On-Chip Tuning of Superconductivity in Fullerides via Current-Driven Rb+ Intercalation Authors: Konstantin P. Shchukin, Oliver N. Gallego Lacey, Baptiste Coquinot, Jacek Jakowski, Jingsong Huang, Patrik Staudenmayer, Yannic Falke, Ram Prakash Pandeya, Alexander Grüneis Published: 2026 ABSTRACT An in-operando electro-intercalation method for the on-chip synthesis of alkali-metal-intercalated materials and their Raman spectroscopic and transport characterization in ultrahigh vacuum (UHV) is developed. We apply this method to synthesize fulleride superconductors via Rb+ intercalation into a C60 film. During the intercalation, we monitor the stoichiometry via UHV-Raman spectroscopy and probe superconductivity via transport measurements. An increase of the superconducting transition temperature from 7.0 K to 14.5 K is observed when the stoichiometry is tuned from Rb2.7C60 to Rb3C60. In our experiment, an ionic Rb+ flux into the host material is induced by an applied electronic current via a Butler–Volmer-type mechanism. Electro-intercalation captivates through improved stoichiometric precision, the ability to smoothly vary stoichiometry via duration of current application, and the absence of a lower limit of the volume of the host material. It represents a powerful concept for the on-chip synthesis of intercalated materials, battery research, and beyond. Introduction Ionic transport is of fundamental interest in materials science and is applied in chemical functionalization of fullerenes and van der Waals (vdW) materials, , iontronics, , and battery technology. , Alkali metal ion intercalation can turn fullerenes into superconductors with an A3C60 stoichiometry (A = Rb, K, Cs). There is a renewed interest in superconducting fullerides for several reasonstheir superconducting properties can be controlled by light, C60 grows epitaxially on topological insulators enabling superconductor-topological insulator interfaces, and the phase diagram of ternary fullerides is still poorly explored. Alkali metal intercalation into fullerenes has been achieved by intercalation from the vapor phase, ,− liquid phase, , and solid-state electrolytes. Vapor-phase intercalation of alkali metal ions requires an annealing step in order to disperse the alkali metal ions throughout the sample where they occupy tetrahedral and octahedral sites in between C60 molecules. For example, a temperature of ∼200 °C is required for synthesizing the bulk Rb3C60 phase ,, which is superconducting around 30 K. Fulleride films have been studied via the van der Pauw method which was applied for vapor-phase-synthesized superconducting K3C60 on Si and for K3C60 and Rb3C60 films on the glass substrate. The traditional intercalation methods have limitations and impede progress in the field toward smaller films and more precise stoichiometries. Vapor- and liquid-phase intercalations have poor control over stoichiometry, and alkali metal intercalation via a solid-state electrolyte often leads to the degradation of the electrolyte. Especially, for small and thin fullerene films, tuning of the precise A3C60 stoichiometry by vapor-phase intercalation is a challenge. The required annealing step is not always possible on-chip, and the intercalation from the liquid phase cannot be performed in a device configuration. Fulleride superconductivity has been explained by the large density of states from the triply degenerate lowest unoccupied molecular orbital (LUMO) combined with strong electron–phonon coupling (EPC) − to the Hg(2) phonon around 440 cm–1. In addition, contributions to superconductivity from Jahn–Teller effects and electronic coupling mechanisms have been noted. , Recently, an in-plane covalently bonded fullerene monolayer has been synthesized via Mg intercalation. In order to address the challenges related to stoichiometrically precise synthesis of fullerides and their in-operando characterization, a new experimental approach is required, which unites intercalation, spectroscopy, and transport measurements in a single setup. In the present work, we perform on-chip synthesis of superconducting Rb3C60 using electro-intercalation of Rb+ into a fullerene film. We monitor the intercalation of air-sensitive Rb+ on-chip via Raman spectroscopy and electrical transport measurements in UHV conditions. , Electro-intercalation is related to a current-driven overpotential across C60 grains that lowers the barrier for electron injection into the C60 grain. The injected electron reduces the energy required to detach and intercalate Rb+. The resulting kinetics are well captured by a Butler–Volmer description. By the applied direct current, we fine-tune the Rb stoichiometry x in-operando and verify x via UHV-Raman spectroscopy. For a given x, we then measure the temperature-dependent resistivity ρ­(T) and the superconducting transition. The fingerprint Ag(2) Raman mode has an energy of 1469 cm–1 in bulk C60 and 1449 cm–1 in bulk Rb3C60, allowing determination of the stoichiometry Rb x C60. We measure the intensity of C60 and Rb3C60 modes during application of an electronic current to prove the Rb+ intercalation. Results and Discussion In-Operando Rb Intercalation and Sample Characterization Figure a depicts the sketch of the experimental setup with the sample plate with four contacts mounted on a cryomanipulator, an inverted flange with a long working distance optical objective (NA = 0.55), the Rb getter source (SAES), and the C60 evaporator inside a UHV chamber. The cryomanipulator is mounted on a stage with encoded motors so that Raman mapping can be performed in UHV. Figure b shows the substrates that have been used in this study: (1) Al2O3 wafers and (2) silicon wafers with a 20 nm thick Al2O3 film grown by Atomic Layer Deposition (ALD) as shown in Figure b (Ia and Ib). Using UHV-compatible conductive silver epoxy, the substrate corners were contacted via 20 μm Au wires to the spring-loaded contact pins. Upon insertion of the sample plate into the cryomanipulator, all pins are electrically connected to the nanovoltmeter and precision current source outside UHV. This setup allows us to transfer the sample out from the cryomanipulator into a heater stage for in situ annealing (not shown). After loading the electrically contacted sample plate into the cryomanipulator, we performed evaporation of 12 monolayers (MLs) of C60 onto the sapphire (Figure b II) followed by Rb evaporation (Figure b III). This entire process is monitored real-time by UHV-Raman spectroscopy and resistivity measurements enabling precise doping control. Figure c shows a photograph of the Omicron-type sample plate with four spring-loaded pin contacts attached to the sample in van der Pauw geometry. The backside of the sample is attached to the sample holder using a droplet of the same UHV-compatible conductive silver epoxy. This floating configuration prevents any electrical shortening of the sample surface toward the holder during or after Rb deposition. In order to determine the location of the laser spot on the sample, we measured the distances relative to the four contacts which we identified by collecting the reflected light in a camera (Figure S1). (a) Sketch of the setup for in-operando synthesis of Rb x C60. (b) Sample contacts and growth of Rb3C60. Ia: sapphire wafer with four contacts. Ib: 20 nm ALD-grown Al2O3 on Si. II: C60 on Al2O3/Si. III: Rb3C60 on Al2O3/Si. (c) Photograph of a sample holder with the contacted sample and the four spring contacts. The red square indicates the scanning window of 0.5 mm × 0.5 mm where the Raman map in Figure was recorded. The UHV-Raman spectrum of the C60 film on sapphire is shown in Figure a along with a typical spectrum of a Rb-doped film with Rb2.85C60 stoichiometry in the range 1430–1470 cm–1 which corresponds to the most intense Ag(2) phonon mode. This sample was synthesized via evaporation of Rb onto the C60 film in UHV at 300 K. The UHV-Raman spectra were recorded at T = 5 K. This temperature was independently verified by measuring the superconducting transition of a Nb film. Due to its large Raman cross section, the spectra of several MLs of doped and undoped C60 can be efficiently measured with a large signal/noise ratio. The Raman spectrum of the alkali metal-doped fullerene molecule depends on its charge state and has distinct frequencies for the observed phases: neutral, x = 1, x = 3, and x = 6. , For a neutral phase, it is a single peak at 1470 cm–1, Rb1C60 has a peak at 1458 cm–1, Rb3C60 at 1449 cm–1, and Rb6C60 at 1432 cm–1. For a mixed phase, several of these peaks appear simultaneously. After determining the relative Raman cross sections for each phase, we used Raman spectroscopy to determine the stoichiometry x of a sample with an arbitrary stoichiometry and Raman components from C60, Rb1C60, and Rb3C60 (see Methods and Figure S2). This procedure was used for the determination of x of the sample shown in Figure a and yields x = 2.85. Doping homogeneity is confirmed by Raman mapping of the Ag(2) peaks inside the energy window shown in Figure a (see Figure S3 for individual scans). The small red square in Figure c indicates the extent of the spatial map. Figure b shows a Raman map of the stoichiometry x in Rb x C60 from a Raman line shape analysis yielding a spatial average of the stoichiometry of x = 2.85. Figure c shows Raman maps of the peak areas from the line shape analysis for the Ag(2) peak of the initial pristine C60, the Ag(2) peak corresponding to Rb3C60, and the peak of pure C60 that remains after evaporation of Rb (indicated as remaining C60). The Raman maps demonstrate high lateral homogeneity of both pristine and doped films, with relative standard deviations of the fitted peak areas in a range between 1.3 and 4.8%. (a) Integrated Raman spectrum of Rb2.85C60 from the maps shown in (b) and (c). (b) Raman map of the stoichiometry x and (c) corresponding Raman maps of the initial Ag(2) C60 peak and two Ag(2) peaks of Rb3C60 and remaining C60. (d) Atomic force microscopy image of a 15 nm thick C60 film on Al2O3 scanned in window 1 μm × 1 μm. (e) Sketches of C60 grains (I) and partially (II) and fully (III) intercalated C60 grains with small amounts of Rb left in between the grains. (f) Raman spectra of Rb x C60 and resistivity ρ characterization. 10 μA current was used to intercalate Rb. (g) ρ versus stoichiometry at T = 300 K. All Raman and transport measurements were carried out in ultrahigh vacuum. The sample morphology of our C60 films was probed by atomic force microscopy (AFM). A typical AFM image of the C60 film on a sapphire is shown in Figure d. The AFM confirms the nanostructured nature of the film reported in previous measurements , of K3C60 and AFM measurements of C60. AFM analysis reveals that the 15 nm thick film consists of C60 grains with a lateral correlation length of 49 nm and an average roughness of 0.693 nm. The results of the AFM analysis are used in the modeling of the Rb+ intercalation described later. The nanostructured nature is crucial for Rb intercalation because it provides a large effective surface, according to which the Rb intercalation is progressing from the outside of the grain toward the inside. Hence, a partially intercalated C60 grain has a C60 core and an outer Rb x C60 shell separated by a moving intercalation front. The progressive Rb intercalation into C60 grains is shown in Figure e. The excess Rb that is not intercalated into a grain forms a network of conducting channels that run in between C60 grains. Because of the relatively poor electrical conductivity of the network of Rb channels, a significant voltage drop can appear across the film. This situation is reminiscent of a metallic wire network across which large voltage drops have been observed. The voltage drop across the sample is important for the intercalation mechanism, as described later. We now discuss Raman and electrical transport measurements of in situ synthesized Rb x C60. Figure f depicts UHV-Raman spectra for sample stoichiometries between undoped (x = 0) and fully doped (x = 6) fullerides. For stoichiometries in the range x = 3, fullerides were metallic, and ρ­(T) was recorded in addition to Raman. For metallic fullerides, we can determine x from Raman and correlate it with an ρ­(T) trace and the appearance of superconductivity. In Figure f, we synthesized the phases Rb1.35C60 and Rb2.71C60 conventionally by two consecutive steps of Rb evaporation and annealing. In order to precisely control the stoichiometry for x beyond x = 2.71, we performed electro-intercalation of Rb+ by passing a 10 μA current through the sample for ∼96 h at T = 300 K which allowed for precise control of the stoichiometry. The electro-intercalation experiment was performed with the identical setup used for the van der Pauw measurements. For electro-intercalation, the applied current was ten times larger than for resistance measurements. In order to supply the Rb, electro-intercalation was performed during the deposition of ∼170 MLs of Rb. We also performed electro-intercalation without Rb evaporation onto the C60 film. In this case, an excess amount of Rb was made available by deposition prior to the electro-intercalation. This deposited Rb acts as a reservoir from which Rb for electro-intercalation was supplied (see the next section). The electro-intercalation shown in Figure f was stopped at x = 2.90 and at x = 2.98 after which the sample was cooled down to T = 5 K. Hereafter, a Raman spectrum and a ρ­(T) trace (right panel of Figure f) were recorded. The Raman fits of stoichiometries of Figure f are shown in the Supporting Information (Figure S4). Both the Raman spectra and ρ­(T) confirm Rb+ intercalation and the increase in stoichiometry x. Our experiments reveal that electro-intercalation at room temperature is a slow process with a rate of stoichiometry change Δx = 3 × 10–4 per 1 h. The rate of stoichiometry change corresponds to an advance of the intercalation front by a distance of 0.83 pm per hour. Hence, stoichiometry tuning is enabled by this method with a high precision that cannot be achieved by other intercalation methods. The ρ­(T) for the three stoichiometries (x = 2.71, 2.90, and 2.98) showed vanishing resistance at low temperatures which is consistent with previous reports. , This indicates that samples prepared by electro-intercalation have consistent superconducting behavior compared to samples intercalated by other methods. The stoichiometry of the samples was independently determined by UHV-Raman. This allows us to establish a precise relation between stoichiometry and ρ­(T), in particular the T at which ρ vanishes. We estimate the transition temperatures from the onset of vanishing resistance (T onset) and the vanishing resistance (T zero). T onset is the temperature of the local resistivity maximum and T zero is the temperature, at which the dρ­(T)/dT < 5 × 10–4 mΩ·cm K–1 in the temperature window shown in Figure f. We find that T onset is weakly dependent on x, while T zero has a strong x dependence. We rationalize this in terms of the Aslamazov–Larkin (AL) model and Berezinskii–Kosterlitz–Thouless (BKT) fluctuations. , The AL model was previously used to describe the excess conductivity of K3C60 and the temperature regime above the critical temperature. We associate this regime with the region around the local maximum in ρ­(T), in which ρ­(T) depends on the sample dimensionality. The BKT fluctuations between temperatures T zero and the critical temperature are sensitive to defects due to the incomplete stoichiometry for x ≠ 3 and may explain the large x dependence of T zero. In the present experiment, the superconducting transition broadens due to BKT fluctuations, which can be enhanced in thin films by disorder, inhomogeneity in the stoichiometry, Coulomb interactions, and surface scattering in systems with reduced dimensionality. Since the transition width also indicates the defect density, it helps to assess how far the stoichiometry is away from the optimal Rb3C60. In the present data, we observe the largest broadening for Rb2.71C60 and the sharpest transition for Rb2.98C60 in agreement to what is expected from BKT theory. We investigated Rb3C60 films of 12 and 30 nm thickness (Figure S5) and found that normal-state and the superconducting response show a clear thickness dependence, consistent with bulk transport. Figure g depicts ρ­(x) at T = 300 K for stoichiometries x in between x = 2.2 and x = 3 with Rb3C60 at the resistance minimum. There is also a residual conduction of electricity through the Rb network in between the grains. The residual conduction acts as a current path in parallel to the conduction through the doped C60 film. The resistance minimum is not affected by the parallel channel because the resistance of the Rb network is almost ten times higher than the resistance of the doped C60 film (see Figure S6). The Raman spectra and ρ­(T) for each data point in Figure g are shown in the Supporting Information (Figure S7). Current-Driven Electro-Intercalation: Experiment and Theory Experimental Observation of Rb+ Intercalation In order to explain the mechanism of electro-intercalation of Rb+ into C60, we performed additional experiments with intermittent DC, continuous AC, and continuous DC currents applied to a C60 film with excess Rb evaporated onto its surface. During electro-intercalation, we also monitored the voltage drop across the sample and determined ρ­(t). Compared to dedicated measurements of the resistivity shown in Figure , where a current of 1 μA was applied, the ρ­(t) data shown here was recorded using a current of 10 μA. The larger current which is dissipated onto a partially intercalated sample and the fact that no compensation of thermoelectric voltages could be performed leads to a higher noise level. Yet, the observed time-dependent trends in ρ are unaffected by the magnitude of the current used which is corroborated by evaluating Raman spectra before and after current application. Figure a shows the Raman spectrum collected before starting the electro-intercalation (at time t = 0). From this Raman spectrum, we evaluate the initial stoichiometry as x = 2.54. The presence of a Rb3C60 peak confirms that a fraction of the deposited Rb has already intercalated. Figure b depicts ρ­(t) measured during an electro-intercalation experiment applying an intermittent 10 μA current for 10 s with a period of 3 h in between two consecutive steps of current application. Figure c depicts the AC electro-intercalation experiment at a frequency of 1 Hz. We observe an increase in ρ­(t) during the intercalation, indicating that no Rb+ is intercalated into the C60 grains. The Raman spectrum (Figure d) taken immediately after AC electro-intercalation attempt confirms a slight reduction of Rb doping of the sample stoichiometry to x = 2.50. We rationalize this observation via a slow deintercalation rate which is independent of an applied current. The next experiment investigates intercalation under continuous DC application (Figure e). Here, ρ­(t) shows a decrease over time, suggesting the Rb+ intercalation into C60 grains. The Raman analysis shown in Figure f confirms Rb intercalation by showing a clear change in stoichiometry from x = 2.50 to x = 2.69 within 2.6 × 105 s application of a 10 μA DC. This experiment confirms our earlier data of Figure f that the application of a 10 μA current indeed intercalates Rb to C60 grains and can be used as a means to fine-tune Rb x C60 stoichiometry at room temperature. We now estimate what fraction of the 10 μA current is ionic. The ionic current required to explain the stoichiometry change Δx ≈ 0.2 in Δt ≈ 2.5 × 105 s (Figure d,f) for the intercalation of N C60 ∼5 × 1014 fullerene molecules is IRb+=eΔxΔtNC60≈60⁡pA (see Supporting Information). This is 5 orders of magnitude lower than the current of 10 μA which was used in the experiment. We thus conclude that the current is mainly electronic. (a) Raman spectrum at t = 0 with Rb2.54C60 stoichiometry. (b,c) Time-dependent resistivity ρ­(t) during intermittent DC current (each 3 h for 10 s) and during AC current (1 Hz). (d) Raman spectrum at t = 5.22 × 105 s (Rb2.50C60). (e) ρ­(t) during continuous DC current. (f) Raman spectra at t = 7.97 × 105 s with stoichiometry Rb2.69C60. The AC and continuous-DC resistivity traces were smoothed using a moving-median filter with an effective time window of 2 × 104 s and a second-order Savitzky–Golay filter. Theoretical Description of Rb+ Electro-Intercalation We explain the observed electro-intercalation of Rb by three essential processes that describe the intercalation of Rb across the vdW gap to a C60 grain (process I), the Rb+ diffusion inside a grain of C60 and on its surface (process II), and the current-driven Rb+ intercalation into C60 grains (process III). Process I dominates until the easily accessible Rb has been used up, and process II is very fast but depends on available Rb. The rate-limiting process III is controlled via the applied current and described by the Butler–Volmer model. Processes I–III are summarized and described in Table . 1Processes That Describe the Intercalation of Rb+ into C60 process noprocess descriptionprocess parameterenergy barrier E kinetic roleIRb intercalation through the vdW gap between the Rb shell and C60 Rb–C60 distance E = 0 (vdW distance) E ≈ 200 meV (10 Å)not rate-limitingIIdiffusion of Rb inside the C60 grain and on its surface, leading to Rb intercalation of C60 diffusion coefficient D of Rb inside C60 E ≈ 200 meV not rate-limitingIIItunneling of an electron followed by Rb+ intercalation described by Butler–Volmer kineticselectron tunneling barrier between Rb and Rb3C60 E ≈ 400 meVrate-limiting and voltage controlledaProcesses I and II describe the intercalation from the vdW Gap and the Rb+ diffusion. Process III is the tunneling of an electron followed by the Rb+ drift and is controlled via the applied voltage and rate-limiting.bEstimated for the Rb-poor case. Processes That Describe the Intercalation of Rb+ into C60 Processes I and II describe the intercalation from the vdW Gap and the Rb+ diffusion. Process III is the tunneling of an electron followed by the Rb+ drift and is controlled via the applied voltage and rate-limiting. Estimated for the Rb-poor case. Process I explains why a fraction of deposited Rb intercalates spontaneously into C60 after Rb deposition. If Rb is within the vdW distance to C60 (i.e., 4.7 Å), the Rb can intercalate without having to overcome an energy barrier. We have performed calculations of the barriers for Rb intercalation into a C60 film with the climbing image-nudged elastic band method (CI-NEB) at three distances: the vdW distance, 10 Å, and 20 Å. For the vdW distance, the intercalation process has no barrier, i.e., it is spontaneous. The barriers for 10 Å and 20 Å are 0.24 and 0.92 eV, respectively (Figure S8). Bader charge analysis along the reaction path shows that the migrating Rb atom is neutral on the left side of the transition state where the driving force arises mainly from the cohesive energy of the Rb slab. It is clear that spontaneous Rb migration will reduce with increasing vdW gap size as all the easily available Rb is used up. Process II describes diffusion of Rb+ on the surface of a C60 grain and inside a C60 grain. In our calculation, we consider a Rb-poor case where one Rb+ is intercalated in an otherwise empty C60 lattice and a Rb-rich case where a single Rb vacancy in Rb3C60 is introduced. The Rb occupies tetrahedral centers of Td point group symmetry (denoted as Td ) and octahedral sites of Oh point group symmetry (denoted as O h ). Our results indicate effective Rb transport via a repeating O h → T d → O h sequence along symmetry-equivalent directions which is more effective than transport via O h → O h and T d → T d hopping. The calculated energy barrier for the T d → O h hop of Rb is ∼0.19 eV for the Rb-poor case and ∼0.49 eV for a vacancy hop for the Rb-rich case (see Figure S9). The energy of the O h site is about 0.05 eV higher than the T d site. For the energy barrier in the Rb-rich case, we obtain a diffusion constant D ≈ 4 × 10–15 m2/s at 300 K. The diffusion time is estimated as t = L 2/(2dD) with dimensionality d = 1 for thickness-limited diffusion and a film thickness L = 15 nm as t ≈ 10–2 s. Hence, t is negligible compared to the observed time scales for the intercalation. Processes I and II provide fast local equilibration of available Rb+ within and around individual grains. However, this easily available Rb is generally not sufficient to synthesize Rb3C60. As we will show in the following, an applied bias is required to reach Rb3C60 stoichiometry. In process III, the intercalation of Rb+ into a C60 grain proceeds via the Butler–Volmer mechanism in the presence of an applied current. This process is rate-limiting and sets the multiday time scale of the experiment. If a current is applied to a network of connected C60 grains surrounded by Rb, there is a voltage drop across each C60 grain. This is similar to metal nanowire networks where a large voltage drop inside the film of interconnected nanowires has been observed. Due to the voltage drop, the barrier for electron tunneling from the Rb metal shell into the C60 grain is reduced. An electron that has tunneled into the C60 grain can attract a Rb+ ion. In the following, we estimate the tunneling barrier. The cross section of the heterostructure is Rb/Rb3C60/C60/Rb3C60/Rb. Process III involves three subprocesses 1–3 that are depicted in detail in Figure a. The first subprocess is an electron tunneling from the Rb metal on the left contact into the center of the C60 grain. The second subprocess is a Rb+ ion being attracted by Coulomb force from the surrounding Rb to the C60 grain. Due to the presence of an electron inside the C60 core, the Rb+ ion that detaches from the Rb metal on the right side has to overcome only a small van der Waals barrier. The third subprocess is that an electron flows from the right contact to the voltage source and closes the circuit. Let us consider the tunneling subprocess 1 in more detail. The tunneling gap between the Rb metal and the Rb3C60 grain is indicated in Figure a. In Figure b, we consider the band diagram across a partially intercalated C60 grain of total stoichiometry x = 2.50. The band diagram was constructed based on the literature values of electron affinity of C60, χC60 = 2.7 eV, the work function of Rb, ϕRb = 2.1 eV, the work function of Rb3C60 is taken as ϕRb3C60=2.7 eV, the HOMO–LUMO gap of C60 is taken to be Eg = 2.3 eV, and the Fermi level of C60 is located 0.9 eV below the LUMO, giving a work function ϕC60 = χC60 + 0.9 = 3.6 eV. In order to evaluate the band bending of C60 at the interface to Rb3C60, we consider the Fermi levels of isolated C60 and Rb3C60 with a common vacuum level. The Fermi level of Rb3C60 lies at higher energy than the Fermi level of C60 and electrons flow from Rb3C60 to C60. In the present system of a partially intercalated grain, there is a common Fermi level across the interface between the Rb3C60 shell and the C60 core. Hence, when electrons are transferred from Rb3C60 to C60, the bands of C60 bend downward close to the interface. In Figure b, the band bending can be seen. The voltage drop is chosen such that the left contact is at negative potential w.r.t. the right contact. (a) Electro-intercalation of Rb+ from outside the vdW distance. Sketch of cross section across a core–shell structure (Rb outside, Rb3C60 inside). Electron tunneling and Rb+ motion are indicated. (b) Band diagram of a Rb/Rb3C60/C60/Rb3C60/Rb heterostructure. Electron tunneling through the gaps occurs along the direction of increasing x. (c) Calculated effective tunneling barrier between Rb and Rb3C60. The inset denotes the barrier height versus the Rb–Rb3C60 separation. Details of the calculations are provided in Figure S10 of the Supporting Information. (d) Calculated ionic current according to the Butler–Volmer model along with the stoichiometry change (right vertical axis). Rectangles indicate the time span probed experimentally. The inset shows the experimentally relevant stoichiometry change versus time. Figure c depicts the effective tunneling barrier for the electron tunneling between two quantum wells resembling the Rb metal and Rb3C60 with a vacuum gap in between (see Figure S10 of the Supporting Information). For the calculation of the tunneling barrier, we considered the Volta potential and effect of image charges. The Rb metal and Rb3C60 are modeled as quantum wells, and bringing them closer together results in a reduction of the tunneling barrier that the electron needs to overcome when tunneling from the Rb metal into Rb3C60. For a distance of 6.06 Å, we calculated an effective barrier height equal to ∼0.4 eV. The inset of Figure c shows the barrier as a function of Rb–Rb3C60 distance. Let us now calculate the ionic Rb+ current using the Butler–Volmer model. This model provides the interfacial ionic current density j(η) = eΓs[k f(η) – k b(η)], where k f, k b are forward/backward rate constants, Γs is the effective number of sites on the C60 surface (here the surface density of C60 Γs ≈ 2/nm2), η is the overpotential, and e is the elementary charge. Denoting l = 50 nm the typical size of the grains of C60 and ΔV the total voltage drop, we estimate η = ( l /L)­ΔV ≪ k B T/e. Using transition-state theory (TST) with transmission factor κ ≈ 1 for a small overpotential η, we get the interfacial ionic current by unit of the Rb/Rb3C60 interface, which we model as a section of the device for each layer of grains, that is, wL × L/ l . The total current in the Butler–Volmer model reads IRb+≈κwLΓse2ΔV2πℏexp(−ΔG0‡kBT)exp(−2κtund) 1 Here, ΔG 0 ‡ is the energy barrier which corresponds to the energy required for promoting the electron into the top of the C60 LUMO. The last term is a tunneling rate of the electron throughout the gap of length d with transmission of κ=2mU/ℏ , where U is the effective barrier and m is the free electron mass. In order to make eq time-dependent, we model the effective number of C60 sites with a phenomenological Avrami model. This model is generally applicable for all the A x C60 material system. Using a current IRb+ that does not depend on the grain size, we can extract an effective ionic current: 2 IRb+eff(t)=edx(t)dtNC60=IRb+·exp(−IRb+3eNC60t) Figure d depicts the plot of IRb+ as a function of the intercalation time t. Here, we used ΔG 0 ‡ = 0.55 eV, U = 0.5 eV, and d = 0.65 nm. Overall, the predicted intercalation dynamics agrees well with the experimental results. In the Supporting Information, we discuss the roles of electromigration and field-induced Rb ionization in Rb+ intercalation. Electron–Phonon Coupling in Rb3C60 Thin Films When intercalating Rb+ into the C60 crystal, the electronic structure changes to a metal with high density of states at the Fermi level which boosts electron–phonon coupling (EPC). The EPC of Raman active Rb3C60 phonons can be determined by Raman spectroscopy from the full width at half-maximum (FWHM) increase when going from C60 to Rb3C60. In the presence of EPC, the phonon lifetime is reduced, which leads to an increase of the FWHM in Rb3C60. ,, Therefore, measuring the increase in the Raman mode’s FWHM upon changing C60 to Rb3C60 allows one to determine the EPC constant λ i of a particular phonon. , There are 37 Raman active phonons in C60, and FWHM Raman analysis has been used to determine the dimensionless EPC constant λR given by 3 λR=∑iλi=∑idiπN(EF)ΔΓiωi2 In eq , ω i , d i , and ΔΓ i are the unrenormalized discrete phonon frequency, mode degeneracy, and mode broadening for the i-th mode, respectively. N(E F) is the DOS at the Fermi level, which is equal to N(E F) = 15 states/eV for Rb3C60. Previous works found a dominant contribution of the Hg(2) phonon mode which contributes λ­[Hg(2)] ≈ 0.2, , and a total EPC constant for all Raman active modes λR ≈ 0.5 was claimed for thick Rb3C60 films. The contributions to the FWHM of Raman modes are EPC, phonon–phonon scattering, and defect scattering. The effects of phonon–phonon scattering depend on temperature, and we find only minimal change in the FWHM of Hg(1) and Hg(2) phonons at 5 and 300 K. The defect scattering contribution is minimized by using phase-pure Rb3C60 in our experiment. In addition to Raman spectroscopy, EPC can also be determined from dρ/dT, the slope of the resistivity with temperature. This determination of EPC includes all phonons (also Raman inactive phonons) whose energy is less than k B·T (T is the temperature in which the linear slope is observed and k B the Boltzmann constant). , Thus, the higher energy optical phonons observed in Raman do not contribute to the EPC constant measured in transport. To highlight the difference to λR, we introduce λtr, the EPC determined from dρ/dT. Previously, λtr was determined in the temperature range between 100 K and 350 K at constant volume from the measured slope dρ/dT , according to 4 λtr=0.246(ℏωp)2dρ/dT In eq , ℏωp (the plasma energy) is to be taken in units of eV and dρ/dT is to be taken in units μΩ·cm/K. The reported values for ℏωp from literature have a wide range, e.g., ℏωp = 0.6 eV (optical spectroscopy) and ℏωp = 1.2 eV (theory). Using dρ/dT ∼2 μΩ·cm/K and ℏωp = 1.2 eV, λtr = 0.65 was obtained. The EPC constant enters the McMillan formula for the critical temperature T C, which reads 5 Tc=ℏωln1.2kBexp(−1.04(1+λ)λ−μ*(1+0.62λ)) For Rb3C60, μ* = 0.1 and an average phonon frequency ωln = 1300 cm–1 from theory have been used. For λR = 0.5, a value of T C = 17 K was predicted (the actual T C of the Rb3C60 sample from transport was T C = 28 K). There is a discussion ,− if the McMillan formula is applicable in the case of Rb3C60 because the bandwidth of the electron energy bands is close to the phonon energy and correlation effects may play a role. Usually, the inclusion of correlation effects lowers T C, but in the case of fullerides, correlation effects and EPC can cooperate to increase T C. , No effective theory exists that predicts T C of Rb3C60 in the presence of correlations and EPC. Let us now determine EPC in Rb3C60 from Raman and transport measurements. We have synthesized a superconducting Rb3C60 film in situ and measured ρ­(T), from which we extracted the transition temperatures T zero and T onset and the slope dρ/dT. We compared the observed transition temperatures to the critical temperatures estimated from EPC constants obtained from Raman and transport. Figure a shows Raman spectra of C60 and Rb3C60 in the wide energy range. We employed eq to determine EPC from the FWHM of each Raman active phonon mode (Figure S11). Overall, our FWHM are in good agreement to previously published works. ,, In Figure b, we plot the contribution of each Raman active phonon mode to the EPC constant λR. It can be seen that, in particular, the Hg(2) phonon mode has a large contribution to the total λR, in agreement to previous works using Raman spectroscopy on Rb3C60 ,, and a recent ARPES study on K3C60. We obtain a total EPC from Raman λR = 0.334 from which we estimate T C via eq to obtain T C = 0.78 K which is considerably lower than the experimental value of T C (c.f. Figures f and c). For estimating T C, we have used an experimental value ωln = 441 cm–1 which is lower than the previously used ωln. We explain our lower value of ωln as follows. In the calculation of ωln, each phonon mode is weighed with its EPC constant, and because of the dominance of H g(2) EPC, the resulting ωln is small in comparison to an ωln weighted by similar EPC constants. The transport EPC λtr is determined from the experimental slope whose fit is shown in Figure c yielding dρ/dT = 0.827 μΩ·cm/K which is a factor ∼2 lower than in previous experiments. The previous experiment was carried out at constant sample volume. The volume of the present thin film sample is constrained by the Si substrate with a ∼10 times lower linear thermal expansion coefficient than Rb3C60 making the present and previous experiments comparable. From dρ/dT, we obtain a value of λtr = 0.293, and from eq , we obtain T C = 0.71 K. We consider the similarity of λR and λtr coincidental because different phonons contribute to λR and λtr. We investigated the superconducting transition of the same sample used for Raman and dρ/dT measurements. Figure c depicts ρ­(T) and dρ/dT yielding T onset = 24.5 K and T zero = 9.2 K which is in disagreement to the estimation of T C from λR and λtr. The following three points are important when discussing this disagreement. First, not all Rb3C60 phonons are Raman active and λR and hence T C are underestimated. Second, in λtr, only low-energy phonons contribute whose energy is below the linear dρ/dT range. The above two points explain the underestimation of T C by considering either λR or λtr. It is thus likely that both λR and λtr contribute to the total EPC. Third, if correlation effects are present and enhance EPC, , we underestimate T C from our EPC analysis. (a) Raman spectra of C60/Al2O3/Si and Rb3C60/Al2O3/Si. The insets show the Hg(2) peaks with a large electron–phonon coupling (EPC) constant in the case of Rb3C60. The total EPC of this sample was λR = 0.334. (b) EPC constant of individual phonon modes versus the phonon energy. (c) Rb3C60 temperature dependence of resistivity and its derivative dρ/dT. In the linear regime, we obtain dρ/dT = 0.827 μΩ·cm/K yielding an EPC constant λtr = 0.293 and T C = 0.71 K. The superconducting transition of this sample is characterized by T zero = 9.2 K and T onset = 24.5 K. Conclusions and Outlook We have introduced electro-intercalation as a method for on-chip intercalation chemistry where ions are driven by electronic currents. Electro-intercalation can be used to control the stoichiometry in order to fine-tune phase transitions. In the present work, we have controlled the superconducting phase transition in fullerides by Rb+ intercalation. Alkali metals with a low work function are particularly suited for electro-intercalation because the corresponding barriers for electron tunneling are small. We conclude that electro-intercalation works well for a nanostructured C60 film that consists of grains because such a film provides a high density of Rb/Rb3C60/C60 interfaces. An applied current leads to a potential drop over a C60 grain which facilitates electron tunneling into the C60 grain followed by the intercalation of a Rb+ ion. The exciting aspects of this experiment are its high level of stoichiometric precision, ultimate cleanliness ensured by UHV conditions, and spatially homogeneous doping. Moreover, there is no lower limit for the volume of the host material that can be intercalated. This doping method could also be applied to the on-chip synthesis of fulleride superconductor heterostructures. In particular, topological insulator (TI)-fulleride interfaces are interesting candidates for superconducting fulleride growth on the TI and can probably be synthesized by electro-intercalation. Electro-intercalation also allows, in principle, for the synthesis of ternary fulleride superconductors such as Rb2CsC60 with a higher critical temperature or polymeric Mg4C60. The presented experimental setup can also be used for experiments on light-induced superconductivity in Rb3C60, where the four-point resistance measurement is carried out during light irradiation. Future experiments could also include a microfabricated on-chip Rb or Cs ion source such as the on-chip Yb source built recently. This ion source would allow for shuffling ions into the sample region. A future theoretical work could consider the unified description of processes I–III by a single theoretical method. We expect that electro-intercalation works not only in C60 but also in a large variety of organic materials such as picene, phenanthrene, coronene, graphene nanoribbons, and metal–organic frameworks and could be used to search for and identify new alkali metal-doped superconducting materials. Experimental Section Sample Preparation In the first step (Figure b II), a C60 film was deposited on the Al2O3 (sapphire) substrate by evaporation from a quartz crucible resistively heated to 400 °C. The base pressure in the UHV system was p ∼ 2 × 10–10 mbar. The deposition rate was 5 Å/min monitored and controlled using a quartz crystal microbalance (QMB). During film growth, the substrate was kept at 120 °C, followed by postannealing at 200 °C for 5 min. Hereafter, the C60 sample was immediately checked by Raman spectroscopy in the range of the strong Ag(2) Raman mode of C60, observed at 1456 cm–1 for C60 at temperatures T > 260 K and 1467 cm–1 at T < 260 K. In the next step, Rb doping of C60 was performed by evaporation of Rb from a well-degassed getter source. The Rb flux, calibrated using a QMB, was approximately 0.1 Å/min and simultaneously monitored throughout the doping process. As the evaporation was carried out with the sample mounted on a cryomanipulator, the resistance evolution could be recorded in situ, enabling the identification of the Rb3C60 phase by the characteristic resistance minimum. During doping, Raman spectra were simultaneously acquired, focusing on the evolution of the Ag(2) mode, which progressively shifted from the Rb1C60 position at 1458 cm–1 toward the single Rb3C60 peak at 1449 cm–1. The doping was continued until the spectrum exhibited only the Rb3C60 feature, indicating completion of the intercalation process. In order to resolve these peaks, all these Raman acquisitions were performed at temperatures less than 10 K. Raman Spectroscopy and Mapping Raman spectroscopy was performed using λext = 532 nm excitation wavelength with a laser power P ≈ 10 mW. The cryomanipulator in our experimental setup was equipped with two encoded stepper motors. It allowed automated sample translations in the plane perpendicular to the laser beam. The minimal stepsize for translation was 250 nm. The mapping spatial resolution was limited by the laser beam spotsize arising from excitation wavelength λext and numerical aperture NA of the objective, which in our case was estimated from the Airy pattern diameter as follows: d ≈ 1.22λext/NA ≈ 1.18 μm. Electrical Transport Characterization Electrical transport measurements were performed in a van der Pauw geometry with electrical contacts precisely placed at the four corners of squared substrate. The temperature-dependent resistivity and superconducting transition were measured with a precision current source Keithley 6220 and nanovoltmeter Keithley 2182A operated in Delta mode to suppress spurious thermoelectric voltages. The Delta acquisition frequency was set at 1 Hz, and the magnitude of the excitation current was 1 μA, ensuring negligible self-heating of the sample in the superconducting state. During the electro-intercalation process, the sample resistance was continuously monitored using a Sourcemeter Keithley 2400 operating in a DC mode with a constant current of 10 μA, while the voltage drop across the film was recorded simultaneously. Electrical transport measurements also allow for the verification of the film homogeneity by interchanging the current and voltage contact pairs and by rotating the measurement configuration by 90°, as permitted by the van der Pauw geometry. Considering the strict square substrate geometry and the accurate placement of the contacts in the corners, the variation in the measured resistances of all configurations did not exceed 5%. Rb x C60 Stoichiometry Estimation Using Raman Spectra We determine the Rb x C60 stoichiometry from the Ag(2) Raman peak components corresponding to C60, Rb1C60, and Rb3C60. For the analysis, we assume that the number of C60 molecules inside the laser spot is conserved during Rb intercalation. The peak areas of components C60, Rb1C60, and Rb3C60 in the i-th doping step are denoted by AiC60 , AiRb1C60 , and AiRb3C60 , respectively. Then x i , the stoichiometry in the i-th doping step, is given by 6 xi=α·AiRb1C60A0C60+3β·AiRb3C60A0C60 The parameters α and β are determined from a regression analysis as α = 5.84 and β = 2.95, respectively (Supporting Information). The relative Raman cross sections for C60, Rb1C60, and Rb3C60 are equal to 1, 1/α, and 1/β, respectively. Let us consider a general sample which has three Raman peaks and consists of C60, Rb1C60, and Rb3C60 fractions with relative Raman cross sections of 1, 0.17, and 0.34, respectively. We assume that the error bar of the Raman peak corresponding to each phase is given by the noise of the Raman spectrum. From the experimentally observed noise, we estimate at most 10% error for the Raman intensities (see Figure f). Considering a stoichiometry x close to x = 3, we typically have an intense Raman peak corresponding to Rb3C60 and a weaker or no Raman peak corresponding to C60. The cross section of C60 is ∼3 times higher than the cross section of Rb3C60. Therefore, if we have a Rb3C60 sample and see no C60 peak, it means the error bar of the stoichiometry x = 3 is about 10%/3 ≈ 3%. For mixed phases, the error can be slightly higher because of the contributions of all three peaks to the total error. Quantum-Chemical Calculations Density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) version 6.4.2. , Interactions between ion cores and valence electrons were treated using the projector-augmented wave (PAW) method. , Exchange–correlation effects were described using the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation (GGA) functional. The Rbsv PAW potential with a valence electron configuration of 4s24p65s1 and the standard C PAW potential with a valence electron configuration of 2s22p2 were used. A plane-wave energy cutoff of 500 eV and a 2×2×2 Monkhorst–Pack k-point mesh were applied. Structural relaxations were converged to 10–5 eV in total energy and 0.01 eV·Å–1 in forces. Reference structures for pristine C60 and fully intercalated Rb3C60 were obtained by relaxing the Fm3̅m cubic phase. The optimized lattice constant was 14.797 Å for C60 and 15.104 Å for Rb3C60 (∼2% expansion upon intercalation). Using the relaxed Rb3C60 structure, minimum-energy migration pathways for isolated Rb+ hopping and vacancy migration between the octahedral (O h ) and tetrahedral (T d ) pockets were calculated with the climbing-image nudged elastic band (CI-NEB) method, thereby providing the bulk migration barriers. To investigate Rb transport across the Rb–C60 interface, slab supercells were constructed containing a Rb slab in contact with two C60 layers. Three interface geometries were considered, including vdW-vdW, vdW-10 Å, and vdW-20 Å, to compare the Rb migration at vdW-sized and large imposed interfacial gaps. Two other interface geometries, 10 Å-10 Å and 20 Å-20 Å, were also constructed to determine the energy barrier for Rb migration across the 10 Å and 20 Å interfacial gaps. The in-plane lattice constants were fixed to those of the relaxed Rb+ structure, and a sufficiently large out-of-plane cell dimension was used to preserve the intended gap sizes. During relaxation, the middle layer of the Rb slab was held fixed to mimic bulk-like behavior of the Rb nanowire. All numerical settings for the slab calculations were identical to those used for the bulk systems, except that a Γ-centered 4 × 4 × 1 k-point mesh was employed and a 0.02 eV·Å–1 force tolerance was used during relaxations and CI-NEB calculations. CI-NEB was then applied to obtain minimum-energy paths and transition states (TS) for Rb migration across various interfacial gaps. Bader charge analysis was used to quantify the charge state of the migrating Rb atom along each CI-NEB pathway. Estimate of the Rb+ Diffusion Coefficient and Intercalation Time To obtain a realistic estimate of the Rb+ diffusion prefactor and diffusion coefficient in the Rb-rich lattice, we complemented the DFT CI-NEB barriers with finite-temperature molecular dynamics using Grimme’s extended tight-binding (xTB) Hamiltonian, , as implemented in the DFTB+ code. , The Rb3C60 Fm3̅m structure (four C60 and 12 Rb per cell) was first reoptimized at the xTB level, and NVT dynamics at T = 300 K were carried out with a Nose–Hoover thermostat for 5 ps simulation time and using a 1 fs time step. Selected Rb ions that remained localized in either O h or T d sites were monitored and their velocities projected onto the geometric hop direction between neighboring O h and T d pockets. From the velocity autocorrelation function and its Fourier transform, we extracted characteristic vibrational frequencies along the hop direction and, by multiplying with the coordination numbers of the corresponding sites, effective attempt frequencies. Supplementary Material