Abstract
Temporal optical solitons have been the subject of intense research due to their intriguing physics and applications in ultrafast optics and supercontinuum generation. Conventional bright optical solitons result from the interaction of anomalous groupvelocity dispersion and selfphase modulation. Here we experimentally demonstrate a class of bright soliton arising purely from the interaction of negative fourthorder dispersion and selfphase modulation, which can occur even for normal groupvelocity dispersion. We provide experimental and numerical evidence of shapepreserving propagation and flat temporal phase for the fundamental purequartic soliton and periodically modulated propagation for the higherorder purequartic solitons. We derive the approximate shape of the fundamental purequartic soliton and discover that is surprisingly Gaussian, exhibiting excellent agreement with our experimental observations. Our discovery, enabled by precise dispersion engineering, could find applications in communications, frequency combs and ultrafast lasers.
Introduction
The fascinating phenomenon of optical solitons, solitary optical waves that propagate in a particlelike fashion over long distances^{1}, has been the subject of intense research during the last decades due to its major role in breakthrough applications such as mode locking^{2}, frequency combs^{3,4} and supercontinuum generation^{5,6} among others^{7,8,9}. Temporal solitons in optical media^{10,11}, as studied to date, arise from the balance of the phase shift due to anomalous quadratic groupvelocity dispersion (GVD), that is, negative GVD parameter β_{2}=(∂^{2}k/∂ω^{2})<0, and the selfphase modulation (SPM) due to the nonlinear Kerr effect.
In practice, higherorder nonlinear and dispersive effects often perturb this behaviour. In silicon (semiconductor) waveguides the most significant higherorder nonlinearities are associated with free carriers (FCs) generated by twophoton absorption (TPA)^{12,13}, which have hampered the observation of solitonbased effects in this material. Recently, some of us achieved higherorder soliton compression of picosecond pulses in silicon^{14} by using a dispersion engineered photonic crystal waveguide (PhCwg). Turning to higherorder dispersive effects, the presence of third order dispersion (TOD; β_{3}=∂^{3}k/∂ω^{3}) leads to soliton instability^{15}, whereas positive fourthorder dispersion (FOD; β_{4}=(∂^{4}k/∂ω^{4})>0), can give rise to radiation at specific frequencies^{16}. In the presence of negative FOD (β_{4}<0), as was shown by a series of theoretical works in optical fibres^{17,18,19,20,21}, solitons can be stable. These studies^{17,18,19,20,21} led to the concept of quartic solitons^{22}, solitary pulses resulting from the interaction of anomalous GVD and SPM but modified by the presence of FOD.
In the following we report the experimental discovery and physical description of an entirely new class of solitons originating purely from the interaction of negative FOD and SPM, as conceptually depicted in Fig. 1a, which can occur even when the GVD vanishes or is normal. Since they arise just from quartic dispersion and SPM, and to distinguish them from the solitary waves studied earlier^{17,18,19,20,21,22}, we propose the name of purequartic soliton for this new class of solitary wave. This experimental discovery is enabled by the unique dispersion properties of PhCwgs, which allow us to combine very large negative β_{4} with small positive β_{2} and negligible β_{3} for the wavelength under study. Though our work directly pertains to pulse propagation in guided wave structures, the same ideas apply to spatial solitons, particularly to subdiffractive matterwave solitons in regimes where the fourthorder diffraction is the dominant diffractive effect^{23,24}. Furthermore, it was shown that in Ti:sapphire lasers FOD ultimately limits the minimum pulse duration in cavities with nearzero GVD and TOD^{25,26,27}, hinting that the pulse shaping behaviour in the laser cavity for ultrashort pulses (below 10 fs) arises from the balance of SPM and FOD^{28}.
Results
Experimental signatures of purequartic solitons
In our experiments we performed time and phaseresolved propagation measurements on the sample using a frequencyresolved electrical gating (FREG) apparatus, depicted in Fig. 1b, which can be modelled using a generalized nonlinear Schrodinger equation (GNLSE). We show shapepreserving propagation and flat temporal phase for fundamental purequartic solitons and temporal compression and convex nonlinear phase for higherorder purequartic solitons. In spite of maintaining these wellknown signatures of solitonlike behaviour, we show that purequartic solitons present remarkably different properties than solitons studied to date^{10,11,17,18,19,20,21}. Importantly, the energy scaling of purequartic solitons suggests much higher energies for ultrashort pulses, which may inspire a new wave of soliton laser developments. Finally, we derive the approximate shape of fundamental purequartic solitons and find that it is close to a Gaussian, which is remarkable given that the solitary wave solutions found to date are always a function of hyperbolic secant^{10,18,20}.
For demonstrating the existence of purequartic solitons we used the 396μmlong dispersion engineered slowlight silicon PhCwg^{29} shown in Fig. 1c (see the Methods section). Figure 1d shows the waveguide dispersion measured using an interferometric technique^{30}. At the pulse central wavelength, 1,550 nm, the PhCwg has a measured group index of n_{g}=30, a GVD of β_{2}=+1 ps^{2} mm^{−1}, corresponding to normal dispersion, a TOD of β_{3}=+0.02 ps^{3} mm^{−1}, and a FOD of β_{4}=−2.2 ps^{4} mm^{−1}. Note that β_{4} is negative in a 6nm wavelength range.
To perform a complete temporal and spectral characterization of the subpicojoule ultrafast nonlinear dynamics in the waveguide we used a FREG apparatus^{31} in a crosscorrelation configuration, as schematically depicted in Fig. 1b (see the Methods section). This setup provides a series of spectrograms, that is, the gated optical power versus delay, for varying input powers. From these spectrograms we then extract the optical pulses’ electric field envelope and phase using a numerical algorithm^{32}. Figure 2 shows the measured intensity (red dashed lines) and phase (black dashed) at the output of the PhCwg, when injecting 1.3 ps Gaussian pulses (fullwidth at halfmaximum, FWHM) at 1,550 nm with different input peak powers, P_{0}. Figure 2a shows the frequency domain and Fig. 2b shows the temporal domain. The physical length scales of the dispersion orders for this pulse duration are: L_{GVD}=T_{0}^{2}/β_{2}=0.615 mm, L_{TOD}=T_{0}^{3}/β_{3}=22.6 mm, L_{FOD}=T_{0}^{4}/β_{4}=0.168 mm, with T_{0}=FWHM/1.665 for Gaussian pulses. These length scales indicate that FOD is dominant, with the total length of the sample being L=2.4.L_{FOD} and the GVD length being L_{GVD}=3.66.L_{FOD}. TOD is negligible in this sample for our pulses.
To understand the origin of the experimental observations in Fig. 2 we employ a GNLSE model to describe the propagation in the silicon PhCwg:
Here A(z,t) is the slowly varying amplitude of the pulse, α_{l,eff} denotes the linear loss, γ_{eff} and α_{TPA,eff} are the effective nonlinear Kerr and TPA parameters, respectively; n_{FC,eff} and σ_{eff} represent the freecarrier dispersion (FCD) and the freecarrier absorption effective parameters for a freecarrier concentration of N_{c}. Since we use a slowlight PhCwg, the effective coefficients vary with the slowdown factor S=n_{g}/n_{0} (ref. 13). We use the measured envelope amplitude of the input pulse as the input to our GNLSE model with the sample parameters detailed in the Methods section. As shown in Fig. 2, the model (solid lines) agrees well with the experimental data (dashed lines) in both frequency and time.
We first focus on the time domain results of Fig. 2b. At the low coupled power of 0.07 W, nonlinear effects can be neglected and we simply observe small temporal broadening, mainly due to quartic dispersion. The different signs of β_{2} and β_{4} counteract each other to some degree, leading to a modest temporal broadening at the output of this short PhCwg (from 1.3 to 1.4 ps). We have verified, by running the GNLSE for longer lengths that the pulse width keeps increasing with the propagation distance in the linear case. Increasing the input power up to 0.7 W, where the nonlinear length L_{NL}=1/(γ_{eff}P_{0}) becomes comparable to L_{FOD}, the pulse preserves its initial shape and duration, as illustrated by the good matching between the measured output intensity (dashed red line) and the normalized input intensity (green solid line). Furthermore, the temporal phase across the pulse duration is nearly flat. These are two signatures of fundamental soliton behaviour^{10}. Simple estimates, confirmed with GNLSE simulations, show that the loss due to TPA at this power level is quite small and that free carriers do not yet play a role. At 2.5 W, where L_{NL}<<L_{FOD,} the phase becomes convex due to the stronger nonlinear Kerr effect and the main peak of the pulse narrows, temporal signatures of a higherorder soliton. At even higher powers, 4.5 W, the main peak of the pulse narrows even more, corresponding to a higherorder soliton with a highersoliton number^{11}. In addition, a long tail develops towards the leading edge of the pulse. We provide an explanation for this effect below.
Next we examine the frequency domain in Fig. 2a. At 0.07 W, since the nonlinearities are negligible and the pulse spectrum is not affected by the dispersion, the pulse spectral shape is maintained. At 0.7 W the pulse preserves its initial spectral shape, again consistent with fundamental soliton behaviour in the spectral domain. At higher powers (2.5 and 4.5 W) the pulse experiences spectral broadening and splits into two peaks, spectral signatures of higherorder solitons. The observed blue shift and asymmetry are associated mainly with FCD. Note that the oscillations in the measured spectra simply correspond to Fabry–Perot reflections at the input and output facets of the PhCwg and disorder in the periodic media^{33}.
Whereas we previously reported shapepreserving fundamental solitons and higherorder soliton compression in silicon^{14}, such behaviour was unforeseen for the normal GVD here. By setting β_{2}=0 in our numerical model we find that the signatures of soliton behaviour are maintained: the shape is preserved and the phase is flat for the fundamental soliton at 0.7 W, and at 2.5 and 4.5 W, the higherorder solitons undergo nonlinear temporal narrowing. This demonstrates that GVD is not important in this system and, since we established that TOD is also negligible, that the soliton behaviour stems purely from the interaction of FOD and SPM. Furthermore, we have verified that the long tail at the leading edge observed at high powers (Fig. 2b), as well as the selfacceleration of the pulse, originate from the interaction of negative FOD and FCD. The FCD generates additional blue components (Fig. 2a) and the negative FOD makes them travel faster than the red components of the pulse, analogous to our earlier results with negative GVD^{13,14,34}.
These observations at the output of the silicon PhCwg suggest the existence of a new type of soliton: purequartic solitons. We use the term soliton here to refer to solitary optical waves that propagate essentially unperturbed over long distances, not to exact localized solutions of integrable nonlinear differential equations^{35}.
As expected in a silicon system at 1,550 nm, purequartic solitons are strongly perturbed by TPA and FC as we just described, and thus the measured behaviour differs from the simple case with just SPM and FOD. Therefore, to elucidate the dynamics of purequartic solitons in the absence of higherorder nonlinearities we next numerically study the propagation of picosecond pulses along the PhCwg neglecting all effects but SPM and FOD.
Propagation behaviour of purequartic solitons
Figure 3a,b depict the propagation dynamics of undistorted purequartic solitons, that is, in the presence of SPM and FOD only. Such a system is governed by the biharmonic nonlinear Schrodinger equation
In our simulations we consider two different power levels: fundamental purequartic solitons occur at moderate powers (Fig. 3a), whereas at high powers higherorder purequartic solitons result (Fig. 3b).
The simulations in Fig. 3a show shapepreserving pulse propagation in time and frequency over five quartic dispersion lengths L_{FOD} for a fundamental purequartic soliton. The very slight increase in the maximum intensity at t=0 corresponds to the pulse adapting itself from the standard Gaussian pulse used as an input to the model, to the soliton form whose approximate shape we provide in the next section. The output, represented by the blue curve to the right of the propagation plot, shows that the pulse maintains essentially the same amplitude, shape (Gaussian), and duration (1.3 ps) as the input pulse. Importantly, the temporal phase at the output, represented by the black solid line, is flat across the duration of the pulse.
Figure 3b reveals a higherorder purequartic soliton, with the pulse experiencing periodic recurrent propagation. In time the pulse undergoes compression and then periodically returns to its initial shape. In frequency the pulse splits into two and then recombines to recover its initial spectral shape after the same period. At the maximum compression point, the pulse reaches the minimum duration of 0.54 ps, a compression factor of 2.4 compared with the initial pulse duration, with a peak intensity of two times that of the initial pulse. Our simulations show the compression factor of the purequartic soliton roughly follows the same trend as conventional solitons^{11}. Specifically, larger intensities lead to higher compression factors and to compression occurring at an earlier spatial position along the waveguide. Crucially, while the trends are superficially similar to the behaviour of conventional solitons, the different scaling of SPM and FOD with pulse length suggests that the wellknown definitions of soliton number and soliton period will not be appropriate for purequartic solitons; further studies are underway to derive the appropriate parameters and physical scaling laws for these field structures.
To understand why the experimental observations of purequartic solitons in Fig. 2 differ from the numerical results of the undistorted system in Fig. 3a,b, we simulate the propagation along the PhCwg including all the effects in the real system indicated in equation 1. The outputs shown in Fig. 3c,d match our experimental measurements at P_{0}=0.7 W and P_{0}=4.5 W. Figure 3c shows that the signatures of the fundamental purequartic soliton remain in realistic simulations: the pulse maintains its shape and width, and the phase at the output remains almost flat. However, the intensity of the pulse decreases due, predominantly, to the linear loss in the slowlight waveguide ∼70 dB cm^{−1}. Since the intensity decreases as the pulse propagates, the linear FOD will eventually dominate over the SPM for longer distances, leading to temporal broadening of the fundamental purequartic soliton. The higherorder purequartic soliton in the realistic scenario of Fig. 3d differs considerably from Fig. 3b. The TPA clamps the intensity in the waveguide from the early stages of propagation. The FCD introduces blue components that lead to the selfacceleration of the pulse and an asymmetry, and the freecarrier absorption induces absorption on the trailing edge. These effects of TPA and FCs on the propagation of higherorder purequartic solitons in silicon are analogous to the effects of FCs on conventional solitons^{14}.
Approximate solution for the fundamental purequartic soliton
We now derive an analytic expression for the fundamental purequartic soliton. The experimental observations and numerical simulations indicate that the central part of fundamental purequartic solitons appears to be Gaussian. Assuming this shape, we look for an approximate solution to equation (2) in two separate ways for verification purposes: using the variational principle and looking for a local approximation near the centre. The complete derivations for the variational and local approximate solutions are described in Supplementary Note 1 and Supplementary Note 2, respectively.
In both cases, after a simple dimensional analysis, we take the central part of the purequartic soliton to be of the form
where, since β_{4}<0, we have written for convenience, and μ and ν are free parameters. The variational approach gives , whereas the local approximation gives , v=1. Thus, these approaches predict pulse widths which differ only by a factor or by <10%. The fact that these different approximations give very similar results reinforces our confidence in them. Based on these results, the argument of Akhmediev and Karlsson^{36} suggests that purequartic solitons do not lose energy due to linear radiation. []ally, taking the Fourier transform of the righthand side equation (3) in time and position leads to a straight line in an ωk diagram. Since this straight line does not intersect the linear dispersion relation of the medium, the soliton cannot lose energy to dispersive waves.
To test the validity of this analytic approximate solution we numerically solve equation (2) with the β_{4} and γ_{eff} of our sample (see Methods section) and obtain the output of the system at the power level corresponding to a fundamental purequartic soliton for a propagation length L=30·L_{FOD}. This long propagation distance ensures convergence of the pulse evolution. The results of this numerical experiment for three different pulse shapes: a Gaussian, a hyperbolic secant (sech), and a super Gaussian of order four, are depicted by a solid blue curve in Fig. 4a–c, respectively. Importantly, the hyperbolic secant and super Gaussian inputs (black solid curve in Fig. 4b,c, respectively) evolve into the solitary wave Gaussian shape, constituting an additional signature of solitonlike behaviour and proving that this type of soliton acts as an attractor. These results are overlapped with the variational approximation (dashed red curve) and the local approximation (green dashed curve), using the same parameters. The agreement between the numerical and the variational solution in the central part of the pulse is remarkable. The local approximation deviates only slightly. In addition we numerically find that μ≈0.63, again very close to the variational result. In the background of Fig. 4a, we show the measured pulse at the output of the chip at P_{0}=0.7 W (cyan dotdashed curve), which matches variational solution perfectly. The wings observed in the numerical solution relate to the fact that the phase shift profile of the FOD has a quartic dependence with time, whereas the SPM varies quadratically, as illustrated in Fig. 4d. This allows both phase shift profiles to perfectly counterbalance each other close to the centre of the pulse, but deviate from each other at the edges. This fact is not captured in the approximate analytic solution since a perfect balance between FOD and SPM is assumed.
Since temporal optical solitons studied to date have some kind of hyperbolic secant shape^{10,18,20}, it is surprising that purequartic solitons are approximately Gaussian. To understand this better, we apply the argument of Dudley et al.^{37} to our system, according to which the dispersive and nonlinear phase components developed during short propagation distances must cancel each other across the pulse duration to lead to the formation of a soliton. Figure 5a shows the FOD (red) and the SPMinduced chirp (blue) after propagating the Gaussian variational solution in equation 3 (dashed black curve) for a distance L_{FOD}/10 and demonstrates how this solution leads to the mentioned cancellation across the central part of the pulse. To highlight the different nature of the solutions found here with respect to the previously studied NLS solitons with FOD, we apply the same verification to the solution found in ref. 18, in the limit β_{2}=0. Figure 5b shows how the sech^{2} solution obtained from ref. 18 (dashed black curve) does not provide the necessary cancellation of the SPM and FODinduced chirp required for the formation of a stable solitary wave in the presence of just SPM and FOD.
Discussion
The experimental results and the numerical simulations presented here have established the existence of a new class of solitons: purequartic solitons, arising from the interaction of SPM and FOD only. In particular, we experimentally demonstrated shapepreservation and flatphase behaviour for the fundamental purequartic soliton, and temporal compression for the higherorder purequartic solitons. We numerically demonstrated that the higherorder purequartic soliton would undergo recurrent periodic propagation in the absence of loss and higherorder nonlinearities. Although we have verified that these signatures of soliton propagation are preserved for long propagation distances in the presence of just FOD and SPM, the disparity between the quartic profile of the FODinduced phase shift and the quadratic profile of the SPMinduced phase shift affecting the edges of the pulse could lead to stability issues that should be studied. Establishing appropriate definitions of concepts such as soliton number and soliton period for purequartic solitons is an open theoretical challenge.
Our discovery was facilitated by the unique dispersion properties of PhCwgs that provide the design freedom to achieve a wide variety of dispersion profiles. However, other guided wave systems such as highly nonlinear fibres^{38}, photonic crystal fibres^{8} or specially designed silicon waveguides^{22,39}, could also be engineered to observe purequartic solitons. The main condition to fulfil is L_{FOD}<<L_{GVD}, L_{TOD} (with β_{4}<0), and, in practice, we have verified that L_{FOD}<<L_{GVD}/3 across most of the pulse bandwidth is enough for a robust observation. The purequartic soliton behaviour starts to become observable when L_{FOD} becomes comparable to the sample length. However, our initial simulations show that the purequartic soliton does not reach a steady state until it propagates for several quartic dispersion lengths, similar to conventional solitons^{40}. Analogous regimes of evolution have been demonstrated in Ti:sapphire laser cavities^{25,26,27,28}, taking advantage of the rich variety of physical regimes offered by the discrete structure of the laser cavity. For example, Zhou et al. demonstrated in ref. 26 8.5 fs pulses from a Ti:sapphire laser operating near zero GVD with the minimum pulse duration limited by FOD, and later Christov et al.^{28} hinted that the ‘solitonlike pulse’ inside such a laser was ‘fourthorder dispersion limited’. Here we experimentally demonstrate that the balance between SPM and FOD gives rise to robust solitonlike behaviour. Hence, the scope of our findings is not just limited to nonlinear guided wave optics, but may provide novel insights into extreme regimes of ultrafast lasers operation.
The Gaussian variational solution provided here constitutes a good approximation to the central form of the fundamental purequartic soliton. The results of our study on the cancellation of the nonlinear and quartic dispersion phase components in short propagation distances proved that no previously found solitary wave solution^{10,18,20} can describe the behaviour of purequartic solitons. This approximate solution, valid only for β_{4}<0, could stimulate new efforts in finding solutions to the biharmonic nonlinear Schrodinger equation^{41,42} also of interest in the field of spatial solitons^{23,24,43,44}. Recent interest in temporal cavity solitons in both microresonators^{3} and optical fibres^{7} with applications in Kerr frequency combs^{45} and lownoise microwave generation^{46} could also benefit from exploring purequartic solitons in their systems. Furthermore, it would be interesting to investigate analytic solutions supported by the purequartic soliton system including the effects of linear loss, TPA and FCs.
Aside from their different physical origin, purequartic solitons present significant potential advantages with respect to conventional solitons. As mentioned, purequartic solitons open the door to soliton functionality in the normal GVD regime of optical media. More importantly, perhaps, the energy of conventional solitons scales like (T_{0})^{−1}, whereas the energy of purequartic solitons scales like (T_{0})^{−3}, which suggests that they are more energetic for ultrashort pulses. We expect that the understanding of purequartic solitons provided in this paper, combined with the previous advances in the laser literature^{25,26,27,28}, will inspire a new wave of ultrafast laser development.
Methods
Device and linear characterization
The present experiment was performed using a silicon photonic crystal airsuspended structure with a hexagonal lattice (p6m symmetry group) constant a=404 nm, a hole radius r=116 nm, and a thickness t=220 nm. A 396μmlong dispersion engineered PhCwg was created by removing a row of holes and shifting the two innermost adjacent rows 50 nm away from the line defect. The airclad devices were fabricated with a combination of electron beam lithography, reactive ion and chemical wet etching. The measured linear propagation loss in this slowlight region is ∼70 dB cm^{−1}, with a total linear insertion loss of ∼13 dB (5 dB per facet). Light was coupled in with tapered lensed fibres to SU8 polymer waveguides with inverse tapers.
Phaseresolved characterization method
For the nonlinear experiments, we used a mode locked laser (Alnair) fed into a pulse shaper (Finisar) generating near transformlimited 1.3 ps pulses at 1,550 nm at a 30 MHz repetition rate. These pulses were then input into the FREG apparatus. The pulses were split into two branches by a fibercoupler, with the majority of the energy coupled into the PhCwg. The remaining fraction was sent to a reference branch with a variable delay, before being detected by a fast photodiode and transferred to the electronic domain. This electronic signal drove a Mach–Zender modulator that gated the optical pulse output from the PhCwg. Using an optical spectrum analyser, we measured the spectra as a function of delay to generate a series of optical spectrograms. We deconvolved the spectrograms with a numerical algorithm (256 × 256 gridretrieval errors G<0.005), to retrieve the pulse intensity and the phase in both the temporal and spectral domain^{32}.
Generalized nonlinear Schodinger equation model
The parameters used in our GNLSE model for the slowlight dispersion engineered PhCwg are: slowdown factor S=n_{g}/n_{0}=8.64, effective linear absorption α_{l,eff}=13.9 cm^{−1} ; β_{2}=+1 ps^{2} mm^{−1}, a TOD parameter of β_{3}=0.02 ps^{3} mm^{−1}, and a FOD parameter of β_{4}=−2.2 ps^{4} mm^{−1}; effective nonlinear parameter ; effective TPA parameter ; effective freecarrier dispersion parameter n_{FC,eff}=−6 × 10^{−27} S m^{3}; effective freecarrier absorption parameter σ_{eff}=1.45 × 10^{−21} S m^{2}. The simulation results in Fig. 2 were obtained by using the measured input pulse as the input to the model. The simulation results in Fig. 4 were obtained using a perfect Gaussian, hyperbolic secant, and a super Gaussian (order 4) pulse of the same width as the experimental pulse, 1.3 ps. The linear loss in the nanowires that couple light into and out of the PhCwg was negligible. Nonlinear absorption in the coupling nanowire (effective area, ∼0.2 μm^{2}) was taken into account in the NLSE model.
Additional information
How to cite this article: BlancoRedondo, A. et al. Purequartic solitons. Nat. Commun. 7:10427 doi: 10.1038/ncomms10427 (2016).
Change history
09 March 2016
The original version of this Article contained an error in the spelling of the author C. Martijn de Sterke, which was incorrectly given as de Sterke C. Martijn. This has now been corrected in both the PDF and HTML versions of the Article.
References
 1
Zabulsky, N. J. & Kruskal, M. D. Interactions of “solitons” in a collisionless plasma and the recurrence of initial states. Phys. Rev. Lett. 15, 240–243 (1965).
 2
Grelu, P. & Akhmediev, N. Dissipative solitons for modelocked lasers. Nat. Photon. 6, 84–92 (2012).
 3
Herr, T. et al. Temporal solitons in optical microresonators. Nat. Photon. 8, 145–152 (2014).
 4
Cundiff, S. T. & Ye, J. Colloquium: femtosecond optical frequency combs. Rev. Mod. Phys. 75, 325–342 (2003).
 5
Husakou, A. V. & Herrmann, J. Supercontinuum generation of higherorder solitons by fission in photonic crystal fibers. Phys. Rev. Lett. 87, 20390 (2001).
 6
Dudley, J. M., Genty, G. & Coen, S. Supercontinuum generation in photonic crystal fiber. Rev. Mod. Phys. 78, 1135–1185 (2006).
 7
Leo, F. et al. Temporal cavity solitons in onedimensional Kerr media as bits in an alloptical buffer. Nat. Photon. 4, 471–476 (2010).
 8
Reeves, W. H. et al. Transformation and control of ultrashort pulses in dispersionengineered photonic crystal fibres. Nature 424, 511–515 (2003).
 9
Foster, M., Gaeta, A., Cao, Q. & Trebino, R. Solitoneffect compression of supercontinuum to fewcycle durations in photonic nanowires. Opt. Express 13, 6848–6855 (2005).
 10
Hasegawa, A. & Tappert, F. Transmission of stationary nonlinear optical physics in dispersive dielectric fibers i: anomalous dispersion. Appl. Phys. Lett. 23, 142–144 (1973).
 11
Mollenauer, L. F., Stolen, R. H., Gordon, J. P. & Tomlinson, W. J. Extreme picosecond pulse narrowing by means of soliton effect in singlemode optical fibers. Opt. Lett. 8, 289–291 (1983).
 12
Yin, L. & Agrawal, G.P. Impact of twophoton absorption on selfphase modulation silicon waveguides. Opt. Lett. 32, 2031–2033 (2007).
 13
BlancoRedondo, A. et al. Controlling freecarrier temporal effects in silicon by dispersion engineering. Optica 1, 299–306 (2014).
 14
BlancoRedondo, A. et al. Observation of soliton compression in silicon photonic crystals. Nat. Commun. 5, 3160 (2014).
 15
Wai, P.K. A., Chen, H. H. & Lee, T. C. Radiations by ‘‘solitons’’ at the zero groupdispersion wavelength of singlemode optical fibers. Phys. Rev. A 41, 426–439 (1990).
 16
Höök, A. & Karlsson, M. Ultrashort solitons at the minimumdispersion wavelength: effects of fourthorder dispersion. Opt. Lett. 18, 1388–1390 (1993).
 17
Blow, K. J., Doran, N. J. & Wood, D. Generation and stabilization of short soliton pulses in the amplified nonlinear Schrödinger equation. J. Opt. Soc. Am. B 5, 381–389 (1988).
 18
Karlsson, M. & Höök, A. Solitonlike pulses governed by fourthorder dispersion in optical fibers. Opt. Commun. 104, 303–307 (1994).
 19
Akhmediev, N., Buryak, A. V. & Karlsson, M. Radiationless optical solitons with oscillating tails. Opt. Commun. 110, 540–544 (1994).
 20
Piché, M., Cormier, J.F & Zhu, X. Bright optical soliton in the presence of fourthorder dispersion. Opt. Lett. 21, 845–847 (1996).
 21
Zhakarov, V. E. & Kuznetsov, E. A. Optical solitons and quasisolitons. J. Exp. Theor. Phys. 86, 1035–1045 (1998).
 22
Roy, S. & Biancalana, F. Formation of quartic solitons and a localized continuum in siliconbased waveguides. Phys. Rev. A 87, 025801 (2013).
 23
Staliunas, K., Herrero, R. & De Valcárcel, G. J. Subdiffractive badedge solitons in BoseEinstein condensates in periodic potentials. Phys. Rev. E 73, 065603 (2006).
 24
Staliunas, K., Herrero, R. & De Valcárcel, G. J. Arresting soliton collapse in twodimensional nonlinear Schrödinger systems via spatiotemporal modulation of the external potential. Phys. Rev. A 75, 011604 (2007).
 25
Lemoff, B. E. & Barty, C. P. J. Quinticphaselimited, spatially uniform expansion and recompression of ultrashort optical pulses. Opt. Lett. 18, 1651–1653 (1993).
 26
Zhou, J. et al. Pulse evolution in a broadbandwidth Ti:sapphire laser. Opt. Lett. 19, 1149–1151 (1994).
 27
Stingl, A., Lenzner, M., Spielman, Ch. & Krausz, F. Sub10fs mirrordispersioncontrolled Ti:sapphire laser. Opt. Lett. 20, 602–6014 (1995).
 28
Christov, I., Murnane, M. N., Kapteyn, H. C., Zhou, J. & Huang, C.P. Fourthorder dispersionlimited solitary pulses. Opt. Lett. 19, 1465–1467 (1994).
 29
Li, J., O’Faolain, L., Rey, I. H. & Krauss, T. F. Fourwave mixing in photonic crystals waveguides: slow light enhancement and limitations. Opt. Express 19, 4460–4463 (2011).
 30
Soller, B. J., Gifford, D. K., Wolfe, M. S. & Froggatt, M. E. High resolution optical frequency domain reflectometry for characterization of components and assemblies. Opt. Express 13, 666–674 (2005).
 31
Dorrer, C. Simultaneous temporal characterization of telecommunication optical pulses and modulators by use of spectrograms. Opt. Lett. 27, 1315–1317 (2002).
 32
Thomsen, B. C., Roelens, M. A. F., Watts, R. T. & Richardson, D. J. Comparison between nonlinear and linear spectrographic techniques for the complete characterization of high bitrate pulses used in optical communications. IEEE Phot. Tech. Lett. 17, 1914–1916 (2005).
 33
Patterson, M. et al. DisorderInduced Coherent Scattering in SlowLight Photonic Crystal Waveguides. Phys. Rev. Lett. 102, 253903 (2009).
 34
Lefrancois, S., Husko, C., BlancoRedondo, A. & Eggleton, B. J. Nonlinear silicon photonics analyzed with the moment method. JOSA B 32, 218–226 (2015).
 35
Ablowitz, M. J., Kaup, D. J., Newell, A. C. & Segur, H. The inverse scattering transform  Fourier analysis for nonlinear problems. Stud. Appl. Math. 53, 249–315 (1974).
 36
Akhmediev, N. & Karlsson, M. Cherenkov radiation emitted by solitons in optical fibers. Phys. Rev. A. 51, 2602–2607 (1995).
 37
Dudley, J. M, Peacock, A. C. & Millot, G. The cancellation of nonlinear and dispersive phase components on the fundamental optical fiber soliton: a pedagogical note. Opt. Commun. 193, 253–259 (2001).
 38
Hirano, M., Nakanishi, T., Okuno, T. & Onishi, M. Silicabased highly nonlinear fibers and their application. IEEE J. Quant. Electron. 15, 103–113 (2009).
 39
CastellóLurbe, D., TorresCompany, V. & Silvestre, E. Inverse dispersion engineering in silicon waveguides. J. Opt. Soc. Am. B 31, 1829–1835 (2014).
 40
Agrawal, G. P. Nonlinear Fiber Optics Academic (2007).
 41
Karpman, V. I. & Shagalov, A. G. Stability of soliton described by nonlinear Schrödingertype equations with higherorder dispersion. Physica D 144, 194–210 (2000).
 42
Baruch, G., Fibich, G. & Mandelbaum, E. Singular solutions of the biharmonic nonlinear Schrödinger equation. SIAM J. Appl. Math. 70, 3319–3341 (2010).
 43
Turitsyn, S. K. Threedimensional dispersion nonlinearity and stability of multidimensional solitons. Teoret. Mat. Fiz. 64, 226–232 (1985).
 44
Cole, J. T. & Musslimani, Z. H. Band gaps and lattice solitons for the higherorder nonlinear Schrödinger equation with a periodic potential. Phys. Rev. A 90, 013815 (2014).
 45
Xue, X. et al. Modelocked dark pulse Kerr combs in normal dispersion microresonators. Nat. Photon. 9, 594–600 (2015).
 46
Savchenkov, A. A. et al. Tunable optical frequency comb with a crystalline whispering gallery mode resonator. Phys. Rev. Lett. 101, 93902 (2008).
Acknowledgements
This work was supported in part by the Center of Excellence CUDOS (CE110001018), Laureate Fellowship (FL120100029) schemes of the Australian Research Council (ARC) and by The University of Sydney and the Technion collaborative photonics research project funded by The Department of Trade and Investment, Regional Infrastructure and Services of the New South Wales Government and The Technion Society of Australia NSW. T.F.K. was supported by EPSRC UK Silicon Photonics (Grant reference EP/F001428/1). C.H. was supported by the ARC Discovery Early Career Researcher award (DECRA—DE120102069).
Author information
Affiliations
Contributions
A.B.R. and C.H. conducted the experiments. A.B.R. performed the retrievals, developed the GNLSE model, analysed the results, contributed with physical insight to the derivation of the approximate analytic solutions, and wrote the paper. C.M. de S. obtained the local approximate analytic solution, contributed to the analysis of the results, and proposed numerous verification tests. J.E.S. obtained the variational approximate analytic solution. T.F.K. fabricated the sample. B.J.E. provided guidance to the project and proposed the verification test for the Gaussian solution. C.H. built the setup and contributed to analysing the results. All authors contributed in editing the paper.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Supplementary Information
Supplementary Notes 12. (PDF 94 kb)
Rights and permissions
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
About this article
Cite this article
BlancoRedondo, A., de Sterke, C., Sipe, J. et al. Purequartic solitons. Nat Commun 7, 10427 (2016). https://doi.org/10.1038/ncomms10427
Received:
Accepted:
Published:
Further reading

Influence of fourthorder dispersion on the Anderson localization
Nonlinear Dynamics (2020)

Highly dispersive optical solitons in birefringent fibers with four forms of nonlinear refractive index by three prolific integration schemes
Optik (2020)

Propagation of dipole solitons in inhomogeneous highly dispersive opticalfiber media
Physical Review E (2020)

Selfsimilar propagation of optical pulses in fibers with positive quartic dispersion
Optics Letters (2020)

Performance of chipscale optical frequency comb generators in coherent WDM communications
Optics Express (2020)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.