Nature Nanotechnology

Published by Springer Nature
Online ISSN: 1748-3395
Print ISSN: 1748-3387
Learn more about this page
Recent publications
Schematic of the nanophotonic circuit
a, Illustration of the nanocircuit enabling the directional routing of K and K′ valleys. The width and height of the waveguide are w = 240 nm and h = 100 nm. The thickness of the SiO2 film is g = 20 nm. b, An atomic force microscopy image of a fabricated circuit sample. c,d, Spin-resolved TPL spectra for a bare WS2 monolayer under σ+ (c) and σ− (d) excitations.
Principle of valley preservation
a, The real parts of effective wavevectors for GM1 and GM2 at different wavelengths. The cross-section Ez distributions of GM1 and GM2 are plotted on the right. Scale bar, 500 nm. b, Normalized electric field intensity distributions in the gap of the waveguide excited by x- and y-polarized dipoles at 630 nm. c,e, Normalized electric field intensity distributions for σ+ and σ− dipoles at wavelengths of 630 nm (c) and 810 nm (e). Scale bars, 100 nm. d,f, Calculated FVP profiles (FVP = (|Eσ+|² − |Eσ−|²) / (|Eσ+|² + |Eσ−|²)) at wavelengths of 630 nm (d) and 810 nm (f).
Demonstration of directional valley router
a,d, Simulated electric field intensity distributions of the valley router when excited by a σ+ (a) or σ− (d) dipole at 630 nm. Scale bars, 1 μm. b,e, Captured TPL images of the valley router when excited by a focused pump laser of σ+ (b) or σ− (e) polarization. Scale bars, 1 μm. The zoomed-in images of output ports A and B are shown on the right. c,f,g, TPL spectra measured from ports A and B for σ+ (c), σ− a.u. (f) or linearly polarized a.u. (g) excitations a.u. h, Experimental and theoretical VSs as a function of the rotation angle φ of the QWP. The error bars indicate the standard deviation of multiple measurements. i, Time sequence of measured VS for the binary code of ‘E’.
Unidirectional valley circulation among multiple channels
a–d, Schematic illustrations (a,b) and measured TPL images (c,d) of the circuitry enabling the unidirectional routing of valley information. The routing is counter-clockwise for K′ valley excitons (a,c) and clockwise for K valley excitons (b,d). Scale bars, 1 μm.
Valleytronics is a promising candidate to address low-energy signal transport on chip, leveraging the valley pseudospin of electrons as a new degree of freedom to encode, process and store information1–7. However, valley-carrier nanocircuitry is still elusive, because it essentially requires valley transport that overcomes three simultaneous challenges: high fidelity, high directionality and room-temperature operation. Here we experimentally demonstrate a nanophotonic circuit that can route valley indices of a WS2 monolayer unidirectionally via the chirality of photons. Two propagating modes are supported in the gap area of the circuit and interfere with each other to generate beating patterns, which exhibit complementary profiles for circular dipoles of different handedness. Based on the spin-dependent beating patterns, we showcase valley fidelity of chiral photons up to 98%, and the circulation directionality is measured to be 0.44 ± 0.04 at room temperature. The proposed nanocircuit can not only enable the construction of large-scale valleytronic networks but also serve as an interactive interface to integrate valleytronics3–5, spintronics8–10 and integrated photonics11–13, opening new possibilities for hybrid spin-valley-photon ecosystems at the nanoscale.
A comprehensive dynamic-metasurface optofluidic platform
a, Schematic of three Si metasurfaces with different dynamic control functions fabricated on the same chip. Integrated microfluidics provide on-demand control over the refractive index (n = 1.0–1.7) of the environment of nanoresonators in a metasurface by flowing liquids with different refractive indices in real time. Inset: scanning electron microscopy (SEM) images of the fabricated metasurfaces that provide the corresponding functions. Scale bars, 10 μm (bottom-left SEM image); 500 nm (all other SEM images). b,c, Reflection optical images (b) and measured reflection spectra (c) of the metasurfaces (height, h = 75 nm; diameter, D = 165 nm; period, p = 250 nm) for dynamic reflectivity tuning. A 530 nm short-pass colour filter is inserted to enhance the contrast in b. Scale bars, 25 μm. d,e, Reflection optical images (d) and measured reflection spectra (e) of the metasurfaces (h = 75 nm, D = 185 nm, p = 280 nm) for dynamic colour tuning. Scale bars, 25 μm. f, Optical images of the back focal plane of a meta-lens (NA = 0.45) (left) and far-field distribution of a meta-hologram (right) illustrating the dynamic diffraction efficiency tuning of optics on demand. Scale bars, 3 μm (left), 1 mm (right). g, Measured diffraction efficiency spectra of geometric phase metasurfaces (h = 120 nm; length, L = 120 nm; width, W = 90 nm; p = 320 nm). Insets: the Fourier plane of a meta-grating that can effectively steer the beam when turned on.
Mechanism of dynamic reflectivity and colour control in Si metasurfaces
a, Electric-field distributions (Ey) of two dominant optical resonances in a Si nanodisk array with symmetric and antisymmetric radiation profiles. b, Illustration of two dominant optical resonances in nanodisks showing different dispersions with the refractive index of the surrounding medium. Inset: the top view of a unit cell of a nanodisk array supporting two dominant optical resonances. c, Analysis for dynamic reflectivity control by Si metasurfaces. The top and middle panels show the amplitude (a, solid lines) and phase (φ, dashed lines) of the scattered plane waves from the symmetric (blue) and antisymmetric (red) modes as a function of the incident wavelength when the Si metasurface is embedded in different dielectric environments (n = 1.0 (1.7) for the top (middle) panel). The bottom panel shows the calculated reflection spectra of the metasurface for dynamic reflectivity control. d, Similar analysis to c but for dynamic colour control. The designed wavelength λd is set as 550 nm.
Mechanism of dynamic diffraction efficiency control and spectral control of phased-array optics by Si geometric phase metasurfaces
a, Illustration of four dominant optical resonances showing different dispersions with the refractive index of the surrounding medium. Depending on the specific resonant wavelength, certain optical resonances are damped by optical absorption in Si. Inset: the top view of a unit cell of a nanoblock array supporting the corresponding four optical resonances. b, The top and middle panels show the amplitude (solid lines) and phase (dashed lines) of the scattered plane waves from two anisotropic resonances for x (blue) and y (red) polarization as a function of the incident wavelength when the Si metasurface is embedded in different dielectric environments (n = 1.0 (1.7) for the top (middle) panel). The bottom panel shows the calculated diffraction efficiency spectra of the geometric phase metasurface. The designed wavelength λd is set as 550 nm. c, Schematic of a multiplexed meta-lens composed of two differently sized Si nanoresonator arrays that only focus red (green) light when embedded in air (high-index oil). Inset: the SEM images of the as-fabricated meta-lens. Scale bar, 1 μm. d, Simulated diffraction efficiency spectra for two nanoresonator arrays when the meta-lens is embedded in air (left) or high-index oil (right). e, Optical images of the back focal plane of the achromatic meta-lens (NA = 0.45) embedded in air (top) or high-index oil (bottom) under different illumination conditions. Scale bars, 5 µm. f, focal length of the meta-lens.
Integration of dynamic metasurfaces with programmable microfluidics on a transparent substrate
a, Schematic of a meta-lens (NA = 0.45) integrated with a programmable microfluidic channel. Top inset: the SEM image of the fabricated meta-lens to be integrated with the microfluidic channels. Scale bar, 500 nm. Bottom inset: the optical image of the fabricated Y-shaped microfluidic channel. Scale bar, 100 μm. b, A series of optical images at the back focal plane of the meta-lens taken with a time step of 25 ms to show the process of turning on (top row) and turning off (bottom row) the meta-lens. Scale bar, 50 μm. c, Modulated intensity in the focal point as a function of time when alternating flow of either high-index oil (n = 1.70) and low-index cleaner (HFE-7500, n = 1.29) over the meta-lens. d, Schematic of a transparent-metasurface digital number display controlled by programmable microfluidics. The inset shows a schematic of the real-time refractive index of the material flowing in the channel. e, Photograph of the integrated-metasurface digital number display. f, Reflection optical images of the display showing different digital numbers. Scale bar, 1 mm. The display is illuminated by unpolarized white light and a 530 nm short-pass filter is inserted to enhance the contrast.
The ability to manipulate light and liquids on integrated optofluidics chips has spurred a myriad of important developments in biology, medicine, chemistry and display technologies. Here we show how the convergence of optofluidics and metasurface optics can lead to conceptually new platforms for the dynamic control of light fields. We first demonstrate metasurface building blocks that display an extreme sensitivity in their scattering properties to their dielectric environment. These blocks are then used to create metasurface-based flat optics inside microfluidic channels where liquids with different refractive indices can be directed to manipulate their optical behaviour. We demonstrate the intensity and spectral tuning of metasurface colour pixels as well as on-demand optical elements. We finally demonstrate automated control in an integrated meta-optofluidic platform to open up new display functions. Combined with large-scale microfluidic integration, our dynamic-metasurface flat-optics platform could open up the possibility of dynamic display, imaging, holography and sensing applications.
Membrane characterization
a,b, Photographs of the unmodified CEM (a) and HMO-CEM (b). c,d, Cross-sectional TEM micrographs for unmodified CEM (c) and HMO-CEM (d) show Mn nanoparticles uniformly embedded within the HMO-CEM matrix with an average particle size of ~79.4 ± 23.1 nm. e,f, XPS spectra (e) show the Mn peak of the HMO-CEM (black line) at ~642.8 eV, demonstrating successful incorporation of Mn into the CEM, and the FTIR spectra (f) show a weak band at 3,400 cm⁻¹ (red box) resulting from the stretching of –OH and peak broadening at 600–700 cm⁻¹ corresponding to MnOx stretching and bending vibrations.
Source data
Membrane performance and selectivity
a, Phosphate concentration and pH in the receiving chamber for an unmodified CEM, high-loading HMO-CEM and low-loading HMO-CEM in the presence and absence of an applied potential. The feed solution was composed of 0.10 M NaH2PO4, whereas 0.05 M Na2SO4 was used as the receiving solution. A potential of 0.8 V versus Ag/AgCl (2.0 V cell potential) was applied across two Pt wires used as electrodes in the feed (cathode) and permeate (anode) chambers. b, Phosphate selectivity over competing anions (Cl⁻, SO4²⁻ and NO3⁻) for an unmodified CEM and HMO-CEM. c,d, Phosphate concentration in the receiving chamber for the unmodified CEM (c) and HMO-CEM (d) for an equimolar solution (1 mM) of NaCl, Na2SO4, NaNO3 and NaH2PO4 as the feed solution and 18 MΩ DI water as the permeate; a potential of 0.8 V versus Ag/AgCl (2.0 V cell potential) was applied across two Pt wires used as electrodes in the feed (cathode) and permeate (anode) chambers. The error bars represent standard deviations.
Source data
Molecular dynamics simulations of phosphate transport
a–f, Example of simulation cell used in the molecular dynamics simulations: all the atoms are shown as van der Waals spheres (a); water molecules are hidden (b); water molecules are hidden and the polymer is represented by the ball-and-stick model (c); only HMO particles and ions are shown (d); example of the H2PO4⁻ ion with its hydration water in the outer-sphere complex geometry (e); illustration of the diffusion pathway of adsorbed H2PO4⁻ (f). g, Molecular representation of the HMO-CEM membrane composed of intertwined charged-polymer chains and embedded HMO particles. h, Proposed mechanism of selective phosphate transport where phosphate ions hop between the adsorption sites on individual and neighbouring HMO particles. The weak outer-sphere complexes H2PO4⁻/HMO are formed within the Stern part of the electrical double layer, which are relatively mobile and can migrate around the HMO particle if subjected to external driving force like flow or weak electric field. The H2PO4⁻ ions jump from one HMO particle to another through the intergel solution phase, that is, through the fluid-saturated micro- and mesopore spaces. The phosphate ions adsorbed onto HMO diffuse within the particle electrical double layer in the direction of flow or the applied electric field.
Comparing experimental values and model predictions
a, Phosphate flux of the three membranes determined experimentally (red) and via the mathematical model (black). b, Comparing the transport number (t) of phosphates across the three membranes obtained from experiments (red) and the mathematical model (black).
Source data
Specific-ion selectivity is a highly desirable feature for the next generation of membranes. However, existing membranes rely on differences in charge, size and hydration energy, which limits their ability to target individual ion species. Here we demonstrate a nanocomposite ion-exchange membrane material that enables a reverse-selective transport mechanism that can selectively pass a single ion species. We demonstrate this transport mechanism with phosphate ions selectively transporting across negatively charged cation exchange membranes. Selective transport is enabled by the in situ growth of hydrous manganese oxide nanoparticles throughout a cation exchange membrane that provide a diffusion pathway via phosphate-specific, reversible outer-sphere interactions. On incorporating the hydrous manganese oxide nanoparticles, the membrane’s phosphate flux increased by a factor of 27 over an unmodified cation exchange membrane, and the selectivity of phosphorous over sulfate, nitrate and chloride reaches 47, 100 and 20, respectively. By pairing ion-specific outer-sphere interactions between the target ions and appropriate nanoparticles, these nanocomposite ion-exchange materials can, in principle, achieve selective transport for a range of ions.
An easily penetrable single-walled carbon nanotube (SWCNT) facilitates the use of nanobionics in living cyanobacterial cells through generations, advancing their fluorescent bioimaging and energy-based photovoltaic applications.
Inspired by the biological processes of molecular recognition and transportation across membranes, nanopore techniques have evolved in recent decades as ultrasensitive analytical tools for individual molecules. In particular, nanopore-based single-molecule DNA/RNA sequencing has advanced genomic and transcriptomic research due to the portability, lower costs and long reads of these methods. Nanopore applications, however, extend far beyond nucleic acid sequencing. In this Review, we present an overview of the broad applications of nanopores in molecular sensing and sequencing, chemical catalysis and biophysical characterization. We highlight the prospects of applying nanopores for single-protein analysis and sequencing, single-molecule covalent chemistry, clinical sensing applications for single-molecule liquid biopsy, and the use of synthetic biomimetic nanopores as experimental models for natural systems. We suggest that nanopore technologies will continue to be explored to address a number of scientific challenges as control over pore design improves. This Review discusses the latest advances in nanopore technologies beyond DNA sequencing.
Sweet spot operation of a silicon hole spin qubit limits the impact of charge noise and improves qubit coherence, circumventing the typical trade-off between operation speed and coherence.
Formation of freestanding multilayer (Pt/Co/Ni/Co) and evaluation of the transfer process
a, Deposition of an SAO/MgO/Pt/Co/Ni/Co epitaxial multilayer on an STO(001) substrate and capped with TaN. b, Immersion of the as-deposited film in deionized water to dissolve the SAO layer after spin coating PMMA. c, Separation of a freestanding multilayer (MgO/Pt/Co/Ni/Co/TaN). d, Transfer of the freestanding multilayer onto a sapphire substrate. e, Out-of-plane θ–2θ XRD patterns of SAO/MgO (i), SAO/MgO/Pt/Co/Ni/Co/TaN (ii) and a freestanding multilayer transferred onto a sapphire substrate (iii). The inset in (iii) shows a freestanding multilayer on a 10 × 10 mm² sapphire substrate. f,g, AFM images of the as-deposited sample (f) and a freestanding multilayer (g) transferred onto a sapphire substrate. The root mean square roughness values in f and g are 0.278 nm and 0.280 nm, respectively.
CIDWM in freestanding racetrack (Co/Ni/Co)
a, Protective PMMA is removed. b, Schematic of device fabrication from a freestanding Pt/Co/Ni/Co heterostructure. The cyan and blue colours correspond to down- and up-magnetized domains, respectively. c, Optical image of a typical racetrack. The device comprises a nanowire (length, 50 μm; width, 3 μm) and two bond pads. DW motion and current direction are along the x axis. d,e, Current-induced DW velocity in freestanding Co/Ni/Co transferred onto a sapphire substrate (blue and olive; square) and as-deposited samples (red and black; circle) without (d) and with (e) a magnetic field applied along the x axis. To examine the longitudinal-field dependence of the CIDWM, a fixed current density of 2.33 × 10⁸ A cm⁻² for the freestanding racetrack and 1.82 × 10⁸ A cm⁻² for the as-deposited sample are used. The insets in d show typical Kerr images of DW motion in response to a series of injected 10-ns-long current pulses (J = 2 × 10⁸ A cm⁻²) composed of three pulses in freestanding multilayers transferred onto a sapphire substrate. The bright and dark regions correspond to down (⊗ or ↓) and up (⊙ or ↑) domains, respectively. The error bars in d and e represent the standard deviation.
Transport properties of freestanding Pt/Co/Ni/Co
a, Schematic of the transport measurement geometry. b–d, Hall conductivity (σxy) in freestanding and as-deposited samples as a function of azimuthal and polar angles (φ, θ) defined in a. e,f, Longitudinal resistivity (ρxx) as a function of temperature (T) (e) and Hall conductivity (σxy) versus longitudinal conductivity (σxx) (f) of the freestanding and as-deposited samples. The blue and red lines in f are linear fits of σxy versus σxx for the freestanding and as-deposited samples, respectively. In these two cases, the fitted line has a slope that is equal to 0.0035 ± 0.0001 (freestanding sample) and 0.0029 ± 0.0001 (as-deposited sample). The slope is the physical coefficient related to the extrinsic contribution of the skew scattering mechanism. The low-temperature (10 K) data are excluded from the fits.
CIDWM across 3D protrusions in HM/FM heterostructures
a, Schematic of freestanding racetracks formed from HM/FM heterostructures transferred onto a pretreated sapphire substrate. The 3D protrusions with different heights and different angles from the DW motion direction are made on the substrate by etching. The angle is defined as the angle between the protrusion and x axis. The blue and red parts of the racetrack correspond to down- and up-magnetized domains, respectively. b, Cross-sectional TEM image of a freestanding racetrack on a 3D protrusion with a height of 700 nm and width of 3 μm (top left). Magnified high-resolution TEM images showing individual layers in the 3D racetrack of the two regions highlighted in the top-left image as an orange rectangle (bottom left) and blue rectangle (right). The horizontal green lines in the bottom-left panel indicate the interfaces between the individual layers. c, DW velocity versus current density in racetracks formed from the 2D freestanding films without protrusions (blue square) and with protrusions perpendicular to the racetrack channel with heights of 20 nm (black triangle) and 900 nm (green triangle). The insets show the typical Kerr images of DW motion in response to a series of injected 10-ns-long current pulses (J = −2.3 × 10⁸ A cm⁻²) composed of ten pulses for the 3D RTM formed on a protrusion with a height of 20 nm and angle of 90°. The bright and dark regions correspond to down (⊗ or ↓) and up (⊙ or ↑) domains, respectively. The tall blue box indicates the position of protrusion. d,e, Threshold current density required to drive a DW across 3D protrusions with various heights (0, 20, 30, 60, 120, 300, 700 and 900 nm) (d) and a height of 20 nm (e). Two types of DW (↑↓ or up–down and ↓↑ or down–up) cross the 3D protrusions at three different angles (45°, 90° and 135°) in d: the ↑↓ DWs cross the protrusions at an angle of 90° (black square), 45° (red triangle) and 135° (cyan diamond); the ↓↑ DWs cross the protrusions at an angle of 90° (green five-pointed star), 45° (blue inverted triangle) and 135° (magenta asterisk). The red and blue bars in e correspond to up–down (↑↓) and down–up (↓↑) DWs, respectively. The positive and negative values are defined with respect to the +x and –x direction, respectively. The error bars in d and e represent the standard deviation.
CIDWM across 3D protrusions in SAF structures
a, Schematic of the freestanding racetracks formed from SAF structures transferred onto a prepatterned sapphire substrate. b,c, Current-induced DW velocity in freestanding SAF films transferred onto a sapphire substrate (blue and olive; square) and as-deposited samples (red and black; circle) without (b) and with (c) a magnetic field applied along the x axis. To examine the longitudinal-field dependence of CIDWM, a fixed current density of 1.85 × 10⁸ A cm⁻² for the freestanding racetrack and 1.75 × 10⁸ A cm⁻² for the as-deposited sample are used. d, DW velocity versus current density in racetracks formed from the 2D freestanding films without any protrusions (blue square) and with protrusions perpendicular to the racetrack channel with heights of 300 nm (black triangle) and 900 nm (green triangle). The insets in b and d show the typical Kerr images of the DW motion in response to a series of injected 10-ns-long current pulses (J = 2.28 × 10⁸ A cm⁻²) composed of one pulse in the as-deposited thin films (b) and 10-ns-long current pulses (J = 1.42 × 10⁸ A cm⁻²) composed of two pulses in freestanding SAF films transferred onto a sapphire substrate with 300-nm-high protrusions and protrusion angle of 90° (d). The bright and dark regions correspond to down (⊗ or ↓) and up (⊙ or ↑) domains, respectively. The blue lines highlight the DW position. The error bars in b–d represent the standard deviation.
The fabrication of three-dimensional nanostructures is key to the development of next-generation nanoelectronic devices with a low device footprint. Magnetic racetrack memory encodes data in a series of magnetic domain walls that are moved by current pulses along magnetic nanowires. To date, most studies have focused on two-dimensional racetracks. Here we introduce a lift-off and transfer method to fabricate three-dimensional racetracks from freestanding magnetic heterostructures grown on a water-soluble sacrificial release layer. First, we create two-dimensional racetracks from freestanding films transferred onto sapphire substrates and show that they have nearly identical characteristics compared with the films before transfer. Second, we design three-dimensional racetracks by covering protrusions patterned on a sapphire wafer with freestanding magnetic heterostructures. We demonstrate current-induced domain-wall motion for synthetic antiferromagnetic three-dimensional racetracks with protrusions of up to 900 nm in height. Freestanding magnetic layers, as demonstrated here, may enable future spintronic devices with high packing density and low energy consumption.
Graphene nanopattern for single-crystal membrane growth and release
a, Schematic of epitaxy on nanopatterned graphene and layer release. b, Plan-view SEM images (left) and EBSD maps (right) of GaAs and Ge grown on GaAs and Ge substrates, showing planarized single-crystalline thin films. Scale bars, 2 μm. c, Three modes of peeling as a function of stressor thickness and epilayer thickness at the Ni stress level of 600 MPa and graphene coverage of 70% on the Ge substrate. The dashed line represents the natural spalling depth without graphene. d, Effect of graphene coverage on the peeling modes at the stressor stress of 600 MPa and epilayer thickness of 1 μm. e, Photograph of an exfoliated GaAs film (left) and remaining two-inch Ge wafer (right). f,g, Plan-view SEM images (f) and AFM image (g) of the substrate after peeling. h, Plan-view SEM image of GaAs substrate in the spalling regime. i, Plan-view SEM images of the sample surfaces in the delamination regime, showing delamination at the Ni/epilayer interface (left) and tape/Ni interface (right).
APB elimination by graphene nanopatterns
a,b, Cross-sectional STEM images of GaAs directly grown on Ge (a) and nanopatterned graphene-coated Ge (b). c,d, Plan-view SEM images of AlGaAs red LEDs grown on bare Ge (c) and nanopatterned graphene-coated Ge (d). e, Cross-sectional SEM image of LED fabricated by exfoliating the LED structure from the substrate and transferring onto the polyimide/silicon substrate. f, I–V curves of the fabricated LEDs on Ge with and without nanopatterned graphene. The error bars represent standard deviation after log transformation. g, Comparison of EL spectra of LEDs on Ge without (left) and with (right) nanopatterned graphene under various injection currents. h,i, Microscopic photographs of EL from red LEDs with different sizes and geometries on nanopatterned graphene-coated Ge (h) and bare Ge (i). Injection currents are 3, 3, 5 and 7 mA (from left to right). Scale bars, 10 μm.
Defect reduction in lattice-mismatched heteroepitaxy by graphene nanopatterns
a–c, MD simulations of heteroepitaxy without graphene (a), on thin and flexible graphene mask (b) and on thick and rigid mask (c). Dislocations are coloured according to their Burgers vector: blue for 1/2 <110>, green for 1/6 <112>, purple for 1/6 <110> and cyan for 1/3 <111>. Atoms are coloured as blue and transparent for the perfect diamond cubic structure and orange and opaque for stacking faults. The carbon atoms are coloured grey and the epilayer/substrate interfaces are indicated as dashed lines. d, Low-magnification STEM image of InAs directly grown on InP without a mask. e, High-resolution STEM image (left) and corresponding GPA maps showing in-plane (centre) and out-of-plane (right) strain. f, Low-magnification STEM image of InAs grown on graphene pattern. g, High-resolution STEM image and corresponding GPA maps at the edge of graphene, showing relaxed InAs film with slightly deformed graphene. h,i, Same set of data for InAs grown on SiO2 pattern, showing severe strain at the interface and at the mask edge. Enlarged atomic-resolution STEM image of i is shown in Supplementary Fig. 23.
Effect of graphene coverage
a, ECCI images of InAs grown on nanopatterned graphene-coated GaAs with different graphene coverages. Scale bars, 200 nm. b, Surface dislocation density as a function of graphene coverage measured by ECCI. c, MD simulations of heteroepitaxy of Ge on Si(100) with different mask widths covering 0%, 20% and 40% of the surface, showing a decrease in defects by increased graphene coverage. The colour coding is the same as that in Fig. 3.
Heterogeneous integration of single-crystal materials offers great opportunities for advanced device platforms and functional systems¹. Although substantial efforts have been made to co-integrate active device layers by heteroepitaxy, the mismatch in lattice polarity and lattice constants has been limiting the quality of the grown materials². Layer transfer methods as an alternative approach, on the other hand, suffer from the limited availability of transferrable materials and transfer-process-related obstacles³. Here, we introduce graphene nanopatterns as an advanced heterointegration platform that allows the creation of a broad spectrum of freestanding single-crystalline membranes with substantially reduced defects, ranging from non-polar materials to polar materials and from low-bandgap to high-bandgap semiconductors. Additionally, we unveil unique mechanisms to substantially reduce crystallographic defects such as misfit dislocations, threading dislocations and antiphase boundaries in lattice- and polarity-mismatched heteroepitaxial systems, owing to the flexibility and chemical inertness of graphene nanopatterns. More importantly, we develop a comprehensive mechanics theory to precisely guide cracks through the graphene layer, and demonstrate the successful exfoliation of any epitaxial overlayers grown on the graphene nanopatterns. Thus, this approach has the potential to revolutionize the heterogeneous integration of dissimilar materials by widening the choice of materials and offering flexibility in designing heterointegrated systems.
Design criteria for ultrahigh-temperature-stable and optically tunable BZHO/MgO photonic crystals
(1) Alternating epifilms of high-melting-temperature BZHO and MgO have low lattice mismatch, which permits high-quality epitaxial growth. Epitaxy prevents grain-growth degradation by mitigating the formation of thermodynamically unfavourable phases in the initial structure. (2) BZHO and MgO are refractory oxides with different crystal structures (perovskite and rocksalt). (3) Experimentally characterized refractive indices (n) of MgO and BZHO demonstrate a weak temperature dependence and negligible absorption (Supplementary Section 1).
Demonstration of thermal stability of BZHO/MgO superlattices
a, Interface between the MgO substrate and first BZHO layer is sharp and maintains coherent atomic columns. b, Interface sharpness and column coherence are preserved with no sign of intermixing post-annealing. c,d, Images of as-fabricated (c) and annealed (d) BZHO/MgO superlattice show constant bilayers of ~8 nm and no discernible degradation after heat treatment. The white arrow indicates growth direction. e–h, Energy-dispersive X-ray spectroscopy linescans illustrate the concentrations of Mg, Ba, Hf and Zr before (e) and after (f) annealing, and that the interface roughness remains between 1 nm and 2 nm before (g) and after (h) annealing. The data are presented as mean values ± standard error (s.e.) (n = 5).
Spectral control and stability of BZHO/MgO PhCs
a, Transmission spectra of BZHO/MgO PhCs with varying bilayer thicknesses (d) demonstrate control over the photonic bandgap. The change in optical transmittance is negligible after annealing at 1,100 °C for 12 h. b, The BZHO/MgO material system (red star) withstands higher temperatures in air than current approaches in nanophotonics. The reference points (letters a–x; Extended Data Table 1) show annealing/operation durations and environments producing a (self-reported) negligible change in optical properties post-annealing.
Spectral control of emission for thermal management and TPV applications
a, Schematic of the experiment with a low-exposure image of a SiC emitter. b,c, BZHO/MgO PhCs with cutoff wavelengths of 1.25 μm (b) and 1.65 μm (c) suppress thermal emission at 1,770 °C (b) and 1,350 °C (c), respectively. d, BZHO/MgO PhCs paired with MgO:NiO emitters surpass the state-of-the-art value in TPV emission control by >10% (relative), regardless of the operating environment. The reference points (letters a–v; Extended Data Table 1) show the data for which the emitters have been heat treated (for >1 h) and measured at high temperatures¹.
Materials screening for high-temperature nanophotonics
The screening process consists of five main steps. (1) An oxide substrate with a high melting point (>2,000 °C) is selected. (2) The Materials Project database is queried to gather properties of candidate oxide pairs. (3) Lattice-compatible pairs are identified by considering stable configurations between the different oxides and a lattice strain cutoff of |ε| ≈ 5% is applied. (4) Candidates are filtered by applying a bandgap cutoff of 2 eV and evaluated based on the criteria of minimal lattice strain and maximum refractive-index contrast. (5) Down-selected candidates are assessed using the available experimental data (for example, melting points). MgO/BaZrO3 (top) and LaAlO3/CeO2 (bottom) are two example systems that have minimal lattice strain, different crystal types, high melting points and contrasting refractive indices.
Nanophotonic materials offer spectral and directional control over thermal emission, but in high-temperature oxidizing environments, their stability remains low. This limits their applications in technologies such as solid-state energy conversion and thermal barrier coatings. Here we show an epitaxial heterostructure of perovskite BaZr0.5Hf0.5O3 (BZHO) and rocksalt MgO that is stable up to 1,100 °C in air. The heterostructure exhibits coherent atomic registry and clearly separated refractive-index-mismatched layers after prolonged exposure to this extreme environment. The immiscibility of the two materials is corroborated by the high formation energy of substitutional defects from density functional theory calculations. The epitaxy of immiscible refractory oxides is, therefore, an effective method to avoid prevalent thermal instabilities in nanophotonic materials, such as grain-growth degradation, interlayer mixing and oxidation. As a functional example, a BZHO/MgO photonic crystal is implemented as a filter to suppress long-wavelength thermal emission from the leading bulk selective emitter and effectively raise its cutoff energy by 20%, which can produce a corresponding gain in the efficiency of mobile thermophotovoltaic systems. Beyond BZHO/MgO, computational screening shows that hundreds of potential cubic oxide pairs fit the design principles of immiscible refractory photonics. Extending the concept to other material systems could enable further breakthroughs in a wide range of photonic and energy conversion applications.
Device, measurement scheme and properties of the first confined hole
a, Simplified three-dimensional representation of a silicon (yellow)-on-insulator (green) nanowire device with four gates (light blue) labelled G1, G2, G3 and G4. Gate G2 defines a quantum dot (QD2) hosting a single hole; G3 and G4 define a hole island used as reservoir and sensor for hole spin readout; G1 defines a hole island screening QD2 from dopant disorder and fluctuations in the source. Using bias tees, both static voltages (VG1, VG2) and time-dependent, high-frequency voltages (MW1, MW2) can be applied to G1 and G2, respectively. The drain contact is connected to an off-chip, surface-mount inductor to enable radiofrequency reflectometry readout. The coordinate system used for the magnetic field is shown on the left side (in the crystal frame, x = [001], y=[11¯0]\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$y=[1\bar{1}0]$$\end{document} and z = [110]). Each axis is given a different colour, which is used throughout the manuscript to indicate the magnetic field orientation. b, Colourized scanning electron micrograph showing a tilted view of a device similar to the measured one. Image taken just after the etching of the spacer layers. Scale bar, 100 nm. c, Rendering of the calculated wave function of the first hole accumulated under G2. d, Measured (dots) and calculated (solid line) hole g-factor as a function of the in-plane magnetic field angle θzy (dots). θzy = 90∘ corresponds to a magnetic field applied along the y axis. e, Same as d but in the xz plane. θzx = 90∘ corresponds to a magnetic field applied along the x axis. BOX, buried oxide.
a, Spin-electric susceptibility with respect to VG2 (LSESG2) as a function of magnetic field angle θzx (symbols), at constant fL = 19 GHz. The LSES vanishes at θzx = 41∘ and 106∘, as indicated by the two arrows. The solid line corresponds to the numerically calculated LSESG2. b, Top: pulse sequence used to measure LSESG1, a voltage pulse of amplitude δVG1 and duration τz is applied to G1 during the first free evolution time of a Hahn-echo sequence. Bottom: spin-up fraction P↑ as a function of τz for δVG1 = 2.16 mV (diamonds), 3.12 mV (stars) and 4.80 mV (squares), at θzx = 90∘. The oscillation frequency varies with δVG1. c, δVG1 dependence of the frequency shift extracted from the Hahn-echo measurements at θzx = 0∘, 42∘ and 90∘. Symbols in the latter data set correspond to the P↑ oscillations shown in b. The solid lines are linear fits to the experimental data whose slope directly yields ∣LSESG1∣. d, Measured (symbols) and calculated (solid line) LSESG1 as a function of θzx, at constant fL = 17 GHz. The negative sign of LSESG1 is inferred from the shift of fL under a change in VG1.
Anisotropy of the hole spin coherence and sweet-spot operation
a, Normalized Hahn-echo amplitude versus free evolution time τwait at fL = 17 GHz. The top-right inset sketches the pulse sequence. The bottom-left inset displays P↑ (τwait = 31.4 μs) versus the phase ϕ of the last π/2 pulse for 100 repetitions. For each τwait, we extract the average amplitude of the P↑(ϕ) oscillations and normalize it to the average amplitude in the zero-delay limit. The resulting normalized echo amplitudes are reported on the main plot. The dashed curve is a fit to exp(−(τwait/T2E)β)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\exp (-{({\tau }_{{{{\rm{wait}}}}}/{T}_{2}^{\mathrm{E}})}^{\beta })$$\end{document} with β = 1.5 ± 0.1. b, Measured T2E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{2}^{{{{\rm{E}}}}}$$\end{document} versus magnetic field angle θzx (symbols). The solid line is a fit to equation (1), using the experimental LSESG1 and LSESG2 from Fig. 2a–d. c, Normalized CPMG amplitude as a function of free evolution time τwait for different numbers Nπ of π pulses (curves are offset for clarity). The solid lines are fits to the same exponential decay function as in a with β = 1.5. d, Extracted T2CPMG\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{2}^{{{{\rm{CPMG}}}}}$$\end{document} as a function of Nπ. The dashed line is a linear fit with slope γ = 0.34. The inset sketches the CPMG pulse sequence: Nπ equally spaced πy pulses between two πx/2 pulses. For the Hahn-echo, we detune the phase of the last pulse.
Free induction decay
a, Collection of 600 Ramsey oscillations as a function of τwait, the free evolution time between two πx/2 pulses, at θzx = 118∘. The applied microwave frequency is detuned by ~700 kHz from the Larmor frequency. Each Ramsey oscillation is measured in ∼5.5 s. The locations of the representative traces shown in b are indicated by a diamond and a dot. b, Selected averages of Ramsey oscillations taken over different measurement times: tmeas = 5.5 s corresponding to a single trace (diamonds); tmeas = 27.5 s, corresponding to five consecutive traces (circles); tmeas ≈ 1 h, corresponding to the full set of 600 traces (squares). The solid lines are fits to Gaussian decaying oscillations. Note that the decay time T2*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{2}^{* }$$\end{document} depends on the chosen subset of consecutive traces (except for tmeas ≈ 1 h), which is a signature of non-ergodicity³⁶ at small tmeas (Supplementary Information, section 4). Hence we observe a distribution of T2*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{2}^{* }$$\end{document} values with mean T¯2*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\overline{T}}_{2}^{* }$$\end{document}. c, Mean T¯2*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\overline{T}}_{2}^{* }$$\end{document} for different tmeas (same symbols as in b) as a function of the magnetic field angle θzx. The solid lines are guides to the eye. The dashed black line is the calculated dephasing time due to hyperfine interactions (Supplementary Information, section 5).
Semiconductor spin qubits based on spin–orbit states are responsive to electric field excitations, allowing for practical, fast and potentially scalable qubit control. Spin electric susceptibility, however, renders these qubits generally vulnerable to electrical noise, which limits their coherence time. Here we report on a spin–orbit qubit consisting of a single hole electrostatically confined in a natural silicon metal-oxide-semiconductor device. By varying the magnetic field orientation, we reveal the existence of operation sweet spots where the impact of charge noise is minimized while preserving an efficient electric-dipole spin control. We correspondingly observe an extension of the Hahn-echo coherence time up to 88 μs, exceeding by an order of magnitude existing values reported for hole spin qubits, and approaching the state-of-the-art for electron spin qubits with synthetic spin–orbit coupling in isotopically purified silicon. Our finding enhances the prospects of silicon-based hole spin qubits for scalable quantum information processing.
Schematic of the SIMS measurements with atomic-layer resolution
a, Before measurement, a beam of primary Cs⁺ ions is deflected as it approaches the surface of the crystal by applying an opposing electric field. This allows the Cs⁺ ions to travel almost parallel to the crystal surface and gently remove only impurity species that may be present on the surface and produce flat surfaces. b, To achieve atomic-depth resolution, Cs⁺ ions with low energy are aimed at the sample surface with a high incident angle so that they interact only with the first exposed layer of atoms. As the primary Cs⁺ ions bombard the surface, atoms from the material are sputtered away as secondary ions and detected. c, Layer-by-layer sputtering generates a depth profile of the layered material, starting from the top. The atomic layers are registered as symmetric peaks, which overlap slightly, but have sufficient resolution to unambiguously distinguish each atomic layer. The structure and representative spectra in b and c correspond to Ti3AlC2 MAX.
Depth profiles of Ti3AlC2 MAX and multilayer Ti3C2Tx MXene
a,c, MAX produced by modified and conventional approaches using excess (a) and stoichiometric (c) amounts of Al precursor. b,d, MXene produced via the aqueous etching of the corresponding MAX in a (b) and c (d) with a mixture of hydrofluoric and hydrochloric acids. Both MAX samples show impurity levels of oxygen in the Al layer and both MXene samples show substantial levels of oxygen as surface terminations in the interlayer space between the Ti3C2Tx sheets. However, only the conventional MAX and MXene samples show elevated levels of oxygen in the carbon layer, qualifying them as oxycarbides.
Analysis of surface termination distribution in multilayer Ti3C2Tx MXene by depth profiling
a, Depth profiles of –O, –OH, –F, and –Cl surface terminations (T), introduced ‘randomly’ during aqueous synthesis, show two peaks corresponding to two adjacent surface termination layers from the lower side of one MXene sheet and the upper side of another MXene sheet. The deconvolution of these two peaks for each species allows for the calculation of surface composition. b, The two surface termination layers may have different compositions, with the upper layer containing more –O and –OH and the lower layer containing more –F as shown in the schematic. c, Depth profiles across 150 surface termination layers for Ti3C2Tx MXene produced from the stoichiometric (left) and conventional (oxycarbide) (right) Ti3AlC2 MAX-phase samples. Depth profiling across many surface termination layers allows us to obtain a more statistically meaningful average of the surface composition compared with other characterization techniques that are typically used for surface termination measurements, such as XPS. The surfaces of Ti3C2Tx MXene produced from conventional MAX were rich in –F and poor in –O and –OH, indicating that the presence of oxygen in the carbon layer affects the surface terminations.
Depth profiles of Mo2TiAlC2 (top) and Cr2TiAlC2 (bottom) MAX samples
A substantial amount of oxygen is present in the carbon layers of both Mo2TiAlC2 and Cr2TiAlC2, where the depth profiles for oxygen are separately shown in Supplementary Fig. 6 for clarity. For the Mo2TiAlC2 sample, there is perfect out-of-plane ordering, where Mo is exclusively present in the outer layers and Ti in the inner layer. For the Cr2TiAlC2 sample, there is some intermixing where Cr and Ti form a solid solution in the inner layer, but only Cr is present in the outer layers.
The MXene family of two-dimensional transition metal carbides and nitrides already includes ~50 members with distinct numbers of atomic layers, stoichiometric compositions and solid solutions, in-plane or out-of-plane ordering of atoms, and a variety of surface terminations. MXenes have shown properties that make them attractive for applications ranging from energy storage to electronics and medicine. Although this compositional variability allows fine-tuning of the MXene properties, it also creates challenges during the analysis of MXenes because of the presence of multiple light elements (for example, H, C, N, O, and F) in close proximity. Here, we show depth profiling of single particles of MXenes and their parent MAX phases with atomic resolution using ultralow-energy secondary-ion mass spectrometry. We directly detect oxygen in the carbon sublattice, thereby demonstrating the existence of oxycarbide MXenes. We also determine the composition of adjacent surface termination layers and show their interaction with each other. Analysis of the metal sublattice shows that Mo2TiAlC2 MAX exhibits perfect out-of-plane ordering, whereas Cr2TiAlC2 MAX exhibits some intermixing between Cr and Ti in the inner transition metal layer. Our results showcase the capabilities of the developed secondary-ion mass spectrometry technique to probe the composition of layered and two-dimensional materials with monoatomic-layer precision.
Bipolar thermoelectric Josephson engine
a, Scheme of the BTJE: two different superconductors S1 and S2 (with zero-temperature energy gaps Δ0,1 > Δ0,2) are tunnel coupled through a thin insulating layer (grey, I). The S1IS2 junction is predicted to generate power when S1 is kept at a higher temperature (Thot) than S2 (that is, T1 = Thot > T2 = Tcold). Remarkably, this structure can produce both positive and negative thermovoltages for the same thermal gradient imposed across the junction. b, Pseudo-colour scanning electron micrograph of a typical BTJE. An aluminium island (S1, red) is tunnel coupled through three AlOx barriers to a Cu/Al bilayer (S2, blue) thus realizing a double-loop SQUID. Additional tunnel-coupled Al electrodes (green) serve as Joule heaters for S1. Replicas resulting from the shadow-mask fabrication procedures are also visible (Methods for details). The charge transport properties of the double-loop SQUID were investigated in a two-wire configuration by recording the tunnelling current (I) while a voltage (V) was applied. ΦA and ΦB represent the magnetic fluxes linked to each loop, so that the total flux through the interferometer is Φ = ΦA + ΦB. The electron temperature in S1 is raised by powering a couple of heaters with a voltage Vh. c, The top panel shows current (I) versus voltage (V) characteristics measured at 30 mK for Φ = 0 (violet) and Φ = 0.33Φ0 (green). The bottom panel shows a zoomed-in view of the IV characteristics around V = 0. d, Experimental (Exp.) Josephson critical current (IC) of the double-loop interferometer versus Φ measured for different values of bath temperature. e, Theoretical (Th.) modulations of IC versus Φ calculated with the model presented in the Supplementary Information.
Bipolar thermoelectric effect
a, Subgap current (I) versus voltage (V) characteristics of the SQUID measured at Φ = 0.33Φ0 for different values of the input power (Pin) injected in S1. In the presence of a thermal bias, the interferometer shows an absolute negative conductance for both polarities of V. The black dashed line is the result from the theory (Th.) for a single junction²⁹ calculated at 10 pW. The temperature bias across the BTJE is deduced by fitting the experimental IV curves with the theory (Supplementary Information for details). b, Thermovoltage (Vth) versus Pin recorded at Φ = 0.33Φ0 (dots). Dashed lines correspond to the theory²⁹. Due to intrinsic PH symmetry, the system provides both positive (Vth+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{th}}}^{+}$$\end{document}, green) and negative (Vth−\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{th}}}^{-}$$\end{document}, purple) thermovoltages for any given Pin. Inset: schematic thermelectric IV characteristics where positive (Vth+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\mathrm{th}}^{+}$$\end{document}) and negative (Vth−\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_{\mathrm{th}}^-$$\end{document}) thermovoltages are shown. c, Seebeck coefficient (S\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{S}}}}$$\end{document}) versus Pin extracted from the data shown in b according to the calibration presented in the Supplementary Information. In the presence of thermoelectricity, the value of S\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{S}}}}$$\end{document} monotonically decreases by increasing Pin. Inset: scheme of the thermoelectric element generating Vth in the presence of a temperature gradient (δT). d, Subgap IV characteristics of the SQUID measured at Pin = 10 pW for selected values of Φ. The top (bottom) inset shows a magnification of the current close to Vth+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{th}}}^{+}$$\end{document} (Vth−\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{th}}}^{-}$$\end{document}). e,Vth versus Φ measured for different values of Pin. Dash-dotted lines are guides to the eye. All measurements were performed at a bath temperature of 30 mK.
Low temperature behaviour of the BTJE
a, Scheme of the electronic circuit used to demonstrate power production. The Joule heaters are powered through a floating voltage source (Vh). A d.c. current source (Ib) biases the parallel connection of the interferometer and a load resistor (RL) while recording the voltage drop VL occurring across RL. b, Sketch of thermoelectric current (I) versus voltage (V) characteristics where a resistive load (blue line) is superimposed. A typical hysteresis loop of the load voltage (VL) is also represented (green dashed curves) together with the thermovoltages (VL+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{L}}}^{+}$$\end{document} or VL−\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{L}}}^{-}$$\end{document}) corresponding to zero biasing current (that is, Ib = 0). Arrows indicate the direction of the Ib sweep. Grey areas represent thermoactive regions. c, VL versus Ib recorded with RL = 2 MΩ at Pin = 10 pW and Φ = 0.33Φ0. The engine can be ignited by ramping Ib from zero towards either positive (red) or negative (orange) values. Thermovoltages measured at Ib = 0 (green dots, VL+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{L}}}^{+}$$\end{document} or VL−\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{L}}}^{-}$$\end{document}) prove power production in the system, that is, the realization of the engine. The thermovoltage polarity can be changed by ramping Ib to large values (green curves). Arrows indicate the direction of the Ib sweep. d,VL versus Ib characteristics measured with RL = 2 MΩ and Φ = 0.33Φ0 for selected values of input power. e, Results from the model for the same parameters as in d (Supplementary Information for details). f, Engine output power (PL) versus input power (Pin) at Φ = 0.33Φ0 for different values of RL. Inset, characteristics of different loads (coloured lines) superimposed on a typical BTJE thermoactive IV curve (grey). g, PL versus Φ at Pin = 10 pW for different values of RL. Dash-dotted lines in f and g are guides to the eye. All measurements were performed at a bath temperature of 30 mK.
Temperature dependence of the BTJE
a, Output power PL at Φ = 0.33Φ0 and RL = 2 MΩ versus input power Pin for different values of bath temperature T. b, PL at Pin = 10 pW and RL = 2 MΩ versus Φ for selected values of T. Dash-dotted lines in a and b are guides to the eye. c, Schematic representation of a series (left) and parallel (right) connection of n S1IS2 thermoelectric elements. The series connection yields a total voltage Vtot = nV (with V the voltage drop occurring across each element) and a total current Itot = I (with I the current flowing through each element), whereas the parallel connection provides Vtot = V and Itot = nI. Therefore, both configurations generate an output power Ptot = ItotVtot = nIV = nP (P denotes the power produced by a single element).
Thermoelectric effects in metals are typically small due to the nearly perfect particle–hole symmetry around their Fermi surface. Furthermore, thermo-phase effects and linear thermoelectricity in superconducting systems have been identified only when particle–hole symmetry is explicitly broken, since thermoelectric effects were considered impossible in pristine superconductors. Here, we experimentally demonstrate that superconducting tunnel junctions develop a very large bipolar thermoelectricity in the presence of a sizable thermal gradient thanks to spontaneous particle–hole symmetry breaking. Our junctions show Seebeck coefficients of up to ±300 μV K–1, which is comparable with quantum dots and roughly 10⁵ times larger than the value expected for normal metals at subkelvin temperatures. Moreover, by integrating our junctions into a Josephson interferometer, we realize a bipolar thermoelectric Josephson engine generating phase-tunable electric powers of up to ~140 nW mm–2. Notably, our device implements also the prototype for a persistent thermoelectric memory cell, written or erased by current injection. We expect that our findings will lead to applications in superconducting quantum technologies.
Virus-like nanoparticles encapsulating functional riboswitches lock-in toxic metabolites.
The distinctive properties of single-walled carbon nanotubes (SWCNTs) have inspired the development of many novel applications in the field of cell nanobiotechnology. However, studies thus far have not explored the effect of SWCNT functionalization on transport across the cell walls of prokaryotes. We explore the uptake of SWCNTs in Gram-negative cyanobacteria and demonstrate a passive length-dependent and selective internalization of SWCNTs decorated with positively charged biomolecules. We show that lysozyme-coated SWCNTs spontaneously penetrate the cell walls of a unicellular strain and a multicellular strain. A custom-built spinning-disc confocal microscope was used to image the distinct near-infrared SWCNT fluorescence within the autofluorescent cells, revealing a highly inhomogeneous distribution of SWCNTs. Real-time near-infrared monitoring of cell growth and division reveal that the SWCNTs are inherited by daughter cells. Moreover, these nanobionic living cells retained photosynthetic activity and showed an improved photo-exoelectrogenicity when incorporated into bioelectrochemical devices.
Excitons in the vdW antiferromagnet NiPS3
a, Magnetic moments of nickel atoms in the hexagonal honeycomb lattice slightly magnetize surrounding sulfur ligands except those located between two nickel atoms with opposite spin. Excitons with highly anisotropic dipole moment (red ellipses) form between spin-polarized nickel d orbitals and unmagnetized sulfur p orbitals. The black dashed rectangle marks the in-plane magnetic unit cell with dimensions 5.8 Å × 10.1 Å (ref. ²⁴). Å, Angstrom. b, An NiPS3 flake with 160 nm thickness and lateral width around 50 μm on top of a SiO2/Si substrate shows narrow PL emission (full-width at half maximum ≲600 μeV) with strong degree of linear polarization (DoP) reaching 83%. Spectra represent minimum (grey) and maximum (blue) emission. PL intensity versus polarizer angle θ, plotted in the inset, is well described by I(θ) = I0sin²(θ), where θ denotes the analyser angle with respect to the minimum, I0, of the PL intensity. Blue and grey spectra respectively represent 90° and 0° analyser angles. c, Optical absorption derived from the linear reflectance contrast (dRc) spectrum reveals three resonances, X1, X2 and X3, below the charge transfer gap.
Strong light–matter coupling in NiPS3 microcavities
a, NiPS3 crystals enclosed by a dielectric bottom mirror and a 35-nm-thick top silver layer form a microcavity. b, Optical microscopy image of a single NiPS3 crystal inside the cavity. Scale bar, 5 μm. c, Angle-resolved optical reflection contrast map at 4 K of the same crystal plotted together with a coupled oscillator model of a single cavity mode and absorption resonances X1, X2 and X3 for negative angles (see also the reflectance map and profile analysis in Supplementary Fig. 7). Anti-crossings at each intersection between the cavity mode (white dashed line) and an excitonic resonance (black dash-dotted lines) are reproduced with good agreement between the model and the experimental data. The resulting multiple branches of the polariton dispersion are depicted as blue solid lines. The Rabi splittings, indicated by grey circles, are ℏΩ1 = 4 meV, ℏΩ2 = 2 meV and ℏΩ3 = 10 meV, as obtained from fits. d, Anti-crossing features are almost absent in the reflectance contrast map recorded at 120 K. e, Integrated PL intensity of X1 as a function of analyser angle. f, Angle-integrated PL emission shows strong linear polarization of more than 50% and pronounced peaks at the X1 and X2 absorption resonances. Blue and grey spectra represent 90° and 0° analyser angles, respectively.
Pronounced bottleneck of polariton relaxation
a, Angle-resolved PL emission recorded at 4 K superimposed with the coupled oscillator model shown in Fig. 2c. b, Schematic illustration of polariton relaxation and the bottleneck effect: excitons from the reservoir scatter with phonons of energy ħωph to weakly populate the lowest polariton branch, while efficient scattering due to long-range exciton–exciton interactions is suppressed. c, Cavity PL emission at 60 K. Intensity scale is indicated in a by the number given in parentheses. d, Comparison of angle-integrated normalized PL spectra at 4 and 60 K (intensity spikes have been removed).
Polariton nonlinearities under increasing density
a, Polariton dispersion for the lowest (P0) and highest (∼360P0) excitation power of the incident broadband laser pulses. Inset: magnified view of the fitted oscillator model for both powers. b, Density-dependent exciton energies directly determined from absorption minima in line-cuts at 0° (X1 and X2) and 20° (X3) with high exciton fraction. c–e, Reduction of Rabi splitting Ω(np) relative to the Rabi splitting Ω0 obtained at the lowest excitation power for c, Ω3; d, Ω2; and e, Ω1. Red solid lines represent a fit based on equation (1) to determine the saturation density.
Strong coupling between light and elementary excitations is emerging as a powerful tool to engineer the properties of solid-state systems. Spin-correlated excitations that couple strongly to optical cavities promise control over collective quantum phenomena such as magnetic phase transitions, but their suitable electronic resonances are yet to be found. Here, we report strong light–matter coupling in NiPS3, a van der Waals antiferromagnet with highly correlated electronic degrees of freedom. A previously unobserved class of polaritonic quasiparticles emerges from the strong coupling between its spin-correlated excitons and the photons inside a microcavity. Detailed spectroscopic analysis in conjunction with a microscopic theory provides unique insights into the origin and interactions of these exotic magnetically coupled excitations. Our work introduces van der Waals magnets to the field of strong light–matter physics and provides a path towards the design and control of correlated electron systems via cavity quantum electrodynamics.
MspA engineered nanopore allows for high-resolution detection of modifications on RNA.
A cation-selective membrane containing hydrous manganese oxide nanoparticles exhibits specific phosphate ion selectivity.
By using microfluidics to control the refractive index of the fluid surrounding dielectric metasurfaces, the optical response and functionality of metasurface devices can be tuned on-demand to enable comprehensive control of the flow of light.
The combination of optical lithography with a transfer technique based on a sacrificial substrate enables the creation of a 3D-racetrack memory made from state-of-the-art spintronic materials.
In this Review we survey the molecular sieving behaviour of metal-organic framework (MOF) and covalent organic framework (COF) membranes, which is different from that of classical zeolite membranes. The nature of MOFs as inorganic-organic hybrid materials and COFs as purely organic materials is powerful and disruptive for the field of gas separation membranes. The possibility of growing neat MOFs and COFs on membrane supports, while also allowing successful blending into polymer-filler composites, has a huge advantage over classical zeolite molecular sieves. MOFs and COFs allow synthetic access to more than 100,000 different structures and tailor-made molecular gates. Additionally, soft evacuation below 100 °C is often enough to achieve pore activation. Therefore, a huge number of synthetic methods for supported MOF and COF membrane thin films, such as solvothermal synthesis, seed-mediated growth and counterdiffusion, exist. Among them, methods with high scale-up potential, for example, layer-by-layer dip- and spray-coating, chemical and physical vapour deposition, and electrochemical methods. Additionally, physical methods have been developed that involve external stimuli, such as electric fields and light. A particularly important point is their ability to react to stimuli, which has allowed the 'drawbacks' of the non-ideality of the molecular sieving properties to be exploited in a completely novel research direction. Controllable gas transport through membrane films is a next-level property of MOFs and COFs, leading towards adaptive process deviation. MOF and COF particles are highly compatible with polymers, which allows for mixed-matrix membranes. However, these membranes are not simple MOF-polymer blends, as they require improved polymer-filler interactions, such as cross-linking or surface functionalization.
Nanoparticle design for photothermal heating of brain tumours
a, Schematic showing the chemical conjugation of Cy5-PEG to the surface of nanoparticles tagged with BPE Raman reporter molecules. b, TEM image demonstrating the star-shaped morphology of the nanoparticles for efficient photothermal heating in the NIR range (scale bar, 100 nm). c, Thermal image showing the photothermal response of the nanoparticles under NIR irradiation (scale bar, 2 mm). d,e, Bioluminescent imaging and T1-weighted MRI (post-Gd contrast) used to estimate the tumour size (human U87-eGFP-fLuc glioblastoma) before the injection of nanoparticles (unit of radiance is the number of photons per second that leave a square centimetre of tissue and radiate into a solid angle of one steradian (p s–1 cm–2 sr–1). f, Thermal images of an excised brain. NIR-triggered heating was only observed in the tumour areas injected with nanoparticles (left), without any noticeable heating effect in normal brain areas (right). g, Quantitative analysis of the thermal images showed significant difference in temperature variation within the tumour and normal brain areas (mean ± standard deviation, n = 5, two-sided Student’s t-test, p = 0.0001). h, Photograph of an excised brain showing sectioning locations in the tumour areas to collect six consecutive brain slices for histology and fluorescent microscopy analysis (scale bar, 2 mm). i, Histological analysis of an H&E-stained tissue slice showing the tumour areas (scale bar, 1 mm). j, Fluorescent imaging of the brain slices prepared in h (scale bars, 2 mm). Comparisons with counterpart histological analysis tissues verified that nanoparticles were only diffused into tumours (Supplementary Fig. 5 shows the control image).
Designing duty-cycled NIR-emitting devices for remote-controlled triggering of nanoparticles’ photothermal effect in the brain over a long-term (15 days) treatment cycle
a, Schematic showing the back- and front-side views of the device, and circuit design and components used for their fabrication. Flexibility of the device enables its curvilinear fitting on the mouse skull. b,c, Representative photograph of a device (scale bar, 3 mm) and SEM imaging at a selected point on the surface of the device showing a 2-µm-thick parylene polymer coating (scale bar, 10 µm; Supplementary Fig. 15). d, Thermal image of a device with received power of 19 dBm at the RX coil (λ = 810 nm emission, resulting in ~3 °C temperature difference in a droplet of nanoparticles) (scale bar, 1 mm). e,f, Conventional approach for wireless powering of the LEDs integrating a matching network, a rectifier and an optional energy-storage capacitor along with a circuit that programs the duty cycle of the LED (f). The duty cycle can be adjusted by employing a switch that allows for programmable duty-cycle generation (red box) using a continuous input or by modulating the incoming RF signal (red and blue waveforms in e). The definition and values of the device power efficiencies are provided in the Supplementary Text. g,h, Single (g) and back-to-back (h) LED circuit designs used for the wireless activation of nanoparticles’ photothermal effect. In the single LED design, the LED was turned on for only half the cycle, achieving a maximum duty cycle of 31.8%. However, in the back-to-back LED topology, each LED was turned on for half the cycle and therefore at least one LED was on during each half-cycle. This effectively doubles the duty cycle. i, Evaluation of nanoparticles’ heating effect during irradiation with 810 and 940 nm devices. The graph shows the variations in nanoparticles’ temperature with voltage amplitude (peak voltage Vp) with a sinusoidal input. A peak voltage of around 4 V (~80 mW input power) resulted in ~3 °C temperature difference in a droplet of nanoparticles deposited on top of a coverslip, due to their photothermal response. TEM analysis of the nanostars (Supplementary Fig. 8) shows their morphological stability after 15 min of photothermal activation for 15 consecutive days, verifying that duty-cycled irradiation of the nanoparticles with our back-to-back LED design helped to maintain the morphology of the nanoparticles and is suitable for long-term and consistent photothermal heating in the brain.
Tuning wireless power transfer efficiency and evaluating safety for photothermal therapy in freely behaving mice
a, Smith chart showing simulated and measured S11 for back-to-back devices with LEDs emitting 810-nm-wavelength light. The measured S11 of the antenna is also shown along with the ideal matching path (dashed line). b, L-match structure and matching components for the back-to-back devices. The 52 pF capacitance was constructed by placing 22 and 30 pF capacitors in parallel. Zant, impedance of the antenna. c, Variation in the matching network efficiency (ηMN) with input power (Pin), assuming capacitor quality factor of Qc = 100 for the matching components. d, Quantitative evaluation of mouse body interactions with 13.56 MHz RF waves to evaluate the safety of the wireless power transfer in our approach during wireless therapy cycles (compare with SAR values calculated for conventional 915.00 MHz RF waves given in Supplementary Fig. 20). The scale bar is logarithmic. e, Effect of tissue on the wireless link efficiency (ratio of the power received at the input of the devices to the transmitted power) after implantation on mouse skull, suggesting ~5 dB degradation in the power delivered to the RX antenna, due to tissue interaction. PLED, power received at the input of the devices; PTX, transmitted power from the transmitter coil. f, In vivo open-skull thermal images of the brain tumour, 24 h after intratumoral injection of nanoparticles. NIR-emitting device was fixed above the brain (Supplementary Figs. 22 and 23). g, Temperature variation along the blue dotted line shown in f, verifying that only the tumour area that contained the nanoparticles was heated during NIR irradiation (~30 mW) due to the nanoparticles’ photothermal response. We did not observe any elevated temperature in the surrounding normal brain tissues.
Wireless photothermal therapy of brain tumours in freely behaving mice
a,b, Schematic showing the computer-controlled wireless power delivery setup used for photothermal therapy of brain tumours. c, Multilevel powering scheme used for continuous therapy on a daily basis for 940 nm devices (Supplementary Figs. 24 and 30). d, Photograph showing a mouse eating during the wireless tumour therapy session, indicating that our approach did not disturb the animal’s normal behaviour (Supplementary Fig. 25 and Supplementary Video 1). e, Plots showing optical power versus the polar position of the NIR-emitting devices (940 nm, pre-implantation) from the centre of the wireless transmitter (TX) coil to evaluate variations in the optical power with height changes to account for the movement of mice. Supplementary Figs. 26–28 show the measurement setup and other related plots for both 810 and 940 nm devices. f, Survival profiles of the mice with human U87-eGFP-fLuc glioblastoma tumours treated with the wireless photothermal approach (treatment groups 1, 2 and 3) compared with control mice (n = 10 per group; total, 60). Nanoparticles (NPs) (1 µl, 0.5 nM) were intratumorally injected and photothermal therapy was started after 24 h (15 min per day for 15 days). (Control 1: NPs(–), implantation(–); control 2: NPs(+), microfibre(+), irradiation(–); control 3: NPs(+), device(+), irradiation(–); treatment 1: NPs(+), microfibre(+), irradiation (810 nm)(+); treatment 2: NPs(+), 810 nm device(+), irradiation(+); treatment 3: NPs(+), 940 nm device(+), irradiation(+)). Significant differences were observed when comparing each treatment group with the control profiles (p < 0.05, using the log-rank test). Supplementary Figs. 41–43 show the survival results for mice with GL26 and GBM39 tumours, as well as the combination therapy results. g, Secondary electron ((i) and (iii)) and backscattered ((ii) and (iv)) SEM images of a brain section, showing photothermal effect of the nanoparticles (porous areas indicated with arrows) in the tumour tissue. All the bright contrast spots in the backscattered images ((ii) and (iv)) represent gold nanoparticles due to their enhanced electron backscattering (Supplementary Figs. 44–48 show a more detailed histological analysis of the tumours at the end of therapies). Scale bars, 50 µm (i), 10 µm (ii) and 100 µm ((iii) and (iv)).
Current clinical brain tumour therapy practices are based on tumour resection and post-operative chemotherapy or X-ray radiation. Resection requires technically challenging open-skull surgeries that can lead to major neurological deficits and, in some cases, death. Treatments with X-ray and chemotherapy, on the other hand, cause major side-effects such as damage to surrounding normal brain tissues and other organs. Here we report the development of an integrated nanomedicine–bioelectronics brain–machine interface that enables continuous and on-demand treatment of brain tumours, without open-skull surgery and toxicological side-effects on other organs. Near-infrared surface plasmon characteristics of our gold nanostars enabled the precise treatment of deep brain tumours in freely behaving mice. Moreover, the nanostars’ surface coating enabled their selective diffusion in tumour tissues after intratumoral administration, leading to the exclusive heating of tumours for treatment. This versatile remotely controlled and wireless method allows the adjustment of nanoparticles’ photothermal strength, as well as power and wavelength of the therapeutic light, to target tumours in different anatomical locations within the brain. Current treatment of brain tumour entails open-skull tumour resection and follow-up X-ray radiation or chemotherapy, with surgery-associated risks and side-effects. Here a photothermal approach is presented that relies on wireless near-infrared stimulation for continuous, on-demand treatment of brain tumours in free-moving animals.
The global emergency caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic can only be solved with effective and widespread preventive and therapeutic strategies, and both are still insufficient. Here, we describe an ultrathin two-dimensional CuInP2S6 (CIPS) nanosheet as a new agent against SARS-CoV-2 infection. CIPS exhibits an extremely high and selective binding capacity (dissociation constant (KD) < 1 pM) for the receptor binding domain of the spike protein of wild-type SARS-CoV-2 and its variants of concern, including Delta and Omicron, inhibiting virus entry and infection in angiotensin converting enzyme 2 (ACE2)-bearing cells, human airway epithelial organoids and human ACE2-transgenic mice. On association with CIPS, the virus is quickly phagocytosed and eliminated by macrophages, suggesting that CIPS could be successfully used to capture and facilitate virus elimination by the host. Thus, we propose CIPS as a promising nanodrug for future safe and effective anti-SARS-CoV-2 therapy, and as a decontamination agent and surface-coating material to reduce SARS-CoV-2 infectivity. While vaccines have curbed the COVID-19 pandemic, effective therapeutic treatments are few, and might be challenged by SARS-CoV-2 variants. A biocompatible, antiviral two-dimensional nanomaterial is now reported that firmly adsorbs the virus by interaction with the spike protein, inducing the conformational changes that lead to inhibition of viral infection in vitro and in animal models.
Topological transition of hybrid polaritons
a,d, Illustration of the graphene/α-MoO3 vdW heterostructure used in this study, supported on SiO2 (a) and gold (d) substrates. b,e, Calculated isofrequency dispersion contours of hybrid polaritons on a 300-nm-thick SiO2 substrate (b) and a 60-nm-thick gold substrate (e) at a fixed incident frequency of 910 cm⁻¹ (λ0 = 10.99 µm) for different graphene Fermi energies ranging from 0 to 0.7 eV and an α-MoO3 film thickness of 150 nm. φ indicates the opening angle of the hyperbolic sectors. kx and ky are the momenta of polariton along the x and y crystal directions of α-MoO3, while k0 is the momentum of light in free space. c,f, Numerically simulated field distributions (real part of the z out-of-plane component of the electric field, Re{Ez}) of hybrid polaritons on SiO2 (c) and gold (f) substrates for several graphene doping levels at a fixed incident frequency of 910 cm⁻¹, as launched by a dipole placed 100 nm above the origin.
Topological transition of hybrid polaritons revealed by nanoimaging
a–c, Experimentally measured polariton near-field amplitude (S3) images with graphene doping EF = 0 eV (a), EF = 0.3 eV (b) and EF = 0.7 eV (c). The polaritons were launched by a gold antenna. The α-MoO3 film was placed on top of a 300 nm SiO2/500 μm Si substrate. d–f, Absolute value of the spatial Fourier transforms (FTs) of the experimental near-field images shown in a–c, respectively, revealing the isofrequency contours of hybrid polaritons. The grey curves represent calculated isofrequency contours. g–i, Experimentally measured polariton near-field amplitude (S3) images with graphene doping EF = 0 eV (g), EF = 0.3 eV (h) and EF = 0.7 eV (i) for an α-MoO3 film placed on a 60 nm Au/500 μm Si substrate. The canalized wavefronts were measured at a graphene Fermi energy close to the value at which the topological transition occurs (h), showing deep-subwavelength and diffractionless polariton propagation. j–l, Absolute value of the FTs of the experimental near-field images in g–i, respectively. The grey curves show the calculated isofrequency contours. The α-MoO3 thickness was 140 nm in all panels. The incident light wavelength was fixed at λ0 = 11.11 μm (900 cm⁻¹). Each colour scale applies to all images in the respective column.
Antenna-tailored launching of hybrid polaritons
a,c, Experimentally measured near-field amplitude (S3) images of hybrid polaritons launched by gold antennas with orientation angle θ in the 0–90° range (Supplementary Fig. 15) for a graphene Fermi energy EF = 0.1 eV (a) and EF = 0.7 eV (c) at a light frequency of 910 cm⁻¹. b,d, Numerically simulated near-field distributions (Re{Ez}, evaluated 20 nm above the surface of the heterostructure) corresponding to the measured results shown in a and c, respectively. e,f, Isofrequency dispersion contours extracted from the experimental data in a and c, respectively (red symbols), compared with the calculated hyperbolic dispersion contour (black solid curves) for an opening angle φ. The green arrows illustrate the direction of the exciting polariton wave vector k perpendicular to the long axis of the gold antenna. The α-MoO3 thickness was 207 nm in all panels. Scale bars in a–d, 2 μm. The error bars were extracted from four sets of measurements on the in situ sample (Supplementary Fig. 15). The artefacts observed in c and not in a can be attributed to the grain boundaries of polycrystalline graphene prepared by chemical vapour deposition⁴³.
Source data
Partial focusing of hybrid polaritons by substrate engineering
a, Schematic of the design, where the heterostructure lies on top of a substrate composed of a Au–SiO2–Au in-plane sandwich structure. b, Isofrequency dispersion contours of hybrid polaritons for Au and SiO2 substrates at 910 cm⁻¹ (λ0 = 10.99 μm). The shaded areas highlight the convex and concave dispersion contours in the region around the x axis on the gold and SiO2 substrates, respectively. With a wave vector inside the shaded area, negative refraction can happen at the Au–SiO2 interface when the polaritons on the gold substrate propagate towards that interface. The scheme for negative refraction is illustrated by further showing the incident wave vector ki and the Poynting vector Pi, together with the resulting transmitted kt and Pt. c, Experimentally measured near-field amplitude (S3) image of hybrid polaritons showing partial focusing in the system shown in a. The central SiO2 film was 1.5 µm wide and served as an in-plane flat lens. The antenna was located 1.0 µm away from the left Au–SiO2 interface. d, Experimentally measured hybrid polaritons on a Au substrate, as a control to c. Scale bar indicates 1.5 μm and also applies to c. e, Near-field profiles for the sections marked by the red (A) and blue (C) vertical dashed lines in c and d, respectively. The black dashed curves are Gaussian fittings. WA and WC indicate the full width at half maximum (FWHM) of profiles A and C, respectively. f, Near-field profiles of the sections marked by red (B) and blue (D) horizontal dashed lines in c and d, respectively. SB1, SB2, SD1 and SD2 represent the electric-field intensity at each fringe. The graphene was doped to EF = 0.6 eV and the α-MoO3 thickness was 320 nm.
Control over charge carrier density provides an efficient way to trigger phase transitions and modulate the optoelectronic properties of materials. This approach can also be used to induce topological transitions in the optical response of photonic systems. Here we report a topological transition in the isofrequency dispersion contours of hybrid polaritons supported by a two-dimensional heterostructure consisting of graphene and α-phase molybdenum trioxide. By chemically changing the doping level of graphene, we observed that the topology of polariton isofrequency surfaces transforms from open to closed shapes as a result of doping-dependent polariton hybridization. Moreover, when the substrate was changed, the dispersion contour became dominated by flat profiles at the topological transition, thus supporting tunable diffractionless polariton propagation and providing local control over the optical contour topology. We achieved subwavelength focusing of polaritons down to 4.8% of the free-space light wavelength by using a 1.5-μm-wide silica substrate as an in-plane lens. Our findings could lead to on-chip applications in nanoimaging, optical sensing and manipulation of energy transfer at the nanoscale. Polaritonic topological transitions of the isofrequency dispersion contour are observed in a graphene/α-MoO3 heterostructure by tuning the graphene doping level, which enables partial focusing at deep subwavelength.
Double ionic gated field-effect transistors
a, Top panel: schematic cross-section of a multilayer WSe2 transistor equipped with an IL top gate and a Li-ion conductive glass ceramic back gate. Also shown are the Pt contacts to the TMD multilayer and an Al2O3/Al/Al2O3 trilayer to decouple electrostatically the top and bottom electrolytes. Bottom panel: expansion of the device channel area (not to scale). When the two gates are biased with opposite polarity, the charges accumulated on the double layers of the two electrolytes (schematically shown in the image as red and blue circles) compensate, and a uniform perpendicular electric field E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}$$\end{document} is established across the TMD (represented by the red arrows in the scheme). b, ISD measured on a 2L WSe2 device for negative VBG values applied to the Li-ion glass gate (with the IL gate grounded, VIL = 0 V), resulting in the accumulation of holes. c, ISD measured for the same device as a function of VIL > 0 V for VBG = 0 V, to bring about electron accumulation (where the applied source–drain voltage is VSD = 0.1 V here and in b). Note that the application of a positive VIL and a negative VBG < 0 V causes the Li ions to be pulled away from the TMD, ensuring that intercalation does not take place. d,e, Band structure of 2L WSe2 computed within density functional theory for a zero perpendicular electric field E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}$$\end{document} = 0 V nm⁻¹ (d) and at the critical field E=Ec\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}} = {\mathcal {E}}_{\mathrm{c}}$$\end{document} (e), showing that quenching of the gap with an electric field is expected theoretically (see Supplementary Note 8 for details). In e, the conduction- and valence-band edges at E=Ec\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}} = {\mathcal {E}}_{\mathrm{c}}$$\end{document} are denoted by the red-shaded lines. Γ, M and K are the high-symmetry points along the Brillouin zone of the hexagonal lattice of WSe2.
Electrical characteristics of a double-gated 2L WSe2 transistor
a, ISD measured for a 2L WSe2 device as a function of VIL for different negative values of VBG. The curves evolve from showing textbook transistor behaviour at a small negative VBG value (see also Fig. 1b) to not showing any sizable current suppression at a large negative VBG. b, Same as for a, with ISD measured as a function of VBG for different positive values of VIL; the evolution of the transistor curves is fully analogous to that seen in a. c, Colour plot of ISD (in logarithmic scale) as a function of VBG and VIL. Note that the simultaneous application of a large negative VBG and an equally large positive VIL causes the current in the transistor to increase by between four and five orders of magnitude, despite leaving the potential of the transistor channel (V* = VIL + VBG) unchanged. d, Evolution of ISD along the A–B–C contour illustrated in c (where the coordinates of A, B and C in the (VIL, VBG) plane are indicated on the bottom axis). Transport in the transistor is mediated by holes at A and by electrons at C; finding that there is no region where the current is fully suppressed implies that, along part of the A–B–C contour, the valence and conduction bands of WSe2 must overlap. In all measurements, VSD = 0.1 V.
Bandgap evolution as a function of the electric field
a, Colour plot of ISD (in logarithmic scale) measured for a 2L WSe2 device as a function of V* = VIL + VBG and E* = VIL − VBG (respectively proportional to the electrostatic potential in the transistor channel and to the electric field perpendicular to the 2L WSe2 crystal). The width of the V* interval over which ISD is suppressed decreases monotonically on increasing E*. b, For a quantitative analysis, we look at horizontal cuts of the colour plot in a for ISD versus V* at a fixed E* (values indicated in the figure). The curve measured at E* = 4.5 V shows a complete suppression of ISD, and the thin black lines illustrate how we determine the threshold voltage for electron conduction (VT-e*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{T}}\textit{-}{\mathrm{e}}}^{*}$$\end{document}) and hole conduction (VT-h*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{\mathrm{T}}\textit{-}{\mathrm{h}}}^{*}$$\end{document}) (by extrapolating ISD to zero for positive and negative values of V*, and the positions of the threshold voltages are marked by the vertical dashed lines). c, Cut of the colour plot in a, taken at V* = 0.5 V, that corresponds to the vertical green dashed line. The data show a transition from a highly resistive state at low E* to a state with a conductivity of the order h/e² at E* > 4 V. This transition, which occurs at fixed V*, is a direct manifestation of the quenching of the bandgap caused by the applied electric field. d, Plot of the difference δ=VT-e*−VT-h*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta ={V}_{{\mathrm{T}}\textit{-}{\mathrm{e}}}^{*}-{V}_{{\mathrm{T}}\textit{-}{\mathrm{h}}}^{*}$$\end{document} as a function of E*, to find the value of Ec*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${E}_{{\mathrm{c}}}^{*}$$\end{document} for which δ = 0 V (Ec*\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${E}_{{\mathrm{c}}}^{*}$$\end{document} = 4 V in our 2L WSe2 device) to determine the condition at which the gap closes.
Quenching the gap in 3L, 4L and 5L WSe2 devices
a,c,e, Colour plots of ISD as a function of V* = VIL + VBG (the sum of the IL and back gate voltages) and the electric field E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}$$\end{document} for 3L (a), 4L (c) and 5L (e) devices. b,d,f, Corresponding cuts (ISD–V* curves) for different fixed values of the electric field E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}$$\end{document} across each WSe2 multilayer for the 3L (b), 4L (d) and 5L (f) devices. g, Comparison between the maximum electric fields that can be applied across nL WSe2 using different types of device. Green dots: field achieved in ionic gated devices for E* = VIL − VBG = 5 V (larger values are possible, as we repeatedly reached 5.5 and 6 V; see Supplementary Note 5 for details on the estimation of E\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}$$\end{document}); blue and orange dots: maximum field reachable with devices based on hexagonal boron nitride (hBN) assuming a breakdown field of 1.1 and 0.6 V nm⁻¹, respectively (1.1 V nm⁻¹ is the ultimate limit reached in ultra-thin hBN; 0.6 V nm⁻¹ is a more realistic estimate for common hBN). For 2L WSe2, the electric field reachable with ionic gating is nearly one order of magnitude larger than the field accessible in hBN-based devices. The black crosses are the field needed to close the gap, as extracted from our measurements; for all thicknesses, it would have not been possible to close the gap with hBN-based devices. h, Comparison between the critical electric field Ec\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}_{\mathrm{c}}$$\end{document} values extracted from experiments (black crosses) and the corresponding theoretical values obtained from first-principles calculations (open diamonds). Ab initio calculations slightly underestimate the gap, resulting in a small underestimation of Ec\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}_{\mathrm{c}}$$\end{document}. To take this into account, the filled diamonds show the value of the critical electric field obtained using the ratio between the known and the calculated gap to rescale Ec\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal {E}}_{\mathrm{c}}$$\end{document} (see Supplementary Note 8). The agreement with the experimental data is excellent.
Perpendicular electric fields can tune the electronic band structure of atomically thin semiconductors. In bilayer graphene, which is an intrinsic zero-gap semiconductor, a perpendicular electric field opens a finite bandgap. So far, however, the same principle could not be applied to control the properties of a broader class of 2D materials because the required electric fields are beyond reach in current devices. To overcome this limitation, we design double ionic gated transistors that enable the application of large electric fields of up to 3 V nm−1. Using such devices, we continuously suppress the bandgap of few-layer semiconducting transition metal dichalcogenides (that is, bilayer to heptalayer WSe2) from 1.6 V to zero. Our results illustrate an excellent level of control of the band structure of 2D semiconductors. Double ionic gated transistors enable excellent control of the band structure of atomically thin semiconductors. Perpendicular electric fields as large as 3 V nm−1 can fully quench the gap of bi- and few-layer WSe2.
Magneto-birefringence effect of 2D h-BN suspension
a–c, Lateral size (a), thickness (b) and volume (c) distributions of 2D h-BN. Inset in c shows that the volume Vs of each h-BN flake is calculated via multiplying its area S and thickness T, both of which are obtained based on atomic force microscopy measurements. Scale bar, 400 nm. d, Schematic view of the optical path for the magneto-optical experiment, and explanation of the optical switching due to the birefringence induced by magnetic alignment of 2D h-BN. e, Real photographs of the cuvette filled with 2.4 × 10⁻³ vol% h-BN aqueous suspensions backlit with white light. The cuvette, with dimensions of 10 × 10 × 40 mm, is sandwiched between two crossed polarizers. The magnetic field is applied perpendicular to the light path and oriented at 45° to the transmission axis of the polarizer. Left: black dominates if the magnetic field is absent; right: a white colour is shown when the magnetic field is switched on. f, Real images of the ‘TBSI SGC’ and panda-patterned h-BN suspension under various magnetic fields. Scale bar, 5 mm.
Source data
Magnetic-field-induced alignment and magneto-birefringence of 2D h-BN inorganic LCs
a, Transmittance of polarized laser light through the 2D h-BN LCs in the polarization direction parallel with (∥) and perpendicular to (⟂) a magnetic field of 0.8 T. The right insets illustrate the transmittance variation due to polarization-dependent optical loss by absorption or scattering from the aligned 2D h-BN. b, Polarization state evolution of the output light. The polarization ellipses are described in terms of the ellipticity (η) and azimuthal rotation (ψ), which give η = −0.1° and ψ = 44.4° at 0 T (left) and η = 10.6° and ψ = 57.4° at 0.8 T (right). x and y are the horizontal and vertical components of the electric-field vector. Dashed line indicates the long axis of ellpitical polarization. c, Real images of a layered h-BN laminate with (right) and without (left) a magnetic field. d, Magneto-birefringence in the range of 0–0.8 T. e, Relationship between birefringence and the square of the magnetic field in the low range of 0–0.06 T.
Source data
Performance of the 2D h-BN inorganic LC-based DUV modulator
a, On–off switching of UV-C light (266 nm). The strength of the turn-on magnetic field is set as 0.8 T. b, Photoluminescence of red, green and blue fluorescent dyes excited by the transmitted 266 nm light. c, Correspondence of the transmitted 266 nm light intensity with the strength of the magnetic field in the range of 0–0.8 T at intervals of 0.2 T. d, Loop test of field-strength correspondence. e, Transient optical signal of transmitted DUV light in response to a magnetic pulse with the peak strength of 1.1 T and a full-width at half-maximum of 7 ms. f, Cycling test for characterizing the DUV stability. g, Comparison of commercialized and research and development (R&D) LC-based optical modulators with proven stability in their respective spectral range.
Source data
Birefringence is a fundamental optical property that can induce phase retardation of polarized light. Tuning the birefringence of liquid crystals is a core technology for light manipulation in current applications in the visible and infrared spectral regions. Due to the strong absorption or instability of conventional liquid crystals in deep-ultraviolet light, tunable birefringence remains elusive in this region, notwithstanding its significance in diverse applications. Here we show a stable and birefringence-tunable deep-ultraviolet modulator based on two-dimensional hexagonal boron nitride. It has an extremely large optical anisotropy factor of 6.5 × 10−12 C2 J−1 m−1 that gives rise to a specific magneto-optical Cotton–Mouton coefficient of 8.0 × 106 T−2 m−1, which is about five orders of magnitude higher than other potential deep-ultraviolet-transparent media. The large coefficient, high stability (retention rate of 99.7% after 270 cycles) and wide bandgap of boron nitride collectively enable the fabrication of stable deep-ultraviolet modulators with magnetically tunable birefringence. A 2D material based liquid-crystal shows an extremely large optical anisotropy factor in the deep ultraviolet region, showing magnetically tunable birefringence.
Concept of direct patterning with dual-ligand QDs
a, Chemical structure of PXLs. Pyrrolidinyl (–N(CH2)4), oxy (–O–) and thio (–S–) groups are at the para positions of benzophenone to modulate its photochemical properties. b, Schematic of dual-ligand QDs. The addition of 1–10 mol% PXLs turns QD films to become crosslinkable on UV irradiation. The DLs (>90 mol%) determine the solubility of QDs. c, Schematic illustrating the photocrosslinking between dual-ligand QDs. On UV irradiation, the carbonyl group yields a radical and forms a covalent bond with the ligands of neighbouring QDs. d, Fluorescent images (top) and a schematic of laterally pixelated and stacked RGB QD patterns (bottom) that are fabricated using dual-ligand RGB QDs. Scale bars, 200 μm. e, A photograph of RGB QD patterns on a six-inch silicon wafer attained by consecutive photolithographic QD-patterning processes using an i-line stepper.
Source data
Structurally engineered PXLs for non-destructive QD photocrosslink
a, Chemical structures of benzophenone and PXLs with different chemical substitutions to the para positions of benzophenone (NS–BP, S–BP and O–BP). Prefixes (O–, S– and NS–) indicate chemical elements substituted for benzophenone. b, Molar extinction spectra for PXLs and unsubstituted benzophenone. The inset shows the semi-log plots of molar extinction spectra for PXLs and the unsubstituted benzophenone between 300 and 450 nm. c,d, Exposure-dose-dependent film retention ratios (c) and fluorescent images of QD films having different PXLs (all the films are exposed to UV radiation with an exposure dose of 630 mJ cm–2 and rinsed with toluene) (d). The error bars in c indicate the standard deviations of the data acquired from five independent runs. Scale bars, 50 μm. e, Normalized PL QYs of QD films (film retention ratio, >0.9) employing different PXLs after the photocrosslinking and rinsing steps under ambient condition. The error bars represent the standard deviations of five independent runs. Exposure-dose-dependent changes in the PL QY of QD films exposed to different wavelengths of UV sources (namely, 365 nm (blue) and 254 nm (purple)) are overlaid for comparison. f, PL spectra of photocrosslinked RGB QD films with NS–BP. InP (core radius, r = 1.9 nm)/ZnSexS1–x (shell thickness, h = 3.2 nm) QDs, InP (r = 1.2 nm)/ZnSexS1–x (h = 2.3 nm) QDs and CdxZn1–xS (r = 2.7 nm)/ZnS (h = 3.6 nm) QDs are adopted as the red, green and blue emitters, respectively. A fixed amount of PXLs (7 mol%) is grafted to each coloured QD. All the QD films are exposed to UV-A (365 nm; exposure dose, 35 mJ cm–2).
Source data
Multicoloured patterns made of dual-ligand QDs
a, Fluorescent image (left, top), scanning electron microscopy image (left, bottom) and atomic force microscopy image (right, top) and the height profile (right, bottom) of QD line patterns (width, 3.6 μm; spacing, 6.6 μm) obtained from a single photolithographic process using an i-line stepper. Line-edge roughness and linewidth roughness are estimated to be 74 and 99 nm, respectively. Scale bars, 10 μm. b, Fluorescent images of RGB QD patterns obtained after consecutive photolithographic processes of primary-coloured QDs using an i-line stepper. Scale bars, 10 μm. The dimensions of the subpixels are 3.8 × 3.8 μm² (left), 1.8 × 1.8 μm² (right, top) and 0.8 × 0.8 μm² (right, bottom), which correspond to the resolution indicated in the images. c–e, Fluorescent images composed of laterally positioned and vertically stacked RGB QD patterns obtained with a contact aligner. Here d is a magnified view of the marked square in c. Scale bars, 1 mm (c and e, left); 200 μm (d and e, right). f, A photograph of dual-ligand RGB QD dispersions in TFT, PGMEA and hexane. TFMBT, MMES and OA are DLs to render QDs dispersed in TFT, PGMEA and hexane, respectively. g, A series of images showing an ejected droplet including dual-ligand QDs with a time interval of 14 μs. h, Fluorescent images (top) and intensities across the indicated green line (bottom) of inkjet-printed RGB crossline patterns attained with photocrosslinked QDs (left) versus pristine QDs (right). Scale bars, 1 mm. Credit: c,d, adapted from JanPietruszka/iStock/Getty Images Plus/Getty.
Source data
Optoelectronic devices implementing photocrosslinked QD patterns
a, Current density−voltage characteristics of an electron-only device (EOD, left) and hole-only device (HOD, right) implementing photocrosslinked QD films. b,c, Schematic of the device architecture (b, top) and energy-band diagram (b, bottom) and current-density-dependent external quantum efficiencies (EQEs) of QD-LEDs implementing photocrosslinked QD films (c). Device characteristics with pristine QDs (oleic acids only) are shown for comparison. CBP and ZnMgO are used as the hole transport layer (HTL) and electron transport layer (ETL), respectively, for EOD, HOD and QD-LEDs. The inset shows a photograph of the operating QD-LED. d, Schematic showing passive-matrix-driven 10 × 10 RGB QD-LED arrays employing patterned QD films. e,f, Cross-sectional schematic (e) and associated electric circuit of RGB pixels (f). g,h, Electroluminescent images of 10 × 10 RGB QD-LED arrays (g) and QD-LED array of each primary colour (h). Scale bars, 2 mm. All the QD films are prepared by spin casting and photolithography. Supplementary Fig. 24 shows the other device applications.
Source data
Colloidal quantum dots (QDs) stand at the forefront of a variety of photonic applications given their narrow spectral bandwidth and near-unity luminescence efficiency. However, integrating luminescent QD films into photonic devices without compromising their optical or transport characteristics remains challenging. Here we devise a dual-ligand passivation system comprising photocrosslinkable ligands and dispersing ligands to enable QDs to be universally compatible with solution-based patterning techniques. The successful control over the structure of both ligands allows the direct patterning of dual-ligand QDs on various substrates using commercialized photolithography (i-line) or inkjet printing systems at a resolution up to 15,000 pixels per inch without compromising the optical properties of the QDs or the optoelectronic performance of the device. We demonstrate the capabilities of our approach for QD-LED applications. Our approach offers a versatile way of creating various structures of luminescent QDs in a cost-effective and non-destructive manner, and could be implemented in nearly all commercial photonics applications where QDs are used. A dual-ligand passivation system comprising photocrosslinkable ligands and dispersing ligands enables quantum dots to be universally compatible with solution-based patterning techniques.
Shamrock phononic insulator
a, SEM (tilted-top view) of the fabricated structure on an SOI substrate with a thickness of t = 220 nm. Inset: schematic illustration of the geometrical parameters of the unit cell (highlighted in red) with lattice constant a = 330 nm, hole radius r = 0.22a, and the distance between the centre of the shamrock and the centre of each circle f=2r/3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f=2r/\sqrt{3}$$\end{document}. b, Simulated three-dimensional phononic dispersion relation of the crystal over the first Brillouin zone. Blue and red curves indicate the symmetric and asymmetric modes with respect to the middle plane of the silicon slab at t/2. c, Calculated phononic density of states (DOS) of the structure. The light-blue region highlights the full mechanical gap spanning 6.7 GHz to 11.4 GHz.
Brillouin light scattering spectroscopy
a, Schematic illustration of Brillouin scattering with the phase-matching condition for the backward configuration used in the experiments. Here, ki\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbf{k}}_{\mathrm{i}}$$\end{document} and ks\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbf{k}}_{\mathrm{s}}$$\end{document} represent the incident and the scattered light, respectively, and q∥\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbf{q}}_{\parallel }$$\end{document} is the parallel mechanic wavevector. The magnitude of q∥\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbf{q}}_{\parallel }$$\end{document} depends on the incident angle, where q∥=2kisinθ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${q}_{\parallel }=2{{k}}_{{\mathrm{i}}}\sin \theta$$\end{document}. b, Measured Brillouin scattering spectrum for an incident angle of θ = 32. 5∘ with p-polarized light. The green central peak stems from elastic Rayleigh scattering. Negative and positive frequency peaks on either side of this large central peak correspond to Stokes and anti-Stokes contributions, respectively. The light-blue regions highlight the mechanical gap. c, Calculated dispersion relation based on the geometrical parameters obtained from SEM images of the fabricated samples that include a 4∘ sidewall angle correction in the vertical profile (inset). The black dots represent the measured frequencies of vibrational modes for different angles and the vertical dotted line indicates the frequencies obtained from the measured spectrum shown in b. The intensity colour scale represents the normalized coupling coefficients for the moving-boundary perturbation. d–f, The direction in which the sample is physically rotated to scan along the highest-symmetry directions ΓK (d), KM (e) and ΓM (f). The green arrows indicate the direction of the incident laser light while the other coloured arrows correspond to the rotation direction during measurements, which represent (and are colour-consistent with) the highest-symmetry direction indicated in c.
Hypersonic phononic waveguide
a, SEM image of a shamrock phononic waveguide with a lattice period of a = 440 nm and waveguide width of w = 184 nm. The thickness and radius of the structure are the same as in previous structures (t = 220 nm, r = 0.22a). b, Measured Brillouin scattering spectra in the waveguide (top) and surrounding phononic crystal (bottom) as is illustrated in the insets, for an incident light angle of 23.8∘. The spectral width of the measured gap is indicated by the blue regions. Two peaks whose frequencies correspond with the guided modes of the system appear inside the gap in the spectrum of the waveguide (top). c, The calculated dispersion relation of the waveguide. The intensity colour scale represents the normalized coupling coefficient for the moving-boundary perturbation. The horizontal and vertical dotted lines indicate the mechanical band edges and phononic wavevector, respectively, while the black dots represent the frequencies of peaks 1 and 2, all for the top waveguide spectrum shown in b. Insets: the mode profiles for the indicated bands where the colour represents the normalized out-of-plane displacement.
Controlling vibrations in solids is crucial to tailor their elastic properties and interaction with light. Thermal vibrations represent a source of noise and dephasing for many physical processes at the quantum level. One strategy to avoid these vibrations is to structure a solid such that it possesses a phononic stop band, that is, a frequency range over which there are no available elastic waves. Here we demonstrate the complete absence of thermal vibrations in a nanostructured silicon membrane at room temperature over a broad spectral window, with a 5.3-GHz-wide bandgap centred at 8.4 GHz. By constructing a line-defect waveguide, we directly measure gigahertz guided modes without any external excitation using Brillouin light scattering spectroscopy. Our experimental results show that the shamrock crystal geometry can be used as an efficient platform for phonon manipulation with possible applications in optomechanics and signal processing transduction. Nanopatterned materials provide control over mechanical vibrations. This allows for the complete damping of vibrations over more than 5 GHz and for the propagation of hypersonic guided modes at room temperature.
A clever phononic crystal design produces a wide band gap for hypersonic phonons.
CrisprZyme assay scheme
Schematic of the combination of a Cas-based reaction with a NLISA proposed in this study. Target RNA is mixed with the gRNA–Cas13 complex and triggers collateral cleavage of reporter RNA. Subsequently, the mixture is added to an immunoassay plate precoated with anti-FAM. The unbound reporter RNA is washed away, and the nanozymes are added to form a complex through the bound reporter RNA. Finally, the substrate is added for colour development.
Pt@Au functionalized with streptavidin shows the best NLISA performance
a, Schematic showing the synthesis of Pt@Au. b, Characterization of the functionalized Pt@Au by S/N ratio and DLS with PBST and PBST supplemented with two different blocking agents: β-casein and BSA. Control corresponds to non-functionalized particles. Data points represent individual experiments. Error bars represent s.d. (n > 3 replicates). A450, absorbance measure at 450 nm. c, Number distribution of the hydrodynamic diameter of the Pt@Au prepared by overgrowing different amounts of Pt onto the surface of AuNP seeds or 120 nm Pt@Au. Data represent the mean ± s.d. (n = 3 replicates). PSD, particle size distribution. d, Sigmoidal regression curve of the reporter RNA with streptavidin-Pt@Au of different sizes. Data represent the mean (n = 2 replicates). e, Structural characterization of non-functionalized and streptavidin-functionalized Pt@Au using TEM. Scale bars (from left to right), 200 nm, 50 nm and 20 nm. f,g, STEM–EDS analysis of streptavidin-functionalized Pt@Au. f, Representative HAADF-STEM image and EDS elemental mapping (Pt and Au). A merged image of Pt and Au maps is shown. Scale bar, 50 nm. g, Representative EDS spectra recorded from the whole area of the individual particle.
CrisprZyme detects synthetic RNA down to picomolar concentration
a, Photograph of CrisprZyme results in a 384-well plate; six replicates were performed for each concentration. Low concentration of a lncRNA target (lnc-LIPCAR) to the left; high concentration to the right. b, Sigmoidal regression of the CrisprZyme of a lncRNA target. Data were obtained measuring the absorbance of each well with a plate reader at 450 nm (A450). c, Sigmoidal regression of the CrisprZyme of a lncRNA target. Data were obtained by taking a photo of the results and recording the blue intensity of each well set as a region of interest. b,c, Data represent the mean ± s.d. (n = 6 replicates). d, Schematic of the combination of a Cas-based reaction with a nanozyme-amplified LFA proposed in this study. Target RNA is mixed with the gRNA–Cas13 complex and reporter RNA to trigger the CRISPR reaction. Subsequently, streptavidin-functionalized nanozymes were mixed with CRISPR reaction product containing the biotinylated reporter RNA to form a complex. A test strip preprinted with anti-FAM was used to draw up the mixture. The uncleaved reporter RNA-nanozymes complexes were captured at the test line. Finally, the substrate was added for colour development. e, Detection of a serial dilution of lnc-LIPCAR with nanozyme-amplified LFA. Photographs show the test bands of the lateral flow test strips after completion of the assay without (top) and with (bottom) the substrate added for enhanced signal. f, Sigmoidal regression of the lnc-LIPCAR target for the nanozyme-amplified LFA. Data represent the mean of test line pixel density normalized to the internal grid lines of the light box. Data represent the mean ± s.d. (n = 3 replicates). g, Sigmoidal regression curve parameters. The data have been extracted from the four-parameter equation (Supplementary equation 1) used to fit the standard curve.
CrisprZyme expands the dynamic range of Cas13-based diagnostics enabling the quantitative sensing of different non-coding RNA species
a, Standard curves for the detection of serial dilutions of lnc-LIPCAR with CrisprZyme (blue) or a combination of RT–RPA and CrisprZyme (orange). Data represent the mean ± s.d. (n = 6 replicates). b, Comparison of the performance of CrisprZyme with different ncRNA targets: lnc-LIPCAR, miR-223, synthetic ncRNA (71-bp RNA) and circ-AURKA. Data represent the mean ± s.d. (n ≥ 3 replicates). c, Detection of miR-223 with CrisprZyme (orange) or RT–qPCR (grey) in five different samples at the indicated concentrations. Ct, cycle threshold for PCR reactions. Data represent individual replicates (n ≥ 2 replicates).
Quantification of different ncRNA species from cell culture and human plasma or tissue
a, Schematic representation of the sample analysis experimental workflow. b, Schematic of upregulated (red) miRNAs (grey) upon differentiation of iPSCs to cardiomyocytes (CMs) in vitro (left). Expression of miR-143-3p in iPSCs and cardiomyocytes as measured by CrisprZyme (N = 4 samples, n = 2 replicates). c, Schematic of upregulated (red) lncRNAs (grey) in plasma of patients with heart failure. Expression of lnc-LIPCAR in patients with heart failure and a control group as measured by CrisprZyme (N ≥ 25 samples, n = 2 replicates). d, Schematic of upregulated (red) circRNAs (grey) in tissue biopsies from patients with prostate cancer. Expression of circ-AURKA in biopsies of ACA and NEPC as measured by CrisprZyme (N = 10 samples, n = 2 replicates). Data points represent the median of two replicates. Box and whisker plots represent median and quartiles. Mann–Whitney test.
CRISPR-based diagnostics enable specific sensing of DNA and RNA biomarkers associated with human diseases. This is achieved through the binding of guide RNAs to a complementary sequence that activates Cas enzymes to cleave reporter molecules. Currently, most CRISPR-based diagnostics rely on target preamplification to reach sufficient sensitivity for clinical applications. This limits quantification capability and adds complexity to the reaction chemistry. Here we show the combination of a CRISPR–Cas-based reaction with a nanozyme-linked immunosorbent assay, which allows for the quantitative and colorimetric readout of Cas13-mediated RNA detection through catalytic metallic nanoparticles at room temperature (CrisprZyme). We demonstrate that CrisprZyme is easily adaptable to a lateral-flow-based readout and different Cas enzymes and enables the sensing of non-coding RNAs including microRNAs, long non-coding RNAs and circular RNAs. We utilize this platform to identify patients with acute myocardial infarction and to monitor cellular differentiation in vitro and in tissue biopsies from prostate cancer patients. We anticipate that CrisprZyme will serve as a universally applicable signal catalyst for CRISPR-based diagnostics, which will expand the spectrum of targets for preamplification-free, quantitative detection.
A series of emergent electronic orders are observed in an antiparallel twisted WSe2 bilayer. The discoveries provide a powerful platform for simulating quantum phenomena in strongly correlated materials.
Delivering light therapy using a remotely controlled bioelectronic device implanted above the brain might complement current glioblastoma therapies, reducing cancer recurrence and improving survival.
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has already infected more than 500 million people globally (as of May 2022), creating the coronavirus disease 2019 (COVID-19) pandemic. Nanotechnology has played a pivotal role in the fight against SARS-CoV-2 in various aspects, with the successful development of the two highly effective nanotechnology-based messenger RNA vaccines being the most profound. Despite the remarkable efficacy of mRNA vaccines against the original SARS-CoV-2 strain, hopes for quickly ending this pandemic have been dampened by the emerging SARS-CoV-2 variants, which have brought several new pandemic waves. Thus, novel strategies should be proposed to tackle the crisis presented by existing and emerging SARS-CoV-2 variants. Here, we discuss the SARS-CoV-2 variants from biological and immunological perspectives, and the rational design and development of novel and potential nanotechnology-based strategies to combat existing and possible future SARS-CoV-2 variants. The lessons learnt and design strategies developed from this battle against SARS-CoV-2 variants could also inspire innovation in the development of nanotechnology-based strategies for tackling other global infectious diseases and their future variants. This Perspective highlights the role that nanotechnology might play in tackling the rise of new SARS-CoV-2 variants.
| The spectrum of perspectives on nanomaterials descriptors. The figure illustrates viewpoints from the contexts of modelling, design, the behaviour of ENMs in a system, the subject matter of encoded ENM features and the sources from which nanodescriptors are obtained.
| The concept of ENM representation. This concept combines the substance identity concept with the definition of a nanoform, adapted from the classic substance paradigm 21 , and includes unique identifiers for the ENMs' composition. These consist of the core components, impurities and coatings and their relationships, as well as their measured or calculated characteristics supported with metadata related to protocols and experimental conditions. The symbols 'S', 'D' and 'P' code chemical structure, molecular descriptors and molecular properties, respectively. The symbol 'f' codes the mathematical function of relationships between the above mentioned elements of the chemical triad.
Engineered nanomaterials (ENMs) enable new and enhanced products and devices in which matter can be controlled at a near-atomic scale (in the range of 1 to 100 nm). However, the unique nanoscale properties that make ENMs attractive may result in as yet poorly known risks to human health and the environment. Thus, new ENMs should be designed in line with the idea of safe-and-sustainable-by-design (SSbD). The biological activity of ENMs is closely related to their physicochemical characteristics, changes in these characteristics may therefore cause changes in the ENMs activity. In this sense, a set of physicochemical characteristics (for example, chemical composition, crystal structure, size, shape, surface structure) creates a unique ‘representation’ of a given ENM. The usability of these characteristics or nanomaterial descriptors (nanodescriptors) in nanoinformatics methods such as quantitative structure–activity/property relationship (QSAR/QSPR) models, provides exciting opportunities to optimize ENMs at the design stage by improving their functionality and minimizing unforeseen health/environmental hazards. A computational screening of possible versions of novel ENMs would return optimal nanostructures and manage ('design out') hazardous features at the earliest possible manufacturing step. Safe adoption of ENMs on a vast scale will depend on the successful integration of the entire bulk of nanodescriptors extracted experimentally with data from theoretical and computational models. This Review discusses directions for developing appropriate nanomaterial representations and related nanodescriptors to enhance the reliability of computational modelling utilized in designing safer and more sustainable ENMs. This Review discusses how a comprehensive system for defining nanomaterial descriptors can enable a safe-and-sustainable-by-design concept for engineered nanomaterials.
Schematics and isofrequency contours of the hybrid plasmon-phonon-polaritons. a, Top: schematic representation of the plasmon-phonon-polariton formed via hybridization between the isotropic graphene plasmon and anisotropic phonon-polariton in α-MoO3 on a dielectric substrate. Bottom: image charges in the metal substrate result in a more compressed image polariton mode. b, Varying Fermi level in graphene leads to the transitioning of the hybrid polariton's isofrequency dispersion contour from the open shape at zero doping (that of the hyperbolic phonon-polariton in α-MoO3) to a closed shape at higher Fermi levels. The higher-momentum image mode (bottom row) exhibits flattened isofrequency contour at Fermi level of ~0.3 eV, providing a diffractionless canalization of the in-plane mode energy. Isofrequency contours are revealed by the imaginary part of the complex reflection coefficient of the heterostructure calculated at 910 cm -1 assuming negligible ohmic loss in graphene.
Heterostructure of graphene and biaxial van der Waals crystal supports a species of plasmon-phonon-polaritons whose isofrequency dispersion contour can be manipulated while experiencing a topological transition.
A ‘dual-ligand passivation system’ is designed and synthesized to functionalize colloidal quantum dots to realize ultra-high resolution patterns by direct photolithography.
Catalytically active, self-propelled microrobots can trap and remove nanoplastics from water.
Pufin ID is a Danish start-up commercializing anti-counterfeiting technology based on nanoscale photophyics.
Intrinsically stretchable QD-based semiconducting nanocomposites enable the realization of electronic retina with multispectral response.
Twisted WSe2 AB-homobilayer
a, Twisted WSe2 bilayer exhibiting three types of high-symmetry stacking sites MX, MM and XX (M = W and X = Se) for twist angle δ ≈ 60°. b, Schematic of a dual-gated WSe2 moiré bilayer device. The right half contains an exciton (bound electron (e)–hole (h) pair) sensor that is made of WSe2 monolayer and separated from the sample by a thin hBN spacer. c, Hole–spin alignment at the −K and K valleys in the top (t) and bottom (b) layers of WSe2 AB-homobilayers. The dashed line indicates the Fermi level. Spin and valley are locked in each layer. Within each valley spins are anti-aligned in two layers. Interlayer tunnelling is spin-forbidden. d, Illustration of the bilayer Hubbard model with intralayer hopping much smaller than intra- and interlayer on-site repulsions. The spin and layer degrees of freedom are addressable by an out-of-plane magnetic (B) and electric field (E), respectively. The arrows denote the positive field directions. e, Reflectance contrast (ΔR/R0) spectrum of the device with the sensor versus top-gate voltage. The back gate is fixed at 0 V. The corresponding doping density is shown on the top axis. The 2s and 1s exciton resonances of the sensor and the moiré exciton resonances of the twisted bilayer are shown in descending order of energy. A series of insulating states, manifested as blueshift of the sensor 2s resonance at integer and some fractional filling factors, are observed for both electron and hole doping.
Source data
Electric-field-controlled layer polarization
a,b, Electric-field dependence of the reflectance contrast spectrum of the sensor 2s exciton at ν = 1 (a) and of the moiré exciton of the twisted bilayer at ν = 2 (b). The 2s resonance energy shifts near E = 0 V nm⁻¹ due to charge transfer within the moiré bilayer at ν = 1. The layer polarization is saturated beyond Ec (white dashed lines). Above 20 mV nm⁻¹ the sensor is doped and the 2s exciton is quenched. At ν = 2, two moiré exciton features (black dashed lines) are replaced by one prominent feature above Ec (white dashed lines). c, Doping dependence of Ec (symbols) determined by the exciton sensor. The dashed lines are guides to the eye. The blue lines are the expected dependence from equation (1). d, Averaged reflectance contrast over 5 meV near the moiré exciton resonance of the twisted bilayer at 1.70 eV. The strong and weak contrast regions correspond to layer polarization |P| = 1 and |P| < 1, respectively. The critical field Ec in c (dashed line) is slightly larger than in d probably due to the larger twist angle (by ~0.1°). The insets illustrate the layer polarization beyond Ec.
Source data
Magnetic properties
a–c, Magnetic-field dependence of MCD at representative temperatures for ν = 1 (a), ν = 1.6 (b) and ν = 2 (c). Top panels, P = 0; bottom panels, P = 1. The legend in a defines the temperature in all panels except the bottom panel of c. The response is PM for all cases with P = 0. With P = 1, the response is PM for ν = 1 and metamagnetic for ν = 2. For 1 < ν < 2 both responses are present with their ratio depending on doping density. Insets: charge/spin configuration with P = 1 at zero magnetic field and 1.7 K. Charges are localized on the MX site (forming a triangular lattice) with random spin orientations at ν = 1. Charges are segregated into clusters that are AF coupled (black arrows) and isolated spins (orange arrows) at ν = 1.6. The dotted circles denote empty sites. Charges form a honeycomb lattice with Néel-type AF order at ν = 2. d–f, Temperature dependence of the inverse magnetic susceptibility for ν = 1 (d), ν = 1.6 (e) and ν = 2 (f). Symbols, experiment; lines, fit to the Curie–Weiss law. Distinct magnetic responses are observed for the P = 0 (blue) and the P = 1 case (black).
Source data
Local moments and antiferromagnetic clusters
a, MCD at B = 2 T as a function of electric field and doping density/filling factor. The MCD is probed at a wavelength near the moiré exciton resonance of the twisted bilayer. It is approximately proportional to the density of nearly isolated local moments. The dashed lines show the threshold electric field determined from Fig. 2. Large local moment density is observed for the layer-unpolarized state. b, Linecuts of a showing doping dependence of MCD with P = 0 (blue) and P = 1 (black). The red dashed lines denote the expected density dependence of the local moment density for both cases.
Source data
Moiré materials with flat electronic bands provide a highly controllable quantum system for studies of strong-correlation physics and topology. In particular, angle-aligned heterobilayers of semiconducting transition metal dichalcogenides with large band offset realize the single-band Hubbard model. Introduction of a new layer degree of freedom is expected to foster richer interactions, enabling Hund’s physics, interlayer exciton condensation and new superconducting pairing mechanisms to name a few. Here we report competing electronic states in twisted AB-homobilayer WSe2, which realizes a bilayer Hubbard model in the weak interlayer hopping limit for holes. By layer-polarizing holes via a perpendicular electric field, we observe a crossover from an excitonic insulator to a charge-transfer insulator at a hole density of ν = 1 (in units of moiré density), a transition from a paramagnetic to an antiferromagnetic charge-transfer insulator at ν = 2 and evidence for a layer-selective Mott insulator at 1 < ν < 2. The unique coupling of charge and spin to external electric and magnetic fields also manifests a giant magnetoelectric response. Our results establish a new solid-state simulator for the bilayer Hubbard model Hamiltonian.
The isQDSN for the intrinsically stretchable phototransistor array
a, Schematic of the isQDSN, which consists of QDs and semiconducting polymer fibrils (PDPP2T) in the SEBS elastomer matrix. The isQDSN is capable of effectively converting electron-hole pairs (EHP) generated by an incident light into an electrical signal during pristine and stretched modes (right). b, Schematic showing the intrinsically stretchable phototransistor array using isQDSN as a photoabsorption layer. Different bandgaps of QDs enable colour selectivity (blue, green and red). c, Photographs of the integrated 5 × 5 × 3 phototransistor array before stretching (left) and after 30% stretching (right). Scale bars, 10 mm. d, Schematic describing the procedure to enhance the accuracy of the phototransistor during mechanical deformations, using the deep learning algorithm.
Material characterization of the isQDSN film
a, Photograph of the stretched isQDSN film attached on an SEBS handling substrate (50% stretching, left). Scale bar, 20 mm. The frames on the top right-hand side show the components of the isQDSN such as the transmission electron microscopy image of red, green and blue QDs with an oleic acid ligand (top) and molecular structure of PDPP2T and SEBS (bottom). Scale bar, 15 nm. b, Cross-sectional HRTEM image of the isQDSN film. Scale bar, 25 nm. c, Linescan energy-dispersive X-ray spectral result of zinc from the ZnS shell of the QDs in the isQDSN. d, Depth XPS spectra of nitrogen (black) and zinc (red) of the isQDSN film. e, Time-resolved PL spectra of green QDs and the isQDSN. f, Normalized TA spectra of the QD, PDPP2T and isQDSN films at 0.5 ps. g, TA spectra of the isQDSN and PDPP2T films at 0.5 ps. ΔT/T, differential transmitted power of the probe with and without the excite pump. h, Schematic of the energy-band diagrams and hole movement by external illumination (hν). T, transmitted power.
Source data
Characterization of the intrinsically stretchable phototransistor
a, Schematic of the intrinsically stretchable phototransistor array. R, G and B arrays are stacked in a misaligned manner. The inset shows a photograph of the intrinsically stretchable phototransistor array. Scale bar, 10 mm. b–d, Photoresponses of the blue phototransistor (b), green phototransistor (c) and red phototransistor (d) under light illumination of 450, 525 and 630 nm wavelength, respectively (Lch = 150 μm; Wch = 1.5 mm). e, Normalized photocurrents of the red, green and blue phototransistors under periodic on/off illumination conditions before stretching (black curve) and after 30% stretching (blue, green and red curves). f, Normalized photocurrents of phototransistors with respect to increasing strains from 10% to 30%. g, Photoresponsivity (left) and photodetectivity (right) of stretchable phototransistors under different strains up to 30%. Scale bar, 10 mm. h, Photograph of the phototransistor array integrated with the deformable lens array. The inset shows the design parameters of a single lens. i, Simulation results of the light-focusing effect by the lens for light irradiation from various angles. j, Simulation result comparing the power delivered by the incident light from various angles, with and without lens integration. k, Photographs of the experimental setup for measuring photocurrents after lens array integration with lights irradiated at various incident angles. l, Comparison of normalized photocurrents for phototransistors with and without lens array integration at various incident angles.
Source data
High-density imaging demonstration using the intrinsically stretchable phototransistor array on a curved surface with deep learning algorithms
a, Schematic and photographic images of 5 × 5 × 3 phototransistor array in flat (top) and deformed states (bottom). Scale bars, 10 mm. b, Schematic showing the image sensing with a deep learning algorithm. c, Diagram of the developed artificial neural network. d, Normalized current signals of the R, G and B image patterns in flat (top) and deformed (bottom) states. e, Histograms and Gaussian fitting curves of weights connected to the fR_F (top) and fR_D (bottom) patterns with respect to the number of training epochs. f, Loss and accuracy values of training and validation with respect to the training epoch. g, Confusion matrix of the classification results. h, Corrected image patterns achieved after the application of DNN (top) from flat and deformed images, and confirmation of the incident-light colour through the colour-sensing algorithm (bottom).
Source data
High-performance photodetecting materials with intrinsic stretchability and colour sensitivity are key requirements for the development of shape-tunable phototransistor arrays. Another challenge is the proper compensation of optical aberrations and noises generated by mechanical deformation and fatigue accumulation in a shape-tunable phototransistor array. Here we report rational material design and device fabrication strategies for an intrinsically stretchable, multispectral and multiplexed 5 × 5 × 3 phototransistor array. Specifically, a unique spatial distribution of size-tuned quantum dots, blended in a semiconducting polymer within an elastomeric matrix, was formed owing to surface energy mismatch, leading to highly efficient charge transfer. Such intrinsically stretchable quantum-dot-based semiconducting nanocomposites enable the shape-tunable and colour-sensitive capabilities of the phototransistor array. We use a deep neural network algorithm for compensating optical aberrations and noises, which aids the precise detection of specific colour patterns (for example, red, green and blue patterns) both under its flat state and hemispherically curved state (radius of curvature of 18.4 mm). Intrinsically stretchable quantum-dot-based semiconducting nanocomposites enable the realization of shape-tunable and colour-sensitive phototransistor arrays.
A graphene–PCM reconfigurable silicon photonic platform
a, Schematic of the device structure. SLG, single-layer graphene; S, signal electrode; G, ground electrode. b, The layered structure of the device. c, Optical micrograph of the waveguide switch. d, False-colour SEM image of the waveguide area, where the GST is patterned. The SLG area is indicated by the black dashed lines. e, Operating principle of the device. k, extinction coefficient; t, time.
Graphene-assisted broadband waveguide switch based on GST
a, Reversible switching of GST on an SOI waveguide using a graphene heater. The switching conditions were 3 V, 100 µs pulse width and 120 µs trailing edge for SET, and 5 V, 400 ns pulse width and 8 ns trailing edge for RESET. Eight consecutive cycles were performed; the shaded area indicates the standard deviation of the cycles and the solid line indicates the average. The device spectrum is normalized to the spectrum of a bare waveguide. b, Cyclability of the switch for 1,500 switching events. The pulse conditions were the same as in a. Each pulse was temporally separated by 2 s to ensure long thermal relaxation. The transmission is normalized to the transmission of a bare waveguide. cGST, crystalline GST; aGST, amorphous GST.
Graphene-assisted phase shifter based on Sb2Se3 in a micro-ring
a, Schematic of the graphene–Sb2Se3 phase shifter in a micro-ring. Note that an additional 10 nm sputtered SiO2 layer is used to encapsulate the Sb2Se3. Such SiO2 capping is not used for GST. b, Optical micrograph of the micro-ring resonator integrated with a phase shifter. c, False-colour SEM image of the micro-ring area, where the Sb2Se3 is patterned. The graphene area is indicated by the black dashed lines. d, Reversible switching of Sb2Se3 using a graphene heater on micro-rings. The switching conditions were 4 V, 100 µs pulse width and 120 µs trailing edge for SET, and 6.8 V, 400 ns pulse width and 8 ns trailing edge for RESET. Three consecutive cycles are plotted; the shaded areas indicate the standard deviation of the cycles and the solid lines indicate the average. The spectra are normalized to the spectrum of a bare waveguide. e, Cyclability of the switch for 2,000 switching events. The switching conditions were 4 V, 100 µs pulse width and 120 µs trailing edge for SET, and 6.4 V, 400 ns pulse width and 8 ns trailing edge for RESET. Each pulse was temporally separated by 2 s to ensure long thermal relaxation. The transmission is normalized to the transmission of a bare waveguide. The data have been filtered by a 50-point moving average to reduce the fluctuation caused by thermal noise.
Quasi-continuous phase modulation using the graphene–Sb2Se3 phase shifter
a, Quasi-continuous tuning of micro-ring resonance by step amorphization. The SET conditions were 4 V, 100 µs pulse width and 120 µs trailing edge. For RESET, the amplitude was increased monotonically from 5.5 V to 6.4 V, and the pulse width and trailing edge were fixed at 400 ns and 8 ns, respectively. b, Temporal trace of a continuous programming iteration with a monotonically increasing RESET pulse amplitude from 5.5 V to 6.9 V followed by a SET pulse. Eight transmission levels are clearly resolved. The pulse width and trailing edge of the RESET pulse were fixed at 400 ns and 8 ns, respectively. The SET conditions were the same as in a. c, Change in phase shift (Δϕ) with programming energy. Fourteen phase levels can be resolved, with the phase shift increasing linearly with programming energy. Exp., experimental. d, Change in transmission caused by the variation in the phase shift with programming energy. Fourteen transmission levels can be resolved. The transmission contrast increases linearly with the programming energy, matching very well with c.
Silicon photonics is evolving from laboratory research to real-world applications with the potential to transform many technologies, including optical neural networks and quantum information processing. A key element for these applications is a reconfigurable switch operating at ultra-low programming energy—a challenging proposition for traditional thermo-optic or free carrier switches. Recent advances in non-volatile programmable silicon photonics based on phase-change materials (PCMs) provide an attractive solution to energy-efficient photonic switches with zero static power, but the programming energy density remains high (hundreds of attojoules per cubic nanometre). Here we demonstrate a non-volatile electrically reconfigurable silicon photonic platform leveraging a monolayer graphene heater with high energy efficiency and endurance. In particular, we show a broadband switch based on the technologically mature PCM Ge2Sb2Te5 and a phase shifter employing the emerging low-loss PCM Sb2Se3. The graphene-assisted photonic switches exhibited an endurance of over 1,000 cycles and a programming energy density of 8.7 ± 1.4 aJ nm–3, that is, within an order of magnitude of the PCM thermodynamic switching energy limit (~1.2 aJ nm–3) and at least a 20-fold reduction in switching energy compared with the state of the art. Our work shows that graphene is a reliable and energy-efficient heater compatible with dielectric platforms, including Si3N4, for technologically relevant non-volatile programmable silicon photonics. A non-volatile silicon photonics switch based on phase-change materials actuated by graphene heaters shows a switching energy density that is within an order of magnitude of the fundamental thermodynamic limit.
Magnetic skyrmions are compact chiral spin textures that exhibit a rich variety of topological phenomena and hold potential for the development of high-density memory devices and novel computing schemes driven by spin currents. Here, we demonstrate the room-temperature interfacial stabilization and current-driven control of skyrmion bubbles in the ferrimagnetic insulator Tm3Fe5O12 coupled to Pt, showing the current-induced motion of individual skyrmion bubbles. The ferrimagnetic order of the crystal together with the interplay of spin–orbit torques and pinning determine the skyrmion dynamics in Tm3Fe5O12 and result in a strong skyrmion Hall effect characterized by a negative deflection angle and hopping motion. Further, we show that the velocity and depinning threshold of the skyrmion bubbles can be modified by exchange coupling Tm3Fe5O12 to an in-plane magnetized Y3Fe5O12 layer, which distorts the spin texture of the skyrmions and leads to directional-dependent rectification of their dynamics. This effect, which is equivalent to a magnetic ratchet, is exploited to control the skyrmion flow in a racetrack-like device. Magnetic skyrmions are topological spin textures that hold potential for the development of post-von Neumann computing schemes. In coupled ferrimagnetic insulators, pinning effects and intentional distortions can lead to a ratchet-like current-driven motion of skyrmion bubbles.
Device structure, transport, magnetic properties and SDE
a, Schematic of the SDE and measurement configuration. The magnetic field is applied perpendicular to both the polar axis and the electrical current. b, Photomicrograph of the device. The wires of the [Nb/V/Co/V/Ta]20 multilayers were electrically connected to the Au (100 nm)/Ti (5.0 nm) metal electrodes. V, voltage. The red scale bar corresponds to 50 µm. c, Temperature (T) dependence of the device resistivity (ρ) at an electrical current density (0.455 kA cm–2) in the multilayers under a magnetic field in the range of 0 to 0.5 T. d, Magnetization loops at several temperatures for the multilayers above and below Tc. e, Magnetic field dependence of the resistivity in the multilayers for several electrical current densities at 1.9 K. The direction of the magnetic field was perpendicular to the polar axis and the electrical current.
Magnetic field and temperature dependences of SDE
a, Current density (J) dependence of the resistivity under various magnetic fields for positive and negative currents at 1.9 K. The magnetic field was swept from +0.5 T to −0.5 T (downward sweep). b, Dependence of the resistivity on J under various magnetic fields for positive and negative currents at 1.9 K. The magnetic field was swept from −0.5 T to +0.5 T (upward sweep). c, Jc as a function of the magnetic field at 1.9 K. First, the magnetic hysteresis of Jc was measured from +0.5 to −0.5 T (downward sweep) for positive (red) and negative (black) currents. Subsequently, the sweep direction was reversed, and the magnetic field was swept from −0.5 to +0.5 T (upward sweep) for positive (blue) and negative (green) currents. The arrows indicate the sweep directions. d, Non-reciprocal component of the critical current density, ΔJc, as a function of the magnetic field at various temperatures. The magnetic hysteresis of ΔJc was measured from +0.5 to −0.5 T (downward sweep, black). Subsequently, the sweep direction was reversed, and the magnetic field was swept from −0.5 to +0.5 T (upward sweep, red). The arrows indicate the sweep direction.
Field-free SDE controlled by magnetization
a, Magnetic field dependences of Jc for positive and negative currents in minor hysteresis loops at 1.9 K. Magnetic field was swept in the order of +0.5, 0, −0.15 T (downward sweep) and then −0.15, 0, +0.15 T (upward sweep). b, Magnetic field dependence of ΔJc in the minor hysteresis loops obtained by Fig. 3a. c, Magnetic field dependences of Jc for positive and negative currents in minor hysteresis loops at 1.9 K. Magnetic field was swept in the order of −0.5, 0, +0.15 T (upward sweep), and then +0.15, 0, −0.15 T (downward sweep). d, Magnetic field dependence of ΔJc in the minor hysteresis loops obtained by Fig. 3c. e, Repeated application of current densities J = 72.7 kA cm–2 and J = −72.7 kA cm–2 at 1.9 K without a magnetic field. f, Non-volatile SDE at 1.9 K. Red and black dots represent the results for negative magnetization (−M) and positive magnetization (+M), respectively. The device shows a superconducting state or normal conducting state depending on the polarity of the current. Note that the polarity of the SDE depends on the direction of magnetization. The −M or +M state is achieved after sweeping the magnetic field in the order of +0.5, 0, −0.15, 0 T or −0.5, 0, +0.15, 0 T, respectively.
Band structure of a bulk Nb/V/Co/V/Ta superlattice
a, Band structure obtained from first-principles calculation for paramagnetic state. b, Partial density of states for paramagnetic state. c, Band structure for ferromagnetic state. d, Partial density of states for ferromagnetic state. The V layer between Nb and Co is labeled with V1 and that between Co and Ta is labeled with V2.
The diode effect is fundamental to electronic devices and is widely used in rectifiers and a.c.–d.c. converters. At low temperatures, however, conventional semiconductor diodes possess a high resistivity, which yields energy loss and heating during operation. The superconducting diode effect (SDE)1–8, which relies on broken inversion symmetry in a superconductor, may mitigate this obstacle: in one direction, a zero-resistance supercurrent can flow through the diode, but for the opposite direction of current flow, the device enters the normal state with ohmic resistance. The application of a magnetic field can induce SDE in Nb/V/Ta superlattices with a polar structure1,2, in superconducting devices with asymmetric patterning of pinning centres⁹ or in superconductor/ferromagnet hybrid devices with induced vortices10,11. The need for an external magnetic field limits their practical application. Recently, a field-free SDE was observed in a NbSe2/Nb3Br8/NbSe2 junction; it originates from asymmetric Josephson tunnelling that is induced by the Nb3Br8 barrier and the associated NbSe2/Nb3Br8 interfaces¹². Here, we present another implementation of zero-field SDE using noncentrosymmetric [Nb/V/Co/V/Ta]20 multilayers. The magnetic layers provide the necessary symmetry breaking, and we can tune the SDE by adjusting the structural parameters, such as the constituent elements, film thickness, stacking order and number of repetitions. We control the polarity of the SDE through the magnetization direction of the ferromagnetic layers. Artificially stacked structures13–18, such as the one used in this work, are of particular interest as they are compatible with microfabrication techniques and can be integrated with devices such as Josephson junctions19–22. Energy-loss-free SDEs as presented in this work may therefore enable novel non-volatile memories and logic circuits with ultralow power consumption.
SENT-seq uses orthogonal capture sequences to generate tunable multiomic readouts
a, The sensitivity of the DNA barcode readouts relative to the biological (that is, mRNA and protein) readouts was controlled by the ratio of two orthogonal capture sequences: the barcode capture sequence and poly-T, which captured mRNA and poly-A tagged cell hash oligonucleotide antibodies. b,c, Beads carrying the barcode capture sequences and poly-T were mixed with the complementary fluorescent barcode probe, fluorescent poly-A probe, both or, as a negative control, none (b), and the MFI of each coupling was quantified (c). d,e, Beads carrying both capture sequences were mixed with varying amounts of LNP barcode or mRNA (d) to generate sequencing read standard curves (e). f, After formulating and injecting N chemically distinct LNPs carrying mRNA and DNA barcodes, tissues were isolated and digested into single-cell suspensions. The workflow shows delivery mediated by all N LNPs, subsequent mRNA-mediated protein production and transcriptome quantification in single cells using next-generation sequencing. BC, barcode; L, linker; UMI, unique molecular identifier.
In vivo multiomic single-cell readouts of the transcriptome, functional LNP-mediated mRNA delivery and LNP-mediated DNA barcode delivery
a, The t-SNE plot of live cells sorted from murine liver. b, aVHH protein expression in the same cells, overlaid on the t-SNE plot, after administration of LNPs carrying mRNA encoding aVHH. The small grey dots represent cells with less than four aVHH reads per cell. c, The most common barcode delivered by each of the 16, out of 24, chemically distinct LNPs, overlaid on the t-SNE plot. The 16 LNPs shown met our initial inclusion criteria and showed up in our sequenced single-cell dataset. The small dark grey dots represent cells with no barcode delivery. The asterisk represents the naked barcode.
Cell subsets differentially uptake LNPs
a,b, Normalized barcode distribution (a) and aVHH expression (b) profiles for endothelial cells, along with violin plots representing the spread of the profiles. c,d, The normalized barcode distribution profiles and violin plots for Kupffer cells (c) and hepatocytes (d). Cell types with narrow distributions are characterized by narrow unimodal peaks of low normalized barcode delivery. Cell types with wide distributions are characterized by wide peaks or bimodal peaks of low and high normalized barcode delivery. In all cases, n = 4 mice per group were used; data are plotted as mean ± standard error of the mean. Histograms in a–d are shown with n = 4 dots per bar, with each dot representing a different mouse. Violin plots in a–d show the median (solid black line) and quartiles (dotted black lines).
Endothelial cell subtypes show transcriptional differences that may dictate LNP-mediated mRNA delivery
a, Schematic of liver vessel morphology. b, Dot map showing the expression levels of 16 genes in hepatic endothelial cell differentiation. cap. venous, capillary venous; exp., expression. c,d, Volcano plots of differentially expressed genes in EC1 as compared with in EC3 (c) and in EC2 as compared with in EC3 (d). FC, fold change; FDR, false discovery rate. Red colour represents upregulated genes. Blue colour represents downregulated genes. e, Differential analysis workflow for EC clusters to identify genes. f, Venn diagram of differentially expressed genes found in EC1 and EC2, compared with in EC3, after separation based on aVHH expression. g, STRING analysis of the 19 differentially expressed genes found in aVHH⁺ cells in EC clusters 1 and 2 yielded 11 genes with statistically significant interactions. h, Dot map of the expression levels of differentially expressed genes in EC1, EC2 and EC3 with significant interactions.
Chemically distinct LNPs exhibit different tropism within the liver microenvironment
a,b, Each LNP was formulated to contain a distinct DNA barcode, which we were able to map onto single cells. LNP barcode counts are represented in each cell cluster as either the average of barcode counts for all single cells within a cluster (a) or the sum of barcode counts for all single cells within a cluster (b). The three negative control naked barcodes are indicated by asterisks. c–f, The distribution and normalized barcode expression of LNP-3 (c), LNP-7 (d), LNP-10 (e) and LNP-12 (f) identified on the basis of their DNA barcode, overlaid on a t-SNE plot of 17 distinct cell subsets, shown alongside each LNPs' composition. 20a-OH, 20a-hydroxycholesterol. g–j, The aVHH expression profiles for LNP-3 (g), LNP-7 (h), LNP-10 (i) and LNP-12 (j), shown alongside a distribution of the cell populations that are delivered to. k–n, The aVHH/barcode ratio for LNP-3 (k), LNP-7 (l), LNP-10 (m) and LNP-12 (n) in all single cells where those LNPs are found.
Cells that were previously described as homogeneous are composed of subsets with distinct transcriptional states. However, it remains unclear whether this cell heterogeneity influences the efficiency with which lipid nanoparticles (LNPs) deliver messenger RNA therapies in vivo. To test the hypothesis that cell heterogeneity influences LNP-mediated mRNA delivery, we report here a new multiomic nanoparticle delivery system called single-cell nanoparticle targeting-sequencing (SENT-seq). SENT-seq quantifies how dozens of LNPs deliver DNA barcodes and mRNA into cells, the subsequent protein production and the transcriptome, with single-cell resolution. Using SENT-seq, we have identified cell subtypes that exhibit particularly high or low LNP uptake as well as genes associated with those subtypes. The data suggest that cell subsets have distinct responses to LNPs that may affect mRNA therapies. Cell heterogeneity might impact the delivery of lipid nanoparticles (LNPs) and efficacy of messenger RNA-based therapies in vivo. Here, the authors propose an approach to measure how various LNPs deliver DNA barcodes and mRNA to cells using single-cell RNA sequencing, providing a correlation between LNP uptake and the expression of specific genes that characterize cellular subtypes.
Illustration of Z-BP measurement modality
a, Three-dimensional schematic of the GETs placed onto the participant’s wrist over the radial artery, with two outer tattoos used for a.c. injection, and two inner tattoos used to measure voltage changes. b, Photograph of 12 GETs, each with a surface area of 25 mm², placed on the radial (six tattoos, comprising Bio-Z1 and Bio-Z2) and ulnar (six tattoos, comprising Bio-Z3 and Bio-Z4) arteries on a participant’s wrist. This multiplexed tattoo placement is essential for effective BP capture. The arteries are pseudo-coloured in pink for visibility, and their locations were tracked using an ultrasound vascular Doppler probe. c, Close-up view of six GETs at the radial artery. The injecting GETs are pseudo-coloured in green, the Bio-Z1 pair in violet and the Bio-Z2 pair in blue, as the graphene is almost invisible. d, Cross-section of the six GETs, with green lines representing the a.c. injected signal and grey lines representing voltage sensing. e, Close-up view of one pair of sensing GETs and the simplified equivalent electrical circuit of the interface, showing Ztissue and Zartery, part of which (ΔZartery) is related to the undulating blood volume. Credit: a, Jo Wozniak, Texas Advanced Computing Center.
Correlation between arterial BP and bioimpedance
a, Illustration of the peripheral arterial BP pulse waveform (red) and correlated arterial volume⁴⁰. The systole and diastole BP regions are highlighted in blue and yellow, respectively. b, The Bio-Z signal (violet) is reciprocal to the BP pulse waveform. c, Two Bio-Z signals recorded by two pairs of GETs are essential for calculating tpt and the interbeat interval, which are used for the machine learning algorithm. The complete machine learning regression analysis is based on four main features: the systolic and diastolic phases (upward and downward triangles), the maximum slope (rhombus) and the inflection point (circle).
Graphene Z-BP measurement results from the HGCP routine
a, Illustration of the HGCP routine that was performed by participants with their right hands. An array of 13 GETs, over radial and ulnar arteries, were placed on the participants’ left forearms. b, Scatter diagram of all DBP and SBP values from six participants, indicating the wide dynamic BP range coverage that is essential for effective BP prediction. c, Six consecutive HGCP-induced BP manoeuvres: DPB, SBP and MAP as measured via the GETs and correlated with the control BP (Finapres NOVA). The dashed areas indicate short (~10 min) breaks between the sets of HGCP manoeuvres. The graphene Z-BP time trace is closely correlated with the control BP. d, The statistical violin plots for graphene Z-BP (DBP (left) and SBP (right)) given in direct comparison with Ag wristband-derived BP (grey) from the same participant. The plot indicates that the accuracy of graphene-enabled BP monitoring is superior to that of the dry Ag wristband. The violin plot includes the box plot (defined as Q1 and Q3 quartiles, and median) with a kernel density estimation over the points. e, Comparison of the DBP estimation accuracy achieved with the GETs (Z-BP, green star) with Ag/AgCl gel-based BP³⁵, photoplethysmography (PPG)²⁵, ultrasound⁵², tonometry⁵³, capacitive sensors²⁴ and various cuff-based methods (white circles)54,55. Only the works with relevant (m.e. and s.d.) data provided were included. The background shading shows the IEEE accuracy categories (Grades A, B and C)⁴⁶.
Source data
Graphene Z-BP model training and performance evaluation
a, DBP and SBP scatter plots for three BP elevation manoeuvres of: HGCP, the Valsalva manoeuvre and cycling. HGCP is a time-consuming routine but raises BP in a usefully wide range; cycling is less effective at elevating BP and the Valsalva manoeuvre is equally effective as HGCP, with a considerably lower number of training points. b, Violin plots representing the accuracies of graphene Z-BP compared with Ag wristbands for different machine learning routines: shuffled HGCP, unshuffled HGCP, Valsalva, post-workout and after a 4 day break. The shuffled HGCP and Valsalva based models yielded the lowest median error of BP estimation. The violin plots include box plots (defined as Q1 and Q3 quartiles, and median) with a kernel density estimation over the points. The corresponding accuracy categories are labelled. c, A time trace of BP changes during three consecutive Valsalva manoeuvres, measured via Finapres NOVA, smoothed with an average window of 20 heartbeats as necessary for accurate machine learning training and as predicted via graphene-enabled Z-BP recordings. Each Valsalva cycle has two peaks and one dip in BP. However, the smoothing algorithm results in waveform alteration and a less pronounced dip. d, Time trace of DBP, SBP and MAP changes for one participant during a series of HGCP exercises as measured by GETs compared with the control BP, followed by an hour-long break including a workout and the model validation session with another HGCP pattern.
Source data
Continuous monitoring of arterial blood pressure (BP) in non-clinical (ambulatory) settings is essential for understanding numerous health conditions, including cardiovascular diseases. Besides their importance in medical diagnosis, ambulatory BP monitoring platforms can advance disease correlation with individual behaviour, daily habits and lifestyle, potentially enabling analysis of root causes, prognosis and disease prevention. Although conventional ambulatory BP devices exist, they are uncomfortable, bulky and intrusive. Here we introduce a wearable continuous BP monitoring platform that is based on electrical bioimpedance and leverages atomically thin, self-adhesive, lightweight and unobtrusive graphene electronic tattoos as human bioelectronic interfaces. The graphene electronic tattoos are used to monitor arterial BP for >300 min, a period tenfold longer than reported in previous studies. The BP is recorded continuously and non-invasively, with an accuracy of 0.2 ± 4.5 mm Hg for diastolic pressures and 0.2 ± 5.8 mm Hg for systolic pressures, a performance equivalent to Grade A classification. Self-adhesive bioimpedance graphene electronic tattoos enable accurate continuous blood pressure monitoring.
NAD(H)-loaded NPs replenished cellular NAD(H) pool and prevented inflammation-induced energy depletion
a, Schematic illustration of NAD⁺ metabolism and cellular uptake. Since NAD⁺ cannot pass through the cell membrane directly, it has to be degraded by extracellular enzymes into several precursors (for example, NAM and NR) which can enter the cells and subsequently enhance the NAD⁺ biosynthesis. Such a conversion process is inefficient as it is regulated and limited by critical enzymes (for example, NAMPT). The NAD(H)-loaded NPs can be taken up by the cells via endocytosis and directly replenish cellular NAD⁺. The CaP or MOF cores can dissolve in acidic endosome, leading to endosome swelling and bursting (due to an increase in osmotic pressure) to release the entrapped payload into cytosol. Created with NMN, nicotinamide mononucleotide; NMNAT, nicotinamide mononucleotide adenylyltransferase; NRK, nicotinamide riboside kinase. b,c, Size and morphology of NAD⁺-LP-CaP (b) and NADH-LP-MOF (c) characterized by dynamic light scattering and transmission electron microscopy (inset). Representative data of three independent experiments. Scale bars, 200 nm. d,e, The NAD(H) release profiles from NAD⁺-LP-CaP (d) and NADH-LP-MOF (e) under different pH values. Data are presented as mean ± s.d. (n = 3). f,g, Intracellular NAD(H) levels (f) and NAD⁺/NAD(H) ratio (g) in BMDMs incubated with free NAD(H) (10 µM) or an equivalent dose of the NPs. Data are presented as mean ± s.d. (n = 5). Statistical significance was calculated via one-way analysis of variance (ANOVA) with Tukey’s post hoc test. h,i, Quantification of intracellular ATP level (n = 5; h) and cell viability (n = 6; i) in an LPS-mediated energy depletion model. LPS-stimulated BMDMs (LPS, 100 ng ml–1) were treated with free NAD(H) (10 µM) or an equivalent dose of the NPs. Data are presented as mean ± s.d. Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test.
NAD(H)-loaded NPs prevented inflammation-induced cell death
a,b, Pro-inflammatory cytokine TNF-α (a) and IL-6 (b) levels in BMDM culture supernatant quantified by ELISA. LPS-stimulated BMDMs (LPS, 100 ng ml–1) were treated with free NAD(H) (10 µM) or an equivalent dose of the NPs. Data are presented as mean ± s.d. (n = 3). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. Statistical analyses were done relative to the LPS treatment group. c–f, Analysis of caspase1 activation (demonstrated by fluorochrome-labelled inhibitors of caspases (FLICA) probe 5-carboxyfluorescein-Tyr-Val-Ala-Asp-fluoromethylketone (FAM-YVAD-FMK) staining; n = 5; c and d) and IL-1β release (n = 3; e and f) indicating the activation of canonical (c and e) and non-canonical (d and f) inflammasome pathways. BMDMs were primed with LPS (100 ng ml–1, 3 h) as signal 1, and then either incubated with ATP (2.5 mM, 1 h; for the canonical pathway) or transfected with lipoLPS (100 ng LPS per well, 3 h; for the non-canonical pathway) as signal 2. Free NAD⁺, empty NPs (denoted as eCaP) or the NAD⁺-LP-CaP NPs were added together with signal 1 or signal 2 or both. Data are presented as mean ± s.d. Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. Statistical analyses were done relative to the positive controls (LPS/ATP or LPS/lipoLPS). g,h, NF-κB p65 nuclear translocation observed by CLSM. BMDMs were pretreated with free NAD(H) or the NPs for 5 h, stimulated with LPS (100 ng ml–1) for 1 h and then immunostained for p65. Representative images of five independent experiments. Scale bars, 20 μm. Data are presented as mean ± s.d. (n = 5). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. Statistical analyses were done relative to the LPS treatment group. i,j, BMDM apoptosis triggered by LPS (100 ng ml–1, 48 h) analysed by annexin V and propidium iodide double staining. FITC, fluorescein isothiocyanate. k, HUVEC apoptosis triggered by TNF-α (80 ng ml–1, 48 h). The cells were treated with free NAD(H) or the NPs at a corresponding dose of 10 μM. l, Fluorescence images of HUVEC monolayer stained for tight junction protein VE-cadherin (green) after incubation with LPS together with free NAD(H) or the NPs for 24 h. Cell nuclei were stained by 4′,6-diamidino-2-phenylindole (DAPI; blue). Representative images of three independent experiments. Scale bar, 50 μm.
The therapeutic efficacy of the NPs in a mouse model of endotoxemia
a, Experimental procedures for the endotoxemia mouse model. Created with b–d, Survival (b and c) and body weight (d) analysis of the mice receiving phosphate-buffered saline (PBS), free NAD(H) (20 mg kg–1) or an equivalent dose of empty NPs and NAD(H)-loaded NPs, with treatment 1 h after LPS (15 mg kg–1) administration. Data are presented as mean ± s.d. (n = 10). Statistical significance was calculated via log-rank test. e,f, Pro-inflammatory cytokine TNF-α and IL-1β levels in the serum of septic mice receiving different treatments were measured 6 h after LPS challenge. Data are presented as mean ± s.d. (n = 6). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. g, Gene expression in the white blood cells of the mice receiving different treatments over negative controls. The colours in the heatmap present the log10 value of relative gene expression. h, Ex vivo fluorescence images representing the biodistribution of Cy5.5-labelled LP-CaP NPs in healthy and septic mice 4 h after NP administration. L, K, S, Lu, H and M represent liver, kidneys, spleen, lungs, heart and thigh muscle, respectively. a.u., arbitrary units. i, Quantitative analysis of the mean fluorescence intensity of the organ or tissue shown in the ex vivo image. Data are presented as mean ± s.d. (n = 3). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. j, The vascular hyperpermeability in LPS-stimulated mice receiving different treatments. Evans blue dye was injected 12 h after LPS challenge, and the amount of the dye retained in the lungs was extracted and quantified. Data are presented as mean ± s.d. (n = 6). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test.
Immune cell population variation, apoptosis and caspase1 activation in the blood, lungs and spleen of endotoxemia mice
a,b, Flow cytometric quantification of monocyte (CD45⁺CD11b⁺Ly6C⁺Ly6G–), neutrophil (CD45⁺CD11b⁺Ly6C⁺Ly6G⁺), CD4⁺ T cell (CD45⁺CD11blowCD3⁺CD4⁺) and CD8⁺ T cell (CD45⁺CD11blowCD3⁺CD8⁺) populations in blood of septic mice 24 h after LPS administration. PBS, free NAD⁺ (20 mg kg–1) or an equivalent dose of LP-CaP and NAD⁺-LP-CaP were i.v. injected 1 h after LPS (7.5 mg kg–1, i.v.) administration. Healthy mice without LPS injection were used as the negative control. Data are presented as mean ± s.d. (n = 5). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. c, Flow cytometric quantification of monocyte and neutrophil populations in the lungs of the septic mice with PBS, free NAD⁺, LP-CaP or NAD⁺-LP-CaP treatment. Data are presented as mean ± s.d. (n = 5). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. d, Representative plots of cell apoptosis for splenic lymphocytes. The grayscale figures on the left show the gating strategy for the CD4⁺ and CD8⁺ T cells. e,f, Quantification of cell apoptosis (e) and caspase1 activation (f) for a variety of immune cells (monocyte and neutrophil in blood, and CD4⁺ and CD8⁺ T cell in spleen) assessed by annexin V/7AAD staining and FLICA assay, respectively. Data are presented as mean ± s.d. (n = 5). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test.
Therapeutic performance of the NAD⁺-loaded NPs in bacteria-induced sepsis models
a, Experimental procedures for the CLP and P. aeruginosa secondary infection model. Mice subjected to CLP received two i.v. injections of PBS, free NAD⁺ (20 mg kg–1) or an equivalent dose of LP-CaP and NAD⁺-LP-CaP at 6 h and 24 h after the surgery and challenged intratracheally with P. aeruginosa (1 × 10⁸ CFU in 50 µl PBS) at day 3. A sham group without CLP and a CLP group without the P. aeruginosa challenge were used as control groups. Created with b,c, Survival (b) and body weight (c) analysis of the mice in the bacteria secondary infection model. Data are presented as mean ± s.d. (n = 14). Statistical significance was calculated via log-rank test. d, Experimental procedures for the model of a polymicrobial blood infection induced by MRSA and P. aeruginosa. Mixed multidrug-resistant bacteria (mixed MRSA and P. aeruginosa with 5 × 10⁷ CFU for each pathogen) were i.v. administrated to induce blood infection and sepsis, and one injection of different treatments, including PBS, NAD⁺-LP-CaP, free Rif (acting as a model antibiotic), Rif-LP-CaP or NAD⁺-Rif-LP-CaP at a dose corresponding to 10 mg Rif per kilogram and 20 mg NAD⁺ per kilogram were i.v. injected 6 h after the infection. Created with e,f, Survival (e) and body weight (f) analysis of the mice in the blood infection model. Data are presented as mean ± s.d. (n = 10). Statistical significance was calculated via log-rank test. g, Bacterial loads in liver, spleen, lungs, kidneys and blood of the septic mice 12 h after the treatments, determined by serial dilution and plate counting. Data are presented as mean ± s.d. (n = 6). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. Statistical analyses were done relative to the PBS treatment group. h, Blood biochemistry analysis demonstrating the liver (alanine transaminase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP)) and kidney (blood urea nitrogen (BUN)) function of the mice with a bacterial blood infection. Data are presented as mean ± s.d. (n = 3). Statistical significance was calculated via one-way ANOVA with Tukey’s post hoc test. Statistical analyses were done relative to the PBS treatment group. i, Representative histological images for tissue sections with haematoxylin and eosin staining and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining, from the healthy mice (control) and the infected mice with PBS or NAD⁺-Rif-LP-CaP treatments. OSOM, outer stripe of outer medulla; ISOM, inner stripe of outer medulla. Representative images of three independent experiments. Scale bar, 100 μm.
Sepsis is a life-threatening organ dysfunction responsible for nearly 270,000 deaths annually in the United States alone. Nicotinamide adenine dinucleotide (NAD+), an immunomodulator, can potentially treat sepsis; however, clinical application of NAD+ is hindered by its inability to be directly taken up by cells. To address this challenge, a family of nanoparticles (NPs) loaded with either NAD+ or the reduced form of NAD+ (NADH), hereafter NAD(H)-loaded NPs, were engineered to enable direct cellular transport and replenishment of NAD(H). The NAD(H)-loaded NPs improved cellular energy supply, suppressed inflammation and prevented inflammation-induced cell pyroptosis and apoptosis. Therefore, the NPs can help maintain immune homoeostasis and vascular function, two key factors in the pathogenesis of sepsis. The NAD(H)-loaded NPs demonstrated excellent therapeutic efficacies in treating endotoxemia and multidrug-resistant pathogen-induced bacteremia. In addition, the NAD(H)-loaded NPs prevented caecal ligation and puncture-induced multiorgan injury and improved outcomes of secondary Pseudomonas aeruginosa infections following caecal ligation and puncture, thus potentially leading to a highly innovative and translational approach to treat sepsis efficiently and safely. Nicotinamide adenine dinucleotide (NAD+) is an immune modulator that was suggested as a potential treatment for sepsis, but its in vivo benefits are contradictory and its low bioavailability as a free drug hampers potential clinical translation. Here the authors show that using a lipid-coated nanoparticle to deliver NAD+ to the cell cytosol can effectively replenish the intracellular content of NAD+ and reduce the extent of the inflammatory response in mouse models of sepsis.
Nanocomplex-decorated microbubbles targeting CD11b on APCs
a, ncMBs are obtained by conjugating MBs with anti-CD11b antibodies and SpeDex and loading negatively charged cGAMP. b, On binding of ncMBs to APCs and under US exposure, cGAMP is delivered directly into the cytosol of the APCs by sonoporation to activate STING and downstream antitumour immunity, a process termed MUSIC. Credit a,b, Erin E. Moore. c, The lipid shell of the MB is partly composed of DSPE-PEG-maleimide, which was conjugated with thiolated SpeDex through a thiol–maleimide coupling reaction. d, Coulter Counter measurements of SpeDex-anti-CD11b MBs show a size distribution of 1–10 μm, with a mean size of 2.6 μm. e,f, A fluorescent analogue of cGAMP (DY547-c-diGMP) was used to verify binding to cMBs (forming ncMBs). Flow cytometry (e) and fluorescence microscopy (f) confirmed binding of the fluorescent analogue to all ncMBs. Scale bar, 50 μm. g, DiD-labelled cMBs were added to EO771 murine breast cancer cells and THP-1 human macrophages to confirm CD11b-specific targeting of cMBs. Confocal microscopy confirmed the binding of cMBs to THP-1 cells. h, Fluorescence microscopy of BMDMs after sonoporation with DY547-c-diGMP-loaded ncMBs indicates cytosolic delivery of the cyclic dinucleotide into all cells. g,h, Scale bar, 100 μm. i, Intensity quantification of DY547-c-diGMP uptake in BMDMs. The data represent mean ± s.d. with n = 3 biologically independent samples. All data are shown as representative from at least three independent experiments (d–i), and were analysed by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test (i).
Sonoporation of APCs targeted by ncMBs (MUSIC) induces activation of STING-IRF signalling in vitro
a,b, BMDMs from C57BL/6J mice were treated with MUSIC, cGAMP or cMBs (+US), respectively. After treatment, the cells were cultured for the indicated periods, followed by western blotting of proteins in the STING-IRF3 pathway. a, The strongest STING-IRF3 activation was induced 6 h after MUSIC treatment. b, STING knockout abrogated MUSIC induced STING-IRF3 activation. c, Confocal fluorescence microscopy images of nuclear translocation of activated IRF3 (phosphorylated IRF3 or pIRF3) in BMDMs at 6 h after treatment. d,e, Western blotting of NF-κB pathway signalling was performed using the same cells treated in a and b. The red box in a,b,d,e refers specifically to MUSIC treatment. f, Confocal fluorescence microscopy images of the nuclear translocation of activated NF-κB (p65) in BMDM. c,f, Scale bar, 20 μm. Right panels indicate quantification of nuclear fluorescent-positive cells done by randomly measuring 500 cells in each group. g, ifna1 and ifnb1 mRNA expression levels in mouse BMDMs treated as above were determined via real-time PCR. h, IFN-α and IFN-β cytokine released in cell culture supernatants from above were measured by ELISA. g,h, n = 4 repeats in each group. i,j, CD8⁺ T cells from OT-I mice and CD4⁺ T cells from OT-II mice were stained with Far Red. T cell proliferation was measured by flow cytometry after co-culture with MUSIC-activated BMDMs with pulsed OVA peptides for 72 h, respectively. Right panels indicate the quantification of proliferated cells as gated. i, n = 6; j, n = 3. All data are representative from at least three biologically independent experiments. c,f–j, data are presented as mean ± s.d.; statistical significance was calculated by one-way ANOVA with Tukey’s multiple comparisons test.
Source data
MUSIC activates STING signalling and T cell response in primary breast cancer in vivo
a, Contrast-mode US images of EO771 breast tumours (13 days) in C57BL/6J mice before ncMBs injection (left, non-treatment), after ncMBs injection (middle, pre-sonoporation) and after US sonoporation (right, post-sonoporation). Loss in signal represents bubbles being destroyed after exposure to US. Images are from the same mouse and are representative of five randomly treated wild-type (WT) mice. b–d, WT and STING−/− mice were inoculated with EO771 breast tumours, and treated with MUSIC, cGAMP or cMBs (+US) following the strategy in Supplementary Fig. 17a. b, At 18 days post-tumour inoculation, immunostaining by confocal microscopy visualized recruited CD11b⁺ cells and pSTING⁺ cells in tumour paraffin section slides. Representative images from random fields of view in one of the three biologically independent mice. Scale bar, 50 μm. c, Fluorescence intensity measurements and comparison by ImageJ software from three randomly selected images of three biologically independent mice, analysed by one-way ANOVA with Tukey’s multiple comparisons test. d, Flow cytometry analysis and quantification of CD8⁺ T or CD4⁺ T cells in tumours of representative mice at day 18 in each group. Data are representative from three biological independent samples (d) and are shown as mean ± s.d. (c), analysed by one-way ANOVA with Tukey’s multiple comparisons test (c).
MUSIC activates STING-mediated antitumour immunity
a–g, WT and STING−/− mice were inoculated with EO771 breast tumours, and treated with MUSIC, cGAMP or cMBs (+US) following the strategy in Supplementary Fig. 17a. a, Representative photographs of mice at 24 days post-tumour inoculation. b,c,e, Tumour volumes were monitored and analysed over the indicated periods. d,f, Survival curves for the mice in the different treatment groups. g, The six living tumour-free mice from the above MUSIC-treated group (b,d) were rechallenged with EO771 cells. Tumour volumes were measured over the following 28 days. n = 10 for all WT mice, n = 7 (PBS) or 8 (MUSIC) for STING−/⁻ mice (a,b,d–f), n = 6 for ncMBs and n = 7 for both MUSIC and IgG-ncMBs (+US) (c). a–g, n means biologically independent animals. h,i, Splenic T cells from mice treated 18 days post-tumour inoculation were assessed by ELISPOT to further verify immune memory enhancement upon MUSIC treatment. n = 3 biologically independent samples. j, IFN-γ and PD-L1 protein expression levels were detected by immunostaining in tumour paraffin section slides. Representative images from random fields of view from one of three biologically independent animals. Scale bar, 50 μm. k, Quantification and comparison of fluorescence intensity using ImageJ software from three randomly selected images of three biologically independent mice. The data represent mean ± s.e.m. (b,c,e,g) or mean ± s.d. (i,k), analysed by two-sided log-rank (Mantel–Cox) test (d,f), or one-way ANOVA with Tukey’s multiple comparisons test (b,c,e,g,i,k). *P and #P in b,d denote the statistical significance relative to the MUSIC group, respectively.
MUSIC activates systemic antitumour immunity to inhibit breast cancer metastasis
a–f, BALB/cJ mice were inoculated with luciferase-expressing 4T1 (Luc-4T1) tumours and treated as indicated in Supplementary Fig. 26a. n = 7 for PBS, cGAMP and cMBs (+US); n = 8 for MUSIC, aPD-1 and aPD-1+MUSIC. Representative IVIS spectrum images (a) and quantified signal intensity (b) showing metastases of Luc-4T1 breast tumours at 28 days. c, Survival curves were plotted and analysed for the mice in each treatment arm. d, Representative photographs and bioluminescence images of organs ex vivo after euthanasia show the presence of tumour metastases. e, Quantification of tumour nodules in lungs from d. f, Quantified bioluminescence signals in lungs from d. g, Accumulation of IFN-α or IFN-β in tumours (determined using ELISA) at 21 days post-tumour inoculation. h–k, Flow cytometry analysis and quantification of CD8⁺ T (h,i) or CD4⁺ T (j,k) cells in tumours at day 21 in each group. i,k, Proportions of CD8⁺ or CD4⁺ T cells. l–o, Levels of activated (phosphorylated) STING and IFN-γ protein in tumour paraffin section slides detected by immunostaining (l,n). Images were collected in a blinded fashion and represent three biologically independent experiments. Scale bar, 20 μm. m,o, Quantification and comparisons of fluorescence intensity using ImageJ software from three randomly selected images. p,q, Flow cytometry of tumour lymphocytes demonstrate that 4T1 tumour-bearing mice treated with aPD-1+MUSIC experienced a shift from naïve to memory CD8⁺ T cells at 21 days post-tumour inoculation (p). Proportions of CD62L+/− or CD44+/− T cells (q). The data represent mean ± s.e.m. (b) or mean ± s.d. (e–g,i,k,m,o,q) from representative experiments of three independent experiments with n = 3 (g,i,k,m,o,q), analysed by two-sided log-rank (Mantel–Cox) test (c), or one-way ANOVA with Tukey’s multiple comparisons test (b,e,g,i,k,m,o,q) or one-way ANOVA with Dunnett’s multiple comparisons test, PBS group as control (f). P, *P and #P in c denote the statistical significance relative to the MUSIC versus PBS group, aPD-1+MUSIC versus MUSIC group or aPD-1+MUSIC versus aPD-1 group, respectively.
The cytosolic innate immune sensor cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) pathway is crucial for priming adaptive antitumour immunity through antigen-presenting cells (APCs). Natural agonists, such as cyclic dinucleotides (CDNs), activate the cGAS-STING pathway, but their clinical translation is impeded by poor cytosolic entry and serum stability, low specificity and rapid tissue clearance. Here we developed an ultrasound (US)-guided cancer immunotherapy platform using nanocomplexes composed of 2′3′-cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) electrostatically bound to biocompatible branched cationic biopolymers that are conjugated onto APC-targeting microbubbles (MBs). The nanocomplex-conjugated MBs engaged with APCs and efficiently delivered cGAMP into the cytosol via sonoporation, resulting in activation of cGAS-STING and downstream proinflammatory pathways that efficiently prime antigen-specific T cells. This bridging of innate and adaptive immunity inhibited tumour growth in both localized and metastatic murine cancer models. Our findings demonstrate that targeted local activation of STING in APCs under spatiotemporal US stimulation results in systemic antitumour immunity and improves the therapeutic efficacy of checkpoint blockade, thus paving the way towards novel image-guided strategies for targeted immunotherapy of cancer. Activation of the STING pathway in antigen-presenting cells has been proposed as a strategy to stimulate the adaptive immune response against tumours, but its clinical application is hampered by the instability, low specificity and low cytosolic entry of natural STING agonists. Here the authors present a platform for targeted ultrasound-mediated cytosolic delivery of STING agonists that shows efficacy in different animal tumour models and improves the response to checkpoint blockade therapies.
Diffusion process of water vapour in nanoporous carbon
a, AWH through microporous sorbents includes a series of nanoscale mass transport phenomena: bulk diffusion (i), surface diffusion (ii), external permeation (iii), internal diffusion (iv) and water release (v). The grey spheres denote carbon; red spheres, oxygen; white spheres, hydrogen; and yellow spheres, nitrogen. b–d, Energy barrier (in eV) for the diffusion processes of H2O in a slit-shaped carbon pore model with various geometric and chemical properties. Based on this model, it is expected that nanoporous carbon with 40% density of adsorption sites and ~1.0 nm pore size has the best sorption dynamics due to the minimized overall diffusion resistance.
Structural and compositional analyses of MOF-derived nanoporous carbon via steam selective etching
a, Synthesis scheme of nanoporous carbon derived from a copper halide MOF ([Cu(4,4′-bipy)2Cl2]n). b, After steam selective etching, the porous products are analysed by XPS. XPS wide-scan surveys confirm that steam etching can widen the pore and simultaneously protect the adsorption sites. c–e, High-resolution XPS data of C1s (c), N1s (d) and O1s (e) for the Steam-80 sample. The presence of graphitic N and phenolic O implies that the synthesized nanoporous carbon is thermally stable and thus particularly suited to capture water vapour. f, Pore-size distributions of MOF-derived nanoporous carbon samples. g, TEM image of Steam-80. There are abundant micropores in the obtained products. Scale bar, 20 nm. h, Left axis, the temperature profile of water-saturated Steam-80 under one-sun irradiation. Right axis, experimental absorption spectra measured by an integrated sphere in the visible and near-infrared regimes of Steam-80. The inset shows the infrared image of Steam-80 under one-sun illumination.
AWH performance and operational stability of the obtained nanoporous carbon
a, Comparison of water uptake of different nanoporous carbon samples at 20–50% RH. The error bars are based on uncertainties of water adsorption measurements (Supplementary Section 9). b, Water adsorption–desorption tests at 25 °C and 20–50% RH for Steam-80. At 200 min, the sample is exposed to one-sun irradiation to release the water. c, Water adsorption at 25 °C and 30% RH for Steam-80 and the carbon sample carbonized at 1,000 °C. The table in the inset shows the relative atomic concentrations of N, O and C, as well as the adsorption rate constant k. The dashed lines are the fitting curves of the first-order kinetic equation. d, Cycling experiments performed on Steam-80 at 25 °C and 50% RH.
Practical AWH
a, Image of the water harvester with MOF-derived nanoporous carbon (0.3 g of Steam-80 and packing porosity of ~0.90). Adsorption mode: the sorbent is sheltered from sunlight irradiation to absorb water vapour. Production mode: the sorbent is exposed to sunlight and then water is released and condensed. Scale bars, 1 cm. b, Formation, growth and coalescence of water droplets on the sidewall as a function of desorption time. c, Schematic showing the water-harvesting productivity of each cycle, solar flux, ambient RH and temperature. d, Comparison of the water production rate of nanoporous carbon and the state-of-the-art adsorbents under one-sun illumination³⁶. Water production rate, defined as the amount of water produced over a certain period, is a comprehensive figure of merit in AWH, which takes into account both capacity and kinetics.
Solar-driven, sorption-based atmospheric water harvesting (AWH) offers a cost-effective solution to freshwater scarcity in arid areas. Creating AWH devices capable of performing multiple adsorption–desorption cycles per day is crucial for increasing water production rates matching human water requirements. However, achieving rapid-cycling AWH in passive harvesters has been challenging due to sorbents’ slow water adsorption–desorption dynamics. Here we report an MOF-derived nanoporous carbon, a sorbent endowed with fast sorption kinetics and excellent photothermal properties, for high-yield AWH. The optimized structure (40% adsorption sites and ~1.0 nm pore size) has superior sorption kinetics due to the minimized diffusion resistance. Moreover, the carbonaceous sorbent exhibits fast desorption kinetics enabled by efficient solar-thermal heating and high thermal conductivity. A rapid-cycling water harvester based on nanoporous carbon derived from metal–organic frameworks can produce 0.18 L kgcarbon⁻¹ h⁻¹ of water at 30% relative humidity under one-sun illumination. The proposed design strategy is helpful to develop high-yield, solar-driven AWH for advanced freshwater-generation systems.
Top-cited authors
Andras Kis
  • École Polytechnique Fédérale de Lausanne
Jonathan Coleman
  • Trinity College Dublin
Aleksandra Radenovic
  • École Polytechnique Fédérale de Lausanne
Qing Hua Wang
  • Arizona State University
Takashi Taniguchi
  • National Institute for Materials Science