The human tongue perceives food through the synergistic sensation of both chemical (taste) and physical (temperature, texture, softness) cues. Inspired by such functionalities, electronic tongues (e-tongues) have been developed for applications ranging from food analysis to biomedical sensing. However, most reported e-tongues primarily capture chemical tastes, overlooking critical physical attributes. In addition, current e-tongues typically rely on heterogeneous sensor outputs (e.g., current, capacitance, impedance), which complicates circuit design and signal processing and leads to high system power consumption. Here, we present a fully potentiometric, monolithically integrated multimodal e-tongue capable of simultaneously sensing chemical attributes (e.g., salinity and acidity) and physical attributes (e.g., temperature, texture, softness) of food. Importantly, all integrated sensors self-generate a unified output─potential difference (mV)─thereby eliminating the need for external power. Such a fully potentiometric sensor integration also greatly simplifies the signal acquisition circuitry and drastically reduces the system power consumption. Assisted by machine learning algorithms, the proposed fully potentiometric e-tongue achieves high accuracy in discrimination of both fruits and beverages, outperforming conventional single-modality e-tongues. This work demonstrates a practical route toward holistic gustatory sensing and offers new opportunities for the development of biomimetic intelligent systems that closely replicate natural taste perception.

Natural skin receptors use ions as signal carriers, while most of the developed artificial tactile sensors utilize electrons as information carriers. To imitate the biological ionic sensing behavior, here, we present a kind of biomimetic, ionic, and fully passive mechanotransduction mechanism leveraging mechanical modulation of interfacial ionic p-n junction (IPNJ) through microchannels. Sensors based on this mechanism do not rely on an external power supply and can encode external tactile stimuli into highly analogous signal outputs to those of natural skin receptors, in terms of both signal type (i.e., ionic potential difference) and signal intensity (≈120 mV). More importantly, the instant interfacial IPNJ regulation characteristic endows the sensors with superior performance when compared to the state-of-the-art piezoionic sensors, including a low detection limit of 0.01 N, fast response/recovery speeds (16 ms/16 ms), ultralow power consumption (pW level), excellent reproducibility (over 100,000 cycles), and good capabilities to resolve both static and dynamic mechanical stimulations. As demonstrations, machine-learning-assisted high accuracy (over 99%) surface texture recognition and object classification are successfully demonstrated with the sensors integrated on robotic hands. This work enriches the family of mechanical sensing mechanisms and provides a path to mimicking natural tactile sensory systems for smart skins, artificial prostheses, and intelligent robots.