Carbon Mass Balance in Arthrospira platensis Culture with Medium Recycle and High CO 2 Supply

: Medium recycling combined with CO 2 recovery helps sustainable use of the alkaline medium in Arthrospira culture. However, high CO 2 supply may cause inorganic carbon accumulation and pH reduction, which could result in low CO 2 recovery and reduced algal growth. This study aimed to elucidate the e ﬀ ect of medium recycling and high CO 2 supply through carbon mass balance analysis in Arthrospira culture. In all CO 2 supply conditions, carbon supply was higher than Arthrospira carbon assimilation, which accounted for 30–58% of supply. However, CO 2 recovery of nearly 100% and 63% for lower (0.20 and 0.39 gC L − 1 d − 1 ) and higher (0.59 gC L − 1 d − 1 ) CO 2 supply rates were achieved, respectively, because of the high concentration of the alkaline agent. The excess carbon accumulated in the medium and ultimately escaped from the system in a form of dissolved inorganic carbon (DIC). Dissolved organic carbon (DOC) contributed to 16–24% of the total photosynthetically assimilated carbon, and the ﬁnal concentration reached 260–367 mgC L − 1 , but there was no signiﬁcant growth reduction caused by DIC and DOC accumulation. This study demonstrated the stability of the medium-recycling process even at high CO 2 supply rates although a balanced supply is recommended for longer operations. the medium. Carbon dioxide recovery rates of 63–100% and steady algal growth of 0.31 gDW L 1 d − 1 were achieved even with the excessive carbon supply compared with the A. platensis carbon assimilation. Carbon mass balance analysis revealed that up of the into the in a form of and thus care be in evaluating the This study demonstrated the e ﬀ ectiveness and stability of the medium-recycling process even at high CO 2 supply rates. Further study is needed to elucidate the mechanisms of DOC excretion and its e ﬀ ects on the growth and biomass characteristics.


Introduction
Alkaliphilic microalgae and cyanobacteria (inclusively referred to as algae here) have been of great interest in a commercial application. Their selective growth conditions (i.e., high pH and alkalinity) prevent overgrowth of other organisms even in outdoor open-pond cultivations and allows relatively easy quality control. Among alkaliphilic algae, Arthrospira (also known as Spirulina) is the most commercially produced genus because of its high nutritional value (protein up to 77%dwt, minerals, and vitamins) [1] and application for blue natural color pigment production [2]. The global production of Arthrospira has been estimated to be approximately 10,000 metric tons in a report in 2013 [2] and is still expanding. Its emerging bio-active compounds have also been well studied [3][4][5][6].
The challenge in mass-cultivation of alkaliphilic microalgae, however, is the large requirement of alkaline agents. For example, the conventional Zarrouk medium requires 16.8 kg of sodium bicarbonate (NaHCO 3 ) for every cubic meter of the medium [7]. Thus, medium recycles after biomass harvest is a preferred operation [8,9]. In a used medium, photosynthetic utilization of bicarbonate (HCO 3 − ) and Figure 1. Experimental set-up of the semi-continuous two-phase CO2 recovery process. After 23.5 hours of CO2 absorption and algal incubation period, the medium was partially recirculated between the two reactors. After recirculation, the headspace of the absorption column was replaced with CO2 containing gas (Table 1) and a newly filled gas bag was attached.
The absorption column was a 1-L glass medium bottle with an active volume of 0.4 L and a gasliquid interfacial area of approximately 73 cm 2 . For the start-up of the experiment, the autoclaved column was filled with a 0.4-L filter-sterilized medium. The headspace of the absorption column was replaced with different concentrations of CO2/N2 mixture for each CO2 supply rates (Table 1). Nitrogen gas of the volume same as that of headspace was injected to all the gasbags so that the headspace pressure remained at the atmospheric pressure (1.013 × 10 5 Pa) even after all CO2 was absorbed. The absorption columns were placed in an incubator at 25 °C under dark, and the absorbent was stirred continuously with magnetic stirrers approximately at 200 rpm.
The PBR was a glass column reactor with an active volume of 1 L and an inner diameter of 106 mm. The PBR was semi-open with 0.2-µm air-filter (Aervent-50, Millipore, USA) attached on top to aseptically alleviate pressure build-up due to oxygen production. The PBR was incubated at 35 °C with 24-hour continuous light with the surface photosynthetic photon flux density (PPFD) of 300 µmol photons m -2 s -1 .  Figure 1. Experimental set-up of the semi-continuous two-phase CO 2 recovery process. After 23.5 hours of CO 2 absorption and algal incubation period, the medium was partially recirculated between the two reactors. After recirculation, the headspace of the absorption column was replaced with CO 2 containing gas (Table 1) and a newly filled gas bag was attached. The absorption column was a 1-L glass medium bottle with an active volume of 0.4 L and a gas-liquid interfacial area of approximately 73 cm 2 . For the start-up of the experiment, the autoclaved column was filled with a 0.4-L filter-sterilized medium. The headspace of the absorption column was replaced with different concentrations of CO 2 /N 2 mixture for each CO 2 supply rates (Table 1). Nitrogen gas of the volume same as that of headspace was injected to all the gasbags so that the headspace pressure remained at the atmospheric pressure (1.013 × 10 5 Pa) even after all CO 2 was absorbed. The absorption columns were placed in an incubator at 25 • C under dark, and the absorbent was stirred continuously with magnetic stirrers approximately at 200 rpm.

Recirculation
The PBR was a glass column reactor with an active volume of 1 L and an inner diameter of 106 mm. The PBR was semi-open with 0.2-µm air-filter (Aervent-50, Millipore, USA) attached on top to aseptically alleviate pressure build-up due to oxygen production. The PBR was incubated at 35 • C with 24-hour continuous light with the surface photosynthetic photon flux density (PPFD) of 300 µmol photons m −2 s −1 .

Experimental Procedures
Twenty-four hours after CO 2 injection into the absorption column and inoculation of PBR, algal harvest and medium recycle was operated ( Figure 1). Firstly, 0.2 L of the algal suspension was withdrawn from the PBR, out of which, 0.05 L was discharged as analytical sample and 0.15 L was filtered through multiple screen meshes and a glass fiber filter with pore size of 0.7 µm (GF/F, Whatman, USA) to obtain filtrate for later transfer into the absorption column. Secondly, 0.2 L of the solution in the absorption column was transferred to the PBR aseptically. Finally, 0.15 L of the filtrate was mixed with 0.05-L new modified SOT medium, filtered through a sterile 0.22-µm membrane filter (Millipore Express ® PLUS, Millipore, San Diego, CA, USA), and aseptically transferred to the absorption column to complete the liquid recycle.
The headspace of the absorption column was then replaced by blowing the respective concentrations of CO 2 gas at 1 L min −1 for 3 min, and a refilled gas bag was attached. The same procedure was repeated every 24 hours. Samples were taken from the algal suspension and the CO 2 absorbent both before and after the media recycle. The hydraulic retention times (HRT) of the absorption column, the PBR, and the entire system were 1.8, 4.75, and 18 days, respectively. The experiment was continued for 18 days.

Analytical Procedures
The algal suspension was used to measure algal dry weight (DW) and algal cell carbon and nitrogen content. The filtrate of algal suspension was used to measure pH, electrical conductivity, nutrients (nitrate, nitrite, and phosphorus), dissolved inorganic carbon (DIC), and dissolved organic carbon (DOC). The gas sample was used to measure headspace CO 2 , N 2 , and O 2 concentration. The amount of CO 2 absorbed into the medium was calculated from the difference of DIC concentration before and after the 24-h absorption period.
Algal DW was measured every day using glass fiber filters. Firstly, algal suspensions were filtered through pre-weighed glass fiber filters with a pore size of 0.7 µm (GF/F, Whatman, USA) and washed with distilled water three times. Then, filters were dried in an oven at 60 • C for over 24 hours and subsequently cooled to room temperature in a desiccator before weighing. Filters were weighed with an ultra-micro balance (XP6U Ultra Micro Comparator, Mettler Toledo, Columbus, OH, USA). Algal cell carbon and nitrogen contents were measured by measuring particulate organic carbon and nitrogen (POC and PON) of filters samples using an elemental analyzer (Flash 2000 CHN, Thermo, Waltham, MA, USA). To measure DIC and DOC, TOC analyzer (TOC-V CSH , Shimadzu, Kyoto, Japan) was used. Nutrients (nitrate, nitrite, and phosphate) were analyzed with an automated nutrient analyzer (SWAAN, BL-TEC, Tokyo, Japan).
Gas composition (CO 2 , N 2 , and O 2 ) was measured using a gas chromatograph (Shimadzu, GC-2014, Japan) equipped with a packed column (Shincarbon ST, 6.0 m long, 3 mm I.D., Shimadzu, Japan) and a thermal conductivity detector. The temperature of the injector and the detector were maintained at 120 • C and 260 • C, respectively. The column temperature was gradually increased from 40 • C to 250 • C. Helium was used as the carrier gas with a flow rate of 40 mL min −1 .

Modeling pH Variation with DIC Change
The composition of DIC was calculated based on the following equations [26]: where K 1 and K 2 are stoichiometric constants for bicarbonate and carbonate, respectively. In this study, pK 1 = 6.38 [27] and pK 2 = 9.8 were assumed. The pK 2 value was derived from NaOH titration (cf. Supplementary material Section S1). Based on the above equations, pH variation with CO 2 absorption/desorption was modeled. The buffering function of phosphate and borate was assumed to be negligible because their concentrations were about 100 times lower than the DIC concentration. In the current experiment, DIC concentration can be expressed as: where c 0 is the initial DIC concentration by the addition of NaHCO 3 (0.27 mol L −1 ), and ∆DIC is the change in DIC concentration due to CO 2 absorption or photosynthetic assimilation (mol L −1 ). Charge balance in the medium can be expressed as: where c b is the final concentration of NaOH in the modified SOT medium that raised pH to 10.5 (0.215 mol L −1 ). Using the ion product constant of water K w = H + [OH − ] = 1.0 × 10 −14 mol 2 L −2 , Equation (12) can be expressed as: Finally, Equations (6), (10), (11) and (13) can be formulated into a quartic equation: In the current experiment, all variables in Equation (7) except for ∆DIC and H + remained constant. Therefore, pH variation with CO 2 absorption/desorption can be estimated based on changes in DIC concentration (∆DIC) with this model. The quartic equation was solved with Mathematica 8.0 (Wolfram Research, Champaign, IL, USA).
Buffer intensity, β, was calculated based on the following approximation for diprotic acids [28]: Appl. Sci. 2020, 10, 228 where dC A is the infinitesimal amount of acids required to decrease dpH. The equilibrated CO 2 gas concentration was calculated according to Henry's law: where p CO 2 is the partial pressure of CO 2 under equilibrium condition (Pa), [CO 2 ] is the aqueous concentration of CO 2 (mol L −1 ), and H cp is Henry's law solubility constant (3.30 × 10 −7 mol L −1 Pa −1 ) (Sander 2015).

Biomass Production, Carbon Fixation, and Carbon Mass Balance
Algal volumetric production rate (P x ; gDW L −1 d −1 ) was calculated with the following equation: The growth steady-state was defined as a period with less than 10% variation in the 3-day moving averages of DW, which was Day 10-18 for all Runs in this study. The carbon mass balance was calculated using the following equations: where C in and C out represent the total incoming and outgoing carbon (gC) after 18 days of the experiment, respectively. CO 2in refers to the carbon dioxide supplied to the process. ABS 0 and PBR 0 refer to the carbon mass in the initial medium of the CO 2 absorption column and PBR, respectively. ABSf and PBRf indicate the final carbon mass of each reactor. ABSs and PBRs represent the carbon mass in each sample taken every day. N represents carbon mass in the new SOT medium.

Statistical Analysis
Results are expressed as means ± standard deviations, where available. The correlation coefficient was obtained using simple regression analysis (Excel software). Results from different conditions were analyzed using the Tukey-Kramer method. Differences with p < 0.05 were considered significant.

Change in pH and DIC Concentration with CO 2 Absorption
The pH in both absorption columns and PBRs repeated reduction and elevation every 24 hours upon the medium circulation and CO 2 gas replenishment ( Figure 2), but the range was smaller in PBRs than in absorption columns. Most of the supplied CO 2 was absorbed in Runs 1 and 2, while the removal efficiency in Run 3 was 63% (Table 2). to approximately 8.5 (Figure 2). Although the DIC in the PBR increased with a longer experimental duration in all conditions, the maximum value was 0.43 mol L -1 (ΔDIC = 0.16 mol L -1 ) in Run 3, which was within the buffer region (0-0.17 mol L -1 ). As a result, the pH in PBR remained over 9 even in Runs 2 and 3 ( Figure 2). Equilibrated CO2 gas concentration was also drawn in Figure 3a based on Equation (6). With a slight increase in ΔDIC less than 0.02 mol L -1 , equilibrated CO2 gas exceeded atmospheric concentration (c.a. 0.04%).  There are two points in a day, representing pH change due to the circulation of medium between PBR and absorption column. To elucidate the pH buffer characteristics of the modified SOT medium, the relationship between CO 2 absorption and medium pH was modeled using Equation (4). The modeled curve showed a close similarity to the experimental data (Figure 3a). The accuracy was confirmed with regression, in which R 2 was 0.8593 and residual mean square error (RSME) was 0.307 with an average pH 9.29, indicating nearly 97% accuracy.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 8 of 14 Figure 3. Estimation of (a) pH, (b) inorganic carbon species, and buffer capacity β, from change in dissolved inorganic carbon (ΔDIC). All data points were from absorption columns. ΔDIC was calculated by subtracting the initial dissolved inorganic carbon (DIC) (ca. 0.27 mol L -1 ) from respective DIC on each day.

Algal Growth and DOC Excretion
The biomass concentrations of A. platensis in PBRs showed similar trends among three Runs; it increased until Day 8-11 and became steady-state (Figure 4a). The average steady-state biomass concentrations were 1.33 ± 0.07, 1.42 ± 0.08, and 1.34 ± 0.10 gDW L -1 for Runs 1, 2, and 3, respectively, and there was no statistical difference (p > 0.05). There was also no statistical difference in the average biomass production rates, Px, among Runs during the steady-state (0.27 ± 0.10, 0.31 ± 0.10, and 0.27 ± 0.09 gDW L -1 d -1 for Runs 1, 2, and 3, respectively). Nitrate and phosphate concentrations gradually decreased during the 18 days of culture, but more than 60 and 70% of the original medium remained at the end, respectively. Dissolved organic carbon concentration kept increasing in PBRs (Figure 4b).
The final concentration ranged from 260 to 367 mgC L -1 , while all Runs had similar initial values (110-128 mgC L -1 ).

Carbon Mass Balance
The carbon content of the dry cell of A. platensis during this experiment was found to be 32 ± 4% from the elemental analysis, and it did not change significantly throughout the experiment. The overall balance between the total incoming carbon mass (Cin; absorbed + remained CO2) and outgoing carbon mass (Cout; differences between initial and final concentrations of DIC, DOC, and POC) was evaluated ( Figure 5). Near closure of the balance was observed with the difference of less than 3-23% between Cin and Cout. The absorbed CO2 was converted into DIC, DOC, and POC. Average biological carbon fixation (DOC + POC) rates were 0.114, 0.131, and 0.103 gC L -1 d -1 for Runs 1, 2, and 3, respectively, and the conversion efficiencies from absorbed CO2 to DOC + POC were 58, 38, and 30%, respectively. The difference between CO2 supply and biological carbon fixation was relatively large Equilibrated CO 2 gas (%) Figure 3. Estimation of (a) pH, (b) inorganic carbon species, and buffer capacity β, from change in dissolved inorganic carbon (∆DIC). All data points were from absorption columns. ∆DIC was calculated by subtracting the initial dissolved inorganic carbon (DIC) (ca. 0.27 mol L −1 ) from respective DIC on each day.
Increase in ∆DIC (CO 2 absorption) lowered the pH of the solution, but a buffer region existed where ∆DIC was 0-0.17 mol L −1 (pH range: 8.9-10.4; Figure 3a), at which buffer intensity β was over half the maximum (>0.10 mol L −1 ; Figure 3b). In Runs 2 and 3 after Day 3, the CO 2 absorbent received so large amount of CO 2 gas that ∆DIC exceeded the buffer region, causing drastic drops of pH down to approximately 8.5 (Figure 2). Although the DIC in the PBR increased with a longer experimental duration in all conditions, the maximum value was 0.43 mol L −1 (∆DIC = 0.16 mol L −1 ) in Run 3, which was within the buffer region (0-0.17 mol L −1 ). As a result, the pH in PBR remained over 9 even in Runs 2 and 3 ( Figure 2). Equilibrated CO 2 gas concentration was also drawn in Figure 3a based on Equation (6). With a slight increase in ∆DIC less than 0.02 mol L −1 , equilibrated CO 2 gas exceeded atmospheric concentration (c.a. 0.04%).

Algal Growth and DOC Excretion
The biomass concentrations of A. platensis in PBRs showed similar trends among three Runs; it increased until Day 8-11 and became steady-state (Figure 4a). The average steady-state biomass concentrations were 1.33 ± 0.07, 1.42 ± 0.08, and 1.34 ± 0.10 gDW L −1 for Runs 1, 2, and 3, respectively, and there was no statistical difference (p > 0.05). There was also no statistical difference in the average biomass production rates, P x , among Runs during the steady-state (0.27 ± 0.10, 0.31 ± 0.10, and 0.27 ± 0.09 gDW L −1 d −1 for Runs 1, 2, and 3, respectively). Nitrate and phosphate concentrations gradually decreased during the 18 days of culture, but more than 60 and 70% of the original medium remained at the end, respectively. Dissolved organic carbon concentration kept increasing in PBRs (Figure 4b). The final concentration ranged from 260 to 367 mgC L −1 , while all Runs had similar initial values (110-128 mgC L −1 ). evaluated ( Figure 5). Near closure of the balance was observed with the difference of less than 3-23% between Cin and Cout. The absorbed CO2 was converted into DIC, DOC, and POC. Average biological carbon fixation (DOC + POC) rates were 0.114, 0.131, and 0.103 gC L −1 d −1 for Runs 1, 2, and 3, respectively, and the conversion efficiencies from absorbed CO2 to DOC + POC were 58, 38, and 30%, respectively. The difference between CO2 supply and biological carbon fixation was relatively large in Runs 2 and 3, causing DIC accumulation (Figures S3 and S4) and pH decrease in the absorption column ( Figure 2). The carbon mass flux analysis also revealed that out of all photosynthetically fixed carbon (increased and discharged POC + DOC), 16-24% was converted into DOC ( Figure 5). The accumulated DIC ultimately escaped from the system in the medium discharge (cf. supplementary data Figure S5 for carbon flux).

Carbon Mass Balance
The carbon content of the dry cell of A. platensis during this experiment was found to be 32 ± 4% from the elemental analysis, and it did not change significantly throughout the experiment. The overall balance between the total incoming carbon mass (C in ; absorbed + remained CO 2 ) and outgoing carbon mass (C out ; differences between initial and final concentrations of DIC, DOC, and POC) was evaluated ( Figure 5). Near closure of the balance was observed with the difference of less than 3-23% between C in and C out . The absorbed CO 2 was converted into DIC, DOC, and POC. Average biological carbon fixation (DOC + POC) rates were 0.114, 0.131, and 0.103 gC L −1 d −1 for Runs 1, 2, and 3, respectively, and the conversion efficiencies from absorbed CO 2 to DOC + POC were 58, 38, and 30%, respectively. The difference between CO 2 supply and biological carbon fixation was relatively large in Runs 2 and 3, causing DIC accumulation (Figures S3 and S4) and pH decrease in the absorption column ( Figure 2).
The carbon mass flux analysis also revealed that out of all photosynthetically fixed carbon (increased and discharged POC + DOC), 16-24% was converted into DOC ( Figure 5). The accumulated DIC ultimately escaped from the system in the medium discharge (cf. supplementary data Figure S5 for carbon flux).

Effect of High CO 2 Supply
According to the carbon mass balance analysis, the CO 2 supply was found to be higher than carbon assimilation by Arthrospira in all conditions ( Figure 5). Nonetheless, the strong pH buffer contained in alkaliphilic algal media allowed stable pH even with high CO 2 supply, and nearly 100% CO 2 recoveries were achieved for Runs 1 and 2 ( Table 2). These results indicate the stability of the CO 2 recovery process using alkaliphilic algae culture. The highest CO 2 supply (Run 3) resulted in lower CO 2 recovery (63%) due to increased [CO 2 ]. Absorption of CO 2 led to a reduction of pH and an increase in DIC, both resulting in increased [CO 2 ]. When [CO 2 ] reached a high level that equilibrated with gas CO 2 concentration, the CO 2 absorption terminated. Similar results were obtained in previous studies that tested CO 2 absorption with high-DIC medium [21,22]. Gonzalez-Lopez et al. [22] tested various concentrations of NaHCO 3 and Na 2 CO 3 for CO 2 recovery. Without the addition of alkaline agents, pH reduced from 10 to less than 7 in the first 2 min. On the other hand, >0.3 mol L −1 of NaHCO 3 and Na 2 CO 3 sustained pH > 9.5 for more than 30 min, resulting in total carbon absorption of up to 2.0 g L −1 . Total absorptivity increased with increasing initial DIC. As such, estimation of the maximal CO 2 absorptivity of the target medium should be necessary for designing an efficient CO 2 recovery system. See Supplementary Material Sections S2 and S3 for the estimation of maximal CO 2 absorptivity of the current experiment.
The effect of high CO 2 supply rates on Arthrospira growth was minimized because of the separation of the CO 2 absorption column and the PBR, as suggested in previous studies [10,21,22]. While pH was reduced to less than 8.5 in the CO 2 absorption column of Runs 2 and 3, it was maintained at around 9.5 in PBRs (Figure 3). Since CO 2 absorption reached oversaturation in the absorption column of Run 3, further CO 2 absorption was prevented, which worked as a protection to avoid excess carbon supply for Arthrospira growth. Since the optimum pH of Arthrospira is 9.8, and its 20% growth reduction is estimated at pH < 9.2 [17], such protection worked effectively to prevent growth reduction.
Carbon mass balance suggested a negligible escape of DIC from PBRs to the atmosphere as gaseous CO 2 , although [CO 2 ] was higher than air-equilibrium level (Figure 3a; cf. Section 3.1). This contradictory result could be explained by the prevention of air exchange between PBR headspace and atmosphere owing to the air-vent filters attached to the PBRs. In practical operations, Arthrospira cultures are majorly conducted in an outdoor pond or in an aerated PBR. To prevent DIC escapes in such conditions, CO 2 supply should be adjusted to maintain ∆DIC close to the air-equilibrium (in the current condition, ∆DIC < 0.02 mol L −1 ). Such ∆DIC level is also preferential for Arthrospira culture, since pH < 9.5 reportedly would increase the risk of contamination [2].

Effect of Medium Recycling
In the current process, 75% of the harvested medium was recycled (Figure 1). According to Cui et al. [24], the cost of alkaline agent reaches 91% of the total medium cost. The current medium recycling process saved 68% of the medium cost, under the assumption that other nutrients are supplemented at every recycling. The medium cost can be further reduced if the medium recycling rate and nutrient supplementation are optimized.
With high CO 2 supply and medium recycling, accumulation of DIC and DOC were observed in all Runs (Figures 2 and 3). Previous studies reported reduced growth rate with high DIC (>0.3 mol L −1 ; [17]) and DOC (>60-100 mg L −1 ; [8,29]). However, in this study, no significant reduction in biomass production rate or in growth rate was observed. The DIC accumulation seemingly did not have much effect at the current level (up to 0.4 mol L −1 ). On the other hand, DOC concentration was 2-3 times higher than the reported inhibition level. The reason this DOC did not show clear inhibition could be the difference in DOC characteristics. A previous study with a green alga Scenedesmus acuminatus reported differential inhibitive levels of DOC depending on growth stages, where DOC from the declining phase had severer inhibition than that from exponential and stationary phases [30]. Differential DOC inhibition levels were also observed between the indoor and outdoor culture of Arthrospira, where DOC in indoor culture inhibited the growth at as low DOC as 65 mg L −1 , while that in outdoor culture did not significantly inhibit the growth even at 900 mg L −1 [24]. Since the current semi-continuous culture maintained continuous growth, cells may not have excreted inhibitive DOC. Further study is needed to elucidate the effect of high CO 2 supply on the characteristics of DOC.
Nevertheless, further DOC accumulation may cause contamination of heterotrophic microorganisms and growth inhibition, and therefore should be avoided. The previous study reports differential amounts of DOC excretion depending on culture conditions. In a culture of A. platensis, DOC release was the most substantial during stationary-phase (> 30% primary production), but it was below 5% during the exponential phase [31]. Excretion of DOC could be decreased through raising the culture-specific growth rate, by maintaining actively growing cells. Since the specific growth rate can be controlled by adjusting the dilution rate [32], a high dilution rate of the PBR may help to keep cells active for minimized DOC production. The excretion of DOC also changes with nutrient availability [33,34], mechanical mixing, and so on. Further study is necessary to elucidate the effect of those conditions on the production and inhibitive characteristics of DOC.

Evaluation of CO 2 Recovery with Media-Recycled Arthrospira Culture
The large CO 2 absorption capacity of the high-DIC medium is an advantage of the current CO 2 recovery process. The large capacity enables pH stability. For example, the pH of the absorbent only decreased from 10.5 to 7.9 with continued CO 2 supply (total absorption of 2.6 gC L −1 ), while it would decrease from 10.5 to 5.2 with merely 0.06 gC L −1 of CO 2 without buffer. The pH reduction limits not only CO 2 absorption but also the growth of algae. Most algae have narrow ranges of suitable pHs, such as 9-10.5 for Arthrospira platensis [17], 7-8 for Dunaliella salina [35], and 6-8 for Chlorella sorokiniana [36], and an abrupt pH change may severely hinder the growth. The pH buffer is especially important with an outdoor algae culture, since algal CO 2 fixation fluctuates depending on the weather and seasons. When algal CO 2 fixation is low, CO 2 supply may lead to a drastic drop in pH without buffer functions. Therefore, CO 2 recovery using Arthrospira culture may be incorporated into the outdoor culture, which would decrease the cost and energy input in CO 2 recovery and algae production.
A similar CO 2 recovery with media-recycled Arthrospira culture was attempted by Cui et al. [24]. They achieved a CO 2 fixation rate of 0.31 gCO 2 L −1 d −1 , which was similar to 0.32-0.36 gCO 2 L −1 d −1 of this study. They revealed a correlation of growth deterioration and DOC accumulation after repeated medium recycle, which was not observed in the current experiment probably owing to the difference in the culture mode (batch or semi-continuous) and/or duration (26 days or 18 days). Another study was performed by Gonzalez-Lopez et al. [22] on CO 2 recovery with an alkaliphilic cyanobacterium Anabaena sp. They achieved nearly 100% CO 2 recovery and productivity of up to 0.4 g L −1 d −1 , which were similar to the current study. They were successful in predicting kinetic reactions of CO 2 recovery using a model. However, neither study reported the carbon mass balance and the fate of carbon incorporated into the system. The fate of carbon may change the interpretation of recovery efficiency. For example, Gonzalez-Lopez et al. [22] reported 95% efficiency of carbon utilization as to explain only about 4% escaped as CO 2 gas, but this efficiency does not take into account the remaining DIC and excreted DOC as unused or unnecessary carbon output. The current study revealed that 24% of assimilated carbon was released as CO 2 ( Figure 5). If this DOC requires further treatment, such as aerobic activated sludge, the carbon footprint would be higher than previously reported algal CCU process. Thus, this study indicated the importance of carbon mass balance analysis in the practical evaluation of CO 2 recovery processes.

Conclusions
This study conducted semi-continuous cultures of A. platensis with medium recycling and CO 2 supply. Medium recycling enabled 68% of total medium cost reduction and did not affect the algal growth although DIC and DOC accumulated in the medium. Carbon dioxide recovery rates of 63-100% and steady algal growth of 0.31 gDW L −1 d −1 were achieved even with the excessive carbon supply compared with the A. platensis carbon assimilation. Carbon mass balance analysis revealed that up to 24% of the assimilated carbon escapes into the medium in a form of DOC, and thus care needs to be taken in evaluating the process. This study demonstrated the effectiveness and stability of the medium-recycling process even at high CO 2 supply rates. Further study is needed to elucidate the mechanisms of DOC excretion and its effects on the growth and biomass characteristics.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-3417/10/1/228/s1: explanations on Figure S1. Modified SOT medium titration with KOH (2) CO 2 partial pressure in the absorption columns, Figure S2. Calculated and experimental amount of the absorbed CO 2 over 24 h (CO2abs) at different initial dissolved inorganic carbon (DIC). Plots: experimental values; solid lines: calculated values; and dashed lines: theoretical 100% absorption of supplied CO 2 Figure S3. Dissolved inorganic carbon concentrations in CO 2 absorption columns; (a,c,e) after CO 2 supply, and (b,d,f) after medium recycle from the photobioreactors. Low pH after CO 2 supply induced decrease of CO 3 2− fraction and increase of HCO 3 − and CO 2 fractions, Figure S4. Dissolved inorganic carbon concentration in photobioreactors; (a,c,e) after 1-day incubation, and (b,d,f) after medium input from CO 2 absorption columns. Slight reduction in total DIC after incubation was observed. Figure  S5. Carbon mass flux during 18-day operation. CO 2 supply rate of (a) Run 1, (b) Run 2, and (c) Run 3. Each box represents total cumulative mass (in milligram) of carbon fraction introduced or discharged over 18 days in the forms of CO 2 , dissolved inorganic carbon (DIC), dissolved organic carbon (DOC) or algal biomass (particulate organic carbon; POC). The mass inside the CO 2 absorption column and photobioreactor represent differences from the initial state.