Cdlokts and Surfaces S (1982) 301-319 301

Elsevier 8cbntiffc Publishing Company, Amsterdam - Printed in The Netherlands

FLOCCULATION OF STIRRED SUSPENSIONS OF CASSKTERITE AND

mUR&UINR. .,-“. .‘.

LEONARD JORN WARRIZN .-

CSIRO civkion ~fMindtui Chemidry. ,P.& Rex 224. Fort Melbourne. Victoria 3207
(A udtmlta)
(Received 8 December 1981; accepted in final form 4 August 1982)

Retativety pure UrnpleA of the minerals casaitcrite and tourmake were separated from
a tin OPB, and were usedto prepare ~uspensfona containing partick between 2 and 8 pm
diameter. Them suspsnstona wets stirred in a baffled beaker under various cowiitions and
ihe chwges in the ltate of aggregatiun measured with t photosedimentometer. Zeta-
potentlab of th- prrtkrler we*e a!so’measurrd. It was found thkhath minerals coagulated
rpo~kneoudy In the pH range of interest (pH 4-9) but that the degree oP ag_yatIon
coufd be incraawd by the shear-ftoti:ulation effect to produe larger, stronger CIQES, The
shear-ftoccuttfon step was made sekctive hy using sodium ftuorositicate aa a selective
dfsperranttoetouranalin~.The mechanbn of the shear-Swculatton effect Ls diwussed and
compared with that of eoagulatiair under shear.

fNTRODUCTKON

One of the pekistent pkabLek in the processing of tin ores is the selective
recovery of the fitiest p&icles oP c&site&e (SnO,). We are investigating a
methqd of improving eovery in which the first step requires aggregation ‘of
the $assiteyite in ttie prbtice .of valueless accesso& mine&Is whfch are kept
dispersed. The ,cas&erite’ flctcs axe then recovetid by flotation [1,23..
Our id.m hi the Cork~iiported in this paper &IS to d&is& a systeni for selec-
tive aggregation of cassiterite from tourmaline, which is one of the key
accessory minerals. The method used was shear-ftoccuIation, which is a\1effect
whereby fine particles suspended in an aqueous solution of a surfactant me
made to aggregate by applying a shear field of sufficient magnitude [l, 31. The
surfactant needs to be one which renders the cassiterite surface hydrophobic
and for this purpose reagents nogm_@ly used as fXotat@n collectors are o-n
suiM$& W&. i&+& &d &IO $&tii’olleiAtir3i:‘&e ‘ctiboxyiic’ &id derivative
‘66 3903’S ‘and’~~~~~~~‘~ho’sphonid i&id.-Th&‘&hird’&b* field k& gei!&rat&
by stirring in a baffGd tank of appropriate geometry- The &die& Were rxuried
out on separate suspensions of the two minerals and form,a basis for further
experiment-s .with mixtures or with cassiterite ore.

6166~622/8Z;UOO_~OQOl$O2.76 8 1982 EIsMer Scfentific Pubbhbg CmnpanY

..
302

EXPERIMENTAL

Three samples of cassitcrite were prepared from a gravity concentrate from

the Clevelaud mine of ir,berfoyle Ltd. Fine grinding was carried out in a stain-
less steel rod mill, Cassiterite 1 (‘Z-- 6 pm, 83% SnOl) was separated from a
ground &my of the concentrate by repeated beaker decantaticn at pH 9.
Cassiterite II (Z+ pm, 94% Sn02), 0.6% Fe, <O,Oti% S) was prepared by up-
grading the concentrate by acid leaching and magnetic separation, before grind-
ing and size fractionation. Cassiterite III was the same purity as cassiterite I
but finer (average size 2.0 pm cf. 2.8 Hm). All samples were stored dry.
TourmaIine was prepaed by crushing a selected rock sample from the
Cleveland mine, and upgraded by repeated magnetic separation and acid leach-
ing. The crushed product, as assessed by optical microscopy, cantained 85%
tourmaline grains, 9% burmaline composite p&i&s, mainly with quutz, and
6% fluorite grains, The tourmaliue was dark green, slightly magrtetic and had
an X-ray diffraction indicating mat it was the variety dravite @&ig,A1,BS8i,0~,-
(OH,F),) . After fine grinding artd sizing the powder (Z-6 Hrn) was stored dry.

The Cyanamid tin flotation reagent S 3803 was supplied as a grey-green

paste, containing a little ethanol. Concentrations in this paper are expressed
in terms of the weight of paste, The active ingredient [41 is 4’asodium salt
of a long chain alkyl derivative of an aspartic acid”, indicating that the molecule
has at least two carboxylic acid groups. Aqueous solutions of S 3903 (1 g/l)
hecame cloudy on standing. Titration of a fresh solution with HNOl showed
one end point, charmteristk of a cachoxylic acid and corresponding to a pK
value of 6.7.
The Union Carbide reagent styrene phoaphonic acid, also a tin flotation
reagent, was suppli@ as a white powder which dissolved to give a clear, stable
aqueous solution, Styrene phosphonic acid has the followigg structure:
0
II
C6H,--CH = CH--F--OH
t
. OH

Titration of the acidic dution wi&NaOH showed two well-qzparated end

points.~ini@r $0 those of &thophospho+c’eM ,md corresponding %i pFvaL
ues’of 2-92 tid 7.42. -

Doubly-distilled water was added to the mineral powder in a 260 mL beaker

.: , - .:;-.-I - ,,,_, -_: , _: ., . . . , :
 303

and the suspension bft standing overnight to ensure rewetting of the particle
surfaces and then-dispersed with dttasound (1 min, 20 kHz, 160 W). The pH
was adjusted, reagents added, ad_ the final votume brought to 2UO ml. Tha
solids content was then 0.08 wt%.
A frame with six glass baffles (6.3 mm wide) was inserted in the beaker
(diameter 63 mm) and the suspension dispersed by ultrasound just before the
start of stirring. The bafffes were required to prevent vortex formation and
air entrainment when the suspension was stirred at high speed, A stainless steel
paddte (26 tim diameter, 16 mm high) was inserted and located about 2 mm
above the base of the beaker. Stirrer speeds of at Ieast 270 rpm were required
to keep the particles (2- 6 pm) suspended. The suspension was insulated
against fluctuations in room temperature by encbsing the beaker in foam
plastic. Drops of suspension were withdrawn periodically and the degree of
wgrel:ation observed with a binocular microscope.
At the end of the chosen stirring time, and with the stirrer stifl running, an
aliquot of suspension (30---IO0 mf) was removd in a gIass tube and transferred
to a photosedimentometer for measurement of the size distribution. If iG was
necessary to diIute the suspension to obtain a suitable initial light trmsmission,
a sotution of the same composition and temperature as that in the test was
used.
ReproducibIe results were obtained only by careful control of the procedure
for preparing the suspension, adjusting the stirrer geometry and withdrawing
the sample of suspension for size analysis.

Measurement of stze distributions

Size distributions of the initial and the stirred suspensions were determined
by following the change in turbidity with time for a sedimenting suspension
E, 6J.
The sine distribution was charactetizeci by: (1) dSO,the equivalent Stokes
diameter (esd) corresponding to 50% by weight undersize, (2) the weight per-
cent of the sampte present in the size ranges <3 pm, 3--6 &rn and >6 pm, (3)
the wt% of t.he sample which is present as f&c8 in the size ranges <3 Hrn, 3- 6
pm >6 pm. (The method of c&dating this is described Mow.) The repro-
ducibUity of~the dSOmeasurements was about + 0.2 pm.
The specific grauity (s.g.) of the particles must be known in order to calcurate
the size distribution._ For dkpersed suspensions, the s.g. value measured by the
s.g. bottle metI@ .ye used(6.35 for ceiterite II, 3.1 for tourmaline). How-
ever, in aggregated suspensions the floe densities are not known and there is no
simpIe,way OE correcting the calculations to aMow for this. It Was assumed,
therefore, that &,fI~.s had t@c! same s.g. as the.disa+e particles, an assumption
which. .wiil tend fo underestimate the actual fIoc sizes, (fn one experiment we
determined w_e_ayerage cross-sectional size of Rats from microphotographs
and found_.. that w.+,~e,most f&qqentIy ocF+ring f&c cross-section was
about 13 pm, the.correspond_ing esd of the most frequently occurring ffoc-siie
304

was about 6 pm. This result fvas not used to corrkct all the size distributions
mearx&d bjq pbottisediinetitometry because tht?‘fIoc density was observed to
vary withreagent type,stirrerspeedandst~ngtlnie.) .'

M~tiurrement of zeta-putenfb&

The particles used In the aggregation tests were too Iarge to be used In the
HamiItolr Stevens mic~electrophoresis cell [fl Wwhich required that while
particles were moving horIzontaIly under the influence of the appIied field,
they did not settle ve&xIly under gravity. 1
U!trafine cas&erite (< 1 pm) ‘was prep- by grinding cassiterite II In m
agak mortar in a cIeti cabinet, Th& powder was stored wet In a phtic beaker
for several days befor& measurements wvar&made. Uttrafine tour-fine
(< 2 pm) was obtained from the Cnes removed during beaker decantation of
the 2-6 Nrn towm&ue sample,
Solutions of the appropriate pH and reagent concentrations were prepared
and sufficient stack suspenaitin (e.g.; 1 mg &‘I) add@d to give a distinct
opalescence. If necessary the suspension was dispersed by ulksound before
measuring the electrophoretic moblities. The ionic strength was normally
O.UOfand the zeta-patentIaIs were caIcuIated from the Smofuchowaki
formula [It].
 2 3 4 S 6 7

Pig,. 1, Varfatbn with pH of the degree of aggregation (4) and the zeta-potential (0) of
cadterite 11 suspensions. ZMa-potentUs measured in 0,061 W KNO, with ~1 pm
particles- Aggrwattan measured after 60 min at 400 rpm.

Fig: 2. Variattoa with pH of’the degree of aggregation (I) and the zeta-potential (m) of
~urmdInesurpenatoM,Zeta-patRntiatsmeuuredinO.OOl~f KNO,wfth<2rmpartictes_
Aggregation measured before rtfrring.

0 20 40 60 eo IO0
0 200 400 sao 600 tmo.1200 ‘. Stirring
time (mm)

sttrrrr rpmd <rpn) I’

Fk. 3. Va$+oa of the,degree ot &ggregatkm‘of touhaline with stirrer speed after stirring
for 60 mio. i pH 5.8.0.601 M YMO,; I pH 6,s. O.Ol+ gll8 3903; A pH 4.2.0.014 d; ’
8 3903, :‘: 1.:I’ -: ..
I I. - :f : _ :.
Ff& 4. Variation of t&e degree of mmtion of t&ma&e kith atirriog time, at a speed
of.490 rpm:* pH 10; * pH 43.6; l pN 6.5.0.014 g/l 8 3603.
306

Increasing the time of stirring generally increased the degree of aggregatkm

for times up to 60 mi’n, after which a Sright decrease in aggregation occurred.
ResuIts for touxmahe am given te Fig. 4,
The dS,, vdue’givm an average measure of the!degree of aggregation of g
stirred suspension but dots not discrimiuate bet&en flats and remaining
discrete particles. To assist in interpieting our results we found it &eful to
have estimates of the proportions of ffocs and discrete particles &I selected
size ranges. Such estimatescould not be obtained without ussuminga rn~&l
for fIac growth. We mumed LhatfL&s in agiuen site fmctlijn were formed
only from dk~fepurficles in the size fnrction below. Thus, in Table 1, the
size distribution has been subdivided into the size fractions < 3,3-6 and
> 6 pm. FIocs in the size fraction 3--G~m are assumed to form from discrete
particles in the fraction < 3 pm, and flocd in the &e fraction > 6 orn from
dizcrete particles in the fmctirln 3-fi pm. (These assumptions are unIikely
to hold when the qfe&re,eof aggregation is high and large floes grow from amall
p&i&s.]
The ccmpletely dispersed tiurmaIine’(Table 1, xow 1) has no ffocs present
in any six&fraction (cohsmns 6 and 7). When stirred for 60 min at pH 5.8 in
OAIOI M KNos, some aggregation occurs (see row 2) and the wt% > 6 pm in-
creases from 3.5 to 12.7, whitih means that the wt.% of flats in the fraction
> 6 pm is 12.7 - 85 = 9.2. Simflaxly, the wt% f&s in the tictlon 3-6 pm
is 815 - (66.5 - - 9.12)= 34.2, as shown in row 2, column 6 of Tabfe 1. The
model rEquirea:that floes formed are always larger than 3 pm, esd, and any
floes present in the fraction < II pm ate igniored.

TABLE 1

V&&ion of Ihe degree of aggregation of bounmdimaauspeudmu with ather weed, after

StIrring fm60min

stirrer ‘.Vk% of samplepresent In Wt% of iampkt prefent as

speed the size fractbm: i fhcsofrIze:~
‘I
crpw --
GSpm 3-6pr. >6pm .3-6mn >6p1i1 ._ :
 307

Shear-floccuiatton
ofstirmdsq&mions
The
Turrnmline: degree of agmkation of tourmaline suspensionswas increased
by adding small amounts of the reagent S 3903 and stirring above a certain
minimum speed (- 3UO rpm, se* Fig. 3) for longer than a certain minimum
time (- 30 mix, Fig. 4). The qxed and time required depended on the pH.
The way in which the weight of ilacsof various sizes ch,wged with stirrer
speed is shown in Table 1 for taurmaline at pH 4.2 in 0.014 g/l S 3903. A cdm-
parison is made in the same table with the behaviour of the Eusgension in the
absence of S 3903 at pH 5.8.
The zeta-potential of tourmake became more negative as the concentration
of 8 3903 was increased, as showerin Fig. 6. The effect of styrene phosphor&
acid was much kss. however.

_J I I I I I I I I I !
0 0.02 0.04 0.05 0.08 0.10

CmIcenlratbn of addrd roagw# (g/l)

Pig. 6* Variation of the zeta-potential of tourmaIine with reagent concentration at pcf 6.0
and 0.001 Jf KNO, (experimental srwx -2 nlV)_ l S 3903; = stytene phwphonic acid.

Increaing the concentration of 8 3903 at firat increased the degree of aggre-

gation of stirred stTspensions, but at concentrations above about 0.01 g/l the
aggregation decre;rsed till at 1 g/i the suspension remained dispersed during
stirring. The op&*rm addition was 0.01gj1 at pH 6.L Add#oons qf styrene
phosphoi& acid,.& t6& Qthkghand, caved !ittle change.‘to ,tho degree of
flocculation at either pH 4 or’6.5.
: _ ..,-., . :
Casitenlte; The degrize of aggsegation’of stiied cassilterite suspensions was iti-
creased by addition of 0.003-0.03 g/‘/1
S 3903. Higher concenttitio’ns prevent-
ed aggregation, as shown in Fig. 6. The effects of styrene phosphonic acid were
generally small, but taxiable e.g. at pH 6, in duplicate experiments it acted as a
mild floccul&t or. 8 mild d&persaM (Fig; 6). . y -; ‘.: .. I . ,.
The.eff=t#of increasingconkmtrations of S 3963 and styx~~ phospha’tic
tiid on .the~zeta_potentialof cassitirite are shown in Fig_ 7;The degree of.
aggre&ti~n of casSi+@ fqr a given addition of 8 3903.or sty,renephosphonic
acid varied with the pH at which the susp&ston wag &red (Fig. 8).
308

-1
_.:’ .’ I_-..

.- :

. \-.- :-.
 309

te&s tequired to optimk~ the performance of each reagent; generally tests

were made at several concentrations and pH vaIues. The reagents were then
: ranked in order of .decreasing aggregating power, as shown in Tabte 2;
*__,I .. ;_ ,_ : 1 -;: .~ .: .. :. !
TABLB2. : .;., .. -; ; j . . -_ : . ;.
*_ -
Degree of aggregation of rtlrrcd casaiterite aurpenatons h the presenceof various ftotatlon
reagents
_
System Average “pacticte” size, d,, hm),
after stirring for 60 min at 400 rprn

Without reagent WiLit reagint

Cl .
PI kf9, sodium oleata (0.016 g/l) 3.3 10-8
PiIf 4, sodium Iauryi sulfate (0,016 gD) 4*4
PIII 4.8 3903 (0.02 91) 4.4 %I:

cl
PIx 4, s 3803 (0.02 gjl) 3*1 7.4
PI3 1. CA Ed0 (0.663 gll)
PI 3 4, AP 864 (0.003 g/l) ::: z-z
+I4, MIMf 3 (0.01 &l) 3:6
2 kI7, Styrene phoaphonic acid (0.01 glfl 3.9
310

Dispersing agents for cassiterite and tourmaiine

The reagentssodium fhorosilicate, sodium fluoride, soditlm metasilicate and

oxalic acid were tested for their ability to disperse stirred ausper~sions of
cassiterite and tourm&ne at pH 4 and pH 6. The resuRs are givc!n in TabIe 3.
Since it appeared that sodium fluorosilicate was a selective dkpersant for
tourmaline, further tests were made in the presence of S 3903 (Fig. 9)-

TABLE 3

The degree of aggregation of athed ~usp@nsIons

of cassiLert&~
and tourmalfrxe
in the presmce
af variousdtspeming agenLd

Reagent Carmen- cl,, (pm) after 60 min at 400 cpm

tr&on
tEW PHI pH6

Cssiterite II TounnaIine %&Write IL TOU&-thtdh~

Na,SiF, 0 442 4.6 4.6 n-d,

0*02 4.4 4.1 4.75 n-d,
0.2 4.3 4_6 n*d.
1.0 ad. YE_ n-d_ 3.3

CyaF :*m 4.2 4.6 f :Z n-d.

4.45 3.0 n.d,
6.2 4.8 n.d. 4.6 n-d.
NafiiO, 0 4.2 4.6 4.6 nA.
0.02 5.05 4.5 6-S n-d.
0.2 4*8 n-d_ 4.a md.
Oxalic acid 0 4-2 4-6 4.6 n-d.
0,uoa 3.9 n.d, 3.5 md.
0.02 a.& 3.9 n.d_ n.d.

n-d.= not determined-

Stirred suspensions of cassiterite and tourmaline were found to coagulate,

at least partLy, over a wide pH range (Figs, 1,2 and 8). Maximum coaguIation
occuxre~i between pH 5 and 63,when about 40 wt% of the cassiterite and
37 wt% of the tourmaline wzw present_* fks (as cak@lated by the r&de1 de-
scribed in Table 1). However, mation was SW significant at pH 2 &d 8
despite Ma-potentials of magnitude 30-50 mir, This ease of coa@ation has
impli+atioF Zpr, b? t h pra$ica ,*d theov. _
Why do su+sm_@, _~ighly,~~~‘l[l~~~~~.~t~-in’the~~b.Gle~t’~~w
__ . .
of a stirred tank? This pmblem is db;umed below ~ler&s 6f ‘hhk’aahke and .’
 311

magnitude of the shear regime in a stirred tank and the effects of increasing
shear rate on the stability of a suspension of charged partictes.
A practical tionsequeuce is that, in mixtures* the tendency towards hetero-
coagulation of the cassiterite and tourmaline particIes wili make clean separa-
tions of the two miner& more difficult_ Also, any aggregation due to the
shear-flocculation effect will be additional to the aggregation caused by
“natural” coagulation of the particles, and will need to he clearly distinguish-
able from such coa&ation. Ways of making this distinction ere discussed be-
low, including looking at the differing effects of increasing stirrer speed (Pig- 3)
and stirring time (Fig. 4) on the aggregation of suspensions by coagulation and
shear-flocculation.
Having established that the degree of aggregation of cassiterite (and
tourmahne) could be increased by the shear-flocculation effect to produce
ktrger, stronger flocq experiments were carried out to improve the seIf&ivity
of the aggregation, e.g., Fig. 9, and to find the optimum physical and chemical
conditions, e.g., Fig, 6, These experiments are discussed in more detail below.

Shear.reginw in Q stirred tank

The system of paddle stirrer in a baffled beaker was chosen because of its *
similarity to stirred tank reactors that might be used in practice and because
of its simplicity of operation. Unfortunately, stirred tank reactors produce
comphcated turbulent flows which are difiicult to cfrar..zterize and which vary
with the dimensions of the hnpeller, bafffes and tank, and the positioning of
the impeller and baffles within the tank [81. Camp et al-[9] have suggested
that the bregular turbulent flow condition can be characterized by the aver-
age energy dissipa&d ifi the fluid per unit volume, e, and that an average
velocity gradient, or shear ratr?,G, may be defined by
G = (Z/u)” (I)
where Y is the kinematic viscosity. For tanks which are broadly similar, but
different in detail; to the one we have used, Camp finds average shear rates
of about 1006 s-* for impeller speeds of 360 rpm. Van ‘t Riet and Smith (163
measured average shear rates in the trailing vortices behind the blades on a
turbine impeller and found a range of 256- 300 s-‘.at 800 rpm.
-We may. expecf, therefore, that the average shear rate in the baffled beaker-
paddle stirrer.system will-be of the order of 206-1000 3-l, and that the average
shear rate.wiRiricrease with increasing stirrer speed. S&ce, in stirred tanks,
VO: W (see, e.g., ref.-[8] ) it follows ~~CII-II eq. (1) that G 0: N3’= and that when
N.is increased fr?m 300 to,,1200rpm, Q will increase to ha of the order .
160(?3QCW sl$.All thesefiguresmust be regarded as, order-of-magnitude
approximationsonly. .’ . .I ’ _.
In a stirredtank the local dissipation of energy is greati& in the small zone
%i;-
nearthe’rtitatirg impeller tip and islowest in the bulk of the fiuidaway from
the impeiIer..T!lmi ind Bagster. [II] suggest that Iocal values of e range from.
312

50 times F in the 0.5% of tank volume near the impeller tip to 0.25 ? in 90%
of the tank volume. This ZOO-fold range in E implies, by eq, (11, a 1CfoId
range in G at any given stirrer speed, It may be that the reiMiveIy small voi-
ume of high shear near the impelier tip is the key factor itiIuencing the degree
of aggregation of suspensions, but we have not yet been able to confirm this.
Into this irregular turbulent flow with its wide distribution of shear rates
wo intro&ce cassiterite or lhumdine particles of averagesizeabout 3 flrn,
The range of particle sizes is such that 92 wt% of the particles have diameters
between 2 and 6 pm, and just a~ the highest and lowest shear rates may have
different effects on the degree of aggrc@ion, so too may the largest and
smallest particles in the size fraction respond differently to a given stirring
rcgi :nr .
Mcause of the complexity of the system, and because we have chosen to
measure fiocculation in krms of the equivalent Stokes’ diameter of the floes
rathlx than their cross-sectional diameter and their porosity, fundsmentaI
anai:Gs of the mechanisms of coIlEon, adhesion, floe growth and fIoc break-
up is not possibte. However, it is possible from our results to distinguish
aggregation by the shear-fioccutation effect from that due to cosgufation,
and to determine chemical conditions for the selective shear-flocculation of
cassiterite and tourmaline,

The coagulation of cassiterite and totirmsline near their isoelectric points

(pH 4-S; see Pig. 1; 2) was expectid, but the coagulation at pH*s giving rise
to relatively high zeta-potentials was not, because the energy barrier resulting
from electrostatic rep&ion is then quite high. Using the method of Wiese and
Heady 1121 the height of the DLVO energy barrier was calculated for various
particle sizes and zeta-potentials, assuming constant surface potential, a
I-Iamaket constant of 3 X 10’ lo J and an electrolyte concentration of 0.0011ci.
The resultsin Fig.10 show that for -1 pm particies with potentiaf 40 mV, the
barrier opposing close approach is 406 M?, rising to 1200 kT for 3 #m-particles
(kT = 4.1 x. 10-Z’ 3) . Su;gensionsof such particle%should be stable towards
coagulation induced by.Brownian motion.
However, it is now’welMst.ablished that suspensions which are stabie at iow
or intermediate shear rates may, nevertheless, coaguIate at higher shear ratis.
Van de Ven and Mason 1131 have shown that-the she&r rate at whicli two 4quai-
sized spheres in shear floiv form permanent touchbig doublets as a result of an
encounter depends on thz bal&e between the van d&r @J&Is attractiuie’force,
the electrostatic repuh$cn and the hy&dynari& for&Using their method,
as adapted by Takaniura[14~ i we caicuiate$ the stibiliw domain dia&aui for
particles of 1 &rn diameter in 6.061 M &xtMyte with a H&n&er &onst.ar&of
3 X 10-V d. The results; given in Pig.. IA, she* that at &6@ tit&A in ihe’ we
lQW-10,066’6’*, which wbuld covkr the en&e rang& eipe&d in’ tie stirred 1.:
beaka system at low s&r speeds,, shear tiM&uIatiijtiis unlikeI+ f& :1,&G .
 313

Fig. 10. Variation of the height of the DLVO energy bankr between spherkal charged
partlctea with the diameter of the particles_ Cdculatio!xs murrxe conatant aurfaco potential.
Curves calculated for zeta-potantiak trom 10 to 50 mV.

Pig. 11. Shear coagutation domain far 1 pm diameter particles as a function of shear rate
(Hamaker
and zeta-potenttal constant 3 x 10dl* J, etecttolyte concentration O.OOt M).
o suspenrbn rtabb; 0 coaguhtion into secondary mInimum; V caagufation Into primaw
minimum.

particles with zeta-potentials above 30 mV. For 3 pm particks, shear coaguta-

tion will be even less likely, according to the theory developal by Van de Ven
and Mason [13], which predicts that increasing parti& size increases the
region of stability.
These predictions are consistent with our obse=&ions on &rred suspensions
of 1 pm scheetite particfes with zeta-potentials of 30-40 mV [lJ, which re-
mained stable at all stirrer speeds and had zeta-potential? of 30. .40 mV, How-
ever, the predictions are not consistent with the observed behaviour of 3 pm
cassiterite particles which coagulated extensively at low stirrer speeds when
the zeta potential was 30--40 mV, and for which the degree of coagulation
decreased with increasing stirrer speed.
The calculations leading to Figs. 10 and 11 were made assuming that the
Hamaker constants for both scheeiite and cassiterite in water were the same.
This was shown td be a reasonable assumption by calculating the Hamaker
constanta froni first principles following the method of Hough and White [15].
As shown in Table 4, the Hamaker constant for cassiteri+water cassitirite
is 4.6 X -lO”* J and that fdr scheeli*water--scheelite is 4.2 X lO”O J.
It is pmbablymot valid to appiy Van de Ven and Mason’s theory [13] to the
growth of multi-particle Rots &on%hreguIar particIes of different sizes in the
turbulent flow of a stir&i tank reactor, The theory might be apolicable if,
: for instance, collisions occurred primarily within eddies in which a laminar
314

TABLE 4

Hamaker constants eaiculated by the method OP Ha& and White [LSI

Syatem Hamaker conatant Sources of data

(IO-” d) (referencee)

Quartz-water--quartz 1.68 I6
C~iterite-water-caiter~te 4.54 16-18
Cassiterite-water-cassiterite 4.63 16.18,19
Scheelitewater-scheelite 4.15 16.19.20
Scheetite-wate~cheelite 4.29 16,ZO. 21

flow regime prevailed, but it is not known whether this is a major collision
mechanism. There is one other factor which may make the Van de Ven and
Mason theory inappropriate; namely, t4at the size and specific gravity of the
paticles aru both sufficiently targe to give the patticcles a substantial momentum,
large enough that inertial effects on the collision process cannot be ignured.
For cassiterite particles with a v&city of 6 cm s-l in the stirred tank, the
kinetic energy and the energy of interaction due to stirface forces are of similar
magnitude for 1 pm particles but for 3 pm particles the kinetic energy is an
order of magnitude larger, Thfs suggests that collisions which lead to attach-
ment through “inertial impaction" [22] will: be more likely for 3 than 1 pm
particles, which is consistent with our results. If inertial effects were important,
it might explain why cassiterite aggregates formed over a wider pH range and
again& hig4er electrostatic potentials than did tourmaline (40 mV cf. 30 mV)_
The difference might then be attributed ta the increased inertia of cassiterite
particles arising from their higher s-g. (6.35 cf. 3.1).

The shear-@eculatian effect

Shear-fIoccuIation may he distinguished from coayfation and polymeric

flocculation [I, 231 by the minimum ehear rate required ti initiate aggrega-
tion, by the strength of the floes wxd by the need for tie particle surfaces to
be rendered hydrophobk by the adsorption of a suitable surfactant. The
appropriate shear rate is determined by the size and the zeta-potential of the
surfachntcrlated particles [l J .
The following characteristics of the flocculation behaviour indicate that
rmspensions of 3 pm tourmaline 02 cassiterite in 0.01 g/l S 3903. stirred at
speeds at or above 400 rpm and at pH values between 4.2 and 6.6 were aggre-
gated by shear-flocculation:
(1) The degree of flocculation achieved when S-3903 was present, was much
greater than when it was &sent. An analysis of the data in Pigs, 3 and 4 in
tonns 0P the madei assume&for flog growth showed that the increasein dSo
arose mainly from the ticrewed proportion of large flocS (> 6 am) which were
formed by the shear-flacculation effect. Thus, atpH 6.6 the wt% sample present
 315

as ffocs > 6 pm increased from 5 to 26, and at pH 4.2 from 9 to 68, when
8 3903 was present,
(2) The resuits also show that Blocs formA by shear-flocculation aze more
resistant to the high shear rates generated at high stirrer speeds, Thus, whereas
the optimum speed for coagulation of tourmatine was 273 rpm, when 43 wt%
coagutated, the optimum speed for shear-flocculation was 800 rpm when 97
wt% IIoccuIated (Fig. 3, Table 1). The strength of the floes, especially those of
escl3-6 pm at 1200 rpm is evident in Table 1.
(3) As was observed previously during work on the shear-flocculation of
uItra.Cne schee1ite [l] , the ftocs are stow to grow, especialty if the reagent-coat-
ed particles are highly charge& For example, for tourmaline at pH 6.5 in
O.OL4 g/l S 3903, the weight of large floes (> 6 pm esd) increased almost ird
proportion to the time oP stirring, with growth still occurring at 90 min where-
as there was virtually no further growth of Iarge coagula (> 6 pm) after 60
min (Fig. 4).
(4) Perhaps the key feature of shear-flocculation is that increased aggregation
with increasing stirrer speed only occurs if the particles are fil?rtm&e hydco-
phobic by the sorption of a suitable reagent which, in this case, was S 3903.
This is best seen by comparing suspensions of highly charged particles. Tour-
maline done 8t pH 10 (I = -+-37 mV) did not aggrE_gati at 400 rpm (Fig. 4)
whereas in the presence of 0.014 g/i S 3903 at pII 6.5 (g = -59 mV) about
half the sample was flocculated (Fig. 3).
The de of the hydrophabic surfaces appears to be twofold: (1) the forma-
tion of agmegates is favoured by an energy of ‘*hydrophobic assaciation” which
comes into effect if the collisions rcsuft in direct contact between the hydro-
phobic particles, and (2) the resistance to thinning and removal of the water
layers separating the approaching particles is tess for hydrophobic particles
than for hydrophilic particles 11).
Pashtey and Israelachvili [24] have shown by direct measurements of the
forces between two mica surfmes in aqueous solutions c;f hexadecyltrimethyl-
ammonium bromide (<=TAB) that when the mica surfaces were coated by a
monolayer of CTAB and were hydrophobic, there was an apparent additional
attractive force between the surfaces which was at least three times iarger than
the expected Van der WaaIs fmce. The resuIt of this effect was that as that mica
surfaces approached each other, they jumped into final adhesive contact from
a separation of 9 n.m instead of from the separation of 4 nm expected from
DLVO theory. In shear-flocculation we are evidently caking advantage of this
extra attractive ‘.*fo~e” between hydrophobic surfaces, which allows closer
approach ,qf surfaces with fairly high zeta-potentials.
The strength of hydrophobzc interactions has been measured by Shchukin
et al. [ZS] . In their experiments, the force necessary to break contact between
two 1 mm diameter methylat& glass spheres was measured and related t4~ the
areal free energy of adhesion. For methylat& glass spheres in water the rfd-
hesion energy was about 40 mJ m’*, which for a contact area of IO.00 nm*
co&responds to an energy of about 10,600 kT.This is of the same rrder of
316

magnitude as the energies we calculated for the hydrophobic interaction be-

tween oleate-coated q&.-&al 1 pm scheelite particles [l] and is further con-
firmation of the strength of the adhesion between hydrophobic particles in
water, and the consequcnl; strength of the floes formed by shear-flocculation.

Opfimfzaftin of 8electiue aggregutibn

r’ype of floccukant: Other reagents, in addition to S 3993, were found to LX-

crease the degree of aggzgation of stirred cassiterite suspensions, As shorun in
Table 2, sodium cr!eate, S 3903 and sodium Iauryl sulfate gave extensive shear-
flocculation. The sulfosuccinamates CA 540 (Allied Coltoids Yty, Ltd.) and
AP $45 (Cyanamid) wem less effective and produced more open fioc structures.
The reagent MlMI 3 (Mitr!!bishi), which is mainly p_toIyl arsenic acid, and
styrene phosphonic acid (Union Carbide) acted as weak disperaants or weak
flaccufant~ or else had no effect on the degree of aggregation.
The reagents 8 3903 and styrene phosphor& acid were studied in detail
because they ate prewntly favoured as collectors in the flotation of cassiterite
ores. Our results show that S 3903 was also a flocculant in stirred suspefisions
whereas styrene phoqhonic acid was not.
it is not clear why styrene phosphonic acid can apprrrentIymake cassiterite
particles hydrophobic enough tc stick to air bubbles but not to each other. At
least two factors may be involved:
(1) The short hydrocarbon chain may be too short to give a siylificant
energy of hydrophobic assaciation when the chains on colliding pariMes over-
fap. However, the work of Shchukin et al. 1261 suggests that even a single
layer of methyl groups on the surface can give a Iarge adhesion energy.
(2) The affinity of phosphonic scids for the cassiterite surface may be too
weak to give sufficient surface coverage and/or a sufficiently strung bond be-
tween the reagent and the surface. Wottgen [Z6l; for example, found that
uptake of n-heptyl phosphonic acid corresponded only to about l!lO mono-
layer at pH 6. In’addition, our zeta-potential measurements (Figs, 6,7)
suggested that uptake of the sin&y charged stytene phoaphonic acid anions at
pH 6 was small because there was little increase in the magnitude of the
negative potential. Adsorption of fractions of a monolayer may be enough
for flotation, but may not be enough for shear-flocculation.

Concentratiun of flocculanf: Maximum aggregation of cassiterite with S 3903,

sodium oleate or sodium lauryl sulfate was obtain&d at floc&l~t cdncentra-
tions of about 0.01 g/l (Pig. 6). At higher concentmtions the &tier&@ Rot size
decreased and the proportion of disc&c particles increaz&l till at concentm-
tions above 0.1 g/l there was no aggregation on stirring.
This behatiiour is consistent with the idea that+ ati the reagent &oncentration
is irtcread past a c&t& v&e,-a s&cotid layer of suifactant is adsorbed with
an orientation opposite to that of thti first I&ye%SiIaJrersOf thi13 type h&ve been
obsemed in several btlier ‘systed [Z4; 271. Itistkad’cif the hydrocarbon taiIs of
 317

the surfactant molecules facing the sotution and making the surface hydro-
phobic, the polar head groups are oriented outwards and make the surface
hydrophilic. Shear-fXoccuIation will be less likely since the hydrocarbon fds
on opposite particles can no ionger easily overlap and the energy of hydro-
phobic association will be small.

Concenfmtbn ofdSspersurt: For the same particte size (3.3 ,WII) and S 3903
concentration (0.01 g/l) the degree of aggregation at pH 6 was greater for
cassiterite than for tourmaline (Fig. 3,6). At pH 4 both minerals were equally
well aggregated. However, the slight selectivity at pH 6 is unlikely to be suffi-
cient to achieve selective flocculation in practice.
Sodium fhtorosilicate proved to be an effective dupersant for stirred tour-
maline suspensions but had no effect on the degree of aggregation of cassiterite
suspensions [Table 3) At concentrations below 0.1 g/l, sodium fluorosilicate
h extensively hydrolyzed, forming sihcic acid and hydrogen and fluoride ions
[281. Sines sodium silicate, at pH 4 or 6, did not disperse tourmaline whereas
sodium ffuorkie did, we concluded that it was the Ctuoride ion which was
selectively adsorbing on tourmaline, hut not an cassiterite, This is consistent
with the observation that the aeta-potential of tourmaline increased from
-15 to -30 mV in the presence of 0.Q4 g/l NaF at pH 6.6, The aluminium
and/or boron sites in taurmaline may be responsible for the specific adsorption
of fluoride,
The selective action of sodium ftuorosilicate was maintained in stirred
suspensions to which sufficient S 3903 had been added to promote good shear-
flocculation. This may be deduced from Fig. 9 and from a test on cassiterite at
pH 6.5 with 0.002 g!l S 3903 snd 0.6 g/l NaxSiFa, for which the dxO was 6-l
pm. Thus cassiterite could be aggregated at the same concentration of fluoro-
silicate which dispersed taurmaline.
Our best results were obtained by adding just enough S 3903 (-0.01 g/l)
to aggregate the cassiterite (Fig. 6) and just enough sodium fluOrOSJ!icate
(-0,3 i’J1) to disperse the tourmaline (Fig, 9). Selectivity was better at pH 6
than 4 (Figs. 3,6). An intermediate stirrer speed (400-800 rpm) and a fairly
long stirring time (30-60 min) were required for goad flocculation (Figs. 3,
4). On the basis of these results for single-mineral suspensions, selective shear-
flocculation in mixtures should be possible.

CONCLUSIONS

(1) Stirred bxspensions of 2--S pm cassiterite and 2-6 pm tourmaline were

coaguIat& over a wide range qf p& Aggregation was not restricted to the
immediate vi&& of thair &oelectric points (pH 4.3 and g.O’resgectively),
but occurred even where the zeta-potentials were about &35 mV. Comparison
of this behaviour with that of 1 pm scheelite particles suggests that inertial
force* @eed to b,e confeidcred in addition to surface and hydrodynamic forces
in ord,e,r‘ti ‘explain the coaguIatien process.
310

(2) The degree of aggregatiiinin stirred suipensiong of th& two minerals was
increased in the presence of sodium oleate, sodium laurjr1sulfate iandthe. ; I .: I
Cyanamid xeagentS 3903;This is caused bJi_th& she&fio&ufation effect; thich
produced h~er;stronger Rots: At high reagent doncentrations there was no :
aggregation, apparently because of the adsorption of a second, oppositely orient-
ed and therefore hydrophilic layer of surfactant.
~(3) .The reagent styrene phosphunic acid; which is reported to be a’good
flotation collector for cassiterite, did not induce sh&r-floctiulation of cassiterite
or tourmaline, probably because of its short hydrocarbon chain and weak
adscrption. i-.. -. - :. . I
(4) Sodium fluorcsiricate was found to absorb selectively on tuurmaline. The
active ingredie-atin the ffuorosilicate appears to be the fluoride ion. Tests on
separate suspensionsindicat& that selective shear-fhxculation of cassiterite
from tourmaline should, other things being equal,-be possible In mixtures of
the two mtierals and in natural cassitetite ore,

The work was supported in part by Abetfoyle Ltd. The author thanks
Mr D.J. Strihley and Mr A.H.D.- Couchman for carrying out most of the ex-
petiental work; Dr T.G.H. Van de Vaii and Dr.-K. Takamura for helpful
discussions and assistance in cabdating shear coagulation domains and Dr
L.R. white for asPistancein calculating Hamaker constants.

REFERENCES
 319

la H.J. vnn De& J. Apple Phys.. 39 (1968) 4467.

19 A.N. Win&e11 and H. Winehell. “Elemenb of Optical Mfnerrlugy.” 4th Edn, Part II,
Whey. New York, lrtS1. pp. 6s. 148.
20 W.8. Browet and P.H. Fang. J. Appl. Phys.. 38 (1967) 2391.
21 A, Chretien and W. Frimudltcfa, Tug&me, in: P. Pascal (Ed.) “Nouveau Traiti de
Chi&ie Miner&,*’ T&me XIV, M-n, Pm& 1969, p. 846.
22 V.G. Lewkh, ‘*Phyrhochemfclt HydradynamIcs,” Prentke-Hall, EntzteoroodCliffs,
NJ., 1962. p- 221.
23 L.J. Warren. Chem. Auk, 44 (1977) 315.
24 R.M. Pa&ley md J.N. braelachvilf, Coltoids Surfaces. 2 (1281) lS9.
25 PD. Shchukinn,I?.A. Amelitm end V.V. Yaminaky. Colloids Surfaces, 2 (1981) 221.
26 E. Wottgen, Tbns. Inat. Min. M~tall. (Sect. C: Mineral Proceor. Extr. Metall.). 76
(1969) cm-
27 P.D. Bisim,J.cf. Carttedge, WM, Keemm and C.J. Radke, J, ColfoW Interface Ski., 78
(1980) 226.
28 L-J_ Wprrsp and J.A. Kltchener, Trans. Inst. Min, Metall, (Sect. C: &herd Process-
Bwr. MctJl.). 81 (lS72) c137.

..‘,