Molecular emission cavity analysis

Analytica ChimicaActa, 153 (1983) 41-51 Elsevier Science Publishers B.V., Amsterdam -Printed

in The Netherlands

MOLECULAR EMISSION CAVITY ANALYSIS Part 25. The Determination of Silicon

MARCELA

BURGUERA*

Chemistry Department, (Gt. Britain)

and S. L. BOGDANSKIb

University

of Birmingham,

PO Box 363, Birmingham

University

of Hull, Hull HU6 7RX (Gt. Britain)

B15 2TT

ALAN TOWNSHEND* Chemistry

Department,

(Received 10th March 1983)

SUMMARY Silicon is rapidly converted to volatile silicon tetrafluoride in 98% sulphuric acid at 135°C in the presence of chloride. The gases are carried by nitrogen to a m.e.c.a. OXYcavity, and the intensity of the white SiO continuum emission is measured. The detection limit is 0.2 fig Si, and linear calibration is achieved up to 500 mg 1-l silicon. Few ions interfere. The procedure is applied to an iron ore.

The determination of silicon is a necessary step in the analysis of natural and synthetic silicates, ores, ferrous and non-ferrous metals, organosilicon compounds and other silicon-containing substances. Silicon compounds are common in air contaminants. Exposure to silica occurs in hard rock mining, and in manufacture of porcelain, pottery, silica firebricks, etc. This paper presents a new method for the determination of silicon, by molecular emission cavity analysis (m.e.c.a.) [l].The formation of an effective population of silicon atoms requires a high-temperature flame [2-6] . The low temperature hydrogen-nitrogen diffusion flame does not provide enough energy to volatilize most inorganic silicon compounds, so that direct injection of such compounds (silicates, for instance) into the m.e.c.a. cavity does not produce any emission. However, an oxy-cavity [7] placed in a hydrogen-nitrogen flame was found to stimulate an emission when gaseous silicon tetrafluoride was swept into it. It was decided therefore to examine the possibility of determining silicon by converting it to volatile silicon tetrafluoride and measuring the emission produced in the m.e.c.a. oxy-cavity.

*Present address: Chemistry Department, University of the Andes, Apartado Postal 542, Merida 5101-A, Venezuela. bPAT Centre International, Royston, Herts., Gt. Britain. 0003-2670/83/$03.00

o 1983 Elsevier Science Publishers B.V.

42 EXPERIMENTAL

Appamtus All the emission measurements were made with a Unicam SP900 flame spectrophotometer, modified to record in the m.e.c.8. mode [8]. A Servoscribe 1s recorder, connected to the output of the instrument, and a premixed hydrogen-nitrogen flame were used. The instrument was operated with a slit width of 0.5 mm (~11 nm) at 540 nm and at gain 5. A duralumin cavity was found to be suitable for these experiments. This cavity, shown in Fig, 1, had two stainless steel tubes (1 mm o.d.) screwed into the rear of the cavity at opposite sides, one for supplying oxygen and the other connected to the vaporisation reactor. The cavity was inserted into a specially designed holder (Fig. 2), which permits adjustments of the vertical and horizontal cavity positions, as well as cavity angle, in order to obtain the optimal cavity position in line with the detector. During the preliminary

(al Fig. 1. Aluminium oxy-cavity:

bl (a) cross section; (b) front view (dimensions in mm).

Fig. 2. Sample holder assembly: the cavity is screwed into hole a. Fig. 3. Volatilization system for silicon: (1) sulphuric acid reservoir; (2) reaction tube, containing sample; (3) oil bath; (4) resistance tape; (6) sulphuric acid trap; (6) duralumin cavity; (7) drying tube.

43

work, the flame composition was 1.0 1 Hz min-‘, 4.5 1 N? min-‘, 80 ml O2 min-’ and carrier gas, 60 ml Nz min-‘. The volatilization system used for the production of silicon tetrafluoride is shown in Fig. 3. The reaction vessel was a teflon tube (10 cm long, 10 mm i.d.) closed by a polyethylene stopper having three holes, one for injecting concentrated sulphuric acid, another for passing nitrogen carrier gas and the third for transport of gases to the cavity. The generator tube containing the aqueous reaction mixture (see below) was placed in a paraffin oil bath maintained at 135°C. Concentrated sulphuric acid was added through the top from an automatic Zippette reservoir. Reagents The water used throughout was double-distilled from glass apparatus. A sodium silicate solution was prepared by dissolving 3.7726 g of sodium metasilicate (Na$iO,* 5H?O) in 100 ml of water and 42.4 ml of 36% hydrochloric acid, and water was added to give a final concentration of 1.0 mg Si ml-’ in 500 ml. A neutral silicate solution was prepared by dissolving the same amount of sodium metasilicate in 500 ml of water. Sodium fluoride solutions were prepared by dissolving 8.102 g of sodium hydrogenfluoride in 500 ml of 1 M hydrochloric acid or water to give a final fluoride concentration of 10.0 mg ml-‘. A 100.0 g 1-l chloride solution was prepared by dissolving 164.89 g of sodium chloride (analytical reagent) in 1 1 of water. Working solutions were prepared daily from these stock solutions. Recommended procedure for de termination of silicon (A) For 50-500 mg L’ silicon. A set of calibration solutions containing O-500 pg Si ml-’ was prepared from the acidic standard sodium silicate solution (1000 pg Si ml-‘) in six 50-ml polyethylene volumetric flasks. To each flask, 6.5 ml of the acidic sodium fluoride solution and 5 ml of the sodium chloride solution were added, the final concentrations of fluoride and chloride being 1.5 mg ml-’ and 10 mg ml-‘, respectively, in each of the flasks. The solutions were diluted to volume with water. For each measurement, a 0.2~ml aliquot of one of these solutions was transferred to the reaction tube and 1 ml of 98% sulphuric acid was injected. The tube was firmly closed and heated at 135°C for 35 s. The gases formed were carried to the cavity by nitrogen at 100 ml mm-‘. The emission intensity was measured at 540 nm for 20 s. The flame composition used was 1.2 1 Hz mind1 and 4.7 1 Nz min-‘, with 80 ml O2 min-’ supplied to the cavity. The peak height was measured for each silicon concentration and a graph of peak height vs. amount of silicon was plotted (Fig. 4). (B) For O-50 mg t’ silicon. A set of calibration solutions containing O-50 pg Si ml-’ was prepared in a similar way to that described above. The concentrations of fluoride and chloride were again 1.5 mg ml-’ and 10.0 mg ml-‘, respectively. The procedure described above was followed, except that the nitrogen carrier gas flow was increased to 140 ml min-' .

44

IO .!

'5,

I a

J

0 Fig. 4. Calibration graphs for silicon: (0) 20-100 (A) 2-10 pg.

30

\\ . 60 I’:

rg; (0) 20-100

. u. . u . 3o’;b 30” “a 30 Time (sl erg (after dehydration);

Fig. 5. Effect of the trap saturation on emission intensity: (a) blank with new trap; (b) emission from 20 fig of silicon; (c) as (b) after 15 experiments using the same trap; (d) blank using the same trap after 25 experiments.

Determination of silicon in iron ore. A 0.5-g or 1.0-g sample of powdered iron ore sinter was placed in a porcelain dish and treated with 10.0 ml of concentrated hydrochloric acid. The solution of the decomposed sample was evaporated to dryness on a hot plate to separate hydrated silica. The residue was heated at 110°C for at least 1 h. The porcelain dish was cooled and the ash moistened with 10 ml of hot 10% hydrochloric acid to dissolve iron, aluminium and alkali metal salts. The residue was filtered off on a Whatman No. 40 filter paper and washed with 5 ml of cold 1% hydrochloric acid. A second evaporation and dehydration of the filtrate was made to recover the small amount of silica that escaped the first separation. The second portion of silica was filtered off onto a fresh filter and washed with 5 ml of cold 1% hydrochloric acid. The two washed precipitates were dissolved in 5 ml of cold 48% hydrofluoric acid. The solution was quantitatively transferred to a loo-ml polyethylene volumetric flask and diluted to the mark with doubledistilled water, after addition of 10 ml of the 100 g 1-l chloride solution. A

45

0.2-ml portion of the final solution was transferred to the reaction tube, and the procedure described above was followed for the measurement of the silicon emission intensity. The same dehydration procedure was applied to standard silicate solutions in order to give a calibration graph from results obtained under the same conditions as the sample (Fig. 4). RESULTS

Emission spectra The spectrum of the silicon emission in the oxy-cavity was obtained from silicon tetrafluoride generated by adding 0.5 ml of 1 M hydrochloric acid containing 0.5 mg of silicon (as NazSi03 5Hz0) and 2 mg of fluoride (as sodium hydrogenfluoride) to the generator. The system was closed, and 1 ml of 98% sulphuric acid was injected. The gas generated when the system was heated at 135°C was carried to the cavity by a steady stream of nitrogen (60 ml min-‘). A constant white emission was maintained in the oxy-cavity for at least 2 min. The spectrum showed a broad, featureless emission band with the apparent maximum intensity between 540 and 620 nm. The only emissions recorded when the spectrum of the flame alone was scanned were from OH around 306 nm and sodium atoms at 590 nm. The region between 320 and 580 nm was practically free of emission bands, but there was a small background increase with increasing wavelength. To avoid the effect of sodium emission on the background, all subsequent measurements were made at 580 nm. Dagnall et al. [9] published the spectra obtained from several volatile fluorides introduced into a nitrogen-hydrogen diffusion flame. The spectrum obtained for silicon tetrafluoride had six principal band heads over the range 400-700 nm. The authors claimed that at least part of the spectrum could be attributed to SiF, which has been obtained previously in discharge tubes [ 10-121. Shanker et al. [ 131 obtained excited SiO in a flame by chemiluminescent reaction of silicon tetrachloride vapour with oxygen atoms in an argon atmosphere. Various SiO band systems were identified in the region 230-460 nm. By comparing the values for the dissociation energy [14] of the possible silicon molecules in the cavity (187 kcal mol-’ for SiO and 115 kcal mol-’ for SiF), it was assumed that the emitting species was the more stable SiO. To prove that the white emission was due only to silicon species, volatile silicon tetrachloride and hexamethyldisilazane (HMDS) were used for obtaining spectra. When about 1.0 ml of SiC14 or HMDS was placed in the reaction tube and the vapours were carried by hot nitrogen through the sealed system to the cavity, a strong emission was obtained in the cavity for long enough to scan the spectrum. In both cases, the spectra were similar to the spectrum obtained from silicon tetrafluoride. l

46

Development of procedure for determination of silicon The reaction taking place in the generator can be represented as follows [ 15-171: SiO:- + 6 F- + 8 H+ -+ HzSiFs + 3 Hz0 H2SiF6 Lheat SiF,, + 2 HF It has been reported [18, 191 that the evolution of silicon tetrafluoride starts when the reaction mixture reaches the temperature of decomposition of sulphuric acid (>15O”C) into sulphur trioxide and water. The time required for qomplete evolution of silicon tetrafluoride is said to vary from several hours to a few minutes. In this investigation, the evolution of the gas was accelerated by the addition of chloride ions, so that silicon tetrafluoride was evolved within a few seconds. The maximum amount of chloride which did not show any effect on the blank response (2.0 mg) was used for all experiments. With the aim of not changing considerably the concentration of sulphuric acid and of preventing its violent reaction with water, mixtures of known amounts of silicon, chloride (2.0 mg in 0.2 ml of solution) and excess of fluoride were prepared; 0.2-ml aliquots of those solutions were pipetted into the generator, and 1 ml of 98% sulphuric acid was injected when the system was closed. The volatile products formed during the heating period (35 s) were carried to the cavity by a stream of nitrogen. Quantitative recovery of volatilized silicon tetrafluoride is complicated, owing to its strong water affinity. Therefore, the construction and use of the vaporization system must be such that the silicon tetrafluoride along with some hydrochloric and sulphuric acid vapours, can be transferred from the sample solution and discharged into the cavity without retention by aqueous condensate along the way. To achieve this, the teflon tube connecting the generator with the cavity was jacketed with a resistance tape controlled by a Variac variable resistance, which maintained the tube above 100°C. Under these conditions, acid vapours and silicon tetrafluoride were generated, which gave rise to an emission that flashed out of the cavity causing an immediate off-scale recorder deflection. This could be prevented if a sulphuric acid trap containing 10 ml of 98% sulphuric acid and maintained at room temperature was placed in the system between the heated outer tube and the cavity. The trap absorbed water and sulphuric and hydrofluoric acid vapours. The emission intensity from a certain amount of silicon started to increase (Fig. 5) after lo-12 consecutive experiments. Thus, the sulphuric acid in the trap was changed every ten experiments. Effect of amount of fluoride. The minimum amount of fluoride required for the maximum response from silicon tetrafluoride was established. The amounts of fluoride calculated to react with 20 pg and 100 pg of silicon for the formation of silicon tetrafluoride are 54 c(g and 271 pg, respectively. The amounts obtained by experiments and extrapolation (Fig. 6) are 55 and 275 pg respectively. For subsequent experiments, an excess of fluoride was

47

always used to ensure complete conversion of silicon to silicon tetrafluoride (i.e., 300 pg of fluoride for
WO, WI 10 pg Si 1OOfigSi

Intensity (mv)

,I 1po 0

&O

go

. 80 Amount

spa

. of

0.2

0.4

0.6

0.8

1.0

1.2

1.5

2.0 0

5.5 5.0

7.0 38.0

7.0 64.0

6.8 65.0

7.1 65.0

7.0 65.0

790

.

120 fluorldd?pgl

opo rpoo 200

Fig. 6. Effect of amount of fluoride on the emission from: (A) 20 ccg of of silicon.

silicon;(0) 100 rg

48

with increasing flame and cavity temperature. This was confirmed by studying the effect of cooling the cavity. Before every experiment, the flame was burned for 60 s, and then the cavity was cooled by an air blower. The time which elapsed between turning off the flame and turning it on again for the measurement of SiO emission was varied. The results in Table 2 show that as the cavity temperature increases (i.e. as the cooling time decreases) the emission intensity becomes greater. The hydrogen flow could not be increased above 2.0 1 min-’ because the aluminium cavity employed became red hot and started to deform. Therefore, different combinations of nitrogen and hydrogen flow rates were used in order to obtain the highest signal-to-background ratio. A combination of 4.7 1 Nz min-’ and 1.2 1 Hz min-’ was best whilst providing a flame sufficiently cool to prevent deformation of the aluminium cavity. The flame was kept burning during the experiments to ensure a constant cavity temperature. The effect of the flow rate of oxygen to the cavity on the emission of silicon is shown in Table 3. In the absence of oxygen, Sz emission appears from sulphuric acid vapours. An oxygen flow rate of 80 ml min-’ was used in all further experiments. The variation in the emission intensity from 20 and 100 r_1gof silicon with carrier gas flow rate is shown in Table 4. The carrier gas flow rate was not critical in the range 140-200 ml mm -’ for 20 pg of silicon, so that a flow of 140 ml mine1 was used for the determination of <20 pg of silicon. For larger amounts of silicon, the flow rate used was 100 ml min-‘. For > 100 pg of silicon, the peaks tended to broaden and peak area measurements are advisable. TABLE 2 Effect of cavity cooling time on silicon emission Time of cooling (8)

0

Peak height (mv)

23.0

20

45

75

18.0

16.5

13.0

105

120

8.5

7.5

TABLE 3 Effect of oxygen flow rate on the emission intensity of 20 ccg of silicon” Oxygen flow rate (ml min-’ )

0

Peak height (mv)

(S,)

60

80

100

120

160

200

140

170

160

155

105

90

aN, carrier = 140 ml min-‘; H, = 1.2 1 min-I; N, = 4.7 1 min-*.

49 TABLE 4 Effect of carrier gas flow rate on net silicon emission intensity Nitrogen flow rate (ml min-‘)”

60

80

100

140

160

200

240

Intensity (mW

11 58

14 64

15 70

16 52

16 46

15 -

11 -

*02 = 80

20rgSi 100pgSi

ml min-’ ; H, = 1.2 1 min-’ ; N, = 4.7 1 min-’ ; 0.2-ml injected.

Analytical performance Linear calibration plots were obtained for both concentration ranges (Fig. 4). From the slope of the calibration graph, the sensitivity for the determination of silicon was 0.5 mV pg -l. The relative standard deviation for seven measurements on 0.2 ml of solution containing 20 pg of silicon was 2.3%. For 1.0 pg of silicon in the same volume, the relative standard deviation increased to 12%. The detection limit, considered as the absolute amount of silicon required to give a signal-to-noise ratio of 2, was 0.2 pg of silicon in 0.2 ml of solution. Several cations and anions were tested for interference. The emission intensity from 20 pg of silicon in 0.2 ml of solution was measured before and after measuring the emission intensity from 0.2 ml containing 20 E.cgof silicon and the ion to be studied for interference. Various amounts of interfering ions (10-1000 pg) were added. In each instance, a solution containing 1 mg of the interfering ion was added in the absence of silicon, to check the emission arising from the blank. An interference was defined as significant if the signal from silicon was different by two standard deviations (i.e., 5%) from the signal obtained for silicon in the absence of any interfering ion. The anions were added as their acids and most of the cations as their nitrates. The results are shown in Table 5. Strong positive interferences were caused by arsenic and ammonium, both giving an emission in the absence of silicon. Under the conditions of this experiment, arsenic probably forms arsenic trifluoride which carried to the cavity, and causes the blue arsenic-oxide emission in the flame [20] ; ammonia causes the whitish NO-O continuum [21]. Both these emissions interfere spectrally with the measurement of the SiO emission, but arsenic and the slight boron interference can be avoided by using a narrower (0.2 mm) slit, and measuring at 580 nm. Most of the other elements had no significant effect on the determination of silicon. Sulphate and sulphite did not interfere at low concentrations. The problem of trap saturation (see above) appeared after every five experiments when more than 100 pg of sulphate or sulphite ions was added. Therefore the sulphuric acid in the trap was changed every two experiments during the study of such interferences.

50 TABLE

5

Effect of other ions on the determination Amount. of ion added (rg)

Change in emission intensity (W) Interfering ion Pbl’

H&+

10

t3

t4

100 1000 Blank

+6 t8 0

+5 t10 0

NH: 10 100 1000 Blank

of 20 pg of silicon

+28 t44 >lOO OS*

co= +6 -4 t8 0

c,o:-5 -7 -30 2

Nia+ -2 +2 +5 0 ClO; t3 t4 +6 0

Mnl+ +6 +6 t7 0

Zn’+ +5 +4 t6 0

PO:-34 -37 -100 0

Baz+ -2 +3 t5 0

so:+2 +6 +lO 0

Cr” 2.5 +5 t6 0

NO; t4 +7 t15 0

Al* 0 +2 +2 0 BO;-

+4 t6 +37 16

Ge’+ 0 -3 4 0 AsO;t80 > 100 09 OF

aOff scale.

Determination of silica in an iron ore Because of difficulty in obtaining standard samples of low silicon content, the m.e.c.a. method was applied to the determination of silica in an iron ore sinter sample (Bureau of Analysed Samples). The sample was also reported to contain Al, Ti, Mn, Ca, Mg, S, P and V. Possible interferences from the elements present were eliminated by the separation of silicon as insoluble silica [17]. The results obtained by m.e.c.a. (8.48% SiO? for a 0.5-g sample, 8.52% for a 1.0-g sample) were similar to the certified value (8.55% SiOz). DISCUSSION

The technique applied was found to be suitable for the determination of silicon at microgram levels. The vaporisation procedure applied for the generation of silicon tetrafluoride was successful when a few parameters were strictly controlled. For example, a constant temperature (135°C) must be achieved during the experiment for providing rapid generation of the gas. The composition of the reaction medium must be carefully controlled, especially the chloride concentration. Condensation must not be present in the exit tubing from the generator to the cavity, lest silicon tetrafluoride be prematurely absorbed. An analogous procedure has been applied to the determination of fluoride [22].

51 REFERENCES 1 S. L. Bogdanski, M. Burguera and A. Townshend, CRC Crit. Rev. Anal. Chem., 10 (1981) 185. 2 W. J. Price and J. F. H. Roos, Analyst, 93 (1968) 709. 3 J. S. Cartwright, C. Sehens and D. C. Manning, At. Absorpt. Newsl., 5 (1966) 91. 4 L. Capacho-Delgado and D. C. Manning, Analyst, 92 (1967) 563. 5 J. J. McAuliffe, At. Absorpt. New&, 6 (1967) 69. 6 G. F. Kirkbright, M. Sargent and T. S. West, Talanta, 16 (1969) 245. 7 S. A. Ghonaim, Ph.D. Thesis, Birmingham University, 1974. 8 R. Belcher, S. L. Bogdanski, I. H. B. Rix and A. Townshend, Anal. Chim. Acta, 81 (1976) 325. 9 R. M. Dagnall, B. Fleet, T. H. Risby and D. R. Deans, TaIanta, 18 (1971) 155. 10 R. C. Johnson and H. G. Jenkins, Proc. R. Sot., 116 (1927) 327. 11 R. K. Asundi and R. Samuel, Proc. Indian Acad. Sci., 3 (1936) 346. 12 E. H. Eyster, Phys. Rev., 51 (1937) 1078. 13 R. Shanker, C. Linton and R. D. Verma, J. Mol. Spectrosc., 60 (1976) 197. 14 A. G. Gaydon, Dissociation Energies and Spectra of Diatomic Molecules, Chapman and Hall, London, 1968. 15 I. M. Kolthoff and E. B. Sandell, Textbook of Quantitative Inorganic Analysis, MacMillan, New York, 1946, pp. 398-406. 16 C. Hodzic, Anal. Chem., 38 (1966) 1626. 17 R. P. Curry and M. G. Mellon, Anal. Chem., 29 (1957) 1632. 18 R. P. Curry and M. G. Mellon, Anal. Chem., 28 (1956) 1567. 19 B. D. Hold, Anal. Chem., 32 (1960) 124. 20 S. A. Ghonaim, Proc. Sot. Anal. Chem., 11(1974) 138. 21 R. Belcher, S. L. Bogdanski, A. C. Calokerinos and A. Townshend, Analyst, 102 (1977) 220;106 (1981) 625. 22 S. L. Bogdanski, M. Burguera and A. Townshend, Anal. Chim. Acta, 117 (1980) 247.