Modulation instabilities in randomly birefringent two-mode optical fibers
Li Jin-Hua1, †, , Ren Hai-Dong2, Pei Shi-Xin1, Cao Zhao-Lou1, Xian Feng-Lin1, ‡,
School of Physics and Optoelectronic Engineering, Nanjing University of Information Science and Technology (NUIST), Nanjing 210044, China
Sunnypol Optoelectronics Company Limited, Hefei 231135, China

 

† Corresponding author. E-mail: lijinhua@nuist.edu.cn

‡ Corresponding author. E-mail: 002642@nuist.edu.cn

Project supported by the Natural Science Foundation of Jiangsu Provincial Universities (Grant No. 14KJB140009), the National Natural Science Foundation of China (Grant No. 11447113), and the Startup Foundation for Introducing Talent of NUIST (Grant No. 2241131301064).

Abstract
Abstract

Modulation instabilities in the randomly birefringent two-mode optical fibers (RB-TMFs) are analyzed in detail by accounting the effects of the differential mode group delay (DMGD) and group velocity dispersion (GVD) ratio between the two modes, both of which are absent in the randomly birefringent single-mode optical fibers (RB-SMFs). New MI characteristics are found in both normal and anomalous dispersion regimes. For the normal dispersion, without DMGD, no MI exists. With DMGD, a completely new MI band is generated as long as the total power is smaller than a critical total power value, named by Pcr, which increases significantly with the increment of DMGD, and reduces dramatically as GVD ratio and power ratio between the two modes increases. For the anomalous dispersion, there is one MI band without DMGD. In the presence of DMGD, the MI gain is reduced generally. On the other hand, there also exists a critical total power (Pcr), which increases (decreases) distinctly with the increment of DMGD (GVD ratio of the two modes) but varies complicatedly with the power ratio between the two modes. Two MI bands are present for total power smaller than Pcr, and the dominant band can be switched between the low and high frequency bands by adjusting the power ratio between the two modes. The MI analysis in this paper is verified by numerical simulation.

1. Introduction

Modulation instability (MI) is a process of exponential growth of weak perturbations imposed on the continuous waves (CWs) as a result of interplay between the dispersion and the nonlinearity.[15] The occurrence of MI is closely related to the soliton existence of the nonlinear system involved, a result of balancing the effects of dispersion and the nonlinearity,[6,7] and is also responsible for the rogue wave formation.[810] MI can be applied for ultrashort pulse generation and supercontinuum generation.[1118] In optics, MI usually occurs for the anomalous dispersion, but under some specific physical settings, MI can also occur in the normal dispersion regime.[1924]

In recent years, space-division multiplexing (SDM) systems based on the few mode and multicore fibers have been proposed as a promising technology to increase the transmission capacity of optical fibers, and has attracted worldwide interests.[2528] Nonlinear effects in these fibers can significantly affect the performance of the SDM systems, and thus have received much attention in recent years.[2932] There are, however, very few studies investigating the MIs in such fibers except for some MI analysis in high-birefringence multimode fibers[33] and two- and three-core fibers.[3436]

In this paper, we give a detailed MI analysis of RB-TMFs by taking the effects of DMGD, GVD ratio, and power ratio between the two modes, which are significant parameters in multimode fibers and has never been thoroughly studied, into consideration in both normal and anomalous dispersion regimes. The main findings are listed below.

The MI analysis in this paper is finally verified by the wave propagation method.

2. Coupling-mode equations

The wave propagation in the randomly birefringent nonlinear multi-mode optical fibers in the weakly coupling regime, where the modes coupling due to various external factors are weak such as the coupling between LP01 and LP11a modes, are derived in detail in Ref. [31]. In this paper, we study the wave propagation in such a TMF, and excite the LP01 and LP11a modes only. The corresponding governing equations derived in Ref. [31] degenerate into

where Aj and Bj (j = x, y) are the slowly varying envelopes of the LP01 and LP11a modes in the j-polarization respectively expressed in a reference moving at the group velocity of LP01 mode. z and t are the propagation distance and delayed time respectively. d0 and d1 denote the average propagation constant and the average group velocity of the LP11a mode in reference to the LP01 mode, of which, d1 is essentially the DMGD parameter between the LP11a and LP01 modes. β2A and β2B are the group velocity dispersion (GVD) coefficients of the LP01 and LP11a modes respectively. γ, fAB, and fBB represent the nonlinear coupling among the spatial modes as defined in Ref. [31].

We normalize Eqs. (1)–(4) into the following equations

with

For a typical TMF with normalized frequency V of 2.405 < V < 3.832, the nonlinear parameters fAB and fBB are calculated respectively in the range of 0.168 < fAB < 0.627 and 0.102 < fBB < 0.964. For T0 = 1 ps, the normalized DMGD parameter δ1 = 0 ∼ 5. The normalized GVD coefficient of the LP01 mode r = β2A/|β2A| = ±1 with the upper and lower signs for the normal and anomalous dispersions respectively. The normalized GVD coefficient of the LP11a mode s = β2B/|β2A| is essentially the GVD ratio of the LP11a and LP01 modes with β2B/β2A = 0.9 ∼ 3.6 for a typical TMF with normalized frequency 2.405 < V < 3.832.[37,38] We emphasize that we do not take into account the twofold degeneracy of the LP11 mode (LP11a and LP11b) in this paper, and assume only one of the two degenerate mode is excited at the input fiber, which will not couple to the other degenerate mode because of the ideally isotropic property of the fiber assumed.[33] The MI of the RB-TMFs are analyzed by solving the normalized equations of Eqs. (5)–(8) in the following sections.

3. MI analysis

The continuous wave (CW) solutions of Eqs. (5)–(8) follow

with

In terms of the standard linear stability analysis, we perturb the exact CW state of Eqs. (11) as

where weak perturbations uk and vk (k = 1, 2) are complex functions of ξ and τ, i.e., uk = uk(ξ, τ) and vk = vk(ξ, τ). The linearization with respect to uk and vk by inserting Eqs. (12)–(15) into Eqs. (5)–(8) gives

Then we assume the perturbations uk and vk in the form of

the nontrivial solutions of Fi and Gi (i = 1,2,3, 4), after one substitutes Eqs. (20)–(23) into the coupled linear equations of Eqs. (16)–(19), lead to the following dispersion relation

where

where P1 = P1x + P1y and P2 = P2x + P2y are the power launched into the LP01 and LP11a modes respectively.

If the solution K of the dispersion relation [Eqs. (24)–(26)] with respect to Ω is complex, the CWs will grow exponentially and the MI gain is defined as

where Im represents the imaginary part of K. It is noted that the average propagation constant difference δ0 does not affect the MI.

With s = r = ±1, δ1 = δ0 = 0, fAB = 0, and fBB = 1, the modes are uncoupled, and the equation models of Eqs. (5)–(8) degenerate into two identical sets of Manakov equations with either set representing an RB-SMF. For either mode, the dispersion relation of Eqs. (24)–(26) degenerate into

where P = P1 = P2 is the total power launched into the either single-mode fiber.

For the general case, the dispersion relation of Eqs. (24)–(27) is quite involved, and can be only treated numerically. The MI analysis for the general case is studied in detail in the following sections.

4. MI results

As a benchmark, we first study the MI in the RB-SMFs. From the dispersion relation of Eq. (29), we see that there is no MI in the normal dispersion regime (r = 1), and MI only exists in the anomalous dispersion regime (r = −1) with the gain

As expected, the instability gain increases with P as shown in Fig. 1. Note that the gain spectra for the negative and positive perturbation frequencies are totally symmetric, and thus only the MI spectrum for the positive perturbation frequency is shown in Fig. 1.

Fig. 1. Variation of MI spectra with total power in an RB-SMF.

We next show MI results in the RB-TMFs in both normal and anomalous dispersion regimes. As in the RB-SMFs (Fig. 1), our results show that the gain spectra in the RB-TMFs for negative and positive perturbation frequencies are also symmetric. Therefore, we only show the MI spectra with the positive perturbation frequency part in the following section.

4.1. Normal dispersion regime

We first consider the normal dispersion. The DMGD plays a crucial role. In the absence of DMGD, as in the RB-SMFs, no MI exists in the RB-TMFs. In the presence of DMGD, MI can be generated in the RB-TMFs.

Figure 2 shows the variation of MI gain with total power P for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P1/P2 = 1, and P = P1 + P2. As the total power increases, this MI band enhances, reduces and finally vanishes after a critical total power. This result is understandable. Physically, DMGD tends to split the pulse, while the nonlinear effects are to suppress the pulse splitting. When the nonlinear effects are strong enough, the effect of DMGD should be completely suppressed, and thus the MI should vanish as the total power exceeds a critical total power.

Fig. 2. Variation of MI spectra with total power in an RB-TMF for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P1/P2 = 1, and P = P1 + P2.

Figure 3 shows the variation of MI gain with DMGD for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, P = 1.5, and P1/P2 = 1. For a given total power, no MI is present for small δ1. As δ1 increases, MI appears, enhances distinctively and eventually saturates with the modulation frequency shifting to the higher frequency at the same time. We remark that the role of DMGD in RB-TMFs is similar with the polarization mode dispersion (PMD) in highly birefringent (Hi-Bi) SMFs,[39,40] which is expected as both the DMGD and PMD denote the group delay only that DMGD denotes the group delay between two distinct spatial modes but PMD denotes the group delay between two polarization components of a given mode.

Fig. 3. Variation of MI spectra with DMGD in an RB-TMF for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, P = 1.5, and P1/P2 = 1.

Figure 4 shows the variation of MI gain with power ratio (P1/P) for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, P = 1.5, and δ1 = 2. When the power is totally launched in the LP11a or the LP01 mode, namely P1/P = 0 or 1, no MI exists for any values of DMGD. As power ratio increases from 0, the MI appears and the MI band broadens with its gain increasing and approaching to maximum at P1/P = 0.5, i.e., equal power distribution between the two modes. Across P1/P = 0.5, the MI of this band changes in the opposite way. We emphasize that only the power distribution between the two modes can affect the MI spectra, and the power distribution in the x- and y-polarizations of each mode does not affect the MI spectra in the RB-TMFs.

Fig. 4. Variation of MI spectra with the power ratio between the LP01 and LP11a modes in an RB-TMF for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, and P = 1.5.

Figure 5 shows the variation of MI gain with GVD ratio (s = β2B/|β2A|) for r = 1, fAB = 0.53, fBB = 0.72, δ1 = 2, P = 1.5, and P1/P2 = 1. Generally, GVD ratio between the LP11a and LP01 modes reduces the instability gain of the MI band, and shrinks the modulation frequency range at the same time. Such effects are dramatic for relatively small values of GVD ratio as shown in Fig. 5.

Fig. 5. Variation of MI spectra with the GVD ratio between the LP11a and LP01 modes in a RB-TMF for r = 1, fAB = 0.53, fBB = 0.72, δ1 = 2, P = 1.5, and P1/P2 = 1.

The dependence of the critical total power (Pcr) for the vanishing of the MI in Fig. 2 on the DMGD (δ1), power ratio (P1/P) and GVD ratio (s) is shown in Fig. 6. This critical total power increases significantly as DMGD increases (Fig. 6(a)), and decreases in a dramatic manner with the increment in the GVD ratio (Fig. 6(c)). The dependence of Pcr on the power ratio P1/P is relatively weak except at P1/P = 0 and P1/P = 1, where the critical power vanishes, namely Pcr = 0.

Fig. 6. The dependence of the critical total power on (a) DMGD for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, and P1/P2 = 1; (b) power ratio for r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, and δ1 = 2; (c) GVD ratio for r = 1, fAB = 0.53, fBB = 0.72, δ1 = 2, and P1/P2 = 1.
4.2. Anomalous dispersion regime

We next consider the anomalous dispersion. It is practical for the value of DMGD being zero by special fiber design.[37,38] Figure 7 shows the variation of MI with the total power in an RB-TMF in the absence of DMGD for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 0, and P1/P2 = 2/3. Without the effect of DMGD, there is only one MI band. Compared to the corresponding RB-SMFs (Fig. 1), MI gain in the RB-TMFs reduces and the MI band also shrinks. Quantitatively, for the given parameters in Fig. 7, the gain and the instability bandwidth are decreased by 20% and 15% respectively when a same total power that is launched into an RB-SMF is now launched into an RB-TMF.

Fig. 7. Variation of MI spectra with the power ratio between the LP01 and LP11a modes in an RB-TMF for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 0, and P1/P2 = 2/3.

In the presence of DMGD, MI spectra is distinctly different as shown in Fig. 8, where the parameters are the same as those in Fig. 7 but δ1 = 2. There exists a critical total power. As power increases from zero, two MI bands appear, enhance, and finally merge into one band across this critical total power. This is also understandable in terms of the physics that DMGD tends to split the pulse and the nonlinear effect tends to suppress the pulse splitting. Therefore, the effect of DMGD can be suppressed as the power increases and the two MI bands should merge into one when power is large enough for a given DMGD. On the other hand, in comparison with the MI in RB-SMFs (Fig. 1), the MI gain in RB-TMFs is decreased further in the presence of DMGD. For δ1 = 2, the gain in RB-TMFs shown in Fig. 8 is decreased by 40% compared with that in Fig. 1.

Fig. 8. Variation of MI spectra with the power ratio between the LP01 and LP11a modes in an RB-TMF for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, and P1/P2 = 2/3.

Figure 9 shows the variation of MI with DMGD for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, P = 1, and P1/P2 = 2/3. Generally, DMGD reduces the MI gain. On the other hand, a new MI band in the low frequency is generated after a critical DMGD, and the original MI band rapidly shifts to the high frequency as DMGD increases. The effect of DMGD is also similar with that of PMD in the anomalous dispersion regime. The same arguments on effect of DMGD in the normal dispersion regime can be also applied here.

Fig. 9. Variation of MI spectra with DMGD in an RB-TMF for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, P = 1, and P1 = P2 = 2/3.

Figure 10 shows the variation of MI with the power ratio (P1/P) between the two modes for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, and P = 1. At P1/P = 0 (or 1), i.e., the total power is totally launched in LP11a (or LP01) mode, there is only one MI band. As power ratio (P1/P) deviates from 0 (or 1), a new high frequency MI band appears, and the gains of both MI bands changes dramatically with P1/P. Specifically, as the power ratio increases from 0, the MI gain of the low frequency band reduces, reaches its minimum value at around P1/P = 0.4, and then rapidly enhances. In contrast, the high frequency band enhances first, begins to reduce after about P1/P = 0.4 and finally vanishes at P1/P = 1. From Fig. 10, we also find that the dominant MI band can be switched between the two MI bands. Quantitatively, the high (low) frequency band dominates the MI for P1/P = 0.4 ∼ 0.6 (P1/P < 0.4 and P1/P > 0.6).

Fig. 10. Variation of MI spectra with the power ratio between the LP01 and LP11a modes in an RB-TMF for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, and P = 1.

Figure 11 shows the variation of MI with GVD ratio for r = −1, fAB = 0.53, fBB = 0.72, δ1 = 2, P = 1, and P1/P2 = 2/3. The effect of the GVD ratio is to shift the two MI bands to the low frequency range as the magnitude of GVD ratio increases. At the same time, GVD ratio enhances the gain of the high frequency band as the increment in the magnitude of GVD ratio, but imposes slight effect on the gain of the low frequency band.

Fig. 11. Variation of MI spectra with the GVD ratio between the LP11a and LP01 modes in an RB-TMF for r = −1, fAB = 0.53, fBB = 0.72, δ1 = 2, P = 1, and P1/P2 = 2/3.

Figure 12 shows the dependence of the critical total power in Fig. 8 on DMGD, power ratio and GVD ratio respectively. The critical total power increases monotonically as DMGD increases, and reduces monotonically as the increment of GVD ratio. However, the critical total power varies in a complicate way with the power ratio. As for the normal dispersion, the critical total power is also zero, i.e., Pcr = 0, at P1/P = 0 or P1/P = 1. As P1/P increases from 0, Pcr is nonzero and experiences a series of reduction and increment. In general, the dependence of the critical power on the DMGD is most dramatic (compare Fig. 12(a) with Fig. 12(b) and Fig. 12(c)).

Fig. 12. The dependence of the critical total power on (a) DMGD for r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, and P1/P2 = 1; (b) power ratio for r = −1, s = −1.2, δ1 = 2, fAB = 0.53, and fBB = 0.72; (c) GVD ratio for r = −1, δ1 = 2, fAB = 0.53, fBB = 0.72, and P1/P2 = 1.

According to above analysis to the effect of DMGD (Figs. 7 and 9), power ratio (Fig. 10), and GVD ratio (Fig. 11), the MI in RB-TMFs, in comparison with RB-SMFs (Fig. 1), has the following distinctive characteristics: first, the MI gain is decreased, the decrement depends on the specific fiber parameters, and the minimum gain can be only 1/4 of that in an RB-SMF for a given total power; second, a new MI band in the high frequency can be generated in the presence of DMGD.

5. Numerical simulation

To verify the MI analysis above, we solve the four coupled nonlinear Schrödinger equations of Eqs. (5)–(8) numerically by launching CWs directly into each polarization of the LP01 and LP11a modes, i.e.,

with j = x, y, where noise = 0.001 × (1 − 2×rand) with ‘rand’ as the uniformly distributed random numbers between [0, 1] generated by MATLAB. We employ the pseudospectral method in the space domain and the fourth-order Runge–Kutta method with the adaptive step control in the time domain. In terms of the MI analysis, the CWs should first evolve into a train of periodic pulses, and the frequency of such a periodic pulse train should be equal to the dominating modulation frequency, which has the maximum gain from the MI analysis. As further propagation, the other frequency components in the noises, which are in the ranges of modulation frequency that generate the MI, should begin to play important roles, which will interact with the dominating modulation frequency component that has the maximum gain. Therefore, the waveform starts to lose its periodicity and becomes increasingly irregular and spiky as further propagation.

We first consider the case of the normal dispersion. The parameters used in the simulation are taken as follows: r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P = P1 + P2 = 1.5, P1 = P2 = 0.75, P1x/P1 = 0.4, P1y/P1 = 0.6, P2x/P2 = 0.7, and P2y/P2 = 0.3. Figure 13 shows the CW propagation in an RB-TMF with above parameters. At about ξ = 30, a train of periodic waves appear with the period of 4.59, which corresponds to the frequency of 0.22. The calculated optimum modulation frequency generating the maximum gain for above parameters selected as shown by the green dash line in Fig. 4(b) is 0.22, in quite good agreement with the results of numerical simulation. As further propagation, the periodic waves vanish and become very irregular, which justify the existence of multi modulation frequency components as predicted in the MI analysis.

Fig. 13. (a) The evolution of the MI from a CW input calculated for the normal dispersion with r = 1, s = 1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P = P1 + P2 = 1.5, P1 = 0.75, P2 = 0.75, P1x/P1 = 0.4, P1y/P1 = 0.6, P2x/P2 = 0.7, and P2y/P2 = 0.3; (b) the waves in the x- and y-polarizations of each mode at ξ = 0 (red line) and ξ = 30 (blue line).

We then consider case of the anomalous dispersion. Figure 14 shows the CW propagation in an RB-TMF for taking the following parameters: r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P = P1 + P2 = 1, P1 = 0.9, P2 = 0.1, P1x/P1 = 0.4, P1y/P1 = 0.6, P2x/P2 = 0.1, and P2y/P2 = 0.9. At about ξ = 13, a train of the periodic waves appears with the period of 5.08, corresponding to the frequency of 0.20, which agrees quite well with the calculated optimum modulation frequency of 0.20 for the parameters selected as shown with the blue dash line in Fig. 10(b). As the wave propagation distance increases, the periodic waves become very spiky, which also verify the existence of the multi-modulation frequency components.

Fig. 14. (a) The evolution of MI from a CW input calculated for the anomalous dispersion with r = −1, s = −1.2, fAB = 0.53, fBB = 0.72, δ1 = 2, P1 = 0.9, P2 = 0.1, P1x/P1 = 0.4, P1y/P1 = 0.6, P2x/P2 = 0.1, and P2y/P2 = 0.9; (b) the waves in the x- and y-polarizations of each mode at ξ = 0 (red line) and ξ = 13 (blue line).

We remark that the selection of random number as the noise is valid.[41,42] On the one hand, the bandwidth of the noise generated by the random numbers has been wide enough to cover all the possible modulation frequencies for the parameters selected in our numerical simulations. The frequencies of the noise in the simulation have covered a range of about ±30 around the optical carrier frequency, while the modulation frequencies calculated in the MI analysis are in a range of about ±0.5 around the optical carrier frequency. On the other hand, the results of our numerical simulation fit well with the MI analysis as discussed above.

6. Discussions and conclusion

In this paper, we studied the MIs in RB-TMFs systematically by solving four coupled nonlinear Schrödinger equations with DMGD and GVD ratio between the two modes. The effects of power ratio between the two modes are also studied. New MI characteristics are observed in both normal and anomalous dispersion regimes.

In the normal dispersion regime, MI, which is absent in RB-SMFs, can occur in the RB-TMFs for any total power smaller than a critical value under the effect of DMGD. The maximum gain is obtained for equal power distribution between the two modes, while the power distribution in the two polarization directions in each mode does not affect the MI. The effect of GVD ratio is to reduce the MI gain and shrink the MI band.

In the anomalous dispersion regime, in comparison with the RB-SMFs, the MI is generally reduced. On the other hand, there also exists a critical total power in the presence of DMGD. Two MI bands are present for the total power smaller than this critical total power, and the dominant MI band can be switched between the two bands by adjusting the power distribution in the two modes. GVD shifts the modulation frequencies of the two bands to the lower frequency. On the other hand, the GVD ratio enhances the gain of the high frequency band, but has little effect on the MI gain of the low frequency band.

MIs in the RB-TMFs are much richer than those in the RB-SMFs. Our results should be quite useful for the evaluation of the nonlinear effects in SDM systems and other applications based on multimode fibers. The effects of the higher order dispersion and higher order nonlinearity in RB-TMFs will be explored in our future studies.

Reference
1Canabarro ASantos Bde Lima Bernardo BMoura A LSoares W Cde Lima EGléria ILyra M L 2016 Phys. Rev. 93 023834
2Zhang JDai XZhang LXiang YLi Y2015J. Opt. Soc. Am. B321
3Wabnitz SWetzel B 2014 Phys. Lett. 378 2750
4Frisquet BKibler BFatome JMorin PBaronio FConforti MMillot GWabnitz S 2015 Phys. Rev. 92 053854
5Zhang LDai XXiang YWen S 2015 Appl. Phys. 121 465
6Agrawal G P 1987 Phys. Rev. Lett. 59 880
7Xu Y ASavescu MKhan K RMahmood M FBiswas ABelic M 2016 Opt. Laser Technol. 77 177
8Chen S HBaronio FSoto-Crespo J MGrelu PConforti MWabnitz S 2015 Phys. Rev. 92 033847
9Walczak PRandoux SSuret P 2015 Phys. Rev. Lett. 114 143903
10Li J HChan H NChiang K SChow K W2015Commun. Nonlinear Sci. Numer. Simul.2828
11Yu MMcKinistrie C JAgrawal G P 1998 J. Opt. Soc. Am. 15 607
12Korobko D AOkhntnikov O GZloltovskii I O 2013 J. Opt. Soc. Am. 30 2377
13Tang D YGuo JSong Y FLi LZhao L MShen D Y 2014 Opt. Fiber Technol. 20 610
14Wei X MZhang CXu S HZhang Z MTsia K KWong K K Y2014IEEE J. Sel. Top. Quantum Electron.207600207
15Nithyanandan KRaja R V JPorsezian KUthayakumar T 2013 Opt. Fiber Technol. 19 348
16Mckinstrie C JRadic SXie C 2003 Opt. Express 11 2619
17Wang H LYang A JLeng Y X 2013 Chin. Phys. 22 074208
18Zhong X QXiang A P 2010 Chin. Phys. Lett. 27 014203
19Zhong X QCheng KXiang A P 2013 Chin. Phys. 22 034205
20Armaroli ATrillo S 2014 J. Opt. Soc. Am. 31 551
21TchofoDindaand PPorsezian K 2010 J. Opt. Soc. Am. 27 1143
22Tanemura TKikuchi K 2003 J. Opt. Soc. Am. 20 2502
23Armaroli ATrillo S 2014 J. Opt. Soc. Am. 31 551
24Millot GWabnitz S 2014 J. Opt. Soc. Am. 31 2754
25Richardson D JFini J MNelson L E 2013 Nat. Photon. 7 354
26Saitoh KMatsuo S 2016 J. Lightwave Technol. 34 55
27Radosavljevic ADanicic APetrovic JMaluckov AHaziewski L2016J. Opt. Soc. Am. B322520
28Sillard PMolin DBigot-Astruc MAmezcua-Correa Ade Jongh KAchten F 2016 J. Lightwave Technol. 34 1672
29Rademacher GPetermann K 2016 J. Lightwave Technol. 34 2280
30Hadzievski LMaluckov ARubenchik A MTuritsyn S2015Light: Science & Applications4e314
31Mumtaz SEssiambre R JAgrawal G P 2013 J. Lightwave Technol. 31 398
32Antonelli CShtaif MMecozzi A 2016 J. Lightwave Technol. 34 36
33Guasoni M 2015 Phys. Rev. 92 033849
34Li J HChiang K SChow K W 2011 J. Opt. Soc. Am. 28 1693
35Shafeeque Ali A KNithyanandan KPorsezian K 2015 Phys. Lett. 379 223
36Shafeeque Ali A KNithyanandan KPorsezian K 2016 J. Opt. 18 035102
37Chiang K S1986J. Lightwave Technol.LT-4980
38Shibata NOhashi MMaruyama RKuwaki N 2015 Opt. Rev. 22 65
39Rothenberg J E 1990 Phys. Rev. 42 682
40Seve EMillot G2000Phys. Rev. E613139
41Ding ELuh KKutz J N 2011 J. Phys. B: At. Mol. Opt. Phys. 44 065401
42Li J HChiang K SMalomed B AChow K W 2012 J. Phys. B: At. Mol. Opt. Phys. 45 165404
43Rogers CMalomed B ALi J HChow K W2012J. Phys. Soc. Jpn.81094005