Upconversion NaYF 4 :Yb:Er nanoparticles co-doped with Gd 3+ and Nd 3+ for thermometry on the nanoscale †

In the present work, the upconversion luminescence properties of oleic acid capped NaYF 4 :Gd 3+ :Yb 3+ :Er 3+ upconversion nanoparticles (UCNP) with pure b crystal phase and Nd 3+ ions as an additional sensitizer were studied in the temperature range of 288 K < T < 328 K. The results of this study showed that the complex interplay of di ﬀ erent mechanisms and e ﬀ ects, causing the special temperature behavior of the UCNP can be developed into thermometry on the nanoscale, e.g. to be applied in biological systems on a cellular level. The performance was improved by the use of Nd 3+ as an additional dopant utilizing the cascade sensitization mechanism in tri-doped UCNP.


Introduction
Temperature is a fundamental parameter of key importance in many different elds of science and technology. Respective sensors are widely used in on a daily basis in climate and marine research as well as in chemistry, biology, metrology, and medicine. [1][2][3] It is estimated that the share of temperature sensors amounts to as much as 75-80% of the world's sensor market. 4 Indeed, current technological demands like micro-uidics or nano medicine have reached a point at which a spatial resolution on the submicron scale is needed. 5 For example the mapping of the temperature in living cells, i.e. cancer cells compared to normal tissue have a higher temperature due to the increased metabolic activity, which strongly improves the perception of their pathology and physiology and in turn can help to optimize diagnosis and therapeutic approaches, e.g., hydrothermal treatment or photodynamic therapy. In particular a thermometer capable of subdegree temperature resolution as well as integration on a cellular level could provide a powerful new tool in many areas of biochemical and biophysical research. 5,6 Many promising approaches for local temperature sensing are being explored at present such as Raman spectroscopy, 6 scanning probe microscopy 6 and luminescent based measurements using organic dyes, 7,8 nanomaterials 9-11 such as quantum dots (QD) or proteins. 8 Many of these already existing methods suffer on several drawbacks such as low sensitivity, low reproducibility or systematic errors due to uctuations in the luminescence rate or local chemical environment. 10 Our approach to nanoscale thermometry uses luminescence upconversion nanoparticles (UCNP), which are suitable for investigation in different biological matrices. The UCNP are excited with light in the near infrared range (NIR), where interferences from background uorescence can be neglected. 12 In the chosen excitation wavelength range between 795 nm < l ex < 976 nm many biological matrices are nearly transparent. 13 Consequently, the penetration depth in this spectral range is very high for a human skin or in blood (see Fig. 1) and a high sensitivity in sensing applications without additional sample preparations can be established. 13 One of the rst promising approaches for upconversion nanoparticle based thermometry Fig. 1 Absorption spectra of diluted blood. on nanoscale was developed by Zink et al. 14 They used a dual core (made of UCNP and smaller superparamagnetic nanocrystal) mesoporous silica nano-particle that acted as nanothermometer as well as nanoheater. An in-depth investigation of the potential of UCNP as nanothermometer was carried out by Woleis et al. using different lanthanide combination and capping agents. 15 Here, the best results were found for NaYF 4 :Yb 3+ :Er 3+ core/shell nanoparticles. Moreover, UCNP can be developed into multimodal probes, e.g., with additives like gadolinium (Gd 3+ ) ions UCNP can be used for magnet resonance imaging or with the appropriate functionalization for optical imaging or assay applications. [16][17][18] Apart from the excitation at l ex ¼ 976 nm Han et al. report on UCNP with additional Nd 3+ doping which could be excited at l ex ¼ 800 nm. This type of UCNP showed a better biocompatibility due to lower impact on biological tissues and a higher penetrability for the excitation light. 19 In the present work, the potential of oleic acid-capped NaYF 4 :Gd 3+ :Yb 3+ :Er 3+ upconversion nanoparticles (UCNP) for thermometry on nanometer scales was studied in the temperature range of 288 K < T < 328 K. Both, the overall luminescence intensity and the ratio of luminescence bands were found to be highly sensitive on temperature. For the temperature determination the intensity ratio of the luminescence signal in the green Er 3+ luminescence bands ( 2 H 11/2 / 4 I 15/2 (G1) and 4 S 3/2 / 4 I 15/2 (G2)) was used. The relative sensitivity (S r ) with 1.49% K À1 of the UCNP under investigation is one of the highest S r value which found in literature for NaYF 4 :Yb 3+ :Er 3+ type nanoparticles or other host lattices like GdVO 4 :Yb 3+ :Er 3+ or Na 2 Y 2 B 2 O 7 :Yb 3+ :Tm 3+ . 20,21 The overlapping of the excitation light at l ex ¼ 976 nm with the absorption of water (see Fig. 2), which may induce heating damage in cells and tissues, is a drawback e.g., for nanoparticlebased imaging or for deep tissue imaging (this experiments require high excitation energy and long-term excitation). To overcome this limitation, Nd 3+ was used as additional lanthanide. Nd 3+ has an absorption maximum around l ¼ 800 nm, which can be used for the initial excitation of the UCNP. At this excitation wavelength the intrinsic water absorption is one order of magnitude lower compared to l ex ¼ 976 nm resulting in less heating of the sample by the laser irradiation (see Fig. 2). Subsequently a better biocompatibility and higher penetration depth in aqueous systems can be envisaged. Colloidal tri-doped (Yb 3+ , Er 3+ and Gd 3+ ions) UCNPs with Nd 3+ were successfully generated and optimized for maximum upconversion efficiency with an excitation wavelength of l ex ¼ 800 nm.
2.2 Synthesis of NaYF 4 :Yb 3+ :Gd 3+ :Er 3+ (UCNP) and UCNP:Nd 3+ (UCNP Nd ) The UCNP were synthesized according to a previously reported procedure in which a thermal decomposition method with oleic acid as a capping agent was used. 22,23 The synthesis was designed to keep a constant overall amount of lattice ions (Y 3+ , Yb 3+ , Er 3+ , Nd 3+ , and Gd 3+ ) in the different samples to study the effect of neodymium. The concentration of the lanthanides (Ln) Er 3+ and Yb 3+ was constant for all samples whereas parts of the Y 3+ ions were replaced by Nd 3+ ions. GdCl 3 hexahydrate (0.61 mmol), ErCl 3 (0.06 mmol), YbCl 3 (0.34 mmol), NdCl 3 (0.02 mmol) and YCl 3 (0.99 mmol) were dissolved under Ar atmosphere in a mixture of oleic acid (13.4 g) and octadecen (35 mL) under stirring. The solution was evacuated (1 mbar) for 45 minutes until evolution of gas had stopped. The reaction mixture was heated to 140 C under Ar atmosphere until a yellowish clear solution occurs. Aer the solution had been cooled to 45 C, ammonium uoride (300 mg) and sodium hydroxide (150 mg) was added to the reaction mixture under stirring until a clear solution was formed. The solution was heated to 310 C for 90 minutes using a heating mantle. A discoloration (yellow/brown) of the reaction mixture as well as a white precipitation occurs. When the reaction mixture reached room temperature, the nanoparticles were separated via centrifugation (6000 rpm for 25 min) and further puried by several redispersion and centrifugation steps in ethanol. The obtained white powder was dissolved in cyclohexane and ltrated using a 0.2 mm PTFE syringe lter. The as-synthesized UCNPs disperse readily in non polar solvents such as cyclohexane, forming a clear colloidal solution. The oleic acid capped UCNP stored at room temperature were colloidal stable for several months.

Synthesis of water soluble AEP-capped UCNP
In a typically ligand exchange reaction 150 mg AEP was diluted in solvents mixture of ethanol and ultrapure water (4 mL/6 mL) and 20 mg UCNP dispersed in 5 mL chloroform were added slowly drop wise. 24 The reaction mixture was stirred for 48 h at room temperature, whereas the UCNP moved obviously from the chloroform to the watery phase. Aer the phase separation the watery phase was centrifuged at 6000 rpm for 25 minutes and the obtained modied UCNP were redispersed in ultrapure water. The AEP-modied UCNP were stored under exclusion of light and different temperature between À10 and 20 C to get information of the long-term colloidal stability.

Structural characterization
The size and morphology of as-prepared UCNPs were observed on a JEM 1011 transmission electron microscope (Jeol Ltd, Tokyo, Japan) (TEM) using a wolfram hairpin cathode, an accelerating voltage of 80 kV and a molybdenum panel. The measurements were recorded using a side-mounted Olympus Mega View G2 (Olympus Germany GmbH, Hamburg, Germany). Particle size characterizations were also carried out with dynamic light scattering (DLS) by using a ZETASIZER Nano ZS (Malvern Instruments Ltd, Herrenberg, Germany) as well. As light source a He-Ne laser at l ¼ 633 nm was used.
X-ray powder diffraction patterns were obtained using a D5005 (Siemens AG, Munich, Germany) in a range of 3-70 /2q with divergence aperture, scattering ray aperture and graphite monochromatized Cu Ka radiation (l ¼ 0.15406 nm). The scanning step was 0.02 /2q with a counting time of 4 s per step. The nanocrystalline domain sizes were calculated using the Debye-Scherrer equation (eqn (1)): D is the domain size to be determined, l is the wavelength of the X-ray, B is the FWHM of the diffraction peak of interest and q is the angle of the corresponding diffraction peak.

Upconversion luminescence measurements at various temperature (288 to 328 K)
In order to study the upconversion luminescence properties at various temperature the experimental set up mentioned above was extended by a water-cooled Peltier element-based temperature adjustable sample holder (temperature controller GR2012 itron 32, JUMO GmbH & Co, Fulda, Germany). In order to ensure the temperature stability the samples were temperated for 15 minutes at the certain temperature. As an additional control a conventional temperature sensor (Testo 945, Testo AG, Lenzkirch, Germany) was used to monitor the temperature of UCNP containing solution under investigation. For all spectroscopic measurements quartz cuvettes sealed with Paralm® were used.

Power dependency of upconversion luminescence at room temperature
The intensity of upconversion emission was measured as function of excitation power at l ex ¼ 976 nm (further details on equipment vide supra). The attenuation of excitation light was achieved by the use of neutral density lters (optical density (OD) 0.1-1.0). The upconversion emission intensity I UC strongly depends on the excitation power I P (see eqn (2)): 21 here n is the number of photons required to populate the emitting state of the lanthanides. The power dependence of the Er 3+ transitions 2 H 11/2 / 4 I 15/2 (G1), 4 S 3/2 / 4 I 15/2 (G2) and 4 F 9/2 / 4 I 15/2 (R) is shown in ESI Fig. 1 † using a log I P À log I UC plot, in which n was calculated from the slope.

Structural investigations
The particle size and morphology of UCNP and UCNP Nd were studied using TEM, XRD and DLS, respectively. The TEM images of the different UCNP showed that the particles were hexagonal in shape (Fig. 3). 25 In order to determine the average particle size from TEM images approximately 200 particles were included in the statistical analysis (see Table 1). In addition, DLS and XRD was used as a complementary method for particle size determination. The results of the DLS, TEM and XRD measurements are also shown in Table 1 (see ESI  Corroborating the results of the TEM images, DLS and XRD showed as well no particle size alteration upon addition of Nd 3+ , which was expected due to the very similar atomic radius of the different lanthanides used. Furthermore, it can be assumed that the UCNP/UCNP Nd are highly crystalline, due to the similar particle diameter obtained from TEM/XRD and the fact that only crystalline parts can be observed in XRD. The AEP capped UCNP Nd are 2.7 times larger than the oleic acid capped UCNP, which could be due to the different ligand on the surface resulting in an increase of the hydrodynamic radius or the possible formation of small aggregates. In comparison to the TEM investigations the particle size obtained from the DLS measurements are nearly similar. The deviations in particles size of TEM in comparison to the DLS are due to the fact that calculation algorithm for DLS is optimized for spherical particles whereas variations in shape leads to a change in the scattering behaviour and nally to an inaccuracy in the calculated particle diameter.

Upconversion luminescence spectroscopy studies
In Fig. 4 the luminescence spectra of the UCNP and UCNP Nd are shown. The optimal excitation wavelengths chosen for the samples under investigation are based on matrices of excitation emission spectra, shown in ESI Fig. 4. † The spectra were recorded in cyclohexane with l ex ¼ 976 nm and are the result of energy upconversion processes between Yb 3+ and Er 3+ ions (see Scheme 1). 26 The three most intense emission bands can be observed in the green spectral region centered at l em ¼ 525 nm ( 2 H 11/2 / 4 I 15/2 transition, G1), 545 nm ( 4 S 3/2 / 4 I 15/2 transition, G2), and in the red spectral region centered at l em ¼ 660 nm ( 4 F 9/2 / 4 I 15/2 transition, R). [27][28][29][30][31][32] The observed ne structure (Stark splitting) is induced by the crystal eld splitting due to small differences in the coordination environment. 29,32,33 The Nd 3+ containing UCNP Nd can be excited at l ex ¼ 976 nm and additionally at l ex ¼ 795 nm. For the excitation at l ex ¼ 795 nm the mechanism of the upconversion processes is extended by an initial energy transfer step between Nd 3+ and Yb 3+ . First the Nd 3+ ions are excited from the 4 I 9/2 to the 4 F 5/2 energy level, followed by a non-radiative relaxation step to 4 F 3/2 level. Originating from this energy level the energy transfer to the 2 F 5/2 level of the Yb 3+ ions can take place. In such cases the Yb 3+ acts like a "relay" between the sensitizer Nd 3+ and the activator Er 3+ . 1,[34][35][36][37][38][39] The subsequent energy transfer steps and relaxation processes from the Yb 3+ to the Er 3+ are identical to the regular UCNP. The cross-relaxation between different excited Er 3+ ions can be neglected in both cases due to the low Er 3+ ion concentration used. 31,40 The Gd 3+ ion doping was chosen to enhance the absolute upconversion luminescence intensity, due to the favoured formation of the b-phase. 41 Since the 6 P 7/2 level, which represents the next electronic state above the ground state of Gd 3+ , its emission is found in the ultra violet spectral region and is therefore much higher in energy than the relevant excited state levels of Er 3+ , Nd 3+ , and Yb 3+ , respectively. Consequently, a Gd 3+ -related luminescence quenching by energy transfer of these ions can be ruled out. The upconversion emission spectra (l ex ¼ 976 nm) shown in Fig. 3 (right) reveal no signicant differences between the UCNP and UCNP Nd investigated, except the G1, G2/R ratio, which is slightly increased. The additional doping of the host lattice with Nd 3+ ions has no signicant impact on the upconversion emission spectra. Also no  Scheme 1 Schematic energy level diagram of the upconversion mechanism of a Yb 3+ and Er 3+ dopant ion system following an excitation at l ex ¼ 976 nm. Furthermore, the upconversion mechanism of a Nd 3+ , Yb 3+ and Er 3+ dopant ion system following an excitation at l ex ¼ 795 nm is shown as well. The full lines pointing upwards represent energy absorption, the dotted lines represent energy transfer, the waved lines represents non radiative relaxation processes and the colored full lines pointing downwards represents the visible emission.
signicant differences in the shape or ne structure of the upconversion emission spectra can be seen for different excitation wavelength, due to the fact that the Nd 3+ ion doping only inuence the population of the emitting energy level of Yb 3+ . On the other hand the absolute upconversion luminescence intensity is much lower (1/10 intensity) for the excitation at 795 nm (see Fig. 4 inset). The lower upconversion intensity at l ex ¼ 795 nm is related to several additional energy transfer steps between Nd 3+ ions and Yb 3+ ions as well as relaxation steps of excited Nd 3+ ions itself. The possibility for non-radiative deactivation channels is increased, subsequently leading to a less effective upconversion.
In Fig. 5 normalized upconversion emission spectra of UCNP Nd with different capping agents at l ex ¼ 795 nm as well as 976 nm are shown. The ratio of the emission bands G1, G2 to R is decreased when the nanoparticles are capped with AEP. The change in the G1, G2 to R ratio could be connected to the new chemical environment at the surface of the nanoparticles with different phonon coupling processes inuencing the luminescence upconversion. Due to fact that the probability for the nonradiative transition 4 I 11/2 to 4 I 13/2 of Er 3+ ions is increased, the population efficiency for the energy level R is increased too, whereas of the population of G1 and G2 is decreased, because it is populated by different upconversion mechanisms. The excitation wavelength has no impact on the spectral intensity distribution including the position of the emission bands or emission band ratio as in the case of oleic acid capped UCNP in cyclohexane (see Fig. 4 above).

Temperature dependent upconversion luminescence
In ESI Fig. 5 † emission spectra of UCNP in cyclohexane (l ex ¼ 976 nm) for the temperature range of 288 K < T < 328 K are shown, which are representative for the observed temperature dependence of all UCNP and UCNP Nd investigated.
The following trends were observed for the different luminescence bands of Er 3+ : (i) the luminescence intensity of the 2 H 11/2 / 4 I 15/2 transition (G1) gradually increased with increasing temperature and (ii) in contrast to the G1 emission band, the intensity of the G2-and R-related emission bands initially decreased slightly with increasing temperature. The difference in the temperature dependence of the G1 and G2 emission bands is connected to the population pathways of the related energy levels 2 H 11/2 and 4 S 3/2 , respectively. In Scheme 2 a detail view of the respective Stark levels of 2 H 11/2 and 4 S 3/2 is shown. The Stark levels are calculated from the excitation and emission spectra of the UCNP under investigation. The population of the green emitting levels G1 and G2 ( 2 H 11/2 and 4 S 3/2 ) usually occurs by successive energy transfer processes from the excited 2 F 5/2 state of Yb 3+ ions to the Er 3+ ions exciting it rst to the 4 I 11/2 state and in a second step to the 4 F 7/2 excited state. Followed by a non-radiative relaxation process the Er 3+ ion deactivates to the 2 H 11/2 anddue to the moderate energy gap between 2 H 11/2 and 4 S 3/2 states (about 700 cm À1 )the Er 3+ ions can relax fast to the 4 S 3/2 state. Finally, the 2 H 11/2 level is repopulated via thermal agitation (see Scheme 2) resulting in the observed two Er 3+ -emission bands G1 and G2 as shown in Fig. 5 (see also Scheme 1). [28][29][30]42,43 The thermal equilibration of the two levels is fast, hence the observed intensity ratio of G1 and G2 will be dependent on the temperature.

UCNP for nanoscale thermometry
For T > 273 K the correlation of the luminescence intensity ratio (Z ¼ I G1 /I G2 ) and the temperature T can be described by an Arrhenius type equation. 34,35,44 Z is the ratio of integrated luminescence intensity originating from band G1 and G2 which are separated by the energy gap DE G1/G2 , k B is the Boltzmann constant, T is the temperature and A is a constant which depends on the spontaneous emission rate and devolution of the energy levels of the emitting states in the host material. From Fig. 6 (le) it can be seen that Z is changing with temperature, because the relative intensity of the G1-related luminescence is increasing. This is a consequence of the thermally induced re-distribution in population between the energy levels 2 H 11/2 and 4 S 3/2 (see Scheme 2). Especially for higher temperatures (T > 273 K) a distinct change with temperature was found, which is connected to the fact that the energy gap DE G1/G2 is in the order of 600-800 cm À1 depending on the host lattice. For UCNP an energy gap DE G1/G2 ¼ 789 AE 9 cm À1 was determined, which correlates very well with the data of Carnall et al. who studied the energy level assignments for Er 3+ in several host lattices. 42 Based on eqn (3) the absolute sensitivity S a and relative sensitivity S r can be obtained. 1,45-48 The temperature sensitive calculated value of S a and S r are shown in Fig. 7. The maximum value of S r of 1.37% K À1 was found at 288 K which is in the range typically found for NaYF 4 :Yb 3+ :Er 3+ or other new host lattices of upconverters like GdVO 4 :Yb 3+ :Er 3+ . 16 The temperature resolution of $0.4 K was obtained from dR/S a where dR is the standard deviation of the residuals in the polynomial interpolation of the experimental data points (temperature vs. Z ¼ I G1 /I G2 ). In Table 2 the absolute and relative sensitivity of oleic acid capped UCNP and AEP capped UCNP are shown. It is obvious that the type of ligand and the surrounding medium (cyclohexane and water) has no signicant impact on S a and S r .

UCNP Nd as optical probes for thermometry
A big challenge for measurements in "real" biological system are the different optical properties of biological tissues like skin, muscles, connective tissue or vertebral column which only enables a sharp window (700-1000 nm) for optical measurements (see Fig. 8). Based on the water and tissue transmission, which are shown in Fig. 2 and 8, wavelengths around l ex $ 800 nm are more suitable for UCNP based thermometry measurements in contrast to the typical used excitation wavelength of l ex ¼ 976 nm, at which also a heating effect by the excitation laser due to water absorption is induced and which is distinctly reduced at l ex $ 800 nm. 13 For UCNP related thermometry the excitation wavelength can be shied to l ex $ 800 nm upon codoping with Nd 3+ (vide supra).
The inuence of the excitation wavelength on the temperature within the observation volume and near surrounding was monitored using a resistance thermometer in an optical cuvette during the irradiation of the sample with light at l ex ¼ 795 nm and 976 nm, respectively (see Fig. 9 and ESI Fig. 6 †). It was found that the temperature aer irradiation at l ex ¼ 795 nm is nearly constant whereas for l ex ¼ 976 nm (excitation power $ 170 mW at both wavelength) an increase in temperature of about DT $    8 Transmission spectra of different biological tissue in a spectral range of 500-1200 nm. 13 0.6 K was found in both cases (pure water and water containing UCNP Nd ). The maximum DT was reached at $200 seconds of irradiation with l ex ¼ 976 nm (since a conventional thermometer was used to measure the temperature change in the nearby bulk phase, the temperature increase at the laser spot is larger).
For biological systems such an increase could induce cellular damage or could signicantly inuence protein-related processes. 6 On the other hand an excitation wavelength of l ex ¼ 795 nm, which is used in the case of UCNP Nd , did not show a comparable heating effect in the sample and therefore offers a promising alternative for investigations of biological systems (using UCNP for sensing purposes as well as nanothermometer). As already shown for UCNP ( Fig. 6 and 7) the absolute and relative sensitivity of UCNP Nd with different ligands and different excitation wavelength (l ex ) were calculated based on luminescent spectra at different temperatures (see Table 3). The absolute sensitivity is unaffected by the l ex applied in the measurement and by the solvent. For the relative sensitivity the same was found when comparing the different solvents (for l ex ¼ 795 nm S r was slightly smaller than for l ex ¼ 976 nm). The obtained values for S r are in good agreement with literature data reported for other host materials. 20

Conclusions
The upconversion luminescence properties of oleic acid and AEP stabilized nanoparticles (UCNP) based on a NaYF 4 host matrix, which was doped with Yb 3+ (UCNP) and Yb 3+ :Nd 3+ (UCNP Nd ) as sensitizer and Er 3+ as activator, respectively, were studied. The focus of this work was the surface modication with AEP to obtain water soluble nanoparticles as well as to investigate possible thermometry applications based on the temperature sensitivity of the upconversion luminescence. The temperature dependence of different luminescence parameters especially the spectral intensity distribution were analysed for the temperature range of 288 K < T < 328 K. Not only the overall luminescence intensity was dependent on the type of capping agent but also the spectral distribution which is effected as well by the phonon coupling possibilities at the surface of the nanoparticles. On the other hand Nd 3+ ions as additional codopant had no effect on the spectral properties of the Er 3+ related luminescence.
The possibility to excite the UCNP at l ex ¼ 795 nm could be realized by the use of Nd 3+ ions as new sensitizer with no further impact on the spectral distribution. The interplay between spectral distribution and the temperature dependence of photophysical parameters was further investigated. The intensity distribution (e.g., ratio of the green emission G1 and G2) was strongly depending on the temperature and can be used in thermometry applications. In practical applications, efficient collection of uorescence signal ensures a high signal to noise ratio (SNR) for improving the sensitivity and resolution of QD, uorescent dyes or upconversion nanoparticle. 49,50 A variety of highly promising approaches of external optical coupling structures for enhancing the excitation and emission of the luminescence from QD or uorescent dyes are being investigated using cascaded photonic crystal surfaces, which could also be a powerful tool for application based on upconversion nanoparticles. 49,50 Furthermore, the heating effect due to the irradiated excitation light in the observed sample volume was investigated. Whereas for l ex ¼ 795 nm only minor increase of the sample temperature was found, the excitation at l ex ¼ 976 nm lead to an increase up to 0.6 K which is rather high especially with regards to possible application in biological tissue. In addition to avoiding heating effects the performance of UCNP Nd are comparable to the regular UCNP with respect to absolute and relative sensitivity of temperature sensing. The intensity ratio of the different Er 3+ luminescence bands in combination with an excitation wavelength of l ex ¼ 795 nm can be envisioned to be used as a nanothermometer, e.g., to measure the temperature spatially resolved in tissues with submicron resolution.