A Comprehensive Dataset on the Immersion Freezing Behavior of Cellulose Particles

We present the results from a laboratory study of immersion mode freezing efficiencies of different types of cellulose particles in simulated supercooled clouds as a function of temperature, T. Two types of chemically homogeneous cellulose samples used as surrogates for ambient predominantly supermicron (giant hereafter) and submicron (~100 nm) ice-nucleating plant structural polymers are micro-crystalline/fibrous cellulose (MCC/FC, Aldrich/Sigma) and nano-crystalline cellulose (NCC, Melodea). Overall, our dataset includes data from twenty different ice nucleation measurement methods that avail at seventeen different institutions around the globe. Experiments with eleven instruments started with the test samples pre-suspended in solvent before cooling, whereas nine other instruments employed water vapor condensation onto dispersed or atomized particles followed by immersion freezing. Without an exception, in all relevant techniques, giant cellulose particles (MCC and FC) act as more efficient ice-nucleating particles than NCC with ~an order of magnitude higher ice nucleation active surface site density, n_s. These particulate samples possess different surface properties (e.g., linear trough density, porosity and size distribution) that may result in different interactions with water vapor and/or super-cooled water droplets and presumably trigger different ice nucleation behaviour. More importantly, we perform systematic accuracy and precision analysis of all twenty instruments by evaluating T-binned (i.e., 1 °C bins) n_s data from each technique over a wide T range of -36 °C< T < -4 °C. While our preliminary inter-comparison results exhibit a reasonable agreement for NCC (on average within T of ~3 °C across the examined T range), T diversities among giant cellulose measurements are extended up to 9 °C. The observed discrepancy of giant cellulose particles is similar to that of chemically heterogeneous illite NX sample described in our previous inter-comparison study (within 8 °C, Hiranuma et al., 2015, ACP), suggesting the compositional heterogeneity may not play a substantial role to explain the diversity. Likewise, the overall difference derived from comparing techniques is significant when compared to the individual uncertainties of each instrument. If it were assumed that all instruments might be reasonably precise, observed deviations could arise from a number of sources, at least for giant cellulose particles, such as solvent, concentration, droplet volume, suspension treatment, ice detection method, aerosolization method and impactor type. There are a number of reasons why given parameters may affect the observed diversity, and we will discuss those in-depth. In addition, the results of complementary physico-chemical analyses via X-ray diffraction technique, scanning electron microscopy (<5keV), transmission electron tomography and single particle mass spectrometry on both cellulose types will also be presented to understand how submicron-sized surface crevices influence cellulose ice activation through droplet freezing. Last but not least, we will also discuss atmospheric relevancy of MCC-alike particles. The data of the ML-CIRRUS aircraft campaign in 2014 over Central Europe suggest that their number fraction is about 1% of all particles between 200 and 900 nm diameter and distributed over the whole tropopause. Our dataset is of special value to the atmospheric modeling community, as it provides experimental data derived by examining multiple experimental parameters. By clearly pointing towards the limits and to possible sources for errors, such a dataset is crucial to improve atmospheric models of cloud feedbacks.

Acknowledgement: We acknowledge support by German Research Society (DFG) and Ice Nuclei research UnIT (FOR 1525 INUIT).