A. W. Espelund Department of Metallurgy University of Trondheim Trondheim, Norway
I I
dluminothermic Reduction An illustrative experiment
In 1900, H. Goldschmidt in Germany performed the first aluminothermic reductions, also named "Thermite reactions," shortly after aluminum became available on a large scale. Batch-type reactions of this kind, involving only condensed phases, are used for the production of chromium, tantalum, niobium, and other reactive metals from their respective oxides, and for the on-site welding of steel rails. The production of uranium from UFI using magnesium also resembles the thermite reactions based on aluminum as the reductant. This well-established process can be made into a very illustrative experiment, involving some typical calculations in high-temperature physical chemistry. Depending on the degree of treatment, the experiment can he used as a tvnical demonstration in high schools or universities. or in laboratory courses for undergraduate students in chemi s t or ~ metallurw. -. When successful. this raoid reaction is an impressive demonstration of the principles involved and. from the author's exoerience. it has been acceoted enthusiastically by hundreds of students over some years. At universitv level. the time allocated for this experi,ment might bedivided into three parts
both elements Al and Fe is governed by the oxygen activity (or pressure), which can have only one value. This may he represented by the equation 3"FeO" + 2Al = 3Fe + &Oz 3 X (-340) 0 0 f-224,O) The Gibhs free energies of formation a t 2318°K are given in the parentheses. For "FeO," which can have a wide composition range, the hypothetical liquid, stoichiometric FeO, was chosen as standard state Total AG",,,
=
-134.3 kcal
=
-RTlnK
where
A.
1) Introductory lecture in the classroom. 2) Calculation of equilibria, enthalpy, and materials balance (foreach gmup of students independently). 3) After calculations are handed in and corrected, weighing, mixing, and ignition are carried out, supervised by an in-
stNct0r. No complex equipment or expensive material is necessaw. In order to reduce analytical work, the production of iron from magnetite or mill-scale Fez04 is recommended. The following treatment was found to he suitable for university students. Thermochemical Aspects
Equilibrium Consideration is characterized by is assumed that the the three typical situations, shown in Figure 1. onceit has heen pointed out that liquid iron and alu. minum oxide will separate into two phases (a metallic phase and a lighter slag), the temperature at which the oxide starts to solidify (state II) is of special interest as the exchange of matter between the two phases must practically stop when the slag solidifies. This implies that the equilibrium at 231PK must he calculated. The liauid metal must contain a trace of dissolved alu,,,inum and the slag phase some ' ' F ~ O , ~being , the oxide with iron, ~h~ distribution of can he in
Figure 1. initial, intermediate, and final stale 400 /
Journal of Chemical Education
With near stoichiometric amounts of reactants, it may be assumed that are = 1anda.\i,o=l, and thus
For the activities it is reasonable in thfs context to assume that Raoult's law is valid. With, for example, 1 wt 70Al in the alloy, corresponding to a i l 0.02, the value of aFeo is about 6 X 10-4. This shows the strong oxygen affinity of aluminum. A trace of aluminum dissolved in the liquid' iron leads to the oracticallv comolete reduction of FeO from the slag phask. The initial assumption, namely that phases consist mainly of iron metal and A1203 slag, with melting points near the pure components, is thus valid. It is illuminating to consider the oxygen activity in this system. From the Gihbs free energy .. of formation of Alz03 from its components, we obtain po, 10-10 atm for the example with 1 wt 70A1 in the alloy. Why then does atmospheric oxygen, being present above the crucible, not completely oxidize the iron? The answer is that the slag cannot dissolve appreciable amounts of oxygen and transport it to the metal below during the short time a t high temperature. It is thus permissible to consider the system as being of A1203(Fe0) and Fe(A1).
--
--
-
~ . ,t~~j~~~~ h ~ forla simoje ~ ~ charme Is i t possihle to perform this reaction in a crucible (after local ignition), without supplying extra heat? Referring to states I and II, is the chemical energy in state I, when released, sufficient to completely fuse the two pmducts of the reaction? The progress of the reaction can also be visualized on a micro scale. When the reaction starts a t low temperatures, two adjoining grains of the reactants FeaOl and Al may soon he separated by a solid layer of A1203 and thus further reaction is prevented. Only when A1203 is formed a t T > 2318°K in the liquid state, which is able to run away, can a continuous reaction throughout the charge occur. Referring to situations I and 11, and assuming a stoichiometric charge, the enthalpy function is a state property and it is only necessary to consider the initial (3Fe30r ~
-
Figure 2. Enthalpy diagram for alurninothermic reduction
m i g h t % A1203
Figure 3. The binary system Ca0-A1203.
+ 8AI a t 298°K) and final states (9Fe 2318°K). Assuming adiabatic conditions
+ 4Alz03 a t T =
3Fe30r + 8AI = 9Fe + 4AIzOj Enthalpies of formation:3 X (-261) 0 0 4 X (-400) Net enthalpy of reaction: AHzss" = -800 kcal Heatingof products: = 198kcal 9Fe : 9 X HZ%3l8 = 9 X 22 = 461 + 26) = 348 kcsl 4AIzOa: 4 H2982318 546kesl Excess -254 kcal The principles of this calculation are illustrated in Figure
-
Z.
As shown in the figure, the liberated heat should be sufficient to raise the temperature not only to 2045"C, but to 2880°C (boiling point of iron), with about 15% evaporation of the metal (assuming an ambient pressure of 1 atm). However, adiabatic conditions are never achieved in a simple experimental set-up. The heat losses are mainly due to radiation from the top and beating of the crucible. Steady state conditions never prevail, but the heat losses are roughly proportional to the time elapsed between states I and 11. Short reaction times are obtained with an intimately mixed fine powder. The relative magnitudes of the heat losses also depend on the size of the charge. From experience, the following empirical "thermal ratio" bas been found available hedenthalpy of reaction + preheating of charge) heat required for complete fusion of reaction products
=
some local spot. The calories supplied, being orders of magnitude below those released by the main reaction, are omitted from this calculation. The experiment can be performed after this theoretical treatment. Practical details are given later. Heat Balance for a Charge with a Complex Slag Melt temperatures greatly in excess of the melting point of iron should be avoided as they lead to extra wear on the cmcihle and entrainment of iron in the slag, resulting in poor separation of slag and metal. By flux additions, the temperature at which the slag is completely molten may be lowered. This leads to interesting calculations and a number of possibilities arise. A brief examination of the binary phase diagrams of the systems A1203 with SiOz. CaO, and MgO shows the extremely good fluxing capacity of CaO. The system A1203CaO contains five compounds (three congruently and two incongruently melting) and four eutectics, as shown in Figure 3. As an example the composition Ca0:Alz03 = 1% will be selected. This has a melting point of 2023°K (CaO.2A1203), slightly greater than the adjoining A1203-rich eutectic. A balanced equation for this slag composition is given by 3Fe30,
1.2
for a charge weighing 0.5-1 kg, with melt temperatures near 2300QK,and with grain size 1mm. Heat in excess of this new requirement may be used to increase the weight of the iron lump by addition of some scrap (e.g., fine nails) to the charge. As the fusion and heating of this iron to 2318°K require 22 kcal/mole, y mole of iron can he fused according to
--
giving y = 5.5 mole. That is, the total charge can thus be composed of 8 mole of Al, 3 mole of Fe304 (e.g., mill scale) and 5.5 mole of iron. In order to allow for a small content of aluminum in the iron and compensate for Alz03 covering the aluminum granules, it is instructional to add deliberately, for example, 2% extra of the reductant. As will be shown later, the charge has to be ignited in
+ 8AI + XaO = 9Fe + 2(Ca02A1,Oa)
For the compound Ca0.2Alz03, the heat capacity in the solid state and its enthalpy of formation from the respective oxides are known, while the enthalpy of fusion is unknown. A value of 2 cal/DK g atom for the entropy of fusion is assumed, corresponding to AH," = 12 X 2 X 2023 = 48.6 kcal/mole. The enthalpy balance can thus be written Enthalpy of reaction AHzss' Heating of products to 2023°K 9 X 20 9Fe Z(Ca0.2 AlzOa): "2 :s:
-803 = 180
(61.72 + 9.58 X lO-=T- 15.30 X lOJT)dT+ 48.6 = 343 523 Excess -280 kcal
At this new temperature the beat losses will be smaller, so that a lowering of the "thermal ratio" to 1.15 is permissiVolume 52, Number 6, June 1975 / 401
Figure 4. Two different reaction paths f r o m S t a t e i
ture. As vapor evolution may cause an explosion, it must be certain that this chemical is practically anhydrous. (Even the label "CaO" does not guarantee this, unless the package has been hermetically sealed.) The Hz0 in "CaO" is chemically bonded and heating to above 450°C is required for efficient calcination. Lump-sized, burned limestone of the kind delivered in plastic bags to steelworks, has been also used with good results. The lumps were ground in a mortar just prior to use. As a crucible, the use of the clay-bound graphite type is suggested (brand name "Salamander," size around 1 1). This type can he used repeatedly. I t is recommended that this be cut in two lengthwise, the two parts being tied together with wire in order to facilitate separation after the reaction is complete. When half buried in dry sand the molten metal and slag will not run out. The crucible should be left overnight in a drying oven to remove absorbed moisture, and preferably still be warm when used.
to S t a t e 1 1 .
ble. z mole of iron can thus be included in the charge, according to
giving z = 8.75 mole of Fe. Alternatively, we may calculate the enthalpy needed to form liquid supercooled A1203 and CaO, and with suhsequent isothermal slag formation. For the two oxides, identical heat capacities in the solid and liquid states are assumed, so that the enthalpies of fusion are independent of temperature within the range of interest. Enthalpy of reaction
-800
By assuming that the enthalpy of slag formation can be neglected, this gives 1563
+ m ~ ) i . i 5= 800
giving z = 6.65 mole of Fe. An attempt to illustrate the two alternatives is made in Figure 4. The lack of precise data for the system CaOA1203 gives rise to the different values for z, namely 8.75 and 6.65 mole of Fe. The charge may thus have the relative composition: 3 mole of Fe3O4, 8 mole (+ excess) of Al, 2 mole of CaO, and 7.7 1mole of metallic Fe. Similar calculations may be performed for any other Ca0:AIzO8 ratio, giving a number of possibilities. The second of the two methods is the simpler and can be used for anv com~osition.including eutectics. It may he instructive to consider the distribution of calcium between the slag and the metal phases. A rough calculation for the Ca-da0-equilibrium gives a very low calcium metal activity at any conceivable oxygen activity created by a small amount of aluminum dissolved in the iron.
*
Practical Aspects Preparation of Materials Mill Scale, Fez04 can be obtained from any steel works. I t should he ground in a cast iron mortar to a grain size of about 1mm. Larger metallic particles should be removed. Aluminum Powder, preferably with grain size less than 1mm, may be obtained commercially. Calcium Oxide, or "quicklime," strongly absorbs mois-
402 / Journal of Chemical Education
Ignition Ignition of the main charge is performed by the reaction
as suggested by Goldschmidt. A stoichiometric mixture of about 20 g of BaOz and 2 g of aluminum dust is poured into a hollow paper tube placed in the center of the C N C ~ ble when about half filled with the main charge, thus ensuring good contact. The tube is pulled out after the crucible has been filled. The BaOz-A1 mixture can be ignited in many ways. It is possible to apply the reaction between sugar and potassium chlorate, i.e.
which can be ignited by one or two drops of concentrated sulfuric acid. A few grams of each chemical is mixed carefully and placed on top of and in good contact with the other igniter in the middle. The sulfuric acid is added through a hole in a simple lid of asbestos. The reaction with sulfuric acid takes just enough time such that the operator may withdraw safely. If the main reaction does not start after a few seconds, it has been found to be due either to some physical reason (poor positioning of the two ignitors or poor mixing of the reactants) or some error in the calculations. For safety reasons, the operator ought to wear gloves and a face mask. When no gas evolution takes place, however, there is no sputtering or ejection from the crucible. The white glow of the charge can be seen through the hole in the asbestos lid, and after a while the crucible becomes red hot. Evaluation of the Experiment The time required for completion of the reaction may be measured (usually less than a minute). After cooling, the slag and metal are removed from the crucible. The two phases can be easily separated by means of a hammer and chisel. The metal may be assumed to consist of pure iron, so that the yield of metal can he calculated. The malleable character of the metal button in contrast to the brittleness of the slag can be noticed. General References Rosenqvisf. T., "Plinelplesof ExVaetive Mcfs1iurgy.l. MeCrsu-Hill Book Co., 1974. Barin, I.. and Knacke, 0 , "Thermochmieal Pmpenier of Inorganic Suhsfsnees," Sp"nger~Vedag. Berlin 1973. Elliott. J. F.. Gleiser, Molly. Ramakrishna, V.. "Th~rmorhcmiafryfor Steelmaking.'' Addison-Wesley. Pexsmon Press, Oxford, W63. Kubarchcwiki. 0 , Evans. E. L., Alcock, C. B., "M~telluxicslThermaeheminfry," 4th Ed.. PolgamanPres.. Oxford, 1967.
