4.1.1.

Metal gratings

The pioneer PCG is metal grating, which is usually fabricated as ACG. Its fabrication is typical holographic technique, which has been well developed for several decades. The grating structures are produced by the intersection of two coherent beams in a photoresponsive material.^89–93 It can be decomposed into four steps, as shown in Fig. 2(a):⁹⁴ (i) deposition of a uniform layer of photoresist on a flat glass substrate; (ii) holographic exposure of a standing-wave interference pattern on the photoresist layer, as shown in Fig. 2(b);³⁶ (iii) elution of the exposed photoresist; and (iv) deposition of a thin gold film on the structure.

Fig. 2

The fabrication process of the ACG: (a) the four steps of ACG fabrication⁹⁴ and (b) the schematic of holographic grating.³⁶

4.1.2.

Multilayer dielectric gratings

To improve the low LIDT, MDGs were proposed to replace ACGs. Its structural diagram is shown in Fig. 3. It is consisted of pairs of dielectric material layers with alternating high and low refractive indices. The DE of MDG can be improved to nearly 100%, and the LIDT can also be improved attributed to the low absorption of dielectric materials, which has been widely evidenced.^95–104 Its performance can be optimized through adjusting the parameters of residual layer thickness, groove depth, and the material of the surface relief.

Fig. 3

Structural diagram of MDG. It was consisted of several pairs of alternating high and low dielectric films. And the groove was etched on the top layer. $h$ is the groove depth. $t_r$ is residual layer thickness. $\mathrm{\Lambda }$ is the grating period. $f$ is the duty cycle. The dark and light color layers represented materials of high and low refractive indices, respectively.

4.1.3.

Metal multilayer dielectric gratings

The key difference between MDG and MMDG is that one metal layer is inserted between the substrate and the dielectric materials in MMDG. Its structural diagram is shown in Fig. 4. It shows good LIDT and DE as that of MDG. However, the DE bandwidth can be effectively broadened. Meanwhile, the metal layer replaces some pairs of dielectric layers, which releases the stress between the whole stack and improves the stability of the structure.

Fig. 4

Structural diagram of MMDG. It was consisted of one layer of metal and several pairs of alternating high and low dielectric films. And the groove was etched on the top layer. $h$ was the groove depth. $t_r$ was the residual layer thickness. $\Lambda$ was the grating period. $f$ was the duty cycle. The dark and light color layers represented materials of high and low refractive indices, respectively.

4.2.1.

Metal gratings

The ACG can show good DE in a broad bandwidth (close to 300 nm), and the ACG is still desired element used as femtosecond compressors for high-intensity laser.^105–108 However, due to the intrinsic intense absorption property of metal, its utilization is seriously restrained by its low LIDT. Therefore, the further research focuses on the improvement of LIDT. In 1995, Boyd et al.¹⁰⁹ systematically investigated the design, fabrication, and LIDT of ACG. The fabricated ACG exhibited DE in excess of 95%. They concluded that the DE of ACG prepared by e-beam evaporated coatings was higher than that of sputtering evaporation. They developed a simple theory to study the relationship between gold thickness and LIDT and systematically investigated the laser damage behavior for different pulses. Figure 5 shows that the theoretical and measured LIDT for 1053-nm pulse laser incident on a gold film. For long pulses, there was an approximately linear dependence on film thickness to 200 nm. For short pulses, similar to that of long pulses, there was also a linear dependence on film thickness, but only to a thickness near the penetration depth. Beyond this thickness, LIDT was predicted to be independent of film thickness. The LIDT of gratings was always lower than that of uniform metal films, which did not agree with the expectation that the ACGs had close LIDT to gold films. The phenomenon maybe aroused by the field enhancement or the plasma formation.

Fig. 5

Theoretical and measured LIDT for short pulse (600 fs) and long pulse (800 ps) of 1053 nm laser. The solid curve and the dashed curve represented the theoretical LIDT of long pulse and short pulse for a flat surface, respectively. The filled symbols showed the measured values for gold films deposited on photoresist, and the open symbols showed the measured values for gold-coated gratings. Circles were long pulse, and triangles were short pulse.¹⁰⁹

In 1996, Stuart et al. investigated the LIDT of ACGs at 1053 and 526 nm for pulse durations ranging from 140 fs to 1 ns. It is found that the LIDT was limited to $0.6\mathrm{J}/{\mathrm{cm}}^2$ for the subpicosecond pulses at 1053 nm. Figure 6 shows that the damage resulted from extremely rapid ablation and vaporization for very short pulses (0.6 ps), and melting, flow, and resolidification for long pulses (900 ps). The heat conduction model was utilized to predict the dependence of LIDT on coating thickness and laser pulse duration. Figure 7(a) shows the theoretical and measured LIDT for gold films of various thicknesses at 1053 nm. It was found that there was a linear dependence on film thickness up to 200 nm for long pulse (800 ps) and linear dependence on film thickness below the penetration depth. Beyond the penetration depth, the LIDT was independent of film thickness. The ${\tau }^{1/2}$ dependence of the LIDT on pulse duration $\tau$ was observed down to 200 ps, whereas the LIDT was nearly independent of pulse duration below 1 ns, as shown in Fig. 7(b). The theoretical model can be well quantitative agreement with pulse duration and wavelength of experimental results.¹¹⁰

Fig. 6

Damage to gold films with 1053 nm pulses: (a) long pulse, 900 ps and (b) short pulse, 0.6 ps.¹¹⁰

Fig. 7

Theoretical and measured LIDT for 1053 nm. (a) LIDT for gold films with different thicknesses. Circles were long pulse (800 ps); triangles were short pulse (600 fs); and curves were the theoretical results. (b) Pulse duration dependence of LIDT of a gold grating and a gold mirror.¹¹⁰

Britten et al. developed a holographically produced ACG. It exhibited high LIDT and 800 to 1100 nm wavelength range for $\mathrm{DE}>91\%$. Maximum $\mathrm{DE}>93\%$ was measured for TM polarization at 1053 nm. Figure 8 shows the DE over the whole grating with size of $40\mathrm{cm}\times 40\mathrm{cm}$. The DE was extremely uniform and $>94\%$ over the central 90% area. The LIDT was $420\mathrm{mJ}/{\mathrm{cm}}^2$ for pulse durations from 100 ps to 200 fs at 1053 nm. Its broad bandwidth testified the potential in stretching and compression of extremely short pulses (10 fs). However, it cannot compare with MDGs in terms of DE or LIDT.¹¹¹

Fig. 8

Scanning photometry map of the first-order DE of a $40\mathrm{cm}\times 40\mathrm{cm}$ $1480\mathrm{l}/\mathrm{mm}$ grating. The measurements were made near Littrow angle with 1047-nm TM polarization. The average DE was 94.5%, neglecting the upper and lower 5 cm of the grating.¹¹¹

4.2.2.

Multilayer dielectric gratings

Dielectric materials barely absorb any energy, so it inherently shows good LIDT. The main focus is the improvement of bandwidth with high DE through optimizing the structure parameters.

In 1995, Perry et al. reported an MDG, consisted of alternating layers of ZnS ($n_{\mathrm{H}}=2.35$) and ${\mathrm{ThF}}_4$ ($n_{\mathrm{L}}=1.52$), and the surface relief was etched on ZnS. The DE of MDG was related to wavelength and polarization of incident light, the shape, and depth of the grooves. The calculations predicted that the peak DE for TE polarization was expected to 98%, whereas it was $<50\%$ for TM polarization. The DE was highly sensitive to the polarization of the incident light, which was widely confirmed in later works. The holographical fabricated MDG showed DE of 96% at 1053 nm in the first order for TE polarization, as shown in Fig. 9. The experimental results were in good agreement with the calculation results.¹¹² Li et al. pointed out that multilayer dielectric (MLD) stack showed high reflectivity, low absorption, and high LIDT. The theoretical results indicated that the peak DE was nearly to 100%. They directly deposited nine layers, $(\mathrm{HL}){}^4\mathrm{H}$, of $\mathrm{ZnS}/{\mathrm{Na}}_3{\mathrm{AlF}}_6$ on the etched photoresist gratings. The test results showed that the DE was 70% and 84% corresponding to incidence from air and substrate, respectively. They inferred that the discrepancy between the theory and the experiment resulted from the non-conformally coated gratings, which was unavoidable for directly deposited grating on photoresist. The key for further improving DE was improving the conformability of the coated dielectric layers. Finally, they pointed out that three considerations answered for the situation all dielectric gratings have not been developed. (i) The coated multilayer gratings were not suitable for spectroscopic applications due to the existence of a large number of diffraction anomalies. (ii) There were no tools for designing coated multilayer gratings. (iii) It was difficult to deposit conformable thin film coatings.¹¹³ In 1996, Britten et al. first proposed that alternating quarter-wave layers of ${\mathrm{HfO}}_2$ and half-wave layers of ${\mathrm{SiO}}_2$ (HLL) can be utilized to fabricate MLD mirror, and the surface relief can be etched into the top layer either ${\mathrm{HfO}}_2$ or ${\mathrm{SiO}}_2$. The calculation predicted that obtaining the same DE, the theoretical groove depth of ${\mathrm{SiO}}_2$ was deeper 300 nm than that of ${\mathrm{HfO}}_2$, which was in good agreement with the experiments, as shown in Fig. 10. Figure 10(a) shows the MDG etched on ${\mathrm{HfO}}_2$ top layer. Its DE was 95% for TE polarization at 1053 nm, which was lower than the designed 99%. This may be attributed to the thinner thickness than the designed 294 nm. The LIDT was $0.21\mathrm{J}/{\mathrm{cm}}^2$ at 1053 nm for 300 fs pulse. The LIDT was disappointing and lower than the gold-overcoated photoresist gratings. They claimed that it may be influenced by the preferential loss of oxygen in the etching process, resulting in reduced LIDT. Figure 10(b) shows the MDG etched on ${\mathrm{SiO}}_2$ top layer with DE of 94%. The LIDT was $0.51\mathrm{J}/{\mathrm{cm}}^2$, which was about 25% higher than that of gold-overcoated photoresist gratings. It was encouraging for the improvement of LIDT. However, more investigations should be executed to further improve the LIDT of MDGs. Meanwhile, because of its good optical property and high laser resistance, ${\mathrm{HfO}}_2/{\mathrm{SiO}}_2$ has been the main stream materials in MDGs.^100,101

Fig. 9

The property of fabricated MDG: (a) SEM image of the MDG and (b) the DE of MDG in (a). The first-order TM polarization: solid curve, open squares; the first-order TE polarization: solid curve, open circles; and zeroth reflection TE polarization: dotted-dashed curve, open triangles. The incident wavelength was 1053 nm.¹¹²

Fig. 10

SEM images of fabricated MDGs. (a) SEM of HLL MDG etched into ${\mathrm{HfO}}_2$ top layer with DE of 95% for TE polarization at 1053 nm. Its groove depth was $\sim 200\mathrm{nm}$ and duty cycle was 0.4. (b) SEM of HLL MDG etched into ${\mathrm{SiO}}_2$ top layer with DE of 94% for TE polarization at 1053 nm.^114,115

In 1999, Hehl et al. confirmed that the theoretical DE of nearly 100% can be obtained. The MLD mirror was deposited with ${\mathrm{Nb}}_2{\mathrm{O}}_5$ ($n_{\mathrm{H}}=2.375$) and ${\mathrm{SiO}}_2$ ($n_{\mathrm{L}}=1.46$) on a plane fused-silica substrate. The grating was etched into the top layer of ${\mathrm{SiO}}_2$ by ion beam etching. Its DE was 97% in the first order for TE polarization at 532 nm. The LIDTs were 4.4 and $0.18\mathrm{J}/{\mathrm{cm}}^2$ under pulse duration of 5 ns and 1 ps at 532 nm.¹¹⁶ In 2003, Wei et al. investigated the basic principle for the generation of high DE of MDG utilizing $S$ matrix method and identified the conditions for achieving high DE. They concluded that the diffraction of the grating can be explained as the interference of a symmetric wave and an antisymmetric wave. The high DE can be achieved when the two left diffracted waves generated by the symmetric and antisymmetric incident waves were in phase. The analysis provided good guidance for the design of high-efficiency MDGs.¹¹⁷

In 2005, Kong et al. designed and analyzed a MLD mirror used in PCGs, the stack of H3L(H2L)^9H0.5L2.03H, where H is ${\mathrm{HfO}}_2$ and L is ${\mathrm{SiO}}_2$. The transmittance was 0.29% at 1053 nm with 51.2 deg for TE polarization. The bandwidth was 70 nm centered at 1053 nm with the reflectivity $>99.5\%$. The experimental transmission was well agreement with the theoretical design, as shown in Figs. 11.^118,119

Fig. 11

The comparison of optical properties between theory and experiment. The stack was H3L(H2L)^9H0.5L2.03H.^118,119

Non-uniform optical near-field distribution was one of the important factors limiting the LIDT of MDGs. Consequently, optimizing the electric field distribution in MDGs was an important means to further improve the LIDT. In 2006, Liu et al. deeply investigated the electric field distributions in gratings and multilayer film region using Fourier modal method. The near-field distributions in the gratings ridge was closely related to the gratings parameters, such as top layer thickness, groove depth, duty cycle, and gratings material. They developed a merit function to balance the DE and the electric field enhancement in the gratings ridge. After optimization, the grating parameters can be obtained with lowest electric field enhancement, thus, higher LIDT, as shown in Fig. 12.^120–122 Néauport et al. fabricated several samples with different electric field distributions inside the MDGs. The close optical and AFM inspection of damage sites showed that damage occurs where the electric field was maximum calculated using the differential method. Figure 13 shows the electric field distributions of two MDGs with different parameters. Figure 13(a) shows obvious electric field enhancement in grating ridge and interface compared to that of Fig. 13(b). The test showed that the LIDTs were 3.6 and $4.5\mathrm{J}/{\mathrm{cm}}^2$ for MDGs of (a) and (b), respectively. It was verified that the LIDT of MDGs was determined by the value of $E^2$.¹²³

Fig. 12

Near-field distributions of MDGs after optimization: (a) ${\mathrm{HfO}}_2$ top layer and (b) ${\mathrm{SiO}}_2$ top layer. The top layer thickness, groove depth, and duty cycle were 700 nm, 540 nm, and 0.22 in (a) and 910 nm, 684 nm, and 0.32 in (b), respectively.¹²⁰

Fig. 13

The electric field distributions in MDGs with different parameters. The residual layer thickness, groove depth, and duty cycle were 0 nm, 303 nm, and 0.53 in (a) and 18 nm, 440 nm, and 0.35 in (b), respectively.¹²³

In 2010, Wang et al. analyzed the restriction factors of widening bandwidth of MDGs, including the reflectivity bandwidth of MLD mirror and the guided-mode resonance (GMR) phenomenon. The existence of GMR in MDG would destroy the pulse spectrum shape and depress the LIDT of MDG. They proved that the bandwidth of MDG was determined by the bandwidth of high-reflectivity mirror for the first order transmitted diffraction. They inferred that reducing grating period was an effective approach to eliminate GMR in MDG, thus, broadening the bandwidth of MDG.¹²⁴ Later, they designed an MDG with the groove depth $<80\mathrm{nm}$ using particle swarm optimization algorithm and Fourier modal method, which was much shallower than the reported MDGs. The shallow groove depth was beneficial for the grating etch process. Its bandwidth was 60 nm with DE higher than 97.5% centered at 800 nm.^125,126

The DE and bandwidth of MDGs have been enough optimized.^127–138 Afterward, the investigations mainly focused on the improvement of LIDT, which can be achieved through optimizing the structure and the fabrication technology, which will be discussed in Sec. 4.2.3.

4.2.3.

Metal multilayer dielectric gratings

The MDG often consisted of several pairs of alternating high and low dielectric film, which would cause mechanical stress among the multilayer stack. The mechanical stress can lead to craze and reduce its properties, as shown in Fig. 14.¹³⁹ For this reason, the MMDG model was first proposed in 2006 by Bonod and Néauport¹⁴⁰ One layer metal was inserted between the substrate and the MLD stack, and some pairs of dielectric bilayers were replaced. The number of bilayers was decreased, and thus the mechanical stress was reduced within the stack, which effectively improves the DE, LIDT, and the stability of MMDG. Theoretical calculation showed that the number of bilayers can be decreased to 7 from 9. This was the first report and prototype of MMDG. Afterward, this research group devoted themselves to the investigation of MMDGs and effectively improved the optical performance of MMDGs.^140–142

Fig. 14

Crazing phenomenon observed on an ${\mathrm{HfO}}_2/{\mathrm{SiO}}_2$ high-reflection e-beam evaporated multilayer dielectric mirror. It was result from the mechanical stress among the multilayer stack.¹³⁹

With the development of MMDGs, it was found that the insertion of metal layer can not only decrease the mechanical stress, but also broaden the bandwidth. When the pulse duration was further compressed to femtoseconds, the bandwidth of pulse can reach to 100 nm and even 200 nm. The work bandwidth of MDG cannot meet the performance requirements.^143,144 Since metal layer always showed good reflectivity over broad bandwidth, Flury et al. reported a high-efficiency wide-band MMDG. It showed DE higher than 95% over 200 nm wavelength range centered at 800 nm for TE polarization. The experimental result verified that the DE of the first order can reach to 98% and was coincide well with theoretical simulation, as shown in Fig. 15. This kind of MMDG was a potential to realize high-efficiency CPA of femtosecond pulses as short as 20 fs. However, they did not test the LIDT of the MMDG.¹⁴⁵

Fig. 15

The experimental first-order DE and zeroth reflected order spectra under incidence of 50 deg for TE polarization.¹⁴⁵

Kong et al. have committed to the work of novel structure design, performance analysis, and optimization. First, Ag layer was inserted between substrate and bilayers, the MMDG showed DE higher than 97% over 130 nm bandwidth centered at 800 nm, and DE higher than 97% over 154 nm bandwidth centered at 1053 nm for TE polarization, respectively.¹⁴⁶ Later, with three kinds of dielectric materials, ${\mathrm{HfO}}_2$, ${\mathrm{SiO}}_2$, and ${\mathrm{Ti}}_2{\mathrm{O}}_5$, the designed MMDG showed DE higher than 97% over 150 nm bandwidth centered at 800 nm,¹⁴⁷ DE higher than 97% over 195 nm bandwidth centered at 1053 nm for TE polarization, respectively.¹⁴⁸ To further decrease the number of dielectric layers, MMDGs, consisted of only single pair of bilayers, were designed. And they showed DE higher than 97% over 120 nm bandwidth centered at 800 nm,¹⁴⁹ and DE higher than 97% over 160 nm bandwidth centered at 1053 nm for TE polarization, respectively.¹⁵⁰ The above MMDGs exhibited good optical performance and favorable fabrication tolerance. It provided good theoretical guidance for the fabrication of MMDGs. However, they did not show the corresponding fabrication technology and product.^151,152

In 2010, Wang et al. reported a new MMDG, consisted of Ag, ${\mathrm{HfO}}_2$, and ${\mathrm{SiO}}_2$, with DE higher than 97% over 200-nm bandwidth centered at 1053 nm for TE polarization. It only consisted of a metal layer, a low-index material layer, and a high-index material with etched groove depth, as shown in Fig. 16. The number of total layers was minimized, which was beneficial for the fabrication.¹⁵³ In 2013, Guan et al. designed an MMDGs centered at 800 nm. The DE, bandwidth, and near-field distributions were theoretically analyzed in detail, especially the effect of single match layer and multimatch layers on the optical performance of MMDGs. The MMDG with single match layer showed that the minimum electric field distributes in the metal layer, and the maximum electric field distributes in grating ridge, which can improve the LIDT of MMDG. If the thickness and refractive index of the match layer were changed, the maximum electric field in the grating ridge, match layer, and metal layer increased with the decrease of DE. For the MMDG with multidielectric match layers, the bandwidth and the maximum electric field in the metal layer decreased with the increase of bilayers, and the maximum electric field in the grating ridge and match layer decrease, as shown in Fig. 17. The maximum bandwidth and minimum electric field should be balanced according to the requirement.^154,155

Fig. 16

(a) Structure of a new MMDG with one metal layer, one low-index dielectric layer, and one high-index dielectric layer with etched groove depth. (b) Ultrabroad top-hat DE spectrum of (a). Its parameters were: groove depth, 315 nm; residual thickness, 173 nm; low-index layer thickness, 140 nm; and duty cycle, 0.25.¹⁵³

Fig. 17

(a) Structure of MMDG with single dielectric match layer. (b) The electric field distribution with parameters of MMDG: groove depth, 632 nm; match layer thickness, 169 nm; period, 931 nm; and duty cycle, 0.26. (c) Structure of MMDG with multidielectric match layers. (d) The changes of bandwidth and maximum electric field in the Ag layer versus the number of alternating bilayers.¹⁵⁴

The previous gratings were all designed aiming at sole either TE or TM polarization. In 2012, Hu et al. designed a polarization-independent wideband MMDG. The MMDG consisted of a metal layer and a connecting layer, and the rectangular groove was etched on the top dielectric layer. The MMDG exhibited the DE higher than 90% over 120 nm bandwidth centered at 800 nm for both TE and TM polarization, as shown in Fig. 18. This MMDG had good potential applications in laser systems and spectrometers.¹⁵⁶

Fig. 18

(a) Schematic of the MMDG: $n_1$ and $n_2$ are refractive indices of grating grooves and ridges, respectively; ${\theta }_i$, incident angle; $d$, grating period; $b$, ridge width; $h_r$, grating depth; and $h_c$, thickness of connecting layer. (b) The first-order diffraction efficiencies for both TE and TM polarizations.¹⁵⁶

4.3.1.

Metal gratings

The systematical investigation of ACGs has suspended for several years imputing its difficult improvement of LIDT.^157,158 However, until now, ACG was difficult to be absolutely replaced for femtosecond pulses (especially for pulse shorter than 100 fs) attributed to its broad bandwidth.^159,160 In 2013, Poole et al. found that the LIDT has no dependence on laser pulse duration for femtosecond laser, but showed clear dependence on Au surface morphology. Figure 19 shows the SEM images of the grating with different coating techniques. Electromagnetic field modeling showed that non-conformal (NC) coating morphology aroused significant local field enhancement near groove edges, lowering the DE and LIDT. And the experimental results verified the effectiveness of the model. As shown in Fig. 20, it was found that the conformal coating showed obvious higher LIDT than NC coating. Meanwhile, the LIDT had no dependence on pulse duration. Finally, they testified that the Ag-coated grating performed higher LIDT than ACG, which maybe a satisfactory substitute.¹⁰⁶ In 2014, Li et al. theoretically investigated the performance of ACGs with different profiles, such as rectangular, sinusoid, and semisinusoid using RCWA.¹⁶¹ It was found that the duty cycle should be excess of 0.5 for the rectangular or semisinusoidal groove, and duty cycle would cause DE to change sharply at the short wavelength. The ACG fabricated with holographic recording method exhibited DE excess of 94% for TM polarization at 808 nm, as shown in Fig. 21. In 2015, Muhutujiang et al. investigated an approach, in which the gratings patterns were generated by directly etching the quartz substrate. Its fabrication process was as follows: (i) preparation of a 220-nm-thick photoresist layer on a quartz substrate; (ii) holographic exposure on the photoresist layer; (iii) developing the grating masks; (iv) ion beam etching the quartz substrate; and (v) depositing gold on the substrate grating. The fabricated ACG showed good DE, the average DE was 89.2% over the wavelength 750 to 850 nm for TM polarization, and the peak DE is 90%. The SEM image of the etched grating is shown in Fig. 22.¹⁶² Recently, Jin et al. have deeply studied the damage process and the effect on LIDT for different fabrication methods, such as magnetron sputtering and e-beam evaporation. The LIDTs were 0.59 and $0.43\mathrm{J}/{\mathrm{cm}}^2$ for ACGs fabricated by magnetron sputtering and e-beam evaporation at pulse duration of 60 fs, respectively. It showed that the adhesion between the gold film and the photoresist determined the damage behavior, and the magnetron sputtering can produce ACGs with better adhesion, thus, improving the LIDT.^94,163

Fig. 19

SEM image of fabricated gratings: (a) conventional sputter-coated grating showing an NC groove structure and (b) energetic sputter-coated grating showing more uniform groove structure.¹⁰⁶

Fig. 20

The dependence of pulse duration on LIDT. While the conformal ($C$) coating showed higher LIDT than the NC coating, and there was no clear pulse duration dependence of the LIDT.¹⁰⁶

Fig. 21

ACGs with different profiles: (a) sinusoid; (b) rectangular; (c) semisinusoid; and (d) the fabricated grating with aperture of $200\mathrm{mm}\times 400\mathrm{mm}$.¹⁶¹

Fig. 22

SEM image of fabricated grating. The sinusoid profile ACG with line densities of 1740 l/mm. Au layer thickness was 200 nm.¹⁶²

4.3.2.

Multilayer dielectric gratings

Previous work has adequately reported the theoretical design and experimental results on DE and LIDT. However, fabrication technology, which was a much important procedure for the realization of the designed properties, has been barely reported. In 2005, Oliver et al. focused on the fabrication of an MLD to allow the fabrication of a grating with higher DE and LIDT and investigated the effect of MLD on the holographic exposure quality. By modifying the MLD structure and suppressing the reflectance of the MLD coating during holographic exposure, the straighter sidewalls of the grating pillars can be obtained, thus, yielding a higher quality MDG with greater control of duty cycle.¹⁶⁴

In 2006, Ashe et al. systematically investigated the effectiveness of some wet-chemical cleaning processes on gratings for improving the DE and LIDT, which were commonly used in semiconductor chemical cleaning processes. It was found that the DE and LIDT can together be improved with the Piranha cleaning process. The SEM images showed no visual contamination after cleaning. This was promising for improving the optical performance of MDGs.^165,166 In 2006, Kong et al. studied the effect on LIDT of MDGs for different fabrication processes. They found that the LIDT of MDG was obviously lower than that of MLD and the cleaning of grating was beneficial for improving LIDT. Finally, they pointed out that the following methods may improve the LIDT of MDGs. (i) Modifying the electric field distributions makes the peak electric field not to distribute in the unsubstantial district. (ii) Modifying the fabrication technology decreases the defects and impurity in the films. (iii) The MLD can be cleaned by surface treatment technology. (iv) Modifying the technology of e-beam etching reduces the change of ${\mathrm{HfO}}_2$ stoichiometric proportion.^167–169

In 2007, Lyndin et al. designed and fabricated an all-dielectric grating with top-hat high DE over a broad spectral. As shown in Fig. 23(a), the grating was etched in high-refractive index layer instead of the traditional low-index silica layer. The experimental result indicated that the groove depth was 123 nm and duty cycle was 0.38. The average DE is 97% over 777 to 815 nm under an incidence angle of 57 deg for TE polarization, which was well agreement with the theoretical simulation, as shown in Fig. 23(b). The zeroth order DE was 2% on the average and exhibited sharp peaks at the band edges.¹⁷⁰

Fig. 23

(a) Cross-sectional view of the dielectric mirror-based leaky mode propagating in the high-index layer and (b) experimental DE spectrum. The first-order average DE was 97% over 777 to 815 nm under an incidence angle of 57 deg for TE polarization.¹⁷⁰

In 2007, Liu et al. proposed that the ${\mathrm{SiO}}_2$ material as the top layer of the MLD mirror for grating fabrication was beneficial for improving the LIDT, and it was experimentally verified. They designed an optimal design of (HLL)^9H as the MLD, the ${\mathrm{SiO}}_2$ top layer was optimized considering the DE and the electric field enhancement. The average LIDT was $10\mathrm{J}/{\mathrm{cm}}^2$ under 12 ns pulses at 1053 nm.^171,172

Liu et al. specially investigated the effect of the photoresist gratings with different profiles as mask on the transferred grating profiles. They concluded that it was necessary to utilize high and steep enough photoresist grating mask to obtain grating with vertical profiles, as shown in Fig. 24.^173,174

Fig. 24

SEM images of photoresist grating and the etched ${\mathrm{SiO}}_2$ grating with different profile masks. (a) Original photoresist grating mask I: groove depth $\sim 420\mathrm{nm}$, duty cycle $\sim 0.22$, and sidewall angle $\sim 85\mathrm{deg}$. (b) Grating etched into ${\mathrm{SiO}}_2$ with mask I: groove depth $\sim 485\mathrm{nm}$, duty cycle $\sim 0.36$, sidewall angle $\sim 85\mathrm{deg}$. (c) Original photoresist grating mask II: groove depth $\sim 540\mathrm{nm}$, duty cycle $\sim 0.19$, and sidewall angle $\sim 90\mathrm{deg}$. (d) Grating etched into ${\mathrm{SiO}}_2$ with mask II: groove depth $\sim 690\mathrm{nm}$, duty cycle $\sim 0.23$, and sidewall angle $\sim 90\mathrm{deg}$. (e) Original photoresist grating mask III: groove depth $\sim 550\mathrm{nm}$, duty cycle $\sim 0.19$, and sidewall angle $\sim 90\mathrm{deg}$. (f) Grating etched into ${\mathrm{SiO}}_2$ with mask II: groove depth $\sim 960\mathrm{nm}$, duty cycle $\sim 0.22$, and sidewall angle $\sim 90\mathrm{deg}$.¹⁷¹

Recently, the investigations for the effect of electric field on LIDT were attractive. These works focused on optimization of the electric field distributions in the grating ridges and the damage process, the LIDT can be improved.^175–185 Meanwhile, much research also verified that the LIDT can also be improved through cleaning the surface, which can effectively reduce the defects and impurity.^186–189 Howard et al. proposed a chemical cleaning process to eliminate the contaminants. Figure 25 shows that BARC and photoresist layers on the pillar tops were effectively removed and the grating pillars were narrowed after the cleaning process.¹⁹²

Fig. 25

SEM images showing MDG cross section (a) before chemical cleaning and (b) after cleaning. The cleaning process removed BARC and photoresist layers form the pillar tops and narrowed the grating pillars.¹⁸⁹

4.3.3.

Metal multilayer dielectric gratings

In 2007, Canova et al. fabricated an MMDG with a silver mirror, a 30-nm protective layer of ${\mathrm{Al}}_2{\mathrm{O}}_3$ and the grating etched in ${\mathrm{HfO}}_2$ layer, as shown in Fig. 26(a). It exhibited a 140-nm broad bandwidth with the average DE of 95%, as shown in Fig. 26(b). They investigated the fabrication technology of MMDG and pointed out that there were the following difficulties in the fabrication process of MMDGs. (i) There was no adhesion layer between metal and dielectric to bond them and avoid delamination. (ii) The dielectric layers cannot be handled by high-temperature annealing for removing possible defects, which may decrease the LIDT. (iii) It was difficult to expose interferogram on the materials before the high-refractive index materials. (iv) To assure the physically and chemically intactness of the metal mirror surface in the process of grating etching, the etching chemistry and sputtering conditions were severely restricted. Although there were too much technical difficulties and influences, the LIDT of MMDG was comparative with that of MDG, $1.1\mathrm{J}/{\mathrm{cm}}^2$, which verified that the introduction of metal layer has no influence on the LIDT.¹⁹⁰

Fig. 26

(a) Cross-sectional view of the MMDG and (b) the experimental and theoretical first-order DE and zeroth reflected spectra under incidence of 57 deg for TE polarization.¹⁹⁰

In 2009, Palmier et al. experimentally compared the reflectivity and LIDT properties of metal multilayer dielectric (MMLD) mirror (Pyrex/20 nm Cr 150 nm Au (246 nm ${\mathrm{SiO}}_2$ 155 nm ${\mathrm{HfO}}_2$)^4) 579 nm ${\mathrm{SiO}}_2$ and MLD mirror (Pyrex/(115 nm ${\mathrm{HfO}}_2$ 311 nm ${\mathrm{SiO}}_2$)^8) 115 nm ${\mathrm{HfO}}_2$ 385 nm ${\mathrm{SiO}}_2$, which were the fundamental component of PCGs. The MMLD, four pairs of dielectric bilayers were replaced by the Au layer, was deposited by e-beam evaporation. The reflectivity is shown in Fig. 27. It showed that MMLD and MLD exhibited high and flat reflectivity centered at 1053 nm, especially the ultrabroad bandwidth of MMLD, which verified the effectiveness of broadening the bandwidth. The experimental result showed that the LIDT of MLD and MMLD were all about $5\mathrm{J}/{\mathrm{cm}}^2$.¹¹³ In 2010, Néauport et al. reported a complete process of MMDG, including design, fabrication, and test. With the goal of high DE and LIDT, to optimize the parameters of the surface relief, the designed grating profile was trapezoidal geometry, instead of traditional rectangle. The MMDG, which was deposited by e-beam evaporation, showed DE reach 96% for TE polarization centered at 1053 nm, and the LIDT was about $3\mathrm{J}/{\mathrm{cm}}^2$ under pulse duration of 500 fs, which was similar to that of MDGs.¹³⁹

Fig. 27

Reflectivity of the designed MLD and MMLD measured under incidence of 70.6 deg for TE polarization.¹³⁹

In 2014, Guan et al. pointed out that the surface relief of ${\mathrm{HfO}}_2$ material showed good bandwidth, but it was difficult to etch. The surface relief of ${\mathrm{SiO}}_2$ material should etch big groove depth and line density, but it had the advantages of high LIDT and easy fabrication. To unite the superiorities of ${\mathrm{HfO}}_2$ and ${\mathrm{SiO}}_2$ materials, a new type of MMDG was designed, as shown in Fig. 28. The grating ridge consisted of one layer ${\mathrm{HfO}}_2$ sandwiched between two layers ${\mathrm{SiO}}_2$. The bandwidth with DE higher than 90% was 200 nm, and 95% was 137 nm. The size of fabricated MMDG was $50\mathrm{mm}\times 50\mathrm{mm}$. The test results showed that the experimental DE was consistent with theoretical simulation, which verified the favorable fabrication tolerance. The LIDT was $0.32\mathrm{J}/{\mathrm{cm}}^2$ under 45 fs pulse duration at 800 nm. The optical performance of the MMDG was not the best among the reported MMDGs. However, its performance was well consistent with the designed MMDG, which built the foundations for the functionization and fabrication technology of MMDG.¹⁵⁵ Wu et al. investigated the influence of annealing temperature on MMLD. It indicated that the roughness of the MMLD changed slightly after annealing. However, the resistance to chemical cleaning damage improved with the annealing temperature increase, whereas the reflectivity decreased. The MMLD annealed at 250°C for 10 h can be an optimal annealing process for the fabrication of MMDG.¹⁹¹ In 2016, Zhang et al. pointed out that the reason for resistance improvement to chemical cleaning damage after annealing was that one transition layer was produced between Au layer and ${\mathrm{SiO}}_2$ layer, and the transition layer enhanced the adhesion between Au layer and ${\mathrm{SiO}}_2$ layer and blocked the infiltration of acid solutions.¹⁹²

Fig. 28

(a) Structure of new MMDG. The grating ridge consisted of an ${\mathrm{HfO}}_2$ center layer sandwiched between two ${\mathrm{SiO}}_2$ layers. (b) Reflectivity of MMLD under incidence of 53 deg for TE polarization. The theoretical reflectivity was coincided with the measured over the range of 600 to 800 nm. (c) SEM image of fabricated MMDG. Its duty cycle, period, and slope angle were 0.44, 565 nm, and 74 deg, respectively. (d) The first-order DE of MMDG under incidence of 53 deg for TE polarization at 800 nm.¹⁵⁵

Chen et al. deeply investigated and explored the fabrication process, including modification of resist masks, the etching speed of different materials, and the cleaning of MMDG. The DE of the fabricated MMDG had a peak DE of 95.1% at 810 nm and the maximum bandwidth is 169 nm with DE higher than 90%, the average DE was 93.71% under incidence of 53 deg for TE polarization, the LIDT was $0.32\mathrm{J}/{\mathrm{cm}}^2$ for 45 fs pulse duration. This was the first particular report for the fabrication process.^{155,193–195}

Recently, Zou et al. experimentally and theoretically investigated the effect of nodular defect on damage of MMDG. It was found that the initial damages occurred in the grating ridge of the perfect MMDG, whereas initial damages will occur in the district of nodular defect if nodular defect exists. The damage process and the relationship between damage degree and the nodular defect size were theoretically analyzed. It was concluded that local electric field enhancement in nodular defect area would lead to the damage.¹⁹⁶ Xu et al. reported a new MMDG with single Au layer and one silica layer. Its LIDT can reach $0.40\mathrm{J}/{\mathrm{cm}}^2$ under 32 fs pulse duration, which was twice more than that of gold-coated gratings. They theoretical predicted that the maximum LIDT of MMDG was $0.60\mathrm{J}/{\mathrm{cm}}^2$.¹⁹⁷

Although MMDG exhibited high diffraction, broad bandwidth, and high LIDT, the technique was not mature enough to fabricate perfect MMDG due to the existence of metal layer. More efforts should be paid to improve the performance of fabricated MMDGs.¹⁹⁸