Aerosol Characterization and Particle Scrubbing Efficiency of Underwater Operations during Laser Cutting of Steel Components for Dismantling of Nuclear Facilities

The goal of this article is to provide results on aerosol particles emissions of a laser Nd:YAG cutting technique used for the decommissioning of nuclear facilities. In particular, the study aims at characterizing the aerosol emitted during the cutting of steel specimens of different thicknesses and to study particulate emissions for cuts in air and under water. To do so, we calculate the emitted aerosol mass per unit area of cut. Overall, it was found that the mass of aerosol per unit area of cut by laser cutting decreases when the laser power and cutting speed increase. We also examine the performance of the height of the water column above the cut on the particle collection efficiency. We found that the driving phenomenon for particle collection is the scrubbing of particles by bubbles present in the water column. When cuts are realized under water, the production of aerosol particles mass per unit area of cut is reduced by a factor of 10 and limited below 70 g m.


INTRODUCTION
When a nuclear facility reaches the end of exploitation, it is dismantled after a more or less long period of waiting for the level of radiations to decrease.The decommissioning consists in cutting the active parts of the installation piece by piece, in order to remove the wastes produced in the corresponding spinner and restore the infrastructure and the soils to a level of radioactivity suitable for a possible reinstatement in the public domain (OECD/NEA, 2014).Assessing the risk of loss of containment of radionuclides during the decommissioning of a nuclear facility requires therefore being able to predict how much radioactive and non-radioactive aerosol particles will be collected on a High Efficiency Particulate Air Filter (HEPA) which may lead to its clogging and rupture.For this, it is important to characterize the aerosol particles -emitted upstream from the HEPA filters -consistent with the type of operation that needs to be performed and the proposed tool (Pilot et al., 2008).Many studies are designed to characterize the kerf geometry from the stainless steel cutting by high power lasers (Choubey et al., 2015;Ghany and Newishy, 2005).
However few works concern the study and characterization of particles emitted in the aerosol form when using such cuttings processes (Pilot, 2008).Indeed, this issue is of prime importance when conducting the decommissioning of nuclear facilities, especially during cutting operations of potentially contaminated equipment (Cesari et al, 2008).In this context, the first part of this work is to provide results on aerosol emissions of a laser cutting technique used for the decommissioning of nuclear installations.In particular, the study aims at characterizing the aerosol particles emitted during the cutting of stainless steel specimens of different thicknesses when cuts are realized in air and under water.Therefore, we calculate the mass of loss material of the specimen as well as the emitted aerosol mass per unit area of cut.Particle characterization in terms of particle size distribution and morphology are also assessed.In a second part, we have compared the scrubbing efficiency of such underwater operations representative of an industrial process to a simple model introduced by Fuchs (1964) and used later on by Pich and Schütz (1991) for the scrubbing of particles by gas bubbles in a fluid.Indeed, bubbles columns capacity to absorb airborne radioactive materials and to operate during an extend period of time at a constant pressure drop makes the process a good candidate for the mitigation of radiological dispersal during dismantling operations.Although the main parameters (bubble size and rising velocity) used in the model of Pich and Schütz (1991) are estimate in the present work using correlations, we will see that data obtained on the collection efficiency of aerosol particles for different heights of water are close to theoretical results.This good agreement between the model and the scrubbing efficiency of bubble columns has previously been observed (Cadavid-Rodriguez et al., 2014) yet at a laboratory scale and for ultrafine particles (Koch and Weber, 2012).

The Laser Cutting Facility
The study was performed at the DELIA facility (French acronym for: in air and underwater laser cutting cell, Chagnot et al., 2010).The facility, which occupies three floors, includes a laser source, a cutting cell which can operate either in air or submerged under 5 meters of water where are placed the test pieces to be cut.At the top floor, an aerosol sampling line is installed allowing the measurement of the aerosol particles concentration and the characterization of their size distribution.The cutting cell is a stainless steel tank with a diameter of 1.2 meter and a length of 1.25 meter with a usable volume of 1.4 m 3 .It is equipped with two chimneys of 5 m in height, the first with an inner diameter of 110 mm, which allows the air inlet, the second with an inner diameter of 300 mm for the exhaust of the aerosol particles and the gas.A HEPA filter is placed at the inlet in order to measure only the aerosol produced by the cut inside the cell.During the test campaign, two heights of water above the cuttable plates of 0.5 m and 4.2 m were studied.The cut pieces are 316 L stainless steel plates (X2CrNiMo 17-12-2 1.4404) of thickness ranging from 10 mm to 60 mm.We show in Fig. 1 an example of a 20 mm thick specimen with various cuts, i.e., cutting speeds and laser powers can be different from one cut to another.The left image corresponds to the front side of the specimen where the laser beam hits the specimen; and the right image corresponds to the rear side of the specimen where one can see the melted dross attached to the plate.

Cutting Tool
The laser source is a continuous wave Nd:YAG laser with a rated power of 8 kW at a wavelength of 1.06 µm from the TRUMPF Company.A flushing air jet surrounding the laser beam evacuates the molten metal sprayed by the laser during the cutting operation.A nozzle head projects a sheath of additional water, to create a dry area around the laser beam when the sample is immersed in order to maintain the fluence of the laser on the part to be cut.The main features of this tool are the following: ■ Power on the plate to be cut: from 3 to 8 kW, ■ Wavelength: 1.06 µm, ■ Length of optical fibers: 120 m, ■ Nozzle diameter: 3 mm and 6 mm for cuts in air, 3 mm for cuts under water, ■ Flushing airflow: 400 L min -1 , ■ Flow of water curtain: 55 ± 5 L min -1 , ■ Position: horizontal beam at right angle to the plate to be cut, ■ Distance between the nozzle and the plate to be cut: between 10 mm and 30 mm for cuts in air, 10 mm under water.
During the tests, laser powers of 3 kW, 6 kW and 8 kW were used.The cutting speed is fixed at 50 per cent of the speed limit to guarantee a total cut of the plate.The speed limit is defined as a function of the thickness of the plate and the laser power, and corresponds to the maximum speed for cutting the workpiece from one edge through its entire thickness.The cutting performances of the laser have been previously studied and, for each plate thickness and laser power intensity, the maximum speed limit has been measured (Chagnot et al., 2010).

The Sampling Line
A schematic diagram of the sampling line for aerosols characterization is presented in Fig. 2. The line is composed of a stainless steel cylindrical pipe with an inner diameter of 110 mm.The duct is connected to the cutting cell at the top of its 5 m height chimney of evacuation with an inner diameter of 300 mm.The air ventilation rate in the facility is maintained at 100 m 3 h -1 .The measurement of the concentration of the aerosol particles in the line is performed with two isokinetic sampling probes.The first probe has an inlet diameter of 12 mm and is connected to a filter holder of 47 mm in diameter that supports a HEPA glass fiber filter produced by Camfil.The second probe has an inlet diameter of 26 mm and is connected to a filter holder of 130 mm in diameter supporting a Bernard Dumas filter  with the same filtering efficiency.The choice for using two different filters originates in the fact that when cuttings are realized underwater, the mass concentration of particles decreases dramatically.Thus, with the smaller filter (47 mm in diameter) which has a sampling flow rate of 15 L min -1 , the amount of particles collected was not sufficient to obtain a noticeable mass.We thus added a bigger filter (130 mm in diameter) with a sampling flow rate of 80 L min -1 in order to collect enough particles during the cuttings realized underwater.The particles size distribution is measured using a DMS 500 (© Cambustion) which provides the number particles size distribution based on their electrical mobility.The DMS 500 is connected to a third isokinetic probe with an inlet diameter of 16 mm.
The aerosol penetrating fraction in the exhaust chimney of DELIA and in the sampling line all the way down to the sampling points was estimated using the software Aerocalc (Aerosol Calculator, Baron and Willeke, 2001) for particles with aerodynamic diameters of 0.1 micron, 1 micron and 10 microns.The deposition mechanisms considered are: inertial effect related to turbulent flow regime (Liu and Argawal, 1974), brownian diffusion in the turbulent flow regime (Friedlander, 1977), gravity in the turbulent flow regime (Schwendiman et al., 1975), inertial effects and gravity in a bend (Pui et al., 1987), inertial effect in an abrupt contraction of the flow (Muyshondt et al., 1996).Calculations show that for aerodynamic diameters up to 1 micron, the penetrating fraction is equal to 1.00 that is to say that the particles deposition is negligible up to the aerosol sampling point.Conversely, for particles with an aerodynamic diameter of 10 µm, the penetrating fraction is equal to 0.91 which corresponds to 9% of particles loss, mainly by impaction and sedimentation in the bends of the horizontal sections.
A test run fulfills these following steps.All sampling filters (Ø 47 mm and Ø 130 mm) are weighed on the same balance before the experiment to obtain the initial mass of the filters.The plate to be cut is weighed before its introduction inside the cell.Sampling filters are set up and volumetric flow meters are noted.The laser is operated simultaneously with the feed motor and the aerosol sampling.Cutting starts on one edge of the specimen and is monitored during the time of the test.The conditions of pressure and temperature of the samples are recorded.After stopping the cut, aerosol sampling is maintained for about 10 minutes.The total sampling time is recorded.The filters are removed from the filter holders.Volumetric flow meters are again noted.The cell door is then opened; the steel plate is removed and weighed.The cutting length is measured.Sampling filters are then weighed to obtain the final mass collected and all recovered samples are sorted and stored.

Isokinetic Sampling
Isokinetic sampling probes were designed in compliance with the international standard ISO2889:2010 and allow obtaining the aerosol volume mass concentration C V (mg m -3 ) in the exhaust duct, at standard ambient temperature and pressure conditions (SATP, T S = 293°K and P S = 1013 hPa): with ∆m the net mass of aerosols collected on the filter (mg), V s the sampled volume reduced to SATP (m 3 ), Q s the sampling flowrate in SATP conditions (in L min -1 ) and t the sampling duration (min).For the filter of 47 mm in diameter, the sampling flow rate Q 47 is equal to 15 L min -1 and for the filter of 130 mm in diameter, Q 130 is equal to 80 L min -1 .The combined standard uncertainty σ Cv associated with the measured volume concentration is composed of standard uncertainties on measured weights, sampling times and flowrates and expressed by: with σ t = 2 s the uncertainty on the sampling time, σ Qs = 1 L min -1 the uncertainty on the sampling flow rate and σ ∆m the uncertainty on the net mass sampled which is the result of two weighting.Its uncertainty is thus expressed by: m m 2.
    with σ m = 0.5 mg the uncertainty of one weighting.The uncertainty on the aerosol mass concentration is expressed by: The confidence interval (CI) at a 95% confidence level is obtained by multiplying the standard uncertainty by the coverage factor k = 2.

Mass of Aerosol per Unit Length of Cut
The mass of aerosol produced per unit length of cut, entrained in the extraction circuit and sampled (M 1 in g m -1 ) is obtained by the relation: with C V47 the volume mass concentration obtained with the Ø 47 mm filter in SATP conditions (mg m -3 ), C V130 the volume mass concentration obtained with the Ø 130 mm filter in SATP conditions (mg m -3 ), Bdf the mean volume mass concentration of the background noise without any cut (mg m -3 ), Q the flowrate in the line in SATP conditions (m 3 h -1 ), t the sampling duration (min) and L the length of cut (m).The combine standard uncertainty of the mass of emitted particle per unit length of cut (M 1 ) is expressed by: where C m is the mean mass concentration of the two sampling and σ L = 1 mm is the uncertainty on the measured length of the cut.We also take the confidence interval at a 95% confidence level, i.e., CI(M 1 ) = 2 × σ M1 .The mean masses of aerosol emitted per unit length of cut for the different thicknesses tested and for both configurations (in air and underwater) are represented in Fig. 3 with their confidence interval.The mass of aerosol emitted per unit length of cut increases with the thickness of the specimen.
For the cuttings realized in air, a minimum value of M 1 ≈ 2 g m -1 is found for the 10 mm plate and a maximum value M 1 ≈ 41 g m -1 for the 60 mm plate.When cuts are performed underwater, the mass of aerosol emitted per unit length of cut remains below 2 g m -1 .

Mass of Aerosol per Unit Area of Cut
The mass of aerosol per unit area of cut M 2 (g m -2 ) is simply expressed by the relation: with e the thickness of the plate in millimeters.The standard uncertainty σ M2 is obtained from the expression:  The confidence interval at a 95% confidence level is obtained by multiplying the standard uncertainty by the coverage factor k = 2, i.e.: CI(M 2 ) = 2 × σ M2 .

Mass Balance per Unit Length of Cut
The remove mass of the cut part per unit length of cut (M p ), expressed in grams, is obtained from the difference in weight before (M i ) and after (M f ) the cutting: The standard uncertainty σ Mp is obtained from the expression: with a confidence interval at 95% confidence level: CI(M p ) = 2 × σ Mp .The mean remove masses of the specimen per unit length of cut for the different thicknesses tested and for both configurations (in air and underwater) are represented in Fig. 4 with their confidence interval CI(M p ).The mass removed per unit length of cut logically increases with the thickness of the specimen with minimum values of M p ≈ 100 g m -1 for the 10 mm plates whether the cut is performed in air or underwater; and maximum values of M p ≈ 1200 g m -1 for the 60 mm plates.Although the removed mass M p increases between the cuts of 10 mm and 20 mm plates in air, it remains under 200 g m -1 when cuttings are performed underwater.This benefit could be another argument for the cutting underwater during dismantling operations by reducing the dross quantity.Nevertheless, considering the uncertainties and the absence of data for the thicker specimens (30 mm and 60 mm), this assumption must be confirmed by other experiments.

Aerosolized Mass Fraction
The aerosolized mass fraction K R is the ratio between the mass of aerosol per unit length of cut M 1 and the mass balance per unit length of cut M p for each plate thickness.The mean aerosolized mass fractions are represented with their confidence interval CI(K R ) in Fig. 5 for each plate thickness and for the two configurations (in air and underwater).For in air cuttings, the mass fraction is between 1% and 4% whatever the thickness of the specimen.For underwater cuttings, the mass fraction drops below 1% for the two thicknesses tested.

Laser Cutting in Air
The first measurement series was carried out in air and 29 tests were performed, with different sample thicknesses and cutting speeds (controlled by the laser power).The cutting capability is about 10 mm per kW but the cutting  speed need to be lowered as the thicknesses of the sample enlarge to be able to cut through the whole sample.Thus the ticker the sample, the lower the cutting speed we use to cut our steel sample.The results for the mass of emitted aerosol per unit area of cut (M 2 ) are depicted in Fig. 6 for a laser power fixed at 3 kW and Fig. 7 for a laser power fixed at 8 kW with respect to the cutting speed.Data are represented with their confidence interval with a 95% confidence level.
Figs. 6 and 7 show that the amount of emitted aerosol particles decreases with the increasing cutting speed.For  cuttings realized with a laser power of 3 kW, the emitted mass of aerosol per unit area goes from 600 g m -2 at 1 cm min -1 to 56 g m -2 at 19 cm min -1 .Five cuts were done with a cutting speed close to 4 cm min -1 in order to test the repeatability of the measurements.Despite one data below 300 g m -2 , the other four tests exhibit similar results between 480 g m -2 and 600 g m -2 .In Fig. 7, the laser power was set to its nominal value, i.e., 8 kW allowing cutting through thicker samples with cutting speeds ranging from 1 to 25 cm min -1 .The maximum mass of aerosol particles emitted per unit area of cut (719 g m -2 ) is obtained with the minimum cutting speed of 1 cm min -1 whereas the minimum mass of emitted aerosol per unit area (141 g m -2 ) is measured for a cutting speed of 20 cm min -1 .One can also notice that for the same cutting speeds, these minimum and maximum values of M 2 are higher when the laser power is 8 kW.
These results are in good agreement with previous studies (Pilot et al., 2008), Bernard et al. (1998) using laser as a cutting tool, and point out that in order to reduce aerosol production, it is preferable to cut as rapidly as possible using the minimal laser power while making sure to cut all the way through.Indeed, it has been shown (Mullick et al.,  2016) that for a constant laser power, a greater cutting speed will cause a lower temperature rise of the piece to be cut thus leading to an increase in the viscosity of the molten layers of the stainless steel.This increased viscosity will mitigate the ejection of molten metal in the aerosol form.Moreover, a faster cutting will reduce the interaction time between the laser and the specimen ensuring a reduction in the energy deposited and thus in the heated affected zone (HAZ) where the vaporization of matter occurs.For a fixed laser power and laser spot diameter, Jebbari et al.  (2008) have shown that the spread of the HAZ is inversely proportional with the cutting speed.

Laser Cutting under Water
The second measurement series was carried out under water and 13 tests were performed, with different sample thicknesses and cutting speeds.For these experiments, there is an additional coaxial water curtain supplied by the laser head in order to create a dry area around the flushing air jet and the laser beam.The tests were performed under two depths of water (0.5 m and 4.2 m) in the DELIA facility for plate thicknesses of 10 mm and 20 mm and results are shown in Fig. 8.In these conditions, performances of the laser were not sufficient to achieve perforating cuts for samples with greater thicknesses.
The graph shows a significant decrease in the total mass of aerosol particles released during experiments per unit area of cut (M 2 ) compared to air cutting results regardless the cutting speeds.For cuts realized with a laser power of 3 kW, the mass of aerosol emitted per unit area drops by a factor of 10 compared to the air cuttings results.For example, at a cutting speed of 1 cm min -1 , M 2 is equal to 69 g m -2 for a cut underwater while M 2 was 600 g m -2 for the same cutting parameters in air.
When the laser power is set to 8 kW and the cutting speed to 60 cm min -1 , the emitted mass of aerosol per unit area cut is at a minimum of 5.2 g m -2 .Indeed, this high power allows faster cutting of specimens (both parameters are obviously related), which has the effect of limiting the HAZ and thus decreases the amount of emitted particles.The water depth also appears to contribute to the reduction of aerosol particles emissions.When the depth of water in the facility is increased from 0.5 m to 4.2 m, the value of M 2 drops under 10 g m -2 for the cut of a 10 mm plate thickness with a laser power of 8 kW.

Particle Morphology
During the cutting of a plate with a thickness of 20 mm,

(2013). The Transmission Electron
Microscope used in this study is a JEOL 100CXII working with an accelerating voltage of 100 kV.In order to avoid high particle concentration on the grids, the sampling time was kept under 20 seconds at a sampling flow rate of 1 L min -1 .The TEM micrographs of particles generated by the laser cuttings are depicted in Fig. 9.The TEM particles images show two different populations, aggregates of nanometer primary particles (images A and C) and bigger spherical particles (image B) with sizes greater than 150 nm for both populations.In contrast, the primary particles composing the aggregates seem to have dimension below 10 nanometers and cannot be observe with the microscope used for this study.During the TEM observations, we carried out a qualitative elemental chemical analysis of the particles by EDS (Energy Dispersive X-ray Spectrometry).Indeed, the TEM microscope (JEOL 100CXII) is equipped with an EDS microprobe (Princeton Gamma-Tech.Avalon).The analysis was performed with a voltage of 100 kV.We have found that spherical and aggregates particles give the same chemical composition corresponding to stainless steel (mostly Fe, Cr, Mn, Ni).Given the results obtained on particle morphology and size, a coagulation-agglomeration mechanism is responsible for the generation of nanoparticles in the aerosol emitted during the laser cutting of the specimen.One hypothesis on the formation process for these particles has been previously assessed (Di Lemma et al., 2014).The bigger particles are related to mechanical phenomena, formed by the ejection of liquid particles from the melted layer of the sample due to mechanical shock caused by the laser heating.The smaller particles formation may be related to rapid vapor condensation as suggest by Dewalle et al. (2010).

Particle Size Distribution for Underwater Cuts
Three particle size distributions measured using the DMS 500 for 1 minute of acquisition are shown in Fig. 10.These three particle size distributions are from cutouts of 10 mm thick plates submerged under 0.50 m of water and for laser powers of 3 kW, 6 kW and 8 kW.The particle size distributions are expressed in mass; for that, given the fact that the DMS 500 provides a raw measurement in number, a global effective density of the aerosol of 1500 kg m -3 was used to convert the electrical mobility diameter into an equivalent volume diameter.The global effective density of the aerosol has been obtained by comparing the mass concentrations given by experimental filter measurements (Eq.( 1)) with the total mass concentration given by the DMS 500 corresponding to the sum of the mass concentration given by each particle size bin of the instrument.Indeed, as it is show by TEM micrographs in Fig. 9, particles are not spherical but porous aggregates thus with an effective density lower than the material density.In order to get rid of the influence of the cutting speed it is chosen to normalize the distribution in g m -1 by multiplying the mass concentration (g m -3 ) by the ventilation rate Q (100 m 3 h -1 ) divided by the cutting speed V L (m h -1 ).
There is a slight influence of the laser power on the particle size distribution of the aerosol emitted with a shift on the particle sizes towards larger values when the laser power increases.For a laser power of 3 kW the main mode is at 0.14 µm, 0.16 µm for a power of 6 kW, and finally 0.18 µm for a power of 8 kW.We also note the presence of a secondary mode between 0.6 µm and 0.8 µm.The mass median diameters (MMD) and the geometric standard deviation (GSD) of these particle size distributions are gathered in Table 1.
Once these particles are produced, they are rapidly subject to environmental conditions surrounding the kerf.During underwater cuttings, a strong flushing airflow is applied around the laser beam in order to maintain the fluence of the laser and to expel the melted metal produce during the cut.We will see in the next paragraph that this flushing airflow produces a large amount of bubbles in the vicinity of the cutting area which leads to improve the trapping of ultrafine particles in the liquid phase.

Particle Collection Efficiency of the Water Column
In the following paragraph, we analyzed the differences between the DELIA experiments realized in air and underwater in order to highlight the benefits of cutting underwater regarding the production of exhaust aerosols in the ventilation circuit.The highest mass of aerosol per unit area of cut is produced with the thicker plate in air (M 2 ≈ 719 g m -2 ) whereas the lowest is for the 10 mm plate cut underwater (M 2 ≈ 5.3 g m -2 ).Underwater cuttings clearly exhibit a strong reduction in aerosol emissions depending on the height of water, the thickness of the plate or the cutting speed.In addition, this reduction is maximized when cuts are done at the highest cutting speed achievable depending on the thickness of the plate for both in air and underwater process.The main mechanism involved in the collection of the particles during the cuts made under water is the collection of particles by bubbles created while cutting.Collection of ultrafine particles is done in the liquid phase and depends on the size of the bubbles, their rising velocity in the water column and the nature of the trapping liquid (Charvet et al., 2011;Koch and Weber, 2012).
Various bubbling regimes exist and will directly influence the characteristics of bubbles and thus their collection efficiency.These regimes depend essentially on the viscosity of the liquid, the gas flow rate and density (and thus pressure) as well as the diameter of the column.The mass flow rate of the gas (Q g ) which passes through the section of the column (S c ) is the superficial gas velocity (u sg ).During the underwater cuttings, the mass flow rate of the gas corresponds to the flushing air used to maintain the laser beam in a dry environment and to expel the molten metal from the specimen.During the experiments, the mass flow rate of the flushing gas is 400 L min -1 and the diameter of the exhaust chimney is 30 cm which gives a superficial gas velocity in the column of u sg = 0.1 m s -1 .With these features, the exhaust chimney of DELIA corresponds to a bubble column with a heterogeneous Churn-turbulent flow regime (Shah et al., 1982).With this assumption, the diameter of the bubbles can be obtained with the Davidson and Schüler (1997) correlation: where g accounts for the gravitational acceleration (m s -2 ).
For an airflow rate of Q 0 = 400 L min -1 , the bubble diameter is approximately 3 mm.The bubble rising velocity is then derivate with the Mendelson (1967) correlation expressed by: 72.10 -3 N m -1 ) and  L the density of the fluid (ρ L = 1000 kg m -3 ), the bubble rising velocity is 0.25 m s -1 .
The most widely use model for particle collection with bubble columns has been developed by Fuchs (1964) and later on by Pich and Schütz (1991).Particles which are dispersed and transported in a stable rising bubble may be absorbed by the surrounding liquid to a certain degree due to several deposition processes.The major processes are particle diffusion, interception and inertial deposition (Kim et al., 2001).Diffusion is a particle capture mechanism based on Brownian motion which is the dominant scrubbing mechanism for small particles (diameter lower than 0.1 µm).Particles with diameter between 0.1 µm and 1 µm may be collected through the interception mechanism if the particle passes within one particle radius from the droplet surface.Finally, inertial impaction is the predominant removal mechanism for particles larger than 5 µm (Bianchini et al., 2016).In this theory, each absorption mechanism is function of the bubble rising velocity, and therefore the bubble size, the gas flow rate and the height of the column.The global efficiency can be calculated for a bubble diameter (d b ), a height of the column (H) and for a particle diameter (d p ). Considering spherical bubbles, the Pich and Schütz model is expressed by:  with ρ p and d p the particle density (kg m -3 ) and the particle diameter (m), µ g the dynamic viscosity of the gas (Pa s), k B the Boltzmann constant (J K -1 ), T the gas temperature (K) and Cu the Cunningham coefficient.Considering the bubble diameter and rising velocity calculated in Eq. ( 10) and Eq. ( 11), we can derivate the collection efficiency of the water column in DELIA for water levels of 0.5 m and 4.2 m regarding the particle diameter.
The calculation gives a global collection efficiency of 86.6% for the height of 0.5 m and close to 99.99% when cuts are submerged under 4.2 m of water.This observation can be explained by the longest residence time of the particles in the column and the resulting improved transfer of particles from the gas to the liquid phase (Charvet et al., 2011).One must also notice that the most penetrating particle size (MMPS) through the water column occurs for particle sizes near 200 nanometers which corresponds to the first mode of the particle mass size distributions presented in Fig. 10.
For sake of clarity, we introduce here the scrubbing coefficient S which is expressed by S = 1/(1 -η Pich ).This coefficient is represented in Fig. 11 for the two water heights used previously with the model of Pich and Schütz (1991).When the collection efficiency η Pich tends to 1 (i.e., 100% of particles collected) the scrubbing coefficient tends toward infinite (no particle passes through the water column).It is therefore obvious that even under a water column of 4.2 m height and collection efficiency close to 99.99%, some particles still passes across the column and reach the top of it.For example, scrubbing coefficient in Fig. 11 indicates that for 1000 particles produced during the cutting with a diameter around 200 nanometers, only one particle will pass throughout the water column of 4.2 m height.On the other hand, for the water height of 0.5 m, one particle on two passes through the water column.

Experimental Collection Efficiency
The particle collection efficiency of the bubbles formed during the cutting increases exponentially with the height of the water column and hence with the residence time.The experimental collection efficiency can be derivate by and is presented in Table 2 for different sample thicknesses, cutting speed and laser power for a height of the water column of 0.5 m and in Table 3 for a height of the water column of 4.2 m.
For the cuttings of the 10 mm thickness specimen, the collection efficiency increases with the height of the water column with a global collection efficiency of 94% for cuts realized under 0.5 m of water and an efficiency of 99% for cuts performed under 4.2 m of water.When the thickness  of the specimen is larger, the global collection efficiency remains above 80% for cuts performed under 0.5 m of water.
It is found that the model of Pich and Schütz (1991) is in good agreement with experimental collection efficiencies determined during the cuttings of various specimens.For maximum water depth of 4.2 m, the efficiency is close to 99% both for the experimental data and the scrubbing model.For height of water of 0.5 m, the overall collection efficiency given by the scrubbing model is 86% thus higher than that for the cuts of specimen thickness of 30 mm but less for cuts with plate thickness of 10 mm.It should be noted here that the Pich and Schütz model gives a good estimate of the collection efficiency for the diffusion regime (d p < 200 nm) but underestimates the collection efficiency in the other regimes (sedimentation and inertial motion) for particle sizes greater than 200 nm (Cadavid-Rodriguez et al., 2014).One hypothesis is that the slight discrepancies observed in Tables 2 and 3 between the model and the data come from the sub-micron mode (d p > 200 nm) present in the particle mass size distributions gathered in Fig. 10 and for which the model of Pich and Schütz (1991) is less adequate (Koch and Weber, 2012).

CONCLUSION
A new experimental set-up has been developed for the characterization of aerosol particles produced by an industrial continuous wave ND:YAG high power laser dedicated to the dismantling of former nuclear equipment.Tests performed on the DELIA facility show that the use of a high power laser, mandatory for rapid cutting of large thicknesses of stainless steel samples up to 60 mm does not imply an increase in aerosol particles emissions.Indeed, we showed that the emission of aerosol particles per unit area of cut decreased when the cutting speed increased.An average gain of one order of magnitude is observed when cuts are done under water, this benefit further increasing with the water depth.With the help of a simple model for the particle collection efficiency of the water column, we have been able to explain this reduction of the aerosol emissions.Bubbles created at the exit of the laser head by the flushing airflow allow scrubbing of ultrafine particles in the liquid phase.Depending on the height of the water column above the piece to be cut, we have found a good agreement between the model and the global collection efficiencies obtained experimentally.The particle size distributions of these aerosol particles emissions were characterized and we found that an important part of these aerosol particles has sizes between 0.1 and 1 micron.Morphological analysis performed with a Transmission Electron Microscope revealed that particles are aggregates of nanometer primary particles.In order to better understand the formation mechanisms of these nanometer primary particles, it is planned to analyze the size and shape of the collected particles and to obtain information on their elemental composition using High Resolution Transmission Electron Microscopy.

Fig. 1 .
Fig. 1.Example of a 20 mm thick specimen with various cuts.The left image corresponds to the front side (where the laser beam impact the plate) and the right image to the back side of the plate.

Fig. 2 .
Fig. 2. Sketch of the DELIA facility and its aerosol sampling line.

Fig. 3 .
Fig. 3. Mean mass of aerosol per unit length of cut (M 1 ) depending on the thickness of the specimen during both in air and underwater cuttings.
per unit length cut (g m -1 ) Thickness of the sample (mm) In air Underwater with σ e = 1 mm the uncertainty on the specimen thickness.

Fig. 4 .
Fig. 4. Mean mass balance per unit length of cut (M p ) depending on the thickness of the specimen during both in air and underwater cuttings.

Fig. 5 .
Fig. 5. Aerosolized mass fraction (K R ) depending on the thickness of the specimen during both in air and underwater cuttings.

Fig. 6 .
Fig.6.Mass of the aerosol emitted per unit area of cut with respect to the cutting speed for tests in air and a laser power of 3 kW.

Fig. 7 .
Fig. 7. Mass of the aerosol emitted per unit area of cut with respect to the cutting speed for tests in air and a laser power of 8 kW.

Fig. 8 .
Fig. 8. Mass of the aerosol emitted per unit area of cut with respect to the cutting speed for tests submerged under water and for laser powers of 3 kW and 8 kW.
12)where, η Pich is the collection efficiency, D (m 2 s -1 ) is the Brownian diffusion coefficient, U b and d b are the bubble rising velocity (m s -1 ) and diameter (m), τ is the particle relaxation time (s) and H is the water column height (m).The particle relaxation time and the Brownian diffusion coefficient are expressed respectively by:

Fig. 11 .
Fig. 11.Scrubbing coefficient for the two water levels in DELIA.

Table 1 .
Mass median diameters and geometric standard deviation of the three particle size distributions.

Table 2 .
Aerosol scrubbing efficiencies when cutting under 0.5 m of water for different plate thicknesses.

Table 3 .
Aerosol scrubbing efficiencies when cutting under 4.2 m of water for a plate thickness of 10 mm.