it-.D. JAMIESON
4084 [ CONTRIBUTIOK FROM
THE
AND
C. R.MASSON
1-01. 84
ATLANTIC REGIOXAL LABORATORY, A-ATIOXAL RESEARCH COUNCILOF CANADA, HALIFAX,X.S . ]
Decarbonization of Liquid Iron-Carbon Alloys BY W. D.
JAMIESON AND
C. R. MASSON
RECEIVEDAUGUST23, 1962 T h e decarbonization of iron-carbon alloys in argon-oxygen mixtures was studied a t oxygen pressures from to 10-2 atm., carbon contents from 2.0 to 4.570and temperatures from 1290 to 1510'. The alloys were held in recrystallized alumina boats and crucibles of varying dimensions. The rate of evolution of CO CO2 could be represented by either of the expressions d(CO CO,)/dt = kpOaA,,A,, or d ( C 0 COs)/dt = k'pO~.4,,(A,, Acm), where A,, and A , , are the areas of the gas-melt and crucible-melt interfaces, respectively, and k and k' are constants the values of which are given b y k = 1.3 x 10*Oe-lO.OOO (=3,50u) R2' molecules/sec. a t m . ~ m and . k' ~ = 9.6 X 10'ge-lo~OOO (' a,600)'KT molecules/sec. atm. (21131.~. The present results are not sufficiently extensive to distinguish between these possibilities. The results support the conclusion t h a t elimination of carbon monoxide occurs solely a t the crucible-melt interface, and a kinetic expression is derived which accounts in part for these observations.
+
+
+ +
Introduction undertaken in an attempt to define more closely The oxidation of carbon dissolved in liquid iron some of the factors which govern the rate under has been the subject of numerous investigations. laboratory conditions. The decarbonization was The results of early work have been summarized studied in argon-oxygen mixtures a t high flow rates fully elsewhere.* The basis of present concepts and a t partial pressures of oxygen in the range regarding the mechanism is due largely t3 Darken3 IO-? to lop4 atm. Rates were determined by whose expression for the rate in terms of carbon measuring the rates of evolution of CO and C02. monoxide evolution (molecules/sec.) under condiExperimental tions favorable for nucleation in the open-hearth Apparatus.-The apparatus is illustrated in Fig. 1. furnace may be written "Welding" grade argon was dried by passage over magnesium perchlorate A and the last traces of oxygen were removed, when required, by passage through "pyrophoric" where Do is the diffusion coefficient of oxygen in the copper B at 200"." Oxygen at known partial pressure was added t o the metered argon stream by allowing the gas to metal, A is the area of the slag-metal interface and pass downward through furnace D, 40 mm. in diameter, ([OIse - [O])!Al is the concentration gradient of packed with cupric oxide wire. T h e temperature of this oxygen across a postulated boundary layer of thick- furnace, as indicated by a noble metal thermocouple located in the wire at the hottest zone, was varied between 600 and ness A1 in the metal adjacent to the slag. 100O0, depending on the desired pressure of oxygen. The The importance of nucleation as a rate-deter- temperature was uniform to within f 5" over a length of 10 mining step was emphasized by L a r ~ e nfollowing ,~ cm. in the hottest zone. The copper oxide was regenerated the work of Korber and 0elsen.j This concept was before each experiment by a stream of oxygen. This furnace further expanded by Richardson,6 Sims7 and served also to oxidize a trace of CO present as a n impurity in the argon; COI mas removed in trap E cooled with liquid Vallet.* Although it has been generally agreed, in oxygen. the light of abundant experimental evidence, that The reaction furnace G was wound with Pt-lOyGRh wire such a process must play a significant role in deter- and was equipped with a horizontal tube of recrystallized mining the over-all rate, i t has been generally alumina of 35 to 40 mm. i.d. The argon-oxygen mixture was preheated by passage through a region P packed with recognized that nucleation alone cannot be rate- alumina chips held in place by plugs of mullite wool. The limiting under open-hearth conditions. D a r k e ~ ~alloy ~ , ~M was contained in a recrystallized alumina crucible";' has suggested that the over-all rate is governed supported on a sled of stabilized zirconia which, in turn, simultaneously by the rates of the two processes of rested on rails of mullite tubing which served to protect the tube from thermal shock. Temperatures were oxygen transfer from slag to metal and from the furnace measured by a Pt-Pt/l3Y0 R h thermocouple located inside metal into expanding bubbles of carbon monoxide; one of the mullite rails, with its junction directly beneath these processes were regarded as being so closely the crucible. The temperature was uniform to within f I" coupled that the rate could be controlled by chang- over a length of 7 cm. and was automatically controlled t o within 0.5" by a saturable core reactor in series with the ing the conditions a t either interface. furnace winding. The impedance of the reactor was varied In spite of the success of this theory in account- by a controller which sensed the temperature of the furnace ing, a t least semi-quantitatively, for the observed by means of the thermocouple and operated through :L rates of decarbonization under industrial condi- magnetic amplifier. The gas whieh issued from the furnace was analysed chemitions, a detailed description of the mechanism has cally for CO and COZ in the apparatus shown in Fig. 2 . remained uncertain.ID The present study was COZwas removed in trap R which iyas cooled in liquid oxygen. Residual CO mas oxidized by cupric oxide a t 400" in (1) Issued a s X . R . C . N o . 7083. furnace B and retained as COZ in trap T, also cooled with ( 2 ) P. 1'.Carter, I v c n ond S f e e l , 2 6 , 427 (1953). liquid oxygen. The remaining gas was vented via ".4sca(3) L. S . Darken in "Basic Open-Hearth Steelmaking," p. 592 rite" in tube U. The COZ retained in either trap was de( A . I . M . E . ,1931). dCO/dt = DoA( [Olse
-
[Ol)/AZ
(1)
(4) B. M . Larsen, Tvans. R.Z.LWM.E., 1 4 5 , 67 (1941). Milt,Kaisev-Wilhelm Insl. / I C Y Eisenforschiing, 17, 39 (1935). (6) F. D. Richardson, Disc. F Q Y O ~ USOC., Y 4, 335 (1918). (7) C. E . Sims, Tvnns. A . I . M . E . , 172, 176 (1947). (8) P. Vallet, Rev. .U?f., 51, 709 (1951); Jmn an3 Steel, 28, 4A3 ( 5 ) F. Kdrber and W. Oelsen,
(193.5). (9) L. S. Darken in "The Physical Chemistry of Steelmaking." ed. J. F. Elliott, Technology Press of M.I.T., Cambridge, Mass., 19.58 p. 101. (10) B. hl. Larsen and I,. 0 . Sordahl in "Physical Chemistry of
Process Metallurgy," Vol. 8, P a r t 2 , ed. G. R . St. Pierre, Interscience Publishers, I n c , S e w I'ork, h-.Y . , 1961, p. 1111. (11) F. R . XIeyer and G. Ronge, Angew. Cliem., 6 2 , 637 (1939). ( l l a ) N o r e A D D ~ IS D Pnoou.-Morganite "Triangle R R " crucibles mere employed. Sei.eral crucibles were suhjected t o porosity measurements in a mercury porosimeter a t pressures u p t o 50,000 p.s.i. S o n e showed any sign of porosity a t this level of pressure, indicating t h a t , if pores exist, t h e y must h e less t h a n 17.q. in radius, T h e authors wish t o thank Dr. n. S. Montgomery, Department of Mines and Technical Surveys, Ottawa, Ont.. for kindly supplying these measurements.
DECARBONIZATION O F LIQCIDIROiX-CIRBON
Nov. 5, 1962
.‘ILLOYS
40S5
Fig. 2.-Analytical section : R,T, liquid oxygen trap‘ ; S, furnace packed with cupric oxide; U, “ . k : i r i t e ” ; \., cold finger; W, multiple-range McLeod gage; X,J’, calibrated volume. G
Fig. l.--lpparatus : .4, magnesium perchlorate; B, pyrophoric copper, 200”; C, flow meter; 11, furnace packed nit11 cupric oxide; E, liquid oxygen trap; F,J, dibutyl phthalate; G , reaction furnace; €7, window; L,L, evacuated bulbs; M, alloy; P, alumina chips, niiillite wool. gassed, condensed in V, and measured with McLeod gage \V in the calibrated volume between stopcocks X and Y. In later experiments samples of gas both entering and leaving the reaction furnace were analysed mass spectrometrically for CO, COZ and 0 2 . T h e samples were withdrawn in previously evacuated bulbs L, Fig. 1. When this procedure was employed, the gas which issued from check value J was allowed to flush the annular space between the alumina tube and the refractory tube of the furnace. Materials.-Alloys were prepared from iron sponge of high purity for which a typical analysis (p.p.m.), as supplied by the manufacturer, was Si, 8; M n , 3; M g and Na, 2 ; Xi, 1; Cu and Ag, each < 1; 43 other elements sought but not detected. This material was melted in argon with “spectroscopic” graphite powder in a graphite crucible also of “spectroscopic” grade. Before use the crucible was further purified by heating in a stream of Hz until HzS could no longer be detected in the effluent. “Pin” samples of liquid alloy were withdrawn in silica tubes and quenched in water. These were crushed, sieved and analysed for carbon as well as for possible contamination by silicon, which never was detected (Si < 0.001 yo). Alloys of lower carbon content were prepared by mixing material prepared in this way with iron sponge. Procedure.-Alloys were charged to or withdrawn from the furnace while it was flushed with argon, and a flow of argon was maintained until the required temperature was established. Zero time was taken as the time at rvhich oxygen was introduced to the argon stream. When analyses were performed chemically, decarbonization was allowed t o proceed for a measured interval during which the products were collected continuously. T h e melt was then held in a slow stream of purified argon until the analysis was completed, whereupon decarbonization was continued for a further interval. This procedure avoided undue depletion of the melt in carbon during an analysis. When analyses were performed mass spectrometrically, the rate of decarbonization a t the time of sampling was determined directly from the flow rate and the composition of the effluent gas. Analysis of samples taken at the inlet allowed rates to be normalized for slight variations in the oxy-
gen pressure and permitted a check on the percentage conversion of oxygen. This technique was employed in most of the experiments reported. At the end of a series of measurements the alloy was removed, quenched in water and analysed for carbon. The carbon content at any time during an experiment was estimated from the final carbon content and the measured rates.
Results Figure 3 shows the results of experiments designed to check the accuracy with which argonoxygen mixtures of known composition could be prepared. The gas from the copper oxide furnace D, Fig. 1, was passed through graphite a t 1000° to convert the oxygen to carbon monoxide which subsequently was analysed chemically. The line in Fig. 3 corresponds to the data compiled by Coughlinl?for the free energies of formation of Cu20 and CuO. The present results are well within the limits of uncertainty of this line. The cornposition of the argon-oxygen mixtures was independent of flow rate up to 60 ~ m . ~ , / s e cthe . , highest flow rate employed, and of the total volume of argon which passed through the furnace up to a maximum of 150 liters. Partial pressures of oxygen were reproducible to within 2.57,. Blank experiments in which oxygen was omitted from the argon stream yielded rates of decarbonig. C/sec., or less than lYc zation of the order of of the lowest rates observed in the presence of oxygen. Allowing for differences in crucible-melt area, blank rates were of the same order as observed by Parlee, et al.13 and were attributed mainly to reaction of the melt with alumina. These rates decreased rapidly with time a t 1375-150O0, the deceleration being greater at the higher temperatures. This may be due to establishment of equilibrium between the melt and oxygen contained by the crucible or to inhibition of the crucible(12) J. P. Coughlin, Bureau of Mines Bulletin 542, U. s. Government Printing O 5 c e , Washington, D. C., 1954. (13) N. A. Parlee, S. R. Seagle and R. Schuhmann, Jr., T r a n s . SOC. A . I . M . E . , ala, 132 (1~5s).
MeC.
mT.D. JAMIESON
4086 I
I
I
AND
c. R. M.4SSON 6
I
I
Vol. 84 I
I
I
I
I
EXPT. 5 --X ri
'
PO, LI
9 IO I$/ TEMPERATURE PKJ
II
Fig. 3.-Log $01in gas us. l / T for cupric oxide: 0, experimental; -, from compilation of Couglilin.'*
(Atm, x
Fig. 4.-Rate vs. p02: X, boat, T = 1380". [CJinitia~ = 4.5070, [C]rinal= 3.77%; 0 , boat, T = 1440°, [C]initlsl = 4.50%, [CIfinal = 3 i2%; 0 , boat, T 1381", [Clinitial 4.3470, [C]rinai = 3.56%; W, cylindrical crucible, T = 1381" [Clinitial = 4.56%, [Clfinal 4.42%. A,, = 1.70 Am = 6.84 0, cylindrical crucible, T = 1381°, [C]initiai= 4.1570, [Clfinal = 3.89%, A g m = 7.55 cm.*, A c m = 13.09
melt reaction by formation of a solid product a t the interface. Higher blank rates were observed when lime crucibles were used in place of alumina. This was attributed to a greater crucible-melt between successive measurements while the teminterfacial area due to the rough texture of the lime perature of the copper oxide furnace was adjusted. I n other experiments (47, 48) the temperature of crucibles as compared with those of alumina. In all experiments the main product of decarbon- the copper oxide furnace decreased continuously ization was COZ while the content of CO in the during the experiment. The effect of carbon concentration was studied products was usually less than 5%. Checks of over the range 2.0 to 4.5% .C. Variation of the the material balance in various experiments showed that the loss of carbon equalled to within 570 the carbon content was accomplished by allowing detotal carbon evolved as gaseous oxides. Similarly, carbonization to proceed for measured intervals the oxygen in the products always corresponded between successive determinations and its vaIue a t within 3% to that consumed. This established that any time was calculated from the measured rates. extraneous effects, such as the deposition of FeO in Data therefore were obtained using the same initial the furnace tube, were insignificant. Some alloys alloy and in the absence of variations in the gaswere analyzed c~lorimetrically~~ for aluminum a t melt and crucible-melt interfacial areas. Values the end of an experiment. n'one was detected of [C] are plotted against time for four experiments in Fig. 5 . (A1 < 0.001%). The effect of temperature was studied over the Rates of decarbonization were independent of 1290-1510°. The same initial alloy was range flow rate above 20 ~ m . ~ / s e under c. all conditions studied. In general, flow rates of 52 ~ m . ~ / s e c .employed in each series of measurements, and were employed. Percentage conversions for oxy- rates were normalized for slight variations in oxygen gen were usually less than 20%. Exceptions were partial pressure. Plots of log [ ( - dC/dt)/pOn] for melts of surface area greater than 5 cm.2, when against the corresponding values of l / T yielded conversions as high as 40% were observed a t flow a series of lines of approximately equal slope but rates of 52 ~ m . ~ / s e cand . oxygen pressures of 0.5 with significantly different intercepts depending on the geometry of the melt. X 10-2atm. Table I shows the results of experiments designed With other variables held constant, the rate of to establish the effect of gas-melt interfacial area. decarbonization was directly proportional to the to 0.9 X Shallow cylindrical crucibles of varying diameter oxygen pressure over the range 1 X atm., as shown in Fig. 4. In some of these were employed. experiments (5, 8, 12) the alloys were held in argon Discussion The results have been interpreted on the basis (14) P. H. Scholes and D. V. Smith, Analyst, 83, 615 (1958); J . Iron Steel Ins;., 195, 190 (1960). that the primary product of the reaction is CO and
Nov. 5, 1962
DECARBONIZATION OF LIQUIDIRON-CARBON ALLOYS I
I
I
4087
K
O
15
10
5
20
cm2 I
A REA(,.
TIME ( S e c o n d s x IO3)
Fig. 5.-Carbon concentration PIS. time for various experi- Fig. 6.-Rate/pOsA,, us. A,, at 1380", from data in Table 1. ments: 0 , cylindrical crucible, T = 1380°, A,, = 7.56 cm.z, The rate thus is given by the expression Ac, = 14.90 cm.2, pop = 6.30 X 10-3 atm.; 0, cylindrical d(C0 COz)/dt = kpOnAgmAcm (2) crucible, T = 13t?0°, A,, = 5.23 em2, A,, = 11.82 CIII.~, 902 = 1.03 X 10-2 atm.; 0 , boat, T = 1454O, A g m = 3.3 where k is the slope of the line in Fig. 6. cm.*,A,, = 6.1 cm.*,pOz = 1.67 X lO-*atm.; -I-,boat, T = A further expression which was also observed 1454'. A , = 3.5 crn.2, A,, = 6.5 ern.*, PO, = 1.51 X lo-* to fit the data within the limits of error is atrn. d(C0 COz)/df = k'pO&gm(Aom Agm) (3)
+
+
+
that COz arises almost exclusively by subsequent oxidation of CO as a homogeneous reaction in the gas phase. CO has been observed by Blyholder and EyringI5 to be the primary product in the reaction of graphite with oxygen a t high temperatures. TABLE I EFFECTOF ALLOYGEOMETRY AT 1380'
Expt.
[Cl, %
cm.*
Mass of alloy, g.
37 38 39 40 41 42 43 44 45 46 47 48
3.71 4.17 4.19 4.20 4.21 4.00 4.15 4.39 4.40 4.47 4.49 4.02
5.23 1.72 7.56 5.19 2.96 7.40 1.77 2.84 4.63 2.90 1.70 7.55
29.29 12.45 39.16 43.73 25.44 53.21 8.20 24.50 22.53 16.17 12.99 29.59
Mean
Agm.
pOz, atm. X 10-2
1.03
0.80 0.63 0.456 1.36 0.913 0.467 0.27 0.227 1.33 0.359 0.246
Rate of decarbonization, molecules
X lO'I/sec.
4.02 0.678 3.92 2.41 2.49 7.31 0.283 0.542 0.668 1.74 0,282 1.38
The most significant observation concerning the rate was derived from the data in Table I. The rate per unit pressure of oxygen was not a simple function of the gas-melt interfacial area, A g m , as expected for a single surface reaction in which diffusion of reactants to this interface is rate-controlling. Inspection of the data revealed that the rate per unit pressure per unit gas-melt interfacial area was directly proportional to the cruciblemelt interfacial area A,, as shown in Fig. 6. Values of A,, were estimated from the weight of the alloys assuming a value of 6.88 g./cm3 for the density of liquid iron. (15) G . Blyholder and H. Eyring, J. Phys. Chem., 61, 682 (1957).
as indicated by a linear plot of [d(CO -I- Cot)/ dt)/A,,pOz against (Agm Acm). The present data are not sufficiently extensive t o distinguish between these possibilities as, fortuitously, the ratio A c m / ( A g m Acm) was approximately constant in all experiments. Values of log k for four experiments are plotted against 1/T in Fig. 7. A single point has been included for the data in Fig. 6. The value of k as a function of temperature is given approximately by
+
+
k = 1.3 X lOzoe--io,ooo(*s.soo)~RT molecules/sec. atm. em.'
A similar plot yields for K' k' = 9.6 X
101se-la.aoo(**,500)irlT
molecules/sec. atm. cm
These plots do not eliminate slight variations in the temperature coefficient observed with different initial alloys, so that the over-all temperature coefficient appears to be complex. The simultaneous dependence of the rate on Agm and A c m must be interpreted to mean that the rate is defined by a t least two steps, neither of which alone is rate-limiting under the conditions employed. This supports the conclusion of Darkeng for the rate under open-hearth conditions. In line with the mechanism proposed by Darken it is assumed that the rate is governed by diffusion of oxygen atoms across boundary layers in the metal a t the gas-melt and crucible-melt interfaces. I t is also assumed that the rate a t which oxygen reacts a t the gas-melt interface is governed by the extent to which available sites are occupied. For diffusion across boundary layers of thickness AIl and AZ2 a t the gas-melt and crucible-melt interfaces, respectively
-
dCO/dt = DGAgm([OI,m
and
[OI)/ALI
- [O]cm)/AL
dCO/dt = DoAom( [O]
(4)
(5)
where [O],,, [ O ] c m and [O] are, respectively, the steady-state concentrations of oxygen a t the gas-
n'. D. JAMIESON
AND
c. R.AIASSON
Vol. 84
where k l and k-1 are the specific rate constants for adsorption and desorption, respectively, and 0 is the fraction of the available surface which is covered. From (10) and ( l l ) ,in the steady state dCO/dt = 2k1Agmfi02(l - e ) - 2k-1Ag,8 (12) 0 is related to [O],, by the expression
e
=
[Ol,m/b
(13)
where b is the value of [ O ] g m which corresponds to complete coverage of surface sites. Halden and Kingery16 have shown that, for melts in alumina crucibles a t 1570°, complete coverage corresponds approximately to a monolayer of FeO. The value of b therefore map be taken as the concentration of oxygen in FeO. Substituting the value of [ O ] g m from (9) in (13) and neglecting the small term mpCO/b [C],, one obtains
0
e
Substituting for 0 in (12) and simplifying yields for the over-all reaction dCO U
P
0"
0.6
6.2
5.8
I04/TEMPERATURE
6.6
(OK.).
Fig. 7.-Log (rate/pOJ,,A,,) os. 1 / T : 0, A,, = 5.19 c m 2 , A,, = 15.08 cm.2, [C]i,iti,i = 4.3870, [Clftnai = 3.9370; X, A,, = 2.96 A,, = 10.58 cm.2, [C]initini= 4.47%, [Clri,,~ = 3.73y0; e,A,, = 7.40 A,, = 17.49 cm.2, [ClLnitili= 4.60'3?0, [C]iin.i = 3.487,; 0,A,, = 1.77 cm.2,A,, = 4.94 cm.2, [Clinitia~= 4.29$&, [Clfinoi = 4.0570.
melt and crucible-melt interfaces and in the bulk of the metal, and Do is the diffusion coefficient of oxygen. It is assumed that concentrations may be used in place of activities in the range under consideration. Elimination of [ O ]between equations (4) and ( 5 ) yields dCO/dt
=
DoAgmAcm( IOIgm
-
[O]om)/
(ALA,,
+ ALAgm)
(6)
The rate of evolution of CO a t the crucible-melt interface is given by
-
k-?pCOaAm
(7)
k?aA,,( [C],,[O]m
- WZPCO)
(8)
dCO/dt = k2[C],m[O]o,aAc,
or dCO/dt
where k p and k-? are the specific rate constants for the formation and dissociation of CO a t this interface, [C],, is the interfacial concentration of carbon and LY is the fraction of the measured crucible-melt area a t which nucleation of CO occurs. Substitution of [O],, from equation (8) in equation (6) yields
If it is assumed that oxygen is adsorbed as a molecular complex a t the gas-melt interface prior to dissociation, the rates of condensation and remoral of oxygen are given by dOs(ads.),/dt = kIA,,pOz(l
and -dOp(ads.)/dt
=
k-lAg,8
- 8)
+ (l/Z)dCO/dt
(10)
(11)
(15)
Further resolution of equation (15) in terms of measurable parameters is possible by substituting for [Clem,using the relationship dCO/dt
DcAo,( [C]
-
[C]crn)/Al~
(16)
where DC is the diffusion coefficient of carbon in the melt. This yields a quadratic expression for which a simple solution is not obtained except under limited conditions I t is readily shown, however, that [C],, in equation (15) may be replaced by [ C J when DCA,,[C]/A/~ >> dCO/dt. In the present study [C] was generally of the order of IO2? molecules 'cc. and < I c m was of the order of 10 cm 2 . Using Morgan and Kitchener'sl' value of 5 X 10-5 cm.*/sec. for DC and assuming that A12 does not differ greatly from the value of 0.003 cm. estimated by Darken for open-hearth conditions, the value of DcA,,[C]IAZ2 is observed to be of the order of 1.7 X loz1rnolecules/sec , which is significantly larger than the rates of 10" to 10" molecules /see. which were generally observed. Replacement of [C],, by [ C ] in equation (15) therefore appears justified under the present conditions. Equation (15) yields the observed dependence of the rate on oxygen pressure if it is assumed that, a t the low pressures employed, k@O2