Exploratory Investigation of the Risk of Desorption from Activated

Jan 10, 2007 - 7.5 L/min, which corresponds to a superficial gas velocity of. 0.064 m/s. .... a function of distance on the moment the flow is interru...
1 downloads 0 Views 75KB Size
Download PDF
4034

Ind. Eng. Chem. Res. 2007, 46, 4034-4039

Exploratory Investigation of the Risk of Desorption from Activated Carbon Filters in Respiratory Protective Devices Marco J. G. Linders,* Piet J. Baak, and Jacques J. G. M. van Bokhoven TNO Defence, Security and Safety, Lange Kleiweg 137, 2288 GJ Rijswijk, The Netherlands

There is a tendency to equip filtering respiratory protective devices with a blower system to lower the breathing resistance. Because there is a continuous flow of air through the filter, the possibility that adsorbed contaminants are released from the filter is enlarged, which evidently is undesirable. The purpose of this work was to explore whether there is a significant risk that physisorbed or chemisorbed contaminants desorb from the filter. Measurements were performed under various conditions with cyclohexane, a physisorbed vapor, and ammonia, a chemisorbed gas. During a certain time, an activated carbon filter bed was exposed to either cyclohexane or ammonia, followed by a period of clean air flow. The experiments showed that respiratory protective devices with a continuous airflow do have a risk to release previously adsorbed contaminants at too high concentrations. Under humid conditions, the release of physisorbed contaminants occurs even more rapidly than under dry conditions. Introduction Individual respiratory protection against gaseous contaminants is often obtained from masks equipped with activated carbon canisters or cartridges. When the contaminant vapor is not too volatile, activated carbon as such offers sufficient adsorption capacity (physisorption). For specific compounds with a low boiling point the activated carbon may carry an impregnant that gives the carbon a chemisorptive capability. Chemisorption of different vapors requires different impregnants, although it is possible to make use of common properties of categories of contaminants, e.g., acid inorganic gases. The vast majority of respiratory protective devices (RPD) are simply operated by the human lungs, resulting in a pulsating flow pattern and thereby affecting the sorption behavior.1-5 The negative influence of this flow pattern in non-power-assisted RPD on the breakthrough performance was discussed earlier.1 In that work the risk of obtaining too optimistic numbers for the protective capacity of a carbon canister on the basis of constant flow tests was quantified. Another risk of using sorbents in RPD is addressed in the present paper. There exists a tendency in industry to equip RPD with a blower device, which turns them into power-assisted devices. The function of the blower is to continuously draw air through the canister and to blow the air into the protective part around the head of the user. The reason for this approach is obvious: in RPD it is important to have as low a pressure drop over the filter as possible. This ensures both a better protection and a higher comfort for the user. Equipping RPD with a blower device entails the hazard of desorption. Because air is continuously drawn through the canister, also during exhalation, the possibility increases that adsorbed contaminants are desorbed from the activated carbon. Intermittent use of filters and changes in the environment (temperature, humidity) may also augment desorption, which is evidently an undesired situation. Up till now, little has been published on this topic. Wood6 has studied a related topic, namely rates of triethylenediamine (which is meant to aid in the sorption of cyanochlorine) desorption from impregnated charcoals to determine whether its release would pose a health * Author to whom correspondence should be addressed. Phone: +31-15-2843521; fax: +31-15-2843963; E-mail: [email protected].

hazard. Ackley7 reported that desorption of 1,3-butadiene, which had been classified as a potential occupational carcinogen, occurred readily when clean air was drawn through cartridges saturated with 1000 ppm 1,3 butadiene. Schmidt8 et al. examined cabin air filters for cars and found that when the input concentration falls, desorption of previously adsorbed pollutants occurs. Surely, many papers9-14 have been published showing that humid conditions negatively influence the breakthrough performance, which is not the same as a desorption risk. It is noted that this well-known phenomenon, as such, is not the central matter of concern in this paper. In this work the extent of the occurrence of desorption is examined under various conditions. The work aims to explore whether or not significant risks exist. To this purpose, experiments were performed in which conditions for the occurrence of desorption were created. The experimental conditions such as flow, concentration, humidity, and contaminant challenge times corresponded to practical situations and were derived from the conditions described in test procedures. These test procedures, either from the United States15 or the European Union Countries,16 are used for official approval of activated carbon canisters for RPD for civil15,16 or military17 use. The contaminant challenge times were 10, 20, and 30 min. These were chosen as representative lengths of attacks (military), of industrial accidents (civil), or of incidents where terrorism is involved. Furthermore, cyclohexane was chosen as a representative physisorbed vapor and ammonia was chosen as an example of a chemisorbed gas. In the extent desorption phenomena pose a risk for the users of RPD, they should be addressed somehow in approval tests for filters. Currently, the European approval standard EN14387 for RPD16 contains a desorption test only for special filters against specific named compounds; for all other filter types, such a test is not required. Experimental Section General. The existence of a desorption risk was examined by means of a series of sorption experiments in which the same procedure was followed. During a certain period of time, a bed of activated carbon was exposed to an air flow containing a contaminant. After this period, the supply of the contaminant was stopped while a flow of clean air was continued. The

10.1021/ie060909b CCC: $37.00 © 2007 American Chemical Society Published on Web 01/10/2007

Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4035

Figure 1. Experimental setup for the cyclohexane measurements, in which the carbon was divided into two separate beds of 1.0 cm each.

effluent contaminant concentration was monitored during the entire experiment. From the viewpoint of a possible occurrence of a desorption risk, it is especially interesting what happens after the supply of contaminant has been stopped. The sorption experiments were carried out in a precision glass tube at a temperature of 23 °C with (a) cyclohexane and Norit R1 carbon, and (b) ammonia and impregnated carbon. The impregnated carbon was taken from a commercially available canister, that carried the official approval mark of the EU. Air was used as the carrier gas. The bed dimensions, i.e., length and diameter, and the flow rate were chosen such that the superficial air velocity corresponds to the conditions encountered in civil gas mask canisters. Essentially the same flow conditions prevail in military canisters. The experimental conditions were as follows: the column diameter was 0.050 m, the bed height was 0.020 m, and the total flow rate was 0.125‚10-3 m3/s or 7.5 L/min, which corresponds to a superficial gas velocity of 0.064 m/s. The carbon was “rained” into the column and then vibrated during 30 min at 170 Hz. This procedure practically ensures that the maximum packing density of the carbon bed is obtained. Procedures that are used by canister producers also aim at the maximum packing density. Cyclohexane. For all measurements with cyclohexane, the entire sorbent bed was divided into two separate beds, in series, of 1.0 cm each, see Figure 1. The cyclohexane concentration was monitored gas chromatographically after each bed. In this way extra information was obtained about the progress of the concentration front. More specifically the following measurements were performed: • Cyclohexane experiments C1, C2, and C3: these measurements were performed under dry conditions; the activated carbon was dried overnight, and the relative humidity of the feed gas was e10%. The cyclohexane concentration was 5.63 g/m3, which was derived from military standards.17 The cyclohexane was supplied during different periods of time: 10, 20, and 30 min, respectively. • Cyclohexane experiments C4 and C5: these measurements were performed under humid air conditions; the activated carbon was conditioned to equilibrium overnight at a relative humidity of 80%, and the relative humidity of the feed gas was also 80%. The cyclohexane concentration17 was 5.63 g/m3 and 1.42 g/m3, respectively. The applied cyclohexane challenge time was 10 min in both cases. • Cyclohexane experiment C6: in this case the activated carbon was dried overnight, but during the measurement, the relative humidity of the feed gas was 80%. The cyclohexane concentration17 was 1.42 g/m3, and the applied cyclohexane challenge time was 10 min.

Ammonia. In contrast to the cyclohexane experiments, the ammonia experiments were performed with single carbon beds. The column diameter was again 0.050 m, but the bed height was 0.038 m in this case. The larger bed height is similar to the situation in practice where canisters apply larger beds of impregnated carbon than beds of non-impregnated carbon. The carbon was not preconditioned but used as received; this carbon has a water content of about 10 wt %. The presence of adsorbed water is needed to facilitate the chemisorption process.18 The relative humidity of the gas stream (air and ammonia) was 70%. Because of the role of water in the chemisorption process, experiments under dry conditions were not performed in this case. The ammonia concentration was 3.54 g/m3 (5000 ppm), which was taken from the European standard EN14387.16 The superficial velocity was 0.064 m/s. In a first experiment, ammonia was supplied until breakthrough occurred, which appeared to be 48 min. Subsequently, a series of experiments were performed with shorter periods of supplying ammonia, i.e., ammonia was supplied during 20, 23, 27, and 35 min. Ammonia was detected with an ammonia specific electrochemical sensor (polytron, Dra¨ger), whose measuring range is between 0.04 and 80 mg/m3. Results Cyclohexane. Figure 2 shows the results of the experiments C1, C2, and C3 where challenge times for cyclohexane were applied of 10, 20, and 30 min, respectively. The concentration is plotted against time on logarithmic scales. The air flow was stopped overnight and during weekends and resumed the next (Mon)day, causing a sudden change in concentration course. The numbers in the figure indicate the length of these periods without flow (in hours). The concentration was monitored halfway through the carbon bed (1.0 cm) and at the end of the carbon bed (2.0 cm). In all three cases cyclohexane was positively detected after the first part of the bed. The point at which the feed of cyclohexane was stopped is visible as well: the concentration rises less fast or decreases even at that moment. The system is flushed with clean air that causes the concentration to rise less fast or even to decrease initially. Subsequently, previously adsorbed cyclohexane starts to release: desorption occurs, and the concentration shows a further increase. A longer challenge time results in a higher concentration. Although at 20 and 30 min challenge time the concentration rises above the breakthrough criterion of 5 mg/m3, this is measured halfway through the carbon bed (1.0 cm). The concentration at the end of the bed is of more importance, as filter canisters are equipped with carbon beds that typically have a bed length of 2.0 cm. Only in case of a challenge time of 30 min was cyclohexane detected at the end of the bed. This occurred, though, only after more than 200 min; nevertheless, the concentration did rise above the breakthrough criterion of 5 mg/ m3. For 10 and 20 min challenge time, no cyclohexane was detected at the end of the bed even after 10-12 h of clean air flow, i.e., the concentration is below the detection limit of 0.01 mg/m3. The concentration jumps that are observed after a prolonged period of air standstill (overnight or during weekends) must be attributed to redistribution of cyclohexane. Depending on the gradient as a function of time in the effluent concentration immediately before the interruption, the jump is positive or negative. If this gradient is positive, the gradient in the bed as a function of distance on the moment the flow is interrupted must be negative. The latter gradient is necessarily a driving force for redistribution during the period of air standstill. So, if the resulting diffusion is fast enough, it leads to a higher load

4036

Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007

Figure 2. Effluent concentrations of cyclohexane on Norit R1. The feed concentration was 5.63 g/m3, supplied during 10, 20, and 30 min; carbon was dried, and feed humidity was e10%. The concentration jumps are a consequence of a period of air standstill.

Figure 3. Effluent concentrations of cyclohexane on Norit R1. The feed concentration was 1.42 g/m3, supplied during 10 min; carbon was conditioned at a relative humidity of 80%, and feed humidity was 80%.

of adsorbate into the direction of the end of the (fractional) bed. Hence, a positive jump arises in the effluent concentration on the moment the air flow is restarted, see experiments C1 and C2 in Figure 2. Conversely, when the gradient in time is negative, the jump is negative; see experiment C3 in Figure 2. The experiments discussed so far were performed under dry conditions: the carbon was dried, and the relative humidity of the feed gas was e10%. In actual use the relative humidity usually is much higher, leading to loading of the sorbent with water during the challenge, and even to an a priori loading in case the RPD is worn for reasons of prevention. Figure 3 shows the results of experiment C5 where the activated carbon was conditioned to equilibrium overnight at a relative humidity of 80%, and the relative humidity of the feed gas was also 80%. Due to the pretreatment, the weight of the carbon increased from 8.0 g (dry) to 11.7 g (humid). The figure shows an almost instant breakthrough of cyclohexane after the first part of the carbon bed, and, after a relatively short time, also at the end of the bed cyclohexane breaks through. The humid conditions have a strong effect on the capacity of the carbon. Under these conditions, the desorption risk is less of a concern than an initially poorly performing carbon bed. Nevertheless, the carbon does adsorb some cyclohexane, which

is much easier desorbed under humid than under dry conditions: the outlet concentration remains high, >10 mg/m3 during a period of about 250 min, indicating a relatively fast desorption. The initially fast breakthrough is primarily a matter of decreased capacity, while the prolonged period of desorption is a matter of both decreased capacity as well as a faster desorption rate. So, humid conditions do intensify the desorption risk. Figure 4 compares sorption experiments under dry and humid conditions. Experiment C1 was performed under completely dry conditions, whereas in experiment C4 the carbon was pretreated at 80% RH, while also the feed was at 80% RH. The challenge time was 10 min. The difference in breakthrough behavior of cyclohexane is evident. Under dry conditions, the concentration is still below the breakthrough criterion after 250 min (just before the overnight stop); the breakthrough criterion is 5 mg/ m3. Moreover, no cyclohexane was detected at the end of the bed. On the contrary, under humid conditions, breakthrough occurs almost immediately after the first part of the bed. Furthermore, in this case breakthrough also occurs at the end of the bed. This comparison clearly demonstrates that desorption under humid conditions may be very rapid. Figure 5 also illustrates the effect of humid conditions on the breakthrough behavior. In the two compared cases (C5 and

Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4037

Figure 4. Effluent concentrations of cyclohexane on Norit R1. The feed concentration was 5.63 g/m3, supplied during 10 min. Experiment C1: dry conditions; experiment C4: carbon was conditioned at a relative humidity of 80%, and feed humidity was 80%.

Figure 5. Effluent concentrations of cyclohexane on Norit R1. The feed concentration was 1.42 g/m3, supplied during 10 min, and feed humidity was 80%. Experiment C5: carbon was conditioned at a relative humidity of 80%; experiment C6: carbon was dried.

C6) the RH of the feed gas was 80%. In experiment C5 the carbon was pretreated at 80% RH, whereas in experiment C6 the carbon was dried. Again, the challenge time was 10 min. The figure shows the concentration profile at the end of the bed (after 2.0 cm). A large difference exists in breakthrough behavior. Humidified carbon has a strong effect on the cyclohexane capacity of the carbon. This effect is well-known and finds its origin in the fact that the sorption capacity of activated carbon usually is largely occupied by water at a RH of 80%. In case of dry carbon (experiment C6) cyclohexane is found in the effluent after about 6 h, although the challenge time of 10 min is relatively short, and the concentration rises even to higher values than the breakthrough criterion. Comparing experiment C6 with C1, see Figures 5 and 2, it is obvious that a humid feed results in an earlier breakthrough inducing a relatively fast desorption. So again, humid conditions do increase the desorption risk. The experiments show that the risk exists that initially adsorbed contaminants may be released at high concentrations. Although an even more rapid desorption is observed under humid conditions, it is clear that dry conditions are certainly not always free from a desorption risk either. In practice the

desorption risk is dependent on the weather conditions. Dry weather conditions mean that the relative humidity typically is 30%, while the humidity can rise to 80% or more in case of wet weather conditions. Moreover, the amount of adsorbed water on carbons typically increases fast when the relative humidity rises from 40% to 75%, and thus the risk of desorption increases as well. Under the conditions of lung-powered RPD, the flow pattern differs from that in the present experiments in that a pulsed air flow passes through the activated carbon filter. For the adsorption phase of the here-studied adsorption-desorption process, the pulsating flow effectively causes an earlier breakthrough, while the mass transfer in the stagnant air film around the carbon beads is less favorable than in the case of constant air flow of the same net value.1 During the desorption phase however, the rate-determining step of the mass transfer is found in the elementary desorption of the adsorbate from the carbon surface. The overall rate of the desorption is much lower than that of adsorption; compare the duration of the desorption process (order 100-1000 min or longer) with that of the adsorption (tens of minutes). This difference can only be attributed to the elementary desorption step, which requires, in contrast to the

4038

Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007

Figure 6. Effluent concentrations of ammonia on impregnated carbon. The feed concentration was 3.54 g/m3, supplied during 20, 23, 27, 35, and 48 min; carbon was used as received, and feed humidity was 70%. The concentration jumps are a consequence of a period of air standstill.

elementary adsorption step, an increase in the energy content of the adsorbate molecules. So desorption is not influenced by the flow pattern, and therefore essentially the same conclusions hold for activated carbon filters in lung-powered RPD. Ammonia. The results of the sorption measurements with ammonia are presented in Figure 6. In a first experiment, ammonia was supplied until breakthrough happened, which apparently occurs after 48 min. The breakthrough curve is rather steep; the concentration exceeds the upper detection limit within 2 min after the initial detection of ammonia. The other curves shown in the figure were obtained while supplying ammonia during a series of shorter periods namely 35, 27, 23, and 20 min. In three out of the four experiments, where loading the carbon was discontinued (while continuing the airflow) before the breakthrough criterion of 3.5 mg/m3 (5 ppm) had been reached, the criterion was exceeded by 1 to 2 orders of magnitude. The fourth experiment (20 min of challenge) was terminated when the criterion was nearly attained. It is noticed that the effluent concentration profiles of the runs with 27 and 23 min supply time, respectively, show a relatively broad gap. This suggests some kind of transition. The cause of this phenomenon must be found in the specific sorption process of ammonia. We suggest the following explanation to this observation. The challenge of the activated carbon bed is a process in which the pulse of ammonia gradually penetrates into the bed. Ammonia in the frontal line of the pulse is initially sorbed physically; part of this ammonia disappears by a chemical reaction with the impregnant, i.e., by chemisorption. The presence of adsorbed water is needed to facilitate the chemisorption process. The adsorbed (or chemisorbed) ammonia is replenished by fresh incoming ammonia. In the successive slices of the bed, equilibrium between physisorbed ammonia and influent concentration will establish itself. If the challenge time is relatively long, the sorption front is close to the end of the bed at the moment the challenge is stopped. From this moment on, physisorbed ammonia will desorb and enter the portion of the bed with pristine carbon. While this portion is small, little ammonia can be removed from the air flow, resulting as yet in a rapid breakthrough and a steep rise in the effluent concentration. If the challenge time is relatively brief, a larger portion of the carbon is vacant to sorb the ammonia that desorbs from the loaded carbon. This results into a more postponed breakthrough.

The concentration slope is less steep in this case because diffusion has broadened the frontal part of the ammonia pulse. Figure 6 indicates the existence of a desorption risk for ammonia. In case of 23 min supply time the ammonia concentration rises to levels higher than 3.5 mg/m3, which is the breakthrough criterion, after 300 min. In case of 20 min supply time the concentration reaches the level of 0.71 mg/m3 (1 ppm) after 300 min. In case of extended use or reuse of filters, the desorption of these levels of ammonia could lead to a nonnegligible exposure. Since physisorption appears to be very important for the overall sorption process of ammonia, the same reasons that hold for the physisorption of cyclohexane suggest that also under pulsating flow conditions the desorption of ammonia could pose a risk. The interruptions of air flow (standstill of air overnight or during a weekend) in the experiments of 27 and 23 min challenge times appear to bring about a downward jump in the effluent concentration. These jumps are probably caused by the chemisorption of ammonia during the period of air standstill, which lasts as long as the impregnant is not depleted. Particularly the experiment in which the challenge time was 27 min supports this explanation. The concentration gradient in the effluent just before the interruption has been practically zero for a relatively long time. So, little diffusion in the carbon bed can occur in that situation, although over even longer periods influence from farther upstream could be felt. Without such influence no explanation other than chemisorption is available for the downward jump in concentration after the restart of the air flow. Conclusions Under activated carbon canister test conditions that are derived from civil and military standards, the following conclusions are drawn from the exploratory experimental investigation: 1. Both physisorbed and chemisorbed contaminants may be released from activated carbon filters in significant concentrations once the influent concentration of the contaminant has been reduced to zero. 2. A redistribution of the physisorbed contaminant over the activated carbon bed occurs in a period of rest (e.g., one night or more), which may lead to augmented release of contaminant when the filter is reused. In the case of a chemisorbed contaminant, such an effect may be overridden by the favorable effect of lagging chemisorption.

Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 4039

3. Under humid conditions, the desorption of physisorbed contaminants occurs more rapidly than under dry conditions. 4. Ammonia is not only adsorbed by chemisorption but partly by physisorption as well. This reversibly adsorbed ammonia desorbs significantly from the sorbent bed. These observations have been made on the basis of experiments under constant flow circumstances. Insight into the details of the sorption process suggests that for activated carbon filters in lung-powered RPD essentially the same conclusions hold. Because desorption phenomena pose a risk for the users of RPD, it is advisable to address this somehow in all approval tests for vapor filters. Literature Cited (1) Linders, M. J. G.; Mallens, E. P. J.; Van Bokhoven, J. J. G. M.; Kapteijn, F.; Moulijn, J. A. Breakthrough of Shallow Activated Carbon Beds under Constant and Pulsating Flow. Am. Ind. Hyg. Assoc. J. 2003, 64, 173. (2) Suzin, Y.; Nir, I.; Kaplan, D. The effect of flow pattern on adsorption of dimethyl methyl phosphonate in activated carbon beds and canisters. Carbon 2000, 38, 1129. (3) Van Bokhoven, J. J. G. M.; Schell, J. M.; Baak, P. J. In Adsorption: science and technology; NATO ASI Series E,; Rodrigues, A. E., LeVan, M. D., Tondeur, D., Eds.; Kluwer Academic Publishers: Dordrecht, 1989; Vol. 158, pp 419-425. (4) Moyer, E. S. Review of influential factors affecting the performance of organic vapor air-purifying respirator cartridges. Am. Ind. Hyg. Assoc. J. 1983, 44, 46. (5) Nelson, G. O.; Harder, C. A. Respirator cartridge efficiency studies IV. Effects of steady-state and pulsating flow. Am. Ind. Hyg. Assoc. J. 1972, 33, 797. (6) Wood, G. O. Desorption of TEDA from Impregnated Respirator and Adsorber Charcoals. Am. Ind. Hyg. Assoc. J. 1984, 45, 622. (7) Ackley, M. W. Chemical Cartridge Respirator Performance: 1,3Butadiene. Am. Ind. Hyg. Assoc. J. 1987, 48, 447. (8) Schmidt, F.; Sager, U.; Da¨uber, E. Dynamic adsorption behaviour of cabin air filters. Filtr. Sep. 2002, 39, 43.

(9) Yoon, Y. H.; Nelson, J. H. Effects of humidity and contaminant concentration on respirator cartridge breakthrough. Am. Ind. Hyg. Assoc. J. 1990, 51, 202. (10) Cohen, H. J.; Zellers, E. T.; Garrison, R. P. Development of a field method for evaluating the service lives of organic vapor cartridges: results of laboratory testing using carbon tetrachloride. Part II: Humidity effects. Am. Ind. Hyg. Assoc. J. 1990, 51, 575. (11) Wood, G. O. A model for adsorption capacities of charcoal beds I. Relative humidity effects. Am. Ind. Hyg. Assoc. J. 1987, 48, 622. (12) Jonas, L. A.; Sansone, E. B.; Farris, T. S. The effect of moisture on the adsorption of chloroform by activated carbon. Am. Ind. Hyg. Assoc. J. 1985, 46, 20. (13) Nelson, G. O.; Correia, A. N.; Harder, C. A. Respirator cartridge efficiency studies: VII. Effect of relative humidity and temperature. Am. Ind. Hyg. Assoc. J. 1976, 37, 280. (14) Lodewyckx, P.; Vansant, E. F. The influence of humidity on the adsorption capacity from the Wheeler-Jonas model for the prediction of breakthrough times of activated carbon beds. Am. Ind. Hyg. Assoc. J. 1999, 60, 612. (15) MSHA Subpart L, Chemical Cartridge Respirators. In Code of Federal Regulations; Title 42, Part 84, Center for Disease Control and Prevention, 1995. (16) Respiratory ProtectiVe DeVices-Gas Filters and Combined FiltersRequirements, Testing, Marking (EN 14387:2004); Comite´ Europe´en de Normalisation (CEN): Brussels, Belgium, 2004. (17) U.S. Army Performance Specification, Canister, Chemical-Biological Mask: C2A1, MIL-PRF-51560A(EA); Edgewood Research, Development and Engineering Center: Aberdeen Proving Ground, MD, 1997. (18) Balieu, E. Fundamental aspects in air filtration and purification by means of activated carbon filters. In Gas Separation Technology; Vansant, E. F., Dewolfs, R., Eds.; Elsevier Science Publishers: Amsterdam, 1990; pp 91-136.

ReceiVed for reView July 13, 2006 ReVised manuscript receiVed November 20, 2006 Accepted November 22, 2006 IE060909B