The K-CPEC includes a cloud chamber and an aerosol chamber, as shown in Fig. 1; the functional specifications of each chamber are presented in Table 1. The cloud chamber, which has a double structure, comprising an outer and inner chamber, can be used for conducting experiments on droplet growth and ice crystal formation by aerosol, by controlling the air pressure, wall temperature of the inner chamber, water vapor content, and aerosol concentration (Cha et al., 2024a). In the case of an expansion-type chamber, rapidly evacuating the air from the chamber lowers the air pressure and causes the air to cool adiabatically. However, once the air becomes cooler than the chamber walls, the positive flux from the walls slows the cooling of the air by the adiabatic process (Wagner et al., 2020). To compensate for the heat, the walls of the inner chamber are cooled by a coolant. Therefore, the air temperature inside the inner chamber may decrease due to a quasi-adiabatic expansion process that simultaneously lowers the air pressure and wall temperature, as observed in the case of the double-structure chamber at the MRI (Tajiri et al., 2013). This structure is similar to the MRI chamber, but the inner-chamber volume of the K-CPEC is 22.4 m³, which is 16 times larger than the chamber volume at the MRI.

The outer chamber is made of 304 grade stainless steel (thickness of 22 mm). Insulation (30 mm thick) is attached to the inside wall of the outer chamber; therefore, heat exchange between the cloud chamber and the inside of the K-CPEC facility is minimized. The inner chamber is made from stainless steel (4 mm thick), and the inside wall of the inner chamber has a meandering pattern with copper pipes (29 mm in diameter) passing from the left to right (at intervals of 40 mm) through each panel of the octagonal prism structure (21 segments in total: 2 for the floor, 2 for the ceiling, 9 for the lower walls – including 2 segments forming the front door – and 8 for the upper walls). A thick copper plate (2 mm) was installed between each copper pipe. The copper pipes lowered the temperature of the wall surface comprising the copper pipes and plates as Novec 7200 heat transfer fluid cooled by the R507 HFC refrigerant of the cooling system flows. The heat transfer fluid was continuously circulated at a flow rate of 55 m³ h⁻¹ through the copper pipes of all the walls of the inner chamber (with an approximate residence time of 7 s per wall segment) and the stainless-steel pipes located between the cooling system and the cloud chamber (with an approximate circulation time of 98 s), using the brine supply pump of the cooling system. All the stainless-steel pipes along the refrigerant path are covered with insulation (100 mm thick) to minimize heat loss.

The inner chamber was not completely sealed, allowing air to be evacuated or supplied through gaps, primarily located on the ceiling, which were created for the installation of thermocouples and measurement instruments. The flow rate of the vacuum pump of the cloud chamber was 1300 m³ h⁻¹ at 1800 rpm (maximum). A solenoid valve (SV) was installed at the top of the cloud chamber to control the flow rate of the vacuum pump by adjusting the opening rate. This flow rate control can generate an updraft velocity ranging from 0.1 to 19 m s⁻¹ inside the cloud chamber, based on an initial pressure of 1000 hPa. The dry-air system included a triple filter, capable of removing particles of 10, 5, and 1 µm or larger, and a compressed air filter, capable of removing particles of 0.01 µm or larger, installed in the pipeline to supply clean, dry air to the cloud chamber. Ultra-pure water with a conductivity of less than 0.055 µS cm⁻¹, produced by the pure water system, was supplied to the chamber, along with dry air passed through a Nafion tube, to control the RH inside the inner chamber; furthermore, water vapor was supplied at a flow rate of 300 L min⁻¹. We used a mixing fan to ensure that the air temperature, water vapor, and aerosol were spatially homogeneous throughout the inner chamber (Vallon et al., 2022).

The measurement instruments installed in the cloud chamber to observe aerosol particles and cloud droplets are listed in Table 2, and their measurement principles are summarized in Table 3. During the experiments, the wall and air temperatures (10 cm from the wall) of the inner chamber were measured using fast-response T-type thermocouples installed at the central points of the top, bottom, left, and right walls of the inner chamber (Dias et al., 2017). The relative humidity (RH) inside the inner chamber was calculated using the dew point temperature measured by a chilled mirror hygrometer (Buck Research 1011C), which drew air from the inner chamber for measurement, and the air temperature measured by the thermocouple (Buck Research Instruments, 2009). The aerosol was injected into the inner chamber at a constant rate using a rotating brush-type aerosol generator (Palas RBG1000). A condensation particle counter (CPC) was used to measure the total number concentration of condensation nuclei (CN) greater than 3 nm in size (in the injected aerosol). An optical particle counter (OPC) and a cloud particle imager (CPI) were used to measure both particle and droplet particles. The OPC measures particles in the 0.3–17 µm size range, whereas the CPI detects particles ranging from 10 µm to 2 mm using a high-speed camera (Connolly et al., 2007). Concerning the CPI, the area of the particle captured by the camera was used to calculate the equivalent diameter. In order to minimize the particle loss of hydrometeors in still air, the OPC and CPI were installed underneath the inner chamber with inlets oriented vertically upwards. As the cloud chamber experiment entailed a rapid decrease in pressure, a mass flow controller (MFC) was used to maintain a constant flow rate. In the case of the CPI, the constant mass flow was preserved using a pump speed control system. Note that all the measurement instruments were connected to an external pump to maintain a constant flow rate. All measurement instruments, except for the CPC and CO₂ / H₂O analyzer, sampled the air from the inner chamber and then vented it to the outer chamber after measurement. The LI-COR CO₂ / H₂O analyzer is an open-path instrument that measures the infrared absorption of water vapor in the air along the optical path between the transmitter and receiver, enabling the calculation of absolute humidity.

Unlike the cloud chamber, the aerosol chamber could not control the temperature directly. However, because dry air was supplied during the chamber-cleaning process, the aerosol chamber was slightly cooler than the ambient air and extremely dry (RH < 1 %), with the temperature typically at 20 ± 5 °C. This enabled us to measure properties of the aerosol particles at a dry state. As in the cloud chamber, the aerosols were suspended using a mixing fan to ensure spatial homogeneity. The total number concentration and particle size distribution (PSD) of aerosol particles with sizes of 11 nm–17 µm were measured using an OPC and a scanning mobility particle sizer (SMPS), respectively (Table 2). A cloud condensation nuclei (CCN) counter was used to measure the concentration of the particles that were activated (condensed) under supersaturation conditions. In this study, the CCNs were measured by dividing the supersaturation range (0.1 %–1 %) into intervals of 0.1 % while considering the typical atmospheric supersaturation level (Loftus and Cotton, 2014). To ensure the stability of the CCN measurements, the data of only the last 3 min for each interval were used. Since no experimental calibration of the CCN counter was conducted in this study, the supersaturation values at each interval may carry an uncertainty of up to ±10 % (Rose et al., 2008). The air temperature and RH in the aerosol chamber were measured using the Vaisala HMM170 sensor.

The PSDs of the NaCl and CaCl₂ powders used in this study are shown in Fig. 2. These distributions were merged with those of OPC values measured after the maximum size measurable by the SMPS; that is, the PSDs of 11 nm–0.48 µm were measured using an SMPS, and the PSDs of 0.48–17 µm were measured using an OPC, with each distribution representing 1 h mean values for each size bin. Notably, the data in Fig. 2 were obtained simultaneously during the same experiment as those shown in Fig. 3 for the CPC and CCN counter measurements. The PSDs of both powders exhibited bimodal distribution, and the number concentration peaks in the submicron size were 0.19 and 0.24 µm (for NaCl and CaCl₂, respectively); the number concentration peaks in the micrometer size for both the powders were 1.84 µm. The mean particle size of NaCl was 0.37 µm, smaller than that of CaCl₂ (0.44 µm). The particles with diameters ≥ 1 µm accounted for 8.75 % and 12.31 % of the total number concentrations of NaCl and CaCl₂, respectively; the particles with diameters ≥ 10 µm accounted for 0.19 % and 0.12 %, respectively. In aerosol chamber experiments, high aerosol concentrations combined with prolonged residence times may promote particle coagulation (Chen et al., 2024). However, in this study, the total number concentration during both experiments ranged from 1100 to 1200 cm⁻³, suggesting that the effect of coagulation on the PSD was minimal. Thus, the sizes of the majority of particles were smaller than 1 µm, but the particles were characterized by a wide range of PSDs. In contrast to this result, the submicron size peak of the burn-in-place-type hygroscopic flare, which is widely used for cloud seeding worldwide, was very small (0.1 µm) (Kuo et al., 2024; Miller et al., 2024). As the flare particles were much smaller in size compared to the typical powder-type materials, their activation as CCNs in low-supersaturation conditions was limited; therefore, an immediate seeding effect was difficult to observe in the cloud seeding experiment (Dong et al., 2023).

The Fact and Dact of NaCl and CaCl₂ powders according to the supersaturation level (S) are shown in Fig. 3. The red points represent the CCN data for each supersaturation interval (with the range being 0.1 %–1 %, at intervals of 0.1 %) considered in this study. The Fact and Dact were calculated using the CN data acquired using the CPC, measured simultaneously with the CCN data. In the case of CaCl₂ powder, the Fact was 92.30 % at S of 0.1 %, while the NaCl powder exhibited a relatively low Fact of 79.33 %. This may be because, as shown in Fig. 2, the NaCl powder contained numerous particles that were not large enough to act as CCN at S of 0.1 %. When S was 0.1 %, the Dact of the NaCl powder was 135.8 nm, and that of CaCl₂ powder was 126.3 nm, with a particle size difference of approximately 10 nm. Notably, hygroscopic materials with relatively large particles can have more particles acting as CCNs, even in low supersaturation conditions (Rose et al., 2010). In this study, since Dact was determined using the method proposed by Hung et al. (2014), it may differ from the critical diameter (Dcrit) measured using a DMA (e.g., Dcrit of NaCl = 100 ± 4 nm at S = 0.1 %; Niedermeier et al., 2008). For the NaCl powder, the Fact value at S = 0.2 % was greater than 90 %, and both the powders exhibited Fact values greater than 95 % at S of 0.3 %. As the value of S increased, the Fact values of both powders increased to almost 100 %, and the Dact values were approximately 70 nm, indicating that the particles could act as CCNs up to size of Aitken mode. Note that for S ≥ 0.4 %, the CaCl₂ powder exhibited a consistent Fact value. Thus, the CaCl₂ powder may be suitable for cloud seeding experiments for relatively low supersaturation conditions. The NaCl is considered to be suitable for cloud seeding experiments in conditions of high supersaturation, i.e., near cumulus clouds with strong updrafts or in the vicinity of mountain ranges with strong orographic updrafts (Cotton et al., 2007; Li et al., 2023).

The results of the cloud chamber experiments conducted using NaCl (NaCl Exp. #1 and #2) and CaCl₂ (NaCl Exp. #1 and #2) are shown in Figs. 4 and 5, respectively. The air pressure, Tair, Tdew, Twall, RH, updraft velocity, cooling rate, lapse rate, absolute humidity, size-dependent number concentration (D > 0.3, > 1, > 5, > 10, and > 20 µm), size-resolved number concentration, and mean diameter of droplets in the inner chamber are shown in the figures. The maximum and mean values of the updraft velocity and cooling rate of the cloud chamber for each experiment are shown in Table 7. In each experiment, the time required for the SV rate to increase from 0 % to 20 % and 50 % was 6 and 10 s, respectively. The time required for the vacuum pump to reach an rpm of 1800 (from 0) was 30 s. In the NaCl Exp. #2, the maximum updraft velocity (18.3 m s⁻¹) could be achieved in 55 s after the beginning of the experiment; the mean updraft velocity was 13.5 m s⁻¹. The maximum cooling rate 85 s after the beginning of the experiment was -7.3 K min⁻¹; the mean cooling rate was -2.3 K min⁻¹. These experimental conditions may simulate cloud seeding experiments in areas with weak-to-strong convection and strong orographic updrafts (Jensen et al., 1998; Field et al., 2001). Within the first approximately 150 s of the cloud chamber experiment, the heat transfer fluid circulated through the inner chamber, rapidly lowering the wall temperature before stabilizing. A full circulation of the heat transfer fluid through the cloud chamber and the cooling system took approximately 98 s. This phenomenon occurred temporarily before the RH in the internal chamber reached 100 % and did not significantly affect the variation in air temperature.

In NaCl Exp. #1 and #2, the total number concentrations of the injected powder were 1052.31 and 1128.55 cm⁻³, respectively (Table 6). Immediately after the beginning of the experiment, particles larger than 0.3 µm accounted for approximately 56 % of the total number concentration for an RH of approximately 55 %. This figure was 22 % higher than the proportion of the number concentration (approximately 34 %) for particles larger than 0.3 µm measured in the aerosol chamber experiment with NaCl powder. Initially, the RH remained below the DRH, and thus, the deliquescence transition was not activated. However, owing to the strong hygroscopicity of NaCl, water vapor might have been taken up even below the DRH, leading to pre-deliquescence hygroscopic growth characterized by the formation of a thin water layer (Tang and Munkelwitz, 1993). After RH exceeded 70 % (Exp. #1: 87 s; Exp. #2: 56 s), the deliquescence transition was initiated and continued until RH reached approximately 85 % (Exp. #1: 224 s; Exp. #2: 90 s) (Peng et al., 2022). During this period, particles smaller than 2.84 µm – corresponding to the intermodal size of the bimodal distribution – grew in size, and the number concentration increased by 13.73 % in Exp. #1 and 7.78 % in Exp. #2. In addition, the number concentration of the particles with PSDs of 2.84–17 µm decreased by 25.81 % in Exp. #1 and 3.61 % in Exp. #2, which may be related to large cloud droplets (several tens of micrometers in size, with the captured by the CPI). After RH exceeded 85 % (Exp. #1: 224 s; Exp. #2: 90 s), post-deliquescence hygroscopic growth continued. This period showed the most significant decrease in mean absolute humidity, with values of 0.012 g m⁻³ s⁻¹ in Exp. #1 and 0.016 g m⁻³ s⁻¹ in Exp. #2. When RH exceeded 100 % (Exp. #1: 430 s; Exp. #2: 316 s), cloud droplet formation occurred through condensational growth. During this phase, the mean lapse rate (-dT/dz) was 4.26 K km⁻¹ in Exp. #1 and 2.28 K km⁻¹ in Exp. #2. This finding indicates that Exp. #1 followed the wet adiabatic lapse rate (Γw, 3.5–6.5 K km⁻¹; Weiner and Matthews, 2003), whereas Exp. #2 exhibited a more stable environmental lapse rate, lower than Γw. Notably, cloud droplets that can be formed in an environment where sufficient water vapor is supplied in a clean atmosphere are typically 20–30 µm in size (Li et al., 2017). In comparison, the artificial injection of hygroscopic materials, such as NaCl, can facilitate the growth of initial cloud droplets. Under super-saturated (RH > 100 %) conditions, cloud droplets of 30–50 µm in size were consistently observed not only in NaCl Exp. #1 but also in the other three experiments. These cloud droplets may have formed earlier at the center of the inner chamber; these droplets may have required time to reach the bottom of the inner chamber, where the observations were conducted using OPC and CPI (Frey et al., 2018).

With respect to the both powders used for the cloud seeding experiment, the cloud DSDs (see Figs. 4 and 5) were not constant due to the fluctuations in supersaturation. In addition, as the air in the inner chamber was evacuated using a vacuum pump, the CCNs or droplets in the chamber may have been lost. Furthermore, condensation also occurred on the inner chamber walls, which could cause a rapid loss in supersaturated conditions (Shao et al., 2022). The observation results clearly indicated that the process through which the particles grew after the RH reached 100 % occurred under supersaturated conditions. The growth process of cloud droplets was divided into the under-saturated stage (RH ≤ 85 %, hereinafter referred to as S1), pre-saturated stage (85 % < RH ≤ 100 %, hereinafter referred to as S2), and super-saturated stage (RH > 100 %, hereinafter referred to as S3), as shown in Fig. 6, and the cloud DSD – possibly including both aerosols and droplets – was expressed. These stages were divided based on the variations in the cloud DSDs observed in different RH ranges; for example, in NaCl Exp. #1, the cloud DSD shown in Fig. 4i for 1–223 s exhibits a monomodal distribution with a peak at 0.58 µm, similar to the observations for S1 shown in Fig. 6a. However, once the RH exceeded 85 %, the monomodal distribution transitioned to a bimodal distribution (with a second mode peak at 7.78 µm), similar to the observations for S2 shown in Fig. 6a. This bimodal distribution persisted until RH exceeded 100 %; in this supersaturated stage, the right tail of the bimodal distribution increased relatively, as observed in the S3 shown in Fig. 6a. The maximum cloud droplet size (Dmax⁡) was 65.71 µm (observed at 832 s) in this experiment. In the case of NaCl Exp. #2, the SV opening rate was increased from 20 % to 50 %, resulting in the evacuation speed being nearly twice as fast as in NaCl Exp. #1. Therefore, the process in the S1 shown in Fig. 6b was shortened to 90 s, and the time required to reach 100 % RH was reduced to 316 s, reaching supersaturation more than 100 s faster compared to that observed in NaCl Exp. #1. Furthermore, in the S3 (Fig. 6b), the peak of the second mode was at 26.43 µm, and the Dmax⁡ was 89.16 µm (observed at 335 s). Although the number concentration inside the inner chamber decreased rapidly compared to that in the NaCl Exp. #1, large cloud droplets continued to form and were observed until the end of the experiment. This phenomenon indicated that as the supersaturation condition was maintained until the end of the experiment, the cloud particles below freezing point were observed by the CPI as spherical supercooled water droplets. The maximum mean cloud droplet size (Dmean) in NaCl Exp. #1 and NaCl Exp. #2 was 27.87 and 40.71 µm, respectively (Fig. 4i and j). These large cloud droplets have a size distribution corresponding to the important diameter for droplets acting as drizzle embryos, leading to rain droplets (Zhu et al., 2024).

In CaCl₂ Exp. #1 and #2, CaCl₂ powder was injected into the inner chamber at total number concentrations of 1040.36 and 1071.92 cm⁻³, respectively (Table 6). Immediately after the beginning of the experiment, the number concentration of particles larger than 0.3 µm was approximately 80 % compared with the injected total number concentration under an RH of approximately 58 %. This figure is 27 % higher than the proportion of number concentration (approximately 53 %) for particles larger than 0.3 µm measured in the aerosol chamber experiment with CaCl₂ powder. Owing to the low DRH of CaCl₂ (28 %), a large number of particles might have undergone deliquescence transition immediately upon injecting CaCl₂ powder into the cloud chamber (Guo et al., 2019). Therefore, before the RH reached 85 %, the mean particle size within the size distribution below 2.84 µm was 1.06 µm, which was larger than the mean particle size of 0.94 µm in the NaCl experiments. In addition, the mean number concentration was approximately 692 cm⁻³, which was 45 % higher than that in the NaCl experiments (Xueling et al., 2021; Peng et al., 2022). Therefore, in cloud seeding experiments conducted using an aircraft, CaCl₂ powder can rapidly increase the number concentration of potential CCNs in low RH conditions. The S1 and S2 stages in CaCl₂ Exp. #1 and #2 progressed faster than those in NaCl Exp. #1 and #2.

The hygroscopicity of NaCl is known to be higher than that of CaCl₂ (Kumar et al., 2011). Note that the stronger the hygroscopicity of a material, the more easily the CCNs can uptake the surrounding water vapor; this may cause a hygroscopic buffering effect that delays the increase in RH in the chamber (Ding et al., 2024). For the NaCl powder (with high hygroscopicity), the time required for the RH inside the inner chamber to reach 100 % was delayed by approximately 50 s. Thus, the process of cloud droplet growth differed depending on the deliquescence and hygroscopicity characteristics of the seeding material. The cloud DSDs in the S1–S3 stages in CaCl₂ Exp. #1 and #2 were similar to those observed in NaCl Exp. #1 and #2. The maximum Dmean observed in CaCl₂ Exp. #1 was 24.64 µm, while that in CaCl₂ Exp. #2 was 38.30 µm, indicating a difference of 2–3 µm smaller compared with the mean diameters observed in the NaCl experiments. The Dmax⁡ values of the droplets observed in CaCl₂ Exp. #1 and #2 were 54.99 (observed at 458 s) and 68.42 µm (observed at 548 s), respectively, 10–20 µm smaller than the Dmax⁡ values observed in NaCl Exp. #1 and #2. In this study, the cloud chamber experiments were performed under similar conditions, using NaCl and CaCl₂ powders. However, the difference in droplet growth indicated greater growth in NaCl powder, which exhibited relatively higher hygroscopicity. The mean Dmax⁡ values for S1–S3 stages in each experiment are summarized in Table 8. In the S1 stage, NaCl and CaCl₂ each showed a small difference (approximately 1 µm) between Exp. #1 and #2, with NaCl consistently exhibiting mean Dmax⁡ values that were 4–5 µm larger than those of CaCl₂. In the S2 stage, where hygroscopic growth was active under pre-saturated conditions, cloud droplets showed a larger mean Dmax⁡ by 15–20 µm compared with that at the S1 stage. Although the largest cloud droplets appeared during the S3 stage, when condensational growth was dominant, the mean Dmax⁡ was smaller than that at the S2 stage. This is because although the RH inside the cloud chamber consistently exceeded 100 %, the absolute humidity decreased as the experiment progressed. In other words, these findings suggest that a larger mean Dmax⁡ can be observed in an environment where sufficient water vapor is supplied.