ARTICLE https://doi.org/10.1038/s41467-019-13791-1 OPEN Ultrafast transient liquid assisted growth of high current density superconducting films L. Soler 1, J. Jareño1, J. Banchewski1, S. Rasi 2,1, N. Chamorro 3,1, R. Guzman 1, R. Yáñez 3, C. Mocuta 4, S. Ricart1, J. Farjas 2, P. Roura-Grabulosa2, X. Obradors1 & T. Puig 1* 1234567890():,; The achievement of high growth rates in YBa2Cu3O7 epitaxial high-temperature super- conducting films has become strategic to enable high-throughput manufacturing of long length coated conductors for energy and large magnet applications. We report on a transient liquid assisted growth process capable of achieving ultrafast growth rates (100 nm s−1) and high critical current densities (5 MA cm−2 at 77 K). This is based on the kinetic preference of Ba-Cu-O to form transient liquids prior to crystalline thermodynamic equilibrium phases, and as such is a non-equilibrium approach. The transient liquid-assisted growth process is combined with chemical solution deposition, proposing a scalable method for super- conducting tapes manufacturing. Additionally, using colloidal solutions, the growth process is extended towards fabrication of nanocomposite films for enhanced superconducting prop- erties at high magnetic fields. Fast acquisition in situ synchrotron X-ray diffraction and high resolution scanning transmission electron microscopy (STEM) become crucial measure- ments in disentangling key aspects of the growth process. 1 Institut de Ciència de Materials de Barcelona, ICMAB-CSIC, Campus UAB, 08193 Bellaterra, Catalonia, Spain. 2 GRMT, Department of Physics, Universitat de Girona, Campus Montilivi, Edif. PII, E17003 Girona, Catalonia, Spain. 3 Departament de Química, Universitat Autònoma de Barcelona, 08193 Bellaterra, Catalonia, Spain. 4 Synchrotron SOLEIL, L’Orme des Merisiers Saint-Aubin BP 48, 91192 Gif-sur-Yvette, France. *email: teresa.puig@icmab.es NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 T he discovery of cuprate superconductors, 30 years ago, to induce transient liquids, i.e., liquids that do not exist in the raised enormous hope for widespread use of super- equilibrium ternary phase diagram for such compositions. This conductivity in many large-scale energy applications and idea results in growth rates of ∼ 100 nm s−1 for epitaxial YBCO powerful magnets. However, huge challenges needed to be films, as demonstrated in this work by in situ synchrotron X-ray addressed to provide a flexible conductor with high current diffraction studies. In addition, the T–PO2 window for hetero- transport capabilities. Coated conductors (CCs) revolutionized epitaxial growth is strongly enlarged and the process is shown to this area by growing epitaxial REBa2Cu3O7 (REBCO, RE = Rare be compatible with scalable solution deposition techniques. We Earth or Y) layers on top of buffered long length flexible metallic further prove that TLAG-CSD nanocomposite films, with high substrates. At present, they are the materials with highest current critical temperatures and current densities, can be grown with the capabilities and the most promising to extend large-scale appli- above depicted approach using colloidal solutions of 5 nm-size cations of superconductivity beyond those achieved with low- BaMO3 (M = Zr, Hf) preformed nanoparticles. These nano- temperature superconductors1–3. Remarkable relevance was composites display relevant nanostructural changes that posi- achieved by the growth of nanocomposites4–7, which strongly tively influence vortex pinning at high magnetic fields. The enhanced performances at high magnetic fields. However, the low synergistic effect between nanoparticles and local nanoscale throughput in the production of CCs remains a severe limitation lattice distortions (nanostrain) observed, brings additional to generate drastic cost reductions of CCs for wide market opportunities to further improved vortex pinning efficiency of implementation. TLAG-CSD nanocomposites. In addition, TLAG-CSD brings Chemical solution deposition (CSD) is a non-vacuum approach opportunities for simplified reactors and precursors flexibility, leading to CCs with very competitive performance, while having thus it is compliant with a high-throughput method for easier a strong potential for achieving the above-mentioned cost market penetration. reductions. Nevertheless, the low throughput owing to small growth rates (∼1 nm s−1) from the most extended trifluoracetate- route (TFA)8 still restricts from full development of CSD Results potentialities. TLAG principles. The TLAG process is based on the capability of The approach identified to reach high film growth rates, is that the ternary system BaCuO2, CuO, and Y2O3 to achieve ultrafast based on liquid-assisted growth9–11. In this case, fast atomic growth of YBCO films. This is possible because the eutectic diffusion and high atomic density encompass a much higher reaction between the first two compounds is able to form a growth rate than in vapor or solid diffusion reaction approa- transient liquid (with the presence of dispersed Y2O3 solid ches12. More recently, few examples of liquid-assisted growth of nanoparticles) in the region of the phase diagram where solid REBCO films are found, all based on vacuum deposition tech- YBa2Cu3O7-x is the equilibrium phase (light gray region of niques10,11,13–16. Fig. 1b). The idea is that there is an energy barrier for nucleation Here, we report on an outstanding process that goes beyond of the crystalline phase (which delays its formation through an the established schemes based on growth from liquids appearing incubation time), whereas the transient liquid phase can emerge in the equilibrium phase diagram. It combines CSD using non- instantaneously, thus enabling the development of an ultrafast fluorinated precursors and a non-equilibrium process grounded non-equilibrium processing approach17. on transient liquids, originating an ultrafast growth approach: Upon dissolution of the Y2O3 nanoparticles in the Ba-Cu-O Transient Liquid-Assisted Growth based on CSD (TLAG-CSD). transient liquid and Y diffusion toward the substrate interface The kinetic hindrances in crystallization are advantageously used (see diagram of Fig. 1a), epitaxial growth of YBCO occurs. The a b T (°C) 1000 900 800 700 600 10–1 Tp C Te uO 10–2 Cu 2O 10–3 PO2 (bar) Iins T route c 100 10–4 Epitaxial fraction (%) 90 (2:3) 60°Cs–1 –1 80 60°Cs (3:7) 60°Cs–1 PO2 route 70 10–5 (2:3) 20°Cs–1 –1 (3:7) 20°Cs 60 (3:7) PO2 route 50 –1 10–6 20°Cs 40 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 0.4°Cs–1 30 1000/T (K–1) 650 700 750 800 850 T (°C) Fig. 1 Transient liquid-assisted growth (TLAG) principles. a Schematic drawing of the TLAG process. b T–PO2 phase diagram for YBCO TLAG growth. The 2BaCuO2-CuO eutectic temperature, Te, the YBCO peritectic temperature, Tp, and the instability YBCO line, Iins, are shown. The T-route and PO2-route are indicated by arrows. Symbols are the final temperature and PO2 values of individual experiments: open symbols (2BaCuO2-CuO, 2:3 composition) or closed symbols (3BaCuO2-4CuO, 3:7 composition) in red (T-route) or blue (PO2-route); all symbols indicate 100% epitaxial growth. Te has been determined from ref. 27 and this article (see details in methods); Tp, Tins, and the CuO-Cu2O line are from ref. 10. c Epitaxial fraction dependence on growth temperature for the T-route at different heating rates for 2:3 (empty symbols) and 3:7 (solid symbols) compositions at PO2 = 10−3 bar. Error bars indicate the accuracy of the method as determined in ref. 7. 2 NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 ARTICLE a b d 800°C 80°C s–1 100 YBCO (005) 350 850°C Intensity (a.u.) 800 intensity (%) 75 Relative 300 600 T 50 YBCO thickness (nm) 250 400 25 PO2 route BaCO3 200 700°C 0 200 100 nm s–1 150 c e T route 10–4 bar 300 thickness (nm) Intensity (a.u.) 300 100 200 YBCO 250 PO2 50 –1 200 100 s 2 nm 150 2×10–6 bar YBCO 0 0 10 20 30 40 16.0 16.4 16.8 0 2 4 6 t (s) 2 (°) t (s) Fig. 2 Fast evolution of phases in TLAG by in situ XRD synchrotron radiation. a YBCO growth rate determined from in situ XRD synchrotron analysis of a 350 nm film demonstrating 100 nm s−1 growth rate for T-route (PO2 = 10−3 bar, dT/dt = 5 °C s−1) and PO2-route (T = 850 °C, PO2,base = 10−5 bar, PO2,final = 10−2 bar). Initial stages for the T-route show a growth rate of 2 nm s−1. b, c Evolution of the diffracted intensity of the (005) YBCO peak for the T-route and PO2-route, respectively, during the transformation for the experiment reported in a. The X-ray energy used for these measurements was 18 keV. The evolution with time of the XRD intensity for d BaCO3 and e YBCO phases in a T-route process in a RTA furnace at 800 °C (PO2 = 10−2 bar, dT/dt = 80 °C s−1) for a film of 300 nm. Notice that BaCO3 consumption to form the transient liquid goes in parallel with YBCO growth, which reaches completion in 3 s in these conditions. fast Y diffusion in the liquid may boost the growth rate. For c axis nucleation and growth window observed. This fact brings instance, the diffusivity of Y at 970 °C in the BaCuO2-CuO liquid unique opportunities for fine tuning of the film manufacturing is 4 × 10−10 m2 s−118, whereas in the YBCO solid is almost two process by employing rapid thermal annealing furnaces. decades lower, of only 8 × 10−12 m2 s−119. This makes liquid- When the two liquids are compared, one realizes that the 3:7 assisted processes potentially so much faster than solid–solid or eutectic composition is able to trigger the formation of c axis solid–gas reactions. We anticipate that the crystallization through YBCO at lower temperatures (solid versus open symbols of the liquid will be mostly limited by the substrate interfacial Fig. 1b, c). This behavior is in agreement with a lower kinetics of Y, Ba, Cu, and O atoms20,21. supersaturation for the 3:7 liquid composition. If CY is assumed Our results demonstrate that crystallization rates of at least similar for both liquids, since Ce increases with Cu content in the 100 nm s−1 are reachable. In addition, we show that the liquid liquid24, supersaturation will indeed be lower for the 3:7 liquid can be formed below the binary BaCuO2-CuO equilibrium composition25. eutectic temperature, Te (Fig. 1b), thus TLAG-YBCO epitaxial The best conditions achieved so far, demonstrating high critical growth is achieved in a wide region of the phase diagram (light current density in the superconducting films, are shown in Fig. 1b gray region of Fig. 1b). In summary, the kinetic crystallization as the dark gray encircled region (for both compositions and hindrances are exploited to induce much faster crystallization different TLAG routes investigated in this work, see details below). through a transient liquid. It is worth noting that reaching this high-temperature region in a In general, epitaxial film growth is ensured by properly tuning very short time lapse is very important because here is where the the supersaturation conditions. In the TLAG case, liquid growth rate is larger: ultrahigh values in the range of 100 nm s−1 composition11,22 and RE element solubility in the melt23 are have been demonstrated for T > Te, whereas two orders of the most relevant parameters. Several process parameters magnitude lower growth rates (2 nm s−1) are reached at lower (temperature, oxygen partial pressure PO2, total pressure of gases temperatures (see Fig. 2a). This is further confirmed in Fig. 2e, in the reactor Ptotal, heating ramp rates) have to be optimized for where the fast growth of TLAG is obtained from rapid thermal epitaxial growth for each composition. In this work, two different annealed samples quenched at 800 °C, demonstrating that 3 s are liquid compositions were studied, 2BaCuO2-CuO (2:3) and enough to grow a 300 nm YBCO film (i.e., 100 nm s−1) with a 3BaCuO2-4CuO (3:7), which correspond to the YBa2Cu3Ox and heating ramp of 80 °C s−1. YBa2Cu4.66Ox film compositions, respectively. The latter matches the Ba-Cu-O eutectic liquid composition and results in the desired phase (YBa2Cu3Ox) together with CuO nanoparticles, TLAG-CSD approach. The potential for high-throughput and mainly observed at the film surface. Both melts give rise to low-cost TLAG-CSD CCs is based on the idea of combining the epitaxial YBCO films. ultrafast growth of TLAG with CSD. A stable and atomically Our results reveal that the temperature window for c axis homogeneous solution of Y, Ba and Cu salts based on propionate epitaxy is widened to temperatures below Te and that this precursors is deposited on a substrate and pyrolyzed to reach a expansion depends on the heating ramp (down to 650 °C for the homogeneous nanoporous layer of Y2O3, BaCO3 and CuO 3:7 composition at dT/dt = 60 °C/s, Fig. 1b, c). In fact, as TLAG is nanocrystalline phases (details given in Methods). Then, the a non-equilibrium kinetically driven process, when high heating sample is heated up to the gray region of Fig. 1b, where upon rates are used, the Y ions dissolution from the Y2O3 nanoparticles BaCO3 reaction, a homogeneous dispersion of Y2O3(s) nano- into the melt might be compromised, and the Y concentration in particles is formed in the BaCuO2-CuO (liquid), to crystallize into the liquid, CY, will decrease. As a consequence, the relative epitaxial YBCO. supersaturation of the liquid, σ = (CY/Ce−1) (where Ce is the The TLAG-CSD process is compatible with multideposition equilibrium concentration), will decrease, leading to the enhanced of pyrolyzed layers and scalable ink jet printing patterning NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 techniques, where layer thickness is not only controlled by the between 10−5 and 10−2 bar. solution concentration but also by drop volume and drop spacing26. Pyrolyzed layers up to 1 µm thick deposited by ink BaCu2 O2 ðsÞþY2 O3 ðsÞ ! ½Ba  Cu  OðlÞþY2 O3 ðsÞ jet printing show a high degree of homogeneity throughout ð3Þ thickness (see Supplementary Figure 1a). These layers are ! YBa2 Cu3 O6:5 ðsÞþxCuOðsÞ transformed after the TLAG growth process to homogeneous, pore free, and epitaxial YBCO films of 550 nm (Supplementary where x = 0 stands for (2:3) and x ≠ 0 for (3:7) liquid Figure 1b), where straight twin boundaries crossing the total film composition. thickness can be identified. Higher magnification with atomic In this PO2-route case, ultrafast growth rates of 100 nm s−1 resolution scanning transmission electron microscopy (STEM) have also been reached (Fig. 2a) by controlling the steepness of confirms the highly crystalline quality of the epitaxial growth. the PO2 change. Direct proof of the transient liquid phase for The versatility and speed of the TLAG-CSD growth process has both, T- and PO2-routes, is given by the results of in situ XRD been further investigated by in situ fast acquisition x-ray experiment reported in Supplementary Figure 2) where the diffraction analysis at SOLEIL synchrotron (Fig. 2 and Supple- crystalline precursor phases disappear at the onset of YBCO mentary Figure 2) and advanced electron transmission micro- growth. scopy from quenched samples (Supplementary Figure 3). Both A detailed TEM analysis of the TLAG process for the T-route techniques enabled us to disentangle the phase transformations approach is shown in Supplementary Figure 3, where the stages of and growth mechanisms. A key point to understand the TLAG- the process with the different reaction phases and their CSD growth mechanism was the identification of the reaction of distribution are presented. The initial nanocrystalline phases BaCO3 with Cu-phases to form the Ba-Cu-O intermediate phases. observed after the pyrolysis (BaCO3, CuO, and Y2O3) are of tens We have investigated two different routes indicated in Fig. 1b of nanometers in size, so fast reaction paths are rendered possible. with arrows, temperature route (T-route, with red arrows) and The CuO nanoparticles, of size in the range of 10–30 nm, are oxygen partial pressure route (PO2-route, with blue arrow). In the distributed within a BaCO3 matrix. At an intermediate tempera- T-route, BaCO3 reacts with CuO during heating at a fixed oxygen ture step, the first heterogeneous nuclei of YBCO grow at the partial pressure (10−4 bar < PO2 < 10−2 bar); then the transient substrate interface whereas BaCO3 grains (size ~ 40-50 nm) liquid forms dissolving Y ions from Y2O3 to give rise to YBCO remain in the bulk of the film. This confirms that the slowest (see in situ X-ray powder diffraction (XRD) synchrotron studies reaction is the BaCO3 elimination. At the final stage, the YBCO reported in Supplementary Figure 2). This reaction process is epitaxial layer is observed and BaCO3 has completely disap- represented in Eq. (1) (not adjusted since we have used two peared. The BaCO3 reaction time is a key point in this process considerably different liquid compositions: 2BaCuO2-CuO (2:3) and strongly depends on heating rate and final temperature. At and the 3BaCuO2-4CuO (3:7)). 80 °C s−1 and 800 °C, for instance, only 3 s are needed to completely remove BaCO3 and to grow a 300 nm thick YBCO film (Fig. 2d, e). In some particular cases, homogeneously BaCO3 ðsÞþCuOðsÞþY2 O3 ðsÞ ! ½Ba  Cu  OðlÞþY2 O3 ðsÞ nucleated YBCO nanocrystals can eventually recrystallize, as ! YBa2 Cu3 O6:5 ðsÞþxCuOðsÞ suggested by synchrotron XRD analysis (Supplementary ð1Þ Figure 2a). We conclude that TLAG-CSD is an ultrafast non-equilibrium and versatile process to grow epitaxial YBCO films at c axis where the labels (s), (l) stand for solid and liquid, respectively, growth rates of ∼ 100 nm s−1 from different routes. This and x = 0 is for (2:3) and x ≠ 0 for (3:7) compositions used. is ~10–100 times faster than most of the thin film processes In the PO2-route (blue arrows of Fig. 1b), BaCO3 reacts during presently used8,29–32 to grow coated conductors and it confirms heating at low (and constant) oxygen partial pressure (10−5 bar < fast growth rates observed in other liquid-assisted techniques, PO2,base < 10−4 bar) in the region where YBCO is not the stable such as in REBCO single crystal growth, REBCO bulk ceramic phase (i.e., below the instability line, Iins, of Fig. 1b), following the melt-textured growth33 or REBCO films14. However here, the reaction process shown in Eq. (2). Then, a steep increase in PO2 process is not restricted to the use of an equilibrium liquid phase (10−4 bar < PO2,final < 10−2 bar) at constant temperature, brings and we demonstrate that the transient liquid formation is the film to the right region of T–PO2 in order to form the compatible with high-throughput CSD film manufacturing. transient Ba-Cu-O liquid that allows fast epitaxial growth of YBCO (Eq. (3)). Hence, in this route, the elimination of BaCO3 is TLAG-CSD film properties. Optimized conditions for the fully decoupled from the YBCO growth. TLAG-CSD process gives rise to highly epitaxial YBCO films of According to Eq. (2), initial BaCO3 and Cu2O react to generate 200–300 nm thickness (out-of-plane and in-plane textures of Δω BaCu2O2, which is the stable solid Ba-Cu phase at this low < 0.3° and Δφ < 0.5°, respectively, Supplementary Figure 4). Also, PO227,28, (Supplementary Figure 2c). very high critical current densities of 5 MA cm−2 at 77 K and 39 MA cm−2 at 5 K with transition temperature ~ 90 K are reached BaCO3 ðsÞþCuOðsÞþY2 O3 ðsÞ ! BaCO3 ðsÞþCu2 OðsÞþY2 O3 ðsÞ (Fig. 3 and Supplementary Figure 5) upon tuning the process ! BaCu2 O2 ðsÞþY2 O3 ðsÞ parameters (Ptotal, PO2, T, ramp rates, substrate). The particular case shown in Fig. 3 is for a 90 nm thick YBCO film using the ð2Þ PO2-route with the reaction mechanism of Eq. (2, 3). Liquid-assisted growth, however, exposes some challenges that After the steep increase in oxygen partial pressure (PO2 = PO2,final) need to be properly addressed for each type of route. Coarsening at constant temperature, the BaCu2O2 solid phase transform to of unreacted precursor phases in the liquid, liquid reactivity with Ba-Cu-O liquid and subsequent YBCO growth (Supplementary the substrates (or the buffer layers) owing to its highly corrosive Figure 2d), following the reaction process given in Eq. (3). nature and, improper substrates wettability with the viscous Again Eq. (2, 3) are not adjusted to generalize in the liquid liquid34 or trapped liquid at grain boundaries, are some of them. compositions employed. Notice that, for the mechanism detailed Nevertheless, we have proved that superconducting films with in Eq. (2, 3), PO2 is the crucial parameter, whereas Ptotal was kept very high critical current densities can be grown if the 4 NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 ARTICLE a 0.4 °C s–1 5K 10 Jc (MA cm–2) 77K 1 10 nm 1 b H 5K 20 °C s–1 0.1 JcIJcsf Pristine 12% BaZrO3 12% BaHfO3 0.1 1E-4 0.001 0.01 0.1 1 10 ab ab 0H (T) ab 0.01 1E-4 0.001 0.01 0.1 1 10 0H (T) Fig. 3 Inductive critical current density, Jc, as a function of the magnetic field for TLAG films. A pristine YBCO film and two nanocomposites of 12% 10 nm molar BZO and 12% molar BHO of composition 3:7 are shown for 5 K and 77 K. The growth conditions followed the PO2-route at temperature of 835 °C and PO2,final = 10−3 bar. Samples are 90 nm thick. The inset shows Fig. 4 STEM cross-section images of TLAG-YBCO films highlighting the the normalized Jc for the self-field value at 5 K for the same three films. differences in microstructure for the T-route. a at a heating rate of 0.4 °C The H* parameter is defined with the criterion of the value of H at 0.9 times s−1, where highly epitaxial film with a low density of defects (Y248 Jc at self-field. The irreversibility line obtained from electrical transport intergrowth) is shown, and b at 20 °C s−1 heating rate, that shows a highly measurements is shown in Supplementary Figure 6, displaying an distorted YBCO matrix with a high density of short Y248 intergrowths and µ0Hirr(77 K) ~ 8.1 T. nanoscale ab grains. Yellow arrows indicate some of the Y248 intergrowths. aforementioned obstacles are overcome, in agreement with previous liquid-assisted CCs growth14,25. conductors, especially for large magnet applications4–7,36,37. In Figure 4 shows a cross-section STEM image demonstrating the CSD, most of the effort has been done on adding metalorganic quality of the film growth. Each route has a distinctive salts to the precursor solution; a method that relies on the microstructure. However, the main defects are the Y2Ba4Cu8Ox spontaneous segregation process of nanoparticles during YBCO (Y248) intergrowths (double CuO chains), similarly to the TFA- growth37,38. However, this approach is rather complex with CSD process35. For the T-route, at low heating rates of 0.4 °C s−1 TLAG-CSD, as one would need to finely control two ultrafast a microstructure with only few Y248 intergrowth (indicated by quasi-simultaneous nucleation processes. Instead, we demon- yellow arrows) is observed (Fig. 4a), in agreement with the low strate that TLAG-CSD is compatible with nanocomposites growth rate identified at low temperatures, where most of the film growth when preformed nanoparticles are stabilized in the pre- would grow. At increased heating rates of 20 °C s−1, we observe a cursor solution. This result represents a second breakthrough for high density of defects, mainly double and triple short CuO chain the TLAG-CSD process, beyond the ultrafast growth. In this case, intergrowths (also marked with yellow arrows) and nanoscale ab a colloidal solution of BaMO3 (BMO, M = Zr, Hf) nanoparticles grains in the range of 5–10 nm distributed in the bulk of the film of 5 nm in diameter (Fig. 5a) was stabilized in the Y, Ba, Cu (Fig. 4b). This is in agreement with the ultrafast TLAG-CSD propionates-based metalorganic solution, in a similar way as for growth process achieved at high temperatures. Overall, we want TFA-CSD39. Cross-section STEM images of a nanocomposite to highlight that different microstructures are reached depending film, grown following the PO2-route, and its associated micro- on the process route and parameters, making TLAG-CSD a very structure are shown in Fig. 5b. Notice that the final nanoparticles tuneable process for vortex pinning. size, identified by STEM measurements, is in the range of 6–10 nm. These values have also been confirmed by XRD crys- tallite size analysis, evidencing a low degree of coarsening. Growth of TLAG nanocomposites. The growth of nano- Nanocomposites with large nanoparticle content (up to 32% mol) composites has demonstrated to be the path to pin vortices and have been prepared maintaining high Tc in the range of 88–92 K reach high critical currents at high magnetic fields in coated and epitaxial growth (Supplementary Figure 4 and 5). The NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 a c that the driving force for Y248 formation in this process is not (110) BZO pole provided by the minimization of large incoherent interfaces like 135° 45° in the TFA-CSD case7. Instead, we propose that the highly 2.6 defective YBCO matrix is induced by the ultrafast growth of 5 nm the nanocomposite film. It is known that the Y248 intergrowth is 180° 10° 20° 0° 2.0 one of the lowest energy defects able to accommodate large 30° 40° distortions in YBCO35. The high density of local nanoscale 50° 60° 70° 1.4 distortions can be characterized by XRD Williamson–Hall 200 nm 225° 80° 315° analysis in terms of nanostrain7. The corresponding increase of 220° the nanostrain with the density of nanoparticles (Supplementary Figure 5) is in agreement with previous analysis on TFA-CSD b nanocomposites, where local strained regions serve as strong artificial pinning centers due to strain induced Cooper-pair suppression7. BZO Figure 3 illustrates the effect of nanoparticles and associated 10 nm microstructure on the magnetic field dependence of the critical current density, Jc, for a 12% mol of BZO and 12% mol BHO BZO nanocomposite films of 90 nm thickness. Notice that the critical 6 nm current density of TLAG-CSD nanocomposites out-performs that BZO of pristine samples already above a magnetic field of 0.1–0.2 T, 6 nm even though the self-field values were not properly optimized yet. The much smoother in-field Jc dependence remains for the whole-temperature range and the crossover field identified by the 2 nm H* parameter (see Fig. 4 inset) can even reach values of 200 mT, BZO which are on the order of the highest results measured by SQUID in CSD YBCO nanocomposite films39. For the same values of nanostrain and therefore of local distortion, nanocomposites present a large increase of H* with respect to pristine samples (Supplementary Figure 5), showing that there is another pinning 10 nm STO mechanism acting beyond the nanostrain effect. We suggest that nanoparticles of 6–8 nm (diameter 2–3 times higher than the Fig. 5 Growth of TLAG nanocomposites. a TEM image of 5 nm BaHfO3 vortex coherence length, ξab) could be directly pinning vortices41, nanoparticles in the colloidal solution. b STEM image of a 12% Molar thus the aforementioned H* enhancement in TLAG nanocom- BaZrO3 (BZO) nanocomposite (in a 90 nm thick YBCO layer) on SrTiO3 posites might be partially governed by nanoparticles acting single crystal substrate, where few BZO nanoparticles are identified. The themselves as core pinning centers. Such a scenario might pave inset shows a high resolution image of the region with a 6 nm BZO the way to radical performance increase in CSD nanocomposite nanoparticle embedded in the YBCO matrix; strong lattice distortions and films upon continuing nanoparticle miniaturization and Y248 intergrowth generated are evidenced. c (110) Pole figure of the 24% the prevention of coarsening at high concentrations. We molar BZO nanocomposite where epitaxial orientation of the nanoparticles emphasize that the fast crystallization kinetics of the TLAG is demonstrated (data are fourfold symmetrized, XRD analysis performed approach can advantageously be used to push the frontier of at SOLEIL Synchrotron). nanocomposite-coated conductor manufacturing. self-field critical current density starts to decrease for nano- composites with nanoparticles contents beyond 24% mol, which Discussion can be explained by some agglomeration of the nanoparticles and We have developed a non-equilibrium process to grow YBCO minor non-oriented grains. films and nanocomposites at ultrafast growth rates (∼ 100 nm s−1), A remarkable feature of TLAG-CSD nanocomposites is that which can be combined with low-cost CSD and large area ink jet BMO nanoparticles tend to grow epitaxially with the YBCO printing deposition techniques. The breakthrough idea is the use of matrix (see (110) pole-figure of BZO in Fig. 5c). This effect is a transient liquid to grow epitaxial YBCO films. Critical current similar to that observed in techniques based on simultaneous densities of 5 MA cm−2 at 77 K and in-field enhanced properties deposition and growth (PLD, MOCVD)36,40 and it is opposite to for nanocomposites based on colloidal nanoparticles solutions the TFA-CSD route where nanoparticles are found randomly have been demonstrated. This growth technique appears very oriented7,37. Quantification of the epitaxial fraction following the promising for high-throughput preparation of YBCO films and if method reported in7 gives an epitaxial fraction in the range from demonstrated for long length thick coated conductors, it could 85 to 95% for the TLAG-CSD nanocomposites. This demon- widely spread the use of high-temperature superconducting strates that the liquid phase allows nanoparticles rotation inside materials for practical applications. Preliminary work on growth of the liquid to minimize interfacial energies. This rotation actually 1 µm thick TLAG-CSD films show no difficulties in eliminating occurs simultaneously to an enhanced crystallinity and some BaCO3 in any of the routes here reported, as well as no reactivity recrystallization of the BMO nanoparticles (Fig. 5b). On the on LaMnO3-buffered coated conductors. In our opinion, this contrary, in the TFA-CSD process, the solid-state hinders the suggests a great future for this methodology, easily scale-up and rotation of the nanoparticles, i.e., nanoparticles remain randomly integrated in simple reel-to-reel furnaces, to create a high- oriented, even if some recrystallization leading to faceted throughput and low-cost CC manufacturing schemes for fabrica- interfaces was also observed7. tion of coated conductors. It is based on very general kinetic and We also point out that the density of Y248 intergrowths is thermodynamic principles and could therefore be applied to grow extremely high in TLAG nanocomposites (Fig. 5b). This indicates other functional nanomaterials. 6 NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 ARTICLE Methods Physical properties. Inductive critical current densities were obtained by applying Sample preparation. Epitaxial c axis-oriented YBCO and BMO (M = Zr, Hf) the Bean critical-state model to thin films. The hysteretic magnetization curves nanocomposites thin films have been grown on single-crystalline (100) SrTiO3 sub- were measured with a commercial superconducting quantum interference device strates. Samples were prepared from metalorganic precursor solutions with two (SQUID) magnetometer from Quantum Design equipped with a superconducting stoichiometries: YBa2Cu3Ox and YBa2Cu4.66Ox. Y, Ba, and Cu anhydrous acetates magnet of 7 T. The H* parameter was determined from the Jc(H) curves with a were dissolved in propionic acid for four hours at 50 °C and afterwards anhydrous criterion of 0.9 × Jc at self-field at 5 K. Critical temperature measurements were methanol (in a 50/50) mixer was added to reach a 1 M–1.5 M concentration determined using a PPMS Quantum Design system using the Van der Pauw depending on desired film thickness. In some cases, a 5%v/v triethanolamine was also method and the derivative criterion. The irreversibility line was measured through added to reach larger thickness. Films of thickness from 90 nm up to 500 nm were a bridge of 40 µm in the same PPMS system using a criterion of R(Tirr)/R(95 K) = studied. Spin coating was used for films up to 300 nm thickness with a pyrolysis of 0.001. Film thickness was determined from a three-dimensional profilometer P16 500 °C for 5 min in an oxygen atmosphere, whereas a home-made ink jet printer with from KLA-Tencor. microfab piezoelectric nozzles was used for film thickness from 300 up to 500 nm. Superconducting YBCO films were grown in a Rapid Thermal Annealing furnace from Annealsys AS.Micro equipped with a SiC coated graphite susceptor, capable of Data availability reaching heating rates up to 80 °C s−1 with calibrated temperatures. The T-route or The data that support the findings of this study are available from the corresponding PO2-route explained in the main text were used as convenient. In the T-route, authors on reasonable request. temperature was increased up to a range between 650 and 850 °C at a heating rate between 0.4 °C s−1 and 80 °C s−1, at constant oxygen partial pressure between Received: 11 October 2019; Accepted: 29 November 2019; 5 × 10−4 and 10−2 bar (as shown in Fig. 1). Best results were obtained around 830 °C at 10−3 bar for a heating rate of 20 °C s−1. In the PO2-route, the film was heated up to the desired temperature (between 750 °C and 850 °C) at a heating rate between 0.4 °C s−1 and 20 °C s−1 at an oxygen partial pressure between 2 × 10−5 and 10−6 bar, and subsequently PO2 was fast (< 1 s) increased up to the range of 10−4–10−2 bar. Best results were obtained heating the film up to 835 °C at 20 °C s−1 at 10−5 bar with a fast change of PO2 up to 10−3 bar. The total pressure was also varied from 10−2 to 1 bar depending on the experiment and route used. Films were subsequently oxygenated in References a tubular furnace at 450 °C for 200 min in oxygen flowing atmosphere. BaMO3 (M = 1. Larbalestier, D., Gurevich, A., Feldmann, D. M. & Polyanskii, A. High-Tc Zr, Hf) nanoparticles were prepared following a polyol solvothermal method by superconducting materials for electric power applications. Nature 414, mixing Zr or Hf alkoxide precursors and Barium Hydroxide into a mixture of 368–377 (2001). alcoholic solvents39. This mixture was treated in an autoclave at 180 °C for 60 min to 2. Obradors, X. & Puig, T. Coated conductors for power applications: materials achieve the crystallization of the nanoparticles. BaZrO3 (BZO) of 5 nm of diameter challenges. Supercond. Sci. Technol. 27, 044003–044017 (2014). and BaHfO3 (BHO) of 4 nm in diameter were obtained in stable colloidal suspensions 3. Marchionini, B. G., Yamada, Y., Martini, L. & Ohsaki, H. High-temperature of 100–120 mM of concentration in ethanol. Afterwards, the colloidal BMO solutions superconductivity: a roadmap for electric power sector applicationh-temperature were stabilized in the YBCO precursor solutions to reach the desired final con- superconductivity: a roadmap for electric power sector applications, 2015–2030. centration of metal salts with different molar percentages of nanoparticles (up to IEEE Trans. Appl. Supercon. 27, 0500907–0500907 (2017). 32% mol). All solutions were characterized by water content, contact angle, viscosity, 4. Haugan, T., Barnes, P. N., Wheeler, R., Meisenkothen, F. & Sumption, M. metal-stoichiometry and dynamic light scattering (DLS). Addition of nanoparticle dispersions to enhance flux pinning of the The eutectic line of Fig. 1b was determined from data on 3BaO-7CuO binary YBa2Cu3O7-x superconductor. Nature 430, 867–870 (2004). pyrolyzed powders and thin films on MgO substrates. They were analyzed by 5. MacManus-Driscoll, J. L. et al. Strongly enhanced current densities in thermogramivetry and differential thermal analysis using the Setsys evolution superconducting coated conductors of YBa2Cu3O7–x + BaZrO3. Nat. Mater. 3, apparatus of Setaram, at subatmospheric conditions Ptotal = PO2 and at a heating 439–443 (2004). rate of 5 °C min−1. 6. Kang, S. et al. High-performance high-Tc superconducting wires. Science 311, 1911–1914 (2006). 7. Llordés, A. et al. Nanoscale strain-induced pair suppression as a vortex- XRD. The phase and texture analysis of the grown films and nanocomposites was pinning mechanism in high-temperature superconductors. Nat. Mater. 11, analyzed with a high resolution Discover D8 Bruker diffractometer (X-ray 329–336 (2012). energy = 8.049 keV) equipped with a Lynxeye XE energy-dispersive 1-D detector. 8. Solovyov, V., Dimitrov, I. K. & Li, Q. Growth of thick YBa2Cu3O7 layers via a The quantitative determination of epitaxial fraction of nanoparticles was investi- gated with a Bruker-AXS D8 Advance diffractometer operating with Cu Kα barium fluoride process. Supercond. Sci. Technol. 26, 013001–013019 (2013). radiation equipped with a General Area Detector Diffraction System. Non-uniform 9. Yun, K. S. et al. Vapor-liquid-solid tri-phase pulsed-laser epitaxy of r.m.s. strain (nanostrain) was determined from XRD integral-breadth measure- RBa2Cu3O7-y single-crystal films. Appl. Phys. Lett. 80, 61–63 (2002). ments through semiquantitative Williamson–Hall plots using the Discover D8 10. Ohnishi, T., Huh, J. U., Hammond, R. H. & Jo, W. High rate in situ YBa2Cu3O7 Bruker system. Debye Scherrer equation was used to determine the nanoparticles film growth assisted by liquid phase. J. Mater. Res. 19, 977–981 (2004). size from the (200) reflection. 11. Kursumovic, A. et al. Hybrid liquid phase epitaxy processes for YBa2Cu3O7 In situ XRD synchrotron analysis were carried out at DiffAbs beamline at SOLEIL film growth. Supercond. Sci. Technol. 17, 1215–1223 (2004). synchrotron with a beam energy of 18 keV, heating ramps of 5 °C s−1 and 1 bar of 12. Scheel, H. J. Liquid Phase Epitaxy of Electronic, Optical and Optoelectronic total pressure. An area detector (X-ray hybrid pixel area detector, XPAD) was used42 Materials, Ed. P. Capper and M. Mauk, John Wiley & Sons, (2007). with acquisition times per data point of 500 ms (for phase evolution before YBCO 13. Ichino, Y. et al. Possibility of high deposition rate in SmBa2Cu3Oy films growth) and 100 ms (for YBCO growth), respectively. To follow random phases, prepared using the Vapor-Liquid-Solid growth mode. Jpn. J. Appl Phys. 45, grazing XRD scans were recorded, whereas to follow YBCO growth the Bragg 758–760 (2006). conditions with the (005) peak were met (θ–2θ geometry). The experiments were run 14. Lee, J. et al. RCE-DR, A novel process for coated conductor fabrication with in a DHS 1100 Anton Paar heater covered with a graphite dome, with maximum high performance. Supercond. Sci. Technol. 27, 044018–6 (2014). heating rates of 5 °C s−1 and equipped with a double connection to the vacuum 15. MacManus-Driscoll, J. L. et al. Strong pinning in very fast grown reactive co- pumps and to the gas inlet system. Electrovalves were used to modulate the total evaporated GdBa2Cu3O7 coated conductors. Appl. Phys. Lett. Mater. 2, 086103 pressure, while an oxygen sensor was used to pre-set the PO2. For PO2-route and (2014). T-route, different combinations of N2 and Air gases were mixed with a mass flow 16. Feighan, J. P. F., Kursumovic, A. & MacManus-Driscoll, J. L. Materials design controller to meet the required conditions. Pole figures (characteristics of the for artificial pinning centres in superconductor PLD coated conductors. nanoparticles embedded in the YBCO matrix) were recorded during synchrotron Supercond. Sci. Technol. 30, 123001–123016 (2017). experiments at DiffAbs using the XPAD area detector. 17. Gamalski, A. D., Tersoff, J., Sharma, R., Ducati, C. & Hofmann, S. Formation of metastable liquid catalyst during subeutectic growth of Germanium Transmission electron microscopy. In addition, the nanoparticle size in the col- nanowires. Nano Lett. 10, 2972–2976 (2010). loidal solutions were determined by TEM studies with a JEOL of 120 kV 1210 TEM, 18. Kursumovic, A., Cheng, Y., Glowacki, B., Madsen, J. & Evetts, J. Study of the with a resolution point of 3.2 Å and DLS Zetasizer Nano Zs with measurement range rate-limiting processes in liquid-phase epitaxy of thick YBaCuO films. J. Cryst. of 0.3 nm–10.0 μm and sensitivity of 0.1 mg mL−1. The microstructure, atomic defect Growth 218, 45–56 (2000). structure of films and nanocomposites, and nanoparticles size in the nanocomposite 19. Chen, N., Rothman, S., Routbort, J. & Goretta, K. Tracer diffusion of Ba and Y films were studied with a STEM using a FEI Tecnai F20 S/TEM operated in TEM in YBa2Cu3Ox. J. Mat. Res. 7, 2308–2316 (1992). mode at 200 kV, and with a X-FEG gun, a CETCOR probe corrector and a Gatan 20. Chernov, A. A. Notes on interface growth kinetics 50 years after Burton, TRIDIEM 866 ERS energy filter operated in STEM mode at 300 kV. Specimens for Cabrera and Frank. J. Cryst. Growth 264, 499–518 (2004). STEM were prepared by conventional methods, by grinding, dimpling and Ar ion 21. Kursumovic, A. et al. High critical current densities in YBa2Cu3O7−x films milling or Focused Ion Beam (FIB). A FIB Zeiss 1560XB Cross Beam was also used to grown at high rates by hybrid liquid phase epitaxy. Appl. Phys. Lett. 87, perform and visualize the cross-section of Supplementary Figure 1a. 252507–3 (2005). NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13791-1 22. Klemenz, C. & Scheel, H. J. Solubility of YBCO and NdBCO the BaO/CuO MICINN and co-financing by the European Regional Development Fund; 2017-SGR flux. J. Cryst. Growth 200, 435–440 (1999). 1519 from Generalitat de Catalunya; and EU COST Action NANOCOHYBRI 23. Krauns, C., Sumida, M., Tagami, M., Yamada, Y. & Shiohara, Y. Solubility of (CA16218). ICMAB authors acknowledge the Center of Excellence award Severo RE elements into Ba-Cu-O melts and the enthalpy of dissolution. Z. für. Phys. Ochoa SEV-2015-0496. Authors thank Dr. Cornelia Pop, Dr. Bohores Villarejo, Dr. B Condens. Matter 96, 207–212 (1994). Bernat Mundet, and Dr Anna Palau for contributions at the initial stages of the 24. Yao, X. & Shiohara, Y. Process for high growth rate and high superconducting project. We thank Diana García, Lavinia Saltarelli, Adria Pacheco, Dr. Max Sieger, and properties of REBCO single crystals. Mater. Sci. Eng. B 53, 11–17 (1998). Dr. Albert Queralto for help in the feasibility study and preliminary results on thick 25. MacManus-Driscoll, J. L. et al. YBa2Cu3O7 coated conductor growth by films and coated conductors. Authors acknowledge support from SuNAM Co Ltd. for Hybrid Liquid Phase Epitaxy. IEEE Trans. Appl. Superconductivity 17, providing the buffered metallic substrates for the initial tests on coated conductors. 2537–2541 (2007). We also acknowledge the Scientific Services at ICMAB, ICTS of IMB-CNM-CSIC, 26. Vilardell, M. et al. Flexible manufacturing of functional ceramic coatings by ICN2 Electron Microscopy Division and LMA-INA from Aragon, and SOLEIL Syn- inkjet printing. Thin Solid Films 548, 489–497 (2013). chrotron for granting beamtime for XRD experiments and the DiffAbs beamline 27. Lindemer, T. B. & Specht, E. D. The BaOCuO system. Solid-liquid equilibria personnel for help. L.S. and J.J. also acknowledge their FPU-MINECO PhD grant, S.R. and thermodynamics of BaCuO2 and BaCu2O2. Phys. C. 255, 81–94 (1995). and N.C. thank University of Girona and Generalitat de Catalunya fort their PhD FI 28. MacManus-Driscoll, J. L., Bravman, J. C. & Beyers, R. B. Phase-equilibria in grant, respectively. the Y-Ba-Cu-O system and melt processing of Ag clad YBa2Cu3O7-x tapes at reduced oxygen partial pressure. Phys. C. 241, 401–413 (1995). 29. Pahlke, P. et al. Influence of artificial pinning centers on structural and Author contributions X.O. and T.P. conceived the idea and the project plan. L.S, J.J., J.B., and S.Ra. synthesized superconducting properties of thick YBCO films on ABAD-YSZ templates. and characterized the films under the supervision of J.F., P.R., S.R., X.O., and T.P. The Supercond. Sci. Technol. 31, 004007–004009 (2018). TEM experiments were realized by R.G. who also carried out the corresponding data 30. Chepikov, V. et al. Introduction of BaSnO3 and BaZrO3 artificial pinning treatment. N.C. synthesized the nanoparticles and stabilized them in the precursor centres into 2 G HTS wires based on PLD-GdBCO films. Phase I of the solution with the guide of R.Y. and S.Ri. The synchrotron experiments were designed and industrial R&D programme at SuperOx. Supercond. Sci. Technol. 30, performed by T.P., J.F., C.M., X.O., J.B., L.S., J.J., and S.Ra. Data treatment was carried 124001–124012 (2017). out by J.B., L.S., J.J., S.R. All contributed to the interpretation. P.R. determined the 31. Abraimov, D. et al. Double disordered YBCO coated conductors of industrial eutectic transition. All authors participated in the discussion and writing of the scale: high currents in high magnetic field. Supercond. Sci. Technol. 28, manuscript. 114007–114008 (2015). 32. Majkic, G. et al. Engineering current density over 5kAmm−2 at 4.2 K, 14 T in thick film REBCO tapes. Supercond. Sci. Technol. 31, 10LT01–16LT01 (2018). Competing interests 33. Shiohara, Y. & Endo, A. Crystal growth of bulk high-Tc superconducting The authors declare no competing interests. oxide materials. Mat. Sci. Eng. R19, 1–86 (1997). 34. Krauns, Ch, Tagami, M., Yamada, Y., Nakamura, M. & Shiohara, Y. Wetting between prospective crucible materials and the Ba-Cu-O melt. J. Mater. Res 9, Additional information 1513–1518 (1994). Supplementary information is available for this paper at https://doi.org/10.1038/s41467- 35. Guzman, R. et al. Probing localized strain in solution-derived YBa2Cu3O7−δ 019-13791-1. nanocomposite thin films. Phys. Rev. Mat. 1, 024801–024807 (2017). 36. Xu, A. et al. Strongly enhanced vortex pinning from 4 to 77 K in magnetic Correspondence and requests for materials should be addressed to T.P. fields up to 31 T in 15 mol.% Zr-added (Gd, Y)-Ba-Cu-O superconducting tapes. Appl. Phys. Lett. Mater. 2, 046111–046118 (2014). Peer review information Nature Communications thanks the anonymous reviewer(s) for 37. Miura, M. et al. Tuning nanoparticle size for enhanced functionality in their contribution to the peer review of this work. perovskite thin films deposited by metal organic deposition. NPG Asia Mater. 9, e447 (2017). Reprints and permission information is available at http://www.nature.com/reprints 38. Gutierrez, J. et al. Strong isotropic flux pinning in YBa2Cu3O7-x – BaZrO3 nanocomposite superconductor films derived from chemical solutions. Nat. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Mater. 6, 367–373 (2007). published maps and institutional affiliations. 39. Obradors, X. et al. Epitaxial YBa2Cu3O7-x nanocomposite films and coated conductors from BaMO3 (M = Zr, Hf) colloidal solutions. Supercond. Sci. Technol. 31, 044001–044016 (2018). Open Access This article is licensed under a Creative Commons 40. Cantoni, C. et al. Strain-driven oxygen deficiency in self-assembled, Attribution 4.0 International License, which permits use, sharing, nanostructured, composite oxide films. ACS Nano 6, 4783–4789 (2011). adaptation, distribution and reproduction in any medium or format, as long as you give 41. Koshelev, A., Sadovskyy, I., Phillips, C. & Glatz, A. Optimization of vortex appropriate credit to the original author(s) and the source, provide a link to the Creative pinning by nanoparticles using simulations of the time-dependent Ginzburg- Commons license, and indicate if changes were made. The images or other third party Landau model. Phys. Rev. B 93, 060508–5 (2016). material in this article are included in the article’s Creative Commons license, unless 42. Mocuta, C. et al. Fast pole figure acquisition using area detectors at the DiffAbs indicated otherwise in a credit line to the material. If material is not included in the beamline – Synchrotron SOLEIL. J. Appl. Cryst. 46, 1842–1853 (2013). article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Acknowledgements licenses/by/4.0/. Authors acknowledge EU for ULTRASUPERTAPE ERC-2014-ADG-669504; COACHSUPENERGY MAT2014-51778-C2-1-R and MAT2014-51778-C2-2-R from MINECO; SuMaTe RTI2018-095853-B-C21 and RTI2018-095853-B-C22 from © The Author(s) 2020 8 NATURE COMMUNICATIONS | (2020)11:344 | https://doi.org/10.1038/s41467-019-13791-1 | www.nature.com/naturecommunications