Artificial olfactory memory system based on conductive metal-organic frameworks

Synthesis and characterization of Ce-HHTP

The synthetic route for Ce-HHTP is shown in Fig. 2a. A facile one-step hydrothermal reaction of H₆HHTP with hydrated Ce(NO₃)₃ produced a dark-blue suspension consisting of hexagonal microcrystalline powders (Supplementary Fig. 1, Supplementary Discussion). As shown in the scanning electron microscopic (SEM) images (Figs. 2b, c), Ce-HHTP has a diameter of about 300 nm, with the length varying between 1 and 300 µm. A high-resolution transmission electron microscopic image showed a characteristic lattice pattern (Fig. 2d). A one-dimensional penetrating hole with a periodicity of ∼ 1.91 nm was observed along the crystallographic c-axis. These pores were accessible to guest molecules, thereby facilitating adsorption reactions that catered to the capturing and retention of molecular chemical information. The selected- area electron diffraction pattern (Supplementary Fig. 2, Supplementary Discussion) also confirmed the high crystallinity. The structure of Ce-HHTP was obtained by Rietveld refinement of powder X-ray diffraction (PXRD) data, which provided an excellent structural model for Ce-HHTP³⁰ (Supplementary Table 2 and Supplementary Discussion). Further information on the structural details was provided in the supplementary Ce-HHTP cif. As shown in Fig. 2e, the experimental PXRD spectrum with isostructural peaks of (100), (200), and (210) at 2θ = 4.7°, 9.3°, and 12.3°, respectively^{33,34,35,36,37}, was qualitatively identical to that of the lanthanide MOFs with differences in unit cell parameters and Ln‒O bond lengths attributable to the smaller size. The structure shows a two-site disorder, each exhibiting two sets of equally occupied ligand and metal sites. The central ions were located between the organic planes attached to the 3D frame by O-Ce-O chains³⁰. Importantly, there were coordinated unsaturated sites between the layers bound to water or the hydroxide groups in the solvent, which is consistent with the conclusion obtained by tracing the formation of Ce-HHTP using Fourier transform infrared (FTIR) spectroscopy (Supplementary Fig. 3 and Supplementary Discussion). The Brunauer–Emmett–Teller (BET) surface areas and pore volume of the Ce-HHTP were calculated as 405 m²·g⁻¹ and 0.615 cm³·g⁻¹, respectively. The experimental pore diameters were ~ 1.4 nm (Supplementary Fig. 4 and Supplementary Discussion). The desorption curve overlapped perfectly with the adsorption curve, indicating a highly reversible gas sorption process.

a Synthetic route of Ce-2,3,6,7,10,11-hexahydroxytriphenylene (Ce-HHTP) via the one-step hydrothermal reaction of H₆HHTP with hydrated Ce(NO₃)₃ and structure of Ce-HHTP. Atoms are colored as follows: O (red), C (black), Ce (light blue), and H (white). b, c Scanning electron microscopic (SEM) images of Ce-HHTP. d Transmission electron microscopic (TEM) image of Ce-HHTP. e Powder X-ray diffraction (PXRD) pattern and the Rietveld refinement of Ce-HHTP. f Energy-dispersive X-ray spectroscopy (EDS) elemental mappings of Ce-HHTP.

Energy-dispersive X-ray spectroscopy (EDS) elemental mapping showed a uniform distribution of C, O, and Ce, further demonstrating the formation of Ce-HHTP (Fig. 2f). Detailed elemental analysis was performed using X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 5 and Supplementary Discussion). The C1s XPS spectra show the coexistence of C-O and C = O, suggesting a small number of hydroxyl groups oxidized to form quinon³⁶. O1s XPS spectra show that almost all of the oxygen is bonded to Ce ions, including the oxygen in adsorbed water³⁸, as well as oxygen coordinated in two different forms, which corresponds to the structure of Ce-HHTP³⁰. In addition, the magnified Ce 3d XPS spectra show eight peaks corresponding to the spin-orbit splitting Ce 3d_5/2 and Ce 3d_3/2 of Ce (III and IV). Ce-HHTP exhibits a higher oxidation state due to bonding with oxygen, indicating the full involvement of Ce ions in the direct bonding with HHTP^38,39. As a measure of the relative content of the corresponding species, we accurately calculated the ratio of peak areas contributed by each of the Ce oxidation states (III and IV), which resulted in IV / III being approximately equal to 57% / 43%. Considering the variability due to sample conditions and different batches, repetition of these tests confirmed the ratios are within a narrow range of 57% / 43% ~ 58% / 42%, with a variability of less than 3%, which is within an acceptable margin of error. Therefore, the Ce oxidation state of the original Ce-HHTP is composed of mixed valences (III and IV). In addition, the electrical conductivity at room temperature was measured by the four-contact probe measurements, and reached 3.6 × 10⁻³ and 0.3 S·m⁻¹ for a polycrystalline pellet and single-crystal, respectively. The thermogravimetric analysis (TGA) curve of the Ce-HHTP was shown in Supplementary Fig. 6 (Supplementary Discussion), a 10% weight loss in the temperature range 30–140 °C is assigned to the H₂O desorption^40,41, and the rather sharp weight loss of the sample took place after 220 °C, corresponding to the structural decomposition³⁰. So, the result showed good thermal stability, as indicated by its PXRD patterns obtained after a variable temperature experiment. (Supplementary Fig. 7 and Supplementary Discussion).

Preparation and evaluation of the artificial olfactory memory system

The fabrication of the artificial olfactory memory system is illustrated in Fig. 3a. Specifically, a sensory memory unit was prepared by drop-casting 20 μL of the suspension onto an interdigital electrode, followed by aging in a constant-temperature oven, which can be combined with a power supply to form a series circuit. Figure 3b shows the workflow of the artificial olfactory memory system. All tests were performed at stable 40 °C (A common temperature that is easy to control thermostatically and close to room temperature⁴².) and 39% ± 1% relative humidity. The sensory memory response was defined as (R_x–R₀)/R₀, where R₀ is the chemical resistance in a normal air atmosphere with a constant relative humidity (RH) of 39% and temperature of 40 °C, and R_x is the chemical resistance under the same conditions with the presence of the analyte. The concentration of the analyte was distributed in liquid gas according to the calculation formula. In addition, the recovery experiments were carried out under the same conditions with a strictly controlled and consistent “target-free” atmosphere to ensure the comparability and reliability of the results. We first tested the stability in a “target-free” atmosphere for 10 h (Supplementary Fig. 8 and Supplementary Discussion), the artificial olfactory memory system is able to keep stable well enough. Three repeated stability tests further show this result has good reproducibility and credibility with CV = 0.6557% for 51500 resistances collected in 36000 s (Supplementary Fig. 9 and Supplementary Discussion), which is the basis for stable operation. In addition, in order to eliminate the effects of factors other than target gas on resistance, additional control experiments, including relative humidity and temperature, were performed, and calibration data and correlation analyses were provided in Supplementary Fig. 10 (Supplementary Discussion). These analyses are essential for achieving higher accuracy and detection range in artificial olfactory memory systems. Figure 3c shows that the artificial olfactory memory system exhibited short-term memory for 100 ppm propanol. The response of the artificial olfactory memory system increased upon exposure to propanol and gradually recovered its initial resistance after the removal of propanol. Figure 3d shows the long-term memory for 100 ppm propionaldehyde, where the response increased upon exposure to propionaldehyde and consistently maintained the final chemical resistance for 9 h after the removal of propionaldehyde. Note that this solitary device, which integrates sensing and memory, is capable of exhibiting two distinct forms of memory (short- and long-term memories), with significant specificity between two chemicals. More interestingly, three consecutive tests of the system showed that the artificial olfactory memory system could stably memorize the response to propionaldehyde for a long time regardless of single or multiple, while the response to propanol remained in the short-term memory (Fig. 3e, f), indicating that this artificial olfactory memory system is capable of efficiently accomplishing continuous memory in complex environments with diverse ever-changing chemical stimuli. In fact, the artificial olfactory memory system has short-term memory for alcohols and long-term memory for aldehydes, as shown in Fig. 3g, h. Therefore, the artificial olfactory memory system based on the electrical signal of Ce-HHTP has two distinct sensory memories for two types of targets, indicating that it has great prospects in the bionic design of human olfactory memory.

a Schematic fabrication of the artificial olfactory memory system. b Schematic workflow of the artificial olfactory memory system. c Short-term memory curve for 100 ppm propanol. d Long-term memory curve for 100 ppm propionaldehyde. e Short-term memory curve of three successive stimuli for 100 ppm propanol. f Long-term memory curve of three successive stimuli for 100 ppm propionaldehyde. g Short-term memory curves for 100 ppm of methanol, ethanol, and propanol. h Long-term memory curves for 100 ppm of formaldehyde, acetaldehyde, and propionaldehyde. i Long-term memory curves for various concentrations of propionaldehyde (50−500 ppm). j Long-term memory grayscale map for varying concentration and time. k Seven-day stability curve for long-term memory.

To further evaluate the performance of this artificial olfactory memory system, we tested long-term memory at multiple concentrations of propionaldehyde (50−500 ppm), as shown in Fig. 3i. The system can make long-term memory response to the different concentrations, indicating a promising advancement over the limitation associated with the threshold confinement observed in primordial memristors. There is no doubt that the memory response increased with increasing concentration and processing time (Fig. 3j). Moreover, the long-term memory response increased linearly with concentration for all tests. For example, when considering the test results recorded at 50 s-intervals spanning from 100 and 400 s, a distinctly linear correlation emerges between long-term memory and concentration within the range of 50−500 ppm (Supplementary Figs. 11, 12 and Supplementary Discussion). This robust and reliable linear response is essential for artificial olfactory memory systems to achieve a higher accuracy and detection range. At the same time, a nonlinear relationship fitting curve between the long-term memory and various concentrations at saturated adsorption showed that the system reached saturation at 480 ppm (Supplementary Figs. 13, 14 and Supplementary Discussion). The LOD for propionaldehyde can be calculated to be about 3 ppm from the simulated linear equation by setting the response to 10%. For formaldehyde, the LOD obtained is about 124 ppm (Supplementary Fig. 15 and Supplementary Discussion). Notably, among the reported chemiresistive sensors based on pure MOFs, Ce-HHTP exhibits a comparable LOD for aldehydes (Supplementary Table 3 and Supplementary Discussion). Even though this result is above the 8 h-weighted average permissible exposure limit (0.016 ppm) in the workplace as well as the ceiling limit of exposure (0.1 ppm) established by The National Institute for Occupational Safety and Health, we will continue to explore new methods and materials for performance enhancement in the future. In order to better evaluate the selectivity, we added the selectivity tests for four classes of VOCs, namely alcohols, aldehydes, ketones, and amines, which are potential analytes that can negatively affect human health. As shown in Supplementary Fig. 16 (Supplementary Discussion). The artificial olfactory memory system has short-term memory for alcohols represented by methanol, ethanol, and propanol, and specific long-term memory for aldehydes represented by formaldehyde, acetaldehyde, propionaldehyde, glyoxal, glutaraldehyde, and benzaldehyde. Interestingly, as the isomer of propionaldehyde, the system exhibits short-term memory for acetone, which is very distinctly different from the former, ensuring its ability to distinguish acetone from propionaldehyde. Notably, although both acetone and propanol exhibit short-term memory, the response of acetone is about 17% of that of propanol under the same concentration of analytes, that is, acetone has a few interferences on the detection of propanol but no effect on the specific detection of propionaldehyde. In addition, it is noteworthy that amines represented by ammonia and ethylenediamine show negative short-term memory diametrically opposed to short-term memory and long-term memory. These results indicate that the system has specific long-term memory for aldehydes, which is clearly different from the other three types of interfering substances. Therefore, the system has the potential to be applied in the future for selective long-term memory detection of potentially carcinogenic aldehydes in a special target area, which is an unusual advantage in the application areas of modeling the health damage of aldehydes to human beings and aldehyde dosimetry, as well as the non-in-situ detection of aldehydes.

Stability is essential for the artificial olfactory memory system. We demonstrated the stability of Ce-HHTP by testing long-term memory responses to 100 ppm propionaldehyde from day one to day seven, as shown in Fig. 3k. During the seven-day-long stability test, the memory response changed by only ~ 6%, suggesting that the degradation was negligible. Therefore, the artificial olfactory memory system can be used reliably for long-term memory.

Mechanism of the artificial olfactory memory system

To understand the contribution of Ce-HHTP to the two distinct forms of memory, FTIR spectroscopy was used to analyze Ce-HHTP before and after exposure to propionaldehyde and propanol. Based on characterization (Supplementary Fig. 17 and Supplementary Discussion), bound water was present in the original Ce-HHTP. With the progressive treatment of propionaldehyde, the characteristic peaks of -CH₃, -CH₂-, and C = O at 2975, 2932, and 1722 cm⁻¹, respectively^{43,44,45,46,47}, in propionaldehyde increased step-by-step, simultaneously the peak of bound water at 3400 cm⁻¹ in Ce-HHTP continued to weaken until it disappeared completely in the final spectrum⁴⁴, as shown in Figs. 4a, b. Subsequently, an in situ FTIR spectroscopy in a vacuum environment showed that the characteristic peaks of the adsorbed propionaldehyde did not weaken for 30 min, and the characteristic peak of water did not recover significantly after 30 min in an atmospheric environment, indicating that propionaldehyde was strongly coordinated with the central metal ions. In fact, these apparent characteristic peaks did not weaken after three hours under vacuum (Supplementary Fig. 18 and Supplementary Discussion). This result was consistent with the conclusion given by XPS results. As shown in Fig. 4c, after combination with propionaldehyde, the peak at 916.7 eV disappeared completely, which is clear evidence that the Ce (IV) oxidation state was considerably reduced. Based on the ratio of the normalized peak areas, the relative content of Ce (IV) and Ce (III) in the sample is from 56.8% / 43.2% before the combination with propionaldehyde to 46.7% / 53.3% after the combination. However, the same experiment with propanol yielded different results, as shown in Fig. 4d, e. With the same progressive treatment with propanol, the characteristic peaks of -CH₃, -CH₂-, and C-O were observed at 2935, 2880, and 1070 cm⁻¹ in propanol^44,45, while the peak of bound water at 3400 cm⁻¹ did not weaken⁴⁴. Simultaneously, an in situ FTIR spectroscopy in a vacuum environment showed that the characteristic peaks of the adsorbed propanol disappeared quickly, and the characteristic peak of water was obviously restored after 30 min in an atmospheric environment, indicating that the molecules were not coordinated with the central metal ions. The comparison experiment is shown in Fig. 4f, after combining with propanol, The relative content of Ce (IV) and Ce (III) only changed from 58.4% / 41.6% to 57.7% / 42.3%. There was almost no change in the oxidation state of the central ions. The same conclusion can be observed by the electron paramagnetic resonance (EPR) results as well. As shown in Supplementary Fig. 19 (Supplementary Discussion), the isotropic signals (g = 2.0038) are obviously observed in the EPR spectra, confirming that Ce-HHTP has remarkable electron delocalization properties⁴⁸. Compared to the original material, Ce-HHTP showed no change after combination with propanol, whereas after exposure to electron-donating propionaldehyde, the interaction of propionaldehyde with Ce-HHTP resulted in a detectable increase in their EPR signal at g = 2.0034 from the Ce-centered radical, indicating that part of the EPR-silent Ce IV in the original Ce-HHTP was transformed into paramagnetic Ce III by coordination with propionaldehyde (4f ⁰ to 4f ¹). Besides, the pre- and post-adsorption PXRD comparisons with different guest molecules demonstrated that there is no significant change in the PXRD patterns before and after the combination (Supplementary Fig. 20 and Supplementary Discussion).

a FTIR spectra of Ce-HHTP treated by propionaldehyde. The direction of the black arrow indicates a step by step increase of propionaldehyde. b In situ FTIR spectra of Ce-HHTP treated by propionaldehyde under vacuum and subsequent atmospheric conditions. The direction of the black arrows indicates the time-resolved FTIR spectra in the corresponding conditions. c XPS results before and after the Ce-HHTP was combined with propionaldehyde. d FTIR spectra of Ce-HHTP treated by propanol. The direction of the black arrow indicates a step-by-step increase of propanol. e In-situ FTIR spectra of Ce-HHTP treated by propanol under vacuum and subsequent atmospheric conditions. The direction of the black arrows indicates the time-resolved FTIR spectra in the corresponding conditions. f XPS results before and after the Ce-HHTP was combined with propanol. g DFT calculations of all the adsorption energy on Ce-HHTP. Atoms are colored as follows: O (red), C (black), Ce (light blue), and H (white). h, i Structural diagram of Ce-HHTP coordination with propionaldehyde and hydrogen binding with propanol. Atoms are colored as follows: O (red), C (black), Ce (light blue) and H (white).

To further explore the molecule-memory relationships at the atomic level, we performed density functional theory (DFT) calculations on Ce-HHTP for all the molecular adsorption energies. As shown in Fig. 4g, all of the molecules adsorption over the Ce site was found to be exothermic, owing to the formation of the Ce−O bond. It binds H₂O with − 0.72 eV, while binds CH₃CH₂CHO and CH₃CH₂CH₂OH with − 1.61 and − 0.38 eV, respectively. Obviously, the adsorption energy describes how stably a molecule binds to the central ion. The smaller the adsorption energy, the easier it was to form a stable complex. Thus, the former is 1.5 times as stable as bound water, while the latter is not, which means that the coordination of propionaldehyde can still occur spontaneously, while propanol is more likely to form weak interactions (hydrogen bonds) underwater coordination. These results indicate that there are two different adsorption behaviors for propionaldehyde and propanol on Ce-HHTP. As shown in Fig. 4h, when Ce-HHTP was exposed to propionaldehyde, C = O was strongly coordinated with the metal central ions, and the original bound water left. Because of the change in the state of the O atom, electrons are transferred to the Ce-O coordination bond, resulting in an irreversible change in the oxidation state of Ce from IV to III. At the same time, the solid-state NMR results showed characteristic peaks corresponding to propionaldehyde, and Ce-HHTP was significantly shifted owing to its strong binding affinity (Supplementary Fig. 21 and Supplementary Discussion), which contributed to the satisfactory long-term memory of propionaldehyde. When Ce-HHTP was exposed to propanol, the –OH group of propanol formed reversible hydrogen bonds with the coordination groups of Ce-HHTP, as shown in Fig. 4i. As this hydrogen bond is weaker than coordination, the short-term memory response recovers when the stimulus is removed.

The interaction region indicator (IRI) method has been used to visually reveal chemical bonds and weak interaction regions, making it convenient for chemists to study various chemical systems and chemical reactions⁴⁹. In Fig. 5a, the sign (λ₂) ρ serves as a good indicator of interaction strength; ρ < 0 indicates attractive interactions (such as hydrogen bonding), ρ > 0 indicates repulsive interactions, and ρ ≈ 0 indicates van der Waals interaction⁵⁰. The calculated visualization results of the binding of Ce-HHTP to propionaldehyde, H₂O, and propanol are shown in Fig. 5b–d, respectively. It can be clearly seen that the force interaction between the central metal ion (Ce) and propionaldehyde, H₂O, and propanol forms a blue color, which represents the chemical bond. To better compare the strength relationships among the three compounds, the calculated scatter plot is shown in Fig. 5e, where black represents propanol, red represents H₂O, and blue represents propionaldehyde. The magnified scatter diagram of the interaction between the metal center (Ce) and propionaldehyde, H₂O, and propanol in Fig. 5f shows that the chemical bond binding strength of Ce with propionaldehyde is the highest, while that with propanol is the weakest. Therefore, the results of the quantitative calculations further reveal the essence of their chemical interactions. Based on the above results, the different combination modes of Ce-HHTP with O atoms in different guest molecules are the fundamental reasons for the specific memory for the two types of gases in the artificial olfactory memory system (Supplementary Fig. 22 and Supplementary Discussion).

a Schematic distribution of non-covalent bonding interaction strengths. Schematic of the interactions of the optimized structure of the unit cells with (b) propionaldehyde, (c) water, and (d) propanol. Atoms are colored as follows: O (red), C (light blue), Ce (brown), and H (white). e Scatter map of stacked IRI quantitative distribution. f Stacked IRI quantitative distribution map with local zoom. Atoms are colored as follows: O (red), C (light blue), Ce (brown) and H (white).

Application of the artificial olfactory memory system

To further demonstrate the potential applications of the artificial olfactory memory system, we can connect the system with possible actuators in series, including indicators, alarm bells, electrical switches, etc., which will greatly expand the scope of application. As an example, we connected it to a light-emitting diode (LED) to obtain a sensing and memory alarm system for alcohols and aldehydes in a specific environment. The logic circuit is illustrated in Fig. 6a. Initially, the system was in a state of low resistance when the targets were absent, and the LED was OFF. When the system was exposed to alcohols or aldehydes, the resistance gradually increased, as shown in Fig. 6b. The local voltage of the LED rose, and the state was changed from OFF to ON with the increase in response. When the external stimulus was removed, the response information was memorized by the system short-term or long-term and displayed on the LED, which is convenient for memory detection in dangerous areas.

a Logic circuit of the memory alarm system for alcohols and aldehydes, photograph of the light-emitting diode. b Response and brightness curves of the memory alarm system. Brightness is limited to the schematic, where light green indicates low brightness, corresponding to a lower response dark green indicates high brightness, corresponding to a higher response, and white indicates off, corresponding to no response. c Schematic diagram of sensory memory of the unmanned aircraft. d Photograph of the memory detection module and the unmanned aircraft. e Photograph of the sensory memory of the unmanned aircraft. f Applications of the artificial olfactory memory system, left: flexible patches, middle: portable detection box, right: telecar.

Aldehydes are of great concern that can cause nausea, dizziness, and fatigue. Most aldehydes are potential carcinogens. However, aldehydes are ubiquitous in industrial settings where chemicals such as plastics and paints are produced, and their concentrations may be much higher than the standard. Therefore, to ensure worker’ safety, we can integrate this system into the detection probe, as shown in Fig. 6d, which is mounted on an unmanned aircraft to carry out memory detection on a pre-set route, as shown in Fig. 6c, e, represents the whole process of memory detection. Specifically, memory detection can be realized when the unmanned aircraft returns to its origin point. If it has passed through an area with a high concentration of dangerous aldehydes, the safety green light will be turned on, and after leaving the monitoring area, the indicator light will remain on. However, alcohol will only lead to short-term memory. The entire process of the telecar is detailed in Supplementary Movie 1. The advantage of this memory detection for dangerous areas is reflected in the fact that once there is a pollution source at any point, the system will sense, recognize, and memory, as shown in Supplementary Movie 2, which ensures the safety of the entire plant environment at all times, not just at one point where leaks are likely to occur. The advantage is also reflected in the fact that it can show distinct memory forms when facing a complex environment with diverse ever-changing chemical stimulations. When it passes by alcohols (propanol) and aldehydes (propanaldehyde), in turn, it can show the corresponding memory forms, as shown in Supplementary Movie 3.

In addition to the combination of various artificial carriers, we also conducted corresponding experiments related to off-site memory detection (Supplementary Fig. 23 and Supplementary Discussion). Ce-HHTP can recognize aldehydes and memory of the entire sensing effect when it is removed from the instrument and not powered on, and the information can be fully represented when powered on again. We have developed portable detection boxes to cope with the scenarios that are not convenient for long-term online detection of instruments (Supplementary Fig. 24 and Supplementary Discussion), such as one-time detection electrodes or flexible patches, as shown in Supplementary Movie 4, which makes it very simple to build a health-injury simulation chip, as shown in Supplementary Movie 5. The flexible patch can fit well in masks, clothes, and even the human body, as shown in Fig. 6f. Through the cumulative recognition and storage of aldehydes at room temperature, it can simulate the health damage of workers in long-term work. Therefore, the proposed system will be a powerful tool for future environmental monitoring and health management in portable and wearable sensing platforms.