2836 MARTIN: RESEARCHES ON SILICON COMPOUNDS, PART VI.CCLXV1.-Researches on Silicon Compounds. Part VI.Preparation o j Silicon Tetrachloride, DisiliconHexachloride, and the Higher Chlorides of Siliconby the Action of Chlorine on 50 p e r cent.Perrosilicon, Together with a Discussion on Theiq-Mode of Formation.By GEOFFREY MARTIN.THE method once exclusively used for preparing silicon tetra-chloride in the laboratory was Oersted's process (Ann. Phys.Chem., 1825, [ii], 5, 132) of passing chlorine over a red-hot mixtureof silica and carbon.A great improvement was introduced by Gattermann andWeinlig in 1894 (Ber., 1894, 27, 1943; see also Gattermann andEllery, Ber., 1899, 32, 1114), when they passed chlorine overcrude silicon contained in a glass tube heated to 300-310°.Theyobtained by this process a producb containing 80 per cent. of silicontetrachloride, 20 per cent. of disilicon hexachloride, Si,Cl,, and0.5 to 1 per cent. of trisilicon octachloride, Si,Cl,.Up to the present time this has proved by far the easiwt methodof preparing, not only silic~n tet'rachloride, but also disilicon hexa-chloride and the higher chlorides of silicon.The cost, however, of preparing disilicon hexachloride inquantity by this method is almost prohibitive, and consequentlydisilicon hexachloride is sold a t an extremely high price.As some kilos. of this costly product were necmsary in orderto carry out some of the research work on which t,he author iMARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI. 2837engaged, it became necessary to devise some cheap and convenientmethod of preparing disilicon hexachloride in quantity.After much preliminary work it was found that commercial 50per cent.ferrosilicon, such as is used for refining steel, can be usedinstead of the expensive silicon itself for the purpose of preparingdisilicon hexachloride in quantity, and that this material formsby far the most economical known method of preparing silicontetrachloride.I n fact there can be no doubt that, in future, 50 per cent. ferro-silicon must prove the starting point for the preparation of allchlorinated compounds of silicon.*Consequently, a description of the method of using this materialfor preparing silicon tetrachloride, disilicon hexachloride, and theother chloridee of silicon, will be of some value to other workerson silicon compounds, as the optimum conditions were onlyarrived a t after much troublesome experimenting and many pre-liminary failures.I n order to give some idea of the extent to which the difficul-ties attached to the preparation of these highly hygroscopic liquidchlorides of silicon were overcome by means of the final form ofapparatus described below, the author may state that he succeededin isolating 3 kilos.of pure disilicon hexachloride, 200 grams oftrisilicon octachloride, and more than 54 kilos. of pure, silicontetrachloride by passing 143 kilos. of chlorine over 50 kilos. offerrosilicon. This was done in an ordinary chemical laboratory,and all leakage of chlorine and of the volatile silicon tetrachloridewas so completely overcome, in the final stages of this preparationthat ordinary class work went on around the apparatus while thepreparation was actively proceeding.Since water causes the instant decomposition of these chlorides,all traces of atmospheric moisture must be carefully excluded fromall parts of the apparatus, and it is this necessity that makes theirpreparation a matter of so much trouble.The large amount of disilicon hexachloride prepared by thisprocess enabled the author to obtain it in a state, of very greatpurity, and so he was able to investigate its properties more care-fully than wa6 possible with earlier workers.It was found, forexample, that Gattermann and Weinlig’s value for the boilingpoint of disilicon hexachloride, Si2C16, namely, 145-146O, wasundoubtedly a little too high, the true boiling point being144--145*5O/760 mm.The boiling points of the substance under* The silicon purchased as 50 per cent. ferrosilicon is nearly eighteen times cheaperthan when purchased as pure silicon ; moreover, 50 per cent. ferrosilicon is readilyavailable, the other grades being made only on the small scale2838 MARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI.pressures ranging from 12 mm. to ordinary atmospheric pressurewere also accurately determined. They are tabulated on p. 2852.The density of pure disilicon hexachloride was found to beDi5 1.5624; Troost and Hautsfeuille gave D 1.58.The refractive index for sodium light (D line) was found to be1.4748 a t 18O.Gattermann and Weinlig (Zoc. cit.) give the re-fractive index for “red light” as 1’45.It was also shown that although a t the ordinary temperaturedisilicon hexachloride does not combine with chlorine (althoughthe latter is very soluble in it), yet a t about 300° it takes fire inthis gas and burns to silicon Getrachloride, thus, Si2C1,+C?l,=2SiC1,.This is a new fact of considerable importance, since it throwslight on the mode of formation of disilicon hexachloride by theaction of chlorine on silicon or ferrosilicon (see below).Although Gattermann and Weinlig (Zoc. cit.) showed that whenwater a c b on disilicon hexachloride there is produced silico-oxalicacid, (Si02H),, yet they Beem to have overlooked the fact thatthere are also soluble colloidal forms of silico-oxalic acid produceda t the same time, as the author proved in the course of this work.These colloidal forms are t o be investigated.After the disilicon hexachloride had distilled, about 200 gramsof crude trisilicon octachloride passed over.This, after carefulfractionation, yielded a b u t 150 grams of pure octachloride, whichboiled a t 210-213O under atmospheric pressure (Gattermann andWeinlig, Zoc. cit., give 210-215O, and Besson and Fournier, Cornpt.rend., 1909, 148, 840, give 215-218O. This value is undoubtedlytoo high). However, it was shown that trisilicon octachloride gradu-ally decompowd when distilled under the ordinary pressure, givingrise t o a dark-coloured residue.To avoid decomposition it wasfound advisable to distil it under diminished pressure. It couldbe repeatedly distilled without decomposition ati pressures below110 mm. (when it boiled a t about 149O). Trisilicon octachloride isconsiderably less &able than disilicon hexachloride. The density isDY 1-61, and the refractive index (D ‘line) 1.5135 a t 14.5O.After the trisilicon octachloride had been removed the liquidremaining was €ractionated under greatly diminished pressure, andhigher chlorides were isolated in small quantities. These weredecomposed by water, giving rise to white, amorphous products,which dissolved in alkalis with the evolution of hydrogen. Alater communication will be made on this subject. Besson andFournier (Compt.Tend., 1909, 148, 839; 149, 34) recentlydescribed higher chlorides, which were isolated by a differentmethodMARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI. 2839The residues left after removal of these chlorides consisted of( a ) about 13 grams of a viscid, black, tar-like mass and ( b ) about160 grams of a black powder like animal charcoal. These productsare now being investigated.It is thus shown that the product obtained by the action ofchlorine on silicon and ferrosilicon is no simple substance, but avery complex mixture of silicon compounds, the different compon-ents of which are now in process of isolation.Gattermann and Weinlig (Zoc. cit.) explained the formation ofdisilicon hexachloride when chlorine passes over silicon a t 300’ byassuming that the chlorine first directly unites with the silicon toform silicon tetrachloride, thus :Si + 2C1, = SiC1,.’Next they supposed that the silicon tetrachloride thus formed a tonce reacts with more silicon to produce the hexachloride, thus:3SiC1, + Si = 2Si2C1,.For the last twenty years this explanation of Gatkermann andWeinlig has been universally accepted as the correct one.However,this explanation is certainly quite erroneous for the simple reasonthat a t the low temperatures employed by Gattermann and Weinlig,and also by the author in the preparation of disilicon hexachlorideby the action of chlorine on ferrosilicon, silicon tetrachloride doesnot react with silicon to produce disilicon hexachloride in noticeablequantity. This is conclusively shown in the experiments quotedbelow. Gattermann and Weinlig made not the slightest attempt toverify their theory experimentally.Indeed, on theoretical groundsthis formation a t a low tempetrature of disilicon hexachloride fromsilicon and silicon tetrachloride would appear to be most improb-able since disilicon hexachloride is an endothermic compound, andits formation requirm the absorption of a considerable amount ofheat.A white heat would favour its formation (as in the similar case ofnitric oxide), but a low temperature would not be expected t o actin this way. Troost and Hautefeuille showed that a t a temperatureapproaching the fusing point of porcelain (that is, a t a white heat)the formation of disilicon hexachloride from silicon and silicontetrachloride does take place to a limited extent (Ann.Chim. Phys.,1876, [v], 7, 459), but the conditions under which the formationtakes place in Troost and Hautefeuille’s experiments are entirelydifferent from those under which it occurs in Gattermann andWeinlig’s experiments, where the temperature is kept quite low,and also in the author’s experiments, whereby disilicon hexachlorideis produced by the action of chlorine on ferrosilicon, where also onl2840 MARTIN: RESEARCHES Oh' SILICON COMPOUNDS. PART VI.low temperatures are employed, so that arguments derived fromTroost and Hautefeuille's experiments do no€ apply in any way t othe case now under consideration.The problem to be explained is how a large proportion, amount-ing to something like 20 per cent.of disilicon hexachloride, isproduced by the action of chlorine on silicon a t the low tempera-tures (about 300O) employed by Gattermann and Weinlig.The conclusive refutation of Gattermann and Weinlig's theory isgiven by the following experimental facts established by the author.(1) When silicon tetrachloride is distilled over either silicon orferrosilicon heated to any temperature between 200° and 340" nonoticeable1 amounts of disilicon hexachloride can be detected inthe resulting silicon tetrachloride ; in other words, disilicon hexa-chloride is Fot formed by the action of silicon tetrachloride onsilicon a t moderately low temperatures, as Gattermann and Weinligsupposed.(2) Silicon tetrachloride was prepared by allowing chlorine toact on ferrosilicon at one end of a long tube packed with ferro-silicon, and the silicon tetrachloride thus produced was passedover the long length of heated ferrosilicon in the later portionsof the same tube.Less disilicon hexachloride was found to bepresent in the resulting silicon tetrachloride than when only shortlengths of ferrosilicon were used. According to the Gattermannand Weinlig theory, the longer the length of ferrosilicon traversedby the vapour of the silicon tetrachloride the better would be theopportunity for the reaction, SSiCl, + Si = 2Si2C1,, to proceed, sothat an increased yield of disilicon hexachloride should have re-sulted.(3) Moreo'ver, the lower the temperatures employed the higherthe yield of disilicon hexachloride.For example, when the tubescontaining ferrosilicon were kept a t 180-200°, in some cases morethan 8.6 per cent. of disilicon hexachloride was produced; a t250-260° about 4.6 per cent., whilst a t 300-310° only about4 per cent. was obtained. This should not be the case if theGattermann-Weinlig theory be correct.It is therefore obvious that some other explanation of theformation of disilicon hexachloride and trisilicon octachloride mustbe sought for.The.theory now advanced, which explains all the known factgin a satisfactory manner, is the following:Ordinary silicon (and also the metallic silicides) consists ofcomplex chains of silicon atoms directly united together. Thefirst action of chlorine on silicon (or metallic silicides) is, con-sequently, a complex one.The chain of silicon atoms is noMARTIN: RESEARCEES ON SILICON COMPOUNDS. PART VI. 2841immediately disrupted by the chlorine, but there are first pro-duced complex chlorinated products still containing chains ofsilicon atoms directly united. These complex chlorides are thenattacked by more chlorine, and decompose into simpler chlorides,such as Si6Cl14, Si5Cl12, Si4Cl10, Si3Cll0, and Si,Cl,, all of whichhave been isolated. Lastly, the chlorine then attacks thesechlorides and produces therefrom silicon tetrachloride. Thussilicon tetrachloride is not the first product of chlorination (asGattermann and Weinlig supposed), but rather is the final pro-duct of chlorination, as indicated in the following scheme :I II I I II I II C1 I c1 c1 c1I I I1 I I II I-Si- C1-Si-C1 SiCI, SiC1, SiC1,-Si- Cl-Si-Cl C1-Si-CI SiC1, SiC1,-Si- C1-Si-Cl SiC1, SiC1, S iC1,-Si- --+ C1-Si-C1 -+ Sic], --+ SiCI, --+ SiCI,-Si- C1-Si-Cl C1-Si-CI SiC1, SiC1,-Si- C1-Si-Cl SiC1, SiC1, SiC1,Chain of silicon First stage of Second stage Third stage Final stageatoms in crude chlorination. showing how showing the s h o w i n gsilicon.the chain of production the com-silicon atoms of disilicon pletechlor-is broken hexachloride ination ofdown by the and silicon the chainschlorine. tetrachloride. of siliconatoms t oSicloConsequently, silicon tetrachloride is the main product of theaction, but small quantities of complex chlorides still containingdirectly linked silicon atoms remain in the silicon tetrachloride,thus indicating its mode of origin.The amount contained in the silicon tetrachloride of disiliconhexachloride, which possesses only two silicon atoms directly linked,is much greater than the amount of chlorides containing longerchains of silicon atoms, such as trisilicon octachloride, because thelonger chains of silicon atoms are the first to be broken under thefurther action of the chlorine.The same considerations apply to the case of the metallicsilicides, and the fact that silicides, such as 50 per cent.ferro-silicon, can be used for preparing disilicon hexachloride and higherchlorides containing directly linked silicon atoms (see above), maybe taken as evidence that in these metallic silicides chains ofdirectly linked silicon atoms are present, and that the smallamounts of higher chlorides containing directly linked siliconatoms produced in theIr chlorination contain parts of the unbrokenchains of silicon atoms which were originally present in thes2842 MARTIN: RESEARCHES ON SILICON COMPOUNDS.PART VI.silicides, but which have been for t,he most part broken down, bythe further action of chlorine, into silicon tetrachloride.The formation of disilicon hexachloride from f errosilicon, forexample, would be very simply explained thus :SiC1,S~CI,’Fe<Si si -.+ c1 FeCI, -I- IThe fact seem6 definitely established that many of these silicides,such as ferrosilicon, are simply mixtures of complex silicides, andin many cases there is reason to believe that the silicides are not‘(compounds” at all, but are merely solid solutions of metal insilicon.The following facts are in favour of this view of the formationof the higher chlorides of silicon:(1) Complex chlorides of silicon are invariably produced whenchlorine acts on silicon or silicides, but (as shown above and below)their formation cannot be accounted for by the action of siliconon silicon tetrachloride.(2) That the disruption in the presence of chlorine of directlylinked silicon atoms, such as exist in disilicon hexachloride, intosilicon tetrachloride does actually take place was proved by pass-ing vapours of disilicon hexachloride mixed with chlorine througha tube heated to about 300°, when the disilicon he’xachloride caughtfire and burnt to silicon tetrachloride, thus : Si,Cl, + Cl, = 2SiC1,.It was also proved that at the ordinary temperature disilicon hexa-chloride does not combine with chlorine to form silicon tetra,-chloride.It is therefore practically certain that the small amounts ofsilicon hexachloride, octachloride, etc., found in the silicon tetra-chloride produced by chlorinating silicon or ferroeilicon, aresimply the residue or debris of much larger quantities of siliconhexachloride or octachloride originally present, this residue havingescaped destruction by the chlorine present owing to the fact thatit was quickly removed from the sphere of action of the latterby quick cooling.(3) This theory also accounts ~~atisfactorily for the fact that thelower the temperature a t which chlorine acts on ferrosilicon thegreater is the yield of disilicon hexachloride and other chloridesof silicon containing chains of directly united silicon atoms in themolecule.On the Gattermann-Weinlig theory the reverse effect wouldrather be expected to take place.(4) Next to carbon, silicon is the element having the most highlydeveloped power of self-combination.Elementary silicon, there-fore, cannot be regarded as a mere aggregate of single silicoMARTIN : RESEARCHES ON SILICON COMPOUNDS. PART VI. 2843atoms, but rather the element must be thought of as composed ofcomplex molecules consisting of many silicon atoms directly unitedtogether, possibly in rings or chains, as in the case of carbon.Itwould, therefore, appear to be unreasonable to suppoee that at themoment the chlorine atoms act on these molecules they immedi-ately fly to pieces with the production of single atoms of silicon,which are hhen acted on by the chlorine to produce silicon tetra-chloride. It is more reasonable to’ suppose that the chlorinationtakes place slowly and the whole complex silicon molecule isattacked, forming first complex chloro-compounds, which later, bythe further action of the chlorine, break down into simplechlorides, the end-product of the whole complex chain of eventsbeing silicon tetrachloride. I n this connexion there is the curiousexperimental fact (see p. 2847) that when chlorine is passed overferrosilicon heated to a suitable temperature, the formation ofsilicon tetrachloride does not begin at once.First of all, a periodelapses (which may last any time from thirty minutes to one hour,or even, under special and not well-understood conditions, to threehours) during which no silicon tetrachloride at all passes over, butin which it is possible that the surface of the ferrosilicon is beingacted on with the production of complex chlorinated silicon com-pounds, by the further chlorination of which the silicon tetra-chloride is produced.Once this initial stage is over, the formation of silicon tetra-chloride takes place with great rapidity.EXPERIMENTAL.Preliminary Experiments.-Some preliminary experiments werecarried out by placing 50 per cent.ferrosilicon in g1,ass tubes andpassing chlorine through them while the latter were heated tovarious temperatures in a Gattermann’s bomb furnace.It wassoon found, however, that glass tubes were quite unsuitable foruse with ferrosilicon. I n the first place, the ferric chloride pro-duced by the chlorination of the iron sublimed down the tube,and soon caused it to block up, and the tubes usually broke whenbeing cleaned out. Moreover, unless the stream of chlorine isvery carefully regulated, the temperature of the reaction may riseso high that the glass may fuse a t certain points.The we of glass tubes, therefore, was abandoned in favour ofordinary iron gas-piping, 30 mm. bore, fitted a t the end withordinary corks.It was found that the iron piping was soon burntthrough by the chlorine a t that end of the tube where the chlorineentered and began t;o react with the ferrosilicon. The expedientA t first a temperature of 300-310° was maintained2844 MARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI.of placing the ferrosilicon on movable iron troughs inside t'he tubewas tried, but was abandoned, as the troughs stuck firmly to theinside of the tube (owing to the ferric chloride acting as a lute),and the latter could not be effectively cleaned out.Although, when the temperature of the furnace was maintaineda t 3OC-310°, the iron tube was very rapidly attacked by thechlorine, it was found that by employing a lower temperature thecorrosive action of the chlorine was very much diminished. A tem-perature of 180--200° was found to be very suitable.A tempera-ture of 170° caused the action of the chlorine on the ferrosilicon tobecome so slow that it was abandoned in favour of the higher tem-perature. Moreover, by keeping the temperatures low, the yield ofdisilicon hexachloride (the substance i t was desired to obtain inquantity) was practically doubled. Thus in the initial experiments,when the furnace was maintained a t 300-310°, the yield of disiliconhexachloride in the crude silicon tetrachloride was about 4 per cent.When the furnace was kept a t 250--260° the yield of disiliconhexachloride rose t o 4.6 per cent., whilst a t 180-200° the yieldrose to 8.6 pe'r cent. A t 170° the action of chlorine on ferrosilicontook place too slowly for effective work.Apparatus for the Production of Silicon Tetrachloride andDisilicon Hexachloride in Quantity.As the result of these preliminary experiments an apparatus forproducing silicon tetrachloride in quantity was built up.I n itchlorine from a cylinder was dried by passing through sulphuricacid, and then, by means of a T-piece, was led alternately throughtwo iron tubes set in a Gattermann bomb iurnace and charged withferrosilicon. The ends of the tubes were fitted with ordinary corksthrough which glass leading-tubes passed.The corks were best coated with paraffin-wax o r bakelite varnish.A plug of glass-wool a t the far end of the tubes arrested the ferricchloride, which slowly distilled down the tube, and tended to blockup the leading tubes.The silicon tetrachloride was collected in aWinchester bottle.This apparatus worked well for the production of a few kilos.of silicon tetrachloride and a few hundred grams of disilicon hexa-chloride. Whea, however, it became necessary to prepare about50 kilos. of silicon tetrachloride, so that about 3 kilos. of disiliconhexachloride, could be isolated therefrom, grave defects soonrevealed its& in the apparatus.I n the first place the corks needed constant replacement andrepair even when well coated with bakelite varnish. Moreover,they often become impregnated with disilicon hexachloride and thMARTIN : RESEARCHES ON SILICON COMPOUNDS. PART VI. 2845higher chlorides, which on contact with atmospheric moisture soondecomposed with the production o€ explosive silicon oxy-compounds.Consequently, the corks, after a time, became unpleasant to handle,since in forcing them into or withdrawing them from the tubeexplosions occurred, which, when much disilicon hexachloride hadaccumulated, could cause injury to the hands.It should be notedthat all these compounds of silicon which contlain silicon atomsdirectly united in the chain are, apparently, formed with theabsorption of heat, and so are capable of explosion under suitableconditions.Frequently in the course of the experiment the corks blew offwithout warning with considerable violence, either as the result ofa sudden blockage in the tube by the sublimed ferric chloride, orpossibly, on certain occasions, by the presence of excess of chlorineexplosively causing the ignition of the higher chlorides formed inthe tube (see p.2859).Leakage of chlorine from the corks could only be prevented withgreat difficulty, and as about 143 kilos. of chlorine were requiredfor the production of the amount of silicon hexa- and tetra-chloridesneeded, the leakage of chlorine became serious. It was, there-fore, essential t o devise chlorine-tight end-pieces, and after someexperimenting, the corks were finally displaced by detachable ironcaps screwed on to the ends of the tubes.The threads were made gas-tight a t first by the use of ordinaryyellow soap (which acted admirably as a lute for chlorine), butlater fine fibres of asbestos introduced into the threads were foundto act better. The difficulty of chlorine leakage was thus sur-mounted, and the ends of the tube could be screwed off with easeand the tube withdrawn, washed out, and recharged when this wasnecessary.Since blockages in the tubes invariably occurred after a certaininterval of time, explosions could easily arise unless the pressureprevailing inside the tubes was properly controlled.This was doneby attaching mercury manometers t o the mouths of the tubes, anyincrease of pressure inside the tubes being indicated by the rise ofmercury in these manometers, which also acted as safety valves.Lastly, it was necessary to absorb chlorine and other corrosivevapours which passed through the apparatus. This was very effi-ciently done by means of a lime absorber.It consisted of a woodenbox, 100 cm. long by 70 cm. wide, fitted with shelves so shapedthat the chlorine passed over them in a zigzag fashion. The wastegases entered a t the bottom, and escaped into the flues a t the topthrough holes of about 3 cm. diameter bored in the shelves. Thelime had to be changed every week, the face of tdhe box being soVOL. cv. 8 2846 MARTIN: RES~ARCEIES ON SILICON COMPOUNDS. PART vr.arranged that the front opened like a door, so that the shelvescould be cleaned out when necessary. The box was made chlorine-tight by luting with ordinary yellow soap. The complete apparatusis Ejhown in Fig. 1. A and A1 are two chlorine cylinders, eachholding about, 42.5 kilos. of liquid chlorine.13 and B1 are the twogeries of sulphuric acid wash-bottles, whilst D and B1 are the two3-cm. gas-pipe ifon tubes, about 120 cm. long, fitted with ironieadiag pipes attached to iron screw-on caps, C and C1. F and F1aye the two mercury manometers attached by means of glass andrubber tubing to the leading tubes of D and D1, so that the pres-sures prevailing inside the pipes D and D1 are accurately known,an increase of pressure being indicated by the rise of mercury inP and F1.The mercury in the reservoirs of F and 3’1 is covered with a layerFIG. 1.of concentrated sulphuric acid, which protects the mercury to alarge extent from vigorous attack by the chlorine gas. 2 is a Gatter-mann bomb furnace, the internal temperaturs of which is indicatedby the thermometer K.In these experiments the tlemperature waskept on the average between 180° and 200O. The tubes D and Dlend in iron screw-on elbow joints E and El, from which a piece ofiron pipe of 12 mm. bore projects into the two receiving vessels,Q and Q1. These consisted of two 2-litre filter-flasks, in whichalmost all the crude silicon tetrachloride passing over from theapparatus condensed, only a very small amount passing awaythrough the condensers ill and MI and collecting in tlie bottle N .An air-pump P provided with a mercury trap R and a dryingtube S is directly united t o the flasks Q and Q’, and by forcing airinto Q and &1 (after closing sundry clips controlling the exits froMARTIN: RESEARCEES ON SILICON COMPOUNDS. PART VI.2847the flasks) the crude silicon tetrachloride could be forced up thetubes X and X1 (which reach almost to the bottom of the filter-flasks @ and Ql), along the pipe X X into the fractionating columnT, and thus into the fractionating flask U, which is heated on awater-bath 7.The fractionating column employed was a Young's three-bulbevaporator still-head, which was fused on to a litre flask. Theleading tube from the top of the distilling column T passes througha condenser H into a receiver J (a Winchester bottle). From J aleading tube L runs to the catch-bottle N , and thence a wide lead-ing tube runs to the lime absorber WW. Here any chlorine andsilicon tetrachloride vapours enter a t the bott'om, and after circu-lating over the lime on shelves in zigzag fashion, escape a t the topinto the flues.The absorption of chlorine and silicon tetrachlorideby the apparatus was almost complete.The iron piping was packed with coarsely crushed ferrosilicon,which commenced a few em. in front of the place where the tubesenter the furnace, and extended for some 20 to 30 em. down thetube, about 1 to 2 kilos. of ferrosilicon being used for each charge.This short length of ferrosilicon was found advantageous in thatthe chlorine was found t o be completely absorbed by the shortlayers, and a, longer length only tended to cause the tube to blockup owing to the sublimed iron chloride condensing in the ferro-silicon a t the far end of the tube. With short lengths of packinga considerable space was left in which the iron chloride could con-dense without choking the tube.The method of working the apparatus was as follows: The tubesDD and DID1 were charged with ferrosilicon, placed in the furnace,and their temperature was raised to 180-200°, the temperaturebeing indicated by the thermometer K.Chlorine was then admittedin a moderately rapid stream, the rate being controlled by thescrew valves attached t o the cylinders A and Al.The action does not take place immediately. Usually the silicontetrachloride begins t o pass over in about thirty minutes afterstarting the operation. Occasionally, however, it was found thatan hour's, two hours', and in some cases three hours' passage of thechlorine through the heated tube was necessary before the silicontetrachloride began to pass over in quantity.Temperature did not ceem to play a great part in shortening thelength of this preliminary period, as even when the furnace wasmaintained a t as high a t'emperature as 300-310° the same pheno-menon was observed, and keeping the temperature a t 190-200'did not cause a sensible prolongation of this period of waitingbefore the silicon tetrachloride began to pass over.Also a very8 2 2848 MARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI.rapid initial stream of chlorine did not shorten sensibly the initialperiod.This effect is, possibly, due to the fact that the chlorine musthave time t o attack the surface of the ferrosilicon and produce inter-mediate complex silicon chlorinated compounds before the produc-tion of silicon tetrachloride takes place.When, however, the actionstarted it took place moderately rapidly with a considerable riseof temperature, and the silicon tetrachloride passed over in a steadystream and collected in the vessels Q and Q1 as a yellow, fumingliquid.The tubes are worked for about three hours a t a time eithersimultaneously or alternately, but as a rule matters were soarranged that one tube was in full action whilst the other one wasbeing pulled out and re-charged. The silicon tet'rachloride con-densed in the filter-flasks Q and Q 1 , and was from time t o timeforced by the pressure of the chlorine from the cylinders up thepipes X and X 1 into the fractionating flask U . When this wasdone the screw clips were once more opened and the action con-tinued.When, for any reason, it was inconvenient to use thepressure of the chlorine from the cylinders A and A1 for forcingthe liquid from Q into U , air-pressure applied by the air-pump Pwas used for this purpose.The silicoil tetrachloride collecting in U was then fractionallydistilled, the distillate being collected in the Winchester bottle J .J when filled is removed and replaced by another Winchesterbcttle. The silicon tetrachloride is stored in them bot'tles, usingordinary corks well boiled in paraffin wax, the corks being takenout of the paraffin bath and while still warm being forced intothe neck of the bottle, and covered over with a layer of meltedparaffin, so as to prevent any danger of atmospheric moisture reach-ing the silicon tetrachloride.When sealed in this way the silicontetrachloride can be stored for months without depreciation.When the flask U became nearly full of residues of high boilingpoint, the water-bath was replaced by an oil-bath heated to ahigher temperature.The disilicon hexachloride passed over a t 147--149OJ and wascollected separately and fractionated in a separate flask, also pro-vided with a fused-on Young evaporator still-head. There wasthus left behind in U a gradually increasing amount of residuesof high boiling point, which were later proved to consist of higherchlorides of silicon (see below), besides a mass of tarry materialand a black, solid residue, much like animal charcoal in appelar-ance.Uothod of Charging and Discharging the Tubes.-When it waMARTIN: RESEARCBES ON SILICON COMFOUNDS. PART VI.2849observed that the supply of silicon tetrachloride dropping into Qor Q1 from one of the tubes diminished, or when the pressure inthe tube began to increase rapidly (as indicated by the mano-meters F or FI), it wae known that either the tube was becomingexhausted, or that i t was becoming choked up by sublimed ferricchloride. This occurred, on the average', every three hours. Con-sequently, it became necessary at the end of this time to withdrawthe tube, clean it out, recharge it with ferrosilicon, and replacei t in the furnace.To do this, the supply of chlorine is cut off from the tube, thenthe end caps C and E are rapidly unscrewed (these caps are coldenough to be touched by the hand, since they project nearly 30 cm.from the furnace), the tube is drawn out over iron rollers (notshown in the illustration), then, while hot, rapidly transferredto a sink, a cmk fitted with a leading tube is attached to oneend (after first withdrawing the plug of glass wool), and a supplyof cold water allowed to flow into the tube.This water, enter-ing the tube, is soon heated to boiling, and largely convertedinto steam, which blows the contents of the tube through the openend of the tube and effectively and rapidly cleans out the ferricchloride in the tube. The stream of cold water is allowed to flowthrough the tdbe until it runs clear, when the ferric chloride hasbeen completely removed.The iron rod with a pointed end isthen applied to clear out any particles of ferrosilicon still adheringto the tube, and the wet, clean tube is then transferred to a com-bustion furnace and dried by heating, while a current of air isblown through it. The tube is then removed from the combustionfurnace, rapidly charged with 50 per cent. ferrosilicon, as describedon page 2847, and replaced in the furnace (being run in over theiron rollers above-mentioned), the caps a t C and D are screwed on,and the stream of chlorine is once more let into the apparatus.As each iron tube corroded very rapidly just at one point,namely, a t the end where the chlorine enters and acts on the ferro-silicon, in recharging care was taken to place the charge a t theend of the tube opposite to that end previously used, so that thecorrosion should take place equally a t each end.However, evenwith this precaution, the average life of each iron tube was notmore than six experiments, the chlorine burning a hole throughthe iron piping a t the point where the action was mostintense.The 50 per cent. ferrosilicon washed out of the tube is wellwashed with water until free from iron chloride, dried in an air-oven, and once more was used for recharging a second tube. Theferrosilicon was thus used over and over again until consumed2850 MARTIN : RESEARCHES ON SILICON COMPOUNDS. PART VI.A t thO same time considerable wastage occurred, owing to particlesof the ferrosilicon escaping as a suspension in the washing water.Too finely powdered ferrosilicon was found not to be suitablefo,r use, as blockages in the tube were thereby easily occasioned.Since traces of higher chlorides accumulate at the ends of thetube and are converted by the washing water into’ explosive oxy-compounds, these, on striking with the iron rod or on scraping,may-explode if in considerable amount, so that a certain amountof care must be taken not to allow these residues to accumulateto any great extent in the tubes.This apparatus, which was gradually evolved out of repeatedfailures and mishaps, worked very smoothly and efficiently, andby means of it about 54 kilos.of silicon tetrachloride were p r epared by the passage of 143 kilos. of chlorine gas over about50 kilos.of 50 per cent. ferrosilicon.From the crude silicon tetrachloride, which was distilled asfast as it was produced in the continuous fractionating apparatus,there were produced about 3 kilos. of disilicon hexachloride andabout 500 grams of residues which contained nearly 200 grams oftrisilicon octachloride.Apparatus for Distilling the Crude Silicon TetrachZom’de so as toSeparate the Higher Chlorides.The problem of dealing with large quantities of the highlyvolatile silicon tetrachloride (b. p. 59O) is complicated by the factthat atmospheric moisture decomposes it, with tlhe formation ofhydrochloric acid and the deposition of silicic acid.Consequently, all the vessels in which the liquid is kept mustbe most carefully dried before allowing the silicon tetrachloride toenter.A description of the apparatus used for distilling and fraction-ating this silicon tetrachloride in the absence of atmosphericmoisture may prove useful to other workers, as the final form wasarrived a t only after much troublesome experimenting.Theapparatus finally used is shown in Fig. 2.A is a vessel filled with coarsely granulated calcium chloride.To this vessel an air-pump, P, is att’ached, connexion being madethrough a mercury trap, T, to prevent the silicon tetrachloridevapours reaching the pump. By this means dry air can be forcedinto the Winchestm bottle, B, contlaining the crude silicon tetra-chloride to be dist4illed. As a result of this increased air-pressure,the crude silicon tetrachloride is forced up the tube K K into thefractionating flask D, which is fitted with a Young’s three-bulbevaporator still-head, C, fused on to D.A t the top of t’he stillMARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI. 2851head C is a thermometer, M . The leading tube from the still-head passes through a condenser, N , as shown. The flask D isplaced on a water-bath, H , and after the proper amount of crudesilicon tetrachloride has been forced into D from the reservoir B,it is fractionated, the residual disilicon hexachloride (b. p. 1 4 5 O ) ,together with the higher chlorides, remaining behind in D, whilsttlhe volatile silicon tetrachloride (b. p. 59O) passes up the columnC and, condensing in N , runs into the receiver E.F is a catchFIG 2.for any silicon tetrachloride vapour that does not condense in E,C being an auxiliary condenser. I n hot weather F should beimmersed in ice. P leads out to the flues.The bottle B, the distilling column C, the bottle E, and theflask F are all fitted with corks which have been boiled in paraffinwax.By means of this apparatus many kilos. of silicon tetrachloridecan be continuously distilled free from contact with atmosphericmoisture, and separated from the residues of high boiling point,which thus accumulate in the flask D2852 MARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI.Pressure. Pressure.mm. B. p. mm. B. p.12 40" 41 61"12.5 43 50 6513 46 53 6614 47 90 8017 48.5 95 8119 49 105 8420 50 110 8622 53.5 119 8927 56 122 9031 60 126 91 .As the liquid concentrates in I) it becomes dark brown, almostblack, whilst the liquid in E consists of almost pure silicon tetra-chloride containing a little dissolved chlorine.It is of a yellowishcolour, but can be rendered colourless by allowing i t to remain forsome days in contact with freshly-ignited animal charcoal, followedby redistillation.When a sufficient quantity of residues has collected in D, thewater-bath, H , is replaced by an oil-bath, and the residues aredistilled, the bulk of the material passing over (after the silicontetrachloride has been removed) a t 147-148O under the ordinaryatmospheric pressure. There remains in I) some black fluid resi-dues, and a black powder resembling animal charcoal.Pressure.mm.B. p.130 92"135 93140 94150 95181 98195 101200 102222 103.5760 144-145.5Redistillation of the Crude DisiZicon Hexachloride.The crude disilicon hexachloride distilled, as above described,from the residues was now purified by keeping it over freshlyignited animal charcoal, and was then fractionally distilled from anapparatus made entJrely of glass, the fractionating column employedbeing a Young's three-bulb evaporator still-head fused to the flask.During this operation the most rigorous precautions had to be takento dry most thoroughly all the vessels used in the distillation, other-wise a turbid distillate would result.Ordinary drying by washing out with alcohol followed by etherand blowing warm air through the apparatus was not efficientenough. The flasks had to be heated nearly to redness (after thepreliminary washing with alcohol and ether), and then, after blow-ing out with hot, air, must be attached still fairly hot t o the receiver.The pure substance boils a t 144-145*5°/760 mm.The boilingpoint, 145--146O, given by Gattermann and Weinlig (Zoc. cit.) isundoubtedly a little too high. The boiling points under diminishedpressures were also determined as follows MARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI. 2853The density of disilicon hexachloride determined with 200‘6684grams was found to be DY 1.5624. Troost and Hautefeuille gave1.58 a t Oo.The refractive index for sodium light (D line) as determined bythe hollow prism method was found to be 1.4748 a t 18O.Anotherdetermination by a drop method gave 1.4775 a t 14.5O. Gatter-mann and Weinlig gave the refractive index for “red light”as 1-45.Although Gattermann and Weinlig showed (Zoc. c i t .) that whenwater acts on disilicon hexachloride, silico-oxalic acid is produced,thus :Si,Cl, + 4H,O = (SiO,H), + 6HC1,in the form of a white precipitate, insoluble in acids, but solublein alkalis with the evolution of hydrogen, yet it seems to haveescaped their notice that soluble colloidal forms of silico-oxalic acidare produced a t the same time. This was proved as follows:disilicon hexachloride was treated with a little water, when awhite precipitate of silico-oxalic acid separated, which was col-lected. The residual clear liquid, however, still contained somesilico-oxalic acid in colloidal solution, as was proved by adding t othe liquid a ferw drops of concentrated ammonia solution, when acmsiderable gelatinous precipitate was obtained, This colloidalform of si1ico:oxalic acid is now being further investigated, and anaccount will be given in another paper.,4 nalysis of Disilicon I€extrddoride.--By means of a small pipettemade of a piece of small-bore glass tubing drawn out a t one endand fitted with a rubber teat a t the’ other end (the whole pipettebeing most carefully dried before use), 1.3029 grams of disiliconhexachloride were transferred to a dry weighing bottle, exactlyweighed, and then decomposed by water rendered alkaline withammonia. The contents of the weighing bottle were finally rinsedout, the precipitated silicic acids collected, the washings exactlyneutralised with nitric acid and titrated with silver nitrate, usingpotassium chromate as indicator. (Found, C1= 78.9.Calc., C1= 78.9per cent.)Isolation and Properties of Trisilicon Octachloride, Si,Cl,.After the crude disilicon hexachloride had been distilled over,there remained in the flask a dark-coloured mass, consisting ofliquid and solid. The liquid was poured into a fractionating flask,and there remained a black, amorphous powder resembling animalcharcoal, and weighing 160 grams. This powder is undergoingexamination.The black liquid, weighing about 377 grams, was now distilled2854 MARTIN: RESEARCHES ON SILICON COMPOUNDS.PART VI.using a rod-and-disk fractionating column fused on to the flask,and heating on it metal bath.After separating the disilicon hexachloride still in the liquidbetween 141O and 147O (atmospheric pressure), the’ temperaturerose rapidly to 170°, and then more slowly to 185O. The weightof the fraction boiling a t 141-185° was 128 grams.The temperature then ros0 rapidly from 185O to ZOOo, when thereceiver was again changed, 64 grams distilling between 1 8 5 O and200O. The bulk of the liquid, amounting to 185 grams, passedover, however, a t 200-220°, a large portion of which distilled a tabout 215-217O.There remained in the flask about 20 C.C. of a dark-colouredliquid boiling at above 220°, which was worked up separately (seebelow).The distillates, which consisted of yellow, fuming liquids, werenow subjected t o careful fractionation under diminished pressure,using a 21 rod-and-disk fractionating column fused on to a glassflask, and heating from an oil-bath.The liquids were easilyseparated into some disilicon hexachloride and trisilicon octa-chloride, the latter being obtained pure after one or two fractiona-tions. The amount of pure trisilicon octachloride was about 150grams, about 30 grams of impure liquid being simultaneouslyisolated.The published accounts of the boiling point of this substancevary considerably. Thus, Gattermann and Weinlig (Zoc. c i t . ) give210-215°, whilst Besson and Fournier (Zoc. cit.) give 215-218O.There is no doubt, however, that Besson and Fournier’s productwas not pure.The boiling point of the above product was210-213O under the atmospheric pressure, agreeing closely withGattermann’s value. The substance slowly decomposes whenboiled under the ordinary pressure, giving rise to a dark-colouredresidue. It is this partial decomposition of the trisilicon octa-chloride that is responsible f o r the fact that the boiling pointis not very sharp under atmospheric pressure. When the liquidis distilled, however, under diminished pressure, no decomposi-tion occurs; the liquid can be repeatedly distilled under dimin-ished pressure to the last drop, without any discoloured residueappearing in the flask. The boiling points were determined asf ollows MARTIN : RESEARCHES ON SILICON COMPOUNDS. PART VI.2855Pressure.mm.17222430475360666972B. p.100"106110113124129133134.5135.5126.5Pressure.111111. B. p.74 137"80 13983 14190 14393 14495 145108 147110 149760 210-213The density was found to be DY 1.61, and the refractive indexfor sodium light (D line) 1.5135 a t 14.5O. Gattermann and Weinlig(Eoc. cit.) gave the refractive index for "red light" as 1.52.The liquid was analysed in the same way as disilicon hexa-chloride (p. 2853). (Found, Cl= 76.99. Calc., C1= 76.97 per cent.)Isolation of Higher Chlorides.After separating the trisilicon octachloride as above describedthere remained in the flask about 20 C.C. of a black liquid whichboiled a t above 220O. This was now subjected to fractional distilla-tion under diminished pressure in a specially constructed smallglass flask, fitted with a rod-and-disk fractionating column, theflask and fractionating column being fused together.After very considerable difficulties, founded principally on thefact that the silicon chlorides must not be exposed to moist air,and the fact that only very small quantities were distilled, a t leastthree distinct substances were separated after repeated distillation :(1) a viscous, colourless liquid (3 grams), boiling a t about 150°/15 mm.; (2) a viscous, colourless liquid (2 grams), boiling a t about190°/15 mm.; and (3) a white, crystalline solid (0.5 gram), whichmelted a t about 21B0, and distilled a t about 210°/12 mm.It wassoluble1 in dry benzene or light petroleum, and could be crystallisedtherefrom.The examination of these substances is being continued.These chlorides, when thrown into water, yield white, amorphousproducts, easily combustible, which are no doubt the higheranalogues of silico-oxalic and mesoxalic acid. These white sub-stances also dissolve in potassium hydroxide t o a clear solutionwith the evolution of hydrogen.The examination of these products is being continued.There remained in the flask a t least 13 grams of a black, viscousmass, like t a r or pitch, which was soluble in ether, insoluble inabsolute alcohol, and evolved hydrogen with sodium hydroxide.The latter product, however, is still undergoing examination2856 MARTIN: RESEARCHES ON SILICON COMPOUNDS.PART VI.Distillatio~~ of Silicon Tetrachloride over Silicon.According to Gattermann and Weinlig (loc. cit.), silicon andsilicon tetrachloride interact as follows :SSiCl, + Si= 2Si2C1,.I n order to test this the following apparatus was employed:FIG. 3.By means of air pressure (from an air-pump) applied to the Win-chester bottle F , pure silicon tetrachloride contained therein couldbe forced into the weighed flask B , which is heated on the water-bath. This silicon tetrachloride was then distilled from the flaskB through an iron pipe C packed with commercial silicon brokeninto a coarse powder and heated in a Gattermann’s bomb furnace A .The iron tube was about 120 cm. long and 3.1 cm. in internaldiameter, with screwed-on iron terminal caps and connecting pieces,the screw-threads being made gas-tight by asbestos packing.Thesilicon tetrachloride passing through the tube Q is condensed in thetwo-litre filter flask D. By means of an air-pump attached to Hthe silicon tetrachloride accumulated in B could (after closingcertain clips connected with the exit tubes from the apparatus) beforced back up the tube EEE into the reservoir F , and thence, ifrequired, up the tube GG back into the flask 23, so that thesilicon tetrachloride in F could be repeatedly distilled over thesilicon in CC as many times as desired.By means of this apparatus 3 kilos. of silicon tetrachloride freefrom disilicon liexachloride were repeatedly distilled over silicoMARTIN: RESEARCHES ON SILICON COMPOUNDS.PART VI. 2857in the tube 0, first of all when the latter was maintained a t ZOOo,then a t 280°, then a t 310°, and lastly a t 340O.However, in no case were noticeable amounts of disilicon hexa-chloride found to have been formed in the distilled silicon tetra-chloride.Hence it is proved that the 20 per cent. yield of disilicon hexa-chloride stated by Gattermann and Weinlig to have been producedby the action of chlorine on silicon at 300-310° could not possiblyhave arisen, as they supposed, from the action of silicon tetra-chloride on silicon.Uistillation of Silicon Tetrachloride over Ferrosilicon.The preceding experiment was repeated, the silicon in the tubebeing now replaced by 50 per cent. ferrosilicon. However, in thiscase, also, no noticeable amounts of disilicomn hexachloride couldbe proved to be produced when 3 kilos.of silicon tetrachloridewere distilled over the mass, even when the tubes were heated to300° and 340O.It was thought that although ordinary silicon tetrachloride whendistilled over ferrosilicon will not give rise t o disilicon hexachloride,nevertheless it might be possible that silicon tetrachloride a t itsmoment of formation might react, with ferrosilicon to produce somedisilicon hexachloride.The easiest way to test this was to pass chlorine over a verylong length of ferrosilicon, so that the silicon tetrachloride pro-duced in the first part of the tube would then react with moreferrosilicon as i t passed in the vaporous condition down the tube.Consequently, the following experiment was carried out :A piece of Jena combustion tube, about 200 cm.long and 2 cm.bore, was drawn out a t one end and bent at right angles. It wasfilled with a long layer of 50 per cent. ferrosilicon in the form ofa coarse powder, and the tube was placed through two bombfurnaces in succession, and was therein heated to 300-310°, whilea slow current of dry chlorine was passed through the tube. Theresulting silicon tetrachloride was received in a distilling flaskimmersed in cold water.It was found that when the action commenced, the chlorine waspractically completely absorbed by the first 15 or 17 cm. of heatedferrosilicon, so that the silicon tetrachloride would have everyopportunity as i t passed over the succeeding lengths of ferrosiliconto react with more silicon to produce disilicon hexachloride.How-ever, the resulting silicon tetrachloride was found on distillationto contain less than 4 per cent. of disilicon hexachlmide. Sincethe average yield of disilicon hexachloride produced by passin2858 MARTIN: RESEARCHES ow SILICON COMPOUNDS. PART VI.chlorine over short lengths of ferrosilicon considerably exceeds this,it is quite certain that the effect of passing silicon tetrachlorideover a long length of heated ferrosilicon is not to increase theyield of disilicon hexachloride. I f anything, i t led to a diminu-tion of the yield.These experiments prove conclusively that the Gattermann-Weinlig reaction certainly does not proceed to a noticeable extenta t low temperatures, and that therefore their explanation of thepresence of disilicon hexachloride and higher chlorides in t,hesilicon tetrachloride produced by chlorinating silicon or ferrosiliconis inadmissible.Action of Chlorine on Didicon Hexachloride.The author’s theory that complicated chlorinated silicon com-pounds are first*,produckd when chlorine acts on silicon or metallicsilicides, and that silicon tetrachloride is foxmed from these by thefurther action of chlorine, was now put t o the test of experiment,and it was definitely poved that chlorine acts vigorously ondisilicon tetrachloride (and no doubt still more vigorously on themore unstable higher chlorides) a t 300-340° so as to break upthe chain of directly united silicon atoms, with the production ofsilicon tetrachloride, thus :Cl,Si=SiCl, + C1, = 2SiC1,.It was proved that disilicon hexachloride burns directly to silicontetrachloride in the presence of chlorine, the experiment beingcarried out as follows :A stream of chlorine (dried by passing through concentratedsulphuric acid) was passed into a 150 C.C.flask containing about50 grams of pure disilicon hexachloride and heated on an oil-bath,the bmperature of which was gradually raised to 165O or 166O.The leading tube of the distilling flask was inserted firmly througha cork a t one end of an iron tube, whilst a t the other end of theiron tub0 there was an iron elbow joint screwed on, which wasfitted with a reducer. A piece of iron piping from this projectedinto a receiving flask, passing in through a rubber stopper. Thereceiving flask had its leading tube projecting into a similar flask,the two latter flasks being immersed in ice.The iron tube was contained in a Gattermann’s bomb furnace,the temperature of which was maintained a t about 340O.A continuous stream of chlorine was now passed through theapparatus. So long as the temperature of the oil-bath in which thefirst flask was immersed was not sufficiently high to cause thedisilicon hexachloride to boil, the t3ilicon tetrachloride was noMARTIN: RESEARCHES ON SILICON COMPOUNDS. PART VI.2859noticed to be passing over rapidly into the receiving flask. Also, ifthe iron tube was kept a t about 120-130° and the disilicon hexa-chloride was allowed to distil through i t in a stream of chlorine, noinflammation followed.If, however, the temperat'ure of the tubewas kept a t 300-340°, as soon as the vapour from the boilingdisilicon hexachloride, mixed with chlorine, reached the iron tubea mild explosion took place, a red flame shot back into the flask,and the disilicon hexachloride could be observed burning with areddish flame all over its surface in the atmosphere of chlorine.When the supply of chlorine was checked the flame rose andburnt round the end of the tube projecting into the flask throughwhich the chlorine entered, the chlorine here burning in an atmo-sphere of disilicon hexachloride vapour. Dense, brown fumesaccompanied the combustion of the disilicon hexachloride, and a tthe same time the silicon tetrachloride produced as the result of thecombustion streamed through the iron tube, and rapidly collectedin the receiving flask, the liquid being of a dark colour.The experiment, however, must be carried out with caution, orit may become dangerous.Although on one occasion about45 grams of pure disilicon hexachloride were burnt to silicon tetra-chloride in this manner with only a mild initial explosion, yet inanother experiment a violent explosion suddenly occurred towardsthe end of th'e operation, the first flask being hurled with greatviolence into the air and shattered with a loud report.The liquid which distilled over was proved to be almost entirelysilicon tetrachloride, distilling almost to the last drop between 50°and 70°, and leaving an inappreciable weight of a brown film inthe flask. Disilicon liexachloride boils a t 1 4 5 O , whereas silicontetrachloride boils a t 5 9 O .Although disilicon hexachloride will thus catch fire and burn ina stream of chlorine to silicon tetrachloride, yet it was proved thatthis action only took place a t temperatures higher than the boilingpoint of t$e hexachloride. At the ordinary temperature disiliconhexachlorida does not combine with chlorine to a noticeable extent.This was proved as follows: Dry chlorine' from a cylinder waspassed first through a wash-bottle containing concentrated sulphuricacid, and then into a distilling flask cont'aining about 270 grams ofpure disilicon hexachloride, the chlorine escaping through anotherwash-bottle also containing concentrated sulphuric acid.A very considerable amount of chlorine was observed t o dissolvein the disilicon hexachloride in the flask, without, however, anyvisible signs of a chemical action taking place; thus no sensibleevolution of heat, could be detected as the chlorine entered theflask. The stream of chlorine was passed through the disilico2860 MARTIN : RESEARCHES ON SILICON COMPOUNDS. PART VII.hexachloride for about four hours, and the liquid containing thedissolved chlorine was then allowed to remain for twenty-four hours.When the liquid in the flask was heated (being connected for thispurpose with a condenser and a receiver rendered moisture-proof bycalcium chloride tubes), on the first application of heat the disiliconhexachloride appeared to boil a t quite a low temperature. Thiseffect, however, was entirely due to the, escape of the dissolvedchlorine, and on distilling the 270 grams of disilicon hexacliloridecertainly less than 1 C.C. of liquid passed over below looo, thusconclusively proving that no appreciable amount of silicon tetra-chloride (b. p. 5 9 O ) was produced by t<he prolonged action of chlorineon disilicon hexachloride.This experiment also proved the very great solubility of chlorinein disilicon hexachloride, a fact which does not appear t o havebeen noted before.The refractive indices of disilicon hexachloride and trisiliconoctachloride were kindly determined for the author by Mr. H. R.Nettleton.The author desires to thank the Senate of London University fora grant from the Dixon Fund which nearly covered the heavyexpenses of the investigation. He also desires to thank the ChemicalSociety for likewise giving him a grant for the same purpose.BIRKBECK COLLEGE,LONDON