2.1. Electrolytic Bubble Generation on IrTaOx Anode and Ti Cathode
Figure 1 shows the morphology and composition of the IrTaO
x (Ir:Ta = 7:3) mixed metal oxide anode employed in this study. The dimensionally stable anodes based on IrO
2 have widely been deployed in EF, due to the supreme electrocatalytic activity for oxygen evolution reaction (OER) as well as corrosion-resistant properties in acidic (anodic) environments [
16,
17]. The catalytic activity for OER has been known to be in a volcano-type relation with the metal-oxygen bond strength, in which IrO
2 and RuO
2 occupied the apex with the minimal kinetic barrier (overpotential) for OER [
16,
17]. The Ta as the secondary component of IrTaO
x could enhance the anodic stability [
17]. IrO
2 is known to carry out the OER by a deprotonation mechanism in near-neutral to acidic pH, exemplified by the following equation:
where –OH and –O are surface-bound reactive oxygen species as intermediates of OER. Scanning electron microscopy (SEM) noted a smooth horizontal surface without a prevailed crack, which would be beneficial for a long-term stability. A penetration of electrolytes through a crack on the surface could induce oxidation of the Ti substrate and passivation by TiO
2. Energy dispersive X-ray spectroscopy (EDX) noted prevailing signals of Ir and Ta with minor contributions of Cl from the residual precursor salts. The X-ray diffraction (XRD) pattern unambiguously indicated a rutile IrO
2 crystalline structure (JCPDS #15870) that was insignificantly altered by the mixed Ta components.
The CE values for bubble generation, prerequisite for G/S ratio calculation in EF, were estimated in 5 mM NaCl solutions whose electrical conductivity (0.61 mS/cm) was comparable with activated sludge mixed liquor (0.60 mS/cm) in this study. To simulate a continuous EF operation, the electrolytes were initially saturated with O
2 and H
2, and mass transfer between rising bubbles and bulk electrolytes could be ruled out [
2]. At
j > 8 mA/cm
2, the CE was consistently near 1 for H
2 bubble generation (measurements with an O
2 absorbent), as shown in
Figure 2. Parallel determination on flow rates of gas mixture (without the O
2 absorbent) estimated the CE of OER near 0.4. These observations were compatible with our previous estimation [
2,
13] on overall gas production efficiency of 0.7 (with assumed H
2:O
2 = 2:1). In comparison, the CE was much lower at
j lower than 8 mA/cm
2. Vogt [
18] suggested that the bubble generation efficiency roughly coincided with the proportion of electrochemical products that were mass transferred into the gaseous phase (bubbles). The increase of
j enlarged the fractional coverage of the adhering bubbles, which would invigorate the transport of dissolved species to the bubbles to a certain extent [
2]. Likewise, the lower surface coverage by bubbles could favor local supersaturation of dissolved gases [
19]. Analogous nonlinear increases of CE in response to
j were reported previously by theoretical and experimental assessments [
18,
19].
Ti has typically been employed as cathode material mainly for cost-effectiveness [
17]. In this study, the Ti cathode further showed an excellent selectivity towards hydrogen evolution reaction (HER), due to poor electrocatalytic activities for the side reactions represented by oxygen reduction reaction (ORR). The type (2 versus 4 electrons transfer) and kinetics of ORR, which determine the consumed charge by ORR, would depend on the dissolved oxygen concentration and the cathode materials. For example, the CE of H
2 bubble generation Pt [
18,
20] was less than 50% due to the active ORR. Additionally, depending on wastewater composition, reduction of nitrate or chlorate could be expected as well, particularly on Sn and Cu cathodes [
21]. The lack of electrocatalytic activity of Ti, however, would elevate the cell voltage more severely as
j increases due to the greater ohmic drop and charge transfer resistance.
The CE for O
2 bubble generation was observed to be much lower (~0.4) than H
2, resulting in nonstoichiometric water splitting (H
2/O
2 > 2). The surface coverage of O
2 on IrTaO
x anode would be by far lower than that of H
2 on Ti cathode as expected by the intrinsic stoichiometry of water splitting (H
2:O
2 = 2:1) and aqueous solubility (1.39 mmol O
2 and 0.81 mmol H
2 per kg of water at 20 °C). Therefore, a supersaturation of O
2 could be preferred over the growth of O
2 bubble nuclei. The observed inferior CE for O
2 bubbles could be further rationalized by the competing chlorine evolution reaction (ClER). The CE for OER and ClER would depend on Cl
− concentration, overpotentials of the electrode materials [
4]. Because the surface-bound reactive oxygen species are known to be the common intermediates both for OER and ClER in circum-neutral pH, the bond strength of O atoms to metal cations could shift the CE of ClER in volcano-type relations [
16]. The rate of OER (charge transfer limited) would be proportional to
j, whereas the diffusion of Cl
− could rate-determine the ClER at relatively low chloride concentrations as in this study. Therefore, the interferences of ClER on anode and potentially chlorine reduction reaction on cathode could be invigorated as the chloride concentration increases and/or
j decreases since the lowered surface coverage by bubbles could enhance the diffusion of dissolved species.
The contributions of O
2 bubbles were neglected in this study, primarily because the measured volume of O
2 bubbles counts for only 20% of the H
2 bubbles. The fraction of O
2 bubbles could be even lower during the separation of bio-solids due to the uptake of O
2 by aerobic microorganisms [
14]. In addition, the O
2 bubbles could be larger than H
2 bubbles [
22] due to the lower generation rate (
vide infra), resulting in inferior collection efficiency. From an engineering point of view, since the density of H
2 (0.09 g/L at 25 °C) is far smaller than O
2 (1.43 g/L), the G/S ratio on a mass basis would be biased significantly by a minor fraction of O
2 in the bubble mixture. As long as the CE of bubble generation remains virtually constant (at
j exceeding 8 mA cm
−2 in this study), the G/S ratio could be controlled linearly by
j, to be an important advantage of EF in practice [
11,
12]. In DAF, the major control factor of the A/S ratio is the pressure of the saturated water [
11,
12]. However, the pressure also affects the air solubility, which makes the control of the air generation rate rather complicated [
11].
2.2. Size Distributions for Electrolytic Bubbles and Activated Sludge Flocs
The operational
j and type of electrolyte marginally affected the number distributions of electrolytic bubbles, with superimposable mean diameters in the range of 34–35 μm (
Figure 3a). The highest fraction (30–35%) was marked by 40 μm diameter, and the bubbles sized from 15 μm to 40 μm occupied the dominant fraction (>75%). In contrast, bubbles smaller than 10 μm were observed with a minor fraction (<4%). The variable bubble sizes in EF could be understood in terms of competing growth (
via mass transfer of dissolved products to the dangling bubbles) and detachment from electrode surface [
22]. The electrostatic interaction between bubbles and electrode surface could be in-turn influenced by
j and ionic strength of electrolytes [
1,
22]. For instance, an augmented
j tends to minify the bubbles by invigorated repulsion of negatively charged bubbles from cathode and sweeping by rising bubbles. Likewise, an elevated electrolyte concentration could decrease the bubble size by reduction in electrical double layer thickness [
22]. Nevertheless, the effects of electrolyte conductivity and
j were marginal in the operational range of this study, as shown in
Figure 3a. Accordingly, a uniform bubble diameter (
db) of 35 μm was assumed in this study for the following theoretical calculations.
The mixed liquor samples showed number distributions of flocs characterized by a distinct peak at ~19 μm with a broad tail up to <200 μm. Primary/colloidal particles sized less than 10 μm and macro-particles bigger than 200 μm occupied tiny fractions. The well-established structure of activated sludge flocs indicated that an excessive growth of filamentous bacteria could enlarge the floc size [
2]. In order to simplify the calculations, the observed size distribution of bio-particles was discretized into five groups with mean floc diameter (
df) of 22.5, 40, 60, 135, and 150 µm, respectively, that fell within the range reported in the literature [
23]. Each group accounted for about 5, 12, 15, 50, and 18% of the total mass of mixed liquor suspended solids (MLSS), respectively.
2.3. Collision-Attachment Efficiency Estimated by Batch Flotation Experiments
The DAF with the injection of pressurized water intrinsically required divided contact and separation zones. In contrast, a continuous bubble generation in EF could allow collision/attachment and separation in single compartment under quiescent conditions. Therefore, the separation zone model widely accepted in DAF [
24], based on the rise velocity of bubble-floc agglomerates (
Vbf), could be employed to predict the overall separation efficiency in EF. The white water bubble blanket model [
24] confirmed that the retention time in EF (23–46 min) could be sufficient for collision/attachment of bubble as a single collector for the bio-floc.
To this end, the efficiency of collision-attachment between flocs and bubbles (α) should be one of the most important parameters in EF. The α is known to be affected by surface properties, such as hydrophobicity and charge density, both for bubbles and particles. Therefore, the range of α for biosolids in EF should be differentiated from more widely known values in DAF [
24,
25,
26]. Accordingly, in this study, the α
0 (α for intact floc without an attached bubble) was estimated from a series of batch flotation experiments. In the absence of downward flow, attachment of a singular bubble was assumed to allow the floating and separation of a floc; i.e.,
Vbf > 0 in Equation (5). We further assumed identical α
0 irrespective of
df, while attachment of multiple bubbles to -single floc was ruled out. The α value also decreases sharply as the number of attached bubbles (
Bn) increases (Equation (3)), since the adhered bubbles hinder the subsequent attachments [
26]. Although the flocs with
df of 135 and 150 µm required multiple bubble attachments (
Bn of 3 and 4, respectively) to float, they accounted for only <4% of the total number of bio-solids. Consequently, the α
0 value could be estimated from the number ratio of bio-particles to the total generated bubbles during the separation time, defined as the electrolysis duration to reach a clear solid/liquid interface [
2].
Under variable
j and initial MLSS concentrations, the total passed charge for separation of unit mass of bio-solids was indeed superimposable (
Figure 4), averaged to 44.7 C/gMLSS. As shown in
Table 1, the total number of flocs per unit mass was computed based on the measured number fraction for each discretized particle group (
Figure 3b) with assumptions of spherical particles with identical
ρf (1.04 g/cm
3) [
23]. The passed charge estimated the total number of generated H
2 bubbles, utilizing measured efficiency (
Figure 2) and mean bubble size (
db = 35 μm,
Figure 3a). Accordingly, the α
0 value was computed to be 0.057 (
Table 1). This estimate was much smaller than the reported values in DAF (up to 0.4) [
24,
25,
26]. The activated sludge flocs are known to have negatively charged surface due to extracellular polymeric substances with developed hydroxyl- and carboxyl-functional groups. To this end, the
α value could be significantly changed by the surface charge (represented by zeta-potential) of bubbles [
25]. The surface of H
2 bubbles, generated from electrostatic repulsion from the cathode in EF are expected to be more negatively charged to give lower attachment efficiency with the bio-particles compared to the air bubbles in DAF.
2.4. Limiting G/S Ratio under Variable Floc Sizes and Hydraulic Loadings
The
Vbf should primarily determine the clarification efficiency for continuous flotation in flow reactors; i.e.,
Vbf should exceed the hydraulic loading (
v) or downflow rate for an effective separation. The size of bubbles represented by
db significantly influences the flotation efficiency [
8]. On one hand, the larger bubbles are the more advantageous with respect to
Vbf at a given
Bn [
25,
26,
27,
28,
29,
30]. A body of literature reported that EF could generate finer bubbles than DAF that produces bubbles with a wider distribution of size even exceeding 200 μm [
1,
30,
31,
32]. Thus, there have been claims that the EF should be limitedly deployed under relatively low
v [
26,
30]. On the other hand, a decrease in
db could substantially elevate the
Bm (the maximum
Bn on a floc with size
df), leading to a net increase in
Vbf under a sufficiently large G/S ratio [
25,
27]. Therefore, the linear control of G/S ratio by
j should be underscored as an important strength of EF, whereas an adjustment of
db is practically infeasible.
Figure 5a clearly depicts that the
Vbf increased nonlinearly with
Bn. The minimum
Bn (
Bnmin in integer form) to overcome the
v could be computed for variable
df as in
Figure 5b. As readily expected, the
Bnmin needs to be raised as
v increases, more sharply for the larger bio-particles. Subsequently, the
Bnmin was translated to total required numbers of bubbles (
nb) and limiting G/S ratio under variable values of
v (
Figure 6), by sequential multiplications by 1/α using Equation (3) with plugging in the estimated α
0 value. As shown in
Figure 5b and
Figure 6, the sludge flocs with
df of 22.5 μm and 40 μm could be captured only under
v below 3 and 8 cm/min due to the limited
Bm of 1.3 and 4.1, respectively. Therefore, the hydraulic loading on the flotation unit should be lowered as the target particle size decreases [
24]. The limiting G/S ratio was negligibly altered by the
v for
df of 22.5 μm (since the only available
Bn was 1) and 135–150 μm (since the floc mass far outweighed the H
2 bubble mass). In contrast, the limiting G/S ratio for solids with
df of 40 μm was the most sensitive to the variations of
v. Consequently, the G/S ratio would be the more effective parameter as the particle size approaches the bubble size (
df/
db~1).
2.5. Limiting G/S Ratio for Continous Clarification of Mixed Liquor
The clarification of activated sludge was assessed in continuous EF experiments under varying G/S ratios, based on effluent suspended solids (SS) concentration (
Figure 7). The upper sludge bed quickly grew up to monotonically descend the solids/liquid interface. The SS in subnatant was comprised of the influent particles and (partly) the solids detached from the top sludge blanket. The bubbles rising from the electrode module at the bottom captured the suspended flocs to prevent them from escaping through the outlet beneath the electrodes.
Figure 7 corroborated the existence of limiting G/S ratios above which the effluent SS concentration was sharply alleviated. A further increase over the limiting G/S ratio insignificantly improved the separation efficiency. The influent MLSS concentration also marginally influenced the clarification efficiency [
2].
Table 2 illustrates the detailed procedure to estimate the limiting G/S ratio (at
v of 0.87 cm/min as an example) for separation of mixed liquor with the array of particles. In brief, the
nb value for each group of flocs was weighted by the estimated number of flocs in unit mass (1 g) of MLSS to determine the total required mass of H
2 bubbles (G/S ratio).
The theoretical estimates of limiting G/S ratio (5.23 × 10
−4 and 5.92 × 10
−4 at 0.87 and 1.74 cm/min, respectively) were in general agreement with the experimental observations, as confirmed in
Figure 7. The
v doubled from 0.87 to 1.74 cm/min actually necessitated similar limiting G/S ratio, as rationalized by
Figure 6. These values were far lower than our previous experimental estimates (7 × 10
−3–1 × 10
−2) that included the contributions of O
2 bubbles [
2]. Although the density of bubble is ignorable compared to water and marginally influences
Vbf, as inferred from Equation (5), the composition of bubble mixture (fraction of O
2) in EF could significantly change the G/S ratio. This study thus claims that neglecting the O
2 bubbles (the minor constituent) would provide more consistent guidelines for the G/S ratio in practical applications. Considering the density of air (1.23 g/L) and H
2, on the other hand, the above criteria correspond to (7.1–8.1) × 10
−3 of A/S ratio, being comparable with the lower limit of the reported range (5 × 10
−3–4 × 10
−2) for sludge clarification by DAF [
11]. Generation of finer bubbles in EF could bring about more efficient separation efficiency than in DAF [
2], although the specific energy consumption and installation cost should be considered for a fair comparison.
A comparison between
Bnmin and
Bm could also preconceive the maximum separation efficiency at a specific
v. For example, it is impossible to separate the particles with
Bnmin exceeding
Bm, and the mass fraction of these flocs could be translated to the effluent SS concentration. In this study, the
Bnmin was always smaller than
Bm in both loading conditions to theoretically allow a perfect separation. However, the observed effluent SS concentration was as high as 40 mg/L with clarification efficiency > 98%. These nonideal behaviors would be partly attributed to a turbulence caused by the incoming flow and the rising bubbles. Our previous report also noted the disruption of an excessively grown sludge bed, especially as the G/S ratio increased [
2]. Thus, both contact and separation in a single compartment EF unit might limit the maximum operational range for
v and G/S ratio [
26] to avoid turbulence. The assumptions of this study could underestimate the limiting G/S ratio, since the bubble-floc agglomerates with
Vbf greater than
v could still consume additional bubbles during flotation. As the
α value diminishes, in addition, interactions not included in the model (e.g., bubble formation at floc surface and/or bubble entrapment) could be more significant [
31].