You are on page 1of 15

Stimulated thermal Rayleigh scattering in

optical fibers
Liang Dong*
The Holcombe Department of Electrical and Computer Engineering, Center for Optical Materials Science and
Technology, Clemson University, 91 Technology Drive, Anderson, South Carolina 29625, USA
*
dong4@clemson.edu

Abstract: Recently, mode instability was observed in optical fiber lasers at


high powers, severely limiting power scaling for single-mode outputs.
Some progress has been made towards understanding the underlying
physics. A thorough understanding of the effect is critical for continued
progress of this very important technology area. Mode instability in optical
fibers is, in fact, a manifestation of stimulated thermal Rayleigh scattering.
In this work, a quasi-closed-form solution for the nonlinear coupling
coefficient is found for stimulated thermal Rayleigh scattering in optical
fibers. The results help to significantly improve understanding of mode
instability.
©2013 Optical Society of America
OCIS codes: (060.2320) Fiber optics amplifiers and oscillators; (060.4370) Nonlinear optics;
(060.3510) Lasers, fiber.

References and links


1. T. Eidam, S. Hanf, E. Seise, T. V. Andersen, T. Gabler, C. Wirth, T. Schreiber, J. Limpert, and A. Tünnermann,
“Femtosecond fiber CPA system emitting 830 W average output power,” Opt. Lett. 35(2), 94–96 (2010).
2. F. Stutzki, F. Jansen, T. Eidam, A. Steinmetz, C. Jauregui, J. Limpert, and A. Tünnermann, “High average power
large-pitch fiber amplifier with robust single-mode operation,” Opt. Lett. 36(5), 689–691 (2011).
3. T. Eidam, C. Wirth, C. Jauregui, F. Stutzki, F. Jansen, H. J. Otto, O. Schmidt, T. Schreiber, J. Limpert, and A.
Tünnermann, “Experimental observations of the threshold-like onset of mode instabilities in high power fiber
laser amplifiers,” Opt. Express 19(14), 13218–13224 (2011).
4. F. Stutzki, H. J. Otto, F. Jansen, C. Gaida, C. Jauregui, J. Limpert, and A. Tünnermann, “High-speed modal
decomposition of mode instabilities in high-power fiber lasers,” Opt. Lett. 36(23), 4572–4574 (2011).
5. C. Jauregui, T. Eidam, H. J. Otto, F. Stutzki, F. Jansen, J. Limpert, and A. Tünnermann, “Temperature-induced
index gratings and their impact on mode instabilities in high-power fiber laser systems,” Opt. Express 20(1),
440–451 (2012).
6. C. Jauregui, T. Eidam, J. Limpert, and A. Tünnermann, “The impact of modal interference on the beam quality
of high-power fiber amplifiers,” Opt. Express 19(4), 3258–3271 (2011).
7. K. R. Hansen, T. T. Alkeskjold, J. Broeng, and J. Lægsgaard, “Thermo-optical effects in high-power ytterbium-
doped fiber amplifiers,” Opt. Express 19(24), 23965–23980 (2011).
8. A. V. Smith and J. J. Smith, “Mode instability in high power fiber amplifiers,” Opt. Express 19(11), 10180–
10192 (2011).
9. B. Ward, C. Robin, and I. Dajani, “Origin of thermal modal instabilities in large mode area fiber amplifiers,”
Opt. Express 20(10), 11407–11422 (2012).
10. K. R. Hansen, T. T. Alkeskjold, J. Broeng, and J. Lægsgaard, “Thermally induced mode coupling in rare-earth
doped fiber amplifiers,” Opt. Lett. 37(12), 2382–2384 (2012).
11. C. W. Cho, N. D. Foltz, D. H. Rank, and T. A. Wiggins, “Stimulated Rayleigh scattering,” Phys. Rev. Lett.
18(4), 107–109 (1967).
12. R. M. Herman and M. A. Gray, “Theoretical prediction of the stimulated thermal Rayleigh scattering in liquid,”
Phys. Rev. Lett. 19(15), 824–828 (1967).
13. D. H. Rank, C. W. Cho, N. D. Foltz, and T. A. Wiggins, “Stimulated thermal Rayleigh scattering,” Phys. Rev.
Lett. 19(15), 828–830 (1967).
14. I. L. Fabelinskii and V. S. Starunov, “Some studies of the spectra of thermal and stimulated molecular scattering
of light,” Appl. Opt. 6(11), 1793–1804 (1967).
15. C. W. Cho, N. D. Foltz, D. H. Rank, and T. A. Wiggins, “Stimulated thermal Rayleigh scattering,” Phys. Rev.
175(1), 271–274 (1968).
16. W. Rother, D. Pohl, and W. Kaiser, “Time and frequency dependence of stimulated thermal Rayleigh
scattering,” Phys. Rev. Lett. 22(18), 915–918 (1969).

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2642
17. N. Bloembergen, W. H. Lowdermilk, M. Matsuoka, and C. S. Wong, “Theory of stimulated concentration
scattering,” Phys. Rev. A 3(1), 404–412 (1971).
18. L. M. Peterson and T. A. Wiggins, “Forward stimulated thermal Rayleigh scattering,” J. Opt. Soc. Am. 63(1),
13–16 (1973).
19. R. C. Desai, M. D. Levenson, and J. A. Barker, “Forced Rayleigh scattering: thermal and acoustic effects in
phase-conjugate,” Phys. Rev. A 27(4), 1968–1976 (1983).
20. H. J. Hoffman, “Thermally induced degenerate four-wave mixing,” IEEE J. Quantum Electron. 22(4), 552–562
(1986).
21. H. J. Hoffman, “Thermally induced phase conjugation by transient real-time holography: a review,” J. Opt. Soc.
Am. B 3(2), 253–273 (1986).
22. R. W. Boyd, “Nonlinear Optics,” third edition, Elsevier, 2008.
23. A. W. Snyder and J. D. Love, “Optical Waveguide Theory,” Chapman and Hall, 1983.
24. M. K. Davis, M. J. F. Digonnet, and R. H. Pantell, “Thermal effects in doped fibers,” J. Lightwave Technol.
16(6), 1013–1023 (1998).
25. C. Jauregui, T. Eidam, H. J. Otto, F. Stutzki, F. Jansen, J. Limpert, and A. Tünnermann, “Physical origin of
mode instabilities in high-power fiber laser systems,” Opt. Express 20(12), 12912–12925 (2012).
26. A. V. Smith and J. J. Smith, “Influence of pump and seed modulation on the mode instability thresholds of fiber
amplifiers,” Opt. Express 20(22), 24545–24558 (2012).

1. Introduction
Fiber lasers are rapidly becoming very an important tool for machining, welding and
materials processing in advanced manufacturing, among many other important applications in
medical, science, sensing, and defense. Recently, mode instability has been observed in
optical fiber lasers at high powers, severely limiting power scaling for single-mode outputs
[1–3]. A fiber amplifier would operate normally at low output powers. Once the output is
over a certain threshold, the output mode pattern starts to oscillate between the lowest two
modes, with no apparent loss of total output power. With a further increase of pump power,
more chaotic mode behavior starts to develop. The process can be repeated by turning pump
off and re-start again. Many experimental [3, 4] and theoretical [5–10] studies have been
conducted. Temperature grating as result of quantum defect heating and mode interference
were identified fairly early on to be responsible for the phase-matched mode coupling [5–7].
A critical piece of the puzzle, the need for a traveling wave and, therefore, a frequency shift in
the two coupled modes, for the stimulated nonlinear process to occur, is first identified in [8].
A stationary grating can cause mode coupling as in the case of long period gratings. If such a
grating is created by mode inference between a fundamental mode and a higher order mode,
the phase of the field in the higher order mode coupled from the fundamental mode will be
π/2 different from the original field in the higher order mode which created the grating in the
first place, prohibiting a stimulated process to take place. A more recent numerical study
based on a dynamic model [9] was used to study transient process of mode instability.
Numerical models used in [8, 9] are very powerful for taking into considerations of a large
variety of effects and very useful for precise quantitative analysis. Many aspects of the
underlying physics are, however, often lost in the numerical process. A simple steady-state
physics model was recently developed in [10]. The difficulty of solving heat transportation
equations in optical fibers prevents it from finding more closed-form solutions for the
nonlinear mode coupling. The simple model does include all key elements of physics and
provides significant insights into the problem.
In this work, a quasi-closed-form solution for nonlinear coupling coefficient is obtained
for the first time for mode instability. One key enabling breakthrough is the finding of a
quasi-closed-form steady-state solution for heat transportation equation. The simple quasi-
analytical model developed as a consequence provides a great deal of insights in mode
instability process.
2. Some historic background
Mode instability in optical fibers is a manifestation of Stimulated Thermal Rayleigh
Scattering (STRS) first observed in the sixties in absorbing liquids with giant-pulse Ruby

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2643
lasers [11–21]. Interference between pump and scattered lights leads to a traveling
temperature wave via absorptive heating, which in turn stimulates further power coupling to
the scattered light [11]. The basic physics was quickly understood [12, 13], including the
nature of the traveling wave and frequency shift in the scattered light. In the following few
years, the effect was thoroughly studied and well understood [14–18]. In early eighties, there
was a resurgence of interests in STRS for phase conjugation and four-wave mixing [19–21].
The similarity between stimulated Rayleigh scattering (SRS) and stimulated Brillouin
scattering (SBS) was noted in [22]. In fact, a simultaneous theoretical treatment of the two
effects together is given in [22] considering both thermal and electrostrictive effects. The
analysis in [22] provides most of the basic physics for understanding mode instability in
optical fibers, which will be referred to as STRS in optical fibers in this paper thereafter. The
simple analytical solutions in [22] are, however, obtained by ignoring any transverse
dependence of electric field and thermal gradient. This may be appropriate for bulk media,
but is not valid for optical waveguides. The analysis in [22] is also more appropriate for
liquid, for which it was developed. As it will be noted later, there are some very important
differences between liquid and solid media.
In optical fibers, co-propagating waveguide modes at slightly different optical frequencies
form moving interference pattern along a fiber amplifier. Quantum defect heating as a result
of the amplification process generates a traveling temperature wave which leads to further
nonlinear coupling between the two waveguide modes via thermal-optics effect. Stimulated
scattering can take place akin to SBS involving a traveling perturbation. It is important to
point out that the traveling temperature wave is not from thermal diffusion along the fiber.
The period of the interference pattern is typically much larger than fiber diameter and thermal
diffusion dominates in the radial direction. The traveling temperature wave is effectively
damped at the rate of thermal diffusion in the radial direction. It is worth noting that the
acoustic wave in SBS is even more severely damped in optical fibers. The recently observed
mode instability is essentially STRS. One variation is the means of heating, i.e. quantum
defect heating in mode instability versus absorptive heating in STRS observed in the sixties.
The second variation is optical waves involved, i.e. co-propagating optical waveguide modes
in mode instability versus two interfering beams in early STRS. .
The analysis in [22] provides quantitatively thermal and electrostrictive contributions for
both SRS and SBS. It is concluded that electrostrictive effect dominates in SBS and thermal
effect dominates in SRS. This conclusion mainly comes from the frequency response of the
two effects. SBS requires higher frequency traveling wave for phase-matching the counter-
propagating optical waves, where the much faster but weaker electrostrictive effect
dominates. SRS requires a much lower frequency traveling wave for phase-matching co-
propagating optical waves, where a stronger traveling temperature wave dominates in
absorbing or amplifying media. Consequently, electrostrictive effect is ignored in this
analysis and the effort is focused on STRS.
Gain saturation effect is not considered in this work for simplicity. Gain saturation can
take place in fiber amplifiers. Since not all signal power is used to generate heat through
quantum defect heating in this case, a higher STRS threshold is expected than that predicated
by this work. This effect can be potentially dealt with by finding the appropriate amplitude for
the traveling optical intensity wave to use in the analysis. Quantum defect heating is assumed
to be the only source of heat and single frequency is assumed for the optical waves. Thermal
lensing is ignored. Constant temperature at the circular fiber surface is assumed, simulating
the case of actively cooled fiber amplifier. The simulations were carried out for step-index
fiber and LP11 modes for simplicity. The basic model can be modified for other fibers. It can
also handle other higher order modes. Threshold condition is developed for amplifiers with
constant gain along its length in this work. This is again for simplicity only. It can be
developed for other amplifier designs.

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2644
3. Fields in optical fibers
Almost all fibers used in high power fiber lasers are operating in the weakly guided regime,
where fields can be well approximated by linearly polarized (LP) modes. The conventional
LP mode representations will be used in this analysis, i.e. LPmn, where m and n is azimuthal
and radial mode numbers respectively. The electric field of LPmn mode can be written as:
f mn ( r )
emn ( r, φ ) =
cos ( mφ ) (1)
2nε 0 cN mn
The reason for the normalization used here will become clear later. In step index fibers,
J m (U mn r / a )
f mn ( r ) = a≥r≥0 (2)
J m (U mn )

K m (Wmn r / a )
f mn ( r ) = r>a (3)
K m (Wmn )

where a is core diameter. Jm represents Bessel functions of the first kind and Km represents
modified Bessel function of the second kind. U and W are defined as in [23] and determined
by the eigenvalue equation. Nmn is normalization factor.
∞ 2π ∞
N 0 n =  rdr  f 02n ( r )d φ = 2π  f 02n ( r ) rdr for m = 0 (4)
0 0 0

∞ 2π ∞
N mn =  rdr  f mn2 ( r ) cos 2 ( mφ ) d φ = π  f mn2 ( r ) rdr for m > 0 (5)
0 0 0
Electric field in a multimode optical fiber can be written as a summation of all
propagating modes.
∞ ∞
E ( r , φ , z ) =  Emn ( r , φ , z ) e (
i β mn z −ωmn t )
(6)
m = 0 n =1

Where

Emn ( r, φ , z ) = Pmn ( z )emn ( r, φ )

and where Pmn is optical power in LPmn mode. It is now clear that the normalization used
previously enables the possibility of writing field with only modal power in the amplitude.
Using Eq. (6), intensity in the optical fiber can be found. Ignoring interfering terms which do
not include fundamental modes,
∞ ∞
f mn2 ( r )
I ≈  Pmn ( z ) cos 2 ( mφ ) +
m = 0 n =1 N mn
(7)
∞ ∞
f 01 ( r ) f mn ( r )
2 P01 ( z ) Pmn ( z ) cos ( mφ ) cos ( β mn − β 01 ) z − (ωmn − ω01 ) t 
m =1 n =1 N 01 N mn
It is easy to see the optical intensity contains traveling waves with wave vector ± (βmn-β01)
and angular frequency ± (ωmn-ω01).
4. Steady-state solution for traveling temperature waves
Unlike traveling acoustic wave involved in SBS, traveling temperature wave cannot exist due
to heat diffusion. In typical optical fiber amplifiers, the period of the traveling wave for phase

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2645
matching co-propagating optical modes is in the orders of few hundred μm to few cm and is
typically much larger than fiber diameter. Heat diffusion, therefore, dominates in the radial
direction. The traveling optical intensity wave as a result of interference between modes will
deposit heat in the fiber core. The deposited heat will diffuse mostly in the radial direction
and eventually lost to the media outside the fiber at a rate determined by heat transportation
equation. This process is continuously driven by the traveling optical intensity wave as a
result of modal inference. The result is a seemingly traveling temperature wave severely
damped by thermal diffusion. At any fixed point in the fiber core, heat is deposited
periodically in time. It is worth noting that the traveling acoustic wave in SBS in optical
fibers is even more severely damped due to high material absorption at GHz regime and there
is a strong analogy between SBS and STRS.
It is worth noting that, in case where thermal diffusion dominates axially, amplitude of the
traveling temperature wave diminishes due to the equalization of temperature along the fiber
over time. STRS in optical fiber, therefore, is stronger when the period of the traveling wave
is significantly larger than fiber diameter. Unfortunately for mode instability, this is typically
the case in fiber amplifiers.
In this section, a steady-state solution for the traveling temperature wave as a result of the
traveling intensity wave and quantum heating is found. It is assumed that temperature is held
constant at the circular outer fiber boundary. This is similar to the case where the fiber is
actively cooled at the surface. Spatial temperature modes of the fiber are found first by
solving the non-driven heat transportation equation. The solution of the heat transportation
equation driven by the traveling intensity wave is then found by summing all the spatial
temperature modes. This method is used for the first time to solve heat transportation
equation to this author’s knowledge.
A traveling temperature wave driven by the traveling intensity wave as result of mode
interference are, therefore, sought.

T ( r, φ , z , t ) = T0 ( r, φ , z ) + T ( r , φ ) e (
i qz −Ωt )
(8)

where the first term on the right is time-independent temperature change; the second term is
from traveling temperature wave; and
Ω = ωmn − ω01 (9a)

q = β mn − β 01 (9b)

The temperature distribution is governed by heat transportation equation. When retaining only
these term involving ei(qz-Ωt), it becomes

∂T ( r, φ ) e ( )
i qz −Ωt

− κ∇2T ( r, φ ) e (
i qz −Ωt )
ρC =
∂t
(10)
1  λs  P01 ( z ) Pmn ( z )
N 01 N mn ( ) 01 ( ) mn ( )
imφ i ( qz −Ωt )
 − 1 g r f r f r e e
2  λ p 
The right hand represents the heat deposited by the interfering modes through quantum
defect heating. Gain coefficient profile g(r) is assumed to be only depend on r. This is
typically the case in fiber amplifiers. λs and λp are signal and pump wavelength respectively.
Note that q and Ω are dependent on mode numbers m and n (not expressed explicitly for
clarity). Note also the ¼ reduction comparing to Eq. (7) for converting the two cosines to
exponential. Also, ρ is density; C is specific heat; and κ is thermal conductivity. Using
T ( r, φ ) = Tm ( r ) eimφ (11)

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2646
Equation (10) can be transformed to
∂ 2Tm ( r ) 1 ∂Tm ( r )  Ωρ C m2 
+ + i − q 2 − 2  Tm ( r ) =
∂r r ∂r  κ
2
r 
(12)
1  λs  P ( z ) Pmn ( z )
N 01 N mn ( ) 01 ( ) mn ( )
−  − 1 01 g r f r f r
2κ  λp 
Firstly, solutions to the non-driven equation will be sought,

∂ 2Tml ( r ) 1 ∂Tml ( r )  2 Ωρ C m 2 
+ +  qml + i − 2  Tml ( r ) = 0 (13)
∂r 2
r ∂r  κ r 

where qml is the eigenvalue and l is the spatial temperature mode number of the heat
transportation equation. The solution is Bessel function of the first kind.
 Ωρ C 
Tml ( r ) = J m  r qml
2
+i  (14)
 κ 
Considering the case of active cooling where the fiber surface r = b is held at a constant
temperature, the boundary condition dictates
Tml ( b ) = 0 (15)

Using the approximation for lower order roots of Bessel functions of the first kind,
π2 Ωρ C
( 4l − 1 + 2m )
2
2
qml ≈ −i (16)
16b2 κ
Substitute Eq. (16) into Eq. (14), solution of each mode can be written as
π 
Tml ( r ) ≈ J m  ( 4l − 1 + 2m ) r  (17)
 4b 
The complete solution can be written as a summation of all spatial modes:

Tm ( r ) =  al Tml ( r ) (18)
l =1

It is possible to prove the spatial temperature modes are orthogonal,


b

 T ( r )T ( r ) rdr = 0
0
ml1 ml2 whenl1 ≠ l2 (19)

π π
b
b2  
 Tml1 ( r )Tml2 ( r ) rdr = −
0
2
J m −1  ( 4l − 1 + 2m ) J m +1  ( 4l − 1 + 2m ) 
4  4  (20)
whenl1 = l2 = l
Using the orthogonality equations, al can be determined by substituting Eq. (18) into Eq.
(12), multiplying both side of the resulting equation by Tml and integrating over the fiber cross
section, the amplitude terms in Eq. (18) can be obtained,

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2647
1  λs  P ( z ) Pmn ( z )
al =  − 1 01
2κ N 01 N mn
 λp 
b

 g ( r ) f ( r ) f ( r ) T ( r ) rdr
0
01 mn ml
(21)

 2 π2 Ωρ C  2
b

 q + 16b2 ( 4l − 1 + 2m ) − i κ   Tml ( r ) rdr


2

 0
5. Coupled nonlinear equations
Taking a similar approach to [22], the nonlinear polarization which is phase-matched for the
mode coupling is found first. Index change due to a change in temperature can be written as,

n = kT T ( r, φ ) e ( = kT Tm ( r ) eimφ e (
i qz −Ωt ) i qz −Ωt )
(22)

KT = dn/dT is thermal optics coefficient, i.e. index change per K. Change in permittivity can
be written as,

ε ≈ 2nn = 2nkT T ( r, φ ) ei ( qz −Ωt ) = 2nkT Tm ( r ) eimφ ei ( qz −Ωt ) (23)


We can now write the relevant nonlinear polarization terms:

p NL = ε 0ε E = 2nε 0 kT Tm* ( r ) eimφ e ( 01 01 ) Emn + 2nε 0 kT Tm ( r ) eimφ e ( mn mn ) E01 (24)


i β z −ω t i β z −ω t

Substituting Eq. (24) and Eq. (6) into nonlinear wave equation, where c is the speed of light,
2
n ∂ E 1 ∂ 2 p NL
2
∇2 E −   = (25)
 c  ∂t
2
ε 0 c 2 ∂t 2
Collecting relevant terms, the nonlinear coupled mode equations can be obtained:
∂P01 ( z )
= − g 01real ( χ mn
*
) P01 ( z ) Pmn ( z ) + g01P01 ( z ) (26a)
∂z
∂Pmn ( z )
= g01real ( χ mn ) P01 ( z ) Pmn ( z ) + ( g mn − α mn ) Pmn ( z ) (26b)
∂z
Loss of fundamental mode is ignored, but a loss αmn is introduced to account for the often
higher loss of the higher order mode. The gains of the two modes g01 and gmn (m>0) are given
as,

 g ( r ) f ( r ) rdr
2
∞ 2π 01
1
g01 =  rdr  g ( r ) f 012 ( r ) d φ = 0

(27a)
N 01 0 0
 f 012 ( r ) rdr
0

 g ( r ) f ( r ) rdr
2
∞ 2π mn
1
rdr  g ( r ) f mn2 ( r ) cos 2 ( mφ ) d φ =
N mn 0
g mn = 0

(27b)
 f ( r ) rdr
0 2
mn
0

The nonlinear coupling coefficient can then be obtained.

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2648
 2Ω 
2 −i

Γ ml  2π kkT  λs 
g01 χ mn = g 01 ( χ mn + i χ mn ) = g01 
r i  χ =  − 1
l =1  2Ω 
2 mnl
ρC  λp 
1+  
 Γ ml 
(28)
 2Ω  d b
2  − i  g ( r ) f ( r ) f ( r ) T ( r ) rdr 0 f 01 ( r ) f mn ( r ) Tml ( r ) rdr
 Γ ml  0
∞ 01 mn ml

 2 b
 2Ω  Tml2 ( r ) rdr

l =1
1+   N 01 N mn Γ ml
 Γ ml  0

where k is vacuum wave vector and d is the radius of the active region. The damping factor is
given by,
2κ  2 π 2 2
Γ ml =q + ( 4l − 1 + 2 m )  (29)
ρ C  16b2 
The nonlinear gain coefficient in Eq. (28) consists of four parts. The first part at the front
consists of mainly materials constants and vacuum wave number. The second part in the
bracket describes the quantum defect heating. The third part immediately after the summation
sign describes the frequency dependence. The last part describes the two overlap integrals
involved in the three-wave interaction. The first integral in the numerator describes the
process where the interference of the two modes deposits heat in the active region to create
the traveling temperature wave. The second integral in the numerator describes the process
where the traveling temperature wave leads to further nonlinear coupling of the two modes.
The terms in the denominator in the last part are all normalization factors except the damping
factor Γml.
The third part describing the frequency dependence deserves some more comments. Phase
of the nonlinear coupling coefficient is determined by this part alone. The real part of the gain
coefficient determines the nonlinear coupling strength. The frequency dependence of the
nonlinear coupling can be clearly seen by examining the real part of this third term, which
achieves a maximum of 1 when Ω = Γml/2. It is easy to see that the there is no coupling at Ω =
0 rad/s.
Equation (26) can be re-written for clarity.

∂P01N ( z )
= − g01 χ mn
r
e( gmn −αmn ) z P01N ( z ) PmnN ( z ) (30a)
∂z

∂PmnN ( z )
= g01 χ mn
r
e g01z P01N ( z ) PmnN ( z ) (30b)
∂z
where the normalized powers are
P01N ( z ) = P01 ( z ) e − g01z (31a)

PmnN ( z ) = Pmn ( z ) e
− ( gmn −α mn ) z
(31b)

6. Characteristics of the nonlinear gain coefficient


The physical constants used in this work are summarized in Table 1, unless stated otherwise.
Refractive index of silica is taken as 1.45. The data are obtained from [24]. There are some
variations in thermal optics coefficient used in the literature. For example, KT = 1.2 × 10−5
K−1 was used in [5].

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2649
The nonlinear gain coefficient in Eq. (28) was first bench marked for the cases in
simulated in [10] for step index fibers. Only V value is give in [10], there is some uncertainty
in what value was used for KT. The bench marking results are summarized in Table 2. It can
be seen that the error is within 3% for all the cases if KT = 1.2 × 10−5 K−1 was used.
Fiber design can now be studied to understand their impact on the nonlinear coupling
coefficient χ. It should be noted that the χ in Eq. (28) is the total nonlinear gain normalized
against amplifier gain coefficient for the fundamental mode g01. A fiber amplifier is usually
designed for a target fundamental mode gain. Varying some fiber parameters such doping
radius d can impact target gain for the fundamental mode. This definition of χ, unlike that in
[10], is normalized against this change, so that fiber designs can be studied while keeping the
target gain for the fundamental mode constant.
Table 1. Coefficients of Silica Used in This Work
Coefficient Symbol Value Unit
Thermal optics coefficient KT 1.1 × 10−5 K−1
3
Density ρ 2.2 × 10 kg/m3
Specific heat C 741 J/kg/K
Thermal conductivity κ 1.38 w/m/K
Thermal diffusivity D = κ/ρC 8.46 × 10−7 m2/s

Table 2. Benchmarking of Maximum Nonlinear Coupling Coefficient to These in [10]


Hansen [10] Equation (27) Equation (27)
KT = 1.2 × 10−5 K−1 KT = 1.1 × 10−5 K−1
a = d = 10μm, V = 3 0.068W−1 @ 5.2KHz 0.066W−1@ 5.2KHz 0.060W−1@ 5.2KHz
a = d = 20μm, V = 3 0.068W1 @ 1.4KHz 0.068W−1 @ 1.4KHz 0.063W−1 @ 1.4KHz
a = d = 40μm, V = 3 0.068W1 @ 0.4KHz 0.066W−1 @ 0.4KHz 0.060W−1 @ 0.4KHz
a = d = 20μm, V = 2.5 0.039W−1 @ 1.1KHz 0.039W−1@ 1.1KHz 0.036W−1@ 1.1KHz
a = d = 20μm, V = 3.5 0.079W−1 @ 1.6KHz 0.080W−1@ 1.6KHz 0.073W−1@ 1.6KHz
a = 20μm, d = 15μm, V = 3 0.043W−1 @ 1.6KHz 0.044W−1@ 1.6KHz 0.040W−1@ 1.6KHz
a = 20μm, d = 10μm, V = 3 0.015W−1 @ 2.0KHz 0.015W−1@ 2.0KHz 0.013W−1@ 2.0KHz

The first fiber in the study is a typical LMA 30/400 fiber, commonly found in many fiber
amplifiers. This fiber has a NA of 0.06, core diameter 2a = 30μm, and cladding diameter 2b =
400μm. Core is uniformly doped, i.e. d = a. Pump wavelength of 976nm and signal
wavelength of 1060nm will be used throughput this paper. Nonlinear coupling between the
fundamental mode and the first higher order mode LP11 is studied thereafter in this paper.
Other higher order modes are expected to have lower nonlinear gain and therefore higher
STRS thresholds. The solutions for the spatial temperature modes are found first. The
nonlinear coupling coefficients amplitude χmnl defined in Eq. (28) and damping factor Γml at
each l are plotted in Fig. 1(a). The largest χmnl is achieved at the mode number l where the
deposited heat overlaps best with the spatial temperature mode. The damping factor Γml
increases with mode number l. This is due to a reduced physical dimension of the spatial
temperature mode features at large l. The simulated χ is shown in Fig. 1(b). The real part of χ
is responsible for the nonlinear coupling. It becomes 0 at Ω/2π = 0 kHz and reaches a
maximum at Ω/2π = ~3.4kHz. It decays slowly towards higher frequency as a Lorentzian
function. The positive sign of real part of χ at Ω/2π<0kHz indicates gain for LP11 mode at the
Stoke frequency. The negative sign of real part of χ at Ω/2π>0kHz indicates loss for LP11
mode at anti-Stoke frequency. This is opposite to what observed in [22] for STRS in liquid.
This is largely due to the fact that index change in liquid and gas is dominated by thermally-
induced expansion with a negative n2, while, in glass, index change is dominated by thermal
optic effect with a possible n2.

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2650
The frequency separation between the LP01 and LP11 mode at the maximum nonlinear
coupling is Ω/2π = ~3.4kHz, which is much smaller than spectral width of almost all seed
lasers. This implied LP11 mode can be seeded by a seed laser at the required Stoke frequency.

Fig. 1. (a) Simulated nonlinear coupling coefficient amplitude χmnl and damping factor Γml for
each l, and (b) real, imaginary and phase of χ for a step index fiber with NA = 0.06, 2a = 2d =
30μm, 2b = 400μm and V = 5.3348. Simulations of 4 fibers with core diameters 2a = 30μm,
25μm, 20μm, and 15μm, with rest of the parameters kept constant are summarized in Fig. 2.

These fibers are good representations of LMA fibers commonly found in high-power fiber
amplifiers. The variations of core diameters lead to different V values at 5.3348, 4.4456,
3.5565 and 2.6674 respectively for the 4 fibers. Nonlinear coupling coefficient amplitude χmnl
are plotted in Fig. 2(a), while real part of χ and its phase is plotted in Fig. 2(b). The l for peak
χmnl moves to larger number for smaller cores to account for the smaller active area. The
damping factors do not change at all with a change of core diameters. The data for various
core diameters essentially overlap in Fig. 2(a). This is due to the fact that the period of
traveling wave is significantly larger than fiber diameter 2b. The second term in the bracket in
Eq. (29), therefore, dominates, which only depends on m, l and b. The peak of the real part of
χ moves towards larger frequency separation for smaller core diameter, accompanied by a
broadening of the spectrum. This move of peak towards higher frequency is a reflection of
movement towards larger temperature mode number l with higher damping factor shown in
Fig. 2(b). It will be shown later on that the smaller nonlinear coupling coefficient at the peak
for smaller core diameter is largely a reflection of smaller V value of these fibers.

Fig. 2. (a) Simulated nonlinear coupling coefficient amplitude χmnl and damping factor Γml for
each l, and (b) real, imaginary and phase of χ for step index fibers with NA = 0.06, 2b =
400μm, 2a = 2d = 30μm, 25μm, 20μm, and 15μm respectively, V = 5.3348, 4.4456, 3.5565 and
2.6674 respectively.

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2651
For the 30μm core step index fiber, the dependence on V value is studied by varying NA
while keeping the rest of parameters unchanged. The real and imaginary part of χ at the real
part peak are plotted in Fig. 3(a), along with the corresponding frequency of the peak fmax.
When V is reduced from 5.5, the peak nonlinear coupling coefficient decreases initially very
slowly and this decrease then accelerates near LP11 mode cut-off at around 2.405. The
absolute value of the peak frequency fmax decreases at smaller V, reflecting the increasing
delocalization of LP11 mode while moving towards its cut_off. A reduction of the doped area
also reduces the peak nonlinear coupling coefficient (see Fig. 3(b)). This reduction is,
however, small when d/a is near 1. Reducing the nonlinear coefficient by 50% requires
d/a≈0.45, i.e. a doped area reduction by ~80%. This would require a significant increase of
doping level to maintain the same level of gain/absorption per unit length. Considering
doping levels are already near their upper limits in many fibers, this may not be possible.
Absolute value of the peak frequency fmax increases with a reduction of d/a, reflecting the
smaller active area.

Fig. 3. Nonlinear coupling coefficient χ at the peak of real part of χ and the corresponding
frequency fmax for a step index fiber with 2b = 400μm, 2a = 2d = 30μm, dependence on (a) V
(NA is varied to change V) and (b) fraction of the doped radius d/a at NA = 0.06.

Fig. 4. Nonlinear coupling coefficient χ at the peak of real part of χ and the corresponding
frequency fmax for step index fibers with 2b = 400μm and 2a = 2d, core diameter is varied while
V is kept constant for each lines (NA is varied to keep V constant).

The dependence of the nonlinear coupling coefficient at fmax is also studied for various
core diameters while V is kept constant for step index fibers with 2b = 400μm and 2a = 2d.
NA is varied to keep V constant. The results are shown in Fig. 4. Absolute value of the peak
frequency fmax increases with a reduction in core diameter as expected. The nonlinear
coupling coefficient remains almost constant at various core diameters. The overlap integrals
in Eq. (28) are essentially dependent only on V. As core diameter gets smaller, peak mode
number l moves to higher mode number with a larger Γml (see Fig. 2(a)), accompanied by a

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2652
broadening of the distribution (see Fig. 2(a)). These effects cancel each other out overall to
keep χ at the peak frequency fmax largely constant.
7. STRS power threshold
Finding out threshold power for modal instability is very important for amplifier designs.
Assuming a fiber amplifier of length L where threshold is reached at z = L, the gain described
in Eq. (30b) can be integrated from 0 to L to find power in the higher order mode, noting that,
in the high gain regime over most part of the fiber, P11N(z)<<P01N(z) and P01N(z)≈P01(0),
assuming uniform gain along the fiber and threshold is reached at Pmn(L) = x P01(L) where
x<<1.

Pmn ( L ) ≈ Pmn ( 0 ) e(
gmn − g01 −α mn ) L χ mn
r
P01 ( L )
e (32)

The threshold power can then be obtained,

1   P01 ( 0 )  
P01th ≈ r 
ln  x  − ( g mn − g01 − α mn ) L  (33)
χ mn   Pmn ( 0 )  
For typical LMA fibers, the second term in the bracket is very small, Eq. (33) is reduced to,
 P (0) 
1
P01th ≈
ln  x 01  (34)
χ  Pmn ( 0 ) 
r
mn

Equation (33) is only applicable in the high gain regime. It is interesting to see the
threshold power depends only on real part of χmn and weakly on input condition P01(0)/Pmn(0)
once x is known in this high gain regime and independent of any other amplifier parameters.
In the low gain regime, the assumption P01N(z)≈P01(0) is no longer true. The lower P01N(z)
leads to a higher threshold power. It can be seen in Eq. (33) that smaller higher-order-mode
gain and larger higher-order-mode loss can also increase the threshold power. This is,
however, very limited, due to the fact that the first term in the bracket in Eq. (33) dominates
in most cases.
To test Eq. (34), amplifier based on a LMA fiber with 30μm core diameter and NA of
0.06 are studied numerically by solving the coupled mode equations described in Eq. (30) for
threshold powers for a range of input conditions. The fiber parameters are NA = 0.06, 2b =
400μm, 2a = 2d = 30μm, V = 5.3348 and α11 = 0. Both g01 and g11 are considered with α11 =
0.The results versus total gain factor g01L (plotted in dBs) are summarized in Fig. 5. The
predicated threshold powers from Eq. (34) are plotted as solid red lines in Fig. 5(a). The
threshold power are independent of gain when g01L>4, i.e. ~17dB. Below this, threshold
power increases with a reduction in gain. For g01L>4, Eq. (34) fits the numerical data very
well. Equation (34) can be modified slightly to account for the threshold increases at lower
gains.

1  P ( 0 )  1.25/ e g 01L
P01th ≈ ln  x 01  e (35)
χ  Pmn ( 0 ) 
r
mn

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2653
Fig. 5. Simulated threshold powers (a) at x = 1% and various input conditions with
P11(0)/P01(0) = 10−5, 10−10, 10−15, 10−20, 10−25, and 10−30 and (b) x = 0.0526, i.e. 5% of total
power in LP11 mode, for P11(0)/P01(0) = 10−2, 10−3, and 10−4. The fiber parameters are NA =
0.06, 2b = 400μm, 2a = 2d = 30μm, V = 5.3348 and α11 = 0. The dashed red lines in (a) are
obtained from Eq. (32) and solid black lines in both figures are from Eq. (35).

The modified threshold powers described in Eq. (35) are plotted as dotted black lines in
Fig. 5(a) and (b). It fits very well with the numerical data through the entire gain range. The
threshold power can be seen in Fig. 5 to increase with a decrease in P11(0)/P01(0) as expected
from Eq. (35). The only two amplifier parameters which can be used to increase the threshold
powers are lower gain and lower P11(0)/P01(0).
Using the FWHM STRS gain bandwidth of ~20kHz for the 30micron core LMA fiber (see
Fig. 1(b)), the quantum noise at 1060nm is estimated to be ~2 × 10−28 W. In practice, the seed
to the higher order mode are likely from the input signal when the input signal spectrum is
broader than just few kHz for core diameter over ~30μm, and, can be, therefore, much higher
than this quantum limit.
8. Mode coupling dynamics
The power evolution along a fiber amplifier is simulated by numerically solving Eq. (30) and
illustrated in Fig. 6. The fiber parameters are NA = 0.06, 2b = 400μm, 2a = 2d = 30μm, V =
5.3348 and α11 = 0. It can be clearly seen in Fig. 6(a) that the LP01 mode is amplified normally
at the first part of the fiber, while the LP11 mode experiences significant nonlinear gain of
>250dB in this case. The fraction of LP11 mode power over the total power is also plotted in
Fig. 6(a), showing the rapid switch over at threshold. Once over the STRS threshold, the
power continues to couple from LP01 mode to LP11 mode, because the sign of nonlinear
coupling coefficient remains the same in Eq. (30). This coupling coefficient diminishes as
power in LP01 mode decreases. The LP11 mode undergoes continued linear amplification in
the second part of the fiber amplifier. In practice, it is expected that LP01 mode can be re-
seeded when its power is below the noise level in the fiber. This can be simulated by
switching the sign of the nonlinear coupling coefficient in Eq. (30). This is done in Fig. 6(b),
showing the reversing of coupling from LP11 to LP01 after the first coupling cycle. This
behavior is repeated with an increasing spatial frequency, driven by the increased total power.
This is confirmed to some extent by the observed chaotic behavior when operating well above
the threshold power [9].

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2654
Fig. 6. Simulated LP01 and LP11 powers along a fiber amplifier, (a) without re-seeding and (b)
with re-seeding. The fiber parameters are NA = 0.06, 2b = 400μm, 2a = 2d = 30μm, V =
5.3348 and α11 = 0. Total amplifier gain is 13.5dB at the STRS threshold. Input power P01(0) =
19.2W and P11(0) = 10−25 × P01(0). Threshold power is 428.3W.

9. Spectral considerations
Up to this point, the optical powers are assumed to be at single frequencies ω01 and ωmn. In
practice, seed lasers can have much broader spectrum than that of the STRS gain. In case
where the input power spectrum is broader than that of the STRS gain spectrum, if the power
in the higher order mode at a given frequency is seeded by the input signal at frequency very
close by (less than few kHz frequency separation for core diameter over 30μm), it reasonable
to assume that phases of the fields in LP01 and LP11 modes with a small frequency separation
of Ω are identical at the amplifier input. This may in fact be true across the entire input signal
spectrum even for the case where the seed is an ASE source as in [9]. In this case, input signal
power at any given frequency can interfere with power in LPmn mode at an adjacent frequency
Ω away to produce an intensity traveling wave described by exp(i(qz-Ωt)). In another word,
power at any frequency within the input power spectrum can interfere with its corresponding
power in the LPmn mode to add to the intensity of the traveling temperature wave described by
exp(i(qz-Ωt)). This is only true when the phase difference of the two fields in the interfering
modes is constant across the power spectrum. This is not the case, for example, for SBS,
where the counter-propagating wave is seeded from quantum noise without any fixed phase
relationship to the input signal. This collaborative effect can lead to the possibility that total
power of the input signal contributes towards nonlinear coupling at any local frequency
within the input signal spectrum in STRS, despite the fact that the power spectrum of the
signal is much larger than the STRS gain spectrum. This effect can lead to threshold being
independent of input signal bandwidth but more dependent on the total power of the input
signal as experimentally observed in [9]. This effect is unique only to STRS due to the fact
that the interfering fields involved are originated from the same source and can, therefore,
have identical phase. In case where the seed laser spectrum is significantly wider than the
STRS gain spectrum, replacing power by power spectrum density represented by S and
keeping the same subscripts and superscripts, the nonlinear coupled mode equations can be
written as,

∂ S01N (ω01 , z )
 Γmn 
−α mn  z
1  g01
Γ
≈ − g 01 χ mn
r
e  01 
∂z 2

(36a)
 2S N
01 (ω01 , z ) S (ω01 − Ω, z )d ω01
N
mn S N
mn (ωmn , z )
0

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2655
∂ Smn
N
(ωmn , z ) 1
≈ g 01 χ mn
r
e g01z
∂z 2

(36b)
 2S N
01 (ω01 , z ) S (ω01 − Ω, z )d ω01
N
mn S N
01 (ω01 , z )
0

If the powers in the two modes have the same relative power density spectrum that is much
broader than the STRS gain bandwidth (this is a reasonable assumption if the power in LPmn
mode is seeded by the signal power in LP01 mode), the integral in Eq. (36) is, in fact, square
root of product of total powers in the two modes. It is easy to see in Eq. (36) that the total
powers in the two modes are contributing towards nonlinear coupling at any frequency within
the power density spectrum. This can also lead to uniform nonlinear gain across the power
spectrum for the LPmn mode.
10. Discussions and conclusions
The mode instability observed recently in high-power fiber lasers is based on the same
nonlinear coupling mechanism as STRS first observed half a century ago. By finding quasi-
closed-form solution for the heat transportation equation in optical fibers, it is able to find
quasi-closed-form solution for the nonlinear coupling coefficient for STRS in optical fibers.
The quasi-analytic solution shines a great deal of insights on the underlying physics involved
as well as on critical design parameters both in optical fibers and fiber amplifiers. Threshold
power can be obtained by a simple analytical formula, surprisingly depends only on the
nonlinear coupling coefficient and seed condition in the high gain regime, not on amplifier
designs. The evolution of power along the fibers are also studied, showing well behaved
power growth when under the threshold power, rapid switch over at the threshold power, and
oscillatory behavior above threshold power.
It should be pointed out that only the steady-state solutions are studied in this work. In
reality, high environmental sensitivity of thermal process can lead to temporal fluctuations in
the nonlinear coupling. It can take several thermal diffusion cycles to reach steady state. It is,
therefore, not surprising to see oscillatory behavior of mode coupling over time when
operating above threshold at a frequency corresponding to the thermal diffusion rate. It is also
expected to be more chaotic when operating well above threshold, due to the higher power
levels in the fiber, especially considering the increased number of higher order modes
reaching thresholds. When a fiber amplifier starts from cold at above threshold powers, it is
expected that there is initially a great deal of temperature fluctuations along the fiber due to
different part of the fiber heating up slightly differently. Assuming the power is mostly in the
fundamental mode, the deposited heat will resemble the fundamental mode spatial pattern at
this early stage. This large temperature fluctuation along the fiber can easily overwhelm those
from much smaller amount of heat deposition from mode interference, and, consequently,
suppresses STRS at startup. STRS can only start when a thermal equilibrium is reached from
a cold start.
After the initial submission of this manuscript, the author was made aware of two
additional references [25, 26]. Some concepts regarding transient process are discussed in
[25], while steady-state is left largely ignored. Mode instability is referred to as STRS in [26]
for the first time to this author’s knowledge. Quantum noise corresponding to one photon per
beat cycle at 2 × 10−16 W is suggested to seed LP11 mode in [26].
Acknowledgments
This material is based upon work supported in part by the U. S. Army Research Laboratory
and the U. S. Army Research Office under contract/grant number W911NF-10-1-0423
through a Joint Technology Office MRI program.

#180850 - $15.00 USD Received 29 Nov 2012; revised 27 Dec 2012; accepted 17 Jan 2013; published 28 Jan 2013
(C) 2013 OSA 11 February 2013 / Vol. 21, No. 3 / OPTICS EXPRESS 2656

You might also like