THE UNIVERSITY OF MICHIGAN INDUSTRY PROGRAM OF THE COLLEGE OF ENGINEERING MASS TRANSFER OF HYDROGEN BETWEEN LIQUID ALUMINUM AND BUBBLES OF ARGON GAS Rob'ert Do Pehlke Arden L. Bement., Jro October, 1961 IP-538

ACKNOWLEDGEMENT The authors gratefully acknowledge the encouragement of the Research Division of Federal Mogal Corporation who supplied some of the materials used in this investigation. ii

TABLE OF CONTENTS Page ACKNOWLEDGEMENT......................o eoo...O O O ii LIST OF FIGURES..................................................00000 iv INTRODUCTION.o o o.................................................... 1 DERIVATION OF THE MODEL.........................................o 1 Mass Transport Control..........ooooooooooooooo 3 Thermal Expansion of the Bubble o........................... 4 Expansion Caused by Reduced Pressure Head................. 5 APPARATUS AND EXPERIMENTAL PROCEDURES............................ 6 EXPERIMENTAL RESULTS AND DISCUSSIONo o.. o o o o o o o o o o 6 Equilibrium Degassing...................................... 9 Loss of Hydrogen from the Melt Surface 00.........000000000. 9 Influence of Bubble Size.......... o........................o o o o o o o o o o o 9 Influence of Flow Rate..................................... 9 SUMMARY.........................0o.oooooooooooo0oooooooooooo.oooo 14 BIBLIOGRAPHY....... ooooooooooooooooooooooooooooooooooooooooooooooo 15 iii

LIST OF FIGURES Figure Page 1 Experimental System,....oo...o... o* OOOoo...,,oo0.o., 7 2 Comparison of Equilibrium Degassing With Experimental Result..oo.o..o.<.eo o... s.......................o..oooo 10 3 Loss of Hydrogen From Melt in Absence of Argon Flushing.. o o *............ oo............ e o o,, o o... o 11 4 Influence of Bubble Size on Removal of Hydrogen by Argon Flushingu.................................. 12 5 Influence of Flushing Rate on Hydrogen Removal for Bubble Radius of 0.5 cmO................. 13 iv

INTRODUCTION The presence of hydrogen in liquid metals in amounts which exceed the solid solubility at one atmosphere hydrogen pressure is highly undesirable, causing unsound ingots or castings, or resulting in difficulties during further fabrication, The removal of hydrogen from liquid metals can be accomplished be several methods, the principal two being vacuum treating and inert flush degassing. The design of a process involving either of these techniques is based on the mass transfer coefficient for hydrogen between the liquid metal and a hydrogen-dilute gas phase. In an effort to extend our knowledge in this area, an investigation of the rate of removal of hydrogen from liquid metals by inert gas flushing was undertaken. In view of the difficulties encountered with gas porosity in the casting of aluminum and aluminum alloys, and because of certain experimental advantages, liquid aluminum was selected for the studyo The removal of gases from liquid metals has received considerable attention in recent years. Sims(l) has reported on the removal of hydrogen from liquid steel by flushing with argon, and Vallet(2) has considered the degassing of an open-hearth bath by carbon monoxide bubbles during the boilo Pehlke and Elliott(3) have studied the kinetics of nitrogen liquid-iron reactions. Machlin(4) has presented an analytical treatment of vacuum induction refining. Considerable work has also been reported on aqueous and organic liquid-gas systems, and an excellent summary prepared by Hovis. 5) The treatment of liquid aluminum by a flushing gas has been an established practice for several years.(6) A number of gases have been used for this purpose, including nitrogen, argon, and chlorine, or mixtures of these gases. Tikkanen and Erko(7) have studied the relative density changes of aluminum castings poured from melts degassed with chlorine, nitrogen, and mixtures of the two gases. Their results show that in the case of chlorinechemical reactions play an important and sometimes detrimental part in the degassing operation. Unfortunately, much of the data reported for the degassing of aluminum melts are not in sufficient detail or in proper form for quantitatively evaluating the mechanism of hydrogen removal. DERIVATION OF THE MODEL Equilibrium Degassing - In view of the high diffusion coefficients for hydrogen in liquid metals, some consideration was given to the possibility that the equilibrium pressure of hydrogen in the inert -1

-2flush gas bubble might be reached during the rise of the bubble in the liquid metal. In the case where equilibrium is assumed, the following mass balance may be written expressing the hydrogen removed: dNg = (dNg + dNf) Pg/(Pg + Pf) (1) where Ng and Nf are the liters of diatomic hydrogen and flushing gas, respectively, and Pg and Pf are their partial pressures. If it is assumed that the purging gas leaves the melt at one atmosphere pressure and is in equilibrium with the residual hydrogen content of the liquid metal, and noting that the inert flush gas enters at the same rate that it leaves, then: dNf = -dNg((l + Pg)/Pg) (2) Converting to concentrations: dCg = -99.0 M dNf (3) ((1 +Pg)/Pg)W where M is the molecular weight of the diatomic gas (2 in the case to be considered), W is the weight of the metal bath in pounds, and Cg is the concentration of dissolved gas in parts per million by weighto From Sieverts' Law: Cg = Kg P (4) where Kg is the equilibrium constant for the solution of hydrogen in the liquid metal under consideration. Combining Equations (3) and (4) and integrating: 0 dC dC V C! 99,0 M f dNf 5) C C2 K2 K2 W O Co g g g and V lo.0 W (Co - C)( + C C)/C Co (6) M where Vg is the volume of flushing gas in liters required to reduce the dissolved hydrogen content from Co to Co

-3Mass Transport Control - Assuming that mass transfer is limited by the transport of hydrogen through the liquid metal phase, the removal of hydrogen by a single bubble may be described by the relation: dng/dt = - kL ab(CB - CM) (7) where ng is the moles of hydrogen being removed, kL is the mass transfer coefficient in cm/sec, ab is the area of the bubble-metal interface in cm2, and CB and CM are the concentrations of hydrogen in the metal at the bubble interface and in the bulk metal, respectively, in moles/cm3. For a system consisting of a crucible of liquid metal exposed to a hydrogen-free atmosphere in which an inert flush gas bubble generator is immersed, mass transfer across the free surface of the melt must also be considered. This portion of the mass transfer can be described by Fick's first law: dng/dt = DA, d (8) dx where D is the diffusion coefficient of hydrogen dissolved in the liquid phase and As is the surface area of the melto Assuming that the hydrogen content of the atmosphere is negligibly small, the concentration gradient can be replaced by CM/5 where 5 is the boundary layer thickness~ In Equation (7) the term CB is approximately zero if the rate of transfer to the bubble is small, since it represents the concentration in equilibrium with the partial pressure of hydrogen in the bubble0 Noting also that ng can be converted to concentration units by dividing by the volure of the melt, Equations (7) and (8) can be combined to give: g = - 1 [D As + kLabn]dt (9) Cg VM 5 where n is the number of bubbles in the melt at any instant, and is given by the relationship: n = FTr/vb (10) where F is the flow rate of the flush gas, Tr is the time required for a bubble to rise through the melt, and vb is the volume of a single bubble, The total instantaneous bubble area, n ab, is given by the relation: Ab = F (11) rb

-4where fb is the bubble radius. Combining Equations (9) and (11) and integrating: C 1 t D 3k FT d. 1 A + A kLF r] dt (12) C Cg Vm o 5 rb or: C = Co exp[- DAs kLF Tr dt] (13) 5Vm Vm o rb Equation (13) now expresses the instantaneous concentration of the melt in terms of the original concentration, several geometry factors, the diffusion coefficient and boundary layer thickness at the exposed melt surface, and the mass transfer coefficient for the rising bubbleso Further simplification of Equation (13) requires a closer examination of the nature of bubble generation and rise in the melto After a bubble of flush gas is released into the liquid metal environment, it can change dimensions due to the following effects: (1) Thermal expansion, if the gas is not at the melt temperature upon release. (2) Expansion caused by reduced pressure head as the bubble riseso (5) Expansion due to hydrogen diffusion into the bubbleo (4) Contraction due to'argon diffusion out of the bubble. Statements (5) and (4) are neglected, assuming that the volume of hydrogen removed is extremely small compared to the volume of flush gas, and assuming that the flush gas is completely insoluble in the liquid metalo Statements (1) and (2), however, require further considerationo Thermal Expansion of the Bubble - Before the bubble is released into the liquid metal, it may be preheated depending upon whether or not the delivery mechanism is immersed in the melt. Under the experimental conditions of this investigation, the gas was preheated while flowing down a vycor tube with an immersed length of approximately 25 cmo Hence, for the case under consideration, heat transfer can be based on forced convection in a heated tube. The Reynolds number for the flow rates used in this investigation, assuming a mean gas temperature of 350~C in the tube, indicate that the gas stream in the delivery tubes was in the region of streamline flowo Assuming a parabolic velocity distribution and a uniform

-5wall temperature, the exit gas temperature may be shown to be essentially equal to the wall temperature of the tube (8) Examination of the thermal resistances involved in heat transfer from the melt to the gas flowing in the delivery tube revealed that transfer of heat to the gas is rate limiting compared to conduction through the tube wallo Hence, it was concluded that the gas reached the melt temperature prior to release from the delivery tubeo Expansion Caused by Reduced Pressure Head - When the diameter of the bubble becomes much larger than the outside diameter of the delivery tube, it becomes unstable and breaks off (9) Hence, in the present study, the initial bubble diameter will be assumed to be equal to twice the diameter of the tube opening. Calculations of the change in bubble diameter during rise to the surface through 20 cm of liquid aluminum showed that the small changes in bubble size caused by a reduced pressure head can be neglected, and that the rise velocity of the bubble can be considered approximately constanto The rise time for the bubble can be calculated from the terminal rise velocity, Vt, which is given by the expression~ Vt = [4(p - pb)gdb] (14) where p, is the liquid density, Pb is the gas density in the bubble, db is the bubble diameter, and f is a friction factor which is defined in terms of the Reynolds numbero(l0) The rise time of the bubble through the melt is then given by the relationship: r - h/Vt (15) where h is the height of the melt above the exit of the delivery mechanism. This expression neglects the time of formation of the bubble and its initial acceleration to the terminal velocity. The time of formation would be extremely short under the flow conditions used in the experiments to be considered below, The period of initial acceleration can also be shown to be very short relative to the rise time. In view of the considerations presented above, Equation (13) reduces to' C [DAs kLFr (6) ^ = ---- + -- -- ]t (16) Co 5Vm rbTVm

-6APPARATUS AND EXPERIMENTAL PROCEDURES The experimental system employed in this investigation is shown in Figure lo The melt consisted of 18 kg of commercially pure aluminum held in a 23 cm diameter, clay-graphite refractory crucible which was heated inductively. Moist wooden sticks were placed in the melt at 800~C to introduce hydrogen. The temperature of the metal, measured with an immersion thermocouple. was lowered to approximately 700~C, and maintained at that level throughout the degassing experiment, Argon gas. which was dried by passing it through an anhydrone tower, was delivered at a constant flow rate into the liquid metal through a straight section of vycor tubingo In an effort to demonstrate that traces of oxygen in the argon flush gas had no influence on the results, separate experiments were run in which the argon was passed over heated copper gauze and heated titanium chips in addition to the drying compound. Samples of the liquid metal were taken periodically for hydrogen analysis by drawing molten aluminum up against copper gauze in a smalldiameter vycor tube and immediately quenching the tube in liquid nitrogen. The sample was then trimmed, wrapped in aluminum foil3 and stored in liquid nitrogen The hydrogen content of the aluminum samples was determined by hot-vacuum extraction at 500C using a standard NRC Vacuum Fusion apparatuso A period of approximately 40 minutes was found sufficient to collect the evolved gases which were found to be 95 percent or more hydrogen. A tenminute circulation period, during which the hydrogen was frozen out of the sample as reacted water vapor, was employed. EXPERIMENTAL RESULTS AND DISCUSSION The experimental results of this investigation are presented in Table Io Vycor delivery tubes with inside diameters of 5, 355 and 2 millimeters were used to evaluate the effects of bubble sizeO The depth of immersion was 20 cm for all of the experiments. The initial hydrogen content varies considerably for the 8 experiments due to the uncontrolled technique of charging hydrogen with wet stickso The data of Ransley and Neufeld(ll) indicate that the equilibrium solubility of hydrogen in aluminum at 700~C and one atmosphere hydrogen pressure is 0o8 parts per million by weight, and at 800~C and one atmosphere hydrogen pressure is lo5 ppmo These data indicate that the charged hydrogen contents in contents in the present experiments approached or exceeded these saturation valueso

-.70 Cd 40) rz4 0.0 ~ ~ ~ L CC L- ~ ~ 0

-8TABLE I HYDROGEN CONCENTRATION OF LIQUID ALUMINUM DURING ARGON DEGASSING Run Tube Opening Flow Rate Degassing Time Concentration (mm) (liters/min) (minutes) Hydrogen (ppm) 1 none 0 0 1.55 8 1.45 2 none 0 0 1.605 9 1.50 35 0.8 0 0.775 5 0.72 6 0.755 12 0.645 4 5 1.6 0 1.05 3 0.855 6 0.83 9 0.695 16 0.555 5 5.5 0.8 0 1.225 6 1.045 *6 5.5 0,8 0 0.965 6 0.825 *7 5.5 0.8 0 1.555 24 0.82 8 2 0.8 0 0.975 6 o.86 12 o.6o * These degassing runs were carried out using argon which had been deoxidized by passing over heated copper gauze and heated titanium chips.

-9Equilibrium Degassing - Equation (6) was derived for the case in which equilibrium is established between the hydrogen dissolved in the melt and the rising gas bubbles before the bubbles leave the le leaemelt. A calculation of the fraction of hydrogen remaining in the melt as a function of flushing gas volume under equilibrium conditions is plotted in Figure 2, assuming a melt saturated with hydrogen at 7000C and neglecting loss of hydrogen from the melt surface, The experimental data for a flow rate of 0,8 liters per minute and an intermediate bubble size are also shown in Figure 2 and reveal that much less hydrogen is removed per unit volume of flushing gas than predicted for equilibrium conditionso Hence, the equilibrium approach defined by Equation (6) does not describe the experimental data. Mass transport control, as defined by Equation (16), can be considered to better represent the removal of hydrogen from the liquid metal phase, as shown below. Loss of Hydrogen From the Melt Surface - Hydrogen is removed from the melt during the experiment by direct evolution into the atmosphere at the exposed melt surface as well as by mass transfer to the argon bubble stream, In order to separate the two paths for hydrogen evolution, Runs 1 and 2 were carried out in a standard manner except that no argon was bubbled through the melto The results of these two experiments are shown graphically in Figure 3, Applying Equation (16), where F (the flow rate of argon) is zero, the slope of a plot of in C/Co versus time will be - DAs/5VMo The mass transfer coefficient at the melt surface, D/5, can be calculated from the melt geometry and the mean slope of Figure 3, and is found to be lo6 x 10-3 cm/sec. This value appears to be low, but may be accounted for by the presence of a thin oxide crust which appeared on the melt surface during the experiments. Influence of Bubble Size - The results of Runs 3, 5, 6, 7, and 8 are presented on a semi-log plot in Figure 4. The slopes of the straight lines drawn through these data points are given in Table II, which also lists the calculated mass transfer coefficients between the melt and the argon bubbleso The mass transfer coefficient is independent of bubble size, at least within the experimental accuracy which is estimated to be 25 percent. An average value for the mass transfer coefficient, kL, at 700~C at a flow rate of 13o3 cm3/sec is 3.9 x 10-2 cm/sec. This result is in agreement with the correlation of Sherwood and Pigford(l2) which predicts a transfer coefficient of about 4 x 10-2 cm/sec for mass transfer in flow past single spheres. Influence of Flow Rate - Flow rates of 0o8 and 1,6 liters per minute were investigated using a 5-mm ID delivery tube, The results of these experiments, Runs 3 and 4, are shown in Figure 5, and the mass transfer coefficients are given in Table II, The mass transfer coefficient was

-101.0 0.8 0.6 C Co 0.4.. EQUILIBRIUM 0 DEGASSING 0 5 10 15 20 25 LITERS OF ARGON Figure 2. Comparison of Equilibrium Degassing With Experimental Result. (Surface losses neglected in calculating equilibrium degassing line.)

-11\ RUN t I A RUN # 2 Ln[ Co; — -0.1-0.2 0 5 10 15 20 25 DEGASSING TIME, MIN. Figure 3. Loss of Hydrogen From Melt in Absence of Argon Flushing.

0 3 FLUSHING RATE =13.3 Cm3/Sec 0 0 RUN # 3 E3 RUN# 5,6,7 -0.1 \A \ RUN 8 - A -0.2 \rb 0.5 Cm rb: 0.35 Cm I I -Q6 \rb O.2Cm -07 0 5 10 15 20 25 DEGASSING TIME, MIN. Figure 4. Influence of Bubble Size on Removal of Hydrogen by Argon Flushing.

-13~0 0 RUN #3 0\ ~A RUN #4 -0. -Q0. Co \ F 26.7 Cm3/Sec -06 -0.7'0 5 10 15 20 DEGASSING TIME, MIN. Figure 5. Influence of Flushing Rate on Hydrogen Removal for Bubble Radius of 0.5 cm.

-14larger for the higher flow rate, which may be due to turbulence in the melto Such turbulence, which becomes noticeable at flow rates above 1 liter/min, undoubtedly increases the loss of hydrogen at the melt surface. Also. increasing the flow rate may result in physical conditions which are not accurately described by the individual-sphere modelo TABLE II CALCULATION OF MASS TRANSFER COEFFICIENTS Bubble Bubble Radius Residence Slope kL D/5 cm cm3/sec time sec sec-1 cm/sec cm/sec - None - -1o03 x 10-4 1X64 x 10-3 05 13o3 0.356 -2~56 x 10-4 357 x 10-2 0o35 13O3 0o425 -4 35 x 10-4 4o5 x 10-2 Oo2 13.3 0.563 -6~79 x 10-4 3.5 x 10-2 0.5 26o7 0.356 -6o72 x 10-4 608 x 10-2 SMMAR'Y 1o The removal of hydrogen from liquid aluminum by inert flush degassing has been shown to be a non-equilibrium process, and to be described in terms of mass transport control in the liquid phaseo 2, The mass transfer coefficient for the removal of hydrogen from liquid aluminum at 700~C is 3.9 x 10~2 cm/sec, as determined for bubbles from 0o4 to lo0 cm in diameter at a flow rate of 1353 cm3/sec (o08 liters/min). 35 The rate of removal of hydrogen is increased by decreasing bubble sizes and by increasing the flow rate~

BIBLIOGRAPHY 1. Sims, C. E, "Flushing Molten Steel with Neutral Gases. " Electric Furnace Steel Proceedings of A.I.M.E., 7, (1949) 302. 2. Vallet, P. "Theory of Degassing of Open-Hearth Bath During Carbon Boil." Physical Chemistry of Steelmaking Proceedings, Dedham, Mass,, (1956) 109. John Wiley and Sons, Inc. New York, 1958. 53 Pehlke, R. D. and Elliott, J. F, "Solubility of Nitrogen in Liquid Iron Alloys, II.o Kinetics, " To be published. 4. Machlin, E. S. "Kinetics of Vacuum Induction Refining-Theory. " A.I.M.E. Transactions, 218, (1960) 514. 5. Hovis, L. S. Mass Transfer in Pure Liquid-Gas Bubble Systems. PhD. Thesis, The University of Tennessee, 1955. 6. Hanson, D. and Slater, I. G. "Unsoundness in Aluminum Sand Castings. Part III - Solidification in Sand Molds Under Pressure," Journ, Inst. Metals, 56, (1955) 105. 7. Tikkanen, M. and Erkko, E. "Avgasning av aluminiumsmaltor medelst gasinblasning." IVA, 27, (1956) 96. 8. McAdams, W. H. Heat Transmission. McGraw-Hill Publishing Co., InCo., New York, 2nd Ed., (1954) 250, 9. Bikerman, J. J. Surface Chemistry Academy Press, New York, (1948) 40. 10, McCabe, W. L. and Smith, J. C. Unit Operations of Chemical Engineering. McGraw-Hill Publishing Co., Inc., New York, (1956) 558. 11. Ransley, C. E. and Neufeld, Hb "The Solubility of Hydrogen in Liquid and Solid Aluminum." Journ. Inst. Metals, 74, (1948) 599. 12, Sherwood, T. K. and Pigford, R. L. Absorption and Extraction, McGrawHill Publishing Co., Inc., New York, 2nd Ed., (1952) 74. -15