Reducing environmental risks in laser cutting: A study of low-pressure gas dynamics

Lasers have emerged as a promising technology for nuclear decommissioning, offering precise and efficient material removal. However, the generation of harmful fumes during the laser-cutting process poses a significant challenge that needs to be addressed.^1,2

These fumes can contain radioactive materials, which pose a risk to the surrounding environment if not properly controlled. Therefore, strict containment and filtration measures must be implemented to prevent their release. However, this limits the maximum processing rate. In addition, the safe disposal of contaminated filter medium creates a secondary waste stream requiring further environmental consideration. Reducing the total volume and increasing the average size of aerosolized by-products (fume) generated during laser cutting, making it easier to filter, will not only aid in increasing the processing rate but will reduce the environmental risk posed by laser cutting in hazardous environments.

Recent studies have focused on the need to cut thick materials (15–100 mm) for the application of nuclear decommissioning.³ While this is an appropriate area of study there are still conditions where thinner material cutting is required, these include many meters for pipework used to transport low level liquors and other waste products around nuclear facilities. However, this topic area has not been as heavily researched.

Gas dynamic interactions in laser cutting are known to play a governing role in melt breakdown and ejection and have been a subject of interest for many years. Pressure and velocity gradients resulting from the gas-melt interaction provide the driving forces, which eject the molten material from the cut.^4,5 As with all melt atomization processes, the exact breakdown mechanism is governed by the respective Weber numbers of the liquid melt and the assist gas.⁶ However, precise determination of these values in a process such as laser cutting is difficult, meaning experimental assessments are necessary for thorough and accurate analysis.

Conventional laser fusion cutting uses high laser powers and assists gas pressures to produce cuts that are free from dross and burrs. However, studies by Pilot et al.⁷ and Lopez et al.³ demonstrated that cutting at high powers and pressures produces larger volumes of fume compared to that of lower powers and pressures. Lopez et al. concluded that lower supply pressures not only reduced the amount of material ejected from the cut but also increased the average size of measured particles.⁸ 

More recently, Lavin et al. demonstrated that even at low pressures, gas dynamic interactions on the surface and inside the workpiece still play a governing role in the size and volume of aerosolized by-products.⁹ The study compared the results of Schlieren imaging in an idealized environment with those from laser-cutting experiments to demonstrate how the complex gas dynamic interactions inside the cut kerf could be used to help control laser fume formation.

Borkmann et al. studied in-kerf dynamics highlighting how variations in the boundary layer flow state impact on shear forces and heat transfer between the gas jet and the molten cut front.^10,11 They estimated an in-kerf heat transfer loss of 24–110 W based on a 1.6 MPa supply pressure and 10 mm thick material. This is considerably higher than the 2.5 W calculated by Vicanek et al., although this was on 3 mm material with a slower airflow velocity estimated at 100 mm/s.⁵ Borkmann et al. concluded that in order to improve laser-cutting models, in-kerf gas dynamics should not be oversimplified.¹⁰ 

Stoyanov et al.,¹² building on studies by Hansmann et al.,¹³ and Miyamoto and Maruo,¹⁴ conducted a comprehensive study on burr formation and melt ejection regimes during high-pressure laser cutting of thick materials. The study employed two high-speed cameras to capture melt ejection in two planes during the cutting process of 6 mm thick stainless steel samples. This was achieved using a 10 kW disk laser and a nitrogen supply pressure of 2.0 MPa. At “optimal,” burr-free, cutting conditions, the study highlighted the formation of a liquid sheet along the bottom edge of the workpiece, which would ligament and breakdown with a periodicity of 100–500 μs⁻¹. The formation of this sheet is possible when the flow has a low Reynolds number and a relatively high surface velocity. This supports the result that low-pressure and low-power cutting would reduce the total volume of ejected material as under these conditions the flow Reynolds number will be relatively high while melt flow tangential velocity will be relatively low due to turbulence in the gas jet caused by boundary layer separation in the lower parts of the cut.^15,16

The present study experimentally assesses the melt ejection regimes in low-pressure, low-power laser cutting. The aim is to identify the changes in melt breakdown regimes as a result of controlling dynamic interactions to provide a better understanding of fume formation in low-pressure laser cutting. The impact of this improved understanding will enable the development of laser-cutting control methods for use in hazardous environments.

A 300 W, JK, Nd:Yb fiber laser was used to cut 1.5 mm thick, 304 stainless steel samples. To keep cutting power as low as possible, the laser was set to only deliver 70% of its total power, providing a continuous wave output of 210 W. The processing head used a Galilean system containing two 76 mm focal length lenses providing a 1/e² diffraction limited spot size of ∼19 μm. A standoff distance of 0.5 mm was selected as previous studies have demonstrated that gas dynamic features such as stagnation bubbles will not be present under this condition.¹⁷ The focal position of the laser was located on the top surface of the workpiece.

The camera’s field of view was configured in the ZX plane, with the X-axis representing the direction of the cut, see Fig. 1. Both the camera and cutting head remained stationary while the workpiece was traversed in the X+ direction. As this study is more concerned with the fraction of the melt that becomes aerosolized, higher magnifications are desired compared to that of previous studies. Table I provides details of the imaging setups used. The spatial resolutions detailed are an order of magnitude smaller than presented in previous studies.^12,14

The assist gas was supplied via a 1.5 mm conical nozzle at pressures ranging from 0.3 to 0.6 MPa. Supersonic nozzles were not used as the benefits they provide were not factors being assessed in this study. This study assessed both nitrogen and air. Previous work¹⁸ has indicated that nitrogen produces less aerosolized by-products than air. However, air is often the assist gas of choice in the nuclear decommissioning industry as it possesses relatively few safety concerns when compared with inert gasses. As a result, any proposed modifications to the utilization of air in this context would require a compelling and well-substantiated rationale.

A cut velocity of 8 mm/s was maintained across all experiments. It is acknowledged that this is lower than conventionally used. However, as the laser power and gas pressures used were considerably lower than standard, a reduction in cut velocity was required.

The cooling effect of the gas jet during laser cutting is often deemed insignificant.¹⁹ However, preliminary studies have demonstrated a significant change in the heat affected zone when cutting at lower pressures or with different nozzle geometries.⁹ As the temperature of the workpiece in and around the cut region will have a significant effect on the adhesion energy and, therefore, the breakdown of the melt, it was deemed necessary to measure the plate temperature to better understand the changes taking place. To achieve this, three 0.5 mm diameter k-type thermocouple probes were mounted to the underside of the workpiece at intervals of 1 mm, in the ZY plane. The thermocouples were seated in holes 0.6 mm wide that terminated in the center of the plate; see Fig. 2 for details.

The thermocouples were read using a thermocouple amplifier capable of sampling at 40 kHz which output, via BNC, to a Picoscope, 2406B oscilloscope. Supply pressures were assessed and monitored using a 0.1–1.1 MPa, 0–5 V, pressure transducer located close to the nozzle exit.

In Fig. 3, the temperature at which the gas jet could maintain the workpiece when the laser was not on is depicted for different supply pressures. As expected, an increase in supply pressure leads to an increase in forced convection and, therefore, workpiece cooling.²⁰ Nitrogen was seen to cool the plate more than air. This is likely because nitrogen was supplied from a compressed cylinder meaning it will be chilled upon entering the system as it expands leaving the cylinder.

At low pressures, the amount of forced convection reduces with distance away from the jet center. However, as pressure is increased above 0.5 MPa in this experiment, the temperature of the workpiece at 3 mm is observed to reduce below that measured at 2 mm. In the process of impingement, the gas jet experiences a rapid deceleration, resulting in a surge of pressure that propels the gas jet at a high velocity across the surface of the workpiece. This high-velocity zone is referred to as the wall region.²¹ The high-velocity gas in the wall region increases the amount of localized forced convection. While its influence may not be notable under the present testing conditions, it perhaps should be considered when operating at elevated pressures.

The typical temperature profile for each of the three probes during laser cutting is shown in Fig. 4. The peak registered at Probe A, located 1 mm away from the cut relates closely to when the laser is in line with the three probes. Due to the conductive properties of the workpiece, the peak at 3 mm is not reached until after the laser has finished cutting. The figure displays over a 410 °C thermal gradient between the peaks of the 1 and 2 mm probes.

As expected, the cuts using nitrogen were at lower temperature compared to air, see Fig. 5. This is due to the additional energy provided by oxygen in the air. At 1 mm from the cut center, the plate temperature for the nitrogen cuts was 115–200 °C cooler than for the air cuts. At 2 mm from the cut center, the nitrogen cuts were 35–43 °C cooler than the air cuts. At 3 mm from the cut center, nitrogen cuts were consistently 28 °C cooler than the results from cutting with air.

Figure 5 shows that as supply pressure is increased the peak temperature reached at each probe decreases. This reduction was expected for the nitrogen data. A combination of an increase in forced convection across the top surface and an increase in pressure gradient across the molten cut front is believed to be the cause of the reduction in temperature with increasing supply pressure. On its own, it is inappropriate to draw further conclusions from these data; however, the potential reasoning for this change is discussed in more detail later following estimations from flow modeling.

Previous studies have identified that oxygen concentration is a key factor affecting the cutting process.²² For this reason, it was thought that increasing air pressure may lead to an increased plate temperature due to the increased flow rate of oxygen. However, the results presented in Fig. 5 demonstrate that, as with nitrogen cutting, as supply pressure is increased, the temperature of the workpiece decreases.

Hence, the increased level of cooling and improved melt ejection efficiency, provided by the increase in supply pressure, outweighs the possible effects of increased oxygen presence. It must be noted that the plate cut here is relatively thin, meaning oxygen consumption in the cut region will be minimal. When cutting thicker materials, where a higher consumption of oxygen occurs, the temperature profiles may differ.

Figures 7 and 8 show high-resolution, high-magnification images of 0.6 and 0.3 MPa air cutting, respectively. The two figures demonstrate the significant changes in breakdown regimes between the higher and lower pressure cuts. Figure 7 shows “explosive” breakdowns, which produce numerous particles with diameters <10 μm. The velocity of these breakdowns is such that they only exist for one frame when processing at 6400 fps. Furthermore, based on the motion blur of the particles, a frame rate of over 200 000 fps would be required to capture this process over multiple frames, something that is not practical at this magnification.

In comparison, the process of breakdown in Fig. 8 is slower, taking orders of magnitude more time than those shown in Fig. 7. In addition, the general size of breakdown constituents is larger too (≫10 μm).

Analysis of the <10 μm particles in Fig. 7 revealed they have a velocity range of 40–80 m/s. While such particles still exist when cutting at low pressures, one can be seen in Fig. 8 at +156 μs, they occur far less frequently. These velocities are an order of magnitude higher than those presented in previous studies, which only assessed ≫10 μm sized particles.^12,26 The violent breakdown of the melt during the higher pressure cutting occurred at a frequency of 600–1000 μs⁻¹. This frequency corresponds to the frequency of the consecutive pattern noted on the top surface of the cuts. This is a lower frequency than that described in previous high-pressure studies.¹² 

Figures 9 and 10 show lower magnification, higher frame rate images of air cutting at 0.6 and 0.3 MPa, respectively. The images in Fig. 9 indicate that the explosive breakdowns described earlier take place over approximately 200 μs. This breakdown process starts by forming a liquid sheet that flows along the underside of the workpiece. This sheet then thins into “ligaments” and breaks down (for the description of melt ligamentation, see Ref. 6) The duration of this process matches closely with that described by Stoyanov et al.,¹² during high-pressure, high-power cutting of thick material.

While the breakdown process recorded here was similar to that described by Stoyanov et al.,¹² the frequency of breakdown under the conditions assessed here was much lower. Stoyanov et al.¹² described a ligamentation frequency of 100–500 μs⁻¹ compared to the 600–1000 μs⁻¹ recorded here. This change in frequency is likely due to the reduced cutting speeds and lower material thickness, leading to a reduction in the volume of material melted per unit time. This implies that, beyond a specific pressure threshold, the mechanism of the breakdown process remains consistent across various cutting circumstances, only differing in terms of frequency at which it takes place.

The melt breakdown and ejection detailed in Fig. 10 show that the melt empties in strands. Stoyanov et al.¹² also noted the formation of temporary drains leading to the melt emptying in strands. During their study, this type of breakdown mechanizing was associated with the formation of large burrs. A similar result is presented here, with significantly larger burrs being present when this type of breakdown mechanism was taking place.

The strand breakdown mechanism, during the study of Stoyanov et al.,¹² occurred at lower cutting speeds and high powers, indicating that it may be a result of excessive energy leading to the formation of a large melt pool. A similar phenomenon could be happening here. As the melt is not being ejected as efficiently, due to the low pressure assists gas, it is allowed to build up before being ejected as a strand. Something not noted during previous studies is that this strand formation was also observed during the initial penetration of the laser. This indicates that the laser may have been on the limit of its ability to cut at this low pressure.

Figures 11 and 12 show images of 0.6 and 0.3 MPa nitrogen cutting, respectively. Comparison of Figs. 11 and 9 demonstrate how much more “stable” the cutting process for 0.6 MPa nitrogen is compared to 0.6 MPa air cutting. The breakdown mechanisms observed during 0.6 MPa nitrogen cutting align more closely with those of the 0.3 MPa air cutting.

It is important to note that the nitrogen and the air data sets have been contrast adjusted by different amounts in order to make them clear. However, viewing the raw images highlights a significant difference in melt brightness. The much lower brightness recorded for the nitrogen data suggests a cooler melt pool. This agrees with the results of the temperature measurements, which show that the temperature 1 mm away from the cut center was 115–200 °C lower for the nitrogen cuts. In addition, the temperature of the melt during air cutting was estimated to be 470 °C hotter than the melt during nitrogen cutting.

This does raise concerns regarding the absence of smaller particles in the nitrogen data, suggesting that their non-detection may be attributed to imaging limitations rather than their actual absence. However, even though the high shutter speed used may mean it was not possible to capture smaller particles due to their reduced temperature, it is evident from the images that nitrogen cutting results in a different breakdown regime compared to air when using the same cutting conditions.

The results presented for 0.3 MPa cutting with nitrogen show the formation of a large dross ball just downstream of the laser center. Figure 11 highlights that this large ball of molten material remains stable for much longer than any of the results presented for the other conditions. While the ball of molten material was present, no ejection or breakdown of particles was noted. It appeared as if the dross ball was “catching” molten material being ejected from the cut preventing it from breaking down into particles. This caused the ball to grow until it eventually fell off. During its existence, the outer surface of the dross ball could be seen to chill and “freeze” over, much like watching the top surface of a molten material cool during casting. However, cyclically the front of this chilling layer would recede, making it possible to see the swirling molten material beneath. The receding of the chill front is hypothesized to take place when new, hot, molten material is captured by the dross ball. The arrows marked in Fig. 12 show the chill front closing over as the particle appears to start solidifying before opening again as more energy is added.

This phenomenon is believed to be due to the uniqueness of cutting at such low pressures and is not something that has been previously presented. Also, note the melt brightness for 0.3 MPa nitrogen cutting was higher than for the 0.6 MPa nitrogen. This indicates that the temperature of the melt and, therefore, the workpiece increases when cutting at lower pressures, supporting the results presented in Fig. 5.

Figure 13 demonstrates the change in dynamic viscosity between the gas compositions does significantly affect the local Reynolds number of the flow. The change in local Reynolds number due to changes in dynamic viscosity is in the same order of magnitude as the change in shear stress due to a 0.1 MPa change in supply pressure. Indicated in Fig. 13 is the critical Reynolds number for the transition to a turbulent boundary layer. The upper horizontal line represents the largest critical value expected for flow transition over a flat plate. The lower line represents the critical value expected for transition over a flat plate with an unsharp edge. Actual laser cutting is expected to be more closely represented by a plate with an unsharp edge meaning the transition is expected to occur somewhere close to this line. This indicates that the boundary layer could be turbulent after as little as 0.25 mm. It is likely a turbulent boundary layer will develop at some point along the cut depth, but with the simplistic approach applied here it is difficult to predict when. For this reason, both lamina and turbulent conditions have been assessed.

Figure 14 details the change in boundary layer thickness along the cut depth based on the Reynolds numbers calculated earlier. As it is unclear exactly when the boundary layer will transition from laminar to turbulent, both laminar and turbulent boundary layer thickness have been calculated along the entire depth of the cut.

Owing to the thickness of the workpiece, even if the boundary layer was turbulent along the entire depth of the cut, the thickness of the boundary layer presented here is relatively thin. This suggests boundary layer separation may not occur during cutting of such thin material. If it were to occur, it would take place near the bottom of the cut where the boundary layer is thickest and where there is likely to be steep pressure gradients not accounted for during this simplified analysis.

Figure 15 shows a significant increase in the wall shear stress for turbulent boundary layers compared to laminar. Furthermore, Fig. 15 demonstrates that shear stress increases with distance along the cut depth. This indicates that the increase in core flow velocity along the cut depth has a greater impact on the wall stress than the increase in flow Reynolds number and the decrease in flow density.

As presented by Yilbas and Aleem, changing the gas supply pressure will influence the velocity profile inside the cut region which will, in turn, affect the melt breakdown.²⁹ However, the change in the size of the turbulent region in the lower section of the cut as a result of changing surface stagnation pressure is also believed to play a significant role in the change in melt breakdown mechanisms when cutting at low pressures. A previous study has demonstrated that at low pressures the size of the turbulent wake downstream of boundary layer separation region can be controlled and changed with relatively small changes in surface stagnation conditions.⁹ Therefore, it would be pertinent to consider it as a key influencing factor under the conditions assessed during this study.

It has been observed at higher gas pressures there is an increase in the forced convection on the workpiece cooling, leading to a greater cooling effect of the gas jet. However, it is also possible that the decrease in measured plate temperature may be due to improved material ejection efficiency evidenced by the change in breakdown regimes when cutting at different pressures. This enhancement in material ejection efficiency is also validated by the increase in shear stress applied to the melt with increasing supply pressure, as detailed in Fig. 15. More investigation is still needed to confirm the exact reasoning for the changes in temperature.

The change in measured temperature between different supply pressures of the same gas composition, Fig. 5, was relatively minimal, suggesting it is unlikely that the resultant change in melt viscosity is the reason for the observed changes in melt breakdown conditions. The shear stress experienced along the cut wall reduces with a reduction in supply pressure, Fig. 15. This reduction in shear stress creates melt breakdown regimes that are more stable and produce larger slower moving particles. Hence, the observed changes in melt breakdown, for a given gas composition, are believed to be the result of the changing shear stress with changing pressure supply pressure and not a result of changes in temperature. This result links well with the results of Arntz et al.,²⁶ who demonstrate a reduction in melt flow velocity with reducing gas pressure.

The difference in dynamic viscosity of the gas jet as a result of differing average boundary layer temperatures did not significantly affect the estimated shear stress applied to the melt. However, based on the equation provided by Kim,²⁴ the change in viscosity of stainless steel as a result of the increased melt temperature is far more significant. Melt viscosities of 2.14 and 1.65 mPa s for the nitrogen and air cuts, respectively, have been calculated. Furthermore, this does not consider any additional reductions in melt viscosity that may result from the formation of iron oxide during air cutting.¹⁴ It is, therefore, believed that the changes in melt breakdown mechanisms observed between the air and nitrogen cuts are the result of an increased melt temperature leading to a reduced melt viscosity.

The change in melt viscosity may also be the reason for the lack of smaller particles noted during the imaging of nitrogen cuts. Melt atomization models, provided by Yule and Dunkley,⁶ highlight that melt viscosity influences melt Webber and Reynolds numbers, which in turn affects the atomization process. An increase in melt viscosity would lead to an increase in average particle diameter and less secondary breakdown.

In the nuclear decommissioning industry, the excitation of surface contaminants may be an important factor to consider.³⁰ Temperature gradients in the workpiece and alterations in workpiece temperature resulting from changes in supply pressure, as noted in Figs. 3 and 5, could impact the volatility and nature of these surface contaminants.³¹ The impact of this should be considered when designing laser-cutting techniques for application in hazardous environments.

The findings from this study offer valuable insights into the reasons behind the observed improvement in laser fume when cutting at reduced gas pressure and using nitrogen. In both nitrogen and air cases, a decrease in supply pressure resulted in breakdown regimes characterized by larger, slower moving particles. Furthermore, the nitrogen cuts exhibited a more stable melt ejection process with fewer occurrences of the “explosive” breakdowns that lead to the formation of particles smaller than 10 μm.

The modeling approach taken here is limited by the fact it assumes a linear transition in velocity and density from the top of the cut to the bottom. In reality, both the velocity and density profiles will vary non-linearly and will contain a steep change in values at the point of boundary layer separation. In addition, the thermal effects of the molten cut front on the properties of the core flow have not been considered.

This study is also limited to one material thickness at low laser powers. It is necessary to apply the techniques used here to the larger material thickness that have been demonstrated in other studies as important to the nuclear industry. This would require higher laser powers but should still focus on keeping gas pressure and laser powers to a minimum.

An experimental investigation into melt breakdown regimes during low-power, low-pressure laser cutting has been presented. High-magnification, high-resolution images have shown that <10 μm sized particles are ejected with velocities in the range of 40–80 m/s.

The measurements of workpiece temperatures emphasize the significance of forced convection on the workpiece surface, challenging the notion proposed by earlier studies that it can be disregarded when evaluating the thermal interaction of a cutting system. Furthermore, the significance of its effect becomes more prevalent with increases in supply pressure.

The breakdown regimes present in low-pressure, low-power cutting are similar to those reported in high-pressure, high-power systems. However, the frequency of breakdown is reduced when cutting at low pressures and there are no cutting conditions where burrs do not form. In addition, the duration for which the features remain stable is significantly increased at lower pressures.

Gas composition has been shown to influence both the workpiece temperature and the breakdown regime taking place. A transition from an explosive, high-velocity breakdown to a more stable breakdown, with changing gas composition, has been presented.

No <10 μm particles were identified for the lowest temperature cut. This may be due to the fact most of the material was collected in the large melt ball that formed on the underside of the plate. However, it is possible that the system was unable to identify these particles due to their low temperature and, therefore, low level of illumination.

Modeling of flow interactions in the kerf region has indicated that the change in melt breakdown regimes for gasses of the same composition is likely the result of changing wall shear stress with changing pressure gradients. However, for gasses of different compositions, where one is reactive and the other is not, the change in melt temperature is believed to be the cause of changes in observed breakdown mechanisms, not changes in dynamic viscosity of the gas jet.

Efficient laser fume control is a critical aspect of enhancing safety and environmental sustainability in nuclear decommissioning. This study has demonstrated that the size, velocity, and temperature of ejected material can be controlled through changes in gas composition and supply pressure. To reduce the environmental risk of laser cutting in nuclear decommissioning via increasing average particle size and reducing the total volume of aerosolized by-products, this study recommends the use of an inert supply gas at pressures as low as the process will allow. The results of the study enable the development of advanced laser-cutting strategies that effectively address fume generation and promote the overall success of nuclear decommissioning efforts.

k

Thermal diffusivity

c

Specific heat capacity

ρ

Density

ω

Interaction area

p

Interaction permitter

x

Distance in x

g

Gas constant (1.4 air, 1.4 nit)

Re_z

Reynolds number at z

u

Local velocity

η

Dynamic viscosity

c_f^’

Skin friction

H

Surface conductance

K

Thermal conductivity

Ma

Mach number

P

Pressure (×10⁶)

T

Temperature

z

Distance in z

R

Specific heat ratio (296.8 air, 287 nit)

δ

Boundary layer thickness

U_∞

Core flow velocity

τ

Local shear stress

*Units for all constants relating to Forbes’s method of liner heat flow are in CGS form. Units for all constants relating to gas jet follow are in SI form.

This work has been funded by the UK Nuclear Decommissioning Authority (NDA) project “Bad Laser Cutting to Get Good Laser Fume” with supervision provided by the National Nuclear Laboratory (NNL).

The authors have no conflicts to disclose.

Jacob J. Lavin: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Jay J. Robus: Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Toby Williams: Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Edward J. Long: Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Visualization (equal); Writing – review & editing (equal). John R. Tyrer: Funding acquisition (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Julian T. Spencer: Supervision (equal); Writing – review & editing (equal). Jonathan M. Dodds: Supervision (equal); Writing – review & editing (equal). Lewis Jones: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).