Advanced Search
RSS LinkedIn Twitter

Journal Archive

Platinum Metals Rev., 2008, 52, (3), 144


Noble Metal Catalysts for Mercury Oxidation in Utility Flue Gas


    • Albert A. Presto
    • Evan J. Granite**
    • National Energy Technology Laboratory, United States Department of Energy,
    • PO Box 10940, MS 58-106, Pittsburgh, PA 15236-0940, U.S.A.



Article Synopsis

The use of noble metals as catalysts for mercury oxidation in flue gas remains an area of active study. To date, field studies have focused on gold and palladium catalysts installed at pilot scale. In this article, we introduce bench-scale experimental results for gold, palladium and platinum catalysts tested in realistic simulated flue gas. Our initial results reveal some intriguing characteristics of catalytic mercury oxidation and provide insight for future research into this potentially important process.

1. Introduction

Coal-fired utility boilers are the largest anthropogenic emitters of mercury in the United States, accounting for approximately one third of the 150 tons of mercury emitted annually (1, 2). In 2005, the U.S. Environmental Protection Agency (EPA) announced the Clean Air Mercury Rule, to limit mercury emissions from coal-fired utility boilers to 15 tons annually, approximately 30% of 1999 levels, by 2018 (3). At the time of publication (July 2008) this measure is under legal dispute. Of alternative legislative proposals to regulate mercury along with other pollutants, most would require a 90% mercury reduction, with deadlines for control varying from 2011 to 2015. Mercury exists in three forms in coal-derived flue gas: elemental (Hg0), oxidised (Hg2+) and particle-bound (Hg(p)) (4). During combustion, mercury is liberated from coal as Hg0. As the flue gas cools, some of the Hg0 is oxidised, presumably to mercury(II) chloride (HgCl2) because of the large excess of chlorine present in coal. Both Hg0 and Hg2+ can enter the particulate phase by adsorption onto fly ash particles (5).

Hg2+ and Hg(p) are relatively easy to remove from flue gas using typical air pollution control devices. Hg(p) is captured, along with fly ash particles, in the particulate control device. Hg2+ is soluble in water, and is therefore removed with high efficiency by wet flue gas desulfurisation equipment (6). Hg0, on the other hand, is difficult to capture. It is insoluble in water and is therefore not removed during flue gas desulfurisation. Activated carbon injection will remove both Hg0 and Hg2+, and currently this is the best method for removing Hg0 from flue gas (6).

In addition to the Clean Air Mercury Rule, the U.S. EPA also enacted the Clean Air Interstate Rule, which requires reductions in NOx and SO2 emissions in twenty-eight states (7). An expected consequence of this law is increased use of wet flue gas desulfurisation for SO2 removal (8). Among the technologies being considered for mercury abatement in coal-fired boilers is therefore the combination of a catalyst and a wet scrubber; the catalyst oxidises Hg0 to Hg2+, and the oxidised mercury is subsequently absorbed by the scrubber solution. Catalysts capable of significant conversion (> 80%) of Hg0 to Hg2+ could have tremendous value because the oxidised mercury can be removed concurrently with acid gases during flue gas desulfurisation.

Mercury oxidation catalysts can be employed in either of two configurations. In the ‘co-benefit’ application, selective catalytic reduction (SCR) catalysts with sufficient activity for mercury oxidation are installed upstream of a flue gas desulfurisation scrubber. The primary function of the SCR catalyst is to reduce NOx concentration in flue gas, and some power plants will need to install SCR to achieve compliance with the Clean Air Interstate Rule (9). During operation, NO is reduced by NH3, which is injected upstream of the SCR, at temperatures above 300°C. When the NH3 is consumed (catalysts are typically oversized to prevent NH3 slip), the catalyst is available for mercury oxidation.

Mercury oxidation catalysts can also be installed specifically for mercury control. In this case the catalyst is located downstream of the particulate control device, where the flue gas temperature is approximately 150°C. The lower temperature favours Hg0 adsorption, and may therefore lead to more efficient mercury oxidation. Catalysts tested in this configuration include gold, palladium and vanadium-tungsten (10).

The catalytic oxidation of mercury to mercury(II) chloride typically assumes an overall reaction between Hg0 and HCl, for example, Reaction (i):

  Hg + 2HCl → → HgCl2      (i)

Cl2 may also play a role in the formation of HgCl2, but the equilibrium concentration of Cl2 is only ∼ 1% of the HCl concentration. Several key questions exist regarding this reaction. Specifically, the reaction mechanism is uncertain. The bimolecular reaction between two species adsorbed to a surface can be described by a Langmuir-Hinshelwood mechanism (11), Reaction (ii), Reaction (iii), Reaction (iv) and Reaction (v):

  A(g) ⇆ A(ads)        (ii)

  B(g) ⇆ B(ads)        (iii)

  Aads + B(ads) → AB(ads)    (iv)

  AB(ads) → AB(g)       (v)

For this mechanism, the rate of reaction is dependent on the concentrations of reactants A and B, the adsorption equilibrium constant (Ki), and the rate constant for the surface reaction (ksurf). Mercury could also react via an Eley-Rideal mechanism, which is the reaction between a surface-bound species and a gas-phase (or weakly adsorbed) species, Reaction (vi) and Reaction (vii):

  A(g) ⇆ A(ads)        (vi)

  A(ads) + B(g) → AB(g)    (vii)

Eley-Rideal and Langmuir-Hinshelwood mechanisms can be inferred by surface analysis of used catalysts to confirm adsorption of specific reactants such as mercury and HCl. Additionally, pre-exposure of the catalyst to an oxidant, followed by mercury oxidation in the absence of the oxidant, would suggest either a Langmuir-Hinshelwood reaction or an Eley-Rideal reaction with the oxidant as the adsorbed species. A Langmuir-Hinshelwood mechanism can also be identified via chemical kinetics, though the relative adsorption behaviour of the reacting species may complicate analysis. In some cases, a Langmuir-Hinshelwood mechanism is characterised by a reaction that is first-order in each of the reactants (for example, Hg0 and HCl). However, if one species saturates the surface, the reaction order with respect to the saturating species can be −1 (11).

Granite et al. (12) proposed that mercury oxidation could occur via a Mars-Maessen (13) mechanism. In this mechanism, adsorbed Hg0 would react with a lattice oxidant (either O or Cl) that is replenished from the gas phase. Reaction (viii), Reaction (ix), Reaction (x), Reaction (xi) and Reaction (xii) show the Mars-Maessen mechanism for the reaction of an adsorbed species (for example, Hg0) with lattice oxygen:

  A(g) ⇆ A(ads)            (viii)

  A(ads) + MxOy → AO(ads) + MxOy–1   (ix)

  MxOy–1 + ½O2 → MxOy        (x)

  AO(ads) → AO(g)           (xi)

  AO(ads) + MxOy → AMxOy+1      (xii)

The Mars-Maessen mechanism can be confirmed by the observation of mercury oxidation in the absence of gas-phase oxygen or chlorine, respectively (through variations of Reaction (ix)).

Medhekar et al. postulated that catalytically active mercury(II) chloride forms on the surfaces of many materials (14). They observed the reaction between elemental Hg and Cl2 catalysed by Inconel® (an austenitic nickel-based alloy), quartz, stainless steel and Teflon®-coated stainless steel. Medhekar et al. found that many surfaces can catalyse the reaction between Hg and Cl2 and that the surfaces are difficult to passivate with oxygen or fluorine. This suggests that the adsorbed HgCl2 product is the actual catalyst. Ariya et al. observed that the Hg + Cl2 reaction proceeded faster when the reactor surface was covered with the reaction products of Hg + Br2 than with a clean surface, suggesting a similar effect (15).

To date, none of the above mechanisms has been verified as the dominant mechanism for catalytic mercury oxidation. Mercury appears to react from an adsorbed state (10), but the phase of the HCl is uncertain. Furthermore, it is unknown whether Hg and HCl react directly, or if another species, such as mercury(II) oxide (HgO), is formed first (16, 17). The role of other flue gas species, specifically NO and SO2, is unclear, and the behaviour of these species in mercury oxidation may depend strongly upon the nature of the catalyst. The deactivation mechanisms for the various mercury catalysts are also unknown.

In a previous article (10), we asserted that further research into the fundamental aspects of catalytic mercury oxidation is required to answer these significant questions. The information presented in this article is part of an ongoing effort toward that goal. We present initial results for mercury oxidation over three noble metal catalysts, Au, Pd and Pt. We envision that these materials could be used downstream of particulate control devices as mercury-specific catalysts. The results reveal several important aspects of the catalysts, and highlight some of the differences between Au and the platinum group metals (pgms) in mercury oxidation.

2. Experimental

Noble metal catalyst samples were exposed to mercury in a bench-scale packed bed reactor that has been described previously (18) and is shown schematically in Figure 1. The bench-scale assembly consisted of a quartz tube reactor, 22 mm internal diameter and 61 cm long, contained in a clamshell tube furnace. A catalyst bed containing approximately 0.5 g of catalyst was placed in the reactor and was supported by a quartz frit. Alumina beads were placed above the catalyst bed to ensure plug flow. A mass spectrometer was located downstream of the packed bed to monitor potential side reactions such as the formation of flue gas halides.

Fig. 1

Schematic diagram of the packed bed reactor

Schematic diagram of the packed bed reactor


The catalysts used in this study were 1 wt.% Au, Pd and Pt, respectively, supported on 2 mm alumina beads (Johnson Matthey PLC). The BET surface area of the alumina beads was approximately 200 m2 g−1. The catalysts were air calcined at elevated temperatures, thereby decomposing the precursor salts. It is very unlikely that the Au particles are of nanometre size because they sinter at the calcination temperatures. Therefore, the Au catalyst should have typical properties of bulk Au.

The mercury concentration and speciation exiting the packed bed were measured using a P S Analytical model 10.525 ‘Sir Galahad’ continuous mercury monitor. A wet conditioning system with two channels for determining elemental and total mercury was placed upstream of the mercury monitor. The elemental mercury channel used an impinger filled with KCl solution to remove Hg2+ from the sample, and the total mercury channel used a SnCl2/HCl solution to reduce Hg2+ to Hg0. Both the KCl and SnCl2/HCl impingers were followed by impingers containing NaHCO3 solution that captured the acid gases SO2 and HCl.

The mass of catalyst was selected to provide a small (10 to 50%) conversion of Hg0 to Hg2+. Very high or very low fractional conversions are unfavourable because they complicate the interpretation of the experimental results. The precision of the mercury monitor used in this study is approximately 10 to 20% (1σ); therefore gas mixtures containing less than 10%, or more than 90%, oxidised mercury are statistically indisinguishable from gas mixtures that contain 0% or 100% oxidised mercury, respectively. The instrument precision is summed in quadrature for kinetics measurements, which require both elemental and oxidised mercury concentrations. The total uncertainty for the kinetics measurements is therefore 15 to 30% (Figure 2), and small fractional mercury conversions are nearly indistinguishable from the noise.

Fig. 2

Fractional mercury oxidation conversion across a gold catalyst as a function of time, from a typical experiment. The open symbols, ◊, indicate the average conversion for the baseline and [HCl] = 75 ppm time periods. Mercury conversion is expected to be nominally constant during these phases of the experiment. The error bars, I, show the 1σ level of precision

Fractional mercury oxidation conversion across a gold catalyst as a function of time, from a typical experiment. The open symbols, ◊, indicate the average conversion for the baseline and [HCl] = 75 ppm time periods. Mercury conversion is expected to be nominally constant during these phases of the experiment. The error bars, I, show the 1σ level of precision


Difficulty in obtaining consistent mercury measurements in real or simulated flue gas at levels of parts per billion by volume is a common problem. For example, results from field studies indicate significant variability in mercury capture efficiency during activated carbon injection. Specifically, during long-term injection tests, individual measurements of mercury capture efficiency (timescale of minutes to hours) can differ significantly from the long-term results (timescale of months) (19). When all of the potential sources of experimental uncertainty are considered, it is our opinion that the precision presented here is appropriate for our experimental system, and is consistent with previous work from this laboratory (20).

The catalysts were exposed to simulated flue gas containing O2, CO2, HCl, SO2, Hg0 and N2. Each catalyst sample was tested using the ‘baseline’ simulated flue gas detailed in Table I. The baseline conditions are roughly consistent with previous work from this laboratory, with the exception that NO was excluded from all but one experiment in this study because its presence interferes with mercury detection by the mercury monitor. The baseline conditions serve two purposes: first, they provide the basis for a like-for-like comparison for each of the catalysts tested. We will refer to the mercury oxidation rate measured in the presence of the baseline simulated flue gas as the ‘baseline reaction rate’. The effects of excursions from the baseline gas composition are measured as deviations from the baseline reaction rate. Second, because the baseline conditions are used at the start of each experiment, they provide a way to measure catalyst deactivation over time.

Table I

Typical Simulated Flue Gas Conditions for the Experiments Conducted in This Study

Parameter Units Baseline value Range

[O2] % 5.25 0–5.25
[SO2] ppm 500 0–1000
[NO] ppm 0 500
[CO2] % 12.5
[HCl] ppm 50 0–100
[CO] ppm 0 0–35
[Hg] μg Nm−3 10 6–18
Temperature °C 149 138–160
Flow rate slpm 8 8–10

Excursions from baseline conditions were undertaken in order to gain a more complete understanding of the reaction order with respect to mercury and/or HCl, potential side or interfering reactions, and apparent activation energy. The effects of other flue gas species, such as SO2 and CO, were also considered. These species can possibly bind to and deactivate the catalyst, or can participate in parallel reactions, such as flue gas halide formation, that may inhibit mercury oxidation.

The catalyst samples were tested for approximately six hours per day for several days. This procedure contrasts with packed bed experiments conducted by this group using mercury sorbents. In those experiments (18) the sorbent was exposed for six hours and removed from the packed bed reactor. The procedural difference between catalyst and sorbent experiments is intended to partially mimic the application of the two technologies in power plants: sorbents are typically injected and subsequently disposed of, whereas catalysts need to stay in place for months or years in order to be an economically viable option for mercury control (21). Exposing the catalysts for multiple six-hour experiments also allows for an initial investigation into the flue gas species and/or processes that can deactivate the catalyst.

3. Results

The data presented in this article were analysed according to the chemical kinetics framework previously outlined by this group (22). The catalysts are compared by considering the overall reaction rate for Hg2+ formation, measured in (mol Hg2+) (g catalyst)−1 s−1. The results presented in this section focus specifically on the roles of HCl and oxygen in mercury oxidation.

A time series from a typical experiment is shown in Figure 3. In this experiment, the HCl and O2 concentrations were changed in successive steps, and the mercury oxidation rate was measured following each change in simulated flue gas composition. Please note that the total concentration of mercury exiting the reactor bed, [HgT], is equal to the concentration entering the reactor. This steady state, with no net adsorption of mercury, is typically referred to as ‘complete breakthrough’. As with other studies of mercury oxidation catalysts, oxidation rate measurements were only made under conditions of complete breakthrough (22, 23).

Fig. 3

Mercury monitor data showing mercury oxidation across a gold catalyst from a typical experiment. The simulated flue gas composition was changed from baseline conditions to high HCl (75 ppm), low HCl (0 ppm) and O2-free conditions in successive steps. The system was allowed to reach an (apparent) steady state after each change to the simulated flue gas composition before another change was imposed. The oxidation rate was measured during the steady state portion of each time interval. [HgT] = total mercury concentration

Mercury monitor data showing mercury oxidation across a gold catalyst from a typical experiment. The simulated flue gas composition was changed from baseline conditions to high HCl (75 ppm), low HCl (0 ppm) and O2-free conditions in successive steps. The system was allowed to reach an (apparent) steady state after each change to the simulated flue gas composition before another change was imposed. The oxidation rate was measured during the steady state portion of each time interval. [HgT] = total mercury concentration


At the start of the test series for each catalyst, the samples adsorbed mercury for two to six hours before reaching complete breakthrough. During most of the subsequent experiments, the catalyst sample adsorbed mercury for a short period, typically one hour, prior to reaching complete breakthrough. The mercury adsorbed during this start-up period likely replaced mercury that was desorbed during the cooldown period of the previous experiment.

During the first experiment for each catalyst, Hg/CO2/N2 and Hg/CO2/O2/N2 gas mixtures were passed prior to the baseline simulated flue gas. All subsequent experiments were initiated with the baseline simulated flue gas (Table I). Complete mercury breakthrough occured quickly in the Hg/CO2/N2 atmosphere, and no mercury oxidation was evident in either of the mixtures. The onset of mercury oxidation coincided with the use of the baseline simulated flue gas.

3.1 Gold Catalyst

The Au catalyst exhibited a consistent baseline rate of Hg2+ formation, as shown in Figure 4. The baseline reaction rate remained constant over a period of seven experiments, suggesting that there was no apparent catalyst deactivation. The mean baseline reaction rate was (2.2 ± 0.3) × 10−10 (mol Hg2+) (g catalyst)−1 s−1. The baseline reaction rate remained constant when the flow rate was raised from 8 standard litres per minute (slpm) to 10 slpm with no change in simulated flue gas composition. This result suggests that the mercury oxidation reaction is not limited by mass transfer under the conditions tested here.

Fig. 4

Baseline reaction rate across the gold (squares) and palladium (circles) catalysts for the experiments presented here. Experiments were conducted on consecutive days (excluding weekends), with one experiment per day. — = the mean baseline reaction rate for the gold catalyst; – – – = one standard deviation

Baseline reaction rate across the gold (squares)  and palladium (circles)  catalysts for the experiments presented here. Experiments were conducted on consecutive days (excluding weekends), with one experiment per day. — = the mean baseline reaction rate for the gold catalyst; – – – = one standard deviation


The HCl concentration was varied from the baseline level of 50 parts per million (ppm) to 75 and 0 ppm. Raising the HCl concentration to 75 ppm had no effect on the reaction rate (Figure 2). Even though the reaction rate, and therefore the fractional conversion to oxidised mercury, were nominally constant during the baseline and elevated HCl portions of the experiment, the data in Figure 2 exhibit considerable scatter. As noted above, the precision for the kinetics measurements is approximately 15 to 30%. Due to the scatter in the data and the difficulty in making precise kinetic measurements with our current system, reaction rates reported here often represent time averages over periods of relative consistency (for example, the diamond symbols in Figure 2).

When the HCl concentration was set to 0 ppm, mercury oxidation continued, but the reaction rate slowed. The reaction rate fell to ∼ 85% of the baseline rate after 1.5 hours and ∼ 45% of the baseline rate after 2.5 hours. Time limitations prevented further testing, though we assume that mercury oxidation would have eventually stopped in the absence of an HCl source.

The mercury oxidation rate was also dependent on the presence of O2. Removing O2 from the simulated flue gas produced a similar effect to removing HCl: mercury oxidation continued at a reduced rate. Upon stopping O2 flow, the Hg2+ formation rate fell to < 50% of the baseline rate. Mercury oxidation in the O2-free simulated flue gas was only monitored for approximately 45 minutes, and as shown in Figure 2, the fractional mercury conversion was trending downward at the conclusion of the experiment. However, we are uncertain whether the reaction rate would have continued to decline as in the case of 0 ppm HCl.

In a separate experiment the temperature was varied from 138 to 160°C. An apparent activation energy of 40 kJ mol–1 for the global reaction (i):

  Hg + 2HCl → → HgCl2  (i)

was calculated for this temperature range. This is consistent with the apparent activation energy of ∼ 30 kJ mol–1 measured by Zhao et al. (24) for the reaction of mercury with Cl2 across a Au catalyst.

3.2 Palladium Catalyst

The baseline reaction rate across the Pd catalyst (Figure 4) declined over the course of the test period, falling from 1.6 × 10−10 to 3.3 × 10−11 (mol Hg2+) (g catalyst)−1 s−1. Because of the rapid decline in the baseline oxidation rate, all comparisons with the baseline rate are limited to a particular experiment. The baseline reaction rate was measured at the start of each experiment, and we assume that the baseline rate is roughly constant during a given experiment. The decrease in the baseline reaction rate with time suggests that the Pd catalyst is deactivated or fouled more readily than the Au catalyst.

The Pd catalyst exhibited similar responses to changes in HCl concentration to those of the Au catalyst. Raising the HCl concentration from 50 ppm to 100 ppm had no impact on the reaction rate. Lowering the HCl concentration to 0 ppm slowed, but did not halt, mercury oxidation. The oxidation rate fell to ∼ 45% of the baseline rate after 80 minutes, and ∼ 42% of the baseline rate after 150 minutes.

NO (500 ppm) was added for one experiment near the end of the test period. The baseline rate for this experiment was 3.3 × 10−11 (mol Hg2+) (g catalyst)−1 s−1, and the mercury was approximately 10 to 20% oxidised downstream of the catalyst bed. Adding NO to the simulated flue gas reduced the sensitivity of the mercury monitor, and the observed total mercury concentration fell from 10 μg Nm−3 to 4 μg Nm−3. With NO present, the mercury downstream of the catalyst bed was 90% oxidised. The reaction rate is not reported here because the measurement for the total Hg concentration is biased low. Regardless of the actual reaction rate, adding NO to the simulated flue gas resulted in significantly greater fractional mercury oxidation downstream of the catalyst bed.

3.3 Platinum Catalyst

As with the Pd catalyst, the baseline reaction rate observed with the Pt catalyst decreased over the course of the test period. The baseline rate fell from an initial maximum of 4.1 × 10−10 to 1.5 × 10−11 (mol Hg2+) (g catalyst)−1 s−1. As with the Au and Pd catalysts, increasing the HCl concentration to 75 ppm and 100 ppm had no impact on the mercury oxidation rate. Reducing the HCl concentration to 0 ppm yielded an immediate halt to mercury oxidation. This behaviour is in contrast to that of the Au and Pd catalysts, both of which continued oxidising mercury at reduced rates when HCl flow stopped.

Prior to one experiment the Pt catalyst was reduced at 165°C with 60 ppm CO in N2. The catalyst was then exposed to a mixture of Hg/CO2/HCl/SO2/N2 at 150°C. No oxidised mercury formation was observed. When O2 was added to the gas mixture to form the baseline simulated flue gas, mercury oxidation was evident. The HCl flow was then stopped, with the expectation that mercury oxidation would again stop. Instead, oxidation continued at a reduced rate, consistent with the behaviour observed for the Au and Pd catalysts.

The apparent activation energy for mercury oxidation across the Pt catalyst was measured for the temperature range 140 to 157°C. The measured activation energy was ∼ 120 kJ mol−1, significantly higher than that measured for the Au catalyst.

4. Discussion

As stated in the Introduction, previous investigations of mercury oxidation over a variety of catalysts indicate that mercury reacts from a bound state, for example, Hg(ads). It is well established that mercury adsorbs to Au, Pt and Pd surfaces. Gold (25, 26), palladium (27) and iridium (28) have all been used as modifiers for improving mercury capture in graphite tube atomic absorption spectrometry. The mercury monitor used in this study removes mercury vapour from the sample gas using Au/sand traps. The captured mercury is thermally desorbed during the analysis step of the instrument cycle. Mercury is also known to adsorb to Pt and form a solid solution with it (29). Therefore, we are confident in assuming that mercury adsorbs to the catalyst surface prior to reacting.

The dependence of the mercury oxidation rate on the presence of HCl in the simulated flue gas suggests that the oxidised mercury species observed by the mercury monitor is indeed HgCl2. Mercury oxidation was not observed when the catalyst was exposed to the Hg/CO2/O2/N2 gas mixture used prior to the simulated flue gas, indicating that the observed oxidised mercury species is unlikely to be HgO. Mercury(I) chloride (Hg2Cl2) is also a possibility, as it can form via the Boliden-Norzink reaction (30), Reaction (xiii):

  Hg(g) + HgCl2 → Hg2Cl2    (xiii)

However, themodynamic calculations indicate that Hg2Cl2 is not stable under typical flue gas conditions (31).

For all three noble metal catalysts tested here, increasing the HCl concentration above 50 ppm had no impact on the reaction rate. In a previous study, we made an initial assumption that the mercury oxidation rate can be described by r = k[Hg][HCl] (22). This assumption is obviously false for HCl concentrations > 50 ppm.

The assumption above required Hg + HCl as participants in the rate-limiting step. At this point, the rate-limiting step is unclear. When HCl was removed from the simulated flue gas during the tests with the Au and Pd catalysts, the reaction rate immediately fell, suggesting that the lack of a chlorine source to replenish the surface reduced the reaction rate. Mulla et al. suggested that the adsorption of O2 to an empty surface site was the rate-limiting step for NO oxidation over a Pt/Al2O3 catalyst (32). Perhaps this step is also rate-limiting for the formation of the presumed HgO intermediate product detailed below.

The data for the Au and Pd catalysts suggest that mercury reacts with HCl that is bound to the catalyst surface. This explains why mercury oxidation continues in the absence of gas-phase HCl, but with a declining reaction rate. Cl2 can chemisorb to Au surfaces and form AuCl3 (33). HCl dissociatively adsorbs to Pt surfaces (34), and similar behaviour might be expected for Pd. Thus, surface-bound Cl should be available for reaction on the Au and Pd surfaces.

The Pt catalyst exhibited different behaviour in the absence of HCl, and mercury oxidation stopped. This might be evidence of an Eley-Rideal mechanism for mercury oxidation across the Pt catalyst, with adsorbed Hg (or an intermediate such as HgO, described below) reacting with gas-phase HCl. Eley-Rideal kinetics would not suggest the zero-order dependence on [HCl] for concentrations greater than 50 ppm, but the overall reaction could exhibit a zero-order dependence on [HCl] if the Hg(ads) + HCl(g) step is not rate-limiting.

In a chlorine- and sulfur-free flame, Schofield (16, 17) observed HgO deposition on Pt and stainless steel surfaces. When HCl was added to the flame, the HgO desorbed as HgCl2. The implication of this observation is that Hg and HCl do not react directly to form HgCl2, but rather form via a HgO intermediate. Pt is an effective adsorber of oxygen (32, 35–37), hence surface-bound oxygen should be present in excess for the conversion of Hg0(g) to HgO(ads). In the absence of HCl(g), the HgO remains bound to the surface because of its low vapour pressure. When HCl is present, the HgO is converted to HgCl2, which desorbs from the surface and allows more Hg to react. The vapour pressure of HgCl2 is sufficiently high (1 Torr at 136°C), and the concentration is sufficiently low, that the simulated flue gas stream can hold HgCl2 as a vapour even at temperatures well below the sublimation point.

We tested the Schofield hypothesis in the experiment that used the reduced Pt catalyst. This test yielded two important results: (a) O2 and HCl (or possibly Cl2) are required for mercury oxidation across a Pt catalyst, possibly because HgCl2 formation is preceded by HgO; (b) the Pt catalyst can display a significant history effect.

In the initial series of experiments, the catalyst was exposed to an O2-containing gas mixture prior to the introduction of HCl. In the experiment with the reduced Pt, the catalyst was exposed to HCl prior to O2. We hypothesise that in the initial tests, O(ads) greatly outnumbers Cl(ads) to the point of exclusion. Thus, HgO is easily formed on the surface, and HCl reacts with the adsorbed HgO from the gas phase. Without HCl, there is no chlorine source for HgCl2 formation. In the experiment using the reduced Pt, the initial exposure to HCl allows for an ample concentration of adsorbed chlorine that is joined by adsorbed oxygen when O2 is introduced. When HCl is removed from the simulated flue gas, there is sufficient surface-bound chlorine to sustain HgCl2 formation at a reduced rate. The reaction mechanism remains unclear. The initial experiments suggest the possibility of an Eley-Rideal mechanism, and the experiments with the reduced Pt catalyst might suggest a Langmuir-Hinshelwood mechanism.

The data suggest that mercury oxidation across the Au catalyst is dependent on the presence of O2. This behaviour, while puzzling, may indicate the formation of the HgO intermediate. HgO binds to Au surfaces, and density functional theory calculations indicate that the binding energy for mercury species on Au(001) decreases in the series: HgO > Hg0 > HgCl2 (38). The predicted energy of binding of HgCl2 to the Au(001) surface is only 17.2 kJ mol−1, suggesting that this species could easily desorb from the catalyst surface at the temperatures tested here.

The mechanism governing the oxygen dependence of mercury oxidation across the Au catalyst is unknown at this time. Unlike with Pt, oxygen is not expected to efficiently adsorb to the Au surface (39), suggesting that adsorbed oxygen for HgO formation is not readily available at the surface. One could postulate an Eley-Rideal reaction between bound Hg and O2(g) to form HgO, but this reaction requires accounting for the second oxygen atom, suggesting either a ternary reaction (xiv):

  2Hg + O2 → 2HgO    (xiv)

which is unlikely, or the migration of an oxygen atom to the Au surface, which is also unlikely. Further research is required to elucidate this behaviour.

The nature of the bonding of mercury, chlorine and oxygen species to the catalyst surface is unknown at this time. One possibility is that each species adsorbs to the surface individually – Hg(ads), Cl(ads), HgO(ads), etc. Mercury is known to interact with other metals to form a variety of oxides and halides (40) that could participate in the surface reactions which lead to the formation of HgCl2. Surface analysis of fresh and used catalyst samples will be conducted in the future in order to gain more insight.

The most significant difference between the performance of the Au and the pgm catalysts, was the loss of catalytic activity for the Pd and Pt over the course of the test period. The presence of adsorbed oxygen on the catalyst surface may offer an explanation for the observed drop in catalyst activity. While testing Pt/Al2O3 catalysts for the oxidation of NO to NO2, Mulla et al. observed that the catalyst deactivated during cooldown and other changes to process conditions (32). The researchers proposed that deactivation may have been the result of the oxidation of the Pt surface. Olsson et al. also observed deactivation of Pt/Al2O3 and Pt/BaO/Al2O3 catalysts during the same reaction and attributed the loss of catalytic activity to the formation of unreactive PtO (41). The formation of Pt and Pd oxides, especially while exposed to O2 during cooldown, may also explain the deactivation observed here. The poor reactivity of PtO might also suggest that the possible HgO intermediate product is not formed via a Mars-Maessen reaction. Au, on the other hand, is typically a poor adsorber of oxygen and is likely not subject to this deactivation mechanism (39).

Mulla et al. found that their Pt/Al2O3 catalyst could be regenerated with CO or H2 (32). This observation agrees well with the data presented here; Hg oxidation proceeded at nearly the initially observed baseline reaction rate after the Pt catalyst was exposed to CO. The enhanced extent of Hg oxidation observed across the Pd catalyst in the presence of NO also suggests that surface oxygen inhibits Hg oxidation. When NO was added to the simulated flue gas used here, it is possible that surface oxygen was removed, allowing for increased conversion of mercury (32, 35–37, 41).

It is unclear at this time whether the Pt and Pd catalysts were deactivated by the formation of oxides (for example, PtO and PdO), surface-bound oxygen, or both. While NO may have removed adsorbed oxygen from the Pd surface, the presence of PtO has been observed to inhibit NO oxidation across Pt catalysts (41). Thermal regeneration of Pt catalysts requires temperatures of 600 to 650°C, significantly higher than the temperatures used here. Surface analysis will be required to confirm the nature of the catalyst surface following exposure to the simulated flue gas.

Of concern with all potential mercury oxidation catalysts are the unwanted side reactions (xv) and (xvi):

  NO + ½O2 → NO2    (xv)


  SO2 + ½O2 → SO3    (xvi)

Stack NO2 concentrations as low as 15 ppm (42) can lead to the formation of a brown plume and enhanced local ozone production. SO3 is captured poorly in most wet flue gas desulfurisation systems, and can lead to sulfuric acid mist in the downstream plume. In each case, a small concentration can produce a significant impact – 15 ppm NO2 corresponds to only 3% NO conversion for flue gas containing 500 ppm NO. To date, both Au and Pd catalysts have been tested at pilot scale, and neither SO3 nor NO2 formation was observed (23). However, formation of these unwanted by-products remains a concern.

The results introduced here may have several implications for the use of noble metal catalysts in mercury abatement schemes. Perhaps most important is the observation that Au exhibits superior resistance to deactivation than Pd and Pt. Contrary to our short-term, bench-scale results, pilot-scale testing of Au and Pd showed similar performance and deactivation over time (23). A likely explanation is the presence of NO in the real flue gas; as shown here, NO appears to regulate the surface oxygen concentration on the Pd catalyst, leading to improved Hg oxidation versus a NO-free simulated flue gas. The oxygen resistance of Au may come into play during periods of down time, when the catalyst beds come into contact with air at ambient temperature. This condition favours oxygen adsorption onto Pt and Pd. However, a preference for Au remains to be seen, and future, longer-term testing may be warranted.

5. Conclusions

The results presented here provide an initial investigation into the mechanisms behind mercury oxidation across noble metal catalysts, and more work is needed. Investigations are needed into the role of other potential flue gas catalyst poisons such as arsenic, selenium and SO3. Selenium is suspected to deactivate Au catalysts tested at pilot scale (23), and initial experiments conducted in our laboratory suggest that high concentrations of SO3 can deactivate Au catalysts. Future work should also focus on the possible beneficial role of different promoters, alloys and supports. At the size and timescale presented here, both Au and pgms show promise for use as mercury oxidation catalysts. As we begin to understand mercury oxidation across these catalysts – specifically the roles of O2 and HCl and the apparent deactivation mechanisms – improvements can be made that will help mercury oxidation catalysts become economically competitive as part of a mercury abatement strategy.



  1.  “Mercury Study Report to Congress”, U.S. EPA, U.S. Government Printing Office, Washington, D.C., December, 1997:
  2.  “Study of Hazardous Air Pollutant Emissions from Electric Utility Steam Generating Units – Final Report to Congress”, U.S. EPA, U.S. Government Printing Office, Washington, D.C., February, 1998:
  3.  ‘Clean Air Mercury Rule’, U.S. EPA, 15th March, 2005: LINK
  4.  K. C. Galbreath and C. J. Zygarlicke, Environ. Sci. Technol., 1996, 30, (8), 2421 LINK
  5.  C. L. Senior and S. A. Johnson, Energy Fuels, 2005, 19, (3), 859
  6.  J. H. Pavlish, E. A. Sondreal, M. D. Mann, E. S. Olson, K. C. Galbreath, D. L. Laudal and S. A. Benson, Fuel Proc. Technol., 2003, 82, (2–3), 89 LINK
  7.  ‘Clean Air Interstate Rule’, U.S. EPA, 10th March, 2005: LINK
  8.  R. K. Srivastava, N. Hutson, B. Martin, F. Princiotta and J. Staudt, Environ. Sci. Technol., 2006, 40, (5), 1385 LINK
  9.  R. K. Srivastava, R. E. Hall, S. Khan, K. Culligan and B. W. Lani, J. Air Waste Manage. Assoc., September, 2005, 55, (9), 1367 LINK
  10.  A. A. Presto and E. J. Granite, Environ. Sci. Technol., 2006, 40, (18), 5601 LINK
  11.  M. J. Pilling and P. W. Seakins, “Reaction Kinetics”, 2nd Edn., Oxford University Press, Oxford, U.K., 1995
  12.  E. J. Granite, H. W. Pennline and R. A. Hargis, Ind. Eng. Chem. Res., 2000, 39, (4), 1020 LINK
  13.  P. Mars and J. G. H. Maessen, J. Catal., 1968, 10, (1), 1 LINK
  14.  A. K. Medhekar, M. Rokni, D. W. Trainor and J. H. Jacob, Chem. Phys. Lett., 1979, 65, (3), 600 LINK
  15.  P. A. Ariya, A. Khalizov and A. Gidas, J. Phys. Chem. A, 2002, 106, (32), 7310 LINK
  16.  K. Schofield, Chem. Phys. Lett., 2004, 386, (1–3), 65 LINK
  17.  K. Schofield, Proc. Combust. Inst., 2005, 30, (1), 1263 LINK
  18.  A. A. Presto and E. J. Granite, Environ. Sci. Technol., 2007, 41, (18), 6579 LINK
  19.  S. Nelson, “Brominated Sorbents for Small Cold-Side ESPs, Hot-Side ESPs, and Fly Ash Use in Concrete”, Progress Report to U.S. DOE/NETL Cooperative Agreement No. DE-FC26-05NT42308, Sorbent Technologies Corporation, 2007
  20.  E. J. Granite and H. W. Pennline, Ind. Eng. Chem. Res., 2002, 41, (22), 5470 LINK
  21.  G. Blythe, “Pilot Testing of Mercury Oxidation Catalysts for Upstream of Wet FGD Systems”, Quarterly Technical Progress Report to U.S. DOE/NETL, U.S. Department of Energy Agreement No. DE-FC26-01NT41185, URS Corporation, 2003
  22.  A. A. Presto, E. J. Granite, A. Karash, R. A. Hargis, W. J. O'Dowd and H. W. Pennline, Energy Fuels, 2006, 20, (5), 1941 LINK
  23.  G. Blythe, K. Dombrowski, T. Machalek, C. Richardson and M. Richardson, “Pilot Testing of Mercury Oxidation Catalysts for Upstream of Wet FGD Systems”, Final Report to U.S. DOE/NETL, U.S. Department of Energy Agreement No. DE-FC26-01NT41185, URS Corporation, 2006
  24.  Y. Zhao, M. D. Mann, J. H. Pavlish, B. A. F. Mibeck, G. E. Dunham and E. S. Olson, Environ. Sci. Technol., 2006, 40, (5), 1603 LINK
  25.  Z. Hladky, J. Rísová and M. Fisera, J. Anal. At. Spectrom., 1990, 5, (8), 691 LINK
  26.  J. P. Matousek, R. Iavetz, K. J. Powell and H. Louie, Spectrochim. Acta Part B.: At. Spectrom., 2002, 57, (1), 147
  27.  X.-P. Yan and Z.-M. Ni, Anal. Chim. Acta, 1993, 272, (1), 105 LINK
  28.  B. Radziuk and J. Kleiner, Spectrochim. Acta Part B: At. Spectrom., 1993, 48, (14), 1719 LINK
  29.  F. L. Fertonani, A. V. Benedetti and M. Ionashiro, Thermochim. Acta, 1995, 265, 151 LINK
  30.  T. Allgulin, Boliden AB, U.S. Patent3,849,537; 1974
  31.  F. Frandsen, K. Dam-Johansen and P. Rasmussen, Prog. Energy Combust. Sci., 1994, 20, (2), 115 LINK
  32.  S. S. Mulla, N. Chen, L. Cumaranatunge, G. E. Blau, D. Y. Zemlyanov, W. N. Delgass, W. S. Epling and F. H. Ribeiro, J. Catal., 2006, 241, (2), 389 LINK
  33.  N. D. Spencer and R. M. Lambert, Surf. Sci., 1981, 107, (1), 237 LINK
  34.  F. T. Wagner and T. E. Moylan, Surf. Sci., 1989, 216, (3), 361 LINK
  35.  J. Després, M. Elsener, M. Koebel, O. Kröcher, B. Schnyder and A. Wokaun, Appl. Catal. B: Environ., 2004, 50, (2), 73 LINK
  36.  R. Marques, P. Darcy, P. Da Costa, H. Mellottée, J.-M. Trichard and G. Djéga-Mariadassou, J. Mol. Catal. A: Chem., 2004, 221, (1–2), 127 LINK
  37.  M. F. Irfan, J. H. Goo, S. D. Kim and S. C. Hong, Chemosphere, 2007, 66, (1), 54 LINK
  38.  E. Sasmaz and J. Wilcox, ‘The Binding of Flue Gas Components on CaO, TiO2, Pd, Au and PdAu Alloy’, Paper No. 409, in“Proceedings of the Air & Waste Management Association's 100th Annual Conference and Exhibition”, Pittsburgh, PA, June, 2007, A&WMA, Pittsburgh, PA,2007
  39.  X. Deng, B. K. Min, A. Guloy and C. M. Friend, J. Am. Chem. Soc., 2005, 127, (25), 9267 LINK
  40.  J. W. Mellor, “A Comprehensive Treatise on Inorganic and Theoretical Chemistry”, Longmans, Green and Company, London, New York, 1952, Vol. IV
  41.  L. Olsson and E. Fridell, J. Catal., 2002, 210, (2), 340 LINK
  42.  A. S. Feitelberg and S. M. Correa, J. Eng. Gas Turbines Power, 2000, 122, (2), 287 LINK


The authors thank Johnson Matthey PLC for providing the catalyst samples. Hugh Hamilton of the Johnson Matthey Technology Centre, U.K., provided excellent insight into the catalyst preparation and properties. The comments of the reviewers are greatly appreciated. Albert Presto acknowledges the support of a postdoctoral fellowship at the U.S. Department of Energy (DOE) administered by the Oak Ridge Institute for Science and Education (ORISE). Funding support from the DOE Innovations for Existing Power Plants (IEP) Program is greatly appreciated. We thank Gregson Vaux, Power/Energy Engineer for the Science Applications International Corporation (SAIC), for his kind help in understanding the regulatory status for mercury emissions as of April 2008.


References in this paper to any specific commercial product, process, or service are to facilitate understanding, and do not necessarily imply its endorsement by the U.S. DOE.

The Authors

Dr Albert A. Presto is currently a Research Scientist and Laboratory Manager for the Air Quality Laboratory in the Center for Atmospheric Particle Studies at Carnegie Mellon University. Prior to his current position he was an ORISE postdoctoral fellow at the U.S. Department of Energy's National Energy Technology Laboratory (NETL) in Pittsburgh. His research interests include mercury removal from coal-derived flue and fuel gas, atmospheric secondary organic aerosol formation, the atmospheric processing of organic aerosol, and atmospheric radical chemistry.

Dr Evan J. Granite is a Research Group Leader at the NETL. His research has focused on mercury and carbon dioxide removal from flue and fuel gases. Dr Granite is the principal investigator for three projects on the capture of mercury, arsenic and selenium from coal-derived flue and fuel gases, and carbon dioxide separation from flue gas. His research interests are in catalysis and surface chemistry, pollution clean-up, electrochemistry and photochemistry.

Read more from this issue »