Temperature Swing Adsorption for Postcombustion CO2 Capture

Jan 14, 2016 - A mathematical model used to describe cyclic adsorption processes is calibrated and validated for the simulation of temperature swing ...
4 downloads 0 Views 976KB Size
Subscriber access provided by UNIV OF CALIFORNIA SAN DIEGO LIBRARIES

Article 2

Temperature Swing Adsorption for post-combustion CO Capture: Single- and Multicolumn Experiments and Simulations Dorian Marx, Lisa Joss, Max Hefti, and Marco Mazzotti Ind. Eng. Chem. Res., Just Accepted Manuscript • DOI: 10.1021/acs.iecr.5b03727 • Publication Date (Web): 14 Jan 2016 Downloaded from http://pubs.acs.org on January 15, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

Temperature Swing Adsorption for post-combustion CO2 capture: Single- and multicolumn experiments and simulations Dorian Marx, Lisa Joss, Max Hefti, and Marco Mazzotti∗ ETH Zurich, Institute of Process Engineering, Sonneggstrasse 3, CH-8092 Zurich, Switzerland E-mail: [email protected] Phone: +41 44 632 2456. Fax: +41 44 632 11 41

Abstract A mathematical model used to describe cyclic adsorption processes is calibrated and validated for the simulation of temperature swing adsorption (TSA) processes applied to the capture of CO2 from a model flue gas (CO2 /N2 ) using zeolite 13X as sorbent material. Three types of experiments are reported in this work, all of them performed in jacketed columns packed with 13X. During these experiments, the temperature was measured at five positions along the central axis of the column, and the exit composition was measured on line by mass spectrometry. The first series of experiments were breakthrough experiments, used to characterize transport phenomena within the packed bed. The second series were heating and cooling experiments, which were used to study the heat transfer from the heating fluid in the column jacket to the bed. Lastly, cyclic TSA experiments were performed to test the model’s ability ∗

To whom correspondence should be addressed

1 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

to predict the cyclic steady state behavior of the column as well as the separation performance of the process.

1

Introduction

Adsorption-based separation processes for the purification of the less retained compound have been in use for decades, achieving purities as high as 99.999 % in the purification of H2 . 1 In the purification of the more retained component, cyclic adsorption processes have not been used as much; more recently, however, they have garnered attention as a possible technology for the capture of CO2 from industrial and power plants within the context of CO2 capture and storage or utilization systems. Within a pre-combustion capture scheme, the conditions of the stream to be separated are such that pressure swing adsorption could be a promising candidate for an efficient separation. 2 For post-combustion capture, on the other hand, research has focused on regenerating the bed by a temperature or a vacuum swing, or a combination thereof. 3–5 These processes benefit from an immense flexibility, as they are constituted of a variety of steps that can be combined in many ways, including opportunities for recycle streams that can be used to sweep the column or to increase recovery. This freedom also provides a strong incentive to use model-based process design, which allows to test different configurations and operating conditions at low expense. The model for adsorption-based gas separation processes that has been presented in Casas et al. 6 was developed with the pre-combustion capture of CO2 in mind. Due to the high pressure of the feed in this scenario and the relatively high CO2 content, a pressure swing adsorption process using activated carbon was chosen as a basis for comparison, 2,6,7 with novel materials being considered as alternatives. 8,9 However, the conditions present in the case of post-combustion CO2 capture are quite different – the feed typically contains a lower mole fraction of CO2 , and is near atmospheric pressure. In addition, the N2 that makes up the bulk of the flue gas, adsorbs on many sorbents itself, and

2 ACS Paragon Plus Environment

Page 2 of 36

Page 3 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

so provides competition for adsorption sites. For these conditions, zeolite 13X is a rather promising candidate material, as it is very selective for CO2 over N2 , and has a high capacity for CO2 ; a number of studies are available that use it in a temperature or vacuum swing adsorption process. 5,10–13 In order to be able to use the model for process design and optimization, a full characterization of the system CO2 /N2 on zeolite 13X was performed. Experimental data for the pure component adsorption equilibrium of both N2 and CO2 on 13X as well as binary adsorption equilibrium data, and the description of this data by a model have previously been reported by Hefti et al. 14 ; the parameters determined there are used throughout this work. In order to simulate dynamic processes, it is also necessary to be able to describe transport processes in addition to the equilibrium. There is some work available regarding the mass transfer of CO2 on 13X, which finds that the mass transfer is limited by a combinations of the heat transfer in the bed and molecular diffusion in the macropores; 5,10,15,16 however the results vary, and it is not always clear how the mass transfer coefficient for CO2 was determined. The heat transfer is influenced by a number of factors, some pertinent to the sorbent material and the process conditions, but some very specific to the experimental setup used. In this work, the transport phenomena are characterized using breakthrough experiments to determine mass transfer coefficients for N2 and CO2 as well as the heat transfer coefficient within the bed. Heating and cooling experiments are performed to study the heat transfer between the column wall and the heat exchange fluid that is used to heat and cool the column. Finally, the model and the transport parameters found in this work are used to design TSA experiments using three different temperature levels for desorption; the experiments are then performed and compared to simulation results as a means to validate both the model and the parameters used. Few published works on the calibration of an adsorption column model for indirect heating TSA experiments are available. In general, a partial validation of the model was carried out by comparing predictive simulations to experimental results of process performance indicators, 17 or by comparing predictive simulations to experimental results of

3 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

internal- and exit profiles. 18–20 In this work, the model is validated by considering all the measured outputs, i.e., internal temperature profiles, exit composition and flow rate profiles, and overall process performance in terms of CO2 purity, recovery and productivity.

2

Experimental

2.1

Sorbent materials and gases

All experiments presented in this work were performed in a two-column laboratory setup packed with spherical pellets of zeolite 13X with a diameter between 1.6 and 2 mm, manufactured by ZeoChem (Uetikon, Switzerland), with a binder content between 15% and 25% on a mass basis 14 . Before the first experiment, the zeolite was regenerated by heating the columns to 250 ◦C and maintaining that temperature over night under a constant low flow of helium. This was done primarily to remove any water that might have adsorbed while the sorbent was exposed to the air in the laboratory. In order to avoid adsorption of water after this thorough regeneration, the columns are kept at or above atmospheric pressure whenever possible. In order to regenerate the sorbent between experiments, the columns were kept at 250 ◦C for two hours under a low flow of helium. The feed gas was obtained from Pangas (Dagmarsellen, Switzerland) with a CO2 content of 12 % on a mole basis (the remainder being N2 ) and with a tolerance of ± 2 % relative (i.e. 0.24 % absolute). The helium used for the regeneration of the sorbent and in the transient experiments was also obtained from Pangas and had a purity of 99.999 %.

2.2

Experimental Setup – changes from PSA setup

The setup used to perform the experiments is based on that described by Schell et al., 21 however it has been modified significantly for this work, as in the schematic shown in Figure 1. The gas piping remained the same, including the two mass flow controllers (MFCs) 4 ACS Paragon Plus Environment

Page 4 of 36

Page 5 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

used to measure and control the feed flow rate into the columns, the back pressure regulators (BPRs) used to control the pressure inside the columns (all Bronkhorst High-Tech BV, The Netherlands), all the automatic gas valves, as well as the thermocouples (at 10, 35, 60, 85, and 110 cm from the bottom end of each column) and pressure sensors used to observe the state of the column. The main difference lies in the temperature control of the columns: The two electrically heated columns were replaced with jacketed columns that allow a heat exchange fluid to externally heat or cool the columns. Tubes passing through both the inside and the outside wall of the columns allow the installation of thermocouples for the measurement of the bed temperature without putting them in direct contact with the heat exchange fluid. The columns were constructed in house and are made of stainless steel. Both the inside and the outside wall are of the same material in order to avoid tensions due to thermal expansion. Automatic valves (GSR Ventiltechnik GmbH&Co, Germany) connected to the inlet and outlet of these jackets allow each column to be connected to one of two thermostats (Huber K¨altemashinenbau GmbH, Germany), and thermocouples at each end measure the temperature of the heat exchange fluid entering and exiting each column. During heating, cooling, and TSA experiments one of the thermostats is set to the desorption temperature (up to 150 ◦C), while the other is kept at a set point of 25 ◦C. The automatic valves are controlled by LabVIEW software (National Instruments, TX, USA), which also controls the gas valves, MFCs, BPRs, and records all temperatures and pressures. The two thermostats, meanwhile, are controlled manually. The physical characteristics of the columns and of the bed are listed in Table 1. The exit composition of the product stream was measured using a mass spectrometer (MS) (Pfeiffer Vacuum Schweiz AG, Switzerland), which can be switched to monitor either the light (N2 -rich) or the heavy (CO2 -rich) product. The flow rate of the light product was measured by a Messglas V-100 rotameter (V¨ogtlin Instruments, Switzerland) after the BPR and before the MS, as illustrated in Figure 1 (labeled “FL”), in conjunction with a webcam and an image analysis script written in MATLAB (Mathworks, Massachusetts,

5 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 36

BPR

FL Vent

Vent

PI TI

PI

TI

TI

Product (N2)

TI

TI

TI Feed TI

TI

TI

TI

TI

TI

TI

TI

to MS

Feed

TI Product (CO2) TI

TI

TI

PI

PI

BPR

PI

Theat

Vacuum pump

Tcool

Figure 1: Flowsheet of the 2-column setup used for the TSA experiments. The five thermocouples in each of the columns are placed along the center axis, at 10 cm, 35 cm, 60 cm, 85 cm and 110 cm from the inlet of the column. USA). The MS was calibrated by using two mixtures with different molar ratios of CO2 to N2 to find the calibration factors necessary to convert the ion current measured to mole fractions, accounting for the fact that they both produce a signal at a mass-to-charge ratio of 28. During the TSA experiments presented here, the flow meter could only be used to measure the light product flow rate. The heavy product is produced during the heating step, which is also the longest step of the cycle. Throughout this step the flow out of the column is determined by the heat transfer into the bed and the consequent desorption of CO2 ; however it is rather slow, indeed for much of the step it is below the lower threshold of the flow meter. Because of this, the light product flow rate was measured, and the heavy product amount was calculated.

6 ACS Paragon Plus Environment

Page 7 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

Table 1: Setup dimensions and physical properties of the adsorbent material. column length

L

[m]

1.2

internal radius

Ri

[m]

12.5 × 10−3

external radius

Ro

[m]

15.0 × 10−3

heat capacity walla

Cw

[J/(K m3 )]

4 × 106

13X material density

ρM

[kg/m3 ]

2359

ρp

[kg/m3 ]

1085

bed density

ρb

[kg/m3 ]

652

particle diameter

dp

[m]

1.6 − 2.0 × 10−3

Cs

[J/(K kg)]

920

particle density

heat capacity sorbent

b

Hefti et al. 14

isotherm parameters at a temperature of 25 ◦C. b from Dantas et al. 22 a

2.3

Breakthrough experiments on 13X

A set of eight breakthrough experiments was performed in one of the columns of the setup. All of them used the same 12/88 CO2 /N2 feed mixture, and were conducted at atmospheric pressure (aside from pressure drop through the piping); they were performed at four initial temperatures (25 ◦C, 45 ◦C, 65 ◦C and 100 ◦C) and using two feed flow rates (200 300

cm3 s

cm3 s

and

at standard conditions, i.e., 25 ◦C and 1 bar).

In general the procedure for the breakthrough experiments was the same as outlined in Marx et al. 23 , with differences only in the method of regeneration. The column was heated to the desired experimental initial temperature while filled with helium. Once that was reached, the feed was switched to the 12/88 CO2 /N2 mixture. The flow rate during the experiment was maintained constant by the MFC, and the BPR was left in the open position to keep the pressure as close to atmospheric conditions as possible. Throughout the experiment, the temperature and pressure inside the column, as well as the exit composition were recorded.

7 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2.4

Page 8 of 36

Heating and Cooling Steps

Two sets of heating and cooling experiments were performed, one using helium to exclude the impact of adsorption, and one with the CO2 /N2 feed mixture. For the experiment using the CO2 /N2 mixture, the column was prepared by letting it reach equilibrium with the feed mixture at a flow rate of 50

cm3 s

(at standard conditions) and at 25 ◦C. This flow rate was

sufficiently low that it did not cause a significant pressure drop through the downstream piping, thus maintaining the column at atmospheric pressure. In each set of experiments, the column was heated from 25 ◦C to 100 ◦C, and then cooled down to 25 ◦C again. Heating: This experiment was performed without a feed, and with the column open at the lower end, as it is during the regeneration step of the TSA cycle. By letting the liquid valves switch such that the column was connected to the hot thermostat, hot heat exchange fluid was pumped through the column’s jacket. Temperature and pressure in the column were measured until the column had reached the regeneration temperature. The final state of the column was the initial state for the cooling experiments. Cooling: For these experiments, the liquid valves were switched back to let cold heat exchange fluid through the column jacket. If the column were closed during these experiments, the adsorption of the CO2 by the cooling bed would create an underpressure, possibly resulting in air (with some humidity) entering the column. In order to avoid that, the column was kept at atmospheric pressure during these experiments by keeping the lower end of the column open and flowing feed past the open valve and through the bypass (see Figure 1) at the lowest flow rate possible with the MFC.

2.5

Cyclic TSA experiments

The TSA experiments were performed using both columns of the laboratory setup (in contrast to the transient experiments). The columns underwent the same three-step process cycle, out of phase by one-half of one cycle, and used the same feed flow rate of 300

cm3 s

(at

standard conditions). As the desorption temperature, TH , has a large impact on the process 8 ACS Paragon Plus Environment

Page 9 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

performance, three different temperature levels were considered: 100 ◦C, 125 ◦C, and 150 ◦C. The three process steps were adsorption, heating, and cooling: During adsorption, feed was passed through the column while it was connected to the cool thermostat (flow direction ↑). In this step, the light product was produced (N2 -rich). During heating, the column was left open at the lower end while it was connected to the hot thermostat (flow direction ↓). In this step, the heavy product was produced (CO2 -rich). During cooling, the column was connected to the cool thermostat again, and was left open at the lower end while feed passed by the open end at minimal flow rate to maintain atmospheric pressure (flow direction ↑). No product was produced in this step. To determine the timing of the steps, the model was used using the transport parameters obtained from the single-column experiments. More specifically, a simple parametric analysis was performed for each value of TH , varying the duration of the adsorption step between 3 and 12 minutes and the duration of the cooling step between 6 and 20 minutes. As the main objective in a post-combustion CO2 capture process is the CO2 product stream, conditions were considered that should yield a CO2 purity larger than 90 %, according to simulations. Other important objectives are the CO2 recovery (

CO2 produced , CO2 fed

often referred to as the cap-

ture rate) and the productivity, i.e. the amount of CO2 that is produced by a unit mass of sorbent in a unit time (

kgCO

2

tzeo h

).

Each of the thermostats was connected to one column at any given time, i.e., there was always one column being heated and one being cooled. As a result, the timing of the process steps is constrained in that theat = tads + tcool . A schematic of the step scheduling can be seen in Figure 2.

column 1 column 2

ads heat

heat cool

ads

cool heat

Figure 2: Scheduling of the process steps in the TSA experiments. The thick vertical lines indicate the switch between heating and cooling each of the columns.

9 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Each of the TSA experiments was performed until cyclic steady-state (CSS) was reached. Simulations performed before the experiments which were used to choose appropriate operating conditions typically converged to CSS in four cycles – as a result, each of the experiments was run for at least five cycles.

3

Modeling and Parameter Estimation

The model is in principle the same as presented by Casas et al. 6 , with the material and energy balance for the bed unchanged. The underlying assumptions are repeated here: • radial gradients of concentration, velocity, and temperature are neglected; • the gas phase and the solid phase are in thermal equilibrium; • the fluid phase is described by the ideal gas law, as shown to be reasonable in an earlier work 6 ; • the mass transfer rate for adsorption is described by a linear driving force model; • mass transfer coefficients, heat of adsorption, viscosity, heat conductivity, and heat capacities are constant; • the pressure drop along the column is described by the Ergun equation; • kinetic and potential energies are neglected in the energy balance. The adsorption equilibrium was described using a Sips isotherm using the parameters determined in Hefti et al. 14 and employing the ideal adsorbed solution theory (IAST) to describe the binary adsorption equilibrium. IAST was shown in the previous work to provide a good compromise between accuracy of description and computational intensity. The heat released during adsorption as well as the external heating and cooling can lead to the formation of a radial temperature profile. While a two-dimensional model could provide higher accuracy by describing such a profile, it would also require greater computational 10 ACS Paragon Plus Environment

Page 10 of 36

Page 11 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

effort. The intended use of the model used in this work is the design and optimization of complex process cycles, which can involve the simulation of thousands of operating conditions with numerous steps each until cyclic steady state is reached. For this reason a one-dimensional model is used, considering separately the heat transfer to the column wall, the thermal inertia of the wall, and the heat transfer to the heating fluid. The only difference to the model used previously is in the energy balance for the column wall. In the new setup the column is heated and cooled externally, and Tamb , the “ambient” temperature, represents the temperature of the heat exchange fluid in the column jacket, which is measured at the inlet and outlet of the jacket. The temperature of the heat exchange fluid at any given time depends on a multitude of factors; predicting this temperature would be challenging and depending on many phenomena that are very specific to the laboratory setup used. The temperature of the fluid surrounding the column is therefore taken as the measured value, interpolated linearly along the length of the column, and used as an input into the model. For the simulation of the breakthrough experiments, where the thermofluid temperature is essentially constant, Tamb is assumed to be constant and equal to the average of the thermofluid temperature over the duration of the experiment. In the case of the selection of the experimental operating conditions in Section 5, the temperatures are obviously not known before the experiments are performed. For these simulations the data from the single column heating and cooling experiment are used and scaled as needed for the different desorption temperature levels. Once the experiments are performed, the simulations are repeated using the experimental temperature data. The parameters that remain to be determined are the mass transfer coefficients for CO2 and N2 , kCO2 and kN2 , and the heat transfer coefficients between the sorbent bed and the wall under conditions of flow (hL ) and under static conditions (h0L ), as well as the heat transfer coefficients between the wall and the heat exchange fluid (hw ).

11 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3.1

Page 12 of 36

Parameter estimation

The estimation of the transport parameters was carried out by comparing measurements taken during the transient experiments (breakthrough, heating, and cooling) to simulations, and using the fmincon routine that is part of the optimization toolbox of MATLAB (The MathWorks, Inc., Natick, MA, USA). The observed variables to be compared were the mole fractions of N2 and CO2 (but not the inert He used to pressurize the column), as well as the temperature inside the bed at five positions along the column, just as in a previous work. 23 The objective function was altered to include weighting factors, and is now defined as

Ncomp

Φobj (p) = ωy

X

 ln 

k=1 Ntemp

+ ωT

X k=1

1 Nobs,y

 ln 

 X  yj,k − yˆj,k (p) 2  y + γ j,k j=1 !2  Nobs,T X Tj,k − Tˆj,k (p)  T j,k j=1

Nobs,y

1 Nobs,T

(1)

where yj,k and Tj,k are the measured output (mole fraction or temperature) k at time j, yˆj,k and Tˆj,k are the corresponding simulated values, Nobs,y and Nobs,T are the number of observed data points for the composition and temperature, respectively, γ = 1 is a constant introduced to avoid dividing by zero when a mole fraction is zero, Ncomp = 2 and Ntemp = 5 are the number of components and temperatures used for the comparison, and ωy and ωT are the weighting factors introduced. As their purpose is to relate the importance of the two measured mole fractions to the five temperatures, an unnecessary degree of freedom was eliminated by constraining the weighting factors to Ncomp ωy + Ntemp ωT = 1. For the breakthrough experiments, which were used to characterize kCO2 , kN2 , and hL , the weighting factors were chosen as ωy = 0.25 and ωT = 0.1 such that the weight given to the mole fractions and the temperatures is the same. For the heating and cooling experiments, which were used to characterize h0L and hw , only the five temperatures inside the bed were used for comparison with simulations. As a result, the objective function consists only of the second 12 ACS Paragon Plus Environment

Page 13 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

term in Eq. (1) (i.e., ωy = 0, hence ωT = 0.2). A slight shift in the timing of temperature and composition fronts can create significant difficulties for the determination of transport parameters. Such a shift can be caused by a small deviation of the flow rate from the set point, even within the specified accuracy of ±2% of the full scale of the MFC’s operating range. The feed velocity was therefore regarded as a model parameter to be estimated along with the transport parameters hL and kCO2 , as discussed previously by Marx et al. 23

3.2

Heat transfer in the bed

The problem of heat transfer in a packed bed is not a new one, and a number of studies are available in literature 24–28 . In Marx et al. 23 , heat transfer coefficients fitted to breakthrough experiments performed with activated carbon were compared to values predicted by an expression by Leva, 24 and it was found that the values were similar. However, the Leva correlation goes to zero for Re = 0. As the heating and cooling steps take place at very low flow velocities and also the breakthrough experiments, which were performed at atmospheric pressure, exhibit relatively low Reynolds numbers, the correlation developed by Leva is no longer applicable. Instead, the values of hL fitted to the experiments in this work were compared to a selection of correlations from literature that each also consider a static component to the heat transfer; two correlations from literature were chosen for this comparison: DeWasch and Froment found a linear relationship between the Nusselt and Reynolds numbers; 27,29 while Specchia et al. worked with a two-dimensional model, considering separately the effective heat transfer within the packed bed, KR , and the heat transfer at the column wall, αw , with each a static and a convective component. 25 The expressions needed for both approaches are summarized in Table 2, while a more detailed discussion of them is available in Section 3 of the supporting information.

13 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 36

Table 2: Heat transfer in a packed bed according to DeWasch and Froment and according to Specchia et al. 25,27,29 The expressions needed to obtain the numerical values in this table are available in Section 3 of the supporting information. The thermal conductivity of the gas was calculated under feed conditions. 30 DeWasch and Froment 27,29

Nu = Nu0 + 0.033PrRe

Pr =

hL dp λg ρνCp λg

1 hL

1 αw

Nu =

Specchia et al. 25

=

+

h0L dp λg vdp ν

Nu0 = Re =

= 1.51

Ri Bi+3 3KR Bi+4

Bi = αw Ri /KR KR = λ0e +

Reρg νCp   d 8.65 1+19.4 2Rp i

αw d p λg

=

Nu0w

+ 0.0835Re0.91

Nu0w = 3.07 thermal conductivity

4

λg = 0.028 mWK

Results of the single column experiments

The aim of this work was to validate a model for fixed bed adsorption processes by calibrating it for temperature swing adsorption processes, such that it may then be used to simulate, design, and optimize these processes for their applications in post-combustion CO2 capture. In the following, the transient experiments performed to determine the transport parameters are presented, followed by a discussion of the phenomena characterized. In Section 5, this characterization is used to determine operating conditions for TSA cycles which are then tested in experiment.

4.1

Breakthrough

In Figure 3, the results of one of the breakthrough experiments are shown alongside simulation results. The experiment was performed with an initial column temperature of 25 ◦C, a feed with a CO2 content of 12 % and a flow rate of 300

cm3 . s

It can be seen that the N2 does

not adsorb very much, as it breaks through almost immediately; some does adsorb, how-

14 ACS Paragon Plus Environment

Page 15 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

ever, as evidenced by the moderate temperature increase throughout the bed. The CO2 , on the other hand, adsorbs significantly more, and breaks through much later. A very notable feature of these results is that the adsorbing CO2 creates a front of very high temperature as it advances through the column, with temperatures in excess of 90 ◦C. The results of all breakthrough experiments performed are shown in Section 1 of the supporting information.

4.1.1

Mass transfer

In general, the uptake of CO2 on 13X is controlled to a large extent by the molecular diffusion in the macropores, which depends on pressure, temperature, and composition. However, in the experiments reported in this work the heat transfer limitation was such that the fitting of the mass transfer of CO2 was inconclusive, as the shape of the CO2 front showed very little sensitivity to values of kCO2 ≥ 0.05 s−1 . For relatively high concentrations of CO2 , the large amount of heat released as the CO2 adsorbs hinders adsorption, to the point where the rate of uptake can be limited by the rate at which the bed cools, rather than the adsorption kinetics of the CO2 . Giesy et al. used a frequency response apparatus to study the mass transfer of CO2 on 13X at 23 ◦C and at pressures up to 1 bar, and found that temperature effects played a significant role in the adsorption kinetics. 15 Hu et al. used a zero length column system to study gas mixtures containing between 0.1 and 10 % CO2 , and found that for the higher CO2 concentrations the heat transfer was indeed what determined the uptake rate. 16 The breakthrough experiments performed at higher temperatures (65 ◦C and 100 ◦C) showed some sensitivity to kCO2 , as the equilibrium adsorbed amount of CO2 is lower, and thus less heat is released. A value of kCO2 = 0.1 s−1 worked quite well to describe these experiments – this value was used for all simulations. The mass transfer of N2 was sufficiently fast that the data obtained did not allow for an accurate estimate for kN2 , as any value over 0.5 s−1 described the data similarly well. Therefore, its value was fixed at kN2 = 0.5 s−1 .

15 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 0.95 mole fractions [-]

N2 0.9

0.1 CO 2

0.05 0 10 cm

100 90 Temperature [°C]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

35 cm 60 cm 85 cm 110 cm

80 70 60 50 40 30 20 0

200

400

600 800 1000 1200 1400 time [s]

Figure 3: Concentration (top) and temperature (bottom) profiles for the breakthrough experiment with an initial temperature of 25 ◦C, a feed composition of yCO2 = 0.12, and a feed 3 flow rate of 300 cms . Experimental data (symbols) are shown together with the simulations (lines) carried out with the mass transfer coefficients kN2 = 0.5 s−1 , kCO2 = 0.1 s−1 , and the heat transfer coefficient fitted to this experiment, hL = 35.3 mW 2 K . The concentration profile is shown with a broken axis to provide better detail on the CO2 concentration.

16 ACS Paragon Plus Environment

Page 16 of 36

Page 17 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

4.1.2

Heat transfer in the bed

Each of the breakthrough experiments was used to fit a value of hL individually to that experiment. The values used for the mass transfer coefficients and the heat transfer on the outside of the column in these simulations were: kCO2 = 0.1 s−1 , kN2 = 0.5 s−1 , hw = 220 mW 2 K. Figure 3 shows that the temperature changes seen in these experiments are rapid and large. While the simulations capture the timing and the height of the temperature peaks fairly well, some differences are evident: The experimental data go slightly higher and show a delay before beginning to cool off. In principle, a better description of the heat transfer could be achieved by considering a more detailed model. The simulated temperature in the bed, e.g., represents an average temperature across the cross-section of the bed at that position, while the temperature measurement takes place at the center of the column axis, where heat dissipation to the column wall takes longer than elsewhere. Nevertheless, the overall description is good, and adding complexities such as the radial dimensions to the model does not seem to be warranted. The heat transfer coefficients determined for each of the breakthrough experiments are shown in Figure 4 along with values predicted from the two correlations from literature which were discussed in Section 3.2. Both correlations obtained values slightly higher than those determined experimentally, but they are quite close in value, and exhibit the same linear increase with Re.

4.2

Heating and cooling

In Figure 5 the results from the heating and cooling experiment using the same CO2 /N2 feed as the breakthrough experiments are illustrated. While these two steps were performed in series, they are shown in separate axes, as the individual steps are truncated to make the details more visible. The full length of the steps was 1800 s for the heating step and 1200 s for the cooling step. Along with the temperatures measured in the bed (symbols) and the corresponding simulation results (lines), the measured temperature of the heat exchange fluid at the jacket inlet and outlet is shown using dots (dark and light gray, respectively). These 17 ACS Paragon Plus Environment

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

heat transfer coefficient, hL [W/(m 2 K)]

Industrial & Engineering Chemistry Research

60

Specchia

40 De Wasch

20

0 0

20

40

60

80

100

Reynolds number, Re [-]

Figure 4: Heat transfer coefficients fitted to each of the experiments performed (symbols), together with two correlations found in literature (lines). The equation of the two correlations are shown in Table 2. ◦ Breakthrough experiments, 4 heating/cooling experiment. data were not simulated, but rather were treated as an input into the model, as discussed in Section 3. From the figure it can be seen that • The thermofluid temperature initially oscillates a bit, and in both cases overshoots past the set point temperature. This is in part due to back-mixing of the fluid from the heating jacket to the thermostat, e.g. the cold fluid going into the hot thermostat. It is also one of the reasons the fluid temperature was used as the input for the model – predicting this behavior would entail extensive calibration that would be very specific to this laboratory setup and was therefore outside the scope of this work. • The temperature of the thermofluid entering the column and exiting the column are fairly similar. This is due to the relatively short residence time of the fluid in the column jacket. If the difference were large, the linear interpolation used to calculate the fluid temperature along the length of the column might not be accurate enough, and the energy balance for the thermofluid might have to be included in the model. • A temperature gradient develops along the length of the bed. As indicated in Figure 5 by the gray arrow, the temperature measured in the lower part of the bed changes 18 ACS Paragon Plus Environment

Page 18 of 36

Page 19 of 36

Heating

Cooling feed = 12/88 CO2 /N2

Tset

100

Temperature [°C]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

80

x

60 x 40 Tset 20 0

200

400

600

800

0

200

heating time [s]

400

600

800

cooling time [s]

Figure 5: Temperature profiles inside the column during the heating and cooling experiment using the CO2 /N2 feed. Symbols are measurements at the five positions:  10 cm, ◦ 35 cm,  60 cm, 4 85 cm, ? 110 cm. The dots are the measured thermofluid temperature at the jacket inlet and outlet. Lines represent simulations. The arrows indicate the axial profile that developed within the column. considerably faster than in the upper portion of the column. This trend is seen in both the heating and the cooling step. The temperature gradient that develops along the column axis during heating and cooling could not be sufficiently explained by the temperature gradient in the heat transfer fluid, as it can be seen that the measured temperatures of the heat transfer fluid entering and exiting the heating jacket are very similar. Nor could it be explained by uneven saturation of the bed, as it was also present in the cooling experiment as well as in the heating and cooling experiment using the inert helium (reported in Section 2 of the supporting information). The most convincing explanation is that hydrodynamically and thermally developing flow in the entry region of the heating jacket introduce a dependency of the heat transfer on position, leading to uneven heating of the bed. While this is specific to the laboratory setup and thus has no bearing on the design and development of TSA processes in general, it may 19 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

be necessary to account for this in order to accurately describe the experiments performed. In the jacket around the column, the heat exchange fluid is flowing in an annular tube. The VDI Heat Atlas (Part G2) provides correlations that describe the heat transfer for this kind of geometry for hydrodynamically developed and developing flow. 31 Considering these entrance effects in the simulation lead to an improved description of the heating and cooling experiments – therefore, all simulations of heating, cooling, and TSA cycle simulations in this work account for them. The expressions and a discussion regarding the necessity thereof can be found in Section 4 of the supporting information. All of the transport parameters that were established via the transient experiments reported above are summarized in Table 3. These are the values that are used for all following simulations of TSA processes. Table 3: Transport parameters used for the simulation of the TSA experiments. mass transfer coefficient CO2

kCO2

[s−1 ]

0.1

mass transfer coefficient N2

kN2

[s−1 ]

0.5

internal (flow)

hL

[ mW 2 K]

33

internal (static)

h0L

22

external (column end)

hw,L

[ mW 2 K] [ mW 2 K]

heat transfer coefficients

5 5.1

220

TSA cyclic experiments Prediction of column behavior during cyclic TSA experiments

Finally, the mass and heat transfer coefficients determined above were used to predict the behavior of cyclic TSA processes; specifically, of the three TSA experiments that were performed. The operating conditions, i.e., step times were chosen based on the parametric analysis for the three desorption temperatures of 100 ◦C, 125 ◦C, and 150 ◦C. Aside from the CO2 purity, objectives important in a post-combustion CO2 capture operation are the CO2 20 ACS Paragon Plus Environment

Page 20 of 36

Page 21 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

recovery (

CO2 produced , CO2 fed

often referred to as the capture rate) and the productivity (

kgCO

2

tzeo h

).

The results are summarized in Figure 6, showing the CO2 recovery and the productivity for all simulations performed; conditions which did not meet the CO2 purity specifications (≥ 90%) are shown as small crosses (×), while those that did are represented by empty symbols. The arrows indicate the trends with the process step times. Longer adsorption step times saturate the bed more thoroughly, leading to a higher cyclic capacity, purity, and productivity at the cost of CO2 recovery. Increasing the heating and cooling time creates a larger temperature swing and a higher cyclic capacity, which allows for higher CO2 recovery, but reduces productivity. The simulations for TH = 100 ◦C show that for this regeneration temperature the purity specifications could only be fulfilled with a rather long adsorption step at the cost of reduced recovery. While this is by no means a comprehensive optimization of the TSA process, it illustrates clearly the trade-off between the CO2 recovery and the productivity for each desorption temperature. From these simulations a set of conditions was chosen that led to a reasonable trade-off between CO2 recovery and productivity . The operating conditions chosen for each of the TSA experiments are reported in Table 4. In the following, the results of the TSA experiment with a desorption temperature TH = 150 ◦C are reported and discussed in detail. Each of the measured quantities as well as the calculated quantities, such as the product purities and component recoveries, are considered and compared with simulation results. The results for the other two TSA experiments performed are reported in Section 5 of the supporting information. Table 4: Operating conditions chosen for the TSA experiments performed. TH

feed yCO

100 ◦C

0.12

125 ◦C 150 ◦C

2

V˙ feed

h

cm3 s

i

step times [s] tads

theat

tcool

300

480

1380

900

0.12

300

240

960

720

0.12

300

360

1080

720

21 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 0.9 0.8 0.7

CO 2 recovery [-]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 36

T = 150°C H

0.6

T H = 125°C

tads

0.5

theat

0.4 0.3 T H = 100°C

0.2 0.1 0

0

20

40

60

80

100

120

productivity [kg CO2 /(tzeo h)]

Figure 6: Simulations (black, orange-red) and experimental results (blue) for the CO2 recovery and productivity of the column for three desorption temperatures:  100 ◦C, 4 125 ◦C, ◦ 150 ◦C. The small crosses (×) represents operating conditions that yielded a CO2 purity < 90 %. Simulations shown in black were re-calculated using measured temperature profiles for the jacket temperature. Note that arrows indicating the trends with the step times have been depicted for the set of data at TH = 125 ◦C; the same trends apply to the other temperature levels. 5.1.1

Temperature profiles

The temperature profiles measured inside one column during the TSA experiment are shown in Figure 7 along with the simulation results. During the adsorption step the heat released by the adsorption of the advancing CO2 front is clearly visible in both experiment and simulation, and the timing of the experimental temperature front is matched rather well by the simulations. The height of the temperature is not described as well – this might be due to the neglect of the radial gradients in the model. During the heating step, the axial gradient seen in the heating and cooling experiments is not quite as evident, as the column does not start out in a homogeneous state, but rather has an initial gradient that 22 ACS Paragon Plus Environment

Page 23 of 36

adsorption

160

heating

cooling

140

Temperature [°C]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

120 100 80 60 40 20 0

200

400

600

800

1000

1200

1400

1600

1800

2000

cycle time [s]

Figure 7: Temperature profiles inside the column during the TSA experiment with a desorption temperature of 150 ◦C. Symbols are measurements at the five positions:  10 cm, ◦ 35 cm,  60 cm, 4 85 cm, ? 110 cm. The dots are the measured thermofluid temperature at the inlet (dark) and outlet (light). Lines represent simulations. The vertical gray lines indicate the switch times between steps. runs opposite of what develops during the heating. Nevertheless, the simulations describe the temperature increase well, and by the end of the heating step simulations and measured data are all within a narrow band. During the cooling step that follows, the axial gradient is much more visible in both experiment and simulation.

5.1.2

Exit profiles

Figure 8 shows the composition and flow rate of the light product (N2 ) over the duration of the adsorption step. Data from both columns are shown, both to illustrate that the two columns operate in the same way, and to verify the reproducibility of the experiment. Along with the data on the product, the feed composition and flow rate are shown. The mole fraction of CO2 in the N2 product is shown in Figure 8a. The measured composition is initially very close to the feed compositions of yCO2 = 0.12, however this is an artifact that 23 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

is due to the fact that the preceding step in the TSA cycle is the open cooling step, wherein the feed is passed through the bypass at a low flow rate, and is then analyzed by the MS. The composition then quickly settles on a low concentration of CO2 that slowly decreases throughout the adsorption step until the last few seconds, where it begins to increase again. The level of this plateau is determined by the isotherm, more specifically the partial pressure of CO2 that is in equilibrium with the amount of CO2 adsorbed at the end of the heating step. It slowly decreases between about 30 s and 300 s during the adsorption step, as the top section of the bed is still cooling (note the data series denoted by ? in Figure 7). Both the level of this plateau and the slow decrease are captured very well by the model. The final increase in yCO2 is due to the imminent breakthrough of CO2 , which simulations predict to be a little earlier than seen in the experiments. The flow rate of the N2 product is shown in Figure 8b. For this TSA experiment, the step timing was such that the CO2 front did not appreciably break through – as a result the product flow rate is essentially constant for most of the step. The exception is the beginning of the step, where the feed flow rate ramps up over the first few seconds, and the product flow rate lags behind it by about 15 seconds. The relatively high flow rate during this step of the process resulted in a significant pressure drop. In fact, even with the BPR open, the downstream piping and valves provided enough resistance so as the pressure at the end of the column was typically around 1.3 bar during the breakthrough experiment and the adsorption step of the TSA experiments. As the heating and cooling step see much lower flow rates, they are essentially at atmospheric pressure. Until the pressure in the column builds up, the exiting flow rate is lower. In addition, the simulations show a slightly lower flow rate than measured. While the difference is not large, it is consistent throughout the entire duration of the adsorption step; it should be noted, however, that it is within the uncertainty of the flow meter. The composition of the CO2 product can be seen in Figure 9. As discussed in Section 2, the flow rate of this stream was not measured, so the composition shown here did not enter

24 ACS Paragon Plus Environment

Page 24 of 36

flow rate [mol/s]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

mole fraction CO 2 [-]

Page 25 of 36

0.12

a)

y feed

b)

feed

0.1 0.08 0.06 0.04 0.02

0.012

product 0.008 0.004 0 0

60

120

180

240

300

360

adsorption step time [s]

Figure 8: Comparison of simulation (lines) and experimental (symbols) results for the light product. a) shows the mole fraction of CO2 in the N2 product throughout the adsorption step, as well as the feed composition. b) shows the flow rate of the light product together with the feed flow rate. To show that the two columns operate consistently, data from both columns are shown:  column 1,  column 2. into the process performance calculations. However, as the course of the composition is very sensitive to both the isotherm and the rate of heating, which in turn depends on the heat transfer into the column, it constitutes a good test of the model and the parameters used. Initially, this product is mostly nitrogen, as it is determined by the (non-selective) gas phase present in the column at the end of the adsorption step. As the column heats, and the CO2 desorbs, it displaces the N2 in the gas phase, until the gas phase consists of pure CO2 . In both columns, it took just over two minutes to reach 80 % CO2 purity, and about five minutes to reach 95 %. The simulations match this development remarkably well. The simulated internal profiles of the CO2 mole fraction inside the column throughout the TSA cycle are reported in Section 6 of the supporting information.

25 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 mole fraction CO2 [−]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

0.8 0.6 0.4 0.2

yfeed 0

120 240 360 480 600 720 840 960 1080 heating step time [s]

Figure 9: Comparison of simulation (lines) and experimental (symbols) results for the composition of the heavy product. Data from both columns are shown:  column 1,  column 2.

5.2

Process performance of TSA experiments

From the data presented in Figure 8, the total amount and composition of the light product can be determined by integration over the duration of the adsorption step. Similarly, the total amount fed to each column is known from the MFC measurements. This allows the calculation of the purity of each product and the recovery of each component. Table 5 summarizes the calculated performances for the TSA experiments performed, together with the corresponding simulation results. The values reported are the mean of the values of both columns during the final cycle performed, while the uncertainties reported are propagated from the uncertainties considered in the measured quantities. It can be seen that the simulations manage to predict all process performance indicators to within the experimental uncertainties. However, it is quite evident that the numbers calculated for the experimental N2 recovery and CO2 purity require close scrutiny, as both of these quantities are calculated to be over 100 %, while the CO2 purity shows excessively large uncertainty (see below). The performance indicators that could be determined, however, show the predictive capabilities of the model. The CO2 recovery and the column productivity are shown in Figure 6 as blue symbols with their respective uncertainty interval, and the simulated performance for those same conditions are shown with black symbols. The energy consumption could not be determined experimentally, as this would require knowledge of the flow rate of the heat exchange fluid, as well as assumptions about the heat 26 ACS Paragon Plus Environment

Page 26 of 36

Page 27 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

losses to the outside wall of the column and the surroundings. In Table 5, the simulated values for the specific thermal energy requirement for regeneration is reported. Two values are reported for each experiment: One considers the column including its wall (sim. ext.), while the other considers only the bed (sim. int.). It can be seen that around four times as much energy is used to heat the column wall as compared to the bed. This is mostly due to the fact that the experimental setup in our laboratory was designed to also be useful for PSA experiments, and the column walls are accordingly thick. In fact, simulations performed with a more appropriate geometry, representative of a shell and tube type adsorber, suggest that heating the wall requires around 70 % as much energy as compared to the bed, resulting in an overall energy consumption of around 5 kgMJ for the processes shown in Table 5. Both CO2

energy consumption and productivity are comparable to values found in literature. Merel et al. 12 studied an indirectly heated TSA process similar to the one presented here, using a different reactor geometry with a smaller specific heat transfer area, and without using a dedicated cooling step. They report a specific heat consumption of 8.8 kgMJ including the reactor and all losses, and a productivity of 39.7

kgCO

2

tzeo h

CO2

. For a different zeolite (NaUSY), and

using direct heating, Ntiamoah et al. 32 reported energy consumption of 3.4 kgMJ considering the bed only, and a productivity of 24

kgCO

2

tzeo h

CO2

. The energy consumption is also in a similar

range as an absorption-based CO2 capture process presented by Abu-Zahra et al. 33 ; they have reported an energy requirement of 3.3 kgMJ for their optimum process. CO2

The cycle performed in this work was simple and not optimized for energy efficiency; more complex cycles and rigorous optimization can further reduce the thermal energy requirement. A useful comparison with other technologies should only be done with an accordingly designed process. The scale-up of this process can then be carried out by increasing the cross-section of the columns while keeping the superficial velocity and column length constant. In the case of thermal regeneration via indirect heating and cooling, the specific heat transfer area must be kept constant as well; this could be achieved through appropriate column design with an integrated heat exchanger.

27 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 28 of 36

Table 5: Comparison of experimental and simulated process performance figures for the three TSA processes run in experiment. All experiments used the same feed composition and flow rate. desorption temperature TH

N2 purity [%] N2 recovery [%] CO2 purity [%] CO2 recovery [%]  productivity

kgCO

100 ◦C

125 ◦C

150 ◦C

sim.

92.0

97.4

98.9

exp.

92.1 ± 0.6

97.4 ± 0.6

98.6 ± 0.6

sim.

99.6

99.1

99.4

exp.

100 ± 4.2

103 ± 4.4

103 ± 4.3

sim.

90.2

90.1

93.9

exp.

106 ± 101

137 ± 71

127 ± 55

sim.

36.5

80.5

91.9

exp.

36.9 ± 4.5

79.6 ± 5.4

89.7 ± 5.6

sim.

37.8

60.4

91.7

exp.

38.0 ± 4.3

58.6 ± 3.1

88.4 ± 4.2

sim. ext.

13.7

16.7

12.6

sim. int.

3.49

3.59

2.85



2

tzeo h

 thermal energy consumption

MJ kgCO



2

5.3

Process performance error analysis

As was seen in Table 5, there were some challenges in determining some of the process performance indicators from experiments, resulting in rather large uncertainties in some cases. To understand where these large uncertainties come from, one has to look at how the quantities are obtained, and what numbers go into the calculation. The N2 purity is calculated directly from the composition and product flow rate measured, and has an accordingly low uncertainty. The component recoveries are determined from the feed composition and flow rate entering the column, and the light product composition and flow rate. Each of the flow rates introduces some uncertainty, thus leading to a larger overall uncertainty. Additionally, the simulations already predicted a N2 recovery of > 99 %, so even a relatively small error could result in a value over 100 %. The determination of the CO2 purity has to contend with the fact that the CO2 product represents a rather small portion of the feed stream. In the

28 ACS Paragon Plus Environment

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

20

change in quantity [%]

Page 29 of 36

10 0 purity

-10 -20 -4

recovery

CO 2 N2

-2

0

2

4

change in flow rate [%]

Figure 10: Dependence of the calculated product purities and component recoveries on the flow rate of the light product. case of TH = 100 ◦C, less than 40 % of the CO2 in the feed are recovered, which was only 12 % of the feed to begin with. With less than 5 % of the feed ending in this stream, and its amount not being measured directly, but via data on a much larger stream, any error in the flow rate of the light product will be magnified strongly in the CO2 purity. To illustrate this, Figure 10 shows the dependence of the calculated purities and recoveries on the flow rate of the light product. It is clear from this figure that even a small error in the light product flow rate has such an impact on the calculated CO2 purity that any value determined with this setup has to be taken with great care. As discussed in Section 2, the direct measurement of the flow rate of the CO2 stream was not possible due to the highly variable flow rate along with large changes in the product composition, both of which would pose a challenge to any flow meter and introduce uncertainty during the integration. The quantities that could be determined with greater accuracy and more certainty, however, show that the model is indeed useful to design cyclic TSA processes. Along with the results of the simulations used to select the experimental operating conditions, Figure 6 shows the experimental results for the CO2 recovery and productivity. The simulations used to select experimental operating conditions as well as those simulating the actual experimental conditions (with the measured jacket temperature) implemented the correlations for the entrance effects on the heat transfer coefficient in the heating jacket. While these complications to the model helped in describing the temperature within the bed and the exit profiles, 29 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

they bore little significance with regards to the overall process performance. In fact, of the performance indicators calculated only the CO2 recovery and column productivity changed somewhat when simulations neglected the entrance effects, and the changes were on the orkgCO

der of 1 % for the CO2 recovery and 1 tzeo h2 for the productivity for all three experiments, i.e., well within the experimental uncertainty. This indicates that the transport coefficients are important to correctly predict the dynamic behavior of the column, and therefore are needed during cycle design to determine appropriate step timing. The overall performance that can be achieved, however, is dictated by the adsorption equilibrium at the conditions at the end of the heating step and at the conditions during the adsorption step. Phenomena such as the entrance effects are very specific to experimental setup, and for general studies of TSA processes, such as comparison of different recycle configurations, it is reasonable to neglect them.

6

Conclusions

In general, the agreement between the simulations and the experiments reported in this work is good. It is clear that for the cycles investigated here the heat transfer plays a central role; therefore its accurate description was an important component to the overall simulation. It could be shown that the column model used for adsorption processes could be used for the description of TSA cycles by determining the parameters relevant for the adsorption equilibrium and kinetics separately in targeted experiments. The adsorption kinetics of N2 were found to be quite fast, making an accurate determination of a mass transfer coefficient impossible; and those of CO2 were found to be strongly impacted by heat transfer. The heat transfer within the adsorption column exhibited a dependence on the flow conditions that is consistent with what is found in literature, 25,27,28 while the heat transfer in the heating jacket around the column exhibited a clear influence of entrance effects that needed to be accounted for.

30 ACS Paragon Plus Environment

Page 30 of 36

Page 31 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

After incorporation of the determined parameters, the model was capable of simulating the behavior of two adsorption columns during three cyclic TSA processes, predicting not only the temperature and composition fronts, but also the separation performance of the operating conditions tested. The entrance effects on the side of the heating jacket had a significant impact on temperature development during the heating and cooling steps; however their effect on the process performance was very minor, so in general during process design these effects should be neglected, as they pertain only to the setup used in this work. Although it was found that the laboratory setup in its current form and with its current experimental uncertainties cannot accurately confirm the CO2 purity, the other performance figures were predicted accurately by the model. This shows that the model can be used to design more complicated cycle configurations, optimize the operating conditions, and assess scale, energy requirements, and feasibility of TSA processes for CO2 capture.

Acknowledgement Support from the Commission for Technology and Innovation through grant no. 12903.1 as well as from the Swiss National Science Foundation through grant NF 200021-130186 is gratefully acknowledged.

Supporting Information Available Complete set of experimental breakthrough results with the simulated profiles; heating and cooling experiments performed with helium; correlations for the heat transfer coefficient of the bed; entrance effects for the heat transfer coefficient of the thermofluid; complete results of the TSA cyclic experiments. This material is available free of charge via the Internet at http://pubs.acs.org/.

31 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

References (1) Voss, C. Applications of Pressure Swing Adsorption Technology. Adsorption 2005, 11, 527–529. (2) Casas, N.; Schell, J.; Joss, L.; Mazzotti, M. A parametric study of a PSA process for pre-combustion CO2 capture. Sep. Purif. Technol. 2013, 104, 183–192. (3) M´erel, J.; Clausse, M.; Meunier, F. Carbon dioxide capture by indirect thermal swing adsorption using 13X zeolite. Environ. Prog. 2006, 25, 327–333. (4) Su, F.; Lu, C. CO2 capture from gas stream by zeolite 13X using a dual-column temperature/vacuum swing adsorption. Energy Environ. Sci. 2012, 5, 9021. (5) Haghpanah, R.; Majumder, A.; Nilam, R.; Rajendran, A.; Farooq, S.; Karimi, I. A.; Amanullah, M. Multiobjective Optimization of a Four-Step Adsorption Process for Postcombustion CO2 Capture Via Finite Volume Simulation. Ind. Eng. Chem. Res. 2013, 52, 4249–4265. (6) Casas, N.; Schell, J.; Pini, R.; Mazzotti, M. Fixed bed adsorption of CO2 /H2 mixtures on activated carbon: experiments and modeling. Adsorption 2012, 18, 143–161. (7) Schell, J.; Casas, N.; Pini, R.; Mazzotti, M. Pure and binary adsorption of CO2 , H2 , and N2 on activated carbon. Adsorption 2012, 18, 49–65. (8) Schell, J.; Casas, N.; Blom, R.; Spjelkavik, A. I.; Andersen, A.; Cavka, J. H.; Mazzotti, M. MCM-41, MOF and UiO-67/MCM-41 adsorbents for pre-combustion CO2 capture by PSA: adsorption equilibria. Adsorption 2012, 18, 213–227. (9) Casas, N.; Schell, J.; Blom, R.; Mazzotti, M. MOF and UiO-67/MCM-41 adsorbents for pre-combustion CO2 capture by PSA : breakthrough experiments and process design. Sep. Purif. Technol. 2013, 112, 34–48.

32 ACS Paragon Plus Environment

Page 32 of 36

Page 33 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

(10) Haghpanah, R.; Nilam, R.; Rajendran, A.; Farooq, S.; Karimi, I. A. Cycle synthesis and optimization of a VSA process for postcombustion CO2 capture. AIChE J. 2013, 59, 4735–4748. (11) Webley, P. a. Adsorption technology for CO2 separation and capture: A perspective. Adsorption 2014, 20, 225–231. (12) Merel, J.; Clausse, M.; Meunier, F. Experimental Investigation on CO2 Post - Combustion Capture by Indirect Thermal Swing Adsorption Using 13X and 5A Zeolites. Ind. Eng. Chem. Res 2008, 47, 209–215. (13) Joss, L.; Gazzani, M.; Hefti, M.; Marx, D.; Mazzotti, M. Temperature Swing Adsorption for the Recovery of the Heavy Component: An Equilibrium-Based Shortcut Model. Ind. Eng. Chem. Res. 2015, 54, 3027–3038. (14) Hefti, M.; Marx, D.; Joss, L.; Mazzotti, M. Adsorption equilibrium of binary mixtures of carbon dioxide and nitrogen on zeolites ZSM-5 and 13X. Microporous Mesoporous Mater. 2015, 215, 215–228. (15) Giesy, T. J.; Wang, Y.; LeVan, M. D. Measurement of Mass Transfer Rates in Adsorbents: New Combined-Technique Frequency Response Apparatus and Application to CO2 in 13X Zeolite. Ind. Eng. Chem. Res. 2012, 51, 11509–11517. (16) Hu, X.; Mangano, E.; Friedrich, D.; Ahn, H.; Brandani, S. Diffusion mechanism of CO2 in 13X zeolite beads. Adsorption 2013, 20, 121–135. (17) Wang, L.; Liu, Z.; Li, P.; Yu, J.; Rodrigues, A. E. Experimental and modeling investigation on post-combustion carbon dioxide capture using zeolite 13X-APG by hybrid VTSA process. Chem. Eng. J. 2012, 197, 151–161. (18) Plaza, M.; Garc´ıa, S.; Rubiera, F.; Pis, J.; Pevida, C. Post-combustion CO2 capture

33 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

with a commercial activated carbon: Comparison of different regeneration strategies. Chem. Eng. J. 2010, 163, 41–47. (19) Clausse, M.; Merel, J.; Meunier, F. Numerical parametric study on CO2 capture by indirect thermal swing adsorption. Int. J. Greenh. Gas Control 2011, 5, 1206–1213. (20) Ribeiro, R. P. P. L.; Grande, C. a.; Rodrigues, A. E. Adsorption of Water Vapor on Carbon Molecular Sieve: Thermal and Electrothermal Regeneration Study. Ind. Eng. Chem. Res. 2011, 50, 2144–2156. (21) Schell, J.; Casas, N.; Marx, D.; Mazzotti, M. Precombustion CO2 Capture by Pressure Swing Adsorption (PSA): Comparison of Laboratory PSA Experiments and Simulations. Ind. Eng. Chem. Res. 2013, 52, 8311–8322. (22) Dantas, T. L. P.; Luna, F. M. T.; Silva Jr, I. J.; Torres, a. E. B.; de Azevedo, D. C. S.; Rodrigues, a. E.; Moreira, R. F. Modeling of the fixed - bed adsorption of carbon dioxide and a carbon dioxide - nitrogen mixture on zeolite 13X. Brazilian J. Chem. Eng. 2011, 28, 533–544. (23) Marx, D.; Joss, L.; Casas, N.; Schell, J.; Mazzotti, M. Prediction of non-isothermal ternary gas-phase breakthrough experiments based on binary data. Adsorption 2014, 20, 493–510. (24) Leva, M. Heat Transfer to Gases through Packed Tubes. Ind. Eng. Chem. 1947, 39, 857–862. (25) Specchia, V.; Baldi, G.; Sicardi, S. Heat Transfer in Packed Bed Reactors With One Phase Flow. Chem. Eng. Commun. 1980, 4, 361–380. (26) Lee, L.-K.; Ruthven, D. M. Analysis of thermal effects in adsorption rate measurements. J. Chem. Soc. Faraday Trans. 1 Phys. Chem. Condens. Phases 1979, 75, 2406.

34 ACS Paragon Plus Environment

Page 34 of 36

Page 35 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

(27) de Wasch, A.; Froment, G. F. Heat transfer in packed beds. Chem. Eng. Sci. 1972, 27, 567–576. (28) Dixon, A. G. An improved equation for the overall heat transfer coefficient in packed beds. Chem. Eng. Process. Process Intensif. 1996, 35, 323–331. (29) Froment, G. F.; Bischoff, K. B. Chemical reactor analysis and design, 2nd ed.; Wiley series in chemical engineering; Wiley: New York, NY, 1990. (30) NIST, NIST Chemistry WebBook. 2011; http://webbook.nist.gov/chemistry/ fluid/. (31) Gnielinski, V. VDI Heat Atlas, 2nd ed.; Springer: Heidelberg, 2010; pp 691–752. (32) Ntiamoah, A.; Ling, J.; Xiao, P.; Webley, P. A.; Zhai, Y. CO2 Capture by Temperature Swing Adsorption: Use of Hot CO2 -Rich Gas for Regeneration. Ind. Eng. Chem. Res. 2015, acs.iecr.5b01384. (33) Abu-Zahra, M. R.; Niederer, J. P.; Feron, P. H.; Versteeg, G. F. CO2 capture from power plants. Int. J. Greenh. Gas Control 2007, 1, 135–142.

35 ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

Graphical TOC Entry 1 0.9 0.8 0.7

CO 2 recovery [-]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

T = 150°C H

0.6

T H = 125°C

tads

0.5

theat

0.4 0.3 T H = 100°C

0.2 0.1 0

0

20

40

60

80

100

productivity [kg CO2 /(tzeo h)]

36 ACS Paragon Plus Environment

120

Page 36 of 36