- Email: [email protected]
Iron oxyhydroxide coating of pyrite for acid mine drainage control
Applied Geochemistry 24 (2009) 1626–1634
Contents lists available at ScienceDirect
Applied Geochemistry journal homepage: www.elsevier.com/locate/apgeochem
Iron oxyhydroxide coating of pyrite for acid mine drainage control Danielle M.C. Huminicki, J. Donald Rimstidt * Virginia Polytechnic Institute and State University, Department of Geosciences, 4044 Derring Hall, Blacksburg, VA 24061, USA
a r t i c l e
i n f o
Article history: Received 31 December 2008 Accepted 21 April 2009 Available online 5 May 2009 Editorial handling by R. Fuge
a b s t r a c t When pyrite oxidizes at near neutral pH in the presence of sufficient alkalinity, Fe oxyhydroxide coatings develop on the surface. As these coatings grow thicker and denser they block oxidant transport from the solution to the pyrite surface and reduce the rate of pyrite oxidation. The authors’ measurements of pyrite oxidation rates in a NaHCO3 solution show that the coating grows in two stages. In the first stage Fe oxyhydroxide colloids form and then attach to the pyrite surface to produce a slight reduction in oxidant transport. In the second stage interstitial precipitation of Fe oxyhydroxide material between the colloidal particles reduces the oxidant’s diffusion coefficient by more than five orders of magnitude. This causes the pyrite oxidation rate to decline as the square root of time. The kinetic predominance diagram, which compares the rates of Fe transformation reactions, shows that when pyrite oxidation releases Fe quickly enough for the total Fe concentration to rise to about 108 m, ferrihydrite forms but lower rates of Fe release will not produce coatings. Extrapolation of the results to longer times predicts that pyrite-bearing materials need to be treated with an extra source of alkalinity for several decades to produce coatings that are thick enough to be sustained by alkalinity levels typical of groundwater. However, once the coatings develop no additional treatment is needed and further pyrite oxidation simply causes the coating to grow thicker and denser until the entire pyrite grain is pseudomorphically replaced by goethite. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Although acid mine drainage (AMD) is a widespread and intensely studied environmental problem, there is a continuous struggle to find scientifically sound and economically viable strategies to mitigate it. Current AMD management practices often treat the symptom, which is the acidic effluent, but not the source, which is the oxidizing pyrite. AMD develops when acid is generated by sulfide mineral oxidation faster than it is neutralized by alkalinity from the surroundings so an important strategy for long-term remediation programs is to supply alkalinity to the mine wastes faster than their rate of acid production. Acid base accounting in mine wastes is a well-established tool for predicting whether there is a sufficient amount of alkalinity generating minerals already in the mine wastes (White et al., 1999) and when a sufficient alkalinity supply does not exist, various alkalinity-generating materials can be incorporated into the wastes (Smith and Brady, 1998). However, having a favorable acid base accounting is not always sufficient to mitigate AMD. In order to be effective, the alkalinity-generating materials must dissolve and supply alkalinity as fast, or faster than acidity is generated by pyrite oxidation. In order to treat the wastes in the most economic way it is necessary to determine just how fast alkalinity must be supplied immediately after the wastes are disposed, when pyrite oxidation rates are fastest, and then later * Corresponding author. Tel.: +1 540 392 8913; fax: +1 540 231 3386. E-mail address: [email protected] (J.D. Rimstidt). 0883-2927/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apgeochem.2009.04.032
on after the oxidation rates have declined. In most cases a decline in acid generation is simply due to the consumption of pyrite but both field and experimental evidence suggests that a more rapid decrease in acid generation occurs if the waste disposal is designed to encourage the growth of a layer of Fe oxyhydroxides on the pyrite surface. As this Fe oxyhydroxide coating grows thicker and denser over time, it becomes an increasingly effective barrier to oxidant transport to the pyrite surface thereby slowing the oxidation rate and the rate of acid generation. Eventually the coating becomes so effective that the rate of acid generation falls below the rate of alkalinity delivery by groundwater and AMD generation stops. This could occur even if the acid base accounting is unfavorable but it does require very rapid alkalinity addition soon after disposal. Because of its abundance, pyrite is commonly the main source of acid production in mine wastes. The pH of freshly exposed pyrite bearing rocks is usually greater than four so that dissolved O2 (DO) is the principal pyrite oxidant and the rate of acid production is relatively low (Fig. 1, py–DO reaction in Table 1) (Williamson and Rimstidt, 1994). However, if this acid is not neutralized as quickly as it is produced the pH drifts downward until it drops below four where dissolved Fe(III) becomes the principal oxidant. A further decrease in pH causes a large increase in Fe(III) concentration leading to even faster pyrite oxidation (Fig. 1, py–Fe(III) reaction in Table 1). This creates a runaway condition where fast pyrite oxidation continually lowers pH and raises the Fe(III) concentration causing the oxidation rate to accelerate. Because microbial activity
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
1627
Notation A surface area (m2) specific surface area (m2/g) Asp b surface area constant (unitless) concentration of species i (mol/m3) Ci D diffusion coefficient (m2/s) Dc and Dp diameter of calcite and pyrite particles groundwater infiltration rate (m/s) Ir flux (mol/m2 s) Ji rate constant ki concentration of species i (mol/kg) mi concentration of species i (mol/L) Mi Mc and Mp mass of calcite and pyrite
quickly regenerates Fe(III) from the Fe(II) released by the pyrite (Williamson et al., 2006), there is always sufficient Fe(III) to sustain the runaway AMD condition. Avoiding or recovering from runaway AMD requires that the acid from the py–DO or py–Fe(III) reaction be neutralized by the addition of alkalinity faster than it is produced. When the rate of alkalinity addition meets or exceeds this rate, the Fe released from the pyrite rapidly oxidizes to Fe(III) and precipitates as an Fe oxyhydroxide coating on the pyrite surface. Coating pyrite grains with a substance that blocks oxidant transport from solution to the surface may be a practical way to reduce the oxidation rate. Several types of pyrite coatings have been proposed including ferric phosphate (Evangelou, 1995), phosphosilicates (Fytas and Bousquet, 2002; Fytas et al., 1999; Fytas and Evangelou, 1998), ferric hydroxide–silica (Zhang and Evangelou, 1998), phospholipids (Kargbo et al., 2004), and iron-8 hydroxy-
Fig. 1. Comparison of the rates of generation and consumption of Fe(II) in AMD solutions using the rate laws and assumptions described in Williamson et al. (2006). At pH > 4, the py–DO reaction rate is faster than the py–Fe(III) rate and the Fe(II) released from the pyrite is abiotically oxidized. The resulting Fe(III) rapidly precipitates from solution. At pH < 4, the py–Fe(III) reaction is fastest and the Fe(II) released from the pyrite is microbally oxidized. The resulting Fe(III) is very soluble and available to react with more pyrite. The grey zone in the upper right corner shows the range of rates for H+ consumption by dissolving calcite (Plummer et al. 1978). All rates are for 1 m2 of mineral surface area per 1 kg of solution.
ni rf ri r 0i t tR VR Vm x u
amount of species i (mol) flow rate (kg/s) (1 kg H2O 0.001 m3) specific rate of reaction of species i (mol/m2 s) apparent rate of reaction of species i (mol/s) time (s) response time (s) volume of reactor (m3) molar volume (m3/mol) thickness of coating (m) porosity, unitless ratio from 0 to 1
quinoline (Lan et al., 2002). These coatings have been shown to reduce oxidation rates in the laboratory but they require relatively expensive reagents, which do not specifically target the pyrite so that a large excess may be required to insure an effective coating. In addition, because they are not self-healing and permanent they require a long-term commitment to site management. In comparison, Fe oxyhydroxide coatings form naturally in high alkalinity environments, they are self-healing and they become more and more effective with time so that the demand for alkalinity addition declines. Nicholson et al. (1990) showed experimentally that the addition of relatively inexpensive bicarbonate alkalinity to oxidizing pyrite produces Fe oxyhydroxide coatings that slow oxidant transport from solution to the pyrite and thus slow pyrite oxidation rates. The effectiveness of this strategy is supported by more recent laboratory studies (Homstrom et al., 1999; Pérez-López et al., 2005, 2007b; Zhang and Evangelou, 1996). We can predict the long-term behavior of Fe oxyhydroxide coatings by considering a natural analogue, limonite pseudomorphs after pyrite (Fig. 2). Limonite pseudomorphs form where pyrite has oxidized under high alkalinity conditions. They are often found in limestone where large amounts of HCO 3 ions are available to neutralize the acid from pyrite oxidation. Under high pH conditions, abiotic oxidation of Fe(II) (Stumm and Lee, 1961) is much faster than pyrite oxidation by dissolved O2 (Fig. 1) so that Fe(II) released from the pyrite quickly oxidizes to Fe(III). Because Fe(III) is very insoluble at near neutral pH, it quickly hydrolyzes and precipitates. Initially formed, metastable Fe(OH)3 and/or ferrihydrite eventually converts to goethite. These coupled reactions produce an approximately 1 for 1 volume replacement of pyrite (23.94 cm3/mol) by goethite (20.82 cm3/mol) creating a pseudomorph with only about 13% porosity. Additional pore filling by the oxidation and precipitation of outward diffusing Fe(II) fills the pores near the outer edge of the pseudomorph to create a dense outer rim. Fig. 2 shows that a typical limonite pseudomorph has a relatively porous interior, often with some remaining pyrite, surrounded by a relatively dense outer rim. Limonite pseudomorphs after pyrite are compelling evidence that Fe oxyhydroxide coatings can be stable and grow to great thickness on pyrite grains. The objective of this paper is to present a conceptual and quantitative model of declining pyrite oxidation rate caused by the formation of a Fe oxyhydroxide coating under high alkalinity conditions. Experiments were performed that bridge the time in the coating process between the initial attachment of Fe(OH)3 colloids and the infilling between the colloids to form a more dense coating. Data from the authors’ and previous experiments (Nicholson et al., 1990; Zhang and Evangelou, 1996) were analyzed to determine the effective diffusion coefficients of oxidants through
1628
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Table 1 Empirical rate laws for chemical reactions discussed in this paper. In the text, the rate laws are named by the most important reactants, which are shown in italics in the left column. Rate laws for solids have been scaled to a system with 1 m2 of mineral surface area per 1 kg of solution. Reaction
r (mol/s)
pH
References
py–H2O2 þ FeS2 ðpyÞ þ 7H2 O2 ¼ Fe2þ þ 2SO2 4 þ 2H þ 6H2 O
dMFeðIIÞ dt
¼ 105:82 M 0:93 H2 O2
1–4
McKibben and Barnes (1986)
py–DO þ FeS2 ðpyÞ þ 7=2O2 þ H2 O ¼ Fe2þ þ 2SO2 4 þ 2H
dmFeðIIÞ dt
DO ¼ 108:19 m0:11
2–10
Williamson and Rimstidt (1994)
py–Fe(III) þ FeS2 ðpyÞ þ 14Fe3þ þ 8H2 O ¼ 15Fe2þ þ 2SO2 4 þ 16H
dmFeðIIÞ dt
¼ 108:58 m0:47FeðIIIÞ m0:32
0.5–3
Williamson and Rimstidt (1994)
Fe(II)–H2O2 Fe2+ + 0.5H2O2 + H+ = Fe3+ + H2O
dmFeðIIÞ dt
¼ k2
6–8
Millero and Sotolongo (1989)
4–8
Stumm and Lee (1961)
6–9
Ninh Pham et al. (2006)
6–8
Yee et al. (2006)
6.8–7.0
Park and Dempsey (2005)
m0:5 Hþ
m0:3
FeðIIÞ
Hþ
mFeðIIÞ mH2 O2 mHþ
log k2 = 11.72 2.14I Fe(II)–DO Fe2+ + 1/4 O2 + H+ = Fe3+ + 1/2H2O
dmFeðIIÞ dt
¼ 1012:96
Fe(III)–Fe(OH)3(s) a FeD + FeT = AFO + nH+
dMFeðIIIÞ dt
¼ kf M FeD M FeT
fh–goe b Fe5O3(OH)9(fh) = 5FeOOH (goe) + 2H2O
a ¼ 1 e410
Fe(II)–fh 5Fe2+ + 5/4 O2 + 19/2 H2O = Fe5O3(OH)9(fh) + 10 H+
dmFeðIIÞ dt
a b
5
1/2
+ 1.38I
mFeðIIÞ mDO mH þ2
ðt3384Þ1:55
¼ 107:3 mFeðIIÞ mFeðIIÞsorbed mDO
FeD = dissolved Fe(III), FeT = dissolved and precipitated Fe(III), AFO = amorphous Fe oxyhydroxide (Fe(OH)3). T = 21 °C; a is the extent of reaction (0–1).
Fig. 3. Schematic diagram of mixed flow reactor experiment.
Fig. 2. Cross-section of a limonite pseudomorph from Bedford Co., Virginia, USA showing a porous center containing a small amount of unreacted pyrite surrounded by a dense outer coating.
the Fe oxyhydroxide coatings and that information was used to predict how alkalinity demand declines over time as the coatings grow thicker. 2. Methods and materials Pyrite from the same source as Jerz and Rimstidt (2004) and Williamson and Rimstidt (1994) was prepared in the manner described in Jerz and Rimstidt (2004) to recover the 250–420 lm size fraction with an estimated specific surface area of 1.13 102 m2/ g (Foust et al., 1980). The grains were sonicated in a 0.1 m HCl solution then rinsed with ethyl alcohol until the supernatant was clear and the grains showed no visible oxidation product layer. These grains were reacted with a mixture of 0.3 m H2O2 and 0.1 m NaHCO3 solution with a pH of approximately 8.5 in order to induce Fe oxyhydroxide coatings. The peroxide–bicarbonate solution was circulated through 5 g of pyrite in a packed bed (Fig. 3). Solution samples were collected at evenly spaced time intervals of 594 s using a fraction collector. The mass of the solution in the reactor during the experiment
was 2.43 ± 0.2 g. Experiments were run for 24 h at 23 ± 1 °C. The flow rate was determined by weighing each sample container before and after sample collection. The average flow rate was approximately 0.5 g/s. The total Fe concentration of each sample was determined using an atomic absorption spectrophotometer (AA). Sulfate concentrations were determined from another set of filtered samples by ion chromatography (IC). To confirm that all S was oxidized to SO4 some ICP-MS sample analyses that measured total S were compared with those from the IC. These values corresponded to each other so it was concluded that all S was in the form of SO4. Reacted grains were observed and characterized using a scanning electron microscope and visible light microscopy. Iron that precipitated in the reactor over the course of the experiment was recovered by leaching the entire reactor by immersing it in 500 mL of a 0.04 M NH2OHHCl in 25% (v/v) CH3COOH solution at 100 °C for 1 h. The concentration of Fe in the leachate was determined by AA. Sulfate was used as the reaction progress variable. The rate of species release was calculated from the solution flow rate multiplied by the change in species concentration
ri ¼ r f ðmout min Þ
ð1Þ
Pyrite oxidation rates by H2O2 and O2 were calculated from the experimental data as well as from Zhang and Evangelou (1996) and Nicholson et al. (1990) based on the stoichiometry of the py–H2O2
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
and py–DO reactions in Table 1. Rates were calculated by numerical differentiation (Pollard, 1977) of their Fe concentration (mol) versus time data (all data are tabulated in Huminicki, 2006). In the experiments relatively little of the Fe precipitated as coatings. The total amount of Fe released (nFe(cal)) was calculated from the amount of SO4 released based on the reaction stoichiometry for the py–H2O2 reaction (Table 1). This Fe was quickly oxidized to Fe(III), which either precipitated onto the pyrite grains (nFe(py)), precipitated onto the reactor walls (nFe(walls)), or was carried out of the reactor by solution (nFe(sol)) so that
nFeðcalÞ ¼ nFeðpyÞ þ nFeðwallsÞ þ nFeðsolÞ
ð2Þ
The amount of pyrite oxidized and the amount of Fe that was carried out of the reactor by the solution were both determined (Huminicki, 2006) at the end of the experiment to be 2.5 104 and 1.5 104 mol, respectively. The difference between these amounts is the total amount of Fe that precipitated onto either the reactor or the pyrite. This was about 40% of the total Fe released. Of this, 37% of the precipitated Fe was recovered by leaching the reactor and tubing (9.12 105 mol) and it was inferred that the rest (3%) had precipitated on the pyrite surfaces. Coating thickness was calculated from the amount of Fe precipitated on the pyrite surface, the molar volume of the coating, the surface area of pyrite, and the porosity of the coating
x¼
nFe fppt V m Að1 /Þ
ð3Þ
The coating thickness was estimated as a function of time for all of the experiments assuming that in each case only 3% of the Fe released precipitated as a coating (fppt = 0.03). It was assumed that the coatings had 10% porosity (u = 0.1) and the coatings consisted of goethite. These thickness values are given in Huminicki (2006). In Eq. (3) the number of moles of Fe released, the molar volume, and the surface area of the pyrite are all well constrained. Because fppt is the difference between two much larger numbers it caries a fairly large uncertainty, which dominates the uncertainty in the coating thickness. For example, if fppt = 0.03 ± 0.01, x = 3(±1) 109 m. On the other hand the uncertainty in u has a relatively minor effect on x; changing u from 0.1 to 0.3 only increases x by 25%. Assuming that the coatings consist of ferrihydrite instead of goethite would increase x by about 25%. After approximately 10 h of reaction, graphs of r versus t1/2 all show that pyrite oxidation rates decreased over time, presumably because the growing coating reduced the rate of oxidant transport to the pyrite surface. The expected relationship between r versus t1/2 can be derived from Fick’s first law of diffusion
Ji ¼
DDC x
ð4Þ
This model allowed determination of the diffusion coefficient, Di, for the oxidant through the coating with a concentration gradient, DC, defined as
DC ¼ 1000ðmðsolÞ mðsfcÞ Þ
ð5Þ
where 1000 is a conversion factor that recasts the molal concentrations into units of mol/m3. It can be reasonably assumed that DC approaches m(sol) as m(sfc) becomes small and that DC becomes nearly constant as soon as a significant coating develops. The flux of oxidant, Ji, to the surface of pyrite can be converted to the rate of release of Fe from the surface using
dnFe 1 dnoxid ¼ ¼ r FeS2 dt mi dt
ð6Þ
where, mi, is the stoichiometric coefficient that relates the rate of oxidant consumption to Fe release from the reactions so mH2 O2 ¼ 7 and mDO = 3.5 (Table 1).
1629
A general model that describes the decrease in the rate of pyrite oxidation over time as a coating forms can be developed by combining these relationships and substituting in J i ¼ mi r FeS2 , DC 1000m(sol), and Eqs. (3) and (4) into Eq. (6), to give
dnFeS2 dnFe 1000mðsolÞ DAð1 /Þ ¼ ¼ dt dt mi nFe fppt V m
ð7Þ
that can be rearranged and integrated for the boundary conditions nFe = 0 when t = 0 to get
n2Fe 1000mðsolÞ DAð1 /Þ t ¼ 2 mi fppt V m
ð8Þ
Solving for nFe gives
nFe ¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t mi fppt V m
ð9Þ
which can be simplified by defining
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2000mðsolÞ DAð1 /Þ c¼ mi fppt V m
ð10Þ
so the total amount of Fe released from pyrite at time, t, is
nFe ¼ ct 1=2
ð11Þ
The time derivative of this equation gives the diffusion-limited rate of pyrite oxidation as a function of time,
rFeS2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t mi fppt V m
1 ¼ 2
ð12Þ
It is expected that as the Fe oxyhydroxide coatings thicken, the rate of pyrite oxidation will decline as a function of t1/2. The slope of the linear regression fit of r versus t1/2 is ½ c so that the diffusion coefficient can be solved for from the relationship
1 1 c¼ 2 2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ mi fppt V m
ð13Þ
to find
D¼
1 mi fppt V m c2 2000 mðsolÞ Að1 /Þ
ð14Þ
3. Results The experiment determined how pyrite oxidation rates declined as Fe oxyhydroxide coatings formed and limited the H2O2 transport to the surface. These results allowed documentation of the initial stage of coating formation and determination of a diffusion coefficient for H2O2 for freshly formed coatings. The rate data are tabulated in Huminicki (2006) along with the rates extracted from Zhang and Evangelou (1996) and Nicholson et al. (1990). All rate data are displayed in Figs. 4–6. Figs. 4–6 all show a linear decrease in rate versus t1/2 that is attributed to the growth of an Fe oxyhydroxide coating acting as a barrier to the transport of oxidant to the pyrite surface. The experiments (Fig. 4) show a transition from a shallow slope of r versus t1/2 to a steeper one, which is the result of the coating becoming a more effective barrier to oxidant transport to the pyrite surface. This transition is not seen in Fig. 5 (Zhang and Evangelou, 1996), because this step of coating formation had already occurred during their pretreatment step. The Nicholson et al. (1990) data show an unexplained initial increase in rates with time (up to 11 days) so those data were not used in the r versus t1/2 fits.
1630
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Fig. 4. Graph of r versus t1/2 data from the MFR experiment. The slope of the line changed from shallow r = 2.19 106t1/2 + 3.42 107 to steep r = 7.72 105t1/2 5.31 108 over the experiment indicating that after about 10 h the coating became a more effective barrier to H2O2 transport. The rate law for the py–H2O2 reaction (Table 1) predicts a rate of 4.57 107 mol/m2 s for the 0.3 m H2O2 solution used in the experiments.
Fig. 6. Graph of r versus t1/2 for data from Nicholson et al. (1990). The equations of the lines are r = 8.91 107t1/2 + 1.37 1010, r = 7.26 107t1/2+3.42 1010, and r = 1.07 106t1/2 + 9.5 1011 for the data sets of the experiments that used 76 (open diamonds), 108 (filled squares), 215 (filled triangles) lm grain sizes, respectively. The chemical reaction-limited rate of pyrite oxidation calculated for py–DO from Table 1 is 9.13 1010 mol/m2 s.
Fig. 5. Graph of r versus t1/2 for data from Zhang and Evangelou (1996). The equation for the line is r = 2.60 107t1/2 + 3.47 1011. The rate law for the py– H2O2 reaction (Table 1) predicts a rate of 2.19 107 mol/m2s for the 0.145 m H2O2 solution used in these experiments. That is about two orders of magnitude faster than their fastest rate because a relatively thick Fe oxyhydroxide coating already existed on the pyrite surface due to pretreatment.
4. Discussion The results (Fig. 4) suggest that Fe oxyhydroxide coatings grow on the pyrite in two stages: (1) an initial stage of colloid deposition followed by (2) the densification and inward propagation of the coating (Fig. 7). In the first stage a thin, highly porous and permeable layer of colloidal Fe oxyhydroxide particles deposits on the pyrite surface. This layer provides only a weak barrier to oxidant transport to the pyrite surface so the rate of pyrite oxidation declines very slowly for the first 10 h. In the second stage, Fe oxyhydroxides precipitate in the pores of this layer making it a more effective barrier to DO transport. This causes the pyrite oxidation rate to decrease more rapidly with time as shown by the line with the steeper slope in Fig. 4. When pyrite oxidizes in the presence of excess alkalinity the acidity produced is neutralized near the pyrite surface so the pH remains high. At high pH, Fe(II) rapidly oxidizes to Fe(III) (Fig. 1), which has low solubility so that it rapidly precipitates as a poorly crystalline colloid (Fe(III)–Fe(OH)3 reaction). Because the point of zero charge (PZC) for the Fe oxyhydroxide colloid is between pH 8 and 9 it has a very slight positive charge (Schick, 2001) and the
Fig. 7. Schematic diagram showing the steps of Fe oxyhydroxide coating formation on pyrite. Stages 1a and b show the initial deposition of a porous and permeable Fe oxyhydroxide coating by the attachment of Fe oxyhydroxide colloidal particles. Stage 2 shows the transition from reaction-limited to diffusion-limited rates as the coating becomes denser and thicker as a result of Fe oxhydroxide precipitation in interstices between the Fe oxyhydroxide colloids.
pyrite with a PZC near pH 1.4 has a strong negative charge (Bebie et al., 1998) so some of the Fe oxyhydroxide colloid is attracted by and attaches to the surface forming a thin, porous and permeable layer (Stage 1b in Fig. 7). During this initial stage the pyrite oxidation rate is reaction-limited because the colloid layer is porous and permeable making it a relatively ineffective barrier to oxidant transport. Caldeira et al. (2003) also report this initial stage of coating formation. In the second stage of development, the coating becomes denser and thicker (Fig. 7). As Fe(II) released by pyrite oxidation diffuses outward and encounters oxidant diffusing inward, the Fe(II) rapidly oxidizes to Fe(III) and precipitates between the attached col-
1631
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
loidal particles. This decreases the porosity and permeability of the coating, which decreases the rate of oxidant transport to the pyrite surface and the rate of Fe(II) transport outward. The presence of Fe(II) encourages the poorly crystalline Fe oxyhydroxide to transform to goethite (Yee et al., 2006). Once the conditions of Stage 2b (Fig. 7) are established, pyrite oxidation rates decline more rapidly with time following the characteristic linear relationship of r versus t1/2 shown in Figs. 3–5. The slopes of the lines in Figs. 3–5 can be used with Eq. (18) to find diffusion coefficients for oxidants through the coatings. The diffusion coefficients for H2O2 are 3.6 1015 m2/s based on the authors’ experiments and 1.7 1020 m2/s based on data from Zhang and Evangelou (1996). These values are much lower than the diffusion coefficient of H2O2 in aqueous solution, which is1.4 109 m2/s (Stewart, 2003). Similarly, the diffusion coefficients for oxygen from the Nicholson et al. (1990) data are 2.38 1017, 2.36 1017, and 9.04 1017 m2/s for the 76, 108, 215 lm grains size, which is also much lower than 2.0 109 m2/s, the diffusion coefficient of DO in water (Han and Bartels, 1996). Nicholson et al. (1990) assumed different values for fppt and u and reported slightly higher values of 3.19 1016, 3.64 1016, and 2.17 1016 m2/s for 76, 108, 215 lm grains sizes, respectively. In any case, it can be concluded that the diffusion coefficients for both H2O2 and DO in Fe oxyhydroxide coatings are lower than in water by more than five orders of magnitude. The conditions where the rates of the governing reactions favor Fe oxyhydroxide coating development can be mapped out in a kinetic predominance diagram. In order for coatings to form on pyrite the rate of Fe oxidation and precipitation must be greater than the rate of pyrite oxidation so that Fe(II) oxidizes on or near the pyrite surface and does not escape to solution. These conditions are met when pH, DO concentration and total Fe concentration fall within a certain range. To find this range the DO concentration is set to 2.7 104 m (air saturation), the rate laws for the governing reactions are made equal to each other, and then solved for the total Fe concentration as a function of pH. The resulting equations are graphed as the lines separating fields of kinetically predominant species in Fig. 8. Fig. 8 is not a thermodynamic phase diagram but rather it shows which Fe species accumulate at various total Fe and pH conditions because of their relative rates of production and consumption. This diagram was constructed by assuming that Fe(II) is constantly being produced by the py–DO reaction (Table 1). For any value of total Fe concentration and pH lying in the field labeled Fe(II), Fe(II) is produced by the py–DO reaction faster than it is consumed by any other reaction so Fe(II) accumulates. If the pH or Fe(II) concentration falls outside of that field, Fe(II) is converted to either Fe(III)(aq) or ferrihydrite faster than it is produced
Fig. 8. Kinetic predominance diagram showing the most abundant Fe species associate with pyrite oxidizing in a solution with a DO concentration of 2.7 104 m. Lines 1, 2 and 3 were found by combining the rate laws in Table 1 and solving for Fe concentration as a function of pH (see text for explanation).
by pyrite oxidation. Line 1 distinguishes between the conditions where the rate of Fe(II) oxidation (Fe(II)–DO reaction) predominates over the rate of pyrite oxidation (py–DO reaction) and was determined by setting
108:19
mFeðIIÞ mDO m0:5 DO ¼ 1012:96 m0:11 m2Hþ Hþ
ð15Þ
and solving for mFe(II) as a function of pH where mDO = 2.7 104 m to get
mFeT ¼ mFeðIIÞ ¼ 106:55 m1:89 Hþ :
ð16Þ
Line 2 separates the conditions where the rate of Fe oxyhydroxide formation by the Fe(III)–Fe(OH)3(s) reaction predominates over the rate of Fe(II) oxidation by the Fe(II)–DO reaction. It was found by setting
1012:96
mFeðIIÞ mDO ¼ kf mFeD mFeT m2Hþ
ð17Þ
and solving for total Fe, mFeT , as a function of pH, where mDO = 2.7 104 m, mFeD is the solubility of ferrihydrite, and mFe(II) is fixed by Eq. (16) so that
mFeT ¼
1012:963 mFeðIIÞ mDO m2Hþ kf mFeD
ð18Þ
The values from Ninh Pham et al. (2006) were used for the rate constant, kf, as a function of pH between 6 and 9.5. The concentration of dissolved Fe(III), mFeD , as a function of pH for ferrihydrite was estimated from the solubility curves in Langmuir (1997) and substituted into Eq. (22) to obtain line 2 in Fig. 8. Line 3 is the lower bound for conditions where the rate of Fe oxyhydroxide precipitation by the Fe(II)–fh reaction (Park and Dempsey, 2005) exceeds the rate of Fe(II) production by the py–DO reaction. This line was found by setting
107:3 mFeðIIÞ mFeðIIÞsorbed mDO ¼
108:19 m0:5 DO m0:11 Hþ
ð19Þ
and solving to get
mFeT ¼ mFeðIIÞ ¼
1015:49 0:11 m0:5 DO mFeðIIÞsorbed mHþ
ð20Þ
This reaction rate depends on the amount of Fe adsorbed, mFe(II)the surface of ferrihydrite; a value of 2 103 was used. This diagram shows that the Fe(II) concentration at the surface of the pyrite must accumulate to the levels indicated by lines 1 or 3 before significant amounts of Fe(III) are produced and the total Fe concentration must exceed the amounts indicated by lines 1 and 2 before significant amounts of ferrihydrite will form. Such levels might not be achieved if flowing solutions sweep the dissolved Fe away from the pyrite surface and lower the fraction of Fe precipitated on the pyrite surface in stage 1 of the coating process. This diagram was not extended to lower pH because the solubility of ferrihydrite increases very rapidly below pH 6 and coatings are unlikely to form at low pH. The effectiveness of coatings for reducing pyrite oxidation rates depends on the properties of the coating. Ostwalds’ step rule predicts that the more soluble amorphous Fe oxyhydroxides will precipitate first and then convert to poorly crystalline ferrihydrite. Then, goethite, which is the most stable phase, will grow at the expense of ferrihydrite. The rate of transition to goethite at high pH depends upon the Fe(II) concentration (Yee et al., 2006). In low pH AMD, Fe(II) is rapidly oxidized to Fe(III) by microbes. Similarly, in the H2O2 experiments Fe(II) was oxidized very rapidly. However, DO oxidation of Fe(II) is much slower so that a significant amount of Fe(II) is likely to exist in the pores of the coating, which will en-
sorbed,to
1632
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
hance the rate of the ferrihydrite to goethite transformation by the Fe(OH)3–goe reaction. Powder X-ray diffraction identified only goethite in the limonite pseudomorphs after pyrite (Fig. 2). 5. Applications The goal of this paper is to report some conceptual and quantitative principles that can be used to implement Fe oxyhydroxide coatings on pyrite as a means of AMD management. For example, both the colloid attachment model (Fig. 7) and kinetic predominance diagram (Fig. 8) suggest that hydrodynamic conditions play an important role in the early stages of coating development. If flow rates are too high, the Fe(II) and colloidal Fe oxyhydroxides will be swept away from the pyrite before they can react and attach. This explains why Fe oxyhydroxide coatings did not develop on pyrite grains in the stirred experiments of Pérez-López et al. (2007a). Although a quantitative model that considers these hydrodynamic effects was not produced, the authors believe that the rate laws from Table 1 can be combined with the concepts in Figs. 7 and 8 to create one. A more pressing problem is to estimate what rates of alkalinity addition are needed, and for how long, in order to develop and maintain Fe oxyhydroxide coatings on pyrite. The experiments show that the initial stage (Stage 1, Fig. 7) of coating development is so brief that it can be ignored in long-term models of coating behavior. In order to provide an example of how much bicarbonate alkalinity addition is necessary to sustain coating development, the following scenario was investigated. A 1 m3 volume of mine waste with a 1 m2 cross-sectional area and 1 m depth was considered. It is assumed that this mine waste consists of coarse sand sized particles containing 10% pyrite under saturated conditions with 1 m2 of pyrite surface area per 1 kg of solution. The rate of H+ production is related to the rate of pyrite oxidation by a summation of the py–DO and Fe(III)–Fe(OH)3(s) reactions (Table 1) þ FeS2 þ 3:75O2 þ 3:5H2 O ¼ FeðOHÞ3 þ 2SO2 4 þ 4H
ð21Þ
so that 4 mol of H+ ions are produced for every 1 mol of pyrite oxidized, i.e. r Hþ ¼ 4r py . Based on this the model can be recast for the rate of pyrite oxidation in the presence of growing Fe oxyhydroxide coatings, Eq. (12), to find the rate of H+ generation as a function of time
r Hþ ¼ 2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t : mi fppt V m
ð22Þ
Fig. 9 shows how this rate decreases with time for a DO concentration of 2.7 104 m (air saturated water), a diffusion coefficient, DDO, of 1 1017 m2/s, and a coating porosity of 0.1. Based on Eq. (21) mO2 is 3.75; fppt, was set as 0.03; and Vm (goethite) is 2.02 105 m3/mol. The initial rate of H+ production by the reaction-limited oxidation reaction at a pH of 8.5 is 108.4 mol/s, so bicarbonate alkalinity must be added to the mine wastes by infiltrating groundwater at a rate of at least 108.4 mol/s in order to neutralize H+ from pyrite oxidation. A typical rate for bicarbonate alkalinity delivery by flowing ground water might involve an average groundwater infiltration rate, Ir, of 1010 m/s (Gleisner, 2005) with a HCO 3 concentration of 0.41 mol/m3 (25 ppm), which is near the minimum for typical groundwater (Langmuir, 1997). This situation would add alkalinity fast enough to consume H+ ions at a rate of
r Hþ ¼ AIr C HCO3 ¼ 1010:4 mol=s
ð23Þ
Fig. 9 shows that this rate is nearly 100 times lower than needed to neutralize the initial rate of H+ production by reaction-limited pyrite oxidation so that additional alkalinity must be added to
Fig. 9. The modeled decrease in the rate of H+ production by coated pyrite. The arrow shows the rate of H+ production for pyrite oxidation with no coatings. The tick marks on the right axis represent the HCO 3 concentration required to neutralize the H+ produced at the corresponding rate of H+ production shown on the left axis when the HCO3 is carried into the mine waste by groundwater with an average infiltration rate of 1010 m/s.
avoid the development of runaway AMD. Initially the infiltrating groundwater would have to have a HCO 3 concentration in excess of 2215 ppm in order to maintain a high pH and initiate coating formation. However, after 10 a the rate of acid generation drops to 109.8 mol/s, which could be neutralized by groundwater with a HCO 3 concentration of 95 ppm and after 50 a groundwater with HCO 3 concentration 35 ppm would be sufficient to maintain coating stability and growth (Fig. 9). The concentration for HCO 3 in natural waters is typically between 25 and 200 ppm (Langmuir, 1997) so it is concluded that natural sources of alkalinity would need to be supplemented over the first few decades of treatment but eventually the rate of H+ generation from pyrite oxidation would decline until the natural rate of alkalinity delivery is sufficient to neutralize any acid produced. Crushed limestone is often used to treat pyrite bearing mine wastes because it is inexpensive and highly reactive. Fig. 1 shows that calcite dissolves 3–4 orders of magnitude faster than pyrite. In order to take advantage of limestone’s neutralization capacity, the limestone dose necessary to maintain the pore water pH in the range where Fe oxyhydroxide coatings form needs to be known. The dose rate can be determined by constructing a kinetic predominance diagram that shows conditions where the rate of alkalinity production by calcite dissolution equals the rate of H+ production by the py–DO reaction. For pH P 5.5, calcite dissolves in the presence of CO2 by the reactions
CaCO3 þ H2 CO3 ¼ Ca2þ þ 2HCO3
ð24Þ
and
CaCO3 þ H2 O ¼ Ca2þ þ HCO3 þ OH
ð25Þ
In both cases the dissolution of 1 mol of calcite releases 2 mol of alkalinity (Plummer et al., 1978). The rate of calcite dissolution by these reactions is
rc ¼ 104:88 PCO2 þ 105:93
ð26Þ
(Plummer et al., 1978). Dissolution of 1 mol of pyrite by the py–DO reaction (Table 1) produces 2 mol of H+ ions and 1 mol of Fe(II), which oxidizes to Fe(III) and precipitates as Fe oxyhydroxide to produces an additional 2 mol of H+ ions. Assuming that the overall rate is controlled by the pyrite oxidation step, the overall rate of H+ production is
rp ¼ 109:34
P 0:5 O2 m0:11 Hþ
!
ð27Þ
(Williamson and Rimstidt, 1994). For this example, pH = 8, PCO2 ¼ 0:1, and P O2 ¼ 0:2 were chosen, which makes rc = 105.60 mol/m2 s and rp = 108.81 mol/m2 s. For this fixed set of
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
chemical conditions, the rate of delivery of alkalinity ðr 0c Þ or acidity ðr0p Þ to the solution is controlled by the surface area of the minerals. The specific surface of spherical mineral grains is
Asp ¼
6
qD
ð28Þ
Based on a calcite density of 2.71 g/cm3 and a pyrite density of 5.01 g/cm3, the specific surface areas of the calcite and pyrite are: Asp(c) = 0.120/Dc and Asp(p) = 0.221/Dp (m2/kg). This makes the rates of calcite and pyrite dissolution
r 0c ¼
0:221Mc ð105:60 Þ Dc
ð29Þ
0:120M p ð108:81 Þ Dp
ð30Þ
and
r 0p ¼
Setting Eq. (29) equal to (30) gives the conditions where the rate of alkalinity release by calcite dissolution equals the rate of acidity release by pyrite oxidation.
Mc Dc ¼ 103:48 Mp Dp
ð31Þ
This relationship is graphed in Fig. 10 in a way that allows rapid estimation of the minimum mass ratio of calcite to pyrite, Mc/Mp, needed in order to insure that the rate of alkalinity generation exceeds the rate of acidity production. For example, if a mine waste containing 1 mm diameter pyrite grains were treated with 1 cm diameter calcite grains, the calcite would consume H+ ions as fast at they are produced by the pyrite if the Mc/Mp ratio equals 102.44 (3.6 103), which is approximately 1 kg of calcite per 275 kg of pyrite. This oversimplified model estimates the smallest limestone dosage needed to avoid immediate development of runaway AMD. If some of the pyrite grains were smaller than 1 mm, a higher initial dosage rate would be needed. Additional models are needed to assess how effectively the alkalinity would be transported from the calcite to the pyrite and how the relative rates of acidity and alkalinity production would change over time. However, this simple model shows that in principle a relatively small amount of calcite could be sufficient to promote the growth of Fe oxyhydroxide coatings and to forestall runaway AMD development. These examples illustrate of how treating mine wastes to produce high alkalinity pore waters could encourage the growth of a
Fig. 10. Kinetic predominance diagram showing the mass ratio of calcite to pyrite that is needed for calcite dissolution to produce alkalinity as fast as pyrite oxidation produces acidity (py–DO reaction) for a pH 8 solution in equilibrium with P CO2 ¼ 0:1 atm and P O0 :2 atm.
1633
passive layer of Fe hydroxides on pyrite surfaces and reduce the pyrite oxidation rate until the natural alkalinity could neutralize the acid produced. Field implementation of this scheme will require simultaneous optimization of several factors. In the present experiments, NaHCO3 solutions were used as the primary source of alkalinity because they buffer the pH between 8 and 8.5, which is below the PZC of Fe oxyhydroxide colloids so they were attracted to the negatively charged pyrite surface. However, using an alkalinity source that produces a solution with a pH higher than the PZC for Fe oxyhydroxides might significantly reduce colloid attachment. Also, if the infiltration rate is too low alkalinity will not be delivered fast enough but if the infiltration rate is too fast dissolved Fe and Fe oxyhydroxide colloids will be flushed away before they can contribute to coating growth. These considerations must be balanced against cost and other possible advantages of various alkalinity sources such as limestone, burnt lime, cememtitious ash, and brucite (Barnes and Gold, 2008; Homstrom et al., 1999; Hossner and Doolittle, 2003; Pérez-López et al., 2007a,b). The models presented in this paper are intended as the first step toward the establishment of the principles for using Fe oxyhydroxide coatings in pyrite-bearing waste management. The goal was to estimate the approximate magnitude of some of the factors that affect acid neutralization and Fe oxyhydroxide coating formation. The reaction rate data used in the models are for mineral grains that are fully immersed in aqueous solution. As a result, they overestimate the rate of delivery of alkalinity by flowing water to pyrite in unsaturated wastes but probably predict the rate of acid production reasonably well. Jerz and Rimstidt (2004) showed that rates of pyrite oxidation in moist air are comparable to those in DO saturated solutions. On the other hand, the models overestimate the rate of pyrite oxidation for conditions were DO transport rates are low but the alkalinity deliver rates by flowing water are high. These models should be viewed as a starting point for well designed bench and pilot scale experiments (e.g. Barnes and Gold, 2008) that can further test the concept and create more refined models. Acknowledgment The authors very much appreciate the thorough reviews by Bob Seal and an anonymous reviewer. References Barnes, H.L., Gold, D.P., 2008. Pilot tests of slurries for in situ remediation of pyrite weathering products. Environ. Eng. Geosci. 14, 31–41. Bebie, J., Schoonen, M.A.A., Fuhrmann, M., Strongin, D.R., 1998. Surface charge development on transition metal sulfides: an electrokinetic study. Geochim. Cosmochim. Acta 62, 633–642. Caldeira, C.L., Ciminelli, V.S.T., Dias, A., Osseo-Asare, K., 2003. Pyrite oxidation in alkaline solutions: nature of the product layer. Internat. J. Mineral Process. 72, 373–386. Evangelou, V.P., 1995. Potential microencapsulation of pyrite by artificial inducement of ferric phosphate coatings. J. Environ. Qual. 24, 535–542. Foust, A.S., Wenzel, L.A., Clump, C.W., Maus, L., Andersen, L.B., 1980. Principles of Unit Operations. Wiley, New York. Fytas, K., Bousquet, P., 2002. Silicate micro-encapsulation of pyrite to prevent acid mine drainage. CIM Bull. 95, 96–99. Fytas, K., Evangelou, B., 1998. Phosphate coating on pyrite to prevent acid mine drainage. Int. J. Surface Mining, Reclamation Environ. 12, 101–104. Fytas, K., Bousquet, P., Evangelou, B., 1999. Application of silicate coatings on pyrite to prevent acid mine drainage. Proc. Mining Environ. II, Sudbury, Ontario 3, 1199–1207. Gleisner, M., 2005. Quantification of Mineral Weathering Rates in Sulfidic Mine Tailings under Water-Saturated Conditions. Stockholm University. Han, P., Bartels, D.M., 1996. Temperature dependence of oxygen diffusion in H2O and D2O. J. Phys. Chem. 100, 5597–5602. Homstrom, H., Ljungberg, J., Ohlander, B., 1999. Role of carbonates in mitigation of metal release from mining waste. Evidence from humidity cell tests. Environ. Geol. 37, 267–280. Hossner, L.R., Doolittle, J.J., 2003. Iron sulfide oxidation as influenced by calcium carbonate application. J. Environ. Qual. 32, 773–780.
1634
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Huminicki, D.M.C., 2006. Effect of coatings on mineral reaction rates in acid mine drainage, Virginia Polytechnic Institute and State University.
sulphide-rich mining waste: column experiments. Chemosphere 67, 1637– 1646. Plummer, L.N., Wigley, T.M.L., Parkhurst, D.L., 1978. The kinetics of calcite dissolution in CO2–water systems at 5° to 60 °C and 0.0 to 1.0 atm CO2. Am. J. Sci. 278, 179–216. Pollard, J.H., 1977. Numerical and Statistical Techniques. Cambridge University Press, London. Schick, M.J., 2001. Chemical properties of material surfaces. In: Hubbard, A.T. (Ed.), Surfactant Science Series. Marcel Dekker, Inc., New York. Smith, M.W., Brady, K.B.C., 1998. Alkaline addition Coal Mine Drainage Prediction and Pollution Prevention in Pennsylvania. Department of Environmental Protection, Pennsylvania. Stewart, P.S., 2003. Guest commentaries, diffusion in biofilms. J. Bacteriol. 185, 1485–1491. Stumm, W., Lee, G.F., 1961. Oxygenation of ferrous iron. Ind. Eng. Chem. 53, 143– 146. White III, W.W., Lapakko, K.A., Cox, R.L., 1999. Static-test methods most commonly used to predict acid-mine drainage: practical guidelines for use and interpretation. In: Plumlee, G.S., Logsdon, M.J. (Eds.), The Environmental Geochemistry of Mineral Deposits. Part A: Processes Techniques and Health Issues. Society of Economic Geologists, Inc., pp. 325–338. Williamson, M.A., Rimstidt, J.D., 1994. The kinetics and electrochemical ratedetermining step of aqueous pyrite oxidation. Geochim. Cosmochim. Acta 58, 5443–5454. Williamson, M.A., Kirby, C.S., Rimstidt, J.D., 2006. Iron dynamics in acid mine drainage. In: Barnhisel, R.I. (Ed.), 7th Internat. Conf. Acid Rock Drainage. American Society of Mining and Reclamation, Lexington, KY 40502, St. Louis MO, pp. 2411–2423. Yee, N., Shaw, S., Benning, L.G., Hien Nguyen, T., 2006. The rate of ferrihydrite transition to goethite via the Fe(II) pathway. Am. Mineral. 91, 92–96. Zhang, Y.L., Evangelou, V.P., 1996. Influence of iron oxide forming conditions on pyrite oxidation. Soil Sci. 161, 852–864. Zhang, Y.L., Evangelou, V.P., 1998. Formation of ferric hydroxide–silica coatings on pyrite and its oxidation behavior. Soil Sci. 163, 53–62.
Contents lists available at ScienceDirect
Applied Geochemistry journal homepage: www.elsevier.com/locate/apgeochem
Iron oxyhydroxide coating of pyrite for acid mine drainage control Danielle M.C. Huminicki, J. Donald Rimstidt * Virginia Polytechnic Institute and State University, Department of Geosciences, 4044 Derring Hall, Blacksburg, VA 24061, USA
a r t i c l e
i n f o
Article history: Received 31 December 2008 Accepted 21 April 2009 Available online 5 May 2009 Editorial handling by R. Fuge
a b s t r a c t When pyrite oxidizes at near neutral pH in the presence of sufficient alkalinity, Fe oxyhydroxide coatings develop on the surface. As these coatings grow thicker and denser they block oxidant transport from the solution to the pyrite surface and reduce the rate of pyrite oxidation. The authors’ measurements of pyrite oxidation rates in a NaHCO3 solution show that the coating grows in two stages. In the first stage Fe oxyhydroxide colloids form and then attach to the pyrite surface to produce a slight reduction in oxidant transport. In the second stage interstitial precipitation of Fe oxyhydroxide material between the colloidal particles reduces the oxidant’s diffusion coefficient by more than five orders of magnitude. This causes the pyrite oxidation rate to decline as the square root of time. The kinetic predominance diagram, which compares the rates of Fe transformation reactions, shows that when pyrite oxidation releases Fe quickly enough for the total Fe concentration to rise to about 108 m, ferrihydrite forms but lower rates of Fe release will not produce coatings. Extrapolation of the results to longer times predicts that pyrite-bearing materials need to be treated with an extra source of alkalinity for several decades to produce coatings that are thick enough to be sustained by alkalinity levels typical of groundwater. However, once the coatings develop no additional treatment is needed and further pyrite oxidation simply causes the coating to grow thicker and denser until the entire pyrite grain is pseudomorphically replaced by goethite. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction Although acid mine drainage (AMD) is a widespread and intensely studied environmental problem, there is a continuous struggle to find scientifically sound and economically viable strategies to mitigate it. Current AMD management practices often treat the symptom, which is the acidic effluent, but not the source, which is the oxidizing pyrite. AMD develops when acid is generated by sulfide mineral oxidation faster than it is neutralized by alkalinity from the surroundings so an important strategy for long-term remediation programs is to supply alkalinity to the mine wastes faster than their rate of acid production. Acid base accounting in mine wastes is a well-established tool for predicting whether there is a sufficient amount of alkalinity generating minerals already in the mine wastes (White et al., 1999) and when a sufficient alkalinity supply does not exist, various alkalinity-generating materials can be incorporated into the wastes (Smith and Brady, 1998). However, having a favorable acid base accounting is not always sufficient to mitigate AMD. In order to be effective, the alkalinity-generating materials must dissolve and supply alkalinity as fast, or faster than acidity is generated by pyrite oxidation. In order to treat the wastes in the most economic way it is necessary to determine just how fast alkalinity must be supplied immediately after the wastes are disposed, when pyrite oxidation rates are fastest, and then later * Corresponding author. Tel.: +1 540 392 8913; fax: +1 540 231 3386. E-mail address: [email protected] (J.D. Rimstidt). 0883-2927/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apgeochem.2009.04.032
on after the oxidation rates have declined. In most cases a decline in acid generation is simply due to the consumption of pyrite but both field and experimental evidence suggests that a more rapid decrease in acid generation occurs if the waste disposal is designed to encourage the growth of a layer of Fe oxyhydroxides on the pyrite surface. As this Fe oxyhydroxide coating grows thicker and denser over time, it becomes an increasingly effective barrier to oxidant transport to the pyrite surface thereby slowing the oxidation rate and the rate of acid generation. Eventually the coating becomes so effective that the rate of acid generation falls below the rate of alkalinity delivery by groundwater and AMD generation stops. This could occur even if the acid base accounting is unfavorable but it does require very rapid alkalinity addition soon after disposal. Because of its abundance, pyrite is commonly the main source of acid production in mine wastes. The pH of freshly exposed pyrite bearing rocks is usually greater than four so that dissolved O2 (DO) is the principal pyrite oxidant and the rate of acid production is relatively low (Fig. 1, py–DO reaction in Table 1) (Williamson and Rimstidt, 1994). However, if this acid is not neutralized as quickly as it is produced the pH drifts downward until it drops below four where dissolved Fe(III) becomes the principal oxidant. A further decrease in pH causes a large increase in Fe(III) concentration leading to even faster pyrite oxidation (Fig. 1, py–Fe(III) reaction in Table 1). This creates a runaway condition where fast pyrite oxidation continually lowers pH and raises the Fe(III) concentration causing the oxidation rate to accelerate. Because microbial activity
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
1627
Notation A surface area (m2) specific surface area (m2/g) Asp b surface area constant (unitless) concentration of species i (mol/m3) Ci D diffusion coefficient (m2/s) Dc and Dp diameter of calcite and pyrite particles groundwater infiltration rate (m/s) Ir flux (mol/m2 s) Ji rate constant ki concentration of species i (mol/kg) mi concentration of species i (mol/L) Mi Mc and Mp mass of calcite and pyrite
quickly regenerates Fe(III) from the Fe(II) released by the pyrite (Williamson et al., 2006), there is always sufficient Fe(III) to sustain the runaway AMD condition. Avoiding or recovering from runaway AMD requires that the acid from the py–DO or py–Fe(III) reaction be neutralized by the addition of alkalinity faster than it is produced. When the rate of alkalinity addition meets or exceeds this rate, the Fe released from the pyrite rapidly oxidizes to Fe(III) and precipitates as an Fe oxyhydroxide coating on the pyrite surface. Coating pyrite grains with a substance that blocks oxidant transport from solution to the surface may be a practical way to reduce the oxidation rate. Several types of pyrite coatings have been proposed including ferric phosphate (Evangelou, 1995), phosphosilicates (Fytas and Bousquet, 2002; Fytas et al., 1999; Fytas and Evangelou, 1998), ferric hydroxide–silica (Zhang and Evangelou, 1998), phospholipids (Kargbo et al., 2004), and iron-8 hydroxy-
Fig. 1. Comparison of the rates of generation and consumption of Fe(II) in AMD solutions using the rate laws and assumptions described in Williamson et al. (2006). At pH > 4, the py–DO reaction rate is faster than the py–Fe(III) rate and the Fe(II) released from the pyrite is abiotically oxidized. The resulting Fe(III) rapidly precipitates from solution. At pH < 4, the py–Fe(III) reaction is fastest and the Fe(II) released from the pyrite is microbally oxidized. The resulting Fe(III) is very soluble and available to react with more pyrite. The grey zone in the upper right corner shows the range of rates for H+ consumption by dissolving calcite (Plummer et al. 1978). All rates are for 1 m2 of mineral surface area per 1 kg of solution.
ni rf ri r 0i t tR VR Vm x u
amount of species i (mol) flow rate (kg/s) (1 kg H2O 0.001 m3) specific rate of reaction of species i (mol/m2 s) apparent rate of reaction of species i (mol/s) time (s) response time (s) volume of reactor (m3) molar volume (m3/mol) thickness of coating (m) porosity, unitless ratio from 0 to 1
quinoline (Lan et al., 2002). These coatings have been shown to reduce oxidation rates in the laboratory but they require relatively expensive reagents, which do not specifically target the pyrite so that a large excess may be required to insure an effective coating. In addition, because they are not self-healing and permanent they require a long-term commitment to site management. In comparison, Fe oxyhydroxide coatings form naturally in high alkalinity environments, they are self-healing and they become more and more effective with time so that the demand for alkalinity addition declines. Nicholson et al. (1990) showed experimentally that the addition of relatively inexpensive bicarbonate alkalinity to oxidizing pyrite produces Fe oxyhydroxide coatings that slow oxidant transport from solution to the pyrite and thus slow pyrite oxidation rates. The effectiveness of this strategy is supported by more recent laboratory studies (Homstrom et al., 1999; Pérez-López et al., 2005, 2007b; Zhang and Evangelou, 1996). We can predict the long-term behavior of Fe oxyhydroxide coatings by considering a natural analogue, limonite pseudomorphs after pyrite (Fig. 2). Limonite pseudomorphs form where pyrite has oxidized under high alkalinity conditions. They are often found in limestone where large amounts of HCO 3 ions are available to neutralize the acid from pyrite oxidation. Under high pH conditions, abiotic oxidation of Fe(II) (Stumm and Lee, 1961) is much faster than pyrite oxidation by dissolved O2 (Fig. 1) so that Fe(II) released from the pyrite quickly oxidizes to Fe(III). Because Fe(III) is very insoluble at near neutral pH, it quickly hydrolyzes and precipitates. Initially formed, metastable Fe(OH)3 and/or ferrihydrite eventually converts to goethite. These coupled reactions produce an approximately 1 for 1 volume replacement of pyrite (23.94 cm3/mol) by goethite (20.82 cm3/mol) creating a pseudomorph with only about 13% porosity. Additional pore filling by the oxidation and precipitation of outward diffusing Fe(II) fills the pores near the outer edge of the pseudomorph to create a dense outer rim. Fig. 2 shows that a typical limonite pseudomorph has a relatively porous interior, often with some remaining pyrite, surrounded by a relatively dense outer rim. Limonite pseudomorphs after pyrite are compelling evidence that Fe oxyhydroxide coatings can be stable and grow to great thickness on pyrite grains. The objective of this paper is to present a conceptual and quantitative model of declining pyrite oxidation rate caused by the formation of a Fe oxyhydroxide coating under high alkalinity conditions. Experiments were performed that bridge the time in the coating process between the initial attachment of Fe(OH)3 colloids and the infilling between the colloids to form a more dense coating. Data from the authors’ and previous experiments (Nicholson et al., 1990; Zhang and Evangelou, 1996) were analyzed to determine the effective diffusion coefficients of oxidants through
1628
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Table 1 Empirical rate laws for chemical reactions discussed in this paper. In the text, the rate laws are named by the most important reactants, which are shown in italics in the left column. Rate laws for solids have been scaled to a system with 1 m2 of mineral surface area per 1 kg of solution. Reaction
r (mol/s)
pH
References
py–H2O2 þ FeS2 ðpyÞ þ 7H2 O2 ¼ Fe2þ þ 2SO2 4 þ 2H þ 6H2 O
dMFeðIIÞ dt
¼ 105:82 M 0:93 H2 O2
1–4
McKibben and Barnes (1986)
py–DO þ FeS2 ðpyÞ þ 7=2O2 þ H2 O ¼ Fe2þ þ 2SO2 4 þ 2H
dmFeðIIÞ dt
DO ¼ 108:19 m0:11
2–10
Williamson and Rimstidt (1994)
py–Fe(III) þ FeS2 ðpyÞ þ 14Fe3þ þ 8H2 O ¼ 15Fe2þ þ 2SO2 4 þ 16H
dmFeðIIÞ dt
¼ 108:58 m0:47FeðIIIÞ m0:32
0.5–3
Williamson and Rimstidt (1994)
Fe(II)–H2O2 Fe2+ + 0.5H2O2 + H+ = Fe3+ + H2O
dmFeðIIÞ dt
¼ k2
6–8
Millero and Sotolongo (1989)
4–8
Stumm and Lee (1961)
6–9
Ninh Pham et al. (2006)
6–8
Yee et al. (2006)
6.8–7.0
Park and Dempsey (2005)
m0:5 Hþ
m0:3
FeðIIÞ
Hþ
mFeðIIÞ mH2 O2 mHþ
log k2 = 11.72 2.14I Fe(II)–DO Fe2+ + 1/4 O2 + H+ = Fe3+ + 1/2H2O
dmFeðIIÞ dt
¼ 1012:96
Fe(III)–Fe(OH)3(s) a FeD + FeT = AFO + nH+
dMFeðIIIÞ dt
¼ kf M FeD M FeT
fh–goe b Fe5O3(OH)9(fh) = 5FeOOH (goe) + 2H2O
a ¼ 1 e410
Fe(II)–fh 5Fe2+ + 5/4 O2 + 19/2 H2O = Fe5O3(OH)9(fh) + 10 H+
dmFeðIIÞ dt
a b
5
1/2
+ 1.38I
mFeðIIÞ mDO mH þ2
ðt3384Þ1:55
¼ 107:3 mFeðIIÞ mFeðIIÞsorbed mDO
FeD = dissolved Fe(III), FeT = dissolved and precipitated Fe(III), AFO = amorphous Fe oxyhydroxide (Fe(OH)3). T = 21 °C; a is the extent of reaction (0–1).
Fig. 3. Schematic diagram of mixed flow reactor experiment.
Fig. 2. Cross-section of a limonite pseudomorph from Bedford Co., Virginia, USA showing a porous center containing a small amount of unreacted pyrite surrounded by a dense outer coating.
the Fe oxyhydroxide coatings and that information was used to predict how alkalinity demand declines over time as the coatings grow thicker. 2. Methods and materials Pyrite from the same source as Jerz and Rimstidt (2004) and Williamson and Rimstidt (1994) was prepared in the manner described in Jerz and Rimstidt (2004) to recover the 250–420 lm size fraction with an estimated specific surface area of 1.13 102 m2/ g (Foust et al., 1980). The grains were sonicated in a 0.1 m HCl solution then rinsed with ethyl alcohol until the supernatant was clear and the grains showed no visible oxidation product layer. These grains were reacted with a mixture of 0.3 m H2O2 and 0.1 m NaHCO3 solution with a pH of approximately 8.5 in order to induce Fe oxyhydroxide coatings. The peroxide–bicarbonate solution was circulated through 5 g of pyrite in a packed bed (Fig. 3). Solution samples were collected at evenly spaced time intervals of 594 s using a fraction collector. The mass of the solution in the reactor during the experiment
was 2.43 ± 0.2 g. Experiments were run for 24 h at 23 ± 1 °C. The flow rate was determined by weighing each sample container before and after sample collection. The average flow rate was approximately 0.5 g/s. The total Fe concentration of each sample was determined using an atomic absorption spectrophotometer (AA). Sulfate concentrations were determined from another set of filtered samples by ion chromatography (IC). To confirm that all S was oxidized to SO4 some ICP-MS sample analyses that measured total S were compared with those from the IC. These values corresponded to each other so it was concluded that all S was in the form of SO4. Reacted grains were observed and characterized using a scanning electron microscope and visible light microscopy. Iron that precipitated in the reactor over the course of the experiment was recovered by leaching the entire reactor by immersing it in 500 mL of a 0.04 M NH2OHHCl in 25% (v/v) CH3COOH solution at 100 °C for 1 h. The concentration of Fe in the leachate was determined by AA. Sulfate was used as the reaction progress variable. The rate of species release was calculated from the solution flow rate multiplied by the change in species concentration
ri ¼ r f ðmout min Þ
ð1Þ
Pyrite oxidation rates by H2O2 and O2 were calculated from the experimental data as well as from Zhang and Evangelou (1996) and Nicholson et al. (1990) based on the stoichiometry of the py–H2O2
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
and py–DO reactions in Table 1. Rates were calculated by numerical differentiation (Pollard, 1977) of their Fe concentration (mol) versus time data (all data are tabulated in Huminicki, 2006). In the experiments relatively little of the Fe precipitated as coatings. The total amount of Fe released (nFe(cal)) was calculated from the amount of SO4 released based on the reaction stoichiometry for the py–H2O2 reaction (Table 1). This Fe was quickly oxidized to Fe(III), which either precipitated onto the pyrite grains (nFe(py)), precipitated onto the reactor walls (nFe(walls)), or was carried out of the reactor by solution (nFe(sol)) so that
nFeðcalÞ ¼ nFeðpyÞ þ nFeðwallsÞ þ nFeðsolÞ
ð2Þ
The amount of pyrite oxidized and the amount of Fe that was carried out of the reactor by the solution were both determined (Huminicki, 2006) at the end of the experiment to be 2.5 104 and 1.5 104 mol, respectively. The difference between these amounts is the total amount of Fe that precipitated onto either the reactor or the pyrite. This was about 40% of the total Fe released. Of this, 37% of the precipitated Fe was recovered by leaching the reactor and tubing (9.12 105 mol) and it was inferred that the rest (3%) had precipitated on the pyrite surfaces. Coating thickness was calculated from the amount of Fe precipitated on the pyrite surface, the molar volume of the coating, the surface area of pyrite, and the porosity of the coating
x¼
nFe fppt V m Að1 /Þ
ð3Þ
The coating thickness was estimated as a function of time for all of the experiments assuming that in each case only 3% of the Fe released precipitated as a coating (fppt = 0.03). It was assumed that the coatings had 10% porosity (u = 0.1) and the coatings consisted of goethite. These thickness values are given in Huminicki (2006). In Eq. (3) the number of moles of Fe released, the molar volume, and the surface area of the pyrite are all well constrained. Because fppt is the difference between two much larger numbers it caries a fairly large uncertainty, which dominates the uncertainty in the coating thickness. For example, if fppt = 0.03 ± 0.01, x = 3(±1) 109 m. On the other hand the uncertainty in u has a relatively minor effect on x; changing u from 0.1 to 0.3 only increases x by 25%. Assuming that the coatings consist of ferrihydrite instead of goethite would increase x by about 25%. After approximately 10 h of reaction, graphs of r versus t1/2 all show that pyrite oxidation rates decreased over time, presumably because the growing coating reduced the rate of oxidant transport to the pyrite surface. The expected relationship between r versus t1/2 can be derived from Fick’s first law of diffusion
Ji ¼
DDC x
ð4Þ
This model allowed determination of the diffusion coefficient, Di, for the oxidant through the coating with a concentration gradient, DC, defined as
DC ¼ 1000ðmðsolÞ mðsfcÞ Þ
ð5Þ
where 1000 is a conversion factor that recasts the molal concentrations into units of mol/m3. It can be reasonably assumed that DC approaches m(sol) as m(sfc) becomes small and that DC becomes nearly constant as soon as a significant coating develops. The flux of oxidant, Ji, to the surface of pyrite can be converted to the rate of release of Fe from the surface using
dnFe 1 dnoxid ¼ ¼ r FeS2 dt mi dt
ð6Þ
where, mi, is the stoichiometric coefficient that relates the rate of oxidant consumption to Fe release from the reactions so mH2 O2 ¼ 7 and mDO = 3.5 (Table 1).
1629
A general model that describes the decrease in the rate of pyrite oxidation over time as a coating forms can be developed by combining these relationships and substituting in J i ¼ mi r FeS2 , DC 1000m(sol), and Eqs. (3) and (4) into Eq. (6), to give
dnFeS2 dnFe 1000mðsolÞ DAð1 /Þ ¼ ¼ dt dt mi nFe fppt V m
ð7Þ
that can be rearranged and integrated for the boundary conditions nFe = 0 when t = 0 to get
n2Fe 1000mðsolÞ DAð1 /Þ t ¼ 2 mi fppt V m
ð8Þ
Solving for nFe gives
nFe ¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t mi fppt V m
ð9Þ
which can be simplified by defining
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2000mðsolÞ DAð1 /Þ c¼ mi fppt V m
ð10Þ
so the total amount of Fe released from pyrite at time, t, is
nFe ¼ ct 1=2
ð11Þ
The time derivative of this equation gives the diffusion-limited rate of pyrite oxidation as a function of time,
rFeS2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t mi fppt V m
1 ¼ 2
ð12Þ
It is expected that as the Fe oxyhydroxide coatings thicken, the rate of pyrite oxidation will decline as a function of t1/2. The slope of the linear regression fit of r versus t1/2 is ½ c so that the diffusion coefficient can be solved for from the relationship
1 1 c¼ 2 2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ mi fppt V m
ð13Þ
to find
D¼
1 mi fppt V m c2 2000 mðsolÞ Að1 /Þ
ð14Þ
3. Results The experiment determined how pyrite oxidation rates declined as Fe oxyhydroxide coatings formed and limited the H2O2 transport to the surface. These results allowed documentation of the initial stage of coating formation and determination of a diffusion coefficient for H2O2 for freshly formed coatings. The rate data are tabulated in Huminicki (2006) along with the rates extracted from Zhang and Evangelou (1996) and Nicholson et al. (1990). All rate data are displayed in Figs. 4–6. Figs. 4–6 all show a linear decrease in rate versus t1/2 that is attributed to the growth of an Fe oxyhydroxide coating acting as a barrier to the transport of oxidant to the pyrite surface. The experiments (Fig. 4) show a transition from a shallow slope of r versus t1/2 to a steeper one, which is the result of the coating becoming a more effective barrier to oxidant transport to the pyrite surface. This transition is not seen in Fig. 5 (Zhang and Evangelou, 1996), because this step of coating formation had already occurred during their pretreatment step. The Nicholson et al. (1990) data show an unexplained initial increase in rates with time (up to 11 days) so those data were not used in the r versus t1/2 fits.
1630
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Fig. 4. Graph of r versus t1/2 data from the MFR experiment. The slope of the line changed from shallow r = 2.19 106t1/2 + 3.42 107 to steep r = 7.72 105t1/2 5.31 108 over the experiment indicating that after about 10 h the coating became a more effective barrier to H2O2 transport. The rate law for the py–H2O2 reaction (Table 1) predicts a rate of 4.57 107 mol/m2 s for the 0.3 m H2O2 solution used in the experiments.
Fig. 6. Graph of r versus t1/2 for data from Nicholson et al. (1990). The equations of the lines are r = 8.91 107t1/2 + 1.37 1010, r = 7.26 107t1/2+3.42 1010, and r = 1.07 106t1/2 + 9.5 1011 for the data sets of the experiments that used 76 (open diamonds), 108 (filled squares), 215 (filled triangles) lm grain sizes, respectively. The chemical reaction-limited rate of pyrite oxidation calculated for py–DO from Table 1 is 9.13 1010 mol/m2 s.
Fig. 5. Graph of r versus t1/2 for data from Zhang and Evangelou (1996). The equation for the line is r = 2.60 107t1/2 + 3.47 1011. The rate law for the py– H2O2 reaction (Table 1) predicts a rate of 2.19 107 mol/m2s for the 0.145 m H2O2 solution used in these experiments. That is about two orders of magnitude faster than their fastest rate because a relatively thick Fe oxyhydroxide coating already existed on the pyrite surface due to pretreatment.
4. Discussion The results (Fig. 4) suggest that Fe oxyhydroxide coatings grow on the pyrite in two stages: (1) an initial stage of colloid deposition followed by (2) the densification and inward propagation of the coating (Fig. 7). In the first stage a thin, highly porous and permeable layer of colloidal Fe oxyhydroxide particles deposits on the pyrite surface. This layer provides only a weak barrier to oxidant transport to the pyrite surface so the rate of pyrite oxidation declines very slowly for the first 10 h. In the second stage, Fe oxyhydroxides precipitate in the pores of this layer making it a more effective barrier to DO transport. This causes the pyrite oxidation rate to decrease more rapidly with time as shown by the line with the steeper slope in Fig. 4. When pyrite oxidizes in the presence of excess alkalinity the acidity produced is neutralized near the pyrite surface so the pH remains high. At high pH, Fe(II) rapidly oxidizes to Fe(III) (Fig. 1), which has low solubility so that it rapidly precipitates as a poorly crystalline colloid (Fe(III)–Fe(OH)3 reaction). Because the point of zero charge (PZC) for the Fe oxyhydroxide colloid is between pH 8 and 9 it has a very slight positive charge (Schick, 2001) and the
Fig. 7. Schematic diagram showing the steps of Fe oxyhydroxide coating formation on pyrite. Stages 1a and b show the initial deposition of a porous and permeable Fe oxyhydroxide coating by the attachment of Fe oxyhydroxide colloidal particles. Stage 2 shows the transition from reaction-limited to diffusion-limited rates as the coating becomes denser and thicker as a result of Fe oxhydroxide precipitation in interstices between the Fe oxyhydroxide colloids.
pyrite with a PZC near pH 1.4 has a strong negative charge (Bebie et al., 1998) so some of the Fe oxyhydroxide colloid is attracted by and attaches to the surface forming a thin, porous and permeable layer (Stage 1b in Fig. 7). During this initial stage the pyrite oxidation rate is reaction-limited because the colloid layer is porous and permeable making it a relatively ineffective barrier to oxidant transport. Caldeira et al. (2003) also report this initial stage of coating formation. In the second stage of development, the coating becomes denser and thicker (Fig. 7). As Fe(II) released by pyrite oxidation diffuses outward and encounters oxidant diffusing inward, the Fe(II) rapidly oxidizes to Fe(III) and precipitates between the attached col-
1631
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
loidal particles. This decreases the porosity and permeability of the coating, which decreases the rate of oxidant transport to the pyrite surface and the rate of Fe(II) transport outward. The presence of Fe(II) encourages the poorly crystalline Fe oxyhydroxide to transform to goethite (Yee et al., 2006). Once the conditions of Stage 2b (Fig. 7) are established, pyrite oxidation rates decline more rapidly with time following the characteristic linear relationship of r versus t1/2 shown in Figs. 3–5. The slopes of the lines in Figs. 3–5 can be used with Eq. (18) to find diffusion coefficients for oxidants through the coatings. The diffusion coefficients for H2O2 are 3.6 1015 m2/s based on the authors’ experiments and 1.7 1020 m2/s based on data from Zhang and Evangelou (1996). These values are much lower than the diffusion coefficient of H2O2 in aqueous solution, which is1.4 109 m2/s (Stewart, 2003). Similarly, the diffusion coefficients for oxygen from the Nicholson et al. (1990) data are 2.38 1017, 2.36 1017, and 9.04 1017 m2/s for the 76, 108, 215 lm grains size, which is also much lower than 2.0 109 m2/s, the diffusion coefficient of DO in water (Han and Bartels, 1996). Nicholson et al. (1990) assumed different values for fppt and u and reported slightly higher values of 3.19 1016, 3.64 1016, and 2.17 1016 m2/s for 76, 108, 215 lm grains sizes, respectively. In any case, it can be concluded that the diffusion coefficients for both H2O2 and DO in Fe oxyhydroxide coatings are lower than in water by more than five orders of magnitude. The conditions where the rates of the governing reactions favor Fe oxyhydroxide coating development can be mapped out in a kinetic predominance diagram. In order for coatings to form on pyrite the rate of Fe oxidation and precipitation must be greater than the rate of pyrite oxidation so that Fe(II) oxidizes on or near the pyrite surface and does not escape to solution. These conditions are met when pH, DO concentration and total Fe concentration fall within a certain range. To find this range the DO concentration is set to 2.7 104 m (air saturation), the rate laws for the governing reactions are made equal to each other, and then solved for the total Fe concentration as a function of pH. The resulting equations are graphed as the lines separating fields of kinetically predominant species in Fig. 8. Fig. 8 is not a thermodynamic phase diagram but rather it shows which Fe species accumulate at various total Fe and pH conditions because of their relative rates of production and consumption. This diagram was constructed by assuming that Fe(II) is constantly being produced by the py–DO reaction (Table 1). For any value of total Fe concentration and pH lying in the field labeled Fe(II), Fe(II) is produced by the py–DO reaction faster than it is consumed by any other reaction so Fe(II) accumulates. If the pH or Fe(II) concentration falls outside of that field, Fe(II) is converted to either Fe(III)(aq) or ferrihydrite faster than it is produced
Fig. 8. Kinetic predominance diagram showing the most abundant Fe species associate with pyrite oxidizing in a solution with a DO concentration of 2.7 104 m. Lines 1, 2 and 3 were found by combining the rate laws in Table 1 and solving for Fe concentration as a function of pH (see text for explanation).
by pyrite oxidation. Line 1 distinguishes between the conditions where the rate of Fe(II) oxidation (Fe(II)–DO reaction) predominates over the rate of pyrite oxidation (py–DO reaction) and was determined by setting
108:19
mFeðIIÞ mDO m0:5 DO ¼ 1012:96 m0:11 m2Hþ Hþ
ð15Þ
and solving for mFe(II) as a function of pH where mDO = 2.7 104 m to get
mFeT ¼ mFeðIIÞ ¼ 106:55 m1:89 Hþ :
ð16Þ
Line 2 separates the conditions where the rate of Fe oxyhydroxide formation by the Fe(III)–Fe(OH)3(s) reaction predominates over the rate of Fe(II) oxidation by the Fe(II)–DO reaction. It was found by setting
1012:96
mFeðIIÞ mDO ¼ kf mFeD mFeT m2Hþ
ð17Þ
and solving for total Fe, mFeT , as a function of pH, where mDO = 2.7 104 m, mFeD is the solubility of ferrihydrite, and mFe(II) is fixed by Eq. (16) so that
mFeT ¼
1012:963 mFeðIIÞ mDO m2Hþ kf mFeD
ð18Þ
The values from Ninh Pham et al. (2006) were used for the rate constant, kf, as a function of pH between 6 and 9.5. The concentration of dissolved Fe(III), mFeD , as a function of pH for ferrihydrite was estimated from the solubility curves in Langmuir (1997) and substituted into Eq. (22) to obtain line 2 in Fig. 8. Line 3 is the lower bound for conditions where the rate of Fe oxyhydroxide precipitation by the Fe(II)–fh reaction (Park and Dempsey, 2005) exceeds the rate of Fe(II) production by the py–DO reaction. This line was found by setting
107:3 mFeðIIÞ mFeðIIÞsorbed mDO ¼
108:19 m0:5 DO m0:11 Hþ
ð19Þ
and solving to get
mFeT ¼ mFeðIIÞ ¼
1015:49 0:11 m0:5 DO mFeðIIÞsorbed mHþ
ð20Þ
This reaction rate depends on the amount of Fe adsorbed, mFe(II)the surface of ferrihydrite; a value of 2 103 was used. This diagram shows that the Fe(II) concentration at the surface of the pyrite must accumulate to the levels indicated by lines 1 or 3 before significant amounts of Fe(III) are produced and the total Fe concentration must exceed the amounts indicated by lines 1 and 2 before significant amounts of ferrihydrite will form. Such levels might not be achieved if flowing solutions sweep the dissolved Fe away from the pyrite surface and lower the fraction of Fe precipitated on the pyrite surface in stage 1 of the coating process. This diagram was not extended to lower pH because the solubility of ferrihydrite increases very rapidly below pH 6 and coatings are unlikely to form at low pH. The effectiveness of coatings for reducing pyrite oxidation rates depends on the properties of the coating. Ostwalds’ step rule predicts that the more soluble amorphous Fe oxyhydroxides will precipitate first and then convert to poorly crystalline ferrihydrite. Then, goethite, which is the most stable phase, will grow at the expense of ferrihydrite. The rate of transition to goethite at high pH depends upon the Fe(II) concentration (Yee et al., 2006). In low pH AMD, Fe(II) is rapidly oxidized to Fe(III) by microbes. Similarly, in the H2O2 experiments Fe(II) was oxidized very rapidly. However, DO oxidation of Fe(II) is much slower so that a significant amount of Fe(II) is likely to exist in the pores of the coating, which will en-
sorbed,to
1632
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
hance the rate of the ferrihydrite to goethite transformation by the Fe(OH)3–goe reaction. Powder X-ray diffraction identified only goethite in the limonite pseudomorphs after pyrite (Fig. 2). 5. Applications The goal of this paper is to report some conceptual and quantitative principles that can be used to implement Fe oxyhydroxide coatings on pyrite as a means of AMD management. For example, both the colloid attachment model (Fig. 7) and kinetic predominance diagram (Fig. 8) suggest that hydrodynamic conditions play an important role in the early stages of coating development. If flow rates are too high, the Fe(II) and colloidal Fe oxyhydroxides will be swept away from the pyrite before they can react and attach. This explains why Fe oxyhydroxide coatings did not develop on pyrite grains in the stirred experiments of Pérez-López et al. (2007a). Although a quantitative model that considers these hydrodynamic effects was not produced, the authors believe that the rate laws from Table 1 can be combined with the concepts in Figs. 7 and 8 to create one. A more pressing problem is to estimate what rates of alkalinity addition are needed, and for how long, in order to develop and maintain Fe oxyhydroxide coatings on pyrite. The experiments show that the initial stage (Stage 1, Fig. 7) of coating development is so brief that it can be ignored in long-term models of coating behavior. In order to provide an example of how much bicarbonate alkalinity addition is necessary to sustain coating development, the following scenario was investigated. A 1 m3 volume of mine waste with a 1 m2 cross-sectional area and 1 m depth was considered. It is assumed that this mine waste consists of coarse sand sized particles containing 10% pyrite under saturated conditions with 1 m2 of pyrite surface area per 1 kg of solution. The rate of H+ production is related to the rate of pyrite oxidation by a summation of the py–DO and Fe(III)–Fe(OH)3(s) reactions (Table 1) þ FeS2 þ 3:75O2 þ 3:5H2 O ¼ FeðOHÞ3 þ 2SO2 4 þ 4H
ð21Þ
so that 4 mol of H+ ions are produced for every 1 mol of pyrite oxidized, i.e. r Hþ ¼ 4r py . Based on this the model can be recast for the rate of pyrite oxidation in the presence of growing Fe oxyhydroxide coatings, Eq. (12), to find the rate of H+ generation as a function of time
r Hþ ¼ 2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 2000mðsolÞ DAð1 /Þ 1=2 t : mi fppt V m
ð22Þ
Fig. 9 shows how this rate decreases with time for a DO concentration of 2.7 104 m (air saturated water), a diffusion coefficient, DDO, of 1 1017 m2/s, and a coating porosity of 0.1. Based on Eq. (21) mO2 is 3.75; fppt, was set as 0.03; and Vm (goethite) is 2.02 105 m3/mol. The initial rate of H+ production by the reaction-limited oxidation reaction at a pH of 8.5 is 108.4 mol/s, so bicarbonate alkalinity must be added to the mine wastes by infiltrating groundwater at a rate of at least 108.4 mol/s in order to neutralize H+ from pyrite oxidation. A typical rate for bicarbonate alkalinity delivery by flowing ground water might involve an average groundwater infiltration rate, Ir, of 1010 m/s (Gleisner, 2005) with a HCO 3 concentration of 0.41 mol/m3 (25 ppm), which is near the minimum for typical groundwater (Langmuir, 1997). This situation would add alkalinity fast enough to consume H+ ions at a rate of
r Hþ ¼ AIr C HCO3 ¼ 1010:4 mol=s
ð23Þ
Fig. 9 shows that this rate is nearly 100 times lower than needed to neutralize the initial rate of H+ production by reaction-limited pyrite oxidation so that additional alkalinity must be added to
Fig. 9. The modeled decrease in the rate of H+ production by coated pyrite. The arrow shows the rate of H+ production for pyrite oxidation with no coatings. The tick marks on the right axis represent the HCO 3 concentration required to neutralize the H+ produced at the corresponding rate of H+ production shown on the left axis when the HCO3 is carried into the mine waste by groundwater with an average infiltration rate of 1010 m/s.
avoid the development of runaway AMD. Initially the infiltrating groundwater would have to have a HCO 3 concentration in excess of 2215 ppm in order to maintain a high pH and initiate coating formation. However, after 10 a the rate of acid generation drops to 109.8 mol/s, which could be neutralized by groundwater with a HCO 3 concentration of 95 ppm and after 50 a groundwater with HCO 3 concentration 35 ppm would be sufficient to maintain coating stability and growth (Fig. 9). The concentration for HCO 3 in natural waters is typically between 25 and 200 ppm (Langmuir, 1997) so it is concluded that natural sources of alkalinity would need to be supplemented over the first few decades of treatment but eventually the rate of H+ generation from pyrite oxidation would decline until the natural rate of alkalinity delivery is sufficient to neutralize any acid produced. Crushed limestone is often used to treat pyrite bearing mine wastes because it is inexpensive and highly reactive. Fig. 1 shows that calcite dissolves 3–4 orders of magnitude faster than pyrite. In order to take advantage of limestone’s neutralization capacity, the limestone dose necessary to maintain the pore water pH in the range where Fe oxyhydroxide coatings form needs to be known. The dose rate can be determined by constructing a kinetic predominance diagram that shows conditions where the rate of alkalinity production by calcite dissolution equals the rate of H+ production by the py–DO reaction. For pH P 5.5, calcite dissolves in the presence of CO2 by the reactions
CaCO3 þ H2 CO3 ¼ Ca2þ þ 2HCO3
ð24Þ
and
CaCO3 þ H2 O ¼ Ca2þ þ HCO3 þ OH
ð25Þ
In both cases the dissolution of 1 mol of calcite releases 2 mol of alkalinity (Plummer et al., 1978). The rate of calcite dissolution by these reactions is
rc ¼ 104:88 PCO2 þ 105:93
ð26Þ
(Plummer et al., 1978). Dissolution of 1 mol of pyrite by the py–DO reaction (Table 1) produces 2 mol of H+ ions and 1 mol of Fe(II), which oxidizes to Fe(III) and precipitates as Fe oxyhydroxide to produces an additional 2 mol of H+ ions. Assuming that the overall rate is controlled by the pyrite oxidation step, the overall rate of H+ production is
rp ¼ 109:34
P 0:5 O2 m0:11 Hþ
!
ð27Þ
(Williamson and Rimstidt, 1994). For this example, pH = 8, PCO2 ¼ 0:1, and P O2 ¼ 0:2 were chosen, which makes rc = 105.60 mol/m2 s and rp = 108.81 mol/m2 s. For this fixed set of
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
chemical conditions, the rate of delivery of alkalinity ðr 0c Þ or acidity ðr0p Þ to the solution is controlled by the surface area of the minerals. The specific surface of spherical mineral grains is
Asp ¼
6
qD
ð28Þ
Based on a calcite density of 2.71 g/cm3 and a pyrite density of 5.01 g/cm3, the specific surface areas of the calcite and pyrite are: Asp(c) = 0.120/Dc and Asp(p) = 0.221/Dp (m2/kg). This makes the rates of calcite and pyrite dissolution
r 0c ¼
0:221Mc ð105:60 Þ Dc
ð29Þ
0:120M p ð108:81 Þ Dp
ð30Þ
and
r 0p ¼
Setting Eq. (29) equal to (30) gives the conditions where the rate of alkalinity release by calcite dissolution equals the rate of acidity release by pyrite oxidation.
Mc Dc ¼ 103:48 Mp Dp
ð31Þ
This relationship is graphed in Fig. 10 in a way that allows rapid estimation of the minimum mass ratio of calcite to pyrite, Mc/Mp, needed in order to insure that the rate of alkalinity generation exceeds the rate of acidity production. For example, if a mine waste containing 1 mm diameter pyrite grains were treated with 1 cm diameter calcite grains, the calcite would consume H+ ions as fast at they are produced by the pyrite if the Mc/Mp ratio equals 102.44 (3.6 103), which is approximately 1 kg of calcite per 275 kg of pyrite. This oversimplified model estimates the smallest limestone dosage needed to avoid immediate development of runaway AMD. If some of the pyrite grains were smaller than 1 mm, a higher initial dosage rate would be needed. Additional models are needed to assess how effectively the alkalinity would be transported from the calcite to the pyrite and how the relative rates of acidity and alkalinity production would change over time. However, this simple model shows that in principle a relatively small amount of calcite could be sufficient to promote the growth of Fe oxyhydroxide coatings and to forestall runaway AMD development. These examples illustrate of how treating mine wastes to produce high alkalinity pore waters could encourage the growth of a
Fig. 10. Kinetic predominance diagram showing the mass ratio of calcite to pyrite that is needed for calcite dissolution to produce alkalinity as fast as pyrite oxidation produces acidity (py–DO reaction) for a pH 8 solution in equilibrium with P CO2 ¼ 0:1 atm and P O0 :2 atm.
1633
passive layer of Fe hydroxides on pyrite surfaces and reduce the pyrite oxidation rate until the natural alkalinity could neutralize the acid produced. Field implementation of this scheme will require simultaneous optimization of several factors. In the present experiments, NaHCO3 solutions were used as the primary source of alkalinity because they buffer the pH between 8 and 8.5, which is below the PZC of Fe oxyhydroxide colloids so they were attracted to the negatively charged pyrite surface. However, using an alkalinity source that produces a solution with a pH higher than the PZC for Fe oxyhydroxides might significantly reduce colloid attachment. Also, if the infiltration rate is too low alkalinity will not be delivered fast enough but if the infiltration rate is too fast dissolved Fe and Fe oxyhydroxide colloids will be flushed away before they can contribute to coating growth. These considerations must be balanced against cost and other possible advantages of various alkalinity sources such as limestone, burnt lime, cememtitious ash, and brucite (Barnes and Gold, 2008; Homstrom et al., 1999; Hossner and Doolittle, 2003; Pérez-López et al., 2007a,b). The models presented in this paper are intended as the first step toward the establishment of the principles for using Fe oxyhydroxide coatings in pyrite-bearing waste management. The goal was to estimate the approximate magnitude of some of the factors that affect acid neutralization and Fe oxyhydroxide coating formation. The reaction rate data used in the models are for mineral grains that are fully immersed in aqueous solution. As a result, they overestimate the rate of delivery of alkalinity by flowing water to pyrite in unsaturated wastes but probably predict the rate of acid production reasonably well. Jerz and Rimstidt (2004) showed that rates of pyrite oxidation in moist air are comparable to those in DO saturated solutions. On the other hand, the models overestimate the rate of pyrite oxidation for conditions were DO transport rates are low but the alkalinity deliver rates by flowing water are high. These models should be viewed as a starting point for well designed bench and pilot scale experiments (e.g. Barnes and Gold, 2008) that can further test the concept and create more refined models. Acknowledgment The authors very much appreciate the thorough reviews by Bob Seal and an anonymous reviewer. References Barnes, H.L., Gold, D.P., 2008. Pilot tests of slurries for in situ remediation of pyrite weathering products. Environ. Eng. Geosci. 14, 31–41. Bebie, J., Schoonen, M.A.A., Fuhrmann, M., Strongin, D.R., 1998. Surface charge development on transition metal sulfides: an electrokinetic study. Geochim. Cosmochim. Acta 62, 633–642. Caldeira, C.L., Ciminelli, V.S.T., Dias, A., Osseo-Asare, K., 2003. Pyrite oxidation in alkaline solutions: nature of the product layer. Internat. J. Mineral Process. 72, 373–386. Evangelou, V.P., 1995. Potential microencapsulation of pyrite by artificial inducement of ferric phosphate coatings. J. Environ. Qual. 24, 535–542. Foust, A.S., Wenzel, L.A., Clump, C.W., Maus, L., Andersen, L.B., 1980. Principles of Unit Operations. Wiley, New York. Fytas, K., Bousquet, P., 2002. Silicate micro-encapsulation of pyrite to prevent acid mine drainage. CIM Bull. 95, 96–99. Fytas, K., Evangelou, B., 1998. Phosphate coating on pyrite to prevent acid mine drainage. Int. J. Surface Mining, Reclamation Environ. 12, 101–104. Fytas, K., Bousquet, P., Evangelou, B., 1999. Application of silicate coatings on pyrite to prevent acid mine drainage. Proc. Mining Environ. II, Sudbury, Ontario 3, 1199–1207. Gleisner, M., 2005. Quantification of Mineral Weathering Rates in Sulfidic Mine Tailings under Water-Saturated Conditions. Stockholm University. Han, P., Bartels, D.M., 1996. Temperature dependence of oxygen diffusion in H2O and D2O. J. Phys. Chem. 100, 5597–5602. Homstrom, H., Ljungberg, J., Ohlander, B., 1999. Role of carbonates in mitigation of metal release from mining waste. Evidence from humidity cell tests. Environ. Geol. 37, 267–280. Hossner, L.R., Doolittle, J.J., 2003. Iron sulfide oxidation as influenced by calcium carbonate application. J. Environ. Qual. 32, 773–780.
1634
D.M.C. Huminicki, J.D. Rimstidt / Applied Geochemistry 24 (2009) 1626–1634
Huminicki, D.M.C., 2006. Effect of coatings on mineral reaction rates in acid mine drainage, Virginia Polytechnic Institute and State University.
sulphide-rich mining waste: column experiments. Chemosphere 67, 1637– 1646. Plummer, L.N., Wigley, T.M.L., Parkhurst, D.L., 1978. The kinetics of calcite dissolution in CO2–water systems at 5° to 60 °C and 0.0 to 1.0 atm CO2. Am. J. Sci. 278, 179–216. Pollard, J.H., 1977. Numerical and Statistical Techniques. Cambridge University Press, London. Schick, M.J., 2001. Chemical properties of material surfaces. In: Hubbard, A.T. (Ed.), Surfactant Science Series. Marcel Dekker, Inc., New York. Smith, M.W., Brady, K.B.C., 1998. Alkaline addition Coal Mine Drainage Prediction and Pollution Prevention in Pennsylvania. Department of Environmental Protection, Pennsylvania. Stewart, P.S., 2003. Guest commentaries, diffusion in biofilms. J. Bacteriol. 185, 1485–1491. Stumm, W., Lee, G.F., 1961. Oxygenation of ferrous iron. Ind. Eng. Chem. 53, 143– 146. White III, W.W., Lapakko, K.A., Cox, R.L., 1999. Static-test methods most commonly used to predict acid-mine drainage: practical guidelines for use and interpretation. In: Plumlee, G.S., Logsdon, M.J. (Eds.), The Environmental Geochemistry of Mineral Deposits. Part A: Processes Techniques and Health Issues. Society of Economic Geologists, Inc., pp. 325–338. Williamson, M.A., Rimstidt, J.D., 1994. The kinetics and electrochemical ratedetermining step of aqueous pyrite oxidation. Geochim. Cosmochim. Acta 58, 5443–5454. Williamson, M.A., Kirby, C.S., Rimstidt, J.D., 2006. Iron dynamics in acid mine drainage. In: Barnhisel, R.I. (Ed.), 7th Internat. Conf. Acid Rock Drainage. American Society of Mining and Reclamation, Lexington, KY 40502, St. Louis MO, pp. 2411–2423. Yee, N., Shaw, S., Benning, L.G., Hien Nguyen, T., 2006. The rate of ferrihydrite transition to goethite via the Fe(II) pathway. Am. Mineral. 91, 92–96. Zhang, Y.L., Evangelou, V.P., 1996. Influence of iron oxide forming conditions on pyrite oxidation. Soil Sci. 161, 852–864. Zhang, Y.L., Evangelou, V.P., 1998. Formation of ferric hydroxide–silica coatings on pyrite and its oxidation behavior. Soil Sci. 163, 53–62.