Abstract
Amplitude and phase ripples of an ultrafast chirped pulse complex envelope are considered. The formula that establishes the relation between the pulse complex envelope and its spectrum amplitude was theoretically derived for both cases of the ripple. Based on current results, a linearly chirped pulse can be retrieved using only a spectrum analyzer and Kerr nonlinear medium.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Nowadays a chirped pulse amplification (CPA) technique is the most straightforward way for the generation of ultrafast high peak power optical pulses [1,2]. A pulse stretcher is the key component in CPA schemes. It is used for broadening pulses in time domain prior to amplification by inducing a group delay dispersion (GDD) and therefore reducing its peak power. Thus, an adverse influence of various nonlinear effects is considerably reduced. However, it is not possible to eliminate absolutely certain nonlinear effects, such as e.g. self-phase modulation (SPM), which have no threshold and therefore affect pulses with any energy. A SPM effect is crucial in the fiber based CPA scheme implementations, where optical pulses propagate in elongated medium with small mode field diameters, and consequently pulses with rather moderate peak powers considerably suffer from SPM distortions [2]. Both output optical pulse quality and its peak power are dramatically reduced due to a group delay ripple (GDR) and an amplitude ripple conditioned by SPM [3,4]. Nevertheless, for many industrial applications it is very promising to develop ripple-free ultrafast high peak power fiber lasers, because such sources are very efficient, robust and easily assembled [5–8]. Eventually, the investigation of the origin of the ripples and its evolution inside CPA lasers is an important task.
The evolution of rippled optical pulses due to SPM inside a CPA laser can be computed by solving numerically a Nonlinear Schrödinger Equation (NLSE). This method works as a black-box computation, a certain rippled pulse is considered as the input data and the resulting pulse is simulated. This procedure cannot provide a physical insight into the phenomenon of the ripple increase due to SPM. Some meaningful questions can be answered only by mathematical analysis: why the spectrum amplitude ripple is changing with power scaling? Are the ripple on different frequencies sense the SPM effect the same way or not? These problems will be addressed in detail in the next sections. Several attempts were made before to explain the SPM effect from the point of view of spectral phase and amplitude ripples [9–11]. In a current paper the approach for the ripple description and analysis is different. We will start to investigate both amplitude and phase ripples of the pulse complex envelope in time domain. Later it will be demonstrated that such approach is natural because these two kinds of the ripple play the same role in the pulse quality degradation at low energies and the approach is fruitful for pulse retrieving problem.
First of all, the formula that establish the relation between the time domain and spectral domain ripples of a chirped pulse is derived. This formula can be interpreted as an application of the spectral filter with a particular shape that acts not in an optical spectrum domain, but in the domain of Fourier harmonics of the pulse spectrum amplitude functions. In detail this is described in the next section. After that a simple quantitative explanation of the evolution of the ripples in Kerr nonlinear medium is provided, influence of group velocity dispersion effect (GVD) is discussed. Moreover, such knowledge of the interrelation between amplitude and phase ripples in time and spectral domains together with the evolution of the ripples in a time domain reveals the solution of a pulse retrieving problem, relying only onto two spectra of the chirped pulse measured at the input of a Kerr nonlinear medium and at its output. According to my knowledge, such chirped pulse retrieving algorithm is so far unknown.
2. Time and spectral domains
A linearly polarized optical pulse propagating in a single mode fiber can be completely described by the carrier frequency value ${\omega _c}$ and the complex envelope function $A\left ( t \right )$. Equivalently in a frequency domain the pulse is characterized by the complex spectrum function $S\left ( \omega \right )$. Further we will use the words “shape“ or “amplitude“ for denoting the absolute values either of these functions. As a rule, the spectrum shape $\left | {S\left ( \omega \right )} \right |$ can be easily measured using conventional laboratory equipment, e.g. optical spectrum analyzer. Typically when measured the spectrum shape function is centered at a carrier frequency. Further we will consider the spectrum function $S\left ( \omega \right )$ of the pulsed laser radiation in reference to an optical carrier frequency, i.e. oscillations at the carrier frequency are excluded from analysis. Then the value $S\left ( 0 \right )$ stands for the carrier harmonic, and for the transform-limited pulse the complex envelope $A\left ( t \right )$ is a pure real function.
The current paper is motivated by the investigation of fiber chirped pulse amplification laser. There stretcher typically is implemented with the diffraction gratings pair, linearly chirped fiber Bragg grating, chirped volume Bragg grating or even a fiber coil. Nevertheless for major variety of implementations stretcher can be mathematically represented as on operator, inducing only a second order dispersion on an optical pulse. The reason is simple — second order dispersion most effectively stretches the pulse in time domain. Third and high order dispersion are most commonly a next order perturbation of the spectrum phase, but sometimes it is possible to obtain a fruitful trade-off between high order dispersion and nonlinear effects [12]. It is well known, that the spectrum amplitude of the laser pulse with a significant accumulated group delay dispersion matches its complex envelope amplitude in a time domain [13]. However, this approximation is correct only to a certain extent and mathematical analysis should be carried out in order to verify the limits of its validity.
Three parameters that characterize optical pulses $A\left ( t \right ),\;S\left ( \omega \right ),\;{D_2}$, where the last one is an accumulated GDD of the laser pulse (measured in ${ps^2}$), are interconnected by the inverse Fourier transform:
In Eq. (1) there is a Fresnel type integral, which can be approximated using the stationary phase method [15]. The main idea of this approximation is that due to a large absolute value of ${D_2}$ the exponent term under the integral rapidly oscillates for every frequency $\omega$ except the small interval near the stationary point ${\omega _s} = t/{D_2}$. An overall contribution of the oscillating intervals into the integration result is zero. Thus, the spectrum shape $\left | {S\left ( \omega \right )} \right |$ can be approximated by its value in the stationary point $\left | {S\left ( {{\omega _s}} \right )} \right |$ and the Eq. (1) can be rewritten as:
Due to the fact that it is impossible to obtain infinitely large group delay dispersion parameter ${D_2}$ the stationary phase approximation is valid only for slowly varying spectrum amplitudes $\left | {S\left ( \omega \right )} \right |$ and, as a consequence, $A\left ( t \right )$ functions. For rapidly varying components of the complex envelope, which are called ripples, this approximation is very rough. Obviously, in a general case the ripples of the spectrum amplitude and the pulse complex envelope amplitude will not coincide. In the next sections the criterion of validity of the Eq. (3) will be introduced and the connection formula between the ripple shapes in time and frequency domains will be derived.
3. Amplitude ripples in a time domain
In this section long chirped pulses with a flat top will be considered in order to avoid dealing with edge effects. However, later it will be shown that the results obtained here are actually valid for the pulses with arbitrary shapes. It is also assumed that optical pulses have only an amplitude ripple of the complex envelope $A\left ( t \right )$ and the chirp of the pulse is strictly linear. It means that the phase-frequency characteristic of the pulse complex envelope is parabolic. In this case the formula that interconnects the pulse spectrum and its shape in a time domain can be written as follows:
Generalizing the result in Eq. (9) for the Fourier transform case, one can obtain that for the linearly chirped pulse with a rippled flat top its spectrum shape $\left | {S\left ( \omega \right )} \right |$ is connected with the pulse shape $\left | {A\left ( t \right )} \right |$ by the formula
The Eq. (10) can be interpreted in such a way that the spectral shape of the rippled pulse coincides with its complex envelope shape modified by the special spectral filter. Here the term “spectral“ does not imply an optical spectral domain, but the domain of the mathematical decomposition of the pulse complex envelope amplitude into harmonic functions.
From the Eq. (10) it also follows that the ripple harmonic components of the complex envelope amplitude with the time period $T \gg 2\sqrt {\pi {D_2}}$ coincide with the ripple harmonics of the spectrum amplitude. It means that the Eq. (10) is applicable for arbitrary shaped pulses, not only flat top ones, because the pulse shape itself can be interpreted as a low frequency amplitude ripple.
Using numerical simulations it is easy to verify the Eqs. (3) and (10), for this purpose the ripple function $a\left ( t \right )$ is needed. Here and elsewhere in this paper ripple function is numerically generated by the following way: random variable with uniform distribution $U\left ( -l, l \right )$ is sampled for every $0.24\;ps$, then this array is low-pass filtered (first-order FIR filter) with cut-off frequency $f_{co}\;[ps^{-1}]$. Thus a random ripple function has two generating parameters: $l$ and $f_{co}$. Note, that ripple function root mean square value or peak-to-peak value linearly depends on $l$ parameter.
The complex envelope shape and the spectrum shape of the rippled pulse are plotted in Fig. 1. Tanh-shaped pulse with pulse duration ${\Delta }{t_{FWHM}} = 500\;ps$, top amplitude ${A_0} = 1\;{W^{1/2}}$, group delay dispersion ${D_2} = - 40\;{ps^2}$, ripple generating parameters $l = 0.1,\; f_{co} = 0.033\;ps^{-1}$ was used for the simulation. It is clear that slowly varying parts of these two functions coincide, Fig. 1(a), according to the Eq. (3). On the contrary the ripple parts of these functions do not match each other, Fig. 1(b), which points that simple statement ”chirped pulse shape matches its spectrum shape” is not valid for rippled pulses.
The Eq. (10) is a more general relation then (3) that connects the rippled complex envelope shape and rippled spectrum shape of the pulse. The result of its application is demonstrated in Fig. 2. The thick yellow curve represents the complex envelope shape with the spectral filter applied $\sqrt {\left | {{D_2}} \right |} {\cal F}_\omega ^{ - 1}\left \{ {\cal F}\left \{ {\left | {A\left ( t \right )} \right |} \right \}\left ( \omega \right )\cos \left ( {{{\omega ^2}{D_2}}}/{2} \right ) \right \}\left ( t \right )$, and the blue one represents the spectrum shape with substituted argument $\left | {S\left ( {t/{D_2}} \right )} \right |$ that was numerically calculated via Fast Fourier Transform of $A\left ( t \right )$ similar to that shown in Fig. 1. It can be clearly seen that in contrast to Fig. 1(b) at this time two curves are in very good agreement, so that it confirms the validity of the relation (10).
For producing Figs. 1 and 2 the ripple generating parameters were $l = 0.1,\; f_{co} = 0.033\;ps^{-1}$, which stands for a rather small ripple of $\sim 3\%$ in respect to the average amplitude ${A_0}$. In this case the assumptions made for the derivation of the Eq. (10) were satisfied, as an excellent match is observed between yellow and blue curves. However, this formula still works well for much greater ripples. For instance, on Figs. 3(a) and 3(b) the same ripple function as for the Fig. 2(b) is scaled to $\sim 60\%$ and $\sim 150\%$ in respect to the average amplitude respectively. It is clear that Eq. (10) continues to provide good qualitative agreement between the pulse shape with the spectral filter applied, and the spectrum shape with substituted argument. If not presented by a thick line, one could only find a difference between this two curves on Fig. 3 at their peaks and depths.
In order to estimate the errors of approximation, used for derivation of the Eq. (10), it should be investigated for the variety of random ripple functions. Then let us define a dimensionless error functional as
The Monte Carlo method was used to find the expected value and the variance of the $err$ functional for various ripple parameters $l \in [0,5]$, $f_{co} = \{ 0.033\;ps^{-1}, 0.1\;ps^{-1}\}$ ($l$ parameters larger than $5$ makes no sense, as the ripple becomes greater then $A_0$). The pulse was again tanh-shaped with pulse duration ${\Delta }{t_{FWHM}} = 500\;ps$, top amplitude ${A_0} = 1\;{W^{1/2}}$, group delay dispersion ${D_2} = - 40\;{ps^2}$. The results are presented on a Fig. 4.
For the $f_{co} = 0.033\;ps^{-1}$ blue curve lies below $0.05$, which means that we could expect maximum $5\%$ root mean square error when applying the Eq. (10) for pulses with such low frequency ripple. For the $f_{co} = 0.1\;ps^{-1}$ the error is worse, but taking in account that increasing $x$ times $f_{co}$ also translates in ripple function RMS growth in $\sqrt {x}$ times, as if it is generated by $\sqrt {x} l$ uniform distribution parameter, one can find, that high frequency ripple practically makes no income to the $err$. In other words frequency of the ripple does not influence the accuracy of the Eq. (10), only root mean square value does.
Obviously, the relation (10) can be also used for the solution of an inverse problem. For example, if it is known that there is only an amplitude ripple of the pulse complex envelope function, then this function can be characterized simply by measuring the spectrum amplitude of the pulse.
Eventually, the following criterion of validity of the Eq. (3) can be formulated: if the Fourier image of the spectrum amplitude of the pulse ${{\cal F}_t}\left \{ {\left | {S\left ( {t/{D_2}} \right )} \right |} \right \}\left ( \omega \right )$ only comprises low frequency harmonics that satisfy the condition $\left | \omega \right | \ll \sqrt {\pi /{D_2}}$, then its complex envelope amplitude $\left | {{A}\left ( {t} \right )} \right |$ can be obtained using the Eq. (3) and the pulse chirp is quite accurately linear.
4. Phase ripples in a time domain
The considerations similar to those introduced in the previous section can be carried out for the pulses with the strictly flat top amplitude ${A_0}$, but with a perturbed linear chirp of the complex envelope, i.e. with the phase ripple of the complex envelope function:
Consequently, the Eq. (14) can be used for the characterization of the pulse complex envelope function simply relying on the measured spectrum amplitude, providing that the pulse complex envelope has only a phase ripple.
It should be noticed, that there is an ambiguity in the definition of the function ${\left | S \right |_{ripple}}\left ( \omega \right )$. There are numerous ways for splitting the pulse spectrum amplitude into slowly and rapidly varying parts. Here the following recommendation can be introduced. As it is known from the previous chapter, the spectrum amplitude harmonics with frequencies $\left | \omega \right | \ll \sqrt {\pi /{D_2}}$ are directly transformed into the complex envelope amplitude harmonics of the pulse. According to this fact we can conventionally refer harmonics with frequencies $\left | \omega \right | \le {1}/{2}\sqrt {\pi /{D_2}}$ to the slowly varying part ${\left | S \right |_{slow}}\left ( \omega \right )$ and, consequently, those with frequencies $\left | \omega \right | > {1}/{2}\sqrt {\pi /{D_2}}$ to the ripple part ${\left | S \right |_{ripple}}\left ( \omega \right )$. Therefore a natural way to extract the rippled part of the spectrum amplitude $\left | {S\left ( {t/{D_2}} \right )} \right |$ is to apply a high-pass filter with the cut-off frequency that is equal to ${1}/{2}\sqrt {\pi /{D_2}}$.
Provided recommendations do not completely eliminate ambiguity, as there are multiple implementations of high pass filters and a cut-off frequency itself was chosen rather frivolously. In practice this ambiguity leads to the discrepancy between the spectrally filtered phase ripple function and the spectrum amplitude ripple function ${\left | S \right |_{ripple}}\left ( {{t}/{{{D_2}}}} \right )$ in those regions, where the spectrum shape changes drastically, e.g. the edges of a tanh-shaped pulse. There is no obvious solution for this problem, because it originates from the fundamental question of how to extract an original signal from the distorted (rippled) measurement data.
5. Evolution of the ripple
In the current section we will investigate ripple evolution in a Kerr nonlinear medium. In order to obtain the propagation equation for the ripple, we shall start from the propagation equation for an optical pulse — Nonlinear Schrödinger Equation. The ripple evolution, discussed in this section, is interesting only as basement for development of the pulse retrieving problem algorithm, where group velocity dispersion effect is unlikely and adverse. Later in this section we will discuss the quantitative parameter showing the influence of the GVD on the ripple, and the approach for the minimization of GVD effect will be presented. Thus in the current section we will consider only a self-phase modulation effect occurring during the propagation of the linearly polarized optical pulse with the complex envelope $A\left ( {t,z} \right )$ along the z-axis. Then the NLSE is simplified to the following propagation equation
GVD effect can be eliminated from pulse propagation by increasing the effective mode area of the radiation inside the Kerr nonlinear medium, as $\gamma$ is inversely proportional to effective mode area. For the bulk medium it can be done by collimating the beam in larger diameter.
Using the fact, that the function in parentheses $r + {r^*} = 2\operatorname {Re}r$ in the Eq. (19) is the integral of motion, the solution of this differential equation is:
6. Pulse retrieval
As it was introduced above, the amplitude ripple of an optical pulse complex envelope leads to the ripple in its spectrum amplitude according to the Eq. (10). In turn, the phase ripple of the pulse complex envelope should be treated using the Eq. (14). In general, when the ripple of the pulse is experimentally observed using a spectrum analyzer, it is impossible to clarify the ripple origin and, consequently, to determine the pulse complex envelope in a time domain. Nevertheless, the theoretical consideration introduced below reveal that the complex envelope of an arbitrary rippled chirped pulse can be retrieved using a spectrum analyzer and a Kerr nonlinear medium. According to the Eq. (21) after the transmission through the Kerr nonlinear medium of the $z$ meters length the difference between two measured spectrum amplitudes can be expressed as:
Eliminating the exponent, using the linearity of Fourier transform and the approximation $\left | {\alpha + \beta } \right | \approx \left | \alpha \right | + \left | \beta \right |\cos \left ( {\arg \alpha - \arg \beta } \right )$, we can rewrite the Eq. (22) as follows:
Zero values of the sine function in (25) formally mean that the infinitely countable discrete set of values ${{\cal F}_t}\left \{ {{{\left | {{A_{slow}}} \right |}^2}\operatorname {Re}{r_{in}}} \right \}\left ( {{\omega _z}} \right )$ is undetermined. Fortunately we can recover the ${{\cal F}_t}\left \{ {{{\left | {{A_{slow}}} \right |}^2}\operatorname {Re}{r_{in}}} \right \}\left ( \omega \right )$ function using its continuity property. For example, ${{\cal F}_t}\left \{ {{{\left | {{A_{slow}}} \right |}^2}\operatorname {Re}{r_{in}}} \right \}\left ( 0 \right )$ is undetermined, it means that we cannot find a constant part of $\operatorname {Re}{r_{in}}$ using the Eq. (26). However, this part is zero by the definition of a ripple. This issue means that in practice an inaccuracy of deriving the ${{\cal F}_t}\left \{ {{{\left | {{A_{slow}}} \right |}^2}\operatorname {Re}{r_{in}}} \right \}\left ( \omega \right )$ function is greater around ${\omega _z}$ arguments.
Naturally, the Eq.s (25) and (26) are valid for a small ripple. As the phase ripple tends to grow with the pulse energy and the length of the medium (see the Eq. (21)), in experiments one should try to minimize these gaining factors of the ripple. In addition, the Eq. (26) makes sense only for the time intervals corresponding to a nonzero pulse amplitude, as the division by ${\left | {{A_{slow}}} \right |^2}$ takes place.
A verification of the Eq. (26) was conducted using numerical simulations for the pulse with the arbitrary assigned amplitude and phase ripple ${r_{in}}$ of the complex envelope $A\left ( {t,0} \right )$. The $A\left ( {t,z} \right )$ function obtained as a numerical solution of the Eq. (16) was used for the calculation of the $\left | {S\left ( {{t}/{{{D_2}}},z} \right )} \right |$ function via Fast Fourier Transform. Modeling results are introduced in Fig. 6, where two curves representing the left hand side and the right hand side of the Eq. (26) are plotted together, $\gamma = 1.25\;W/km,\;z = 0.25\;km$. Except the constant value ambiguity and the edge effects, discussed earlier in this section, it can be seen, that a good matching is observed between two curves. This fact actually proves the validity of the Eq. (26).
It is clear, that the right-hand side of the Eq. (26) comprises parameters that can be measured experimentally. As the low-frequency components of the spectrum amplitude $\left | {S\left ( \omega \right )} \right |$ and the pulse amplitude $\left | {{A_{slow}}\left ( t \right )} \right |$ coincide, according to the Eq. (3), then the input amplitude of the chirped pulse complex envelope follows the formula
The solvability of the pulse retrieving problem using only two spectra of the pulse measured before and after the transmission through a Kerr nonlinear medium was demonstrated earlier using a complex iterative algorithm [18–20] or even a cross-phase modulation [21]. Especially it should be noted, that in [18] it is shown, that group velocity dispersion impact is small for the pulse retrieving procedure, and it also could be even more reduced by the proper choice of Kerr nonlinear medium. However, an analytic approach for the pulse retrieving problem, which is introduced in the present paper, was made for the first time, as a result two simple Eqs. (26) and (27) were derived.
7. Conclusions
A theoretical analysis of the chirped optical pulse amplitude and phase ripples was performed. It was demonstrated that an amplitude ripple of the pulse complex envelope leads to the ripple of its spectrum amplitude. The interrelation between these ripples is described by the Eq. (10). In turn, a pure phase ripple of the pulse complex envelope also induces the spectrum amplitude ripple that is described by the Eq. (14). An evolution of the amplitude and the phase ripples of the chirped pulse in Kerr nonlinear medium was considered. As a result, for the first time to my knowledge an experimentally simple retrieving procedure of the linearly chirped pulse complex envelope was analytically described. The facilities required for the pulse retrieval are a spectrum analyzer and the Kerr nonlinear medium, which induces only a self-phase modulation of the transmitted pulse. The absolute value of the pulse complex envelope can be determined using the Eqs. (3), (26) and (27). A principle opportunity of the application of the pulse retrieving procedure is theoretically proved by modelling, still experimental verifications of the introduced approach should be carried out in future.
References
1. T. Eidam, J. Rothhardt, F. Stutzki, F. Jansen, S. Hädrich, H. Carstens, C. Jauregui, J. Limpert, and A. Tünnermann, “Fiber chirped-pulse amplification system emitting 3.8 gw peak power,” Opt. Express 19(1), 255–260 (2011). [CrossRef]
2. J. Limpert, F. Roser, T. Schreiber, and A. Tunnermann, “High-power ultrafast fiber laser systems,” IEEE J. Sel. Top. Quantum Electron. 12(2), 233–244 (2006). [CrossRef]
3. I. C. M. Littler, L. Fu, and B. J. Eggleton, “Effect of group delay ripple on picosecond pulse compression schemes,” Appl. Opt. 44(22), 4702–4711 (2005). [CrossRef]
4. I. C. M. Littler, L. Fu, M. Lee, and B. Eggleton, “Investigation of single harmonic group delay ripple on picosecond pulses using frog: Tailoring pulse bursts,” Opt. Commun. 265(1), 147–152 (2006). [CrossRef]
5. A. Yusim, I. Samartsev, O. Shkurikhin, D. Myasnikov, A. Bordenyuk, N. Platonov, V. Kancharla, and V. Gapontsev, “New generation of high average power industry grade ultrafast ytterbium fiber lasers,” Proc. SPIE 9728, 972839 (2016). [CrossRef]
6. S. Frankinas, A. Michailovas, N. Rusteika, V. Smirnov, R. Vasilieu, and A. L. Glebov, “Efficient ultrafast fiber laser using chirped fiber bragg grating and chirped volume bragg grating stretcher/compressor configuration,” Proc. SPIE 9730, 973017 (2016). [CrossRef]
7. Z. Xin, Y. Zhi, Q. Liang, F. Li, Y. Xiaojun, W. Yishan, and Z. wei, “Pulse duration tunable fiber cpa system based on thermally dispersion tuning of chirped fiber Bragg grating,” Optik 127(20), 8728–8731 (2016). [CrossRef]
8. J. Limpert, T. Clausnitzer, A. Liem, T. Schreiber, H.-J. Fuchs, H. Zellmer, E.-B. Kley, and A. Tnnermann, “High-average-power femtosecond fiber chirped-pulse amplification system,” Opt. Lett. 28(20), 1984–1986 (2003). [CrossRef]
9. D. N. Schimpf, E. Seise, J. Limpert, and A. Tünnermann, “The impact of spectral modulations on the contrast of pulses of nonlinear chirped-pulse amplification systems,” Opt. Express 16(14), 10664–10674 (2008). [CrossRef]
10. D. N. Schimpf, E. Seise, J. Limpert, and A. Tünnermann, “Decrease of pulse-contrast in nonlinear chirped-pulse amplification systems due to high-frequency spectral phase ripples,” Opt. Express 16(12), 8876–8886 (2008). [CrossRef]
11. C. Dorrer and J. Bromage, “Impact of high-frequency spectral phase modulation on the temporal profile of short optical pulses,” Opt. Express 16(5), 3058–3068 (2008). [CrossRef]
12. Y. Zaouter, J. Boullet, E. Mottay, and E. Cormier, “Transform-limited 100 $\mu$j, 340 mw pulses from a nonlinear-fiber chirped-pulse amplifier using a mismatched grating stretcher–compressor,” Opt. Lett. 33(13), 1527–1529 (2008). [CrossRef]
13. A. Mahjoubfar, D. V. Churkin, S. Barland, N. Broderick, S. K. Turitsyn, and B. Jalali, “Time stretch and its applications,” Nat. Photonics 11(6), 341–351 (2017). [CrossRef]
14. G. Rousseau, N. McCarthy, and M. Piché, “Description of pulse propagation in a dispersive medium by use of a pulse quality factor,” Opt. Lett. 27(18), 1649–1651 (2002). [CrossRef]
15. N. Bleistein, Asymptotic Expansions of Integrals (Holt, Rinehart and Winston, 1975).
16. G. P. Agrawal, Nonlinear Fiber Optics (Academic Press, 2007).
17. J. R. Fienup, “Phase retrieval algorithms: a comparison,” Appl. Opt. 21(15), 2758–2769 (1982). [CrossRef]
18. E. A. Anashkina, A. V. Andrianov, M. Y. Koptev, and A. V. Kim, “Complete field characterization of ultrashort pulses in fiber photonics,” IEEE J. Sel. Top. Quantum Electron. 24(3), 1–7 (2018). [CrossRef]
19. E. A. Anashkina, V. N. Ginzburg, A. A. Kochetkov, I. V. Yakovlev, A. V. Kim, and E. A. Khazanov, “Single-shot laser pulse reconstruction based on self-phase modulated spectra measurements,” Sci. Rep. 6(1), 33749 (2016). [CrossRef]
20. E. T. J. Nibbering, M. A. Franco, B. S. Prade, G. Grillon, J.-P. Chambaret, and A. Mysyrowicz, “Spectral determination of the amplitude and the phase of intense ultrashort optical pulses,” J. Opt. Soc. Am. B 13(2), 317–329 (1996). [CrossRef]
21. J. J. Ferreiro, R. de la Fuente, and E. López-Lago, “Characterization of arbitrarily polarized ultrashort laser pulses by cross-phase modulation,” Opt. Lett. 26(13), 1025–1027 (2001). [CrossRef]