G as-liquid bubble columns have a wide range of applications (absorption, catalytic slurry reactions, bioreactions, coal liquefaction, waste water treatment, etc.) because of their simplicity of operation, low operating costs, ease of maintenance and sufficiently high gas-liquid mass transfer rates. The satisfactory design of bubble columns invariably requires a knowledge of the volumetric liquid-phase mass transfer coefficient k L a which characterizes the rate of interfacial mass transfer and determines the amount of gas transferred from bubbles into the liquid. Two approaches have been used for k L a prediction: (1) separate correlations are obtained for the liquid-phase mass transfer coefficient, k L , and the interfacial area per unit volume, a, and (2) an empirical correlation is obtained for the overall k L a.In our previous papers (Nedeltchev et al., 1999(Nedeltchev et al., , 2000 we used the first approach and estimated k L by making use of both the classical penetration theory put forward by Higbie (1935) and the subsequent correction introduced by Miller (1974) for ellipsoidal bubbles. It will be shown, however, that the experimental local k L a values are different from the theoretical k L a value calculated by the penetration model. Only the arithmetic mean k L a value for the overall bubble bed (BB) is equal to the theoretical one.The present paper basically deals with the physical absorption of carbon dioxide (CO 2 ) in deionized water where the overall mass transfer coefficient is essentially equal to the liquid-phase mass transfer coefficient (Houghton et al., 1957;Hughmark, 1967;Hatton and Lightfoot, 1982;Lee and Tsui, 1999). Houghton et al. (1957) underlined that as much as 90 % of the entering CO 2 may be absorbed in a deep BB and hence the gas composition may change from 85 % down to 30 % CO 2 . Thus, the superficial gas velocity u g through the deep BB may vary considerably as absorption proceeds. The optimal design of bubble columns should take into account the variable u g since it influences almost all fluid dynamic properties (Deckwer, 1976 1978, 1980 Buchholz et al., 1980;Herbrechtsmeier et al., 1981). Alvarez-Cuenca et al. (1980) observed two distinct absorption zones (distributor and bulk zones) during mass transfer of oxygen (O 2 ) between air and water. High k L a values were observed in the distributor zone which extended up to a maximum height of about 5-15 % of the BB volume (Deckwer et al., 1974(Deckwer et al., , 1978. The bulk zone extended from the end of the distributor zone to the top of the BB. The axial liquid concentration C L gradients were much