Sixteen sensors turn invisible gases into safety warnings.
BERKELEY, United States | June 2026
Researchers at the University of California, Berkeley have developed an electronic nose capable of identifying spoiled food and detecting common allergens more accurately than human smell. The proof-of-concept device combines 16 miniature gas sensors with machine learning to recognize chemical signatures released by different foods. During laboratory tests, it distinguished fresh products from samples left to deteriorate and detected extremely small quantities of allergenic ingredients. The technology could eventually be integrated into refrigerators, supermarkets and food distribution systems to provide earlier and more objective safety warnings.
The device does not attempt to imitate the shape or biological structure of a human nose. Instead, it reproduces the basic logic of olfaction by using multiple sensors that respond differently to gases in the surrounding air. Each sensor is coated with a distinct material and converts its interaction with airborne molecules into an electrical signal. When the signals are combined, they produce a pattern that functions as a digital fingerprint for a particular food or condition.
Carla Bassil, a doctoral researcher in electrical engineering and computer sciences at Berkeley, led the development of the system. She described the sensors as a collection of digital taste buds, each reacting in its own way to the molecules presented to it. A machine-learning model then compares the combined response with patterns learned during training. This allows the device to classify odors without depending on a single chemical compound or one highly specialized sensor.
The researchers trained the electronic nose to recognize strawberries, blueberries, bananas, walnuts, hazelnuts, cashews and peanuts. It also analyzed raw chicken, milk and eggs under different conditions. Samples were tested when fresh and after remaining at room temperature for 24 and 48 hours. The system successfully differentiated the stages of deterioration by identifying changes in the gases released as bacteria and other processes altered the food.
The technology may be especially useful because human smell is subjective and often unreliable. A product can look acceptable and produce no obvious odor while still containing harmful microorganisms or chemical changes associated with spoilage. Some people also have a reduced sense of smell because of age, illness or other conditions. An electronic system could provide a consistent evaluation that does not depend on individual perception.
Food labels and expiration dates also have limitations. Dates are generally based on estimated shelf life under expected storage conditions, but real products may deteriorate faster or remain usable longer depending on temperature, handling and packaging. Consumers often discard food that may still be safe or consume products that have spoiled before the printed date. A sensor that measures the actual chemical condition of food could reduce both health risks and unnecessary waste.
The same platform showed significant potential for identifying allergens. During testing, it detected only 0.05 grams of isolated walnut, approximately one hundredth of an average shelled walnut. That sensitivity could become valuable for people with severe allergies to nuts and other ingredients. Even trace quantities can trigger dangerous reactions, particularly when cross-contamination occurs during food preparation or manufacturing.
The current results, however, come from controlled laboratory conditions rather than complex real-world environments. The device has not yet been fully tested when several foods and odors are present at the same time. A refrigerator, restaurant kitchen or supermarket contains a mixture of gases from many products, cleaning materials and packaging. The machine-learning system will need to distinguish relevant signals from that background before the technology can be considered reliable for everyday use.
One of the project’s most important innovations involves the use of carbon nanotubes as the conductive material inside the sensors. Electronic noses have existed as a research concept since the 1980s, but combining many different sensing films on a single compact chip has remained difficult. Conventional metal-oxide sensors often require heating, which limits the materials that can be used and increases power consumption. Carbon nanotubes operate effectively at room temperature and can form layers only a few nanometers thick.
Their extremely large surface area increases sensitivity to chemical interactions. Because the system does not require high temperatures, researchers can use polymers and other materials that would degrade under heat. The design also allows multiple sensing films to be deposited in a single manufacturing step through a relatively simple process. That combination could make future versions more scalable and less expensive to produce.
Bassil has already developed a portable version that can be controlled through an iPhone application, although it was not part of the published study. The mobile prototype suggests that the technology could eventually move beyond laboratory equipment and become accessible to consumers or food professionals. A user might place the sensor near a product and receive an immediate freshness or allergen assessment. Further development will focus on improving reliability, sensitivity and performance in mixed environments.
Smart refrigerators are among the most frequently discussed applications. A refrigerator equipped with gas sensors could monitor food continuously and warn that broccoli is beginning to spoil or that raw chicken should be consumed immediately. Such a system could reduce guesswork and help households organize meals before products become unsafe. It could also identify unexpected allergens introduced through cross-contamination.
The commercial potential extends across the food supply chain. Producers could monitor products during processing, warehouses could identify deterioration before distribution and supermarkets could check freshness without opening packages. Restaurants and institutional kitchens could use the technology as an additional safety layer alongside established hygiene procedures. Earlier detection could prevent contaminated products from reaching consumers and reduce the cost of large recalls.
The research also reflects the growing convergence of materials science, artificial intelligence and public health. The sensors generate raw chemical information, but machine learning gives that information practical meaning by recognizing patterns too complex for manual interpretation. As more samples are collected, the models could become capable of identifying a wider range of products, allergens and stages of deterioration. The electronic nose would therefore improve not only through better hardware but also through accumulated data.
The device remains experimental and cannot yet replace laboratory analysis or established food-safety protocols. Its performance must be validated across different temperatures, humidity levels, packaging systems and real commercial conditions. Researchers must also determine how often sensors require calibration and how long their materials remain stable. These challenges will decide whether the technology becomes a useful tool or remains a promising laboratory demonstration.
Even at this early stage, the electronic nose offers a different way to evaluate what people eat. Rather than relying only on appearance, expiration dates or subjective smell, it measures the invisible chemical changes occurring around food. Its ability to detect deterioration and allergens could make kitchens and supply chains safer while reducing avoidable waste. The broader promise is simple but significant: turning odor into data before danger becomes visible.
La verdad es estructura, no ruido. / Truth is structure, not noise.