Вы находитесь на странице: 1из 14

European Journal of Soil Science, March 1997, 48, 101-1 14

Effects of crystallinity of goethite: 11. Rates of sorption and desorption of phosphate


R . S T R A U S S a * , G . W . B R U M M E R a & N.J. B A R R O W b
aDepartment o f Soil Science, Universiry of Bonn, NuJallee 13, 53115 Bonn: and bCSIRO Division of Land and Water, Wembley, Western Australia 6014, Australia

Summary
Eight samples of goethite ranging in surface area from 18 to 132 m2 g-' were mixed with phosphate at a range of pH values for periods which ranged from 0.5 h to 6 weeks. The sample with a surface area of 18 m2 g- had been hydrothermally treated to improve its crystallinity. Its rate of reaction with phosphate depended on pH but was complete within a day. Its maximum observed reaction was close to the theoretical maximum for surface adsorption of 2.5 pmol m-2. For the other samples, phosphate continued to react for up to 3 weeks and exceeded the value of 2.5 pmole mP2. The duration and extent of the reaction depended on the crystallinity of the goethite. The results were closely described by a model in which the phosphate ions were initially adsorbed on to charged external surfaces. The phosphate ions then diffused into the particles. This was closely described using equations for diffusion into a cylinder. Samples of goethite which had been loaded with phosphate dissolved more slowly in HC1, and had a longer lag phase, than phosphate-free goethite. For the hydrothermally treated goethite, HC 1 removed much of the phosphate when only a small proportion of the iron had been dissolved. For a poorly crystallized goethite, it was necessary to dissolve much more of the iron to obtain a similar removal of phosphate. Brief treatment with NaOH removed most of the phosphate from the hydrothermally treated goethite but only half the phosphate from a poorly crystallized goethite. These results are consistent with the idea that phosphate ions were not only bound on external surface sites but had also penetrated into meso- and micro-pores between the domains of the goethite crystals and were then adsorbed on internal surface sites. This penetration tied the domains together more firmly thus increasing the lag phase for dissolution. Differences between sites for phosphate adsorption are therefore caused mainly by their location on either external or internal sites. Models that ignore this are incomplete.

'

Introduction
It has been known for some time that the effectiveness of phosphate fertilizers decreases with time even when there is no removal of phosphate in produce. This decline is largely caused by a slow reaction between phosphate and soil (Barrow, 1980; Parfitt et al., 1989; Mendoza, 1992). As there is certainly a fast initial adsorption reaction the magnitude of which is affected by such factors as ionic strength and pH, it follows that there must be at least two distinct reactions involved when phosphate reacts with soil. Iron oxides are important in the reaction of phosphate with soil. However, in many studies with iron oxides it has been

*Present address: Fichtner. Postfach 101454, 70013 Stuttgart, Germany. Correspondence: G. W. Brilmmer. Email: bobo@uni-born.de Received 3 July 1995; revised version accepted 13 September 1996

assumed that reaction is confined to an initial adsorption reaction. Values for the maximum phosphate adsorption have been calculated on that basis. Schwertmann (1988) calculated that the maximum surface phosphate adsorption on goethite is 2.51 pmol m-'. Many of the published values for goethite, as summarized by Goldberg & Sposito (1984), are close to this value. Yet, in many cases, the maximum adsorption may have been underestimated because the pH was too high. Several authors have used pH 3.5 (Atkinson, 1969; Golden, 1978; Ainsworth & Summer, 1985), while Torrent et al. (1990) estimated maximum sorption at pH 6. In the present work we show that adsorption is greater at pH 2 than at pH 3.5 and much greater than at pH 6 and suggest that previously published values underestimate maximum sorption. We further crysta11ized goethite that the show that it was Only for a maximum observed sorption was close to a value of 2.5 p o l m-2. For less well crystallized samples it was greater.
~~

0 1997 Blackwell Science Ltd.

101

102 R. Strauss et al. Despite the emphasis on a presumed surface adsorption, there is appreciable evidence that reaction with iron oxides is not always rapid. For example, Torrent et al. (1990) showed that sorption after 75 days was greater than after one day for a range of goethites of differing crystallinity. Analogous studies of the reaction between goethite and several heavy metals have also shown a fast initial reaction followed by a continuing reaction (Gerth & Briimmer, 1983; Briimmer et al. 1988; Gerth et al., 1993; Bibak et al. 1995). The explanations for this continuing reaction listed by Torrent (1991) include diffusion through crystal defects (Barrow, 1983, 1987), diffusion through surface micro- or meso-pores (Madrid & de Arambarri, 1985), and migration to surface sites within aggregated particles (Willett et al., 1988). Analogously, Torrent et al. (1992) suggested that phosphate diffused to surface sites in micro-pores. There are some similarities amongst these suggestions, for if a defect is large enough it can be considered to be a pore and the surface of such pores inside the crystals would contain surface sites. Meso- and micropores between goethite domains have been emphasized as pathways for the diffusion process by Briimmer et al. (1988), Gerth et al. (1 993), Bibak et al. (1995) and Fischer ef al. (1996). In the present work we studied the reaction of the goethites with phosphate for up to six weeks. We think that our observation of little continuing reaction with a sample of wellcrystallized goethite supports explanations based on slow diffusion into pores. We also think that they do not favour the other explanations listed by Torrent (1991): precipitation of iron phosphate on the surface (Martin et al., 1988), diffusion through a coating of iron phosphate (van Riemsdijk et al., 1984, Van der See & van Riemsdijk,l991), and burial of phosphate by progressive coagulation (Anderson et al., 1985). We provide two further lines of evidence for our conclusion that phosphate has diffused into pores. One is based on our ability to describe closely the observations using a model in which diffusion into pores is assumed. The other is based on more direct tests. Two approaches were used. In one, samples of goethite which had reacted with phosphate were dissolved slowly in hydrochloric acid. We measured and compared the rates of release of phosphate and iron. In the other approach sodium hydroxide was used to desorb phosphate from the surface, but the period of reaction was short so that reverse diffusion of any previously penetrated phosphate was small. Both approaches were used to estimate the amount of phosphate penetrated into inner surfaces of goethite crystals. which contained aluminium are designated as, for example, Al-Goe-44. The crystallinity was reflected by the surface area with the best crystallized samples having the smallest specific surface area. The sample Goe-18 was especially well crystallized as it had been hydrothermally treated. Effects of time and amount of phosphate application on sorption Adsorption measurements were made using 20 mg of goethite in the presence of 10 ml of 0.01 M NaN03 and at pH between 2 and 11. There were two series of treatments. In one, both 10 and 20 mg of P 1 - I were used as initial concentrations, and the samples were shaken for 0.5 h, 1 h, 2 h, 8 h, 1 d, 3 d, 1 w, 3 w, and 6 w. In this series, the possible surface loading of Goe-18 at 10 mg of P is the same as that for Al-Goe-36 at 20 mg of P1-. In the other series, reaction was for 1 h and a range of initial concentrations were used. The range depended on the adsorption by the particular goethite samples (see later). For both series there were 12 differing pH treatments for each combination of time and P concentration. The tubes were then shaken in an end-over-end shaker at 20 cycles per minute at 20C. After centrifuging, an aliquot of 1 to 5 ml was taken for phosphate determination. The pH was measured in the remaining solution. The amount of phosphate sorbed was calculated from the difference between the initial and the measured concentration. For the series for which time was varied, sorption at given times and at given values of pH was interpolated using curves fitted visually to plots of sorption against pH. Interpolation was necessary both to enable presentation of the results through time and to decrease the computations required in modelling them. Visual fitting of curves was preferred because of the complicated shapes of the sorption curves. For the series for which reaction time was 1 h, the original data are presented and were used for modelling. Phosphate desorption and goethite dissolution in 5 M HCI The four samples of goethite used for the studies on dissolution in acid were: Goe-23, Goe-46, Goe-60 and Goe-132. Preliminary experiments showed that a concentration of 5 M HC1 provided a good compromise between too rapid and too slow dissolution. In order to load the goethites with phosphate samples of 2 g of goethite were mixed with 250 ml of 0.01 M NaN03 containing phosphate at a pH close to 4. The initial phosphate concentrations were proportional to the surface area of the four goethite samples. After mixing, the suspensions were incubated at 40C with vigorous shaking once a day for 7 d. A sample of the suspension was then analysed for pH and phosphate in solution. The measured values for phosphate sorption in pmol m-2 were: Goe-23, 2.58; Goe-46, 2.62; Goe-60, 2.81; Goe-132, 2.56.
101-1 14

Methods
Eight samples of goethite were investigated. Details of their preparation and of their crystal properties were given by Strauss et al. (1997). The samples are identified using their surface area. Thus Goe-18 indicates a sample with a surface area of about 18 m2 g-. The three samples of goethite

0 1997 Blackwell Science Ltd, European Journal of Soil Science, 48,

Sorption and desorption o f phosphate on goethite

103

After the adsorption stage the addition of HCI followed without prior separation of the equilibrium solution by centrifuging. This was to prevent the dispersed goethite particles coagulating. Reproducibility of the results decreased when coagulation occurred. An aliquot of the suspension was transferred to a centrifuge tube, HC1 was added, the centrifuge tubes were closed and shaken end-over-end at 20C for various periods and were then centrifuged. The supernatant solutions were analysed for phosphate and iron. All investigations were in duplicate. The data for goethite dissolution were compared with those reported for unphosphated goethite by Strauss et al. ( 1997).
Phosphate desorption by sodium hydroxide

pH and phosphate concentration were measured. A volume of 10 rnl of 0.1 M NaOH was added, and the oxide was resuspended. The centrifuge tubes were shaken for 30 minutes at 20C, centrifuged, and phosphate was determined in the supernatant solution. The phosphate which was calculated as being sorbed includes that which remained in solution in the centrifuge tubes after decanting the original solution. The maximum error this would cause was estimated to be only 2%.

ReSUltS
Effects of time and amount o f phosphate application on sorption

For the studies on phosphate desorption at high pH Goe-18 was also included. Samples of 20 mg of goethite were suspended in 10 m1 of 0.1 M NaN03 solution at pH values between 2 and 1 1. The initial phosphate concentrations were 10 and 20 mg 1-'. Reaction was for one week at 40C. The tubes were centrifuged, the supernatant was removed, and
Goe-18

For the well-crystallized samples, Goe- 18 and Goe-23, reaction with phosphate stopped within a day. For the other samples reaction continued for much longer (Fig. 1 and 2). The larger the specific surface area, the longer the reaction continued but, in all cases, reaction had reached an end-point after three weeks (504'h). Continuation to 6 weeks produced no further change. Modelling of the reaction from three weeks
Goe - 60

0
1
10

0
100
100

10

100

100
80

80 60
40

20

20
0
1
10 Timelh 100
1

10 Time/h

100

Fig. 1 Time course of phosphate reaction with Goe-18 and Goe-60. The mass of goethite used was 2.01 g 1 - ' and the initial concentration of phosphate was 20 mg 1 - I . The upper figures express sorption as a percentage of the amount of phosphate added; the lower figures express it as a percentage of the final adsorption for each given pH value. The lines in the upper figures were obtained by fitting the model to the data of this figure only; the lines in the lower figures are polynominals fitted to these data

0 1997 Blackwell Science Ltd, European Journal of Soil Science, 48,

101-114

104

R. Strauss et al.
r
2
OT

I
Goe -23 Initial P
2

il
P
n
U

E -

10

n "

10

OT

AlGoe - 30 Initial P 0 2

R t

AIGoe-36 Initial P 0 2

E 2

fi
n

51'
0
2 4

10

" 2

10

AIGoe-44 Initial P
OT

B
51
n
2 4

E -

.. .

0 2

10

10

Goe - 60 Initial P
2

G08-132

Initial P
a 5 0 7 v 15
0

-:I

25

10

10

PH

PH

Fig. 2 Effect of the indicated amounts of addition of phosphate (mg P 1 -') on phosphate sorption after 1 h for the eight samples of goethite. In all cases, the mass of goethite used was 2.01 g 1-'. The lines indicate the fit of the model using the parameters shown in Table I. The model was simultaneously fitted to this data and to the data in Fig. 3 and 4.

0 1997 Blackwell Science Ltd, European Jounral of Soil Science, 48,

101-114

Sorption and desorption o f phosphate on goethite

105

to six weeks would have required many more time steps. It was therefore not done, and the data for 6 weeks are not presented. Reaction was initially slower at high pH. This is shown in Fig. Ic and Id by expressing sorption as a percentage of the final value. At low pH reaction with Goe-18 was almost complete after 2 h; at high pH it was still incomplete at the 8 h sampling. However, reaction was apparently complete by the next sampling at one day. We therefore conclude that only the fast, initial, adsorption reaction was present for this sample. The effect of pH on kinetics shows the difficulty of separating the postulated fast and slow reactions (when both are present) by attempting to choose particular sampling times. If one were to choose, say, 1 h, the fast reaction would be incomplete at high pH. Yet, if one chose longer periods, the slow reaction may have become increasingly important-depending on the crystallinity of the goethite. Figure 2 shows that, after 1 h reaction, there was a steady decrease in sorption with increasing pH up to about pH 6. The decrease was much steeper at pH values higher than about 7. P a r t of the decrease at high pH occurs because, after 1 h, the initial reaction was still incomplete at high pH. Maximum sorption was not achieved at pH 6 and a concentration of 6 mg P 1-' (194 p ~ (Figs ) 2 and 3). These were the conditions chosen by Torrent et al. (1990) who, however, measured sorption after one day. Sorption increased with both decreasing pH and increasing phosphate concentration. Table 1 presents values for sorption interpolated from curves fitted to sorption data at given pH values. On the average, values for pH 6 and 6 mg P 1 are about two thirds of those at pH 2 and 30 mg P 1-'. Figures 2 and 3 both show that sorption at pH 2 was greater than at pH 3. This supports our contention that measurements made at pH 6 or even at pH 3.5 underestimate the maximum sorption. Short-term sorption curves such as those in Fig. 3 approach a smaller maximum sorption as the pH increases. This is caused by a decrease in the electrical potential of the reacting sites. The maximum amount of long-term sorption was similarly affected by pH. This is shown by the trend of the curves in Fig. 4 to approach an end-point which decreases with increasing pH. That these curves have a similar shape is shown by plotting sorption as a percentage of the final value for Goe-60 (Fig. Id). After the 8 h sampling, points fell close to a common line. This result shows that the region into which the phosphate diffused, and the surface on which it was initially adsorbed, were affected by pH in a similar way. If this were not the case, and diffusion was into an uncharged region, it would have continued for much longer at high pH.

2.5

DH

'

-V

V 8

200

400

600

800

P concentration/pM

Fig. 3 The sorption curves (lines) at constant pH derived from the fitted model for Goe-18 together with interpolated values (points) for sorption at the specified pH values.

the curves fitted to describe the rate of reaction. The function used was:
Ct = C , - Crexp(-(kt)a),
t, C ,

(1)

-'

Phosphate desorption and goethite dissolution in 5 M HCI

Figure 5 shows that the release o f phosphate was much faster than the release of iron f r o m the goethites. Furthermore, the shapes of the release curves were different: the curves for iron were sigmoid; those for phosphate were not. This is shown by

where C, is the amount of phosphate or iron dissolved at time indicates the maximum amount dissolved, C,-Cr indicates the amount dissolved at zero time, and k is a measure of rate. The greater the value of k, the faster the dissolution. The value of a reflects the degree to which the curve is sigmoid. A value greater than unity indicates a sigmoid curve; a value of unity indicates an exponential curve; and a value less than unity indicates a curve that rises mpre steeply than an exponential curve (Barrow & Mendoza, 1990). Table 2 shows that, for iron dissolution in the presence of phosphate, the values of a were greater than unity, whereas for phosphate dissolution they were smaller than unity. The larger values for the k coefficients for phosphate than for iron, and also the curves in Fig. 5 , show that the release of phosphate was faster than the release of iron. Similarly, the rate of release of both iron and phosphate tended to increase as the surface area of the goethites increased. In all cases the release of iron from the phosphated goethites was appreciably slower than from unphosphated goethites (Fig. 5). Not only were the k terms smaller for the phosphated goethites but the a terms were also larger (Table 2). The differences in the a terms show that the curves for dissolution of the phosphated goethites were more sigmoid-that is, there was a longer lag phase. In Fig. 6 the release of phosphate after increasing periods is plotted against the release of iron from the four different goethites. For Goe-23 most of the phosphate was released when only small amounts of iron had been dissolved; when

0 1997 Blackwell Science Ltd, European Journal of

Soil Science, 48, 101-1 14

106 R. Struuss et al.


15

20
15

ci
c

k i 5 n
0
n l
I

AlGOe - 30

30

40

- AlGoe - 36
-

8
20
c

30

0 Q

20
I -

.
1 10 100
I

10

+
0
1 10

10

+I

0 -

100

40

20

0
1
10
100

10

100

80 t

Goe - 60

100
80

8
0
v)

60

60
40 40

20

.
I

20

0'

0'
1 10 Tirneh

10 Tirneh

100

100

Fig. 4 Time course of phosphate sorption for all eight samples of goethite. In all cases the mass of goethite used was 2.01 g 1-', and the initial concentration of phosphate was 20 mg 1 - I . The open circles indicate pH 2 and the other symbols indicate pH 3 to 11 in sequence down the graph. The lines indicate the fit of the model using the parameters shown in Table 1. The model was simultaneously fitted to these data and to the data in Figs 2 and 3.

0 1997 Blackwell Science Ltd, European Journal of Soil Seience, 48,

101-1 14

Sorption and desorption of phosphate on goethite


Table 1 Measured phosphate sorption and parameters for the model fitted the describe phosphate sorption by eight samples of goethite
Goe-18
Measured properties P sorbed 2a Ipmol m- P sorbed 6b Ipmol rn- P sorbed 3wc Ipmol m- Excess over 2.5d Ipmol m- Fitted parameters Max. internal sorption /pmol m-2 Capacitance GSA IF m-2 Binding constant K, / 1 pmol m-I Rate constant k l / h 1 pmol- Diffusion term (D/,a2)Ih- R2

107

Goe-23 2.86 2.00 2.94 0.44 0.62 2.99 52.1 25.9 0.0856 0.997

Goe-46 2.89 1.66 3.31 0.8 1 2.73 2.43 12.6 4.03 0.0 103 0.995

Goe-60 3.12 2.09 3.96 1.46 2.95 3.33 7.18 2.91 0.0101 0.994

Goe-132 3.23 2.30 n.d. n.d. 5.70 3.19 19.05 7.59 0.0017 0.995

Al-Goe-30 2.96 1.70 3.28 0.78 1.72 2.80 2.58 4.60 0.047 0.992

AI-Goe-36 A1-Goe-44 2.86 1.80 3.59 1.09 2.5 1 2.83 11.54 4.68 0.0122 0.995 2.99 2.07 3.32 0.82 1.48 2.84 42.03 20.10 0.0293 0.997

2.30 1.42 2.40


-0.1

0.16 2.89 5.01 3.38 0.0 138 0.993

aPhosphate sorbed at pH 2 at a concentration of 30 mg P I- after 1 h. bPhosphate sorbed at pH 6 at a concentration of 6 mg P I- after 1 h CPhosphate sorbed at pH 2 with an initial concentration o f 20 mg P I - and 2.01 g of goethite I- after 3 weeks dValue in previous line minus 2.5 pmol m-.

Table 2 Values of the C and a parameters and for R2 for the equationa used to describe
the rate of dissolution of four samples of goethite with and without sorbed P in 5 M HC 1 Pb Goe-23 Goe-46 Goe-60 Goe-132

R2

Fe(+P)

Fe(-P)

Fe(+P)

Fe(-P)

P
0.999 0.997 0.998 0.997

Fe(+P) Fe(-P)

0.280 0.354 0.293 3.073

0.0202 0.0223 0.470 0.0322 0.0573 0.359 0.0666 0.1072 0.492 0.703 1.360 2.065

1.81 1.42 1.59 1.14

1.15 1.31 1.82 0.98

0.904 0.998 0.998 0.999

0.996 0.998 0.998 0.997

aThe equation used was: C, = C , - C,exp[ -(kry), where C, indicates the amount of P or Fe dissolved and f is the time in hours. bP indicates phosphate released from phosphated goethites, Fe (+P) indicates iron released from phosphated goethites, Fe( -P) indicates iron released from unphosphated goethites.

only 10% of the iron had been dissolved, about 90% of the phosphate had already been released. At a given value for iron dissolved, the phosphate released decreased in the sequence Goe-23 > Goe-46 z Goe-60 z Goe-132. For Goe-132, when 10%of the iron had been dissolved, only 45% of the phosphate had been released. Thus, the proportion of the phosphate bound within the crystal increased with increasing specific surface area. This in turn indicates decreasing crystallinity and increasing porosity (Strauss et al., 1997).
Phosphate desorption by sodium hydroxide

little affected by either the pH of the initial sorption reaction or by the level of addition of phosphate (Fig. 8). For some samples, there was a slight downward trend with pH, for others a slight upward trend.
Development of a model to describe the observations

Figure 7 shows that for Goe- 18 most of the phosphate initially sorbed was desorbed after exposure for 30 min to 0.1 M NaOH. With increasing surface area of the goethites the proportion that was desorbed decreased. For Goe-132 only about half of the previously sorbed phosphate was desorbed. When the pH at which the sorption occurred was varied both the phosphate initially sorbed, and the phosphate subsequently desorbed, decreased with increasing pH (Fig. 7). The proportion of the sorbed phosphate which was desorbed was

In order to model sorption of phosphate it is necessary to augment the model used by Strauss et al. (1997) to describe charge. In that model the mean charge on the adsorbed protons is allocated to a plane ( S plane) close to the surface arid the charge on the counter ions to a plane slightly further from the surface. When phosphate is present its adsorption is assumed to be into a plane (A plane) between the protons and the counter ions (Barrow, 1987). The position of this plane is effectively determined by the capacitance between the two planes, GSA. As the concept is that this is the mean position of the charge on the phosphate ion, the effect is similar to that recently obtained by Hiemstra & Van Riemsdijk (1996) who distributed the charge on the adsorbed phosphate between the other two planes.

0 1997 Blackwell Science Ltd, European Jountal

of Soil Science, 48, 101-1 14

108 R. Struuss et al.


100

80

60

40

20 Goe - 46
0

0
0

f
P

20

40

60

80

100

120

20

40

60

80

100

120

100

P e
LL
9,

80

b
60

40

20
Goe - 60

0
0

"
20 40 60 Time/h

Fig. 5 Effect of time on the dissolution of iron and of phosphate by 5 M HCI from four samples of goethite pre-treated with phosphate (0 phosphate, 0 iron from samples pre-treated with phosphate, c ]iron from samples not pre-treated with phosphate). The lines are drawn from fitted curves and some of the parameters are given in Table 2.

It is proposed that the initial adsorption reaction is relatively rapid and that this is followed by a diffusive penetration towards the center of the particle. Although the initial reaction is relatively rapid, it is nevertheless not instantaneous and it is necessary to include equations to describe its rate.

7
0 Goe-23 0 Goe-46 v Goe-60 v Goe-132
20

The results show that the rate of the initial reaction was slower at high pH. This indicates that its rate was controlled by the rate of reaction at the charged surface rather than by the rate of transfer through a liquid film. It has been suggested by Barrow et ul. (1981) that Equation (9.24) of Bockris & Reddy (1970) is appropriate:
Rate ofrhe initial reaction.

dO _- klapyc,m, exp( %FYa/RT)


100

60 iron dissolved/%

40

80

dt
- k20, exp(-dFY',/RT),

(2)

Fig. 6 The release of phosphate relative to the release of iron (both as a percentage of the total present) when four samples of phosphated goethite were reacted with 5 M HCI for a range of times.

where 0, is the concentration of Occupied Sites at time t and is therefore a measure of the amount of sorption (pmol 1 -'), k, and k2 are the rate constants (h l2 pmol-? and h 1 pmol-'),

0 1997 Blackwell Science Ltd, European Journal of Soil Science, 48,

101-114

Sorption and desorption of phosphate on goethite


I

109

50

Goe - 23
40

25 20

30
15

20
10

5
n

10

"2
100 I

6
PH

10

6
PH

10

100

80
60

80

b
40

60

40

20

b
0
-20 2

20

0 4
6
DH

10

6
PH

10

100

80

0
0

Observed sorption Observed desorption

60

Upper line is modelled total sorption Lower line is modelled external sorption

..a@
40

a a a

20

8
PH

10

Fig. 7 Effects of the pH of initial sorption of phosphate from a solution containing 10 mg P 1 and on the subsequent desorption of phosphate in a solution of NaOH after 1 h. The lines were fitted to the data in Figs 2, 3 and 4. The upper lines show the modelled values for total phosphate sorption; the lower lines show the modelled values for the amount of phosphate adsorbed on the external surface and therefore more likely to be rapidly adsorbed.

0 1997 Blackwell Science Ltd, Eumpean Journal of Soil Science, 48,

101- 1 14

110 R. Strauss et al.


60 1
1
I

5 0 1 1
a
's

a
.m

$?

30

a 20

z s !

10

decreased with increasing pH suggests that the reacting surface adopted an electric potential determined by the external pH. The observation that the rate of the second reaction was much slower than the initial reaction suggests that the rate was limited by a different mechanism-probably by the rate of diffusion to the reacting sites rather than the rate of reaction with the sites. The observation that reaction was completed within the experimental period suggests that phosphate had penetrated to the end of the diffusion pathways of the goethite particles. Models for which penetration was small relative to the radius of the particles were therefore inappropriate. We argued from the results of Strauss et al. (1997) and Fischer e f ~ l(1996) . that the goethite particles could be approximated to cylinders. The problem is then one of diffusion into cylinders from a stirred solution of limited volume, and Equation 5.33 of Crank (1964) is appropriate:

10

PH

where M,, the amount of the solute in the cylinder after time
t , is expressed as a fraction of the amount at equilibrium,

Fig. 8 The measured values for the proportion of the sorbed phosphate which was retained after brief treatment with NaOH. Hollow symbols indicate an initial concentration for phosphate sorption of 10 mg P I - ' , filled symbols 20 mg P 1 - ' .

up is the fraction of the total phosphate dissociated to HPOi-, c, is the total concentration of phosphate in solution (pmol 1-I), m, is the concentration of vacant sites at time t (pmol I-'), Ya is the electrical potential in the plane of adsorption at time f (mV), and F is the Faraday, R the gas constant and T (K) the temperature. The terms % and d are called transfer coefficients and are described by Bockris & Reddy (1970) in their Equation (9.25). The magnitude of these coefficients is determined by the position of the ratedetermining step in the sequence of steps involved in the overall reaction (Barrow et al., 1981). The observed behaviour of decreasing rate of reaction with increasing pH is reproduced when % is 2 and d is zero. These values were used here. They indicate that the rate-determining step precedes the electron transfer steps and does not itself involve an electron transfer. The adsorption at equilibrium is given by:

M , , j ? is the final fractional uptake by the cylinder; a is the radius of the cylinder; the qns are the positive, nonzero roots of Equation (5.34), the values of which are tabulated in Table 5.1 of Crank (1964); and D is the diffusion coefficient. It is convenient to regard the ratio D/a2 as a single parameter.
f the model. The results had suggested that Formulation o diffusion of phosphate occurred into a region carrying an electric potential which was influenced by the pH of the solution. It was therefore necessary to allocate an electric potential to this surface. The resuits of Fig. 8 were important in constraining the model in this regard. They could only be reproduced if it was assumed that the internal phosphate experienced the same changes in potential with changes in pH as the external phosphate. For the external surface the parameters estimated were: the , the rate constant for the forward reaction binding constant Ki k , , and the capacitance GSA.The maximum adsorption was taken as 2.5 pmol m - 2 . For the internal surface the parameters estimated were: the maximum adsorption and the diffusion The binding constant far the internal surface was term D/d. set equal to that for the external surface. The program was written in BASIC. Details of the program and copies of it are available from N.J. Barrow at E-mail jimb@ccmar.csiro.au. The data sets are shown in Figs 2 and 4. For each sample of goethite there were typically 160 observations through time (10 pH values by 8 times by 2 levels of P) plus 84 observations at 1 h (12 pH values by seven levels of added P). There were 200 iterations allowed in the simplex procedure used to optimize the choice of parameters, and the
Science Ltd, European Journal o f Soil Science, 48, 101-114

O=

+ N KiapyC, exp(2YaF/RT) '

N KiupyC, exp(2YaF/RT)

(3)

where N is the maximum adsorption, C, is the total concentration at equilibrium (in pmol I-') and the binding constant Ki is equal to the ratio of kllk2. This is equivalent to the adsorption equation of Bowden et al. (1977).
Rate of the continuing reaction. The observation that the amount of phosphate reacting via the continuing reaction

0 1997 Blackwell

Sorption and desorption o f phosphate on goethite

111

procedure was restarted twice at the previous best values, giving 600 iterations in total.
Output from the model

loo

/ -

The model described the observations very closely (Figs 2, 3 and 4)and the average value for R2 was 0.995 (Table 1). When the internal surface was excluded from the model, the changes in the value of R2 for Goe-18 were on fourth decimal place only. This shows that internal penetration was unimportant for this sample. As the maximum external adsorption N was set at 2.5 pmol m-' for all goethites, this meant that, for Goe-18, the effects of pH, initial concentration of P, and time were all described using only three adjustable parameters. The importance of internal penetration increased with increasing surface area. Thus for Goe-23, excluding internal penetration from the model significantly (P i 0.05) decreased the value of RZ. That is, there was a significant though small effect even for this well-formed sample (Fig. 4 ) .This figure shows that the slow reaction continued for much longer for the goethites with larger surface area. For these goethites, the model was also very efficient in describing the observations as only five adjustable parameters were needed to closely describe the effects of different pH, initial concentration of P, and time. The maximum internal sorption was estimated in the model in units of pmol of phosphate. In order to compare the values with those for external sorption it is expressed relative to the BET surface area. The regression estimates of the internal sorption per unit of surface area increased as the surface area increased (Table 1). These values should be interpreted with caution because they depend on the values allocated to the other parameters and especially the value of the binding constant. The values allocated by the model (V,) exceed the observed excess sorption over the calculated surface maximum of 2.5 pmol rn-' (V,) (Table l), but were correlated with them (V, = 0.25 1.97 V , , r = 0.90). A small part of the larger value for V,,, arises because the observed maxima (V,) were calculated from a particular amount added, rather than from a specified final concentration. Especially for the samples with a large surface area, slightly larger values might be obtained with a greater amount added. However, the main reason is the assumption in the model that penetration of phosphate increases the negative surface charge and therefore decreases the sorption on the external surface at a given pH. Hence, the observed excess sorption over 2.5 pmol m-' underestimates the internal surface. Two factors influence the other model parameters in Table 1. One is that values for the diffusion term for Goe-18 and Goe-23 are imprecise because the continuing reaction was small. The second is that some of the parameters are correlated. This means that changing values for one can be partly compensated by change in another. Thus, because of the way the terms are defined, large values for the binding constant K i are associated with large values of the rate constant k l :

10 Timelh

100

Fig. 9 Effects o f time on the modelled split of phosphate between the extemal and internal surfaces. The symbols show the observed values of total sorbed phosphate for an initial concentration of 20 mg P I - ' at pH 8; the solid lines show the fit of the model using the parameters of Table 2; the broken lines show the modelled extemal phosphate. The gap between the solid and the broken lines therefore represents the internal phosphate.

kl = 0.15 0.473 Ki, r = 0.905. Similarly, increases in the rate term k l can be partly compensated by changes in the diffusion term. Because of these effects, individual values of the parameters should be treated with caution. Figure 7 shows that the phosphate modelled as being sorbed on external surfaces was smaller than that measured as desorbed after 30 min. This is to be expected because some of the phosphate desorbed within this period would come from internal surfaces. Nevertheless, the model reproduced the effects of the pH on the amounts of phosphate desorbed by sodium hydroxide. Figure 9 shows the modelled partition of phosphate between the internal and external surfaces at a common level of addition of phosphate for three samples of goethite. For Goe-18 the modelled total sorption and the modelled external sorption were virtually the same. For Goe-132 modelled total sorption increased with time, but modelled external sorption decreased partly because increasing internal sorption decreased the solution concentration of phosphate and partly because it increased the negative charge as described above. At the end of the experimental period modelled external phosphate was less than half of modelled total sorption. These trends were similar but less marked for Goe-60.

Discussion
It was only for the well-crystallized Goe- 18 that no continuing reaction could be detected after 24 hours shaking and it was only for this sample that the maximum observed sorption was

0 1997 Blackwell Science Ltd, European Journal of Soil Science, 48, 101-1 14

112 R. Struuss et al. slightly lower than the theoretical maximum surface adsorption of 2.5 p o l m-. Furthermore, most of its sorbed phosphate was removed by 0.5 h treatment with sodium hydroxide or by brief treatment with hydrochloric acid. We think this shows that it was only for this sample that the phosphate reacted solely with the surface. All of the less wellcrystallized samples differed in all of these respects: reaction continued for up to 3 weeks; the maximum exceeded 2.5 pmol m-*; and recovery by brief treatment with acid or alkali was incomplete. All of these observations are consistent with the conclusion that there was a continuing reaction with imperfections or meso- and micro-pores between crystal domains (Briimmer et al., 1988; Torrent et al., 1992; Gerth et al. 1993; Fischer et al., 1996). In comparing this conclusion with those of previous reports questions of semantics arise. If an imperfection or a defect can be invaded by a foreign molecule it is reasonable to refer to it as a pore. Once it is so invaded reaction can be considered to be with sites on an internal surface (Briimmer et al., 1988; Gerth et al., 1993). We therefore think that our conclusion is compatible with the ideas put forward by Barrow (1983, 1987) and Madrid & de Arambarri (1985) as indicated earlier. Further, it is compatible with the idea of Anderson et al. (1985) in so far as they suggested that phosphate was buried. However, they associated this burial with observed changes in the coagulation behaviour. We think that changes in the coagulation behaviour were caused by changes in the surface charge as a result of reaction with phosphate and were not the cause of the continuing reaction. On the other hand, we think that our results are incompatible with ideas of precipitation. Again semantics arise, but precipitation suggests that some threshold exists below which no continuing reaction occurs and above which continuing reaction is marked. This is not consistent with our results. Nor are our results compatible with formation of a sequence of layers of iron phosphate which progress inwards (van Riemsdijk et al., 1984, van der Zee & van Riemsdijk, 1991) for we would expect this process to also occur with Goe- 18. These conclusions are supported by the fact that the results were closely described by a model that involved an initial adsorption step on to the surface of the oxide followed by diffusive penetration into the particle. We are not aware of any other successful description of the long-term rate of reaction of anions with oxides. The reaction of the cations nickel, zinc and cadmium with goethite were closely modelled by assuming that the initial adsorption reaction was followed by diffisive penetration of the surface (Briimmer et al. 1988; Barrow et al., 1989). However, Yiacoumi & Tien (1995a, b) could obtain only a poor description of the rate of reaction of cadmium with aluminium oxide despite the presentation of some 90 equations. Our purpose in developing the model was to test the idea of diffusive penetration in a quantitive manner and for this purpose, the ability to describe the data is more important than the values of the parameters. This purpose differs from that of Hiemstra & Van Riemsdijk (1996) who wanted to test whether they could describe phosphate adsorption from detailed knowledge of the bonds formed and of the distribution of charge. For this purpose, the values of the parameters are more important. They distribute the charge on the adsorbed phosphate ion between the plane in which protons are thought to adsorb (S plane) and a plane in which counter ions adsorb (C plane). The effect is very similar to that of our approach in which the charge on the phosphate ions is allocated to a mean position between these two planes by the value of the capacitance term (Table 1). To illustrate this consider Goe-18. For this sample, the fitted value for the capacitance between the S plane and the plane of phosphate adsorption (A plane) was 2.89 F m- (GsA, Table 1). The value for the capacitance between the S plane and the counter ions was 1.75 F m- (Gsc from Table 1 of Strauss et al., 1997). If we assume that the dielectric properties are uniform in this region, the relative position of the planes is proportional to the reciprocal of the capacitances. The A plane is therefore about 0.6 of the distance from the S plane to the C plane. Hiemstra & Van Riemsdijk (1996) obtained a similar description of the effects of pH, phosphate concentration and of background electrolyte as had been obtained using our approach. However, they assumed that the only reaction is with the surface. They note that large values for sorption can occur with some samples of goethite but they associate this with large positive surface charge. We suggest that our results show that phosphate can react with more than the surface layers and that this needs to be incorporated in more fundamental models of the reaction and that large values for charge are caused by analogous diffusion of protons. The results also show that the initial reaction was slower at high pH. The remarkably slow reaction at pH 10 was indeed compatible with the faster reaction-at lower pH provided that the initial rate of reaction was limited by the rate of reaction of the phosphate ions with differently charged surfaces. Indeed, for Goe-18, this was sufficient to explain the entire time course of the reaction. It is possible that this result was obtained because the reactants were mixed fairly vigorously. If mixing were not as vigorous it is possible that diffusion through a surface layer would then become limiting and that the rate would not depend on the charge on the surface. The sigmoid dissolution curves arise because dissolution by HC1 is a two-stage process (Strauss et al., 1997 and references therein). Dissolution is initially slow until attack by the acid causes the goethite domains to separate, thus exposing a bigger surface to further reaction (Cornell et al., 1976; Schwertmann, 1984; Ruan & Gilkes, 1995). We think that the more-markedly sigmoid curves for the phosphated goethites indicate that separation of the domains was slower and that this is direct evidence that phosphate had indeed penetrated between the domains. This is consistent with the observations of Biber et al. (1994) that addition of anions such 88 phosphate to the solution

0 1997 Blackwell Science Ltd, European Journal o f SON Science, 4 8 , 101- 1 14

Sorption and desorption o f phosphate on goethite

113

phase slowed the dissolution of iron oxides. Similarly, Willett & Cunningham (1983) found that phosphate stabilized the surface of ferrihydrite at a range of pH and Eh values. Previous investigations of desorption of phosphate with sodium hydroxide have used a longer period of desorption. McLaughlin et al., (1977) extracted phosphate from ferrihydrite using two extractions with 0.1 M NaOH over a total period of 18 h. They found that when the phosphate reacted with the ferrihydrite for 30 d prior to desorption, only 88% of the phosphate was desorbed. In contrast, Willett et a l . (1988) used the same procedure and desorbed all of the phosphate that had reacted with ferrihydrite for 90 d. However, femhydrite forms very small crystals and could have only short pores. Consequently results with ferrihydrite may not be transferable to goethite. The influence of crystallinity on the proportion of the phosphate which could not be desorbed from goethite and lepidocrocite (y-FeOOH) was investigated by Cabrera et al. (1981). They desorbed phosphate that had reacted with the oxides for six days using sodium hydroxide with six treatments over a total period of 17 h. They found a significant proportion of the phosphate remained within the crystal. They attributed this to diffusion of the phosphate into micropores. Because of the slow diffusion, the desorbability of the phosphate was very slight. Torrent et a l . (1990) extracted phosphate that had reacted with samples of goethites for 75 d by shaking for 16 h with sodium hydroxide. Up to 0.66 pmol m P 2 of phosphate remained unextracted. We have shown that amounts of phosphate extracted by 0.5 h treatment with sodium hydroxide were larger than our modelled values for surface phosphate. We suggest that longer extraction with sodium hydroxide would extract an even greater proportion of the phosphate that has penetrated into the particles. In summary, this work has shown that differences between sites for phosphate adsorption are mainly caused by their location on either external or internal sites. Models that ignore this are incomplete.

References
Ainsworth, C.C. & Sumner, M.E. 1985. Effect of aluminum substitution in goethite on phosphorus adsorption: 11. Rate of adsorption. Soil Science Society of America Journal, 49, 1149-1 153. Anderson, M.A., Tejedor-Tejedor, M.I. & Stanforth, R.R. 1985. Influence of aggregation on the uptake kinetics of phosphate by goethite. Environmental Science and Technology, 19, 632-637. Atkinson, R.J. 1969. Crystal Morphology and Suflace Reactivity of Goethite. Ph.D. Thesis, University of Western Australia. Barrow, N.J. 1980. Differences amongst a wide-ranging collection of soils in the rate of reaction with phosphate. Australian Journal of Soil Research, 18, 215-224. Barrow, N.J. 1983. A mechanistic model for describing the sorption and desorption of phosphate by soil. Journal of Soil Science, 34, 733-750. Barrow, N.J. 1987. Reactions with Variable Charge Soils. Martinus Nijhoff, Dordrecht.

Barrow, N.J. & Mendoza, R.E. 1990. Equations for describing sigmoid yield responses and their application to some phosphate responses by lupins and by subterranean clover. Fertilizer Research, 22, 18 1 - 188. Barrow, N.J., Gerth, J. & Briimmer, G.W. 1989. Reaction kinetics of the adsorption and desorption of nickel, zinc and cadmium by f goethite. 11. Modelling the extent and rate of reaction. Journal o Soil Science, 40, 437-450. Barrow, N.J., Madrid, L & Posner, A.M. 1981. A partial model for the rate of adsorption and desorption of phosphate by goethite. Journal of Soil Science, 32, 399-407. Bibak, A., Gerth, J. & Borggaard, O.K. 1995. Retention of cobalt by pure and foreign-element associated goethites. Clays and Clay Minerals, 43, 141-149. Biber, M.V., Dos Santos Afonso, M. & Stumm, W. 1994. The coordination chemistry of weathering: IV. Inhibition of the dissolution of oxide minerals. Goechimica et Cosmochimica Acta, 58, 1999-2010. Bockris, J.OM. & Reddy, A.K.N. 1970. Modem Electrochemistry: An Introduction to an Interdisciplinary Area. Plenum Press, New York. Bowden, J.W., Posner, A.M. & Quirk, J.P. 1977. Ionic adsorption on variable charge mineral surfaces. Theoretical-charge development and titration curves. Ausfralian Journal of Soil Research, 15, 121 - 136. Briimmer, G.W., Gerth, J. & Tiller, K.G. 1988. Reaction kinetics of the adsorption and desorption of nickel, zinc and cadmium by goethite. I. Adsorption and diffusion of metals. Journal of Soil Science, 39, 37-52. Cabrera, F., de Arambarri, P., Madrid, L. & Toca, C.G. 1981. Desorption of phosphate from iron oxides in relation to equilibrium pH and porosity. Geoderma, 26, 203-216. Cornell, R.M., Posner, A.M. & Quirk, J.P. 1976. The kinetics and mechanisms of the acid dissolution of goethite cc-FeOOH. Journal of Inorganic Nuclear Chemistry, 38, 563-567. Crank, J. 1964. The Mathematics o f Difusion. Oxford University Press, London. Fischer, L., Zur Miihlen, E., Briimmer, G.W. & Niehus, H. 1996 Atomic force microscopy (AFM) investigations of the surface topography of a multidomain porous goethite. European Journal of Soil Science, 47, 329-334. Gerth, J. & Briimmer, G.W. 1983. Adsorption und Festlegung von Nickel, Zink und Cadmium durch Goethit (cc-FeOOH). Fresenius Zeitschrifi fur Analytische Chemie, 316, 616-620. Gerth, J., Briimmer, G.W. & Tiller, K.G. 1993. Retention of Ni, Zn and Cd by Si-associated goethite. Zeitschrqt fdr Pflanzenerndhrung und Bodenkunde, 156, 123- 129. Goldberg, S. & Sposito, G. 1984. A chemical model of phosphate f adsorption by soils: 11. Noncalcareous soils. Soil Science Society o America Journal, 48, 779-783. Golden, D.C. 1978. Physical and Chemical Properties of AluminiumSubstituted Goethite. Ph.D. Thesis, North Carolina State University, Raleigh. N.C. Hiemstra, T. & Van Riemsdijk, W.H. 1996. A surface structural approach to ion adsorption: the charge distribution (CD) model. Journal o f Colloid and Inteflace Science, 179, 488-508. Madrid, L. & de Arambarri, P. 1985. Adsorption of phosphate by f Soil Science, two iron oxides in relation to their porosity. Journal o 36, 523-530.

0 1997 Blackwell Science Ltd, European Journal ofsoil Science, 48,

101-1 14

114

R. Strauss et al.
Torrent, J. 1991. Activation energy of the slow reaction between phosphate and goethites of different morphology. Australian f Soil Research, 29, 69-74. Journal o Torrent, J., Schwertmann, U. & Barr6n, V. 1992. Fast and slow phosphate sorption by goethite-rich natural materials. Clays and Clay Minerals, 40, 14-21. Van der Zee, S.E.A.T.M. & Van Riemsdijk, W.H. 1990. Model for the reaction kinetics of phosphate with oxides and soil. In: Interactions at the Soil Colloid-Soil Solution Interface, (eds G.H. Bolt et al.), pp. 205-239. Kluwer Academic Publishers, Dordrecht. Van Riemsdijk, W.H., Boumans, L.J.M. & De Haan, F.A.M. 1984. Phosphate sorption by soils: I. A model for phosphate reaction with metal-oxides in soil. Soil Science Society of America Journal, 48,537-541. Willett, LR. & Cunningham, R.B. 1983. Influence of sorbed phosphate on the stability of ferric hydrous oxide under controlled f Soil Research, 21, eH and pH conditions. Australian Journal o 301 -308. Willett, I.R., Chartres, C.J. & Nguyen, T.T. 1988. Migration of phosphate into aggregated particles of ferrihydrate. Journal of Soil Science, 39, 275-282. Yiacoumi, S. & Tien, C. 1995a. Modelling adsorption of metal ions from aqueous solution. I. Reaction-controlled cases. Journal o f Colloid and Interface Science, 175, 333-346. Yiacoumi, S. & Tien, C. 1995b. Modeling adsorption of metal ions from aqueous solution. 11. Transport-controlled cases. Journal of Colloid and Interface Science, 175, 347-357.

Martin, R.R., Smart, R.St.C., & Tazaki, K. 1988. Direct observation of phosphate precipitation in the goethite/phosphate system. Soil f America Journal, 52, 1492- 1500. Science Society o McLaughlin, J.R., Ryden, J.C. & Syers, J.K. 1977. Development and evaluation of a kinetic model to describe phosphate sorption by hydrous ferric oxide gel. Geodermu, 18, 295-307. Mendoza, R.E. 1992. Phosphorus effectiveness in fertilized soils evaluated by chemical solutions and residual value for wheat growth. Fertilizer Research, 32, 185- 194. Parfitt, R.L., Hume, L.J. & Sparling. G.P. 1989. Loss of availability of phosphate in New Zealand soils. Journal of Soil Science, 40, 371-382. Ruan, H.D. & Gilkes, R.J. 1995. Acid dissolution of synthetic aluminous goethite before and after transformation to hematite by heating. Clay Minerals, 30, 55-56. Schwertmann, U. 1984. The influence of aluminium on iron oxides: IX. Dissolution of Al-goethites in 6M HCI. Clay Minerals, 19, 9- 19. Schwertmann, U. 1988. Some properties of soil and synthetic iron oxides. In: Iron in Soils and Clay Minerals. (eds J.W. Stucki, B.A. Goodman and U. Schwertmann), pp. 203-250. North Atlantic Treaty Organisation, Advanced Studies Institute Series C, Vol. 217, D. Reidel, Dordrecht. Strauss, R. Brilmmer, G.W. & Barrow, N.J. 1997. Effects of crystallinity of goethite: I. Preparation and properties of goethites of differing crystallinity. European Journal o f Soil Science, 48, 87-99. Torrent, J., Barr6n, V. & Schwertmann, U. 1990. Phosphate adsorption and desorption by goethites differing in crystal morphf America Journal. 54, 1007- 1012. ology. Soil Science Society o

0 1997 Blackwell Science Ltd, European Journal of Soil Science, 48, 101-114

Вам также может понравиться