CA1188107A - Removal of alkali metals and alkaline earth metals from molten aluminium - Google Patents

Abstract

A B S T R A C T

A method of removing dissolved contaminant alkali. metals and alkaline earth metals from molten aluminium by reaction of the contaminants with solid particulate materials comprising aluminium fluoride, wherein a body of the molten aluminium to be decontaminated is stirred, in the presence of the aluminium fluoride particles, in such manner as to generate and maintain a stable vortex in the molten body.

Description

--1~
"R~loval o~ ~lkali ~etals and Alka].ine Earth ~Jet~ls from ~Jo l t en AlumiTIium'l This invention relates to the rernoval of contamltlant quantities of alkali metal.s and S alkaline eart~ metal~ frorll molten aluminium by reaction with aluminium ~luoride~
Molten aluminium withdrawn ~rom electro lytic reduc~ion cells Gontains slnall amo~lnt~ of alkali metals such as lithlum ~nd sod:iu,~l and al~alitle earth liletals such as magnesium and calci.um~ The preselloe of: these contarniliant alkall metals and alkaline earth meta].s is ~lelet:erlous for v~ricus use~
to wlllch the primary metal may be put~
For example, i.n the produc~ion o~ magnesium-lS conta.ining alumlnium alloy sheet or platep sodium isamounts o~ approxi.lnately 2 p.p.m. or mo~e can call~e "hGt shortness" or edge cracking during hot rolling~
The presence oE trace quantities of lithiu~G and/o~
sodi~n increases the rate o cxi~atior of ~olten ~0 aluminiulll. This inc~eases the melt loss and generates rl thlck dross layer which cr~n ~lock _asting mrlchine nozzles and diminl.sh metal 1uidity.
T~iere~ore.9 economic and teohnical considerat.lons requi.re th~t these elements be remove~ as soon as possible after withdrawal of pri~ary aluminium from the recluction cells to re~uce ~he time period during which lithium~ and~or sodi.um~containing n~olten alumin~um is ~posed to thc ztmosphere~ Magne3Lum i~.
smal:L quant,l'cie s ls de~rlment~l. to e~le~trical ~D~

eonductivity and should be removed rom primary aluminium ~hich is to be used or products in whi.ch this property is importallt.
Accordingly, it has been recognized as desirable to reduce the concentration of alkali metal and alkali.ne earth metal contaminants to

2 p,p,m. or preferably even less. Such removal is also desirable in other circumstances where aluminium or an aluminiumobasecl alloy (the ter~l "alumlnium" being used h~rein broadly to embrac.e these ai.loys as well as pure aluminium metal) is contaminated with mino~ amounts of alkali met~l ar.ld/
or alkali~e earth metal.
It 1~ k~own that the content o dissolved ~lkali metal.~ andJor alkaline earth metals ~an be reduced ~y bringing molten aluminium into contact with ~lumi.nium ~luoride (ALF3~ (or a ma~erial cont~ining it). The contaminant reacts with aluminium fluoride to fo~ mixed compounds (e.g.
cryolithionite oornpounds such as 3Li~.3NaF.2AlF3)~
Typically, aluminitlm fltltor.ide irl the form of solid partl.cle~ is brought lnto contact wl.~h t'ne molten aluminium. The treatment material may consi~t essentially of aluminlu~ f~.uoride9 or may be ~omposed whol7y or ln part of alkaJ.i metal fluo~l~mir~tes ~,7hich are solid at the temperature o the molten metal~ An example of the latter type of material (useul f or removal of llth:lum, magnesium~ and calcium~ is parti~ulate sodium cryolite or lithi.urJ~rree reduction ~ ~ ~ 8 ~ ~ 7 cell electrolyte having a l.ow ratio by weight of ~o~ium fluoride to alumlnium ~luorlde so ~s to cont~in aluminium fluorlde in excess o~ ~he stoichiometric require~ents of Na3AlF6 with a composition SUC;l that a major proportion ~em~ins solid ~t the treatment temperature, as i5 usually the case provided the afore~
menti.oned ratio remains within the range of 1.3 ~ 1.5.
Indeed, it is not essential that the addition re~ain in solid formg a low (approxlmately 725~C) melting~
ternperature compound conta.ining a large excess of AlF3 (e.gO h~-v-irlg a ve-~y low ~7~F~AlF3 r~tio by weig~lt o~
0.6 ~ 0.7~ which rl~elts on introducti.on to molten aluminill~ would be equally ef~ect-Lve in removing ~lkali ~et~ls or ~llc~line earth metals. The active fluoride materi~l may also conta~n inert material suc,h as ~luminillm oxide, in a proportion even as high as 50% by weight, although 7 ~ 20% i~ the more usual &lUm~ ~liUm oxide contenk of commercial alumini~lm fluor:lde.
Treatments with a1uminium fluoride ~re considered advantageous or removal of alk~ etals and ~lkRline earth metals, as compared with fluxing with chlorine gA5 or chlorine/inert g~s mi~tures, beca~se the gas flw~ing operations yield deleterious gaseous by-products and a~e otherwise inc~nvenienta In prior treatments employing AlF39 as described in U.S~ ~tents No. 3,305,351, No. 3,528,8019 and ~0. 4313~9 246, the molten aluminium was passed through a p~cked ~ilter be~ of solid particulMte mater1~1 containing aluminium fluoride, alone or in mixture with oarbonaceou~ material, such as coke, In U~S. Patent No. 4,277,?.80 a similar effect is achieved by pas~ing molten aluminium upwardly throllgh a reactive bed of coarse gratlular AlF3-containlng materi.al which is not S a fllter. However, the use of reactive beds or bed filter~ comprised o~ reactive materials has several d1sadvantag2s, A substantial proportion of the products of reaction of the alkali ~etal (L1D Ma, M~) wlth aluminium fluorlde remains trapped on or within the re~ctive bed or associ~ted filter ~aterl~sl to cause premature plugging~ electrolyte ~rom the reductioll cell 7 sludge and/Qr ot.her solid or liquld impuritie~ carri.ed over wi~h the mol~en metal frOM the electrolyt:ic cell h~ve th~ same efects~ For slmllar reasonC7~ a prefererltial Inetal path or "c~annel" can appear within the reactive bed ~d ~eriously reduce the ~lk~3.i metal removal efficiency. The aluminium fluoride material is consumed durin~ the treatment of molten inetal and consequently the performance of a reactive bed is no~
~0 constant durlng it~ s~rvice life~
To prevenS alusminium 1uorlde pyr~hydroiysis and met~L losses in reacti~e beds, it is prefera~le to keep ~lusnirsium ~luoride always completely submerged in molten Aluminium9 but this requir~s a constant heating ~nd uel consurnption, even when the installation is not oper~ting7 which adds to the cost of the treatment.
Change of the composition of the metal treated through ~uch a system is il~va..iably associaLed -w-ith sri2tal losse~

~ 8 ~

Also, during initial prehee~ting of the AlF3 bed, decomposition ~)y pyrohydrolysis (i.e. reaction wi.th water vapour in cornbustivn products) tend3 to occur.
I~ is difficult to achieve eff~ctive contact be~ween loose aluminium fluoride particles and molten alu~Qinlum metal. The reason for this i5 that, due t~ the high surace tension of molterl aluminlum ancl the sm~ll di ~erence in densi~y between alumini~ml fluoride po~.~der and molten aluminium, AlF3 powder will float on the surface of molten aluminiu~l.
Addition~lly aluminium fluor-lde powder is not easily wetted by molten alumi.n-LuM, and is the~nally very stable, i.eO it does not melt under atmospheri.c pressu-fe! and it has a sublimation temperature of approxi~.ately 1,27~ C, so that reaction betweerl liquid~
li~uid or g~s-liquid ph~.~es is impossible at ~he treatlnerlt temperature o~ molten aluminium (660C - 900GC~o Th2se physical characterlstics explain the poor performances of previous attempts to introd~lce discrete ~articulate aluminiu~ fluoride into mol~en alumlniul~.
It is pc~ssible to ir~ct aluMinlum fluoride particles into ~olten alu~inium in a jet o~ carri.er ga59 such a~ al, or nitrogen, ~y means of an injectlon lances Injection operations, however, have be.en ound to require substantial periods of time7 and there are saety hazards associated with the high gas pressure in the metal; ln addition, use o air as the carrier gas can l~ad to excessive dross and oxide fi.~m o~.l~t~ong ~6~
It is also possible to make a large addition of alumin~.um fluoride powder to the bottom of an empty cruclble before metal addition. However~ i.t has been observed th~t the aluminiu~n fluoride powder reaots ; preercntially with the cell elec~rolyte (which is invariably siphoned from the reduction cell alon~ with the mol~en Al metal) to fo~l a solid mass which remains attached to the cr~cible lining. T~us effective cont~.ct witlt the molten ~luminium is prevente~., As wilt be understood from th~ foregoing discussi.on a sub~tantial contact time is ,eq~eiLed ~or the eff~ciellt reactlon be~ween loose particulate AlF~o containirlg material and alkAli metal a~d alkaline ea~th metal contamil~ants in molteR Al metal.
The method o~ the invention requires the addition of an ~ppropriate charge of the treatment material ~AlF3 or AlF3 containi.ng mate~lal) to the molten Al metal under conditions which involve re eirculati.on o~
the treatmellt materi.al within the molte~ metal while avoiding excessive ~istltrbance o~ the molten metal surface, to hold down oxldation of the metal. In the method o~ the lnvention the treatment ~aterlal is entrained in the molten aluminium by sttpplying the treatment material to a vortex generated in a body of the 2~ molten metal h~ld within a container. The vortex generatsr al~o ser~es to generate u~wardly spiralling current~ in the molten metal ln th~ region Qf t.he boundarie~ of the container to maintain prolonged cont~ct of the par~lcl.es o~ the ~reaS:rrlent materi~l with the molten me~al. The circulation of the molten met~ll by vortex ~eneration i9 continued for a sufficient length of time to reduce the alkali metal and al~aline earth me~al conten~ o~ ~he moltel1 metal to a desired low value, after which the circulatiol1 is discontinued.
Some of the reaction products, whi.ch are admixed with residual trea~ment material, will rise to the surface as R dross, f.om which the molten metal can be separated by dro~s skinm1in~ or metal ~iphoning or other conventional means~ However the greater part tends to adhere to the c~lcible lining cluri~ the stirring proce~s~ whence it can be remove~ when the cruc;ble is empty .
It is well kno~n in metallurgical process~s to ~ntroduce ~eactl~e materials into molten me~al vort~ces~ generated ln ves~els from which the molten metal i.s dlschar~ed as a continuou~ str~am. In th~
present procedure ~er!eration of a vortex serves both as a mezns or bringing a finely powdered particulate materi~ relatl~ely ].ow bulk ~enslty into oontact with molten metal ~nd as a means for ma.intaining the particles o such m~terial d~persed within saicl mol~en metal and in intimate contact therewith over an extended perlod until g~neratlon OL the vortex i5 terminated.
The vortex is preferabl.y generated and malntained by using a rotating stîrrer having a multi~
bladed rotor immer~ed within a body of molten metal COTItained 1n a c~lcible ~nd rota~ed a~out a vertical axis 3 wi.th the blades pitched so that c~ch blade h~s a ~8~7 ^ & -ma~or surface facing downwardly at an acute angle to the vertical. The impeller rotor is preferably arranged in the crucible eccentrically with respect to the vertical centre line of the crucible.
Electromagneti.c induction stirring may also be employed to generate a vortex. Appropriately arran~ed induction coils may be disposed externally of a ~rucible or other vessel containing the molten metal.
The lnvention also provides apparatus or mlxing parti.culate AlF3 treatment material with molten aluminium, i.ncluding a crucible for the molten metal~ and ~n impeller or rotor having pitched blades and disposed eccentrlcally of ~he vertical centre line lS of thP c~ucible5 with v~ri.ous dimensional. and positional relationships maintained within speciied ranges or li~its described below~
In the accompanying drawings:
Fig. 1 is a simplifled sec'cional plan ~iew o~ ~n apparatus or perfonming the method of the invent~on, and Fig, 2 is a vertical section on line 2~2 of Fig. 1, In the drawings a cylindrical crucible 10 contains a body of molten al~inium 11, A separate lid 12 supports an eccentr~cally ~ounted impelle~ 147 ~riven by a motor 16. The impeller 14 has a shaft 18 which ~arries b].ades 20 for immer~ion in the molten ~lumin~um body 1~. The lid 12 also includes a ~ 7 duct 22 for supply ~f treatment ~ateri.al to the crucible; and an exhaust conduit 24 for exhausting fumes fro~ the crucible. Typically9 the crucibl~
comprises A steel shell) with a refractory lining i.nert to molten aluminiu. The lid 12 and associated items comprise a vortex generator assembly which may be transferred to permlt the ~ame stirrin~ apparatus to be used to stir batches o~ molten metal contained in ~ ser~es of differenL mobile crucibles.
For removal o contami~ant alk~li metals arld~
or alkaline earth metals ~rom mol~en al~minium~ th~
crucible 10 is ch~rged with an approprlate quantity o molten Al metal~ The lid 12 is then pl~ced on the crucible to immer~e. the ~lad~d portion of the impeller 14 P~rticulate treatment material comprisirlg or con~isting of Aluminium fluo~ide (AlF3), which is solid at the temperature of molten alumini.urn~ is then ed b~ gravi.ty throu~h the duct 2~. Rotatiol.l of the impeller should prefer~bly be col~tenced before introduction of the treatment material (but m~y be commenced after such introduction) and maintains a stable vortex (indica~ed at 26 in Fig, 2) in the molten body 11, Generation o~
the vortex results in ~ ~ombln~tion of ax~al and radi~l flow compon2nt in the molten metal~ The AlF3 particles are dra~n into the vortex and then circulated through the molten body along flow paths genera3.1y indicated at 28.
It is not necessary to char~e aluminitlm fltloride directly into the vortex~ slncP the mat~rial wl 11 be rapidly mQved thereto by th~ hi~,h rate o mPtal ~o~
clrcula~-ion at the melt s~lrface.
Rotation o the impeller is continued, with m~intenance o~ the vortex 26 and recirculatlon cf the aluminium ~luoride particles until there ha5 been sufficient reaction between the al.uminium fluoride and the d.issolved contamin~nt alkali rnetals ~nd~or ~lkaline earth metals to reduce the content of these contaminants in the mel~ to a desired low valueO Typically, the time required ~o a~hieve thi~ result is no more t}~n abo~t tPn minutes, and indeed o~te suhstantially les~ than ten n~inute~, ~ompounds, ~uch as c~yolithionit2 compo~mds, produced by reactio~ o the contaminaTIt alkali metals and alkalin2 earth metals with the alumini~n fluori~e, float on the surface o the mol ten body9 and may be readily removed by skimming or o~her means when the rotation o the impeller is ended and the lid i~ lited away fro~ the crucible. ~rhe decontaminated moltel~l"etal may then be poured o~ othel~ise withdrawn from the cruclble~
By this methody it is possi~le to reduce ~he l~vel o~ cont~minants from a typical level. o~ about 20 p.pOm~ lithium and abou~ 30 - 60 p~p.m, sodi~m to less than 1 pOp~m. wi~hin a period of ten minutes or less of continuou~ stirring with the impeller. Since some reduction in the levels of these c.on~aminants occurs inherently durir.g the performance of other steps comm~nly employed in h~ndlin~ molt.en aluminium, it ~s frequently possible to achieve satisfact.ory reduction in content of the contaminant metals, e.g. to 2 p.p.m.
Li, with even shorter periods. Even though the aluminium ~luoride may ~ontain a proportion of S alumina, the fluxing action of the fluoaluminate reaction product serves to remove the insoluble alumina.
In fact, i~ is observed that the process of the inventlon has the incidental eff ect of removing inclusion ~orming materials, such as aluminium carbide (A14C3) 9 ~.~hich were present in the melt before treatment.
The optimum colnbina~ion of axial and radial flow components for attaining a high mixing efficiency of the solid AlF3 particles :lnto the molten alulslinium ~5 is achieved by appropriate disposition of the impeller relative to the crucible and/or by the dimen~i.ons and deslgn o the i~npeller blade. To this end, the impeller ~ay comprise a plurality of e~lliangularly spaced, pitched blades 20 each having a major sur~ace 20a that faces down~a~:dly at an ~cutP angle to the vertlcal. The axis of the impeller shaft is dispo~ed eccentrically of the ~eometric axis o the crucible, and the direction o~ impeller rotation is such that the blade surfaces 20a are the leading ~5 surfaces of the blades, exertlng a forc.e having a downwar~ component on the molten aluminlum. In the drawing, 0 designates the pitch angle o the blade s~r~aces 20a, d designates the overall. diameter of the bladed po.rtioll o the lm~ell er, h designates th~

-12~
height o~ the i;npeller blades, x designates the eccentrlci.ty o~ t:he impel~er shaft, y designa~es the vertlcal distance from the bottom of the c~lcible interior to the midpoint of the impeller blades, H designates the vertical distance from the bo~tom of the c~lcible interior to the quiescent level of molten metal in the ~lcible, D i~ the internal diameter of ~he crucible, and th~
arrow R represents the direction of impeller rotation~
In accordance wlth the inventjon, as particular or preferred features thereo, the following ranges of relationships and dimenslons are observed iTI the de~ign ~nd disposltion o~ ~n impeller o the illustrated type:
Relatiolship Preferred or dime~sion Outside ra~ Ran~
d/~ 0.1 - 0~6 0.15 ~ 0.40 h/H 0.1 0.7 0~2 ~ 0040 Y ~.25~1 ~ 0.75~ 0.4 ~ ~ _ 45 30 ~ 40 Although s~ti~facto-ry results can be obtained with a centrally located impeller or impeller ~vlng little eccentr~c.i.ty, the eccentricity7 X9 of the impeller ~ha~t.
1S USUA11Y in the range OL 0.1 O. 25D ~nd more preferably in the ~ange of 0.25 0.7 d. It is e~pecial~y pr~ferred to utilize three blades spaced ~5 120b ap~r~ with a pi.tch ang~e of 3Q~ - 35, and a rati~
d/D of abuut 0. 250 The impeller e~e~tricity9 x, i~
most preerably O.S d.
The de~oribed i¢~peller arrangement i5 adv~ntageous ln creati.ng a stable vortex without us~
o~ ~ertic~l bafle~ ich would be im~r~cticabl.e ln interch~ngeabl.e transfer cr~cibles. The funrtion of conventional baffles in generating vortices by maintainir,g a high rate of relative rotation bet-~een the impeller and the liquid is achieved wi~h the present i~peller arrange.nent by the combination o radial and axial 10w components produced by the impeller G
Since blade pitch angles ~ as large as 45 or more tend to cause ~plas~ring and surface w~ves, lt is preferred ~o ~ e a smaller pitcl~ all~le? suc~ a~
30 ~ 35 to force the metal downwardly to drag the ~luori.de powder lnto the ~nolten aluminiumO
The requisite axLal component vf molten Metal ~low can bc achieved9 even with a ~ertical-b3.aded i~pellel, by 3.ncating the lmpeller eccentricall.y with respect to the geometr-Lc axiY of the crucible~
HowevPr it is greatly preferred ~o employ an eccentrically loca~ed pitched~bladed irnpeller, in order to minimize rnetal waves atld o~cillatiolls at the surface of the rnetal. It is found that eccentric location o~
the lmpell~r pe~nits the crucîble to be filled to a greater ~xtent without rlslc of spJ.ashing during the ~tirring of metal in transer c~uci~le~ of large size.
T~e e~centric locatlon o~ the impeller consti~utes an important ~eature o a preferred arrangement in accordanc~ with the inventi.on since it pe~mit~ the treatme-llt o a ~ubstanti~lly laI~er ba.cn of metal :in a cruci~le ~f given size.

The minimum r~te of rot~tion, for a glven impeller, is that which will gellerate and maintain a stable vortex, ~hile the maximum rotation rate ls that above which ~ir is in~ested i.nto the molten body being stirred, These values are determined by the impeller diameter d. The opti~um rotation rate is that whioh produce~ a good vortex without causing exc~ssive metal splashing and loss or being responcible for ero~ion of either the crucible ref~actory or impeLler constr~ction material~ Referrin~ to an impeller provid~ng a d~D
ratlo wi thin the p~eferred range of O.lS ~ 0.40, it i5 ~t present preferred to operate such an impeller at a rotation rate of about 100 tn about 300 r~pO~.
~owever? rates of rotat;ion outside this range ~y al~o be used~ so long a~ th y produce th~ desired ~ort.ex without excessive s~lashing. The u~e o an eccentrically ~isposed impPller havin~ tilted or pitched blades rotated in the above-defined dirertion is found to be e~pecially ~atisfactory in generating the stable vortex w~th a highly effective combinaticn of axial and rad~al flow components or enhancing penetr~tion of the solld alumini~l fluor~de particles into the mol~en me~al.
During the trea ment of molten aluminiu~ with AlF3 powd~r, alkall ~nd alk.alinP earth metals react wiSh AlF3 to ~o~Q mixed alkali cxyolithionite compounds, ~-g- Na~13F14~ N~7L1~1~6, and L13Na3A1~ These compounds~ havl.ng ~ relatively ~.ow mel~in~ po~n~, can easily ~e agglomPrated or stick to the c~lcible walls or float to the ~elt surface where they re~ct with rnetal oxide or particles of cell electrolyte ~lwa~s present after the siphoning o electrolyt~c cells, During subsequent metal transfer by siphoning, most of these compounds will remaill ins;de the c~cible alld are thus separated from the molten Al.
AlthQugh a high grade of AlF~ is desira~le or a fast reaction with alkali metal, a higher ratio of addition of AlF3/Al could compensate for a lower~grade powder.
Other mixtures whicll could be usecl are lower-grade AlF~
(e.g. AlF3 mi-.~ed ~ith A12033 or elect~olyti.c bath m~teri.al containing a large excess of AlE'3 (i.e. Na3AlF6 ~ith excess AlF3), By ~way of further i.llustration o the inventlon9 reference may be made to the following e.xamples:

130 kilogr~m samples of 99~7% pu~ity molten aluminium ContaiTIing ~etween about 20 and 25 p.v.m. Li were treat.ed with solid AlF3, grolmd to -35 mesh particle slze, usin~ each of the following proce~lures:
A~ 300 ~rams of AlF3 particle~ deli~ered to the melt surfac~, without a~itat:ion.
B. 200 ~rams of AlF3 particles delive-Led to the melt- surface, with the melt agitated by a rotor rotating at 900 r.p.m.; without vortex generation~
C~ 300 grams of AlF3 particles injected into the m lt below the surface through a ~r~phite lance, using nitrogen as the carrier gas~
D~ 20~ &r~ms of AlF3 particl~s injected a~

~8~

in C, but ~7ith the melt agitated by ~ rotor (posit.i.oned above the lance outlet) rotating at 900 r.p.m.
E. 200 gr~ms (E 1) and 300 grams ( E-2) of AlF3 particl~s delivered to the melt surface while a stable vortex is generated and maintained in the rnelt, by a stirrer rotating at 225 r.p.m., in accordance with the present method.
In a urther procedur,~ (F), no A1~3 was used, b~t the melt was agitated (without creatioll of a vortex~
by a rotor rotating at 900 rOp.m. Results were as follows:
Li (p.p.m.) remaining a~ter treatment Procedure ~ ~
0 min.
(start) 3 minO 6 min. 9 mln~ l~.rninO
__ _ A ~414 9 6 4 B 221~ 7 5 C ~3 9 ~ 2 D lS 7 2 F 2~16 14 12 1~
The procedures embodying the present znethod (E 1 a~d E~2) achieved signlficantly ~ore rapid reduction in lithi.um content ~han any other proGedures 7 and the l~thi~n level rea-~hed after nine minutes (1 p.p.m~) with the procedures of the present zaethod was equalled only by the procedl~re (D) involving combined use of i.njection and agitation, wherein the i.niti~l le~el of lithi~
contamlnation was much lower~

88~7 The dlmension~ ~n~ disposition of the i.mpeller sti.rrer in thi~ example are as ~ollows:-The i.mpeller diameter9 d, was 12.5 cms and theimpeller was a four-bladecl lmpeller havin& a blade height of 8.8 cms, with the blade3 inclined at ~n angle o 35~
The diame~er of the crucible was 50 cms and the values of H and x were respectively 37Q5 cms and d/2, Using equipment of the type illustrated ln the drawing, includlng ~ cylindrical cruclble having an internal dlameter of about 160 cms with a nomin~l capacit~r ~ about 4500 kg of molten alumini~, a series of quanti.ties of molten alumi~ium contamlnated with lithlum ~nd sodiurn were treated in a~cordancc with the present method. In each instaIIce~ the crucible was filled ~o a deptll of 100 cms with the rnolten ~luminlum, and ~lF3 powder at the rate of about 1.7 kg per metric tonne of ~l~Lllnium was supplied to the molten alumlni~n surface~
A pitched~bladed ~mpe~ler havin~ a blade height o~ 2~...cms and diameter of 45 cms was lmmer~ed in the rnolten aluminium w~th an eccentricity o ~0-30 cm3 ~preferably 22.5 cms~
with the centre of the impeller ~lades 37.5 cms a~ove th~
crucible bottom9 ~uch that the top edges o the blades were locat~d halfway between the top of the melt ~nd the crucible bottom (the blades thus b~ing diQposed entirely w~thin the lower half of the mclten metal body); the lmpeller was rot~ted, ln each case, for 10 minute~ at a rate o between about 130 and about 135 r.p~m. to create and main~ain a stable vortex, The mode of addlti.on of AlF3 p~zti~le~, and the molten me~tal temper~tule, were ~arled rom test to tQst. Results of twenty succesYi~Je te~t~ were~ as follows:

~, t) r~ o u~ C ~ o o ~ O n In o o ~:
o q~ CC O ~ ~ ~ ~ ~ O ~ r~ # #
r~ r~ r ~ r ~ r ~
E

.~
dP ~ ~P ~ dP c3P dP ot dP dP o~ dP dP dP ~ dP c~ dP df' d3 O O O O O O O O O O ~t O O C:~ O O O O O 0 1 E~ c, o o o o o o c:) o o o o o c:~ o ~ o o o a~
~ r~ r~ ~ r~ r-l ~ ~ r~ r~t r--t r~t r~ ~ r~ l r~l ~~~ r~
El cP
P~ .
_ ~ ~ t --t r~l r-l ~t --t ~1 -t --t --t --t --t r-t ~ r-l ~t --t t~
~ ~J ~ ~ ' V V ~ V ~/ V V ~ r _,_1 ~ O
~1 U~

U~

o ~ o ~ ~1 0 _ ~
. ~

. ~P .

_ ~.~ a~ o ~ ~ o ~ r~
~.~ ~ o ,~ o ~i o o o ~ ~ _I o ~ ~ ~ ~ ,~ ~ o ~ ~ o ~ .
O ~ 1-- ~1 ~ W ~ r- o~ ~ o ~ ~- c~
....... , ...... ~ ....
u~
*
a~ ~
~ o ~ ~ ~l O JJ O

n 0'~ ..
E~ :

~ not measured ** a - 50% of AlF3 added at start, 50% added after 1 minute of impeller o~eration b - 33% added at start, 33% after 30 seconds, 33% after 1 mimlte of impeller op~ration c continuous feed for 1.5 minutes ~rom start of impeller operation d - conti.nuous feed for 1.0 minutes from start of impeller operatlon e ~ continuous feed for 0.5 minutes from s~Art o i.mpeller operation f - ~3% ~dded at start, 33% after 15 seconds~ 33 after 30 seconds of impeller operation g - 3~% added at start, 33% after 10 seconcls, 33%
after 20 ~econds These data illustrate the adverse Pfect o~
increased metal tefnperature Oli efficienry of lithi~n remo~al, attributable to a the~modynamic~lly controlled lithium equilibrium betwe~ the fluoride material and the metal which prevents 1~0% efficient l.ithlum removal rom the hot metal; a s..milar effect could not be observed for sodium because o the higher vapour pressure o~ SO~iUIIl, which assists in its rernoval.
The average lithit~n removal efficiency~ after 10 minutes' treatment time, was 93% for the twenty tests included ln the fore~oing table~ This corresponds to a lithium level (for the treated metal) which is satisfactory~ i~e. below the maximum ~cceptable limit for most purposes.
3G EXA~LE 3 Sev~ral series of ~reatments were per~ol~ed 20~
ln transfer crucibles on aluminium which had been siphoned into the crucibles frorn electrolytic reduction cells. The aluminium fluoride powder used (92% ~lF3, about 8% A1203 by weigh~) had a bulk density o~ 1.5 - 1.7 g/cm and a particle size distribution as follows: 25% larger than 100 microns, 50% larger than 80 microns, 75% larger than ~5 microns.
In tnese trea~ments, the c~ucibles contained approximately 3,500 kg. of molten alu~.inium each.
A three-bl.aded impeller having a blade pitch (angle ~) of 35, dlameter (d~ of 4~ cm,, and blade hei.ght ~h) of 25 cm. was employed, and rotated to establish and maintaln a stable vortex; the ratios d/~
and h/H w~re each 0.259 and the MaXimUm trPatmellt time lS was 9iX minutes. Tl~e rotor eccentriGity9 X9 was d/2.
Several crucibles were treated in each s~ries~ For purpose~ o comparison, one series (Series 1) W&~
run without use o aluminium fluori~e. The remaining six series of treatments embodied t~e process o~ the invention, In seri.es 2 5, all the alu~nini~m fluori~e was ~dded at or before the start of stirring; i~
series ~ and 7~ one third of the alum~ni~m fluoride was added at the start, one third a~r one half mlnute~ and one third after sne minute of stirringO
The metsl in the crucibles of series 7 initially contained 101 p.p.m. of magnesi~; the metal in the other six series contained less than 10 p~p.m.
magnesi~n, Results were as follows:

;~ 9 ~udd v~
ad ~ _-. ~ ~ , _ ~udd ~ ) r~ ~ r- ~ i _ -~ ~ r~ - --l ~
~ L I ~ % ! ~ a~ ~ ~ ~ cn ~ i ! I ~ I . . , . . ~ _.. _ j j j Illdd ¦ ~ ~P ~ ~ ~ ~ ~ -~ . _.
i :~ I ~ P
D~ l-eAot~ I % ~1 ~ 1 ~ t`l r-l --_ JJ~ ~ ~: ~ ~:
o . ...... . __ _ . _ --,.
U~ . I~
~ ~ . ~ . . . .
uldd~ ~" ~ ~ ~ ~ r- ~-. ~ ---r.) . ~ ~ ~D O ~ N 1~ t~
.~ ~ ulddtn ~ ~ ~ ~ r~ ~7 = ~~k --_ ---- ___= I ..
d~ ~P ~P ~P ~ d~ ~P _--, . l e~ou~3~ % 0~
. _ _, . .~ . --=
.~ . . I o t~ r _.
. ¦ ~o u~dd ~ ~ r __.
_. _ _ _ ' . ._ ._ .- .. .____._. - _.
~ C'.
1 ~ o ~ouIa~ % ~ ~ ~ ~ p ~.-D ~,-.
~rl - _ -- _ . .. _ - - C
~ ~ . O~ ~,, dd , ~ Ln I r- - ~ 1- -- --- I --- - - I r-

3 1 h ~ O r ~
.~) I a ~ r) ~ ~ ,c -.

I r l ~ l L'l G

r ~U n d ~ o o o o u~ u~ =,-p~3~dS fju~ S ~

auuO~ I 0 1- 0 ~ I .--.
~ ~ o ~ ¢ ~ a~

s a ¦ q ~ F=
p~sa~L ~ aq.~ O ~ c oNsa~la~sa~ L~
L -- ----- ~ ~ ~ - ~~ --~ _--$~

-22~
Series 1 illustrates the removal of alkali metals due to the aluminium stirring effect only.
The greater sodium removal after 3 and 6 minutes (61%
and 72%) compared to the lower lithium removal efficiency (15% and 19%) is attributable to the much lower vapour pressure of lithium than sodium.
In e~fect, sodium has a boiling point at atmospheric pressure of ~82 ~ compared with 19 329C for lithi~, Series 2, 3 and 4 compare the effec.t o~ the AlF3 quantity on sodium and lithium removal. It oan be seen that increasing the ratio kg AlF3~met~ic tonne Al ~rom ~.7 to 3.3 had a maxked effect on lithiu~
removal. The effect is not so apparent for sodium due to the faster sodium removal by oxidation only~
Series 5 i5 identical ~o serles 3 except or an increased r.p.m. from 100 to 150. This increased the sodium and lithium removal efLi.ciency ~rom 89~/o to 92% and from 74% to 85%9 respectively.
Serie3 6 illustrates, for 7 trAns~er crucibles, th2 influence of a sequential addition of AlF3 powder Gn the removal r~te of alkall metals. It can be ~een that this also helps in increasing the removal rate, probably by increasing the interfacial area between the powde~ and the aluminium (the ~ddition o a large quantity of AlF3 in one "shot99 can cause powder agglomeratlon and decrease ~he effective contact with the alumi.nium).
Serles 7 illustrates the i.nfluence of Mg metal pre~ent in additior. to ~i, Na and Ca~ I~e Mg 23~
content after 3 mi.nu~es stirring was 46 p.p,mO (5~%
removal) and after 6 minutes as 30 p.p.m. ~70% remo~al~.
It can b~ concluded that the presence of Mg, even in a concentration larger than otheL alkali metals, does not signi~icantly affec~ process efficiency. The presence of Mg in these tes~s wa~ due to the u~e o a mixed LiF~
MgF2 electrolyte in the reduction cells. The presence of magnesium in the metal due to other causes (e.g.
contamination from Al-Mg alloys) could also be tolerated. However, if Mg conGentration increases p the addition o~ AlF3 would have to be ad~usted accordingly, to ensure a constant lithium and sodium removal ef lclellcy 0 In two further series o tests u3ing the same equipment as ln Example 3, ~roups of transfer crucibles each containing about 3,400 kg~ of molten alumlnlum wer~
treated in accQrdance wi~h the pre~ent method. AlF3 powder was added at the rate o 2.0 kgr AlF3~metrio tonne Al to each crucible in three equal lncrements, viz~
at the start; after 30 seconds of stirring; ater one mlnute stirrirlg. The stirring was performed for six minutes at 175 r~pOm., produing and maintaining a stable vortex as in Test 6 of Example 3. The ~rea~ed metal ?5 fro~ one series was used to prepare a irst alloy ~havlng the Aluminum Associat~on designation AA~1350) and the treated metal o the second series was used to prepare second alloy (Alumin~n Association designation AA-5154~o Alkall ~etal and alkaline ~arth metal content was 3~ mea~ured again af er alloy.ingO P.~sults webe as follows, ~38 _ .~ ~
U) O ~ ~D
~ ~ . .
;~ ~ ~ ~ O r~ ~ O ~
~ ~ V V ~, V
a~ S,~, s~ ~
o ,~ ~
a)~ .
. . .
U~ .
~1 ~1 Ul dP ~ dP ~
d~ O ~ ) ~ ~ .-a~
~ ~ .
sd ~
h 1:
:~ ~ r~ r~
~¢~ Q. . . . .
~ ~ 4 ~
~¢ ~ V
.
~ rC
a~ ~ r~ r~ r~
0 N--1 N1~~IS~r) n- I I I I I I
C) ~ dt~
E~ Z
.
~_ l ~1 ~ tn -,~ I
Jl a) u rl ' C) ., s 0 O ~
' ~ e.) u~ .-~i ~ J ,~:
~1 ~ C) ~ ~ t.
.sC,~,I _ ~ Q r O tr S-l ~_ t:-~rl ~ ~ ~-a) a~ . u~ -3 e~ r- ~
_ E
~,s h ¦
I o ~r a~ In m h r~ ~1 -Pi ~1 In -l l O ~ ~
s . 1~-_ r ~ 7 It can be seen that the efficiency is o~ tne s~me order as in Test 6 of Example 3. It can ~lso b~
~oncluded that the sodium and lithium concentration oontinues to decrease ater the ~reatment. This can be attributed to various metal operations and treatment (transfer, alloying, stirring, heating, holding, etcl~ which ~ccelerate the oxidation of ~lkali metals in the rurnace.

_ _ Again using the same equipment and the s~me AlF3 powder ~s in Examples 3 and 4, molten alumlnlum in transfer cruciba~sea~h containing 3,500 kgo of ~lumi.nium was stirred or 10 minu e~. at 100 r.p.m., a stable vortex being created and maintained. Ater treatment~ the metal ~tood for 10 mi~utes and the alkali metal content was reme~sured. ~esults were as ~ollows.

.
__ ~ -----7 ~
During Treatment after Tes~ AlF3 Temp. (Min.) Trea~me~t No. kgs/tonne ( C) ~~ rF~~~
_ ~ .
2.0 760 Lil 18.3 6.0 3.6 3Dl Na 43 9.0 4.0 3 B 2~6 825 Li¦ 15 4.3 2.7 2.5 N2 _ _ _ _ 1 ~8~7 2~-The observ~d further decrease in alkali metal content after stirring was ended may be explained by the high ]evel of activity of the AlF3-rich re~ction products in contact with molten aluminiurn, Even if this alkali metal decrease on standlng after treatment is not signiicantly important as compared to the reduction during the treatmetlt itself, it nevertheless indicates that there is no risk of back~reaction (alkali metal pi~k-up) during holding in ~he transfer cl~cible betweenthe treatment and the transfer to the rasting fu~n~ce.
This would not be the case if the alkali metal were removed using a treatment with chlorine gas reaction only.

To illustrate the effect of impeller blade angle, a series of tests were performed on 125 kgo sample~ of molten ~luminium of 99.7% purity at a temperature o 825C, u~ing 35 mesh aluminium fluoride powder in a proportion of 0.~ kg.Jmetric tonne Al. Impellers with bl~des of various pitches were employed; in each case9 d - 12~5 cm, h = 11 ~m, d/D = 0.25, and h/~l - 0O25 and x = d/2. Stirring was performed for six minutes at 250 r.pOm4 Result~
were a~ follows:

.. . .... ..

Angle to Lithium Concentration Number of Vertical ppm (% Removal) Bl~des 0 ~
___ ., ~ , ,, Start 3 6 0 27 13 (52%)11 (~0%)

4 30 ~0 1~ ~70%)8 (80%) 35 26 5 (81%)3.5 (~6%) ~ 45 26 4 (85%)3 (88%) 1~ 3 35 Increase in pitch angle incre~sed the percentage removal of Li after three and six minutes, and the number of blades appeared also to affe~t the removal ef~iciency.

A synthetic mixture containing 50~/O each (by weight) of oryolite (Na3AlF6) and AlF3 (weight ratio of NaFjAlF3 8 0,43) was prepared by fusion of the tw~
compounds, ground to -35 mesh particle size, and employed for treatment of molten aluminium in accordance with the present method. Two 150 kg. samples of aluminium, both at 825C, were treated~ using ~ stirrer heving four blades, with a pitoh angle ~ of 30, diameter (d) of 12.5 cm., blade helght h of 11 cm. 9 in a crucible so d~mensloned tha~ the ratios dJ'D and h/H
were each e~ual to O.~S; vne of the two ~ests employed ,, ,, . ., .. , ~ . . . , . , " . .. .. . .... ..... .. .. ... .

~8 a fluoride--containing material constituted o 85% AlF3, 1~% A1~03 by welght, ~nd the other ~mployed the aforementioned cryolit~-AlF3 mixture9 both in a proportlon of 2~0 kg per metric tonne of aluminium~
Results were as follows:
Lithîum Concentr~tlon (ppm~
. YS. Stirrlng Time (min~) AlF3-containing ~ ~ .
_ mrc 85% A1~3 21 6.3 3~8 ~.5 105 1.0 15% A1203 (88%) ~95%) 50~ AlF3J 3V 7~5 3.6 2.4 1.5~1 50~ ( >97%) .
~

The high efficiency o the AlF3/~a3AlF6 mixture i3 possîbily attributable to the srmation of low melting lS point (about 700G) phasesJ It thereore melt~ at~r cvntact with the li.quid Aluminl~ pro~idlng a liquid-liquid reaction rqther than the solid-liquid reaction ~ith the AlF3 powder whic~ compellsates ~or the ~lumi.ni~m 1uoride dilution.
In addition, al~min~um ~luoride powder in mixtures of a wide ran~e of partlcle sî~ distrlbut~on have been used, with the average partlcle siz~ dlmension vary~ng be~ween 1 and 0005 mm~

Claims (12)

THE EMBODIMENTS OF THE INVENTION IN WHICH AN EXCLUSIVE
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:

1. A method of removing contaminant alkali metals and alkaline earth metals from molten aluminium by reaction with aluminium fluoride to form fluoaluminates of such contaminants which comprises (1) placing a charge of the contaminated aluminium metal within an upright essentially cylindrical vessel (2) stirring molten aluminium in said vessel under conditions to establish a vortex therein and flow currents in said molten aluminium having both downward and lateral components at the bottom of said vortex and upwardly spiralling currents in the region of the periphery of said vessel (3) supplying particulate aluminium-fluoride containing material for entry into said vortex (4) continuing the stirring of the molten aluminium until the alkali metal and alkaline earth metal content is reduced to a desired low level (5) separating the molten Al from the molten fluoalumi-nate reaction products.

2. A method according to claim 1, further characterized in that said vortex is established and maintained by means of a multi-blade impeller having blades which are inclined in relation to the axis of rotation.

3. A method according to claim 1, or 2, further character-ized in that the vortex is established eccentrically in relation to the axis of the container.

4. A method according to claim 1, or 2, further character-ized in that the molten metal is treated with powdered AlF3 or NaF.AlF3 having a low NaF/AlF3 ratio by weight.

5. A method according to claim 2, further characterized in that the vessel has an internal diameter D, and is filled with the molten body to a height H, and the impeller has a diameter d and a blade height h, such that the ratio d/D is between 0.1 and 0.6 and the ratio h/H is between 0.1 and 0.7.

6. A method according to claim 5, further characterized in that the axis of impeller rotation is eccentric in relation to the vessel axis by a distance, x, having a value 0.1 - 0.25D.

7. A method according to claim 6, further characterized in that the midpoint of said blades is spaced above the bottom of said vessel by a distance, y, between 0.25H and 0.75H.

8. A method according to claim 5, further characterized in that d/D is between 0.15 and 0.40 and the impeller is rotated at a rate between 100 and 300 r.p.m.

9. A method according to claim 2, further characterized in that the treatment material is fed to the molten metal in separate quantities or continuously during a short period after establish-ment of the vortex.

10. A method according to claim 5 further characterised in that the axis of impeller rotation is eccentric in relation to the vessel axis by a distance, x, having a value of 0.25 - 0.6 d.

11. Apparatus for mixing particulate aluminium fluoride-containing material with molten aluminium to remove dissolved contaminant alkali metals and alkaline earth metals from the molten aluminium, said apparatus comprising (a) a cylindrical vessel, having a vertical geometric axis and an internal diameter D, for containing a body of molten aluminium to a height H above the floor of the vessel; said vessel being essentially free from internal baffles and (b) a cover for said vessel supporting a multi-bladed impeller and means for driving said impeller about a vertical axis and means for rotating the impeller, said impeller having a diameter, d, and its blades having a height, h, the midpoint of said blades being spaced above the floor of the vessel by a distance, y, the axis of impeller rotation being spaced from said geometric axis by a distance x, and said blades having major surfaces pitched downwardly at an angle .theta. to the vertical;
(c) the values of d, D, h, H, x and .theta. being such that d/D is between 0.1 and 0.6, h/H is between 0.1 and 0.7, x is between 0.1 - 0.25 D, y is between 0.25H
and 0. 75H, and .theta. is between 0° and 45°;
(d) said cover also supporting means for feeding said particulate material and for discharging fumes.

12. Apparatus according to claim 10, further characterised in that d/D is between 0.15 and 0.40, h/H is between 0.2 and 0.409 x is 0.25 - 0.7 d, y is between 0.4H and 0.6H, and .theta. is between 30°
and 40°.