Four-Wave Mixing Stability in Hybrid Photonic Crystal Fibers With Two Zero-Dispersion Wavelengths

Four-wave mixing stability in hybrid
photonic crystal fibers with two

zero-dispersion wavelengths
Benoit Sévigny,1,∗ Olivier Vanvincq,1 Constance Valentin,1 Na Chen,1,2

Yves Quiquempois,1 and Géraud Bouwmans1
1 Laboratoirede Physique des Lasers, Atomes et Molécules, UMR 8523,
IRCICA Research Institute for Advanced Communication,
Parc Scientifique de la Haute Borne, CNRS USR 3380,
50, avenue Halley, 59658 Villeneuve d’Ascq cedex, France
2 KeyLab of Specialty Fiber Optics and Optical Access Networks,
School of Communication and Information Engineering,
Shanghai University, Shanghai 200072, China
∗ benoit.sevigny@ed.univ-lille1.fr
Abstract: The four-wave mixing process in optical fibers is generally

sensitive to dispersion uniformity along the fiber length. However, some
specific phase matching conditions show increased robustness to longitu-
dinal fluctuations in fiber dimensions, which affect the dispersion, even
for signal and idler wavelengths far from the pump. In this paper, we
present the method by which this point is found, how the fiber design
characteristics impact on the stable point and demonstrate the stability
through propagation simulations using the non-linear Schrödinger equation.
© 2013 Optical Society of America
OCIS codes: (060.2280) Fiber design and fabrication; (060.5295) Photonic crystal fibers;
(190.4380) Nonlinear optics, four-wave mixing.
References and links

1. N. Shibata, R. Braun, and R. Waarts, “Phase-mismatch dependence of efficiency of wave generation through
four-wave mixing in a single-mode optical fiber,” IEEE J. Quantum Electron. 23, 1205–1210 (1987).
2. G. P. Agrawal, Lightwave Technology: Telecommunication Systems (Wiley. com, 2005).
3. P. V. Mamyshev and N. A. Mamysheva, “Pulse-overlapped dispersion-managed data transmission and intrachan-
nel four-wave mixing,” Opt. Lett. 24, 1454–1456 (1999).
4. J. E. Sharping, M. Fiorentino, A. Coker, P. Kumar, and R. S. Windeler, “Four-wave mixing in microstructure
fiber.” Opt. Lett. 26, 1048–1050 (2001).
5. R. W. Boyd, Nonlinear Optics, vol. 5 of Electronics & Electrical (Academic Press, 2003).
6. G. P. Agrawal, Nonlinear Fiber Optics (Academic Press, 2006), 4th ed.
7. M. Levenson, R. Shelby, A. Aspect, M. Reid, and D. Walls, “Generation and detection of squeezed states of light
by nondegenerate four-wave mixing in an optical fiber,” Phys. Rev. A 32, 1550–1562 (1985).
8. J. Chen, X. Li, and P. Kumar, “Two-photon-state generation via four-wave mixing in optical fibers,” Appl. Phys.
Lett. 72, 033801 (2005).
9. C. Bennett, F. Bessette, and G. Brassard, “Experimental quantum cryptography,” Journal of Cryptology 5, 3–28
(1992).
10. T. Jennewein, C. Simon, G. Weihs, H. Weinfurter, and A. Zeilinger, “Quantum cryptography with entangled
photons,” Phys. Rev. Lett. 84, 4729–4732 (2000).
11. J. A. Slater, J.-S. Corbeil, S. Virally, F. Bussières, A. Kudlinski, G. Bouwmans, S. Lacroix, N. Godbout, and
W. Tittel, “Microstructured fiber source of photon pairs at widely separated wavelengths.” Opt. Lett. 35, 499–
501 (2010).
12. A. Furusawa, “Unconditional quantum teleportation,” Science 282, 706–709 (1998).
#199150 - $15.00 USD Received 8 Oct 2013; revised 27 Nov 2013; accepted 28 Nov 2013; published 6 Dec 2013
(C) 2013 OSA 16 December 2013 | Vol. 21, No. 25 | DOI:10.1364/OE.21.030859 | OPTICS EXPRESS 30859
13. A. F. Abouraddy, B. E. A. Saleh, A. V. Sergienko, and M. C. Teich, “Role of entanglement in two-photon imag-
ing,” Phys. Rev. Lett. 87, 123602 (2001).
14. J. Hansryd, P. Andrekson, M. Westlund, J. Li, and P. Hedekvist, “Fiber-based optical parametric amplifiers and
their applications,” IEEE J. Sel. Top. Quantum Electron. 8, 506–520 (2002).
15. R. Stolen and J. E. Bjorkholm, “Parametric amplification and frequency conversion in optical fibers,” IEEE J.
Quantum Electron. 18, 1062–1072 (1982).
16. A. Kudlinski, V. Pureur, G. Bouwmans, and A. Mussot, “Experimental investigation of combined four-wave
mixing and Raman effect in the normal dispersion regime of a photonic crystal fiber,” Opt. Lett. 33, 2488–2490
(2008).
17. S. Watanabe, T. Naito, and T. Chikama, “Compensation of chromatic dispersion in a single-mode fiber by optical
phase conjugation,” IEEE Photon. Technol. Lett. 5, 92–95 (1993).
18. S. Watanabe and M. Shirasaki, “Exact compensation for both chromatic dispersion and Kerr effect in a transmis-
sion fiber using optical phase conjugation,” J. Lightwave Technol. 14, 243–248 (1996).
19. G. J. Dunning and R. C. Lind, “Demonstration of image transmission through fibers by optical phase conjuga-
tion,” Opt. Lett. 7, 558–560 (1982).
20. T. Andersen, K. Hilligsøe, C. Nielsen, J. Thøgersen, K. Hansen, S. Keiding, and J. Larsen, “Continuous-wave
wavelength conversion in a photonic crystal fiber with two zero-dispersion wavelengths: erratum.” Opt. Express
13, 3581–3582 (2005).
21. T. Torounidis, M. Karlsson, and P. A. Andrekson, “Fiber optical parametric amplifier pulse source: Theory and
experiments,” J. Lightwave Technol. 23, 4067 (2005).
22. F. Abdullaev and S. Darmanyan, “Modulational instability of electromagnetic waves in media with varying non-
linearity,” J. Opt. Soc. Am. B 14, 27–33 (1997).
23. M. Karlsson, “Four-wave mixing in fibers with randomly varying zero-dispersion wavelength,” J. Opt. Soc. Am.
B 15, 2269–2275 (1998).
24. F. Abdullaev, B. Umarov, M. Wahiddin, and D. Navotny, “Dispersion-managed solitons in a periodically and
randomly inhomogeneous birefringent optical fiber,” J. Opt. Soc. Am. B 17, 1117–1124 (2000).
25. B. P.-P. Kuo and S. Radic, “Highly nonlinear fiber with dispersive characteristic invariant to fabrication fluctua-
tions.” Opt. Express 20, 7716–7725 (2012).
26. B. P.-P. Kuo, M. Hirano, and S. Radic, “Continuous-wave, short-wavelength infrared mixer using dispersion-
stabilized highly-nonlinear fiber.” Opt. Express 20, 18422–18431 (2012).
27. J. Lægsgaard, “Phase-matching conditions for single-pump parametric amplification in photonic crystal fibers,”
J. Opt. A: Pure Appl. Opt. 9, 1105–1112 (2007).
28. Y. Chen, W. J. Wadsworth, and T. A. Birks, “Ultraviolet four-wave mixing in the LP02 fiber mode,” Opt. Lett.
38, 3747–3750 (2013).
29. B. Øksendal, Stochastic Differential Equations, 6th ed. Universitext (Springer, Berlin, 2003).
30. E. Bibbona, G. Panfilo, and P. Tavella, “The Ornstein-Uhlenbeck process as a model of a low pass filtered white
noise,” Metrologia 45, S117 (2008).
31. D. Gillespie, “Exact numerical simulation of the Ornstein-Uhlenbeck process and its integral.” Phys. Rev. E 54,
2084–2091 (1996).
32. Y. Ould-Agha, A. Bétourné, O. Vanvincq, G. Bouwmans, and Y. Quiquempois, “Broadband bandgap guidance
and mode filtering in radially hybrid photonic crystal fiber.” Opt. Express 20, 6746–6760 (2012).
33. T. A. Birks, J. C. Knight, and P. S. Russell, “Endlessly single-mode photonic crystal fiber,” Opt. Lett. 22, 961–
963 (1997).
34. P. M. Moselund, M. H. Frosz, C. L. Thomsen, and O. Bang, “Back-seeding of higher order gain processes in
picosecond supercontinuum generation.” Opt. Express 16, 11954–11968 (2008).
35. A. Bétourné, Y. Quiquempois, G. Bouwmans, and M. Douay, “Design of a photonic crystal fiber for phase-
matched frequency doubling or tripling,” Opt. Express 16, 14255–14262 (2008).
36. K. Inoue, “Four-wave mixing in an optical fiber in the zero-dispersion wavelength region,” J. Lightwave Technol.
10, 1553–1561 (1992).
37. K. Inoue, “Arrangement of fiber pieces for a wide wavelength conversion range by fiber four-wave mixing,” Opt.
Lett. 19, 1189–1191 (1994).
38. J. Laegsgaard, “Mode profile dispersion in the generalized nonlinear Schrödinger equation,” Opt. Express 15,
16110–16123 (2007).
39. J. Hult, “A fourth-order Runge–Kutta in the interaction picture method for simulating supercontinuum generation
in optical fibers,” J. Lightwave Technol. 25, 3770–3775 (2007).
40. P. L. François, “Nonlinear propagation of ultrashort pulses in optical fibers : total field formulation in the fre-
quency domain,” J. Opt. Soc. Am. B 8, 276–293 (1991).
41. N. Smith, F. Knox, N. Doran, K. Blow, and I. Bennion, “Enhanced power solitons in optical fibres with periodic
dispersion management,” Electronics Letters 32, 54–55 (1996).
42. O. Yaakobi and L. Friedland, “Autoresonant four-wave mixing in optical fibers,” Phys. Rev. A 82, 023820
(2010).
1. Introduction
Four-wave mixing (FWM) in optical fibers has been extensively studied, increasingly so in
the last two decades [1–3] as it presents a nuisance in WDM communication systems but also
because it provides a leveraging platform for a range of interesting applications [4–6]. For in-
stance, generation of correlated light at different wavelengths through this process, giving rise
to entangled photons [7, 8], allows for quantum cryptography, [9–11] and a range of other ap-
plications necessitating entangled light sources [12, 13]. Additional applications, also based on
FWM, include parametric amplification [14–16], phase-conjugation [17–19], frequency conver-
sion (FOPO) [15, 20] and pulse generation [21] amongst others. It has been reported in various
publications that the efficiency of this process is generally impaired by random longitudinal
fluctuations of the dispersion along the fiber, which can limit the gain bandwidth in FOPAs
and substantially diminish wavelength conversion efficiency [22–24]. Furthermore, this degra-
dation is usually worse as the phase-matching condition occurs farther from the pump [23].
Highly nonlinear fiber designs have been studied which stabilize the dispersion in the presence
of scale fluctuations [25, 26]. The idea relies on negating the scale-sensitivity of the dispersion
by compensating the contributions of different waveguiding structures in a multi-layered index
configuration. The effect of this is the flattening of the waveguide dispersion contribution in a
region of interest and, for instance, can result in a fixed zero-dispersion wavelength invariant to
scale fluctuations, which can be of use in a number of applications. This method, however, may
prove challenging when the wavelength range becomes quite large (well in excess of 100 THz)
and higher-order dispersion becomes important. It was also noted elsewhere that some phase-
matched sets of wavelengths are invariant with small scale perturbations for some PCF designs
with two ZDW [27] and could potentially be used as stable degenerate FWM (DFWM) points
but a full investigation on the properties of these phase-matching conditions and a gain analysis
to validate the usability of these points remained to be performed. In this paper, we investigate,
with the help of an elaborate DFWM model and numerical simulations, stable phase-matching
points, both in wavelength and gain, that do not require the ZDW (or dispersion) to remain
fixed and are particularly useful in terms of fiber manufacturing process, especially for large
frequency spacings [28]. The procedure by which the dispersion can be analyzed to find the
point of stability and a numerical investigation that shows the stability of the gain under these
conditions (in a normal dispersion pumping regime) will be presented. As well, a method by
which the design parameters of a hybrid photonic-crystal fiber (HPCF) can be tuned to have
this stable point occur at a wide range of different sets of pump, signal and idler wavelengths
will be discussed.
2. Four-wave mixing and longitudinal dispersion fluctuations

To begin this study, let us first consider the general case of quasi-continuous-wave (CW) de-
generate FWM where three waves (an intense, undepleted pump “p” and weak signal “s” and
idler “i” with wavelengths λs < λ p < λi ) interact together through the third-order susceptibility
tensor in an isotropic medium:
∂ Ap in2 ω p 2
= f pp A p A p (1)
∂z c
∂ As in2 ωs 2
= 2 fsp A p As + fsipp A2p A∗i exp(−iΔkz) (2)
∂z c
∂ Ai in2 ωi 2
= 2 fip A p Ai + fispp A2p A∗s exp(−iΔkz) (3)
∂z c
where A denotes the envelope of the wave, fi j and fi jkl are the overlap integrals (considered
constant for small variations in fiber scale and which can be readily derived, in vectorial form,
by keeping the full fields at pump, signal and idler wavelengths in the non-linear polarization
term of the wave equation) [6], n2 is the nonlinear refractive index of the material (assumed
constant over the cross-section), ω is the angular frequency of the wave and Δk is the linear
phase-mismatch term defined as:
Δk = βs + βi − 2β p (4)
where βm is the propagation constant at wavelengths λm . Under these conditions, the pump can
1/2
be trivially integrated to A p = Pp exp(iγ Pp z) with γ = n2 ω p f pp /c. The coupled equations for
the signal and idler waves can thus be recast as follows:
∂ As in2 ωs
= [2 fsp Pp As + fsipp Pp A∗i exp(−iθ z)] (5)
∂z c
∂ A∗i in2 ωi ∗
= − 2 fip Pp A∗i + fispp
∗
Pp As exp(iθ z) (6)
∂z c
where θ = Δk − 2γ Pp . At this point, it is necessary to consider that the dispersion fluctuates
along the fiber length. In this case, the phase argument can readily be z generalized from θ z to
z (ζ )dζ ]/2} and simi-
0 θ ( ζ )d ζ . Taking this into account and substituting Bs = A s exp{i[ 0 θ
larly for the idler wave, we can reform the system of coupled differential equations as:
∂ Bs κsp
i = Bs − γsipp Pp B∗i (7)
∂z 2
∗
κip
∂ B∗
i i = − B∗i + γispp∗
Pp Bs (8)
∂z 2
where γmnpp = n2 ωm fmnpp /c. If we consider that the fluctuation of the dispersion can be mod-
eled as a stochastic process around a constant mean value along the fiber length, we can
write the global phase-mismatch term, κi j , as the sum of a constant term and a fluctuating
(z-dependent) term:
κsp = 2(γ − 2γsp )Pp − Δkmean − Δkvar (z) (9)

κip = 2(γ − 2γip )Pp − Δkmean − Δkvar (z) (10)
where γmp = n2 ωm fmp /c. Grouping the constant terms together, κmp /2 = Δmp + W (z), where
Δmp = −[2(2γmp − γ )Pp + Δkmean ]/2 and W (z) = −Δkvar (z)/2, the whole system of differential
equations can be re-written in matrix form:
∂B
= [G + HW (z)]B (11)
∂z
where the matrices B, G and H are given by

Bs −iΔsp iγsipp Pp −i 0
B= , G= ∗ P , H= . (12)
B∗i −iγispp p iΔ∗ip 0 i
Since W (z) is a stochastic variable, Eq. (11) is now considered to be a stochastic differential
equation, or SDE [29], and the solution must also be a stochastic process. It is to be noted that
the analysis proposed here is a generalization of the derivation presented in [23], keeping the
loss terms and full overlap integrals as the case where signal and idler far from the pump will
be considered, so the constant γ approximation becomes invalid. In order to push the analysis
further, more information is needed on W (z). The first thing to mention is that if several com-
ponents of very different statistical nature were to contribute to the stochastic fluctuation of Δk,
several terms Wi (z) should be added to the SDE. However, in this case we will consider that
the use of one process is enough and such formulation has been shown to produce accurate
predictions [23]. In first approximation, we can assume that Δk is linearly related to physical
stochastic processes that are occurring along the fiber. Let us thus consider a number of contri-
butions X to the fluctuations of the phase-mismatch. The fluctuation of those parameters, δ X(z),
can be characterized in terms of their correlation length Lc,X , which can be derived from the
autocorrelation function RX (ζ ) = δ X(z)δ X(z + ζ ) (where “. . .” denotes the expected value
operator with RX (0) = δ X(z)δ X(z) = σX2 , σX being the standard deviation of X) as:

∞
Lc,X = σX−2 RX (ζ )dζ (13)
0
If we make the assumption that all contributions to the fluctuation in the dispersion are rela-
tively small (i.e. can be expressed as a first order expansion), have zero mean and are symmetric
around a nominal value Xi,0 , have the same correlation length Lc and are statistically indepen-
dent, they can be combined in a single stochastic variable W (z) with standard deviation given
by:
2
N
1 ∂ Δk
σW2 = ∑ σX2i (14)
i=1 2 ∂ Xi Xi =Xi,0
For instance, lets assume that the change in the phase-mismatch is primarily related to a change
in fiber dimensions (for instance, through the pitch Λ), which is a major contributor [28], ne-
glecting other contributions, we can write the standard deviation of the stochastic variable W (z)
as: 2

1 ∂ Δk
σW =
2
σΛ2 (15)
2 ∂ Λ Λ=Λ0
A representative process that can be applied to optical fiber drawing is the Ornstein-
Uhlenbeck process (OUP) [30], which is basically used to model noisy relaxation and can
be easily implemented [31]. It can be noted that this process has an autocorrelation function
given by
|ζ |
R(ζ ) = R(0) exp − . (16)
Lc
The parameter of interest in this study is the power at the signal and idler wave-
lengths (Ps = |As |2 and Pi = |Ai |2 ) for which a new set of four coupled SDEs for P =
[Ps , Pi , 2Re(Bs Bi ), 2Im(Bs Bi )]T can be derived from the amplitude differential equations. This
system of equations can be written exactly in the same form as Eq. (11) where G and H are
now 4×4 matrices. To simplify the following calculations, one can derive that fispp = fsipp
and is real. To further simplify the remaining calculations and illustrate the contribution of the
frequencies ωs and ωi , it is convenient to define h = n2 fispp /c = n2 fsipp /c. The complete SDE
system for power is thus given by
⎡ ⎤
2Im(Δsp ) 0 0 ωs hPp
⎢ 0 2Im(Δip ) 0 ωi hPp ⎥
G=⎢ ⎣
⎥ (17)
0 0 Im(Δsp + Δip ) Re(Δsp + Δip ) ⎦
2ωi hPp 2ωs hPp −Re(Δsp + Δip ) Im(Δsp + Δip )
with H having only two non-zero elements H34 = −H43 = 2. Careful integration of Eq. (11)
can be performed, yielding
P(z) = exp (Gz + HB(z)) P(0), (18)
z )dz .
with B(z) = 0 W (z Taking the average, as shown in [23], we get

H2
P(z) = exp Gz + f (z) P(0). (19)
2
with, assuming the same type of autocorrelation function as Eq. (16),

z z
f (z) = 2Lc2 σW2 exp − −1+ . (20)
Lc Lc
It is to be noted that the complete effect of the perturbations is contained in f (z); the behavior
of this factor is thus of critical importance in characterizing the impact of fluctuations.
This solution is generally complicated but has a simpler form in some cases. Let us first
consider that Δsp and Δip are real in the absence of loss since, in this case, the propagation
constants are real. Furthermore, γsp and γip are also real. Thus G11 = G22 = 0. For further
simplification, we consider the gain at the nominal phase-matching condition such that Δsp +
Δip = 0. For the case of spontaneous amplification of a small noise at the signal wavelength,
equivalent to P(0) = [Ps0 , 0, 0, 0]T , the expected gain G = Ps /Ps0 is calculated to be (as
shown in [23])

1 f (z)
G = 1 + cosh[g(z)] + sinh[g(z)] exp[− f (z)] (21)
2 g(z)
1/2
where, however, g(z) is now defined as g(z) = ωs ωi (2hPp z)2 + f 2 (z) . Equation (21) is
always maximum for f (z)=0, which occurs if σΛ = 0 or ∂ Δk/∂ Λ = 0. This implies that maxi-
mum gain should be obtained at the phase-matching condition, in the presence of fluctuations
(σΛ = 0), when ∂ Δk/∂ Λ = 0. This formally illustrates, in other words, that if the phase match-
ing does not change with small changes in the scale of the structure, then the gain should be
stable as well.
3. Investigation of FWM stability in the case of a hybrid PCF design with two ZDW
Let us consider the case of a fiber with two zero-dispersion wavelengths (ZDW). Such fibers
are pertinent in this study as they can show two sets of phase-matched wavelengths and as we
will show, under certain conditions, a stable point can be found at one set but not at the other.
A particular fiber design will serve for our study, which exhibits two ZDWs; the hybrid PCF
presented in [32], and shown on Fig. 1, represents a good example of such a fiber as the funda-
mental mode of the first all-solid photonic bandgap (ASPBG, at short wavelengths) smoothly
transitions to the LP02 mode of a 7-cell defect PCF (at long wavelengths), both of which exhibit
a ZDW [32]. More specifically, the design considered here consists of six Ge-doped inclusions,
with maximum index difference of n0 −nSiO2 = 0.032 and dGe /Λ = 0.7, embedded in the seven-
cell defect of a PCF structure of air holes with dAir /Λ = 0.6. The nominal pitch Λ0 is selected
to be around 1 μ m. Simulations were performed to compute the dispersion, at the nominal con-
ditions described above, from the point where the mode becomes the fundamental PBG mode
(its effective index neff < nSiO2 ) to the point where neff falls below nFSM , the effective index of
the fundamental space-filling mode of the 7-cell hole lattice [32, 33]. These simulations were
carried out by finite elements method, using commercially available software, and material dis-
persion was added pertubatively [6]. Although other fibers with similar two-ZDW have been
studied [20, 34], the advantage of such a design is that both ZDW can theoretically be tuned
by changing the parameters of the germanium inclusions and air-holes separately. More com-
plicated structures giving additional design flexibility could also be envisaged, as suggested
in [35].
Fig. 1. Illustration of the hybrid PCF fiber design under study (Fiber A). The gray
area represents pure silica, the white areas represent air holes and the blue areas rep-
resent graded-index (parabolic) Ge-doped inclusions with a maximum index difference
n0 − nSiO2 = 0.032. In this particular design, dAir /Λ = 0.6 and dGe /Λ = 0.7. The pitch is
Λ0 = 1.03 μ m.
From Eqs. (15) and (20), one can see that the condition ∂ Δk/∂ Λ = 0 represents a point
where the noise on the phase-matching vanishes (σW = 0, thus f (z) → 0) and maximal gain
is obtained. If material dispersion is taken into account, this stable point will be a function of
several design parameters. The concept of phase-matching stability can be further visualized
through the dynamics of how the signal and idler wavelengths vary with a change in pitch.
Figure 2 represents the linear phase-matching condition for fiber A (Fig. 1) at the nominal pitch
Λ0 as well as Λ0 ± 1%. As it can be seen, if the pitch is changed by a percent, the phase-
matched wavelengths generally move around except at one point, where the Δk and ∂ Δk/∂ Λ
are simultaneously equal to zero. It is to be noted too that the trivial condition λs = λi = λ p
also satisfies this condition but is not of interest in this study as our focus is on wavelength
pairs far from the pump. For such a fiber design, assuming fixed dGe /Λ, the stable point will be
a function of relative hole size dAir /Λ, pitch Λ and pump wavelength λ p . For a given design,
the pump, signal and idler wavelengths at the stable point will be identified as λ̃ p , λ̃s and λ̃i
respectively. The trajectory of the stable point in (Λ,λ̃ p ,λ̃s )-space can be plotted to reveal trends
on its behavior as a function of the design choices, as seen on Fig. 3. From this data, it can
be noted that smaller dAir /Λ will generally increase the signal wavelength (λ̃s ). It can also be
noted that the relationship between pitch and pump wavelength at the stable point is roughly
proportional to λ̃ p at high values (panel A of Fig. 3). One last trend to be observed is that λ̃s
becomes less sensitive to changes in Λ, as seen on panel B of Fig. 3 (or λ̃ p , as seen on panel C)
for higher pitch values. Indeed, at high pitch, λ̃s (Λ) is, for all intents and purposes, a vertical
line on panel B. This is very interesting from a manufacturing point of view as it suggests that,
for a selected pump wavelength and dAir /Λ, obtaining a draw at the right diameter to generate
the right stable point with a small deviation on the generated wavelength could be practically
achieved by, for instance, making a very long taper in the fiber and then retrieving the segment
with maximum gain.
However, an additional consideration has to be taken into account and it is related to the
type of dispersion at the pump wavelength. Indeed, the stable point may lie in the normal or
abnormal dispersion regime of the pump. In the case where the pump is in the abnormal regime,
supercontinuum generation from the pump fission into solitons [6] is likely to drain the pump
1.4
1.2
1
Δk = 0 (Fiber A)
λ s,i (μm)
∂Δk/∂Λ = 0
0.8
0.6
0.4
0.58 0.59 0.6 0.61 0.62 0.63 0.64 0.65 0.66

λ p (μm)
Fig. 2. Linear phase matching curve obtained for fiber A (Fig. 1). The thick solid line is the
phase matching at the nominal pitch while the two thinner solid lines illustrate the phase-
matching for variations in the pitch Λ = Λ0 ± 1 % with nominal pitch Λ0 = 1.03 μ m. The
dashed line represents the stability condition ∂ Δk/∂ Λ = 0. For this design, the stable point
is at pump λ̃ p = 602.6 nm. The outer set of phase-matched wavelengths are λ̃s =1,218.9 nm
and λ̃i = 400.2 nm. The inner set of phase-matched wavelengths are λs = 906.7 nm and
λi = 451.2 nm. The first ZDW for fiber A is located at 617.5 nm
energy and reduce the efficiency of the phase-matched FWM at the stable point, although this
remains to be investigated experimentally. An example of how the regime changes with pump
wavelength for dAir /Λ = 0.6 in our design is shown on Fig. 4. It is to be noted that the abnormal
dispersion regime is located between the two ZDW of any given curve. If the stable point lies in
the normal dispersion regime however, it is generally at a point where two sets of phase-matched
wavelengths exist, with only one (the outer set in our case) showing stability, as shown on Fig. 2.
In order to perform this investigation, let us focus on DFWM in the normal dispersion regime
which is achieved by selecting the design of fiber A (with λ̃ p around 600 nm) as, generally, the
gain bands of the outer set are very narrow and far from the pump, which suggests they should
be more sensitive to fluctuations of the dispersion. Furthermore, comparison with numerical
propagation simulations will be much easier to interpret as no supercontinuum will be generated
at the pump wavelength, thus greatly helping with visualization and comparison with the simple
gain model presented in the previous section.
In general, there are three factors that need to be examined to verify the practicality of the sta-
tionary phase-matching point: the correlation length of the dimensional noise Lc , pump power
Pp and pump wavelength λ p . In order to evaluate the impact of those factors on the expected
gain, Eq. (17) will be used, without approximations (i.e. full phase-mismatch, overlap integrals
and loss terms included) to compute the expected gain G = Ps /Ps (0) from Eq. (19), seeding
with noise of one photon per frequency and random phase at the fiber input. The following
subsections contain computational results showing the influence of those three parameters on
G.
3.1. Impact of correlation length

The impact of correlation length on gain degradation for FWM interaction in standard fiber has
been reported in several publications [22,23,36] and is known to be worse for longer correlation
Stable point parameters dAir /Λ =0.6
dAir /Λ =0.62
dAir /Λ =0.64
dAir /Λ =0.66
dAir /Λ =0.68
1.4
A
1.3
1.2
1.1 B
Λ (μm)
1
0.9
0.8
0.7
0.6
C
0.55
0.5
0.65
0.45
0.6
0.4
0.55
λ̃i (μm)
λ̃p (μm)
Fig. 3. Mapping of the stable point as a function in (Λ,λ̃ p ,λ̃s )-space for different dAir /Λ and
Λ, all other parameters being the same as for fiber A. The projection of the 3D trajectories
on each couple of parameters is also plotted on panels A, B and C to help illustrate the
different trends.
2.2
2 λ̃ p = 0.65 μm, Λ = 1.49 μm
λ̃ p = 0.60 μm, Λ = 1.05 μm

1.8
λ̃ p = 0.54 μm, Λ = 0.70 μm
1.6
1.4
λ i,s (μm)
1.2
0.8
0.6
0.4
0.5 0.6 0.7 0.8 0.9 1 1.1

λ p (μm)
Fig. 4. Example of the phase matching curve with the stable point at different pump wave-
lengths (and corresponding pitch) for dAir /Λ = 0.6. The black circles denote the position
of the stable set of wavelengths λ̃i,s for each pump wavelength λ̃ p and the black “×” indi-
cate the position of the ZDW. The dashed line represents the trivial solution λi,s = λ p . We
notice that, as the wavelength λ̃ p gets shorter, the stable point moves closer to the ZDW
as the pump then crosses into the abnormal dispersion regime. Missing points in the phase
matching curves lie outside the simulation domain and thus have been discarded. It is to be
noted that the pump is in the abnormal dispersion regime when it is located between the
two ZDW for a given curve.
lengths. Experiments have also been performed to effectively reduce the correlation length by
cutting a fiber into short pieces and randomly splicing them back together to notice an improve-
ment in FWM gain [37]. Close inspection of Eq. (20) also shows that long correlation lengths
will have a worse impact on gain as, for fixed z, f (z) is a monotonically increasing function
of Lc . One thing to note, however, is that f (z) is always proportional to σW2 , which means that
the degradation of gain, at the phase-matched condition, is always related to the amplitude of
the underlying perturbation. Computation of the gain spectrum for fiber A is shown on Fig. 5,
all plots sharing the same scale, and confirms the degradation of the gain as Lc increases, es-
pecially for the inner set of phase-matched wavelengths while the (stable) outer set is virtually
unaffected. It is also interesting to note the broadening of the gain bands. In more detail, we
note that the Lc = 0 graph shows lower gain at the outer wavelengths, which is caused both
by the frequency spacing Ω = ωi − ω p = ω p − ωs and monotonically decreasing overlap inte-
gral term for FWM, h. Although this drop in h will diminish gain, its impact is reasonable as
fully-vectorial computation of fmp and fmnpp from simulation data yields values that stay within
the same order of magnitude across the band, as seen on Fig. 6(a), suggesting that the drop in
gain is not very large when going from the inner to the outer set of phase-matched wavelength
pairs. Compared with the standard gain term γ Pp = n2 f pp ω p Pp /c, usually considered constant
√
for all wavelengths (i.e. ωi ≈ ωs ≈ ω p and fispp = fsipp ≈ f pp ), the effective gain, ωi ωs hPp
has to drop also due to the frequency spacing Ω increase when moving away from the pump
as ωs ωi = ω p2 − Ω2 ≤ ω p2 . But as the effect of Lc is introduced, the stability of the outer set of
wavelengths to perturbations changes the balance and the gain at the inner set of phase-matched
wavelengths becomes lower than that of the outer set (to the point of vanishing, whereas the
gain on the outer set drops by less than 15 % for Lc = 10 m) and constitutes the key property
of the stability point. Indeed, this property indicates that a set of phase-matched wavelengths
located very far from the pump and relying on higher-order dispersion (generally meaning
narrow gain bandwidth) can be virtually insensitive to perturbations in the fiber scale, which
L =0m
log 10(P /P0 )
10 c
0
0.4 0.6 0.8 1 1.2 1.4
Lc = 1e−03 m
log 10(P /P0 )
10
0
0.4 0.6 0.8 1 1.2 1.4
L = 1e−01 m
log 10(P /P0 )
10 c
0
0.4 0.6 0.8 1 1.2 1.4
L = 1e+01 m
log 10(P /P0 )
10 c
0
0.4 0.6 0.8 1 1.2 1.4
λ (μm)
Fig. 5. Illustration of the dependance of the expected gain G to spatial noise correlation
length Lc . In these calculations, σΛ = 0.5 %, z = 0.2 m and Pp = 1 kW. It is to be noted
that all graphs share the same scale in order to better appreciate the change in gain as
Lc is increased. One can also notice that the outer set of wavelengths, where the stability
condition occurs, suffers very little gain degradation whereas the inner set vanishes.
11
x 10 10
10
3.5 |fmnpp |
|fmp | 8
10
3
Overlap integrals (m −2 )
6
2.5 10 Ideal gain
G
2 4
10
1.5
2
10
1
0
10 −4 −3 −2 −1 0 1
0.4 0.6 0.8 1 1.2 1.4 10 10 10 10 10 10
λ (μm) σΛ (μm)
(a) (b)
Fig. 6. (a) Values of the overlap integrals fmnpp and fmp as a function of wavelength for a
fixed λ p = 602.6 nm. The four-wave mixing term (blue) is symmetric (in frequency) and
both curves cross at the pump wavelength with the value fmnpp = fmp = f pp = 1/Aeff,p .
Black circles have been added on the curves at the set of wavelengths under study. As we
can see, both factors vary by less than an order of magnitude over the entire domain. The
abscissa represents λm , λn being the conjugate wavelength. (b) Gain at the stable point as a
function of perturbation amplitude σΛ for Lc = 10 m and z = 0.2 m.
appears counter-intuitive. Let us take as an example a standard PCF design operated in the
LP02 mode, such as described in [28]. With d/Λ = 0.9 and nominal pitch Λ0 = 3.27 μ m, such
a design would have the same phase matched wavelengths as the stable set of fiber A for the
same pump wavelength. Comparing peak gain for both the hybrid and standard PCF, calculated
using Eq. (19), we note an improvement of over four orders of magnitude in scale fluctuation
tolerance σΛ , favoring the hybrid PCF for any given Lc . This tolerance criteria is defined where
the logarithm of the expected gain, i.e. log10 (G), falls to half of its ideal value [25]. Indeed, if
compared numerically, this improvement should be in the ratio of ∂ Δk/∂ Λ for each fiber, which
yields 1.075 × 104 at the phase-matched wavelengths. In terms of actual fluctuations amplitude,
the hybrid PCF should be able to have a tolerance in excess of 3.5% in σΛ /Λ for Lc = 10 m and
z = 0.2 m, as seen on Fig. 6(b), which is easily achievable. This gives hope to increase conver-
sion efficiency in FWM applications involving wavelengths far from the pump, regardless of
the presence of small fluctuations in scale, which can otherwise become very demanding from
the fiber manufacturing point of view.
Another feature to notice is the presence of gain where there is normally no phase-matching,
for instance, the bumps around the pump for large values of Lc . There is no analytical way to
write P for non-zero phase-mismatch (Δip + Δsp ), as expressed before, however looking at the
solution of the SDE, Eqs. (17) and (18), can provide some insight on the cause. For small values
of z, there is a probability that the stochastic part of the integrated fluctuation over z, H34,43 B(z)
compensates the uniform phase-mismatch term, G34,43 z, thus giving rise to non-zero net gain.
The likelihood of this happening is higher for small values of phase-mismatch and the gain
generated thus is always going to be relatively small as it relies on the probability of a non-
nominal condition to be satisfied, keeping in mind that the highest probability is that B(z) is
zero. This very well explains the broadening of the gain and the apparition of bumps around
phase-matched conditions but in places where there is always a non-zero phase-mismatch term.
This reasoning also implies that this condition will be a function of σW and Lc , i.e. a fluctuating
process is necessary for those bumps to appear as, otherwise, the phase-mismatch term could
never be compensated. The small feature located around the pump and present on all graphs
is attributed to numerical inaccuracies coming from the quasi-singular nature of the matrix
being exponentiated in Eq. (19), which is related to vanishing values of both Δmp and σW and
interpolation errors in the computation of Δk.
3.2. Impact of pump power

Pump power has both an impact on the magnitude of the gain as well as the width and position
of the generated bands, as seen on Fig. 7 (the noisy feature around the pump is attributed to the
same numerical phenomena mentioned in the previous section). One can notice that the position
of the inner set of phase-matched wavelengths tends to shift away from the pump as the power
increases compared to the outer set of wavelengths, which remains relatively stationary. This
can also be easily visualized by plotting the phase matching curve at different power levels,
as seen on Fig. 8. Furthermore, as expected, the gain increases with increasing pump power.
Taking this into account, it is theoretically possible to compensate for gain degradation by
increasing pump intensity, depending on how important the fluctuations are. One can also note
that the position of the outer set of wavelengths is much more stable as a function of pump
power, since the slope of linear phase- mismatch Δk as a function of detuning is much more
important in that area, given that this solution is a consequence of higher-order dispersion terms.
Thus, a higher parametric shift (i.e. Pp ) is required to significantly move the phase-matching
condition than it is at lower values of Ω. The behavior of gain as a function of pump power
also suggests that a differential gain analysis with respect to Pp (for different fiber lengths)
might help in quantifying the amplitude of stochastic variable W (z). Such quantification may
log 10(P /P0 )
P = 1e+02 W
p
0.5
0
0.4 0.6 0.8 1 1.2 1.4
log 10(P /P0 )
Pp = 5e+02 W
0
0.4 0.6 0.8 1 1.2 1.4
log 10(P /P0 )
Pp = 1e+03 W
5
0
0.4 0.6 0.8 1 1.2 1.4
log 10(P /P0 )
P = 5e+03 W
p
20
0
0.4 0.6 0.8 1 1.2 1.4
λ (μm)
Fig. 7. Illustration of the dependance of the expected gain G to pump power Pp . In these
calculations, σΛ = 0.5 %, z = 0.2 m and Lc = 5 × 10−3 m.
actually prove to be quite useful, in particular when trying to decouple the effects of multiple
contributions to the stochastic part of Δk.
3.3. Impact of pump wavelength

As discussed on Fig. 2, the condition for stability, at a given nominal pitch, occurs for a specific
set of pump and phase-matched wavelengths. However, inaccuracies in the fiber fabrication
1.4
Pp =0e+00 W
1.2
Pp =1e+02 W
Pp =5e+02 W
1
Pp =1e+03 W
λ s,i (μm)
Pp =5e+03 W
0.8
0.6
0.4
0.58 0.59 0.6 0.61 0.62 0.63 0.64 0.65 0.66

λ p (μm)
Fig. 8. Evolution of the mean phase-matching condition (Δsp + Δip )/2 = 0 for different
pump powers. The thicker solid line represents the linear phase-matching condition and
“×” marks the position of the ZDW. The vertical line illustrates the position of the stable
pump λ̃ p .
process could very well throw off the ZDW by a few nanometers, thus compromising both
the intended phase-matching and stability. To characterize the impact of pump wavelength on
stability, the expected gain was computed for a wide set of pump wavelengths, spanning the
range around the stable point. A 3D plot of the gain map is shown on Fig. 9. One can notice
that, generally, the gain falls as one gets further away from the pump λi,s ≈ λ p or the stable point
(λ̃ p , λ̃i,s ) in the phase-matching diagram. Around the pump, where the lower dispersion orders
dominate, ∂ Δk/∂ Λ vanishes as the phase-matched wavelength pair gets closer to the pump,
explaining the increase in gain of the inner set of phase-matched wavelengths as the pump gets
closer to the ZDW. As for the outer set of wavelengths, in a small range of λ p around the stable
point, the gain remains high with a FWHM (in log-space) of about 6 nm. This can be easily
explained by looking at Fig. 2; indeed, as the pump gets closer to the stable point, all three
phase-matching curves tend to pinch together, thus indicating that the global σW falls to zero,
following the phase-matching curve, as one moves closer to the point of stability. Combining
this effect with a careful selection of the phase matched wavelengths and fiber design, one
could simultaneously benefit from a very low sensitivity to nominal pitch as well as some
amount of leeway given by the low sensitivity to the pump wavelength. However, a change in
the generated wavelength would still occur to some extent, but could be used, alongside other
fiber characterization procedures, to determine how far one is from the intended design point.
4. Propagation simulations using noisy longitudinal pitch profiles with different corre-
lation lengths
In this section, we investigate the validation of the theory developed in the first section of this
paper through propagation simulations using the Modified Generalized Nonlinear Schrödinger
Equation (MGNLSE) integrated using a fourth-order Runge-Kutta in the interaction picture
method (RK4IP) [38–40] with adaptive step-size and interpolation of the fiber parameters as
a function of pitch (Aeff , β2 , etc.) This equation being a deterministic differential equation
8
log 10(P /P0 )

4
0
0.61
0.605
0.4
0.6
0.8 0.6
1
1.2 0.595
1.4 λ p (μm)
λ s,i (μm)
Fig. 9. Plots of the expected gain G as a function of pump wavelength λ p and signal/idler
wavelength λs,i for fiber A. In these calculations, σΛ = 0.5 %, z = 0.2 m, Pp = 1kW and
Lc = 5 mm. The FWHM (in log-space) of the stable point gain in terms of λ p is about 6 nm.
(by opposition to stochastic), several noisy pitch profiles Λ(z) have to be generated, propa-
gated through and the average gain taken in order to compare these results with the solution of
Eq. (11). First and foremost, OUP profiles for Λ have to be generated for different correlation
lengths. The OUP presents a normal distribution around its mean and has an autocorrelation
function given by Eq. (16). A method by which to generate such processes has been detailed
in [31] and can be used to generate the noisy longitudinal pitch profiles needed for the simu-
lations. Examples for two different correlation lengths Lc are shown on Fig. 10(a). Given the
typical drawing parameters, response times of the control systems and the physics of the pro-
cess, realistic values for Lc are expected to range from tens of centimeters to several meters or
more [23].
Generally speaking, numerical results obtained using this method are in fairly good agree-
ment with the ones described in the previous section. In particular, the fact that the gain at the
outer set of phase-matched wavelengths is more robust to longitudinal scale variations, thanks
to the stability condition, is corroborated. This confirms the validity of the analytical approach
developed in section 2 and that this original robustness property is preserved even when taking
into account several effects disregarded in the model such as SPM, XPM, Raman, dispersion
(walk-off) and supercontinuum generation [6].
More precisely, the results presented on Fig. 10(b) represent the average gain for several
simulations (10 for Lc = 1 mm, 30 for Lc = 10 cm) using different random Λ(z) profiles having
the same correlation length Lc for λ p = 603 nm, z = 200 mm, a pulse length of 200 ps and a peak
power Pp =1 kW. The choice of the pulse duration and power is intended to be a compromise
between representing a CW pump in a broad spectral window with sufficient spectral resolution
and keeping the number of points for each simulation reasonable (220 points). A noise of one
photon per mode with random phase was used as a seed. As we can see, the general trend
confirms that the gain on the inner bands decreases with increasing Lc while the gain of the
outer bands remains high in spite of the other effects considered.
0.015
Lc = 0.001m Lc = 1e−03 m
Lc = 1m 0
0.01 −20
−40
P (dB)
−60
0.005
−80
(Λ − Λ0 )/Λ0
−100
0 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Lc = 1e−01 m
−0.005 0
−20
−40
P (dB)
−0.01
−60
−80
−0.015 −100
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
z (m) 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
λ (µm)
(a) (b)
Fig. 10. Numerical simulation data. (a) Example of noise profiles used for pitch perturba-
tions in the MGNLSE simulations (the same step-size resolution has been used to generate
both profiles). The standard deviation used for the OUP is 0.5% of the nominal pitch and
the mean is zero. Initial conditions are random and normally distributed with the same stan-
dard deviation as the process and zero mean. (b) MGNLSE simulations for Lc = 1 mm (10
simulations) and Lc = 10 cm (30 simulations). The results shown are the average of the
gain for simulations with different random scale profiles with the same correlation length
Lc with z = 200 mm, a pulse duration of 200 ps and a peak power of Pp =1 kW.
5. Conclusion
In this paper, we demonstrated the existence of a phase-matched FWM condition that is insen-
sitive to scale fluctuations along the fiber length. The method by which to identify this point
of stability has also been discussed. We noticed a number of different trends on the behavior
of the stable point as a function of the design parameters, which can be of use as a starting
point for design optimization. It has also been discussed that some solutions lie in a region
where the pump is in abnormal dispersion regime, which may prove less useful as supercon-
tinuum generation would occur. Propagation simulations were performed and agree fairly well
with the analytical prediction of expected gain developed in this paper. Furthermore, the gen-
eral steps for solving the FWM equations could be extended to other models with moderated
longitudinally varying dispersion (such as dispersion-oscillating fibers [41] or tapered fibers
where autoresonance is observed [42]) to assess the impact of fluctuations on such processes.
Combined with dispersion engineering, this work can help on improving designs for nonlinear
applications previously difficult to achieve due to dispersion instabilities.
Acknowledgments
This work was partly supported by French Ministry of Higher Education and Research, the
Nord-Pas de Calais Regional Council and FEDER through the “Contrat de Projets Etat Région
(CPER) 2007-2013”, the “Campus Intelligence Ambiante” (CIA), and the FLUX Equipex
Project (“Programme Investissement d’Avenir”).

Four-Wave Mixing Stability in Hybrid Photonic Crystal Fibers With Two Zero-Dispersion Wavelengths

Uploaded by

Copyright:

Available Formats

Four-Wave Mixing Stability in Hybrid Photonic Crystal Fibers With Two Zero-Dispersion Wavelengths

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Four-Wave Mixing Stability in Hybrid Photonic Crystal Fibers With Two Zero-Dispersion Wavelengths

Uploaded by

Copyright:

Available Formats

Four-wave mixing stability in hybrid

photonic crystal fibers with two

Benoit Sévigny,1,∗ Olivier Vanvincq,1 Constance Valentin,1 Na Chen,1,2

Abstract: The four-wave mixing process in optical fibers is generally

References and links

2. Four-wave mixing and longitudinal dispersion fluctuations

κsp = 2(γ − 2γsp )Pp − Δkmean − Δkvar (z) (9)

P(z) = exp (Gz + HB(z)) P(0), (18)

0.58 0.59 0.6 0.61 0.62 0.63 0.64 0.65 0.66

3.1. Impact of correlation length

2 λ̃ p = 0.65 μm, Λ = 1.49 μm

λ̃ p = 0.60 μm, Λ = 1.05 μm

0.5 0.6 0.7 0.8 0.9 1 1.1

3.2. Impact of pump power

3.3. Impact of pump wavelength

0.58 0.59 0.6 0.61 0.62 0.63 0.64 0.65 0.66

log 10(P /P0 )

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.