1. Introduction

The diffraction gratings in the compressor of a CPA (Chirped Pulse Amplification) Ti:Sapphire laser system [1] are exposed to extreme conditions: the energy and the intensity are at the highest level compared with the other parts of the laser system The pulse energy is at its maximum and its duration may be as short as a few femtoseconds.

To enable high performances and a reliable operation of the laser system, the diffraction gratings of the compressor must present the following characteristics:

The search for grating configurations and technologies offering all at once high diffraction efficiency over a wide bandwidth and high damage threshold is therefore still a subject of intense interest. The search for an alternative to gold gratings began in the early nineties as first suggested by Sychugov et al [5]. All-dielectric grating structures composed of a multilayer mirror and a corrugated superstructure have been developed in a number of laboratories [3,6] and have been shown to withstand larger flux at 1050 nm wavelength [7]. The possibility to achieve 100% diffraction efficiency was shown to be related with the excitation of a leaky mode of the mirror-based diffractive superstructure [8]. The polarization is preferably TE since the TM polarization exhibits the Brewster effect which limits the field reinforcement in a leaky mode. Such gratings are today available commercially with impressive performances and size [9]. With the development of Ti:sapphire industrial lasers for machining in the 800 nm wavelength range [10], a new set of specifications has emerged with moderate bandwidth (about 30–40 nm) and ultimate efficiency to limit the loss of a 4-pass scheme. An all-dielectric solution was proposed and demonstrated experimentally exhibiting a close to 100% top-hat diffraction efficiency over a 20 nm bandwidth [11,12]. Although the corrugation is made in the last index layer unlike in the state of the art, the demonstrated damage threshold was larger than 1 J/cm².

The search for ever shorter pulses in advanced high energy physics places new bandwidth specifications [13] which all-dielectric gratings can not fulfill because the condition of leaky mode excitation [8] can not be satisfied over a sufficiently broad spectral range, and, furthermore, true guided modes are likely to be excited over such a broad wavelength range with the detrimental consequence of having the diffraction spectrum chopped by sharp amplitude dips and sudden phase jumps. The rationale prevailing in the solution of Ref. [12] will nevertheless be resorted to for pushing the diffraction efficiency as close as possible to 100%; a plane metal mirror will be used instead of a multilayer mirror since the former ensures an almost constant reflection phase shift which permits to broaden the spectral domain around 800 nm where leaky mode resonance can be satisfied. This type of mixed metal-dielectric grating is likely to exhibit a higher damage threshold since the metal mirror is directly deposited on a non-organic substrate of large thermal conductivity which can be polished at very low residual roughness with an adhesion layer inbetween. Since the TE polarization is used, there is much less risk for destructive plasmon excitation.

However, the problem of the mixed metal-dielectric interface at the incidence side is a difficult one which may lead to unforeseen causes of damages: no adhesion layer such as Cr of Ti may be used to make the metal-dielectric bond strong enough to avoid delamination; moreover, the dielectric layers on top of the metal mirror can not be submitted to high temperature annealing for removing possible defects on which destructive electronic discharges could start. Although the dielectric layers and the corrugation grating are very thin, therefore close to the metal plane, it is not known yet whether the proximity of a electron sink will decrease the probability of discharges. Finally, this type of mixed structure is difficult to fabricate: the projection of an interferogram in the presence of a highly reflective background remains a big challenge, the more so if an unusual line/space ratio is needed. Using an ARC layer does not solve this problem completely. Furthermore, the very process of grating etching must leave the metal mirror surface intact physically and chemically which severely restricts the etching chemistry and sputtering conditions.

The present paper also describes the operation of the proposed metal-dielectric grating and brings the experimental evidence of its high diffraction efficiency and wide band in the case of a structure fabricated on small substrates. In spite of the fabrication difficulties, it also gives very encouraging results as to the flux resistance: 1 J/cm² damage threshold was measured in gratings where the processes haven’t by far been optimized and where evident defects were still present.

2. All-dielectric grating

The basic compression grating structure is represented in Fig. 1. It comprises a multilayer mirror on a substrate and a dielectric superstructure with a corrugation in the last high index layer at the air side. The diffraction is produced by the sole -1^st order in a contradirectional scheme away from the Littrow condition. As shown in Ref. [8, 12] the condition for high efficiency is given as the dispersion equation of a TE leaky mode excited by refraction of the incident wave in the corrugated layer. When the leaky mode resonance condition is fulfilled, the Fresnel reflection coefficient in a gratingless structure is real and the contribution to reflection of the air/layer interface and of the leakage of the field enhancement layer are of opposite sign. Degrading the quality factor of the field enhancement layer by means of a -1^st order grating at the layer/air interface has two consequences: first, the Fresnel reflection can be cancelled out by destructive interference in the Fresnel reflection direction, secondly, the energy has nowhere else to propagate but to be diffracted along the -1^st order. 100% diffraction efficiency is therefore ensured by properly adjusting the grating strength.

 

Fig. 1. Cross-sectional view of the dielectric mirror based leaky mode propagating structure with binary corrugation in the last high index layer. TE incidence is under angle θ_i, diffraction along the sole -1^st order.

Download Full Size | PDF

Figure 2, dotted curve a) represents the -1^st order diffraction efficiency spectrum of a typical dielectric structure of the state of the art comprising a quarter-wave mirror and a corrugated silica layer of t_b unetched thickness and etched depth d. The incidence angle in air is 57 degrees, the central wavelength is 800 nm. The quarter-wave layers are made of silica (n₁ = 1.48) and HfO₂ (n_h = 2.12) and are quarter-wave for the incident wave, therefore t₁ = 164 nm, and t_h = 103 nm respectively. The chosen period is ʌ = 570 nm. The mirror’s last layer is of high index. Curve a) shows the best diffraction efficiency profile under the above conditions when the grating is a binary corrugation etched in a low index layer as in state of the art all-dielectric compression gratings. The optimisation code of Lyndin [14] was used with the objective of maximizing the diffraction efficiency of the TE polarization in a range of 40 nm width centered at 800 nm wavelength. It can be seen that the efficiency can not reach 100% and that there is no top hat character achievable. The present approach also uses quarter-wave layers, but the superstructure is composed of a low index buffer layer of thickness t₁ and a high index layer of thickness t₂ in which a binary corrugation of depth d is etched with a line/space ratio ρ = w_L/w_S between the width w_L of the grating line and the groove width w_S.

For the electric field to be strong enough in the corrugation to ensure maximum strength with minimum depth, the two-layer system at the air side is made to satisfy the dispersion equation of the fundamental TE leaky mode given as [8]

where k₀ = 2π/λ at vacuum wavelength λ, θ_i is the incidence angle in air. The phase terms (ϕ_m and ϕ_a are the reflection phase shifts at the mirror boundary and at the air side with incidence from the leaky mode propagating layer side. ϕ_a is 0 since the transmission medium (air) has lower index, and ϕ_m is zero too since the last layer of the multilayer mirror is of high index.

The dispersion equation must be satisfied in a structure which is corrugated, therefore an equivalent index n_eq must be used for the top layer instead of the high index n_h [8].

 

Fig. 2. Optimized -1^st order diffraction efficiency spectra. a) according to the state of the art with corrugated silica layer on a quarter-wave dielectric mirror. b) of the optimised broad band character of the present all-dielectric structure.

Download Full Size | PDF

The introduction of a broad band character relies upon electromagnetic considerations which will be reported separately. Submitting the structure to the optimisation code[14] delivers the diffraction efficiency of curve b) in Fig. 2. The efficiency spectrum exhibits a top-hat character with close to 100% over a broad band and quickly falls to zero at the band edges. Defining the corrugation in the last high index layer requires quite shallow grooves of hardly more than 100 nm depth whereas grooves defined in a silica layer (curve a)) must be between 400 and 600 nm deep to give rise to a sufficient grating strength.

 

Fig. 3. AFM scan of a typical hafnia grating of an all-dielectric grating.

Download Full Size | PDF

The test structures were fabricated on thick fused quartz substrates of 25 mm diameter, 6 mm thickness by the deposition of alternating layers of silica and hafnia. The lithography of the 580 nm period grooves was made by exposing a 200 nm thick resist layer to the interferogram produced by a stabilized Mach-Zehnder scheme at the 442 nm wavelength of a HeCd laser. The problem of the reflection of the exposure beams by the multilayer is known to be a big problem, sometimes imposing a modification of the multilayer mirror [15]. The present structure is made to be little reflecting under specific exposure conditions which permits to skip any ARC resist or absorptive layer, thus making the exposure on a high reflection mirror easy.

The physical transfer of the resist grating into the last hafnia layer was made by RIBE. Special care was taken not to broaden the corrugation walls. Figure 3 illustrates a typical corrugation obtained in hafnia which closely corresponds to the specifications given by the modelling with a groove depth of 123 nm and line/space ratio ρ = 0.62.

The corrugated wafers were tested by means of a CW tunable Ti:sapphire laser under an incidence angle of 57 degrees between 730 and 850 nm wavelength. As shown in Fig. 4, the - 1^st order diffraction efficiency is remarkably high: 96% on the average and is remarkably flat between 777 and 815 nm wavelength. The 0^th order is 2% on the average and exhibits sharp peaks at the band edges. Making the measurement at different locations of the grating surface shows a few percent variation only which means that the technology can be scaled up to larger substrates. Although the diffraction efficiency is already quite high, these results do not represent a limit. There is room for improvement by a better control of the corrugation line/space ratio.

 

Fig. 4. Experimental diffraction efficiency spectrum under 57 degree TE incidence.

Download Full Size | PDF

3. Metal-dielectric grating

The basic grating structure is represented in Fig. 5. It comprises a metal mirror and a dielectric multilayer with a corrugation in the last layer at the air side. The diffraction conditions are the same as in the all-dielectric structure. There is not much benefit to draw from a large number of dielectric layers on top of the metal mirror because the grating would then excite a number of true guided modes of the multilayer waveguide bounded by the metal substrate and the air cover, the more so if the femtosecond pulses are very short (for instance 20 fs), therefore the equivalent bandwidth is very broad. A spectral band without waveguide mode excitation is therefore needed. This can easily be achieved with a restricted number of dielectric layers. The minimum number of layers is actually 1 in which the grating corrugation can be etched. However, a metal thin film may have to be protected by a specific dielectric coating, and it is advantageous to make the grating in a high refractive index layer so as to increase its strength without having to etch too deep a corrugation.

 

Fig. 5. Cross-sectional view of the metal-mirror based multilayer with binary corrugation in the last high index layer. TE incidence is under angle θi, diffraction along the sole -1st order.

Download Full Size | PDF

The theoretical modeling of the present structure comprising the metal mirror, a protection layer and the corrugated layer can be found in Ref. [8].

In its application with a silver mirror (ε_m = -28 - j1.5 at λ = 800 nm), a 30 nm thick protective layer of Al₂O₃ (n₁ = 1.65) is required, and the corrugation was made in the thicker HfO₂ layer (n_h = 2.12 at λ = 800 nm). The structure is illustrated in Fig. 5.

 

Fig. 6. Optimized diffraction efficiency spectrum of a high index hafnia grating on a silver mirror with protective layer.

Download Full Size | PDF

Submitting the phenomenologically designed structure to Lyndin’s optimisation code [14] provides the final structure achieving the highest diffraction efficiency over the requested spectral width which is here 200 nm centered at 800 nm wavelength. The optimisation code must be somewhat assisted to deliver a structure whose line/space ratio is still fabricable. Figure 6 is the resulting optimised spectrum.

The chosen metal was silver for its slightly lower losses. The silver protection layer is aluminum oxide. There are two big technological difficulties: the first one is to make the lithography on so highly reflecting a surface, and to achieve a small line/space ratio. The second one is the rather deep dry etching of the hafnia layer down to the protective layer without physically and chemically damaging the silver surface. These difficulties found provisionally an appropriate solution. Instead of using a thick ARC to isolate the resist layer from the highly reflecting silver layer [16], a 30 nm-thick layer of CuO was used. This decreased the reflection to below 40% which was sufficient to adjust the two lithographic steps leading to the requested line/space ratio. Figure 7 is the AFM scan of a resist grating obtained under the adequate exposure conditions. The resist ridges are slightly rounded due to the presence of a standing wave field node close to the top of the resist layer. The CuO thin layer at the bottom of the resist grooves were opened by wet etching in vinegar. The RIBE etching conditions were adjusted to further reduce the thickness of the resist walls down to the hafnia layer.

 

Fig. 7. AFM scan of a small line/space ratio resist grating on top of the CuO layer.

Download Full Size | PDF

The 25 mm diameter, 6 mm thickness corrugated samples were measured by means of a CW tunable Ti:sapphire laser under an incidence angle of 50 degrees between 710 and 850 nm wavelength which was the available tuning range. As shown in Fig. 8, the -1^st order diffraction efficiency is 95% on the average and is remarkably flat. The 0^th order is 2% on the average. Although the diffraction efficiency is already quite high, these results do not represent a limit. A slightly different ratio ρ could lead to a better extinction of the 0^th order and to a suppression of the slope. Furthermore, the silver layer was slightly damaged during deposition by a mechanism which was identified and can be corrected with an expectable loss decrease.

 

Fig. 8. Experimentally measured -1^st order diffraction efficiency and 0^th reflected order spectra under 57 degree TE incidence.

Download Full Size | PDF

4. Damage threshold measurements

The experiments were performed with a 1 kHz, 10 mJ Ti: Sapphire laser system at the central wavelength of 800 nm based on the chirped pulse amplification technique. In this system, the oscillator beam is first temporally stretched to 500 ps pulse duration, then amplified by a regenerative and a 4-pass amplifier. Amplified pulses are then compressed by means of a standard grating based compressor. A beam splitter is used to probe a few percent of the overall energy and send it in the damage threshold measurement setup (Fig. 9). The maximum available energy is 80 μJ at a repetition rate of 1000 Hz.

A 12 bit CCD camera was used to record the normal transverse beam profile and to determine the beam size. The laser fluence F was estimated by using the pulse energy E_L and the radius R at 1/e² of the maximum intensity on a cross-section normal to the beam, by means of the formula F = 2 E_L/πR². With this configuration, the focal spot diameter, defined as the diameter at 1/e² of the maximum intensity on the beam transverse cross-section, has an average of 55 μm. The energy damage threshold is defined on a cross-section normal to the beam. The beam trace on the sample surface is elliptical; the energy density there is the energy density normal to the incident beam multiplied by the cosine of the incidence angle, i.e. approximately a factor 2 smaller. The pulse duration was optimized by adjusting the compressor parameters. The pulse duration was measured by the SPIDER technique [17] to be about 40 fs.

 

Fig. 9. Laser damage threshold measurement setup: Incident laser beam characteristics: 800 nm center wavelength, 1 kHz repetition rate, about 40 fs duration (FWHM) and 80 μJ available. a: Pockels cell, b: polarizer, c: λ/2 wave plate, d: polarizer, e: Joule meter, f: 150 mm focal length lens, g: 12 bits camera. The grating normal makes an angle α with the beam axis.

Download Full Size | PDF

The experimental setup of Fig. 9 for laser damage measurement comprises the following elements: a Pockels cell (a) and a polarizer (b), the former being synchronized with the laser clock to select the repetition rate from the single shot regime. A half-wave-plate (c) and a polarizer (d) control the laser energy delivered onto the grating sample. Real time energy measurement was performed by sampling a small part of the laser beam power (~ 1%) with a calibrated microscope plate; the measurement instrument is a microjoulemeter (model PE9 pyroelectric sensor with NOVA 1 Display from Ophir Laser measurement Group). The laser pulse is focused onto the sample surface with a 150 mm focal length lens (f). The angle α between the sample and laser beam direction is adjusted to optimize the diffraction efficiency. The sample was installed on a motorized XYZ translation stage allowing displacements with a precision of 10 μm. To ensure the reproducibility of the damage threshold measurement, attention was paid to position the sample surface at the beam focus. To that end a 12 bits CCD camera (g) (Scorpion camera from Point Grey) is used to visualize the beam reflected from the sample.

For the determination of the grating damage threshold the laser was used in the 1 kHz mode. The main goal of these experiments was to measure the grating damage threshold under conditions analogous to the operation conditions. Instead of measuring a single shot threshold, we opted for the testing of the ageing effect. Ageing measurements are usually defined at 10000 shots of the same energy on the same area. The sample is translated stepwise, keeping the same distance between the beam impact and the focusing lens, to repeat the test on a fresh area with a different energy. We first used the maximum energy available (80 μJ per pulse), and decreased it with an energy step of 1 μJ. The damage threshold is defined as the energy level at which no damage occurs any more. Damages were a posteriori observed using an optical microscope of one micrometer resolution. Figure 10 illustrates a typical sequence of 10000 shot exposures with decreasing pulse energy. The damaged zones have a ellipticity roughly determined by the cosine of the incidence angle.

 

Fig. 10. Optical microscope picture of the grating surface after 15 exposures to 10000 shots of decreasing energy of 1 μJ increment with a beam waist of about 55 μm normally to the incident beam.

Download Full Size | PDF

Table 1 summarizes the damage threshold measurement results. We found a damage threshold fluence of about 1.1 J/cm² for the metal-dielectric mixed grating, and about 1.1 J/cm² as well for the all-dielectric grating. The 0.1 J/cm² error accounts for the uncertainty on the focus size and pulse energy. The damage threshold of the gratingless multilayer used in the all-dielectric grating was measured as a reference; the damage threshold is 1.5 times higher than that of the etched grating. This infers that the groove etching process does not provoke a strong degradation of the flux resistance. This is in contrast with the existing all-dielectric gratings having a very high aspect ratio corrugation etched in the last low index layer where the ratio between the non-etched and the etched damage thresholds is about 2 to 2.5.

Table 1. Damage threshold, incidence angle and grating period of the mixed and all-dielectric gratings, and of the gratingless multilayer.

View Table

5. Conclusion

The evidence is shown that, unlike in the state of the art, compression gratings of high efficiency and wide band with the corrugation made in a high index layer can be fabricated to exhibit very interesting optical characteristics with rather shallow groove depth of less than 150 nm in the 800 nm wavelength range.

First, an all-dielectric compression grating of close to 100% efficiency over a 40 nm wide spectral width suitable for femtosecond laser machining of high average power can be fabricated with sufficient control.

Secondly, a compression grating associating a flat metal mirror and a corrugated high index dielectric overlay can be fabricated to match the demands of femtosecond CPA laser down to 20 fs pulse duration. The fabrication technology is difficult; however, the present work has permitted to identify fabrication steps which can lead to a better control and to a major reduction of the fabrication costs.

A damage threshold slightly above 1 J/cm² was measured under the incidence of 50 fs TE-polarized pulses in both grating types. This result may have a major impact on the whole laser system as it equalizes the limitations set by the amplifier saturation level and by the compressor’s flux resistance. As from now it is possible to envisage laser systems delivering ultrashort, high energy pulses with optical elements of reduced area. These results must now be confirmed with large gratings operating at the system scale.