The use of artificial intelligence (AI) has become ubiquitous in various fields, from data analysis to image recognition. Its performance has surpassed that of humans in many areas, but its energy consumption is vast and will increase exponentially in the upcoming years. To address this issue, scientists are researching physical systems that could support or partially replace electronic systems for certain tasks.

Artificial neural networks (ANNs) inspired by the brain are one such option. They consist of nodes linked in a complex structure, most commonly implemented using digital connections. However, recent experience with training large language models has made it clear that their energy consumption is substantial. Therefore, scientists are exploring alternative physical systems, including optics and photonics.

Optics and photonics have several advantages over conventional electronic systems, such as high bandwidths and information encoding in high-dimensional symbols. They also allow parallel processing and connection to established systems like the optical fiber-based world-wide internet. When scaling up, photonics holds the promise of lower energy consumption for complex problems.

Researchers at the Stiller Research Group at the Max Planck Institute for the Science of Light (MPL) and Leibniz University Hannover (LUH), in collaboration with Dirk Englund at MIT, have now experimentally demonstrated an all-optically controlled activation function based on traveling sound waves. This development is essential for photonic computing, a physical analog computing alternative that promises to be able to realize energy-efficient AI in the long term.

The nonlinear activation function is crucial for deep learning models to learn complex tasks. In optical neural networks, these parts should ideally be implemented in the photonic domain as well. For the weighted sum – a matrix operator – a plethora of photonic approaches already exist. However, few approaches have been demonstrated experimentally for the nonlinear activation function.

The scientists’ approach uses sound waves as a mediator to introduce nonlinearity into photonic computing systems. The optical information does not need to leave the optical domain and is directly processed in optical fibers or photonic waveguides. Via the effect of stimulated Brillouin scattering, the optical input information undergoes a nonlinear change depending on the level of optical intensity.

“Our photonic activation function can be tuned in a versatile way: we show the implementation of a sigmoid, ReLU, and quadratic function,” says one of the lead authors, Grigorii Slinkov. “An interesting advantage comes from a strict phase-matching rule in stimulated Brillouin scattering: different optical frequencies – for parallel computing – can be addressed individually, which may enhance the computational performance of the neural network.”

Including a photonic activation function in an optical neural network preserves the bandwidth of the optical data, avoids electro-optic conversion, and maintains the coherence of the signal. The versatile control of the nonlinear activation function with the help of sound waves allows the implementation of the scheme in existing optical fiber systems as well as photonic chips.

The long-term prospect of creating more energy-efficient optical neural networks depends on whether it is possible to scale up physical computing systems, a process potentially facilitated by a photonic activation function. As Birgit Stiller, head of the research group “Quantum Optoacoustics,” notes, “The long-term prospect of creating more energy efficient optical neural networks depends on whether we are able to scale up the physical computing systems, a process potentially facilitated by a photonic activation function.”

The power of community cohesion is a vital factor in determining how well a community recovers from extreme events such as earthquakes, hurricanes, floods, or wildfires. According to Jose Ramirez-Marquez, associate professor at Stevens Institute of Technology, tightly connected communities tend to be more resilient when facing these events. This solidarity and togetherness are key to bouncing back, and can even help prevent the collapse of community structures during intense stress periods.

Scientifically, this togetherness is defined as community cohesion, which encompasses a sense of belonging, mutual support among members, and shared values or sentiments. However, whether this cohesion directly influences how well a community recovers from extreme events was not known until recently. To address this issue, Ramirez-Marquez and his research partner Alexander Gilgur developed mathematical techniques to measure community cohesion and its resilience.

The researchers investigated two case studies of the same San Francisco Bay Area community during 2020 wildfires and during 2022-23 rainstorms. They found that during less intense adverse events such as rainstorms, the community performance improved despite increasing stress levels. However, in high-stress disturbances like the wildfires, the community’s performance suffered due to a negative correlation between resilience and the strength of disturbance.

Furthermore, the scientists discovered that emotional intensity has a strong effect on community cohesion. For helping communities be more resilient, emotional engagement is a very important factor, regardless of whether emotions are positive or negative. On the other hand, people’s economic level does not have a direct effect on community cohesion because disasters can affect everyone equally.

Developing metrics to assess community cohesiveness and resilience offers practical benefits. Establishing a causal link between cohesiveness and resilience allows policymakers to set targets and implement policies that aim to improve resilience. Community cohesiveness is essentially a social glue that holds people together, and quantifying this glue can help indicate whether a given community is resilient or can be stronger.

In conclusion, the power of community cohesion is a vital factor in determining how well a community recovers from extreme events. By understanding the relationship between community cohesion, resilience, and emotional intensity, policymakers can develop targeted policies to improve resilience and ultimately save lives during natural disasters.

The self-assembly of molecules into complex structures is a crucial phenomenon in material science. Researchers from Japan have made a significant breakthrough by elucidating a technique where a small amount of residual aggregates drastically alters the self-assembly process of photo-responsive molecules. This study, led by Professor Shiki Yagai and his team, was published online in Nature Nanotechnology on April 11, 2025.

In recent years, there has been an increasing focus on controlling the size and hierarchical structures of self-assembled aggregates to achieve materials with desired properties. However, self-assembly is a dynamic process that requires precise control. As Professor Yagai explains, “During the process of self-assembly, molecules repeatedly associate and dissociate,” making it challenging to predict the final structure of the formed aggregates.

The research team focused on the self-assembly of a chiral, photoresponsive azobenzene molecule that naturally forms left-handed helical aggregates. They discovered that the presence of a small amount of residual aggregates within the solution induces a drastic change in the assembly process and leads to the formation of right-handed helical aggregates instead.

The team found that when the scissor-shaped azobenzene molecule is dissolved in an organic solvent at room temperature, it forms a closed scissor-like folded structure that further extends into a helical assembly. This is due to the chirality of the molecule, which causes it to fold like left-handed scissors and twist to form a left-handed helical stacking of the assembly.

The molecules are photoresponsive, meaning they can change their structure in response to light. When exposed to weak ultraviolet (UV) light, the helical assembly disassembles back into individual molecules, and upon subsequent exposure to visible light, the molecules reassemble into helical structures again.

Interestingly, under certain conditions, the resulting helical aggregates were found to be right-handed instead of left-handed, and exposure to stronger UV light followed by visible light led to the regeneration of the original left-handed helical aggregates. The team attributed this phenomenon to “secondary nucleation,” which explains why meta-stable right-handed aggregates are preferably formed instead of left-handed aggregates.

The researchers also discovered that the intensity of visible light affects the timing of the assembly, with strong visible light promoting rapid assembly while minimizing the influence of residual aggregates. By optimizing the intensities of UV and visible light, the team successfully controlled the switching between left- and right-handed helical structures which were dependent on the influence of residual aggregates.

Furthermore, it was found that the stable left-handed aggregates and meta-stable right-handed aggregates also exhibit opposite electron spin polarization, signifying the tuning of electronic characteristics of the helices. This study aims to explore the critical role of residual aggregates and explained how light-enabled fine-tuning can result in the fabrication of novel functional materials, giving promising insights into the field of material science.