Removal of mercury contamination on primary mass standards by hydrogen plasma and thermal desorption

Without any cleaning, the historical Pt reference mass C3 was first analysed by XPS. Figure 1 shows a widescan spectrum. It proves that the standard is pure platinum metal. Surprisingly, we found, in addition to the natural contamination with carbon and oxygen compounds, a huge amount of mercury. Angle-resolved XPS (ARXPS) measurements revealed the following layer structure: mercury is directly adsorbed on the Pt metal surface with a sharp interface between mercury and platinum. The natural contamination by carbon and oxygen compounds is at the top. As assumed earlier for all reference kilograms [18], this old Pt standard, conserved at the mass laboratory for over 150 years, is completely contaminated by mercury. The observed layer structure is in good agreement with earlier findings [18]. From our XPS measurements we calculated a layer thickness of (2.0 ± 0.5) nm. Using the conversion formula from [18] we estimate a total mass of mercury of (200 ± 30) µg for the standard, which is much more than 22 µg to 32 µg found previously [18]. As we will show later, this unexpectedly high amount of mercury can be explained by the extremely long exposure time of more than a century.

Zoom In Zoom Out Reset image size

In order to remove the mercury by thermal desorption, the standard was heated twice for 2 h at 200 °C to 220 °C in a vacuum furnace under an argon atmosphere at 5 × 10⁴ Pa. Figure 2 shows the evolution of the platinum, mercury and carbon photoelectron lines upon individual steps of treatment. The standard was first analysed by XPS (figure 2, (i)) and then cleaned by hydrogen plasma to reduce the natural carbon and oxygen contamination, followed by an additional XPS analysis (figure 2, (ii)). There was a significant decrease in carbon after hydrogen cleaning resulting in an increase in the mercury and platinum intensities, due to the reduction of carbon and oxygen. After each heating, the standard was cleaned by hydrogen plasma to reduce again the natural contamination of carbon and oxygen arisen during the heating process, and to reach a similar level of contamination for a better comparison of mercury removal (figure 2, (iii+iv)). The measurements show a remarkable reduction of mercury after heating. We estimate a layer thickness of 0.6 nm after the second heating, which corresponds to a reduction of 1.4 nm.

Zoom In Zoom Out Reset image size

We attribute this reduction to thermal desorption of mercury at elevated temperatures. The remaining contamination is still located at the surface. With ARXPS measurements we found a sharp interface between Hg and Pt, but no indication of Hg diffusing into Pt. This is in good agreement with earlier findings [18]. The authors outlined that the diffusion of Hg into bulk Pt at room temperature was extremely small; the time needed for a mercury atom to move one atomic spacing exceeds 100 years. Thus, interdiffusion of Hg into bulk Pt can be ruled out also in our experiment. Figure 3 shows the intensity of bulk Pt and Hg obtained from XPS measurements as a function of the emission angle. The dashed lines are best fits to the data based on a physical model for bulk (Pt) and layer (Hg, C) materials following strictly an approach by Cumpson and Seah [18]:

$$\begin{eqnarray} &&\fl I_{{\rm Pt}}(\Theta)=I_{{\rm Pt}}^{\infty}\times \exp \left( -\frac{d_{{\rm Hg}}}{\lambda_{{\rm Hg}}\times \cos \left( \Theta \right)} \right)\nonumber\\&&\times\, {\rm \exp}\left( -\frac{d_{{\rm C}}}{\lambda_{{\rm C}}\times \cos \left( \Theta \right)} \right) \end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray} &&\fl I_{{\rm Hg}}(\Theta)=I_{{\rm Hg}}^{\infty}{\rm \times}\left[ {\rm 1-\exp}\left( -\frac{d_{{\rm Hg}}}{\lambda_{{\rm Hg}}\times \cos \left( \Theta \right)} \right) \right]\nonumber\\&&\times\, {\rm \exp}\left( -\frac{d_{{\rm C}}}{\lambda_{{\rm C}}\times \cos \left( \Theta \right)} \right) \end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray} &&\fl I_{{\rm C}}(\Theta)=I_{{\rm C}}^{\infty}{\rm \times}\left[ {\rm 1-\exp}\left( -\frac{d_{{\rm C}}}{\lambda_{{\rm C}}\times \cos \left( \Theta \right)} \right) \right] \end{eqnarray} \tag{ 3 }$$

The layer thicknesses of mercury, d_Hg, and carbon, d_C, are fitting parameters using attenuation lengths of λ_Hg = 3.05 nm and λ_C = 2.8 nm which are in good agreement with published values [25, 26].

Zoom In Zoom Out Reset image size

Our findings are in good agreement with results obtained by thermogravimetry (TG) and scanning electron microscopy (SEM) on a Pt–20% Rh alloy [27]. The authors found the desorption of mercury occurs in three well-defined temperature ranges. From ambient to 175 °C mainly bulk mercury is desorbed very rapidly, resulting in a loss of 80.8% of the total mercury. The remaining mercury was found to form an intermetallic PtHg₄ surface film. In a second step from 175 °C to 224 °C the intermetallic PtHg₄ is decomposed into PtHg₂ and Hg which desorbs and leads to a further reduction of 11% of the total mercury deposited on the surface. The remaining intermetallic film does not completely cover the surface. The final removal of mercury in the range from 224 °C to 305 °C is attributed to the thermal decomposition of intermetallic species PtHg₂ and RhHg₂. For the maximum temperature of 200 °C to 220 °C used in our experiment we assume a thermal desorption of bulk Hg and an incomplete intermetallic layer of PtHg₂ and RhHg₂.

In parallel, we tried to measure gravimetrically the loss in mass under vacuum due to the evaporation of mercury. Unexpectedly, we found huge changes in mass of 2 mg to 3 mg which cannot be explained by the removal of mercury. Similar effects have been reported for standards manufactured by forging before the metre convention. Such standards, including our standard C3, exhibit a density which is up to 5% smaller than for cast Pt, indicating the presence of porosities and inclusions which are held responsible for loss in mass when measured under vacuum [28]. Historically, as a consequence vacuum measurements were stopped around the year 1870.

In order to verify the results with high-quality materials used in today's mass laboratories, we continued this study with materials that are currently used for mass prototypes standards such as PtIr as well as a material for potential future standards made of an AuPt alloy [23]. First, we focused on the contamination mechanism to find out the speed of action. Therefore, we exposed polished samples of both materials to saturated mercury vapour in air. The samples were placed in a glass exsiccator which contained 1 cm³ of spilled mercury and monitored the level of contamination on the samples by XPS.

Figure 4 shows the rapid increase in mercury layer thickness with time. Within a few minutes the intensity of the photoelectron lines of mercury increased dramatically. The speed of contamination is unexpectedly fast and initial saturation is reached after 1 h with a layer thickness of (0.63 ± 0.05) nm.

Zoom In Zoom Out Reset image size

Surprisingly, the mercury contamination on the AuPt alloy was much lower, and it starts with a delay of 1 h as shown in figure 5. The reason for this delay is not yet understood, but might be caused by natural contamination with carbon and oxide compounds. Surface reactions are quite complex and impurities can play an important role, as shown for Au [29–32]. For other materials such as Pd it is known that oxygen can promote or prevent chemical reaction [33, 34].

Zoom In Zoom Out Reset image size

It is generally accepted that the growth of natural contamination by carbon and oxygen compounds on a clean mass standard is initially very rapid for the first few days and weeks, but then slows down and further accumulation of contaminants is small. A rich summary of observation and results can be found elsewhere [35]. Very recently, a simple two-step model was presented which can explain the short-term as well as the long-term growth of contaminants [20]. It is the first model that is based on physical principles, following concepts of Langmuir [36]. For the natural contamination with carbon and oxygen compounds it is in good agreement with numerous observations [35], and furthermore, as we will show in the following, it can also be applied to the mercury contamination. Molecules of mercury vapour have a certain probability S (e.g. S = 1) to stick on a clean metal surface. The amount of material accumulated on the surface in time, δm/δt, and the loss of clean surface, δA_c/δt, are consequently proportional to the clean surface A_c itself.

$$\begin{equation} \frac{\delta m}{\delta t}\sim \frac{\delta A_{\rm c}}{\delta t}\sim A_{{\rm c}} \end{equation} \tag{ 4 }$$

One solution to this differential equation is an exponential function in time (5). Integration over time delivers the total mass Δm_tot of the saturated contamination layer (6):

$$\begin{equation} m(t)={\rm const} \times {\rm e}^{-\frac{t}{\tau}} \end{equation} \tag{ 5 }$$

$$\begin{equation} \Delta m(t)=\Delta m_{\rm tot}\times (1-{\rm e}^{- \frac{t}{\tau}} ) \end{equation} \tag{ 6 }$$

Since the clean surface area is limited and decreasing, the gain in mass for this first step of contamination is self-limited. The dashed line in figure 4 shows the best fit to our data using (6). The initial and self-limited growth model is in fair agreement with our data. Once the surface is covered, the sticking probability for mercury on mercury is completely different from the initial sticking factor of mercury on metal, but does not further change with time in the simplest case. Hence, this second step gives rise to an additional but small gain in mass, which never stops, and which is superimposed on the initial growth of the first layer. Under this assumption the sticking probability is unchanged upon increasing layer thickness; the multilayer growth would be linear in time. As reported earlier [27] the first layer of mercury forms intermetallics. Therefore, it is most likely that the sticking probability for Hg on intermetallics is different from the initial one. Also for further layers it can change with each layer. In this more general case the growth of individual layers Δm_i(t) can be described as (7), and the total mass of mercury increases as a series of Langmuir layers as described in (8).

$$\begin{equation} \Delta m_{i}(t)={\Delta m}_{i,{\rm tot}}\frac{\Delta m_{i-1}(t)}{{\Delta m}_{i-1,{\rm tot}}}\times (1-{\rm e}^{- \frac{t}{\tau_{i}}}) \end{equation} \tag{ 7 }$$

$$\begin{equation} {\Delta m}_{\rm tot}(t)=\sum\limits_i {\Delta m_{i}(t)} \end{equation} \tag{ 8 }$$

The fraction Δm_i−1(t)/Δm_i−1,tot represents the coverage factor of the layer (i − 1).

Very recently the short-term adsorption of natural contaminants was studied gravimetrically with 1 kg PtIr artefacts [20] and with a gold-coated surface of a QCM [37]. For both measurements the data are in almost perfect agreement with our exponential model. For the long-term contamination the adsorption is small but uniformly continuous. Cumpson and Seah analysed weighing data of the UK national standard No 18 and found that the mass of the carbonaceous contamination grows as the square-root of time over a period of 60 years [38]. Based on data taken in the period from 1 month to 6 years after cleaning, the same standard shows good agreement with the time-root dependence for the first 15 months, but is in poorer correlation for the long-term data [35, 39]. However, with a thickness-dependent sticking probability also a time-root increase in mass can be well explained by our model.

As explained earlier, the standard C3 was stored in the mass laboratory of METAS in different buildings over the years. The level of mercury contamination is not known, but was most likely not the same at different places. Therefore, it is difficult to determine the long-term growth mechanism and we can only deliver an average value. From our historical standard C3 we can estimate an initial increase in mass of 25 µg right at the beginning of Hg exposure, followed by a long-term average mass gain of roughly (1.3 ± 0.1) µg/year over 120 years, resulting in a total mass of the mercury layer of 200 µg. Since this long-term growth is very small, it is invisible on a short time scale.

In 2003 the mass laboratory moved into a newly built facility which was proven to be completely free of Hg contamination. Two new PtIr standards presented in [20] were stored for 8 years under identical conditions next to C3. By XPS we found no evidence for mercury contamination on these standards. As a matter of fact we found 200 µg Hg on the C3 standard by XPS, after 10 years of storage in this mercury-free environment. This high amount of mercury accreted on the surface is consistent with earlier findings [19], where no asymptotic limit to the mass increase is predicted. Furthermore, it clearly shows that the build-up of a Hg multilayer cannot be described by an equilibrium process, such as outlined in the Brunauer, Emmett and Teller (BET) theory [40], where a rapid loss of Hg would be expected when the exposure to Hg vapour is stopped. In the period from 2003 to 2013 C3 was used as a check-standard for the calibration of primary stainless steel standards, which are not subject to mercury contamination. From these gravimetric measurements we can exclude a loss in mass for C3 with an uncertainty of 12 µg. This is in good accordance with earlier findings [19] on a PtIr sputter-coated QCM, where the Hg mass remains constant also after interruption of exposure to Hg vapour [19]. The authors attribute this to a diffusional process into sputtered PtIr film which exhibits a high density of grain boundaries and dislocations.

In order to determine the amount of mass accumulated on the surface we exposed a 0.1 mm thick PtIr foil to mercury vapour and studied the increase in mass gravimetrically and the surface chemical state by XPS. The foil was used as rolled, only cleaned by rubbing with ethanol/ether soaked chamois. Surprisingly, the contamination on the foil by mercury is much slower than for the polished samples analysed before. To investigate the influence of the initial surface conditions on the growth of mercury, half of the foil was polished with diamond paste. The entire foil was cleaned again with solvents and the analysis was continued. On the polished area the mercury could completely be removed whereas on the unpolished area it remained unchanged as illustrated in figure 6.

Zoom In Zoom Out Reset image size

XPS spectra clearly showed that Hg contamination can be removed by polishing. On the other hand, rubbing with ethanol/ether soaked chamois does not, as expected and observed earlier [18], affect the mercury film, but reduces the natural contaminants slightly. However, this reduction of carbon and oxygen compounds does not noticeably influence the further growth of the mercury layer as can be seen on the unpolished area. In contrast, on the polished area the speed of contamination and the level of saturation are increased, showing the same amount of mercury as observed earlier for the polished bulk samples (see figure 4). For the speed of contamination right at the beginning of exposure we calculated (0.15 ± 0.02) nm h⁻¹ for the polished area and (0.07 ± 0.02) nm h⁻¹ for the unpolished area. This clearly indicates that the surface can be activated by polishing, resulting in a much higher speed and saturation level of mercury contamination than for the unpolished surface as rolled. On the foil sample we found no evidence that the natural contamination by species of carbon and oxygen can delay or suppress contamination.

Earlier studies clearly demonstrated that the chemical activity of metal surfaces can be strongly enhanced by mechanical treatment [41, 42] as shown first for steel [43], Ni [44, 45] and Au [41, 46]. The authors attributed this mechanochemical or tribochemical enhanced activity to surface defects, such as kink sites and step edges, which are energetically favourable [41].

Since our first mass determination of the PtIr foil was not successful, the gravimetric measurements were continued with another set of polished small samples with a total surface area of 20 cm². After 2 h of exposure to Hg vapour we found a gain in mass of (5.0 ± 0.2) µg, and saturation is reached at (7.8 ± 0.2) µg after 18 h. The corresponding evolution of the photoelectron line intensities of Pt_4f and Hg_4f of the bulk and layer material, respectively, is illustrated in figure 7. Using the gravimetric results above, the total gain in mass can be scaled for a 1 kg PtIr reference standard with a total surface area of 72 cm² to (18 ± 1) µg after 2 h, and (28 ± 1) µg for a saturated surface. The gravimetric measurements were performed under ambient conditions, where also natural contamination by species of carbon and oxygen takes place. The mass increase for a 1 kg PtIr due to this natural contamination was estimated to be (9.4 ± 1.5) µg [18]. Recent gravimetric measurements reported a much smaller value of (1.63 ± 0.1) µg [20]. Taking this into account the mass of the saturated mercury layer can be estimated to be 19 µg to 26 µg. This amount is in good agreement with earlier studies [19] on a PtIr coated QCM. The authors report about a very rapid initial adsorption of roughly 28 µg extrapolated/corresponding for a PtIr standard, followed by a slow time-root-dependent diffusional growth of mercury on a long term.

Zoom In Zoom Out Reset image size

In addition, ARXPS measurements were subsequently made of the PtIr sample to verify the layer structure and to determine the layer thickness of mercury and carbon (figure 8). The measurements revealed the same layer structure as for the mass standard C3 with a layer thickness of (0.6 ± 0.1) nm for mercury and (1.2 ± 0.2) nm for carbon which is in good agreement with the result above. The fitted curves using equations (1) to (3) are in good accordance with the measured data, indicating a sharp interface between the individual layers.

Zoom In Zoom Out Reset image size

In order to perform ARXPS of the mercury layer itself without any natural contamination, we tried to remove the carbon and oxygen compounds by cleaning the samples for 1 h with hydrogen plasma. Surprisingly, and completely unexpectedly, the mercury concentration on the PtIr and AuPt surfaces decreased upon this treatment. Subsequent steps of hydrogen plasma cleaning clearly demonstrated that mercury can be removed by hydrogen plasma.

For a more systematic study of the mercury removal by hydrogen plasma we exposed Hg-saturated PtIr and AuPt samples to hydrogen plasma. Figure 9 shows the XPS spectra of the bulk material and mercury layer right after the exposure to mercury vapour (solid black line) and after each individual step of cleaning (dashed coloured lines). The measurements show equal contamination levels of Hg for both materials. Figure 10 shows the decrease in the mercury layer thickness as a function of cleaning time.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Hydrogen plasma is well known for its capability to remove organic substances, such as carbon and oxygen compounds, but normally it does not react with inorganic compounds. It is really surprising that it can remove mercury from the surface of metals. Mercury is a hazardous air pollutant and was therefore intensively studied in the past. It is generally accepted that group 12 metals such as Hg, Cd, Zn in their ground state do not react with molecular hydrogen, because large energy barriers exist and the overall reactions are endoergic [20, 21]. But in the presence of an electrical discharge, gaseous HgH₂ can be generated from the direct reaction of mercury vapour with molecular hydrogen, as demonstrated previously [20, 21]. The authors used conditions which are almost identical to our hydrogen plasma: in a low-pressure hydrogen atmosphere of 67 Pa to 330 Pa, mercury vapour was synthesized to mercury dihydrate (H–Hg–H) by dc electrical discharge. The cleaning effect of a hydrogen plasma is due to surface chemical reactions [16, 17], since the sputter yield of ionized hydrogen can be neglected [16, 47]. Therefore, we attribute the removal of mercury in our experiment to the chemical reaction with hydrogen and the formation of mercury dihydrate; the generated gaseous HgH₂ is pumped away.