Water is one of the most abundant resources on earth, and hydrovoltaic technology1–8 is promising to address energy crises and industrial pollution issues. Among which the most delicate and microscopic form is droplet-induced electricity generation (DEG)9–18, mainly via an electrokinetic mechanism to generate power based on diverse interactions between nanomaterials and water molecules. For the last decade, one of the main trends of DEG is to harvest environmental energy via various methods, such as drawing18, flowing11, splashing10, etc. The induced voltage originates from the moving, establishing, or vanishing of the electric double layer (EDL)18,19. Graphene/polymer composite membrane is the most typical DEG device form20–23, with the generation voltage regulated by the droplet velocity18, the ion species14,18, graphene layers10, and the substrates11. In principle, hydrovoltaics originate from the trilateral interactions11,16,24 among ions (from the drops), electrons (from the conductors), and charges (of the substrate). It is governed mainly by two sequentially occurring interaction pairs: the one between the surficial substrate charges and the ions in the liquid, denoted as wi−c; and the other between the ions and the electrons, denoted as wi−e. In detail, in the moving drops, counterions are firstly preferentially adsorbed out due to the attraction (wi−c, electrostatic force) from the substrate charges and keep the EDL ever evolving. Then the ions drag electrons (wi−e, Coulomb drag) in the conductors to move in accord with the macroscopic drop movements (Fig. 1a). This electricity generation induced by in-plane ion movements above conductors, is termed in-plane electrokinetics (IPE, Fig. 1a). Its primary mechanism is the direct forcing of carriers by fluctuating Coulumbic fields, more commonly known as streaming currents25.
For most reported DEG, conductors need to be sandwiched by the liquid drops and the substrates, thus the ion-charge paring (wi−c) is always separated by the conductors and unnecessarily suppressed, no matter how thin. Additionally, the conductors have almost no choice but only monolayer graphene, as the suppressed wi−c significantly further decays or even vanishes at elevated conductor thickness larger than the two-dimensional range (Fig. 1b). To sum, the compromise between wi−c and wi−e results in the suppressing wi−c, and also the case of Hobson’s choice when selecting the conductors. Recently the selection principle of DEG’s conductors has been concluded as adequate electron conduction within a 2D-confined thickness2,16. Although with few alternatives reported13,17, 26–29, this principle severely constrains DEG's study and application scenarios.
Hereby a novel DEG paradigm was proposed to override the compromise between wi−c and wi−e, by geometrically reconfiguring ion movements, termed out-of-plane electrokinetics (OPE). Firstly, we designed a porous ultrathin polyethylene membrane-substrate to establish direct wi−c between ions diffusing interiorly inside the substrate and the negative charges of substrates (Fig. 1c). Then the preferentially adsorbed ions interact with the conductor’s electrons to generate hydrovoltaics. Both intimate contacts of ion-charges and ion-electrons, annihilated the compromise between wi−c and wi−e. Thus, the long pending wi−c suppression issue in DEG could be solved (Fig. 1d). As a bonus, OPE considerably expands the available spectrum of material species and dimensions for conductors of electrokinetic power generation.
Out-of-plane Electrokinetics Of Gpe Nanomembrane
Based on previous work30, surficial negative charging of ultrathin polyethylene (PE) membranes was confirmed, satisfying the basic criteria for DEG substrate. The zeta potential of the PE nanomembrane with a thickness of 109 nm was − 81.2 mV (Supplementary Fig. 1) in 0.001 mol/L NaCl solution, which represented the EDL built on the surface of the PE nanomembrane. We then engineered them to be facilitated with the channel diameter designed to distribute around tens of nanometers. This channel configuration could drive microfluids to diffuse in Knudsen mode31 with intensified interactions between ion and channel walls, enabling fine control of the fluid movement inside the membranes. Detailed information on the fabrication, morphology, thermodynamics, optical and mechanical property of PE nanomembrane were introduced in Supplementary Fig. 2–6.
In spite of the newly unlocked multiple choices of DEG conductors, we still started with graphene as a control group. The CVD-grown monolayer graphene with the size of 5 × 5 cm2 was transferred to the PE nanomembrane by polymer annealing and catalyst layer etching32 (Supplementary Fig. 7). And interestingly, the semi-freestanding monolayer graphene with centimeter square area was obtained (Fig. 2a), as considerable share of it suspends in air self-standingly with maximum spanning length over 25 nm (Supplementary Fig. 8c). This was achieved by suspending the graphene above the PE ultrathin nanoporous structure (Fig. 2b and Supplementary Fig. 9), with the thickness of only 109 nm (Supplementary Fig. 10). This graphene/PE ultrathin composite nanomembrane was denoted as GPE (the photograph can be seen in Supplementary Fig. 11a). The quality of monolayer graphene after the transfer was characterized by Raman spectroscopy (Supplementary Fig. 12). Two Ag electrodes were constructed at the two ends of GPE nanomembrane and locally encapsulated with silicone to avoid direct contact with water (Supplementary Fig. 11b). The Cu wires were linked to the electrodes to detect the electrical signal (Supplementary Fig. 11c), and the resistance of the GPE device was 5.90 kΩ (Supplementary Fig. 13). Before studying OPE, IPE was firstly tested with the device, by flowing ionic droplets on the surface of the GPE (Supplementary Fig. 14), and the results were in line with existing reports18.
Being different from the IPE which converts the momentum of external dynamic energy into electricity, the OPE mainly harnesses surface tension (chemical potential) to generate power. Specifically, OPE utilizes the spontaneous and continuous capillary micro-pumping effects of PE ultrathin nanomembranes to actuate the inner-membrane ionic microflow (Fig. 2c left). E.g., by simply floating the GPE above the 0.6 M NaCl aqueous solution in the ambient environment without inputting any external work, a persistent voltage over 6 hours was recorded as shown in Fig. 2d. Power generation via this continuous micro-pumping phenomenon was coined as pumping potential, serving as the basis for the OPE in this work. Yet the modulation of purely spontaneous capillary pumping can be challenging due to numerous determining factors and subtle regulation effects. Thus, we artificially intensified this process by ethanol dripping to instantly provoke much more violent voltages to quantitatively investigate the OPE mechanism (Fig. 2c right). Ethanol or other volatile solvent drops functioned as the primer to thoroughly purge the air in PE capillary channels, rapidly mixed with the aqueous solution, and then evacuated off during vaporizing to accelerate the micro-pumping.
The experimental set-up of primer-induced intensified pumping potential is shown in Fig. 2e, where the GPE device floated on the 0.6 M NaCl solution. First of all, a location dependence of micro-pumping on OPE voltages was reported. Three droplets of 50 µL ethanol were sequentially dripped on along the straight line connecting the electrodes on GPE. The dripping spots were labeled as a, b, and c, where b was at the center of the device while a and c were 2.0 cm left or right to the center. Their corresponding voltages were 4.03 mV, ± 0.3 mV, and − 4.18 mV, respectively (Fig. 2f). The results showed that micro-pumping at two bilateral symmetric locations (a & c) around the device center produced almost opposite numbered voltages. And the center-voltage (b) was seemingly neutralized by two counter potentials. It is the first time that a static position dependence of DEG voltage was discovered, which suggests a unique mechanism of OPE compared to IPE. More details can be found in Supplementary Fig. 15 with the mechanism further explained below.
The relatively faint yet persistent spontaneous pumping potential could be treated as a background signal in this work. To validate this assumption, we switched off and on this background signal by simply taking up and releasing back of the floating GPE off/on the NaCl solution surface. The pumping potential quickly dropped to null when taking up GPE from the solution (Fig. 3a). When releasing back to floating, the pumping potential peaked to ~ 3 mV, followed by the restoration of spontaneous pumping potential of persistent 0 ~ 1 mV voltage. The peak voltages were attributed to the impulse momentum of releasing GPE. It has been found that the peak voltages were proportional to the volume of the primer dripping and thus the micro-pumping flux (Supplementary Fig. 16c). As a result, the pumping potential elevated from 3.17 mV to 5.76 mV when primer dripping increased from 25 µL to 100 µL (Fig. 3b). The signal duration increased linearly with the primer volume (Fig. 3c). Thus, the peak voltage and background voltage were attributed to primer-induced intensified micro-pumping and the purely spontaneous micro-pumping, respectively. The former was the main research object to explore the mechanism of OPE.
And here, it was further assumed that each intensified micro-pumping site functions as an independent OPE power source——one pumping, one power. We found the intensified pumping potential decayed from 2.12 mV to 0.90 mV as three 15 µL primer droplets dripped sequentially on the same site of the GPE nanomembrane (Fig. 3d). If the primer droplets were dripped on different spots, three sharp peaks ranged from 2.39 mV to 2.15 mV were produced (Fig. 3e). This proved that the intensification effects on exactly the same power source (pumping spot) by the primer decayed dropwise. While with the same usage of a primer, triggering new micro-pumping on different spots to establish multiple new individual power sources with higher efficiencies.
Thickness and porosity of PE nanomembrane were also identified as the factors determining pumping potential. As shown in Fig. 3f, the voltage dropped from 8.47 mV to 0.88 mV as the thickness increased from 109 nm to 1028 nm (Supplementary Fig. 10) (150 µL primer dripping). And to shut down the pores, the PE nanomembrane was annealed at 140oC for 10 min before transferring the graphene (the morphology can be seen in Supplementary Fig. 17). The transformation of the porous semi-crystalline structure of PE was based on the thermally induced recrystallization of polyethylene (Supplementary Fig. 4), the detailed explanation can be found in Supplementary Fig. 17. The voltage decreases from 4.12 mV to 0.95 mV after the pores shutting down (Fig. 3f, inset) (150 µL primer). Thus modulating the thickness or porosity of the GPE nanomembrane can also regulate the intensified micro-pumping to adjust the pumping potential. To sum up, putting aside the background purely spontaneous pumping potential, the intensified pumping potential could be modulated by pumping flux, the number of intensified pumping sites, thickness, and porosity of PE substrate.
Mechanism
To uncover the mechanism of this out-of-plane electrokinetic effect, we further in-depth investigated the capillary micro-pumping (kinetic part), and the induction of electrons (electronic part). We first focused on the purely spontaneous pumping potential by simulating the capillary micro-pumping process in Fig. 4a. The PE nanomembrane was simplified as a model with three layers of fiber bundles lapped together. The Red in the picture represents the water phase, along with the blue gas phase, and white solid PE phase. Under the capillary force, the water phase gradually spread to the solid phase and finally covered the solid surface at 1.045 µs. Therefore, trace ionic solutions could spread through the substrate via pores of the GPE nanomembrane even without primmer intensification (Fig. 4d), resulting in the persistent voltage.
Then the mechanism of intensified pumping potential was studied. A series of four experiments were performed. The first experiment was designed to directly uncover the trans-membrane movements of the capillary micro-pumping process. As shown in Fig. 4b, the ionic solution was substituted by the red ink, and by spontaneous micro-pumping, it could not penetrate through PE nanomembrane to dye DI water due to the hydrophobicity of PE membrane (Fig. 4b i). However, when the primer was dripped into the water the red ink underneath the membrane was quickly pumped up to color the above water (Fig. 4b ii, Supplementary Fig. 18). The second experiment was to test the intensified pumping potential without ionic solution. And no voltage was detected with NaCl solution removed (Supplementary Fig. 19a). These two sets of results elucidated more of the physical picture of the capillary micro-pumping. That is after dripping, the ethanol droplet spread on the surface of graphene (the contact angle on ethanol on PE was about 20°, Supplementary Fig. 20) and partially penetrated through the defects of graphene into PE nanomembrane. When ethanol came into contact with the permeated NaCl solution, they quickly dissolve. Then the solution was forced to move upwards to accelerate the micro-pumping, and eventually generated the peak voltage. Then physical parameters of micro-pumping intensification were investigated. In the third experiment, the ethanol was replaced by acetone as the primer to alter the chemical potential configuration, the resulting intensified pumping potential decreased by 83.2% due to less miscibility (Supplementary Fig. 21b). In the fourth experiment, we found that the micro-pumping could not only be intensified by incorporating primer (chemical potential driven), but also could be achieved by external mechanical work. E.g., the intensified pumping potential could reach 3.53 mV by blowing air to the middle of the GPE surface (Fig. 4c). As the macroscopic pressure was evenly distributed microscopically to drive NaCl solution moving through substrates. Therefore, distinct forms of forces including capillary pumping, miscibility process, and mechanical pressing functioned equivalently to initiate DEG based on GPE nanomembranes. The last two sets of experiments proved that the primary driving force of the micro-pumping could be modulated and even shifted, as long as the trans-membrane ion movement was maintained, which should be the core process for the pumping potential.
In Fig. 4f, we verified the electron migration in graphene by fabricating four equally spaced electrodes on the GPE nanomembrane. The total voltage (V1,4) was 1.05 mV, approximately the sum of all the individual voltages (~ 0.93 mV) of three subsections on the GPE nanomembrane. It demonstrated that as a result of the accumulation of Na+, the electrons of graphene moved towards micro-pumping spots (wi−e). This result was distinct from similar experiments of other DEG, such as waving potential and streaming potential. As the wi−e of OPE is not instantly induced by the blinking EDL fluctuating potentials, but by the enduring potential from the accumulated counterions. Thus, during electrokinetics in this work, the kinetics part and the electronic part got decoupled at a centimeter scale.
The electron migration has also been elucidated based on the ‘one pumping, one power’ hypothesis. As shown in Fig. 4g, when two primer droplets (50 µL) were dripped simultaneously parallel to the inter-electrode line, the produced voltage (7.84 mV) was nearly double that induced by single primer droplet dripping (3.79 mV). By contrast, the voltage kept constant (Supplementary Fig. 22b), while the current doubled (Fig. 4h) when two primer droplets (50 µL) dripped perpendicular to the inter-electrode line. In the equivalent circuit, the above corresponded to the series and parallel power supply models, respectively, with each pumping spot equivalent to an independent power source.
Previously reported DEGs were all based on IPE where ions in-plane moved above substrates, such as drawing potential, flowing potential, or waving potential. Here, the ions of pumping potential uniquely travel through the substrate. In this way, e.g., Na+ was dragged by Coulomb force and moved through negatively charged channels of GPE nanomembranes (wi−c) during pumping, to induce the directional migration of the electron of graphene (wi−e) (Fig. 4e). Thus, the mechanism of OPE was distinguished in DEG by the through-plane ion moving direction perpendicular to electron migrating.
For traditional DEG devices based on graphene, the resistance inevitably reaches tens of kΩ due to the doping of impurities adsorbed on the surface, and the irreversible rupture during the transferring and handling of large-area monolayer graphene. This undoubtedly suppresses the electrokinetic conversion of the DEG, with the manifestation of only a few hundred nanoamps of current. The community long expects the utilization of other materials other than graphene, but is impeded by the shielding effect in IPE. As we discussed above, the OPE eliminates the shielding effect of the conductive material on the negative charge of the substrate, with a wide range of conductor utilization. We then employed three typical conductive materials such as Ag nanowires (AgNWs), carbon nanotubes (CNTs), and Pt nano-particles on the PE nanomembrane (the morphology can be seen in Supplementary Fig. 23) to assemble the DEG devices. The conductor layer thickness was 77 nm, 5286 nm, and 301 nm, respectively (Supplementary Fig. 24). As shown in Fig. 5 (a-c), the pumping potential could reach 0.27 mV, 3.16 mV, and 40.78 mV, respectively, and its sign depended on the position where the primer dropped on the membrane (the pumping current can be seen in Supplementary Fig. 25). i.e., when primer droplet dripped on the left and right ends of the membrane, the signs of the voltages were opposite. It could be seen that the potential peaks of the three conductors other than graphene were hump-shaped and the signal duration significantly increased, especially for Pt, which lasted 56 times longer than that of graphene. As graphene screened out most of the primer infiltration while for the three porous conductors, a considerable part of primer directly interacted with the ionic solution to reinforce micro-pumping in a persistent and intensified manner.
As shown in Fig. 5d, compared with graphene, the improved conductivity of these three conductors produced much higher pumping currents and thus boosted electrokinetic conversion. In Fig. 5e, we compared the output power of the OPE devices based on four types of conductors. A drop of 50 µL of primer could generate approximately 407.6 nW of power for Pt, which was 357 times higher than that of graphene. In contrast with the specific power of previously-reported DEG devices typically below 10 nW/µL, the Pt exhibited a leading specific power of 177.2 nW/µL with a minimum ionic solution volume of 2.3 µL (Fig. 5f). The above findings enabled stereo-configuration of OPE devices, paving the road to three-dimensional design of environmental energy-harvesting system.