Of these impurities, those which are found to have the most detrimental effect are oxide of iron and alumina, as these, especially on keeping, form compounds with phosphoric acid insoluble in water, causing the superphosphate to "go back" or "revert." Hence, as in England the price of superphosphate is fixed solely by the percentage of soluble phosphate which it contains, it is important to the manufacturer to obtain his raw phosphates as free as possible from oxide of iron and alumina.

For this reason it has of late years become customary in certain grades of phosphate to estimate their value not only by the percentage of phosphate of lime contained in them, but also to take into account the percentage of oxide of iron and alumina, it being generally considered that these substances will prevent double their weight of phosphate from becoming soluble. As, however, it is impossible to find any mineral phosphates free from these impurities, a limit of 3 per cent. is usually allowed, beyond which amount the excess counts double as against the phosphate. For example, a sample of phosphate testing in the dry condition 80 per cent. phosphate of lime and 3 per cent. oxide of lime and alumina together, would be charged for on the basis of 80 per cent., while a similar phosphate containing 4 per cent. would be invoiced at only 78 per cent. It will thus be seen that the accurate estimation of these two substances is a matter of the greatest possible commercial importance to all those dealing in these commodities, whether as buyers or sellers; and, unfortunately, as is often the case in newly-developed necessities, the results have been anything but satisfactory.

In framing a contract it is now customary to allow a margin of 1 per cent. on the phosphate of lime and per cent on the aggregate of the oxide of iron and alumina between the analyses of the two chemists nominated by the two parties, and if any greater differences occur the matter has to be referred to a third chemist previously agreed upon; the mean of the two most nearly-agreeing results to be taken for the purposes of valuation.

As greater care is now exercised in the taking of samples, it is not often that those specially engaged in the analyses of phosphates are found to differ beyond the stipulated 1 per cent. in the phosphate of lime, but it can scarcely be said that the same uniformity exists with regard to the estimation of the oxide of iron and alumina. During the last few years, and especially since a conference appointed by the Chemical Manure Manufacturers' Association to consider the subject, much greater uniformity has prevailed, and among English chemists the wide discrepancies which were formerly very common are now of rare occurrence; but it is to be regretted that the same agreement is not observed between our tests at home and those made by some leading Continental and American analysts.

The accuracy of some of my estimations of oxide of iron and alumina having been called in question, I determined to pat the processes I have in use to a critical examination, and with this end in view I started the series of experiments which I wish to bring before your notice.

Of the large number of processes described with greater or less exactitude in the various books on chemical analysis, few yield accurate results or are adapted to the requirements of a busy laboratory. Practically the processes used by those chemists who devote their attention specially to phosphate analysis may be classed under two heads-1st the acetate of ammonia process; and 2nd, the Glaser method and its modifications-and it will be to a discussion of these that I shall limit myself.

The first or acetate method is based upon the fact that phosphate of iron and phosphate of alumina are insoluble in a cold solution of acetic acid, while phosphate of lime is soluble. It would thus appear at first sight that a ready and simple means was at our disposal to effect a perfect separation, but unfortunately, as is often the case in seemingly simple operations, sources of error arise which render the accurate estimation of the oxide of iron and alumina a matter of considerable difficulty, the chief cause of danger being the fact that phosphate of alumina is appreciably soluble in excess of acetic acid. Although this fact has long been known to some analysts, and was clearly pointed out by W. C. Young in The Analyst of 1890, pp. 61

and 83, it has been very widely overlooked, even by those whose duty it is to analyse phosphates almost daily, and it is to this neglect that many of the discrepancies between the results of eminent men are attributable. It is true that the paper just referred to did not treat directly of the estimation of alumina in phosphates, and it may be partly for that reason that its teachings have remained in certain quarters so long unheeded.

Another source of error in working the acetate method has been the widespread assumption that the acetate of ammonia precipitate consisted solely of phosphates of iron and alumina, and that these phosphates were normal orthophosphates. This was an error which received some sort of semi-official sanction in a report issued in 1878 by a committee appointed by the British Association to investigate the methods in use for the estimation of phosphoric acid in commercial phosphates, and, in spite of a warning which I gave in a letter to the Chemical News of August 23rd of that year, I fear the assumption of the purity of the precipitates was too widely relied upon and the percentage of alumina obtained by deducting the FePO, calculated from the amount of Fe2O3 found in the precipitate from the total weight, no account being taken of the solubility of Al PO in the excess of acetic acid used, nor of the presence of lime in the precipitate, nor of the fact that the phosphates present are not necessarily normal. Hence results obtained in this manner were at best but a balancing of errors, and sometimes errors of considerable magnitude. It is not to be wondered at, therefore, that with such possibilities of error the results obtained by different chemists were widely divergent, and were such as to call forth from dealers in phosphate a vigorous protest, and an attempt, by means of the conference before alluded to, to remedy what undoubtedly a crying shame to the analytical profession. Fortunately, the publication of the Glaser process gave to the world an alternative method, which enabled chemists to place an effective check upon the results obtained by the older acetate process. If matters had then gone on to what should have been the natural termination of such a discussion, and if the information obtained had spread to other countries, there can be little doubt that greater uniformity in results would have occurred; but unfortunately such was not the case, and we still have to deplore too frequently wide variations in analysis, especially between the English tests and those made for the shippers abroad.

was

It therefore seemed to me eminently desirable that the processes we have in use should be put to a critical examination in order to ascertain to what extent they are reliable.

With this object in view, I have made test experiments under known and well-defined conditions, with practically three distinct processes-first, the Glaser in a modified form, which depends for its success on the complete insolubility of sulphate of lime in alcohol acidified with sulphuric acid and the solubility of phosphates of iron and alumina in the same menstruum; second, the acetate method, as I have been in the habit of using it for many years past, and thirdly, the acetate method as described by Dr. F. Wyatt in a book recently published (1892) in New York, entitled "The Phosphates of America."

As this process is evidently largely in use among chemists in the United States, and as it is in this direction that discrepancies are often observed, I have examined it in the same way as I have done the other two processes.

Before, however, giving the test experiments for each of these methods it will be necessary to describe in detail the processes as carried out by me in actual practice; and this will perhaps best be done under the three heads given above.

The Glaser Method. This process as now practised by me is substantially the modification given by my friend Mr. H. H. B. Shepherd in the Chemical News 63, 251, and is as follows:-2.5 grms. of the finely prepared sample, previously dried at 100° C., are treated with from 20-30 cc. of pure strong HCl in a beaker, and the solution, when effervescence has ceased, is evaporated to hard dryness in a water-bath. By this means the fluorine present is expelled, and any soluble silica converted into the insoluble condition. The mass is then dissolved in

about 10 cc. of of dilute HCl (1--4), and when all but the silicious matters has dissolved the contents of the beaker are washed into a 250 cc. flask with the least possible quantity of hot water. It should here be noted that aqua regia, either for the original attacking of the phosphate, or the subsequent solution is not admissible, as it decomposes any pyrites which may be present, and thus includes in the oxide of iron, the iron combined with the sulphur, which, being unattacked by the sulphuric acid in the manufacture of superphosphate, has no detrimental effect. When the solution of phosphate is cooled, 10 cc. of pure sulphuric acid are added and the contents of the flask shaken by a rotatory motion, and the mixture allowed to cool. When cold the flask is filled up with alcohol (redistilled methylated spirit will do) to the mark on the neck, and the whole well shaken. When the contraction is complete the liquid in the flask is again adjusted by the addition of more alcohol, and allowance is made for the volume of the precipitated sulphate of lime. It is again well shaken and put aside for the sediment to settle, when it is passed through a dry filter into a 200 cc. flask. The 200 cc. (= 2 grms, of substance) are then evaporated in a platinum basin at a low temperature, and the evaporation is continued until vapours of sulphuric acid are given off, and the whole of the organic matter present (obtained either from the substance or from the spirit used) thoroughly charred. This is very important, otherwise there is a danger of a portion of the oxide of iron and alumina being held in solution by the organic matters in the subsequent precipitation. When cold the residue is washed into a beaker, bromine or peroxide of hydrogen is added in good excess, and the liquid is heated nearly to boiling for some time. Ammonia is then added to the liquid to distinctly alkaline reaction, it is then boiled gently for a short time, a few drops more of dilute ammonia added and the precipitate filtered and washed with boiling water, to which preferably an ammonium salt has been added. This precipitation needs considerable care, as on continued boiling the sulphate of ammonia in solution is resolved into an acid sulphate which dissolves more or less of the precipitated phosphates. It is important therefore, even when the instructions I have just given are followed, to test the filtrate by adding a little more ammonia and warming to ascertain that the whole is precipitated. Unless an ammonium salt be used in the washing there is a tendency for the precipitate to split up into a basic phosphate, and a portion of the iron and alumina may wash through the filter.

An alternative method of precipitation, which I have lately adopted with advantage, and which obviates entirely the necessity of searching for magnesia in the precipitate, is as follows:-After the bromine has been expelled the liquid is made distinctly alkaline with ammonia, a little solution of acetate of ammonia is added, and the solution made slightly acid with acetic acid, and then well boiled. When precipitated in this way the mixed phosphates filter more rapidly and the washing is easier, while at the same time there is no risk of loss by solubility, as in the ordinary method of precipitation.

The precipitate, which consists of oxide of iron and alumina in combination with phosphoric acid, contains, if the instructions I have given are implicitly followed, no lime whatever, but if precipitated with ammonia only may contain a trace of magnesia. As the composition of this precipitate varies in almost every analysis, it is not, in my experience, safe to calculate the percentage of oxide of iron and alumina for the weight of the whole, but it must in every instance be carefully analysed. The method I employ is as follows:-After ignition, the precipitate is weighed, dissolved in dilute HCl, when a perfectly clear solution should be obtained, citric acid added (usually about 2 grms.), and the solution made ammoniacal, so that about one-quarter of the bulk shall be strong ammonia (0.880 sp. gr.). If sufficient citric acid has been added, the liquid should retain its light yellow colour and remain clear for some time. It is then set aside, with occasional stirring, for at least an hour, when, if no precipitate forms, magnesia is absent, or present only in negligible quantity. If magnesia is present, the solution is again well stirred, and after settling, the precipitate must be filtered off, washed

with dilute ammonia (1 in 4), and ignited to Mg,PO. To the filtrate or solution, magnesia mixture in moderate excess is added, the liquid well stirred for a period of not less than two hours, and the precipitate collected in the manner described for the magnesia, if present. (There is no need to redissolve this precipitate, as there are no impurities which can come down with it and contaminate it. If there was any doubt as to whether lime was present and oxalate of ammonia has been used, the phosphate precipitate must be dissolved and reprecipitated with ammonia.) To the filtrate, sulphide of ammonium is added in small excess, the liquid warmed, and the precipitated FeS collected with the ordinary precautions, and it can then be burnt directly to FeO3. By subtracting the weight of the phosphoric acid and magnesia (if present) from the total weight, the weights of the oxide of iron and alumina are readily obtained. The process is a long one, but with due care very satisfactory results are obtained.

In testing this and the other processes I prepared standard solutions of iron and alumina by dissolving known quantities of steel wire and alumium foil in acid-the aluminium foil being carefully analysed to correct for the impurities which were contained in it in small quantity. Solutions were also prepared of phosphate of lime (or phosphoric acid and chloride of calcium) and of chloride of magnesium.

The following are the results obtained :—

When ammonia only was used the precipitate was found to contain 0.007 grm. of MgO, a fact which shows the necessity of searching for it in every analysis. In the second case no MgO was present.

A long series of duplicate tests by the Glaser method and by the acetate process about to be described has convinced me that with due care and a close observance of the con. ditions I have stated will enable duplicate tests by the two processes to agree usually within 0.1 per cent. on the total material taken. I prefer to have my duplicate tests by the two methods, especially as the primary separations are so dissimilar.

Acetate of Ammonia Process.—Under this head I propose to treat only of the process as I have been accustomed to use it for many years past.

Two grms. of the finely-prepared sample are evaporated with strong HCl in the same manner as in preparing the solution for the Glaser process, the residue is taken up with about 10 cc. of dilute HCl, digested until the whole is soluble, diluted with water, and the silicious matter filtered off. The solution is then boiled with peroxide of hydrogen, or bromine, to be quite sure that the whole of the iron is as peroxide; the solution is allowed to cool, and dilute ammonia is added until a slight permanent precipitate is obtained. Dilute HCl is then added, drop by drop, until the liquid is again quite clear, and then sufficient acetate of ammonia solution, with an inclination to acidity rather than alkalinity, to form chloride of ammonia with the whole of the free HCl. An excess of acetate of ammonia does not affect

the accuracy of the results. After well stirring, the beaker is set aside for some time, and the precipitate is then collected, washed first with cold and then with hot water, dried, burnt, and weighed. By this means the solution contains free acetic acid equal to only the quantity of free hydrochloric acid necessary to keep the phosphates of iron and alumina in solution, and in this solution the alumina is not appreciably soluble. The precipitate thus obtained consists of phosphates of iron and alumina, and varying quantities of phosphate of lime, and it is, therefore, necessary that a careful analysis of it be made, in order to arrive accurately at the percentage of alumina. This I prefer to do as follows:-The precipitate is dissolved in dilute HCl, citric acid added, and then oxalate of ammonia in slight excess. The solution is boiled, and then dilute ammonia added, drop by drop until the liquid is neutral to litmus paper and then acetic acid to distinctly acid reaction. The liquid is then kept simmering for some time to ensure the complete precipitation of the whole of the lime; it is then filtered, the precipitate washed with hot water, dried, and burnt to carbonate, care being taken to keep the temperature sufficiently low to prevent the formation of any caustic lime. In the filtrate the phosphoric acid is precipitated with magnesia mixture, the solution well stirred, and preferably allowed to stand overnight. The supernatant liquid is then passed through a filter, the filter washed into the original beaker, and the precipitate dissolved in the least possible amount of dilute HCl, and then reprecipitated with strong ammonia. It is then allowed to stand for at least an hour with occasional stirring, collected, washed with dilute ammonia, dried, ignited separately from the filter, and weighed as Mg,PO7.

To the filtrate sulphide of ammonium is added in slight excess, the solution warmed, and the iron estimated with the usual precautions.

The CaCO3 is calculated to CaO; the Mg.PO, to P O ; the iron is weighed as FeO3 and the Al2O3 is obtained by difference.

The following experiments will show the necessity for a close observance of every detail:

The filtrate from a similar test was evaporated to dryness with the addition of hydrochloric acid, the residue was taken up with a little dilute HCl, and washed with the least possible quantity of water into a 250 cc. flask, and treated by the Glaser method. On adding ammonia and boiling for some time no trace of any precipitate was observed, from which it may be concluded that no alumina is soluble.

The weak point of both the Glaser and the acetate processes is that the alumina has to be obtained as the difference in the analysis of a comparatively complex precipitate, which renders the nicest work necessary to obtain strictly accurate results; but after having made a very great number of experiments to obviate this fault, I am free to confess that I have failed to hit upon any plan wholly without objection, and I still believe the difference method to be the best at present at our disposal.

Bearing upon the formation of phosphate of lime in combination with the iron and alumina, the following results, although not obtained under quite similar conditions to the acetate of ammonia precipitation, may prove of interest, as showing the tendency which exists for lime to be precipitated in preference to iron and alumina.

0 087 grm. Fe2O3 was precipitated in the presence of 0.565 PO with ammonia, the solution boiled, and the resulting precipitated weighed in the usual manner. After weighing, it was dissolved in HCl, diluted, CaCl, 2.2 grms. added, and the solution again precipitated with ammonia, dissolved in HCl, and precipitated with ammonia. The following were the results obtained:

A repetition of these experiments, so far as the precipitations were concerned, gave substantially the same results. It would seem from the results of the analyses of the final precipitates that, whereas there is little tendency on the part of phosphate of iron to part with phosphoric acid on washing, in the corresponding phosphate of alumina a considerable loss (0·012 grm.) has taken place.

The fact that the whole of the alumina and lime precipitate dissolved readily in very dilute HCI in the cold, and that a

considerable portion (more than 3) of the iron and lime was soluble, renders it very probable that the precipitates are, chiefly at all events, double phosphates of the two

bases.

The Wyatt process.-This process as described by Dr. Wyatt is as follows:-

Fifty cc. of filtrate (substance dissolved in aqua regia) from the siliceous matter, equalling 1 grm. of the phosphate, are placed in a beaker and made alkaline with ammonia.

"The resulting precipitate is redissolved by the addition of just sufficient hydrochloric acid, and the liquid is again inade alkaline with ammonia in very slight excess. 50 cc. of concentrated and pure acetic acid are now added; the mixture is stirred and allowed to stand in a cool place until perfectly cold. It is then filtered on an ashless filter and the beaker and residue are carefully washed twice with boiling water. The flask containing the filtrate is then removed from beneath the funnel and replaced by the beaker in which the first precipitation was made. The substance on the filter is now carefully dissolved in a little hot, 50 per cent. solution of hydrochloric acid and the filter is washed twice with hot water. The filtrate in the beaker is next made alkaline with ammonia in slight excess, then made strongly acid with pure concentrated acetic acid, well stirred up and again allowed to stand until absolutely cold. The flask containing the first filtrate is now replaced under the funnel, the liquid in the beaker is filtered into it, the filter is washed twice with cold water containing a little acetic acid and then three times with boiling distilled water. The contents are calcined and weighed as phosphates of iron and alumina in 1 gramme of the material. In one-half of this precipitate the phosphoric anhydride is determined by the molybdate method. The remaining half is reduced and titrated with N. permanganate solution."

In the experiments, the details of which follow, I religiously pursued the process as here given, except so far as that relating to the analysis of the mixed phosphates.

IN the Chemiker Zeitung of 2nd and 9th November last, there appears under the title "Briefe aus Egypten," an article, in two parts, by Prof. E. Sickenberger of Cairo, describing a visit to the Wady Atrun (Natron or Soda Valley) in Egypt, and containing a theory of the manner of the formation of the carbonate of soda contained in the water of the lakes there, and in the salt deposit which covers their banks.

At Kafr Dawar station, on the left Nile railway, large heaps of the natron brought there on camel-back, and containing 60 per cent. and upwards of carbonate of soda, are accumulated for further transport, the Egyptian Government receiving annually on contract about 500 tons of the material. The valley lies on the left bank of the Nile, and about 36 miles from it, and 60 miles from the Mediterranean. Its level generally is below that of the Nile and of the sea. It forms a long trough covered at intervals by ridges of drift sand. Its edges are formed of rounded sand hills, and nearly in the middle there projects a solitary rock (El Melouk), whose naked sides show alternate layers of limestone, ferruginous clay containing loose well-defined crystals of rock-salt and gypsum, and chalky sandstone, and judging from the levels the probability is that the edges of the valley below the sand would show a continuation of these strata.

It has been generally assumed that the formation of the Egyptian sesqui-carbonate takes place in the following manner:-The water of the Nile on its way underground to lakes lying at a lower level passes through strata containing sodium chloride, calcium sulphate, and calcium carbonate. Under the influence of the water mutual decomposition takes place, the water becomes charged with sodium carbonate, and issuing in the form of springs which supply the lakes, leaves on evaporation an impure sodium carbonate behind.

The springs supplying six of the lakes-there are 16 lakes altogether-were examined, and not one of these showed an alkaline reaction. They were neutral. They gave off no gas, the taste was pure bitter, slightly salt, with no trace of alkali, and tests showed the presence of sulphates and chlorides. The springs frequently form at their outflow small basins, and both springs and basins are filled with algae, consisting of a small proportion of an oscillatoria and much of a green conferva. Soon after issuing from the ground the water begins to evolve hydrogen sulphide. The smell of it becomes very strong, and is quite lost again as the distance from the point of outflow increases, while at the same time the algae assume a currant-red, then a brown colour, and gradually dying, change into a brown mud which at last becomes deep black. The black colouring matter is iron sulphide. This mud covers the bottom of the lakes under the crystallised soda. In proportion as the hydrogen sulphide was dissipated and alga and mud assumed a red colour, alkalinity appears, faint at first, but gradually becoming stronger and attaining its maximum where the mud was black and the water of a red colour. On stirring the red or black mud with a stick, numerous large bubbles of gas ascended which proved to be carbonic acid. Microscopical examination showed the presence in the red, brown, or black mud of large numbers of a micrococcus; and samples of the mud in which the confervæ were dead, when mixed with water, fermented with evolution of much carbonic acid and considerable increase of the micrococcus, showing its importance as a producer of carbonic acid in the formation of the sodium sesquicarbonate.

The process, therefore, is probably completed as follows:The water of the Nile on filtering through strata containing calcium carbonate, gypsum, and rock-salt, becomes charged with sodium sulphate. This is decomposed by the alge (in order to supply their requirement of oxygen) into sodium sulphide, and this again under the influence of the carbonic acid produced by the micrococcus, is converted into sesquicarbonate, hydrogen sulphide being driven off,

The sodium chloride, in so far as it has not been previously converted into sulphate, takes no part in the reaction, and crystallises on evaporation of the water in the lakes, forming an upper crust, separate from the sesquicarbonate of soda.

This view of the formation of the sesquicarbonate in this valley is supported by the fact that it is entirely a superficial one. The ground below the covering of iron sulphide is without any alkaline reaction, and plants such as halfagrass grow here in a soil the surface of which is covered to the depth of several inches with sand often containing as much as 80 per cent. of sodium carbonate, because, owing to the rainless climate, that salt is not washed down to their rocts, which are further protected by the covering of iron sulphide at the bottom of the springs and lakes.

Prof. Sickenberger promises further information as to the possible use to be made of these deposits, and the industries to which they may eventually give rise.

In the "Berichte der Deutschen Chemischen Gesellschaft" of January 9th of this year, E. W. Hilgard has a paper on "The mode of formation of alkaline carbonates in nature." In this he bases his conclusions apparently on observations made chiefly in California and in Western America generally. Beginning by recalling the fact that simple carbonate is never found in nature, but always sesquicarbonate, he concludes that it is formed by mutual reaction between the neutral salts, sodium sulphate or chloride and calcium carbonate, in presence of an excess of carbonic acid. In this way, from sodium sulphate and calcium carbonate, gypsum and sodium bicarbonate are formed, and on evaporation and exposure to the air the latter would remain behind as sesquicarbonate.

The reaction may be shown by passing carbonic acid gas at the ordinary temperature into a solution in which calcium carbonate is suspended. The change takes place also with sodium chloride, and in that case calcium chloride and sodium bicarbonate are found in the solution existing side by side.

Hilgard shows that in very dilute solutions-up to 1 grm. of sulphate per litre-both potassium and sodium sulphate are completely converted into their bicarbonates, but in stronger solutions the conversion becomes less and less complete as the strength of the solution increases, so that at 8 grms, per litre only about 25 per cent. of the sulphate is so converted.

After filtering these solutions to remove the excess of calcium carbonate and any calcium sulphate which may have been formed, and on evaporation in the laboratory, it is found that a reverse action takes place between the calcium sulphate in solution and the bicarbonate, and in the dry residue only one-tenth of the bicarbonate is to be found, nine-tenths on the average reappearing again as sodium sulphate together with calcium carbonate.

Hilgard says, however, that " undoubtedly if the evaporation were effected at ordinary temperature this reverse action would not take place," and that therefore a residue of bicarbonate, gradually becoming converted into sesquicarbonate, would remain.

But he gives no reason for this belief, and it is not easy to imagine his grounds for it. In any case, in view of his experiments, it would be next to impossible to conceive the formation, in the manner indicated by him, of deposits containing absolutely more carbonate than sulphate, except under the supposition that after the formation of the deposit, water has again acted upon it in such a manner as to remove the excess of sulphate and leave the less soluble sesquicarbonate behind. And natural deposits containing as much as 80 per cent. of carbonate, and therefore little sulphate, are not uncommon.

One interesting fact mentioned by Hilgard, tends, I think, much to the destruction of his theory. It is, that in certain districts of Western America, where the presence of alkaline carbonate in the soil renders its cultivation impossible, he has succeeded in improving this condition by the application Under the influence of water

of

gypsum as a manure.

say after irrigation-calcium carbonate and the neutral and harmless sodium sulphate are formed, thus showing that at ordinary temperature the "reverse action" between calcium sulphate and sodium sesquicarbonate actually does take place.

BY CARL OTTO WEBER, PH.D., F.C.S.

Ir is a well-known fact that in the manufacture of guncotton, as well as of the lower nitrated pyroxilines, the greatest attention must be paid to remove, in the washing following upon the nitration, every trace of free acid, otherwise the nitrocellulose is liable to explode spontaneously during the drying. The complete removal of the free acids is a rather tedious operation, and in preparing one of the lower nitrocelluloses some time ago it occurred to me that the washing out of the last traces of acid might be dispensed with by finally treating the cotton in water containing a very small quantity of ammonia. Carrying this idea out I observed the pyroxiline to assume a slightly yellowish tinge, a sure sign of alkalinity prevailing. The pyroxiline was then removed from the water. and as much as possible dried between filter-paper. To dry it completely the substance was put in an oven at 70° C., when after about three hours a terrific explosion ensued, literally tearing to pieces the strong copper oven, the fragments of which were hurled all over the room. The quantity of the pyroxiline was about 1 oz.

This explosion is remarkable from two points of view. In the first instance the pyroxiline was dinitrocellulose, which is scarcely considered an explosive; and further, the temperature at which the explosion occurred was very much below the temperature at which even gun-cotton (hexanitrocellulose) ignites, which temperature, according to Cross and Bevan, lies at from 160° to 170° C. I have since tried the ignition-temperature of dinitrocellulose by placing test-tubes, filled to about one-fifth with mercury, in a sand-bath. In the test-tubes were placed thermometers, the bulbs of which were submerged in the mercury. By throwing in these test-tubes little balls of dinitrocellulose at varying temperatures the ignition-point can be estimated with perfect safety and great accuracy. Thus I found that dinitrocellulose, if pure, ignites at from 194° to 198° C.

It is therefore evident that that explosion must be due to the treatment of the partially-washed dinitrate with ammonia, and I believe that its explanation is given by an observation I made some years ago, and which I believe to be new, as I never saw it recorded anywhere. If we take a concentrated solution of nitrate of ammonia, add to it a little acetic acid, and then proceed to evaporate the solution on the water-bath, as soon as a certain concentration is reached the whole mass ignites spontaneously, and the