TECHNICAL PAPER Price

ISSN 1047-3289 J. Air & Waste and

Manage. Schmidt
Assoc. 48:1135-1145
Copyright 1998 Air & Waste Management Association

VOC Recovery through Microwave Regeneration of Adsorbents:

Process Design Studies
David W. Price and Philip S. Schmidt
Center for Energy and Environmental Resources, University of Texas at Austin, Austin, Texas

ABSTRACT in the effluent. Economical recovery is therefore prob-

Process design studies are described for a new type of VOC lematic because low condensation temperatures are re-
recovery system which uses microwave heating to regen- quired in the case of hot gas stripping, or because a subse-
erate adsorbents. Microwave regeneration systems create quent liquid separation step is needed with steam strip-
a highly concentrated effluent from which the VOCs can ping. Using microwave heating to regenerate adsorbents
be recovered by condensation at near-ambient tempera- may significantly enhance recoverability. Unlike conven-
tures. Important design considerations, predicated on tional regeneration methods, microwave regeneration
experimental work and model development, are identi- deposits heat directly and rapidly within the adsorbent
fied and discussed. Parametric studies are then described bed, producing a highly concentrated effluent so that the
that identify the optimal adsorbent selection, operating VOCs can be easily recovered by condensation at near-
cycle, recovery configuration, regeneration pressure, re- ambient temperatures.
generation final coverage, and column configuration. In During the 1980s a number of researchers investigated
general, it was found that microwave regenerated adsorp- the potential for using microwave heating to regenerate
tion systems favor the use of low dielectric loss-factor adsorbents. Burkholder et al.1 found that applying micro-
polymeric adsorbents and operation under low pressure wave energy enhances the desorption of ethanol from
conditions (about 5 torr absolute pressure). silicalite because it selectively heats the ethanol without
heating the adsorbent. Furthermore, the researchers recog-
INTRODUCTION nized the possibility for using MW regeneration to facili-
One common method for VOC abatement consists of tate solvent recovery. Singh2 examined the feasibility of
passing the exhaust stream through a suitable adsorbent regenerating water from types 5A and 13x molecular sieves
to strip out organics, then regenerating the adsorbent to in a domestic-type microwave oven. Roussy et al.,3 in an
concentrate the VOCs. Conventional regeneration of experimental and theoretical study, concluded that water
adsorbents entails passing a hot gas, such as steam, air, or molecules are desorbed from 13X molecular sieves directly
nitrogen, upward through the bed. The stripping gas plays by the electromagnetic field; they found that the desorp-
the dual role of providing the necessary heat for desorp- tion kinetics were first order, varied with the square of the
tion and sweeping the desorbed VOCs from the bed (main- electric field intensity, and were independent of the tem-
taining a low VOC partial pressure). However, the large perature of the solid. Gibson et al.4 determined that micro-
volume of stripping gas dilutes the concentration of VOCs wave heating reduced the desorption times of ethylene
oxide from polyvinyl chloride (PVC) by up to 400% over
conventional heating for the same macroscopic tempera-
IMPLICATIONS
ture. All of these studies, consisting of small laboratory set-
The emission of VOCs poses a serious environmental
control problem for many industries using organic solvents.
ups with test-tube quantities of adsorbents, were limited to
In particular, VOC emissions are of concern to the print- characterizing the heat-up and diffusion rates; they did not
ing, metal fabrication, plastics, semiconductor, and textile address the applicability or feasibility of full-scale imple-
industries. As restrictions on VOC emissions continue to mentation of microwave regeneration processes.
increase, there is a growing market for suitable control tech- The use of microwave heating to regenerate
nologies. Most control systems presently employ destruc-
adsorbents has been studied at the University of Texas at
tive abatement techniques; relatively few options exist for
economic recovery and reuse of these valuable solvents, Austin Center for Energy Studies since 1991. A series of
particularly those that are water-soluble. bench-scale experiments have shown that very high heat
transfer rates to the bed can be achieved, limited only by
the capacity of the microwave generator.5 Moreover, mass

Volume 48 December 1998 Journal of the Air & Waste Management Association 1135
Price and Schmidt

transport out of the adsorbent is enhanced by a pressure- Finally, the system process model incorporates the
driven flow (“expulsion”); in fact, the mass transfer rates output from these programs along with other relation-
are so high that desorption follows a quasi-equilibrium ships to determine the overall performance characteris-
process.6 By contrast, conventional regeneration is typi- tics of the system. These include sizing calculations for
cally rate-limited by heat and mass transfer resistance, the adsorbent bed and other components as well as rela-
resulting in longer regeneration cycle times. This study tions for determining pressure drop, blower requirements,
extends this body of work by addressing the technical fea- and power consumption of the microwave generators,
sibility and identifying the most important process con- vacuum pumps, and refrigeration system. The model also
siderations for a full-scale microwave regeneration plant: incorporates an economic analysis to estimate the capital
adsorbent selection, column flow configuration, regen- and operating costs of the individual equipment items
eration purge method, regeneration system pressure, and and the total plant based on vendor quotations and cor-
final regeneration coverage. relations in the literature.
The underlying assumptions and methodology of the
OVERVIEW OF PROCESS MODELS economic analysis are discussed in detail by Price.8,9 Mi-
Two VOC emission streams were chosen as the basis for crowave generator costs are based on quotes from two
study, typical of large-scale industrial printing operations, industrial generator suppliers for 50-75 kW, 915 MHz units.
an important area of application. In the first case, the Other microwave-related costs were estimated following
concentration and flow rate are representative of a plant the procedure developed by Sanio and Schmidt.10 Oper-
in which the VOCs are emitted into and collected from ating costs are predicated on electric power rates of $0.059/
the building permanent total enclosure (PTE flow condi- kWh for 5880 hr per year. In addition to the cost of elec-
tions: 144,000 cfm at 500 ppm). The second stream con- tricity, the microwave operating costs include power tube
tains the same volume flow of VOCs, but concentrated in replacement at $0.031/kWh. Equipment costs for the re-
a much smaller air stream (22,500 cfm at 3220 ppm); this frigeration system, condenser, heat exchangers, vacuum
is consistent with close-capture (CC) collection at the print pumps, adsorption columns, and other equipment are
nip. The VOC was assumed to be methyl ethyl ketone based on vendor quotes and correlations in the literature.
(MEK), a common industrial solvent with representative All costs are updated to January 1995 dollars using the
dielectric and sorption properties. appropriate Chemical Engineering equipment cost indices.
Figure 1 overviews the set of process models devel- Most of the studies discussed below focussed on the ap-
oped to predict the performance of the adsorption and plication of hydrophobic adsorbents such as Dowex
regeneration cycles. First, relationships are developed for Optipore (a polymeric adsorbent produced by Dow Chemi-
the relevant dielectric, sorption, and thermodynamic cal) and UOP Molsiv High Silica Zeolite (MHSZ).
properties. Detailed knowledge of the sorption equilib-
rium behavior over the entire regeneration cycle is cru- SYSTEM CONFIGURATIONS
cial to accurately estimating the heating and purge re- The most common types of adsorption systems employ
quirements for efficient desorption. The microwave des- fixed-beds in batch operations. Here, the adsorbent remains
orption kinetics program is predicated on experimental stationary in the column while the bed is cycled between
observations which indicate that the column follows a adsorption and desorption modes. Efficient desorption re-
quasiequilibrium process.6 An adsorption kinetics program quires two elements: heating to provide the necessary en-
has also been developed following the method of ergy for desorption and to lower the equilibrium capacity
Michaels7 to predict the shape and location of the break- of the adsorbent, and purging of the desorbed VOCs to pre-
through curves and the working capacity of the bed. vent build up of the gas-phase concentration. For micro-
wave regeneration, adequate purge can be achieved by ei-
ther pulling a vacuum on the bed or flowing a purge gas
through it. Both methods reduce the concentration of VOCs
in the vapor-phase in the bed; for vacuum-purge regenera-
tion the desorption effluent is a pure solvent vapor stream
at low pressure while the stream is a mixture of inert and
VOC at ambient pressure for gas-purge regeneration.
Figures 2a-b show two possible solvent recovery con-
figurations for microwave regeneration systems operating
at low pressure. A large “dry” type (non-contaminating)
vacuum pump is used in Figure 2a to directly pull the
Figure 1. Modeling organization for microwave regeneration studies. vacuum on the column. The desorbed VOCs are then

1136 Journal of the Air & Waste Management Association Volume 48 December 1998
 Price and Schmidt

condensed at ambient pressure with cooling water. A heat source, much lower stripping gas flow rates can be
precooler upstream of the vacuum pump cools the desorp- employed for MW regeneration than in conventional
tion effluent to near-ambient temperature in order to in- regeneration. Air is not considered a suitable stripping
crease the density and lower the operating temperature in agent since the elevated VOC concentrations would
the pump. Figure 2b depicts a system where low pressure is exceed the lower explosion limit. The inert gas at the
achieved by solvent condensation at low temperature. A outlet of the condenser is saturated with VOCs and,
small vacuum pump is required for startup and for remov- therefore, cannot be vented directly to the environ-
ing non-condensibles. The vacuum capability of the con- ment. It can either be sent to one of the on-line ad-
denser is limited by the temperature of the coolant; for some sorption columns before venting or recycled back to
solvents, a small refrigeration system would be necessary. the desorber as the purge gas stream. The first of these
The configuration decision strongly depends on the vapor two options is not attractive for two reasons. First, the
pressure curve of the solvent. For low-boiling solvents, such solvent content of this stream would still be relatively
as acetone (B.P. 56 °C at 1 atm), the vacuum-pump system high and readsorbing it would entail additional des-
(Figure 2a) would be preferred, while less volatile solvents, orption stages, and second, the inert gas (probably ni-
such as toluene (B.P. 110 °C), favor the refrigerated con- trogen) is relatively expensive in large quantities and
denser configuration (Figure 2b). must be recycled.
For low concentration emission streams the adsorp- Incorporating a heat-recovery heat exchanger and
tion cycle is much longer than desorption, so there are steam preheater reduces the microwave heating and refrig-
typically multiple columns on-line adsorbing and a single eration requirements for the inert-purge case. The gas-gas
column being regenerated. Therefore, the desorption heat exchanger serves the dual purpose of cooling the des-
modes would be staggered in such a way that the micro- orption effluent and preheating the inert recycled back to
wave generator could be time-shared among the several the bed. Still, the recycled inert temperature is significantly
columns, thus maximizing its duty cycle and minimizing lower than the bed temperature, so a steam-preheater is
generator capital costs. added to increase the temperature to about 100 °C. Both of
An alternative method for removing the desorbed VOCs these devices are relatively inexpensive, but as will be shown,
from the bed is to sweep them out with a flow of inert gas they have a significant favorable impact on the operating
(Figure 2c). Because the stripping gas is not the primary costs of the refrigeration and microwave subsystems.

Figure 2. Microwave regeneration configurations: (a) vacuum-purge MW regeneration with a non-contaminating vacuum pump with atmospheric
pressure condensation, (b) vacuum-purge MW regeneration with a refrigerated condenser and auxiliary vacuum pump, and (c) inert-purge MW
regeneration with heat recovery and steam preheating.

Volume 48 December 1998 Journal of the Air & Waste Management Association 1137
Price and Schmidt

There are important differences between Table 1. Comparison of performance and economics among vacuum and inert-purge microwave
regeneration systems for the PTE flow conditions (144,000 cfm at 500 ppm).
vacuum and inert purge with respect to the des-
orption thermodynamics, kinetics, and overall Vacuum Purge Inert Purge Inert Purge
system performance. Table 1 compares the per- (Figure 2b) With No Heat With Heat Recovery
Recovery (Figure 2c)
formance and economics of vacuum purge re-
generation (Figure 2b) and inert-purge regenera- Adsorbent Optipore Optipore Optipore
tion with and without heat recovery. The PTE Adsorbate MEK MEK MEK
flow conditions were assumed and the key oper- Condensation temperature (°C) -26 -26 -13
Pvoc at bed inlet (torr) 5 5 12
ating variables (condensation temperature, inert P at bed outlet (torr) 5 45 38
voc
flow rate, and final regeneration temperature) Nitrogen flow/VOC flow (kg/kg) — 6.9 10.6
were chosen based on separate optimization stud- Final regeneration temperature (°C) 120 150 150
Microwave power consumption (kW) 243 387 277
ies for each configuration. The levelized cost per
Recovery system power (kW) 111 250 100
pound of treated solvent (COS) in Table 1 reflects Makeup nitrogen costa ($/yr) — 260,000 260,000
b
the total cost of ownership, including both capi- Total capital investment ($) 3,063,000 3,333,000 3,159,000
Adsorption subsystem cost ($) 1,129,000 1,129,000 1,129,000
tal and operating costs, and is normalized based
Adsorbent inventory ($) 1,335,000 1,335,000 1,335,000
on the annual mass of treated solvent. Microwave system cost ($) 403,000 607,000 451,000
The results indicate that vacuum-purge of- Recovery system cost ($) 196,000 263,000 245,000
c
fers substantial performance and economic ben- Total operating costs d ($/yr) 472,000 882,000 762,000
Levelized cost (COS) ($/lbm VOC) 0.206 0.303 0.271
efits for MW regeneration. Most importantly, the
vacuum-purge system has a lower microwave a Makeup nitrogen costs are predicated on the fact that at the beginning of each desorption cycle, the volume
power consumption since, for a given conden- of b
air in the column must be displaced with inert.
The total installed cost of the system.
sation temperature, there is a lower average va- c
Include all utilities, operating and maintenance costs, overhead, property taxes, insurance, and administra-
por-phase concentration in the bed. That is, the tive costs.
d
vacuum pulled by the condenser at a given con- The levelized COS reflects the total cost of ownership, including both capital and operating costs. To
densation temperature establishes a uniform VOC calculate the levelized cost ($/lbm VOC), the capital costs are annualized using a capital recovery factor of
16.28% (10-year equipment life at a 10% interest rate) and added to the annual operating costs. This total
partial pressure in the bed at the corresponding annual cost ($/yr) then is divided by the total mass of treated solvent per year.
saturation pressure. On the other hand, if the
same condenser temperature is available for an
inert-purge system, this partial pressure is maintained only OPTIMAL REGENERATION PRESSURE
at the inlet; the desorbed VOCs increase the concentra- For vacuum-purge batch MW regeneration systems, the
tion of VOCs in the inert stream with bed height. Con- pressure is perhaps the most significant process variable
sequently, the sorption equilibrium curves indicate a since it strongly affects desorption thermodynamics and
higher final regeneration temperature and higher micro- kinetics as well as the size and power consumption of the
wave power consumption for the inert-purge system to recovery system. Lowering the pressure demands larger
achieve the same degree of regeneration. This effect is, vacuum pumps that consume more power. However, mi-
in fact, large enough to outweigh the effect of supple- crowave power consumption and generator capacity both
mental steam heating. decrease since reducing the pressure lowers the final tem-
The cost of the adsorbent inventory, adsorption perature to which the bed must be heated to achieve a given
columns, and recovery system are roughly the same degree of reactivation. Because the system pressure affects
for the three cases. Most of the variation in capital cost the performance of both the desorption and solvent recov-
stems from the microwave system cost, which is pro- ery subsystems, the total impact can only be assessed by
portional to the power consumption. The inert-purge examining the total system performance and economics.
configurations also require a much larger condenser Figure 3 plots the cost per unit of recovered solvent as
because of the higher film resistance caused by the a function of regeneration pressure for both of the vacuum-
inert. The operating costs resulting from the make-up purge configurations described previously. The very clear
nitrogen and the higher microwave and refrigeration decrease in system cost with decreasing pressures reflects
power are mostly responsible for the higher cost of re- the fact that mechanical vacuum pump energy is generally
covered solvent. By adding heat recovery and steam cheaper and has a stronger effect on desorption equilibria
preheating to the inert purge configuration, both the than microwave heating. This variation in system cost with
microwave generator and recovery system power con- pressure apparently stems entirely from the difference in
sumption are reduced substantially. The remaining equipment costs between microwave and vacuum produc-
studies will focus on vacuum-purge regeneration since tion equipment, per unit of power; the sum of the operat-
it is clearly the more attractive configuration. ing costs associated with the vacuum pump and microwave

1138 Journal of the Air & Waste Management Association Volume 48 December 1998
 Price and Schmidt

overall system economics. Moreover, there are several in-

0.25 herent differences in the bed dynamics between micro-
0.24
wave and conventional regeneration.
Levelized Cost, COS

In conventional regeneration systems, it is seldom

0.23 Refrigeration
economical to completely reactivate the bed. First, these
($/lbm)

System

0.22
Vacuum Pump systems exhibit distributed moving heat and mass trans-
fer zones. As the fronts of these zones approach and break
0.21 through the top of the bed, an increasing portion of the
stream’s heat content is thrown away. Therefore, desorp-
0.2
0 20 40 60 80 100 120 140 tion is typically terminated soon after the leading edge of
System Pressure (torr) the zone breaks through the top of the bed, leaving a
“heel” of partially unregenerated bed. Furthermore, com-
Figure 3. Optimal regeneration pressure level for vacuum-purge MW plete regeneration of the bed for conventional adsorbents
regeneration. such as activated carbon and molecular sieves frequently
requires gas temperatures in excess of 250 °C, particularly
generator power consumption does not appreciably change for high-boiling organics. Low-pressure steam, the most
with pressure. Furthermore, regenerating at lower pressures common stripping agent, is generally not available at tem-
implies lower final heating temperatures; this makes the peratures above 200 °C. This problem is compounded by
process less likely to thermally damage the solvent. the fact that the heat of desorption increases sharply at
It appears that using a vacuum pump is slightly more low coverage.
attractive than a refrigerated-condenser at all but the low- Experimental studies have demonstrated that it is
est pressures for MEK. However, dry-type vacuum pumps much more attractive to fully regenerate the bed for
are generally not available as single units for flow rates vacuum-purge microwave regeneration. First, the micro-
exceeding 12,000 acfm, which corresponds to the given waves heat the entire bed volumetrically while the vacuum
mass flow of VOC at 5 torr. The sharp increase in the sys- pump maintains a uniform gas pressure. Consequently,
tem cost at 1 torr stems from the necessity of purchasing there are no moving heat and mass transfer zones and
four separate vacuum pumps (cost exponent of 1) as op- the bed can be efficiently desorbed to completion. Fur-
posed to simply increasing the capacity of a single pump thermore, high bed temperatures are readily achieved with
(typical cost exponent of 0.45). microwave heating since there is no limiting heat source
Figure 3 suggests that further reduction in pressure temperature. Finally, process conditions suggest that it
would be cost-effective for the refrigerated condenser sys- may be desirable to regenerate the bed to near-comple-
tem. However, at pressures below 1 torr the condensation tion in order to minimize the possibility of early break-
temperatures become extremely low for MEK and other through in the subsequent adsorption cycle.
low boiling solvents. Also, further reducing the final re- Figure 4 plots the microwave generator power re-
generation temperature by lowering the pressure reduces quirement as a function of final coverage. As will be
the microwave heating rate and, therefore, may eventu- discussed later, the microwave power consumption is
ally shut down the microwave-enhanced mass transfer a strong function of the working capacity of the ad-
effects and the quasi-equilibrium process observed in the sorbent; regenerating the adsorbent more fully (lower
experimental studies. For example, a 1 torr regeneration coverage) increases the desorption efficiency by put-
pressure corresponds to a final heating temperature of only ting more energy into the desorption phase change
90 °C. Therefore, the optimal pressure for vacuum-purge and less into the sensible heat of the bed. The results
microwave regeneration is probably in the 1–5 torr range shown in Figure 4 correspond to the PTE case flow
for MEK. While these results were for MEK on Dowex conditions, but the same trends are seen for other flow
Optipore at the PTE flow conditions, the same trends have conditions. Note that while the lower pressure curves
been observed in calculations for the MHSZ adsorbent and reflect lower microwave power requirements, the power
with other solvents under different flow conditions. consumption of the vacuum pump and/or refrigera-
tion system are correspondingly higher. Thus, opera-
OPTIMAL FINAL REGENERATION COVERAGE tional and process considerations suggest that the
Once the regeneration pressure has been selected, the Dowex Optipore and MHSZ adsorbents should be re-
degree to which the adsorbent is regenerated depends only generated to near-zero coverage. Consequently, a low
on the final heating temperature. Because of the high unit- enough pressure should be chosen so that the corre-
cost of microwave energy relative to conventional energy sponding final regeneration temperature does not ex-
sources (e.g., steam), this selection strongly affects the ceed the temperature limit of the adsorbent.

Volume 48 December 1998 Journal of the Air & Waste Management Association 1139
 Price and Schmidt

benign species to pass through the bed. Since most ad-

1000
sorption applications involve humid gas streams (e.g.,
MW Generator Power (kW)

150 torr
800 ambient air) it is necessary to either choose hydrophobic
25 torr adsorbents or make provisions for dehydrating the stream;
600 excessive water adsorption hampers performance by rap-
10 torr idly depleting the adsorbent’s capacity, wasting regenera-
400
tion energy to desorb the water, and producing a recov-
ered solvent with a high water content. The Dowex
200
Optipore and MHSZ adsorbents exhibit excellent hydro-
0 phobicity— only adsorbing water at high relative humidi-
0 2 4 6 8 10 ties. Activated carbon is moderately hydrophobic, but
Coverage
(g MEK/100 g adsorbent) exhibits a dramatic increase in its affinity for water at rela-
tive humidities over 50%. Molecular sieve 13X is hydro-
Figure 4. Optimal final regeneration coverage for vacuum-purge philic and only applicable for dry gas streams or in pro-
regeneration at different regeneration pressures.
cesses where the moisture has been removed upstream
using, for example, a rotary desiccant wheel.
ADSORBENT SELECTION Tables 2 and 3 also list the equilibrium sorptive capaci-
Choosing an appropriate adsorbent is one of the most ties for MEK at low and high VOC concentrations, respec-
important process decisions for VOC recovery systems, tively. Dowex Optipore and activated carbon both exhibit
either conventional or microwave-regenerated. In many relatively high sorptive capacities for MEK— about twice
ways, it dictates the performance of both adsorption and that of MHSZ and MS 13X. The actual capacity of the bed,
desorption cycles. This study considers four adsorbents however, depends in part on the sharpness of the isotherm
commonly employed for adsorbing volatile organic com- and the adsorbent’s kinetic performance. Dynamic adsorp-
pounds from waste air streams: Dowex Optipore 11,12 (a tion occurs in a mass transfer zone with a finite length.
hydrophobic, polymeric adsorbent produced by Dow When the leading edge reaches the top of the bed, the col-
Chemical), UOP hydrophobic Molsiv High Silica Zeolite umn must be regenerated despite the fact that much of the
(MHSZ),13,14 Calgon BPL activated carbon,15 and Grace/ capacity of the adsorbent in the mass transfer zone is not
Davison molecular sieve 13x.16 utilized. Dowex Optipore and MHSZ exhibit much better
Tables 2 and 3 list some of the properties and perfor- kinetic performance (i.e., very steep isotherms and high
mance of these four adsorbents operating in low and high mass transfer rates) and therefore a shorter mass transfer
VOC (MEK) concentration air streams, respectively. These zone than the molecular sieve and activated carbon
cases studies were conducted for vacuum-purge microwave adsorbents. The values for the equilibrium sorptive capaci-
regeneration at a pressure of 25 torr in order to corre- ties in these tables correspond to a low humidity environ-
spond with the available lab data for the different ment (i.e., less than 40% relative humidity).
adsorbents.6 Although unit costs of the MHSZ and Dowex The other factor that dictates the working capacity of
Optipore adsorbents are considerably higher than the the adsorbent is the regenerability. While the Dowex
molecular sieve and activated carbon, a more relevant Optipore and MHSZ may be easily regenerated to comple-
number is the cost per unit of adsorbed VOC, which takes tion, much higher temperatures are required to completely
into account the sorptive capacity. The high cost of the desorb polar solvents from activated carbon and molecular
Dowex Optipore polymeric adsorbent is, in part, balanced sieves. In practice, it is not economical to completely re-
by the fact that its resilience is expected to give it a longer generate these adsorbents. Thus, the actual “working ca-
lifetime than the three years typical of the other adsorbents pacity” of the adsorbent is diminished considerably from
in VOC removal service.11,13,17 the theoretical equilibrium capacities listed in Tables 2 and
The relative performance of each of these adsorbents 3. Activated carbon, for example, typically has a working
from the sorption, process, and dielectric heating perspec- capacity of about half of its equilibrium capacity.18
tives is discussed below. Because the adsorbent affects the
performance of virtually every plant component, the true Process Considerations
effect can be gleaned only by examining the total system One of the largest operating costs stems from the power
performance and economics. requirements of the system fans. While increasing the
adsorbent particle size reduces the pressure drop across
Sorption Considerations the bed, it also increases the mass transfer resistance and,
First, the adsorbent must possess the selectivity to adsorb therefore, reduces the working capacity of the adsorbent.
the VOC(s) of interest while allowing environmentally Table 2 lists the pressure drop for each of the adsorbents.

1140 Journal of the Air & Waste Management Association Volume 48 December 1998
 Price and Schmidt

The average particle size used in this analy- Table 2. Effect of adsorbent selection for the PTE flow conditions (144,000 cfm @ 500 ppm). a,b
sis is 1.5 mm for the Dowex Optipore and 3
Dowex Activated Molecular
mm for the others. In this study, a bed depth Optipore MHSZ Carbon Sieve 13X
of 1 ft was used for the Dowex Optipore and
Adsorbent unit cost ($/lbm adsorbent) 25.00 7.10 2.50 2.50
MHSZ adsorbents because of the high unit
Specific adsorbent cost ($/lbm MEK) 192 101 17 36
cost and short length of the mass transfer Equil. coverage (g MEK/g adsorbent) 0.13 0.07 0.15 0.07
zone; a bed depth of 1.5 ft was more suit- Pressure drop/ft of bed (in. H2O/ft) 12.2 5.0 4.1 8.7
able for the other two adsorbents because System fan power (kW) 343 140 171 366
Specific heat (kJ/kg K) 1.26 0.84 1.05 1.00
of their lower unit cost and larger mass trans- Heat of desorption (kJ/kg) 576 662 752 814
fer zone. As shown in the tables, the system Final regen. temp. (°C) (coverage) 150 (0.00) 193 (0.01) 360 (0.05) 350 (0.01)
fan power is directly proportional to the Specific microwave heat (kJ/kg MEK) 2,071 3,276 6,139 8,575
Microwave generator power (kW) 298 537 1,038 1,840
pressure drop and air flow rate.
Total capital investment ($) 3,039,000 2,549,000 3,015,000 4,268,000
The variation in system fan power is Annual operating costs ($/yr) 477,000 718,000 936,000 1,459,000
rather small compared to the variation in Levelized cost, COS ($/lbm VOC) 0.207 0.223 0.295 0.445
microwave power with adsorbent selection. a
These parameters are based on the available sorption isotherms from the adsorbent manufactures, microwave
All of the microwave heating essentially regeneration experimental data, and the sorption equilibria models.
b
ends up in two places: sensible heat of the The process flow diagram for these systems is shown in Figure 2b.
bed (and VOC) and latent heat of desorp-
tion. The bed and VOC sensible heat is basi- Table 3. Effect of adsorbent selection for the CC flow conditions (22,500 cfm @ 3,220 ppm).a
cally “wasted” energy; it must eventually be
removed in the bed cooling step prior to Dowex Activated Molecular
Optipore MHSZ Carbon Sieve 13X
adsorption and by the cooling water in the
condenser. The microwave power consump- Specific adsorbent cost ($/lbm MEK) 104 55 11 19
tion is a complex function of several vari- Equil. coverage (g MEK/g adsorbent) 0.24 0.13 0.22 0.13
System fan power (kW) 39 14 20 42
ables:
Specific microwave heat req.
(kJ/kg MEK) 1,483 2,133 4,117 4,927
(1) Adsorbent/VOC Heat Capacity. For Microwave generator power (kW) 197 302 620 804
a given final regeneration tempera- Total capital investment ($) 1,321,000 1,197,000 1,578,000 1,878,000
Annual operating costs ($/yr) 239,000 357,000 535,000 666,000
ture, adsorbents with large specific Levelized cost, COS ($/lbm VOC) 0.097 0.116 0.166 0.202
heats require more microwave
a
power. Identical information from Table 2 is not repeated.
(2) Final Regeneration Temperature.
The microwave power and system economics at 500 ppm the molecular sieve requires 8575 kJ
depend strongly on the temperature to which the of microwave heating per kg of desorbed MEK
adsorbent must be heated to achieve the desired compared to only 4927 kJ/kg at 3220 ppm.
degree of regeneration for the given vacuum con- The Dowex Optipore adsorbent has all of these factors
ditions. When comparing different adsorbents, working in its favor. Its high equilibrium capacity and small
it is crucial to employ the optimal final regen- mass transfer zone give it the largest working capacity of
eration coverage. the four adsorbents. It also has the lowest heat of desorp-
(3) Heat of Desorption. Roughly half of the micro- tion and final regeneration temperature. These factors com-
wave heating is consumed by the latent heat of bine to give it the lowest specific microwave heating re-
desorption. Table 2 indicates that there is signifi- quirement. The MHSZ, while having a low equilibrium
cant variation in this value among the four capacity, has a moderately high working capacity because
adsorbents. of its sharp isotherms; it also has low heat of desorption,
(4) Working Capacity. The lower the adsorbent work- heat capacity, and final regeneration temperature. The
ing capacity, the higher the ratio of bed sensible carbon adsorbent has a moderate working capacity, but
heat to latent heat of desorption. That is, more requires high regeneration temperatures and has a high
adsorbent must be heated up to desorb the same heat of desorption. Molecular sieve 13X has all of these
amount of VOC. The effect of the working ca- factors working against it: low working capacity and high
pacity alone can be examined by comparing the heat of desorption, specific heat, and final regeneration
specific microwave heating requirement for a temperature. As a result, six times more microwave heat-
given adsorbent between the high and low VOC ing is required than for Dowex Optipore.
concentration cases (Tables 2 and 3). For example, Tables 2 and 3 also summarize overall system

Volume 48 December 1998 Journal of the Air & Waste Management Association 1141
Price and Schmidt

economic parameters for the prototypical fixed-bed sys- heating takes place in this area. Thus, heating non-unifor-
tems. As shown, the economics are very sensitive to the mities are, to some degree, self-correcting in a manner analo-
microwave generator capacity, so the Dowex Optipore, gous to “moisture leveling” in drying applications. Table 4 il-
while 10 times more expensive (per unit), yields the most lustrates the variation in loss-factor and penetration depth
cost-effective system. The effect of VOC concentration in among adsorbents for MEK adsorption.
the feed stream can be seen by comparing the results from Finally, the selection of an adsorbent may also be in-
Table 3 with that from Table 2. While the performance is, fluenced by the dielectric properties of the solvent. For
in general, better for the CC case because of the higher example, if a very low loss-factor solvent is to be regener-
sorptive capacities, the same trends appear. ated, an adsorbent with a high loss-factor might need to
Process safety considerations may also restrict the use be chosen in order to achieve reasonable bed heat-up rates.
of certain adsorbent/VOC combinations. For example, ke- In short, while the polymeric and high-silica adsorbents
tones, aldehydes, and organic acids can react exothermi- are considerably more expensive, they exhibit superior
cally on activated carbon adsorbents, possibly leading to sorptive performance and their dielectric properties make
bed fires. Finally, if microwave heating non-uniformities them more suitable for microwave heating.
occur, it may be possible to mix these out by intermittently
fluidizing the bed during the desorption cycle. Dowex COLUMN CONFIGURATION
Optipore’s resilience and resistance to attrition make it par- The configuration of the adsorption column for fixed-bed
ticularly attractive if such a scheme is employed. systems represents the heart of the design problem since
it must satisfy both sorption and electromagnetic con-
Dielectric Considerations straints. Figure 5 illustrates the electromagnetic and gas
The property that describes how well a material absorbs flow regimes for three possible column configurations:
microwave energy and converts it into heat is the dielec- axial-flow columns, horizontal columns with an internal
tric loss factor (ε”). In general, the heat-up rate of a mate- rectangular bed, and radial-flow columns. For axial or
rial in an applied electric field is proportional to the di- horizontal-bed configurations, the feed stream is usually
electric loss factor, frequency, and the square of the directed downward through the bed during adsorption
strength of the electric field inside the material. The pen- in order to prevent fluidization of the adsorbent particles
etration depth (δ), defined as the distance from the sur- at the high velocities typical of adsorption. Counter-cur-
face of the bed at which the power decays to 1/e of its rent flow during regeneration (upward) prevents
value at the surface, is inversely proportional to the di- readsorption in the unsaturated bottom portion of the
electric loss factor20 bed.21 In radial-flow columns, the adsorbent is held in
λ0 ε' the annular space between the two cylinders. The follow-
δ= (1) ing discusses some of the important design considerations
2πε" for column selection.

Here, λ0 is the wavelength of the radiation in free space, ε’ Flow Uniformity

is the dielectric constant, and ε” is the dielectric loss factor. A non-uniform air flow distribution can result in portions
Experimental studies have indicated that the dielec- of the bed that are unsaturated or lead to early break-
tric loss-factor of the adsorbent dominates the heat-up rate through. This has a negative effect on the operating costs
when regenerated with microwaves.6 However, eq 1 sug- since the bed must be regenerated more frequently. Axial
gests that adsorbents with very high loss-fac-
tors, such as activated carbon, exhibit short Table 4. Comparison of the dielectric loss factor a (ε”) and penetration depthb (δ) among dry and saturated
penetration depths, making it difficult to (MEK) adsorbents.
uniformly heat a large bed of material. For
this reason, it may be more attractive to Adsorbent ε”dry ε”sat δdry δsat δdry δsat
employ adsorbents with low loss-factors, 2,450 MHz 2450 MHz 915 MHz 915 MHz
such as Dowex Optipore and MHSZ. Roy and (cm) (cm) (cm) (cm)

Schmidt indicate that for such adsorbents,

Molecular sieve 13X 0.25 0.25 15 15 39 39
the loss factor of the bed becomes a strong
Dowex Optipore 0.03 0.24 103 13 276 34
function of the concentration of solvent on
High silica zeolite 0.03 0.15 120 24 324 64
the adsorbent.19 When the bed is saturated,
the bed has a high loss factor; but as the VOC a The loss factor was measured at 2,450 MHz and assumed to not change significantly at 915 MHz.19
is driven off a portion of the bed, the local b The penetration depth is defined as the distance from the surface of the bed at which the power decays to 1/e of
loss-factor decreases and little additional its value at the surface.

1142 Journal of the Air & Waste Management Association Volume 48 December 1998
 Price and Schmidt

and horizontal-bed columns both exhibit excellent flow The bed area perpendicular to the flow is the air flow rate
uniformity, but there are several potential sources for flow divided by the superficial velocity through the column.
maldistribution in radial flow columns:22-24 Pressure drop and residence time constraints limit the
(1) Flow Bypass. Some empty space must be left in superficial velocity to 75–100 ft/min while transportation
the top of the column to provide for backwashing limitations restrict column sizes to about 12 ft in diam-
and adsorbent swelling. The feed stream may eter and 30 ft in length.18 A 144,000 cfm VOC emission
tend to bypass the bed and flow through this stream (PTE conditions) at a superficial velocity of 85 ft/
empty space, resulting in poor bed utilization or min would require a total of 6 horizontal-bed columns,
early breakthrough. 17 axial-flow columns, or 9 radial-flow columns. Thus,
(2) Channeling. Flow may tend to favor either the for large flow rate systems such as those considered in
top or the bottom of the bed if the pressure gra- this study, horizontal-bed columns are usually recom-
dient across the bed is not uniform. It is relatively mended.17,25 The number of required columns is impor-
difficult to achieve a uniform pressure gradient in tant not only because of the large capital cost per column,
radial-flow columns because the finer adsorbent but also because of the added cost of associated microwave
particles settle toward the bottom, resulting in a delivery components (waveguide, applicator, etc.).
higher packing density (lower permeability). While radial-flow columns have been applied exten-
(3) Fluidization. Typical adsorption velocities are on sively for applications such as methanol synthesis, cata-
the order of the minimum fluidization velocity lytic reforming, desulfurization, and catalytic mufflers,
for smaller adsorbent particles. For radial-flow they have not been particularly successful for adsorption
columns, the velocity is even higher at the inner applications. Kovach states that annular beds have been
annulus. Fluidization is avoided with axial and designed but have seen only limited use because of high
horizontal-bed columns by directing the flow attrition losses.25 They also have more severe structural
downward through the bed. problems due to the increased stress acting on the adsor-
(4) Bed Swelling. As the bed rises due to swelling, bent which settles to the lower end of the column while
the top of the bed will move relative to the loca- the bed expands and contracts in successive adsorption/
tion of the top row of flow holes on the inner desorption cycles. Because of the large air flow rates, the
annulus. This will result in poor utilization of inner annulus will have to be about 2 ft for a bed flow
the top layers of the bed. rate of 12,000 scfm.

Electromagnetic Considerations CONCLUSIONS

During microwave regeneration, non-uniformities in elec- Design studies were conducted for several process configu-
tric field distribution will result in bed hot spots and, con- rations to evaluate the effects of various design and operat-
sequently, incomplete or inefficient regeneration. To mini- ing parameters on the performance of fixed-bed MW re-
mize the effect of field strength decay, the microwaves generation systems for VOC recovery. Computer models
will be applied in the direction of the smallest bed di- were developed to simulate the performance of various MW
mension. One possibility for the axial and horizontal-bed regeneration configurations and systematically investigate
columns, is to use the entire column as a multimode cav- the effects of the following key operating variables on per-
ity and direct the microwaves into the bed from both the formance and economics: adsorbent selection, column flow
top and bottom; the bed support structure would be con- configuration, regeneration purge method, regeneration
structed out of fiberglass, alumina, or some other micro- system pressure, and final regeneration coverage. Table 5
wave-transparent material. For radial-flow columns, it may presents the optimal set of operating conditions for the PTE
be attractive to employ a leaky waveguide applicator; di- and CC flow conditions, respectively.
recting the microwaves radially-inward would balance the While the cost of the adsorbent inventory is a large
decay in the electric field intensity with decreasing cross- percentage of the total system cost, the high capacity and
sectional area. However, locating a discrete number of ease of regeneration of the Dowex Optipore and MHSZ
applicators around the periphery might produce hot-spots adsorbents make them the most cost-effective. Both
due to the irregular geometry, constructive/destructive adsorbents are presented as optimal cases because their
interference of the waves, and high field intensities near performance and economics were nearly the same. Acti-
the waveguide slots. vated carbon and conventional molecular sieves are con-
siderably less expensive, but the high desorption tempera-
Column Cost tures, heats of desorption, and low working capacities
The size and number of columns required for adsorption entail several times more microwave power consumption.
is dictated by the volumetric flow rate of air to be treated. The Dowex Optipore and MHSZ adsorbents are also more

Volume 48 December 1998 Journal of the Air & Waste Management Association 1143
Price and Schmidt

attractive from the operational viewpoint Table 5. Optimal plant configuration and operating variables for fixed-bed, vacuum-purge microwave regener-
because they are hydrophobic and com- ated adsorption system (Figure 2a) under both PTE and CC flow conditions.
pletely non-reactive. Carbon is moderately
hydrophobic but impurities within it can PTE Flow CC Flow
144,000 cfm, 500 ppm 22,500 cfm, 3,220 ppm
catalyze exothermic reactions with ke-
tones. Molecular sieves are hydrophilic
Dowex MHSZ Dowex MHSZ
and, therefore, not suitable for this appli-
Optipore Optipore
cation. The polymeric and high-silica
adsorbents also possess relatively low di- Purge method Vacuum Vacuum Vacuum Vacuum
electric loss-factors which implies longer Regeneration pressure (torr) 5 5 5 5
microwave penetration depths and there- Final regeneration coverage 0.00 0.00 0.00 0.00
fore more uniform heating. Final regeneration temp. (°C) 120 150 120 150
Large air flow systems favor horizon- Column configuration Horiz. bed Horiz. bed Horiz. bed Horiz. bed
tal-bed columns beds since they present a Total number of columns 6 6 2 2
larger cross-sectional area and, therefore, Bed height (ft) 1 1 1 1
fewer columns are necessary. This reduces Bed superficial velocity (ft/min) 85 85 85 85
Pressure drop (“H2O) 12.2 5 8.8 3.3
the system complexity and lowers the
Cycle times: ads/reg/cool (hr) 6.4/1.1/0.13 4.7/0.7/0.13 2.4/2.1/0.16 1.7/1.5/0.17
capital and installation costs of not only
Microwave generator capacity (kW) 243 368 168 235
the columns and gas piping, but also the
Microwave power consumption (kW) 254 355 189 252
microwave power delivery components. System fan power (kW) 343 140 39 14
Both horizontal-bed and axial columns are Vacuum pump power (kW) 80 80 74 78
commonly manufactured and exhibit Vacuum pump capacity (scfm) 13,000 13,000 13,000 13,000
good flow distribution. Axial-flow col- Total capital investment ($) 3,099,000 2,450,000 1,408,000 1,230,000
umns would be applicable for smaller air Adsorption subsystem costa 36% 45% 34% 39%
flow systems (<20,000 cfm) because they Adsorbent inventory cost 43% 20% 32% 14%
are easier to fabricate and implement. Ra- Microwave subsystem costsb 13% 24% 15% 29%
c
dial-flow columns are less common and Recovery subsystem costs 7% 10% 16% 19%
d
suffer from flow maldistribution problems. Total operating costs ($/yr) 462,000 639,000 238,000 331,000
Levelized cost, COS ($/lbm) 0.206 0.221 0.099 0.113
In addition to adequate heating, effi-
cient desorption requires that the VOCs a
Includes the cost of the vessels, instrumentation, fans, recovery tank, condenser, decanter, and internal piping.
be quickly removed from the gas space in
Does not include adsorbent inventory.
the bed. It was found that microwave re- b Includes cost of microwave generator, power measurement and control, applicator, and waveguides.
generation of fixed beds favors vacuum c Includes cost of the vacuum pump (condenser and other costs are included with adsorption system costs).
purge rather than flowing an inert stream d Includes cost of cooling water, electricity, microwave power tube replacement, adsorbent replacement, mainte-
through the bed because of lower micro- nance, overhead, property taxes, and insurance.
wave and refrigeration costs. For vacuum-
purge systems, a regeneration pressure of about 5 torr pro- Environmental Solutions Program. Adsorbent samples and
vides the optimal balance between the cost of the micro- technical information were provided by Dow Chemical
wave and vacuum equipment. Either a dry vacuum pump Company and UOP Molecular Sieves. The authors would
or a refrigerated condenser may be employed as the like to thank Dr. H. Robert Goltz of Dow Chemical for
vacuum source. In contrast with conventional regenera- reviewing a draft of this paper.
tion systems, the adsorbent bed should be regenerated to
near-zero coverage. NOMENCLATURE
COS Levelized cost per pound of
ACKNOWLEDGEMENTS treated solvent [$/lbm]
This project has been funded in part with Funds from the LMTZ Length of the mass transfer zone [m]
U.S. Environmental Protection Agency (EPA) as part of PPM Parts per million gas concentration [—]
the program of the Gulf Coast Hazardous Substance Re- q Sorption coverage [wt. %]
search Center. The contents do not necessarily reflect the T Temperature [°C]
views and policies of the EPA nor does the mention of v Velocity [m/s]
trade names or commercial product constitute endorse- δ Depth of microwave penetration [m]
ment or recommendation for use. Additional grants were ε’ Dielectric constant [—]
provided by Borden, Inc., and the University of Texas

1144 Journal of the Air & Waste Management Association Volume 48 December 1998
 Price and Schmidt

ε” Dielectric loss factor [—] 17. U.S. Environmental Protection Agency. Handbook: Control Technolo-
gies for Hazardous Air Pollutants, June 1991, EPA/625/6-91/014.
λ0 Wavelength of microwave radiation 18. Vatavuk, W.M. Estimating Costs of Air Pollution Control; Lewis Publish-
ers/CRC Press: Chelsea, MI, 1990.
in free space [m] 19. Roy, P.; Schmidt, P.S. Dielectric Loss Factors of Industrial Adsorbents
Loaded with Volatile Organic Compounds. In Proceedings of the 30th
International Microwave Symposium; International Microwave Power
REFERENCES Institute: Denver, CO, July, 1995.
1. Burkholder, H.R.; Fanslow, G.E.; Bluhm, D.D. “Recovery of ethanol 20. Metaxas, A.C.; Meredith, R.J. Industrial Microwave Heating; Peter
from a molecular sieve by using dielectric heating,” Ind. Eng. Chem. Peregrinus: London, 1983.
Fund. 1986, 25, 414-416. 21. Ruthven, D.M. Principles of Adsorption and Adsorption Processes; John
2. Singh, V.P. Microwave Regeneration of Molecular Sieves. Canadian Fu- Wiley & Sons: New York, 1984.
sion Fuels Technology 1984; Report No. F84023. 22. Genkin, V.S.; Dil’man, V.V.; Sergeev, S.P. “The Distribution of a Gas
3. Roussy, G.; Zoulalian, A.; Charreyre, M.; Thiebaut, J.M. “How micro- Stream Over the Height of a Catalyst Bed in a Radial Contact Appara-
waves dehydrate zeolites,” J. Phys. Chem. 1984, 88, 5702-8. tus,” Int. Chem. Eng. 1973, 13, 24-28.
4. Gibson, C.; Matthews, I.; Samuel, A. “Microwave enhanced diffusion 23. Ponzi, P.R.; Kaye, L.A. “Effect of Flow Maldistribution on Conversion
in polymeric materials,” Microwave Power and Electromagnetic Energy and Selectivity in Radial Flow Fixed-Bed Reactors, AIChE J. 1979, 25,
1988, 23 (1), 17-28. 100-108.
5. Weissenberger, A.P.; Schmidt, P.S. Microwave Regeneration of 24. Chang, H.S.; Calo, J.M. “Design Criterion for Radial flow Fixed-Bed
Adsorbents. In Proceedings of Fourth Biennial Symposium on Microwave Reactors, AIChE J. 1983, 29, 1039-1041.
Processing of Materials; Materials Research Society: San Francisco, April 25. Kovach, J.L. Gas-Phase Adsorption. In Handbook of Separation Tech-
1994. niques for Chemical Engineers; Schweitzer, P.A. Ed.; McGraw-Hill: New
6. Price, D.W.; Schmidt, P.S. “Microwave Regeneration of Adsorbents at York, 1979, p. 3-39.
Low Pressure: Experimental Kinetics Studies,” J. Microwave Power and
Electromagnetic Energy 1997, 32 (3), 145-154.
7. Michaels, A.S. “Simplified Method of Interpreting Kinetic Data in
Fixed-Bed Ion Exchange,” Ind. Eng. Chem. 1952, 44 (8), 1922.
8. Price, D.W. Recovery of Volatile Organic Compounds by Microwave
Regeneration of Adsorbents, Ph.D. Dissertation, University of Texas
at Austin, 1996.
9. Price, D.W.; Schmidt, P.S. Design Analysis of Microwave-Regenerated
Adsorbent Systems for Recovery of Volatile Organic Compounds. In
Proceedings of the Microwave and High-Frequency Heating Conference,
Cambridge, UK, September 1995; pp. C4.1-4.
10. Sanio, M.R.; Schmidt, P.S. Cost Estimation for Industrial Dielectric Heat-
ing Systems, Process Energetics Program Report 89-01, Center for En-
ergy Studies, Univeristy of Texas at Austin. 1989.
11. Goltz, H.R. Dow Chemical, Liquid Separation Systems, personal com-
munication, 1994-96. About the Authors
12. Goltz, H.R.; Jones, K.; Tegen, M. High Surface Area Polymeric
Adsorbents for VOC Capture and On-Site Regeneration. Presented at Dr. David W. Price is a post-doctoral research associate and
the Air & Waste Management Association 87th Annual Meeting & Dr. Philip S. Schmidt (corresponding author) is the Donald J.
Exhibition, Cincinnati, OH, June 19-24, 1994, 94-TA42.04.
Douglass Professor of Engineering and Head of the Pro-
13. Southworth, J. UOP Molecular Sieves, personal communications,
1993-1995. cess Energetics Program at the University of Texas at Aus-
14. UOP “Molsiv™ High Silica Zeolites: Product Information,” Des Plaines, tin, Center for Energy and Environmental Resources, Aus-
IL, 1993.
15. Xtrusorb™ 600 and 700 Pelletized Activated Carbons. Calgon Car- tin, TX 78712. Dr. Schmidt is the author of two books and
bon Corporation, Technical Bulletin, 1993. numerous journal articles on industrial thermal process de-
16. Davison Molecular Sieves Adsorption Equilibria. Davison Chemical sign and dielectric heating of materials.
Division, W.R. Grace & Co., Technical Bulletin, 1994.

Volume 48 December 1998 Journal of the Air & Waste Management Association 1145