Project No. 14-6309

Zeolite Membranes for Krypton/ Separation from Spent Reprocessing Off-

Fuel Cycle Research and Development

Sankar Nair Georgia Institute of Technology

Patricia Paviet, Federal POC Robert Jubin, Technical POC Final Project Report (Oct 2014-Dec 2017)

Project: Zeolite Membranes for Krypton/Xenon Separation from

Reprocessing Off-Gas

(Project 14-6309, PICS WP NU-14-GA-GT-0201-04, CID NE0008298)

List of Graduate and Undergraduate Students Supporting Project

Yeon Hye Kwon, South Korea, Chemical Engineering, expected graduation in 2018, pursuing Ph.D.

Byunghyun Min, South Korea, Chemical Engineering, expected graduation in 2019, pursuing Ph.D.

Shaowei Yang, China, Chemical Engineering, Post-doctoral Fellow.

Kiwon Eum, South Korea, Chemical Engineering, graduated in 2016, Ph.D.

Emily Benjamin, USA, Chemical Engineering, graduated in 2016, B.S.

Christine Kiang, USA, Chemical Engineering, expected graduation in 2018, pursuing B.S.

Abstract

The overall focus of this project is to develop and understand SAPO-34 zeolitic membranes that can separate mixtures of radioisotope krypton-85 and xenon released as off- during used nuclear fuel recycling. The primary advantage of separating 85Kr from Xe is to reduce the volume of radioactive waste for storage. The second advantage is the revenue generated from the sale of high-purity Xe. Zeolite membranes are attractive because of their much lower energy requirements relative to cryogenic distillation, and their high resistance to radiation degradation. We report the detailed study of silicoaluminophosphate zeolite SAPO-34 materials and membranes for this application, due to hypothesized favorable molecular sieving properties. In the 3-year Mission Support project, we developed a novel, high-performance, low- energy intensity, lower-cost zeolite membrane process for Kr/Xe separation during SNF processing; and investigated the underlying molecular adsorption and transport processes in both

‘idealized’ and ‘realistic’ operating conditions to develop reliable synthesis-structure-property relationships for such membranes. Adsorption and diffusion measurements on SAPO-34 crystals indicate their potential for use in Kr-Xe separation membranes, but also highlight competing effects of adsorption and diffusion selectivity. SAPO-34 membranes synthesized on α-alumina substrates via steam-assisted conversion seeding and hydrothermal growth are characterized in detail, with Kr permeances 26 GPU and ideal Kr/Xe selectivities >20 at 298 K after thickness reduction. Post-synthesis cation exchange shows large (>50%) increases in selectivity at ambient or slight sub-ambient conditions. In addition, we confirm that SAPO-34 membrane is stable under radiation exposure and the impact of radiation exposure on membrane performance would not be substantial. We also successfully synthesized hollow-fiber SAPO-34 membranes with the same performance levels as the disk-type and tubular membranes. This important development will allow a very compact and low-cost Kr/Xe separation system. Finally, a detailed process calculation for techno-economic analysis was performed by integrating Maxwell-Stefan model into cross-flow membrane system, in order to estimate the required number of membrane stages and the total cost.

Background of Program Objectives

Nuclear energy is an efficient and economical means of providing carbon-free energy, but requires effective management of radioactive waste products [1, 2]. As part of the overall process for recycling spent nuclear fuel (SNF), it is necessary to develop technologies of advanced and cost-effective SNF recycle and waste minimization for long-term energy security. Recycling of zircaloy (the second largest mass in SNF) has been investigated to maximize reuse and minimize waste volume requiring geological disposal. This involves an integrated process (e.g., Figure 1), including chemical zircaloy decladding, removal of fission products by voloxidation, and dissolution. separate the off-gases formed during the fission of and in nuclear power reactors [3]. Separating and capturing each gas is critical to reduce the volume of radioactive

Figure 1: Process diagram for off-gas treatment. Red arrows indicate solid and aqueous streams (uranium and plutonium), blue arrows are gas streams (process off-gas). Kr/Xe separation, which is the main focus of the proposed work, is highlighted at the bottom left. waste for storage. The long-lived radioisotope 85Kr and stable 136Xe are released as process off- gas and Kr/Xe separation is challenging because they are and physically very similar. Based on data from a US national laboratory, the Kr/Xe mixture is approximately 10/90 by volume [4] and hence the separation of Kr would reduce the volume of radioactive 85Kr waste by a factor of 10. One way of separating them is by cryogenic distillation [3, 5] wherein they are condensed out as pure streams at very low temperature (Kr : -153°C, Xe: -108°C).

However, cryogenic distillation has intensive power requirements and a large system volume, resulting in high capital, operating, and disposal costs. Membrane-based separation using radiation-stable materials is an attractive alternative [6, 7], if membranes with sufficient separation properties are successfully fabricated and scaled up. Chabazite (CHA) framework type zeolites have a nominal pore size of 0.38 nm, which is in between the kinetic diameters of Kr (0.36 nm) and Xe (0.396 nm), and it has been recently shown that silicoaluminophosphate CHA (also referred to as SAPO-34) membranes are good candidates for Kr/Xe separation [7, 8].

Main Objectives and Results

Our proposed work was divided into 5 main objectives. In this report we describe our accomplishments in each objective over the project performance

.

Objective 1: Understanding and control of intrinsic Kr and Xe adsorption/diffusion properties of

SAPO-34.

The main goals of this work were to synthesize SAPO-34 crystals of appropriate dimensions and obtain the adsorption and diffusion properties of Kr and Xe as a function of temperature and pressure from a well-defined set of SAPO-34 materials of tuned SixAlyPzO2 (x + y + z = 1) compositions and containing exchanged cations. Due to limited information on intrinsic Kr and Xe adsorption and diffusion properties which determines the membrane performance, we synthesized SAPO-34 crystals and obtained its adsorption and diffusion parameters using pressure-decay measurement over 10 – 400 kPa and at 308 – 343 K.

Figure 2. Single-gas Kr and Xe adsorption isotherms on (a) H-SAPO-34 at different temperatures and (b) H-SAPO-34 and -exchanged M-SAPO-34 materials. Symbols: experimental data, dashed lines: Langmuir fits.

SAPO-34 crystals were synthesized from a reactant molar composition of 1.0 Al2O3:1.0

P2O5:0.32 SiO2:1.0 TEAOH:1.6 CHA (cyclohexylamine):52 H2O with aluminum hydroxide

(Al(OH)3) as the Al source. The average crystal sizes and standard deviations are 10.7±2.3 µm

(SAPO-34-CHA) and the edge dimension of the cuboid crystals is used as the effective crystal diameter, since this achieves the required equivalence of surface area-to-volume ratio with a spherical crystal of the same diameter. The crystals were immersed in the ethanol solution of acetate salts of monovalent (Li+, Na+, and K+) cations for ion exchange.

Single-component gas adsorption isotherms of Kr and Xe were collected and fitted to

Langmuir model (Equation 1) by a global fit of the isotherms over all temperatures (Figure 2a),

푞 푏 푝 푞 = 푠푎푡 (1) 1 + 푏 푝 where q is the adsorbate concentration in the sample, qsat the saturation coverage, p the pressure and b the Langmuir constant which can be expressed in terms of the enthalpy of

o adsorption (ΔHads) and the pre-exponential entropic factor b = exp (ΔSads/R). The enthalpy and entropy of Xe adsorption are larger than those of Kr, leading to the stronger adsorption of Xe.

Table 1 also shows the M-S diffusion parameters obtained from the kinetic uptake data. The diffusion activation energy of Kr is much lower than that of Xe which is consistent with the of Kr and Xe in relation to the crystallographic pore size of SAPO-34.

Table 1. Langmuir Adsorption Parameters and Maxwell-Stefan Diffusion parameters for Kr and Xe in H- SAPO-34 0 qsat -ΔHads b0 D MS Ea Gas (mmol/g) (kJ/mol) (*10-6 kPa-1) (*10-10 m2/s) (kJ/mol) Kr 6.1 ± 0.2 16.1 ± 0.3 1.4 ± 0.3 2.5 ± 0.3 13.9 ± 1.0 Xe 4.2 ± 0.1 18.3 ± 0.4 3.8 ± 0.2 2.0 ± 0.1 23.6 ± 0.8

Adsorption isotherms on H-SAPO-34 and the three ion-exchanged M-SAPO-34 materials were measured at 298 K over 10-300 kPa (Figure 2b) which showed all the materials preferentially adsorb Xe over Kr. Table 2 summarizes the textural properties obtained by N2 physisorption at 77

K, and the results of fitting the Kr and Xe adsorption isotherms with the Langmuir model. The

BET surface area and micropore volume decrease monotonically (by 10 – 36 %) as the cation size increases (H+ < Li+ < Na+ < K+). The Langmuir constants (b) are not strongly influenced by ion exchange, indicating that the van der Waals interactions between the atoms and SAPO-

34 are not substantially affected by the cations. This is consistent with a previous report that lithium cation exchange does not notably change the heats of adsorption of CO2 and CH4 in SAPO-34 [9].

However, the Kr and Xe saturation capacities (qsat) decrease in the same order as the micropore volume, consistent with the decreased pore volume available as larger sizes and numbers of cations are introduced.

Table 2. Gas adsorption and textural properties of SAPO-34 materials as obtained from Kr and Xe adsorption (298 K) and physisorption (77 K) isotherms.

H-SAPO-34 Li-SAPO-34 Na-SAPO-34 K-SAPO-34

BET surface area [m2/g] 480 429 323 306

Micropore volume [cm3/g] 0.23 0.21 0.20 0.18

qsat Kr 6.7 6.1 5.6 5.2

[mmol/g] Xe 5.2 4.8 4.4 4.1

b Kr 9.0 8.3 8.2 8.2

[10-4 kPa-1] Xe 39 37 36 39

The bulk compositions of the as-synthesized H-SAPO-34 and cation-exchanged M-SAPO-

34 (M = Li, Na, K) materials were determined using ICP-MS (H-SAPO-34 and Li-SAPO-34) and

EDX analysis (H-SAPO-34, Na-SAPO-34, and K-SAPO-34) of the Si, P, and Al content. H-

SAPO-34 was used as a reference to investigate the consistency between ICP-MS and EDX analysis. Based upon the measured atomic content of these elements, the H stoichiometry was calculated using a charge balance based upon the crystallographic SAPO-34 structure [9] with a fixed/known O stoichiometry. The composition (expressed on a unit-cell basis) of the H-SAPO-

34 membranes was estimated as H1.5Si4.3Al16.6P15.1O72 (ICP-MS) and H1.7Si4.5Al16.6P14.9O72 (EDX).

The equilibrium degree of alkali cation exchange is notably different for the three cations (Li+,

Na+, and K+). After Li+ ion exchange, only 61% of the protons were exchanged to yield a

+ composition of H0.7Li1.1Si3.6Al17.1P15.3O72, whereas the level of Na ion exchange was 82% + (H0.3Na1.3Si4.3Al16.7P15.1O72) and that of K was 92% (H0.1K1.5Si3.8Al16.9P15.3O72). These results are found to be consistent with a previous computational study [10] that suggested a cation ion exchange selectivity order of K+ > Na+ > Li+. We verified that the amount of cations exchanged did not increase after repeated ion exchange, which creates confidence that the measured compositions represent the equilibrium/maximum extent of ion exchange for each cation.

Since membrane separation occurs under mixture conditions, we also used the Langmuir- fitted single-component isotherms to predict the binary adsorption behavior via ideal adsorbed solution theory (IAST) [11]. The IAST equilibrium relationship for solid-fluid equilibrium is analogous to Raoult’s law for vapor-liquid equilibrium. We hypothesized essentially ideal mixing in the adsorbed , which is an excellent assumption for noble gas mixtures in non-cationic zeolites. The adsorbed amounts of gas as well as total adsorbed amounts were calculated at different conditions of Kr-Xe mixture composition, total pressure, and temperature. A computer code was written to generate these predictions, according to the details of practical application of

IAST as described in detail elsewhere [11, 12]. The double Langmuir approximation was also used

Figure 3. IAST predictions of (a) Xe and Kr adsorption on SAPO-34 from a 10/90 Kr/Xe binary feed mixture; and (b) the corresponding Xe/Kr adsorption selectivity; at 255 and 298 K. (c) Comparison between double Langmuir approximation and IAST predictions at 298 K. for comparison to the IAST predictions. Figure 3 shows the IAST predictions of binary adsorptions from 10/90 Kr/Xe feed gas mixture at two different temperatures for pressures up to

500 kPa. The IAST predicted binary adsorption selectivity for Xe is higher than that predicted by the weighting of single-component isotherms. Thus, Xe is preferentially adsorbed in a binary mixture and blocks the adsorption site of Kr. The IAST predictions were then used for comparison with double Langmuir model. The double Langmuir model is a very convenient model for use in

Maxwell-Stefan modelling of membrane transport due to its simple mathematical form as well as the ease of deriving analytical expressions for the thermodynamic correction factors. As shown in

Figure 3c, the double Langmuir isotherm matches well with the IAST predictions over the entire pressure range.

Objective 2: Control of membrane thickness and defects in tubular geometry.

High manufacturing cost is a key factor limiting the commercial applications of zeolite membranes [13]. The most straightforward way to reduce the membrane cost is to decrease the required membrane area, which in turn can be accomplished by increasing the flux (throughput) via a reduction in the membrane thickness. However, thin (~ 1 m) zeolite membranes are challenging to fabricate since the control of defects becomes more difficult when growing thin zeolite layers on low-cost supports. The goal of this work is to develop approaches for SAPO-34 membrane thickness reduction (i.e., Kr permeance increase) without reduction of Kr selectivity.

To study the Kr and Xe permeation properties of SAPO-34 membranes, our first goal was to synthesize good-quality SAPO-34 membranes on low-cost α-alumina supports of both disk

(obtained from Coorstek) and tubular (obtained from Ceramco, Inc.) types. Most previous work on SAPO-34 membranes for CO2/CH4 separation have used secondary hydrothermal growth process with conventional seeding method, such as dip coating or rub coating [14]. However, our

choice of low-cost (and lower-quality) supports requires the use of different seeding processes in

order to promote the growth of a continuous membranes. Our initial efforts using dip-rub seeding

method consistently led to membranes with high permeances and low (< 5) Kr/Xe selectivities.

Based upon detailed SEM examination of the seeded substrates and the resulting membranes, as

well as analysis of the permeation data, it was determined that the above seeding methods were

unable to provide sufficient coverage of seeds on the support surfaces because of the

inhomogeneous and low-porosity character of the support surface. Therefore, a different seeding

approach was desired. We decided to adapt the steam-assisted conversion (SAC) method reported

by Zhou et al [15]. In this method, a seed layer was nucleated on the entire support surfaces by a

two-step process: first, a paste/gel containing SAPO-34 crystal seeds as well as unreacted Si/Al/P

precursors is coated on the support; and second, the paste was transformed into a continuous seed

layer by crystallization in the presence of steam. The SAC process was then carried out at 473 K

for 24 h to convert the paste into a continuous seed layer (Figure 4b). SAPO-34 membranes were

then grown from the SAC layer using the same procedures used earlier for dip/rub-coated supports.

Figure 4c shows the formation of a well-intergrown polycrystalline SAPO-34 membrane.

Figure 4. Top view SEM images of (a) bare tubular α-alumina substrate, (b) SAC seed layer, and (c) final SAPO-34 membrane. Figure 5. Example surface and cross-sectional SEM images of M1 (a, c) and M2 (b, d) membranes, and (e,f) the corresponding cross-sectional P/Al ratio profiles of each membrane as determined by EDX analysis. The yellow lines define the thickness of the membrane.

To achieve this objective, reduced amounts of seed-containing paste for SAC conversion was used to fabricate thinner SAPO-34 membranes. Thin SAPO-34 membranes were synthesized on disk type α-alumina support (Coorstek) and the thickness of membrane was determined with

P/Al ratio analyzed by EDS line scanning. Figure 5 shows the SEM micrographs and EDS analysis of the cross-section of SAPO-34 membrane for determination of the location of membrane layer. Among four elements present in membrane layer or support layer (Si, Al, P, and O), P element profiles are localized to the membrane layer, thus ratio of P/Al intensity was used for measuring thickness more accurately. When using 0.05 mL SAC paste/cm2 support, the average thickness of resulting membranes (M1) was found to be 5.6 ± 0.5 μm (Figure 5c, 5e). The thickness of membranes fabricated with reduced (0.02 mL SAC paste/cm2) amount of SAC gel

(M2) were twice to thrice thinner than that reported in M1, which were determined as 2.1 ± 0.5

μm (Figure 5d, 5f). 14 membranes for total were synthesized as “thinner” SAPO-34 and the reproducibility was confirmed by repeated synthesis in separate batches. Top-view and cross- sectional SEM images of membranes verified that 2.4 μm thick membranes are continuously grown and are free of visible defect.

Figure 6. Single gas Kr permeances and ideal Kr/Xe selectivities of M1 (average thickness = 5.6 µm) and M2 (average thickness = 2.1 µm) membranes at 298 K. Each membrane sample is represented by the small symbols, and the large symbols are the averaged values.

Figure 6 shows the single-gas Kr and Xe permeances of a larger number of SAC-seeded

SAPO-34 membranes at 298 K and 140 kPa feed pressure. The average Kr permeance of M1 membranes was obtained as 7.5 ± 0.8 GPU with Kr/Xe ideal selectivity of 16±2. After thickness reduction (M2 membranes), the permeances systematically increased. The Kr permeance increased greatly to 26.3 ± 8.3 GPU, with ideal selectivity 23±3. While the increase in permeance is a direct result of thickness reduction, the accompanying increase in Kr ideal selectivity is also significant and interesting. This can be explained by the size of crystals comprising the membrane. M1 membranes have crystal sizes of ~5 µm on the membrane surface whereas ~ 1.5 µm crystals are packed in M2 membranes (Figure 5a-5b). Small crystal grains pack better than larger crystals which results in improvement in selectivity.

Objective 3: Impact of radiation exposure on SAPO-34 membrane performance.

An important aspect of this work is to perform targeted experiments that shows the impact of radiation exposure due to radioactive off-gas streams on membrane performance. In practice, the membranes would be exposed to 85Kr, which is a lower-energy gamma emitter. The total activity of 85Kr per unit membrane mass is limited by the gas-phase concentration and the relatively small amount of 85Kr present in the membrane due to low concentration of 85Kr in feed mixture.

However, it is necessary to evaluate the radiation stability since the zeolite membrane will receive a high gamma radiation dose over months/years of operation, which may result in performance change caused by radiation-induced crystal defects or change. We hypothesized that silicoaluminophosphate zeolite would have good stability under radioactive Kr/Xe since some zeolites are known to function effectively as adsorbents in other off-gas separations [16-18].

To obtain a reliable characterization of membrane performance under long-term radiation exposure, radiolytic treatment was performed using a 60Co beta-gamma irradiator at Oak Ridge National Laboratory (ORNL). The Co-60 irradiator employed for this test was JL Shepherd &

Associates Model 109-68. It has a cavity with the dimensions of 6 inches in diameter and 8 inches tall, which generates an unshielded exposure rate at 12,000 R hr-1. Membranes were prepared with

Kr permeance of 10.4 ± 1.7 GPU and ideal selectivity of 15.5 ± 1.5 for radiolytic treatment. Since the inside of irradiator is under the same condition as ambient air, two control samples were prepared to investigate the effect of humidity (~ 40 % RH) under non-radioactive ambient condition.

Table 3 shows the effect of radiation exposure on SAPO-34 membranes by the comparison of membrane performance. After 30 – 60 days of exposure, the permeances of both Kr and Xe increased but the ideal selectivity dropped by 34 – 37 %. Reduction in ideal selectivity can be explained by the defect formation over time. In order to investigate the effect of humidity and other gases existing in air, the performance change in control sample was characterized. The ideal selectivity of control sample reduced by 33 %, which means that the degradation after radiolytic treatment was mainly due to water vapor in the air. It is reported that water vapor causes slow degradation of SAPO-34 due to breakage of Si-O-Al and Al-O-P bonds [9]. The continuous off- gas stream in practice, however, will be under dry condition without water vapor or acid gas.

Considering the dosing amount of exposure to 0.1 atm of 85Kr, 30 days of exposure to accelerate

60Co gamma irradiation simulates about one year exposure to the off-gas containing 85Kr.

Therefore, we were able to confirm the radiation stability of SAPO-34 membrane via the simulated radiation exposure for years of operation.

Table 3. Effects of radiolytic treatment of SAPO-34 membranes using Co-60 irradiator

Before radiolytic treatment After radiolytic treatment Permeance Ideal Permeance Ideal Selectivity Exposure [GPU] Selectivity [GPU] Selectivity dropped time [days] Kr Xe (Kr/Xe) Kr Xe (Kr/Xe) by [%] R1 30 11.2 0.6 17.8 12.9 1.1 11.7 34.2 R2 60 13.1 0.8 16.1 15.4 1.5 10.2 36.6 Control – ambient 9.50 0.7 14.6 10.7 1.1 9.8 32.8 for 60 days 10.4 0.67 15.5 Average ± 1.7 ± 0.1 ± 1.5

Objective 4: Detailed membrane transport measurements and hollow fiber SAPO-34 synthesis

In this study, we performed a detailed evaluation of Kr and Xe single-gas and binary mixture permeation properties of ion-exchanges SAPO-34 membrane. The main aim of this objective was to conduct detailed binary transport measurements with SAPO-34 membranes, to comprehensively understand their behavior, validate the fundamental-data based models.

Single-component permeation measurements were conducted using in-house built permeation units operating in dead-end mode [19]. Kr and Xe permeation measurements were carried out from 254 K to 298 K. For sub-ambient temperature (254 K), the membrane module was submerged in a salt water-ice cooling bath. Binary mixture permeation was measured in

Wicke-Kallenbach mode at a system pressure of 110 kPa. The feed was 10 mol% Kr/90 mol% Xe mixture, generated by two mass flow controllers. A sweep (10 mL/min) was used to collect the permeate, whose composition was then analyzed by a gas chromatograph (Shimadzu GC-TCD- 2014) equipped with a HP-PLOT 5A column (Agilent). The permeance of each component is obtained as its flux normalized by its transmembrane partial pressure. The Kr/Xe selectivity of the membrane is given as the ratio of permeances of each gas. For all reported permeation measurements, the permeances and selectivities are average values with standard deviations recorded from more than three membrane samples synthesized independently.

Figure 7a shows single-gas Kr permeance and Kr/Xe ideal selectivity data from the ion- exchanged M2 membranes at 298 K and 140 kPa feed pressure. In Figure 7b we also plot the Kr permeabilities against the ideal selectivities to gain better insight on the effects of ion exchange.

The Kr permeability of each membrane sample is obtained by dividing the Kr permeance by the apparent thickness of that membrane, the latter being determined from multiple SEM-based measurements along the cross-section. In Figures 7a-7b, the single-component Kr permeances and permeabilities of all the three types of ion-exchanged membranes decrease relative to the as- synthesized M2 (H-SAPO-34) membranes, but there are significant differences between the ion- exchanged membranes. At a feed pressure of 140 kPa, Figure 2b shows that the differences in Kr adsorption are not large (~25%) and cannot by themselves explain the reduction in permeability relative to the H-SAPO-34 membranes. Therefore, it is attributed mainly to the reduction in diffusivity as a result of a lower effective pore size after alkali cation exchange. However, the Xe adsorption isotherms show larger differences and hence may play a more significant role along with diffusivity decreases. To separate these effects formally, we use the Maxwell-Stefan equation for the single-component permeability (P) of a nanoporous membrane [20]:

표 −퐸푎 1+푏푝푓 푃 = 휌푞푠푎푡퐷푀푆 exp ( ) ln ( ) (2) 푅푇 1+푏푝푝 Figure 7. (a) Single-component Kr permeances versus selectivities, and (b) Single-component Kr permeabilities versus selectivities, for M1, M2, and ion-exchanged M2 membranes at 298 K and 140 kPa feed pressure. Each membrane sample is represented by small symbols and corresponding large symbols are the averaged values for each type of membrane. (c) Percent changes of permeabilities and Kr/Xe selectivities as a function of cation radius.

3 Here, pf and pp are the pressures of the feed and permeate, ρ is the (1800 kg/m ) of the

SAPO-34 zeolite, and qsat and b values are adsorption parameters given in Table 2. Using the experimental permeability data (Figure 7b) and the Langmuir adsorption parameters (Table 2), we obtain the effective M-S diffusivity of Kr and Xe for each membrane sample, and then average these to obtain statistically valid M-S diffusivities for each of the four types of M2 membranes (as- synthesized and ion-exchanged). These values are shown in Table 4. It is seen that the M-S diffusivities of Kr and Xe decrease monotonically as the size of the cation increases. This is attributed to the reduction in effective pore size which hinders the passage of molecules through the 8MR windows [21]. M-S diffusivities of both Kr and Xe decrease after ion exchange, but the effect of cation exchange on diffusivities is more significant for Xe (50-58% reduction) than Kr

(23-26%), thereby leading to a large increase in diffusion selectivity. Combined with the adsorption behavior, we also obtain large increases in permeation selectivity as well. The ideal

(single-component) permeation selectivity of cation-exchanged membranes increases by 40-63% relative to H-SAPO-34 membranes. Although the permeances are reduced after ion exchange, the

resulting values are still considerably higher than the baseline M1 membranes due to the effect of

thickness reduction in M2 membranes.

Table 4. M-S diffusivities and Kr diffusion selectivities in SAPO-34 membranes.

-14 2 -14 2 DKr [×10 m /s] DXe [×10 m /s] Diffusion Selectivity M2 172 ± 12 2.78 ± 0.5 65.2 ± 8.7 Li-M2 133 ± 13 1.51 ± 0.1 94.3 ± 6.2 Na-M2 130 ± 12 1.26 ± 0.2 107 ± 10.3 K-M2 128 ± 12 1.04 ± 0.1 122 ± 8.9

Among the ion-exchanged membranes in Figure 7, K-SAPO-34 appears to be the best

candidate due to its high ideal selectivity (~37) and only modestly reduced Kr permeance (18.8 ±

0.9 GPU). The separation properties of K-M2 and M2 membranes for a binary 10/90 (molar ratio)

Kr/Xe mixture were characterized as a function of temperature at a feed pressure of 100 kPa

Figure 8. Temperature dependence of binary Kr and Xe permeances and Kr/Xe separation factors in disk H-SAPO-34 (M2) and ion- K-SAPO-34 (K-M2) membranes. (Figure 8). The binary Kr permeances increase with decreasing temperature whereas Xe permeances decrease. As a result, the binary Kr/Xe separation factor increases with decreasing temperature. Clear molecular sieving effects are seen. The Kr/Xe separation factor in K-SAPO-34 membranes is 48±3 at 253 K and 30±5 at 298 K.

Hollow fibers have a small diameter less than 1 mm, which allow high packing volume as high as 1000 m2 area/ m3. The other advantage of using hollow fiber substrates is that they can be produced by spinning process which is much simpler and cheaper than tubular substrate fabrication. With home-made α-alumina substrates, the SAC and hydrothermal growth methods for M2 membranes were successfully transferred to α-alumina hollow fiber supports of 800 µm

OD. Figure 9 shows representative SEM images of H-SAPO-34 membrane fabricated on the outer surface of a hollow fiber. The average crystal grain size is ~3 µm and the apparent (SEM) membrane thickness is determined as 3.2±0.4 µm. The average single-gas Kr permeance is 19.4 ±

1.1 GPU with ideal selectivity of 31.7 ± 13.3, which is comparable to the permeation performance of K-M2 disk-type membranes. We then investigated the separation of a binary Kr/Xe mixture with hollow fiber SAPO-34 membranes before and after ion exchange with potassium cations.

Figure 10 shows the temperature dependence of Kr/Xe binary mixture permeation. The binary mixture separation showed improved selectivities after ion exchange at all temperatures with clear molecular sieving effects. Comparison with Figure 8 shows similar separation characteristics of the hollow fiber membranes and the disk-type membranes.

Figure 9. (a) Top view and (b), (c) cross-sectional SEM images of a SAPO-34 membrane fabricated by SAC seeding, hydrothermal secondary growth, and ion exchange on an -alumina hollow fiber.

Figure 10. Temperature dependence of binary Kr and Xe permeances and Kr/Xe separation factors in hollow fiber H-SAPO34 and ion-exchanged K-SAPO-34 hollow fiber membranes.

Objective 5: Membrane process system calculations

To investigate the permeation characteristics further, we modeled the binary Kr/Xe mixture permeation using Maxwell-Stefan equations for two components [20]. M-S formulation generalizes the relationship between the flux and the gradients of fractional occupancies:

−1 (푁) = −휌[푞푠푎푡[퐵] [Γ](∇휃) (3)

1 휃2 휃푖 퐵푖푖 = + ; 퐵푖푗 = − 퐷푖 퐷푖푗 퐷푖푗 where i and j refer to the two diffusing species (Kr and Xe). The diagonal components (Di) of the diffusivity matrix correspond to the single-component diffusivities of the two species. The off- diagonal terms (Dij) correspond to the exchange of species i and j in the pores. In the case of zeolites such as CHA and DDR which have narrow pore windows connected by larger cavities, it is expected that exchange of species can easily take place at low or moderate loadings [22]. As a result, the off-diagonal terms of the B matrix can be neglected. The binary adsorption equilibrium can be estimated using the double Langmuir isotherm, which has a convenient mathematical form.

Figure 11 shows the temperature dependence of permeation properties of M1 with a binary

Figure 11. Temperature dependence of Kr and Xe mixture permeances and Kr/Xe mixture selectivity for tubular SAPO-34 membranes. The feed mixture is at 1 atm with 10 vol % Kr/90 vol % Xe. mixture of 10 % Kr and 90 % Xe with M-S predictions. The binary Kr permeance decreases slightly with increasing temperature whereas the Xe permeance is small and increases with temperature. A Kr/Xe selectivity of 30 could be obtained at the lowest measured temperature (255

K) and 12 at the highest measured temperature (343 K). Overall, the model predictions are in good agreement with the experimental data and they correctly predict the temperature dependence of permeances of both Kr and Xe without any fitting to the experimental permeation data.

We integrated M-S equations into a cross-flow model, which allows the permeance and selectivity to change over the stage length as the composition of feed stream is changing from one end of the stage to the other. A tubular membrane design is considered here and a schematic and required specification is illustrated in Figure 12. A multi-stage membrane unit is shown and inter-

Figure 12. Schematic of membrane stage model for a cross-flow membrane system. stage pump and compressor is used to maintain the pressure of feed at 100 kPa and permeate pressure at 100 Pa.

Using the adsorption and diffusion properties of each membrane (M1, M2, Li-M2, Na-M2, and

K-M2), the process modeling simulations were carried out. A plug flow model is used on the retentate and the permeate sides, thus the governing mass balance equations for each components are as follows:

푛푖(푧) − 푛푖(z + 푑푧) + 2휋푟퐽푖푑z = 0 (3) where ni is molar flow rate of component i along the stage length, Ji is flux of component i passing through the membrane. The feed flow rate is 1 L/min with 10 % Kr and 90 % Xe at 1atm pressure.

Retentate stream of each stage has 99.9 % purity and the permeate stream from the final stage requires > 90 % Kr purity.

Figure 13. Numbers of stages and membrane area for desired purity for each type of membrane.

The thickness of M1 membrane was assumed as 5.6 µm, and all types of M2 membranes as 2.1 µm, which were determined based on our EDX analysis. The results are summarized in

Figure 13. Large membrane area (~ 15 m2) with 5 stages is required for M1 in order to achieve desired product purity. After thickness reduction (M2), the membrane system requires 4 stages with smaller system area (~ 5 m2) due to increased permeance and selectivity. Post-symthesis ion exchange (M-M2) leads to reduced number of stages because of huge improvement in selectivity.

Even though the required area is increased to ~ 8 m2, it is desirable to decrease the number of stages since the system cost caused by energy cost can hugely be reduced.

Overall Accomplishments

• Adsorption and diffusion measurements on SAPO-34 crystals indicated their high potential for

use in Kr-Xe separation membranes.

• Thinner SAPO-34 zeolite membranes (~2 µm) were successfully synthesized using thin

precursor gels to control membrane thickness.

• Post-synthesis cation exchange resulted in excellent selectivity (30-50) with only modest (30-

40%) permeance drop: higher permeances than base case membranes were obtained.

• The impact of radiation exposure on membrane performance was not significant.

• Hollow-fiber SAPO-34 membranes were successfully fabricated with Kr permeance ~25 GPU

and selectivities of 30-50: this important development will result in a very compact, low cost

system.

• Process modeling showed K-SAPO-34 membranes will require only 3 stages with ~8 m2 total

area to handle a 1 L/min Kr/Xe feed stream.

List of Publications

1. Y. H. Kwon, B. H. Min, S. Yang, D.-Y. Koh, R. R. Bhave, S. Nair, "Ion-Exchanged SAPO-34 membranes for Krypton-Xenon Separation: Control of Permeation Properties and Fabrication of

Hollow Fiber Membranes", ACS Applied Materials & Interfaces, 10, 7, pp. 6361-6368 (2018). 2. Y. H. Kwon, C. Kiang, E. Benjamin, P. Crawford, R. Bhave, S. Nair, "Krypton-Xenon

Separation Properties of SAPO-34 Zeolite Materials and Membranes", AIChE Journal, 63, pp.

761-769 (2017).

3. Y. H. Kwon, B. H. Min, C. Kiang, R. R. Bhave, “Chabazite Zeolite Membranes for

Krypton/Xenon Separation: Control of Membrane Properties and Process Modeling”, ANS

Transactions, 116 (1), pp. 126-129 (2017).

4. Y. H. Kwon, E. Benjamin, V. Pisharodi, J. Hwang, R. R. Bhave, S. Nair, “Adsorption and

Transport Properties of Zeolite SAPO-34 for Krypton/Xenon Separations”, ANS Transactions,

114(1), pp. 215-218 (2016).

References

[1] Strategy for the Management and Disposal of Used Nuclear Fuel and High-Level Radioactive

Waste. US Department of Energy. Washington, DC2013.

[2] Marques JG. Evolution of reactors: Third generation and beyond. Energy

Conversion and Management 2010;51:1774-80.

[3] Global Nuclear Energy Partnership Conceptual Design Studies: Summary (Non-Proprietary)

GE Hitachi Nuclear Energy 2008.

[4] Jubin RT, DelCul GD, Patton BD, Owens RS, Ramey DW, Spencer BB. Advanced Fuel Cycle

Initiative Coupled End-to-End Research, Development, and Demonstration Project: Integrated

Off-Gas Treatment System Design and Initial Performance-9226. Waste Management

Conference 2009. Phoenix, AZ2009.

[5] Goloubev D. Method and apparatus for the cryogenic separation of air. Unites States: Goloubev

Dimitri; 2016. [6] Stern SA, Leone SM. Separation of Krypton and Xenon by Selective Permeation. AIChE

journal 1980;26:881-90.

[7] Kwon YH, Kiang C, Benjamin E, Crawford P, Nair S, Bhave R. Krypton-xenon separation

properties of SAPO-34 zeolite materials and membranes. AIChE Journal 2017;63:761-9.

[8] Feng X, Zong Z, Elsaidi SK, Jasinski JB, Krishna R, Thallapally PK, et al. Kr/Xe Separation

over a Chabazite Zeolite Membrane. Journal of the American Chemical Society

2016;138:9791-4.

[9] Hong M, Li S, Funke HF, Falconer JL, Noble RD. Ion-exchanged SAPO-34 membranes for

light gas separations. Microporous and Mesoporous Materials 2007;106:140-6.

[10] Civalleri B, Ferrari AM, Llunell M, Orlando R, Me´rawa M, Ugliengo P. Cation Selectivity

in Alkali-Exchanged Chabazite: An ab Initio Periodic Study. of Materials

2003;15:3996-4004.

[11] Murthi M, Snurr RQ. Effects of Molecular Siting and Adsorbent Heterogeneity on the Ideality

of Adsorption Equilibria. Langmuir 2004;20:2489-97.

[12] Nugent P, Belmabkhout Y, Burd SD, Cairns AJ, Luebke R, Forrest K, et al. Porous materials

with optimal adsorption thermodynamics and kinetics for CO2 separation. Nature

2013;495:80-4.

[13] Li M, Zhang J, Liu X, Wang Y, Liu C, Hu D, et al. Synthesis of high performance SAPO-34

zeolite membrane by a novel two-step hydrothermal synthesis + dry gel conversion method.

Microporous and Mesoporous Materials 2016;225:261-71.

[14] Poshusta JC, Tuan VA, Falconer JL, Noble RD. Synthesis and Permeation Properties of

SAPO-34 Tubular Membranes. Industrial & Engineering Chemistry Research 1998;37:3924-

9. [15] Zhou L, Yang J, Li G, Wang J, Zhang Y, Lu J, et al. Highly H2 permeable SAPO-34

membranes by steam-assisted conversion seeding. International Journal of Energy

2014;39:14949-54.

[16] Pillay KKS. A Review Of The Radiation Stability Of Ion Exchange Materials Journal of

Radioanalytical and Nuclear Chemistry 1986;102:247-68.

[17] Daniels EA, Puri M. Physico-Chemical Investigations Of Gamma-Irradiated Zeolite-4a

International Journal of Radiation Applications and Instrumentation Part C Radiation Physics

and Chemistry 1986;27:225-7.

[18] Wang LM, Chen J, Ewing RC. Radiation and thermal effects on porous and layer structured

materials as getters of radionuclides. Current Opinion in Solid State and Materials Science

2004;8:405-18.

[19] Crawford PG. Zeolite membranes for the separation of krypton and xenon from spent nuclear

fuel reprocessing off-gas. SMARTech Georgia Tech Theses and Dissertations: Georgia Institue

of Technology; 2013.

[20] Krishna R, Broeke LJPvd. The Maxwell-Stefan description of mass transport across zeolite

membranes. The Chemical Engineering Journal and the Biochemical Engineering Journal

1995;57:155-62.

[21] Chew TL, Ahmad AL, Bhatia S. Ba-SAPO-34 membrane synthesized from microwave

heating and its performance for CO2/CH4 gas separation. Chemical Engineering Journal

2011;171:1053-9.

[22] Krishna R, Li S, van Baten JM, Falconer JL, Noble RD. Investigation of slowing-down and

speeding-up effects in binary mixture permeation across SAPO-34 and MFI membranes.

Separation and Purification Technology 2008;60:230-6.