## Abstract

Supercontinuum generation with femtosecond pulses in photonic crystal fibers with two zero-dispersion wavelengths (ZDWs) is investigated numerically. The role of the higher ZDW is examined for 5 fiber designs with a nearly constant lower ZDW. It is found that the resulting spectrum is mainly determined by self-phase modulation in the first few mm of fiber, followed by soliton self-frequency shift and amplification of dispersive waves. It is demonstrated how femtosecond soliton pulses can be generated with any desired center wavelength in the 1020–1200 nm range by adjusting the fiber length. Further, the generation of a bright-bright soliton-pair from an initial single red-shifted soliton is found. The soliton-pair has one color in the anomalous dispersion region and the other color in the normal dispersion region, which has not previously been described for bright-bright soliton-pairs.

© 2005 Optical Society of America

## Corrections

Michael H. Frosz, Peter Falk, and Ole Bang, "The role of the second zero-dispersion wavelength in generation of supercontinua and bright-bright soliton-pairs across the zero-dispersion wavelength: erratum," Opt. Express**15**, 5262-5263 (2007)

https://www.osapublishing.org/oe/abstract.cfm?uri=oe-15-8-5262

## 1. Introduction

Supercontinuum generation in photonic crystal fibers (PCFs) has attracted much attention in the past few years. One of the reasons is the possibility to fabricate small core sizes (~ 1 μm), leading to the high effective nonlinearities necessary to generate broad spectra. Another reason is the possibility to design the dispersion profile of the PCF, which is very important for particularly parametric processes [1, 2, 3]. An interesting example of the possibilities offered by this design freedom, is the fabrication of PCFs with two zero-dispersion wavelengths (ZDWs) within the optical spectrum. In standard optical fibers there is only one ZDW in the optical spectrum, typically at ~1300 nm. It is well known that the gain bandwidth for four-wave mixing (FWM) is widest in the vicinity of the ZDW due to phase matching conditions [1, 2], and that solitons in the vicinity of the ZDW can amplify dispersive waves in the normal dispersion region (NDR) [4, 5, 6]. Therefore, PCFs with 2 (or more) ZDWs could prove advantageous for efficient supercontinuum generation and spectral shaping.

Previous investigations of supercontinuum generation in PCFs with two ZDWs have provided two rather distinct explanations of the underlying physical mechanisms. Hilligsøe *et al*. [7] examined a PCF with a spacing of 165 nm between the two ZDWs and explained the super-continuum as a result of self-phase modulation (SPM) and FWM. Genty *et al*. [5] investigated supercontinuum generation in PCFs with a spacing of more than 700 nm between the two ZDWs and found the most important mechanisms to be amplification of dispersive waves and soliton self-frequency shift (SSFS). Both [7] and [5] pumped near the lower ZDW. Recently, Efimov *et al*. [6] considered a PCF pumped close to the higher ZDW, and investigated the interaction between solitons and dispersive waves generated above the higher ZDW. It was found that close to the higher ZDW, SSFS is counter-balanced by spectral recoil from these dispersive waves, thus stabilizing the soliton by halting its red-shift [8, 9]. This spectral recoil effect could prove beneficial in the quest to achieve better control of the supercontinuum generation.

The resulting spectrum in ref. [7]consisted mainly of two separated spectral peaks. Such a spectrum could be used in e.g. differential absorption imaging using optical coherence tomography[10]. In this context it is advantageous to be able to control the spectral location of the two spectral peaks.

For these reasons, it is of fundamental interest to investigate pulse propagation in PCFs with two ZDWs, to clarify the role of the second ZDW and the mechanisms behind the supercontinuum generation. This could allow for better spectral shaping using appropriate fiber design, and therefore lead to light sources especially suited for particular applications such as optical coherence tomography. In this work, we numerically examine supercontinuum generation in 5 different triangular structure PCFs where we change the separation between the two ZDWs from 165 nm to 870 nm by altering the pitch and hole size of the triangular hole structure [11, 12]. We pump close to the nearly constant lower ZDW, which allows us to study the influence of the SSFS, as the soliton can then be formed close to the lower ZDW, followed by a red-shift in frequency towards the higher ZDW, due to SSFS. It is shown that this SSFS can be used to generate wavelength-tunable femtosecond solitons, and we demonstrate a spectral shift of more than 30% of the optical frequency. The location of the higher ZDW can limit the red-shift of the soliton, and is found to determine whether dispersive waves or a soliton pair is formed when the red-shifted soliton approaches the higher ZDW. Specifically, we found that for one of the fiber designs, a bright-bright soliton pair is generated across the ZDW. To the best of our knowledge, this has not been previously described.

## 2. Numerical simulation of pulse propagation

The pulse propagation is simulated using the split-step Fourier method to solve the generalized nonlinear Schrödinger (NLS) equation [1]

where *A*(*z*,*t*) is the pulse envelope variation in a retarded time frame *t* moving with the group velocity of the pump, along the fiber axis *z*. The dispersion parameters *β¯*_{m}
are estimated from a polynomial fit to the fiber dispersion profile *β*
_{2}(*ω*) given by [1],

where

and we include up to *β*¯_{15} to obtain a good polynomial fit to the dispersion profile over the wavelength range of interest. *β*(*ω*) is the mode-propagation constant. The dispersion profile for each fiber is calculated numerically using fully-vectorial plane-wave expansions [13]. We consider triangular PCFs with pitch ʌ and hole diameter *d* (see Fig. 1). *γ* = *n*
_{2}
*ω*
_{0}/[*cA*
_{eff}] is the nonlinear parameter, where *n*
_{2} = 2.6 · 10^{-20} m^{2}/W is the nonlinear-index coefficient for silica, *ω*
_{0} is the center angular frequency of the pump pulse, *c* is the speed of light in vacuum, and *A*
_{eff} is the effective core area. *R*(*t*) is the Raman response function [1].

Contrary to the traditional slowly varying envelope approximation, the model is valid for pulses with spectral widths up to about 1/3 of the carrier frequency [14]. The model accounts for SPM, FWM, stimulated Raman scattering (SRS) and self-steepening. We simplify the investigation by assuming a polarization maintaining fiber pumped along one polarization axis. Our model includes the wavelength dependence of the effective core area*A*
_{eff}(*λ*) using the more general definition suitable for fibers where some of the field energy may reside in the air-holes [15]. For all simulations presented in this work the relative change in the total photon number, a measure of the numerical error which is ideally zero in the absence of loss [14], was less than 0.2%. We have implemented the adaptive step-size method outlined by Sinkin *et al*. [16], since it reduces the total number of Fourier transforms, and thus increases computation speed.

We used 2^{15} points, unless otherwise stated, and a temporal resolution of 1.4 fs, giving a time window of 46 ps. The local goal error used in the adaptive step-size method was set to *δ*
_{G} = 10^{-8} for the fibers with a pitch ʌ ≤ 1.2 μm, and *δ*
_{G} = 10^{-9} for the fibers with a pitch ʌ ≥ 1.3 μm. For the ʌ = 1.0 μm PCF, this resulted in a step size of ~ 50 nm in the beginning of the fiber which increased gradually along the fiber until it was ~6 μm at the end of the fiber. As is shown in the next section, much of the power launched into the ʌ = 1.0 μm PCF is rapidly moved to the NDR, such that only a small amount of power is left for the solitons. This causes the pulse to spread temporally, resulting in a lower peak power and lower nonlinearities, which allows the adaptive step-size algorithm to automatically increase the step-size, and thereby also the computation speed. On the other hand, in the ʌ = 1.2 μm PCF the soliton retains a large peak power and the step size is initially ~60 nm in the beginning of the fiber and increases to only ~0.6 μm at the end of the fiber.

To better understand the pulse dynamics, we consider the pulse simultaneously in the time and spectral domain using spectrograms calculated as [7]

where we have used a window size of *α* = 16 fs. The spectrogram displays the relative temporal positions of the frequency components of the pulse, and is similar to a cross-correlation frequency-resolved optical gating measurement (X-FROG) [17]. We plot the spectrograms on a logarithmic color scale, normalized to *S*(0,0, *ω*
_{0}).

## 3. The role of the higher ZDW

As shown in Fig. 1 we have found that by modifying the pitch ʌ and hole size *d* of the triangular hole structure the lower ZDW remains constant at ~780 nm, while the higher ZDW varies between 950 nm and 1650 nm [11, 12]. The dispersion profiles of the fiber with ʌ = 1.0 μm and the fiber with ʌ = 1.4 μm are similar to the dispersion profile of the fiber examined by Hilligsøe *et al*. [7] and Genty *et al*. [5], respectively. *A*
_{eff}(*λ* = 804 nm) is 1.38 μm^{2} for the fiber with ʌ = 1.0 μm and increases with the pitch ʌ. For the ʌ = 1.4 μm fiber, *A*
_{eff}(*λ* = 804 nm) is 1.97 μm^{2}.

To focus this investigation on the influence of the higher ZDW, we have used the same input pulse parameters for all simulations: the pump wavelength is *λ*
_{0} = 804 nm, the pulse is Gaussian shaped with an intensity full-width-half-maximum (FWHM) of *T*
_{FWHM} = 13 fs, and the peak power is *P*
_{0} = 15 kW. When calculating the spectral average power density, we assume a repetition rate of 80 MHz; the power spectral densities *S*(*λ*) presented here are thus normalized so that ∫ *S*(*λ*)d*λ* = *P*
_{av}, where *P*
_{av} is the average pulse power of the input pulse. The used parameters are realizable with commercially available femtosecond lasers. Hilligsøe *et al*. [7] used a fiber length of 5 cm. Genty *et al*. [5] used a fiber length up to 1.5 m, but found that the continuum generation was complete after 50 cm of propagation. We have simulated propagation up to a length of 60 cm, 52 cm, and 60 cm for the fibers with ʌ ≤ 1.2 μm, ʌ = 1.3 μm, and ʌ = 1.4 μm, respectively.

A soliton can transfer energy to a dispersive wave when (1) the soliton and the dispersive wave have equal wave numbers, and (2) a significant part of the soliton spectral power is at the dispersive wave wavelength [4]. First, we estimate the wave number (eigenvalue) *k*
_{sol} of a fundamental soliton with temporal width *T*
_{sol} and carrier frequency *ω*
_{sol}. We assume that the soliton spectrum is narrow enough for *k*
_{sol} to only be slightly perturbed by higher-order dispersion, and neglect this perturbation. One can then use the simple NLS equation

insert the soliton solution *A* = √*P*
_{sol}sech(*t*/*T*
_{sol})exp(*ik*
_{sol}
*z*) and find *k*
_{sol} = |*β*
_{2}(*ω*
_{sol})|/[2${T}_{\text{sol}}^{2}$] [1].

Since the dispersive wave can be generated far from *ω*
_{sol}, we use the NLS equation with all the higher-order dispersion terms (up to *β*
_{15}) to estimate the wavenumber *k*
_{lin} of the linear dispersive wave. However, since the dispersive wave initially has neglible power, the nonlinearity can be neglected by setting *γ*=0:

By inserting the expression for a dispersive (linear) wave *A*
_{lin} = √*P*
_{lin} exp{*i*[*k*
_{lin}
*z* - (*ω*
_{DW} - *ω*
_{sol})*t*]} into Eq. (6), we obtain for *k*
_{sol} = *k*
_{lin}

where *ω*
_{DW} is the dispersive wave angular frequency. Since only second-order dispersion is considered for the soliton wavenumber, this equation is, strictly speaking, not valid in the immediate vicinity of the ZDWs. However, as shown below, the predictions given by this equation show agreement with the numerical simulations.

When the pump peak power *P*
_{0} is sufficiently high for a higher-order soliton to form, the soliton can break up into *N* fundamental solitons with different peak powers and temporal widths. The shortest fundamental soliton has the highest peak power, and its width is given by [18]

where *T*
_{0} = *T*
_{FEHM}/1.665. This is inserted into Eq. (7) which is then solved for each fiber dispersion profile. We assume that the fundamental soliton temporal width *T*
_{sol} is unchanged as the soliton is down-shifted in frequency due to SSFS; *T*
_{sol} given by Eq. (8) is therefore calculated using the pump peak power *P*
_{0}, the input pulse temporal width *T*
_{0}, and the second-order dispersion *β*¯_{2} (*β*
_{2}(*ω*
_{sol}) = *β*¯_{2} at *ω*
_{sol} = (*ω*
_{0}). The result is shown in Fig. 1, which shows the wavelengths *λ*
_{DW} at which dispersive waves can be amplified, as a function of the soliton wavelength *λ*
_{s}. From this figure we expect that a soliton initially launched at *λ*
_{0} = 804 nm will amplify dispersive waves at ~ 600 nm in all the 5 fibers investigated here. Dispersive waves in the infrared are not expected to be amplified in the beginning of the fiber, because there is initially not enough spectral power from the soliton in the infrared [5].

Equation (8) gives the width of the shortest fundamental soliton, which was then used for Fig. 1. We have calculated that the temporally longer fundamental solitons are phase-matched to amplify dispersive waves slightly closer to the soliton wavelength than the shortest soliton. More importantly, the frequency shift per unit propagation length caused by SSFS is smaller, because it scales inversely with *T*
_{sol} [19]. This makes the shortest fundamental soliton move quicker towards the higher ZDW and be the first to amplify dispersive waves. It is therefore more important for the generation of dispersive waves than the longer solitons.

Due to small differences in *A*
_{eff}(*ω*
_{0}) and *β*¯_{2}, *N* varies from 3.8 to 3.3 for the ʌ = 1.1 - 1.4 μm fibers. Compared to these fibers, *β*¯_{2} for the ʌ = 1.0 μm fiber is ~ 4 times smaller numerically, giving *N* = 8.2.

#### 3.1. Propagation up to 6 mm

Figure 2 shows spectrograms of the pulse evolution during the first 6 mm in the ʌ = 1.0 μm and ʌ = 1.1 μm fibers. For this short propagation distance, we only used 2^{10} computation points, but the same temporal resolution of 1.4 fs. The relative change in photon number was less than 0.015%. From the spectrograms it is seen that the initial spectral broadening during the first ~ 2 millimeters of the fibers is due to SPM, since the leading edge of the pulse is red-shifted while the trailing edge is blue-shifted [1]. The observed continued spectral broadening in the next few millimeters could potentially be caused by continued SPM, FWM [7] and/or SSFS combined with amplification of dispersive waves [5]. For the ʌ = 1.0 μm fiber we tested this by making two additional simulations: (a) one without the delayed Raman response (*R*(*t*) = *δ*(*t*)) and (b) one where the summation in Eq. (1) only contains *β*¯_{2} and *β*¯_{3}. In case (a) there can be no SSFS, since there is no Raman loss to phonons. In case (b) the wavelengths at which the phase match
*κ* [1]:

where Ω is the angular frequency shift from the pump in the FWM process. It is seen in Fig. 3 (right) that in case (b) the phase matched (*κ* = 0) wavelengths are closer to the pump. Furthermore, the solution to Eq. (7) for case (b) predicts wavenumber match to a dispersive wave at ~ 400 nm, instead of ~ 600 nm. The simulation results for case (a) and (b) are shown in Fig. 3 (left).

For case (a) the red- and blue-shifted peaks are at almost the same location as for the full simulation. This shows that SSFS has neglible influence during the first 6 millimeters of propagation. The results for case (b) show that the red- and blue-shifted peaks are even further away from the pump than in the full simulation. This is contrary of what is expected, if degenerate FWM plays a significant role. We can therefore conclude that FWM is negligible. The blue-shifted peak is more blue-shifted in case (b) than in the full simulation, as predicted from the calculated dispersive wave wavenumber match so amplification of dispersive waves may play a role. The initial spectral broadening is therefore mainly due to SPM, possibly assisted by amplification of dispersive waves.

In the ʌ = 1.0 μm fiber, the anomalous dispersion region (ADR) is so narrow that the SPM has rapidly moved most of the pulse energy into the normal dispersion region (NDR) (Fig. 2). In the ʌ = 1.1 μm fiber, the red-shifted peak remains in the ADR and can thus form a soliton. It is seen from Fig. 2 that the soliton starts to amplify dispersive waves at 1600 nm at *z* ~ 3 mm. The dispersive wave wavelength of ~ 1600 nm is correctly predicted from Fig. 1. We note that the dispersive waves immediately spread temporally, as expected.

The power spectra after 6 mm of propagation in each of the 5 examined fibers are compared in Fig. 4. The dispersive wave generated at ~1600 nm in the ʌ = 1.1 μm fiber has too low power to be seen on the linear scale used in the figure. The comparison shows that in all of the 5 fibers there are two distinct peaks, a red-shifted and a blue-shifted, which arise from SPM, as outlined above. Except for the ʌ = 1.0 μm fiber, in all cases the red-shifted peak is still in the ADR after 6 mm and is able to form solitons.

#### 3.2. Propagation beyond 6 mm

The solitons are gradually red-shifted by SSFS towards the higher ZDW. When a soliton has red-shifted to the vicinity of the higher ZDW it starts to amplify dispersive waves in the NDR [5]. As shown in the previous section, this occurs already after approximately 3 mm of propagation in the ʌ = 1.1 μm fiber. For fibers with a larger pitch , the soliton can propagate several centimeters before amplifying dispersive waves in the infrared. This happens at approximately *z* = 35 cm for the 1.2 μm fiber (Fig. 5). The spectral ripples in the dispersive waves at ~600 nm can be explained by cross-phase modulation (XPM) between the soliton and the dispersive waves [20]. Note that the pulse generated in the infrared NDR of the ʌ = 1.2 μm fiber, does not immediately disperse (as was the case for the ʌ = 1.1 μm fiber, see Fig. 2) but retains its temporal width over several centimeters (Fig. 6). The pulse is generated at ~ 1600 nm where *β*
_{2} = 2 · 10^{-25} s^{2}/m, and has an estimated temporal width (3 dB) of less than 50 fs. The dispersion length is thus [1] *L*_{D}
= ${T}_{0}^{2}$/ l|β_{2}| ≈ 1 cm. Since the temporal width of the radiation in the NDR is almost constant during 20 cm of propagation (Fig. 6), it is clear that some nonlinear effects counter-act the dispersion. This is similar to the effect of SPM in the ADR, giving rise to solitons. We believe that the pulse generated in the NDR forms a soliton-pair with the soliton in the ADR.

Soliton-pairs across the ZDW are known in the form of bright-dark and bright-gray soliton pairs made possible by XPM between the solitons [1]. However, in our case a bright-bright soliton pair across the ZDW is apparently formed, which has not been previously described. The co-propagation of a sech-pulse in the NDR with a sech-pulse in the ADR was investigated in ref [21], but only the propagation of the pulse in the ADR was considered, and the influence of the ADR-pulse on the NDR-pulse was neglected. A bright-bright soliton pair can be formed within the same dispersion region [22], but XPM alone cannot allow a bright-bright soliton pair across the ZDW. We therefore expect that the Raman effect also plays a role in this new observation.

It is known that a region of modulational instability (MI) can exist near the ZDW in the NDR, and that this MI can possibly allow bright solitons to form in the NDR [23]. One requirement for the MI region to exist at the angular frequency *ω* is *β*
_{2}(*ω*) + *β*
_{4}(*ω*)Ω^{2}/12 < 0 [24], where Ω is defined as in Eq.(9). Near the higher ZDW of the ʌ = 1.2 μm fiber we have *β*
_{4}(*ω*) < 0, so there does not exist an MI region in the NDR (*β*
_{2} < 0), and this can therefore not explain the apparent soliton formation in the NDR.

Since one part of the soliton pair is formed in the NDR with higher wavelength than the ADR, the red-shift of the soliton in the ADR is halted due to spectral recoil from the higher ZDW [8, 9]]. The soliton in the NDR therefore does not have to continuously shift its wavelength to match the group-velocity of the soliton in the ADR; if the NDR-pulse had a lower wavelength than the ADR-pulse the two pulses would separate spectrally, due to SSFS of the ADR-pulse and spectral recoil from the lower ZDW. Instead, the result is a spectrally stable soliton pair. The condition of group-velocity matching for generation of the soliton-pair is examined analytically in the next section.

For the 1.3 *ε*m fiber, the soliton has shifted to ~1200 nm at *z* = 52 cm. Since this is still far from the higher ZDW, the soliton has not yet lost power to dispersive waves in the infrared (Fig. 5). Similarly, in the 1.4 *ε*m fiber the soliton has shifted to ~1170 nm at *z* = 60 cm without losing power to infrared dispersive waves. In the simulations with a relatively large distance between the ZDWs (e.g. Fig. 5), we note that the soliton seems to be quite stable: it does not split into multiple solitons. In [5] one can observe immediate splitting into multiple solitons which have roughly equal peak power. We believe this difference is due to our use of a shorter pulse length (13 fs) compared to [5] (200 fs); the soliton number *N* scales as (*P*
_{0}
${T}_{0}^{2}$)^{1/2}, and is almost 6 times smaller in our case compared to [5]. Furthermore, the soliton number *N* ~ 3 in the beginning of the fibre, but since the peak power decreases due to SPM combined with group-velocity dispersion after a few millimeters of propagation, *N* could decrease to ~ 1 so that only one fundamental soliton remains.

A potential application of these fibers could e.g. be the generation of fs pulses in the infrared from an initial pulse at *λ* = 804 nm. It is further demonstrated in Fig. 7(a) that the SSFS can be used to produce fs pulses with an almost Gaussian spectral shape. By choosing a suitable fiber length, the central wavelength of the Gaussian spectrum can be freely selected in the range ~ 1020–1200 nm. The FWHM of each Gaussian spectrum is on the order of 30 nm. The use of the SSFS to generate wavelength-tunable soliton pulses has previously been demonstrated in tapered PCFs [25]. For comparison, in [25] a SSFS of 20%of the optical frequency (*λ* = 1.3 → 1.65 μm) was experimentally demonstrated in a 15 cm long tapered fiber, where our results correspond to a SSFS of more than 30% in 52 cm of untapered fiber.

We expect that for longer fiber lengths, the solitons in the 1.3–1.4 μm PCFs will also continue to red-shift until they are in the vicinity of the higher ZDW, followed by an amplification of dispersive waves in the infrared. We note, however, that fiber losses in the infrared region (OH-absorption and confinement loss) may ultimately limit the spectral extension into the infrared [5, 7].

#### 3.3. Group-velocity matching condition for soliton-pair across ZDW

In this section we derive an equation describing the condition for group-velocity matching. This is a necessary condition if radiation initially formed from amplification of dispersive waves in the NDR, leads to the formation of a soliton-pair consisting of a soliton in the ADR (at *λ* = *λ*
_{A}) co-propagating with a ‘soliton’ in the NDR (at *λ* = *λ*
_{B}).

A necessary condition for the two pulses to be co-propagating is that the group-velocity *v*
_{g} is matched, i.e. 1/*v*
_{g} = *β*
_{1}(*ω*
_{A}) = *β*
_{1}(*ω*
_{B}). From the definition of *β*(*ω*) [1] we have,

where *β*
_{m}(*ω*) is given by Eq. (3) and we only include up to *β*
_{3}(*ω*
_{A}). The condition *β*
_{1}(*ω*
_{A}) = *β*
_{1}(*ω*
_{B}) then gives

where Δ*ω*= *ω*
_{B} - *ω*
_{A}. From Eq. (11) we have plotted *λ*
_{B} as a function of *λ*
_{A} for the dispersion profiles belonging to the fibres with ʌ = 1.1 μm and ʌ = 1.2 μm (Fig. 7(b)). From the previous section, we know that the SSFS is cancelled for the soliton in the ʌ = 1.1 μm fiber when the soliton wavelength reaches ~ 1000 nm. It is seen in Fig. 7(b), that for *λ*_{A}
= 1000 nm there is no group-velocity match in the vicinity of *λ*_{A}
. Since the amplification of dispersive waves transfers energy to ~ 1400 nm, a soliton-pair can not be formed. For the ʌ = 1.2 μm fiber, we know that the soliton is red-shifted to ~ 1250 nm before cancellation of SSFS. According to Fig. 7(b) at *λ*
_{A} = 1250 nm there is group-velocity match to *λ*
_{B} ~ 1500 nm. The ADR-pulse can therefore match its group-velocity to the NDR-pulse formed by the transfer of energy to ~ 1600 nm. As observed in Figs. 5 and 6, this leads to the formation of what appears to be a bright-bright soliton pair across the higher ZDW.

## 4. Conclusions and discussion

We have numerically investigated femtosecond pulse propagation in 5 different photonic crystal fibers with an almost equal lower ZDW at ~ 780 nm and with higher ZDWs ranging from 950 nm to 1650 nm. The fibers are all pumped close to the lower ZDW. Our results show that SPM is dominant in the first ~ 6 mm of the fibers. Contrary to [7], we found that FWM does not play a part in the initial broadening of the spectrum. In [7] the influence of FWM was tested by shifting the dispersion of the fiber to normal dispersion everywhere,*β*
_{2}(*ω*) → *β*
_{2} (*ω*) + *β*′, thus removing the possibility of phasematched FWM. A significantly weaker spectral broadening was then observed, and from this it was concluded that FWM plays an important role. However, in the normal dispersion regime SPM and group-velocity dispersion will act together to broaden the pulse temporally significantly faster than in the anomalous dispersion regime [1]. Shifting the dispersion to normal dispersion everywhere will therefore weaken the SPM induced spectral broadening, and this could explain the observations made in [7].

For the fiber with the most narrow ADR, SPM caused most of the power to be moved out of the ADR, and further broadening into the infrared was halted. For fibers with a larger separation between the ZDWs a soliton is formed, which gradually red-shifts along the fiber due to soliton self-frequency shift. The larger the higher ZDW is, the further into the infrared the soliton can red-shift without losing power to dispersive waves above the higher ZDW. We demonstrated how this could be used to generate fs soliton pulses with a center wavelength anywhere between 1020–1200 nm from pulses at 804 nm, corresponding to a shift in optical frequency up to 30%, simply by varying the fiber length. This could potentially find an application in e.g. optical coherence tomography, where it is desired to have a light source with a center wavelength above 1000 nm, for increased penetration in highly scattering tissue such as skin.

We also demonstrated the generation of a bright-bright soliton-pair, from one single red-shifted soliton. This occurs when spectral recoil is sufficiently weak to allow the soliton in the anomalous-dispersion region to red-shift close enough to the higher ZDW so that group-velocity matching to the pulse generated in the normal-dispersion region occurs. The colors of the soliton-pair are in the normal and the anomalous dispersion regime, respectively. This has not previously been observed for bright-bright soliton-pairs.

## Acknowledgments

P. Falk acknowledges financial support from the STVF BIOLASE grant No. 26-02-0020. The authors would like to thank Jesper Lægsgaard for his helpful guidance on calculating the dispersion curves. We would also like to thank Anders Bjarklev, Jes Broeng and Peter E. Andersen for fruitful discussions.

## References and links

**1. **G. P. Agrawal, *Nonlinear Fiber Optics*, 3rd ed. (Academic Press, San Diego, CA, USA,2001).

**2. **J. M. Dudley, L. Provino, N. Grossard, H. Maillotte, R. S. Windeler, B. J. Eggleton, and S. Coen, “Supercontinuum generation in air-silica microstructured fibers with nanosecond and femtosecond pulse pumping,” J. Opt. Soc. Am. B **19**, 765–771 (2002). [CrossRef]

**3. **N. I. Nikolov, T. Sørensen, O. Bang, and A. Bjarklev, “Improving efficiency of supercontinuum generation in photonic crystal fibers by direct degenerate four-wave mixing,” J. Opt. Soc. Am. B **20**, 2329–2337 (2003). [CrossRef]

**4. **N. Akhmediev and M. Karlsson, “Cherenkov radiation emitted by solitons in optical fibers,” Phys. Rev. A **51**, 2602–2607 (1995). [CrossRef] [PubMed]

**5. **G. Genty, M. Lehtonen, H. Ludvigsen, and M. Kaivola, “Enhanced bandwidth of su-percontinuum generated in microstructured fibers,” Opt. Express **12**, 3471–3480 (2004).http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-15-3471. [CrossRef] [PubMed]

**6. **A. Efimov, A. J. Taylor, F. G. Omenetto, A. V. Yulin, N. Y. Joly, F. Biancalana, D. V. Skryabin, J. C. Knight, and P. S. J. Russell, “Time-spectrally-resolved ultrafast nonlinear dynamics in small-core photonic crystal fibers: Experiment and modelling,” Opt. Express **12**, 6498http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-15-3471.6507 (2004). [CrossRef] [PubMed]

**7. **K. M. Hilligsøe, T. V. Andersen, H. N. Paulsen, C. K. Nielsen, K. Mølmer, S. Keid-ing, R. Kristiansen, K. P. Hansen, and J. J. Larsen, “Supercontinuum generation in a photonic crystal fiber with two zero dispersion wavelengths,” Opt. Express **12**, 1045–1054 (2004).http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-6-1045. [CrossRef] [PubMed]

**8. **D. V. Skryabin, F. Luan, J. C. Knight, and P. S. J. Russel, “Soliton self-frequency shift cancellation in photonic crystal fibers,” Science **301**, 1705–1708 (2003). [CrossRef] [PubMed]

**9. **F. Biancalana, D. V. Skryabin, and A. V. Yulin, “Theory of the soliton self-frequency shift compensation by the resonant radiation in photonic crystal fibers,” Phys. Rev. E **70**, 016615 (2004). http://dx.doi.org/10.1103/PhysRevE.70.016615. [CrossRef]

**10. **J. M. Schmitt, S. H. Xiang, and K. M. Yung, “Differential absorption imaging with optical coherence tomography,” J. Opt. Soc. Am. A **15**, 2288–2296 (1998). [CrossRef]

**11. **M. H. Frosz, O. Bang, A. Bjarklev, P. E. Andersen, and J. Broeng, “Supercontinuum Generation in Photonic Crystal Fibers: The Role of the Second Zero Dispersion Wavelength,” presented May 25th 2005, CWC1, at CLEO/QELS 2005, Baltimore, Maryland, USA, 22–27 May 2005.

**12. **M. H. Frosz, P. Falk, L. T. Pedersen, O. Bang, and A. Bjarklev, “Supercontinuum generation in untapered and tapered photonic crystal fibers with two zero dispersion wavelengths,” talk #5733–36 presented at SPIE Photonics West, San Jose, California, USA, 22–27 January 2005.

**13. **S. G. Johnson and J. D. Joannopoulos, “Block-Iterative Frequency-Domain Methods for Maxwell’s Equations in a Planewave Basis,” Opt. Express **8**, 173–190 (2001). http://www.opticsexpress.org/abstract.cfm?URI=OPEX-8-3-173. [CrossRef] [PubMed]

**14. **K. J. Blow and D. Wood, “Theoretical description of transient stimulated Raman scattering in optical fibers,” IEEE J. Quantum Electron. **25**, 2665–2673 (1989). [CrossRef]

**15. **J. Lægsgaard, N. A. Mortensen, and A. Bjarklev, “Mode areas and field-energy distribution in honeycomb photonic bandgap fibers,” J. Opt. Soc. Am. B **20**, 2037–2045 (2003). [CrossRef]

**16. **O. V. Sinkin, R. Holzlöhner, J. Zweck, and C. R. Menyuk, “Optimization of the Split-Step Fourier Method in Modeling Optical-Fiber Communications Systems,” J. Lightwave Technol. **21**(1), 61–68 (2003). [CrossRef]

**17. **A. Efimov and A. J. Taylor, “Spectral-temporal dynamics of ultrashort Raman solitons and their role in third-harmonic generation in photonic crystal fibers,” Appl. Phys. B **80**, 721–725 (2005).http://dx.doi.org/10.1007/s00340-005-1789-2. [CrossRef]

**18. **J. K. Lucek and K. J. Blow, “Soliton self-frequency shift in telecommunications fiber,” Phys. Rev. A **45**, 6666–6674 (1992). [CrossRef] [PubMed]

**19. **J. Herrmann and A. Nazarkin, “Soliton self-frequency shift for pulses with a duration less than the period of molecular oscillations,” Opt. Lett. **19**, 2065–2067 (1994). [CrossRef] [PubMed]

**20. **G. Genty, M. Lehtonen, and H. Ludvigsen, “Effect of cross-phase modulation on supercontinuum generated in microstructured fibers with sub-30 fs pulses,” Opt. Express **12**, 4614–4624 (2004). http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-19-4614. [CrossRef] [PubMed]

**21. **C. S. Aparna, S. Kumar, and A. Selvarajan, “Suppression of the soliton frequency shifts by nonlinear pairing of pulses,” Opt. Commun. **131**, 267–273 (1996). http://dx.doi.org/10.1016/0030-4018(96)00350-1. [CrossRef]

**22. **V. V. Afansyev, Y. S. Kivshar, V. V. Konotop, and V. N. Serkin, “Dynamics of coupled dark and bright optical solitons,” Opt. Lett. **14**, 805–807 (1989). [CrossRef]

**23. **F. K. Abdullaev, S. A. Darmanyan, S. Bischoff, P. L. Christiansen, and M. P. Sørensen, “Modula-tional instability in optical fibers near the zero dispersion point,” Opt. Commun. **108**, 60–64 (1994). http://dx.doi.org/10.1016/0030-4018(94)90216-X. [CrossRef]

**24. **J. D. Harvey, R. Leonhardt, S. Coen, G. K. L. Wong, J. C. Knight, W. J. Wadsworth, and P. S. Russell, “Scalar modulation instability in the normal dispersion regime by use of a photonic crystal fiber,” Opt. Lett. **28**, 2225–2227 (2003). [CrossRef] [PubMed]

**25. **X. Liu, C. Xu, W. H. Knox, J. K. Chandalia, B. J. Eggleton, S. G. Kosinski, and R. S. Windeler, “Soliton self-frequency shift in a short tapered air-silica microstructure fiber,” Opt. Lett. **26**, 358–360 (2001). [CrossRef]