ti10 WILLIAMS : DETERMIKATIOX OF FLUORIDE BY ETCHISG [Vol. 75 PROCEEDINGS OF THE SOCIETY OF PUBLIC ANALYSTS AND OTHER ANALYTICAL CHEMISTS Determination of Fluoride by Etching BY H. AMPHLE‘CT T31‘ILLIAMS (Read at the meeting of the Society on Wednesday, May 3rd, 1950) SuNoPsIs-The reactions involved in the etching of glass hat-e been studied with a view- to the development of a specific quantitative method for the determination of small quantities of fluoride. Optimum conditions for high sensitivity have been found by trial, and a method capable of detecting 0.1 pg. of fluoride ion is described. The reactions which take place under these conditions have been found to comprise the dissolution of silica from the glass by the hydrogen fluoride, and the deposition of hydrated silica through hydrolysis of the silicon tetrafluoride formed, with the production of a matt “etching” and the regeneration of hydrogen fluoride.It is concluded that a process in which the fluoride acts as a catalyst provides no sound basis for quantitative analysis. THE etching test, in which the fluoride ion is identified by its property of attacking glass, is the oldest known test for fluorides; it is still one of the most widely used. The reason for its popularity lies in its specificity, for the acids of fluorine are believed to be the only acids that will dissolve glass rapidly and in dilute solutions. Numerous attempts have been made to derive a reliable quantitative method for the determination of fluoride ion by iitilising this property, a few with some measure of success.Thirty of the more promising methods published during this century have recently been studied by the author,* but only brief reference can be made here to the literature. The majority of published methods afford only an approximate estimate of quantity, since they depend upon visual comparison of the depth of an etching produced on 3 piece of glass that has been exposed to the vapours resulting from treatment of the fluoride with concentrated sulphuric acid with that produced under similar conditions by a known weight of fluoride ion. The minimum quantity of F’ which will give a visible etching has been variously reported as between 10 and 0.1 mg., the average being about 0.5 mg.; by using modified techniques higher sensitivities have been claimed by some analysts, notably Woodman and Talbotl and Gautier and Clausmann,2 who found that by heating the reaction mixture to 220” C .in suitable apparatus a few micrograms of F’ could be detected. With quantities of that order, however, these workers were unable to obtain consistent etchings. Several workers have attempted to estimate fluoride ion from the loss of weight observed when a piece of glass is exposed to the hydrogen fluoride evolved by treating the sample, or a purified concentrate obtained from the sample, with sulphuric acid. In Brauns’ m e t h ~ d , ~ for instance, for the determination of F’ in organic compounds, the material is digested in a small Pyrex flask with sulphuric acid and potassium nitrate until decomposition is complete, when the fluorine content is calculated from the loss of weight of the flask.But, unfor- tunately, almost all minerals containing fluoride ion also contain silica, the presence of which might vitiate an etching test through combining with the fluoride ion as silicon tetrafluoride; and the complete separation of the F’ from the Si”” is a lengthy and unsatisfactory operation. With skilful manipulation, however, Mayrhofer and Wasitzky4 obtained good results with known weights of sodium fluoride; the average loss in weight of a small, inverted Jena glass funnel, fitted into the top of a crucible in which the purified fluoride residue was heated with concentrated sulphuric acid, being 0.96 mg. of glass per mg. of F’. In another ingenious method,2 the fluoride residue after removal of silica is treated with sulphuric acid in a closed platinum crucible containing a small basket of fragments of * A full review of the literature of etching and other tests for fluoride appears in the author’s Doctorate Thesis entitled “A Study of Existing Methods of Fluorine Analysis and the Development of an Improved Method of General Application” (University of London, 1948), in which is also included the greater part of the experimental work upon which this paper is based.October, 19501 WILLIAMS : DETERMINATION OF FLUORIDE BY ETCHING 51 1 optical glass (containing 42 per cent.of lead) suspended above the reaction mixture; after treatment for 5 hours at 140” C. the lead fluoride, which is formed on the glass in proportion to the hydrogen fluoride liberated, is dissolved in potassium chlorate solution and the lead determined colorimetrically.Quantitative etching methods of this type have been used where the quantity of fluoride ion to be determined exceeds 0.1 mg.; where the F’ is less than this, they have been found generally unreliable owing to errors inherent in the procedures available for concentration of the F‘ and its separation from Si““. The object of the present investigation was to ascertain whether the attractive, if elusive, possibilities suggested by the solvent action of hydrogen fluoride on glass could be utilised as a basis of a specific quantitative method for the determination of very small quantities of fluoride in diverse materials. Although no such object has been attained, it is hoped that the conclusions and the sensitive etching test which follow may be of interest to other analysts seeking the same goal.Owing to the practical drawbacks of weighing microgram quantities, gravimetric methods were regarded as undesirable and consideration was first given to the principle of producing a permanent etched mark, the extent of which could be determined by optical methods. But before the development of any such method is justified two questions must be answered: whether sufficient sensitivity could be obtained, and whether proportionate and comparable manifestations of the extent of attack on glass could be obtained with known amounts of fluoride. Since the only way in which these questions could be answered was by the method of trial and error, a very large number of etching tests have been performed, under various conditions.The results summarised below show the effect of varying the experimental conditions. EXPERIMENTaL The method usually given in the textbooks of analysis was first tried, a known weight of fluoride being heated with sulphuric acid in a platinum crucible covered with a waxed watch glass engraved with a distinctive mark. In the majority of these trials, the fluoride ion was introduced in the form of weighed quantities of intimately ground mixtures of purified calcium fluoride and potassium sulphate, free from silica, containing 1, 0.1 or 0.01 per cent. of fluorine ; measured volumes of standard sodium fluoride solution, gently evaporated in the reaction vessel, were substituted occasionally as a check.In the general textbook method the reaction mixture cannot be strongly heated owing to the wax melting, and further trials were carried out with an unwaxed watch glass at a higher temperature. Sensitivity appeared to be slightly higher, as shown in Table I. TABLE I ETCHINGS OBTAINED BY “TEXTBOOK” METHOD -1. Apparatus-Platinum crucible with waxed and marked watch glass. F’ introduced, Time of heating, Etching produced mg. min . 10 20 deep 1 20 fairly deep 0.5 30 slight 0-3 40 very faint 0.1 40 nil (bis) I3. Apparatus-Platinum crucible with unwaxed watch glass ; higher temperature. 0.5 0-3 0.2 0-2 0.1 0.1 30 definite but irregular 30 slight but irregular 15 faint 60 faint 30 nil 60 nil Etchings were patchy and the methods insufficiently sensitive, probably owing to the imperfect fitting of the watch glass on the crucible rim and the large area of glass exposed.To ensure better fitting and exactly comparable conditions, a rectangular lead block was cast with two parallel cylindrical cavities. The top of the block was ground flat and the cavities,512 WILLIAMS DETERMISATION OF FLUORIDE BY ETCHING [Vol. 75 after introduction of the fluoride and acid, were covered with a microscope slide held in position by a lead cover over a felt pad. The tests were carried out by heating the lead block in an oil-bath and varying the temperature, time of heating and quantity of acid. TABLE I1 EFFECT OF VARIATION OF TIME AND TEMPERATURE OF HE.4TING AND QUANTITY OF -4CJ.D Apparatus-Lead block with two caxTities, covered with unwaxed microscope slide.First four trials with cavities 7/16 inch in diameter, remainder with new lead block having smaller cavities 5/16 inch in diameter. Time of F' taken, H,SO,, heating, Teniperatu re, Etching mg. ml. min . O c. 0.3 2 30 "0 very distinct 0.2 2 30 2000 definite 0.1 3 30 800 faint 0.05 2 30 800 nil 0.2 1.5 30 300 well defined 0.1 1.5 30 200 slight but definite 0.05 1.5 30 200 very faint (bis) 0.03 1-5 30 300 nil 0.05 1.0 30 200 very faint 0-05 2.0 30 200 very faint 0.05 1.5 60 150 very faint 0.03 1.5 60 200 nil 0.03 1.5 9 0 200 nil Sensitivity was increased in this way to about 0-05 mg. of fluoride ion, but could not be further increased by variations of time of heating, temperature, volume of acid, nor by B Cooling Water c A Lead Cover Fig.1 G I yce ri n-Li n t Pad / ,GI ass Cove r-si i p ,Lead Block ,Medicinal Parafi n insertion of lead washers having holes of 1/16, 118 and 3/16 inch diameter over the cavities and under the glass slide, which was tried with a view to reducing the area of glass exposed to attack. The temperature could not be raised appreciably above 200" C. owing to the risk of acid creeping up and covering the glass, but 150" C. appeared to be sufficient for the etching.October, 19501 WILLIAMS : DETERMINATION OF FLUORIDE BY ETCHISG 5 13 This is in general agreement with Caley and Ferrer's conclusions,j using a similar technique in 1937. For the next series of trials, on the recommendation of certain of the earlier workers, direct cooling of the glass was introduced.A simple cooling apparatus consisting of flat- bottomed test tubes through which cold water circulated was employed, as shown in Fig. 1 ; the bottoms of the test tubes resting upon the glass, with pads of lint soaked in glycerin inserted between the glass surfaces to assist conduction. Microscope cover-slips were sub- stituted for the thicker slides, to avoid cracking through the wide difference of temperature on the two sides of the glass. TA4BLE 111 EFFECT OF COOLING THE GLASS -4pparatzts-Lead block : cooled, unwaxed , cover-slips. F' taken Time of heating, min. 30 30 3 0 30 30 60 90 30 30 30 30 45 60 90 45 45 45 Temperature, c. 150 150 150 150 150 150 150 165 180 180 180 180 180 180 180 180 200 Etching very deep deep definite slight nil nil nil faint definite faint nil just perceptible very faint very faint nil (gave smear when breathed upon) nil 2 spoiled, 2 faint but very irregular The results, given in Table 111, show that a remarkable increase in sensitivity can be achieved by keeping the glass cool throughout the test.Since it is difficult to conceive that the rate of attack of hydrogen fluoride on glass would be greater at low temperatures, the increased action is ascribed to concentration of the hydrogen fluoride in the droplets of condensation which form on the glass, and to a prolonged opportunity for attack through avoiding the evaporation from these droplets that probably occurs when the glass becomes heated. The trials show that by heating for 45 minutes at 180" C. a perceptible etching can be obtained from as little as 1 pg.of fluoride ion, and they suggest that still less might be detect- able if the temperature could be raised above 180" C. Difficulties were encountered when this was attempted, the causes being (i) flooding of the cover-slip by creeping or distillation of sulphuric acid, or both, and (ii) deposition of sulphur and water on the cover-slip owing to increased action of the sulphuric acid on the lead. Even at 180" C., etchings were very irregular, the attack producing deep pits in some parts of the glass and a rough, matt surface in other parts, and critical visual comparison of one etching with another was impossible. Moreover, in the course of further trials, a third source of error was encountered, viz., (iii) it was found that after an etching test had been carried out with a relatively large quantity of fluoride ion (20 to 100 pg.), subsequent etchings from very small quantities tended to be erratic and often high, although the cavities were washed out thoroughly between tests.This was ascribed to adsorption of hydrogen fluoride or formation of lead fluoride in the upper part of the cavity, and it was found that distinct etchings sometimes appeared with no added fluoride. Satisfactory "blanks" were only obtained in such circumstances after the lead block had been strongly heated in sulphuric acid, a procedure which caused further corrosion. These three difficulties render the estimation of microgram quantities of fluoride by etching with lead apparatus impossible, and an alternative material was sought.Eventually a new apparatus was constructed in the following manner.514 WILLIAMS : DETERMINATION OF FLUORIDE BY ETCHING [Vol. 75 To contain the reaction mixtures and to provide satisfactory seating for the cover-slips, three platinum tubes, 1& inches long by $ inch internal diameter, with flat 1-inch diameter flanges at the open end, were made. The oil-bath, which had proved troublesome owing to the creeping of the fluid (whether medicinal paraffin, wax, or glycerin), was replaced by a solid block of Duralumin, 3 x 24 x 2& inches, with three equidistant holes, 5/16 inch in diameter and 12 inches deep, drilled to take the platinum tubes, and a fourth hole drilled obliquely to the centre of the block to take a thermometer. A few drops of medicinal paraffin, sufficient to cover the bulb of the thermometer, were inserted in the oblique hole, and a small coiled wire spring in each of the other holes; the platinum tubes when inserted fitted snugly into the holes and no conducting liquid was found to be necessary or desirable.Heating was effected by means of an adjustable electric hot plate upon which the block was placed, and cooling by means of a conical flask condenser, the bottom of which was ground flat. The fluoride and acid were placed in the tubes, covered with No. 3, $-inch circular cover-slips, and the flask, supported by a retort stand, was gently lowered into position. Contact with the cover-slips was effected by the use of glycerin-soaked lint washers, # inch in diameter. By means of this apparatus, shown in Fig.2, three tests may be carried out simultaneously under identical conditions and interference from extraneous reactions and the creeping of oil is obviated. U Platinum Tube C 0 0 0 0 Plan of Block Fig. 2 The most suitable amount of acid, to leave only a small air space but to avoid risk of creeping up to the cover-slip, was found to be 0.7 ml., and this volume of pure A.R. sulphuric acid was used in each tube. The first 30 trials, with quantities of fluoride ion ranging from 0-1 t o lopg., produced no etchings whatever, no matter what the temperature, time of heating, type of glass or surface condition; even the possibility of lead acting as a catalyst was considered and some lead turnings were added, but without effect. Results with the platinum apparatus proved disappointing and puzzling.October, 19501 WILLIAMS DETERMISATION OF FLUORIDE BY ETCHING 515 Eventually another supply of sulphuric acid was tried; this gave a slight etching with 2pg. of fluoride ion.Four other supplies were then tried, one of which, from the same bottle of ordinary “pure” acid that had been used in the trials with the lead block, gave much better results. Since the principal variant in the different supplies of sulphuric acid was probably the moisture content, it seemed that this might be the essential factor. On determining the strength of various acids by titration and on trying dilutions prepared from the AnalaR acid which had failed to produce any etchings, this proved to be so. The results are shown in Table IV, from which it will be seen that the optimum concentration is apparently 96 to 96.5 per cent.w/w of sulphuric acid (H,SO,). TABLE IV EFFECT OF CONCENTRATION OF SULPHURIC ACID ,-1PPnl.n2us-Platinurn tubes ; cooled cover-slips ; heating for 1 hour at 200” C . F’ taken, /*g- 5 5 5 5 > Grade of sulphuric acid =Inalali (B.D.H.) “Sitrogcn-free” (B.U.H.1 ‘‘Technical” (B.D. H.) “Pure” (T. & 31.) “Pure” (B.D.H.), as used with lead block -4nalaR (B.D.H.), diluted ditto ditto ditto ditto ditto H,SO,, per cent. w/w 97.6 96.8 92.3 98.8 96.3 Etching nil slight nil slight definite very faint definite (ter) definite (bis) faint doubtful nil ’The trials with different supplies of acid were conducted at 200” C. for 1 hour. At this temperature droplets of moisture condense on the cooled cover-slip and the evolved hydrogen fluoride dissolves in them.The character of these droplets is critical, since if they do not form rapidly enough the hydrogen fluoride may be evolved first and escape under the cover- slip without being dissolved, and if they are too large the solution of hydrofluoric acid formed will be too dilute to attack the glass. According to Olivier,6 solutions of hydrofluoric acid containing less than 0.00033 per cent. of fluorine do not visibly corrode glass. Moreover, whilst a very fine mist produces an even, matt etching of a type suitable for comparison, larger droplets produce irregular craters in the glass. This seems to be the explanation of the vital importance of the moisture content of the sulphuric acid. \Wh the exclusive use of 96.5 per cent.sulphuric acid, further trials were carried out with the platinum apparatus to ascertain the optimal conditions, firstly as to temperature and time of heating. Results were not entirely consistent, and even when simultaneous triplicate tests were performed differences in the area or depth of the etchings would occur; but from a large number of trials the following inferences appeared to be justified: (i) that the hydrogen fluoride is not completely evolved at 180” C.; (ii) that a temperature of 220” t o 240” C. is necessary; (iii) that the action is not completed by the time the tubes reach 220” or 240” C,., a further period at that temperature being necessary; (iv) that no further action occurs at higher temperatures ; (v) that gradual heating increases the sensitivity, the best method apparently being to raise the temperature to 120” C.in 45 minutes, to 220” C. in the next 45 minutes, and to maintain at 220” to 230” C. for 45 minutes more; (vi) that no appreciable increase in the sensitivity can be achieved by raising the temperature of the cover- slip during the test; and (vii) that effective contact between the cover-slip and the cooling surface is important. To secure this, the glycerin-soaked lint pad, which tended to enclose air bubbles, was replaced by a small drop (0.02 ml.) of glycerin containing about 2 per cent. of agar, and a fine wisp of cotton wool, placed in the centre of the upper side of the cover-slip. The experimental requirements of thinness and flatness practically restricted the choice to microscope cover- slips, but careful selection of the grade of cover-slip was found necessary, as some supplies showed a finely pitted surface when examined microscopically with oblique illumination, and variations in sensitivity to attack were also found.Regarding the preparation of the glass surface for etching, it was thought possible that destruction of the molecular symmetry of the surface layer, for example bj- grinding or by pi-eliixiiiar>- attack, might render the The type and condition of the glass to be etched was next considered.516 WILLIAMS : DETERMISATIOS OF FLUORIDE BY ETCHISG [Vol. 75 substance of the glass more vulnerable. Superior etchings of greater depth and clarity were in fact obtained with the ground and polished glass cover-slips with optically plane surfaces which are specially made for use with haemocytometer cells ; the sensitivity, however, was not greatly enhanced.Pre-treatment of cover-slips by immersion in hydrofluoric acid baths of a strength just too low to cause pitting (> 1 per cent.) did not appear to render the glass any more susceptible. Further trials indicated that the state of hydration of the glass surface exerted a marked influence on sensitivity. The best results were obtained when the glass was in a highly hydrated state, and the cover-slips, selected after careful examination from a single packet, were cleaned in chromic acid mixture, soaked in distilled water, and air-dried immediately before use. On applying the technique and precautions outlined above, a marked increase in sensitivity was obtained with the platinum apparatus, as little as 0-2 pg.of fluoride ion giving a visible etching. The platinum tube overcame the difficulties encountered with the lead block at the higher temperatures, and produced, in the centre of the exposed area of the cover-slip, round matt etchings of a type which would readily admit of comparison, but replicate etchings from similar quantities of fluoride were not always consistent. Before examining further means of increasing the sensitivity, the question was considered whether the maximum degree of attack theoretically possible had not already been attained. Could it be obtained, a measure of the quantity of silicon separated from the glass in the process of etching or of the amount of combined fluoride ion in the end-products might show whether the whole of the F' taken had been induced to attack the glass.Attempts to extract and determine the silicon in the deposit of condensed moisture and end-products that formed on the cover-slip during the etching, by solution in alkalies, etc., were unsuccessful owing to high blanks from unetched cover-slips, and determinations of the residual combined F' by distillation of the deposit on the cover-slip from perchloric acid, followed by thorium nitrate titration, yielded only from one-fifth to one-quarter of the F' originally taken. I t seemed possible, however, that a proportion of both Si"" and F' might have escaped in the form of silicon tetrafluoride. In these circumstances the only practicable way to ascertain how much glass had been attacked appeared to be by direct weighing of the cover-slip before and after attack.A further series of etching tests, with larger amounts of fluoride, was therefore carried out, the cover-slips being weighed on a microgram balance after standardised cleaning and pre-weighing technique. T-~BLE V LOSS OF WEIGHT OF GLASS COVER-SLIPS Apparalus-Platinum tubes; cooled cover-slips; heated to 120" C. in 45 minutes, to &-O b'') ' C. in next 45 minutes, and a t 220" to 230" C. for 45 minutes more. Weight of cover-slip before F' taken, etching, mg. mg. nil* 148.535 nil* 145.824 nil* 138-867 nil* 129.086 nil* 139.005 Weight after etching and cleaning, mg. 148.533 145-8 19 138.856 129*080 139.001 Loss corrected for Glass to Loss, ' 'blank' ' F' ratio mg. 0.002 0.005 0.01 1 0.006 0.004 Mean cleaning blank., .. .. 0-006 1.000 139.900 138.862 1 *038 1.000 200.760 199.747 1.013 1.000 11 1.808 110.822 0.986 0.500 128.804 128.2 84 0.520 0.200 109.432 109.2 16 0.216 * Cover-slip cleaned only. 1.032 1.032 1.007 1.007 0.980 0.980 0.514 1.028 0.210 1.050 From the results shown in Table V it can be seen that the average loss of weight by etching was equal to 1.02 mg. per 1 mg. of fluoride ion. This ratio of loss, it may be observed, is higher than any found in the literature (0.5 to 0.96 rng, per 1 mg. of F'). Enquiries elicited the information that the glass used approximated in composition to the "formula,"October, 19501 U'ILLIAMS : DETERMISATIOS OF FLUORIDE BY ETCHISG 517 Na,O.CaO.SSiO,; assuming this to be true, the theoretical loss of weight, on the basis of the "equation"- should be 0-916 mg. per 1 mg.of F'. Even allowing for considerable variation in the com- position of the glass, it would seem evident that the attack effected by the F' had been not less than 100 per cent. efficient. Since no appreciable increase of sensitivity could therefore be expected, attention was directed to the effect caused by the presence of traces of silica in the fluoride tested, a question Sa2O.Cs0.5SiO, +- 24HF -+ Sa,SiF, + CaSiF, - 3SiF, -+ lPH,O, TABLE VI ETCHING IN THE PRESEXCE OF SILIC.~ ~ppnl.atzfs-Platinnin tubes; cooled cover-slips; heated to 120" C. in 43 minutes, to 220" C. in next 45 minutes, and at 220" to 230" C. for 46 minutes more.F' taken 1 CLg. 1 CLg- 1 Ph". 1 ELg. 1 P.9. 1 LLg. 1 pg. 1 tLg. 2 pg. 5 pg. 0.4 pg. 1 pg. 1 nil nil 1 mg. 1 mg. 0.2 mg. 1 CLg. 0.1 pg. 1.5 pg. 0.4 pg. 0.2 pg. 0.1 pg. 0.1 pg. 0.1 pg. 0.1 pg. 0.2 pg. 0.06 pg. 0.6 pg. 0.3 pg. 0.7 pg. 0.5 pg. 0.5 pg. 0.8 pg. 0.8 pg. 0.8 pg. 0.4 pg. 0.3 pg. 0.26 pg. 0.13 pg. Source of F' CaF, CaF, CaF, CaF, NaF S a F CaF, CaF, CaF, CaF, CaF, CaF, CaF2 Ca 17, - - c a F, CaF, Sa,SiF, Xa,SiF, Sa,SiF, Sa,SiF, Sa,SiF, Sa,SiF, Sa,SiF, Sa,SiF, Sa,SiF, CaF, CaF, CaF, CaF, CaF, XaF XaF well water well water, H,SiF, distillate ditto well water, ditto ditto ditto S O , added 2 rug. 4 pg. 3 pg. 20 pg. 100 pg. 2 pg. 4 PLg. 20 pg. 1 ing. 1 mg. 2 mg. I mg. 10 mg. 10 mg. 25 mg. 1 mg. r 1 ing. { Img. I 1 mg. 1 Illg. - - - - - - 3 1ng. -- 2 nig. -- 2 mg.L 1 ing. t 1 mg. 0.2 pg. 0.5 pg. 2 pg. 10 ing. 10 ing. lmg. 26 mg. - - -- 10 ins. - 10 ing. - - - Source of SiO, Sa,SiO, soh. ditto ditto ditto ditto dltt0 ditto pure SiO, pure SiO, purified sand ditto ditto ditto ditto ditto ditto powdered glass SiO, 7 glass } sand ,I _- suid >ant1 sand hand >and Sa$iO, S O h l . cl i t to ditto sand sand sand sand _- sand sand - - - Etching very strong very strong none strong weak weak normal very strong normal very strong none strong normal weak slightly strong normal strong weak normal slightly weak strong just visible strong strong strong normal none strong definite definite normal none normal normal normal normal normal normal normal strong strong of fundamental importance to the general application of an etching process to the determina- tion of fluoride.If sufficient silica is present, the F' must be evolved, partly at least, in the form of silicon tetrafluoride, and it has been shown by Germanii and Booth' that pure silicon tetrafluoride does not attack glass, though Roses (in 1850) observed that a slow attack occurred in the possible presence of moisture, and Reichg and Olivier6 also noticed some attack on moistened glass.518 WILLIAMS : DETERMINATION OF FLUORIDE BY ETCHING pa. 75 To ascertain the degree of separation from silica that would be necessary before microgram quantities of fluoride could be estimated by etching, a number of trials were conducted in which different amounts of silica, in various forms, were introduced into the platinum reaction tubes before adding the sulphuric acid.Representative results are shown in Table VI, the etchings being designated by comparison with the average result produced by a similar weight of F’ without silica. As the first trials with added silicate recorded in Table T I showed no reduction in the depth or area of the etching, the quantities of silica were increased, ultimately to proportions far in excess of those likely to be encountered in ordinary analytical work. Sext, a standard silicofluoride solution was tested, the measured volumes being gently evaporated to dryness in the platinum tubes before addition of sulphuric acid; finally, a sample of a London well water containing 1.3 parts of fluorine per million was tested, with and without additional silica. In many of these tests, some with proportions of Si”” to I?’ exceeding 1000 times the theoretical equivalent, there can be little doubt that much, if not all, of the F’ must have been evolved as silicon tetrafluoride ; yet, contrary to expectation, definite etchings were obtained.Indeed, although somewhat erratic, the etchings tended to be more marked than those without silica; in a few instances remarkably so and of a rather superior character. Since silicon tetrafluoride could, presumably, only attack glass after conversion to H,SiF,, it seems probable that the larger etchings observed with sub-microgram quantities of fluoride ion when silica was present are connected with the humidity, which, as shown earlier, was found to be necessary for high sensitivity; for silicon tetrafluoride, being more hydrophilic than hydrogen fluoride, as indicated by its vigorous reaction with water, would probably be more quickly and completely absorbed in the moisture condensing on the cover- slip than would hydrogen fluoride, and the H2SiF, formed would then be favourably placed for the production of hydrogen fluoride by dissociation in direct contact with the glass.The proportion of F’ converted to H2F2 might thus be from a third to the whole of the F’ absorbed, according to either or both of the reactions- H,SiF, -+ H2F, -+ SiF, H,SiF, -y- 4H,O --+ 3H2F2 -+ Si(OH),, the course of which would depend upon local conditions of temperature, moisture, etc. In this connection the importance of cooling the cover-slip, at least in the early stages, may also be explained, for JacobsoiilO has shown that the liberated SiF, and H,O do not react together if the temperature exceeds 30” C., although they immediately combine when cooled to form H2SiF,, which subsequentl? dissociates into H,F, and SiF, when the temperature rises.The possibility of increasing the sensitivity by placing a very small drop of water on the cover-slip before the test, to accelerate and localise the hydrolysis, was suggested by the above hypotheses. It was found, however, that small drops of water, or a thin film, on the under- side of the cover-slip quickly disappeared, owing to the desiccating action of the sulphuric acid. A 10 per cent. solution of glycerin was therefore employed instead, a droplet of 0.5mm. diameter being placed in the centre of the cover-slip.\Vith this solution, the droplet was somewhat reduced in volume at the commencement of the test and did not materially interfere with subsequent comparison of the etchings. The use of the glycerin droplet was found to stimulate attack in its immediate vicinity, 0.2 and 0.1 pg. quantities of fluoride ion giving distinct etchings round the droplet that were several times as great as without the droplet, with or without the addition of 2 mg. of sand, but the sensitivity was not reduced appreciably below 0.1 pg. of F’. This quantity of F’ is probably approaching that required to produce the minimum concentration of free H,F, in the condensed water necessary to attack glass. The discovery that silicon tetrafluoride produces an attack at least as vigorous as that of hydrogen fluoride under the conditions of the test, however welcome from some points of view, raises questions of fundamental importance to the quantitative application of the etching test, So far as is known to the author, SiF, can only attack glass by hydrolysis, 3SiF, -+ 4H,O --f 2H2SiF6 + Si(OH), .. .. . . . . (i) thus- \1 2H,F, -+ 2SiF,October, 19501 WILLIAMS : DETERMISATIOS OF FLUORIDE BY ETCHISG 519 The residual SiF, shown in the equation may be volatilised and gradually lost, or it may continue to undergo hydrolysis until the reaction becomes ultimately- 3SiF, + 12H,O --f 6H2Fz + 3Si(OH), . . . . . . . . (ii) The next stage comprises the attack by the liberated H,F, on the glass; this reaction cannot be precisely specified, but the production of more SiF, is to be expected, as in the simplified equation usually quoted- 2H,F, + SiO, -+ SiF, + 2H,O .. . . . . . . . . (iii) Hydrolysis of the secondary quantity of SiF, thus formed would lead to a chain reaction, with the F' assuming the role of catalyst. It cannot be assumed, however, that the liberated H2F, would confine its attack to the glass, and leave the precipitated Si(OH), untouched. To some degree, therefore, it may reverse the reaction shown in equation (ii). In any case there can be no prospect of finality. For these reasons any hydrolysis of silicon tetrafluoride would have disastrous consequences to quantitative analysis ; consequences, moreover, which could not be avoided by complete elimination of silica from the fluoride residue to be tested.In view of the evident importance of establishing more clearly what reactions take place during the etching, the end-products remaining on the cover-slip were micro-analysed. For this purpose, 1-mg. etchings were made, with and without added sand, and the moist white deposits found on the exposed areas carefully removed by wiping with fragments of wet filter-paper. Both deposits proved to consist almost entirely of sodium and calcium silico- fluorides and water, the weight of the dry silicofluorides in each test being approximately 400pg. A slight scum was observed on the surface of the sulphuric acid in the platinum tubes after the tests, both with and without sand; this consisted of silicic acid. Except for this small quantity (about 30 pg.of SO,), no significant amount of silicic acid or silicate was found in the deposits analysed. But having regard to the loss of weight of the glass, and the loss of fluoride ion unaccounted for, it is evident that a relatively large proportion of dissolved silica must have escaped under the cover-slip in the form of silicon tetrafluoride. To reduce such losses, further trials were carried out with the cover-slips sealed to the platinum tubes. The under-sides of the cover-slips were moistened with fodr O.5-mm. diameter droplets of the glycerin - agar solution, near to the periphery, forming nearly complete seals by capillary attraction when the cover-slips were placed upon the platinum flanges; also the time of heating was increased to 34 hours, the temperature being very slowly raised to 260" C.during this time. The resultant etchings exhibited a remarkable form, being apparently raised above the surface of the glass and covered with a powdery micro-crystalline deposit. Repeated cleaning treatment with boiling chromic and sulphuric acid mixture failed to remove this deposit, though silicofluoride would have been dissolved. It was found, however, that much of the deposit could be removed by scraping and that, operating under a dissection microscope with a scalpel, some of the matt surface of the etching could also be removed. On re-examina- tion of some of the previous etchings, obtained from microgram quantities of fluoride ion, it was found that much of the apparent etching could also be removed from these cover-slips in the same way, although this did not apply to the larger quantities employed before the platinum apparatus was devised, where the cover-slips were deeply pitted.Micro-analysis of the substance removed showed-that it consisted entirely of hydrated silica. The loss on ignition was 11.3 per cent., suggesting the hemihydrate, 2Si0,.H20 (13.0 per cent. of H,O), but the degree of hydration in the silica when deposited may not have been uniform, and may have differed from that existing after the cleaning treatment. The formation of the matt surfaces on the cover-slip by the deposition of hydrated silica is analogous to commercial matt etching processes, which have been shown to depend upon the growth of fine crystals of silicofluorides or of insoluble double fluorides on the surface of the glass, so protecting the glass underneath from further attack; the pitted appearance is in fact due to minute elevations, the size of which is determined by the degree of saturation of reacting ions and undissociated salts in the etching r n i x t ~ r e .l l ~ ~ ~ The evidence of a deposit of hydrated silica confirms the supposition that hydrolysis of silicon tetrafluoride takes place, and a means of ascertaining the extent of the hydrolysis was sought. Since in the previous experiments the silica remained firmly attached to the glass, it is clear that the loss of weight shown in Table V does not represent the full extent of the attack. Direct determination of the weight of silica deposited could not readily be effected520 WILLIAMS DETERMIKATIOK OF FLUORIDE BY ETCHING [Vol.i 5 as the silica could not be quantitatively removed by scraping, and alkali extraction proved unsatisfactory owing to dissolution of the glass, but it was thought that, with the knowledge of the water content of the silicic acid remaining after the cleaning and drying treatment, a sufficiently close approximation might be obtained by finding the loss on ignition of the cover-slips. Corrected for a small “ignition blank,” some representative results are shown in Table 1711, together with the estimated ratio of glass attacked per 1 mg. of fluoride ion. With the larger etchings, obtained from 0.5 to 1.0 mg. quantities of fluoride ion, a further, cloud-like, area of attack was noticeable round the periphery of the first sharp etching in trials where the heating was prolonged; this appeared to be due to secondary attack caused by hydrolysis, with further liberation of hydrogen fluoride, of the deposit of silicofluorides first formed; its area extended as the heating was continued.TABLE 1711 TOTAL WEIGHT OF GLASS ATTACKED Afiparatus-Platinum tubes, with liquid-sealed cover-slips ; heated gradually to 260” C. F’ taken, mg. . . . . . . Silica added . . . . . . Duration of test . . . . Loss of weight of cover-slip, mg. Weight of silicic acid, mg. (further loss on ignition, x 100/’88*7) . . . . . . Total glass attacked, mg. . . Glass to F’ ratio . . . . 1.000 0.200 0.100 1.000 none none none 10 mg. glass 31)hr. Z$hr. 34hr. 39 hr. 1.001 0.210 0.194 0.118 0.772 0.193 0.174 0.489 1.773 0.403 0.368 0.607 1-8 2.0 3.7 0.6 0.200 1 mg.each sand, glass and silica 29 hr. 0.069 I 0.100 0.169 0.8 0*100 10 mg. sand 39 hr. 0.163 It is noteworthy that the glass attacked per unit weight of fluoride ion was greater with the smaller quantities of F’; this is ascribed to the greater protection afforded by the thicker coating of deposit which forms with larger quantities of F’, and it is probable that with microgram quantities of F’ the ratio of glass attacked would be higher still. It is now apparent that the losses of weight shown in Table V are understated by the weight of water held by the silicic acid, since the cover-slips were not ignited before being weighed. Moreover, the strange circumstance that, whilst sub-microgram etchings showed inconsistencies of as much as 100 per cent. on visual comparison, the losses of weight from milligram etchings showed fairly close agreement, becomes explicable ; the visible etching represents the amount of glass changed into silicic acid, but the loss of weight represents the amount of glass converted into sodium and calcium silicofluorides and silicon tetrafluoride (plus any silicic acid dissolved or dislodged by the cleaning treatment).If the whole of the fluoride ion were evolved in the form of SiF,, and the “reaction”- In the presence of added silica the losses of weight were extremely variable. proceeded to completion, no appreciable loss of weight would be anticipated. But part of the F’ in the substance under test may be evolved as H2F2, although an excess of silica is present, as postulated by Schneider13; part of the SiF,, and part of the deposited silica, may ’ be lost in various ways, and since, moreover, the reactions taking place appear to depend upon fortuitous local concentrations of SiF, and H,O at the glass surface, the variations of weight are not surprising.To summarise the results of the investigations, an extremely sensitive etching test for fluoride, reacting to 0.1 pg. F’, even in the presence of silica, has been developed; but it has also been shown that the etching of glass, under the conditions necessary for high fluoride sensitivity, comprises a catalytic structural transformation of the silica, which, being capable of proceeding indefinitely, appears to provide no reliable basis for the determination of fluoride.REFERENCES 1. 2. 3. 4. 5. Woodman, A. G., and Talbot, H. P., J . Amer. Chew Soc., 1906, 28, 1437; 1907, 29, 1362. Gautier, A., and Clausmann, P., Bull. Soc. Chirm., 1912, 11, 872. Rrauns, D. H., J . Res. Nut. Bur. Stand., 1941, 27, 105. Mayrhofer, A., Schneider, C., and Wasitzky, A,, Biochew. Z . , 1929, 204, 62. Caley, E. R., and Ferrer, J. M., jun., Mikrochem. Actu, 1937, 1, 160.October, 19501 COX: OBSERVATIONS ON VITAMIN D 6. Olivier, E., Rev. Universelle Mines, 1921, 14, 25. 7. Germann, A. F. O., and Booth, 14. S., J . Phys. Cheni., 1917, 21, 81. 8. Rose, H., Pogg. Ann., 1850, 80, 406. 9. Reich, J. A., Z h e m Zeit., 1896, 20, 985. 10. Jacobson, C. A,, J . Phys. Chem., 1923, 27, 517-80 am1 761-70, 11. Spencer, C. D., and Ott, L., J . Amev. Ceram.Soc., 1927, 10, 403. 12. Schiilz, H., Glushutte, 1938, 68, 828-30. 13. Schneider, L., Uesterr. Z. Berg. u. Hiittenwesen, 1913, 61, 365; Anulyst, Xbst., 1913, 38, 475. THE LABORATORY, 1, SOUTHWARK STREET, LONDON, S.E.1. 521 DISCUSSION DR. J . R. NICHOLLS said that this test had always been applied to the etching of glass and it appeared that secondary reactions caused complications. I t might, however, be possible to use material other than glass. Suppose the sample were mixed with silica and sulphuric acid and the cover-slip were made of plastic. The liberated silicon tetrafluoride would react with the film of moisture on the cover-slip leaving a deposit of silica and secondary reactions should not take place. The difference in the amount of light passing through the cover-slip before and after the test might be proportional to the deposited silica. There seemed no reason why a cover-slip of clear fused silica should not be used since that material is very unreactive to hydrofluoric acid. DR. J. G. A. GRIFFITHS said that, bearing in mind that a t least part of fluorine present as hydrofluoric acid or silicon fluoride appears to become fixed on the glass microscope cover-slip as sodium and calcium silicofluorides, and therefore no longer available to “catalyse” the etching process, it would be interesting to know what happens if a silica cover-slip is used in place of a glass slip, and whether the possibility of an apparently interminable cycle of reactions involving hydrofluoric acid would lead to an even smaller quantity of fluorine ion than the quite remarkable minute figure of 0.1 pg. being detected if the glass slip is replaced with a silica slip, DR. WILLIAMS said, in reply to Dr. Nicholls, that the deposition of silicic acid when silicon tetrafluoride reacted with water provided a very sensitive test for fluorine, and a large number of modifications of this method had been published. In the quantitative variations the silicic acid formed was usually determined colorimetrically by development of the molybdenum blue complex, but he did not know of any procedure by which microgram quantities or less of fluorine could be estimated. Dr. Griffiths’ suggestion was an interesting one, but he did not think the sensitivity could be increased much further because a certain minimum strength of hydrofluoric acid, said by Olivier to be 0*0003 per cent., was necessary for the solution to attack glass a t all.