In the segment 'A Look Around Innovation,' we introduce research and development (R&D) sites that have led to social implementation. In this 14th installment, we introduce the efforts of Shinya Ohmagari, senior researcher at the Sensing Systems Research Center of the National Institute of Advanced Industrial Science and Technology, who has developed an ultrasensitive portable "electronic tongue sensor" that leverages the physical properties of diamond. He is also working to realize an "AI sommelier" that can determine information about liquids in one minute using just a single drop.
Utilization of electrodes that do not self-elute: Fingerprinting acquired liquid information
Hearing that there is an "AI sommelier" that can instantly detect taste of liquids using a very small amount of the sample and as keenly as an expert sommelier, JST News visited the Sensing Systems Research Center of the National Institute of Advanced Industrial Science and Technology (AIST) in Tosu City, Saga Prefecture. The sommelier is actually an ultrasensitive and compact "electronic tongue sensor," which fits in the palm of one's hand (Figure 1).
Figure 1:Developed electronic tongue sensor
Diamond is used as the electrodes of the sensor. When the sensor is immersed in a small amount of a sample, analysis results are displayed on a monitor. Shinya Ohmagari, who is a senior researcher and the developer of this sensor, explained that "Just by immersing the sensor in a drop of a sample for one minute, the information of the liquid is converted into a fingerprint. This is an electrochemical measurement method does not measure a specific component in a liquid; rather, it captures the state of the liquid as a spectral signal. The diamond electrode at the tip senses the oxidation-reduction current produced when a voltage is applied to the liquid sample."
While this measurement method has existed for quite some time, the problem was the "self-elution reaction," in which the sensor dissolved in liquid samples during measurement and peaks related to it were also detected. Even gold and platinum, which are generally considered stable substances, produce such peaks. "However, diamond is extremely stable chemically and never dissolves. Therefore, diamond electrodes do not produce background peaks that become noise," said Ohmagari emphatically. Moreover, the developed sensor is ultrasensitive and has a much larger measurement voltage range than those of conventional sensors. Even minute differences between liquids can be detected without the use of multiple sensors (Figure 2).
Figure 2:Comparison of a conventional sensor and the developed electronic tongue sensor with a diamond electrode
Interviews conducted with over 50 companies — Machine learning to connect with taste data
Diamond is the hardest of all substances found on earth. In addition, it shows high light transmittance, thermal and electrical conductivity, radiation resistance, and chemical stability. Ohmagari was fascinated by these excellent physical properties of diamond since his high school days. After entering Kyushu University, he started conducting research on diamond crystal growth and its use as a semiconductor material. During this time, he felt that it was a waste that diamond was only used in tools and jewelry and its characteristics other than hardness and thermal conductivity were not fully utilized.
After moving to AIST, he continued to develop power devices and sensors that take advantage of diamond's physical properties. However, these properties were difficult to put to practical use. A pivotal point came in 2021: The project was selected for the SCORE Team Promotion Type, JST's commercialization verification program. The program involved a thorough needs assessment and interviews with more than 50 companies. "During the interviews, a German researcher specializing in data science advised us that 'there is a need for wine quality testing,' which led to the creation of the AI sommelier concept," says Ohmagari.
The global wine market is worth 43 trillion yen, according to a survey. Diamond electrodes, which are not affected by high acid or alcohol contents, were ideal for analyzing alcoholic beverages. Accordingly, Ohmagari accelerated his research to realize an AI sommelier. Information of alcoholic beverages, which changes based on fermentation level, maturity, region of origin, temperature, and other factors, was measured by the electronic tongue sensor and linked to human taste data through machine learning. Machine-learning-based classification of red wine by grape variety revealed that the same grape variety has a similar waveform in the spectrum. It was also found that it may be possible to distinguish between different regions of wine production.
The sensor analysis capabilities are reported to highly match the taste evaluation capabilities of an experienced sommelier. There are so many varieties of wine that it is difficult for a single sommelier to statistically classify them; however, Ohmagari's eyes lit up when he explained that the AI sommelier can be used to map a wide range of wine varieties and production regions. In addition to wine, the sensor can be used for coffee, sake, and shochu samples, and it is expected to be useful in preserving traditional flavors for future generations, for example, by archiving the unique tastes of sake breweries.
Establishment of a large-area, high-quality synthesis technology: Expanding applications to healthcare and disaster prevention
Ohmagari applied to, and got selected for, the A-STEP industry-academia collaboration (training type) in 2022 to promote the implementation of the AI sommelier. Aiming to develop a portable and easy-to-use device, he worked with Associate Professor Suguru Ueda of the Faculty of Engineering at Saga University, an expert in machine learning, to develop an algorithm that can highly mimic human sensibilities.
In the course of sensor development, they aimed to create a disposable sensor part that does not need to be washed so that anyone can easily obtain data on liquid samples. To achieve this, a significant cost reduction pertaining to electrodes was necessary; hence, the "thermal filament CVD method" was applied, which produce uniform, high-quality, large-area diamond chips. In the photo at the top of the article, Ohmagari shows a diamond thin film prepared on a 12-inch substrate.
Along with this R&D, they established "ExtenD," a venture company originating from AIST, in July 2022. Ohmagari was appointed as the director and chief technical officer of the company. The name of the company means "expanding the possibilities of diamonds," he said. They have already been commissioned by several sake brewing companies to perform sensory testing and have received orders for sensors from a major beverage company.
Authenticity and quality certification of alcoholic beverages are not the only areas in which Ohmagari and his colleagues are looking to expand their business. If the sensor becomes widely used, it is expected to have applications in healthcare, such as blood and urine tests for humans and animals. Furthermore, if used for monitoring water quality in rivers, the sensor can be used for disaster prevention purposes, such as issuing an alarm when water quality changes due to heavy rainfall. We look forward to seeing how the possibilities of this electric tongue sensor will expand in the future (Figure 3).
Figure 3:Expected applications of the sensor
