Radium is present naturally in soil, minerals, groundwater, and can incorporate in foodstuffs, and several common materials, including many used in construction. Exposure to radium isotopes by ingestion can lead to significant committed radiation dose considering their affinity to bone structures as well as rapid series equilibrium of the progeny resulting in several alpha-emitting radionuclides . This radiation dose is known to cause bone cancer . Owing to solubility of radium found naturally in rocks and soil and the alpha-recoil mechanism, it can get into groundwater supplies . In communities where wells are used, drinking water can be an important source of radium ingestion to the public. According to the U.S. Environmental Protection Agency (EPA), long-term consumption of water containing 185 mBq/L (5 pCi∙L−1) radium may cause 44 additional cancer deaths for every million people exposed . The risk increases as the level of radium increases. Therefore, radium isotopes are considered listed contaminants in drinking water in many countries. From the regulatory health perspective, 226Ra (T1/2 = 1600 y) from the uranium series and 228Ra (T1/2 = 5.75 y) from the thorium series are the most significant radionuclides. 224Ra is already included in the thorium series as a progeny of 228Ra; it has also been studied separately as it can deviate from the series equilibrium  . In detailed studies, 223Ra from the actinium series has also been investigated , and its main significance is in geochemistry. For these reasons, we investigate 226Ra and 228Ra only in this work.
There exist several methods for the determination of radium isotopes in water, the ones utilizing gamma spectrometry with Germanium (Ge) detectors are the most efficient among them. The sensitivity of radium determination by gamma spectrometry depends on several sample-related factors, such as sample volume , preconcentration on Mn-based medium  , coprecipitation with BaSO4  , coprecipitation with PbSO4  or nondestructive analysis . On the instrumental side, the sensitivity depends on the efficiency of the Ge spectrometer, extent and purity of shielding materials, use of active cosmic-muon rejection , Compton suppression , as well as counting time.
Under the Safe Drinking Water Act, the U.S. Environmental Protection Agency (EPA) has promulgated maximum contaminant level goals (MCLGs) as zero for radioactivity in drinking water. However, the maximum contamination level (MCL) for combined 226Ra and 228Ra in drinking water is equal to 185 mBq/L (5 pCi∙L−1) and both radium isotopes are required to be measured separately .
Among many methods for Ra determination, it is sometimes not clear how sensitive they are. In this work, we use the Currie’s detection limit  as a measure of detection capability, also referred to as the minimum detectable activity (MDA) by the National Environmental Laboratory Accreditation Conference Institute Standard . On the other hand, the EPA approved methods for radium require regulatory detection limit (DL) of 37 mBq/L (1 pCi∙L−1) as a measure of detection capability . This required DL is only five times smaller than the MCL. Even if a given detection capability can be calculated from the data, it does not necessarily follow that this capability can be achieved in repetitive analytical measurements. In order to verify the achieved DL, a Chi-square test has been recommended .
The purpose of the present work was to develop a method for the determination of two major natural radium isotopes: 226Ra and 228Ra in drinking satisfying the regulatory requirements. At the New York State Department of Health, we have been performing low-background gamma spectrometry for more than three decades laboratory . The applied uses of gamma spectrometry include monitoring of environment, food, air, and water, and surveillance of energy and manufacturing facilities, as well as health physics and radiological emergency applications. Low-background gamma spectrometry is applicable to very low activity matrices, such as water or chemically separated samples. Speciﬁc projects include mandated analysis of radium in drinking water and monitoring of radioactive iodine and cesium at nuclear facilities. In this work we have investigated two radium detection methods in drinking water, referred as destructive and non-destructive. We described in detail the achieved MDA and DL using the newly developed procedures , supported by the newly developed methods for Chi-square testing .
2. Experimental and Calculation Procedures
2.1. Gamma Spectrometry
In this study, we have utilized p-type high purity Germanium (HPGe) detectors having 102% relative efficiency (Model GC10020) and 134% relative efficiency (Model GC13021), as well as an XtRa-type 140% Ge detector (Model GX13023), all by Mirion Technologies (Canberra) Inc. (Meriden, CT, USA). The spectrometers are situated inside a 15-cm-thick wall steel room made of pre-World War II steel (Dixie Manufacturing Co., Baltimore, MD, USA), which is located under a 47-story building providing 33 meters of water-equivalent (mwe) shielding from cosmic rays in the vertical direction. The 102% and 134% detectors have general low-background lead shields and are equipped with custom active muon shields made of plastic scintillators positioned on the top, which reduce cosmic muon background by a factor of 2. The 140% detector has a custom 3-layer ultra-low background lead shield of a 17 cm total thickness, which is surrounded by plastic scintillators for muon rejection . This spectrometer has an integrated background rate of 2.4 counts per min (cpm) in the gamma-ray energy range between 50 - 2700 keV, corresponding to a background rate of 15 counts ks−1∙kg−1 Ge. This background compares well with the lowest-background gamma spectrometers in the world .
Considering the facilities at this institution, we refer to low-background gamma spectrometry when using the 102% and 134% detectors, and to ultra-low background gamma spectrometry when using the 140% detector.
2.2. Radium Detection Methods
In this work we investigate two radium detection methods, referred to as destructive and nondestructive. In the destructive method, we start with a 2.5 L of water sample and coprecipitate radium isotopes with BaSO4 carrier with an addition of 133Ba recovery tracer, as reported elsewhere  . The precipitate was settled, centrifuged, dissolved in ethylenediaminetetraacetic acid (EDTA) solution, and transferred to a 50-mL counting jar [Figure 1(a)] calibrated for gamma spectrometry. After allowing 4 weeks for to reach secular equilibrium with 222Rn and progeny, 226Ra was measured and assayed by means of a 609.3-keV gamma peak from 214Bi (Iᵧ = 0.4549). While it is possible to use and average other peaks of lower intensity, we did not include them because they are often poorly defined in the spectrum for low-level applications. It is also possible to degas 222Rn prior to its ingrowth from the sample , which enables calculation of ingrowth factor and shorter waiting time. This procedure, however, decreases the sensitivity while earlier reporting has not been necessary for radium monitoring at this institution. Therefore, we waited 30 days for full equilibrium. 228Ra activity was assayed using a 911.2-keV gamma peak from its progeny, 228Ac (Iᵧ = 0.258), which reaches equilibrium in less than 3 days. Similarly, we did not include less intense peaks. The chemical recovery was determined with a 133Ba tracer (T1/2 = 10.6 y), using a 356.0-keV gamma peak (Iᵧ = 0.6205). This peak and its Compton continuum do not interfere with the detection of either 214Bi or 228Ac. The measured chemical recoveries for 133Ba tracer varied between 0.75 and 0.96.
Figure 1. (a) 50-mL jar for destructive, and (b) 800-mL Marinelli beaker for non-destructive, 226Ra and 228Ra measurements.
In the nondestructive method, 800 mL of the sample was transferred to 800-mL Marinelli beaker [Figure 1(b)] without any processing and counted after waiting 30 days for equilibrium. The counting time for gamma spectrometry measurements of either 50-mL or 800-mL geometry was 1000 min for the sample spectrum and 4000 min for the background spectrum.
The absolute efficiencies for the 50-mL and 800-mL geometries were calibrated using a mild acidic solution spiked with a mixed-gamma standard traceable to the National Institute of Standards and Technology (NIST, Gaithersburg, MD, USA) as described elsewhere . Since the calibration standards and samples were water solutions, no density corrections were applied. The coincidence-summing corrections calculated using the Gespecor software  ranged from 0.95 to 0.96.
2.3. Sample Containers
It is well recognized that radon distributes between air and water by a ratio of 3 to 1, according to Henry’s law . Therefore, to maintain counting geometry for the 50-mL and 800-mL containers described in Section 2.2, they must be filled with water with no air gap above the solution. This creates practical problems with filling procedure and leaking of both radon and water. Several techniques were developed for sealing of counting containers using a variety of sealants  .
In this work, we modified common sample containers in house to prevent leaks. The 50-mL jar screw lid (white in [Figure 1(a)]) was made of Teflon, and an aluminized washer was inserted between the lid and the container. Then, the rim of the lid was sealed with a silicone rubber. The solution entry port on the top had a screw cap with a Teflon washer. A standard 800-mL Marinelli baker’s lid was modified by thermal fusing two ports to it with screw caps equipped with Teflon washers [Figure 1(b)]. One port is for filling the solution, and the second port for degassing. The lid was then sealed to the beaker with the silicone rubber.
A radon leak test was performed on the smaller container, which was filled with water saturated with 222Rn gas from a historical radium-ore Revigator . After establishing equilibrium with the progeny, the decay of radon was followed on a gamma spectrometer for 45 days.
2.4. Calculation Procedures
Define As described in the Introduction, in this work we study quantitative measures of sensitivity of 226Ra and 228Ra detection in drinking water. One such measure is MDA  . We calculated MDA using Genie 2000 algorithms in the gamma spectrometry software (Mirion Technologies (Canberra), Inc.). Another measure of sensitivity is the EPA’s DL defined as a “concentration which can be counted with a precision of plus or minus 100 percent at the 95 percent confidence level (1.96σ, where σ is the standard deviation of the net counting rate of the sample)  ”. The formula for calculation of DL is given in  for paired counting. In gamma spectrometry, its calculation is more complicated as it involves gamma peak fitting and subtraction of background spectrum . We have recently revised the formula for DL by better quantifying Poisson fluctuations in the counting data  and use it in this investigation.
It has been guided by EPA (2017)  to verify detection performance at the required DL using a Chi-square test. We provide the verifications at the regulatory DL of 37 mBq/L (1 pCi∙L−1) as well as MCL of 185 mBq/L (5 pCi∙L−1). From 7 to 10 samples were spiked with NIST-traceable standards of 226Ra and 228Ra at activities close to the above levels. Then either destructive or nondestructive procedures were followed. The results were evaluated with a combined location/variance right-tail (RT) Chi-square test at 0.01 significance (99% confidence level, CL), as described elsewhere . The observed Chi-square variable is defined as , where is the measured activity, is known spike activity, and is the measured uncertainty, for , where n is the number of measurements. The calculated Chi-square is performed using standard algorithm in Excel (Microsoft) for n degrees of freedom, since there are no constraints in the data.
Another measure of verification is the En score defined as , where is mean measured activity, is sample standard deviation, is spike standard deviation, and is the coverage factor for 95% CL .
3. Results and Discussion
The results of radon leak test from the 50-mL jar are shown in [Figure 2]. The measured counting rates are depicted as points. The fitted curve consisted of an exponential term and a constant term for background. The fit resulted in the best estimate for 222Rn half-life of 3.8079 days, slightly smaller than the literature value of T1/2 = 3.8235 days , indicating at most a minor leak. However, it affects the relative saturation factor in 30 days by about 10−4 and, therefore, can be
Figure 2. 222Rn decay curve for leak testing of the 50-mL container.
neglected. Radon leak has not been explicitly tested on the 800-mL Marinelli beaker. Nevertheless, the verification experiments described below do not indicate any leaks.
The results of the DL calculations are given in Table 1, which is subdivided into two major groups by destructive and nondestructive methods. The DL results are given in mBq/L. They were obtained for 2.5 L of sample, 1000 min sample counting time, 4000 min background counting time, and 30-day radon ingrowth period. The DL values were calculated using statistical methods on statistical data, including variability in chemical recovery for the destructive method, therefore some variability in the results is expected.
It is seen that, for the destructive method and low-background detectors (102% and 134%), the DL values range from 14.9 to 22.6 mBq/L for 226Ra from 28.4 to 30.4 mBq/L for 228Ra. All these values are below 37 mBq/L, therefore they satisfy the regulatory DL under these experimental conditions for both radionuclides. Significantly lower detection limits were calculated for the destructive method and the ultralow-background detector (140%) with the DL values of 5.3 and 6.2 mBq/L for 226Ra as well as 7.4 and 9.7 mBq/L for 228Ra. Since they are significantly lower than the required DL, there is room for decreasing of the sample quantity and decreasing counting time, as reported earlier , or for shorter radon ingrowth time. Also given in Table 1 are the MDA results. It is seen that they are slightly higher than DL, which is the result of different definition.
Within the nondestructive-method group in Table 1, none of the low-background detectors satisfied the required DL, with the values between 68.1 and 77.9 mBq/L for 226Ra as well as 80.7 and 97.7 mBq/L for 228Ra. These
Table 1. Detection limits for 226Ra and 228Ra in method blanks by low-level gamma spectrometry (T-top muon shield, F-full muon shield).
results could be lowered by increasing sample volume to over 4 L, as done by earlier study , and/or by increasing sample counting time above 1000 min. However, using the ultralow-background gamma spectrometry we obtained satisfactory results of 26.0 and 26.9 mBq/L for 226Ra as well as 27.6 and 28.6 mBq/L for 228Ra. In addition, the nondestructive method used only 0.8-L sample instead of 2.5 L.
The performance of the developed methods was tested by spiking 7 to 10 samples simultaneously with 226Ra and 228Ra at the levels close to the DL or MCL in separate experiments. The actual spiking activities and their standard deviations are given in the heading of Table 2 for the destructive method and of Table 3 for the nondestructive method. The measured mean activities and their standard deviations are reported in Table 2 and Table 3 as well.
A simple measure of performance verification is the En score, as defined in Section 2.4. It is seen that for both destructive and nondestructive methods, the En scores are between −1.0 and +1.0. This verifies the performance of both methods at the two spiking levels, at 95% CL. A more sophisticated performance verification is the Chi-square test. A version of Chi-square was adopted which tests for a combined location and variance, as described in Section 2.4. The observed Chi-square values were below the calculated Chi-square values for all cases including two radium isotopes, destructive and nondestructive methods, as well as two spiking levels, as listed in Table 2 and Table 3. Therefore, the Chi-square tests passed at 99% CL, and the methods in question are considered verified.
Table 2. Verifications of 226Ra and 228Ra determinations using destructive low-level gamma spectrometry (T-top muon shield, F-full muon shield).
Table 3. Verifications of 226Ra and 228Ra determinations using non-destructive ultralow-level gamma spectrometry for 140% detector with full muon shield.
4. Summary and Conclusions
In this work, we have developed a method for the determination of both 226Ra and 228Ra for the purpose of regulatory radium monitoring in drinking water. We used two approaches to the method: destructive, based on coprecipitation of barium sulfate, as well as nondestructive. Both versions utilized gamma spectrometry. While coprecipitation and gamma spectrometry have been studied extensively in the literature, the sensitivities of many investigations vary and/or are difficult to judge. Therefore, we paid emphases to rigorously quantifying the results in terms of DL and MDA. In addition, we used statistical measures such as En score and Chi-square test to verify performance at DL and MCL.
It was shown that by using a 2.5-L sample quantity, 1000 min of sample counting, 4000 min of background counting, and allowing for a full 30-day radon ingrowth, we were able to achieve the regulatory DL of 37 mBq/L (1 pCi∙L−1), as stipulated by the U.S. EPA, with the destructive method which involved low-background gamma spectrometry. The DL obtained using the ultralow-background gamma spectrometry is significantly lower than the required DL. Therefore, in the ultralow-background gamma spectrometry technique, there is room for decreasing of the sample quantity and/or decreasing sample counting time, or for shorter radon ingrowth time.
For the nondestructive version of the method, only ultralow-background gamma spectrometry was able to achieve the regulatory DL, and as such for 0.8-L sample volume only, whereas the low-background gamma spectrometry failed to do so under the experimental conditions studied.
The verifications performed at two spiking levels: around DL and MCL were followed for the destructive method as well as low-background and ultralow-background gamma spectrometry. For the nondestructive method, the verifications were performed using ultralow-background gamma spectrometry only. All the verifications described above passed using sophisticated statistical tests such as En score as well as location/variance Chi-square.
The methods developed here are judged fit for radium monitoring as required by the SDWA.
Thanks are due to Dr. M. Kitto for valuable suggestions.
 BEIR IV (1988) Health Risks of Radon and Other Internally Deposited Alpha-Emitters. Committee on the Biological Effects of Ionizing Radiations, National Academy of Sciences, National Research Council, Washington DC.
 Szabo, Z., de Paul, V.T., Fischer, J.M., Kraemer, T.F. and Jacobsen, E. (2012) Occurrence and Geochemistry of Radium in Water from Principal Drinking-Water Aquifer Systems of the United States. Applied Geochemistry, 27, 729-752.
 Parekh, P., Haines, D., Bari, A. and Torres, M. (2003) Non-Destructive Determination of 224Ra, 226Ra, and 228Ra Concentrations in Drinking Water by Gamma Spectroscopy. Health Physics, 85, 613-620.
 Parsa, B., Obed, R.N., Nemeth, W.K. and Suozzo, G. (2004) Concurrent Determination of 224Ra, 226Ra, 228Ra, and Unsupported 212Pb in a Single Analysis for Drinking Water and Wastewater: Dissolved and Suspended Fractions. Health Physics, 86, 145-149.
 Condomines, M., Rihs, S., Lloret, E. and Seidel, J.L. (2010) Determination of the Four Natural Ra Isotopes in Thermal Waters by Gamma-Ray Spectrometry. Applied Radiation and Isotopes, 68, 384-391.
 Michel, J., Moore, W.S. and King, P.T. (1981) γ-Ray Spectrometry for Determination of Radium-228 and Radium-226 in Natural Waters. Analytical Chemistry, 53, 1885-1889.
 Kahn, B., Rosson, R. and Cantrell, J. (1990) Analysis of 228Ra and 226Ra in Public Water Supplies by a γ-Ray Spectrometer. Health Physics, 59, 125-131.
 GIT (2004) Method for the Determination of Radium-228 and Radium-226 in Drinking Water by Gamma Ray Spectrometry Using HPGE or Ge(Li) Detectors. Environmental Resource Center Report V.1.2, Georgia Institute of Technology, Atlanta, GA.
 Khan, A.J., Semkow, T.M., Beach, S.E., Haines, D.K., Bradt, C.J., Bari, A., U.-F. Syed, U.-F., Torres, M., Marrantino, J., Kitto, M.E., Menia, T. and Fielman, A. (2014) Application of Low-Background γ-Ray Spectrometry to Monitor Radioactivity in the Environment and Food. Applied Radiation and Isotopes, 90, 251-257.
 James, W.D., Boothe, P.N. and Presley, B.J. (1998) Compton Suppression Gamma-Spectroscopy in the Analysis of Radium and Lead Isotopes in Ocean Sediments. Journal of Radioanalytical and Nuclear Chemistry, 236, 261-265.
 TNI (2016) Quality Systems for Radiochemical Testing. TNI Standard EL-V1M6 2016-Rev2.0, Vol. 1, Module 6, The National Environmental Laboratory Accreditation Conference Institute, Weatherford, TX.
 Semkow, T.M., Khan, A.J., Menia, T.A., Li, X., Chu, L.T., Torres, M.A. and Bari, A. (2019) Detection Limit for Ra-228 in Drinking Water by Gamma Spectrometry. In: Brisson, M.J., Ed., Detection Limits in Air Quality and Environmental Measurements, American Society for Testing and Materials International, West Conshohocken, PA, 146-160.
 Semkow, T.M., Freeman, N., Syed, U.-F., Haines, D.K., Bari, A., Khan, A.J., Nishikawa, K., Khan, A., Burn, A.G., Li, X. and Chu, L.T. (2019) Chi-Square Distribution: New Derivations and Environmental Application. Journal of Applied Mathematics and Physics, 7, 1786-1799.
 Cagniant, A., Douysset, G., Fontaine, J.-P., Gross, P. and le Petit, G. (2015) An Introduction to γ3 a New Versatile Ultralow Background Gamma Spectrometer. Background Description and Analysis. Applied Radiation and Isotopes, 98, 125-133.
 Semkow, T.M., Bradt, C.J., Beach, S.E., Haines, D.K., Khan, A.J., Bari, A., Torres, M.A., Marrantino, J.C., Syed, U.-F., Kitto, M.E., Hoffman, T.J. and Curtis, P. (2015) Calibration of Ge Gamma-Ray Spectrometers for Complex Sample Geometries and Matrices. Nuclear Instruments and Methods in Physics Research A, 799, 105-113.
 Sima, O., Arnold, D. and Dovlete, C. (2001) GESPECOR: A Versatile Tool in Gamma-Ray Spectrometry. Journal of Radioanalytical and Nuclear Chemistry, 248, 359-364.
 Nielson, K.K. and Rogers, V.C. (1982) A Mathematical Model for Radon Diffusion in Earthen Materials. Report NUREG/CR-2765, U.S. Nuclear Regulatory Commission, Washington DC.
 Parekh, P.P., Bari, A., Semkow, T.M. and Torres, M.A. (2002) A New Method for Sealing Containers with Liquid Samples for Radioactivity Measurements. Journal of Radioanalytical and Nuclear Chemistry, 253, 321-325.
 Bonczyk, M. and Samolej, K. (2019) Testing of the Radon Tightness of Beakers and Different Types of Sealing Used in Gamma-Ray Spectrometry for 226Ra Concentration Determination in NORM. Journal of Environmental Radioactivity, 205-206, 55-60.
 Kitto, M.E., Judd, C.D., Torres, M.A. and Bradt, C.J. (2013) Radionuclide and Chemical Hazards of a Radium Ore Revigator. Journal of Radioanalytical and Nuclear Chemistry, 296, 57-62.
 ISO (2015) Statistical Methods for Use in Proficiency Testing by Interlaboratory Comparison. International Standard ISO 13528:2015(E), Corrected Version 2016-10-15. International Organization for Standardization, Geneva.
 NNDC (2019) NuDat Version 2.7. National Nuclear Data Center, Brookhaven National Laboratory, Upton, NY.