## Abstract

Several nuclear facilities are currently being decommissioned in France, on CEA and EDF sites. The decommissioning process as a whole, with cutting operations and nuclear waste management presents workers with significant risks of exposure. The aim of this study was therefore to improve risk assessments of internal exposure in operations where the main uncertainties are the removal factor and the airborne release factor. A new method is presented to assess the risk of internal exposure and optimize the choice of personal protective equipment to use during nuclear dismantling operations. The main forces and parameters influencing the level of labile contamination and particle resuspension were identified from a literature review and feedback. A manageable set of parameters was then obtained based on literature data and on-site information. The effects of the drying temperature, relative humidity, roughness of the contaminated surface, and wiping pressure were thus investigated on labile contamination and removal factor measurements. Successive wipings with cotton pads were performed on surfaces contaminated with simulated contaminants to estimate the influence of the different parameters. Results highlight the importance of surface roughness over the other considered parameters.

## 1 Introduction

A major task for Electricity of France (EDF) and the French Alternative Energies and Atomic Energy Commission (CEA) is and will be to decommission a growing number of nuclear facilities over the coming decades. In the course of these operations, workers' health is the main priority. Improved measurement methods for labile radioactive contamination are therefore sought to optimize personal protective equipment choices.

Labile contamination is a form of contamination that is readily transferred simply by contact, wiping, brushing, or light washing. It is therefore easily detached from contaminated surfaces, leading to airborne contamination and creating a risk of internal exposure. Following international standards (ISO 7503-2) [1], removable contamination is defined as all the quantifiable activity (per unit mass) that can be extracted by successive dry smears of a surface.

According to the literature, the main parameters governing the removal factor in smearing processes are the type contamination [2], the nature [3], roughness [4], and size of the wipe [5], the pressure applied [6,7], the recovering solvent [3], the roughness [8,9] and nature of the surface [4,5,7,10], the age of the contamination [10], the adhesion force [9], the concentration of contamination on the surface [11], the adherence to the recommended swiping procedure [7], the temperature [11], the relative humidity [11] and the form in which the contamination was deposited [4,11] (dried liquid contamination or particulate contamination).

The four parameters that were deemed most feasible to evaluate according to on-site constraints and available methods were:

The drying temperature;

The drying relative humidity;

The roughness of the contaminated surface;

The swiping pressure.

The other parameters mentioned above kept were fixed.

Shoji et al. studied the effect of temperature and humidity on the removal factor [11], showing that, for a low deposited concentration of 1.0 *μ*g/mL methyl-14C-thymidine, the removal factor depended more on the nature of the surface than environmental conditions. However, they observed the opposite when the contaminant concentration was 500 times higher. At low contaminant concentrations indeed, a greater proportion of the contaminant is in direct contact with the surface, such that the nature of the surface has a stronger influence. For higher contaminant concentrations on the other hand, the adhesive forces between contaminant particles depend only on environmental conditions, which consequently govern the removal factor [11]. Understanding the effect of these parameters relative to others seems important to accurately estimate internal exposure risks.

Surface roughness is a straightforward parameter to measure on worksites but can affect measurements of labile contamination. Jung et al. [8] studied the influence of roughness by comparing the contamination levels measured by two operators. They observed that the interoperator variability in removal factor estimates was higher for rough surfaces than for smooth ones. Verkouteren et al. [4] found that roughness had no effect on the adhesion of fluorescent polystyrene latex spheres 42 *μ*m (1.7 × 10^{−3} in.) and 9 *μ*m (3.5 × 10^{−4} in.) in size. Roughness should affect adhesion forces and thus the removal factor, but since the range of van der Waals interactions between particles and interfaces is just a few nanometers, if the scale of the roughness is much less than the particle size, less of each particle is in direct contact with the surface, resulting in weaker adhesion [9].

The pressure applied on the wipe is also an important parameter. To improve surface wipe sampling, Verkouteren et al. [6] compared a wide range of pressures (14.9 to 69.6 kPa, 2.2 to 10.1 psi) applied on the wipe using a force-sensing resistor. They found that the collection efficiency increased linearly as the pressure increased from 5.0 kPa (0.7 psi) to 54.7 kPa (7.9 psi). However, Warren et al. [10] found no increase in recovery with the pressure applied on dry wipes, but at lower pressures (1.0 kPa (0.1 psi) to 7.8 kPa (1.1 psi)). In fact, the latter results are within the estimated uncertainties of Verkouteren's et al. [6] at low applied pressures. Robinson et al. [7] found that the collection efficiency with trinitroperhydro-1,3,5-triazine increased with pressure over a load range of 3.6 kPa (0.5 psi) to 14.7 kPa (2.1 psi). However, on ABS plastic surfaces, for instance, the removal factor does not increase indefinitely with the applied pressure [7]. Once the applied pressure is sufficient to move the labile contamination, the removal factor can only be increased further by increasing the adhesion force of the wipes [4].

In this study, the effects on the labile contamination and the removal factor of the drying temperature, the relative humidity, the roughness of the contaminated surface, and the wiping pressure were investigated using fluorescein. Dried liquid fluorescein contamination was chosen because it can be deposited in very small amounts and can be detected at very low concentrations (detection limit, 10^{−10} g/L).

The objective of this paper is to provide a better understanding of which parameters need to be investigated to experimentally measure internal exposure risks.

## 2 Materials and Methods

The methods described in the ISO 7503-2 standard [1] were followed to obtain the best possible estimates of the level of labile contamination and the removal factor by analyzing successive smears on different stainless steel surfaces. The goal was to find out which parameter has the greatest influence on the variability of the labile contamination and the removal factor measurements.

### 2.1 Experimental Design.

In this study, the result of an experiment, “*Y*”, represents the proportion of the labile contamination removed or the removal factor. The set of investigated parameters retained above are likely to affect the outcome of the experiments, noted “*Y*”. These parameters are studied over a given range: the domain of the study. The conclusions of this kind of study are valid provided the parameters are within their domain.

The four parameters were varied between two levels each, a low value coded “−1” and a high value coded “1”. Sampling the entire domain of the study would therefore have required 2^{4} = 16 experiments in full factorial design. However, the experiments are too time-consuming so a fractional factorial design was chosen instead, whereby the principal effects of each parameter were obtained with just 2^{4−1} = 8 experiments.

A fractional factorial design to study four parameters in eight experiments can be represented by a 2^{4−1} = 2^{3} matrix (Table 1). The first three columns, A, B, and C are constructed by alternating between −1 and 1 on every row in column A, every two rows in column B, and every four rows in column C. The other columns are obtained by multiplying the first three columns together in every possible combination. For instance, column AB is obtained by multiplying column A with column B.

A | B | C | AB | AC | BC | ABC |
---|---|---|---|---|---|---|

−1 | −1 | −1 | 1 | 1 | 1 | −1 |

1 | −1 | −1 | −1 | −1 | 1 | 1 |

−1 | 1 | −1 | −1 | 1 | −1 | 1 |

1 | 1 | −1 | 1 | −1 | −1 | −1 |

−1 | −1 | 1 | 1 | −1 | −1 | 1 |

1 | −1 | 1 | −1 | 1 | −1 | −1 |

−1 | 1 | 1 | −1 | −1 | 1 | −1 |

1 | 1 | 1 | 1 | 1 | 1 | 1 |

A | B | C | AB | AC | BC | ABC |
---|---|---|---|---|---|---|

−1 | −1 | −1 | 1 | 1 | 1 | −1 |

1 | −1 | −1 | −1 | −1 | 1 | 1 |

−1 | 1 | −1 | −1 | 1 | −1 | 1 |

1 | 1 | −1 | 1 | −1 | −1 | −1 |

−1 | −1 | 1 | 1 | −1 | −1 | 1 |

1 | −1 | 1 | −1 | 1 | −1 | −1 |

−1 | 1 | 1 | −1 | −1 | 1 | −1 |

1 | 1 | 1 | 1 | 1 | 1 | 1 |

Four columns need to be chosen from Table 1 to study four parameters and create the experiment matrix (Table 2). Choosing columns A, B, C, and ABC minimizes interaction effects that might perturb the experimental setup. Defining the parameters now as A = *X*_{1} = Roughness, B = *X*_{2} = Applied pressure, C = *X*_{3} = relative humidity, and ABC = *X*_{4} = temperature. The values “1” and “−1” correspond to the extrema of the ranges of each parameter, such that for the temperature, “−1” corresponds to 25 °C and “1” corresponds to 35 °C. The *X*_{0} column filled with 1 s corresponds to the case in which all the parameters are set to the middle of their predefined ranges. The *Y* column corresponds to the results of each experiment, with one row per experiment.

X_{0} | X_{1} | X_{2} | X_{3} | X_{1}X_{2} | X_{1}X_{3} | X_{2}X_{3} | X_{4} = X_{1}X_{2}X_{3} | Y |
---|---|---|---|---|---|---|---|---|

1 | −1 | −1 | −1 | 1 | 1 | 1 | −1 | Y_{1} |

1 | 1 | −1 | −1 | −1 | −1 | 1 | 1 | Y_{2} |

1 | −1 | 1 | −1 | −1 | 1 | −1 | 1 | Y_{3} |

1 | 1 | 1 | −1 | 1 | −1 | −1 | −1 | Y_{4} |

1 | −1 | −1 | 1 | 1 | −1 | −1 | 1 | Y_{5} |

1 | 1 | −1 | 1 | −1 | 1 | −1 | −1 | Y_{6} |

1 | −1 | 1 | 1 | −1 | −1 | 1 | −1 | Y_{7} |

1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | Y_{8} |

X_{0} | X_{1} | X_{2} | X_{3} | X_{1}X_{2} | X_{1}X_{3} | X_{2}X_{3} | X_{4} = X_{1}X_{2}X_{3} | Y |
---|---|---|---|---|---|---|---|---|

1 | −1 | −1 | −1 | 1 | 1 | 1 | −1 | Y_{1} |

1 | 1 | −1 | −1 | −1 | −1 | 1 | 1 | Y_{2} |

1 | −1 | 1 | −1 | −1 | 1 | −1 | 1 | Y_{3} |

1 | 1 | 1 | −1 | 1 | −1 | −1 | −1 | Y_{4} |

1 | −1 | −1 | 1 | 1 | −1 | −1 | 1 | Y_{5} |

1 | 1 | −1 | 1 | −1 | 1 | −1 | −1 | Y_{6} |

1 | −1 | 1 | 1 | −1 | −1 | 1 | −1 | Y_{7} |

1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | Y_{8} |

Each column is corresponds to a parameter: *X*_{1}: roughness; *X*_{2}: pressure applied on the wipe; *X*_{3}: drying relative humidity; *X*_{4}: drying temperature. The *X*_{0} column corresponds to the case where all parameters are in the middle of their ranges. The *Y* column corresponds to the results of each experiment, represented by each row in the table. The *X _{i}X_{j}* columns are used to calculate the interaction coefficients between parameters

*i*and

*j.*

Note that the X_{4} column is confounded with column *X*_{1}*X*_{2}*X*_{3}. This is known as aliasing and is written $X4\u2009\u2261\u2009X1X2X3$. Similarly, $X1\u2009\u2261\u2009X2X3X4$; $X2\u2009\u2261\u2009X3X4X1$; $X3\u2009\u2261\u2009X4X1X2$ and $X0\u2261\u2009X1X2X3X4$.

$X0$ is also called the independent generator of the 2^{4−1} fractional factorial matrix. This is an identity vector that helps account for all aliases. For instance, for $X1$, we have:$\u2009X1X0\u2009\u2261\u2009X1X1X2X3X4\u21d2X1\u2261X2X3X4$, meaning that X_{1} is confounded with $X2X3X4$.

Each column in Table 2 corresponds to a specific parameter (*X _{i}*) and highlights its influence,

*I*, on the result.

_{i}*I*are the estimators of the (aliased) coefficients of a fractional factorial matrix and

_{i}*b*are the estimators of the (nonaliased) coefficients of a complete factorial matrix. For example, for column

_{i}*X*

_{1}, we have

*I*

_{1}=

*b*

_{1}+

*b*

_{234}since

*X*

_{1}is confounded with $X2X3X4$, with

*b*

_{1}being the coefficient related to

*X*

_{1}, characterizing the impact of the roughness on the results, and b

_{234}the coefficient quantifying the impact of the interaction of the applied pressure (

*X*

_{2}), the relative humidity (

*X*

_{3}), and the temperature (

*X*

_{4}). This last coefficient (

*b*

_{234}) can be neglected because interactions between three parameters are rare. We obtain

*I*

_{1}=

*b*

_{1}.

The same approach leads to the same conclusions for the other columns.

For column AB = *X*_{1}*X*_{2}, we have *I*_{12} = *b*_{12} + *b*_{34}. In this case, neither interaction term can be neglected because interactions between two parameters are common.

*X*(= −1 or 1) corresponds to the state of parameter, and the

_{i}*b*coefficients represent the weight of parameter

_{i}*i*in the regression model. These coefficients can be calculated by multiplying each column of matrix

*X*by each row of

*Y*. This leads, for instance, to Eq. (2) for

*b*

_{1}

Table 3 lists all the experiments required to characterize the parameters. After calculating the integer coefficients corresponding to the midrange of the response between the limits of the experimental domain, experiments in the middle of the domain need to be performed to interpret them, bearing in mind that in the studied domain, roughness is a discrete parameter. The responses at maximum and minimum roughness are key to estimating the confidence intervals of the coefficients in the middle of the domain (*T* = 30 °C (86 °F), RH = 60%, *P* = 15.5 kPa (2.25 psi), Ra). Equations (3), (4), and (5) can then be solved

Exp | Ra (μm) (in.) | P (kPa) (psi) | RH (%) | T (°C) (°F) | Y |
---|---|---|---|---|---|

1 | 4.69 | 6.2 | 50 | 25 | Y_{1} |

1.57 × 10^{−4} | 0.9 | 77 | |||

2 | 0.05 | 6.2 | 50 | 35 | Y_{2} |

1.97 × 10^{−6} | 0.9 | 95 | |||

3 | 4.69 | 24.8 | 50 | 35 | Y_{3} |

1.57 × 10^{−4} | 3.6 | 95 | |||

4 | 0.05 | 24.8 | 50 | 25 | Y_{4} |

1.97 × 10^{−6} | 3.6 | ||||

77 | |||||

5 | 4.69 | 6.2 | 70 | 35 | Y_{5} |

1.57 × 10^{−4} | 0.9 | ||||

95 | |||||

6 | 0.05 | 6.2 | 70 | 25 | Y_{6} |

1.97 × 10^{−6} | 0.9 | 77 | |||

7 | 4.69 | 24.8 | 70 | 25 | Y_{7} |

1.57 × 10^{−4} | 3.6 | 77 | |||

8 | 0.05 | 24.8 | 70 | 35 | Y_{8} |

1.97 × 10^{−6} | 3.6 | 95 |

Exp | Ra (μm) (in.) | P (kPa) (psi) | RH (%) | T (°C) (°F) | Y |
---|---|---|---|---|---|

1 | 4.69 | 6.2 | 50 | 25 | Y_{1} |

1.57 × 10^{−4} | 0.9 | 77 | |||

2 | 0.05 | 6.2 | 50 | 35 | Y_{2} |

1.97 × 10^{−6} | 0.9 | 95 | |||

3 | 4.69 | 24.8 | 50 | 35 | Y_{3} |

1.57 × 10^{−4} | 3.6 | 95 | |||

4 | 0.05 | 24.8 | 50 | 25 | Y_{4} |

1.97 × 10^{−6} | 3.6 | ||||

77 | |||||

5 | 4.69 | 6.2 | 70 | 35 | Y_{5} |

1.57 × 10^{−4} | 0.9 | ||||

95 | |||||

6 | 0.05 | 6.2 | 70 | 25 | Y_{6} |

1.97 × 10^{−6} | 0.9 | 77 | |||

7 | 4.69 | 24.8 | 70 | 25 | Y_{7} |

1.57 × 10^{−4} | 3.6 | 77 | |||

8 | 0.05 | 24.8 | 70 | 35 | Y_{8} |

1.97 × 10^{−6} | 3.6 | 95 |

For each experiment, the roughness, the pressure applied on the wipe, the drying relative humidity and the drying temperature were set to the following values.

where $se(0)2$ is the mean square deviation of the result in the middle of the domain, *s _{Y}*

_{1}

^{2}is the mean square deviation of the result for rough surface experiments,

*s*

_{Y}_{2}

^{2}is the mean square deviation of the result for smooth surface experiments, ν

_{1}is the degrees of freedom for rough surface experiments, ν

_{2}is the degrees of freedom for smooth surface experiments,

*s*

_{b}_{i}

^{2}is the mean square deviation of coefficient

*b*

_{i}, quantifying the effect of parameter i on the result,

*N*

_{0}is the number of experiments in the middle of the domain,

*s*is the standard deviation of coefficient

_{bi}*b*

_{i}characterizing the influence of parameter i on the result,

*t*is the t-value from Student's t-distribution.

_{α}### 2.2 Parameter Management.

Parameter management is a crucial part of any scientific experiment. For the present study, since many parameters influence the level of labile contamination and the removal factor and because they cannot all be studied, those that cannot study will be kept fixed.

#### 2.2.1 Fixed Parameters.

*Contaminant*: The removal factor is estimated by wiping the same surface multiple times, with a risk of contamination for the operator [1]. To avoid this problem, fluorescein was used as a nonradioactive surrogate. Fluorescein is widely used as a simulant in atmospheric studies because it can be detected at very low concentrations by fluorimetry (detection limit, 10^{−10} g/L) [13]. An Optoprim USB ESElog fluorescence detector (E470/D520/E550/D600) was used to quantify fluorescein in the collected samples. Since the removal factor was not known beforehand, at least 7.12 × 10^{−9 }mol fluorescein was deposited on the surfaces to ensure it remained detectable throughout the experiments. Note that this is about 77 times more than the number of radioactive substances likely to be found on decommissioning sites, based on an activity of 400 Bq/cm^{2} (*β*/*γ*) for ^{137}Cs.*p*H: the *p*H of the fluorescein sample was fixed at 10.6 to keep the fluorescence yield constant.

*Wipes*: CEPOVETTE tear-proof cotton wipes (200501PACK) were used. Although many different types of wipes are used in the nuclear industry, this one was created specifically to estimate levels of labile contamination and is routinely used by EDF. Note that, only pristine cotton wipes were used because the random fiber arrangement increases the removal factor [4].

In addition to these fixed parameters, the four variable parameters considered in this study are described below.

#### 2.2.2 Variable Parameters.

*The drying temperature:* Drying experiments were performed in a Firlabo steam room calibrated with a PCE-HT 114-ICA thermohygrometer to avoid deviations.

*The drying relative humidity*: The same setup was used to control the relative humidity as a function of the temperature.

*The roughness of the contaminated surface*: The experiments were performed on 316 L stainless steel (S.M.T.S. Company, L'Ardoise, France), with different roughness grades. Roughness was measured using a Handysurf+ E-35 roughness meter with a 2 *μ*m (7.87 × 10^{−5} in.) roughness probe. The average roughness (Ra) of the substrate was obtained from measurements in the middle and in each corner of the surface, in five measurement zones. Three Ra measurements were performed in each zone according to ISO 97/09, with a cutoff value of 2.5 mm (0.08 in.). The measurement step was 0.6 mm/s (0.02 in./s) with a total sample length of 5.00 mm (0.20 in.). The 15 measured values were then averaged to obtain Ra.

*Pressure*: A device specifically designed to apply the same pressure on all the wipes during the smearing process was produced by 3D printing (Fig. 1). The area of the device was equal to the size of the wipes (45 mm (1.8 in.) diameter) to control the pressure applied. To determine a representative pressure range, the pressure applied by different operators was calculated from the mass of the wipe and the surface area of their fingerprints

where *P* is the pressure (Pa), *m* is the mass of the applied wipe (kg), *g* is the gravitational acceleration (m/s^{2}), *S* is the the surface area of the wipe (m^{2}).

The pressures applied by the operators were found to be of the same order of magnitude as reported elsewhere [6,7,10], with values clustering around 6.2 kPa (0.9 psi) and 24.8 kPa (3.6 psi). We used these results since they correspond to hand-applied pressures, rather than using the devices that are typically used on decommissioning sites.

### 2.3 Experiment Steps.

where *A* is the total labile activity (Bq), *m* is the mass of the sample (g), *n* is the number of wipes performed.

Since fluorescein, a nonradioactive surrogate, was used as a contaminant, it is important to consider the molar amount of contaminant present. To estimate the total labile contamination, one hundred successive wipes were performed on the same surface as stipulated in ISO 7503-2 [1] and the removal factor associated with the considered parameter was calculated from the value obtained for the first wipe.

The surface area of the samples was 100 cm^{2} (15.5 in.^{2}) and was chosen to limit operator uncertainties. However, performing a homogeneous wipe over 100 cm^{2} is challenging, even for simple surface geometries [6]. We nevertheless excluded this subjective parameter, assuming that mitigation measures would be used in practice to limit any possible uncertainties [4].

The smearing experiments were performed as follows:

The stainless-steel substrate was warmed up in the steam room to a stable 25 °C (77 °F) at 50% RH.

Twenty-one 32

*μ*L spots of fluorescein were deposited within a 100 cm^{2}(15.5 in.^{2}) circular zone (Fig. 2). The fluorescein concentration was fixed at 1.06·10^{−5}mol/L in ammonia water at*p*H = 10.6 to deposit a total amount of 7.12 × 10^{−9}mol fluorescein on the surface;The surface was dried for two hours in a box with holes in each corner in the steam room. The box served to shield the surface from aeraulic stress;

The temperature and the relative humidity were adjusted according to the experimental design. As soon as the designated temperature and relative humidity were reached, the sample was left to dry for 40 min;

The surface was then dry-wiped 100 times using the wiping device to control the applied pressure (Fig. 1). The wiping was done along a spiral path, as described in Fig. 3, from the edge to the center. Note that no standard wiping path was defined and the geometry of the cotton wipes constrained the paths that could be adopted. Dry wipes were used to avoid having to account for the effects of a solubilizing solvent because some species are more soluble than others in any given solvent [3]. In general, moreover, the use of solvents to wipe surfaces on decommissioning sites is prohibited to avoid spreading the contamination.

After smearing, the fluorescein was extracted from the wipes using 50 mL of ammonia water. Each wipe was placed in a closed vessel and then sonicated (two periods of 30 min without shaking);

Finally, the stainless-steel surface was rinsed with 150 mL of ammonia water to calculate the substance balance and the amount of non-labile contamination.

### 2.4 Distribution of the Contamination on the Studied Surface.

The stainless-steel substrate used to measure the level of labile contamination and the removal factor was shaped to fit in a scanning electron microscopy (SEM) instrument.

After depositing fluorescein as described in Sec. 2.3, samples of each roughness grade were observed using a FEI Inspect S50 SEM device in high vacuum mode, at a working distance of 9.8 mm (0.4 in.), with an acceleration voltage of 10 kV and a current intensity of 50 nA. The distribution of fluorescein before and after the smearing process was monitored via the fluorescein counterion (sodium) in energy dispersive X-ray (EDX) maps obtained with a Bruker X-flash SDD detector (10 mm^{2}), using standardless analysis with the P/B-ZAF algorithm.

## 3 Results and Discussion

The experiments (Table 3) were designed to provide information on the influence of the different parameters on the removal factor and the level of labile contamination. The results from labile contamination measurements will be presented and discussed first, before focusing on those for the removal factor.

### 3.1 Labile Contamination.

To obtain the most accurate possible estimates of the labile contamination loads, series of 100 successive wipes were performed on substrates with different levels of roughness. The values of Ra obtained for the rough and smooth stainless steel surfaces were respectively 4.69 *μ*m (1.57 × 10^{−4} in.) and 0.05 *μ*m (1.97 × 10^{−6} in.). To simplify the experiments, labile contamination was extracted from 50 wipes, and the amounts of labile contamination in the remaining 50 were estimated by regression. However, the substance balance was always calculated after the final rinsing of the surface, which recovered all remaining contamination. The sum of the estimates from the wipes and of the amounts recovered from the final rinsing gave a mean labile concentration amount of 7.0 × 10^{−9 }mol over all the experiments, with a standard deviation of 0.3 × 10^{−9 }mol. The relative error between the final substance balance and the initial deposition was 2%.

Figure 4 shows the effect of substrate roughness on the percentage of labile contamination removed from the surface after a certain number of wipes. The amounts of contamination removed from the smooth surface (blue, orange, gray, and yellow points) are at least three times greater than from the rough surface, indicating that the contamination is more accessible on the smooth surface. The contamination was deposited as a liquid, which penetrates more readily into the cracks and features of a rougher surface. To confirm this interpretation, a drop of contamination was dried and smeared with a wipe and the sodium distribution on the surface was mapped by EDX, the amount of sodium being proportional to the amount of fluorescein present. Figure 5 shows an SEM image of a rough stainless steel surface contaminated with dried fluorescein. A red halo is observed because fluorescein is hydrophilic and remains in the water until the end of the drying process. Figures 6 and 7 show the same substrate after wiping. The halo is partly missing, indicating that a large part of the contamination was removed by wiping. The contamination on the top (smooth) parts of the surface is almost all gone but more fluorescein remains between the ridges on the surface, as confirmed by EDX analysis of 3 points from smooth parts of the surface and 3 points inside ridges (Appendix 1), which show large amounts of sodium in the ridges. Using iron as a reference ion, the Na/Fe ratio in the ridges is about 0.80 compared with just ∼0.01 on top. Our observations differ from Verkouteren et al.'s [4], who worked with particles and noticed no strong correlation between the roughness of the surface and the diameter of the particles. Indeed, since the range of van der Waals interactions is only a few nanometers, when the surface features are much smaller than the particles, less of the particle is in direct contact with the surface, reducing the adhesion force [9]. In contrast, when the contamination is deposited as a liquid and left to dry, as done here, the liquid is in complete contact with the surface and penetrates into any features present, leading to strong adhesion and reducing the level of labile contamination.

We then calculated the coefficients of the experimental design (Table 4) to understand the influence of each parameter on the level of labile surface contamination (Table 2). The values for the amount of removed labile contamination in Fig. 4 correspond to the Y column in Table 2, expressed as a percentage of the deposited amount.

Figure 8 highlights the strength of each parameter's effect on the level of labile contamination. The confidence interval of ±4 percentage points (orange line) on these values was calculated from the experiment performed in the middle of the experimental domain (Fig. 9). Figure 8 shows that the roughness of the surface is by far the most important parameter to consider in order to obtain reasonable estimates of the level of labile contamination removed in the form of dried liquid contamination. The level of labile contamination was on average 63.2 percentage points higher with a surface roughness of 0.05 *μ*m (1.97 × 10^{−6} in.) than with a roughness of 4.69 (1.57 × 10^{−4} in.).

The Pareto diagram in Fig. 10 demonstrates the importance of roughness in labile contamination measurements, with 95% of the variability in labile contamination being due to variations in surface roughness. Note that the pressure applied on the wipes had a much weaker influence on the amount of contamination removed, indicating that fixed contamination remains on the surface for moderate applied pressure and stress, and only the labile contamination can be removed.

The second important result is that relative humidity only had a weak effect on the level of labile contamination removed. However, since fluorescein becomes reactive above 80% RH [12], meaning that above this RH threshold, fluorescein absorbs water, potentially increasing its adhesion to the surface via capillary attraction, the influence of the RH should only be seen for RH values above 80%. Different results would therefore probably have been obtained if a different simulant had been used.

where $b0\u2212theoretical$ is the theoretical b_{0} calculated algebraically, $y0\xaf$ is the mean value of Y at the center of the domain. This value is identical to $b0\u2212experiment$, $t\alpha \u2212DDL\u2009(0)$ is the t-value of Student's t-distribution for the degrees of freedom (DOF) at the center of the domain. This value is based on the number of repeat measurements at the center of the domain, *N*_{0}, and $se(0)2$ is the variance of the results at the center of the domain, *N* is the number of measurements performed according to the experimental design without taking into account the experiments at the center of the domain, *N*_{0} is the number of measurements at the center of the domain.

## 4 Removal Factor

The 100 successive wipes provide good estimates of the level of labile fluorescein contamination according to ISO 7503-2 [1]. These results can also be analyzed to evaluate the removal factor of the first wipe. The amounts of labile contamination and the removal factors obtained for the different experiments are compared in Fig. 11.

Table 2 and the experimentally determined removal factors (Fig. 11) can be combined to obtain a regression model. The model is the same as Eq. (1), but with different values for the coefficients quantifying the impact of each parameter on the removal factors listed in Table 5. The confidence interval of ± 2 percentage points (orange line) shown in Fig. 12 was calculated using experiments performed in the middle of the domain (Fig. 13).

B_{0} | B_{1} | B_{2} | B_{3} | B_{4} | B_{12}+B_{34} | B_{13}+B_{24} | B_{23}+B_{14} |
---|---|---|---|---|---|---|---|

16.5 | 7.8 | 4.1 | −1.7 | −4.6 | 4.4 | −4.1 | −4.0 |

B_{0} | B_{1} | B_{2} | B_{3} | B_{4} | B_{12}+B_{34} | B_{13}+B_{24} | B_{23}+B_{14} |
---|---|---|---|---|---|---|---|

16.5 | 7.8 | 4.1 | −1.7 | −4.6 | 4.4 | −4.1 | −4.0 |

Results show that a linear model is inadequate to estimate the removal factor using four parameters.

As described above for the labile contamination measurements, the roughness of the surface also has a strong effect on estimates of the removal factor. However, Fig. 12, also shows that the temperature and the pressure applied on the wipe also affect the values obtained for the removal factor. The removal factor is on average 15.6 percentage points higher when the surface roughness is 0.05 *μ*m (1.97 × 10^{−6} in.) rather than 4.69 *μ*m (1.57 × 10^{−4} in.). Indeed, more labile contamination remains trapped by surface features on a rougher surface. According to Verkouteren et al. [4], this does not hold when the particles are much larger than the length scale of the surface features. Besides, the difference of 8.1 percentage points in the removal factor between pressures of 6.2 (0.9 psi) and 24.8 kPa (3.6 psi) could be explained by the fact that higher applied pressures make the wipes behave like squeegees on the smooth surface and like brushes across the features of the rougher surface. Finally, the removal factor is on average 9.1 percentage points lower at 35 °C (95 °F) than at 25 °C (77 °F). This is expected because surface adhesion forces increase with temperature, reaching a maximum at ∼100 °C [13], due to capillary forces. Capillary nucleation is enhanced when the diffusivity of water increases, creating larger and more liquid bridges. However, this increase remains moderate below 60 °C, which is why the effect of temperature in this study was smaller than the effect of surface roughness.

Finally, Fig. 14 compares the importance of the different parameters in estimating the removal factor. Most (60%) of the variability of the removal factor is still due to variations in surface roughness. Moreover, if the pressure is kept fixed, more than two-thirds of the variability of the removal factor are due to changes in surface roughness.

Element | Series | Atom % |
---|---|---|

Top point 1 EDX results | ||

Sodium | K-series | 62.19 |

Iron | K-series | 0.21 |

Top point 2 EDX results | ||

Sodium | K-series | 62.76 |

Iron | K-series | 0.36 |

Top point 3 EDX results | ||

Sodium | K-series | 64.36 |

Iron | K-series | 0.34 |

Ridge 1 EDX results | ||

Sodium | K-series | 2.13 |

Iron | K-series | 3.52 |

Ridge 2 EDX results | ||

Sodium | K-series | 25.01 |

Iron | K-series | 0.69 |

Ridge 3 EDX results | ||

Sodium | K-series | 4.41 |

Iron | K-series | 3.14 |

Element | Series | Atom % |
---|---|---|

Top point 1 EDX results | ||

Sodium | K-series | 62.19 |

Iron | K-series | 0.21 |

Top point 2 EDX results | ||

Sodium | K-series | 62.76 |

Iron | K-series | 0.36 |

Top point 3 EDX results | ||

Sodium | K-series | 64.36 |

Iron | K-series | 0.34 |

Ridge 1 EDX results | ||

Sodium | K-series | 2.13 |

Iron | K-series | 3.52 |

Ridge 2 EDX results | ||

Sodium | K-series | 25.01 |

Iron | K-series | 0.69 |

Ridge 3 EDX results | ||

Sodium | K-series | 4.41 |

Iron | K-series | 3.14 |

The results obtained for the removal factor as a function of the drying temperature, the drying relative humidity, the pressure applied during the smearing process, and the roughness of the stainless-steel surface, indicate that the relationship in the studied domain is nonlinear (Eq. (8)). This means that the regression model presented above cannot be used to estimate the removal factor. Further experiments and calculations are therefore required.

## 5 Conclusion and Prospects

Understanding the behavior of labile contamination and removal factors on nuclear decommissioning sites is essential to better assess risks of internal exposure for operators. The influence of the type of contamination on the removal factor [2] and the influence of the nature of the wipe [3] and its roughness [4] on the recovery of the contamination have been studied previously. The influence on the removal factor of the pressure applied on the wipe [6,7] and of the recovering solvent [3] have also been investigated, as have the effect of the roughness of the contaminated surface on the adhesion of the contamination [8,9] and the influence of the nature of the surface on the recovery of the contamination [4,5,7,10]. Other factors whose effects on the removal factor have been studied include the concentration of contamination on the surface [11], the adherence to the recommended wiping procedure [7], the drying temperature [11], and the drying relative humidity [11].

A notable feature of this study is the use of experimental design optimization, an approach that has never previously been used to understand the effects of various parameters on the removal factor and proportions of labile contamination. This method provided a quantitative comparison of the effects of the drying temperature, the drying relative humidity, the pressure applied during the smearing process, and the roughness of the stainless-steel surface on the removal factor and on the level of labile contamination. The results of the study show that the roughness of the contaminated surface is the most important factor to consider when measuring the level of labile contamination and the removal factor. The removal factor is also affected to a lesser extent by the pressure applied on the wipe and the temperature. The first step to optimize estimates of the amount of labile contamination and the removal factor should therefore be to determine the roughness of the studied surface and to use a smearing device to control the applied pressure.

The experimental design also allowed us to obtain and validate a linear model for the level of labile contamination. This is the first model of its kind, allowing the labile contamination on a surface to be estimated from just four parameters and the total contamination. For the removal factor, however, more information is required to build a second-order model. This could be achieved using central composite design. This approach typically involves three stages:

Experiments based on an orthogonal design, typically according to a fractional factorial design. This was the approach followed here to study four parameters. Two further steps are required.

Experiments in the middle of the studied domain. Three such experiments were performed in this study, but nine more are required;

Eight new star experiments encoded “+

*α*” and “−*α*”with $\alpha =N4$.

where *N* is the number of measurements performed according to the experimental design without taking into account the experiments at the center of the domain.

These experiments should be performed for different contaminations and different contaminated surfaces and will provide useful information in the course of decommissioning projects to optimize the assessment of internal exposure risks. A validated second-order model for the removal factor would be a tremendous improvement in the nuclear field. While for now, the removal factor is considered fixed at 10%, this model would allow it to be adjusted as a function of the drying temperature, the drying relative humidity, the pressure applied during the smearing process, and the roughness of the contaminated surface, to optimize estimates of the level of labile contamination without knowing the total amount of contamination present.

Moreover, the behavior of fluorescein could be compared to that of ^{137}Cs in hot-lab experiments to better translate the results obtained with the simulant into the expected behavior of the radionuclide on decommissioning sites. These observations should also be completed with on-site measurements of labile contamination during dismantling operation in operators' breathing zones. Finally, the nature of the contaminated surface is also essential to consider and further research is required on other surfaces, for example, linoleum and concrete, which are common materials in nuclear facilities. Warren et al. [10] and Passo et al. [3] found indeed that removal factors vary depending on the nature of the contaminated surface.

## Acknowledgment

The authors thank Virginie Frémy for her help with fluorimetry experiments, Benoît Meilleray and Dr. Vanessa Proust for the SEM observations, and Fabien Frances for his advice and interesting discussions. The authors also thank Amory Di Campo and Thomas Roque for their help with the 3D modeling and printing of the smearing device.

## Funding Data

French Alternative Energies and Atomic Energy Commission (CEA).

Electricity of France (EDF) (Funder ID: 10.13039/501100006289).

## Data Availability Statement

The authors attest that all data for this study are included in the paper.

## Nomenclature

### Acronyms and Abbreviations

- CEA =
French Alternative Energies and Atomic Energy Commission

- EDF =
Electivity of France

- EDX =
energy dispersive X-ray spectroscopy

- exp =
experiment

- IRSN =
radioprotection and Nuclear Safety Institute

- Ra =
roughness, arithmetic average of profile height deviations from the mean line,

*μ*m*Rf*=removal factor, %

- RH =
relative humidity, %

- SEM =
scanning electron microscopy

- S.M.T.S =
welding piping maintenance company

*T*=temperature, °C or °F

*Y*=result of an experiment

### Appendix 1 EDX analysis of rough surface contaminated by a drop of fluorescein and smeared by a wipe

EDX blank on a uncontaminated surface to obtained the background signal from the stainless steel surface.

EDX analysis point: 3 at the bottom, in the surface features, and 3 points at the top of the surface.

EDX results for the different points. The results for iron and sodium, used to calculate the Na/Fe ratio, are listed below.