Structure and desolvationCd2(TTFTB) was synthesized based on the previous literature (see Methods)37. Optical microscopy and scanning electron microscopy revealed hexagonal rod-like single crystals with regular shapes, millimeter-scale size, and smooth surfaces, which facilitate fabrication of single-crystal devices (Fig. 1d, e; Supplementary Fig. 1a). Single-crystal X-ray diffraction (SC-XRD) of the as-synthesized Cd2(TTFTB) revealed π-stacked TTF columns formed by π − π stacking and S···S interactions between TTF moieties along the crystallographic c-axis (Fig. 1b). The side chain consists of two alternating and crystallographically independent Cd2+ ions bridged by carboxylate groups (Fig. 1c). One type of Cd2+ is six-coordinated and is bound to two terminal water molecules; the other type is five-coordinated and does not coordinate to water. The π-stacked TTF columns and side chains together delineate one-dimensional (1D) quasi-elliptic cylindrical pores (Fig. 1a). This structure is different from the previously reported Cd2(TTFTB)37 yet analogous to several other M2(TTFTB) (M = Mn2+, Co2+, and Zn2+)37,38. SC-XRD showed solvent molecules in the pores and elemental analysis revealed a formula of [Cd2(TTFTB)(H2O)2]·(DMF)1.34(H2O)3.73 (DMF = N,N-dimethylformamide). The guest DMF and water molecules may assist with the terminal water molecules to form hydrogen-bonded networks within the pores.Fig. 1: Structure and desolvation of Cd2(TTFTB).a, b Portions of structures viewed parallel or perpendicular to the crystallographic c-axis showing nanoscale pores or the π-stacked TTF columns, respectively. Gray, red, yellow, and purple spheres represent C, O, S, and Cd, respectively. Solvent and H atoms are omitted for clarity. c Portion of the side chain highlighting terminal water molecules. d, e Micrographs of single crystals viewed perpendicular or parallel to the long axis, respectively. f PXRD patterns of the as-synthesized Cd2(TTFTB) and a sample treated at 353 K and 90% RH for 72 h in comparison with a pattern simulated from the crystal structure. Miller indices of primary PXRD peaks are indicated. g IR spectra of evaporated gaseous products generated at various temperatures from TGA. IR features of CO2 appeared due to instrumental artefacts.Electrical characterization of Cd2(TTFTB) involves maintaining its crystal in the air at temperature up to 363 K and relative humidity (RH) up to 90% or purging it in dry N2 for 1 h at a fixed temperature that varies from 298 K to 363 K (see Methods). The structural stability of Cd2(TTFTB) under elevated temperature and humidity was confirmed by powder X-ray diffraction (PXRD), which revealed consistent diffraction pattern after subjecting crystals to the air at 353 K and 90% RH for 72 h (Fig. 1f; Supplementary Figs. 2 and 3). The diffraction angles and widths of PXRD peaks persisted after this treatment, confirming the integrity of the framework structure.We further investigated the desolvation of Cd2(TTFTB) with thermogravimetric analysis coupled with Fourier transform infrared spectroscopy (TGA-FTIR). To simulate the electrical characterization procedures, the sample was heated slowly to 363 K in a flow of dry N2, maintained at this temperature for 1 h, and heated rapidly to 473 K. TGA shows a mass loss during the whole treatment, which is attributed to desolvation (Supplementary Fig. 4). In the first two stages, the degas rate was too slow to meet the detection limit of spectrometer, resulting in negligible features of water or DMF in IR spectra (Fig. 1g). In contrast, rapid degassing took place during the fast-heating stage, revealing intense IR features of DMF (e.g., C = O stretch at approximately 1700 cm−1) as well as a weak and broad band centered at approximately 3400 cm−1 that signifies water. The former shows up at above 417 K that is comparable to the boiling point of DMF (426 K), whereas the latter was only observed at 473 K that is much higher than the boiling point of water (373 K) likely due to a strong binding of water to Cd2+. Thus, the N2 purging process involved in electrical characterization desolvates Cd2(TTFTB) partially—only a portion of DMF and water may be removed from the pores.DC electrical characterizationTwo-contact probe single-crystal devices of Cd2(TTFTB) were fabricated to characterize the electrical conductivity along the crystallographic c-axis (Fig. 2a and Supplementary Fig. 1b). To elucidate the atmospheric modulation of electrical conductivity in Cd2(TTFTB), we conducted direct-current (DC) electrical characterization at room temperature (298 K) and under various atmospheres including humid air (38% − 45% RH), humid N2 (100% RH), dry air (<2% RH), and dry N2 (<0.02% RH) (Supplementary Fig. 1c). The comparison between humid and dry atmospheres reveals the influence of water on the DC conductivity, whereas that between air and N2 atmospheres shows the influence of oxygen (Fig. 2b). Current−voltage (I − V) curves are linear in all tested atmospheres, allowing extraction of the apparent electrical conductivity values with the Ohm’s law (Fig. 2c).Fig. 2: DC electrical characterization and analysis of Cd2(TTFTB).a Fabrication of a single-crystal two-contact probe device. Pink, pale green, orange, yellow, gray, and blue objects represent a single crystal of Cd2(TTFTB), carbon paste, gold wires, gold electrodes, a piece of glass slide, and electrical probes, respectively. b Protocol for elucidating the influence of O2 and H2O on the apparent electrical conductivity with variable-atmosphere electrical characterization. c I − V curves acquired at 298 K under humid air, humid N2, dry air, and dry N2, where the latter two overlap under the scale of this figure. d Comparison of the DC conductivity in different atmospheres. e Dynamic variation of the electrical current through a device under an applied bias of 0.1 V in various atmospheres.For a representative device, the DC conductivity in humid air is σDC, humid air = 9.44 × 10−5 S·cm−1, which is consistent with previously reported values of Cd2(TTFTB)25,37. The crystal was then purged in dry N2 at 363 K for 1 h to desolvate it, cooled down to 298 K, maintained in dry N2 for 1 h, and exposed to dry air for 1 h (see details in Methods and Supplementary Fig. 5). As discussed above, this treatment can only desolvate the framework partially with a significant amount of residual DMF and coordinating water in pores. Nonetheless, it reduced the DC conductivity to σDC, dry N2 = 1.03 × 10−6 S·cm−1, which remained in dry air (σDC, dry air = 1.06 × 10−6 S·cm−1). Finally, exposing the crystal to humid N2 for 1 h improved the DC conductivity to σDC, humid N2 = 5.24 × 10−5 S·cm−1. This is slightly lower than σDC, humid air, which may be attributed to desolvation-induced changes of the crystal structure (vide infra) and guest composition in pores. Nonetheless, such recovery indicates that the low DC conductivity values under dry atmospheres were not caused by an accidental device damage. Thus, the room-temperature DC conductivity of Cd2(TTFTB) shows the following trend: σDC, humid air > σDC, humid N2 » σDC, dry air ≈ σDC, dry N2 (Fig. 2d).We further conducted variable-atmosphere DC electrical characterization for crystals from different batches and N2-purged at 298 K. The order of atmospheres was switched for some devices and two cycles of measurements were performed for one device. Characterization of four-contact probe single-crystal devices was also conducted to eliminate the influence of contact resistance (Supplementary Fig. 1d). Although the exact electrical conductivity values differ among these devices, their atmospheric dependencies are consistent with the above trend (see examples in Supplementary Figs. 6 and 7). In addition, we monitored dynamic changes of the current through a device (Fig. 2e and Supplementary Fig. 8). Purging the single crystal of Cd2(TTFTB) with dry N2 reduced the current immediately, reaching a plateau after 700 s. Switching dry N2 to dry air caused a negligible change in current. We then switched the purging gas to humid N2. The current first slightly dropped for 70 s, then increased sharply, and finally levels off after 900 s. Notably, both the decline of current in dry N2 and the rise of current in humid N2 exhibit exponential decays with rate constants of 0.126 s−1 and 0.019 s−1, respectively (Supplementary Fig. 9). These trends are consistent with the first-order dynamics of desorption and adsorption39. Thus, the adsorbed guests should play a major role in the variation of electrical conductivity.Atmospheric modulation of charge mobility and hole densityThe DC electrical conductivity is a product of elementary charge (e), charge mobility (μ), and charge density (n), i.e., σ = eμn2. Therefore, it is viable to examine charge mobility and charge density separately to understand their modulation by the atmosphere.Directly measuring the charge mobility of Cd2(TTFTB) is challenging due to technical difficulties of fabricating field-effect transistors or Hall bars. Hence, we probed it indirectly with structural characterization. Previous studies on a series of M2(TTFTB) (M2+ = Mn2+, Co2+, Zn2+, Cd2+) materials showed that their electrical conductivity increases with decreasing S···S distances. This was rationalized by enhanced overlap between 3pz orbitals of adjacent TTF moieties and in turn improved charge mobility37,38. As the S···S contact is in line with the crystallographic c-axis, a shorter S···S distance should manifest as a reduced unit cell parameter in the c direction.Driven by this hypothesis, we conducted in situ PXRD measurements on Cd2(TTFTB) using CaCO3 as an internal reference for calibration (Supplementary Fig. 10). The PXRD patterns were analyzed by the Le Bail refinement to extract unit cell parameters (Supplementary Figs. 11−15). A PXRD pattern of the as-synthesized Cd2(TTFTB) was first acquired in humid air, revealing unit cell parameters of a = b = 19.6463 Å and c = 21.0196 Å, which are consistent with those obtained from SC-XRD (Supplementary Table 1). The same sample was then evacuated at 298 K for 1 h. Its PXRD peaks shifted towards high angles with the most representative (006) diffraction peak shifting from 25.39° to 25.96° (Fig. 3b, c). Although the crystal symmetry remained, the evacuation caused a contraction of the unit cell to a = b = 19.5446 Å and c = 20.5834 Å. The sample was then allowed to stay in the humid air for 1 h, which shifted the PXRD peaks back towards low angles. The (006) diffraction peak was at 25.52°, and the unit cell parameters were a = b = 19.6411 Å and c = 20.8910 Å. This unit cell is slightly smaller than that of the as-synthesized sample, which is likely due to partial removal of the adsorbed solvent (vide supra). The evacuation and air-refilling process was repeated once, displaying nearly consistent PXRD patterns with the (006) peak showing at a slightly higher angle possibly due to further removal of solvent during the second evacuation process. These observations indicate that the S···S distance decreases because of partial desolvation, which should lead to an increase in charge mobility.Fig. 3: Atmospheric modulation of charge mobility and charge density of Cd2(TTFTB).a CW-EPR spectra collected in humid air and in vacuum. b in situ PXRD patterns collected in humid air (process 0) and then during two cycles of evacuation and air-refilling treatment (process 1 and 2). Patterns were calibrated by the CaCO3 standard (the intense peak at 29.4°). c in situ PXRD patterns zoomed at the region of (006) diffraction. Dashed lines are to guide eyes. d π-stacked TTF columns in the as-synthesized and e evacuated Cd2(TTFTB) highlighting the shortest S···S contact. f, g Electronic band structures of the as-synthesized and evacuated frameworks. M–Γ and A–H are in-plane vectors, while Γ–A samples the TTF π-stacking direction.To confirm the shortening of the S···S distance, we evacuated Cd2(TTFTB) at room temperature for 15 h and acquired its crystal structure by SC-XRD. This treatment removed the adsorbed solvent and shrank the unit cell to a = b = 19.6184 Å and c = 20.6160 Å (Supplementary Table 2), matching well with the parameters obtained from PXRD. Importantly, the S···S distance decreases from 3.71 Å in the as-synthesized framework to 3.62 Å in the evacuated one (Fig. 3d, e). Density functional theory (DFT) calculations were performed on both the as-synthesized and evacuated forms of Cd2(TTFTB) (Fig. 3f, g), revealing nearly identical electronic band features. Aligning with the previous report37, the valence band maximum is a six-folded carbon and sulfur band (Γ–A) with a bandwidth of 293 meV for the as-synthesized material. Evacuation results in contraction in the π-stacked TTF column by approximately 3%, which is quantitatively consistent with the experimentally observed contraction, and a corresponding increase in the bandwidth to 333 meV due to the increased S 3pz-orbital overlap. Since TTF is a hole acceptor40 and the curved bands are associated with the π-stacked TTF columns, both forms of Cd2(TTFTB) should behave as p-type semiconductors. The increase in valence band curvature should lead to a reduction of hole effective mass and an enhancement of hole mobility.Cd2(TTFTB) is known to contain TTF·+ radical cations37, which are likely formed through spontaneous oxidation of the TTF moiety by O2 during the high-temperature synthesis of the precursor or framework. If the relatively high σDC, humid air had been caused exclusively by TTF oxidation, the concentration of TTF·+ radical cations would be significantly higher in aerobic atmospheres than that in anaerobic atmospheres. To investigate this possibility, we conducted continuous wave electron paramagnetic resonance (CW-EPR) spectroscopic characterization on the same sample of Cd2(TTFTB) under both aerobic and evacuated conditions at 298 K. Both CW-EPR spectra exhibit single axial peaks with \({g}_{\parallel }\) and \({g}_{\perp }\) centered at 2.0068 and 2.0013, respectively (Fig. 3a). These are close to the free electron value (g = 2.0023), confirming the presence of TTF·+. Notably, the evacuation barely changes the concentration of radicals as indicated by nearly identical peak intensities and shapes in the two spectra. Thus, the presence of O2 does not promote further oxidation of Cd2(TTFTB) at room temperature and in turn does not improve the hole density. This is consistent with the abovementioned observation that σDC, dry air ≈ σDC, dry N2.Therefore, upon evacuating Cd2(TTFTB), its charge mobility increases, and its charge density persists, which together should enhance the electrical conductivity. Although technical limitations prevented us from performing these measurements in dry N2, the same trend is expected as dry N2 purging also removes portions of guest molecules. This is contrary to the experimental observation that σDC, humid air » σDC, dry N2. It implies that in humid air, besides TTF·+-based hole conduction, other charge transport mechanisms and/or charge carriers are involved, which are likely mediated by coordinating water and adsorbed solvent molecules.AC electrical characterizationTo investigate the contribution of water to apparent electrical conductivity in Cd2(TTFTB), we performed alternative-current (AC) electrochemical impedance spectroscopy (EIS) and DC I − V measurements, which reveal AC and DC conductivity, respectively, for the same device at 298 K under humid air, humid N2, dry air, and dry N2. The Nyquist plot obtained from EIS characterization under each atmosphere appears to be a semicircle. It can be well fitted by an equivalent parallel circuit consisting of a resistor and a capacitor. The resistance was used to calculate the AC conductivity (Supplementary Fig. 16)32. In each atmosphere, the AC and DC conductivity values are nearly identical — they differ by less than 1%. Take the abovementioned single-crystal device for example. Its AC conductivity in humid air was σAC, humid air = 9.38 × 10−5 S·cm−1, dropped to σAC, dry air = 1.06 × 10−6 S·cm−1 in dry air and σAC, dry N2 = 1.03 × 10−6 S·cm−1 in dry N2, and recovered to σAC, humid N2 = 5.26 × 10−5 S·cm−1 in humid N2 (Fig. 4a, b). These values match closely with DC counterparts (Fig. 2d).Fig. 4: AC electrical characterization of Cd2(TTFTB).a, b Nyquist plots acquired at 298 K under various atmospheres. c DC and AC conductivity acquired at 298 K under various relative humidities for the same device. d Schematic illustration of the hole transport pathway and the proton conduction channel coupled with redox reactions at the crystal−electrode interface.We further conducted EIS and DC I − V measurements for several devices at 298 K and in the air with the relative humidity ranging from 20% RH to 90% RH. Both AC and DC conductivity values increase exponentially with increasing relative humidity, and they match well with each other (Fig. 4c and Supplementary Fig. 17). For instance, in an exemplary device, both started with 1.31 × 10−4 S·cm−1 at 20% RH and rose to 1.75 × 10−4 S·cm−1 at 90% RH. Extrapolating the conductivity value to 0% RH gives 1.30 × 10−4 S·cm−1, which is significantly higher than the expected value in dry air and dry N2 likely because the material was desolvated to a higher degree in these dry atmospheres. These observations confirm the key role of water in the apparent electrical conduction. Hence, we tentatively assigned the AC conductivity to proton conductivity. Considering the ratio between σhumid air and σdry air observed from both DC and AC electrical characterization, the proton conductivity is at least 1 − 2 orders of magnitude higher than the electrical conductivity in humid atmospheres.Electrical conduction mechanismsWater may contribute to the apparent electrical conduction in Cd2(TTFTB) through two mechanisms. On one hand, water could provide protons and help form their conduction pathways32. Meanwhile, although the TTF moiety is not oxidized by the air at room temperature, it could undergo redox reactions at the crystal−electrode interface under an applied electric field, giving rise to so called interfacial pseudo-capacitance41. The coupling between proton conductivity and interfacial pseudo-capacitance may improve the apparent electrical conductivity value, resulting in an overestimation of the latter (Fig. 4d). This phenomenon has been observed in [(CH3)2NH2]In(H4TTFOC) (H8TTFOC = tetrakis(3,5-dicarboxyphenyl)-tetrathiafulvalene) in which proton conduction and interfacial redox reactions of the TTF moiety improve the apparent DC conductivity by five orders of magnitude41. On the other hand, water could be deprotonated under an applied electric field, which facilitates the formation of holes through proton-electron coupling. Such self-doping process improves hole density and thereby electrical conductivity, as evident by molecular conductors comprising of TTF moieties and protonic functional groups (e.g., hydroxy)31,42,43.Both mechanisms require redox-active components and hydrogen-bonded networks. Although there are terminal water molecules in the structure, their locations and orientations prevent direct formation of long-range hydrogen-bonded networks (Fig. 1c). This explains the relatively low DC and AC conductivity in dry atmospheres. These distant coordinating water molecules might be connected by adsorbed water and DMF to form hydrogen-bonded networks in the pores to promote proton conduction41,44. The proton conduction may couple with interfacial redox reactions of TTF moieties to contribute to the DC conductivity41. Meanwhile, such hydrogen-bonded networks could also promote self-doping that improves hole-based electrical conduction through π−stacked TTF columns. The latter is unlikely to be dominant because non-conjugated linkages (e.g., Cd2+−carboxylate coordination) between hydrogen-bonded networks and TTF columns should lead to weak proton-electron coupling. Although contributions differ, both mechanisms may be applicable to Cd2(TTFTB), rendering it as a potential candidate of proton−electron dual conductor (Fig. 4d)45,46.
