Partial hydrolysis of ferric chloride salt. Structural investigation by

Partial hydrolysis of ferric chloride salt. Structural investigation by photon-correlation spectroscopy and small-angle x-ray scattering. D. Tchoubar,...
0 downloads 0 Views 494KB Size
Langmuir 1991, 7, 398-402

398

Partial Hydrolysis of Ferric Chloride Salt. Structural Investigation by Photon-Correlation Spectroscopy and Small-Angle X-ray Scattering D. Tchoubar,? J.-Y. Bottero,*J P. Quienne,+and M. Arnaudt Laboratoire de Cristallographie, UA 810 CNRS, G S "Traitement Chimique des Eaux", Universitb d'Orlbans, B P 6709,Orlbans Cbdex, France, and Equipe de Recherche sur la Coagulation-Floculation,UA 235 C N R S , GS "Traitement Chimique des Eaux", CRVM ( E N S G ) ,B P 40,54501 Vandoeuure Cedex, France Received March 13, 1990.I n Final Form: June 25, 1990 Speciation resulting from fast hydrolysis of dilute ferric chloride salt (0.1 M) by NaOH at 25 "C was studied versus the hydrolysis ratio n = NaOH/Fe from 0 to 3.0 at different aging time from t = 400 s up to 1 day by using quasi-elastic light scattering and small-angle X-ray scattering. Before the flocculation threshold (n 2.7), large colloids are formed during the hydrolysis. Photon correlation spectroscopy showed that at t = 400 s the size range varied from -10 nm ( n = 1.0) up to 700 nm (n = 2.7). At larger times the colloid size rapidly decreased during the first 30 min. The mean sizes were centered around 10 nm. Small-angleX-ray scattering experiments carried out with n = 1.0ton = 3.0 permitted description of the semilocal structure of the colloids. The scattering curves are characteristic of colloids formed by aggregates of subunits. Whatever the value of n and the time ( t I 1 h), the subunit size is equal to 16 A in diameter. It is not modified by pH variation for t I 1 h). The semilocal structure of the colloids depends on the time and n. For n = 1.0 and t = 400 s, locally the colloids are formed of linear aggregates. This local shape is originated from long-range magnetic dipolar interactions, which are known to linearize the fractal aggregates. The same sols aged for 1 h exhibited slightly more branched aggregates. From n = 2.0 up to n = 3.0 fractal structures occurred in the suspensions with fractal dimensions Df, which was varied from -1.7 ( n = 2.0 and 2.5) up to -2 ( n 3.0).

-

-

-

Introduction T h e speciation of hydrolyzed ferric chloride complexes is certainly less known than for A1 salts in spite of the amount of literature since 1960. However an important study completed by Flynn' and Schneider2 showed that the stoichiometry, size, and structure of polycations are not yet totally determined. The size of the polymers generally determined from electron m i c r o ~ c o p yvaried ~-~ with time from 2-4 nm in diameter to 7 nm." T h e existence of monomers and chloride complexes has been proven by many studies with UV spectrometry or recently by XANES and EXAFS s p e c t r o s ~ o p i e s . Studies ~~~ carried out on fully hydrated precipitates using X-ray d i f f r a ~ t i o n or ' ~ EXAFS14 showed that a structural continuity exists. The evolution of the local range order progressively tends to that of Goethite when the hydrolysis ratio n = (base)/(Fe) is increased.14 Fe is always oct Universit.6 d'Orl6ans. t

CRVM.

(1) Flynn, C. M. Chem. Reu. 1984,84, 31-41. (2) Shneider, W. Comments Inorg. Chem. 1984, 3 (4), 205-223. (3) Magini, M. J . Inorg. Nucl. Chem. 1977, 39, 409. (4) Eggleton, R. A.; Fitzpatrick, R. W. Clays Clay Miner. 1988,36 (2), 111-124.

(5) Combes, J. M.; Manceau, A.; Calas, G.; Bottero, J. Y. Geochim. Cosmochim. Acta 1989. (6) Murphy, P. J.; Posner, A. M.; Quirk, J. P. J . Colloid Interface Sci. 1976, 56, 270. (7) Murphy, P. J.;Posner, A. M.; Quirk, J. P. J. Colloid Interface Sci. 1976, 56, 284.

(8)Murphy, P. J.; Posner, A. M.; Quirk, J. P. J . Colloid Interface Sci.

1976, 56, 312.

(9) Dousma, J.; De Bruyn, P. L. J . Colloid InterfaceSci. 1978,64,154. (10) Dousma, J.; VanDenHove, T. J.; DeBruyn, P. L. J. Inorg. Nucl. Chem. 1978,40, 1089. (11) Spiro, T. G.; Allerton, S. E.; Renner, J.; Terzis, A.; Bils, R.; Saltman, P. J . Am. Chem. SOC.1966,88, 2721. (12) Sidall, T. H.; Voburgh, W. C. J. Inorg. Nucl. Chem. 1951, 4270. (13) Mulay, L. N.; Selwood, P. W. J. Inorg. Nucl. Chem. 1955, 2693. (14) Knight, R. J.; Sylva, R. N. J. Inorg. Nucl. Chem. 1975, 37, 779. (15) Ciavatta, L.; Grimaldi, M. J. Inorg. Nucl. Chem. 1975, 37, 163.

0743-7463/ 9 1/ 2407-0398$02.50/ 0

tahedrally coordinated and the condensed phases are formed by Fe sharing at both edges (olation) and corners (oxolation). These structural determinations were made with high Fe concentrations (0.5-2 M). For kinetic reasons the results cannot be extrapolated to dilute solutions, where condensation is diffusion limited and dependent on the polycation number. In this work we have investigated both the size and the structure of the colloids formed versus n = (NaOH)/(Fe) and time in partially dilute (0.1 M) hydrolyzed ferric chloride sols by dynamic light scattering (photon correlation spectroscopy or PCS) and small-angle X-ray scattering (SAXS).

Materials and Methods Materials. Solutions of crystallized FeClyGH20 (Merck ref 3904),0.2 M, were prepared by dissolving in HCL, 0.1 M. The

solutionsbefore adding NaOH solution only showedthe presence of Fe(H20)e3+or FeCl complexes as indicated by UV spectroscopy data (Figure 1). It seems (Table I) that a large part of the monomers are FeC12+ complexes. The partially hydrolyzed solutions are obtained by NaOH addition under vigorous stirring at a sufficient concentration to obtain a desired hydrolysis ratio n from 1 up to 2.7. This last value corresponds to a fast flocculation threshold. The final Fe concentration was 0.1 M or 0.05 M for PCS experiments. The kinetics of addition of NaOH was varied from 0.05 mol of OH/ mol of Fe (n = 1.0) up to 0.15 mol of OH/mol of Fe (n = 2.7). The sols were studied at various ages between t = 0 (just at the end of the NaOH addition) and t = 24 h, with most intensive study during the first hour. Methods. The turbidity of the solutions (suspensions) was followed during the first 60 min with a Hach X ratio turbidimeter, which allows absorption to be accounted for. Quasi-elastic light scattering using photon correlation spectroscopy at 90° was used to measure the translational diffusion coefficient DT of the particles. In the case of noninteracting particles, the first-order autocorrelation function g(r) is directly 0 1991 American Chemical Society

Partial Hydrolysis of Ferric Chloride Salt

7

0.4 O t \ *

Langmuir, Vol. 7, No. 2, 1991 399 Small-angle X-ray scattering experiments were carried out on 0.1 M solutions and suspensions. Synchrotron radiation of the DCI storage ring of LURE (Universit6 de Paris sud, Orsay) was used for its high intensity associated with point collimation. Very small angles were attainable, and the collected data covered the Q range from -0.77 to -2.6 A-l, where Q is the scattering vector amplitude; Q = 47r sin 8 / X with 28 the scattering angle and A the wavelength. Because fluorescence is important for X 5 1.6 A, a wavelength of 1.8 A was chosen. X-ray D a t a Analysis. The scattered intensity by aggregates, 4Q), is I(Q) = K IoCQ)G(Q)

where K is a scale constant of the experiment, Io(Q) is the scattering intensity by the subunit particle, and G(Q) is the interference function depending on the intersubunit arrangement within an aggregate. G(Q) is only valid within the Q range 2 00 300 4 00 F i g u r e 1. Ultraviolet spectrum of 0.2 M FeC133+before hydrolyzing.

Table I. Ultraviolet Spectroscopy Data for Different Simple Fe(II1) Species species wavelength (A), nm refs 12-15 Fe2(0H)z4+ 335 12-15 Fe3+, FeOH2+ 240 12-15 Fe(OH)2+ 300 FeC1416 317, 365 FeCI2+ 16, 17 220,335

1/L 5 Q 5 l/Ro

where L is the characteristic length of the aggregate and Ro the characteristic size of the subunit. If the aggregates are mass fracta116J7 G(Q)

-

Q-Df

where Df is the fractal dimension, which can be calculated from the log Z(Q) versus log Q plot within the convenient Q range. Another parameter, of structural interest, is the distance distribution function P(r). P(r)is obtained from the correlation function y ( r ) in the direct space. It is calculated from the Fourier transform of ZN(Q)'~

turbidity

30

y ( r ) = 7 r 2 / 2 S Q 2 1 N ( Q )(sin QrlQr) dQ

where ZN(Q) is the scattered intensity per aggregate divided by the invariant PO, the scattering power of the sample, given by the relation

3

2 0-

0 0 0 0 0

PO = (1/27r2)SQ2Z(Q)dQ

0 0 n

The distance distribution function P(r) is then

10-

ti me(mn)

0 0 20 B'O 60 F i g u r e 2. Turbidity evolution versus time for n = 1.0 (A), n = 2.0 ( X ) , and nn = 2.5 (0). '

'

related to the translational diffusion coefficient D T g( T) = exp(-DT q2r) The momentum transfer q = 47r sin (0/2)/A, where n is the refractive index of the medium, 8 the scattering angle, and X the vacuum wavelength of the incident beam. Polydisperse systems are analyzed by using the method of the cumulants in which a distribution of exponential terms in g(r) is assumed g(T) = F(r) exp(-rr) d r where r = D T 92. F(r),the distribution function, is expanded in a power series of T. The first moment (cumulant) in the expansion is the average which defines an effective coefficient, Deft, decay constant, ravg, for the particle or aggregate size distribution. D,ff is converted to an effective hydrodynamic radius, r H , by using the StokesEinstein relationship De, = k T/67rtrH

where 1is the viscosity of the medium, k the Boltzmann constant, and 7' the absolute temperature of the experiment.

P(r) = r2 y ( r ) The shape of P(r) indicates the organization of the aggregate and also the coordination number of the particles within the aggregate.

Results

The evolution of the t u r b i d i t y with t i m e of the samples [Fe] = 0.1 M and n = 1.0, 2.0, and 2.5 shows that the size and/or the number of particles vary with t i m e up to t 50 min. The decrease of the t u r b i d i t y is fastest for n = 2.5 (Figure 2). The evolution of the turbidity versus t i m e corresponds here to a decrease of the size of the particles as the PCS experiments show. At [Fe] = 0.05 M and n = 2.5, the size distribution is bimodal at t = 10 m i n and unimodal f r o m t = 30 m i n (Figure 3). Therefore the mean size is -10 nm at 90°. At the same concentration and various n from 1.0 up to 2.7 and t = 24 h, the average r H values increase from 12 nm ( n = 1.5) to 200 nm ( n = 2.7) (Figure 4). These results show that the size decreases with t i m e and increases with n at constant time. The scattering intensity f r o m SAXS experiments for samples with [Fe] = 0.1 M and n = 1.0,2.0,and 2.5 increases as n increases. Condensed species would be expected to increase in size and n u m b e r as n increases. The evolution of the scattering intensity with n seems to reflect the

-

(16)Asakura, K.; Nomura, M.;Kuroda, H. Bull. Chem. SOC.Jpn. 1985, 58, 1543. (17) Combes, J. M.; Manceau, A.; Calas, A.; Bottero, J. Y. Geochim. Cosmochim. Acta 1989. (la) Glatter, 0.; Kratky, 0.Small-Angle X-Ray Scattering;Academic: New York, 1982.

400 Langmuir, Vol. 7,No.2, 1991

Tchoubar et al.

-

mass distribution

/o

30 I

Figure 3. Size distribution calculated from PCS (0 = 90') versus time for n = 2.5. mass

distribution

,

2or

this solution the log Z(Q)-log Q plot shows an oscillation a t log Q -1.5. It is now well-known that when aggregates of subunit particles are observed, the region relative to the largest Q values provides information on the shape, size, interface, and polydispersity of the subunits. The part of the curve corresponding to the lowest Q values provides information on the spatial organization of the subunit particles. The curve of Figure 5a relative to n = 1.0 exhibits the same shape as that corresponding to the aggregation of monodisperse silica colloids by organic polymers,21122surfactant^,^^ aluminum hydroxide, or iron o x y h y d r ~ x i d e .The ~ ~ suspensions ~~~ n = 2.0 and 2.5 do not exhibit the same scattering curves as n = 1.0 (Figure 5b,c). The log I-log Q plots between -1.2 I: log Q I-0.6 show a linear part with a slope d log Z/log Q -1.74 corresponding to aggregates with an apparent fractal dimension of 1.74, characteristic of clustering of c1usters.s.n Compared with n = 1.0, the organization of the subunits within the aggregates is very different. Concerning the size of the subunits we observe that the different curves are superimposed from log Q I-0.7. This part of the curve would indicate that particles as small as 1.6 nm exist. In this case the subunits could be aggregates of small particles. The portion of the curve relative to the subunits in n = 1.0 suspension is theoretically limited at the lowest Q values by a horizontal asymptote whose value ZN(O) is proportional to the volume of the particle.l8 The limit has been defined as shown in Figure 6a. The part of the scattering curve as defined in Figure 6a was fitted by using the Debye equation18 for composite particles of centrosymetric subunits and was used to calculate Z(Q) of each model. The theoretical intensity was compared with the experimental one. Linear aggregates of subunits of 16 8, of diameter led to the best fit. When the fit was satisfactory the distance distribution function P(r) was calculated by prolonging Z(Q) toward the largest Q values by the scattering of the subunits whose the size was previously determined in order to prevent the cut-off effects on the Fourier transform. The experimental p(r) presents five maxima characteristic of linear-shape aggregates. Then to increase the precision of the model, P(r) could also be fitted by essay/error tests by semirigid chainlike particles formed by subunits of 16 8, diameter (Figure 6b and Table 11). The comparison between model and exprimental data shows that the aggregates are very linear even if the fit is not completely satisfactory. The colloids in n = 1.0 are constituted by the aggregation of linear aggregates of small subunits of 16 A diameter. The suspensions n = 2.0 and n = 2.5 exhibit scattering curves of self-similar structures of 16-8,subunit particles. If a superstructure exists, it is not detected because it scatters at very low Q values. Time Effect. The time effect corresponds to a size decrease of the largest particles detected by PCS experiments (Figure 3). With SAXS, at the local range the evolution of the curves is complex and depends on n =

-

-

Figure 4. Size distribution calculated from PCS (Q = 90') at t = 24 h for n = 1.0, 2.0, 2.5, 2.6, and 2.1.

iDg*

I

-3

-2

I

log Q

-1

Figure'fi. SAXS results in log I-log Q plot versus log Q for n = 1.0 (O),n = 2.0 (A), and n = 2.5 (O),t = 400 s. increasing size of the particles as shown from PCS experiments. The spectra recorded during the first 400 s (Figure 5) are very different from those obtained with aluminum hydroxide.18J9*20 These three curves could not be fitted by the characteristic laws relative to homogeneous elongated cylinders or platelets. That means that colloids are not homogeneous particles. The use of models of aggregates was necessary to interpret the data. The scattering curves present two characteristic parts especially in the case of the solution where n 1.0. For

-

(19) Axelos, M. A. V.; Tchoubar, D.; Bottero, J. Y.; Fiessinger, F. J. Phys. (Paris) 1985. (20) Axelos, M. A. V.; Tchoubar, D.; Jullien, R. J.Phys. (Paris) 1986, 47, 1843.

(21) Wong, K.; Cabanes, B.; Somasundaran, P. Colloids Surf. 1987. (22) Wong, K.; Cabanes, B.; Duplessix, R. J. Colloid Interface Sci. 1988,123, 466. (23) Wong, K.; Cabanes, B.; Duplessix, R.; Somasundaran, P. Langmuir 1989, 5, 1346. (24) Bottero, J. Y.;Tchoubar, D.; Axelos, M. A. V.; Quienne, P.; Fiessinger, F. Langmuir 1990, 6, 596. (25) Bottero, J. Y.;Tchoubar, D.; Cases, J. M. Paper presented at the International Congress of Pacific Basin Societies, Honolulu, HI, 1989. (26) Meakin, P. Computer Simulation of Growth and Aggregation Processes; Stanley, H. E., Ostrowsky, N., Eds.;M. Nijhoff Publishers: Boston, MA, 1986. (27) Jullien, R.; Bottet, R. Aggregation and fractal aggregates; World Scientific: Singapore, 1987.

Langmuir, Vol. 7, No. 2, 1991 401

Partial Hydrolysis of Ferric Chloride Salt 10s I

log1 r

,

I

-2

I

toga

-1

-3

Figure 6. (a) log I-log Q plot relative to n = 1.0, t = 400 s, and the hypothetic asymptotic limit corresponding to the subunit. (b)The correspondingexperimentalP(r) (0) and calculated P(r)

,

,

-2

-1

1oga

Figure 8. (a) log I-log Q plot relative to n = 1.0 and t = 1h and the hypothetic asymptotic limit corresponding to the subunits. and calculated (-) P(r). (b)The correspondingexperimental(0)

(-).

Table 11. Comparison between Experimental, Calculated P(r), and Theoritical Distances in a Linear Chain of Spheres of 8 A Radius A theoretical distances in a chain distances exptl calcd of spheres 16 A in diameter, 8, p(r)i

D1 D2 D3

-8 18

8 18

33

33

32

D4

48 64

48 64

48 64

D5

5 r

Log1

t

8 16

log1

-3

-2

-1

Figure 9. log I-log Q plots relative to n = 2.5 and t = 400 s (0) and t = 1 h (0).

Figure 7. Influence of the aging time, log I-log Q plot relative

means that the probability of having a first neighbor is increased; i.e. the coordination number is i n ~ r e a s e d . ' ~ J ~ In effect the best fit of I ( Q )and P(r) is obtained by slightly branched aggregates of eight subunits of 16 A diameter. The aging of the suspensions n = 2.0 or 2.5 ( t = 1h) leads to a decrease of the scattering intensity for log Q I -1.7 (Figure 9). The part for log Q 2-1.7 is not modified. This means that the local organization of the aggregates is not modified. Only size is affected as shown in Figure 3.

(NaOH)/(Fe). For n = 1.0 we observe that the scattering power is increased (Figure71, which means that the volume of matter is increased. The size of the subunits does not seem to be modified because the intensity a t the largest Q values is the same (Figure 7). The corresponding P(r) for t = 1h exhibits a different probability in the distance distribution function (Figure 8). The second maximum is increased in Figure 8b compared with Figure 6b, which

Discussion The size scales investigated with PCS and SAXS are not the same. PCS data are relative to the hydrodynamic radius of the aggregates and large particles are preferentially detected. SAXS technique detects the local structure and sometimes can measure the size of the aggregates, if not too large. For ferric polycation aggregates, a part of the structure relative to the largest distances is not detected. Two kinds of local organization are detected:

-3

I

I

-2

-1

l o ga

to n = 1.0: t = 400 s (0) and t = 1 h (0).

Tchoubar et al.

402 Langmuir, Vol. 7, No. 2, 1991

very linear aggregates, which can be partly branched with time, and fractal aggregates of 16-A subunits. The size of the subunits is determined only from the modeling of the scattering curves. The value of 16 A is realistic if we compare it with other results obtained by electron microscopy.3-7 Evidently the shape of the subunits cannot be revealed by SAXS because they are too small. But we believe that the subunits are not small akaganeite particles because the stoichiometry of the subunits is different from that of crystallized akaganeite or goethite as revealed by EXAFS,28which shows that the ratio between edge and corner sharing Fe octahedra is larger in 0.1 solution (-2) than in a- or P-FeOOH (1.6). Within the suspension n = 1.0, linear aggregates of subunits of 16 A in diameter exist. The same polycations are arranged in fractal aggregates in n = 2.0 or 2.5 suspensions. The polycations are invariant in size in a large scale of n as previously postulated29 from titration and pH-free relaxation of ferric solutions along the titration curve. Such shapes of aggregates are attributable to local interactions and diffusion laws, which detrmine the collisions and sticking rules.27 Linear shapes have been observed on different systems. When dipolar interactions exist, the aggregates are more linear.30 Recently the same effect has been observed in experiments done with magnetic (cobalt and iron) aerosol^.^^^^^ The nucleation concerns the formation of the polycation of 16 A diameter. It forms aggregates of various structure and size with n and time. When n increases, the sticking of (28) Bottero, J. Y.; et al. Manuscript in preparation. (29) Van der Woude, J. H. A.; De Bruyn, P. L. Colloids Surf. 1983,3,

55.

(30)Mors, P. M.; Botet, R.; Jullien, R. J . Phys. A: Math. Gen. 1987,

20, 975.

(31) Kim, S. G.; Brock, J. R. J. Appl. Phys. 1986,60, 509. (32) Kim, S. G.; Brock, J. J. Colloid Interface Sci. 116,431.

Log1

Figure 10. log I-log Q plot relative to precipitates formed at pH 7 at t = 400 s and 1 h.

the polycations follows the usual rules of clustering of clusters generated by the van der Waals interactions. The 3 leads to precipitates hydrolysis increasing up to n that exhibit a characteristic scattering curve with an apparent fractal dimension of 2 relative to denser aggregates (Figure 10). In n = 1.0 solution, the polycations have a high positive charge. Consequently the mechanism of aggregation could depend on the long range dipolar magnetic interactions, each subunit acting as a dipole which can align.30 In this case the aggregates are more linear. When n increases, more polycations or subunits are formed, but the positive charge at the subunit surface decreases, especially a t n 2 2.5, which is just before the threshold of flocculation of the sols and when flocculation occurs. In this case electrostatic repulsion decreases and electrostatic attractive interactions become responsible for the fractal geometry of the colloids.

-

Acknowledgment. We thank the staff of LURE (P. Vachette and J. P. Benoit) for help during the experiments. Registry No. FeCl3, 7705-08-0; iron hydroxide, 11113-66-9.