Integrating temporal and spatial control of electronic transitions for bright multiphoton upconversion

The applications of lanthanide-doped upconversion nanomaterials are limited by unsatisfactory brightness currently. Herein, a general strategy is proposed for boosting the upconversion efficiency in Er3+ ions, based on combined use of a core−shell nanostructured host and an integrated optical waveguide circuit excitation platform. A NaErF4@NaYF4 core−shell nanoparticle is constructed to host the upconversion process for minimizing non-radiative dissipation of excitation energy by surface quenchers. Furthermore, an integrated optical microring resonator is designed to promote absorption of excitation light by the nanoparticles, which alleviates quenching of excited states due to cross-relaxation and phonon-assisted energy transfer. As a result, multiphoton upconversion emission with a large anti-Stokes shift (greater than 1150 nm) and a high energy conversion efficiency (over 5.0%) is achieved under excitation at 1550 nm. These advances in controlling photon upconversion offer exciting opportunities for important photonics applications.


S2
Synthesis of thick shelled (7 and 14 nm) NaYF 4 :Er@NaYF 4 nanoparticles. The NaYF 4 :Er@NaYF 4 coreshell nanoparticles with thick NaYF 4 shells were synthesized by a successive hot-injection method. NaYF 4 shell precursors were first prepared by mixing an aqueous solution of Y(CH 3  for the NaYF 4 :Er@NaYF 4 (14 nm) coreshell nanoparticles is similar, except that using 0.2 mL of as-synthesized NaYF 4 :Er@NaYF 4 (7 nm) coreshell nanoparticles as the seed, reducing OA/ODE amount to half and adjusting the injection repetition to 30 times. And finally redisperse the nanoparticles in 0.2 mL of cyclohexane.
Synthesis of ligand-free nanoparticles. The as-prepared coreshell nanoparticles in cyclohexane (4 mL) were extracted and re-dispersed in 8 mL HCl solutions (0.1 M). The slurry solution was then sonicated at room temperature for 1 h and kept still for overnight in order to remove the surface oleate ligands. After the reaction, the ligand layer was discarded, and the nanoparticles were collected via centrifugation at 14000 rpm for 30 min and re-dispersed in ethanol. The washing process was repeated twice and the ligand-free nanoparticles were finally re-dispersed in 4 mL ethanol.
Synthesis of NaErF 4 microrods. The NaErF 4 :Lu (20%) microrods were prepared via a hydrothermal method. Typically, Er(CH 3 CO 2 ) 3 (1.6 mL, 0.2 M), Lu(CH 3 CO 2 ) 3 (0.4 mL, 0.2 M) and trisodium (1.6 mL, 0.3 M) was added to 15 mL of deionized water under vigorous stirring to form a milky suspension. The resulting suspension was then transferred to a 30 mL Teflon-lined autoclave and heated to 200 o C for 36 h. The product was collected by centrifugation, washed with deionized water and ethanol several times, and finally dispersed in 10 mL ethanol.
Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) images were taken on a JEOL JEM 2100F transmission electron microscope at an acceleration voltage of 200 kV. Scanning electron microscopy (SEM) imaging of SU-8 pattern on the microring resonator was performed using a Hitachi S-4800 field emission SEM. Before SEM imaging, the surface of the device was coated with a gold layer to prevent surface charging.
The luminescence spectra in the visible range were recorded with a Hitachi F-4600 spectrophotometer, in conjunction with a 980 or 1532 nm diode laser as the excitation source.
The luminescence spectra in the NIR range and decays of cyclohexane dispersion of nanoparticle were acquired on an Edinburgh FLSP920 spectrometer equipped with a 808, 980, or 1532 nm pulsed laser and a nitrogen-cooled NIR photomultiplier rod (Hamamatsu R5509-72). Optical micrographs of all the devices were recorded with an advanced research microscope (ECLIPSE Ni-U, Nikon). All measurements were performed at room temperature. The InputThrough and InputDrop port transfer function of the microring resonator was measured by the integration of a tunable laser, polarization synthesizer and optical power meter all connected with a Keysight Photonic Application Suite.
Fabrication of integrated optical circuit excitation platform. The integrated optical circuit excitation platform was fabricated from a high refractive index glass. The waveguide core of both the microring resonator and the waveguide loop are made of low-loss, high index doped silica glass that is semi-buried within a SiO 2 . The waveguide dimensions and refractive indices are 1.5 m x 1.5 m and 2.0 m x 2.0 m, and n=1.70 and n=1.60, for the microring resonator and the waveguide loop respectively. The bus waveguide that couples to the ring waveguide is 0.9 m x 1.5 m. In the fabrication, high-index silica-glass films were first deposited using standard chemical vapour deposition. Subsequently waveguides were formed using photolithography and reactive ion etching, producing extremely smooth sidewalls. The waveguides were then buried in fused silica glass. To allow strong interaction between the nanoparticles and the optical fields in the waveguide, the top cladding of the device was removed by chemical-mechanical polishing to the top of the core waveguide. Unlike the devices reported in ref. [2][3][4][5][6], the devices reported here do not have a mode-transformer as the chemicalmechanical polishing process does not allow their formation, resulted with an approximately 10 S4 dB addition coupling loss than that reported in ref. [2]. We reduced the radius of the microring resonator from 135 m (FSR=200 GHz) to 47 m (FSR=573 GHz) to further localize the excitation from the ring surface.
Excitation of NaYF 4 :Er@NaYF 4 nanoparticles with the integrated optical circuit platform.
The nanoparticles in an ethanol dispersion was drop cast on the surface of the pigtailed device and allow the full evaporation of ethanol. The excitation energy pumped by a tunable laser (81960A, Agilent.) first goes through a mini manual variable attenuator (Operating center wavelength at 1550 nm, Advanced Fiber Resources) to lower the excitation power and then a polarization controller (ProtoDel) was used to tune the polarization of excitation light. A 99:1 coupler was then utilized to assist reading 1% of the input power at 1550 nm before the laser finally arrived at the input port of microring resonator. To make sure that as much as possible of laser was coupled into microring resonator, Drop and Through ports were then monitored by using another power meters at 1550 nm (Figure 4a). The wavelength and polarization of the input laser were slowly adjusted until the power at the Drop port was maximized and the power at the Through port was minimized.
Experimental procedure for overlaying polymer waveguide on the microring resonator. 20 μL of NaErF 4 @NaYF 4 nanoparticles in ethanol dispersion (0.01 M) was first drop-casted on the microring-resonator. After the ethanol was evaporated, a 2 μm thick SU-8 2002 was spin-coated on the device followed by baking at 90 o C for 10 min. The device was then placed on an alignment stage to align the input/output fiber array with the device. To align the wavelength and polarization of the laser with the microring resonator, we first use a very low input laser power and slowly tune the wavelength as well as adjust the polarization controller until power at the drop port is maximized. Using a very low laser power for finding the resonance wavelength was to avoid too much light exposure to the straight bus waveguide and microring resonator at this stage. Once the resonance and polarization are aligned between the input and the resonator, the laser was then increased to 20 μW to start the curing process. After 10 min the device was removed from the alignment stage and transferred to a 90 o C hotplate for 10 min of postexposure bake procedure. After the device was cooled down to room temperature, then remove the unexposed part of SU-8 by immersing the device into the SU-8 developer for 1 min and finally resined in water.

S5
Fabrication and excitation of upconversion microdisk. The upconversion microdisk was fabricated through a sequence of photolithography, inductively coupled plasma (ICP) etching, photoresist removing and spin coating. [7] Typically, the oxidized silicon wafer was first cleaned by acetone, isopropanol and deionized water. Then the photoresist (AZ2020, Microchem) was spin-coated onto the wafer, followed by a standard photolithography protocol to pattern microstructure within the photoresist. The microstructure was then transferred onto the silica by performing the ICP etching procedure with C 4 F 8 as the anisotropic etching gas. Subsequently, the photoresist was thoroughly removed by a soaking process. And finally, the mixture of 2.9 wt% NaErF 4 @NaYF 4 nanoparticles and silica resin was spin-coating onto the preformed SiO 2 substrate to form the upconversion microdisk.
Calculation setup for relative oscillator strength. For calculating the energy states, we used an efficient formulation of time-dependent linear response density functional theory for the use within the plane wave basis set framework developed by Hutter. [8] Regarding the time-dependent density functional theory calculation functionalities, the related modules and codes have been recently implemented within the density functional perturbation theory (DFPT) framework within CASTEP package. [9] Under the excitation theoretical calculation, consensuses have been approached that the excitation light source can usually be simulated as external frequencydependent electric field and the electronic orbital level-variation are described by the electronic density response. Therefore, this approach can effectively compute all the corresponding response by valence electrons regarding the targeted external electric field of a set frequency.
Different from the Gaussian-based computational manner, we imported the theoretical method introduced by the pioneering work of Hutter. [8] This can reach computationally more efficient and the non-local pseudopotential technique can better approach the scaling linearly for the size with accuracy maintained. Moreover, Hutter's method avoids the transformation of the Kohn-Sham orbital matrix into the canonical basis and projects the orbitals into occupied and virtual parts. Further applied Tamm-Dancoff approximation (TDA) [10] with Lagrangian formulation, the geometry optimization is feasibly conducted under the excited states.
The above equation is the equation 17 of the Hutter's theoretical development work. [8] In this equation, the term ) 0 ( H is the ground state total energy (described by Kohn-Sham orbitals) Hamiltonian of the targeted system, i  represents the ground state eigenvalue for the band index The Eq. 2 is the essential equation written in the coding implemented for the electric-field-based (simplified laser light source) response in the CASTEP package. Moreover, our recent efforts have been startup for developing the tunable ω of the chosen external frequency and meanwhile the implementation of the double-laser-light-source excitations referred to the theoretical works of Lin et al. [11][12] It is also worth to mention the advantage of this method. To model the overall excited state levels for the targeted nanoparticle structural models, accurate calculations have met the bottleneck limited by the computation-loading based on the current method. In detail, the small solid particle systems usually contain over thousand electronic states that urges tremendous computation cost over several hundreds of gigabytes in memory size for data input/output (I/O), by using conventionally improved wavefunction-based method.
According to the above detail explanations on the theoretical fundaments, the relative oscillator strength (ROS) that we used here have been further derived from the as-calculated During the performed excited state calculation, we chose the two-electron based Tamm-Dancoff approximation imported from self-consistently corrected ground state wavefunctions. [10] To guarantee the stability of the electronic diagonalization process, currently we only consider using the block-davidson solver. Regarding different doped nanoparticle modeling system, the hexagonal phase β-NaYF 4 comprising different amount of Er 3+ dopants (12.5% to 100%) were The ground state wavefunction and the related electronic structure were calculated by simplified rotational invariant DFT+U method using the CASTEP source codes. [9] Hexagonal lattice with the 6 P space group is modeled for β-phase NaREF 4 (RE=Y and Er). The Na, Y, Er, and F norm-conserving pseudopotentials are generated using the OPIUM code in the Kleinman-Bylander projector form, [13] and the non-linear partial core correction. [14] A scalar relativistic averaging scheme [15] were applied to treat the spin-orbital coupling effect. In particular, we treat the (4f, 5s, 5p, 5d, 6s) states as the valence states of Er atom. The RRKJ method was chosen for the optimization of the pseudopotentials. [16] For more accurate calculations of the excited electronic states in β-NaREF 4 (RE=Y and Er), we used the self-consistent determination for the U correction on the localized 4f orbitals to correct the on-site Coulomb energy of the electron spurious self-energy. This is a two-way crossover linear response that generally searches for the optimal Hubbard U parameter to minimize the residue of the counteracting electronic self-energy and the orbital relaxation in the excited states. In order to obtain accurate orbital eigenvalues for electronic structures and transition levels, we have established an algorithm to determine the on-site electronic self-energy S8 and related wavefunction relaxation in the semicore d or f orbitals in heavy elements with mixed valence. [17][18][19][20][21][22][23][24][25][26] The algorithm is based on ab-initio two-way crossover linear response searching calculations by two different sets of functionally compiled CASTEP-17 developing source codes. [22,[27][28] The detail process is described in a previous work, [24] and the schematic theoretical determination process is demonstrated in ref. [29]. With the self-consistently determination process, the on-site Hubbard U parameters for 4f of Er, 4d of Y, and different 2p of F-sites were obtained respectively.  Figure S4d. To study the effect of Er 3+ concentration on the time decay, a high energy pulse (1 mJ mm -2 ) was thus consistently used to minimize the interference of energy transfer to the measured decay curves and to enhance the signal-to-noise ratio. The decay curves as a function of Er 3+ concentrations for the 4 G 11/2 and 2 H 9/2 states for nanoparticles with a mean shell thickness of 3 nm, 7 nm, and 14 nm, respectively. Decay curves were measured under excitation of a 980 nm pulsed laser at an energy density of 1 mJ mm -2 . Note that the decay curves of the 4 G 11/2 and 2 H 9/2 states for the core nanoparticles could hardly be recorded due to the extremely low photon counts. for reference and upconversion nanoparticles, respectively. The absorption of upconversion nanoparticles was then can be calculated to 590 μW at off resonance. (b) The emission is collected by an integrating sphere (Labsphere) of 3inch in diameter and detected by an Ocean Optics USB 2000 spectrometer (spectral range 180880 nm). The recorded spectra were calibrated with a standard halogen lamp and converted into spectral power distribution. The total radiant fluxes (Φ) of device were determined to be 6.6 μW and 20.1 μW at the off-and on-resonance wavelengths (1550.00 and 1549.47 nm), respectively. Accordingly, the off-and on-resonance energy conversion efficiencies (η = Φ / P smp ) are 6.6 μW / 590 μW = 1.1% and 20.1 μW / 400 μW = 5.0%. (c) TEM micrographs showing close size of the NaErF 4 @NaYF 4 upconversion nanoparticles and the NaYF 4 reference nanoparticles. Emission spectra and optical micrographs recorded from single NaErF 4 :Lu (20%) microrods on the microring resonator. Note that Lu 3+ dopant was introduced to alleviate energy migrationinduced concentration quenching by increasing the particle size and by diluting the erbium sublattice. The slightly weaker UV emission from the microrod with respect to that from the core−shell nanoparticles are due to: i) The microrod was lack of a surface protection layer. ii) The Er 3+ concentration was lower in the microrod. iii) The microrod was less efficiently interacted with the evanescent field of waveguides due to the larger size.

S25
Supplementary Figure 17. Performance of the waveguide circuit in enhancing upconversion luminescence in nanoparticles of various compositions and at different input powers. (a) Emission spectra recorded in the on-and off-resonance states. Emission spectra for each sample were normalized to its maximum emission peak at 20 mW excitation. (b) The ratio of UV emission at 382 nm between the on and off states. In general, the enhancement effect weakens with increasing excitation power and Er 3+ concentration due to saturation of upconversion emissions in the high-power regime. The results also show that a conventional waveguide excitation scheme needs to use ~20-fold higher excitation power than the microring resonator (i.e.; 20 mW versus 1 mW) to render the same emission profile. The waveguide core is embedded in the SiO 2 substrate but having the top surface exposed to environment. In the experiment, the tunable laser was set to a fixed wavelength at 1550 nm. The polarization controller is tuned until maximum power was recorded at the power meter connected to the output of waveguide loop. The laser power was then set to the designed power level to cure the SU-8 pattern.