The polymers found in plant cell walls represent the largest reserves of renewable carbon on Earth - it has been estimated that the United States has approximately 1.3 billion tons of plant biomass available per year. Ionic liquids (ILs) are salts with melting points less than 100oC, typically containing organic cations, that possess a wide range of extraordinary properties as compared to traditional organic solvents. The unusual catalytic and solvation properties of ionic liquids have attracted the biofuel and biomass communities because IL-pretreatment provides higher fermentable sugar yields than those obtained by acid- or alkali-pretreatment under similar severities. This research used lignin, a major by-product of the biofuel and pulping industries, as raw material for IL synthesis. The goal is to lower the cost of IL production, and enable a more "closed-loop" biorefinery. The lignin is oxidatively depolymerized, purified, and used to prepare a new class of renewable ILs. Biomass pretreatment was the target application for the lignin-derived ILs, while we continue to investigate additional applications within the growing IL market - process chemistry, biotechnology and materials recycling. When comparing our corn stover pretreatment data to recently published reports, we have shown comparable glucose and xylose yields by using neat and 10% aqueous solutions of lignin-derived ILs.

Spiking neural network (SNN) is broadly deployed in neuromorphic devices to emulate brain function. In this context, SNN security becomes important while lacking in-depth investigation. To this end, we target the adversarial attack against SNNs and identify several challenges distinct from the artificial neural network (ANN) attack: 1) current adversarial attack is mainly based on gradient information that presents in a spatiotemporal pattern in SNNs, hard to obtain with conventional backpropagation algorithms; 2) the continuous gradient of the input is incompatible with the binary spiking input during gradient accumulation, hindering the generation of spike-based adversarial examples; and 3) the input gradient can be all-zeros (i.e., vanishing) sometimes due to the zero-dominant derivative of the firing function. Recently, backpropagation through time (BPTT)-inspired learning algorithms are widely introduced into SNNs to improve the performance, which brings the possibility to attack the models accurately given spatiotemporal gradient maps. We propose two approaches to address the above challenges of gradient-input incompatibility and gradient vanishing. Specifically, we design a gradient-to-spike (G2S) converter to convert continuous gradients to ternary ones compatible with spike inputs. Then, we design a restricted spike flipper (RSF) to construct ternary gradients that can randomly flip the spike inputs with a controllable turnover rate, when meeting all-zero gradients. Putting these methods together, we build an adversarial attack methodology for SNNs. Moreover, we analyze the influence of the training loss function and the firing threshold of the penultimate layer on the attack effectiveness. Extensive experiments are conducted to validate our solution. Besides the quantitative analysis of the influence factors, we also compare SNNs and ANNs against adversarial attacks under different attack methods. This work can help reveal what happens in SNN attacks and might stimulate more research on the security of SNN models and neuromorphic devices.

This scale-up study demonstrated the feasibility of an ionic liquid (IL) pretreatment process at 40 kg scale, using the IL 1-ethyl-3-methylimidazolium acetate ([C2C1Im][OAc]) as the solvent. The pretreatment was followed by enzymatic hydrolysis through which the process efficiency for biomass conversion to monomeric sugars was determined. The results show that 43 wt% of switchgrass was dissolved in IL after 2 h of pretreatment at 160 °C with 15 wt% solid loading. A 120 h enzymatic hydrolysis of the pretreated switchgrass results in 96% glucan and 98% xylan conversion. [C2C1Im][OAc] pretreatment has been successfully scaled up to 40 kg with improved sugar titers and yields relative to bench scale (6 kg). The mass flow of the overall process was established and the major scale-up challenges of the process were identified.

As well known, the huge memory and compute costs of both artificial neural networks (ANNs) and spiking neural networks (SNNs) greatly hinder their deployment on edge devices with high efficiency. Model compression has been proposed as a promising technique to improve the running efficiency via parameter and operation reduction, whereas this technique is mainly practiced in ANNs rather than SNNs. It is interesting to answer how much an SNN model can be compressed without compromising its functionality, where two challenges should be addressed: 1) the accuracy of SNNs is usually sensitive to model compression, which requires an accurate compression methodology and 2) the computation of SNNs is event-driven rather than static, which produces an extra compression dimension on dynamic spikes. To this end, we realize a comprehensive SNN compression through three steps. First, we formulate the connection pruning and weight quantization as a constrained optimization problem. Second, we combine spatiotemporal backpropagation (STBP) and alternating direction method of multipliers (ADMMs) to solve the problem with minimum accuracy loss. Third, we further propose activity regularization to reduce the spike events for fewer active operations. These methods can be applied in either a single way for moderate compression or a joint way for aggressive compression. We define several quantitative metrics to evaluate the compression performance for SNNs. Our methodology is validated in pattern recognition tasks over MNIST, N-MNIST, CIFAR10, and CIFAR100 datasets, where extensive comparisons, analyses, and insights are provided. To the best of our knowledge, this is the first work that studies SNN compression in a comprehensive manner by exploiting all compressible components and achieves better results.

Wet anaerobic storage of corn stover can provide a year-round supply of feedstock to biorefineries meanwhile serving an active management approach to reduce the risks associated with fire loss and microbial degradation. Wet logistics systems employ particle size reduction early in the supply chain through field-chopping which removes the dependency on drying corn stover prior to baling, expands the harvest window, and diminishes the biorefinery size reduction requirements. Over two harvest years, in-field forage chopping was capable of reducing over 60% of the corn stover to a particle size of 6 mm or less. Aerobic and anaerobic storage methods were evaluated for wet corn stover in 100 L laboratory reactors. Of the methods evaluated, traditional ensiling resulted in <6% total solid dry matter loss (DML), about five times less than the aerobic storage process and slightly less than half that of the anaerobic modified-Ritter pile method. To further demonstrate the effectiveness of the anaerobic storage, a field demonstration was completed with 272 dry tonnes of corn stover; DML averaged <5% after 6 months. Assessment of sugar release as a result of dilute acid or dilute alkaline pretreatment and subsequent enzymatic hydrolysis suggested that when anaerobic conditions were maintained in storage, sugar release was either similar to or greater than as-harvested material depending on the pretreatment chemistry used. This study demonstrates that wet logistics systems offer practical benefits for commercial corn stover supply, including particle size reduction during harvest, stability in storage, and compatibility with biochemical conversion of carbohydrates for biofuel production. Evaluation of the operational efficiencies and costs is suggested to quantify the potential benefits of a fully-wet biomass supply system to a commercial biorefinery.

Postconsumer absorbent hygiene products (AHPs), one of the most intractable municipal solid waste (MSW) streams, are generally subjected to landfills, causing urban waste management difficulties and environmental pollution. This study investigated the viability of recycling the cellulose component from postconsumer AHPs, demonstrated the fermentability of obtained cellulosic sugars for bioethanol production, and evaluated the technical feasibility of the process through scale-up. A set of integrated unit operations was developed to effectively upcycle the recovered AHPs into bioproducts. The processes were further optimized and scaled up 50-fold. Up to 83.5% glucose and 46.0% xylose yields were achieved during enzymatic saccharification. The sugars were readily fermentable through separate hydrolysis and fermentation (SHF) as well as simultaneous saccharification and fermentation (SSF) with maximum 89.9% ethanol theoretical yield.

Sixteen cellulose rich municipal solid waste (MSW) blends were developed and screened using an acid-assisted ionic liquid (IL) deconstruction process. Corn stover and switchgrass were chosen to represent herbaceous feedstocks; non-recyclable paper (NRP) and grass clippings (GC) collected from households were chosen as MSW candidates given their abundance in municipal waste streams. The most promising MSW blend: corn stover/non-recyclable paper (CS/NRP) at 80/20 ratio was identified in milliliter-scale screening based on the sugar yield, feedstock cost, and availability. A successful scale-up (600-fold) of the IL-acidolysis process on the identified CS/NRP blend has been achieved. The sugar and lignin streams were recovered and characterized. Mass and material energy flows of the optimized process were presented. Feedstock cost for MSW blends was also discussed. Results suggest the promising potential of using MSW as a feedstock blending agent for biorefineries while maintaining sufficient performance and low feedstock cost. The bench scale (6 L) study is an essential step in demonstrating the scalability of this IL technology.

Background

Lignocellulosic biorefineries have tonnage and throughput requirements that must be met year round and there is no single feedstock available in any given region that is capable of meeting the price and availability demands of the biorefineries scheduled for deployment. Significant attention has been historically given to agriculturally derived feedstocks; however, a diverse range of wastes, including municipal solid wastes (MSW), also have the potential to serve as feedstocks for the production of advanced biofuels and have not been extensively studied. In addition, ionic liquid (IL) pretreatment with certain ILs is receiving great interest as a potential process that enables fractionation of a wide range of feedstocks. Acid catalysts have been used previously to hydrolyze polysaccharides into fermentable sugars following IL pretreatment, which could potentially provide a means of liberating fermentable sugars from lignocellulose without the use of costly enzymes. However, successful optimization and scale-up of the one-pot acid-assisted IL deconstruction for further commercialization involve challenges such as reactor compatibility, mixing at high solid loading, sugar recovery, and IL recycling, which have not been effectively resolved during the development stages at bench scale.

Results

Here, we present the successful scale-up demonstration of the acid-assisted IL deconstruction on feedstock blends of municipal solid wastes and agricultural residues (corn stover) by 30-fold, relative to the bench scale (6 vs 0.2 L), at 10% solid loading. By integrating IL pretreatment and acid hydrolysis with subsequent centrifugation and extraction, the sugar and lignin products can be further recovered efficiently. This scale-up development at Advanced Biofuels/Bioproducts Process Demonstration Unit (ABPDU) will leverage the opportunity and synergistic efforts toward developing a cost-effective IL-based deconstruction technology by drastically eliminating enzyme, reducing water usage, and simplifying the downstream sugar/lignin recovery and IL recycling.

Conclusion

Results indicate that MSW blends are viable and valuable resource to consider when assessing biomass availability and affordability for lignocellulosic biorefineries. This scale-up evaluation demonstrates that the acid-assisted IL deconstruction technology can be effectively scaled up to larger operations and the current study established the baseline of scaling parameters for this process.

Achieving low cost and high efficiency lignocellulose deconstruction is a critical step towards widespread adoption of lignocellulosic biofuels. Certain ionic liquid (IL)-based pretreatment processes effectively reduce recalcitrance of lignocellulose to enzymatic degradation but require either costly separations following pretreatment or novel IL compatible processes to mitigate downstream toxicity. Here we demonstrate at benchtop and pilot bioreactor scales a separation-free, intensified process for IL pretreatment, saccharification, and fermentation of sorghum biomass to produce the sesquiterpene bisabolene, a precursor to the renewable diesel and jet fuel bisabolane. The deconstruction process employs the IL cholinium lysinate ([Ch][Lys]), followed by enzymatic saccharification with the commercial enzyme cocktails Cellic CTec2 and HTec2. Glucose yields above 80% and xylose yields above 60% are observed at all scales tested. Unfiltered hydrolysate is fermented directly by Rhodosporidium toruloides-with glucose, xylose, acetate and lactate fully consumed during fermentation at all scales tested. Bisabolene titers improved with scale from 1.3 g L-1 in 30 mL shake flasks to 2.2 g L-1 in 20 L fermentation. The combined process enables conversion of saccharified IL-pretreated biomass directly to advanced biofuels with no separations or washing, minimal additions to facilitate fermentation, no loss of performance due to IL toxicity, and simplified fuel recovery via phase separation. This study is the first to demonstrate a separation-free IL based process for conversion of biomass to an advanced biofuel and is the first to demonstrate full consumption of glucose, xylose, acetate, and lactic acid in the presence of [Ch][Lys].