A wavefront sensor has been used to measure the Kerr nonlinear focal shift of a high intensity ultrashort pulse beam in a focusing beam geometry while accounting for the effects of plasma-defocusing. It is shown that plasma-defocusing plays a major role in the nonlinear focusing dynamics and that measurements of Kerr nonlinearity and ionization are coupled. Furthermore, this coupled effect leads to a novel way that measures the laser ionization rates in air under atmospheric conditions as well as Kerr nonlinearity. The measured nonlinear index n2 compares well with values found in the literature and the measured ionization rates could be successfully benchmarked to the model developed by Perelomov, Popov, and Terentev (PPT model) [Sov. Phys. JETP 50, 1393 (1966)].
© 2012 Optical Society of America
The optical Kerr effect gives rise to many of the phenomena in nonlinear optics. Its associated nonlinear index of refraction n2 leads to self-focusing  in high power laser beams through index variation n(I) = n0 + n2I over the spatially varying intensity profile I. This in turn leads to a plethora of effects such as filamentation [2, 3], self-phase modulation , spectral broadening , and self-compression [6, 7]. Measurements of n2 in air have been made but these values vary significantly, in part due to differences in wavelength and pulse duration (see table 1). The variations due to wavelength are the result of dispersion, with n2 decreasing at increasing wavelength . Variations due to pulsewidth stem from the combination of pure Kerr effect (dominant at short pulsewidths ≤ 100 fs) and molecular alignment from rotational Raman effects (noticeable at a few 100 fs ), causing n2 to increase with pulsewidth. As seen from table 1, wide scatter in measured data can mask these effects. A direct measurement of the index of refraction is therefore highly desirable for a given wavelength, pulse duration, and pressure. We employ the self-focusing of a high power ultrashort pulse laser beam in order to measure the nonlinear index of refraction. Previous work by [9,10] in a focusing beam geometry used the focal shift at P > Pcrit to calculate the critical power for self focusing and hence the nonlinear index of the medium. For Liu et al , as the laser beam power increases, the focal shift with respect to the geometrical focus is determined by measuring the spatial shift of the plasma fluorescence in air. The power at which a shift is first observed is then identified with the critical power. A fundamental drawback of this method is the fact that the plasma will have an effect on the focusing behavior which is not taken into account (as pointed out by ). In fact, as we show, plasma effects occur at much lower powers than the n2 based Pcrit. Hence the onset of self-focusing can be difficult to isolate in observations. In this paper, we have appropriately modeled plasma de-focusing in conjunction with self-focusing and as a result we can extract both laser ionization rates and n2 from our measurements. While n2 measurements are readily available at 800 nm for sub-picosecond pulses (see Table 1), little n2 data exists in the literature for our wavelength (1054 nm) and pulsewidth (540 fs) of interest. Even though the present paper discusses measurements at 1054 nm only, its technique should be applicable to a broad range of wavelengths and pulsewidths.
2. Experimental setup
We propose a new method for measuring n2 and ionization rates which allows us to detect beam shifts at much lower power (P ≪ Pcrit). In our setup (see Fig. 1), a wavefront sensor is used to directly sense a shift in focal length as opposed to indirectly measuring shifts via a secondary effect such as plasma fluorescence. This is critically important because plasma induced focal shifts occur well before plasma fluorescence can be detected.
A short pulse laser at low power P ≪ Pcrit is focused with a focal element M1 (focal length f1) into a gas ionization cell and re-collimated with a second focal element M2 (focal length f2). A wavefront sensor (WFS) is placed at a distance d behind M2 and a flat wavefront is recorded as a reference (see Fig. 2(a)) at minimum powers. Finally, to avoid any nonlinear effects in the WFS filters, the front surface reflections of two uncoated wedges were used to attenuate the beam energy prior to any filters. For the same reason, great care was taken to not use any transmissive optics in the short pulse beam train. Energy was adjusted through a wave-plate/polarizer combination in the temporally stretched part of the laser system. Furthermore, only spherical mirrors were used in the focusing and re-collimation of the laser beam (see M1 and M2 in Fig. 1) in order to avoid nonlinear phase accrual in the lens material which may compromise data. The use of spherical mirrors in down-collimation requires a small off-axis angle which could induce astigmatism. As a result of our large f/#, the angles were kept to less than 1°, yielding a negligible measured astigmatism with the wavefront sensor.
As the laser beam power increases, the beam focus will shift by a distance Δ toward M1. Correspondingly, the light emanating from the new self-focused position f1 – Δ or f2 + Δ will not be collimated any more but rather be imaged a certain distance D behind M2. This distance D is just the radius of curvature R of the wavefront (as measured by the wavefront sensor; see Fig. 2(b)) plus the distance d of the sensor to the lens M2: D = R+d. The focal shift Δ can then be calculated from the thin lens equation:
The wavefront radius R at position d can be calculated from the measured defocus Zernike term (see Fig. 2(b)). In our case , where a is the pupil radius. For a measured pupil of radius a on the wavefront sensor and a defocus of in waves one calculates:
The validity of this approach was cross-checked by ZEMAX  modeling. A test measurement with the following parameters was taken: f1 = 1 m, f2 = 0.3 m, d = 0.3 m, and a = 1.75 mm. A defocus term of waves was recorded which corresponds to a focus shift of Δ = 9 mm. This result has been corroborated using a ZEMAX model where M1 was modeled to have a focal length of f1 – Δ = 99.1 cm. All other parameters mirrored the experimental setup. The model shows indeed a defocus of 0.068 waves when the modified M1 focal length was used. One can conclude that the above method should be valid for determining the focal shift as a function of laser beam power.
3. Experimental results
A laser beam at λ = 1054 nm with a pulsewidth of τ = 540 fs FWHM and a spatial first order Gaussian with a 1/e2 waist radius of wi = 6.2 mm was focused by spherical concave mirrors of f/120 and f/40. A larger f/# (greater than f/120) was used as well but led to the onset of filamentation at low laser energies. Each data point represents an average of 200 measurements at 10 Hz on the wavefront sensor (Phasics SID-4, see http://www.phasicscorp.com/) as well as the energy meter. The peak to valley “defocus” Zernike term was then fitted to the averaged wavefront and converted to focal shift as discussed earlier. The error bars in Fig. 3 represent the standard deviation of the energy spread over 200 shots and the achievable focal shift accuracy is based on the sensitivity of the wavefront sensor. Note that the wavefront sensor amplitude maps (beam near fields) were concurrently measured with the phase maps. Care was taken to assure that the amplitude maps remained Gaussian and that the near field diameter remained self-consistent throughout a scan. Measurements were stopped once changes to amplitude profile or beam size were detected.
Figure 3 shows the measured focal shift versus laser input energy for the two f/# mentioned earlier. One can see that, depending on f/#, the onset (elbow in the graph) of the focal shift is different for any given focus geometry. Two main processes contribute to focal shift:
- Nonlinear self-focusing: As the laser beam power P becomes comparable to the critical power Pcrit = 3.77λ/(8πn0n2), the nonlinear index of refraction n2 leads to additional focusing in beams with Gaussian (and also other) intensity profiles. For a collimated input beam entering a nonlinear medium and an unchanged amplitude profile during self-focusing, one can write the self-focusing distance zsf due to the nonlinear index n2 as :
Based on Eq. (3), a focal shift can only occur at P ≥ Pcr which would be independent of focus geometry (see dashed lines in Fig. 3). Liu et al  uses this argument when he identifies this elbow as the onset of P = Pcr in order to calculate n2. According to measured data in Fig. 3, the elbow is clearly dependent on beam focusing geometry and hence one cannot consider the effects of nonlinear self-focusing alone.
- Ionization: It is clear that ionization must occur as the laser beam propagates towards the minimum beam waist. This ionization will be highest around the focal region and will profoundly affect the self-focusing behavior . As the beam collapses towards the focus the electron density increases as a result of ionization due to increased beam intensity. This leads to plasma de-focusing before the geometrical focus is reached. Hence ionization will lead to a focal shift toward the input focal lens as will nonlinear self-focusing due to the nonlinear index n2 (see Fig. 4).
It can be seen that one has to consider both processes in order to adequately explain the results presented in Fig. 3. One should note that for f/40, where ionization is dominant, the focal shift begins at lower power but shows smaller shifts at larger powers as compared to the regime where nonlinear self-focusing is dominant (f/120).
4. Experimental analysis
In order to extract values for both n2 and the ionization rates, one has to model the beam waist as a function of laser power P and laser induced ionization. We used the semianalytical model employed in [4,11,14] and it was shown that it adequately reproduces the beam waist evolution until the minimum waist is reached  which is the regime of interest. The laser beam waist w(z,t) as a function of time and propagation distance z can be written as:15]: Eq. (5) simplifies to:
An initial rough estimate of the 11th order ionization coefficient σ(K=11) can be calculated based on the weak field approximation of the Keldysh theory :
For a square temporal pulse shape of full width τ, Eq. (5) can be integrated to yield:Equation 9 can be solved numerically with initial conditions: wi = w(z = 0) and dw(0)/dz = −wi/f. Note that Eq. (9) now accounts for saturation effects by using the integrated solution of Eq. (5) as inserted into Eq. (4). The assumption of a square pulse has allowed the generation of Eq. (9) in a relatively simple analytical form. A Gaussian pulse profile would be more realistic but yields a complex form for the integrated solution to Eq. (4) which causes numerical integration issues in Eq. (9).
Figure 4 depicts an example for a calculated laser beam waist versus propagation distance for a f/120 focusing geometry at various laser input energies. It can be seen that for a low energy of 2 mJ the geometrical focus coincides with the minimum beam waist. For larger energies/powers, the focus moves toward the input optic as a result of self-focusing and ionization and the minimum waist grows as a result of laser generated plasma de-focusing. Based on Eq. (9), one can now calculate the position of the minimum beam waist (and hence the focal shift Δ) as a function of laser energy/power and compare the model to the measured data in Fig. 3.
Since there exist two data sets and two unknowns (n2 and σ(K)), one should be able to find a unique solution that fits the measurement. This is accomplished rather easily since the physics in the f/40 focusing geometry is largely dominated by ionization which sets the limit for σ(K). Conversely, self-focusing plays a dominant role in the f/120 focusing geometry which leads to a determination of n2. The higher influence of ionization in the f/40 case is evidenced by the fact that the onset of a focal shift (elbow) occurs at much lower powers.
The importance of ionization for the correct estimation of n2 and σ(K) is illustrated by the dashed lines in Fig. 3. They are based on theoretical curves that do not include ionization. As expected, those curves are shifted to the right because the focal shift contribution from the laser plasma is omitted. If only one f/# measurement is considered (as is the case in ), one is inclined to identify the elbow in the curve as Pcrit, leading to an erroneously low Pcrit (or conversely high n2).
The solid lines in Fig. 3 show the calculated focal shift versus energy for n2 = (2.6 ± 0.2) × 10−23 m2/W and σ(11) = (3 ± 1.5) × 10−191m22W−11s−1. While the value for n2 compares well with other data from the literature (see Table 1), one should note that, at our pulsewidth of 540 fs, the measured n2 is a combination of pure optical Kerr effect and Raman contributions. This is also evidenced by the fact that the 1 ns result for 1053 nm in table 1 is larger than our number due to its greater Raman contribution. On the other hand, one should be aware that the measured ionization coefficient for the 11th order process is 10 orders of magnitude lower than expected from Eq. (7). The reason for this discrepancy will become apparent at the end of this section.
In order to gain a better understanding of the validity of the measured ionization coefficient, one has to investigate the electron density contribution in some more detail. Figure 5 shows the calculated peak laser focal intensity that corresponds to each input energy in the experiment. As expected, one can see that intensity clamping occurs for each f/# almost at the same level. Furthermore, the intensity rises more strongly for the f/40 case than for f/120. This is ultimately the reason why the f/40 case is ionization dominated, because the early rise in intensity causes ionization to be more prominent at lower energy. Furthermore, one can see that the intensity clamping of Icl = 3.5 × 1018 W/m2 occurs at the elbow in both measured data sets.
At this clamping level, one has to consider the contribution of tunneling ionization. To see this one must have a closer look at the Keldysh (or adiabatic) parameter γ defined in Appendix A. For γ ≫ 1 MPI is the dominant ionization process; for γ ≪ 1 tunneling ionization is more important. At the above mentioned Icl = 3.5 × 1018 W/m2, γ corresponds to γ < 0.5. In order to obtain a clear picture on how MPI and tunneling ionization contribute for our intensity regime of interest, we have modeled the ionization rates for oxygen and nitrogen based on the PPT theory [4, 21–23] (see Appendix A). Figure 6 shows the ionization rate R versus laser intensity for oxygen and nitrogen. Note, that dNe/dt = (NO2 – Ne) × R.
One strong feature of the PPT theory is the fact that it is able to show the transition of ionization rates from an MPI dominated regime into the tunneling regime. To show that the model is self-consistent, we have plotted the ionization rate for oxygen at 800 nm and 1054 nm. One can see that the slope for the 800 nm case is lower than the 1054 nm case due to the fact that ionization at 800 nm requires K = 8 photons as opposed to K = 11 in the other case. Consequently, one observes that the ionization rate is higher for 800 nm (higher photon energy and lower ionization order) versus 1054 nm. In addition, both curves merge as they approach the tunneling ionization regime. This due to the fact that tunneling ionization is independent of laser wavelength (photon energy) but rather is driven by the amplitude of the electric field. Comparing ionization rates of oxygen and nitrogen at 1054 nm, one can see that the slope for nitrogen is steeper than for oxygen. This is again due to the fact that nitrogen requires K = 14 photons to ionize at 1054 nm rather than K = 11 for oxygen. In addition, we have plotted the ionization rate (dashed line) that corresponds to the Keldysh approximation in Eq. (7). As expected, for a weak field approximation, the Keldysh rate is the same as the PPT rates for weak electric fields (meaning low intensities). The excellent agreement between the Keldysh number and the PPT model at low intensities shows that both models are self-consistent. Looking at the ionization rate for oxygen in Fig. 6 one can observe that an 11th order slope (see dashed line) is only valid for very low laser intensities I ≪ 1017 W/m2. However, looking at the ionization dominated f/40 focusing geometry in Fig. 5, one can see that intensities above 1018 W/m2 are reached at very low laser input energies. This means that tunneling ionization is the dominant effect (see Fig. 6) and that MPI alone does not fully apply. However, for the commonly used intensity regime (≤ 5×1017 W/m2) in the semi-analytical model of Eq. (9), MPI is the electron density source term. Ideally, one should use the intensity dependent PPT model calculated ionization rates to obtain the electron density source term in Eq. (9). This however complicates matters to the point where the semi-analytical model would become unusable. As a result, we have kept the simple MPI source term in Eq. (9) which means that the measured result produces an “effective MPI coefficient” corresponding to the intensity clamped regime of 3.5× 1018 W/m2 (see Fig. 5) only. We call this coefficient “effective” since MPI does not fully apply to this regime. However, this “effective coefficient” produces the correct electron density in Eq. (9) at the clamping intensity which allows one to correctly fit the data and get a reliable n2 value. Because this value was measured in the Tunneling Ionization dominated intensity clamped regime, the “effective MPI coefficient” (dotted black curve in Fig. 6) is 10 orders of magnitude lower than a value that would be predicted by simple MPI Keldysh theory (dotted red curve in Fig. 6). This perceived discrepancy actually makes sense because Keldysh theory does not account for the effect of Tunneling Ionization at higher intensities (see dashed red curve versus solid red curve in Fig. 6). The PPT model supports this argument by exactly predicting our measured ionization rate for the clamping intensity of 3.5 × 1018 W/m2 (see intersection of solid red curve with dotted black line and dashed black line in Fig. 6). Please note:
- Our measured “effective MPI coefficient” should not be confused with a σ(K=11) coefficient from Eq. (7). This latter coefficient is truly valid in the MPI regime whereas our measured number (the “effective MPI coefficient”) only applies to the intensity clamped regime at our specific intensity clamped value.
- On a much broader note, we would like to point out that for λ = 1054 nm, only measurements taken at I ≤ 1016 W/m2 can yield a true MPI coefficient. When larger intensities are used, the ionization order (slope) changes to smaller values and MPI becomes less valid. At this point, the slope of the ionization rate is not constant but varies with intensity. This implies that, in principle, each of our focusing geometries have different ionization rates. However, since the intensity clamping is almost the same for both f/# (see Fig. 5), the resulting ionization coefficient is almost the same as well. This is the reason that we can fit the same ionization coefficient to both curves in Fig. 3.
- The PPT model has been validated for oxygen in ambient air at atmospheric pressure at λ = 1054 nm by showing agreement with measured data at high intensities and by demonstrating excellent agreement with the Keldysh approximation for low intensities. This model can now be used over a large range of laser intensities, e.g. for simulation purposes.
We have presented a novel method for measuring the nonlinear index of gases and their corresponding ionization rates (in this case air/O2) under atmospheric pressure at λ = 1054 nm. It has been pointed out that measured focal shifts result from a combination of self-focusing and plasma-defocusing which has been appropriately treated in the semi-analytical model. Based on this model, we have calculated effective MPI coefficients for O2 in the intensity regime of 1018 W/m2. The measured coefficient is in excellent agreement with the PPT model presented in Appendix A. This result has broader implications, because it allows one to safely use the presented PPT curves for modeling efforts on laser propagation in air and other gases. In principle, the above technique is applicable to a broader range of wavelengths and pressures. However, one should make some initial estimates of the expected focal shift in order to verify that the wavefront sensor is sensitive enough to register the expected wavefront deformations.
The PPT model used here is based on the work by Perelomov et al.  and is presented in a similar form as in . For a linearly polarized beam with electric field E, the ionization rate R(E) (measured in Hz) can be expressed as follows:Eq. (10) are defined as: WH = 13.6 eV is the ionization potential of hydrogen, and E0 = EH (Wi/WH)3/2 where GV/m is the electric field in the hydrogen atom. ɛ0 is the dielectric constant in vacuum and h = 2πh̄ is Planck’s constant. Taking the quantum defect into account leads to an effective principle quantum number with Z being the residual charge state of the ion (e.g. Z = 1 for a single ionized atom like Ar). Correspondingly, l* = n* – 1 is the effective orbital quantum number. l is the orbital quantum number and m is the magnetic quantum number (used below in Eq. (12)). The dimensionless constant |cn*,l*|2 is: Eq. (13), Φm(x) would be replaced by . The other functions of γ are:
The ionization rates in Fig. 6 were produced using the following parameter inputs:
- For molecular O2 and N2, l = m = 0  and Z is measured experimentally  to account for the difference in charge state for a molecule versus an atom.
- – N2: Wi = 15.58 eV, Z = 0.90.
- – O2: Wi = 12.55 eV, Z = 0.53. The ionization potential of Wi = 12.55 eV (as opposed to the more common 12.07 eV) was chosen according to  in order to account for the various vibrational levels in the oxygen molecule.
We would like to thank Prof. Miroslav Kolesik for fruitful discussions. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
References and links
2. J. Kasparian, R. Sauerbrey, and S. L. Chin, “The critical laser intensity of self-guided light filaments in air,” Appl. Phys. B: Lasers Opt. 71, 877–879 (2000). [CrossRef]
3. J. Schwarz and J. C. Diels, “Long distance propagation of UV filaments,” J. Mod. Opt. 49, 2583–2597 (2002). [CrossRef]
4. A. Couairon and A. Mysyrowicz, “Femtosecond filamentation in transparent media,” Phys. Rep. 441, 47–189 (2007). [CrossRef]
5. M. Ghotbi, P. Trabs, and M. Beutler, “Generation of high-energy, sub-20-fs pulses in the deep ultraviolet by using spectral broadening during filamentation in argon,” Opt. Lett. 36, 463–465 (2011). [CrossRef] [PubMed]
6. A. A. Voronin, S. Ališauskas, O. D. Mücke, A. Pugžlys, A. Baltuška, and A. M. Zheltikov, “High-energy-throughput pulse compression by off-axis group-delay compensation in a laser-induced filament,” Phys. Rev. A 84, 023832 (2011). [CrossRef]
7. C. Brée, A. Demircan, J. Bethge, E. T. J. Nibbering, S. Skupin, L. Bergé, and G. Steinmeyer, “Filamentary pulse self-compression:the impact of the cell windows,” Phys. Rev. A 83, 043803 (2011). [CrossRef]
8. V. Fedorov and V. Kandidov, “Interaction/laser radiation with matter filamentation of laser pulses with different wavelengths in air,” Laser Phys. 18, 1530–1538 (2008). [CrossRef]
10. O. Bukin, E. Bykova, Y. Geints, S. Golik, A. Zemlyanov, A. Ilyin, A. Kabanov, G. Matvienko, V. Oshlakov, and E. Sokolova, “Filamentation of a sharply focused ultrashort laser pulse at wavelengths of 800 and 400 nm: measurements of the nonlinear index of air refraction,” Atmos. Oceanic Opt. 24, 417–424 (2011). [CrossRef]
12. ZEMAX, “Software for Optical System Design,” http://www.zemax.com/
13. W. Koechner, Solid-State Laser Engineering, 5th ed. (Springer-Verlag, 1999).
14. J. Schwarz and J. C. Diels, “Analytical solution for UV filaments,” Phys. Rev. A 65013806 (2001). [CrossRef]
15. P. R. Yurii, “Breakdown and heating of gases under the influence of a laser beam,” Sov. Phys. Usp. 8, 650–673 (1966). [CrossRef]
16. L. V. Keldysh, “Ionization in field of a strong electromagnetic wave,” Sov. Phys. JETP 20, 1307–1314 (1965).
17. Á Börzsönyi, Z. Heiner, A. P. Kovács, M. P. Kalashnikov, and K. Osvay, “Measurement of pressure dependent nonlinear refractive index of inert gases,” Opt. Express 18, 25847–25854 (2010). [CrossRef] [PubMed]
18. V. Loriot, E. Hertz, O. Faucher, and B. Lavorel, “Measurement of high order kerr refractive index of major air components: erratum,” Opt. Express 18, 3011–3012 (2010). [CrossRef]
19. E. T. J. Nibbering, G. Grillon, M. A. Franco, B. S. Prade, and A. Mysyrowicz, “Determination of the inertial contribution to the nonlinear refractive index of air, N2, and O2 by use of unfocused high-intensity femtosecond laser pulses,” J. Opt. Soc. Am. B 14, 650–660 (1997). [CrossRef]
21. A. M. Perelomov, V. S. Popov, and M. V. Terentev, “Ionization of atoms in an alternating electric field,” Sov. Phys. JETP 50, 1393–1409 (1966).
22. L. Bergé, S. Skupin, R. Nuter, J. Kasparian, and J.-P. Wolf, “Ultrashort filaments of light in weakly ionized, optically transparent media,” Rep. Prog. Phys. 70, 1633–1713 (2007). [CrossRef]
23. A. Talebpour, J. Yang, and S. L. Chin, “Semi-empirical model for the rate of tunnel ionization of N2 and O2 molecule in an intense Ti:sapphire laser pulse,” Opt. Commun. 163, 29–32 (1999). [CrossRef]
24. R. Nuter and L. Bergé, “Pulse chirping and ionization of O2 molecules for the filamentation of femtosecond laser pulses in air,” J. Opt. Soc. Am. B 23, 874–884 (2006). [CrossRef]