Use of Chlorine to Remove Magnesium from Molten Aluminum

Estéfano Aparecido Vieira1, Jose Roberto de Oliveira1, Gianni Ferreira Alves1, Denise Crocce Romano Espinosa2 and Jorge Alberto Soares Tenório2

1Federal Institute of Espírito Santo Department of Metallurgical and Materials Engineering, Av. Vitória, 1729, Vitória-ES, 29040-780 Brazil 2Polytechnic School University of São Paulo Department of Metallurgical and Materials Engineering, Av. Prof. Mello Moraes 2463, São Paulo SP, 05508-030 Brazil

Removal of Mg from aluminum scraps, known as demagging, has been widely applied in the aluminum industry. This work discusses bubble-formation theories and magnesium kinetic removal from aluminum scraps using chlorine and inert gas fluxing. The interfacial area of the bubbles and residence time were estimated using a mathematical model. To inject gaseous chlorine, three types of nozzles were used with varying internal diameter. In addition, a porous plug, as well as varying input chlorine flow and concentration were used. The use of lower chlorine concentration improves efficiency because the interfacial tension is reduced therefore, more and smaller bubbles are formed. The model proposed herein is consistent with the experimental data. [doi:10.2320/matertrans.M2011256]

(Received August 19, 2011; Accepted December 13, 2011; Published February 25, 2012) Keywords: aluminum, reﬁning, reactions rates

1. Introduction The first reaction to occur when chlorine is purged in molten aluminum is reaction (1); next, aluminum chloride Magnesium is introduced as an alloy element to produce reacts with dissolved magnesium following reaction (5). No aluminum cans; however, most aluminum alloys do not gas emissions are observed when the aluminum chloride contain Mg as an alloying element. Consequently removal of bubbles have enough time to react completely with Mg may be necessary if different types of scraps are mixed magnesium. The reaction kinetics will depend on parameters during the aluminum recycling process. The removal of such as temperature, magnesium concentration, metal stirring contaminants is frequently necessary in the secondary and contact area. According to Lagowski12) the reaction aluminum industry.13) Chlorine may also be helpful in between pure chlorine and dissolved magnesium has a removing alkaline metals such as Na, K, and Li. Another use maximum efficiency at around 710°C. 1) of Cl2(g) in the Al foundry is related to the removal of Fu et al. observed the same behavior described by hydrogen and also to improving the removal of solid particles Lagowski, and also stated that below 710°C emissions of 4,5) such as TiB2,Al2O3, MgO and Al4C3. chlorine and aluminum chloride occur, but above this Several techniques have been developed to remove the Mg temperature only aluminum chloride emissions were ob- from molten Al. The use of Cl2(g) flow, electrolysis and served. According to these authors this behavior is explained reactive powders are the main alternatives to decrease Mg because of the highly negative Gibbs free energy for the concentrations in Al baths.2,6,7) Muñoz-Arroyo et al.8) used aluminum chloride and also because of the high amount of rich silica based compounds such as Ca(Si7Al2)18·6H2O, liquid aluminum in direct contact with the chlorine gas. KAlSi3O8 and SiO2. In such cases the goal is to form Magnesium chloride melts at 710°C, hence, below of this MgAl2O4. temperature, this compound forms a solid phase on the The chlorine injection technique is essentially performed bubbles that restrain Cl2(g) reacts with Al(l) so this could introducing the gas inside the molten Al through a graphite or explain the percentage increase of aluminum chloride stainless steel nozzle. Usually a mixture of gases is applied. formation. Under such conditions nitrogen or argon are the bases of the mixture and chlorine is the active gas. Nitrogen typically is 2. Objective employed more commonly because of its lower cost. An improvement to this process is to use a rotor. The main In the present work, the aim was to study the kinetics of advantage of this device is that it produces a dispersion of Mg depletion from Al can scraps using Cl2(g) (gaseous small bubbles (less than 5 mm) inside the bath. chlorine) and Ar(g) (argon). Thus, a kinetic model is proposed Table 1 shows the main reactions occurring in the process to explain the experimental data obtained. of Mg removal by the chlorine introduction. Gibbs free energy equations are also presented. Data for the reactions 3. Methodology involving moisture and oxygen are also exhibited. These compounds may also be found under industrial conditions. All tests were performed by melting 1.75 « 0.2 kg of an Through these calculi, all the reactions investigated are Al1.5 mass% Mg alloy in graphite crucibles. The charged feasible, since at, for example, at 790°C; all Gibbs free graphite crucible was placed inside a hot chamber of a energies are less than zero. Furthermore, it is possible to laboratory electric furnace. The furnace was set at 790°C, conclude that the Mg removal by chlorine is thermodynami- because under this condition the magnesium chloride is a cally feasible. liquid phase. The gas purging was carried out through an 478 E. A. Vieira, J. Roberto de Oliveira, G. F. Alves, D. C. R. Espinosa and J. A. S. Tenório

Table 1 Standard free Gibbs energy as a function of temperature of the mainly reactions that occur in the Mg removal from the aluminum.911)

Reaction # ¦G° (J·mol¹1) «kJ

Al(l) + 3/2Cl2(g) ¼ AlCl3(g) 1 ¹586,872 + 10.45T log T + 29.47T 12.54

Mg(l) ¼ Mg 2 ¹14,538 ¹ 1.254T ®

Mg(l) + Cl2(g) ¼ MgCl2(l) 3 ¹618,013 ¹ 56.76T log T + 304.18T 6.27

Mg + Cl2(g) ¼ MgCl2(l) 4 ¹603,475 ¹ 56.76T log T + 305.43T Min 6.27

2/3AlCl3(g) + Mg ¼ MgCl2(l) + 2/3Al(l) 5 ¹212,227 ¹ 63.75T log T + 285.79T ®

2Al(l) + 3/2O2(g) ¼ Al2O3(s) 6 ¹1,696,077 ¹ 15.68T log T + 385.48T 16.72

Mg + 1/2O2(g) ¼ MgO(s) 7 ¹593,025 ¹ 1.00T log T + 113.49T 6.27

3H2O(g) + 2Al(l) ¼ Al2O3(s) + 6H8¹1,229,940 ¹ 16.64T log T ¹ 186.08T ®

3Mg + Al2O3(s) ¼ 3MgO(s) + 2Al(l) 9 ¹82,998 + 12.67T log T ¹ 45.02T 18.81

Mg(g) ¼ Mg(l) 10 ¹129,455 + 95.05T 1.67

Table 2 Tests performed.

Gas Nozzle Flow (mL·s¹1) Designation ®® ®no purged Ar 1.8* (mm) 1.00 Ar1810 Ar 1.8* (mm) 2.50 Ar1825 Ar 1.8* (mm) 4.17 Ar1842

Cl2 0.4* (mm) 2.50 Cl20425

Cl2 1.8* (mm) 2.50 Cl21825

Cl2 4.7* (mm) 2.50 Cl24725

Cl2 Plug (**80 µm) 1.00 Cl2P10

Cl2 Plug (**80 µm) 2.50 Cl2P25

Cl2 Plug (**80 µm) 4.17 Cl2P42

Cl2 + N2 Plug (**80 µm) 2.50/2.00 Cl2N2P2520 *internal diameter **pore diameter

Fig. 1 Experimental apparatus.

alumina duct, which was immersed 115 mm beneath the 1.54 1.2 molten aluminum surface. The gas flows were measured at 1.53 room temperature but the devices corrected the gas flows to 1.0 the Standard Condition for Temperature and Pressure (STP). 1.52 0.8 Two types of nozzles were applied: (i) a graphite porous 1.51 plug with external diameter of 15 mm and mean pore 1.50 0.6 diameter of 80 µm. (ii) an alumina nozzle with three different 1.49 internal diameters: 4.7, 1.8 and 0.4 mm. The samples from 0.4 molten aluminum were extracted after 0, 1800, 3600, 5400, 1.48 mass of Mg, m/g 0.2 7200, 9000 and 10800 s. Table 2 presents the tests performed 1.47 percentage of Mg, (wt%) and their particular designation. The magnesium concen- 1.46 0.0 0 1 2 3 4 5 6 7 8 9 10 11 trations were measured using atomic absorption (AA) 3 spectrophotometry in accordance with the ASTM standard time x 10 , t /s procedure.13) Fig. 2 Test performed without gas injection. Another four tests were performed. In the first one, no purging gas was used, and the bath was just laid resting inside the hot furnace atmosphere. The others one was of Table 1. In such conditions the oxide layer of aluminum executed through the purging of argon. Figure 1 shows the formed on the surface of the liquid reacts with magnesium. sketch of the laboratory apparatus. It is proposed that the rate of Mg depletion is given by the following equation: 4. Results and Discussion dc ¼kðc c Þn ð1Þ dt e As aforementioned before, the liquid AlMg alloy was kept at 790°C to verify the loss of Mg under such conditions. Where c is Mg concentration in the liquid aluminum and may ¹1 Figure 2 presents the Mg concentration as a function of be expressed in mol·kg , ce is the Mg concentration at elapsed time. In this case Mg removal follows the reaction 9 equilibrium and k is the proportionality constant which is Use of Chlorine to Remove Magnesium from Molten Aluminum 479 dependent on interfacial area. Lastly dc/dt may be expressed 1.7 ¹1 ¹1 % ¹1 in mol·s , kg·s or mass ·s and a choice may be made 1.6 adjusting the corresponding k. According to thermo dynamic 1.5 calculi, ce is very low compared with concentrations obtained in this work. Consequently if n is equal to 1, then one can 1.4 write: 1.3 dc ¼ kc ð2Þ 1.2 no purged dt -1 1.1 1.0 mL.s and 2.5 mL.s-1 1.0 -1 ¼ kt ð Þ percentage of Mg, (wt%) 4.2 mL.s c cie 3 0.9 0 1 2 3 4 5 6 7 8 9 10 11 Using eq. (3) and data from Fig. 2, a R2 equal to 0.987 was -3 determined. Although this model matches very well with time x 10 , t /s experimental results, another model can be proposed. In this Fig. 3 Test performed with injection of argon 1.0, 2.5 and 4.2 mL·s¹1. case, one can consider Higbie’s model14) where the reaction is controlled by the thickness of the boundary layer. d μ during the initial stages of the argon injection. A period of c A melt ¼ hcðc ceÞð4Þ 1800 s was considered to stabilize the bath surface. This time dt M T is necessary to stabilize the opened area on the bath surface 2 Where A is the reaction Area (m ), MT is total mass of alloy in for oxidation to take place. Hence the initial time for the ¹3 (kg), μmelt is density of the melt (kg·m ) and hc is the mass model was 1800 s. transfer coefficient of dissolved Mg (m·s¹1) and may be According to Fig. 4 the removal of magnesium by surface expressed by: oxidation can be described by eq. (6). The same equation D 1=2 also describes the removal of magnesium using inert gas. h ¼ 2 ð5Þ Table 3 presents the values of k and it was assumed n = 1. c ³t As seen in Table 3, eq. (6) matches well with experimental Where D is the magnesium diffusion coefficient in the liquid results. In Fig. 3 one can notice that the removal of Mg was aluminum (m2·s¹1) and t is the time of exposure (s) so: far superior when argon was purged than when no gas was d purged. c ¼ 0 1=2ð ÞðÞ d k t c ce 6 This result shows that the bubbles appearing on the surface Z t Z C t creates a discontinuity in the oxide layer, and consequently dc 0 ¼ k t 1=2 dt ð7Þ the oxidation process occurs in these regions. Additionally, c Ci 0 Fig. 5 shows that k varies as the flow increases; such ð Þ¼ 0 1=2 þ ð ÞðÞ ln c 2k t ln ci 8 behavior can be attributed to the increase on the oxidation Using these equations and the experimental results and interface on the surface of the molten bath as the flow 2 also considering ce ¹ c, the obtained R is equal to 0.996. increases. Consequently, one can state that both models may describe Experiments Cl20425, Cl21825, Cl24725 and Cl2P25 were the magnesium loss under such conditions. However, this carried out to understand the kinetics of Mg removal as a second model matches better with experimental results than function of bubble size and chlorine injection. Figure 6(a) the first one. shows the results obtained for the test Cl20425. These The magnesium concentration remained nearly unchanged experimental data agree with the mass balance, and no during the entire test, so very little magnesium oxidation emissions of Cl2(g) or Al2Cl6(g) are expected. For this test the occurred and also an insignificant amount of magnesium was experimental data are close to the stoichiometric curve lost due to evaporation. Thermodynamic evaluations dem- throughout the entire test. Figure 6(b) exhibits the results onstrated that the amount of Mg vapor that may be present in from the test in which the 1.8 mm nozzle was used, and one the bubbles is in the magnitude of 0.0003% of the total Mg can observe that the experimental data follow the stoichio- spent in the process. According to Tenório & Espinosa15) metric curve until the Mg concentration is around under these conditions MgO·Al2O3 is the oxide appearing 0.12 mass%. on the molten surface. Magnesium removal throughout the At this point, the kinetics of Mg removal changes, and this oxidation process was very slow; consequently, the oxide point is identified as the critical Mg concentration and is also layer was able to protect the molten aluminum from further characterized by the beginning of the Al chloride Al2Cl6(g) magnesium oxidation. Only 5.2% of the initial magnesium emissions. For the 4.7 mm nozzle, Fig. 6(c) Al2Cl6(g) was lost after 3 h. emissions appeared at Mg concentration around 0.80 mass%. Figure 3 presents the results obtained when argon was This fact indicates that the bubbles are greater than previous purged in the aluminum bath. The use of argon increased the ones. rate of magnesium removal compared to the previous test. On the other hand, according to Fig. 6(d) the critical Mg However, eq. (3) did not match these experimental results, so concentration is around 1.0 mass% when the porous plug was eq. (4) was tried. Therefore, the opening of the aluminum used. Consequently, the greatest part of the Al chloride did oxide on the surface caused by bubbles was considered not react. 480 E. A. Vieira, J. Roberto de Oliveira, G. F. Alves, D. C. R. Espinosa and J. A. S. Tenório

(a) (b) 0.6 0.5 no purged 0.5 0.4 -1 1.0 mL.s 0.4 0.3 2.5 mL.s-1 4.2 mL.s-1 0.3 0.2 0.2 0.1

ln (c) 2 0.1 no purged ln (c) R =0.9932 -1 0.0 0.0 1.0 mL.s -1 2.5 mL.s -0.1 -0.1 -1 4.2 mL.s -0.2 -0.2 20 40 60 80 100 120 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 1/2 2k + ln(c ) time, t /s t i

Fig. 4 Experimental curves of magnesium concentration as a function of time (a) relationship between experimental data and (b) adjustment using Higbie’s Model.

-1/2 4 Table 3 Kinetic parameter obtained using the diffusion model. , -k/s ¹2kB © 104 ¹2kB © 104 3 3 Test R2 Test R2 (s¹1/2) (s¹1/2)

No gas 6.34 1.00 Cl2P25 440.4 0.97 2

Ar1810 53.05 0.98 Cl2P42 24.7 1.00

Ar1825 63.94 0.99 Cl24725 180.6 1.00 1

Ar1842 87.10 0.98 ®®® constant x 10 0 0 1 2 3 4 5 flow, Q/mL.s-1

Fig. 5 k as a function of argon ﬂow using a nozzle with 1.8 mm internal diameter.

(a) (b) 1.6 1.6 o Cl 0425 o Cl 1825 1.4 2 1.4 2 stoichiometric stoichiometric 1.2 1.2 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 percentage of Mg, (wt%) percentage of Mg, (wt%) 0.0 0.0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 time x 10-3, t /s time x 10-3, t /s (c) (d) 1.6 1.6 o Cl P25 o Cl 4725 2 1.4 2 1.4 stoichiometric stoichiometric 1.2 1.2 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 percentage of Mg, (wt%) 0.0 percentage of Mg, (wt%) 0.0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 time x 10-3, t /s time x 10-3, t /s

¹1 Fig. 6 Influence of orifice diameter on the kinetics of Mg concentration as a function of time using flow of 2.5 mL·s of Cl2(g). Use of Chlorine to Remove Magnesium from Molten Aluminum 481

The rate of Mg removal from Al is directly related to the nozzle inside the liquid, fluid dynamic characteristics of the bubble size. The prediction of the size of bubbles formed liquid and gas. Koide et al.17) proposed the expression (11) when a gas is injected inside a liquid was made by the to predict the mean bubble diameter when a porous plug is Reynolds (R ), Weber (W ) and Froude (F ) orifice used: eo eo ro 2 3 numbers.16) Table 4 presents the calculi of bubbles diameter 0:16 V 2 for the tests performed. An expression to define the bubble 6 7 6 ¾2gd 7 size can be derived as a function of the Reo,Weo and Fro ¼ ð · μ Þ1=36sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffip 7 ð Þ db 1:65 dp =g ‘ 6 7 11 numbers. 4 2μ 5 dpV ‘ The diameters of the bubbles were estimated through the ¾2· expression:16) Where ¾ is the void fraction in the porous plug, d is the d ¼ 3:8ð·d =μ gÞ1=3 ð9Þ p b 0 ‘ pore diameter and V is the velocity of the gas. The calculi for In this case db is the bubble diameter, do is the orifice the test Cl2P25 are present in Table 4. The data showed that diameter, · is the interfacial tension, do is the nozzle external in this test, the velocity imposed on the system was superior diameter, μl is the Al density, and g is the gravity to the critical velocity. Thus, the bubbles coalesced, which acceleration. The calculi for tests Cl20425, Cl21825 and explains the behavior observed in Fig. 6(d). Cl24725 showed that the calculated bubble diameter increases Figure 7 presents the results of the second set of trials as much as the nozzle internal diameter increases. carried out using the porous plug. The Cl2 fluxes employed ¹1 ¹1 The increase in bubble size explains the behaviors were 1.0, 2.5 and 4.2 mL·s of Cl2. When 1.0 mL·s of Cl2 observed in Fig. 6 in which the critical Mg concentration was purged, the results were better than at 4.2 mL·s¹1.As increases following the higher diameter of the bubbles. In this shown in Table 4, in the first case the critical velocity was not respect, the increase of bubble size causes an increase in the reached, so a large amount of small bubbles was formed. As a Al chloride emission or a premature deviation from the result the stoichiometric behavior was reached, and therefore stoichiometric behavior. In the case of the porous plug, there no Al chloride was emitted. ¹1 is a critical injection velocity Vc where the bubbles start to On the other hand, when 4.2 mL·s of Cl2 was injected, coalesce. The critical velocity can be calculated through the the emissions started from the beginning of the experiment. expression (10).17) Each time the critical Mg is reached, Mg removal may occur by both the reaction with the Al chloride and by the oxidation V ¼½ð2³·Þ=ð0:6d μ Þ1=2 ð10Þ c b ‘ on the surface of liquid Al. Since the bubbles that rise to the The critical flux or velocity is a function of the pore liquid surface generate an opening in the protective oxide diameter, shape and size of the porous plug, depth of the layer. In the test Cl2N2P2520, a mixture of Cl2 and nitrogen was used. The results of this experiment are plotted in Fig. 7. Table 4 Experimental and calculated parameters. The Mg was removed and low Al chloride emissions fi ¹1 ¹1 occurred. However, in the test Cl2P25, the same ef ciency Test V (m·s ) db (mm) Vc (m·s ) was not observed. The decrease in the Cl2 concentration in ® Cl20425 27.95 8.0 the purged gas resulted in the decrease in the critical Mg ® Cl21825 2.30 13.0 concentration. ® Cl24725 0.46 20.0 This effect is in close agreement with the results of Cl2P10 0.25 10.5 0.49 Fu et al.;1) although in their experiments the initial Mg Cl2P25 0.62 12.0 0.46 concentration was around 0.1 mass% and flow constant. Cl2P42 1.07 13.2 0.44 Thus, in the test Cl2N2P2520 low Al chloride emissions Cl2N2P2520 1.10 14.6 0.42 happened even though the flux was above the critical

(a) (b) 1.8 -3 -1 stoichiometric 2.5 x 10 mL.s 2.0 -3 -1 stoichiometric 1.0 x 10-3 mL.s-1 1.6 stoichiometric 4.2 x 10 mL.s 1.8 stoichiometric mix(2.5Cl + 2.0Ar) x 10-3 mL.s-1 1.4 1.6 2 1.2 1.4 1.0 1.2 Cl P42 2 1.0 Cl P10 0.8 Cl P25 2 2 0.8 Cl N P2520 0.6 2 2 0.6 0.4 0.4 percentage of Mg, (wt%) 0.2 percentage of Mg, (wt%) 0.2 0.0 0.0 0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 10 11 time x 10-3, t /s time x 10-3, t /s

Fig. 7 Concentration of Mg as a function of time using a pore plug. Mixture with argon and low rate of injection improves the process performance. 482 E. A. Vieira, J. Roberto de Oliveira, G. F. Alves, D. C. R. Espinosa and J. A. S. Tenório

velocity. Table 4 shows the calculated data for this test. One 0.6 Cl P42 - 4.2 x 10-3 mL.s-1 2 possible reason for this higher Mg removal is because of the -3 -1 0.3 Cl 4725 - 2.5 x 10 mL.s 2 stirring increase caused by the mixture of Cl2(g) and N2(g) Cl P25 - 2.5 x 10-3 mL.s-1 decreasing the diffusion path for Al2Cl6(g) in the boundary 0.0 2 layer. -0.3 Finally, the experiments show that the Mg removal rate can follow two distinct kinds of behaviors, the ﬁrst being -0.6 stoichiometric. In this case, the Al chloride has enough time ln (c) -0.9 2 to react before reaching the surface, and the total Cl ﬂow R =0.9937 2 -1.2 determine the velocity of Mg removal. In this case, the simple expression (12) describes this behavior: -1.5 Q -1.5 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 c ¼ c PM 100t ð12Þ i Mg 2k + ln(c ) 22:4MT t i

Where, ci is the initial Mg content in mass%, Q is the total Fig. 8 Relationship between experimental results for chlorine injection and ’ Cl2 flux in L, MT is the Al mass, t is the time, and PMMg is the adjustment curve of Higbie s Model. Mg molecular weight. In the second case, the bubbles reach the surface before all reactions have finished. The bubble holding time in the liquid is not enough to 5. Conclusions ensure the reaction between the Al chloride and the Mg. Under such conditions, the removal of Mg is caused by two (1) The mg removal from liquid aluminum on the tests separate phenomena. The first one refers to the Mg removed when no gas is purged is controlled by diffusion of Mg in the by the reaction with the Al chloride (Table 1 and reaction 5), bath. and the second refers to the oxidation on the surface (2) Argon purging increases the Mg removal because the (Table 1 and reaction 9). In both cases, the controlling step bubbles break the protective layer. of the process is the Mg diffusion in the boundary layer. (3) Chlorine injection increases the removal of Mg. The Furthermore, the overall rate of Mg removal is given by the controlling step is the diffusion on the boundary layer to form eq. (13): Al2Cl6(g). When the bubbles reach the surface, the oxidation d d of Mg by the atmosphere also occurs. dc cðAl Cl ð ÞÞ cðO ð ÞÞ ¼ 2 6 g þ 2 g ð13Þ (4) The use of a mixture of inert gas and chlorine enhances dt dt dt Mg removal. This effect may be explained by the decrease in The Mg removal is established by the sum of two states. the boundary layer caused by the bath stirring.

The transitory state is expressed by dcðAl2Cl6ðgÞÞ refers to the amount of Mg reacted with Al chloride. The permanent state REFERENCES dcðO2ðgÞÞ is the parcel of Mg that reacts with oxygen on the surface. Consequently, the eq. (14) describes the Mg removal 1) Q. Fu, D. Xu and J. W. Evans: Met. Mat. Trans. 29B (1998) 971978. kinetics: 2) Q. Fu and J. W. Evans: Met. Mat. Trans. 29B (1998) 979986. dc 3) D. V. Neff and B. P. Cochran: Light Metals, Denver: TMS. (1993) ¼þk t 1=2c þ k t 1=2c ð14Þ dt 1 2 pp. 1053 1060. d 4) T. A. Utigard, K. Friesen, J. L. Roy, A. Silny and C. Dupuis: JOM 50 c 1=2 (1998) 3843. ¼þðk1 þ k2Þt c ð15Þ dt 5) J. A. S. Tenório, M. C. Carboni and D. C. R. Espinosa: J. Light Met. 1 (2001) 195198. Where k1 and k2 are constants. 6) B. L. Tiwari, B. J. Howie and R. M. Johnson: Trans. AFS 94 (1986) Thus, eq. (15) describes the Mg removal kinetics. The 385390. calculi to obtain the reaction rate constants (k1 + k2) were 7) G. M. Crepeau, M. L. Fenyes and J. L. Jeanneret: Modern Cast. 82 (1991) 2830. performed for tests Cl24725, CL2P25 and CL2P42 and results 8) R. Muñoz-Arroyo, J. C. Escobedo-Bocardo, H. M. Hernández-García, are shown in Table 4 and Fig. 8. D. A. Cortés-Hernández, M. Terrones-Maldonado, A. Rodríguez- This agreement indicates that the diffusion of Mg in the Pulido and J. L. Hernández-Piñero: Rev. Met. 46 (2010) 351359. boundary layer is the rate controlling step. Under industrial 9) G. K. Sigworth and T. A. Engh: Scand. J. Met. 11 (1982) 143149. conditions, the critical Mg concentration is a very important 10) A. T. Engh: Principles of Metal Refining, (Science Publications, New parameter to be determined. The time to reach the critical Tork, 1992) pp. 171 218. 11) Z. Moser: J. Phase Equilib. 19 (1998) 3847. concentration is close to the time to start the chloride 12) B. Lagowski: Trans. AFS 77 (1968) 205207. 1) emission. The results showed that the critical Mg concen- 13) American Society for Testing and Materials: Annual Book of ASTM tration and the time to reach this concentration are influenced Standard, (2000). by the system adopted to inject Cl2(g), by the gas flux, and 14) H. Higbie: Trans. AIChE 31 (1935) 365 388. 15) J. A. S. Tenório and D. C. R. Espinosa: Oxid. Met. 53 (2000) 361373. also by the Cl2(g) concentration in the gas. The developed 16) J. Szekely: Fluid Flow Phenomena in Metals Processing, (Academic equations showed that the Mg removal kinetic is controlled Press, London, 1979) pp. 305350. by the Mg transport in the boundary layer either for the 17) M. Iguchi, M. Kaji and Z. Morita: Met. Mat. Trans. 29B (1998) 1209 oxidation parcel or by the reaction with Al chloride. 1218.