Search for: All records

Plastic waste has been a major social and environmental problem. Chemical upcycling of plastic waste into transportation fuels or lubricant oil has emerged as an attractive and promising approach to convert plastic waste into valuable products. However, conventional catalyst systems often require high-pressure hydrogen, prolonged residence time, and the use of noble metal catalysts. Here, we present zero-valent nickel single atoms on Mo-based MXenes (Ni/Mo2TiC2Tx) for effective hydrogenation of polyolefins. In situ spectroscopic and microscopic characterizations demonstrate the formation of Ni–Mo and Ni–C bonds at a low Ni loading, resulting in the formation of dispersed intercalated Ni atoms. The catalysts were then employed to convert polyethylene into fuels and lubricant molecules under atmospheric hydrogen pressure, short vapor residence time (τ < 1 s), and mild reaction temperature (300 °C). We further demonstrated the possibility of tuning the Ni structure by changing the Ni loading, attributed to the strong metal–support interactions (MSIs) with the Mo2TiC2Tx support. Under optimized conditions, 22.5 C% gasoline, 12.0 C% diesel, 19.2 C% jet fuel (JP-8), and 37.4 C% lubricant were produced over 0.5% Ni/Mo2TiC2Tx. This work highlights the potential of utilizing MSIs of Mo2TiC2Tx MXenes to synthesize single-atom catalysts (SACs) for upcycling plastic waste into higher-value products. 

This letter proposes a Capacitor Transformer (CT), a non-magnetic, electric-field-coupled converter building block that utilizes a stacked printed circuit board (PCB) structure to achieve high-density voltage conversion. By duality between electric and magnetic circuits, we derive a capacitive turns ratio model and integrate it into a resonant network to facilitate soft switching. A prototype utilizing a multi-layer PTFE-PCB structure was constructed. It achieves a peak DC-DC efficiency of 97.3% and an outstanding volumetric power density of 26.4 kW/dm3, and a competitive weight power density of 15.38 kW/kg, delivering 9.6 kW with a 1000-V input and ~400-V out-put at 1-MHz. These results demonstrate the system-level ad-vantages of proposed capacitor transformer, confirm it as a lightweight, core-less, and highly efficient alternative to magnetic transformers for next-generation data center power delivery. 

Reinforcement Learning from Human Feedback (RLHF) has become the predominant approach for language model (LM) alignment. At its core, RLHF uses a margin-based loss for preference optimization, specifying ideal LM behavior only by the difference between preferred and dispreferred responses. In this paper, we identify a common pitfall of margin-based methods -- the under-specification of ideal LM behavior on preferred and dispreferred responses individually, which leads to two unintended consequences as the margin increases: (1) The probability of dispreferred (e.g., unsafe) responses may increase, resulting in potential safety alignment failures. (2) The probability of preferred responses may decrease, even when those responses are ideal. We demystify the reasons behind these problematic behaviors: margin-based losses couple the change in the preferred probability to the gradient of the dispreferred one, and vice versa, often preventing the preferred probability from increasing while the dispreferred one decreases, and thus causing a synchronized increase or decrease in both probabilities. We term this effect, inherent in margin-based objectives, gradient entanglement. Formally, we derive conditions for general margin-based alignment objectives under which gradient entanglement becomes concerning: the inner product of the gradients of preferred and dispreferred log-probabilities is large relative to the individual gradient norms. We theoretically investigate why such inner products can be large when aligning language models and empirically validate our findings. Empirical implications of our framework extend to explaining important differences in the training dynamics of various preference optimization algorithms, and suggesting potential algorithm designs to mitigate the under-specification issue of margin-based methods and thereby improving language model alignment.