Stable Isotope Ratios of Mercury in Commercially Available Thermometers and Fluorescent Tubes.

Five stable isotope ratios of mercury (199Hg/198Hg, 200Hg/198Hg, 201Hg/198Hg, 202Hg/198Hg, and 204Hg/198Hg) in commercially available thermometers and fluorescent tubes were analyzed to characterize their potential anthropogenic emission source to landfills, manufacturing factories, and our daily lives. The results for the liquid metal mercury yielded from the thermometers showed similar mass-independent fractionation values to those in the literature. The analysis of fluorescent tubes resulted in that more than 96% of mercury in the fluorescent tubes was found in the adsorbed state, and up to 3.5% of mercury was in the gas-phase. Unique mass-independent isotope fractionation values were found in the gaseous and adsorbed mercury in the fluorescent tubes. This fractionation is distinct from other emission sources and systematic; therefore, it can potentially be used to fingerprint mercury in fluorescent tubes in environmental samples.

Introduction

Mercury (Hg) is a toxic and unique metal element among metal elements because as the elemental form it is the only common metal liquid at STP conditions;[1] thus, it has higher vapor pressure than other metal elements have.[2] Once turned into a gaseous form, Hg easily spreads out over the whole Earth via the atmosphere[3] and enters the terrestrial and aquatic ecosystems where inorganic Hg is converted to methyl Hg, resulting in its bioaccumulation in aquatic organisms, and eventually in humans and wild animals through the food web.[4] In August 2017, the United Nations started implementing international regulations on the use of man-made Hg, the so-called Minamata Convention on Mercury, to mitigate the burden of anthropogenic Hg emissions on the natural environment.[5] Among man-made Hg emission sources to the atmosphere, Hg emission from wastes of mercury-added products accounts for 7% approximately.[6] This includes fluorescent tubes and thermometers, which are the most widely used commercial goods containing Hg worldwide. However, the detailed breakdown of Hg emission from the waste of Hg-contained products is not clear.[6] In addition, mercury-added products sometime turn to be a large emission source of Hg accidentally. For example, the Hg release to the ambient air from the fire in a fluorescent tube factory in Vietnam[7] was large and its impact on the air quality was a concern of the resident neighbors. Atmospheric Hg is, however, a mixture of gaseous Hg from a variety of emission sources, and the quantitative understanding of source contributions by concentration measurements only at a receptor site (i.e., the top-down analysis) is not an easy task. Therefore, it is important to establish a method to trace Hg emitted from every emission source. A recent technique using stable Hg isotope ratio (δHg) measurements allows identification of the origin of Hg and gaining insight into source identification and apportionment.[8,9] The initial δHg values of Hg from every emission source (i.e., the emission inventory accompanied by stable mercury isotopic compositions) are the necessary information for source apportionment analysis. To date, scientists have been making effort to characterize the initial δHg inventory, such as cinnabar mining,[10−14] coal combustion,[15−19] natural gas production,[20] volcano,[21] Hg in plant and soil,[22−27] biomass burning,[28] ocean,[29] and permafrost.[30] However, all sources have not been covered yet. Commercially available Hg-added products are one of the emission sectors that the initial δHg values are not evaluated in detail yet. Fluorescent tubes and thermometers are not major man-made emission sources of Hg currently (<7%),[6] but those may significantly be responsible for the emissions from waste disposal (i.e., incinerators and landfills). To the best of our knowledge, there has been no reports of investigating the δHg of Hg in commercially available thermometers, and, to the best of our knowledge, there is only one report for Hg in fluorescent bulbs.[31] This fluorescent bulb study, however, focused on Hg trapped inside the glass material of fluorescent bulbs, and the δHg values of the easily releasable forms of Hg to environment, gaseous and adsorbed Hg, have not been studied and reported yet. Here, we attempted to characterize five δHg values of Hg (199Hg/198Hg, 200Hg/198Hg, 201Hg/198Hg, 202Hg/198Hg, and 204Hg/198Hg) in thermometers and fluorescent tubes manufactured and used in Japan. The stable isotope ratios of gaseous Hg from fluorescent tubes and of bulk Hg from thermometers are reported for the first time.

Results and Discussion

Validation of the Analytical Method

Measurement tests that validate the analytical method and isotope measurements, described in subsection , Materials and Methods, are discussed in the subsection S1 in the Supporting Information (SI). Briefly, 99.9% of the total gaseous mercury (TGM) trapped in the fluorescent tubes was successfully transferred to the Tedlar bags as gaseous elemental Hg (GEM), and more than 99.99% of the GEM enclosed in the Tedlar bag was successfully converted to Hg2+ in a trapping solution. Two-day dissolution tests of the adsorbed Hg to the fluorescent tube glass walls showed no variation in the Hg concentration after heating the dissolution solutions, indicating that the dissolution of the adsorbed Hg was complete (SI Figure S1).

Even though no analytical test was performed during the preparation of the sample solution for liquid elemental Hg (Hg0(L)) from the thermometers due to the lack of a reference isotopic standard available for Hg0(L), we expect that the visually confirmed complete dissolution of a Hg drop in 40% reversed aqua regia would not cause an artifact. Thus, we conclude that the stable Hg isotopic compositions in the sample solutions reflect the original isotopic compositions.

The routine δHg measurements of SRM 8610 (n = 20) demonstrated that the precision (2× standard deviation or SD) and accuracy (the average difference between the measured δHg and the recommended δHg by NIST) were better than 0.09 ‰ and 0.04 ‰, respectively (SI Table S1). These precision and accuracy values are good enough to gain δHg values we aimed.

Hg Yields from Thermometers and Fluorescent Tubes

Sampling bulk elemental mercury from 11 samples of three different thermometers resulted in the retrieval of 1.2 to 3.5 g of liquid mercury. The quantitative analysis of mercury found in the gas phase and the adsorbed mercury to the tube wall of fluorescent tubes showed that the mercury concentrations in the trapping solution from the GEM Tedlar bag extraction, the dissolution solution from the tube wall extraction, and the dissolution solution from the electrode extractions were in the range of 36 to 237 ng mL–1, 807 to 16440 ng mL–1, and 242 to 68567 ng mL–1, respectively (Table ). These values were high enough to enable the stable mercury isotope measurements, which required 10 ng mL–1 for our instrumentation. A comparison of mercury found in the gas phase and from the tube wall showed that more than 97% of the mercury in the fluorescent tubes was found on the tube wall and on the electrodes.

Table 1

Results of the Quantitative Measurements of Fluorescent Tube Mercury

	TGM		adsorbed Hg on the tube wall		adsorbed Hg on electrode 1		adsorbed Hg on electrode 2
sample ID	solution concn (ng mL–1)	Hg mass (μg)	solution concn (ng mL–1)	Hg mass (μg)	solution concn (ng mL–1)	Hg mass (μg)	solution concn (ng mL–1)	Hg mass (μg)
32w-1	124	12.4	5105	876	n/aa	n/aa	n/aa	n/aa
32w-2	36	3.6	807	143	n/aa	n/aa	n/aa	n/aa
32w-3	66	6.6	6140	1120	n/aa	n/aa	n/aa	n/aa
32w-4	52	5.2	7070	1252	n/aa	n/aa	n/aa	n/aa
32w-5	87	8.7	6122	1074	n/aa	n/aa	n/aa	n/aa
32w-6	49	4.9	4105	730	n/aa	n/aa	n/aa	n/aa
40w-1	87	8.7	4209	736	n/aa	n/aa	n/aa	n/aa
40w-2	237	23.7	5097	869	n/aa	n/aa	n/aa	n/aa
40w-3	127	12.7	16440	2946	n/aa	n/aa	n/aa	n/aa
40w-4	161	16.1	10701	1876	n/aa	n/aa	n/aa	n/aa
40w-5	133	13.3	4842	846	n/aa	n/aa	n/aa	n/aa
40w-6	73	7.3	4245	787	n/aa	n/aa	n/aa	n/aa
40w-7	111	11.1	6877	1298	n/aa	n/aa	n/aa	n/aa
40w-8	144	14.4	11773	1064	8199	410	11003	550
40w-9	75	7.5	9162	827	1775	89	1751	88
40w-10	514	51.3	6831	624	3942	197	11809	591
N40w-1	42	4.2	30860	2886	8462	423	5450	273
N40w-2	70	7.0	4058	374	68567	3428	863	43
N32w-1	34	3.4	17640	1599	26455	1323	242	12

The data are not available.

Variation Range of δHg for Thermometer and Fluorescent Tube Hg

The δxHg values from the Hg0(L) in the thermometers exhibited a small variation between the samples (Figure ): the variation ranges over the 11 samples were 0.19, 0.24, 0.41, 0.44, and 0.63 ‰ for δ199Hg, δ200Hg, δ201Hg, δ202Hg, and δ204Hg, respectively. Testing the analysis of variance for the evaluation of significant differences in δHg between samples resulted in F values, the coefficients of the analysis of variance, between 0.10 and 0.24. Compared to the upper critical F value (F-statistic) of 1.8307 at α (significance level) of 0.05 for F(∞,10), the determined F values were smaller. Thus, we concluded that there was no significant difference between δHg values in our thermometer samples. For this reason, the δHg data from the 11 samples for each isotope were combined together. The overall average ±2 × standard error, or SE, of the 11 samples was −0.12 ± 0.02, −0.32 ± 0.02, −0.48 ± 0.04, −0.65 ± 0.04, and −0.99 ± 0.06 ‰ for δ199Hg, δ200Hg, δ201Hg, δ202Hg, and δ204Hg isotope ratios, respectively.

Figure 1

Box plot for five stable isotope ratios of Hg found in electrodes 1 and 2 (n = 6), the gas phase (n = 19), and on the glass wall surface (n = 19) of fluorescent tubes, and in liquid metal Hg enclosed in thermometers (n = 11) (from left to right). A box represents the range of the upper and lower quartiles, a red horizontal bar indicates the median, and the upper and lower vertical bars attached to the top and bottom of the box indicate the maximum and minimum values of the observed isotope ratios.

In contrast, δHg values of Hg found in the three locations of fluorescent tubes (TGM, adsorbed Hg on the glass wall, and adsorbed Hg on the electrodes) showed a large variation: The largest variation was for electrode 1, and relatively small variations for electrode 2, TGM, and the glass wall (Figure ). The large variation in electrode 1 was due to the extremely high δHg values of the 40w-2 sample, 7.3–44.8 ‰. Since the concentration of the sample solution was high enough and there is no reason to identify this sample as an outlier, the very high δHg presumably indicates that such a high value can occur.

Comparison of Thermometer δHg with Literature Values

We compared the values obtained in this study to those in the literature (Table ) for identification of their origins. The literature values include isotope ratios of Hg0(L) from cinnabar ore,[10] which is the original material of Hg0(L), and cinnabar ores from a variety of mining locations, such as Almadén, Spain,[10] Wanshan, China,[17] New Idria, USA,[11] and Terlingua and McDermitt, USA.[12]Table also includes the results of the early cinnabar characterization work,[13] which covers mining locations around the world. However, it should be noted that they measured the isotope ratios of their samples against a different reference standard, SRM 1641. Even though the original material of liquid Hg used for SRM 1641 and 3133 are the same, using SRM 1641 may result in larger uncertainties in δxHg values due to the interference of Hg2+ reduction by an Au additive in SRM 1641.[32] It should also be noted that the isotope ratios that Hintelmann and Lu[13] reported were relative values to a different denominating isotope, 202Hg. To validate this comparison, we converted their reported δHg values to δHg values relative to 198Hg on the SRM 3133 scale. This conversion was made using the reported raw isotope ratios under the condition that all the true δHg values from SRM 1641 on the SRM 3133 scale correspond to 0 ‰.

Table 2

Comparison of Stable Hg Isotope Ratios from the Literature

	δ199Hg	δ200Hg	δ201Hg	δ202Hg	δ204Hg
thermometer Hg0(L) (this study, n = 11)a	–0.12 ± 0.02	–0.32 ± 0.02	–0.48 ± 0.04	–0.65 ± 0.04	–0.99 ± 0.06
Cinnabar
cinnabar ore in Wanshan area, China (n = 14)b	–0.18 ± 0.08	–0.37 ± 0.10	–0.55 ± 0.14	–0.74 ± 0.22	n/ai
Almadén cinnabar ore, Spain (n = 7)c	–0.21 ± 0.08	–0.31 ± 0.16	–0.48 ± 0.18	–0.56 ± 0.26	n/ai
cinnabar from other locations in the Almadén district, Spain (n = 10)d	–0.17 ± 0.24	–0.24 ± 0.52	–0.38 ± 0.72	–0.48 ± 1.02	n/ai
New Idria cinnabar ore, CA, USAe	–0.05 ± 0.11	n/ai	–0.18 ± 0.11	–0.26 ± 0.10	n/ai
cinnabar mined from the Terlingua district, TX, USA (n = 3)f	–0.22 ± 0.04	–0.80 ± 0.04	–1.14 ± 0.07	–1.66 ± 0.06	n/ai
Cinnabar mined from McDermitt, ND, USA (n = 11)f	–0.14 ± 0.02	–0.29 ± 0.05	–0.42 ± 0.06	–0.58 ± 0.08	n/ai
Red Devil mine (Alaska)g	–0.45	–0.66	–0.98	–1.30	n/ai
Zlatna (Romania)g	0.02	0.01	0.03	–0.01	n/ai
Nikitowka (Ukraine)g	–0.18	–0.40	–0.49	–0.63	n/ai
Almadén (Spain)g	–0.55	–0.68	–0.72	–1.13	n/ai
Sonora (California)g	–0.16	–0.31	–0.55	–0.66	n/ai
Wolfenstein (Germany)g	–0.25	–0.64	–0.99	–1.33	n/ai
Punitaqui (Chile)g	–0.37	–0.69	–0.99	–1.33	n/ai
New Almaden (USA)g	–0.04	–0.06	–0.09	–0.09	n/ai
Chinag	–0.31	–0.42	–0.62	–0.75	n/ai
Stahlberg (Germany)g	–0.36	–0.61	–0.94	–0.89	n/ai
Avala (Serbia)g	–0.52	–0.73	–1.01	–1.22	n/ai
New Idria (California)g	–0.12	–0.33	–0.53	–0.62	n/ai
Giftberg (Czech)g	–0.32	–0.62	–0.88	–1.15	n/ai
Szlaniz (Hungary)g	–0.41	–0.66	–1.03	–1.18	n/ai
Liquid Elemental Hg (Hg0L)
Hg0(L)-1 in ore, El Entredicho, Spainh	0.05	0.23	0.04	0.26	n/ai
Hg0(L)-2 in ore, Almadén, Spainh	–0.12	–0.36	–0.57	–0.84	n/ai
Hg0(L)-4 in ore, Las Cuevas, Spainh	–0.25	–0.4	–0.48	–0.84	n/ai
Hg0(L)-5, Almadén from Hg flask, Spainh	–0.2	–0.34	–0.49	–0.67	n/ai
Hg0(L)-6, chlorine-alkali plant, Italyh	–0.03	–0.05	–0.15	–0.16	n/ai

This study, average ±2SE.

Yin et al.,[17] average ±2SD.

Gray et al.,[10] average ±2SE.

Gray et al.,[10] average ± SD.

Wiederhold et al.,[11] average ±2SD propagated.

Stetoson et al.,[12] average ±2SD.

Hintelmann and Lu,[13] single value.

Gray et al.,[10] single value.

Not available.

Comparing δHg values reported by others for cinnabars from the same origin may indicate the size of this bias. Therefore, δHg values of cinnabar ores from the Almadén district, Spain, and New Idria, USA, reported by Gray et al.[10] and Wiederhold et al.,[11] were compared with those values reported by Hintelmann and Liu.[13] The comparison shows that the δHg values from Almadén cinnabar reported by Hintelmann and Lu were lighter than other values by 0.34–0.65 ‰, but those were within the uncertainty ranges. Meanwhile, the δ201Hg and δ202Hg values in cinnabar from New Idria were significantly lighter by 0.35 ‰ and 0.36 ‰, respectively. Considering the level of bias in their reported values, the cinnabar ores from Zlatna (Romania) and New Almaden (USA) have uniquely heavier δHg signatures. The δHg values of cinnabar from other locations are similar, demonstrating the limitation of δHg use for the identification of mining locations.

The comparison of our data with values from the literature shows that our δHg values agree with values from Almadén (Spain), Wanshan (China), and McDermitt (USA). The comparison of δHg values from Hg0(L) in Table also demonstrates that the values were similar for Hg0(L) contained in Almadén cinnabar ores and for Hg0(L) calcined from the Almadén cinnabar. The Hg used in the thermometers may also be derived from Almadén, a major Hg mining region in the world. However, identifying the origin of Hg0(L) using δHg values was not conclusive due to similar δHg values from other locations and uncertainty of isotope fractionation during the calcination process. The application of δHg to the identification of mining location of cinnabar and Hg0(L) seems difficult.

Hg in the Fluorescent Tubes

Comparing the yields of TGM, the adsorbed Hg to the glass wall, and the adsorbed Hg on the electrodes demonstrates that the majority of extractable Hg was in the adsorbed state (Table ). Only 0.1–3.5% of the total Hg found in the fluorescent tubes was partitioned in the gas phase, regardless of the brand and of new or used. However, it is likely that less TGM initially existed inside the tube before the tube was broken because the total volume inside the tube that the TGM occupied was smaller than the volume including the fluorescent tube and a 5 L plastic bag used for the TGM sampling. It is reasonable to assume that more adsorbed Hg instantaneously partitioned into the gas phase when the tube was snapped due to the sudden increase of the total volume. This phenomenon probably influences the isotopic compositions of TGM and thus is discussed later. A comparison between the adsorbed Hg to the glass wall and on the electrodes showed a random distribution of yielded Hg masses between the glass wall and the electrodes, regardless of whether tubes were brand new or used (Table ). The Hg adsorbed to the glass wall varied from 10 to 82% relative to the sum of all Hg found in the fluorescent tube. Although there was no systematic trend in the quantitative data, the isotope ratios showed some interesting trends.

Table 3

Comparison of Yielded Mass and δHg of Hg Found in the Gas-Phase, Adsorbed State on the Glass Wall and Electrodes of Six Fluorescent Tubes

	yielded Hg mass (μg)	δ199Hg (‰)	δ200Hg (‰)	δ201Hg (‰)	δ202Hg (‰)	δ204Hg (‰)
Glass Wall
40w-8	1177	–0.62	–2.17	–0.43	–3.34	–1.47
40w-9	916	–0.42	–2.23	–0.94	–3.61	–2.55
40w-10	683	–0.17	–0.38	0.17	–0.54	0.11
N40w-1	3086	–0.05	–0.20	–0.30	–0.39	–0.59
N40w-2	406	–0.08	–0.20	–0.37	–0.52	–0.75
N32w-1	1764	–0.15	–0.33	–0.54	–0.67	–1.06
Electrodes (Overall Electrode 1 and Electrode 2)
40w-8	960	1.27	3.33	0.83	5.06	2.58
40w-9	176	3.37	14.51	8.30	24.37	20.48
40w-10	788	0.43	0.55	–1.12	0.58	–0.80
N40w-1	696	–0.01	0.02	–0.03	–0.01	0.04
N40w-2	3471	–0.18	–0.26	–0.46	–0.53	–0.83
N32w-1	1335	–0.12	–0.24	–0.42	–0.44	–0.65
TGM
40w-8	14	0.15	0.84	–2.54	0.40	–1.29
40w-9	7	–1.16	–1.45	–2.98	–3.03	–3.53
40w-10	51	–0.96	–1.23	–1.67	–2.53	–3.96
N40w-1	4	–0.30	–0.94	–1.21	–1.84	–2.72
N40w-2	7	0.14	–1.28	–0.75	–2.33	–3.01
N32w-1	3	–0.13	–1.09	–0.99	–2.09	–2.81
Adsorbed (Overall Glass Wall and Electrodes)
40w-8	2137	0.23	0.30	0.14	0.43	0.35
40w-9	1092	0.19	0.48	0.55	0.91	1.17
40w-10	1471	0.15	0.11	–0.52	0.06	–0.38
N40w-1	3782	–0.04	–0.16	–0.25	–0.32	–0.47
N40w-2	3877	–0.17	–0.25	–0.45	–0.53	–0.82
N32w-1	3099	–0.14	–0.29	–0.49	–0.57	–0.89
Whole Tube (Glass Wall + Electrodes + Gas)
40w-8	2152	0.23	0.31	0.12	0.43	0.34
40w-9	1100	0.18	0.46	0.52	0.88	1.14
40w-10	1522	0.11	0.07	–0.55	–0.03	–0.50
N40w-1	3786	–0.04	–0.16	–0.25	–0.32	–0.47
N40w-2	3884	–0.17	–0.25	–0.45	–0.53	–0.82
N32w-1	3102	–0.14	–0.29	–0.49	–0.57	–0.89

The δHg values for the Hg adsorbed onto the glass wall and electrodes in the brand-new fluorescent tubes displayed similar values, while those δHg values for the used fluorescent tubes showed a large variation (Table ). The large differences and variations for the used fluorescent tubes can be explained by that electric discharge under the normal operation of fluorescent tubes causes Hg to undergo physical and chemical processes, which may cause inhomogeneous distribution of Hg within the fluorescent tube. If the processes are accompanied by some extent of isotope fractionations, undergoing such processes may result in a large difference and variation in δHg. In contrast, the Hg in new fluorescent tubes has not yet undergone such processes, resulting in a relatively homogeneous distribution of Hg isotopes within the tube. This fractionation phenomenon is implied by the δHg values of the electrodes from the used fluorescent tubes, which were substantially higher (i.e., heavier isotopic composition) than the δHg values for the glass wall Hg and TGM. Although significant extent of isotope fractionation during being trapped in the glass medium of fluorescent tubes has been reported, trapped Hg in the glass material is 1% or less.[31] Assuming the extent of isotope fractionation by −30 ‰,[31] corresponding the fractionation factor of 1.03, and δHg values for all isotope ratios are 0 ‰, the estimated δHg shift for the residual Hg (the majority is the adsorbed Hg) is only by 0.3 ‰, which does not explain the largely fractionated δxHg we observed.

Overall, the δHg values of TGM in the fluorescent tubes, except for δ199Hg in N40w-2 and N32w-1 and δ200Hg in 40w-8, were significantly lighter than those for the adsorbed Hg, which acts as the reservoir of TGM. This is likely due to isotope fractionations that occur during the evaporation and condensation processes of Hg. Significant Hg isotope fractionations during the evaporation of Hg from bulk Hg with a fractionation factor of 1.0067 and during the equilibrium between the evaporation and condensation with a small fractionation factor of 1.00086 have been reported.[33] Since the fractionation factor for the reverse process is a reciprocal of the fractionation factor for the forward process,[34] the condensation fractionation factor is predicted to be 0.9933. The value smaller than unity indicates isotope fractionation in which heavy isotopes preferentially condense. As pointed out previously, the TGM inside the tubes was initially under equilibrium between the gas and condensed phases. The TGM was then evaporated from the glass wall right after the tube was broken due to the sudden increase of total volume. This process probably made some light Hg isotopes be preferentially evaporated from the glass wall and electrodes. The lighter δHg values for the TGM than the δHg values for the adsorbed Hg with the magnitude of the isotope fractionation smaller than the evaporative isotope fractionation, but larger than the equilibrium isotope fractionation at the 3.5% or less extent of the evaporation process, can be explained by the break of equilibrium.

Analysis of Mass Dependent and Independent Fractionation

For further characterization, the series of δHg values was evaluated if the observed isotopic compositions were resultant mass-dependent fractionation (MDF) or mass-independent fractionation (MIF). ΔHg for the mass x Hg isotope is defined as follows using the delta notations of the isotope ratios for mass x and 202, as the δHg ranges are below 10 ‰:[32]where β is the MDF factor for the mass x Hg isotope reference to 202Hg. This analysis evaluates how far the observed δH deviates from the predicted δH by MDF theory. The values of β199, β200, β201, and β202 used are 0.2520, 0.5024, 0.7520, and 1.493, respectively.[32] If any significant difference from the predicted value is observed, this indicates the influence of MIF.

Plots of Δ199Hg versus Δ201Hg for Hg emissions from fossil fuel production and combustion[15,16,18,19] or for the Hg contained in the fuel itself[17,18] have been reported to exhibit deviations from the origin with a slope of approximately unity. And our observations for thermometer Hg0(L) followed this trend (Figure ). That is, fingerprinting our thermometer Hg0(L) samples using ΔHg analysis is still difficult due to the similar trend. No unique feature was observed in other ΔHg plots (Figure ). In contrast, the observations for the Hg adsorbed to the glass wall of the fluorescent tubes displayed a different trend of MIF: high correlations between Δ200Hg and Δ201Hg or Δ204Hg (r2 = 0.70–0.87) with negative slopes of −0.16 to −0.26; a high correlation between Δ201Hg or Δ204Hg (r2 = 0.78) with a positive slope of 0.6 (Figure ). The plots for the same ΔHg values obtained from the literature in the figure clearly show distinct trends, suggesting that multiple ΔHg plots potentially fingerprint Hg from fluorescent tubes. A negative trend in the plot of Δ200Hg versus Δ204Hg was also observed in atmospheric studies,[26] but the slope was −0.51, which was more negative, and the mechanism of this fractionation was unknown. Photolysis or photochemistry may cause such a unique fractionation,[31] but our data set cannot reveal the mechanism. Similarly, high correlations with the same sign for the slopes were observed in the ΔHg plots for the TGM (Figure ) and adsorbed Hg onto the electrodes (Figure ), but the magnitudes of the slopes were different. It is worthwhile to note that the ΔHg plots for the electrode Hg showed very high correlations (r2 higher than 0.79), regardless of new or used fluorescent tubes. The series of observations likely imply the existence of unusual fractionation that is independent of the operational hours.

Figure 2

Plot of Δ199Hg versus Δ201Hg for thermometer Hg in this study (blue circles) and a variety of Hg emission sources reported: world natural gas (yellow triangles),[20] coal from China (red triangles),[17] Almadén cinnabar (green triangles),[10] liquid elemental Hg (pink squares),[10] and cinnabars collected worldwide (black crosses).[13] For comparison, a linear regression with a slope of unity (dotted green line), a frequently observed relationship line in source and atmospheric studies, is also shown.

Figure 3

Plot of ΔHg versus ΔHg for thermometer mercury (blue circles) and worldwide natural gases (yellow triangles).[20]

Figure 4

Plot of Δ199Hg versus Δ201Hg for Hg adsorbed to the glass wall surface of fluorescent tubes in this study (orange circles) and a variety of Hg emission sources reported: world natural gas (yellow triangles),[20] coal from China (red triangles, Δ200Hg and Δ204Hg are unavailable),[17] Almadén cinnabar (green triangles, Δ204Hg is unavailable),[10] liquid elemental Hg (pink squares, Δ204Hg is unavailable).[10]

Figure 5

Plot of Δ199Hg versus Δ201Hg for TGM found in the fluorescent tubes in this study (red circles) and a variety of Hg emission sources reported: world natural gas (yellow triangles),[20] coal from China (red triangles, Δ200Hg and Δ204Hg are unavailable),[17] Almadén cinnabar (green triangles, Δ204Hg is unavailable),[10] liquid elemental Hg (pink squares, Δ204Hg is unavailable).[10]

Figure 6

Plot of Δ199Hg versus Δ201Hg for Hg found in electrodes 1 and 2 of the fluorescent tubes in this study (purple circles) and a variety of Hg emission sources reported: world natural gas (yellow triangles),[20] coal from China (red triangles, Δ200Hg and Δ204Hg are unavailable),[17] Almadén cinnabar (green triangles, Δ204Hg is unavailable),[10] liquid elemental Hg (pink squares, Δ204Hg is unavailable).[10]

Regardless of the unknown mechanism for this fractionation, the observed unique MIF can be used for source identification. If a substantial amount of Hg from fluorescent tubes is discharged to the natural environment, the analysis of multiple ΔHg values (correlations, sings of linear regressions, and magnitudes of MIF) potentially provide source identification of Hg from fluorescent tubes. The study here, however, was made with the limited number of fluorescent tubes. Collection of more data will strengthen this conclusion.

Conclusion

We analyzed the stable isotopic compositions of Hg from 11 thermometer and 19 fluorescent tube samples to characterize potentially important sources of man-made Hg. The results demonstrated that the isotopic compositions of Hg0(L) from thermometers had similar δHg values and MIFs to the literature values; thus, unique fingerprinting information was not found. Meanwhile, the results from the fluorescent tube analysis showed very unique MIF trends, likely due to unique isotopic fractionation that occurs during the operation of the fluorescent tubes, independent of the operational hours. This feature can possibly be used for source identification of Hg from the used fluorescent tubes.

Materials and Methods

Sampling

Availability of Hg thermometers are currently limited in Japan due to the regulation on the production of Hg-contained products. Eleven brand-new thermometers were purchased from a lab-ware company for the analysis. Sixteen used fluorescent tubes were collected from the recycle waste in the municipal recycle center. Those samples were randomly chosen from the pile of used straight fluorescent tubes. Three brand-new straight fluorescent tubes were randomly selected and purchased at a neighborhood home center. Some major electric manufacturers have ceased production of fluorescent tubes, thus, fluorescent tubes manufactured by only several companies were available. These samplings resulted in the choice of the brand-new thermometers manufactured by two different companies (Table S2 in the SI), the 16 used fluorescent tubes manufactured by four different companies, and the three brand new fluorescent tubes manufactured by two different companies (Table S3 in the SI).

Preparation of the Sample Solutions for Isotope Measurement

To analyze stable Hg isotope ratios, the Hg needs to be in a stable form in solution (i.e., Hg2+). Therefore, Hg0(L) sampled from the thermometers, TGM, and adsorbed Hg to the glass walls and electrodes of the fluorescent tubes were sampled and oxidized in acid mixtures. Those methods are described below.

Hg0(L) in the thermometers was sampled by breaking the glass tips of the thermometers. A half drop of mercury from each thermometer was pipetted into a 20 mL glass vial (AS ONE Co. Ltd., Osaka, Japan) containing 20 mL of 40% inversed aqua regia, which is the highest concentration that allows for direct measurements of the stable mercury isotope ratios without the reoxidation problem,[35] and then the sample solution was left at room temperature until the complete dissolution of Hg0(L) was visually confirmed. The complete dissolution took nearly two months under this setup. Last, the solution was warmed in a 313 K water bath and sonicated overnight before the measurements were taken.

TGM and adsorbed mercury on the glass wall of the fluorescent tubes and both sides of the electrodes were sampled and oxidized using acid solutions in the following manner. First, a 5 L plastic bag with a double zip used for food storage (0.07 mm thickness, 20 cm × 25 cm, REED Freezer bag, Lion Corp., Tokyo, Japan) was used to create a closed environment for sampling TGM from the fluorescent tubes (Figure ). The outside of the bag was reinforced with curling tape to prevent immediate damage by the sharp edges of the broken glass from the fluorescent tubes when the tubes were snapped. The two bottom corners of the plastic bag were cut off, and the fluorescent tube, which was scratched by a glass cutter in advance so that the fluorescent tube can be snapped in the plastic bag easily, was inserted into the bag through one of the holes created. As the TGM sampling port, a 5 cm length PFA tubing (6 mm o.d. × 4 mm i.d., Yodoflon, Yodogawa Hu-Tech Co. Ltd., Osaka, Japan) was inserted into another opening. Those openings were then sealed with Parafilm. The pressure inside the fluorescent tubes is usually below the atmospheric pressure, and this pressure difference causes instantaneously vacuuming of the bag into the tube immediately after snapping the tube. This often results in the bag tearing. To avoid this problem, mercury-free air was introduced into the bag in advance, as an addition to the reinforcement of the bag with curling tape. The fluorescent tube was then snapped, and no air was further introduced into the bag until the bag expanded. The air containing TGM from the fluorescent tube was then sampled through a handmade gold-coated sand trap (a 200 mm length ×10 mm o.d. × 8 mm i.d. quartz tube (COSMOS VID, Fukuoka, Japan) stuffed with approximately 1 g of gold-coated silica sand grains (Nippon Instruments Co, Inc., Osaka, Japan) in the center) under a flow rate of 0.3 L min–1. The homemade traps were needed for a larger trapping capacity than commercially available gold-coated sand traps. Breakthrough TGM from the homemade sampling trap during a TGM sampling from a fluorescent tube was previously tested using double sampling traps connected in series, and typical breakthrough TGM (i.e., TGM trapped in the backup trap) was 12–37 pg per sample. These amounts were negligibly small compared to the TGM trapped in the front trap, typically on the order of micrograms.

Figure 7

Schematic illustration of total gaseous mercury sampling method from fluorescent tubes.

After sampling TGM, the electrode ends of the fluorescent tube were removed, and then the glass of the fluorescent tubes was smashed into smaller pieces in the plastic bag. Following the method suggested for the analysis of a solid sample,[36] half or the full weight of the smashed glass was transferred into a 1 L beaker, and the glass pieces were submerged in 100 or 150 mL of concentrated sulfuric acid mixture (sulfuric acid/nitric acid/perchloric acid = 5:1:1, ACS grade, 95%, 61%, and 70% purity, respectively, Kanto Chemical Co. Inc., Tokyo, Japan) and left on a 503 K hot plate for 1 h to oxidize and dissolve adsorbed Hg on the glass into the solution. The efficiency of this extraction was tested by observing the concentration changes over the two-day submerging period (see SI subsection S1). This extraction method was also applied to electrode samples with 50 mL of the sulfuric acid mixture. But it should be noted that the analysis of the electrodes was done only for six fluorescent tubes (40w-8, -9, -10, N40w-1, -2, and N32w-1 listed in Table S3).

The TGM collected in the homemade sampling trap from the fluorescent tube sample was oxidized and dissolved in an acid solution. First, 100 mL of 40% (v/v) inversed aqua regia trapping solution, which was made of 2:1 nitric and hydrochloric acids (ACS grade, 61% and 36% purity, respectively, Kanto Chemicals Co. Inc.), was poured into a 5 L Tedlar bag through a PTFE stopcock (AS ONE Co. Ltd.). The sampling trap, which captured the TGM, was attached to the Tedlar bag, and then the sampling trap was heated to 873 K for approximately 10 min under the 0.5 L min–1 flow of Hg-free air so that the majority of the captured TGM was converted to GEM and then flushed into the Tedlar bag. The background mercury concentration in the zero air was 13 pg m–3 or less. The efficiency of this transfer was evaluated by collecting the residual GEM in the sampling trap after this transfer using a conventional gold-coated sand trap (4 mm i.d. × 160 mm length, Nippon Instruments Corp.) for another 10 min. The residual GEM in the conventional trap was then quantitatively analyzed using a cold-vapor atomic fluorescence spectrometer or CV-AFS (WA-5F, Nippon Instruments Co, Inc.).

The Tedlar bag enclosing the GEM and the 100 mL trapping solution was shaken one to three times per day (a few minutes each time), and the bag was left for 11 days in total at room temperature. This extraction period is sufficient to oxidize and dissolve nearly 100% of the enclosed GEM into the trapping solution.[35] Prior to retrieving the trapping solution from the bag, the residual GEM inside the Tedlar bag was sampled using a conventional gold-coated sand trap through a soda lime (Kanto Chemicals Co. Inc.) water trap to evaluate the trapping efficiencies. This residual GEM was also quantitatively analyzed by a CV-AFS.

Stable Isotope Ratio Measurements by a CV-MC-ICP-MS

A sample solution and a reducing reagent of 5% (w/w) tin(II) chloride dehydrate (97% purity, Kanto Chemicals Co Inc.) in 10% (v/v) hydrochloric acid (ACS grade, 35% purity, Kanto Chemicals Co Inc.) solution were introduced into a cold-vapor generator (CV) (HGX-200, Teledyne CETAC Technologies, Inc., Gaithersburg, NE, USA) by a peristaltic pump (Perimax, Spetec GmbH, Erding, Germany) under a flow rate of 0.58 mL min–1. Hg2+ in the sample solution was instantaneously reduced to GEM by Sn2+, and the produced GEM was flushed into a multicollector inductively coupled plasma mass spectrometer or MC-ICP-MS (Neptune Plus, Thermo-Fisher Scientific GmbH, Bremen, Germany) for five stable mercury isotope ratio measurements (199Hg/198Hg, 200Hg/198Hg, 201Hg/198Hg, 202Hg/198Hg, and 204Hg/198Hg) using the cup settings in SI Table S4. Thallium aerosols were produced from the 25 ng g–1 standard reference material (SRM 997, NIST, Gaithersburg, MD, USA) solution using a dried aerosol generator (Aridus II with SP820A nebulizer, Teledyne CETAC Technologies) to correct for artificial mass-dependent isotope fractionations occurring in the ICP system, based on the shift in the reference 205Tl/203Tl ratio. The generated aerosols were introduced at the end of the CV system, and then the mixture of GEM and thallium aerosols was introduced into the MC-ICP-MS together. Prior to the isotope ratio measurement, a sample was analyzed by the CV-MC-ICP-MS to determine the concentration first. The sample solution was then diluted to an equivalent concentration of the working standard, a 10 ng g–1 diluted solution of the standard reference material 3133 (SRM 3133, NIST). In the isotope measurement method, an operational software of the CV-MC-ICP-MS was programmed to preinject a sample solution for 3 min of flushing, to measure the isotope ratios in 60 cycles with 3 blocks (180 measurements in total), and then to average out the measured isotope ratios for the sample. The SRM 3133 of 10 ng g–1 was measured before and after the sample run (i.e., the standard bracketing method), and the averaged isotope ratios over the two runs were used as the reference for determining the sample isotope ratio on the SRM 3133 scale. The typical sensitivity of our CV-MC-ICP-MS under the conditions above was 0.19 V min g ngHg–1 mL–1 for 202Hg. The typical uncertainty determined from the replicate Hg isotope ratio measurements under our measurement conditions (180 measurements per sample) was better than 0.08 ‰ (the expanded uncertainty, corresponding to 2× SE). Herein, Hg isotope ratios are expressed using the delta notation:where “x” stands for the stable Hg isotope with a mass x, and the bracketed quantitative isotope ratios with subscripts “sample” and “3133” stand for the stable Hg isotope ratios of mass x relative to mass 198 for the sample and SRM 3133, respectively. To determine the measurement precision and accuracy another standard, SRM 8610 (NIST), was routinely analyzed. The referenced isotope ratios of both SRM 3133 and 8610 are provided by NIST.