Ambient humidity, the overlooked influencer of radioactivity measurements

Data from a temperature (T) and relative humidity (H_r) monitor inside the radionuclide metrology laboratory 35 A in building 010 of the JRC in Geel were extracted to calculate the absolute humidity (H_a) through the ideal gas equation [40]

$$\begin{equation}{H_{\text{a}}}[{\text{g}}\,{{\text{m}}^{ - 3}}] = \frac{{2.1674\,{P_{\text{s}}}[{\text{hPa}}]\,{H_{\text{r}}}[\% ]}}{{273.15 + T[^\circ {\text{C}}}]}\end{equation} \tag{ 1 }$$

in which the quantities must be expressed in the indicated units, and P_s represents a heuristic formula for the saturation vapour pressure above water as a function of temperature [41]

$$\begin{equation}{P_{\text{s}}}[{\text{hPa}}] = 6.112\exp \left( {\frac{{17.67\,T[^\circ {\text{C}}]}}{{243.5 + T[^\circ {\text{C}}}]}} \right).\end{equation} \tag{ 2 }$$

The equation (2) is considered accurate enough (≈0.1% for T between −30 °C and +35 °C) without applying an enhancement factor (for moist air compared to pure water vapour, as a function of atmospheric pressure) or a different approximation below 0 °C [42, 43].

Starting from the relationship H_r/100 = P_s(T_D)/P_s(T), the dew point T_D can be derived from the measured temperature and relative humidity through [44]

$$\begin{equation}{T_{\text{D}}}\left[^\circ {\text{C}}\right] = \frac{{243.5\,X\left({H_{\text{r}}},T\right)}}{{17.67 - X\left({H_{\text{r}}},T\right)}}\end{equation} \tag{ 3 }$$

with

$$\begin{equation}\,X\left({H_{\text{r}}},T\right) = \left( {\ln \left( {\frac{{{H_{\text{r}}}[\% ]}}{{100}}} \right) + \frac{{17.67\,T[^\circ {\text{C}}}]}{{243.5 + T[^\circ {\text{C}}}]}} \right).\end{equation} \tag{ 4 }$$

These data measured inside the laboratory in Geel will be compared with external data from a weather station [45] in the Antwerp airport, located about 43 km away.

The JRC laboratories use an air circulation system that establishes underpressure in certain containment areas and regulates the temperature typically within (21 ± 1) °C. However, there have been some variations in the system's settings over the years, as well as exceptional periods where the system (or sensors) did not function adequately. Figure 1(a) displays the relationship between inside and outside daily average temperatures over the past decade. The inside temperature remains relatively stable during colder days, but there is a slight positive quadratic trend overall. Despite the strong correlation between activity measurements in the ionisation chamber IC1 [1, 2, 46, 47] and outside temperature, they are only weakly correlated with the inside temperature. Therefore, the inside temperature is not the main driving force behind IC1's seasonal effect, and air humidity in the room is more likely the culprit.

Zoom In Zoom Out Reset image size

Figures 1(b) and (c) show the relationships between inside and outside absolute humidity and dew point, respectively. There is a linear relationship between the humidity data, except at the most humid days (outside H_a> 10.3 g m⁻³) when the inside humidity is visibly flattened by the air-conditioning system. A similar plot is obtained for inside versus outside dew point, since part of the humidity in the incoming air condensates as soon as the dew point exceeds the temperature of the cooling element. The mechanism is more pertinently visible in figure 1(d), which depicts the deviation of the inside humidity H_a from a linear relationship with outside humidity H_a as a function of the outside dew point T_D. The graphs also indicate that weather data collected tens of kilometres away are representative of the external conditions surrounding the laboratory.

The decay curves of a single ²²Na source [46] and two ¹³⁴Cs sources [47] have been regularly monitored in the ionisation chamber IC1 from 2010 to 2022. The residuals of the measured activity rates to an exponential decay function exhibit a high negative correlation (ρ= −0.55 and ρ= −0.59) with the absolute ambient humidity. Figure 2 shows the averaged residuals of the two ¹³⁴Cs sources binned by corresponding absolute humidity values outside the laboratory, with a bin width of 0.2 g m⁻³. There is a clear downward linear trend of the residuals between 3 g m⁻³ and 11 g m⁻³, which appears to break at higher humidity levels. This can be attributed to the drying effect of the laboratory's air-conditioning system. A similar trendline has been observed in the ²²Na data [46].

Zoom In Zoom Out Reset image size

The ambient humidity undergoes a seasonal cycle that follows the average ambient temperature, as heat drives evaporation of water and increases the humidity capacity of air. Due to their sensitivity to humidity, the ²²Na and ¹³⁴Cs activity measurements in IC1 exhibit an annual cycle with an amplitude of 0.0055% and a phase shift of 50 d. Figure 3 confirms that this cycle is well synchronised with ambient humidity, and less so with the Earth–Sun distance (as seen in the variation in solar neutrino flux) or local temperature.

Zoom In Zoom Out Reset image size

The amplitude A of the annual cycle in the residuals from exponential decay is determined from a weighted fit of a sinusoidal function f(x) = Ax, in which x= sin((t+ a)/365), t is the day of the year, and a the phase shift. To account for uneven distributions of x, the amplitude is normalised to A' = A $\sqrt {2s_x^2}$. The statistical uncertainty on A' (when generated by normally distributed noise) is easily calculated from the uncertainty of the weighted mean of the residuals, $\sigma _{A^{^{\prime}}}^2 = 2\sigma _{{y_w}}^2$ [48]. The ratio $k = A^{^{\prime}}/{\sigma _{A^{^{\prime}}}}$ determines the power S= k²/2 and the probability P= e^−S that a power above S may be produced from random noise.

The method used in previous work [7–10] differs slightly as the data were binned in time zones and the uncertainty of the amplitude was derived by changing its value until the chi-squared (χ²) of the fit varied by unity. In this work, the power of the annual cycle in decay series will be compared before and after compensation for a humidity effect. For example, the power of the ¹³⁴Cs and ²²Na cycles dropped from 124 and 54 to statistically acceptable levels of 3.8 and 2.9, respectively.

Other nuclides, including ⁵⁴Mn, ⁶⁵Zn, ¹⁷⁷Lu, ¹⁰⁹Cd, ¹²⁴Sb, and ²²⁶Ra, have been measured in the same IC [49–53]. Although these data sets are less elaborate, they also exhibit a humidity effect. Extensive decay rate series of two ²⁰⁹Po sources measured with Si detectors show no significant correlation with humidity [34, 35]. The same conclusion applies to repeated activity measurements of ³H and ¹⁴C in a liquid scintillation detector (LSC) [9], ⁵⁵Fe in an x-ray counter with defined solid angle using a gas counter [54] and a large pressurised gas proportional counter [55], as well as a combination of ²³⁰U decay rates taken with a CsI sandwich spectrometer, LSC, pressurised proportional counter, and planar Si detectors [56]. The findings are summarised in table 1.

Additional evidence has been collected from metrology institutes in various locations. The obtained decay rate series were optimised to detect annual cycles by removing outlying data points and compensating apparent discontinuities or systematic trends with time using linear functions covering relatively large time periods. Weather data were collected from nearby weather stations (Weather Underground, 2023) and a linear regression analysis was conducted to establish a linear relationship ${e_A} \propto a + b{H_{\text{a}}}$ between the absolute ambient humidity, ${H_{\text{a}}}$, and the residuals from exponential decay, ${e_A}$. Activity data without corresponding humidity measurements were excluded from analysis. The amplitude and power of annual cycles in activity series were determined before and after compensation for the hypothesised humidity effect (${e_A} \leftrightarrow {e_A} - a - b{H_{\text{a}}}$), and the results are summarised in table 1.

The NIST has provided historical decay measurements of various nuclides, including ²²Na, ⁶⁰Co, ⁸²Sr, ¹⁰⁹Cd, ¹³³Ba, ¹³⁷Cs, ¹⁵²Eu, ²⁰⁷Bi, ²²⁸Th, and ²²⁶Ra, obtained with ICs [57, 58]. Additionally, a decay series of ⁶⁰Co was measured by coincidence counting, ³H by internal gas counting, and ⁸²Sr by γ-ray spectrometry with a HPGe detector. Humidity data were extracted from the weather station at the Washington Dulles Airport (Sterling). To suppress the influence of instabilities on the measurements obtained from the 'IC A' chamber, the source current was taken relative to that of a ²²⁶Ra reference source. As a result, a mild annual cycle was observed in the measurements, which was reduced after compensation for the hypothesised humidity effect.

Direct evidence of the sensitivity of instruments to environmental factors was collected in the new laboratory wing of the NIST. The behaviour of the 'AutoIC' chamber was continuously monitored by performing source and background measurements while the climate control system generated short-term cyclic variations in temperature and humidity in the laboratory. Figure 4 illustrates how the ionisation current for a ²²³Ra source in the 'AutoIC' chamber follows the environmental changes with a delay of approximately an hour. This information was persuasive in efforts to prioritise improved climate control as part of the ongoing construction project.

Zoom In Zoom Out Reset image size

Over the years, the PTB has applied three different current measurement methods for their IC (type IG12, 20th Century Electronics, UK); a Townsend balance until 1999, a Keithley electrometer until 2020, and currently an Ultrastable Low-Noise Current Amplifier (ULCA). As a result, the detector has shown different sensitivity to environmental influences. It has already been demonstrated in previous work [8, 59–62] that the Townsend balance appeared to be susceptible to radon concentration in air, which explains the annual cycles observed in the pre-1999 data sets. These data are not further considered in the paper.

More recent data sets were provided for 11 radionuclides: ⁵⁴Mn, ⁶⁰Co, ⁸⁵Kr, ⁹⁰Sr, ^108mAg, ¹³³Ba, ¹²⁴Sb, ¹³⁷Cs, ¹⁵²Eu, ¹⁵⁴Eu, and ²²⁶Ra. Those obtained with the Keithley electrometer generally show some humidity effects, which reduce to an acceptable level after compensation based on humidity data from the Langenhagen weather station. A peculiar case is the beta emitter ⁹⁰Sr measured in the IC, which manifests a significant annual cycle with power S= 52, as shown in figure 5. It would appear that the cycle is synchronised with ambient humidity to a large extent, but there is additional structure due to local instabilities in the data set. There is a gradient in the residuals of −0.0045% per g m⁻³ humidity. Besides that, other influential factors in the measurement conditions may have played a role. As a result, a linear correction proportional with humidity reduces the power of the cycle to S = 15, but does not completely cancel the annual trend.

Zoom In Zoom Out Reset image size

The ULCA, developed by PTB [63, 64], has been introduced to measure small currents with low noise and high gain stability. While the 2020–2022 data sets appear to be independent of humidity, there is still some variance present that can be partly attributed to imperfections in geometrical replication of the source positioning using a robot arm. In general, the amplitudes of the fitted annual cycles are statistically insignificant, but reside at the high side of the Rayleigh probability distribution. It should be noted that in order to mitigate the influence of environmental effects in normal praxis, the PTB conducts its calibrations relative to a ²²⁶Ra reference source.

The PTB [65, 66] generated particularly stable decay series measurements of ³⁶Cl, ⁹⁰Sr, and ²⁴¹Am by means of liquid scintillation counting, applying TDCR in the case of ³⁶Cl and ⁹⁰Sr and a commercial LSC for ²⁴¹Am. The data show no correlation with humidity. Undisclosed results of ongoing ³²Si decay rate measurements promise exceptional stability, in contrast with the seasonal variations observed in the BNL/Purdue data [67, 68], which were found to be correlated with ambient humidity [17, 18].

At the NPL (UK), a long history of ²²⁶Ra check source measurements was provided for two well-type re-entrant ICs: a PA782 (2 MPa argon gas, steel inner well) and a Vinten (1 MPa nitrogen gas, 3 mm Al inner well), as well as a more recent series of ²⁴¹Am measurements in the Vinten IC. Only data obtained since July 1996 could be used in the study, upon the availability of weather information from the Hounslow weather station. The residuals show the presence of a significant annual sinusoidal deviation from exponential decay, which clearly correlates with ambient humidity. Evidence is shown in figure 6. In addition, a series of gamma-ray measurements with lower precision showed a hint of a humidity effect which disappeared after compensation.

Zoom In Zoom Out Reset image size

The BIPM (located in Sèvres) houses the SIR, a system of two Centronics IG11 ICs used as long-term reference instruments by which mono-radionuclide solutions with standardised activities can be compared for international equivalence. The ionisation current is measured with a Keithley electrometer with external capacitors in a sealed box and using the Townsend balance technique. The laboratory room is temperature and humidity controlled. The stability of the IC #389 has been monitored with the ²²⁶Ra check source #4 and the residuals from exponential decay between 2001 and 2022 show no annual cycle, and no significant correlation with humidity in the Toussus-le-Noble weather station.

Additional evidence was collected in the southern hemisphere, at ANSTO (Australia) [8]. Several sets of ²²⁶Ra measurements in a TPA Mk-II IC with a Keithley 6517 A electrometer have been combined and turned out to be free of an annual cycle. Records of the response of a medical calibrator (TPA IC with Keithley electrometer) to a ²²⁶Ra reference source show also no sign of annual oscillations.

A series of ¹⁴C measurements by TDCR in NMISA (South-Africa) showed no significant annual cycle [9].

At the IRA (Switzerland), a ¹³⁷Cs source was measured between 1980 and 2012 in a Centronic IG11 IC and automated readout based on the principle of the Townsend balance. A typical measurement cycle corresponds to the time needed to load a given capacitor up to 0.1 V. Data after 1996 could be compared with humidity values from the Lausanne weather station. There was no notable annual cycle.

From NRC (Canada) three small data sets were made available from ¹³⁷Cs in three different ICs (Capintec, Vinten and TPA). The individual data sets seem correlated with humidity in the Ottawa weather station. In table 1, values are provided for a combination of the data sets.

At the laboratory in Casaccia, thousands of measurements have been performed between 24 March and 12 May 2016 to investigate correlations between the Centronic IG11 IC readouts with a Keithley 617 electrometer for a ²²⁶Ra source and various environmental conditions obtained with an AlphaGUARD monitor in the laboratory [8]. The results of the study show a significant correlation (ρ = 0.84) between the averaged IC currents and temperature, indicating that temperature variations of several degrees Celsius in the laboratory may be the primary cause of seasonal variations in the IC signals at ENEA. The impact of other environmental conditions, including ambient pressure, relative humidity, and radon concentration in air, was found to be lower (−0.2 < ρ < 0.2).

In the frame of quality control of the γ-ray spectrometry service at the SCK CEN (Belgium), spectra of mixed ¹⁵²Eu-²⁴¹Am reference sources were collected on multiple HPGe spectrometers at a weekly rate. In previous work, data from 2008–2016 were analysed and a strong correlation was found in the annual cycles of characteristic lines of ¹⁵²Eu and ²⁴¹Am, however the amplitude of the cycles differed among the detectors [7, 10]. This was considered proof that the instabilities were of instrumental nature and not caused by 'variations in the decay constants due to solar neutrinos'. At that time, the detectors were kept in a temperature-controlled room, without humidity control. Through comparison with climate data from a weather station in Antwerp, evidence suggests that sensitivity of the spectrometers to ambient humidity was the most likely culprit for these seasonal deviations.

In the current work, data sets obtained with 14 HPGe detectors up to Oct 2017 have been (re)analysed. The amplitude of the annual cycle was determined for the peak areas of the 59 keV line of ²⁴¹Am and for the sum of the 122 keV, 779 keV and 1408 keV peak areas of ¹⁵²Eu. Weighted mean values for both nuclides are presented in figure 7, showing the amplitude height of the fitted annual cycle as a function of the correlation factor between the decay rates and ambient humidity. The highest annual cycles seem to appear in data sets with a relatively high correlation with humidity. Their amplitudes drop to a statistically insignificant level when the individual data are compensated for a hypothesised linear relationship with the average outdoor absolute air humidity at the day of measurement.

Zoom In Zoom Out Reset image size

By April 2019, the detectors have been moved to a new laboratory equipped with temperature and humidity control. The data sets since that date are less extensive, not necessarily more stable, but show no statistically significant annual cycles. The associated amplitudes remain almost invariant for humidity compensation (figure 8).

Zoom In Zoom Out Reset image size

The JSI (Slovenia) provided quality assurance measurements of ⁶⁰Co and ¹⁰⁹Cd on different HPGe detectors. Results from 5 detectors were combined for ⁶⁰Co and 3 for ¹⁰⁹Cd, and they were compared with weather data from Cerklje. No annual cycle or humidity effect could be identified.

The IAEA environmental laboratory in Monaco conducted quality control measurements of three radionuclides (²⁴¹Am, ¹³⁷Cs, ⁶⁰Co) on four HPGe detectors. Weather data were obtained from the Oceanographic Museum of Monaco and supplemented with data from the airport in Nice. The laboratory room was equipped with a dehumidifier, which effectively suppressed the inside humidity level whenever the ambient humidity exceeded 10 g m⁻³. Three detectors showed minimal annual cycles, while the measurement results of HPGe #15 exhibited indications of anti-correlation with ambient humidity. For this detector, a mean annual cycle of 0.33 (7)% amplitude was observed, which reduced to 0.17 (7)% after compensating for humidity.

The PSI reported severe instabilities in the percent range in decay rate series of ¹⁷¹Tm, ¹⁴⁵Sm, and ⁴⁴Ti measured on a HPGe detector with imperfect vacuum in the crystal housing [69, 70]. It was found that the variations in the count rates over the measurement campaign were highly correlated with the absolute humidity outside the laboratory. A correlation plot is shown in figure 9. The count rate variations went hand in hand with shifts in the energy calibration.

Zoom In Zoom Out Reset image size