Parallel photon avalanche nanoparticles for tunable emission and multicolour sub-diffraction microscopy

SJ_Zhang
Jun. 2, 2025

Abstract

Photon avalanche (PA) can generate upconversion luminescent emission that grows steeply as a function of excitation power, effectively exhibiting a high order of nonlinearity (N) that is attractive for applications ranging from photophysics studies to biophotonics. Besides the limitations in available material systems, PA is typically sustained by a single reservoir level, limiting the ability to modulate the chromaticity of the emission as well as leading to small values of N and large excitation thresholds. Here we report a parallel PA mechanism in holmium (Ho³⁺)-doped nanoparticles for tunable emission at room temperature. The intermediate ⁵I₇ and ⁵I₆ levels of Ho³⁺ serve as dual reservoir levels that create two parallel energy loops. This activates multiple emissive levels and enables red, green and blue PA emission under 965 nm continuous-wave excitation. By rationally engineering transition kinetics through controlling doping concentration and core/shell configuration, we demonstrate multicolour PA with large N values of 17–22 and mild excitation threshold of ~22 kW cm⁻². Moreover, emission can be tailored from almost pure red to intense red, green and blue by modifying the host lattice and introducing additional cross-relaxation pathways by doping with Ce³⁺/Tm³⁺. When using the nanoparticles to label biological cells, we demonstrate multicolour imaging on a single-continuous-wave-beam microscope with lateral spatial resolution of 78 nm and 102 nm in the green–blue and red channel, respectively. These findings open the way for manufacturing nonlinear multicolour fluorophores for versatile optical and biological applications.

Main

Nonlinear optics that exhibit high-order nonlinearity (N) have profound significance in diverse fields, such as infrared detection¹, frequency-upconverted lasing², three-dimensional data storage³, lithographic microfabrication⁴ and multiphoton microscopy⁵. By virtue of the ladder-like energy levels in 4f shells, lanthanide ions (Ln³⁺) are ideal nonlinear photon converters⁶. In particular, photon avalanche (PA) based on the positive-feedback interionic energy loops has been reported for Ln³⁺ to generate upconversion emissions with a giant N (ref. ⁷). Recently, PA of Tm³⁺ (ref. ⁸), Pr³⁺–Yb³⁺ (ref. ⁹) and Nd³⁺ (ref. ¹⁰) have been achieved at the nanoscale at room temperature (RT), which advances the application scenarios^11,12,13 compared with the previous ones operated at low temperature, especially in sub-diffraction imaging^8,9,14. Notably, √N-fold improved imaging resolution can be fulfilled directly on a canonical single-continuous-wave (CW)-beam microscope, which greatly simplifies the optical configuration of super-resolution microscopy^15,16.

Multicolour sub-diffraction imaging holds great significance as it allows for the clarification of the relative spatial organization and dynamical interactions of the targets of interests¹⁷. The multi-level configurations of Ln³⁺ make it applicable to produce various emissions under single-beam excitation¹⁸, which is especially attractive for PA. The intermediate reservoir level, which reserves energy for the burst of PA, links with the emissive level via excited-state absorption (ESA), and thus plays an important role in determining the emission colours. Conventionally, PA is sustained by one reservoir level from either monotypic (Tm³⁺ (ref. ⁸), Nd³⁺ (ref. ¹⁰); Fig. 1a, left) or heterotypic (Pr³⁺–Yb³⁺ (ref. ⁹); Fig. 1a, right) emitters, leading to the activation of one primary emissive level. To construct multicolour PA emissions, tandem PA–ESA–energy transfer upconversion (ETU) mechanism¹⁹, for example, Tm³⁺–Gd³⁺–A³⁺ (A³⁺ = Eu³⁺, Tb³⁺ and so on), and transitions from one emissive level to different lower ones^9,20, for example, Pr³⁺, were developed. Nonetheless, these strategies were accompanied by inherent deficiencies. The tandem PA–ESA–ETU mechanism endows the short-wavelength emissions of Tm³⁺/A³⁺ with low efficiency, small N and large excitation threshold (P_th)¹⁹, while the PA of Pr³⁺ suffers from non-adjustable chromaticity⁹, exhibiting hybrid emissions with an invariable spectral ratio. Modulation of the PA emission chromaticity and its implementation in multicolour sub-diffraction imaging is greatly desirable.

Fig. 1: Principles of the PPA.

a,b, Schematic showing the conventional PA (in monotypic (left) and heterotypic (right) emitters; a) and the PPA (b) mechanisms. Dual reservoir levels in light red and light green colours are illustrated in PPA for clarity. c, Proposed PPA in Ho³⁺ under 965 nm excitation. The ⁵I₇ and ⁵I₆ levels are highlighted with light red and light green colours, respectively. Typical CRs, obtained from differential rate equation modelling, are shown for clarity and are indicated by the thickness of the arrows based on their rates. The crucial CRs for the PPA are depicted in Supplementary Information. d, Absorption spectrum of HoCl₃ aqueous solution at RT. e, PA emission spectrum of NaGdF₄:10%Ho@NaYF₄ nanoparticles under 965 nm excitation at RT. f, Schematic of the PPA-enabled multicolour sub-diffraction microscopy. Here dual-channel imaging was demonstrated, with the red and green–blue PA emissions being collected in red (R) and green–blue (GB) channels, respectively. I_em, emission intensity; I_ex, excitation intensity. NIR, near-infrared.

Herein we propose the parallel photon avalanches (PPA) in monotypic emitters (Fig. 1b) for tunable emission and multicolour sub-diffraction imaging. In PPA, there are two reservoir levels that link with different emissive levels under single-beam excitation, facilitating the simultaneous operation of two PA loops to generate multicolour emissions. Moreover, the spectral components can be modulated for tunable chromaticity. Holmium ion (Ho³⁺) with dual long-lived intermediate levels, ⁵I₇ and ⁵I₆, is used as the PPA emitter. In a fluoride host, the lifetime of ⁵I₇ and ⁵I₆ levels lies in the order of milliseconds²¹, which is suitable for energy reservation^22,23. To meet the excitation criterion of PPA, an excitation light of 965 nm is used (Fig. 1c), which matches well with the ESA of ⁵I₇ → ⁵F₅ and ⁵I₆ → ⁵S₂/⁵F₄, but is ~1,750 cm⁻¹ above the barycentre of the ground-state absorption (GSA) of ⁵I₈ → ⁵I₆ (Fig. 1d and Supplementary Fig. 1). The dual reservoir levels, ⁵I₇ and ⁵I₆, associate with the red (⁵F₅ → ⁵I₈, PA-I) and green (⁵S₂/⁵F₄ → ⁵I₈, PA-II) emissions, respectively, and ESAs and extra interionic cross-relaxations (CRs) can activate the blue emission (F_2,3/⁵K₈ → ⁵I₈), leading to sustainable multicolour PA (Fig. 1e). Moreover, by altering the host lattice, co-doping with Ce³⁺/Tm³⁺ and so on, tunable PA chromaticity can be achieved for multicolour sub-diffraction imaging (Fig. 1f).

Results

PA emissions of Ho³⁺-doped nanoparticles

Because of the low phonon energy and broadband transparency²⁴, hexagonal structured NaREF₄ was adopted for hosting the PA of Ho³⁺. NaGdF₄:10%Ho@NaYF₄ core/shell nanoparticles with 13.5 nm core and 3.1 nm shell were prepared (Fig. 2a), in which the core affords PA, while the shell minimizes non-radiative (NR) loss. The excitation–emission contour map shows two excitation bands ranging from 860 nm to 1,010 nm (Fig. 2b and Supplementary Fig. 2), one locates at 895 nm, corresponding resonant GSA/ESA upconversion (Supplementary Figs. 3 and 4), while the other one centres at 965 nm, corresponding to PA. Under 965 nm excitation, prominent red (645 nm, ⁵F₅ → ⁵I₈), green (540 nm, ⁵S₂/⁵F₄ → ⁵I₈) and blue (485 nm, ⁵F_2,3/⁵K₈ → ⁵I₈) emissions are generated (Fig. 1e). The multicolour PA of Ho³⁺ provides new excitation/emission wavelengths for the PA systems^8,9,10.

Fig. 2: PA emissions of Ho³⁺-doped nanoparticles.

a, High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of NaGdF₄:10%Ho@NaYF₄ nanoparticles. The nanoparticles are monodispersed with a diameter of 19.6 ± 1.0 nm. The core/shell structure can be disclosed from the contrast difference in the HAADF-STEM image. Inset in a: the high-resolution TEM image of the nanoparticle, in which the hexagonal phase can be identified. b, Excitation–emission contour map of NaGdF₄:10%Ho@NaYF₄ nanoparticles at RT. c,d, Logarithmic plot (c) and rise time (d) of the red, green and blue emission intensity versus 965 nm excitation intensity at RT. e, Photostability of PA emissions under continuous laser irradiation (100 kW cm⁻²) for 1 h. Photons were collected with an avalanche photon detector. The binning time for the data points is 1 ms. Inset: corresponding intensity distribution, showing no photoblinking or photobleaching. f, PA emission spectra collected under continuous laser irradiation for 1 h with a time interval of 60 s, showing no photoblueing.

The PA emissions exhibit well-defined signatures, including large N, critical P_th and long rise time that has an up-and-down dependence on the excitation intensity²⁵. For Ho³⁺-doped nanoparticles, N of 17, 22 and 22 were noticed for the red, green and blue emissions, respectively, along with mild P_th of 22 kW cm⁻², 23 kW cm⁻² and 23 kW cm⁻² (Fig. 2c and Supplementary Fig. 5). The long and volcano-shaped rise time with peak values of 343 ms (red), 352 ms (green) and 371 ms (blue) further consolidates the PA character (Fig. 2d and Supplementary Figs. 6–8). Compared with previous PA (Supplementary Table 1), the PA of NaGdF₄:10%Ho@NaYF₄ nanoparticles are attractive in terms of the giant N, mild P_th and non-overlapped emissions. The PA emission intensity was stable after 1 h of continuous laser irradiation (Fig. 2e), without photoblinking or photobleaching, even the binning time was down to 1 ms. In addition, the PA emission wavelength keeps invariant, showing no hypsochromic photoblueing (Fig. 2f), which is inevitable for dyes²⁶ and quantum dots²⁷ under intense irradiation.

PPA mechanism of Ho³⁺

To investigate the PA mechanism of the Ho³⁺ under 965 nm excitation, the PA signatures were evaluated. The green and blue emissions have a relatively larger N and longer rise time than that of the red one, suggesting different looping pathways for these emissions. To elucidate the PA mechanism, differential rate equation modelling was performed based on the energy levels of Ho³⁺ (Supplementary Fig. 9 and Supplementary Tables 2–9). The simulated PA (Fig. 3a) agrees with the experimental results, with a slight deviation of N for the blue emission, which could be ascribed to the influence from weak emissions around P_th on fitting. In addition, NR losses in nanoparticles rendered a slightly higher P_th than the simulated one.

Fig. 3: PPA mechanism and kinetics modulation of Ho³⁺-doped nanoparticles.

a, Simulated population density on the emissive levels as a function of the excitation intensity. Slopes of the PA linear fits (N) are shown (inset), and the P_th is 18 kW cm⁻². b, Simulated logarithmic plot of the population density on the red emissive level versus excitation intensity with different σ_ESA/σ_GSA. c, Effect of σ_ESA/σ_GSA on the N (top) and P_th (bottom) of the PA emissions from simulation. σ_ESA refers to the absorption cross-section of ⁵I₇ → ⁵F₅, which was set as 2.1 × 10 ⁻²¹ cm² in the modelling. d, Simulated population density on the ground state (⁵I₈), reservoir (⁵I₇, ⁵I₆) and other intermediate (⁵I₅, ⁵I₄) levels as a function of the excitation intensity. Slopes of the PA linear fits (N) are shown (inset), and the P_th is 18 kW cm⁻². e–g, TEM images (e), logarithmic plot of the red PA emission intensity versus excitation intensity (f), and experimental N and P_th of the PA emissions (g) of NaGdF₄:10%Ho naked nanoparticles of different sizes. h–j, HAADF-STEM images (h), the logarithmic plot of the red PA emission intensity versus excitation intensity (i), and experimental N and P_th of the PA emissions (j) of NaGdF₄:10%Ho@NaYF₄ core/shell nanoparticles. k,l, Logarithmic plot of the red PA emission intensity versus excitation intensity (k) and experimental N and P_th of the PA emissions (l) of NaGdF₄:Ho@NaYF₄ core/shell nanoparticles with different contents of Ho³⁺.

According to the modelling, the PA mechanism can be elaborated. The GSA pathway was determined to be ⁵I₈ → ⁵I₆, while the population of the ⁵I₅ level is neglectable (Supplementary Fig. 10). As the N and P_th are closely related to the absorption cross-section of GSA and ESA⁸, a reasonable σ_ESA/σ_GSA should be around 75,000 (Fig. 3b,c and Supplementary Fig. 11), which is larger than the criteria of 10,000 for PA. The interionic CR, which is decisive for the population of intermediate reservoir levels, was evaluated with a knockout strategy in simulation²⁸. Through careful knockout of CRs separately and collectively (Supplementary Figs. 12–17), the key ones were identified (Supplementary Fig. 15). In addition, the role of ⁵I₇ and ⁵I₆ as reservoir levels was recognized by their obviously larger population density than that of other intermediate levels (Fig. 3d), and was further verified from the disappearance of PA if CRs populating ⁵I₇ and ⁵I₆ levels were knocked out (Supplementary Fig. 14).

The contribution of the two reservoir levels to PPA is analysed (Extended Data Fig. 1). Typically, the ⁵I₇ reservoir level is populated by CRs multiplicatively (CR i; Supplementary Fig. 15), and the subsequent ESA of ⁵I₇ → ⁵F₅ triggers the population of the ⁵F₅ level for the red PA emission. Meanwhile, the ⁵I₆ level is populated through the cooperation of multiple CRs (CRs ii–v; Supplementary Fig. 15) along with proper NRs, which activates the ⁵S₂/⁵F₄ level with the ESA of ⁵I₆ → ⁵S₂/⁵F₄, resulting in green PA emission. Despite the fact that CR ii reduces N (Supplementary Fig. 13b), as one would expect as an ETU process, the high nonlinearity of the other processes that collectively contribute to the loop allows a high nonlinearity for the PPA. Besides, the blue PA emission can be produced simultaneously through ESAs and CRs, some of which (CRs viii, x, xi; Supplementary Fig. 15) are participated by the reservoir levels. The energy-level configuration of Ho³⁺ supports the unique PPA mechanism in getting multicolour emissions, which is more efficient than the tandem PA–ESA–ETU of Tm³⁺–Gd³⁺–A³⁺ (ref. ¹⁹) and provides tunable chromaticity compared with Pr³⁺ (ref. ⁹). The quantum yield of the PPA of Ho³⁺ was inferred with a kinetic computational model²⁹, and an overall quantum yield of ~35% was deduced (Supplementary Fig. 18), which is comparable to that of Tm³⁺ (ref. ⁸) and Pr³⁺ (ref. ⁹).

Kinetics modulation of PA of Ho³⁺

We then investigated the dependence of the PA kinetics of Ho³⁺ on the NR and CR rates, which is primarily controlled by the density of surface defects³⁰ and the concentration of Ho³⁺ (ref. ³¹), respectively. It can be found that as the size of nanoparticles increased, the N increased gradually, while the P_th decreased and reached a plateau (Fig. 3e–g and Supplementary Fig. 19). For the 48.5 nm nanoparticles, PA with an N of 16–21 can be produced. This suggests that a small NR rate should be favourable to PA, which was further verified by the simulation (Supplementary Fig. 20).

Surface passivation with inert shells is effective to reduce the NR rate by protecting Ln³⁺ from surface losses^32,33. We show that as the shell thickened, the N increased and levelled off, while the P_th decreased correspondingly (Supplementary Figs. 21–23). This implies that thick shells can sustain PA with a minimized influence from NR. To validate this hypothesis, we deposit thick shells onto nanoparticles of different sizes (Fig. 3h and Supplementary Figs. 24 and 25). As expected, these core/shell nanoparticles exhibit very similar PA kinetics (Fig. 3i,j and Supplementary Fig. 26). Therefore, with rationally engineered core/shell structure, giant N and low P_th can be achieved in small-sized nanoparticles, for example, the 19.6 nm core/shell nanoparticles with an N of 17–22 and mild P_th of ~22 kW cm⁻² (Fig. 2).

The influence of CR rate on PA kinetics was studied by tuning the content of Ho³⁺. To exclude the interference from NR, thick NaYF₄ shells were grown (Supplementary Figs. 27 and 28). Prominent PA can be observed as the Ho³⁺ concentration exceeds 2% (Supplementary Fig. 29). The avalanching N shows a rise–fall dependency and reaches a maximum with 10% Ho³⁺ (Fig. 3k,l and Supplementary Fig. 30), while the P_th varies in a fall–rise manner and reaches a minimum with 20–40% Ho³⁺. This suggests that heavy doping of Ho³⁺ is essential to trigger adequate CR for PA. However, overloaded CR may lead to energy dissipation and is deleterious to PA, which is further validated by the simulation (Supplementary Fig. 31).

Single-CW-beam sub-diffraction microscopic imaging

To study the sub-diffraction imaging capability of the PPA of Ho³⁺, single NaGdF₄:Ho@NaYF₄ nanoparticles were imaged on a single-CW-beam scanning multiphoton microscope (Fig. 4a–c and Supplementary Figs. 32 and 33). The green and blue emissions were collected simultaneously in the GB channel for enhanced signals and were separated from the red emission in the R channel by a dichroic mirror, enabling synchronous detection. Considering the giant N and high brightness, 54.8 nm NaGdF₄:10%Ho@NaYF₄ nanoparticles (Fig. 3h) were used. A high-numerical aperture (NA 1.45) objective lens was applied, suggesting a diffraction limit of ~333 nm. It can be found that the lateral resolution, determined by the full-width at half-maximum (FWHM) of the point spread function (PSF) profile, was improved as the excitation intensity approached to the P_th (refs. ^8,9). Moreover, because of the larger N, the GB emission enables a higher imaging resolution than the R one at the same excitation intensity. In the P_th regime, the imaging resolution can be promoted to 74 ± 3 nm (19 kW cm⁻²) and 85 ± 6 nm (25 kW cm⁻²) in R and GB channels, respectively.

Fig. 4: Single-CW-beam sub-diffraction imaging enabled by PPA.

a, Imaging of single NaGdF₄:10%Ho@NaYF₄ nanoparticles with individual PA emissions in the R and GB channels. b, Enlarged views of the selected nanoparticle in a and corresponding resolution diagram in the R and GB channels. The dimension of the images in a and b is 512 × 512 pixels. c, PSF profiles and corresponding Gaussian fits along the dashed line crossing the nanoparticle in b. d, Subcellular imaging of the intercellular actin filaments (tunnelling nanotubes) of BS-C-1 cells labelled with NaGdF₄:10%Ho@NaYF₄ nanoparticles in the R and GB channels. The image dimension is 800 × 800 pixels. e, Close-up imaging of the region of interest in d. f, Enlarged views of the two regions of interest in e. The dimension of images under 634 kW cm⁻² excitation is 512 × 512 pixels, and that of others is 800 × 800 pixels in e and f. The pixel dwell time is 200 μs for all images. g, PSF profiles and corresponding Gaussian fits along the dashed line crossing the actin filaments in f.

Subcellular imaging of the intercellular actin filaments (tunnelling nanotubes), which play a crucial rule in the cell-to-cell communication³⁴, was then performed. NaGdF₄:Ho@NaYF₄ nanoparticles were modified with polyacrylic acid (PAA) to render aqueous dispersity and conjugated with phalloidin (Supplementary Fig. 34) to achieve targeting of actin filaments. After hydrophilization, the giant N maintained (Supplementary Fig. 35), while the P_th increased slightly to 27 kW cm⁻², which should be ascribed to the energy losses via hydroxyl vibration³⁵. The intercellular actin filaments of BS-C-1 cells can be well stained by the PA nanoparticles (Fig. 4d and Supplementary Figs. 36 and 37). At a large field of view (FOV; 63.2 × 63.2 μm²) involving a whole cell, the resolution of the actin filaments is improved obviously with a decreased excitation intensity. The higher resolution enabled by PA emissions is more prominent in the close-up scanning of the local actin filaments with a relatively small FOV (15.2 × 15.2 μm²; Fig. 4e). Sharply contrasted images with additional details can be obtained with excitation intensity approaching the P_th. From the further enlarged views of the ultrastructure of the actin filaments, drastically improved imaging resolution can be noticed (Fig. 4f,g). The FWHM was promoted to 95 ± 5 nm and 89 ± 6 nm in R and GB channels, respectively, reinforcing the multicolour potential from PPA of Ho³⁺.

Tunable emission for multicolour sub-diffraction imaging

The multicolour PA emissions endow Ho³⁺-doped nanoparticles with prospects for multicolour sub-diffraction imaging. Besides the influences from doping concentration (Supplementary Fig. 29) and excitation intensity (Supplementary Fig. 5), the effect of host lattice and lanthanide co-doping on the PA chromaticity of Ho³⁺ was studied. Cubic NaYF₄, LiYF₄ and Y₂O₃ core/shell nanoparticles were prepared as counterparts of hexagonal NaGdF₄ (Fig. 5a, Supplementary Fig. 38 and Supplementary Table 10). It was found that hexagonal NaGdF₄ was favourable for the green/blue emissions of Ho³⁺, while LiYF₄ was beneficial to the red one³⁶ (Fig. 5a and Supplementary Fig. 39). The PA of Ho³⁺ exhibits a smaller N and larger P_th in cubic NaYF₄ and LiYF₄ compared with that in hexagonal NaGdF₄ (Supplementary Fig. 40), which should be related to the larger phonon energy that aggravates NR. The spectral splits of Ho³⁺ are more prominent in LiYF₄ (ref. ³⁷), which can be used to distinguish the luminescent sources.

Fig. 5: PPA-enabled tunable chromaticity and multicolour sub-diffraction imaging.

a, PA emission spectra of Ho³⁺ in different host materials, including hexagonal (β) NaGdF₄, cubic (α) NaYF₄, LiYF₄ and Y₂O₃. b, PA emission spectra of NaGdF₄:10%Ho,Ce@NaYF₄ nanoparticles with various concentrations of Ce³⁺. Inset: the CR process between Ho³⁺ and Ce³⁺. c, PA emission spectra of NaGdF₄:10%Ho,Tm@NaYF₄ nanoparticles with various concentrations of Tm³⁺. Inset: the CR processes between Ho³⁺ and Tm³⁺. d,e, PA emission spectra of LiYF₄:20%Ho,2%Ce@LiYF₄ (d) and NaGdF₄:5%Ho,5%Tm@NaYF₄ (e) nanoparticles. The excitation power density is 500 kW cm⁻² for a–d and 3,500 kW cm⁻² for e. f, Logarithmic plot of the B (left), G (middle) and R (right) PA emission intensity versus excitation intensity for LiLuF₄:20%Ho,0.5%Ce@LiYF₄ (LLF:Ho,Ce@LYF) and NaGdF₄:10%Ho,10%Tm (NGF:Ho,Tm) nanoparticles. g, Multicolour imaging of single LLF:Ho,Ce@LYF and NGF:Ho,Tm nanoparticles with respective R and GB PA emissions under 965 nm excitation. h, PSF profiles and the corresponding Gaussian fits along the dashed line crossing the nanoparticle in g. i, Red PA emission spectra of the NGF:Ho,Tm (1, 2) and LLF:Ho,Ce@LYF (3) nanoparticles in g. j, Multicolour subcellular imaging of the phalloidin-modified LLF:Ho,Ce@LYF and PAA-modified NGF:Ho,Tm nanoparticles with respective R and GB PA emissions in a large FOV encompassing a whole cell under 965 nm excitation. k, Enlarged multicolour subcellular imaging of the upper left region of interest in j. l, PSF profiles and the corresponding Gaussian fits along the dashed line in k. Connecting lines between data points are shown in the red PSF profiles (h,l) for clarity. m, Enlarged multicolour subcellular imaging of the lower right region of interest in j. The pixel dwell time is 200 μs for all images.

To further modulate the chromaticity, additional CRs were introduced through lanthanide co-doping. Ce³⁺ with an excited level of ²F_7/2 (~2,300 cm^–1) was co-doped to increase the ratio of the red emission through the CR of ⁵I₆ (Ho³⁺) + ²F_5/2 (Ce³⁺) → ⁵I₇ (Ho³⁺) + ²F_7/2 (Ce³⁺)³⁸ (Fig. 5b and Supplementary Fig. 41), favouring the population of the red-emissive level (⁵F₅) via a subsequent ESA. Meanwhile, the introduction of Ce³⁺ leads to a decreased N and increased P_th (Supplementary Fig. 42). Tm³⁺, with a long-lived ³F₄ reservoir level⁸, was co-doped to increase the ratio of the green/blue emissions (Fig. 5c and Supplementary Fig. 43) through the CRs of (⁵F₅ (Ho³⁺) + ³H₆ (Tm³⁺) → ⁵I₆ (Ho³⁺) + ³F₄ (Tm³⁺) and ³F₄ (Tm³⁺) + ⁵F₅ (Ho³⁺) → ³H₆ (Tm³⁺) + ⁵F_2,3/⁵K₈ (Ho³⁺)). Moreover, the introduction of Tm³⁺ leads to a larger N and smaller P_th compared with Tm³⁺-free nanoparticles (Supplementary Fig. 44), which further consolidates the significance of reservoir levels in PA.

Taken together, LiYF₄ co-doped with Ce³⁺ and hexagonal NaGdF₄ co-doped with Tm³⁺ favour the red and green/blue PA emissions of Ho³⁺, respectively. Enlightened by this, almost pure red (Fig. 5d) and similarly intense red, green and blue (Fig. 5e) PA emissions can be obtained, respectively (Supplementary Fig. 45). For multicolour sub-diffraction microscopy, LiLuF₄:20%Ho,0.5%Ce@LiYF₄ (LLF:Ho,Ce@LYF) with stronger red emission and a smaller P_th and NaGdF₄:10%Ho,10%Tm (NGF:Ho,Tm) with prominent green/blue emission and a larger P_th were optimized (Fig. 5f and Supplementary Fig. 46). This enables the collection of red and green/blue emissions from LLF:Ho,Ce@LYF and NGF:Ho,Tm nanoparticles at low- and high-power excitation, respectively.

The multicolour sub-diffraction imaging was first demonstrated with single nanoparticles in R and GB dual channels (Fig. 5g–i and Supplementary Fig. 47). Single LLF:Ho,Ce@LYF and NGF:Ho,Tm nanoparticles, exhibiting distinctive spectral profiles, were imaged on the single-CW-beam scanning multiphoton microscope. Enabled by the respective stronger green/blue and red emissions under high- and low-power excitation, dual-channel imaging was achieved with a resolution of 112 ± 5 nm (GB)/125 ± 13 nm (R). The imaging resolution of NGF:Ho,Tm nanoparticles can be further improved to 73 ± 9 nm (GB)/58 ± 5 nm (R) at P_th (Supplementary Fig. 47). Subsequently, multicolour subcellular imaging was performed. LLF:Ho,Ce@LYF and NGF:Ho,Tm nanoparticles were modified with phalloidin and PAA molecules (Supplementary Fig. 48) to label the intercellular actin filaments and enter the cytoplasm (Fig. 5j–m, Extended Data Fig. 2 and Supplementary Fig. 49), respectively. In both the large FOV encompassing a whole cell and the small FOV in proximity to the membrane, the intercellular filaments stained by LLF:Ho,Ce@LYF and intracellular NGF:Ho,Tm nanoparticles can be imaged in R and GB channels, respectively, with an FWHM of 78 ± 8 nm (GB)/102 ± 13 nm (R). Interestingly, the intercellular structure of the LLF:Ho,Ce@LYF-stained tunnelling nanotubes (R), containing NGF:Ho,Tm nanoparticles (GB), was captured (Fig. 5m). Moreover, additional parallel subcellular images (Extended Data Fig. 3 and Supplementary Fig. 50) corroborate the multicolour sub-diffraction imaging capability of PPA.

Discussion

By utilizing the dual-reservoir-level configuration of Ho³⁺, we report the PPA in Ho³⁺-doped nanoparticles for tunable multicolour emissions with large N. We unravelled that the less NR events and the adequate CR rate are favourable to PA emissions. Benefiting from the ~20th order of nonlinearity, the PPA of Ho³⁺ can improve the imaging resolution towards 74 nm, about 1/13 of the excitation wavelength, on a diffraction-limited single-CW-beam multiphoton microscope. Moreover, through adjusting the excitation intensity, host lattice and CR pathways, the PA chromaticity of Ho³⁺ can be modulated. The resulted dominant red and green/blue PA emissions, exhibiting different excitation thresholds, facilitate multicolour sub-diffraction imaging.

Compared with the previously reported single-loop PA^8,9,10, the proposed PPA of Ho³⁺ exhibit unique advantages (Extended Data Table 1 and Supplementary Table 11), including the highly efficient multicolour PA emissions with a large N and mild P_th, largely tunable chromaticity and the applicable multicolour sub-diffraction imaging with a resolution down to sub-100 nm. Moreover, compared with the ~585 nm/~750 nm excitation for the PA of Ho³⁺-doped bulk materials (Supplementary Table 1), the 965 nm excitation is distinctive in generating multicolour PA emissions and should facilitate deep tissue imaging. In the near future, multifocal parallel excitation scheme³⁹ and shortening of the PA rise time from enhanced local electric field by metal nanostructures⁴⁰ would be in favour of fast imaging scenarios. To further reduce the P_th of PA, the sensitization from organic dyes or quantum dots as well as the coupling with metal nanostructures might be helpful, which is capable of increasing the absorption⁴¹ (Supplementary Fig. 51) and amplifying the local electric field⁴², respectively. It is noteworthy that the PA emissions are compatible with other sub-diffraction imaging modalities⁴³ to achieve flexible applications. Owing to the great sensitivity to microenvironments, PPA holds great promise for subcellular sensing and detection^{20,44,45,46,47}. Our investigations establish a paradigm for exploring novel PA nanoparticles and provide an interesting platform for multicolour sub-diffraction imaging.

Methods

Synthesis of hexagonal NaGdF₄:Ho and NaGdF₄:Ho@NaYF₄ nanoparticles

Hexagonal NaGdF₄:Ho nanoparticles were synthesized with a thermal decomposition method. Two steps are involved, namely the formation of cubic NaGdF₄:Ho nanoparticles and subsequent phase transition towards the hexagonal counterparts. The NaGdF₄:Ho@NaYF₄ core/shell nanoparticles were synthesized by growing NaYF₄ shells epitaxially over the hexagonal NaGdF₄:Ho nanoparticles. Detailed procedures are as follows. (i) Cubic NaGdF₄:Ho: 1 mmol CF₃COONa and 1 mmol RE(CF₃COO)₃ (RE = Gd, Ho) were added into a three-necked flask containing 10 mmol OA, 10 mmol OM and 20 mmol ODE. The slurry was heated to 110 °C under vigorous stirring to remove water and oxygen till a clear solution formed. Then, the solution was heated to 310 °C for 15 min under N₂ atmosphere with a heating rate of 15 °C min⁻¹. Upon cooling to RT, an excess amount of ethanol was added to precipitate the products. Nanoparticles were collected by centrifugation at 7,800 rpm for 10 min. Cyclohexane (10 ml) was used to disperse the nanoparticles. (ii) Hexagonal NaGdF₄:Ho: 5 ml colloidal solution of cubic NaGdF₄:Ho nanoparticles, 0.5 mmol CF₃COONa and 0.5 mmol RE(CF₃COO)₃ were added into a three-necked flask containing 20 mmol OA and 20 mmol ODE. The reaction procedures and after-treatments were the same as that in step (i), except for the 30 min reaction time at 310 °C. Hexagonal NaGdF₄:Ho nanoparticles were dispersed in 10 ml cyclohexane for further use. (iii) NaGdF₄:Ho@NaYF₄: 2 ml as-synthesized hexagonal NaGdF₄:Ho nanoparticles, a given amount of CF₃COONa and Y(CF₃COO)₃ were added into a three-necked flask containing 20 mmol OA and 20 mmol ODE. The reaction procedures and after-treatments were the same as that in step (ii). NaGdF₄:Ho@NaYF₄ core/shell nanoparticles were dispersed in 5 ml cyclohexane for further use.

Cell incubation with phalloidin-modified nanoparticles

BS-C-1 cells were cultured in 96-well confocal plates overnight. Approximately 15,000 cells were contained in each unit. Then, the cells were washed with PBS buffer three times and fixed at RT with immunostaining fixative solution for 20 min. The fixed cells were washed by PBS three times and permeabilized with immunostaining permeate solution for 20 min. Subsequently, the immunostaining permeate solution was removed, and the immunostaining blocking solution was added for another 20 min. After removing the immunostaining blocking solution, 200 μl HEPES solution containing phalloidin-modified nanoparticles was added for labelling. The labelling of the actin filaments was kept at 4 °C overnight. Cells were rinsed with HEPES buffer three times for further imaging.

Cell incubation with PAA- and phalloidin-modified nanoparticles for multicolour imaging

The BS-C-1 cells were cultured overnight in 20 mm glass bottom dish containing approximately 15,000 cells. The cells were fixed with 100% ice-cold methanol at –20 °C for 10 min and were washed with 0.05% Triton X-100 once. Then, 200 μl solution of as-prepared phalloidin-modified nanoparticles was added for staining overnight at RT, and the cells were washed with PBS buffer three times. The fixed cells were subsequently permeabilized with 0.5% Triton X-100 for 20 min at RT, followed by the addition of 200 μl PAA-modified nanoparticles for staining for 4–6 h at RT. The cells were then rinsed with PBS buffer three times for subsequent imaging.

PA emission spectra measurements

Thin nanoparticle films were prepared for measuring the pump power-dependent PA emission spectra on a modified confocal microscope (IX71, Olympus). The thin nanoparticle films were transparent and made by dropping 10 μl diluted colloidal solution of Ho³⁺-doped nanoparticles (approximately 1 μmol ml⁻¹ (Ln)) on a glass coverslip. The 965 nm CW laser is provided by a wavelength-tunable Ti-sapphire laser (Mira 900, Coherent). Necessary filters are used to exclude the interference from undesired laser lines. The laser beam was directed to the back aperture of an oil-immersed 100× objective (UPLFLN100XOI2, Olympus, NA 1.30) by a single-mode fibre. The PA emissions passed through the same objective and were collected by a visible spectrometer after filtration by a short-pass filter. For excitation wavelength-dependent emission spectra measurements, a 750 nm short-pass filter (F2, ET750SP-2p8, Chroma) and a Horiba spectrometer were used. For excitation intensity-dependent emission spectra measurements, a 694 nm short-pass filter (F5, FF02-694/SP-25, Semrock) and a visible spectrometer (QE65000, Ocean Optics) were used. The power of the 965 nm laser is tuned by the combination of the half-wave plate (Thorlabs) and the polarized beam splitter (Thorlabs). By rotating the half-wave plate, the polarization direction of the laser is modulated, which further leads to the adjustment of the relative proportion of the two perpendicular beams as the laser passes through the polarized beam splitter. Since only one of the perpendicular beams was used for exciting the nanoparticles, the pump power can be flexibly tuned with the variation of the relative proportion. The lasing output wavelength from the Ti-sapphire laser is stable throughout the tuning of the excitation power. The diameter of the laser spot was defined as the full width at 1/e² of maximum of the PSF profile. The excitation intensity was calculated with the measured laser power divided by the area of the laser spot. The P_th of the PA emission was determined by the intersection of the linear fits of the data points below and above the avalanching threshold.

PA rise curve measurements

The 965 nm CW laser beam was modulated by a chopper (model SR540, Stanford). A red (FF01-655/40-25, Semrock), green (FF01-530/23-25, Semrock) or blue (ET480/20 X, Chroma) filter was added to obtain the rise curve of specific PA emissions before the emissions enter the PMT detector (7ID101-CR131, SOFN Instruments). The PA rise curve is available on the oscilloscope (SDS1000X-C, Siglent).

Sub-diffraction imaging of PA nanoparticles and subcellular actin filaments

For sub-diffraction imaging of single PA nanoparticles, 10 μl diluted colloidal solution of Ho³⁺-doped nanoparticles (approximately 0.1 μmol ml⁻¹ (Ln)) was spin-coated on a clean glass coverslip. For subcellular sub-diffraction imaging and multicolour subcellular imaging, the 96-well confocal plate and 20 mm glass bottom dish were used, respectively. The R/GB dual-channel sub-diffraction imaging was performed on a multiphoton laser scanning microscope (FV1000MPE-S, IX81, Olympus). A single-mode fibre is used to direct the 965 nm CW laser beam into the microscope via a dichroic mirror (T690spxxf-UF1, Chroma). The laser beam was focused by an oil-immersed 100× objective (UPLXAPO100XO, Olympus) with an NA of 1.45. Red band-pass (FF01-660/30-25, Semrock) and green + blue short-pass (FF570-Di01-25 × 36, Semrock) filters were used to allow the transmission of R and GB PA emission signals, respectively. Two PMTs were used to detect the dual-channel PA emission signals.

Differential rate equation modelling

Numerical solution of the differential rate equation was simulated by RStudio with the Runge–Kutta 4th-order method.

Other experimental details, including the materials, instrumentation, synthesis, surface modification of nanoparticles, rate equation modelling and characterizations, are provided in Supplementary Information.

News