The SAPOS facility consists of a test vessel, gas supply system, and aerosol generation and measurement systems (Figure 1). The test vessel is a cylindrical pressure vessel 5 m high and 1 m in diameter. The pool comprises resistance temperature detectors, located at eight different levels, with three detectors placed 120° from each other at every level. A downward vertical vent pipe with six holes (1 cm in diameter) is used as the multihole injector. The six holes are divided into two groups and are symmetrically distributed on the pipe. The bottom of the multihole injector is 50 cm above the bottom of the vessel.

The gas supply system consists of a nitrogen supply system and a steam generator. A mass flow meter is used to measure the flow rate of nitrogen and steam. The nitrogen supply system provides two gas supplies, one for the carrier gas of the aerosol generator and the other for the noncondensable gas in the experiment, which is directly passed into the mixer. Electric heat tracing is arranged on the steam pipe and nitrogen pipe. The steam leaves the steam generator at 180°C and 7 bar and is superheated at the outlet to 180°C for the injecting mixer.

TiO₂ particles are used to stimulate an experimental aerosol generated by the aerosol generator RBG 2000 SD (Palas GmbH). The aerosol generator disperses the particles using pure nitrogen as the carrier gas, and the powder is conveyed onto a rotating brush at a precisely controlled feed rate and pulled out of the brush by the nitrogen flow. When the aerosol powder flow rate is precise, the aerosol generator produces stable and repeatable particle diameter distributions.

The nitrogen, steam, and aerosol are combined in a mixer. The aerosol mixture is then passed into the pool through a submerged pipe.

The temperature, pressure, and flow of the carrier gas at the inlet and outlet and the water temperature, liquid level, concentration, and particle diameter of the aerosol at the inlet and outlet are measured and recorded. An aerosol spectrometer (Promo 3000 HP) with two probes is attached to the inlet and outlet of the pool to continuously monitor the aerosol concentrations and size distribution. The measured particle diameter is 0.15–10 μm, and the measured concentration is 10⁶ P/cm³. The inlet aerosol mass median diameter (MMD) measured by using the spectrometer is 0.5 μm.

The average concentration is used to calculate the DF because of the steady inlet flow rate. The average DF is expressed as follows: D F = S i n S o u t = ∫ Q i n ( t ) C m − i n ( t ) d t ∫ Q o u t ( t ) C m − o u t ( t ) d t ≈ Q i n ¯ C m − i n ¯ Q o u t ¯ C m − o u t ¯ , (1) where S _in and S _out represent aerosol mass in and out of the test vessel, respectively. C _m-in and C _m-out denote the inlet and outlet aerosol mass concentrations, respectively. Q _in and Q _out represent the inlet and outlet volume flow, which can be calculated by the mass flow rate and carrier gas density. When S _in and S _out are the mass concentrations of aerosol for a particular particle diameter, the obtained value represents the DF of the pool for the aerosol of this particle diameter.

The function f (x₁,x₂,. . ., x_n) is calculated from the measured quantities x₁, x₂,. . ., x_n with uncertainties σ_x1, σ_x2, . . ., σ_xn. The total uncertainty, σ_y, is calculated by the error transmission formula proposed by Moffat, 1988: σ y = ( δ f δ x 1 ) 2 σ x 1 2 + ( δ f δ x 2 ) 2 σ x 2 2 + ⋯ + ( δ f δ x n ) 2 σ x n 2 = ∑ j = 1 n δ f δ x i σ x i 2 . (2)

The DF is calculated as shown in Eq. 1, and the detailed uncertainty calculation for the DF is expressed in the following equations: σ D F = ( δ D F δ Q i n ⋅ σ Q i n ) 2 + ( δ D F δ C m − i n ⋅ σ C m − i n ) 2 + ( δ D F δ Q o u t ⋅ σ Q o u t ) 2 + ( δ D F δ C m − o u t ⋅ σ C m − o u t ) 2 . (3) σ D F = ( D F Q i n ⋅ σ Q i n ) 2 + ( D F C m − i n ⋅ σ C m − i n ) 2 + ( D F Q o u t ⋅ σ Q o u t ) 2 + ( D F C m − o u t ⋅ σ C m − o u t ) 2 . (4)

The TiO₂ concentration at the inlet and outlet is measured after calibration of the aerosol spectrometer with the membrane filter measurements. The stainless-steel filter holder into which the PTFE filter (0.1 μm) is installed is at the back end of the measuring probe of the aerosol spectrometer, and the flow rate of 5 L/min is controlled by the mass flow controller. The calibration results show that the measurement error of the aerosol mass concentration of aerosol spectrometer is approximately 20%.

The inlet and outlet volume flow are calculated using Eq. 5, and the uncertainty calculations for the volume flow are expressed in the following equations: Q = Q m ρ = Q m R T M P , (5) σ Q = ( δ Q δ Q m ⋅ σ Q m ) 2 + ( δ Q δ T ⋅ σ T ) 2 + ( δ Q δ P ⋅ σ p ) 2 , (6) σ Q = ( R T M P ⋅ σ Q m ) 2 + ( Q m R M P ⋅ σ T ) 2 + ( Q m R T M P 2 ⋅ σ p ) 2 , (7) where Q represents the volume flow, Q _m is the mass flow with an accuracy of 1%, T and P denote the temperature and pressure of the fluid, and the error is ±1°C and ±1.0% full scale, respectively. R is the ideal gas constant. M is the molar mass fraction.

In addition, a repetitive experiment was carried out to check the experimental uncertainty. As shown in Table 1, SF1 and SF1-R were carried out under the same experimental conditions. The average experimental DFs (averaged over the whole experiment) of two results are 39 and 42, respectively. As shown in Figure 2, the DFs of the two experiments are consistent for different particle diameters.

In SF1, the steam mass fraction is 0.8, pool depth is 1.2 m, initial temperature of the pool is 50°C, and the total mass flow of nitrogen and steam at the inlet is 70 kg/h. The experiment lasted for 4,200 s from the time the aerosol entered the pool. The nitrogen flow and steam flow at the inlet are stable at 56.30 kg/h and 14.07 kg/h, respectively, and the average value of the outlet flow is 16.44 kg/h, as shown in Figure 3. This indicates that the outlet contains a small amount of steam. The injection gas temperature is maintained at approximately 170°C, as shown in Figure 4. The average injection pressure is 32.71 kPa (g), and there is a slight increase in pressure at the inlet, which is primarily caused by the rise of pressure above the vessel.

Figure 5 shows the variation in the aerosol mass concentration at the inlet and outlet, which indicates that the aerosol concentration at the outlet decreases slightly when the aerosol feed is halted at 3,100 s. This also shows the accumulation of aerosol above the water surface. Nitrogen and steam continue to be injected until the concentration of aerosol at the outlet reaches below 500 P/m³, and then, the experiment stops. The average mass concentrations of the inlet and outlet are calculated to be 13.27 mg/m³ and 1.73 mg/m³, respectively. Eq 1 is used to obtain the average experimental DF of 39 ± 14, along with a 95% confidence interval.

Figure 6 demonstrates the size distribution of aerosol for the tests, depicting a log-normal distribution, and the MMD and geometric standard deviation (GSD) are 0.50 μm and 1.255, respectively. The particle diameter of aerosols is primarily between 0.2 and 1 μm. The larger particles are retained during pool scrubbing. The MMD and GSD of the aerosol at the outlet are 0.43 μm and 1.225, respectively, which are lower than those at the inlet. Owing to the high resolution of the Promo 3000 HP, the particle-diameter–dependent DF for each experiment can be obtained. Such information is essential in determining the range at which the various aerosol removal mechanisms become dominant. For SF1, the DFs of 11 different particle diameters are shown in Figure 7.

The effect of different steam mass fractions on average DFs of aerosol is investigated in experiments SF1, SF2, SF3, and SF4. As shown in Table 1, the average DF increases from eight at a steam mass fraction of 0.5 to 197 at a steam mass fraction of 0.9 with the same MMD of 0.5 μm (the retention efficiency increases from 87.5 to 99.5%).

The DFs versus different steam mass fractions at different particle diameters are shown in Figure 8. The higher the steam fraction, the greater the influence of the steam condensation mechanism on aerosol retention and the larger the DF. The steam condensation effect leads to the destruction of the uniformity of gas mixing in the carrier gas, and the steam is constantly migrated to the gas–liquid interface, driving the aerosol to migrate to the gas–liquid interface and strengthening the removal of aerosols. The higher the steam mass fraction, the stronger the heat transfer between the carrier gas and the pool, the more the aerosols are carried to the gas–liquid interface by steam, and the stronger the enhancement effect of the steam condensation mechanism on DF. In addition, as the steam mass increases, the amount of aerosol deposition increases due to the condensation. The results of data analysis show that log(1/(1−X_m)) has a linear relationship with log (DF).

The effect of different pool depths (1.2, 1.8, 2.4, and 3.0 m) on the average DFs of aerosol is investigated by fixing the carrier gas flow rate at 70 kg/h. It can be seen from Table 1 that the average DFs increased from 15 at a pool depth of 1.2 m to 260 at a submerged depth of 3.0 m with the same MMD of 0.5 μm and the same steam mass fraction of 0.64 (the retention efficiency increases from 93.3 to 99.6%).

The DFs versus different pool depths at different particle diameters are showed in Figure 9, which shows a linear relationship between log (DF) and pool depth that means the DF increases exponentially with the growth of pool depth. The larger the particle diameter, the greater the slope of the growth curve, which indicates that the pool depth has a greater influence on the retention of aerosols with a larger particle diameter (0.4–0.7 μm). The higher the pool depth, the more the time it takes for the carrier gas to escape from the pool and the longer the inertial impact and gravitational settling mechanisms act, which leads to a greater possibility of the large particles being trapped.

Figure 10 shows the DFs of different particle diameters at different pool depths (1.2, 1.8, 2.4, and 3.0 m). The DF increases with the growth of particle diameters at all pool depths. This is because of the effect of inertial impaction and gravity settling caused by the particle diameter during injection and rise regions. In the case of particle diameter less than 0.52 μm, the DF hardly changes, while for particle diameter greater than 0.52 μm, the DF increases rapidly, which reveals that the influence of inertial impaction and gravity settling on the DF is dominant for large particles (＞0.52 μm). For particle diameters in the range of 0.2–0.52 μm, the particle diameter has little effect on the DF.

At low pool depth, most particles are removed in the injection region due to steam condensation. It can be seen from the experimental results that for a low-depth pool scrubbing, the particle diameter has little effect on the DF.

In the injection region, condensation of steam occurs around the injection point, and aerosol is removed by condensed steam in the pool water. A simple model can be used to represent the steam condensation removal effect. The SPARC code assumed that the DF caused by steam condensation equals the fractional loss in gas volume caused by condensation, and the aerosol concentration in the carrier gas is uniform throughout the scrubbing process (Owczarski et al., 1985).

As found by these tests, the estimation of particle retention due to steam condensation in this way alone is insufficient. Steam condensation in a pool occurs at the gas–liquid interface between the bubble and water. Steam condensation on the surface of the bubble causes the steam in the bubble to flow continuously to the gas–liquid interface, and the noncondensable gas is squeezed to the center of the bubble, which leads to the formation of the density and temperature gradients in the bubble. Under the influence of temperature gradient, density gradient, and gas flow in the bubble, particles are constantly moved to the gas–liquid interface, which leads to the inhomogeneity of aerosol concentration in the bubble. The higher the steam mass fraction, the more intense the steam condensation and the greater the inhomogeneity of the aerosol concentration in the bubble. Steam condensation intensifies the process of aerosol migration into the pool.

A new formula for DF_CD is given as follows: DF CD = f c ( 1 − X s,eq 1 − X s ) n , (8) X s , e q = p s p 0 + ρ w g h , (9) where X _s is the mole fraction of steam in inlet gas and X _s,eq is the mole fraction of steam after it attains thermal and steam equilibrium in the pool at the inlet depth. The constant n is introduced to increase the influence of the steam mass fraction according to the experiment results.

Aerosols are retained not only by the steam condensation mechanism but also by the inertial impingement mechanism. This mechanism can be represented by the Stokes number Stk, the function of which is the DF caused by inertial collision (Escudero et al., 1995): S t k = ρ p v i n d p 2 18 μ g D o , (10) where ρ _p is the aerosol density; ν _in is the gas injection rate for a single hole; d _p is the particle diameter; μ _g is the aerodynamic viscosity; and D ₀ is the aperture of the injector.

In the bubble rising process, the aerosol is gradually transferred from the bubble to the water owing to gravity settling, centrifugal deposition, and Brownian diffusion (Owczarski et al., 1985). For these mechanisms, the corresponding deposition velocities, v g , v c , and v d , are now defined.

Gravity deposition refers to the deposition behavior of particles in a bubble owing to its own gravity, which has a significant effect on particles larger than 1 μm in diameter. In addition, the final free settling velocity of spherical particles can be obtained by applying the Stokes law and introducing the Cunningham slip correction coefficient C _c (Escudero et al., 1995): v g = ρ p g d p 2 C c 18 μ g , (11) C c = 1 + 2.492 l d p + 0.84 l d p exp ( − 0.435 d p l ) , (12) l = μ g π 2 p B ρ g , (13) where l is the mean free path of gas molecules and P _B is the pressure in bubbles.

When the bubble rises in the pool, it performs a relative motion with the surrounding liquid. The viscous shearing action of the liquid makes the surface and interior of the bubble rotate frequently. Particles in the bubble are captured because of the centrifugal force acting on the bubble interface; the centrifugal deposition velocity can be expressed as follows (Owcazrski and Burk, 1991): v c = v s 2 v g r c g , (14) where v s is the tangential velocity of the bubble surface, which is closely related to the shape of the bubble and the relative velocity of the rising bubble. r c is the radius of the surface curvature of the bubble.

Based on the permeation theory of the mass transfer process and considering the possible influence of gas flow on the bubble interface during rising, the particle velocity caused by Brownian diffusion can be estimated by introducing a correction factor (Owcazrski and Burk, 1991): v d = ξ D π t e , (15) D = k T C c 3 π μ g d p , (16) where k is Boltzmann’s constant; T is the diffusion ambient temperature gas; and t _e is the temperature inside the bubble.

The product of DFs calculated by different mechanisms is the final cumulative DF (Escudero et al., 1995). From the abovementioned discussion on aerosol deposition modes, the DF for pool scrubbing is expressed as follows: D F = D F C D ⁡ exp ( A × S t k + λ × Δ t ) , (17) λ = 1 V B ∫ ( v c + v d − v g ⋅ cos ⁡ β ) d A , (18) Δ t = h v s w , (19) where λ is the retention efficiency coefficient of particles in the bubble; v sw is the average rising velocity of the bubble group; Δt is the time required for bubbles to rise; h is the submerged depth; V _B is the stable bubble volume; A is the bubble surface area for particle deposition; and β is the included angle between the normal and vertical directions of the bubble surface.

Because the bubble size and its rising velocity were assumed to be constant, Eq. 17 can be simplified as a function of steam mass fraction, particle diameter, and submerged depth. Subsequently, the simplified model expression is given as follows: D F = f c ( 1 − X s , e q 1 − X s ) n e A ∗ d p 2 e ( B d p 2 + C d p - 0.5 ) h . (20)

The coefficients A, B, C, and f _c are obtained from the experimental data, and the result is shown in Eq. 20. D F = 0.37 ( 1 − X s , e q 1 − X s ) 1.7 e 3.7 d p 2 e ( 0.035 d p 2 + 0.86 d p - 0.5 ) h . (21)

Figure 11 shows the correlation between the calculated and measured DFs under the test conditions. This figure reveals that a significant portion of the measured DFs is within −50% or +100% of the calculated DF. Therefore, it can be considered that the calculated results are in good agreement with the experimental results. This simplified model is also used to calculate the values in accordance with the ACE experiment with a multihole injector. The results show that only the calculated DF for the AA4 case (Escudero et al., 1995) with the initial pool temperature of 83°C and the submerged depth of 4.6 m has a large discrepancy from the experimental DF, and this deviation is not included in the range of −50% or +100%. The high pool temperature in the AA4 case reduces the effect of the steam condensation removal mechanism. The influence of the pool temperature on the DF is also being considered a topic for future studies.