1. Introduction

Since the 1996 demonstration of ultrafast laser refractive index modification inside bulk glasses,¹ considerable effort has been made to enhance the technique, not only in terms of the quality of the optical structures and the fabrication speed, but also advancing its applicability to a broader range of optical materials. Along this path, Nolte and coworkers² were the first to demonstrate laser formation of optical waveguides in a crystal (quartz), while Gui et al.³ made further extension to lithium niobate (LN), one of the most widely used nonlinear optic media today. Gui et al. demonstrated single-mode guiding of visible light, exhibiting waveguide propagation losses of ≈1 dB/cm.

Later, several groups,^4–7 including ours⁸ extended the LN work by optimizing various exposure conditions amongst a seven dimensional parameter space of pulse energy, scan speed, wavelength, focusing geometry, repetition rate, pulse width and polarization. Thomson et al.,⁴ Burghoff et al.⁵ and Lee et al.⁷ used linearly polarized laser pulses at 1–5 kHz repetition rate having tens to hundreds of femtosecond duration, energy of hundreds of nJ to tens of µJ, and scan speeds of 10–100 µm/s to generate waveguides with propagation losses of 9, 2.4 and 2.3 dB/cm, at respective wavelengths of 650, 1064 and 1550 nm.

Gui et al.³ reported that waveguides in z-cut samples were formed directly in the laser interaction zone. Thomson et al.⁴ found two types of waveguides in similar samples. The first type were formed in volumes immediately adjacent to the laser damage tracks, guiding both polarizations—perpendicular to the sample surface, hereafter referred to as transverse magnetic (TM), and parallel to the sample surface, referred to as transverse electric (TE). Their second type of waveguides were reported to guide in the laser interaction zone. Burghoff et al.⁵ also demonstrated TM modes in z-cut samples, located in the fabrication laser focal spot, while two TM modes were observed in x-cut samples on both sides of the laser interaction zone. Waveguides of Lee et al.⁷ were made in z-cut periodically poled lithium niobate (PPLN), supporting the TM polarization at regions around the laser interaction zone. The laser damage tracks are believed to cause compressive stresses in the surrounding crystalline media that lead to these polarization dependent refractive index changes.⁶

Further optimization within the four dimensional subspace of pulse energy, scan speed, pulse width and polarization led us to reduce the waveguide losses to 0.7 dB/cm and improve the fabrication speed 10 fold.⁸ Longer pulse duration of 1 ps together with application of circularly polarized light were key to these improvements, while single pass scanning of 700 nJ pulses at 0.9 mm/s scan speed defined a net 100–1000 fold reduction of total laser fluence exposure compared with other studies. Such moderate writing speed of ≈1 mm/s is suitable for applications needing only short lengths of waveguides, but falls short when complex circuits or high numbers of waveguides are required. For example, an excessive exposure time of approximate three hours would be required to fabricate a typical 100 waveguide array for quadratic soliton experiments.⁹ A second area for improvement is the modal symmetry of the present waveguides, which is of paramount importance in reducing coupling losses to optical fibers and in providing uniform vertical and lateral coupling in two dimensional arrays of waveguides in nonlinear studies.

One promising approach that we investigate in this paper to address these two issues— fabrication speed and waveguide mode symmetry—is to harness repetition rate as a new optimization parameter for crystals in the aforementioned seven dimensional space.

Previous studies in a variety of glasses ^10–27 have shown that three characteristics are associated with waveguide writing with high repetition rate lasers. First, with appropriate pulse energy, scan speed, and repetition rate, heat can be accumulated over a number of pulses, leading to structures significantly larger than the geometrical size of the focal spot.^{11, 15, 19} Second, the temperature cycling of the medium during the laser processing can be drastically reduced.^{11, 19} In this way, one can more effectively control the heat accumulation effects, henceforth, leading the material through more controlled and desired thermal cycles. The third point is simply the availability of more pulses per unit of time to speed up the fabrication time over that provided by more conventional 1 kHz laser systems.^{11, 13, 15, 19, 22}

This paper presents, to our best knowledge, the first application of high repetition rate lasers in inscribing buried waveguides in lithium niobate. Building on the optimization parameters discovered in our previous 1 kHz study,⁸ a variable repetition rate fiber-amplified femtosecond laser was applied to z-cut LN, to effectively harness heat accumulation and low thermal cycling effects for possible improvements in waveguide morphology, writing speed, mode profile and overall optical quality. This extension over prior 1 kHz laser studies of LN,^3–8 have provided 50 fold faster writing speed, while obtaining low propagation loss values similar to our previous report.⁸ Definitive evidence of thermal cumulative effects are presented that appear responsible for smooth and crack free modification tracks and nearly circular waveguide mode profiles. Finally, we present what is to our knowledge, the first observation of periodic structures in bulk ultrafast laser interaction with a non-glass material.

2. Experimental arrangement

A commercial fiber-chirped-pulse-amplified (FCPA) laser (IMRA m Jewel D-400-VR) with variable repetition rate in the range of 0.1–5 MHz, provided 1045 nm wavelength pulses with a repetition rate dependant minimum pulse duration of ≈400 fs (full width at half maximum; FWHM). The grating-prism compressor was detuned to provide up to 600 fs pulse duration (FWHM) of Lorentzian profile, and thereby best match the 1.0 ps pulse duration (FWHM) found optimal in our previous 1 kHz study of LN,⁸ where pulse profiles were Gaussian. Such long pulse duration helps suppress unwanted nonlinear pre-focusing effects and thereby deliver a larger portion of the pulse energy to the desired focal volume for waveguide formation. However, attempts to stretch the present laser beyond 600 fs led to highly distorted non-Lorentzian profiles and were therefore avoided. Linearly (parallel and perpendicular to the scan direction) and circularly polarized beams were examined in forming waveguides with 400 and 600 fs pulse durations, and eventually only the combination of circularly polarized light and 600 fs duration were fully optimized due to evidence of better quality waveguides. Circular polarization was also found to be optimal in generating low loss waveguides in our previous 1 kHz study of LN,⁸ possibly owing to weaker nonlinear interactions that suppress detrimental pre-focus nonlinear effects. In addition, utilization of circular polarization decouples the direction of sample velocity vector from the laser polarization, which is of paramount importance when curved waveguides are to be formed.

The material used in this work was z-cut congruent LN (Crystal Technology 99-60011-01). The laser beam was focused ≈110µm below the surface of LN samples, with an aspheric lens (New Focus 5722-C) having a numerical aperture of 0.55. The samples were placed on two-axis air-bearing motion stages (Aerotech ABL1000) having bidirectional repeatability of 50 nm, and straightness/flatness of better than 400 nm over a 10 cm travel range. Modification tracks were then made by scanning the motion stages orthogonal to the laser propagation (perpendicular to LN z-axis), with the expectation of forming birefringent waveguides. Multiple tracks were made at each exposure condition with laser repetition rates of 0.1, 0.25, 0.7, 1, and 1.5 MHz, and respective on target energy in ranges of 170–700, 170–700, 150–561, 150–390, and 120–264 nJ/pulse. The uppermost energy for the three latter repetition rates was limited by the maximum output power available from the laser. Scan speeds were varied between 1 and 80 mm/s.

After processing, the sample facets were ground and polished to optical quality to access the laser-formed waveguides. The waveguides were then probed in free space by a lens firing method, exploiting two external cavity tunable laser sources (Photonetics, Tunics) emitting wavelengths in the telecommunication bands of 1300 and 1550 nm. An IR detector (Newport, 818-IG) and a vidicon camera (Electrophysics, 7290A) were used respectively for measuring power and recording waveguide mode profiles. As previously described,⁸ a Fabry-Perot technique²⁸ was used for assessing waveguide propagation losses.

 

Fig. 1. Cross section (first row) and top section (second row) optical microscope images of laser-formed tracks in z-cut lithium niobate at 46 mm/s scan speed and pulse energy of ≈270 nJ for different values of repetition rates. Laser exposure direction is from the bottom in the top row images.

Download Full Size | PDF

3. Results and discussion

3.1. Morphology

Optical microscope top and cross-section views of several modification tracks formed at the aforementioned repetition rates are shown Fig. 1. The scan speed was fixed at 46 mm/s and the pulse energy was held at ≈270 nJ. Circular polarization and ≈600 fs pulse duration was applied. Dramatic changes are seen in the morphology and the structural modifications as repetition rate increases from 100 to 1,500 kHz. However, these changes are counterintuitive to the expectation of higher modification contrast and damage as the net fluence exposure rose 15-fold over this increase in repetition rate. The weakest contrast and smoothest modification structures are noted for the 1.5 MHz sample, which we attribute later to a heat accumulation and thermal annealing mechanism. The most significant damage is seen in the lowest exposure sample of 100 kHz, where nonuniform structural changes, together with microcracks, are apparent in both top and cross-section views. An angled crack is also prominent in the side view. The modification features get slightly smaller at 250 kHz, while at 700 kHz, the structure size, microcracks, and overall contrast are dramatically reduced. At 1 MHz the laser modified zone becomes almost cylindrically symmetric and consists of a high-contrast zone surrounded by a weak modification envelope (not visible in the figure) of ≈9.6 µm diameter. These features exceed the theoretical ≈1.2 µm (1/e²) focal diameter, and ≈4.8 µm depth of focus of the writing laser. This contrasts with an absence of any weak modification envelope at lower repetition rates, where only high contrast ≈4.3 µm structure—alongated in the laser propagation direction—are noted.

Higher magnification images of structures formed at 250 and 700 kHz are shown in Fig. 2. The tracks reveal a very strong laser modification zone of ≈ 1.4 µm that is consistent with the calculated—Gaussian optics and ABCD matrices—laser spot size of ≈ 1.2 µm. Higher laser exposure of 700 nJ energy at 60 mm/s speed and 500 nJ energy at 46 mm/s speed, respectively, were applied to increase the modification contrast. The angled cracks noted in the 100 and 250 kHz cross-section images of Fig. 1 are now strongly apparent in the 250 kHz cross-section image of Fig. 2. The extended dark lines are seen to form at an angle of ≈ 56°, which agrees well with cracking along the 55°52′ crystalline angle of the rhombohedron unit cell of lithium niobate.²⁹ On the other hand, at the repetition rate of 700 kHz, such cracks appear greatly diminished and laser formation of a new periodic structure is reported for the first time in the bulk of a crystalline material. A micro-grating of ≈1.2 µm period is noted with alignment also matched to the 55°52′ crystal plane. The period of this structure closely matches the free space wavelength of the laser (1045 nm), and is also weakly observable with similar period in the 1,500 kHz track of Fig. 1. Because theses periodic structures follow the crystalline planes, are formed with circularly polarized light, and their period seem uncorrelated to the refractive index of the medium, they appear unrelated to the nanogratings formed in glass with a linearly polarized light,^{30, 31} and we are further investigating the underlying formation mechanisms.

 

Fig. 2. First column depicts the cross section and top views of a track formed at laser exposure conditions of 250 kHz, 700 nJ/pulse, and 60 mm/s. Cross section and top views of a track formed at 700 kHz, 500 nJ/pulse, and 46 mm/s are depicted in the second column. All images were obtained with a 100× microscope objective (0.95 NA).

Download Full Size | PDF

3.2. Waveguiding

Good optical waveguiding was observed only in a small number of the laser formed tracks over the exposure conditions presented in Sec. 2. At 100 kHz, some form of 1300 nm guiding could be observed across a wide range of exposure conditions for pulse energy in the tested range of 170 to 700 nJ. However, propagation losses were too high in all samples for quantitative analysis. This correlates with the observations of Fig. 1 (100 kHz sample), where crystal damage and disorder are seen in the vicinity of the laser exposed track. For tracks made at high repetition rates of 1 and 1.5 MHz, very weak guiding was observed at 1300 nm wavelength, only for samples exposed at the maximum available pulse energies of 390 and 264 nJ, respectively. We therefore anticipate that higher pulse energy would generate stronger guiding at these high repetition rates, or alternatively, multi-line scanning, as demonstrated in Ref. 7, may be used to build up refractive index change in a larger volume, but at the cost of lower net processing speed. At 250 and 700 kHz repetition rates, well-confined and low loss guiding for the TE polarization was observed in the telecom band of 1300 nm. This TE guiding conforms well with our previous observations,⁸ and is attributed to the stress induced increase of ordinary refractive index. The TE-mode guiding profile for the 700 kHz inscribed track of Fig. 2 is shown in Fig. 3, where waveguiding is seen to take place in the pre-focus zone immediately before the largest laser modification track, located on the incoming side of the writing laser.

 

Fig. 3. The cross section microscope view of the 700 kHz track of Fig. 2, together with the 1300 nm TE-guided mode profile, showing its relative position with respect to the laser modified zone. The writing laser was applied from the bottom in this figure.

Download Full Size | PDF

For orthogonal polarization (TM), guiding was observed in the regions between the three separate laser interaction zones (dark spots seen in Fig. 3), but propagation losses were too high for quantitative analysis. This TM guiding is attributed to the increase of extraordinary refractive index in the region sandwiched between two laser interaction zones, as previously reported by Burghoff et al.⁵ All samples were also probed by 1550 nm light, but guiding was too weak for quantitative analysis.

For TE polarization, the 1300 nm guiding quality was high enough to apply the Fabry-Perot technique²⁸ for measuring the propagation loss. The recorded transmission spectra of the lowest loss waveguides written at 250 and 700 kHz repetition rates, have clearly resolved Fabry-Perot fringes as shown in Fig. 4, while their respective mode profiles are depicted in Fig. 5. The mode size diameter of ≈20µm here is similar to that observed in waveguides made previously with our 1 kHz laser,⁸ but with greatly improved mode symmetry (near-circular profile) over an ellipticity of 0.75 in the previous study. Since the waveguides are positioned adjacent to the main laser interaction zone, the guided light partially overlaps with the heat affected zone, as seen in the right side of Fig.2. Here, heat accumulation effects may indirectly modify the refractive index profile and improve the waveguide mode symmetry. The improved mode symmetry offers better prospects for coupling into and out of telecom optical fibers, and also suggests better overlapping of multi-propagating beams to improve the efficiency of nonlinear optical interactions.

Fabry-Perot spectra as noted in Fig. 4 were applied extensively to assess the TE-mode propagation losses of waveguides written at 250 and 700 kHz and results are mapped in Fig. 6(a) and (b), respectively, across a widely explored energy-speed parameter subspace. At 250 kHz, two regions of exposure conditions (high energy, high speed) and (low energy, low speed) resulted in good guiding with moderate losses of several dB/cm. The lowest waveguide losses of 2.0 and 2.2 dB/cm were observed at laser exposure conditions of 700 nJ and 60 mm/s, as well as 320 nJ and 1 mm/s. Clearly, the former is the preferred process window, owing to the 60-fold faster fabrication speed.

 

Fig. 4. The measured transmission spectra of the lowest loss lithium niobate TE-mode waveguides, written at repetition rates of (a) 250 kHz and (b) 700 kHz (depicted in Fig. 2 and 3), showing Fabry-Perot fringes. Both waveguides were 9.2 mm long.

Download Full Size | PDF

 

Fig. 5. TE polarization mode profiles recorded at 1300 nm wavelength in waveguides fabricated at 250 kHz (a) and 700 kHz (b), whose transmission spectra are depicted in Fig. 4(a) and 4(b), respectively.

Download Full Size | PDF

At 700 kHz, the low-loss process window in Fig. 6(b) is significantly altered from the 250 kHz case. Up to pulse energies of 360 nJ, no guiding was observed, while for energies of 400 nJ and higher, good guiding (losses <5 dB/cm) existed for more than 5 octaves of scan speeds, ranging from 1 to 80 mm/s. Guiding at higher speeds might be possible, but with the present translation stages, our maximum achievable speed was 80 mm/s. At the speed of 46 mm/s, losses below 0.7 dB/cm were measured across a fairly wide energy range of 400–500 nJ/pulse, with a best result of 0.6 dB/cm at 500 nJ, closely matching the best 0.7 dB/cm value of our previous study.⁸

 

Fig. 6. The 2-D maps for TE-mode waveguides indicating low-loss laser processing window for a repetition rates of (a) 250 kHz and (b) 700 kHz. Normalized pseudo color (scale on top) indicates low loss guiding with dark colors (black to dark red) and high loss or weak guiding with light colors (yellow to white).

Download Full Size | PDF

3.3. Heat accumulation and annealing

The waveguide morphology trends of Fig. 1 is closely connected with the shifting of low loss processing windows observed in Fig. 6 for different laser repetition rates. We assess these trends in terms of heat accumulation effects that become more pronounced with increasing repetition rate. The relevant thermal parameters of lithium niobate at room temperature are listed in Table 1. The wide variance in thermal conductivity and diffusivity values here, together with unknown changes expected at high laser-heated temperatures, will provide only cursory insight into the underlying physics that, nonetheless, will be instructive.

To verify whether heat accumulation has occurred in our experiments, we examine the relevant temporal and spatial thermal conditions.³² The effective cooling time, τ_c=d²/D, diffusing heat out of a laser interaction volume of diameter d,¹⁸ yields values of τ_c=1.4-4.8 µs, for the present case of d=1.4 µm and the diffusivity values from Table 2. This suggests that a transition to heat accumulation effects is possible for laser repetition rates exceeding R_p=1/τ_c=200 to 700 kHz, which we have transitioned in the present experiments. In other words, the temporal condition required for the heat accumulation has been satisfied under our experimental conditions.

To verify the satisfaction of the spatial condition, we estimate the spatial extent of the heat accumulation zone. The 2-D thermal diffusion length associated with the time interval between two consecutive laser pulses is obtained from

with predicted values listed in Table 2 for the 100 kHz to 1.5 MHz range of repetition rates, studied here. The table further provides values for the distance traveled by the sample between two consecutive laser pulses, calculated from

Table 1. Thermal parameters of lithium niobate33

View Table | View all tables in this article

for the case of v_s=46 mm/s scan speed, which matches the experimental conditions in Fig. 1. The total heat loading also depends on the effective number of laser pulses overlapping within the laser focal diameter, which we estimate from

Values for N are also presented in Table 2 for v_s=46 mm/s.

Examination of the data in Table 2 clearly shows the calculated diffusion length exceed the laser focal diameter (d=1.4 µm) for repetition rates of 250 kHz and lower, therefore suggesting that no significant heat accumulation is expected for this range of repetition rates. In contrast, at 700 kHz and beyond, the calculated diffusion lengths are comparable with the size of the laser interaction zone, and one therefore expects the onset of cumulative thermal effects to be between these two rates. Further driving this thermal scaling is a decreasing step size, h, and a commensurate increase in the number of effective overlapping pulses, N, that is expected to strongly heat the sample at increasing repetition rate, while reducing the overall degree of temperature cycling.

The evidence for such cumulative heat effects is given in the optical microscope images of Fig. 2, for the transitional zone of 250 to 700 kHz repetition rates. The top view of the 250 and 700 kHz tracks both clearly show a thin central region of 1.4 µm width, associated with the original laser focal spot size, and the formation of an extended laser affected zone of 4.3 µm width. The size of these two regions are equal, despite a factor of almost 3 increase in total fluence for the higher repetition rate case. However, an extended heat affected zone is also faintly visible, with diameter increasing from 6.1 to 9.0 µm for 250 and 700 kHz, respectively. Although the absolute size of these extended melt zones are different from previously reported measurements in glass,^{11, 15, 19} the observations are in qualitative agreement, with visible evidence of an extended heat affected zone that grows in size with increasing fluence and increasing repetition rate.

The results of Fig. 1 and Table 2 suggest that high repetition rates in the range of 250 to 700 kHz mark an important transition in lithium niobate, where laser pulses are interacting with media heated above the melt temperature, and undergoing little thermal cycling due to the rapid arrival of pulses faster than the effective cooling time calculated above. Such nearly continuous heating avoids the shock-induced damage apparent at lower repetition rates in Fig. 1, thereby providing final structures with much reduced damage through processes that possibly resemble a combination of epitaxial regrowth and rapid thermal annealing (RTA). At this stage, it is not known whether the poor quality of waveguides formed at higher repetition rates of 1 and 1.5MHz is due to weakened refractive index contrast resulted from overly extended heat zones, or the lack of enough energy in individual interacting laser pulses. More extensive thermal modeling, together with temporally and spatially resolved spectroscopy of laser interactions are underway to better elucidate this physics.

Table 2. Thermal effects in LN waveguides for 46 mm/s scan speed

View Table | View all tables in this article

3.4. Trends and comparison

For the case of 250 kHz repetition rate, Fig. 6 presented a diagonal processing window where increasing pulse energy facilitates higher writing speed, that collectively load a somewhat constant net fluence for low loss waveguide formation. A similar trend was noted in our previous 1 kHz experiments,⁸ suggesting that low loss guiding should also have been observed at intermediate repetition rate of 100 kHz, in contrast to the no-guiding observation in Sec. 3.2. Cumulative thermal effects do not play major roles (Sec. 3.3) for these lower repetition rates (<250 kHz). Here, high exposure generates strong lattice damage, while diffusion-only processes with high temperature cycling is not able to anneal this damage as inferred for the case of the low loss waveguides seen at 250 kHz or higher repetition rates. However, low loss waveguides were also observed at 1 kHz where such heat accumulation has no role. If we scale the optimum 0.9 mm/s scan speed found for 1 kHz exposure⁸ (for 1300 nm waveguides) to 100 kHz rate, a speed of 90 mm/s appears optimal for equal net fluence exposure. However, this speed just exceeds the maximum 80 mm/s available in the present apparatus. Other reasons could include differences in the writing wavelengths of 800 nm and 1045 nm, the Gaussian versus Lorentzian temporal profiles, or pulse width differences of 1 ps and 600 fs, respectively, for the 1 kHz and 100 kHz approaches. As a result, further investigation of exposure conditions, particularly pulse width, scan speed, and wavelength, may open a 100 kHz processing window for waveguide formation in LN.

On the other hand, the gradual onset of heat accumulation effects at 250 kHz and beyond (Sec. 3.3) opens a new processing domain, where thermal annealing appears to reduce the formation of scattering centers and microcracks that otherwise contribute to linear loss in the waveguides. At 250 kHz repetition rate, the laser dissipation transitions from diffusion-only formation of heat affected zones at lower rates to the cumulative regime at higher rates, where continuous heat and reduced temperature cycling manifest in lower loss waveguides. Following this transition is a transformation of the low loss processing window from a diagonal band of constant net laser exposure for 250 kHz, as seen in Fig. 6(a) (low energy and low scan speed to high energy and high scan speed), to a flat onset threshold of 400 nJ energy per pulse at 700 kHz in Fig. 6(b) that, surprisingly, is largely independent of scan speed over a wide 1 to 60 mm/s range. That is the characteristic of material structuring in the high repetition rate regime, where past certain energy threshold, the fluence loses its significance to some extent, because of the nature of processing heated material. Although, in this regime, the combination of pulse energy and scan speed still plays a key role in the characteristics of the waveguides, especially its loss figures, the overall behavior of guiding is almost independent of them.

At higher repetition rates, 1 and 1.5MHz, similar morphological trends are seen, as described in Sec. 3.1, but due to the lack of guiding, further conclusions cannot be drawn at this stage. One possible explanation may be self-healing effects such as epitaxial regrowth or RTA of the laser-heated volume, as described in Sec. 3.3.

All waveguide samples in this study were examined more than two months after waveguide

Table 3. Comparison of laser-formed waveguides in lithium niobate

View Table | View all tables in this article

fabrication, suggesting good room temperature stability, comparable to our previous 1 kHz study of 5 week old samples.⁸ The elevated temperature stability of the guides were not studied at this stage.

Table 3 summarizes our results against the aggregate data associated with all prior studies of ultrafast laser waveguide writing in LN, of which we are aware. The studies prior to this work were all completed in the kilohertz repetition rate domain, with a maximum writing speed of ≈1 mm/s and a minimum propagation loss of 0.7 dB/cm, reported for circularly polarized light in Ref. 8. Prior works show a wide 500-fold variance in laser exposure conditions that leads to widely varying degrees of crystal damage, waveguides stability and propagation loss. The present high repetition rate results provide waveguides having losses equal to the best values reported in previous kilohertz work⁸ with the substantial benefit of 50 fold faster fabrication speed. Using this new high repetition rate process, a 10 by 10 array of 10-cm long waveguides would take less than 4 minutes to form, opening new opportunities for nonlinear studies and fabrication of complex optical circuits in lithium niobate.

4. Conclusions and outlook

In conclusion, we presented what is to our knowledge the first study of high repetition rate bulk ultrafast laser matter interaction in lithium niobate.

The heat accumulation effects associated with these higher repetition rates are definitely observed for the first time in a non-glass material, and appear responsible for annealing of laser modified tracks, so that morphologically smooth waveguides were attained.

We also presented the first time observation of periodic structures in bulk interaction of an ultrafast laser with a non-glass material.

Using high repetition rate (250 to 700 kHz) ultrafast lasers to form optical waveguides in bulk lithium niobate, we have achieved propagation losses of 0.6 dB/cm, that match the best kHz results in the telecom band, while reducing the fabrication time by almost a factor of 50, and offering improved transversal mode symmetry. Megahertz rate femtosecond lasers henceforth present a maskless and chemical free process, well suited for rapid fabrication of large size waveguide arrays, therefore, possibly enabling a new generation of previously unexplored quadratic nonlinear experiments, which has been hindered so far by the complexity of guided wave optics formation in lithium niobate.

Similar to our low repetition rate study, circular polarization and stretched pulses resulted in extending low loss waveguiding intoMHz repetition rates. Further studies are required to better understand the mutual effects between polarization and pulse width as well as their interplay with the fabrication laser wavelength, and their aggregate effects on the morphology and optical characteristics of the formed structures.