Kemira Inc.

Atomistic detail molecular dynamics (MD) simulations were used to examine molecular level dependencies of water interactions and permeability of amphiphilic di-block copolymer assemblies. Four different linear amphiphilic di-block copolymers, with chemically identical hydrophilic blocks but varying hydrophobic blocks, were studied. The simulations show that while all examined di-block copolymers formed assemblies highly efficient in blocking water, the ability of the assemblies to sustain their water blocking character when subject to lateral strain had significant differences. The assemblies retaining high internal order under lateral strain were most efficient in moisture blocking. In this, flexible side chains with branching filling the surroundings with hydrophobic groups to prevent molecular level water penetration, were important. Moreover, styrene rings provided efficient, space-filling moisture protection blocks, demonstrating further the importance of flexible spatial orientation hydrophobic side chains. The simulations also connect the roughness and interfacial fluctuations of the block-copolymer assembly with water penetration. The work shows molecular level design principles for enhancing moisture preventing coatings and gives insight into engineering the performance of other filtration systems.

Process runnability and end-product quality in paper and board making are often connected to chemistry. Typically, monitoring of the chemistry status is based on a few laboratory measurements and a limited number of online specific chemistry-related measurements. Therefore, mill personnel do not have real-time transparency of the chemistry related phenomena, which can cause production instability, including deposition, higher chemical consumption, quality issues in the end-product and runnability problems. Machine learning techniques have been used to establish soft sensor models and to detect abnormalities. Furthermore, these soft sensors prove to be most useful when combined with expert-driven interpretation. This study is aimed at utilizing a hybrid solution comprising chemistry and physics models and machine learning models for stabilizing chemistry-related processes in paper and board production. The principal idea is to combine chemistry/physics models and machine learning models in a fashion close to white box modeling. A cornerstone in the approach is to formulate explanations of the findings from the models; that is, to explain in plain text what the findings mean and how operational changes can mitigate the identified risks. The approach has been demonstrated for several different applications, including deposit control in the wet end, both raw water treatment and usage, and wastewater treatment. This approach provides mill personnel with knowledge of identified phenomena and recommendations on how to stabilize chemistry-related processes. Instead of using close to black box machine learning models, a hybrid solution including chemistry/physics models can enhance the performance of artificial intelligence (AI) deployed systems. A successful way of gaining the trust from mill personnel is by creating a plain text explanation of the findings from the hybrid models. The correlation between the likelihood of a phenomena and disturbance and the explanations are derived and validated by application and chemistry and physics experts.

Anaerobic digestion of organic waste into methane and carbon dioxide (biogas) is carried out by complex microbial communities. Here, we use full-length 16S rRNA gene sequencing of 285 full-scale anaerobic digesters (ADs) to expand our knowledge about diversity and function of the bacteria and archaea in ADs worldwide. The sequences are processed into full-length 16S rRNA amplicon sequence variants (FL-ASVs) and are used to expand the MiDAS 4 database for bacteria and archaea in wastewater treatment systems, creating MiDAS 5. The expansion of the MiDAS database increases the coverage for bacteria and archaea in ADs worldwide, leading to improved genus- and species-level classification. Using MiDAS 5, we carry out an amplicon-based, global-scale microbial community profiling of the sampled ADs using three common sets of primers targeting different regions of the 16S rRNA gene in bacteria and/or archaea. We reveal how environmental conditions and biogeography shape the AD microbiota. We also identify core and conditionally rare or abundant taxa, encompassing 692 genera and 1013 species. These represent 84–99% and 18–61% of the accumulated read abundance, respectively, across samples depending on the amplicon primers used. Finally, we examine the global diversity of functional groups with known importance for the anaerobic digestion process.

Loose nanofiltration (LNF) membranes with a molecular weight cutoff (MWCO) of about 1000 Da have great potential for high selectivity between natural organic matter (NOM) and mineral salts. Therefore, they are interesting for treatment plants for purifying oligotrophic lake waters with an elevated NOM concentration. This study was conducted to determine the design and operational expenses as well as the environmental impact of an LNF-based drinking water treatment process that removes NOM from Finnish surface water. Two LNF membranes with similar MWCOs were selected, and the results were compared to ultrafiltration, nanofiltration, and conventional treatment. One LNF membrane demonstrated a rejection rate of NOM higher than 95 % and a low rejection rate of hardness at about 40 %, while the other LNF membrane performed similarly to an ultrafiltration membrane. The LNF-based treatment process would increase operational expenses by 53—69 % but at the same time decrease the environmental impact by >18 %. Improved removal of aromaticity and low molecular weight compounds could lower the disinfection byproduct formation potential and microbial growth in the network, thus reducing the required chlorine dosage in the network and possibly bringing further cost savings.