Precisely tailored shell thickness and Ln3+ content to produce multicolor emission from Nd3+-sensitized Gd3+-based core/shell/shell UCNPs through bi-directional energy transfer

Lanthanide (Ln3+)-doped upconversion nanoparticles (UCNPs) have been paid great attention as multiplexing agents due to their numerous uses in biological and clinical applications such as bioimaging and magnetic resonance imaging (MRI), to name a few. To achieve efficient multicolor emission from UCNPs under single 808 nm excitation and avoid detrimental cross-relaxations between the Ln3+ activator ions (positioned in either the core and/or shell in the core/shell), it is essential to design an adequate nanoparticle architecture. Herein, we demonstrate the tailoring of multicolor upconversion luminescence (UCL) from Nd3+-sensitized Gd3+-based core/shell/shell UCNPs with an architecture represented as NaGdF4:Tm3+(0.75)/Yb3+(40)/Ca2+(7)/Nd3+(1)@NaGdF4:Ca2+(7)/Nd3+(30)@NaGdF4:Yb3+(40)/Ca2+(7)/Nd3+(1)/Er3+(X = 1, 2, 3, 5, 7) [hereafter named CSS (Er3+ = 1, 2, 3, 5 and 7 mol%)]. Such UCNPs can be excited at a single wavelength (∼808 nm) without generation of any local heat. Incorporation of substantial Nd3+-sensitizers with an appropriate concentration in the middle layer allows efficient harvesting of excitation light which migrates bi-directionally across the core/shell interfaces in sync to produce blue emission from Tm3+ (activator) ions in the core as well as green and red emission from Er3+ (activator) ions in the outermost shell. Introduction of Ca2+ lowers the local crystal field symmetry around Ln3+ ions and subsequently affects their intra 4f–4f transition probability, thus enhancing the upconversion efficiency of the UCNPs. By simple and precise control of the shell thickness along with tuning the content of Ln3+ ions in each domain, multicolor UCL can be produced, ranging from blue to white. We envision that our sub-20 nm sized Nd3+-sensitized Gd3+-based UCNPs are not only potential candidates for a variety of multiplexed biological applications (without impediment of any heating effect), but also can act as MRI contrast agents in clinical diagnosis.


Introduction
In view of the recent research upsurge, multicolor luminescence from photoluminescent probes is of great interest due to its widespread use in chemical and biological sciences. [1][2][3][4][5][6][7] The multicolor luminescence provides remarkable insight into molecular biological processes, and, using a number of colorencoded probes, makes a large number of biochemical entities feasible for high-throughput sequencing technologies together with brief sequencing chemistry. [8][9][10] Along with multicolor luminescence, multiplexing experiments are adequately used in human genome decoding, and biological and clinical applications. [11][12][13] To date, there are several reports available on photoluminescent probes that have been used as multiplexing agents in bio-system analysis and therapy applications. 14-18 Among them, Ln 3+ -doped UCNPs hold promise because of their unique anti-Stokes (known as upconversion) optical properties. [19][20][21][22][23][24] Typically, these Ln 3+ -doped nanoparticles are composed of sensitizers (e.g., Yb 3+ /Nd 3+ ) and activators (Er 3+ /Tm 3+ /Ho 3+ ) which are spatially distributed in an appropriate host matrix or in a dilute host-guest system. Successive absorption of near infrared (NIR) photons and/or energy transfer (ET) generally promote the activators to higher excited states and concurrently lead to the emission of higher energy radiation spanning over a wide region, from the ultraviolet (UV) to NIR spectral regions. [25][26][27][28][29] The relatively high penetration depth of NIR light in biological structures causes a minimal autouorescence background signal and low biotoxicity, which again make UCNPs ideal candidates for a wide range of biological applications such as sensing, imaging and therapy, among others. [30][31][32][33][34][35][36][37][38] Moreover, compact, inexpensive continuous-wave NIR diode lasers can be used to excite UCNPs, leading to the generation of multiphoton UCL which is an important additive to conventional multiphoton microscopy. This characteristic physiochemical property of UCNPs also allows their extensive use in biological domains. [39][40][41] Doping different types of activators (e.g., Er 3+ /Tm 3+ /Ho 3+ ) together with sensitizers (e.g., Yb 3+ ) with a precise concentration in a suitable host matrix can produce multicolor UCL under single NIR wavelength excitation at 980 nm, which corresponds to the absorption of Yb 3+ ions. 42,43 However, this strategy can lead to cross-relaxation processes between the different types of activators, resulting in lower UCL efficiency from the UCNPs. [44][45][46] Moreover, in the Yb 3+ -sensitized upconversion process, the excitation wavelength (980 nm) is strongly absorbed by water molecules in biological environment which causes local heating problems in the vicinity of cells and tissues, thus damaging them. 42,47 To address this problem, recent research has been focused on the extension of the upconversion excitation spectrum to shorter wavelengths (especially, in the vicinity of 808 nm), where water molecules do not absorb signicantly. [48][49][50][51][52] Nevertheless, efficient multicolor UCL output from Nd 3+ -sensitized UCNPs remains vague until this point.
Furthermore, the multicolor UCL from Nd 3+ -sensitized UCNPs, obtained through a combination of different types of activators (such as Er 3+ and Tm 3+ ) in a single domain, can introduce deleterious cross-relaxation quenching processes, leading to low UCL efficiency from UCNPs. Therefore, it is necessary to separate different types of activators at a nanoscale distance. Moreover, to date, Nd 3+ -sensitized core/shell UCNPs have typically been exploited to transfer the energy from a Nd 3+doped light-harvesting layer across the core/shell interface to a Yb 3+ /X 3+ (X ¼ Er, Ho or Tm) co-doped layer in one single direction. This uni-directional ET process provides no plausible scope to sensitize two different types of activators in separated layers which again leads to the decrease in UCL efficiency, knowing that the efficiency depends on the distance between the Nd 3+ -doped layer and the activator-doped layer. It is worth mentioning that the efficiency of Nd 3+ / Yb 3+ non-radiative ET is directly proportional to the number of Yb 3+ ions per Nd 3+ ion in the doped region. In fact, the ability to harvest light most efficiently by UCNPs at 808 nm is also directed by the total number of doped Nd 3+ ions. Thus, to achieve efficient multicolor UCL, it is essential to optimize the concentration of Yb 3+ ions and Nd 3+ ions in their dened domains. Recently, Hao et al. reported that the Tm 3+ and Er 3+ ions doped into sandwichstructured UCNPs of core/shell/shell NaYbF 4 :Tm 3+ @CaF 2 :-Nd 3+ @CaF 2 :Er 3+ effectively produce multicolour UCL via a bidirectional ET process. 53 However, to our knowledge, there is no report available on the use of a bi-directional ET process along with precise control of the shell thickness and proper tuning of the content of Ln 3+ ions in each domain to produce bright multicolour emission from Nd 3+ -sensitized Gd 3+ -based core/ shell/shell UCNPs. Such strategies could be a potential tool for bio-imaging and MRI dual modal applications in the near future.
In this context, we report Nd 3+ -sensitized Gd 3+ -based core/ shell/shell (CSS) UCNPs to achieve multicolor UCL under single wavelength excitation at 808 nm. In particular, NaGdF 4 has been selected as a host matrix (both as the core and the shell layer) as it allows the formation of small sized nanoparticles. In small sized UCNPs, the surface-to-volume ratio is very high and a large portion of dopant ions are located at the surface. Moreover, with proper functionalization, Gd 3+ -based core/shell/ shell UCNPs could effectively serve as MRI contrast agents in medical diagnosis. Ca 2+ was chosen as it has lower valence than Gd 3+ ions. Upon replacement of some Gd 3+ ions by Ca 2+ ions, a charge imbalance in NaGdF 4 occurs which lowers the local crystal eld symmetry around Yb 3+ , Er 3+ and Tm 3+ ions and subsequently affects the intra 4f-4f transition probability of the Ln 3+ ions, thus enhancing the upconversion efficiency. 34,[54][55][56] The Nd 3+ sensitizer is doped at a high concentration (30 mol%) in the middle layer to effectively and efficiently absorb the excitation light in this region and simultaneously conveys it to core and outermost shell in a bi-directional manner to emit efficient multicolor emissions from the nanoparticles. Tm 3+ and Er 3+ activators are separated into the inner core and the outermost shell, respectively, to prevent deleterious crossrelaxation processes between them. Moreover, the core/shell/ shell nanoconguration allows simultaneous non-radiative ET from the middle layer to both the Tm 3+ -doped core and Er 3+doped outermost shell. Since high concentrations of Yb 3+ ions (40 mol%) are incorporated in both the outermost shell and core domain, efficient core/shell interface energy migration takes place from the Nd 3+ ions to the Yb 3+ ions, which then sensitizes the activator ions (both Er 3+ and Tm 3+ ) to produce bright UCL. A lower doping of Nd 3+ (1 mol%) in the outermost shell and in the core region is expected to assist the energy migration across the core/shell interfaces through a resonant ET process between Nd 3+ ions. We demonstrated that the CSS (Er 3+ ¼ 1, 2, 3, 5 and 7 mol%) UCNPs are able to emit efficient multicolor UCL, ranging from blue to white under laser excitation at 808 nm. Tailoring the shell thickness, along with a precise control of the concentration of sensitizers and activators in their dened domains, allows us to achieve such tunable multicolor UCL from our designed sub-20 nm sized core/shell/shell UCNPs.

Methods
Synthesis of core NaGdF 4 :Tm 3+ (0.75)/Yb 3+ (40)/Nd 3+ (1)/ Ca 2+ (7). The core NaGdF 4 :Tm 3+ (0.75)/Yb 3+ (40)/Nd 3+ (1)/Ca 2+ (7) UCNPs were synthesized following a modied co-precipitation method according to a literature report. 57 In a typical experiment, 4 mL aqueous solution (0.2 M) of LnCl 3 $6H 2 O (Ln ¼ Gd, Yb, Nd and Tm) and CaCl 2 were added to a 100 mL three-necked ask containing oleic acid (6 mL) and 1-octadecene (14 mL). The resulting mixture was heated to 150 C and kept for 1 h to form lanthanide oleate complexes and then cooled down to room temperature. Subsequently, a methanol solution (10 mL) containing NH 4 F (3.25 mmol) and NaOH (2.0 mmol) was added and the mixture was stirred at 50 C for 30 min. The temperature of the reaction mixture was then increased to 110 C to remove methanol. Upon removal of methanol and keeping the reaction mixture at 110 C under vacuum for 15 min, the solution was heated to 300 C under a N 2 atmosphere and kept for 30 min before cooling down to room temperature naturally. The resulting nanoparticles were precipitated out by the addition of ethanol, collected by centrifugation, washed with cyclohexane, ethanol and methanol and nally redispersed in 4 mL cyclohexane.
Synthesis of core/shell/shell [CSS (Er 3+ ¼ 1, 2, 3, 5 and 7 mol%)]. In a typical experiment, a 4 mL solution (0.2 M) of LnCl 3 .6H 2 O [Ln ¼ Gd, Nd, Yb, Er (X ¼ 1,2,3,5,7)] and CaCl 2 in water was added to a 100 mL three-necked ask containing oleic acid (6 mL) and 1-octadecene (14 mL). The mixture was then heated to 150 C and kept for 1 h with magnetic stirring and then cooled down to 50 C naturally. The CS (Nd 3+ ¼ 30 mol%) UCNPs in 4 mL cyclohexane were then added along with a 10 mL methanol solution of NH 4 F (3.25 mmol) and NaOH (2.0 mmol). The resulting mixture was stirred at 50 C for 30 min. The temperature of the reaction mixture was increased to 110 C to remove the methanol. Upon removal of the methanol and keeping the reaction mixture at 110 C under vacuum for 15 min, the solution was then heated to 300 C under a N 2 atmosphere and kept for 1 h before cooling down to room temperature naturally. The resulting nanoparticles were precipitated out by the addition of ethanol, collected by centrifugation, washed with cyclohexane, ethanol, and methanol and nally redispersed in 4 mL cyclohexane. Different shell thickness controlled CS (Nd 3+ ¼ 30 mol%) and CSS (Er 3+ ¼ 5 mol%) UCNPs were synthesized following identical procedures to those mentioned above, except for a change in the amount of shell precursors and methanolic solution of NH 4 F and NaOH.

Characterization
Structure and morphology. Fourier transform infrared (FTIR) spectra were collected using a PerkinElmer Spectrum 1000 FTIR spectrometer with a resolution of 2 cm À1 and averaged over four scans. Room temperature optical absorption spectra of the samples were recorded on a Varian model Cary 5000 spectrophotometer. The samples were taken in a 3 mL quartz cuvette (path length, 2 cm). The XRD measurements were performed on a powder X-ray diffractometer (Rigaku-RINT 2000) using Ni-ltered Cu Ka X-ray radiation (1.540Å) from a rotating Cu anode operated at 30 mA and 30 kV. The diffracted X-rays in the 2-theta range of 10-60 were monitored with a linear D/Tex detector at a scan rate of 1 min À1 . TEM analyses of the samples supported on holey carbon-coated copper grids were carried out using an FEI TECNAI (G2 F20) transmission electron microscope operated at an accelerating voltage of 200 kV. Before the imaging, each sample was further diluted in cyclohexane and sonicated for 1 min. One drop of the dispersion was deposited on a carbon-coated copper grid and the solvent was allowed to slowly evaporate.
Optical properties. Upconversion emission spectra of the core, core/shell and core/shell/shell UCNPs were acquired using a Horiba Jobin Yvon spectrouorometer (model uorolog-3 FL3-122) equipped with a photomultiplier tube (model R 928 P, Spex) sensitive between 250 and 850 nm. An 808 nm laser diode (from DMC, Brazil) coupled with an optical ber (200 mm diameter) was utilized as the excitation source. All spectra were acquired in identical fashion under the same experimental conditions. The laser power used in all measurements was 7 W cm À2 , determined using a Newport power meter, model 2935-C.
Monodispersity in the size of these core, CS and CSS UCNPs is evident from the size distribution histograms shown in Fig. S1. † Along with the size increase (core / CS / CSS), there is hardly any change in the morphology observed, indicating the homogeneous nature of shell growth over core and CS UCNPs. The high resolution TEM results further supported the formation of high quality hexagonal phase core, CS and CSS UCNPs (Fig. S2 †). The qualitative chemical composition of these UCNPs was studied by energy-dispersive X-ray spectroscopy (EDS) (Fig. S3 †). The EDS spectra of the samples show the presence of Gd, Nd, Ca, Yb, and Tm, Gd, Nd, Ca, Yb, and Tm and Gd, Nd, Ca, Yb, Tm, and Er for core, CS and CSS UCNPs, respectively, implying the layer-by-layer growth of shells on core nanoparticles. The capping of core, CS and CSS UCNPs with oleic acid onto the surface of the same is conrmed by the appearance of strong carbonyl stretching vibration near 1500 cm À1 (Fig. S4 †). This frequency is much lower than that observed for the free oleic acid molecules (1700 cm À1 ), indicating strong binding of the ligands (oleic acid) onto the surface of the nanoparticles.

Study of optical properties
The presence of Yb 3+ ions (along with 1 mol% Nd 3+ ions) in core, (Yb 3+ + Nd 3+ ) ions in CS (30 mol% Nd 3+ ) and (Yb 3+ + Nd 3+ ) ions in CSS (5 mol% Er 3+ ) UCNPs is conrmed by the presence of their characteristic peaks in the electronic absorption spectra of the samples (Fig. S5 †). In comparison to core UCNPs, the absorption intensity of the characteristic peaks of Nd 3+ ions increases with increase in the Nd 3+ concentration in the CS and CSS UCNPs, suggesting efficient incorporation of Nd 3+ ions in the shell of CS and in the middle layer of CSS UCNPs. To justify the importance of a Nd 3+ -enriched shell to efficiently harvest light at 808 nm for simultaneous sensitization of both Tm 3+ and Er 3+ activators to produce bright multicolor UCL, we rst prepared the core-NaGdF 4 :Tm 3+ (0.75)/Yb 3+ (40)/Nd 3+ (1)/Ca 2+ (7), CS (30 mol% Nd 3+ ) and CSS (5 mol% Er 3+ ) UCNPs. Fig. 2 shows the upconversion emission spectra for these core and CS UCNPs obtained using 808 nm laser excitation. The UCL peaks centered around 451 and 477 nm are assigned to the 1 D 2 / 3 F 4 and 1 G 4 / 3 H 6 transitions of Tm 3+ ions, respectively. The calculated integrated UCL intensity of the blue (442-505 nm) emission of CS UCNPs is 77 times higher than that of core UCNPs. The reason behind this observation might be explained on the basis of two factors. First, surface related harsh quenching of the core NaGdF 4 :Tm 3+ (0.75)/Yb 3+ (40)/Nd 3+ (1)/Ca 2+ (7) is diminished by coating the NaGdF 4 :Nd 3+ (30)/Ca 2+ (7) shell over the former. Secondly, the high concentration of Nd 3+ ions (30 mol%) in the shell of the core/shell UCNPs might harvest the light at 808 nm more efficiently and subsequently transfer it to the core region to sensitize Yb 3+ ions and simultaneously the Tm 3+ ions (activator) to produce blue emission.
To understand the effect of shell thickness on the UCL intensity from CS (30 mol% Nd 3+ ), the core particles were coated with a shell of variable thickness (i.e., 1.3, 1.8, 2.5, 3.9 and 4.8 nm) using epitaxial growth ( Fig. 3C and D). As shown in Fig. 3A, the luminescence intensity rst increases, reaching a maximum Scheme 1 Schematic illustration of the synthesis of CSS UCNPs following a seed-mediated layer-by-layer growth method.  for a shell thickness of 2.5 nm, and then decreases. From  Fig. 3B, it could be noted that the calculated integrated intensity of the blue (462-505 nm) emission of CS UCNPs with 2.5 nm shell thickness is 7.6 and 3.5 times higher than that of CS nanoparticles with 1.3 and 4.8 nm shell thickness, respectively. This result clearly suggests that shell thickness plays a pivotal role in the UCL intensity of CS UCNPs. Thinner shells (i.e., shell thickness < 2.5 nm) allow the activator (Tm 3+ , 0.75 mol% in the core) and the sensitizer (Nd 3+ , 30 mol% in the shell) to come closer in proximity. Hence, a good overlap between emission of Tm 3+ ions and the absorption of Nd 3+ ions could be expected which assures an efficient back ET from Tm 3+ ions to Nd 3+ ions, leading to a decrease in blue UCL from the CS UCNPs. On the other hand, with an increase in the shell thickness (>2.5 nm), though the shell with 30 mol% Nd 3+ efficiently harvests excitation light, it does not allow it to migrate effectively to Yb 3+ ions (in core) across the core/shell interfaces to produce efficient blue emission from Tm 3+ (activator) ions in the core. In fact, such CS UCNPs with a thicker shell (>2.5 nm) enable excitation energy to travel a long Nd 3+ -Yb 3+ distance, which again leads to the decrease in the UCL intensity from the nanoparticles.
Similar to CS (30 mol% Nd 3+ ), to examine the effect of shell thickness on the UCL intensity from CSS (5 mol% Er 3+ ), CS (30 mol% Nd 3+ ) was used as a seed for the epitaxial growth of outermost shells with different thickness (1.2, 2.5, 3.5, 5 and 5.7 nm) (Fig. 4C and D). From Fig. 4A, a shell thickness of 3.5 nm results in the highest UCL. From Fig. 4B, it can be clearly observed that the calculated integrated intensity of the green (536-570 nm) emission of these CSS UCNPs with 3.5 nm shell thickness is 2.3 and 4.5 times higher than that of CSS nanoparticles with 1.2 and 5.7 nm outermost shell thicknesses, respectively. A similar trend was observed for red (630-680 nm) emission too. Thinner outermost shells (i.e., shell thickness 1.2 and 2.5 nm) allow the activator (Er 3+ , 5 mol% in the outermost shell) and the sensitizer (Nd 3+ , 30 mol% in the middle shell layer) to come closer in proximity. Therefore, in this case too, a good overlap between emission of Er 3+ ions and the absorption of Nd 3+ ions can be expected which manifests itself in the form of an efficient back ET from Er 3+ ions to Nd 3+ ions, leading to the decrease in UCL (green + red) from the CSS UCNPs. On the other hand, with an increase in the outermost shell thickness (>3.5 nm), the shell with 30 mol% Nd 3+ , despite efficiently harvesting excitation light, does not allow it to migrate efficiently to Yb 3+ ions (40 mol%, outermost shell) concurrently to produce green and red emissions from Er 3+ (activator) ions in the outermost shell.
Moreover, like the CS UCNPs, the CSS UCNPs with a thicker shell (>3.5 nm) make the excitation energy travel a long Nd 3+ -Yb 3+ distance, which again leads to a decrease in the green and red UCL intensity from the nanoparticles. The UCL spectra and integrated emissions in Fig. 4A and B, respectively, are
increases in the shell. This enhanced interaction consumes the excitation energy, leading to less efficient energy migration across the core/shell interface to the Tm 3+ -doped core. We believe that a similar phenomenon might occur in the Er 3+doped outermost shell in CSS UCNPs. Thus, multicolor UCL intensity from the UCNPs is greatly reduced. Moreover, a good overlap between the emission of Tm 3+ and Er 3+ ions with the absorption of Nd 3+ ions ensures an efficient back ET from Tm 3+ or Er 3+ ions to Nd 3+ ions which again leads to a decrease in UCL from the nanoparticles. Upon variation in the concentration of Nd 3+ ions in the shell, the crystalline phase and size of the CS UCNPs remain intact (Fig. S9 †). It is worth mentioning that the coating of at least one inert shell over CS UCNPs is necessary to preserve excitation energy and simultaneously achieve strong luminescence from the nanoparticles. In this work, we did not follow the growth of any inert shell over CS (30 mol% Nd 3+ ) UCNPs as our goal was to allow the excitation energy to migrate along the core/shell interface to both the core and the outermost shell to produce efficient multicolor UCL. Though Nd 3+ -Nd 3+ energy migration is not reported in the literature, using lower doping of Nd 3+ ions (1 mol%) in the core and outermost shell is due to two reasons: (i) to systematically follow the UCL spectra of UCNPs upon growth of shells in a layer-by-layer fashion, under single 808 nm excitation and (ii) to help the energy transfer (Nd 3+ -Yb 3+ ) across the core/shell interfaces through a resonant ET process between Nd 3+ ions.
It is interesting to note that, in the core and outermost shell, a relatively high concentration (40 mol%) of Yb 3+ ions was used in our designed CSS (5 mol% Er 3+ ) UCNPs. To examine the impact of Yb 3+ concentration in the outermost shell on the complete UCL spectrum of these CSS UCNPs, we varied the concentration of Yb 3+ ions in the outermost shell, while the Yb 3+ concentration (40 mol%) in the core was kept xed. As shown in Fig. 7A, maximum green and red emission intensities are observed with an increase in Yb 3+ ions concentration (in the outermost shell) up to 40 mol%, whereas the intensity of blue emissions remains almost unchanged. From Fig. 7B, it is clear that the calculated integrated intensity of the green (536-570 nm) emission of CSS UCNPs with 40 mol% Yb 3+ concentration in the outermost shell is 4.4 and 1.2 times higher than that of CSS nanoparticles with 10 and 50 mol% Yb 3+ concentration in the outermost shell, respectively. The change in the calculated integrated intensity of red (630-680 nm) emission was similar to that of green emission.
Furthermore, we examined the impact of Yb 3+ ions in the core on the complete UCL spectrum of our designed CSS UCNPs. We varied the concentration of Yb 3+ ions in the core while xing its concentration at 40 mol% in the outermost shell. As shown in Fig. 7C, as the concentration of Yb 3+ ions is increased, the UCL intensity of blue emission increased concomitantly, reaching a maximum at 40 mol% concentration of Yb 3+ ions, while hardly any change in the green and red emission intensity was observed.
From Fig. 7D, it is quite clear that the calculated integrated intensity of the blue (462-502 nm) emission of CSS UCNPs with 40 mol% Yb 3+ concentration in the core is 2.5 and 1.6 times higher than that of CSS nanoparticles with 10 and 50 mol% Yb 3+ concentrations in the core, respectively. In fact, the use of a high concentration of Yb 3+ ions is expected to strongly affect the UCL intensity of the nanoparticles as Yb 3+ -mediated sub-lattice enables excitation energy migration from the primary sensitizer to the activator. Upon increase in Yb 3+ concentration, the total absorption cross-section of the same increases which again leads to the enhanced efficiency of ET from Nd 3+ (middle layer domain) to Tm 3+ (core region) and Er 3+ (outermost shell). Moreover, a higher concentration of Yb 3+ ions resulted in shorter Yb 3+ -Yb 3+ interionic distance within the sublattice ensuring an efficient ET from Nd 3+ to both the Tm 3+ -doped core and Er 3+ -doped outermost shell in CSS UCNPs. To our knowledge, literature reports suggest that the ET rate is proportional to r À6 (where, r is the average donor-acceptor distance) within the dipole interaction. 58,59 Based on this fundamental, we assume that a suitable shell thickness (2.5 nm for CS and 3.5 nm for CSS, in our work) and high concentration of Nd 3+ ions in the middle layer along with a high concentration of Yb 3+ ions in both the core and outermost shell not only allow our designed UCNPs to migrate excitation energy directionally across the core/shell interface but also prevent the excitation energy from travelling long Nd 3+ -Nd 3+ , Yb 3+ -Yb 3+ and Nd 3+ -Yb 3+ distances. However, when the Yb 3+ -Yb 3+ interionic distance exceeds the optimal value, back ET might occur from Tm 3+ or Er 3+ to Yb 3+ ions, leading to the reduction of the UCL intensity from the nanoparticles. In this point of view, a high concentration of Yb 3+ ions played a pivotal role in limiting the possibility of producing unwanted cross-relaxation pathways, while also affecting the process of bridging efficient ET from Nd 3+ (middle layer) to Tm 3+ (core region) and Er 3+ (outermost shell) to produce efficient multicolor UCL from CSS UCNPs. Now, to check whether the use of a higher concentration of Yb 3+ ions leads to the formation of bigger sized nanoparticles and simultaneously affect the UCL intensity of the same, we performed TEM measurements of core nanoparticles where the Yb 3+ concentration was varied from 20 to 50 mol% while the concentration of other ions was kept unchanged. Almost no change in the size of the core nanoparticles was observed (Fig. S10 †). Hence, from Fig. 7, we believe that the change in the UCL intensity of CSS (5 mol% Er 3+ ) UCNPs is not related to the variation in size of the core but with the variation of Yb 3+ ions in the core as well as in the outermost shell.
In addition, in our work, the Ca 2+ ion has been chosen as dopant in each domain as the UCL intensity (438-503, blue emission) of the CS (30 mol% Nd 3+ ) UCNPs with 7 mol% Ca 2+ (in both core and shell) showed an around 1.4 times enhancement as compared to the same sample prepared without Ca 2+ (Fig. S11 †).
Very small sized NaGdF 4 UCNPs possess a very high surfaceto-volume ratio and, at the same time, a large portion of dopant ions occupy positions at the surface of the nanoparticles. Thus, the excitation energy is easily quenched either by surface defects or surface capping ligands. Therefore, it is essential to modify the host lattice and this is possible while Ca 2+ ions are introduced into the NaGdF 4 core (and even into the shell). Due to the lower valence, when Ca 2+ is introduced to replace Gd 3+ , a charge imbalance occurs in NaGdF 4 . In addition, F À vacancies play an important role in charge compensation inside the crystal lattice. This induces the formation of transient electric dipoles with positive poles pointing outward on the grain surfaces. These transient electric dipoles accelerate the diffusion of ions for the crystal growth, needing F À ions from the solution to the grain and thus promoting the growth of NaGdF 4 . In this case, the growth of NaGdF 4 nanoparticles is indeed a diffusion-controlled process, resulting in better size-uniformity and fewer surface defects which lead to the enhancement of the UCL intensity of the UCNPs. Moreover, the formation of F À vacancies induced by the substitution of Gd 3+ with Ca 2+ in the host lattice lowers the local crystal eld symmetry around Yb 3+ , Nd 3+ , Er 3+ and Tm 3+ ions, which in turn affects the intra 4f-4f transition possibility of Ln 3+ ions and subsequently enhances the UCL intensity from the CSS UCNPs.
Precise control of the concentration of Er 3+ (activator) ions in the outermost shell of CSS UCNPs makes them potential candidates to produce efficient multicolor UCL output. There is hardly any change in the size (Fig. S12 †) and size distribution (Fig. S13 †) observed for CSS UCNPs with varying Er 3+ concentration in the outermost shell. Moreover, it is evident from Fig. 9A that the appropriate tuning of the concentration of Er 3+ ions in the outermost shell plays a vital role in the change in relative intensities of the green ( 4 S 3/2 / 4 I 15/2 ) and red ( 4 F 9/2 / 4 I 15/2 ) emissions of Er 3+ ions, resulting in the continuous production of UCL colors ranging from blue to white. Such spectral changes could also be observed by the naked eye (Fig. 9B, right side). Furthermore, the calculated chromaticity coordinates (Fig. 9B) reveal a similar trend of the emission colors as perceived through spectroscopic study and naked eye visualization. It is worth mentioning that control in the use of each activator (e.g., Tm 3+ or Er 3+ ) in the CSS UCNPs allows manipulating and producing different UCL spectra with different colors. From Fig. S14, † it is quite clear that the core/ shell/shell UCNPs (analogues to CSS and without Er 3+ in the outermost shell) show pure blue emission at 451 and 477 nm from Tm 3+ ions positioned in the core. However, core/shell/ shell UCNPs (analogues to CSS and without Tm 3+ in the core) exhibit both green (540 nm) and red (650 nm) emissions, which originate from Er 3+ ions present in the outermost shell. On the other hand, our designed CSS (5 mol% Er 3+ ) shows a three primary color (blue, green and red) blended spectrum responsible for generating white UCL light.
to that of Er 3+ ions. However, the ratio between the integrated area of green + red (510-570 nm + 630-680 nm, Er 3+ ) and blue (436-500 nm, Tm 3+ ) regions in the UCL spectrum of the same CSS UCNPs was found to be 2.4 indicating higher contribution from Er 3+ . This is only possible if the ET from Nd 3+ to Er 3+ is more compared to Nd 3+ to Tm 3+ in the UCNPs.
Furthermore, to investigate the number of photons involved to produce white UCL, the power dependence of UCL was measured from CSS (5 mol% Er 3+ ) UCNPs. Fig. S16 † shows that the number of photons involved in multicolor UCL is 3 (slope, 3.01), 2 (slope, 1.98) and 2 (slope, 1.76) for blue, green and red emissions, respectively. These results are in good agreement with the literature reports.

Conclusion
In summary, we have successfully illustrated multicolor UCL from Nd 3+ -sensitized Gd 3+ -based core/shell/shell UCNPs under biologically relevant, single wavelength excitation (808 nm) which, unlike 980 nm excitation, does not cause biological heating problems that could damage cells and tissues. The separation of Tm 3+ (in the core) and Er 3+ (in the outermost shell layer) in the core/shell/shell UCNPs effectively alleviates the deleterious cross-relaxation processes between them. The doping of a Nd 3+ sensitizer (30 mol%) in the middle layer, on the other hand, enables efficient harvesting of excitation light and concurrent migration of the energy across the core/shell interfaces to both Tm 3+ in the core and Er 3+ in the outermost shell layer. Precise control of the shell thickness and judicious manipulation of Ln 3+ ions (activators and sensitizers) along with their spatial distribution in their dened layer domain provide a broad range of multicolor UCL (from blue to white) through a bi-directional ET process. Such tuneable multicolor UCL makes these sub-20 nm sized Nd 3+ -sensitized Gd 3+ -based core/shell/shell UCNPs potential candidates for bioimaging and MRI dual modal applications.

Conflicts of interest
There are no conicts to declare.