- Email: [email protected]
Determination of low-level carbon in tungsten wire by combustion gas chromatography
DETERMINATION OF LOW-LEVEL CARBON IN TUNGSTEN WIRE BY COMBUSTION GAS CHROMATOGRAPHY PETER CUKOR@, CARMINE PERSUNI and ARTHUR RUSSELL General Telephone and Electronics Laboratories, Waltham, Mass. 02154, U.S.A. (Received 3
October
1974. Accepted 22 ~oue~~r
1974)
Summary-A
combustion gas-chromatographic technique for the determinatton of trace amounts of carbon in tungsten wire is descrrbed. The method involves the oxidation of the tungsten wire m a quartz oven at looo”. The liberated gases are swept mto a cooled sample-loop in a gas-samplmg valve. Upon completion of the oxidation process, the contents of the sample loop are introduced into a gas chromatograph. The use of a 3-ft long column of silica gel allows separation of carbon dioxide and oxygen. The presence of oxygen requires that the hot-wire detector used be equipped with filament-protecting circuitry. Calibration curves are constructed by using organic and tungsten carbide standards. A limit of detection of @2pg carbon can be achieved with a precisron of better than 10%.
The presence of low level impurities can profoun~y affect the chemical and physical properties of tungsten wire and consequently influence its behaviour as a filament in an incandescent lamp. Carbon is an element of particular concern since large amounts of it in the form of graphite or as a viscous organic compound are used to prevent oxidation in the hot drawing of tungsten. The result of this treatment produces a significant carbon deposit on the surface of the wire with the formation of tungsten carbide below the surface. In the subsequent process of converting such tungsten wire into lamp filament, most of the carbon is removed by various heat-treatment operations. It has been shown that large quantities of carbon adversely affect the performance of a filament and reduce its life.’ It has also been proposed, however, that trace quantities of carbon are beneficial to filament performance because they minimize oxidation and deterioration due to the presence of traces of water vapour in the iamps In order to investigate the effect of various amounts of carbon on filament performance, a method capable of detecting traces of carbon in tungsten wire and filaments was
required. In addition to possessing high ~nsitivity, this method must be reliable and have the capability of detecting carbon whether it is present as surface contamination or as bulk impurity. It must also be applicable to the analysis of tungsten at each step of the filament manufacturing process. A literature search indicated that the determination of carbon in metals is normally carried out by methods which employ either direct instrumental analysis or combustion followed by instrumental analysis (see Table 1). The sensitivity attainable by the spark-source mass spectrometer is extremely high. Unfortu~tely, the carbon monoxide and dioxide which are to be detected are present either in the residual gases of the vacuum system or as surface contaminants in the mass spectrometer. Thus, the background reduces the sensitivity for carbon detection. Recent work indicates that the background can be reduced considerably by using a cryogenic accessory panel to lower the background noise during measurement.s The emission spectrograph has been shown to be capable of performing carbon determination at the
Table I. Methods of carbon analysis Combustion-lnstfumentot
/
Spark source moss spectrometry
methods
RBrsut_s~~e
Indfub$oan~Rf)
I .I
1 Methods of detectton
Grovimetric
571
Volumetric
Conductonce
-
572
PETERCUKOR.CARMINEPERSIAN]and ARTHURRUBELI
1Oppm level. The sample is excited in a d.c. arc and the CN bands are used for measurement.4 Malamand describes the determination of carbon by measuring the ultraviolet emission spectrum in the region 50100 nm.5 Neutron activation is unsuitable for the determination of carbon because of the element’s low neutroncapture cross-section. Charged-particle activation analysis was used, however, by Rook and Schweikert.‘j who reported an ultimate detection limit at the ppM (parts per milliard) level for carbon. This method is essentially restricted to surface analysis because of the limited penetration of a charged particle. A more attractive nuclear method uses 1%MeV photons to activate carbon. The resultant radioactivity of “C is measured with a coincidence-counting detector. Lutz and Masters’ have performed the determinations in high-purity metals and reported carbon values of less than 1 ppm. Some interferences can be encountered if oxygen and other contaminants are present at higher levels. Combustion methods involve the thermal decomposition of a sample in a stream of oxygen in which the carbon is converted into carbon dioxide, which in turn is measured by a suitable detection system. The combustion apparatus most often used consists of a radiofrequency induction furnace. Sometimes, however, a wire-wound tube furnace is preferred. The more frequently used detection systems are summarized in Table 1. Gravimetric procedures have been used extensively. The sample, with an appropriate flux, is combusted in a “Globar” or similar type of furnace. The resulting carbon dioxide is collected in an absorption bulb and weighed.’ This method 1s mainly suitable for samples containing not less than about 1OOppm of carbon. Volumetric methods employing the collection and measurement of carbon dioxide gas utilize similar combustion techniques, but the carbon dioxide collected is measured manometrically in gas collection vessels of known volumes.’ The carbon content is calculated by using gas-pressure relatlonships. Lower levels of carbon may be determined by using
the various instruments manufactured by the Laboratory Equipment Corporation (LECO). In these instruments, combustion takes place in pure oxygen in a crucible heated by high-frequency induction, and the liberated carbon dioxide is measured in various ways. One model of the instrument uses conductometric detection by passing the carbon dioxide through a solution of barium hydroxide and measurmg the change in resistivity of the solution due to the formation of barium carbonate. Other LECO models are equipped either with gas-chromatographic detection systems utilizing a thermistor-type Wheatstone bridge as a sensor, or with infrared detection systems employing gas cells. These instruments provide rapid. reliable analysis for carbon at trace level (2pg) and above in a variety of matrices. One of the problems encountered in applying these instruments to refractory metals is the incomplete loading in a radiofrequency induction coil. As a result, the necessary temperature for complete combustion is not obtained. Accelerators, such as tin or iron, can be used to promote complete combustion; however, these materials always contain some carbon. Thus. a blank is introduced which lowers the sensitivity for carbon detection. The combustion systems which do not utilize induction furnaces are not subject to these errors, but they are severely limited in terms of the maximum temperature they can attain. For this reason, combustion in a conventionally-heated resistance furnace is usually restricted to the determination of carbon in organic materials. Often the furnace is coupled with a gas chromatograph, to take advantage of the high detection-sensitivity of this instrument.” Tungsten wire and tungsten carbide oxidize at temperatures between 600 and 1000”. It should therefore be possible to use a conventionally heated quartz furnace in conjunction with a gas chromatograph to determine trace amounts of carbon in tungsten wire samples. This paper reports the development of such a method. In its final form. the method can detect 0.2 pg of carbon (in a 05-g sample) with a precision of better than lo:<.
Gas chromotograph
0, He
FIG.
1. Schematic diagram of the apparatus.
-hot
ware detector
Carbon in tungsten
573
wire
length
Overall
GC. 24/40 Standard taper with teflon sleeve
InjectIon
joint quartz
i+
cylinders
J
To
Quartz
varlac
sample
boat
timer i”
Power
FIG. 2. D&&d
drawing
EXPERIMENTAL
Apparatus The schematic diagram of the apparatus used is shown in Fig. 1. The combustion tube and boats were fabricated from “Supersil” quartz and are shown in Fig. 2. Heating was accomplished by means of a small tube furnace heated with nichrome wire surrounding the quartz tube and controlled by a “Variac” and timer. Copper tubing with “Swagelock” fittings was used for all connections to the gassampling valve and gas chromatograph (GC). Temperature readings were checked with a Leeds & Northrup potentiometer. Gas-chromatographic analyses were performed with a Perkin-Elmer Model 900 gas chromatograph equipped with a six-port injection valve Model 154-0068 containing a 14-cm long, @3-cm bore copper sample-loop. It is absolutely essential that the gas-sampling valve be free from leaks m spite of the extremes in temperatures (from - 180 + 150 to + 50”) it IS subjected to durmg the procedure. An Alltech Associates six-port valve was also evaluated and was found to be reliably leakproof. The GC detector was a hot-wire thermal-conductivity cell equipped with a protective circuit to prevent filament burn-out.
The column used was a 3-ft long, 3-mm bore, stainlesssteel tube packed with Biorad silica gel (100-200 mesh). Injection port temperature: 60”. Column temperature: 60” isothermal. Manifold temperature: 150”. Detector: operated at 150” and 225 mA filament current. Helium carrter-gas flow-rate: 40 ml/min; Oxygen purifier trap: molecular sieve 5A maintained at - 7s” during operation, purged overnight with hehum at 50”. Oxygen flow-rate: 100 ml/min. Hehum purge-gas Flow-rate: 100 ml/mm. Volume of sampling loop: 1 cm3.
Tungsten carbide secondary standard, previously found by classical gravimetric combustion technique to contain 69; carbon. Potassium hydrogen phthalate NBS standard, prepared so as to contain 5 fig of carbon per 100 ~1 of solution. Etch solution consisting of 50 parts of hydrofluoric acid and 1 part of nitric acid. Matheson “Ultra High Purity Grade” oxygen and “High Purity Grade” helium.
of combustion
oven.
Procedure
Weigh the sample into a quartz boat, put this into the quartz furnace tube and purge for at least 5 mm with purified oxygen, venting it through the sample injection valve (see Fig. 1). Cool the sample loop with liquid nitrogen (a polystyrene cup can be used Instead of a Dewar flask). After cooling the sample loop, heat the furnace around the combustion boat. The furnace temperature should be above 900” and attained m less than 1 min. Burn the sample for Smin in oxygen (more or less combustion time may have to he used, depending on sample size, matrix, and oxygen Aow). After the heat cycle, immediately purge with helium for 30~~. Turn the sample valve to the injection position and remove the liquidnitrogen trap. Heat the sample loop to drive off the sample. The attenuation and detector-current settings will have to be monitored because of the great difference between the amounts of oxygen and carbon dioxide passing through the detector. Upon completion of the run, heat the column at 120” for 10 min. Determine the carbon dioxide from the peak area and calculate the amount of carbon by means of the calibration curve. Sample preparation Calibration curves. Ahquots of potassium hydrogen phthalate solution containing l-2Opg of carbon were transferred by micropipette into quartz boats. The water
FIG. 3. Plastic
holder used m etchmg (diagrammatic).
tungsten
wire
574
PETER CUKOR, CARMINE PERSIANI and ARTHUR RU~.YGELL
was evaporated under an infrared lamp. The boats were placed in the furnace and their contents were combusted and analysed. Various amounts of tungsten carbide powder, from 5 to 2Opg in carbon content, were weighed into quartz boats on a microbalance. The boats were loaded into the furnace and their contents were analysed. Wire A. The wires were free from splits (hairline cracks) and were chopped to uniform length, blended, etched wrth potassium hydroxide solution, and electropolished. Etching and electropolishing removed about 20% of the original weight. Wire B. The wire was heated for 2min in a reducing atmosphere at 2500”. Wire C. The wire was etched with a mixture of 50 parts of hydrofluoric acid and 1 part of nitrrc acrd. The amount of materral removed was determined by weighmg the samples before and after etching. The diameter of the wire was measured with a precision micrometer before and after etching. Etching was accomphshed by using a
jig prepared by cutting off the top and bottom portions of a square cross-section polyethylene reagent bottle. Equally spaced slots were cut in opposite sides of the jrg and a single strand of wtre was woven back and forth between them (Fig. 3). The whole assembly was placed m a plastic beaker contaming the etching solutron. Etching was quenched after the desired time by copious washing wrth water. The wire was dried and removed from the jig. The portions of wire which were in contact with the plastic during etching were cut out and discarded. Thus etching procedure yielded considerably better precisron than the conventional approach in which small strands of the wire were Immersed m the etching solution. RESULTS Typical Figure
AND DISCUSSION
chromatograms 4(a)
shows
the
are
presented
separation
in
which
Fig. can
4. be
go-
(a) 0070 -
60-
a2
Oxygen
:
: ; L z
50-I
40
; 0 9,
30-
[r 20IO-
OL-
la;:,:/, II
X8
2
3
4
Time,
90
5
6
7
8
m,n
-
6070-
60Oxygen 5040-m 3020-
10-i
0'
I
2
1
3
I
4 Time,
FIG.
4. (a) Typical gas chromatogram
I
5
6
7.
min
obtained in analysis of samples; (b) typical gas chromatogram
for a blank run.
Carbon in tungsten wire Table 2. Results obtained on typical tungsten samples SampleSLZC, Sample Wm A Wxe A Wne B Wue C, unetched WIR C, etched
Carbon
content,
AVerage,
9
PPf=
PPm
0150 0500 0250 O-100 0100
12. 12. 17. 18, 19 13, 14, 15, 14. 15, IS, 14 IO, II. IO, 11 1400 109, 110, I58
156 143 IO.5 1400 126
achieved between the carbon dioxide and oxygen. The carbon dioxide is readily detected in the presence of a relatively high level of oxygen. Figure 4(b) is a chromatogram representing the total background signal of the system. A calibration curve obtained by using tungsten carbide and potassium hydrogen phthalate standards was a straight line passing through the origin and exhibited a slope of 11.5 cm2 peak area per pg of carbon. The results of the analyses of wire samples are summarized in Table 2. It can be seen that, for low carbon content, optimum precision is attained when using sample weights of at least 0.25 g. With a sample weight of 0.15 g, the relative standard deviation is 22%, while at sample weights of 0.25 and OS g, it is 6% The determination of the carbon content of tungsten wire is complicated by the inhomogeneity of the carbon distribution in the wire, as shown by the dependence on the degree of etching to which the sample has been subjected. It is likely that this circumstance is responsible for discrepancies found in the comparison of analytical results from different laboratories. It is essential that this be understood and that a standardized etching procedure be adopted as part of sample preparation, in order to obtain agreement among different laboratories. Furthermore, agreement must be reached on the extent of etching to be performed before analysis. In the course of generating the calibration curves, it was found that the background signal corresponded to 0.1 pg of carbon. The limit of detection is often defined as the signal that is twice the noise level. According to this definition, the limit of detection of the method is 0.2pg of carbon. One very important factor in maintaining the blank at a low level is the low-temperature molecular-sieve trap used to remove organic impurities from the oxygen gas. Mass spectrometric analysis of the oxygen used showed the presence of about 20ppm of hydrocarbons in the gas. These organic materials are converted into carbon dioxide while passing through the heated oven and yield a high blank having a magnitude proportional to the heating time. The use of a molecular-sieve trap eliminated this large source of background error. A crucial operating variable was found to be the combustion time. Ignition for too short a time produces low results, probably because of incomplete liberation of the carbon from the tungsten matrix. Heating for a prolonged period also leads to slightly lower results, probably owing to aerosol formation
575
in the sample loop. The optimum heating time has to be determined for each type of sample. Tungsten carbide powder, for instance, yielded a relative response of 73, 110 and 100 when 25 5.0 and 7.0min heating times were used with similar sample sizes. The optimum heating time for most samples was between 5 and 7 min. The gas-sampling valve also serves as a preconcentration cold-trap. During combustion, the sample loop is maintained at - 180”, while during sample introduction, its temperature is elevated to about 150”. The Perkin-Elmer gas-sampling valve did not give completely leak-free operation when subjected to these extreme temperatures. Alltech Associates manufacture a gas-introduction valve free from O-rings and containing a plastic rotor resistant to high temperatures. At an advanced stage of this project, this valve was evaluated and was found to be superior in performance. The geometry of this valve required the establishment of new optimum operating conditions, i.e., flow-rates, column temperature. The size and shape of the sample loop has been optimized. The use of larger volume traps yielded larger amounts of condensed oxygen, which was then difficult to separate from the carbon dioxide. Smaller traps did not provide enough volume to condense all the carbon dioxide; consequently, low results were obtained when they were used. In the course of the development of this method, a sample loop without a packing was found to be optimal, for it produced relatively narrow carbon dioxide peaks. Packed sample loops yielded flat wide peaks. The use of an empty sample loop required the application of liquid-nitrogen cooling. Dry iceacetone mixture (-80”) failed to condense all the carbon dioxide while ethanol-liquid nitrogen mixture (- IlO>) did not prove to be reliable in providing constant temperature. The use of liquid-nitrogen cooling resulted in the condensation of some oxygen in the sample loop together with the carbon dioxide. The oxygen-carbon dioxide mixture was subsequently introduced into the gas chromatograph, where the two gases were separated. Passage of the oxygen through the column and detector demanded certain precautions. The columns found most satisfactory for the separation of oxygen and carbon dioxide were packed either with silica gel or with Porapak T. Either of these columns had a useful life of about 40 determinations but exhibited deterioration. The Porapak T packing, which is a styrene-divinylbenzene copolymer, was slowly attacked by oxygen, and carbon dioxide was formed, leading to an increasingly large blank value. The silica gel column’s performance also slowly deteriorated, smaller and smaller responses being obtained for the same size of sample. A set of calibration curves obtained at various times during the useful life of a silica gel column, clearly showed the deterioration. These observations make it mandatory that a working curve be prepared on each day of operation. In our experiments a new column was
576
PETERCUKOR,CARMINE PERSIANI and ARTHURRUSTELI
prepared after every forty analyses. It is possible, however, to recondition a column by heating it overnight at 300”. The hot-wire thermal-conductivity detector used has to be of the type which is equipped with a protective circuit so that relatively large amounts of oxygen passing through it will not burr; out the tungsten filaments. The sensitivity of the method was dramatically increased by using the flame-ionization detector instead of the thermal-conductivity detector. The carbon dioxide formed during combustion was converted into methane, followed by measurement with a flameionization detector.” The conversion into methane was achieved by using a nickel powder catalyst at 280”. A stainless-steel tube 12.7 cm long and 3 mm in diameter was packed with lOO-mesh nickel powder and inserted in the manifold of the Perkin-Elmer 900 GC between the flame-ionization detector and the T-fitting used for mixing the column effluent with the hydrogen detector-gas. The manifold oven was maintained at 280”. Thus, the hydrogen gas served a dual purpose; it reacted with carbon dioxide to form methane and also provided the fuel for the flame jet of the detector. This scheme was, however, abandoned because the oxygen passing through the gas chromatograph poisoned the catalyst after a few determinations, necessitating lengthy regeneration in hydrogen gas. Should the trapped oxygen and carbon dioxide be separated before sample introduction, for instance by gradual elution of the gases from a packed trap, the catalytic conversion approach would become feasible. It should be pointed out. however, that no tungsten sample has so far been found which required greater sensitivity of detection than that attainable with the hot-wire thermal-conductivity detector.
Although this work is primarily concerned with the determination of carbon in tungsten, in principle the method could be extended to the determination of carbon in other matrices. The method in its present form may be used for the analysis of any metal that oxidizes in an oxygen atmosphere at temperatures beI& 1100”. This maximum temperature may be increased by substituting a platinum-wound alumina furnace for the present combustion tube, provided that the new furnace does not introduce a serious blank problem. Acknowledgements-The authors most gratefully acknowledge Mrs. Amelia Fermin’s help in doing some of the experimental work. Mr. Kenneth Early is thanked for design and fabrication of the combustion oven. The authors are also extremely appreciative of the wealth of useful information provided by Dr. Robert Long of GTE Sylvama Chemical and Metallurgical Division, Towanda, Pa., and by Dr. Edward Passmore of GTE Sylvania Lighting Division, Danvers. Mass.
REFERENCES 1 R. Reid, Sylvania Technologist, 1952. 5, 75 2. B. Kopelman, ibid., 1949, 2, 13. 3. F. Konishi and N. Nakamura, Advances WIMuss Spctrometry, Vol 5, ed. A. Quayle, p. 547. Institute of
Petroleum, London, 1971. 4. A. V. Kozlova and P. D. Korzh, Sb. Nauchn. Tr. Magnitogorsk Gorno-met. Inst., 1968, 5, 145.
5. F. Malamand, O&e Nat. Etud Rech. Aerospatlalrs Tech. Notr 105,
p. 41, 1967.
6. H. L. Rook and E. A. Schweikert. Anal. Chem.. 1969. 41, 958. 7. G. J. Lutz and L. W. Masters, ibid., 1970, 42, 948. 8. J. Kuch, Encycl. lnd. Chem. Anal.. 1966. 2. 516. 9. H. Goto. T. Watanabe and K. Suzuki, &i.‘Repts. RPS. Insts. Tohoku Univ., 1958.
10, 2.
10. E. Pella and B. Colombo, Mlkrochim. Acta, 1973, 637. 11. K. Porter and D. Volman, Anal. Chem., 1962. 34, 790.
October
1974. Accepted 22 ~oue~~r
1974)
Summary-A
combustion gas-chromatographic technique for the determinatton of trace amounts of carbon in tungsten wire is descrrbed. The method involves the oxidation of the tungsten wire m a quartz oven at looo”. The liberated gases are swept mto a cooled sample-loop in a gas-samplmg valve. Upon completion of the oxidation process, the contents of the sample loop are introduced into a gas chromatograph. The use of a 3-ft long column of silica gel allows separation of carbon dioxide and oxygen. The presence of oxygen requires that the hot-wire detector used be equipped with filament-protecting circuitry. Calibration curves are constructed by using organic and tungsten carbide standards. A limit of detection of @2pg carbon can be achieved with a precisron of better than 10%.
The presence of low level impurities can profoun~y affect the chemical and physical properties of tungsten wire and consequently influence its behaviour as a filament in an incandescent lamp. Carbon is an element of particular concern since large amounts of it in the form of graphite or as a viscous organic compound are used to prevent oxidation in the hot drawing of tungsten. The result of this treatment produces a significant carbon deposit on the surface of the wire with the formation of tungsten carbide below the surface. In the subsequent process of converting such tungsten wire into lamp filament, most of the carbon is removed by various heat-treatment operations. It has been shown that large quantities of carbon adversely affect the performance of a filament and reduce its life.’ It has also been proposed, however, that trace quantities of carbon are beneficial to filament performance because they minimize oxidation and deterioration due to the presence of traces of water vapour in the iamps In order to investigate the effect of various amounts of carbon on filament performance, a method capable of detecting traces of carbon in tungsten wire and filaments was
required. In addition to possessing high ~nsitivity, this method must be reliable and have the capability of detecting carbon whether it is present as surface contamination or as bulk impurity. It must also be applicable to the analysis of tungsten at each step of the filament manufacturing process. A literature search indicated that the determination of carbon in metals is normally carried out by methods which employ either direct instrumental analysis or combustion followed by instrumental analysis (see Table 1). The sensitivity attainable by the spark-source mass spectrometer is extremely high. Unfortu~tely, the carbon monoxide and dioxide which are to be detected are present either in the residual gases of the vacuum system or as surface contaminants in the mass spectrometer. Thus, the background reduces the sensitivity for carbon detection. Recent work indicates that the background can be reduced considerably by using a cryogenic accessory panel to lower the background noise during measurement.s The emission spectrograph has been shown to be capable of performing carbon determination at the
Table I. Methods of carbon analysis Combustion-lnstfumentot
/
Spark source moss spectrometry
methods
RBrsut_s~~e
Indfub$oan~Rf)
I .I
1 Methods of detectton
Grovimetric
571
Volumetric
Conductonce
-
572
PETERCUKOR.CARMINEPERSIAN]and ARTHURRUBELI
1Oppm level. The sample is excited in a d.c. arc and the CN bands are used for measurement.4 Malamand describes the determination of carbon by measuring the ultraviolet emission spectrum in the region 50100 nm.5 Neutron activation is unsuitable for the determination of carbon because of the element’s low neutroncapture cross-section. Charged-particle activation analysis was used, however, by Rook and Schweikert.‘j who reported an ultimate detection limit at the ppM (parts per milliard) level for carbon. This method is essentially restricted to surface analysis because of the limited penetration of a charged particle. A more attractive nuclear method uses 1%MeV photons to activate carbon. The resultant radioactivity of “C is measured with a coincidence-counting detector. Lutz and Masters’ have performed the determinations in high-purity metals and reported carbon values of less than 1 ppm. Some interferences can be encountered if oxygen and other contaminants are present at higher levels. Combustion methods involve the thermal decomposition of a sample in a stream of oxygen in which the carbon is converted into carbon dioxide, which in turn is measured by a suitable detection system. The combustion apparatus most often used consists of a radiofrequency induction furnace. Sometimes, however, a wire-wound tube furnace is preferred. The more frequently used detection systems are summarized in Table 1. Gravimetric procedures have been used extensively. The sample, with an appropriate flux, is combusted in a “Globar” or similar type of furnace. The resulting carbon dioxide is collected in an absorption bulb and weighed.’ This method 1s mainly suitable for samples containing not less than about 1OOppm of carbon. Volumetric methods employing the collection and measurement of carbon dioxide gas utilize similar combustion techniques, but the carbon dioxide collected is measured manometrically in gas collection vessels of known volumes.’ The carbon content is calculated by using gas-pressure relatlonships. Lower levels of carbon may be determined by using
the various instruments manufactured by the Laboratory Equipment Corporation (LECO). In these instruments, combustion takes place in pure oxygen in a crucible heated by high-frequency induction, and the liberated carbon dioxide is measured in various ways. One model of the instrument uses conductometric detection by passing the carbon dioxide through a solution of barium hydroxide and measurmg the change in resistivity of the solution due to the formation of barium carbonate. Other LECO models are equipped either with gas-chromatographic detection systems utilizing a thermistor-type Wheatstone bridge as a sensor, or with infrared detection systems employing gas cells. These instruments provide rapid. reliable analysis for carbon at trace level (2pg) and above in a variety of matrices. One of the problems encountered in applying these instruments to refractory metals is the incomplete loading in a radiofrequency induction coil. As a result, the necessary temperature for complete combustion is not obtained. Accelerators, such as tin or iron, can be used to promote complete combustion; however, these materials always contain some carbon. Thus. a blank is introduced which lowers the sensitivity for carbon detection. The combustion systems which do not utilize induction furnaces are not subject to these errors, but they are severely limited in terms of the maximum temperature they can attain. For this reason, combustion in a conventionally-heated resistance furnace is usually restricted to the determination of carbon in organic materials. Often the furnace is coupled with a gas chromatograph, to take advantage of the high detection-sensitivity of this instrument.” Tungsten wire and tungsten carbide oxidize at temperatures between 600 and 1000”. It should therefore be possible to use a conventionally heated quartz furnace in conjunction with a gas chromatograph to determine trace amounts of carbon in tungsten wire samples. This paper reports the development of such a method. In its final form. the method can detect 0.2 pg of carbon (in a 05-g sample) with a precision of better than lo:<.
Gas chromotograph
0, He
FIG.
1. Schematic diagram of the apparatus.
-hot
ware detector
Carbon in tungsten
573
wire
length
Overall
GC. 24/40 Standard taper with teflon sleeve
InjectIon
joint quartz
i+
cylinders
J
To
Quartz
varlac
sample
boat
timer i”
Power
FIG. 2. D&&d
drawing
EXPERIMENTAL
Apparatus The schematic diagram of the apparatus used is shown in Fig. 1. The combustion tube and boats were fabricated from “Supersil” quartz and are shown in Fig. 2. Heating was accomplished by means of a small tube furnace heated with nichrome wire surrounding the quartz tube and controlled by a “Variac” and timer. Copper tubing with “Swagelock” fittings was used for all connections to the gassampling valve and gas chromatograph (GC). Temperature readings were checked with a Leeds & Northrup potentiometer. Gas-chromatographic analyses were performed with a Perkin-Elmer Model 900 gas chromatograph equipped with a six-port injection valve Model 154-0068 containing a 14-cm long, @3-cm bore copper sample-loop. It is absolutely essential that the gas-sampling valve be free from leaks m spite of the extremes in temperatures (from - 180 + 150 to + 50”) it IS subjected to durmg the procedure. An Alltech Associates six-port valve was also evaluated and was found to be reliably leakproof. The GC detector was a hot-wire thermal-conductivity cell equipped with a protective circuit to prevent filament burn-out.
The column used was a 3-ft long, 3-mm bore, stainlesssteel tube packed with Biorad silica gel (100-200 mesh). Injection port temperature: 60”. Column temperature: 60” isothermal. Manifold temperature: 150”. Detector: operated at 150” and 225 mA filament current. Helium carrter-gas flow-rate: 40 ml/min; Oxygen purifier trap: molecular sieve 5A maintained at - 7s” during operation, purged overnight with hehum at 50”. Oxygen flow-rate: 100 ml/min. Hehum purge-gas Flow-rate: 100 ml/mm. Volume of sampling loop: 1 cm3.
Tungsten carbide secondary standard, previously found by classical gravimetric combustion technique to contain 69; carbon. Potassium hydrogen phthalate NBS standard, prepared so as to contain 5 fig of carbon per 100 ~1 of solution. Etch solution consisting of 50 parts of hydrofluoric acid and 1 part of nitric acid. Matheson “Ultra High Purity Grade” oxygen and “High Purity Grade” helium.
of combustion
oven.
Procedure
Weigh the sample into a quartz boat, put this into the quartz furnace tube and purge for at least 5 mm with purified oxygen, venting it through the sample injection valve (see Fig. 1). Cool the sample loop with liquid nitrogen (a polystyrene cup can be used Instead of a Dewar flask). After cooling the sample loop, heat the furnace around the combustion boat. The furnace temperature should be above 900” and attained m less than 1 min. Burn the sample for Smin in oxygen (more or less combustion time may have to he used, depending on sample size, matrix, and oxygen Aow). After the heat cycle, immediately purge with helium for 30~~. Turn the sample valve to the injection position and remove the liquidnitrogen trap. Heat the sample loop to drive off the sample. The attenuation and detector-current settings will have to be monitored because of the great difference between the amounts of oxygen and carbon dioxide passing through the detector. Upon completion of the run, heat the column at 120” for 10 min. Determine the carbon dioxide from the peak area and calculate the amount of carbon by means of the calibration curve. Sample preparation Calibration curves. Ahquots of potassium hydrogen phthalate solution containing l-2Opg of carbon were transferred by micropipette into quartz boats. The water
FIG. 3. Plastic
holder used m etchmg (diagrammatic).
tungsten
wire
574
PETER CUKOR, CARMINE PERSIANI and ARTHUR RU~.YGELL
was evaporated under an infrared lamp. The boats were placed in the furnace and their contents were combusted and analysed. Various amounts of tungsten carbide powder, from 5 to 2Opg in carbon content, were weighed into quartz boats on a microbalance. The boats were loaded into the furnace and their contents were analysed. Wire A. The wires were free from splits (hairline cracks) and were chopped to uniform length, blended, etched wrth potassium hydroxide solution, and electropolished. Etching and electropolishing removed about 20% of the original weight. Wire B. The wire was heated for 2min in a reducing atmosphere at 2500”. Wire C. The wire was etched with a mixture of 50 parts of hydrofluoric acid and 1 part of nitrrc acrd. The amount of materral removed was determined by weighmg the samples before and after etching. The diameter of the wire was measured with a precision micrometer before and after etching. Etching was accomphshed by using a
jig prepared by cutting off the top and bottom portions of a square cross-section polyethylene reagent bottle. Equally spaced slots were cut in opposite sides of the jrg and a single strand of wtre was woven back and forth between them (Fig. 3). The whole assembly was placed m a plastic beaker contaming the etching solutron. Etching was quenched after the desired time by copious washing wrth water. The wire was dried and removed from the jig. The portions of wire which were in contact with the plastic during etching were cut out and discarded. Thus etching procedure yielded considerably better precisron than the conventional approach in which small strands of the wire were Immersed m the etching solution. RESULTS Typical Figure
AND DISCUSSION
chromatograms 4(a)
shows
the
are
presented
separation
in
which
Fig. can
4. be
go-
(a) 0070 -
60-
a2
Oxygen
:
: ; L z
50-I
40
; 0 9,
30-
[r 20IO-
OL-
la;:,:/, II
X8
2
3
4
Time,
90
5
6
7
8
m,n
-
6070-
60Oxygen 5040-m 3020-
10-i
0'
I
2
1
3
I
4 Time,
FIG.
4. (a) Typical gas chromatogram
I
5
6
7.
min
obtained in analysis of samples; (b) typical gas chromatogram
for a blank run.
Carbon in tungsten wire Table 2. Results obtained on typical tungsten samples SampleSLZC, Sample Wm A Wxe A Wne B Wue C, unetched WIR C, etched
Carbon
content,
AVerage,
9
PPf=
PPm
0150 0500 0250 O-100 0100
12. 12. 17. 18, 19 13, 14, 15, 14. 15, IS, 14 IO, II. IO, 11 1400 109, 110, I58
156 143 IO.5 1400 126
achieved between the carbon dioxide and oxygen. The carbon dioxide is readily detected in the presence of a relatively high level of oxygen. Figure 4(b) is a chromatogram representing the total background signal of the system. A calibration curve obtained by using tungsten carbide and potassium hydrogen phthalate standards was a straight line passing through the origin and exhibited a slope of 11.5 cm2 peak area per pg of carbon. The results of the analyses of wire samples are summarized in Table 2. It can be seen that, for low carbon content, optimum precision is attained when using sample weights of at least 0.25 g. With a sample weight of 0.15 g, the relative standard deviation is 22%, while at sample weights of 0.25 and OS g, it is 6% The determination of the carbon content of tungsten wire is complicated by the inhomogeneity of the carbon distribution in the wire, as shown by the dependence on the degree of etching to which the sample has been subjected. It is likely that this circumstance is responsible for discrepancies found in the comparison of analytical results from different laboratories. It is essential that this be understood and that a standardized etching procedure be adopted as part of sample preparation, in order to obtain agreement among different laboratories. Furthermore, agreement must be reached on the extent of etching to be performed before analysis. In the course of generating the calibration curves, it was found that the background signal corresponded to 0.1 pg of carbon. The limit of detection is often defined as the signal that is twice the noise level. According to this definition, the limit of detection of the method is 0.2pg of carbon. One very important factor in maintaining the blank at a low level is the low-temperature molecular-sieve trap used to remove organic impurities from the oxygen gas. Mass spectrometric analysis of the oxygen used showed the presence of about 20ppm of hydrocarbons in the gas. These organic materials are converted into carbon dioxide while passing through the heated oven and yield a high blank having a magnitude proportional to the heating time. The use of a molecular-sieve trap eliminated this large source of background error. A crucial operating variable was found to be the combustion time. Ignition for too short a time produces low results, probably because of incomplete liberation of the carbon from the tungsten matrix. Heating for a prolonged period also leads to slightly lower results, probably owing to aerosol formation
575
in the sample loop. The optimum heating time has to be determined for each type of sample. Tungsten carbide powder, for instance, yielded a relative response of 73, 110 and 100 when 25 5.0 and 7.0min heating times were used with similar sample sizes. The optimum heating time for most samples was between 5 and 7 min. The gas-sampling valve also serves as a preconcentration cold-trap. During combustion, the sample loop is maintained at - 180”, while during sample introduction, its temperature is elevated to about 150”. The Perkin-Elmer gas-sampling valve did not give completely leak-free operation when subjected to these extreme temperatures. Alltech Associates manufacture a gas-introduction valve free from O-rings and containing a plastic rotor resistant to high temperatures. At an advanced stage of this project, this valve was evaluated and was found to be superior in performance. The geometry of this valve required the establishment of new optimum operating conditions, i.e., flow-rates, column temperature. The size and shape of the sample loop has been optimized. The use of larger volume traps yielded larger amounts of condensed oxygen, which was then difficult to separate from the carbon dioxide. Smaller traps did not provide enough volume to condense all the carbon dioxide; consequently, low results were obtained when they were used. In the course of the development of this method, a sample loop without a packing was found to be optimal, for it produced relatively narrow carbon dioxide peaks. Packed sample loops yielded flat wide peaks. The use of an empty sample loop required the application of liquid-nitrogen cooling. Dry iceacetone mixture (-80”) failed to condense all the carbon dioxide while ethanol-liquid nitrogen mixture (- IlO>) did not prove to be reliable in providing constant temperature. The use of liquid-nitrogen cooling resulted in the condensation of some oxygen in the sample loop together with the carbon dioxide. The oxygen-carbon dioxide mixture was subsequently introduced into the gas chromatograph, where the two gases were separated. Passage of the oxygen through the column and detector demanded certain precautions. The columns found most satisfactory for the separation of oxygen and carbon dioxide were packed either with silica gel or with Porapak T. Either of these columns had a useful life of about 40 determinations but exhibited deterioration. The Porapak T packing, which is a styrene-divinylbenzene copolymer, was slowly attacked by oxygen, and carbon dioxide was formed, leading to an increasingly large blank value. The silica gel column’s performance also slowly deteriorated, smaller and smaller responses being obtained for the same size of sample. A set of calibration curves obtained at various times during the useful life of a silica gel column, clearly showed the deterioration. These observations make it mandatory that a working curve be prepared on each day of operation. In our experiments a new column was
576
PETERCUKOR,CARMINE PERSIANI and ARTHURRUSTELI
prepared after every forty analyses. It is possible, however, to recondition a column by heating it overnight at 300”. The hot-wire thermal-conductivity detector used has to be of the type which is equipped with a protective circuit so that relatively large amounts of oxygen passing through it will not burr; out the tungsten filaments. The sensitivity of the method was dramatically increased by using the flame-ionization detector instead of the thermal-conductivity detector. The carbon dioxide formed during combustion was converted into methane, followed by measurement with a flameionization detector.” The conversion into methane was achieved by using a nickel powder catalyst at 280”. A stainless-steel tube 12.7 cm long and 3 mm in diameter was packed with lOO-mesh nickel powder and inserted in the manifold of the Perkin-Elmer 900 GC between the flame-ionization detector and the T-fitting used for mixing the column effluent with the hydrogen detector-gas. The manifold oven was maintained at 280”. Thus, the hydrogen gas served a dual purpose; it reacted with carbon dioxide to form methane and also provided the fuel for the flame jet of the detector. This scheme was, however, abandoned because the oxygen passing through the gas chromatograph poisoned the catalyst after a few determinations, necessitating lengthy regeneration in hydrogen gas. Should the trapped oxygen and carbon dioxide be separated before sample introduction, for instance by gradual elution of the gases from a packed trap, the catalytic conversion approach would become feasible. It should be pointed out. however, that no tungsten sample has so far been found which required greater sensitivity of detection than that attainable with the hot-wire thermal-conductivity detector.
Although this work is primarily concerned with the determination of carbon in tungsten, in principle the method could be extended to the determination of carbon in other matrices. The method in its present form may be used for the analysis of any metal that oxidizes in an oxygen atmosphere at temperatures beI& 1100”. This maximum temperature may be increased by substituting a platinum-wound alumina furnace for the present combustion tube, provided that the new furnace does not introduce a serious blank problem. Acknowledgements-The authors most gratefully acknowledge Mrs. Amelia Fermin’s help in doing some of the experimental work. Mr. Kenneth Early is thanked for design and fabrication of the combustion oven. The authors are also extremely appreciative of the wealth of useful information provided by Dr. Robert Long of GTE Sylvama Chemical and Metallurgical Division, Towanda, Pa., and by Dr. Edward Passmore of GTE Sylvania Lighting Division, Danvers. Mass.
REFERENCES 1 R. Reid, Sylvania Technologist, 1952. 5, 75 2. B. Kopelman, ibid., 1949, 2, 13. 3. F. Konishi and N. Nakamura, Advances WIMuss Spctrometry, Vol 5, ed. A. Quayle, p. 547. Institute of
Petroleum, London, 1971. 4. A. V. Kozlova and P. D. Korzh, Sb. Nauchn. Tr. Magnitogorsk Gorno-met. Inst., 1968, 5, 145.
5. F. Malamand, O&e Nat. Etud Rech. Aerospatlalrs Tech. Notr 105,
p. 41, 1967.
6. H. L. Rook and E. A. Schweikert. Anal. Chem.. 1969. 41, 958. 7. G. J. Lutz and L. W. Masters, ibid., 1970, 42, 948. 8. J. Kuch, Encycl. lnd. Chem. Anal.. 1966. 2. 516. 9. H. Goto. T. Watanabe and K. Suzuki, &i.‘Repts. RPS. Insts. Tohoku Univ., 1958.
10, 2.
10. E. Pella and B. Colombo, Mlkrochim. Acta, 1973, 637. 11. K. Porter and D. Volman, Anal. Chem., 1962. 34, 790.