Ice nucleation process

Ice nucleation is one of the key factors in ASC phase inversion. By turning off all heterogeneous nucleation processes (OFF in Table 4), no ice is nucleated (Figure 22a, Figure 23a, & Figure 24a), inferring that heterogeneous nucleation processes are crucial to ice formation. Deposition nucleation and immersion freezing are the two more well-known heterogeneous nucleation processes for dust particles to nucleate, and their roles in ice nucleation are discussed in this section.

.4.2.1.1 Deposition nucleation versus immersion freezing

To know how ice particles are nucleated in ASC, sensitivity tests of the ice nucleation process by remaining either deposition nucleation (DE) or immersion freezing (IM) are designed (Table 4).

From the offline calculation of the nucleation rate (Figure 1), the deposition nucleation rate is larger than the immersion freezing rate by about 5-7 orders at the temperature range (-14 ~ -26°C) of consideration. From the liquid and ice distributions pattern in the clouds (Figure 22b,c & Figure 23b,c), DE and IM both have a significant area of liquid and ice presence, but the latter has a larger liquid area, less ice, and has no ice out of the liquid layer above 900 hPa. From the time series of average ice water path (IWP) and LWP in domain 3 (Figure 24), DE is very similar to CTL, while IM has much less IWP and more LWP. All these indicate that deposition nucleation is the main ice nucleation process in this case, implying that no liquid formation is needed before ice formation. Such results are different from previous studies (de Boer et al. 2010; Prenni et al. 2009) that emphasized the role of immersion freezing in ASC.

Different from deposition nucleation, immersion freezing requires liquid to form first, which can be seen in Figure 24 that the value of LWP rises earlier than that of IWP in IM. Although immersion freezing is weaker in this case, IM still produces a small amount of ice (Figure 22c, Figure 23c, & Figure 24a). However, LWP and IWP in DE are nearly the same as those in CTL (Figure 22b, Figure 23b, & Figure 24b), suggesting that immersion freezing is not effective when deposition nucleation is working. The immersion nucleation is limited not only due to its low nucleation rate but also because deposition nucleation was initiated earlier and thus consumed a large portion of dust particles with a fewer amount left for immersing into droplets (Figure 25a & Figure 26a).

In contrast, IM has more dust particles available for immersion in droplets (Figure 25c &

Figure 26c), although the number of ice nucleated is still less than DE. Also, the earlier liquid formation and the longer liquid lifetime in IM comparing with DE and CTL (Figure 24b & c) indicate that the early ice formation by deposition nucleation in DE and CTL delays the liquid formation and shortens the liquid cloud’s lifetime. Both ice growth and immersion freezing compete for liquid. Consequently, when deposition nucleation nucleates ice in advance, the liquid formation would be harder and later, and the liquid layer lifetime would be shorter because more liquid is consumed by the ice WBF growth in advance. As a result, deposition nucleation has a crucial influence on the liquid cloud formation and persistence, as well as the phase inversion structure. Thus, how many dust particles are in the air plays a key role by engaging in deposition nucleation in this case.

.4.2.1.2 Immersion due to activation or collection

How dust particles get into liquid influences not only the rate of immersion freezing but also the amount of remaining in the air. In our microphysical scheme, dust can get

activation process requires that the dust particles have a certain degree of hygroscopicity κ (based on the Köhler theory) or can acquire water by surface adsorption (Sorjamaa and Laaksonen 2007). To know how dust particles go into droplets primarily and how this affects ice nucleation subsequently, sensitivity tests of dust immersion process by turning off the collision between dust and droplets (NOCOL), and raising κ of dust from 0.001 to 0.1125 (K0.11) are conducted (Table 4). The higher κ represents a condition of

“polluted” dust (dust mixed internally with soluble material). The adsorption activation process tends to be relatively weak for dust lager than about 2 μm in radius (roughly the size for efficient ice nucleation) as judged from its higher Köhler-curve critical supersaturation comparing to that for κ = 0.001 above this size (Figure 27); therefore, this mechanism is ignored in this study.

First, we want to know whether collision or activation is the dominant process. From the profiles of dust in cloud liquid (Figure 28) and the time series of average dust in liquid in domain 3 (Figure 29), NOCOL has the least dust in liquid, while K0.11 has the same order of dust in liquid as CTL, implying that collision is dominant in dust immersion process and κ has little effect on the number of activated particles.

Next, the effect of dust in liquid on ice nucleation is investigated. From the time series of average IWP in domain 3 (Figure 30a), IWP is similar to CTL in both NOCOL and K0.11, which indicates that the amount of dust in liquid has only little effect on ice nucleation and implies two things. First, immersion freezing is too small to create significant ice nucleation difference although CTL and K0.11 have more dust in liquid than NOCOL about 2 orders. Second, the effect of immersed dust on ice nucleation takes time to evolve since immersion freezing can happen only after the formation of cloud drops, which then take time to collide with dust. Nevertheless, deposition nucleation has already happened earlier and decreased the available dust for immersion, together with

the small immersion freezing rate, leading to a rather small effect of the dust immersion process on ice nucleation.