that make them of potential interest as op-
tical amplifiers and lasers. The results are insightful into how
this molten core process can be a valuable processing ap-
proach to realizing optical fibers with cores comprising ma-
terials that are either highly dissimilar from the cladding or
would not themselves be amendable to the direct fabrication
into fiber form.
II. EXPERIMENTAL PROCEDURES
Commercial grade single crystals of YAG 共undoped
0.25 wt %Er doped and 50 wt %Er doped兲for use in the
core of the preforms were provided by Northrop Grumman
Synoptics 共Charlotte, NC兲. The YAG boule was core-drilled
共Ceramare, Piscataway, NJ兲into rods measuring 3 mm in
diameter and approximately 40 mm in length. Optical quality
silica glass 共F300兲for use as the cladding was purchased
from Heraeus Tenevo 共Buford, GA兲.
A series of the silica glass capillary tubes was drawn to
differing sizes and sleeved within one another to yield a pre-
form having an outer diameter of 51 mm and an inner diam-
eter of 3 mm. This preform was then fire polished using an
internal procedure. The YAG core rod then was placed in the
center of the capillary cladding glass preform.
For each of the YAG compositions 共undoped, 0.25%,
and 50% Er:YAG兲from which the preform was prepared a
series of large fibers 共canes兲was drawn to varying diameters
in order to quantify diffusion effects on the core composi-
tion. Following the collection of these canes, each YAG-
derived fiber was drawn down to 125
m, coated with a
single layer of UV curable polymer 共DSM Desolite 3471–3-
14兲, and about 1 km was collected on a spool for a variety of
subsequent measurements. All fibers were drawn at approxi-
mately 2025 °C using the Heathway fiber draw tower at
Clemson University.
A. X-ray diffraction
Powder x-ray diffraction patterns were collected using a
Scintag XDS 2000-2 powder diffractometer with Cu K
␣
ra-
diation 共=1.5418 Å兲and a solid-state Ge detector. Diffrac-
tion patterns were collected from 5° to 65° in 2-theta in 0.03°
steps at rates of 2–10 s/step. Samples were analyzed both as
intact fibers and as pulverized powders. Further, in some
cases both the core and the silica cladding were analyzed and
in selected cases, the silica cladding was partially removed
by etching in dilute HF acid.
B. Microscopic and elemental analysis
Prior to examination the fiber ends were mechanically
polished with 600 grit silicon carbide 共SiC兲bonded paper.
Uncoated samples were investigated using a Hitachi 3400 N
scanning electron microscope. EDX spectroscopy was per-
formed to examine the distribution of major elements 共Si, O,
Al, Y, and Er兲. Elemental compositions were measured at
several locations along a line traversing the core. The micro-
scope was operated at 20 or 30 kV and 10 mm working
distance under variable pressure. EDX typically has a spatial
resolution of about 1
m.
C. Fiber attenuation and spectroscopic measurements
Baseline fiber attenuation measurements were performed
using a Photon Kinetic 2500 optical fiber analysis system. In
this case the attenuation wavelength dependence was mea-
sured with 10 nm wavelength increments. The 5 and 10 m
cutback lengths were used for the fiber drawn using the un-
doped YAG crystal and 0.25 wt %erbium-doped YAG crys-
tal, respectively.
In order to characterize the fibers spectroscopically, with
the purpose of determining whether they have crystalline or
amorphous structure, significantly higher 共than 10 nm兲spec-
tral resolution is required. For this purpose fiber absorption
was measured by using the standard “cutback” technique,
whereby the light from the fiber-coupled “white light” source
共Yokogawa AQ4305 white light source兲transmitted through
the fiber under investigation was spectrally analyzed by the
AQ6370 optical spectrum analyzer 共OSA兲. Spectrally dis-
persed, with the 1 nm OSA resolution, light intensities 共log
scale兲separately measured on the samples of 2.688 m and
0.653 m long were subtracted and then divided by the differ-
ential length 共2.035 m cutback兲in order to infer the true
dB/m core absorption value.
The 4I13/2−4I15/2Er3+ fluorescence spectra were mea-
sured using an Acton 2500i monocromator, thermoelectri-
cally cooled InGaAs detector, and lock-in amplifier. The
monochromator slits were chosen to provide a resolution bet-
ter than 0.35 nm. Fluorescence was always collected perpen-
dicular to the fiber axis. This collection mode, even though it
does not offer a higher signal-to-noise ratio, does avoid spec-
tral distortions due to fluorescence emission reabsorption of
waveguided light. It also helps us to avoid systematic errors
in lifetime measurements associated with fluorescence radia-
tion trapping. All obtained spectra were corrected for the
spectral response of the entire system. For spectral scanning
fluorescence was excited by electrically chopped 976 nm
fiber-coupled GaAs diode laser. The same setup was utilized
for fluorescence lifetime measurements, except a short-pulse
共⬃10 ns兲excitation source, frequency-doubled Nd:YAG la-
ser at 532 nm, was used in this case. Time resolution for the
lifetime measurements was 3
s. The lifetime was measured
by a Tektronix digital oscilloscope with the averaging used
for signal-to-noise improvement.
D. Thermal conductivity measurements
The geometry and size of optical fibers is not conducive
to conventional four-probe steady state thermal conductivity
measurement methods. Therefore, the parallel thermal con-
ductance system was used to measure the thermal conduc-
tance along the length of the samples at temperatures ranging
from 10 K to room temperature.10 Samples were cooled with
a cryocooler to the desired temperature set by the user at an
adjustable cooling rate then warmed up while the system
begins to take the data with preset temperature intervals. For
053110-2 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
the optical fibers discussed here, the cooling rate was 1
K/min and the samples were mounted on the stage with Ag
paste 共DuPont 4924 N兲.
E. Fiber strength measurements
It is well known that any crystallization in glasses can
dramatically decrease the strength since it leads to sharp fea-
tures that are effective stress concentrators. Fused silica op-
tical fiber must be strong in order to survive the stresses it
experiences during component fabrication and during ser-
vice. Typical fiber is very strong 共⬃5 GPa兲but occasional
weak defects are removed by proof testing at typically 350 or
700 MPa. Any crystallization in the fiber could degrade the
strength below the proof stress and so would make the fiber
unusable in conventional component designs. The mechani-
cal properties of the fibers described here have therefore
been characterized using industry standard test methods us-
ing both two-point flexure11 and uniaxial tension.12 Fused
silica exhibits fatigue or stress corrosion cracking due to en-
vironmental moisture. For this reason all measurements have
been made in a controlled environment at 23.0⫾0.2 ° C and
50%⫾3%humidity.
III. RESULTS AND DISCUSSION
Figure 1provides an optical micrograph of the conven-
tional 共125
m兲fiber after fabrication. As can be seen, the
core is central, circular, and guides light. The concentric
cladding tubes that had been layered to provide the desired
core/clad ratio were well fused.
Earlier iterations led to noncircular cores which most
likely were due to ovalities in the starting cladding tubes.
Since the core is molten at the draw temperatures, it would
tend to flow and take on the shape of its “container.” Light
from a He–Ne laser 共632.8 nm兲propagated through a several
meter length of 125
m diameter fiber exhibited a far-field
pattern consistent with a low-moded fiber 共4⬍V⬍6兲with a
numerical aperture 共NA兲of approximately 0.38 and ⌬n
⬃0.049. Given that the index of the core precursor YAG
crystal exceeds 1.8 the measured NA clearly indicates a con-
siderable amount of diffusion from the silica cladding into
the 共molten兲core during draw. This is quantified in greater
detail later in this paper.
In order to make an estimation of the spectral attenua-
tion, a cutback measurement was made using a Photon Ki-
netics 2500 unit on selected lengths of 125
m undoped and
0.25 wt %doped fibers. The results are shown in Fig. 2.A
5 m length of the undoped fiber exhibited a minimum loss
共background attenuation兲of about 0.33 dB/m at a wave-
length of 1300 nm 共where there is no OH or Er3+ absorption
lines兲. A 10 m length of the lightly doped fiber exhibited a
lower loss: about 0.14 dB/m at a wavelength of 1300 nm.
These loss values are roughly 100 times lower than those
achieved in Ref. 8for a Cr-doped analog. The absorption due
to the erbium dopant also is observed and is discussed in
more detail below. In both cases, the peak at approximately
1385 nm is due to residual OH in the as-grown crystal from
which the core rod was cut.
A higher resolution absorption spectrum for the fiber
made from the 0.25 wt %Er:YAG is shown in Fig. 3. The
figure is indicative of about 4.5 dB/m Er3+ peak core absorp-
tion 共above the ⬃0.15–0.2 dB/m background transmission
loss兲at 1532.5 nm for this low concentration fiber. The ob-
served difference between the relatively low peak absorption
FIG. 1. 共Color online兲Optical 共a兲and electron microscopic 共b兲images of
different representative fibers drawn from a YAG starting core crystal. The
optical micrograph 共a兲is of a fiber with a 125
m outer diameter. The
electron micrograph 共b兲is of a fiber drawn to a diameter such that the core
was about 250
m. The points marked across the core region 共b兲indicate
where elemental analysis was performed 共see Figs. 6–8兲.
FIG. 2. Spectral attenuation of the undoped and lightly doped 共0.25 wt %
Er:YAG in the preform兲“YAG-core” fiber. Absorption band peaked at 1385
nm is due to OH groups in the as-grown crystal.
053110-3 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
values 共⬃1.6 dB/m兲in the Photon Kinetics 2500 attenua-
tion measurement 共Fig. 2兲and true absorption measurements
共Fig. 3兲can be attributed to the highly under-resolved nature
of the former. This difference is, as expected, the most
prominent for the sharpest and narrowest 1532.5 nm absorp-
tion peak. Better resolution of the cutback measurement al-
lows observation of the true spectral width of the major Er
absorption peak in the fiber as well as any spectroscopic fine
structure on the long-wavelength side of the absorption
curve.
This is a quite reasonable level of transparency for such
a nonoptimized initial study. In order to further characterize
the optical properties of these fibers, the luminescence spec-
trum and luminescence lifetime were evaluated.
Figure 4provides the absorption spectra of the YAG-
derived fiber compared to that of single crystalline Er:YAG
as well as two erbium-doped fiber amplifier 共EDFA兲fibers.
The broad and “smooth” Er3+ absorption features in the
drawn fiber versus those of the original Er:YAG in the pre-
form implies a noncrystalline environment in the core of the
drawn fiber. However, the observed linewidth is somewhat
narrower than that of Er3+ doped into typical telecom-type
aluminosilicate and germanosilicate glass hosts. These fea-
tures can be seen in Fig. 4, which provides the normalized
absorption spectra from the fiber derived from the Er:YAG
containing preform compared to those from an Er:YAG
single crystal and two commercial EDFA fibers.
Figure 5provides the normalized 4I13/2→4I15/2Er3+
fluorescence and absorption spectra 关Fig. 5共a兲兴and the mea-
sured fluorescence intensity as a function of time 关Fig. 5共b兲兴.
Like Fig. 4, the spectra in Fig. 5共a兲are indicative of an
amorphous structure though the linewidth of the central peak
is narrower than found in conventional EDFA glasses. Addi-
tionally, the ⬃10 ms first e-folding lifetime from the fiber is
somewhat longer than the typical lifetime for Er-doped
silica-based glasses. In order to better understand if there
were any crystallographic considerations at play, x-ray dif-
fraction and elemental analysis was performed.
Thicker fibers, due to the larger amount of core material
relative to a thinner fiber, were analyzed using powder x-ray
diffraction. In all cases, the cores were found to be x-ray
FIG. 3. Core absorption spectrum of the fiber drawn from the preform
containing a 0.25 wt %Er:YAG crystal; 1 nm spectral resolution. Absorp-
tion peak at 1385 nm is also attributed to residual water absorption from the
as-grown crystal.
FIG. 4. Normalized absorption spectrum of the optical fiber derived from
the Er:YAG containing preform compared to that for a YAG single crystal
and two commercial EDFA fibers.
FIG. 5. Spectroscopic properties of the fiber drawn using the 0.25 wt %
Er:YAG crystal in the core: 共a兲fluorescence corrected for spectral response
of the spectrometer overlaid with the absorption of the fiber; 共b兲fluores-
cence kinetics of the 0.25 wt %Er3+-doped fiber measured at 1532.5 nm
with the ⬃10 ns pulse excitation at 532 nm.
053110-4 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
amorphous. This is in contrast to a previous report in the
literature for a fiber fabricated using a fiber draw process,9
which cites results on laser-heated pedestal grown fibers,13
even though the formation processes and thermal histories
are entirely different. The authors note the formation of crys-
talline
␥
-Al2O3having an anomalously large unit cell param-
eter derived from selected area electron diffraction 共0.856
nm versus the accepted value of 0.790 nm兲, which they at-
tribute to the Cr doping. Such a large lattice expansion seems
unlikely based on the low dopant concentrations reported
共0.5 mol %兲and the crystallite analyzed is more likely ex-
tremely Cr heavy or an alternative phase. The work does not
provide sufficient detail to understand the chemical identity
of the observed crystals or how they might result from the
composition or fiber fabrication process.
In the present study, a manifestation of the amorphous
nature of the core is found in the overlay of the absorption
spectra of Er-doped YAG single crystal and Er-doped YAG-
derived fibers 共Fig. 3兲. Single-crystalline Er:YAG has very
sharp, well defined, and mostly fully resolved, even at room
temperature, absorption peaks which belong to different
inter-Stark transitions between the 4I15/2and 4I13/2manifolds
of Er3+ ions in YAG. Meanwhile, Er-doped YAG core fiber
exhibits a smooth absorption contour with much less struc-
ture, and inter-Stark transitions mostly manifest themselves
by little “shoulders” in the wings of the major absorption
peak at 1532.5 nm, which is similar to Er-doped glass ab-
sorptions.
In order to further understand this, elemental mapping
using the electron microscope was performed. The series of
squares noted in Fig. 1共b兲marks the locations where elemen-
tal analysis using EDX spectroscopy was performed. EDX
provides the relative amounts of each element present in the
region evaluated. Figures 6–8provide the Si, O, Y, Al, and
Er elemental profiles across the core and part of the glass
cladding from the YAG 共Y3Al5O12兲-derived fibers at each
erbium doping level and a variety of fiber diameters. The
specific fiber diameters at which the elemental analysis was
performed were chosen arbitrarily.
There are a few interesting points to be made from Figs.
6–8. First, the presence of silicon 共in the form of silica兲in
the core implies diffusion from the cladding into the core.
This is reasonable to expect since one has a melt 共the YAG
core兲in contact with a soft glass 共the silica cladding兲at el-
evated temperatures 共⬃2050 °C兲. Diffusion is a thermally
activated process and so the high processing temperatures for
silica fibers further facilitate interdiffusion. Second, the
smaller the core, the more silicon 共i.e., silica兲diffuses in
from the cladding. This also is expected: the smaller the
fiber, the shorter the diffusion length, the greater the concen-
tration of the solute 共diffusing species from cladding兲in the
solvent 共core兲. Last, the compositional profiles are more
steplike than expected for a diffusion-related process. This is
likely due to the fiber fabrication process which promotes
homogenization of the molten core as it transitions from the
bulk preform through the neck down region and into the fiber
where the composition is quenched into constancy.
An additional effect of silica diffusion from the cladding
is that the core composition is “diluted.” For example, the
fiber drawn from a preform containing 50 wt %Er:YAG ex-
hibited about 20–25 wt %Er3+ in the final fiber, depending
on the core size. While this dilution obviously reduces the
amount of dopant in the fiber relative to that in the preform,
FIG. 6. Elemental profiles 共relative elemental composition as a function of
position across the fiber兲for the undoped fiber. The figures 共a兲,共b兲, and 共c兲
were drawn to core sizes of 230, 191, and 57
m, respectively.
053110-5 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
nonetheless there is a considerably higher doping level than
could be achieved under the more conventional methods em-
ployed to add rare earths into a fiber, e.g., solution doping
and vapor-phase doping. Accordingly, this molten core ap-
FIG. 7. Elemental profiles 共relative elemental composition as a function of
position across the fiber兲for the for drawn from a preform containing
0.25 wt %Er:YAG. The figures 共a兲,共b兲, and 共c兲were drawn to core sizes of
369, 248, and 24
m, respectively. Note that the erbium concentration is
too small to be measurement in this particular experiment.
FIG. 8. Elemental profiles 共relative elemental composition as a function of
position across the fiber兲for drawn from a preform containing 50 wt %
Er:YAG. The figures 共a兲,共b兲, and 共c兲were drawn to core sizes of 600, 500,
and 260
m, respectively.
053110-6 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
proach could be used to yield doped fibers with unusually
high rare-earth doping levels, even higher potentially than
was achieved in Ref. 2.
Figure 9is a compilation of the data from Figs. 6–8and
shows the concentration of silicon 共as silica兲at the center of
each core of diameter noted. The lines are provided as a
guide for the eyes and to show trends for the doped and
undoped samples. Needless to say, the diffusivity of silica in
each composition of molten YAG is slightly different and
likely accounts for the measurable differences in concentra-
tion. If one was to take an average concentration at each
diameter, then the silicon concentration decreases monotoni-
cally with increasing core size, hence diffusion length, as
would be expected. A more in-depth evaluation of the inter-
diffusion between the molten YAG core and the glass clad-
ding is underway and will be reported separately. The pur-
pose here is to provide an initial quantification of the
diffusion effects on core composition such that postulations
could be made on the crystallography, or lack thereof, in
these resultant fibers.
Even in the case of the largest core fiber, where the
diffusion distance is longest, the center of the core possesses
approximately 10 wt %silicon 共as silica兲from the cladding.
The amorphous nature of the core then is likely due to some
combination of two effects: 共1兲the presence of the silica in
the YAG leads to a more stable yttrium aluminosilicate glass
and 共2兲the significant quench rate 共⬃2000 °C/s兲of the melt
to the solid during the fiber fabrication process prevents crys-
tallization of the core. It is more likely that the composition
共effect 1兲plays the dominant role since equally high quench
rates have recently been shown to still permit highly crystal-
line fibers utilizing similar draw techniques.7
Based on this recent success in fabricating crystalline
core fibers via fiber draw techniques,7an original hope for
this study was that silica-clad crystalline YAG core fibers
could be realized. This would be of value to higher power
amplifiers and lasers where thermal management and heat
dissipation are critical. Even though the fibers were found to
be amorphous, the thermal conductivity 共Fig. 10兲was mea-
sured as a function of temperature alonga4cmlength of the
large core fiber. A bare glass silica fiber of equivalent diam-
eter was drawn for direct comparison.
At low temperatures, the fiber derived from the YAG
core preform exhibits equal thermal conductivity to that of a
conventional silica-based optical fiber. This is not particu-
larly surprising because of the small core-to-clad ratio; hence
the fiber is principally silica. However, at temperature ap-
proaching room temperature, there is a deviation and the
fiber derived from YAG exhibits a thermal conductivity
along its length that is ⬃20%higher than a conventional
silica fiber. The reason for this enhanced thermal conductiv-
ity is unclear to date but is reproducible and is a subject for
continued study.
In order to fully characterize these fibers, their mechani-
cal properties also were investigated. Figure 11 shows the
results of two-point bend strength measurements on the un-
doped YAG fiber. The results are graphed on a Weibull prob-
ability plot ln ln关1/共1−Pf兲兴 versus ln
f, where Pfis the
FIG. 9. Silicon concentration 共indicating SiO2content兲in the center of the
fiber core taken from the data of Figs. 6–8. The lines are guides for the eyes.
FIG. 10. Thermal conductivity as a function of temperature measured along
the fiber length.
FIG. 11. Weibull probability plot of the strength of undoped fiber measured
in two-point bending at four different faceplate speeds.
053110-7 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
cumulative probability of failure and
fis the failure stress.
This plot is widely used to visualize strength distributions. In
this case all four distributions are narrow as is typical of
fused silica optical fiber. The strength is measured at four
faceplate speeds in order to characterize the stress corrosion
cracking behavior which occurs in the humid environment.
The stress corrosion susceptibility parameter, n, calculated
from these data is 20.9 with a 95% confidence interval of
20.2–21.6. Again, this behavior is very typical of fused silica
optical fiber.
However, two-point bending is not sensitive to the pres-
ence of large defects in the core of the fiber, such as might be
produced by crystallization or residual bubbles from the core
melt, since the stress at any point in the bent fiber is propor-
tional to the distance from the neutral axis, i.e., the distance
from the center of the core. For a ⬃6
m diameter core and
a 125
m diameter fiber, the stress at the core/clad interface
is only ⬃0.05%of the maximum stress at the surface of the
fiber. For this reason the strength has been characterized us-
ing uniaxial tension using a 0.5 m gauge length in which the
entire volume of the fiber is subjected to the same stress.
Figure 12 shows the results for the strength of fiber derived
from the 50% Er:YAG measured at a stress rate of 30 MPa/s.
Except for one weaker specimen at ⬃3 GPa, the strength
distribution is narrow with values typical of fused silica. The
broken ends of the weaker specimen were recovered and
examined in the optical microscope. The two fracture sur-
faces were not mirror images of each other meaning the fiber
shattered locally at failure, as is typical for strengths above
⬃1 GPa. Fractographic analysis was therefore unable to de-
termine the position of the flaw causing failure. This weaker
specimen could have been caused by a defect at the surface
of the glass, which can occasionally occur, or at the core/clad
interface. In either case the impact on reliability is not
great—the strength of 3 GPa is much higher than typical
proof stresses 共typically 0.36–0.72 GPa兲so the strength lim-
iting defect would not cause failure during either proof test-
ing or subsequent service. Further, if the flaw is indeed
caused by the presence of a defect in the core, it is less
severe than a flaw of an equivalent size at the surface be-
cause it is not exposed to environmental moisture and so will
not fatigue due to stress corrosion.
While no severe defects were detected in the ⬃15 m of
fiber tested here, there is a possibility of very large flaws if
much longer lengths of fiber were examined. Unlike the
other properties of the fiber described here, which are aver-
age properties along the length of the fiber, strength depends
on the extreme values of flaw sizes in the test length and so
is well known to be length dependent. However, these early
results on this fiber do conclusively show that the rapid
quenching of a molten core during draw does not necessarily
degrade the strength to the point where the fiber is unusable,
at least on the length scale of tens of meters.
Residual issues and considerations
The purpose of this paper was to provide a more in-
depth analysis of the materials and physical properties of
glass-clad optical fibers derived from a crystalline 共YAG兲
core containing preform. As with any initial study, there are a
number of future efforts worth pursuing. More specifically,
the following studies and developments would enable such
fibers to be of greater value to the community:
共i兲Determine through time, temperature, and transforma-
tion thermal studies if it is possible to develop YAG
crystallinity 共or polycrystallinity兲of the core. If poly-
crystallinity is achievable, then develop approaches to
minimizing the grain size in order to optimize fiber
transparency. Here previous work on glass ceramic
optical fibers might be useful.14
共ii兲There is a need for better understanding of the advan-
tages and disadvantages of diffusion and other ways
to control it 共e.g., draw temperature, draw time, inter-
mediate cladding layers as diffusion barriers, etc.兲.
Given the inevitability of diffusion, how might it be
minimized or, at least, insightfully used to end with a
core composition that provides desirable optical prop-
erties. The work in Ref. 7proves that high degrees of
crystallinity can be obtained in molten core-derived
optical fibers even given moderate degrees of diffu-
sion.
共iii兲Determine the application-specific balance between
larger core size 共i.e., less diffusion兲and fiber flexibil-
ity.
共iv兲Develop approaches for tailoring the fiber NA for
single mode operation as well as achieving more com-
plex fiber designs.
共v兲Continue to investigate the broadened range of mate-
rials and compositions permitted by such nonequilib-
rium molten core approaches to optical fibers.
IV. CONCLUSIONS
An in-depth analysis of optical fibers derived from a
preform containing crystalline YAG as the starting core ma-
terial. Conventional fiber draw parameters led to a core that
was amorphous, which likely results from in diffusion of
silica from the cladding. Regardless, the resultant fibers ex-
FIG. 12. Weibull probability plot of the strength of fiber drawn from the
50% Er:YAG containing preform measured in uniaxial tension at a stress
rate of 30 MPa/s.
053110-8 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
hibited background loss figures of ⬃0.1–0.3 dB/m, room
temperature thermal conductivity along its length that was
roughly 20% higher than conventional optical fiber, and me-
chanical properties equivalent to standard optical fiber. The
spectroscopic properties of Er3+ doped into the precursor
YAG corroborated the amorphous nature of the core material
though the linewidth is somewhat narrower and fluorescence
lifetime is somewhat longer than that in conventional silicate
glasses. Efforts should continue to further evaluate the influ-
ence of processing conditions on core crystal composition
共i.e., diffusion兲and the potential development of a crystalline
or polycrystalline core.
ACKNOWLEDGMENTS
This work was supported in part by the Joint Technology
Office 共JTO兲through their High Energy Laser Multidisci-
plinary Research Initiative 共HEL-MRI兲programs at Clemson
University: “High Power Fiber Lasers” under an ARL
supplement to USARO under Contract No. W911NF-05-1-
0517 and “Eye-Safe Polycrystalline Lasers” under US-
AFOSR under Contract No. FA9550-07-1-0566. Addition-
ally, the authors wish to thank Northrop Grumman-Synoptics
共Charlotte, NC兲for providing at no cost the undoped and
doped YAG samples, Dr. Bob Rice of Northrop Grumman
Space Technology 共Redondo Beach, CA兲for insightful com-
ments, and Dr. Larry McCandlish of Ceramare 共Piscataway,
NJ兲for core-drilling the YAG samples.
1E. Snitzer and R. Tumminelli, Opt. Lett. 14, 757 共1989兲.
2J. Ballato and E. Snitzer, Appl. Opt. 34, 6848 共1995兲.
3N. Goel, R. Stolen, S. Morgan, J.-K. Kim, D. Kominsky, and G. Pickrell,
Opt. Lett. 31, 438 共2006兲.
4M. Fejer, J. Nightingale, G. Magel, and R. Byer, Rev. Sci. Instrum. 55,
1791 共1984兲.
5G. A. Magel, D. H. Jundt, M. M. Fejer, and R. L. Byer, Proc. SPIE 61B,
89 共1986兲.
6C. Lo, K. Huang, J. Chen, S. Tu, and S. Huang, Opt. Lett. 29,439共2004兲.
7J. Ballato, T. Hawkins, P. Foy, R. Stolen, B. Kokuoz, M. Ellison, C.
McMillen, J. Reppert, A. M. Rao, M. Daw, S. Sharma, R. Shori, O.
Stafsudd, R. R. Rice, and D. R. Powers, Opt. Express 16, 18675 共2008兲.
8Y. Huang, Y. Lu, J. Chen, Y. Hsu, Y. Huang, S. Huang, and W. Cheng,
Opt. Express 14, 8492 共2006兲.
9Y.-C. Huang, J.-S. Wang, Y.-K. Lu, W.-K. Liu, K.-Y. Huang, S.-L. Huang,
and W.-H. Cheng, Opt. Express 15, 14382 共2007兲.
10B. M. Zawilski, R. T. Littleton, and T. M. Tritt, Rev. Sci. Instrum. 72,1770
共2001兲.
11TIA/EIA TSB 62–13, Telecommunications Industry Association, ITM,
Washington, DC, 2000.
12TIA/EIA-455–28-C, Telecommunications Industry Association, FOTP,
Washington, DC, 1999.
13C.-Y. Lo, K.-Y. Huang, J.-C. Chen, C.-Y. Chuang, C.-C. Lai, S.-L. Huang,
Y.-S. Lin, and P. S. Yeh, Opt. Lett. 30, 129 共2005兲.
14B. N. Samson, P. A. Tick, and N. F. Borrelli, Opt. Lett. 26,145共2001兲.
053110-9 Ballato et al. J. Appl. Phys. 105, 053110 共2009兲
Downloaded 16 Mar 2009 to 130.127.56.12. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp