Pulse Burst Generation and Diffraction with Spatial Light Modulators for Dynamic Ultrafast Laser Materials Processing

A pulse burst optical system has been developed, able to alter an energetic, ultrafast 10 ps, 5 kHz output pulse train to 323 MHz intra-burst frequency at the fundamental 5 kHz repetition rate. An optical delay line consisting of a beam-splitting polariser cube, mirrors, and waveplates transforms a high-energy pulse into a pulse burst, circulating around the delay line. Interestingly, the reflected first pulse and subsequent pulses from the delay line have orthogonal linear polarisations. This fact allows independent modulation of these pulses using two-phase-only Spatial Light Modulators (SLM) when their directors are also aligned orthogonally. With hybrid Computer Generated Holograms (CGH) addressed to the SLMs, we demonstrate simultaneous multi-spot periodic surface micro-structuring on stainless steel with orthogonal linear polarisations and cylindrical vector (CV) beams with Radial and Azimuthal polarisations. Burst processing produces a major change in resulting surface texture due to plasma absorption on the nanosecond time scale; hence the ablation rates on stainless steel with pulse bursts are always lower than 5 kHz processing. By synchronising the scan motion and CGH application, we show simultaneous independent multi-beam real-time processing with pulse bursts having orthogonal linear polarisations. This novel technique extends the flexibility of parallel beam surface micro-structuring with adaptive optics.


Introduction
Ultrashort, femtosecond and picosecond laser pulses are ideally suited for the study of linear light-matter interactions, for example, during pump-probe spectroscopy [1], non-linear laser lithography for large area nano-structuring [2], and for industrial applications such as selective ablation for thin film solar cells [3,4], micro and nanoscale surface modification for control of chemistry and wettability [5], generation of superhydrophobic and antibacterial surfaces [6], and creation of complex Laser Induced Periodic Surface Structures (LIPSS) [7,8]. For Industrial applications involving micro-machining of deepmilled structures, volume ablation rates need to be high (δV/δt > 3 mm 3 /min), so kHz, low power, multi-Watt ultrafast laser systems are simply not adequate. This limitation has been addressed recently by scaling ultrafast laser output to the multi-hundred Watt level, achieved primarily by increasing repetition rates from tens of MHz to GHz [9,10], although kHz, ultrafast, high energy thin disc systems with pulse energy E > 200 mJ, 1 kW average power and pulse length τ < 1 ps have been developed [11]. The flexibility of high-power ultrafast laser technology has also been expanded by introducing burst modes at much higher intraburst frequencies (GHz-THz) [12], which alter the laser-material interactions, leading, for example, to improved surface finish. Burst mode ablation can improve (PBS), dielectric reflecting mirrors, and internal wave plate (HWP2) within the loop. The relative setting of the HWP1,2 fast axes creates reflected and circulating pulses or a pulse burst. Figure 1. Schematic of the optical setup. The attenuated and expanded beam is directed to the delay line, HWP1/PBS plus mirrors, and HWP2. The pulse burst is controlled by altering the relative settings HWP1 and HWP2 (within the delay line). The resulting bursts are modulated by the SLMs with a 4f system (L1/L2, f = 300 mm) re-imaging diffracted spots to SLM2. A second 4f system (L3/L4, f = 400 mm) re-images all diffracted beams to the input aperture of scanning galvo. A flip mirror with a long focal length lens (L5, f = 750 mm) focuses resulting intensity distributions to a CCD camera. Samples are brought to the lens focal plane by a 3−axis stage. For dynamic control of CGH application, two PCs with a trigger signal from the shutter controller and NI input/output were employed.
The burst pulses are then incident at low AOI < 10° on SLM1 addressed with appropriate CGH and re-imaged via a 4f system (f1 = f2 = 300 mm) to SLM2 where, after further modulation, the beam was again re-imaged (via 4f system, f3 = f4 = 400 mm) to the input aperture of a scanning galvo system (Nutfield) and focused with a 100 mm focal length ftheta lens to the substrate. The scan software employed was SCAPS GmbH. The beam profile of pulse bursts prior to materials micro-structuring could be observed using a flip mirror and imaged to a Spiricon camera (SP620U) by a long focal length lens (L5, f = 750 mm). The SLMs are reflective phase-only type, Hamamatsu X13138-5785 (1280 × 1024 pixels) and X10468-03 (800 × 600 pixels), respectively, and SLM2 was mounted with its director vertical, orthogonal to that of SLM1. The CGHs shown are simple hybrid holograms used for beam shaping [27] of the quasi-Gaussian laser beam (elliptical) to circular. The inner circular regions have a fixed phase, while the outer region with high spatial frequency diffracts away from the peripheral elliptical component. The first SLM1 modulates |H〉 polarised pulses from the delay line while SLM2 modulates only |V〉 polarised reflected pulses from the PBS. Pulse bursts of diffracted horizontal and vertical polarised arrays could be generated with appropriate hybrid CGHs while arrays of orthogonal vector beams [24] could be created by inserting a nano-structured S-wave plate (SWP) [28,29]  The pulse burst is controlled by altering the relative settings HWP 1 and HWP 2 (within the delay line). The resulting bursts are modulated by the SLMs with a 4f system (L 1 /L 2 , f = 300 mm) re-imaging diffracted spots to SLM 2 . A second 4f system (L 3 /L 4 , f = 400 mm) re-images all diffracted beams to the input aperture of scanning galvo. A flip mirror with a long focal length lens (L 5 , f = 750 mm) focuses resulting intensity distributions to a CCD camera. Samples are brought to the lens focal plane by a 3-axis stage. For dynamic control of CGH application, two PCs with a trigger signal from the shutter controller and NI input/output were employed.
The burst pulses are then incident at low AOI < 10 • on SLM 1 addressed with appropriate CGH and re-imaged via a 4f system (f1 = f2 = 300 mm) to SLM 2 where, after further modulation, the beam was again re-imaged (via 4f system, f3 = f4 = 400 mm) to the input aperture of a scanning galvo system (Nutfield) and focused with a 100 mm focal length f-theta lens to the substrate. The scan software employed was SCAPS GmbH. The beam profile of pulse bursts prior to materials micro-structuring could be observed using a flip mirror and imaged to a Spiricon camera (SP620U) by a long focal length lens (L 5 , f = 750 mm). The SLMs are reflective phase-only type, Hamamatsu X13138-5785 (1280 × 1024 pixels) and X10468-03 (800 × 600 pixels), respectively, and SLM 2 was mounted with its director vertical, orthogonal to that of SLM 1 . The CGHs shown are simple hybrid holograms used for beam shaping [27] of the quasi-Gaussian laser beam (elliptical) to circular. The inner circular regions have a fixed phase, while the outer region with high spatial frequency diffracts away from the peripheral elliptical component. The first SLM 1 modulates |H polarised pulses from the delay line while SLM 2 modulates only |V polarised reflected pulses from the PBS. Pulse bursts of diffracted horizontal and vertical polarised arrays could be generated with appropriate hybrid CGHs while arrays of orthogonal vector beams [24] could be created by inserting a nano-structured S-wave plate (SWP) [28,29] just beyond (and in close proximity) to SLM 2 . Polished 304 stainless steel samples (roughness Ra = 50 nm) were mounted on a three-axis (x,y,z) motion control system (Aerotech) to bring the sample surface to the focal plane.

Pulse Burst Generator
Referring to Figure 2, HWP 1 rotates the incident linear polarisation at the PBS interface; hence, the amplitude of reflected, |V polarised and transmitted |H polarised components are determined by the HWP 1 fast axis setting. The internally transmitted pulse circulates around the delay line, and if no internal waveplate were added, it would exit the delay line at the PBS with |H polarisation, thus creating an orthogonally polarised pulse pair. The addition of HWP 2 within the delay line alters internal polarisation so that the circulating pulse is again split into transmitted |H and reflected |V components at the PBS. The relative setting of external and internal waveplates thus allows the creation of a pulse burst with burst number N P ≥ 2 with polarisations, |V , |H , |H , |H . . . .and so on. The optical path length from the PBS centre around the loop was measured to be ∆L = 91.7 ± 0.5 cm. Allowing for the PBS dimensions and its refractive index (2.5 cm cube, RI = 1.5), the effective optical path delay of the loop OPD = 93.0 ± 0.5 cm and hence the burst frequency ν IB = 322.8 ± 1.6 MHz determined primarily by the speed of light in air. We designate pulse bursts as (1,1,0) for two pulse orthogonal polarisations with a 3.1 ns delay, while other pulse sequences can be designated (1,0,1) or (0,1,1) for 2-pulse burst orthogonal polarisations (6.2 ns delay) or identical |H polarisations (3.1 ns delay) respectively. Similarly, (1,1,1) represents a 3-pulse burst with the two delayed pulses, |H polarised, orthogonal to the first pulse |V polarised. just beyond (and in close proximity) to SLM2. Polished 304 stainless steel samples (roughness Ra = 50 nm) were mounted on a three-axis (x,y,z,) motion control system (Aerotech) to bring the sample surface to the focal plane.

Pulse Burst Generator
Referring to Figure 2, HWP1 rotates the incident linear polarisation at the PBS interface; hence, the amplitude of reflected, |V〉 polarised and transmitted |H〉 polarised components are determined by the HWP1 fast axis setting. The internally transmitted pulse circulates around the delay line, and if no internal waveplate were added, it would exit the delay line at the PBS with |H〉 polarisation, thus creating an orthogonally polarised pulse pair. The addition of HWP2 within the delay line alters internal polarisation so that the circulating pulse is again split into transmitted |H〉 and reflected |V〉 components at the PBS. The relative setting of external and internal waveplates thus allows the creation of a pulse burst with burst number NP ≥ 2 with polarisations, |V〉, |H〉, |H〉, |H〉….and so on. The optical path length from the PBS centre around the loop was measured to be ΔL = 91.7 ± 0.5 cm. Allowing for the PBS dimensions and its refractive index (2.5 cm cube, RI = 1.5), the effective optical path delay of the loop OPD = 93.0 ± 0.5 cm and hence the burst frequency νIB = 322.8 ± 1.6 MHz determined primarily by the speed of light in air. We designate pulse bursts as (1,1,0) for two pulse orthogonal polarisations with a 3.1 ns delay, while other pulse sequences can be designated (1,0,1) or (0,1,1) for 2−pulse burst orthogonal polarisations (6.2 ns delay) or identical |H〉 polarisations (3.1 ns delay) respectively. Similarly, (1,1,1) represents a 3−pulse burst with the two delayed pulses, |H〉 polarised, orthogonal to the first pulse |V〉 polarised. Figure 2. Schematic of the pulse burst generator setup with external HWP1, PBS, dielectric mirrors, and internal HWP2. An incident pulse linearly polarised with E field direction, θ to the horizontal axis is split at the PBS, creating reflected, |V〉 polarised and transmitted, |H〉 polarised components. The polarisation of the transmitted pulse is then rotated within the delay line by HWP2 splitting again at the PBS so generating the pulse burst. Polarisation vectors on reflected/transmitted beams are shown, while the setting of the fast axes of HWP1 and HWP2 determines the pulse burst characteristics. Careful adjustment of the delay line mirrors allowed spatial overlap of the pulse burst in both the near and far fields. Temporal pulse separation Δt = 3.1 ns.
A schematic diagram of a 3−pulse burst output at 5 kHz fundamental frequency with νIB = 323 MHz intra-burst frequency is shown in Figure 3. A schematic diagram of a 3-pulse burst output at 5 kHz fundamental frequency with ν IB = 323 MHz intra-burst frequency is shown in Figure 3.

Hybrid CGHs
The use of a flip-up mirror with a long-focus lens (f = 750 mm) allowed the delay line burst output to be imaged to a Spiricon CCD camera to check relative spot intensities/spatial overlap prior to surface micro-machining experiments. Figure 4a shows a 3−pulse (1,1,1) CCD image from the delay line with nearly equal pulse energies, spatially separated using a slight misalignment of the delay line, then spatially overlapped, Figure 4b. No correction with hybrid CGHs are used here, and the result of blind drilling on stainless steel with the overlapped pulse burst is shown in the optical image, Figure 4c, with near elliptical cross-section and eccentricity ε = 0.72. The laser mode from the regenerative amplifier is only quasi-Gaussian and so eccentricity ε = ω0 (min)/ω0 (max) where ω0 (min) and ω0 (max) are the measured ablation semi-minor and semi-major axes, close to the 1/e 2 radii at intensity focus. Figure 4d shows the effect of using hybrid holograms (Figure 1), which diffract away the outer intensity wings from the ellipse major axis. Figure 4e shows the overlap intensity profile, significantly improving beam roundness, while Figure 4f demonstrates (1,1,1) pulse burst processing with hybrid CGHs yielding eccentricity ε = 0.88, a great improvement but at the cost of a reduction in overall transmission efficiency.

Hybrid CGHs
The use of a flip-up mirror with a long-focus lens (f = 750 mm) allowed the delay line burst output to be imaged to a Spiricon CCD camera to check relative spot intensities/spatial overlap prior to surface micro-machining experiments. Figure 4a shows a 3-pulse (1,1,1) CCD image from the delay line with nearly equal pulse energies, spatially separated using a slight misalignment of the delay line, then spatially overlapped, Figure 4b. No correction with hybrid CGHs are used here, and the result of blind drilling on stainless steel with the overlapped pulse burst is shown in the optical image, Figure 4c, with near elliptical cross-section and eccentricity ε = 0.72. The laser mode from the regenerative amplifier is only quasi-Gaussian and so eccentricity ε = ω 0 (min)/ω 0 (max) where ω 0 (min) and ω 0 (max) are the measured ablation semi-minor and semi-major axes, close to the 1/e 2 radii at intensity focus. Figure 4d shows the effect of using hybrid holograms (Figure 1), which diffract away the outer intensity wings from the ellipse major axis. Figure 4e shows the overlap intensity profile, significantly improving beam roundness, while Figure 4f demonstrates (1,1,1) pulse burst processing with hybrid CGHs yielding eccentricity ε = 0.88, a great improvement but at the cost of a reduction in overall transmission efficiency.

Hybrid CGHs
The use of a flip-up mirror with a long-focus lens (f = 750 mm) allowed the delay line burst output to be imaged to a Spiricon CCD camera to check relative spot intensities/spatial overlap prior to surface micro-machining experiments. Figure 4a shows a 3−pulse (1,1,1) CCD image from the delay line with nearly equal pulse energies, spatially separated using a slight misalignment of the delay line, then spatially overlapped, Figure 4b. No correction with hybrid CGHs are used here, and the result of blind drilling on stainless steel with the overlapped pulse burst is shown in the optical image, Figure 4c, with near elliptical cross-section and eccentricity ε = 0.72. The laser mode from the regenerative amplifier is only quasi-Gaussian and so eccentricity ε = ω0 (min)/ω0 (max) where ω0 (min) and ω0 (max) are the measured ablation semi-minor and semi-major axes, close to the 1/e 2 radii at intensity focus. Figure 4d shows the effect of using hybrid holograms (Figure 1), which diffract away the outer intensity wings from the ellipse major axis. Figure 4e shows the overlap intensity profile, significantly improving beam roundness, while Figure 4f demonstrates (1,1,1) pulse burst processing with hybrid CGHs yielding eccentricity ε = 0.88, a great improvement but at the cost of a reduction in overall transmission efficiency.   Figure 5a,b show the hybrid CGHs used to create a 2 × 2 spot orthogonally linearly polarised Gaussian array, while Figure 5c shows the resulting CCD image of the 2 × 2 Gaussian array. By adding the nano-structured SWP to the optical line just after SLM 2 , the orthogonal linearly polarised Gaussians were converted to a vector beam array with orthogonal (Radial/Azimuthal) polarisations and ring intensity distributions, Figure 5d. One can see evidence of low-energy ghost beams here, expected from patterns with a high degree of symmetry when calculating inverse Fourier Transforms (IFTs) [30].
Materials 2022, 15, x FOR PEER REVIEW 6 of 17 Figure 5a,b show the hybrid CGHs used to create a 2 × 2 spot orthogonally linearly polarised Gaussian array, while Figure 5c shows the resulting CCD image of the 2 × 2 Gaussian array. By adding the nano-structured SWP to the optical line just after SLM2, the orthogonal linearly polarised Gaussians were converted to a vector beam array with orthogonal (Radial/Azimuthal) polarisations and ring intensity distributions, Figure 5d. One can see evidence of low-energy ghost beams here, expected from patterns with a high degree of symmetry when calculating inverse Fourier Transforms (IFTs) [30]  The high purity of the vector beam polarisations was confirmed by adding a wedged beam splitter (B/S) to the CCD camera while rotating the CCD by 90°. As the B/S angle of incidence (AOI) is close to the Brewster angle, it acts as a polarisation analyser. Vector beams with radial/azimuthal polarisations, |Ra〉, |Az〉 and their superpositions, |Ψ〉 = α|Ra〉 + βe iδ |Az〉 (where δ is a phase angle) are shown in Figure 6. The double lobe structure with an intensity null on the axis is expected from vector beam transmission through a linear polariser [7].
The double lobes are well-defined, indicating a high vector beam purity. These vector polarisation states can be represented on the equatorial axis of a high-order Poincare sphere (HOPS) while scalar ring modes with pure optical angular momentum OAM (e.g., m = ± 1, phase e ± iϕ ) and |R〉, |L〉 circular polarisations appear at the poles [31]. In between the poles and the equator, the states are cylindrical vector vortex (CVV) beams. The Laguerre-Gaussian, LG (0,1)* spiral phase mode can be represented by a superposition of two orthogonal degenerate Laguerre-Gaussian (LG) modes, given by [32], The high purity of the vector beam polarisations was confirmed by adding a wedged beam splitter (B/S) to the CCD camera while rotating the CCD by 90 • . As the B/S angle of incidence (AOI) is close to the Brewster angle, it acts as a polarisation analyser. Vector beams with radial/azimuthal polarisations, |Ra , |Az and their superpositions, |Ψ = α|Ra + βe iδ |Az (where δ is a phase angle) are shown in Figure 6. The double lobe structure with an intensity null on the axis is expected from vector beam transmission through a linear polariser [7].
The double lobes are well-defined, indicating a high vector beam purity. These vector polarisation states can be represented on the equatorial axis of a high-order Poincare sphere (HOPS) while scalar ring modes with pure optical angular momentum OAM (e.g., m = ± 1, phase e ± iφ ) and |R , |L circular polarisations appear at the poles [31]. In between the poles and the equator, the states are cylindrical vector vortex (CVV) beams. The Laguerre-Gaussian, LG (0,1)* spiral phase mode can be represented by a superposition of two orthogonal degenerate Laguerre-Gaussian (LG) modes, given by [32], where e x , e y are unit vectors along the x, y axes, r and ϕ are cylindrical coordinates where ρ = 2 r 2 /ω 2 and ω is the 1/e 2 radius of the Gaussian from which the LG mode can be generated. The intensity distribution reflected from the wedged plate (transmission axis along horizontal axis e x ) is then given by (e x • e y = 0) which has a double lobe structure and intensity null at the centre due to the exponential term.
where ex, ey are unit vectors along the x, y axes, r and φ are cylindrical coordinates where ρ = 2 r 2 /ω 2 and ω is the 1/e 2 radius of the Gaussian from which the LG mode can be generated. The intensity distribution reflected from the wedged plate (transmission axis along horizontal axis ex ) is then given by (ex • ey = 0) * • sin (2) which has a double lobe structure and intensity null at the centre due to the exponential term.

Fast Photo-Diode Outputs
Pulse burst amplitudes from the delay line were checked using a fast photo-diode (PD, Thorlabs DET025A/M, risetime τ ~ 100 ps) connected to a wide bandwidth digital oscilloscope (Tektronix 3054, 500 MHz, 5 GS/s). Figure 7 shows oscilloscope traces from output pulse bursts when adjusting the external and internal half-waveplate in the delay line. Figure 7a shows a (1,1,0) pulse pair with a 3.1 ns delay, while Figure 7b shows the (1,0,1) pulse pair with a 6.2 ns delay. Figure 7c shows the (1,1,1) 3−pulse burst at intraburst frequency νIB = 323 MHz with equal amplitude. Figure 7d shows a 4-pulse burst with decaying amplitude. It was not possible to generate equal amplitude pulse bursts for Np > 3. The temporal pulse shape does have a low-intensity component arriving approximately 2 ns before the main pulse. Surprisingly, adjustment of the Regen amplifier Pockels cell voltage and intracavity waveplate did not reduce this component appreciably, while the seed oscillator pulses showed no such temporal structure.

Fast Photo-Diode Outputs
Pulse burst amplitudes from the delay line were checked using a fast photo-diode (PD, Thorlabs DET025A/M, risetime τ~100 ps) connected to a wide bandwidth digital oscilloscope (Tektronix 3054, 500 MHz, 5 GS/s). Figure 7 shows oscilloscope traces from output pulse bursts when adjusting the external and internal half-waveplate in the delay line. Figure 7a shows a (1,1,0) pulse pair with a 3.1 ns delay, while Figure 7b shows the (1,0,1) pulse pair with a 6.2 ns delay. Figure 7c shows the (1,1,1) 3-pulse burst at intra-burst frequency ν IB = 323 MHz with equal amplitude. Figure 7d shows a 4-pulse burst with decaying amplitude. It was not possible to generate equal amplitude pulse bursts for N p > 3. The temporal pulse shape does have a low-intensity component arriving approximately 2 ns before the main pulse. Surprisingly, adjustment of the Regen amplifier Pockels cell voltage and intracavity waveplate did not reduce this component appreciably, while the seed oscillator pulses showed no such temporal structure.

Ultrafast Ablation with Orthogonal Linear Polarisations
With equal energy (1,1,0) pulse bursts, orthogonally linearly polarised and spatially diffracted, the resulting surface structuring observed on polished stainless steel with a 2 × 2 Gaussian spot array is shown in the SEM (JEOL 7001FEGSEM) images of Figure 8a-c. Pulse exposure/spot was Np = 100 with pulse energy/spot E ~ 3 μJ (F0 ~ 0.3 Jcm −2 ); hence, total burst energy on target Ep ~12 μJ. While Figure 8b shows the complete symmetric ablation pattern, Figure 8a,c with higher magnification exhibit low-frequency orthogonal LIPSS with pitch Λ ~ 1 μm at the outer edge of the exposed regions.

Ultrafast Ablation with Orthogonal Linear Polarisations
With equal energy (1,1,0) pulse bursts, orthogonally linearly polarised and spatially diffracted, the resulting surface structuring observed on polished stainless steel with a 2 × 2 Gaussian spot array is shown in the SEM (JEOL 7001FEGSEM) images of Figure 8a-c. Pulse exposure/spot was N p = 100 with pulse energy/spot E~3 µJ (F 0~0 .3 Jcm −2 ); hence, total burst energy on target E p~1 2 µJ. While Figure 8b shows the complete symmetric ablation pattern, Figure 8a,c with higher magnification exhibit low-frequency orthogonal LIPSS with pitch Λ~1 µm at the outer edge of the exposed regions.

Ultrafast Ablation with Orthogonal Linear Polarisations
With equal energy (1,1,0) pulse bursts, orthogonally linearly polarised and spatially diffracted, the resulting surface structuring observed on polished stainless steel with a 2 × 2 Gaussian spot array is shown in the SEM (JEOL 7001FEGSEM) images of Figure 8a-c. Pulse exposure/spot was Np = 100 with pulse energy/spot E ~ 3 μJ (F0 ~ 0.3 Jcm −2 ); hence, total burst energy on target Ep ~12 μJ. While Figure 8b shows the complete symmetric ablation pattern, Figure 8a,c with higher magnification exhibit low-frequency orthogonal LIPSS with pitch Λ ~ 1 μm at the outer edge of the exposed regions.

Pulse Burst Ablation with Orthogonal Radial and Azimuthal Polarisations
With the addition of the cylindrically birefringent SWP to the optical line, orthogonal linear polarisations were converted to ring mode Radial and Azimuthal polarisations. This depends only on the direction of the linear polarisations relative to the direction to the SWP axis. Thus, two incident orthogonal linear polarisations from a modulated pulse pair created simultaneous Radial and Azimuthal polarisations. Figure 9a shows SEM images of a 2 × 2 array micro-machined simultaneously on 304 stainless steel with orthogonal Radial and Azimuthal polarisations. Figure 9b shows the higher magnification images of the resulting orthogonal complex LIPPS which are well-defined, supporting the view that the vector polarisation states are quite pure ( Figure 5). Exposure pulse number N p = 100/spot with total pulse energy E~7 µJ/spot (F 0~0 .26 Jcm −2 ) and average power <P> = 140 mW exposure at the material surface. The ring mode peak fluence = (1/e) F 0 (Gaussian).

Pulse Burst Ablation with Orthogonal Radial and Azimuthal Polarisations
With the addition of the cylindrically birefringent SWP to the optical line, orthogonal linear polarisations were converted to ring mode Radial and Azimuthal polarisations. This depends only on the direction of the linear polarisations relative to the direction to the SWP axis. Thus, two incident orthogonal linear polarisations from a modulated pulse pair created simultaneous Radial and Azimuthal polarisations. Figure 9a shows SEM images of a 2 × 2 array micro-machined simultaneously on 304 stainless steel with orthogonal Radial and Azimuthal polarisations. Figure 9b shows the higher magnification images of the resulting orthogonal complex LIPPS which are well-defined, supporting the view that the vector polarisation states are quite pure ( Figure 5). Exposure pulse number Np = 100/spot with total pulse energy E ~ 7 μJ/spot (F0 ~ 0.26 Jcm −2 ) and average power <P> = 140 mW exposure at the material surface. The ring mode peak fluence = (1/e) F0 (Gaussian)

Scanning with Orthogonal Linear Polarisations, Fixed CGHs Polarisations
LIPSS are quasi-periodic surface undulations whose source can be regarded as due to the superposition of the incoming wave with a surface scattered wave which affects the spatial energy deposition on the surface [33,34]. The key parameters for producing LIPSS are the temporal pulse length (fs-ps), laser wavelength, fluence, pulse overlap, pulse exposure, and linear polarisation direction relative to scan direction. There is, therefore, a relatively wide processing window over which LIPPS can be created; however, overexposure leads to a reduction in the finesse of the LIPSS. On s. steel, a fluence in the range F = 0.3-0.4 Jcm −2 was found to be effective with a 10 ps pulse length.
Galvo scanning on the surface with orthogonal polarisations allowed fast surface LIPSS texturing. Figure 10a,b shows optical images of WL illumination of a diffractive chess board, micromachined simultaneously at 5 kHz with orthogonal linear polarisations from the 2−pulse (1,1,0) burst with Gaussian beams, inscribing LIPSS approximately at right angles when scanned in the x direction. The spot separation was adjusted to 1 mm, galvo scan speed was s = 40 mm/s, pulse overlap ~33%, and hatch distance d = 20 μm. Pulse energy/spot E = 3 μJ (F0 = 0.3 Jcm −2 ). Illumination directions are indicated with red

Scanning with Orthogonal Linear Polarisations, Fixed CGHs Polarisations
LIPSS are quasi-periodic surface undulations whose source can be regarded as due to the superposition of the incoming wave with a surface scattered wave which affects the spatial energy deposition on the surface [33,34]. The key parameters for producing LIPSS are the temporal pulse length (fs-ps), laser wavelength, fluence, pulse overlap, pulse exposure, and linear polarisation direction relative to scan direction. There is, therefore, a relatively wide processing window over which LIPPS can be created; however, overexposure leads to a reduction in the finesse of the LIPSS. On s. steel, a fluence in the range F = 0.3-0.4 Jcm −2 was found to be effective with a 10 ps pulse length.
Galvo scanning on the surface with orthogonal polarisations allowed fast surface LIPSS texturing. Figure 10a,b shows optical images of WL illumination of a diffractive chess board, micromachined simultaneously at 5 kHz with orthogonal linear polarisations from the 2-pulse (1,1,0) burst with Gaussian beams, inscribing LIPSS approximately at right angles when scanned in the x direction. The spot separation was adjusted to 1 mm, galvo scan speed was s = 40 mm/s, pulse overlap~33%, and hatch distance d = 20 µm. Pulse energy/spot E = 3 µJ (F 0 = 0.3 Jcm −2 ). Illumination directions are indicated with red arrows. The optical image in Figure 10d shows that surface texturing is highly polarisationdependent, while Figure 10c,e show SEM images of the LIPSS formation, which are not orthogonal. This is due to an E field rotation within the Galvo and LIPSS formation is sensitive to scan direction. When the E field vector is nearly parallel to scan direction, we observe stronger coupling so that low-frequency LIPSS form normal to the major E x component. On the other hand, with E tilted away from the y direction (E y major component) while scanning in x-direction, LIPSS form normal to the E field direction, not that of the E y vector. This can be resolved by rotating scan directions by approximately 20 • .
arrows. The optical image in Figure 10d shows that surface texturing is highly polarisation-dependent, while Figure 10c,e show SEM images of the LIPSS formation, which are not orthogonal. This is due to an E field rotation within the Galvo and LIPSS formation is sensitive to scan direction. When the E field vector is nearly parallel to scan direction, we observe stronger coupling so that low-frequency LIPSS form normal to the major Ex component. On the other hand, with E tilted away from the y direction (Ey major component) while scanning in x-direction, LIPSS form normal to the E field direction, not that of the Ey vector. This can be resolved by rotating scan directions by approximately 20°.

Dynamic Polarisation Modulation and Micro-Structuring
As nematic liquid crystal SLMs can be addressed in real-time, the application of CGHs can be synchronised to the scan motion control system. Two series of CGHs were designed (lens and gratings algorithm) to demonstrate independent real-time modulation of a two-pulse burst (1,1,0) with near Gaussian spots. A TTL trigger signal from the laser shutter was sent to a NI input/output controller (NI USB 6501), which signalled a Labview environment to synchronously start the CGH application to the two SLMs. The shutter was opened for 20 ms (50 pulses) and closed for 980 ms every second, which matched the CGH frequency, set to 1.0 Hz. Hence, we demonstrate the synchronous application of CGHs to both SLMs while independently modulating their relative spatial separations during a pulse burst. Figure 11a,b shows microscope images of the resulting surface micro-structuring when CGHs modulate the (1,1,0) orthogonal two-pulse burst with sinusoidal functions, which have a π/2 phase shift while the stage was translated at v = 30 μm/s. The surface LIPSS are clearly orthogonal, as expected, Figure 11b.

Dynamic Polarisation Modulation and Micro-Structuring
As nematic liquid crystal SLMs can be addressed in real-time, the application of CGHs can be synchronised to the scan motion control system. Two series of CGHs were designed (lens and gratings algorithm) to demonstrate independent real-time modulation of a twopulse burst (1,1,0) with near Gaussian spots. A TTL trigger signal from the laser shutter was sent to a NI input/output controller (NI USB 6501), which signalled a Labview environment to synchronously start the CGH application to the two SLMs. The shutter was opened for 20 ms (50 pulses) and closed for 980 ms every second, which matched the CGH frequency, set to 1.0 Hz. Hence, we demonstrate the synchronous application of CGHs to both SLMs while independently modulating their relative spatial separations during a pulse burst. Figure 11a,b shows microscope images of the resulting surface micro-structuring when CGHs modulate the (1,1,0) orthogonal two-pulse burst with sinusoidal functions, which have a π/2 phase shift while the stage was translated at v = 30 µm/s. The surface LIPSS are clearly orthogonal, as expected, Figure 11b.

Comparison of Burst Mode with Single Pulse Processing
Ablation characteristics of pulse burst processing compared to low-frequency (kHz-MHz) have been shown to vary significantly in the literature [15][16][17]. We confirm that this is indeed the case, firstly during blind drilling. Typical surface texture with 5 kHz exposure, Np = 150 pulses, Ep = 4 μJ/pulse (F0 = 0.4 Jcm −2 ) is shown in the SEM image, Figure  12a, with clear low-frequency LIPSS and little evidence of material melting, expected at fluences a few times ablation threshold. On the other hand, 3−pulse burst (1,1,1) ablation at the intraburst frequency νIB = 323 MHz with the same fluence/pulse (Np = 150 pulses) demonstrates a major difference in surface texture with evidence of melting and surface smoothing, Figure 12b. LIPSS (orthogonal to those with 5 kHz exposure) are still apparent in an outer ring but effectively disappear within a large region of the exposed area. This remarkable change in surface texturing is likely due to plasma absorption and heating by the |H〉 polarised pulses, coming within 3.1 ns delays, the time scale on which the plasma plume expands during ultrafast laser ablation [35][36][37], resulting in additional plasma surface heating [38,39].

Discussion
There is currently great interest in pulse burst processing of materials with ultrafast lasers [9]. Here, we converted a low repetition rate 5 kHz, 10 ps, ultrafast laser source to high-frequency pulse bursts at the fundamental frequency using a burst generator delay line consisting of a polarisation beam splitter with external and internal HWPs combined with 3 dielectric mirrors so that reflected and delayed pulses could be brought collinear. The relative setting of the two HWPs altered the energetic 10 ps laser pulses (f0 = 5 kHz) to a pulse burst with intra-burst frequency νIB = 323 MHz or 162 MHz. Up to 3 pulses with equal energy could be generated, allowing the study of the pulse burst laser-materials interactions and comparison with kHz processing. As the pulses reflected from the PBS were vertically |V〉 polarised while transmitted components were horizontally |H〉 polarised, this fact allowed independent polarisation modulation and diffraction of the pulse bursts with the aid of two phase-only SLMs whose directors were also orthogonally aligned.

Discussion
There is currently great interest in pulse burst processing of materials with ultrafast lasers [9]. Here, we converted a low repetition rate 5 kHz, 10 ps, ultrafast laser source to high-frequency pulse bursts at the fundamental frequency using a burst generator delay line consisting of a polarisation beam splitter with external and internal HWPs combined with 3 dielectric mirrors so that reflected and delayed pulses could be brought collinear. The relative setting of the two HWPs altered the energetic 10 ps laser pulses (f 0 = 5 kHz) to a pulse burst with intra-burst frequency ν IB = 323 MHz or 162 MHz. Up to 3 pulses with equal energy could be generated, allowing the study of the pulse burst laser-materials interactions and comparison with kHz processing. As the pulses reflected from the PBS were vertically |V polarised while transmitted components were horizontally |H polarised, this fact allowed independent polarisation modulation and diffraction of the pulse bursts with the aid of two phase-only SLMs whose directors were also orthogonally aligned.
By employing hybrid CGHs on each CGH, beam shaping of the elliptical laser mode was achieved, altering eccentricity from ε = 0.72 to ε = 0.88; hence, spot roundness was improved significantly, although at the cost of lower transmission efficiency. Consequently, pulse burst arrays of spots with orthogonal linear polarisations were modulated simultaneously, imprinting surface LIPSS with orthogonal directions. By introducing a nanostructured SWP into the optical line, CV beam arrays with ring modes and orthogonal polarisations (e.g., Radial/Azimuthal) simultaneously imprinted more complex low-frequency LIPSS, also confirming the high purity of these interesting polarisation states, regarded as classically non-separable in spatial and polarisation modes [41,42]. As nematic liquid crystal SLMs can be addressed in real-time, synchronisation of CGH application to the SLMs was demonstrated at 1 Hz in a Labview environment, allowing independent dynamic diffraction of the (1,1,0) orthogonally polarised pulse pairs.
A comparison of 5 kHz multi-pulse (1,0,0) and 3-pulse burst processing (1,1,1) at 323 MHz intraburst frequency on fixed spots demonstrated a remarkable change in surface texturing of stainless steel during blind drilling. While low-frequency LIPSS are characteristic of multi-pulse ultrafast kHz ablation at fluences just above the ablation threshold, pulse bursts with nanosecond delays introduced significant surface melting due to plasma absorption, which, using the pump-probe technique, can be maximum after a few ns delay [34]. With pulse bursts on stainless steel, ablation efficiency decreased with increasing pulse burst number; 2-pulse (1,1,0) ablation efficiency η = 46% while 3-pulse, (1,1,1) η = 39% of that measured at the fundamental 5 kHz frequency, respectively. This is strong evidence of plasma absorption due to the following pulses in a burst which arrive within 3.1 ns of each other. The residual surface texture with pulse burst processing appeared much smoother than 5 kHz, so burst processing could be regarded as a valuable finishing process to improve surface finish after a lower repetition rate kHz micro-machining of metals.
With an optical setup involving two SLMs, a 4f optical system, and a delay line burst generator, it is worth reporting the overall efficiency of laser light used at 5 kHz compared to a pulse burst with and without beam shaping. With the (1,0,0), 5 kHz output and no beam shaping, the overall transmission efficiency from the PBS to the Galvo input aperture was measured to be η = 63%, while with the hybrid CGHs applied to both SLMs, this dropped to η = 50%. With the (1,1,0) two-pulse burst without beam shaping, η = 61%, while with hybrid CGHs applied, efficiency dropped to η = 43%. These results compare favourably with those measured by Hasegawa and Hyasaki [22] with two SLMs who quoted η~8% with an fs laser source and 2 SLMs. However, they also demonstrated up to 20 diffracted spots with orthogonal linear polarisations.
Very recently, a delay line consisting of a non-polarising beam splitter (40:60, 20:80) and three mirrors was used to create pulse bursts, all with the same polarisation at 1 GHz intra-burst frequency for fs laser processing of steel and copper [43]. Bursts with up to 5 pulses and decaying amplitudes could be generated, also showing a reduction in ablation rate on steel with burst mode. Domke et al. [16] used a linear delay line with a PBS, 2 mirrors, and quarter-wave plates (QWPs) to cleverly double their burst frequency from a femtosecond laser at 77 MHz to 154 MHz. Although not stated, a pulse burst with up to 28 pulses and decreasing amplitude had orthogonal polarisations from pulse to pulse. Fraggelakis et al. [44] demonstrated large area, 2D nano-structured LIPSS on stainless steel with fs double pulse cross polarised exposure using interpulse delays in the range 0.1 ps ≤ ∆t ≤ 50 ps.
The work presented here also demonstrates dynamic polarisation and spatial modulation of orthogonally polarised pulse bursts with 2 reflective SLMs, increasing the flexibility of laser processing of materials. If the output laser beam were truly Gaussian, the losses incurred by beam shaping would be reduced, hence, increasing light use efficiency to η~63%, although this will be CGH dependent. The technique developed in this work would allow simultaneous processing at kHz repetition rates combined with pulse burst processing, dependent only on CGH application, and might well be applied in the spatial control of surface materials wettability, surface bio-compatibility, or encoding complex surface structures for security marking of valuable components. In addition, it may be possible to directly write high-quality reflective optical gratings or plasmonic sensors into metallic surfaces. By optimising the synchronisation scheme, orthogonal burst polarisation modulation at frequencies >10 Hz should be possible, limited only by the SLM bandwidths <30 Hz [45]. With current cooled SLMs able to handle high average powers >100 W [26], then very high throughput large area pulse burst processing with orthogonal polarisations becomes possible.

Conclusions
An optical pulse burst optical system operating at intra pulse burst frequency of 323 MHz from a 5 kHz fundamental frequency has been developed to study the differences in ultrafast laser-material processing of stainless steel. At 5 kHz, micro-structuring produces clear low-frequency LIPSS both for fixed spots and during scanning when the beam overlap is approximately 33%. On metals, these low-frequency LIPSS, with period Λ~λ (laser wavelength), appear near right angles to the local E vector. On the other hand, pulse bursts with 3.1 ns between pulses lead to plasma absorption, which re-heats the surface, leading to melting, so that LIPSS disappear. As a two-pulse (1,1,0) burst has orthogonal polarisations, each pulse can be modulated independently by an SLM with directors oriented appropriately. This has allowed independent, real-time, two-beam surface processing with orthogonal and complex LIPSS. The current Galvo system with a 100 mm focal length f-theta lens has a 60 × 60 mm 2 flat scan field, while the translation stages have a 100 mm range. By combining these motions, the processing area can be increased to approximately 130 × 130 mm 2 , industrially relevant. The novel optical system developed expands the current window utilising techniques available for high throughput, dynamic ultrafast laser processing, particularly when combined with kHz, high energy, high average power systems currently under development [11].

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.