RaDoM: a lung dosimeter for radon progeny

This paper discusses a novel Timepix-based on-line radon dose monitor called RaDoM. The detector provides a direct estimate of the effective dose to the lung via a system of filters having the same absorption characteristics as this organ and a measurement of the PAEC (the potential alpha energy concentration). Measurements performed in a reference radon chamber and in normal environmental conditions showed that RaDoM well follows changes in the radon concentration and is accurate in both long- and short-term measurements. The paper describes the detector and associated software for data acquisition and analysis, and compares the measurement of the lung dose with the value derived from a measurement of the radon concentration.

measurements, as far as dose assessment is concerned, this paper discusses a novel on-line radon dose monitor (RaDoM). The detector has a response that is directly proportional to the effective dose. This result is obtained by 1) sampling air through to a system of filters having a particulate absorption characteristic that mimics the one of the extra-thoracic airways, and 2) by measuring the potential alpha energy concentration (PAEC). The RaDoM is the evolution of two radon detection systems called RADONPIX and RADONLITE developed at CERN in the past few years [5] and shares with them the Timepix hybrid silicon detector as sensor.

The RaDoM
RaDoM is made of a Timepix sensor (see for instance ref. [6] and references therein quoted), a system of aerosol filters and a pump for air sampling. The pump draws the air and the radon decay products through the system of filters. The Timepix counts the alpha particles from the radon daughters attached to the filter. In order to obtain a direct estimate of the effective dose, the system of filters was designed so as to have absorption characteristics similar to those of the respiratory tract. RaDoM responds well to changes in environmental conditions and is accurate in both longand short-term measurements. Its response is not affected by natural background radiation thanks to the Timepix pattern recognition capability and to the cluster analysis algorithm developed in this work. Furthermore, RaDoM solves the 210 Po contamination problem, which was a major issue with RADONPIX and RADONLITE [ figure 1] as well as with most commercial active radon monitors. 210 Po is a long-lived radon decay product with a half-life of 22 years. In RaDoM the decay products attach to the filter, which can be easily replaced, rather than to the detector surface, as it occurs in existing radon monitors.

Detector design
The working principle and the geometrical configuration of RaDoM is sketched in figure 2. The pump draws the air containing radon and the radon decay products through the system of mesh screens. Depending on the type of mesh and the chosen configuration of the system, either single or four mesh screens, it is possible to build different versions of the dosimeter. The paper filter, placed above the Timepix silicon sensor, stops the radon decay products that have not been collected by the mesh screens. The alpha particles from the radon daughters attached to the filter are then counted by the Timepix. The size distribution of the particles collected on four 400-mesh (400 holes per inch, corresponding to a wire diameter of 27 µm) screens matches the dose conversion factor (DCF) [7] curve of the tracheobronchial region, which has high susceptibility to radiation-induced cancer, with a response proportional to the effective dose. Figure 3 shows the final design of the RaDoM. The materials are PVC for the case and aluminum for the filter holder. The paper filter is a Millipore AA 0.8 µm thick, chosen for its high collection efficiency for radon progeny (99.99%) [8]. The pump is a NMP 05 M model produced by KNF in Germany [9], with maximum air volume sampling rate of 0.3 l/min. The pump is connected to the detector though the central hole in the upper side of the detector. The others holes are for the 3 mm screws used to tighten the various parts of the RaDoM together. The filter drawer (dark grey in figure 3) keeps the Millipore filter exactly aligned to the silicon sensor at a distance of 13.4 mm. Figure 4 shows the filter fixing mechanism. The holder (A) fixes the Millipore filter (B) to the drawer (C). A standard 3 mm threaded hole (D) allows to easily -2 -

JINST 13 P07030
extract the drawer from the RaDoM in order to change the filter without disassembling the detector. Figures 5 and 6 show exploded and assembled views of the detector prototype. By removing the mesh screens, RaDoM can be used as a PAEC detector, in which configuration it counts the alpha particles from the radon progeny that attached to the filter surface. Thanks to the high collection efficiency of the filter, almost all radon decay products are stopped [8] and their alpha particles counted. In radon monitoring, the potential alpha energy concentration is mostly expressed in Working Level [WL]. The WL is defined as any combination of short-lived radon products in one litre of air that will result in the ultimate emission of potential alpha particle energy equal to 1.3 × 10 5 MeV. In air, this energy is released when an equilibrium concentration of 3700 Bq/m 3 is present. Using expression 2.1 [10],  The PAE (potential alpha energy) is the total energy emitted in the alpha decay of the radionuclide to 210 Pb. Therefore it is 13.69 MeV (6.0 MeV+ 7.69 MeV) for 218 Po and 7.69 MeV for 214 Pb and 214 Bi. In the RaDoM, the Timepix counts the alpha particles coming from both 218 Po and its decay product 214 Po, because both radon daughters remain attached to the filter surface. This leads to an increase in the 214 Po counts. By using a value of 6.0 MeV for the PAE of 218 Po rather than 13.69 MeV the real value of the PAEC is obtained. The RaDoM geometrical efficiency is 15.6%, defined by the ratio between the solid angle between the filter and the silicon surface and 4π. A second prototype, which uses a quad Timepix, was also developed. The quad Timepix has four chips and 2.8 · 2.8 cm 2 active area. This prototype has a geometrical efficiency of 33.2% thanks to the larger sensor area and shorter distance between the filter and the silicon surface.

Acquisition and analysis software
In the Timepix detector, different types of ionizing radiation create different types of clusters because of their different dE/dx. Thereby by analysing specific properties of the cluster, such as its morphology or the energy released, the cluster can be classified and assigned to a specific particle category. A software code in Python language for offline analysis, PyPix, was written. A software for online analysis, RADON MONITOR, has been described in a previous paper [5]. The principle of the PyPix algorithm is as follows. A loop over the pixel matrix looks for aggregate pixels that have a much higher number of hits than the average; the found clusters are compared with the default parameters registered in the algorithm. For example, circular clusters with eccentricity up to 0.1 are classified as heavy blobs, random paths are categorized as curly tracks, linear paths as straight tracks, etc. The eccentricity is a parameter associated with a conic section. It can be thought of as a measure of how much a conic section deviates from being circular (the eccentricity of a circle is zero). Figure 7 shows a one-minute frame acquired in the radon chamber at Politecnico di Milano with the RaDoM. Dust inside the chamber has been generated by the combustion of incense. In figure 7 different kinds of tracks are distinguishable, such as heavy blobs, curly tracks, dots and straight tracks. The algorithm recognizes and categorizes the clusters. Comparing the ToT (Time over Threshold) counts (which correspond to the energy deposition if the Timepix is calibrated) and the size of the cluster with the kind of track, the algorithm assigns a specific particle to each cluster. For example, heavy charged particles lose energy in silicon according to the Bethe-Bloch equation, and therefore an alpha particle creates a heavy blob because of its short range. An electron loses its energy by collisions or by Bremsstrahlung so its path is random, and this leads to the creation of a curly track. Therefore in figure 7 cluster number 1 is a muon, cluster 2 is an alpha particle, cluster 3 is an electron and cluster 4 is a photon. In particular the overlapping alpha particle and electron (cluster 5) is due to the decay of 214 Po, which occurs 164 µs after the decay of 214 Bi producing a blob (alpha particle) with a tail (beta particle). In fact, some of the -6 -particles attach to the sensor surface rather than to the Millipore filter because most of the dust generated by combustion is electrically charged [11] and it is attracted by the electric field generated by the detector bias voltage.Thanks to the spectroscopic capabilities of the Timepix, the script distinguishes the alpha particles coming from 218 Po, 214 Po and 210 Po, and removes the background. Some examples of the PyPix interface are shown in figures 8 and 9. Figure 8 is a scatter plot of energy deposition as a function of the cluster size for all particles (A) and for alpha particles (heavy blobs) only (B): it is also possible to distinguish the two peaks due to 218 Po and 214 Po both in the energy and in the size spectra. Figure 9 shows the counts of 218 Po (blue) and 214 Po (red) as a function of time with their statistical uncertainties. WALKABOUT [12] is a python-based server we have developed, to which the RADON MONITOR plugin can be connected. The functioning principles of the server are schematized in figure 10. The RaDoM is connected by an USB cable to a computer running the Pixelman [13] software and the RADON MONITOR plugin [5]. RADON MONITOR displays the results of the measurement in real time, saves the data and sends them to the WALKABOUT server via internet, Wi-Fi or locally if the server is installed on the same computer -7 - Thanks to this system, it is also possible to remotely connect to WALKABOUT using portable devices, such as a smartphone or a tablet, observing the results of the measurement in real-time from anywhere (figure 10). The present results indicate that RaDoM is a good candidate for realtime radon dose monitoring, due to the features of the Timepix detector combined with a proper design of the instrument and an appropriate cluster analysis and pattern recognition algorithm. The three different software solutions, RADON MONITOR, WALKABOUT and PyPix make RaDoM user-friendly, in spite being a rather powerful and sophisticated analysis tools.

Effective dose measurements
RaDoM in the dose configuration was first tested in the calibration laboratory of the Radiation Protection group at CERN [16], where an ambient radon concentration of about 100 Bq/m 3 is present. The duration of the measurement was 15.45 days. The configuration chosen was four copper mesh screen in series with a wire diameter of 30 µm. According to Cheung et al. [7] and Hopke et al. [17], four 400-mesh screens simulate the collection characteristic of the tracheobronchial region. In order to count the alpha particles from the radon decay products attached to the four 400-mesh screen, parallel measurements with two detectors are needed, as the Timepix is only able to count the alpha particles coming from the nearest mesh. Figure 11 shows the set-up of the two detectors for the simultaneous measurements. Detector A has the four 400-mesh screen whereas detector B only has the Millipore filter. The latter is RaDoM in the PAEC configuration with the quad Timepix chip. Figure 12 shows the detectors A and B during measurement in the calibration laboratory at CERN. The collection efficiency [7] of the wire screen series is obtained by expression 3.1: -9 -2018 JINST 13 P07030 Figure 10. Functioning principle of the WALKABOUT server.
-10 - PAEC A and PAEC B (in Working Level) are calculated by expression 2.1. The DCF for this configuration is 2.66 · 10 −4 mSv/(s·WL) [7]. Therefore the effective dose is obtained by expression 3.2, where t is the duration of the measurement in seconds and ∈ geometry the geometrical efficiency of the detector. The integrated effective dose over the 15.45 days of exposure is 0.0283 ± 2.3% mSv.
The uncertainty is assessed composing: (a) the statistical uncertainty supposed Poisson distributed (Type A), (b) the systematic (Type B) uncertainty in the collection efficiency on the wire mesh -11 -

JINST 13 P07030
screen (12.4%) and (c) the systematic (Type B) uncertainty associated to the unattached fraction of PAEC (8%) [7]. Since the two systematic uncertainties are not correlated, the overall systematic uncertainty is 14.7% (quadratic sum of uncertainties). The effective dose calculated from the radon concentration is 0.0278 ± 0.0018 mSv assuming an indoor occupancy factor of 0.2 (workplace) and an equilibrium factor of 0.4 (dwellings). The effective dose was calculated by using the value of the committed effective dose rate E [4,18], expression 3.3, multiplied by the exposure time.
Here C Rn is the radon concentration expressed in Bq/m 3 , D is the conversion factor equal to 9.0 · 10 −6 mGy/h per Bq/m 3 [4], F is the equilibrium factor equal to 0.4 (indoor radon), T is the indoor occupancy factor equal to 0.2 for workplaces (2000 h/y) or 0.8 (7000 h/y) for domestic places, W R = 20 and W T = 0.12 [19] are the radiation weighting factor for alpha particles and the tissue weighting factor for lung, respectively. The result of the present measurement is in good agreement with the value calculated from the radon concentration (the standard method).

Test in radon chamber
Inside the radon chamber the effective dose is generally lower than in a real environment for the same concentration. This is because of two phenomena. First, the value of the equilibrium factor is much lower than 0.4 due to the small volume of the chamber (1 m 3 ) and because the radon progenies tend to stick to the walls of the chamber rather than lingering in the air. Second, the air inside the chamber is filtered twice, so it is dust-free. For these reasons the radon chamber is not the ideal place to test the RaDoM, since the chamber lacks a "dust generator". In order to generate particulate, incense was burnt inside the chamber. This certainly cannot give precise information on the capability of RaDoM to determine the dose, because the generated dust is very different from the particulate found in a normal environment. However this test was useful to understand the response of the detector to changes in the amount of dust. Figure 13 shows the measurement set-up. Figure 14 shows the effective dose rate estimated directly with the RaDoM and calculated from the radon concentration assuming the standard equilibrium factor of 0.4. As soon as incense started to burn, the counts of the radon decay products started to increase. After 35 hours the counts rate began to decrease because the generated particles started to deposit. During the first 21 hours the effective dose rate estimated directly by the RaDoM is about 5 times lower than the effective dose rate calculated from the radon concentration, due to the low value of the equilibrium factor between the radon and its progeny (about 0.1 rather than 0.4) and the amount of dust inside the chamber. As soon as the incense begins to burn (hour 21) the effective dose rate estimated by the RaDoM reaches the value of the dose rate calculated from the radon concentration and then decreases when the generated particles start settling (after 35 hours). In figure 14, the error bars associated to the dose evaluation based on the radon concentration (red line) have been assessed from the concentration uncertainty as reported by the Alphaguard [20]. The error bars associated to the dose measured by the RaDoM include type A and type B uncertainties as described in section 3.1. The effective dose over 45 hours obtained by the RaDoM and from the radon concentration are 0.81 ± 0.037 mSv and 1.86 ± 012 mSv (type A uncertainties), respectively.

Conclusions
The first experimental results indicated that the RaDoM has a good capability of assessing the effective dose in situations where the environmental conditions are characterized by a standard equilibrium factor (F=0.4) and a standard unattached fraction of the PAEC (8%). On the other hand, in environments like the radon chamber characterized by a low equilibrium factor, the RadoM correctly assesses a lower effective dose with respect to the value calculated from the radon concentration. By changing the particulate concentration, the RaDoM response varies accordingly. The RaDoM has proved his capability to behave as a dosemeter with a response function that is reasonably independent of the concentration of the airborne particulate. It is worth underlining that the RaDoM can perform long-term measurements and can be useful in borderline situations where the radon concentration is close to the action level and a dose assessment can be needed. The RaDoM can also play a key role in situations where the equilibrium factor is expected to change during the measuring period [21]. Another feature of RaDoM is that it avoids the 210 Po contamination, which is a major problem with most active radon monitors like the ALPHAGUARD, as well as in RADONPIX and RADONLITE. In RaDoM the radon decay products attach to the filter surface, which can be easily replaced, rather than to the detector. However the problem recurs partially when the particles are electrically charged, such as in generation by combustion, because they are attracted by the positive charge of the silicon sensor. By setting the aluminum filter drawer at the same potential as the silicon sensor the contamination problem will be fully resolved. Future improvements of the dosimeter entails the replacement of the Timepix by a single silicon diode. This reduces the manufacturing cost of the detector. By using a silicon diode it is also possible to build a stand-alone device, integrating a microcontroller and a display unit, thus overcoming the need of a computer connected to the detector.