1 Introduction

Ozone (O3) is a reactive triatomic allotrope of oxygen that is well known for shielding the Earth from the dangerous solar ultraviolet rays and its high oxidation potential, which is second only to fluorine. Recently, industrial ozone generation has witnessed an increasing need in various applications including wastewater and drinking water treatment, swimming pool water and aquarium water disinfection, bleaching of fabrics and wood pulp, food processing, flue gas denitrification as well as indoor air cleaning and purification [1, 2]. Since ozone is an unstable compound that decomposes to oxygen under ambient conditions, it cannot be transported or stored and must be produced at the point-of-use and at the time it is needed.

Ozone can be generated through electrolysis, ultraviolet (UV) irradiation or dielectric barrier discharge. The latter method is the most effective and widely used to produce ozone industrially. Although advanced discharge-based reactors have been proposed, such as coplanar discharge reactors [3] and packed bed reactors [1], corona discharge reactors have so far represented the mainstream technology [4, 5]. In a corona discharge reactor, a current flows from an electrode with a high potential into a neutral fluid, usually air or oxygen, by ionizing that fluid so as to create a region of plasma around the electrode. The ions generated eventually pass the charge to nearby areas of lower potential, or recombine to form neutral gas molecules such as ozone [4]. It is well established that the efficiency of the ozone generation depends upon the oxygen concentration of the gas subjected to the corona discharge and rises with increasing oxygen concentration. When pure oxygen is used as the feed for the ozone generator the concentration of ozone product ranges between 6 and 15% by weight (balanced with oxygen). Air may also be used as the source of oxygen resulting in an ozone concentration between 2 and 6% by weight (balanced with air) [6]. While air-based ozone generators are less efficient due to the formation of NOX that destruct ozone, a disadvantage of using high-purity oxygen is the cost associated with the separation of oxygen from air. However, for medium-to-large production rates of ozone, the total capital plus operating costs as well as the footprint are lower when pure oxygen is used as the feed [6, 7]. Some works in the literature have also assessed the benefits of having small quantities of nitrogen in the oxygen feed in order to maximize the ozone yield but an agreement on the optimal feed composition has not been reached yet [2, 5].

Given that most industrial-scale ozone generators are fed with expensive high-purity oxygen from liquid oxygen systems (LOX) or pressure swing adsorption systems (PSA), it is essential to recover the oxygen from the outlet ozone–oxygen mixture for recycling to the ozone generator in order to substantially improve the process economics. Cyclic adsorption processes have been proposed to effectively carry out this separation due to their unique hazard-free operation when handling reactive ozone [8,9,10,11]. In an adsorption process ozone is selectively adsorbed at high pressure from the ozone generator outlet stream, and the recovered ozone-depleted oxygen is compressed in a blower and recycled to the ozone generator. A portion of the recovered oxygen could also be vented to adjust the oxygen composition in the ozone generator feed. The adsorbed ozone is then desorbed by a sweep gas such as air or nitrogen, and the mixture of ozone and sweep gas is provided as a product to the downstream user. When air is used, it must be compressed and dried prior to desorbing ozone. An ozone destruct unit that converts all venting ozone to oxygen is also required for process safety. A complete block flow diagram of an ozone generator integrated with an adsorption process to recover oxygen is shown in Fig. 1.

Fig. 1
figure 1

Block flow diagram of an ozone generator integrated with an adsorption process to recover oxygen

Full size image

Some early works in the literature studied the ozone–oxygen separation by adsorption on refrigerated silica gel in the temperature range between – 90 and − 10 °C. Cook et al. [8] investigated experimentally the removal of ozone from oxygen and its transfer to dry air using a 3-bed, 4-step vacuum pressure swing adsorption (VPSA) cycle including an adsorption step, an oxygen evacuation step, a desorption step and an air evacuation step. Starting with a stream of oxygen containing 2 wt% of ozone the authors claimed that the overall recovery of ozone with the silica gel beds operating at − 10 °C was about 90%. Grenier and Petit [12] patented a 3-bed, 3-step adsorption cycle in which ozone was adsorbed on a silica gel bed at − 90 °C and desorbed by a stream of dry air before the bed was purged from the nitrogen by a portion of the ozone–oxygen feed gas. The effluent of this step was then combined with the dry air. Koike et al. [13] designed and manufactured an adsorption process using refrigerated silica gel beds operated in temperature-swing mode. The separation of ozone from oxygen was carried out in a 3-bed, 4-step cycle incorporating an adsorption step at − 60 °C, a heating step, a desorption step at − 30 °C and a cooling step. The results indicated that the ozone generator coupled with the adsorption process could supply concentrated ozone up to 30 mol% at a constant flowrate.

The major drawback of using silica gel as an adsorbent is that it exhibits a low adsorption capacity for ozone at ambient temperature, thus requiring the adsorption process to be carried out at temperatures close to the cryogenic range. In turn, producing the refrigeration necessary for operating at these low temperatures is expensive. Moreover, silica gel catalyses the decomposition of ozone to oxygen, which reduces the overall ozone recovery and increases the cost of ozone generation [8, 14]. To overcome these issues, zeolitic adsorbents have also been investigated for ozone adsorption, including ZSM-5, modernite, Y zeolite and L zeolite, with an appropriate amount of pre-adsorbed water that is not reactive with ozone. It was shown that these zeolitic adsorbents can provide approximately the same ozone adsorption capacity at ambient temperature as silica gel operating at about − 60 °C, along with lower decomposition rates of ozone [14]. Among the patents using zeolitic adsorbents at ambient temperature, Weist Jr. [15] proposed a 2-bed, 4-step VPSA cycle to recover ozone from an ozone–oxygen mixture. The cycle steps were adsorption, evacuation, desorption with sweep gas and feed pressurization. The same author [10] also patented an improved PSA cycle that was designed to reduce the variation in ozone concentration and produce a higher average ozone concentration in the product gas. The configuration was a 3-bed, 4-step cycle including an adsorption step, a product rinse step, a desorption step with dry air and a feed rinse step. Fitch and Huertas [7] disclosed a 4-bed, 4-step adsorption cycle to recover oxygen from an ozone–oxygen mixture produced in an ozone generator. After the adsorption step, a sweep gas was passed through the bed to desorb the ozone with an optional vent after which the cycle was terminated with a purge step by the feed gas.

The previous patents proposed several ideas to improve the ozone separation from an ozone–oxygen mixture but the potential benefits were discussed only qualitatively, without providing any separation performance indicators. Thus, this paper aims for the first time to design, simulate and assess in detail cyclic adsorption processes for recovering oxygen from ozone generators at ambient temperature. To this end, various cycle configurations and operating conditions were investigated to process a 10:90 wt% ozone–oxygen stream produced by a commercial ozone generator with a capacity of 30 kg h−1 of ozone [11, 16]. The adsorbent selected for the present study was L zeolite, which so far has proven to be the best zeolitic adsorbent for this kind of separation [15].

2 Adsorption process model

A rigorous mathematical model was used to accurately simulate the cyclic adsorption processes investigated in this study. The model couples mass, energy and momentum balances over a packed bed with appropriate boundary conditions for each cycle step [17, 18]. The main model equations are reported in Table 1 while the boundary conditions associated with the cycle steps investigated in this study are reported in Table 2. A detailed explanation of the equations along with the underlying model assumptions are reported in previous works [19,20,21].

Table 1 Main model equations for the adsorption processes
Full size table
Table 2 Boundary conditions of the cycle steps of the adsorption processes
Full size table

Beyond refrigerated silica gel, the only ozone isotherms available in the literature are on H-modernite, H-ZSM-5 and L zeolite [14, 15]. While the ozone adsorption capacities are comparable on these zeolitic adsorbents, ammonium-exchanged L zeolite exhibits lower oxygen and nitrogen capacities, which represents a clear advantage in the adsorption cycles. In the patent by Weist Jr. [15], experimental isotherms of ozone in terms of ozone partial pressure in oxygen on L zeolite were measured at − 10 °C, 10 °C and 30 °C while experimental isotherms of oxygen and nitrogen on L zeolite were measured at 23 °C with pure gas. The respective isotherms are shown in Figs. 2 and 3.

Fig. 2
figure 2

Experimental isotherms of ozone in terms of ozone partial pressure in oxygen on L zeolite at − 10 °C, 10 °C and 30 °C [15]. Solid lines represent the Langmuir model fits

Full size image
Fig. 3
figure 3

Experimental isotherms of oxygen and nitrogen on L zeolite at 23 °C [15]. Solid lines represent the Langmuir model fits

Full size image

Based on the above isotherms, pure component equilibrium adsorbed amounts were fitted according to the Langmuir model using Origin software [22]. Multicomponent adsorption equilibria were then predicted by the extended Langmuir model as follows:

$$q_{i}^{*} = \frac{{q_{s,i} \cdot b_{i} \cdot P_{i} }}{{1 + \sum\limits_{j = 1}^{n} {b_{j} \cdot P_{j} } }}$$
(1)

where

$$b_{i} = b{}_{i,0} \cdot \exp \left( {\frac{{ - \Delta H_{i} }}{R \cdot T}} \right)$$
(2)

As oxygen and nitrogen adsorption isotherms were only given at one temperature (Fig. 3), heats of adsorption for these components were evaluated as those of reference zeolites used for air separation [23].

Due to the lack of available information on diffusivities of ozone, oxygen and nitrogen in L zeolite, equilibrium theory was assumed in the model, for which:

$$\overline{q}_{i} = q_{i}^{*}$$
(3)

In addition, physical properties of L zeolite were taken the same as those of modernite since both zeolites have a 12-membered oxygen ring system, a similar pore size and a similar Si/Al ratio [24, 25]. Eventually, a heat capacity value typical of zeolite pellets was assumed in the model while the selected pellet diameter was in the range of industrial PSA applications including air separation [23, 26]. From the literature it is well known that zeolitic adsorbents used to produce industrially oxygen from air such as A type zeolite, X type zeolite, L type zeolite and modernite are usually characterized by negligible mass transfer limitations [23, 25]. Thus, it is expected that the kinetics of adsorption and desorption of ozone, oxygen and nitrogen in L zeolite will not be critical for the proposed separation. A complete list of physical properties and equilibrium parameters for the L zeolite considered in this study along with the relevant references can be found in Table 3.

Table 3 Physical properties and equilibrium parameters for L zeolite
Full size table

In all adsorption processes investigated in this study the cyclic steady state was reached after 20–30 cycles with differences of O2 purity and O2 recovery between the new and previous cycles both less than 10−5. In all simulations, the differential–algebraic equations (DAEs) were solved using gPROMS software [27]. Axial domains were discretized by the second order centered finite difference method (CFDM) with 100 elements along the bed length. The absolute and relative numerical tolerances were both set to 10−6.

3 Design basis and cycle configurations

As already mentioned, the aim of this study was to design, simulate and assess cyclic adsorption processes to recover oxygen from the outlet stream of a commercial ozone generator with a capacity of 30 kg h−1 of ozone and a typical composition of 10:90 wt% or 6.9:93.1 mol% ozone–oxygen [11, 16]. Thus, the resulting feed flowrate was 300 kg h−1 that is approximately equivalent to 2.52 mol s−1 or 0.031 m3 s−1. Since industrial ozone generators produce the ozone–oxygen mixture in the pressure range of 1.5–3.0 bar, an intermediate feed pressure of 2.0 bar was selected for this study. In designing the adsorption processes the bed diameter was determined to have a superficial velocity in the vicinity of 0.1 m s−1 that is typical of industrial PSA units [19]. The Lb/Db ratio was then fixed at 2 in order to have a cycle time of 8–12 min and pressure drops along the bed less than 0.01 bar. A full list of bed parameters and operating conditions for the adsorption processes is reported in Table 4.

Table 4 Bed parameters and operating conditions for the adsorption processes
Full size table

The simplest adsorption cycle to allow a continuous oxygen recovery and ozone production is the 2-bed, 2-step configuration depicted in Fig. 4. In the adsorption step, the ozone–oxygen mixture is fed to the adsorption bed where ozone is selectively adsorbed on L zeolite while oxygen passes through and is then recycled to the ozone generator (Fig. 1). In the desorption step, dry air at the same feed pressure flows counter-currently through the bed and strips out the ozone that has been adsorbed in the previous step. An ozone-air mixture leaves the bed around the same concentration as the incoming ozone so that it can be used by the downstream user (Fig. 1). This method of desorbing ozone by gas stripping using a sweep gas such as dry air is the most widespread amongst the patents in the literature and has been recently demonstrated by Linde’s OZORA™ system [11]. However, it was found that the performance of this adsorption cycle was unsatisfactory in that the O2 purity in the recycle stream was too low, up to a maximum of 93%, because of the nitrogen build up from air. Therefore, two additional adsorption cycles were assessed with the aim of boosting such O2 purity by removing the nitrogen after the desorption step. The target purity for the recovered oxygen was fixed at 98+%, similarly to Linde’s technology demonstration [11]. Higher amounts of nitrogen, in fact, could adversely affect the performance of the ozone generator.

Fig. 4
figure 4

Configuration of the 2-bed, 2-step cycle (AD: adsorption, DE: desorption with air; tAD = tDE = 240 s)

Full size image

The second adsorption cycle investigated in this study was a modification of the configuration patented by Grenier and Petit [12] and Fitch and Huertas [7]. The idea is to incorporate in the cycle an oxygen purge step after the desorption step where a portion of the recovered oxygen is co-currently introduced in the bed in order to strip out the nitrogen. Thus, the O2 purity in the recycle stream can be improved with a sacrifice of the O2 recovery. It is essential to operate such oxygen purge step co-currently so as to maximize the nitrogen removal and prevent the non-desorbed ozone from escaping in the air vent given that it will be concentrated near the bed feed end. For continuous operation, the oxygen purge step needs to have the same duration as the adsorption and desorption steps. The resulting configuration is the 3-bed, 3-step purge gas stripping cycle shown in Fig. 5.

Fig. 5
figure 5

Configuration of the 3-bed, 3-step cycle (AD: adsorption, DE: desorption with air, PU: co-current purge with oxygen; tAD = tDE = tPU = 240 s)

Full size image

Another method to remove the nitrogen after the desorption step is to incorporate in the adsorption cycle a co-current vacuum blowdown step followed by a counter-current light product pressurization step. In this way, the O2 purity in the recycle stream can be improved with an increase of energy consumption in running the nitrogen vacuum pump, as proposed by Cook et al. [8] and Weist Jr. [15]. Also in this case it is essential to operate the blowdown step co-currently so as to maximize the nitrogen removal and avoid ozone losses in the air vent. In terms of pressurization, light product pressurization is more suitable for purification processes [28] such as the oxygen recovery investigated in this study. In addition, the pressurization step needs to be operated counter-currently otherwise the non-desorbed ozone will be pushed near the bed product end and could breakthrough during the following adsorption step. For continuous operation, both vacuum blowdown and pressurization steps need to have half the duration of the adsorption and desorption steps. The resulting configuration is the 3-bed, 4-step VPSA cycle shown in Fig. 6.

Fig. 6
figure 6

Configuration of the 3-bed, 4-step cycle (AD: adsorption, DE: desorption with Air, BD: co-current blowdown, PR: counter-current pressurization with oxygen; tAD = tDE = 240 s, tBD = tPR = 120 s)

Full size image

For the 2-bed, 2-step configuration the total cycle time was 480 s (8 min) while for the 3-bed, 3-step and the 3-bed, 4-step configurations the total cycle time was 720 s (12 min). In the 3-bed, 4-step cycle the valve flow coefficient for both co-current blowdown and counter-current pressurization steps was set to 5 × 10−7 kg s−1 Pa−1/2. In all simulations it was assumed that the beds were initially filled with pure oxygen at the adsorption/desorption or blowdown pressure depending on the cycle schedule and at the temperature of 296.15 K.

Since ozone is selectively adsorbed on L zeolite compared to oxygen (Figs. 2, 3), oxygen will be recovered in the raffinate stream (oxygen recycle) during the adsorption step while ozone will be desorbed by dry air in the offgas stream (ozone product) during the desorption step. The key performance indicators for this separation are O2 purity, O2 recovery and specific energy consumption, which are calculated by the following equations:

$$O_{2} \;Purity = \frac{{\int_{0}^{{t_{AD} }} {c_{{O_{2} }} \cdot MW_{{O_{2} }} \cdot \left. u \right|_{{z = L_{b} }} dt} }}{{\sum\limits_{i = 1}^{n} {\int_{0}^{{t_{AD} }} {c_{i} \cdot MW_{i} \cdot \left. u \right|_{{z = L_{b} }} dt} } }}$$
(4)
$$\;O_{2} \;Recovery = \frac{{\int_{0}^{{t_{AD} }} {c_{{O_{2} }} \cdot \left. u \right|_{{z = L_{b} }} dt} - \int_{0}^{{t_{PU} }} {c_{{O_{2} }} \cdot \left. u \right|_{z = 0} dt} - \int_{0}^{{t_{PR} }} {c_{{O_{2} }} \cdot \left. u \right|_{{z = L_{b} }} dt} }}{{\int_{0}^{{t_{AD} }} {c_{{O_{2} }} \cdot \left. u \right|_{z = 0} dt} }}$$
(5)
$$Specific\;Energy = \frac{{\int_{0}^{{t_{{DE}} }} {Power_{{compressor}} dt} + \int_{0}^{{t_{{BD}} }} {Power_{{vacuum}} dt} }}{{\left( {\int_{0}^{{t_{{AD}} }} {c_{{O_{2} }} \cdot \left. u \right|_{{z = L_{b} }} dt} - \int_{0}^{{t_{{PU}} }} {c_{{O_{2} }} \cdot \left. u \right|_{{z = 0}} dt} - \int_{0}^{{t_{{PR}} }} {c_{{O_{2} }} \cdot \left. u \right|_{{z = L_{b} }} dt} } \right) \cdot MW_{{O2}} \cdot A_{b} }}$$
(6)
$$Power_{compressor} = \frac{1}{{\eta_{c} }} \cdot F \cdot \frac{\gamma }{\gamma - 1} \cdot \left( {\frac{{P_{atm} }}{{\rho_{g} }}} \right) \cdot \left[ {\left( {\frac{{P_{AD} }}{{P_{atm} }}} \right)^{{\frac{\gamma - 1}{\gamma }}} - 1} \right]$$
(7)
$$Power_{vacuum} = \frac{1}{{\eta_{v} }} \cdot F \cdot \frac{\gamma }{\gamma - 1} \cdot \left( {\frac{{P_{BD} }}{{\rho_{g} }}} \right) \cdot \left[ {\left( {\frac{{1.1_{{}} bar}}{{P_{BD} }}} \right)^{{\frac{\gamma - 1}{\gamma }}} - 1} \right]$$
(8)

Note that the O2 purity calculated by Eq. (4) is expressed in mass basis. In the numerator of Eq. (5) and in the denominator of Eq. (6) the second and third integral terms are non-zero when the adsorption cycle incorporates an oxygen purge step and a light product pressurization step, respectively. Likewise, the second integral term in the numerator of Eq. (6) is non-zero when the adsorption cycle incorporates a vacuum blowdown step. Both the co-current purge step in the 3-bed, 3-step cycle and the counter-current pressurization step in the 3-bed, 4-step cycle make use of the average O2 concentration in the oxygen recycle stream, which was produced during the adsorption step. Equations (6)–(8) relate to calculation of the specific energy required for the separation, which is obtained by dividing the power consumption in running the air compressor, and in case the nitrogen vacuum pump, by the amount of oxygen recovered in the adsorption cycle after the cyclic steady state (CSS) is achieved. Similarly to Eq. (4), Eq. (9) is used to calculate the component purity by mass in the ozone product stream:

$$Purity\;in\;O_{3} \;product\;stream = \frac{{\int_{0}^{{t_{DE} }} {c_{i} \cdot MW_{i} \cdot \left. u \right|_{z = 0} dt} }}{{\sum\limits_{i = 1}^{n} {\int_{0}^{{t_{DE} }} {c_{i} \cdot MW_{i} \cdot \left. u \right|_{z = 0} dt} } }}\quad i = O_{3} ,O_{2} ,N_{2}$$
(9)

In calculating the specific energy consumption [Eqs. (6)–(8)], an efficiency of 85% was considered for the air compressor [29], in line with industrial compressors used in air separation plants, while an efficiency of 75% was considered for the nitrogen vacuum pump [30], given the moderate levels of vacuum investigated in this study (Table 4).

4 Results and discussion

4.1 2-bed, 2-step cycle

The separation performance indicators for the 2-bed, 2-step cycle were assessed by means of the air purge-to-feed (P/F) ratio, defined as the ratio of the air mass flowrate during the desorption step to the feed mass flowrate during the adsorption step. The simulation results at cyclic steady state are given in Table 5. It is expected that with an increase of the air P/F ratio the adsorption bed will be better regenerated so that most of the ozone will be desorbed in the ozone product stream and only a small fraction will be recovered in the oxygen recycle stream. Likewise, higher air P/F ratios will incur greater energy consumptions as more air will need to be compressed from atmospheric pressure to the desorption pressure of 2.0 bar (Table 5). Figure 7 shows the ozone average adsorbed phase concentration along the dimensionless axial bed distance at the end of the desorption step. It can be noted how the profiles drop significantly at higher air P/F ratios starting from the bed product end, thus confirming the efficacy of the desorption.

Table 5 Performance results of the 2-bed, 2-step cycle
Full size table
Fig. 7
figure 7

Profiles of ozone average adsorbed phase concentration along the dimensionless axial bed distance at the end of the desorption step for the 2-bed, 2-step cycle

Full size image

Run 2 in Table 5, having an air P/F ratio of 0.85, showed two clear performance advantages in that the O3 purity in the recycle stream was very low, less than 0.1%, and, at the same time, the O3 purity in the product stream was 10.5%, which is very close to the ozone composition in the feed (10%). Therefore, an air P/F ratio of 0.85 along with a specific energy consumption of 76.6 kJ kgO2─1 was optimal for the separation. These results are confirmed by Fig. 8a and b showing the component gas phase mass fraction profiles at feed end (z = 0) and product end (z = Lb), respectively, over a cycle at cyclic steady state. In fact, the ozone does not break through during the adsorption step (Fig. 8b) while the ozone mass fraction is kept around the feed composition during the desorption step (Fig. 8a). Moreover, it is clear from Fig. 8a that nitrogen mass fractions reach values up to 70% within the first 25 s into the desorption step, thus the gas will be pushed out of the bed at the beginning of the adsorption step before pure oxygen can be produced (Fig. 8b). As anticipated, the maximum O2 purity achieved using the 2-bed, 2-step cycle was only 93% (Table 5), which was deemed too low for the recycle stream.

Fig. 8
figure 8

Profiles of component gas phase mass fraction at a feed end and b product end over a cycle at cyclic steady state for the 2-bed, 2-step cycle (Run 2)

Full size image

4.2 3-bed, 3-step cycle

In this cycle configuration the O2 purity in the recycle stream was raised at the expense of O2 recovery as a portion of the recovered oxygen was used to purge the nitrogen from the bed. Similarly to the air P/F ratio, the 3-bed, 3-step cycle was assessed with respect to the O2 P/F ratio, defined as the ratio of the oxygen mass flowrate during the purge step to the feed mass flowrate during the adsorption step. The simulation results at cyclic steady state are reported in Table 6 and exhibited in Fig. 9. Keeping the same air P/F ratio of 0.85, the O2 purity in the recycle stream increased with higher O2 P/F ratios because nitrogen is purged from the adsorption bed at a higher extent, while the O2 recovery decreased as more oxygen is consumed in the purge step. The specific energy consumption also increased as a result of the lower O2 recovery (Fig. 9). Figure 10 shows the nitrogen average adsorbed phase concentration along the dimensionless axial bed distance at the end of the purge step. In this case the profiles are abated at higher O2 P/F ratios starting from the bed feed end as the nitrogen purging becomes more effective.

Table 6 Performance results of the 3-bed, 3-step cycle
Full size table
Fig. 9
figure 9

Evolution of O2 purity, O2 recovery and specific energy consumption with the O2 P/F ratio for the 3-bed, 3-step cycle

Full size image
Fig. 10
figure 10

Profiles of nitrogen average adsorbed phase concentration along the dimensionless axial bed distance at the end of the purge step for the 3-bed, 3-step cycle

Full size image

Run 7 in Table 6, having an O2 P/F ratio of 0.075, achieved the separation target, i.e., an O2 purity of 98.8% in the recycle stream along with an O2 recovery of 91.8% and a specific energy consumption of 79.1 kJ kgO2─1. The high O2 purity in the recycle stream can be justified by inspecting the component gas phase mass fraction profiles at feed end over a cycle at cyclic steady state (Fig. 11a). It can be noted that at the end of the purge step the nitrogen mass fraction approaches zero. In turn, at product end the residual nitrogen in the bed that will contaminate the oxygen recycle stream is significantly reduced at the beginning of the adsorption step (Fig. 11b). Compared to the 2-bed, 2-step cycle (Run 2), the 3-bed, 3-step cycle (Run 7) reached the target O2 purity with a 3% sacrifice of O2 recovery.

Fig. 11
figure 11

Profiles of component gas phase mass fraction at a feed end and b product end over a cycle at cyclic steady state for the 3-bed, 3-step cycle (Run 7)

Full size image

4.3 3-bed, 4-step cycle

In this cycle configuration the O2 purity in the recycle stream was raised at the expense of additional energy consumption in running the vacuum pump used to remove the nitrogen during the blowdown step. The simulation results at cyclic steady state are reported in Table 7 and exhibited in Fig. 12, where the separation performance indicators were assessed by varying the blowdown pressure in the range of 0.25–1.0 bar. It should be noted that the nitrogen vacuum pump consumes energy also at 1.0 bar since it was assumed that the turbomachinery operated with a discharge pressure of 1.1 bar [31], as reported in Eq. (8). It is clear from Table 7 and Fig. 12 that a higher degree of vacuum during the blowdown step improves the O2 purity in the recycle stream while keeping a similar O2 recovery, since more nitrogen is removed from the bed. This is also corroborated by the profiles of nitrogen average adsorbed phase concentration along the dimensionless axial bed distance at the end of the blowdown step, which drop throughout the bed at lower blowdown pressures (Fig. 13). However, on the one hand the O2 purity varies linearly with the blowdown pressure following the Langmuir isotherm model in the low pressure region, while on the other hand the specific energy consumption sharply raises at lower blowdown pressures (Fig. 12). This rise in energy consumption is only due to the vacuum pump since the air compressor consumes nearly a constant energy (Table 7).

Table 7 Performance results of the 3-bed, 4-step cycle
Full size table
Fig. 12
figure 12

Evolution of O2 purity, O2 recovery and specific energy consumption with the blowdown pressure for the 3-bed, 4-step cycle

Full size image
Fig. 13
figure 13

Profiles of N2 average adsorbed phase concentration along the dimensionless axial bed distance at the end of the blowdown step for the 3-bed, 4-step cycle

Full size image

Run 12 in Table 7, having a blowdown pressure of 0.25 bar, achieved the following optimal separation performances: O2 purity in the recycle stream of 98.8%, O2 recovery in the recycle stream of 93.3% and total specific energy consumption of 111.7 kJ kgO2−1. Also in this case the high O2 purity in the recycle stream can be justified by Fig. 14a and b showing the component gas phase mass fraction profiles at feed end and product end, respectively, over a cycle at cyclic steady state. At product end the nitrogen mass fraction increases during the blowdown step as a significant amount of nitrogen is evacuated from the bed while at feed end the nitrogen mass fraction decreases towards the end of the pressurization step as the gas is pushed inside the bed by introducing pure oxygen. Consequently, in Fig. 14b at the beginning of the adsorption step the oxygen mass fraction is close to 1, it then briefly drops because of the residual nitrogen being pushed out of the bed and eventually rises again to 1 after around 30 s into the step. Compared to the 2-bed, 2-step cycle (Run 2), the 3-bed, 4-step cycle (Run 12) reached the target O2 purity with a 45.8% increase in energy consumption.

Fig. 14
figure 14

Profiles of component gas phase mass fraction at a feed end and b product end over a cycle at cyclic steady state for the 3-bed, 4-step cycle (Run 12)

Full size image

Confronting now the 3-bed, 3-step and the 3-bed, 4-step cycles, both Run 7 and Run 12 managed to achieve a satisfactory O2 purity in the recycle stream of 98.8%, which can also be confirmed by the nitrogen average adsorbed concentration along the bed at cyclic steady state prior to the adsorption step. In fact, the area under the curve with an O2 P/F of 0.075 in Fig. 10 (Run 7) and that under the curve with a blowdown pressure of 0.25 bar in Fig. 13 (Run 12) are approximately equivalent. Compared to Run 7, Run 12 exhibited a 1.5% higher O2 recovery in the recycle stream but a 41.2% higher specific energy consumption mainly because of the nitrogen vacuum pump. Therefore, an in-depth economic analysis would be required to identify the best configuration between the 3-bed, 3-step purge gas stripping cycle and the 3-bed, 4-step VPSA cycle.

5 Conclusions

The aim of this work was to design, simulate and assess for the first time cyclic adsorption processes applied to the outlet stream of a commercial ozone generator fed with high-purity oxygen for recovering oxygen from the resulting ozone–oxygen mixture and recycling it to the ozone generator. The target purity for the recovered oxygen was fixed at 98+%, so as not to adversely affect the performance of the ozone generator. The separation was carried out using L zeolite as an adsorbent at ambient temperature. The adsorption processes were rigorously simulated by a mathematical model integrating mass, momentum and energy balances along with the extended Langmuir model and the equilibrium theory.

First, a 2-bed, 2-step cycle was investigated where, after the adsorption step, the ozone was desorbed via purge gas stripping using dry air as a sweep gas. It was found that an air P/F ratio of 0.85 was optimal to desorb the ozone from the bed and concentrate it in the product stream around the feed composition (10 wt%), thus making it suitable for the downstream user. However, this configuration could not achieve an O2 purity of 98+% in the recycle stream because of the nitrogen build up from air, so two additional adsorption cycles were proposed and assessed to remove the nitrogen. In the 3-bed, 3-step configuration an oxygen purge step was incorporated in the cycle to boost the O2 purity at the expense of O2 recovery while in the 3-bed, 4-step configuration a vacuum blowdown step was incorporated in the cycle to boost the O2 purity at the expense of additional energy consumption in running the nitrogen vacuum pump. Both cycles enhanced such O2 purity in the recycle stream up to 98.8% but the O2 recovery and the total specific energy consumption in the 3-bed, 4-step cycle were 1.5% and 41.2% higher, respectively, than the 3-bed, 3-step cycle. Given the overall high recovery rates of oxygen in the recycle stream, the adsorption processes devised in this study could substantially improve the process economics of industrial ozone generators.