Immersion mode heterogeneous ice nucleation by an illite rich powder representative of atmospheric mineral dust

Atmospheric dust rich in illite is transported globally from arid regions and impacts cloud properties through the nucleation of ice. We present measurements of ice nucleation in water droplets containing known quantities of an illite rich powder under atmospherically relevant conditions. The illite rich powder used here, NX illite, has a similar mineralogical composition to atmospheric mineral dust sampled in remote locations, i.e. dust which has been subject to long range transport, cloud processing and sedimentation. Arizona Test Dust, which is used in other ice nucleation studies as a model atmospheric dust, has a significantly different mineralogical composition and we suggest that NX illite is a better surrogate of natural atmospheric dust. <br><br> Using optical microscopy, heterogeneous nucleation in the immersion mode by NX illite was observed to occur dominantly between 246 K and the homogeneous freezing limit. In general, higher freezing temperatures were observed when larger surface areas of NX illite were present within the drops. Homogenous nucleation was observed to occur in droplets containing low surface areas of NX illite. We show that NX illite exhibits strong particle to particle variability in terms of ice nucleating ability, with ~1 in 10<sup>5</sup> particles dominating ice nucleation when high surface areas were present. In fact, this work suggests that the bulk of atmospheric mineral dust particles may be less efficient at nucleating ice than assumed in current model parameterisations. <br><br> For droplets containing ≤2 × 10<sup>−6</sup> cm<sup>2</sup> of NX illite, freezing temperatures did not noticeably change when the cooling rate was varied by an order of magnitude. The data obtained during cooling experiments (surface area ≤2 × 10<sup>−6</sup> cm<sup>2</sup>) is shown to be inconsistent with the single component stochastic model, but is well described by the singular model (<i>n</i><sub><i>s</i></sub>(236.2 K ≤ <i>T</i> ≤ 247.5 K) = exp(6.53043 × 10<sup>4</sup>− 8.2153088 × 10<sup>2</sup><i>T</i> + 3.446885376<i>T</i><sup>2</sup> − 4.822268 × 10<sup>−3</sup><i>T</i><sup>3</sup>). However, droplets continued to freeze when the temperature was held constant, which is inconsistent with the time independent singular model. We show that this apparent discrepancy can be resolved using a multiple component stochastic model in which it is assumed that there are many types of nucleation sites, each with a unique temperature dependent nucleation coefficient. Cooling rate independence can be achieved with this time dependent model if the nucleation rate coefficients increase very rapidly with decreasing temperature, thus reconciling our measurement of nucleation at constant temperature with the cooling rate independence