Optimization of Temperature Sensing with Polymer-Embedded Luminescent Ru(II) Complexes

Temperature is a key parameter in many fields and luminescence-based temperature sensing is a solution for those applications in which traditional (mechanical, electrical, or IR-based) thermometers struggle. Amongst the indicator dyes for luminescence thermometry, Ru(II) polyazaheteroaromatic complexes are an appealing option to profit from the widespread commercial technologies for oxygen optosensing based on them. Six ruthenium dyes have been studied, engineering their structure for both photostability and highest temperature sensitivity of their luminescence. The most apt Ru(II) complex turned out to be bis(1,10-phenanthroline)(4-chloro-1,10-phenanthroline)ruthenium(II), due to the combination of two strong-field chelating ligands (phen) and a substituent with electron withdrawing effect on a conjugated position of the third ligand (4-Clphen). In order to produce functional sensors, the dye has been best embedded into poly(ethyl cyanoacrylate), due to its low permeability to O2, high temperature sensitivity of the indicator dye incorporated into this polymer, ease of fabrication, and excellent optical quality. Thermosensitive elements have been fabricated thereof as optical fiber tips for macroscopic applications (water courses monitoring) and thin spots for microscopic uses (temperature measurements in cell culture-on-a-chip). With such dye/polymer combination, temperature sensing based on luminescence lifetime measurements allows 0.05 °C resolution with linear response in the range of interest (0–40 °C).


Introduction
Temperature is one of the most important parameters in industrial process control, environmental monitoring, medicine practice, and biology, to name a few areas [1]. Although measurement of temperature by mechanical or electrical methods is widely established, quantification of this physical magnitude can be sometimes extremely difficult due to hostile media for the operator, restricted accessibility, and/or the presence of strong electromagnetic fields that preclude the use of conventional thermometers. Besides, when studying temperature at the micro-and nanoscale, the latter may lack spatial resolution [2]. Commercially available non-contact IR thermometry may be used under some of those particular situations provided that there is a straight free path between the measuring device and the sample point, and the sample is large enough [3]. When these requirements are not met, fiber-optic sensing offers a convenient alternative. Among the various temperature-dependent optical properties that can be used to develop a temperature optode (optical sensor), luminescence thermometry, encompassing fluorescence, and phosphorescence, is probably the most sensitive and From Equation (1), it can be inferred that the temperature sensitivity of the luminescent Ru(II) complexes increases with the accessibility to their 3 MC state. However, a very efficient crossing to this excited state leads to photolabile species, due to the anti-bonding nature of σ-type 3 MC states (e(d*)) [31]. The relative energy of the metal-centered level basically depends on the σ-donor features of the chelating ligands (average ligand field strength), while the position of the luminescent 3 MLCT manifold is determined by the energy of the lowest-lying π* orbital of the ligand set. Therefore, we propose that the synthesis of photostable yet temperature-sensitive Ru(II) indicator dyes should be feasible by selecting appropriate polyazaheterocyclic ligands for engineering heteroleptic (i.e., made of different ligands) metal complexes.

Luminescence Lifetimes
Emission decay measurements in solution were performed with a single photon timing spectrometer (SPT, Edinburgh Instruments FLS-980, Livingston, UK) equipped with a 450 nm LED head (Nanoled 450L, Horiba, Kyoto, Japan) ns-pulsed at 20 or 40 kHz as the excitation source (through a 467 nm interference filter) and a red-sensitive Hamamatsu R-955 photomultiplier cooled at −20 °C. Solutions of the ruthenium complexes in PC (7-16 μM) were deoxygenated with argon (>99.99%, Praxair, Madrid, Spain), sparging them for at least 30 min before collecting the emission decays. The luminescence lifetimes of the sensor tip were also measured as a function of temperature with the SPT spectrometer. The cuvette sample holder of the FLS-980 was replaced by a homemade fiber-optic adapter to which a (randomly) bifurcated silica waveguide bundle was connected (FiberGuide, Township, NJ, USA, 6.5 and 4.5 mm diameter at the common end and at each branch, respectively). The emission decays were fitted to a single exponential function (solutions) or to the minimum number of summed exponentials (doped polymers) to achieve a reduced χ 2 value lower than 1.1 using the Edinburgh Instruments proprietary software based on non-linear least squares fitting. Caution must be exercised with multi-exponential decay functions because they might not

Luminescence Lifetimes
Emission decay measurements in solution were performed with a single photon timing spectrometer (SPT, Edinburgh Instruments FLS-980, Livingston, UK) equipped with a 450 nm LED head (Nanoled 450L, Horiba, Kyoto, Japan) ns-pulsed at 20 or 40 kHz as the excitation source (through a 467 nm interference filter) and a red-sensitive Hamamatsu R-955 photomultiplier cooled at −20 • C. Solutions of the ruthenium complexes in PC (7-16 µM) were deoxygenated with argon (>99.99%, Praxair, Madrid, Spain), sparging them for at least 30 min before collecting the emission decays. The luminescence lifetimes of the sensor tip were also measured as a function of temperature with the SPT spectrometer. The cuvette sample holder of the FLS-980 was replaced by a homemade fiber-optic adapter to which a (randomly) bifurcated silica waveguide bundle was connected (FiberGuide, Township, NJ, USA, 6.5 and 4.5 mm diameter at the common end and at each branch, respectively). The emission decays were fitted to a single exponential function (solutions) or to the minimum number of summed exponentials (doped polymers) to achieve a reduced χ 2 value lower than 1.1 using the Edinburgh Instruments proprietary software based on non-linear least squares fitting. Caution must be exercised with multi-exponential decay functions because they might not correspond to the presence of a discrete number of luminescent species, but rather to a plurality of luminophore populations of different widths, as shown by the Edinburgh Instruments (Livingston, UK) advanced Fluorescence Analysis Software Technology (FAST).

Phase-Sensitive Luminescence Measurements
Phase-sensitive monitoring was carried out with a field-deployable fiberoptic phase-sensitive luminometer developed in our research group. The instrument is able to determine simultaneously up to four target parameters (e.g., temperature, O 2 , moisture, CO 2 ), or a single parameter in different locations, using common electronic and optoelectronic components, but specific Ru(II) polypyridyls immobilized in tailored polymer supports. The system is equipped with four 470 nm LEDs (9600 mcd, RS Amidata, Madrid, Spain) as excitation light sources, digitally modulated at 39, 78, or 156 KHz (user selectable). The optical module contains a 450 nm CS5-60 colored band-pass filter (100 nm FWHM) (Kopp Glass, Pittsburgh, PA, USA) in the excitation channel, a 503 nm dichroic filter (Edmund Optics, York, UK), and a 570 nm long-pass filter (CS2-73, Kopp), plus a plano-convex lens with 4 mm effective focal length (Edmund) in the emission channel. The excitation light is focused onto a 1000 µm multimode plastic optical fiber (Mitsubishi Cable, Tokyo, Japan) fitted with a SMA connector (Ratioplast Optoelectronics, Lübbecke, Germany). The luminescence from the four sensitive terminals is monitored with an H11901-01 compact photomultiplier (Hamamatsu Photonics, Hamamatsu, Japan). The reference signal is obtained directly from the excitation blue LEDs with one Hamamatsu S5821-03 photodiode per optical channel. The instrument configuration and data are stored in the unit and can be transferred at any time to and from a laptop computer via the RS232 or USB ports, using our Windows-based software.
The measured phase shift (φ) and the luminescence lifetime (τ), for a particular modulation frequency of the excitation light source (f ), are related through the well-known Equation (2) [40]: If the emission decay is perfectly exponential (e.g., in isotropic solutions of the luminophore), the observed φ can be used to determine the τ instead of the single photon timing technique. However, if the excited state decay is an ensemble of luminophore molecules in different microenvironments, the value of τ calculated from φ is an averaged one, which also depends on the modulation frequency and is not always coincident with the pre-exponentially weighted emission lifetime determined from SPT (see Section 3.4) [41]. All spectroscopic, emission lifetime, and luminescence phase shift measurements have been performed on fully cured (>5 days) polymer materials (see below) unless otherwise stated.
Temperature sensing with the phase shift-based luminometer was carried out in two ways: on the one hand, the temperature-sensitive polymer layers (see below) were attached to the optical fiber tips, for realizing environmental/industrial monitoring instrumentation. On the other hand, sensitive layers were deposited into organ-on-a-chip devices for monitoring the internal temperature from the outside with the fiberoptic luminometer.

Temperature Control
For the macroscopic monolithic sensor tips, temperature was kept within ±0.25 • C with a digital water bath/circulator (mod. 9110, Polyscience, Niles, IL, USA). When working at temperatures below 0 • C, a water/ethylene glycol (1:1 v/v) mixture was employed as cryogenic fluid. For testing the microscopic temperature-sensitive films, temperature was controlled with a heating plate (IKA ® Polymers 2018, 10, 234 6 of 16 RCT basic, Staufen, Germany) and monitored with a commercially available thermistor (LM335, Texas Instruments, Dallas, TX, USA).

Fabrication of the Sensor Tips
A copolymer of EHMA and MMA doped with the temperature indicator dye was prepared by dissolving 0.23 mg of [Ru(phen) 2 (4-Clp)]Cl 2 in 0.1 mL of EHMA, and mixing the solution with 3 mL of MMA containing 0.23 mg of AIBN. Round-bottomed glass tubes (50 × 6 mm 2 , 5 mm i.d.) were filled with ca. 0.5 mL of the mixture, sonicated for 10 min and heated at 60 • C for 72 h. A piece of ca. 3 mm from the bottom of the Ru-doped polymer matrix was cut after drawing the cured material out of the mold, polishing several times with sand paper of increasingly fine grain the flat side to be in contact with the optical fiber.
Sol-gel glasses were produced from a mixture of ethanol (1.2 mL), deionized water (0.56 mL), and tetraethoxysilane (1 mL), after increasing the drying temperature up to 700 • C with a laboratory oven (0.5 • C/min). After formation of the monolith, differential absorbance measurements of the supernatant showed that 4 mol of the ruthenium complex, dissolved in ethanol, was incorporated into 300 mg of this porous matrix (72 h contact time).
Sensor tips prepared from a commercial epoxy resin (see above) were fabricated by dissolving ca. 0.2 mg of the dye in 1.5 mL of 2-butanone (99%, Merck, Darmstadt, Germany, DE), and adding the same amount of the hardener and the epoxy components to this solution. Borosilicate round-bottomed glass tubes were filled with this mixture, which was allowed to polymerize at room temperature for 72 h. The Ru-doped monoliths were cut and polished, as described before for EHMA-MMA copolymer.
Poly(ethyl cyanoacrylate) (PCA) monoliths containing the temperature indicator complex were prepared by filling the above mentioned borosilicate molds with ca. 500 µL solution of the Ru(II) complex in the monomer (0.20-0.35 mg in 4 mL). Due to the relatively large solution volume, CA polymerizes slowly, so that changes in the luminescence from the dyed material were observed over the first five days (see Results and Discussion section). Therefore, all of the monoliths were allowed to cure for at least five days to obtain sensors showing a stable response. After this period, the borosilicate tube was broken to extract the solid matrix. Then it was cut to ca. 1.5 mm and thoroughly polished according to the procedure described above to yield 1-mm high domed monoliths. The latter are placed in direct contact with the distal end of the optical fiber, mounted in home-made stainless-steel ferrules after covering the outermost part of the monolith with aluminum foil to eliminate the interference of the external light on the luminescence from the dyed polymer ( Figure S1, Supplementary Materials). The application of a black painting overcoat was also successfully tested.

Temperature-Sensing Film in Microfluidic Devices
Organ-on-a-Chip devices were kindly provided by Micronit Microtechnologies (Enschede, The Netherlands) (https://www.micronit.com/products/organ-on-a-chip.html). These chips consist of three layers: a thin glass membrane-carrier sandwiched between two glass slides equipped with integrated gaskets ( Figure S2a, Supplementary Materials). The gaskets are made of a perfluorinated elastomer, and are fabricated by means of a dispensing computer numerical control machine. The flow chambers are created by mechanically compressing the three layers using the dedicated (but customizable) Micronit Fluidic Connect Pro clamp ( Figure S2c, Supplementary Materials). This sealing strategy yields flow channels without the need of thermal or chemical processes, and allows for re-opening the chips at any time just by opening the clamp. Temperature-sensitive [Ru(phen) 2 (4-Clp)] 2+ -dyed PCA films, with a thickness lower than 100 µm to match the size of the channel, are deposited in the internal part of the top glass slide, as described below. Sheets of polyimide tape (Kapton ® , DuPont) of 50 µm thickness were cut into 15 mm × 45 mm pieces with three holes (4 mm dia.) along the long axis, using a cutting plotter (Graphtec CE6000-60, Irvine, CA, USA). The patterned tape sections were then transferred onto the glass slide and used as a deposition mask. Solutions of the dye (67 µM) in the pure monomer were knife-coated onto the Kapton ® tape-coated glass, and the tape was immediately removed leaving behind three round-shaped spots of dyed solution. In some cases, the latter were covered with a piece of 15 µm-thick aluminum foil, which remained glued onto the dyed cyanoacrylate film. The luminescent spots were allowed to cure for 24 h and fiber-optic interrogation was always made from the uncovered side of the polymer film ( Figure S2b, Supplementary Materials).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru(II) complexes containing different chelating ligands, namely [Ru(bpy) 3 ](PF 6 3 ]Cl 2 , [Ru(phen) 2 (4-OHp)](PF 6 ) 2 , and [Ru(phen) 2 (4-Clp)]Cl 2 (Scheme II), were chosen. The electronic features of the selected coordination ligands span from the significant σ-donor ability of 4-OHp to the highly π-accepting bpz. The intermediate σ-donating properties of the other chelating ligands decrease in the sequence bpy > phen > dip > 4-Clp, while their π-accepting character increases from bpy to 4-Clp. The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz) 3 ] 2+ complex for which a biexponential fit had to be used at temperatures above 35 • C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz) 3 ] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru(II) complexes containing different chelating ligands, namely [Ru(bpy , [Ru(phen)2(4-OHp)](PF6)2, and [Ru(phen)2(4-Clp)]Cl2 (Scheme II), were chosen. The electronic features of the selected coordination ligands span from the significant σ-donor ability of 4-OHp to the highly π-accepting bpz. The intermediate σ-donating properties of the other chelating ligands decrease in the sequence bpy > phen > dip > 4-Clp, while their π-accepting character increases from bpy to 4-Clp. The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru(II) complexes containing different chelating ligands, namely [Ru(bpy , [Ru(phen)2(4-OHp)](PF6)2, and [Ru(phen)2(4-Clp)]Cl2 (Scheme II), were chosen. The electronic features of the selected coordination ligands span from the significant σ-donor ability of 4-OHp to the highly π-accepting bpz. The intermediate σ-donating properties of the other chelating ligands decrease in the sequence bpy > phen > dip > 4-Clp, while their π-accepting character increases from bpy to 4-Clp. The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru(II) complexes containing different chelating ligands, namely [Ru(bpy , [Ru(phen)2(4-OHp)](PF6)2, and [Ru(phen)2(4-Clp)]Cl2 (Scheme II), were chosen. The electronic features of the selected coordination ligands span from the significant σ-donor ability of 4-OHp to the highly π-accepting bpz. The intermediate σ-donating properties of the other chelating ligands decrease in the sequence bpy > phen > dip > 4-Clp, while their π-accepting character increases from bpy to 4-Clp. The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru ( The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru ( The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).

Indicator Dye Selection
The development of a novel indicator-based luminescent sensor involves an initial stage of selecting the optimum dye for a suitable analyte detection. To that aim, six Ru(II) complexes containing different chelating ligands, namely [Ru(bpy)3](PF6)2, [Ru(bpz)3](PF6)2, [Ru(phen)3](PF6)2, [Ru(dpp)3]Cl2, [Ru(phen)2(4-OHp)](PF6)2, and [Ru(phen)2(4-Clp)]Cl2 (Scheme II), were chosen. The electronic features of the selected coordination ligands span from the significant σ-donor ability of 4-OHp to the highly π-accepting bpz. The intermediate σ-donating properties of the other chelating ligands decrease in the sequence bpy > phen > dip > 4-Clp, while their π-accepting character increases from bpy to 4-Clp. The luminescence decay profile of these complexes was recorded as a function of temperature to establish the influence of the structure and number of specific polypyridyl ligands on the 3 MLCT excited state deactivation kinetics of the corresponding metal complexes. The emission decays in propylene carbonate (PC) could be fitted to single-exponential functions, except that shown by the [Ru(bpz)3] 2+ complex for which a biexponential fit had to be used at temperatures above 35 °C. Moreover, an exponential fit could not be recovered after cooling down the solution containing this dye, pointing out to an irreversible thermal transformation of the luminescent indicator in the organic solvent. The known instability of [Ru(bpz)3] 2+ is attributable to the very weak σ-donation properties of the bipyrazine when compared to either bipyridine or phenanthroline ligands [42,43]. The weak ligand field strength of bpz and related ligands approaches the 3 MLCT and the (dissociative) 3 MC excited states, with deleterious results for the luminescent probe design.
Fitting the experimental emission lifetimes obtained at different temperatures ( Figure 3) to Equation (1) allows determination of the parameters of the fit (Table 1).    Due to the significant covariance observed between the B and ∆E parameters of the fit, they cannot be analyzed separately: only their order of magnitude and the rate constant of the temperature-activated crossing (k t = Bexp(-∆E/kT)) are meaningful quantities [44]. In this way, B > 10 12 s −1 and ∆E > 3000 cm −1 values (at 298 K) have been attributed to thermal population and decay via the 3 MC excited state, whereas B < 10 8 s −1 and ∆E < 800 cm −1 indicate a non-radiative deactivation via the higher-lying (fourth) 3 MLCT state [45]. Taking these values into account, the collected data for the Ru(II) luminophores investigated in this work (Table 1) indicate the predominance of the former deactivation pathway in all cases. The long emission lifetime of the [Ru(dpp) 3 ] 2+ dye, together with the low efficiency of its thermally-activated deactivation process (η t ), probably mean that the observed decrease of its luminescence upon increasing temperature is due to both decay pathways. This fact would explain the intermediate values of B and ∆E obtained for this dye. The conjugated aromatic structure of dpp increases its π-acceptor features when compared to the unsubstituted phen ligand, leading to stabilization of the metal complex 3 MLCT excited state with respect to its ligand-field 3 MC level and, therefore, displays longer emission lifetimes and much lower temperature sensitivity. Nevertheless, [Ru(bpy) 3 ] 2+ and [Ru(phen) 3 ] 2+ are the ruthenium complexes most used in the literature for temperature optosensing, probably taking advantage of the vast amount of information about their luminescent features put forward for fiber-optic O 2 measurements with these dyes [15]. Interestingly, [Ru(phen) 3 ] 2+ is the most sensitive amongst the investigated dyes in the −30 to +20 • C temperature range. However, above 20 • C its luminescence is appreciably quenched so that, at 25 • C, the efficiency of thermally activated promotion from the 3 MLCT excited state is the lowest amongst the values obtained in this study.
The photoexcited [Ru(bpz) 3 ] 2+ is the most sensitive complex in the temperature range of interest due to its high η t value. Unfortunately, this efficient promotion to the dark 3 MC state leads to fast photodecomposition, making it unsuitable for luminescent temperature sensing. The luminescence lifetime of [Ru(phen) 2 (4-OHp)] 2+ also shows a strong dependence with temperature but only in acidic PC (Table 1) due to the almost complete loss of its luminescence when deprotonated. The high η t value found for this indicator (Table 1), close to that of [Ru(bpz) 3 ] 2+ , also leads to a high photolability. Similarly to the pyrazine moieties (see above), but in a lower degree, the electron withdrawing effect of the chlorine atom on the phenanthroline moiety leads to stabilization of the 3 MLCT excited state in [Ru(phen) 2 (4-Clp)]Cl 2 . However, this complex also contains two unsubstituted phen ligands coordinated to the metal core, the strong σ-donation of which results in destabilization of its 3 MC level (strong ligand field). Therefore, this metal complex shows good photostability as a consequence of its high ∆E value (4215 cm −1 ). At the same time, its very large pre-exponential factor (1.5 × 10 14 s −1 ) leads to significant temperature dependence, making it the best choice for a lifetime-based luminescent temperature indicator dye to be used both in the 0 to 50 • C range (water monitoring applications) [46], and the 25 to 40 • C range (cell culture applications) [47,48].

Photophysical Characterization of [Ru(Phen) 2 (4-Clp)] 2+ in Solution
The absorption and emission maxima of [Ru(phen) 2 (4-Clp)] 2+ in different organic solvents and in water, together with its luminescence lifetimes in air-equilibrated and deoxygenated solutions, are summarized in Table 2. The selected organic solvents try to simulate the environment around the indicator dye in the best polymer matrix (see Section 3.3). The absorption spectra show intense UV bands corresponding to intraligand (IL) π → π* transitions and a broad band, with two maxima in the visible region, assigned to a metal-to-ligand charge transfer transition (d → π*) [31]. No significant solvent effect has been observed either on the maxima or on the intensity of the absorption bands within the relatively polar solvents used in this work. This complex also displays, as expected, an unstructured emission band centered at 610-619 nm, depending on the solvent polarity and viscosity. The differences found in the emission lifetime as a function of the solvent can be explained in terms of the corresponding oxygen solubility (higher in organic solvents), solvent viscosity, and energy transfer to the solvent O-H oscillators.

Polymer Sensor Tip Development
After photophysical characterization of the selected Ru(II) dye in solution, four polymeric matrices were tested in order to find the best solid support for the dye immobilization without altering its temperature-sensing features. A facile manufacturing of the dye-doped polymer material, together with a good thermal conductivity and low oxygen permeability were the main searched characteristics for the temperature sensor matrix. The investigated polymers were a sol-gel material, an epoxy resin, poly(ethyl cyanoacrylate) (PCA), and a copolymer of an hydroxylated acrylate monomer (EHMA) and MMA (see Experimental section). The scarce solubility of the Ru(II) complex in pure MMA led us to copolymerize this monomer with its hydroxylated derivative where the dye was soluble in.
The temperature sensitivity of the dye entrapped in the polymeric supports, in terms of the relative change in the luminescence intensity displayed by the manufactured sensor tip when temperature was dropped from 40 to 0 • C, has been summarized in Table 3. The intensity change for other temperatures is shown in Figure S3 (Supplementary Materials). Supports containing hydroxyl groups in their structure, such as the sol-gel and p(EHMA-MMA), showed a lower temperature sensitivity, probably due to competitive deactivation of the luminophore via energy transfer to the surrounding O-H oscillators. Moreover, the sensor tip manufactured with the epoxy resin displayed a slower and unstable response to temperature changes, owing to a lower thermal conductivity. On the contrary, luminescently-doped polycyanoacrylate monoliths showed a fast and stable response (Figure 4), allowing, at the same time, fabrication of readily-made reproducible sensor tips.
Supports containing hydroxyl groups in their structure, such as the sol-gel and p(EHMA-MMA), showed a lower temperature sensitivity, probably due to competitive deactivation of the luminophore via energy transfer to the surrounding O-H oscillators. Moreover, the sensor tip manufactured with the epoxy resin displayed a slower and unstable response to temperature changes, owing to a lower thermal conductivity. On the contrary, luminescently-doped polycyanoacrylate monoliths showed a fast and stable response (Figure 4), allowing, at the same time, fabrication of readily-made reproducible sensor tips.

Photophysical Characterization of the Dye Entrapped into a Poly(Cyanoacrylate) Matrix
Different authors have studied the effect of the polymer support on the photophysical properties of Ru(II) complexes entrapped into rigid matrices. Thus, it has been demonstrated the inhibition of the emission decay through the 3 MC pathway for photoexcited Ru(bpy)3 2+ (and related complexes) immobilized in cellulose acetate [43], zeolites [49], or metal-organic frameworks (MOFs) [50,51], so that its small temperature dependence in the region of interest is then controlled by the additional (fourth) MLCT state. This effect is due to the anti-bonding nature of the excited state 3 MC, because its population is expected to lead to the elongation of the Ru-N bonds that is inhibited in the rigid medium.
We have monitored the changes of the luminescence spectrum of [Ru(phen)2(4-Clp)] 2+ dissolved in the cyanoacrylate monomer during its polymerization. The metal complex displayed a blue shift of its emission band as the monomer became harder (during the first three days, Figure 5). The emitting 3 MLCT cannot be stabilized by solvent reorganization in the solid matrix (rigidochromism), and, therefore, the radiative deactivation takes place from a more energetic excited state leading to the emission blue shift. The luminophore also undergoes a strong enhancement of its luminescence intensity ( Figure 5), largely due to the less accessibility of the oxygen to the indicator, which decreases the deactivation pathway that is caused by this effective quencher. The observed changes in the intensity and maximum emission wavelength of the dye dissolved in the monomer also allow for a continuous monitoring of the polymerization process to assess its progress. Table 4 lists the emission lifetimes obtained from the analysis of the luminescence decay kinetics of the [Ru(phen)2(4-Clp)]/PCA sensor tip. Under the two evaluated temperatures (0.0 and 40.0 °C), the emission decays could only be fitted correctly to 3-exponential functions. Several environments for the dye into the solid matrix are the origin of the different lifetimes found: the longest one is associated to very diluted ruthenium complexes entrapped in the rigid polymer medium, while the shortest one corresponds to aggregated dye molecules dwelling in other regions of the polymer monolith. The bright aggregates undergo nonetheless self-quenching of their emission lifetime, as it has been shown by luminescence lifetime imaging microscopy in O2-sensitive polymer layers that are A B

Photophysical Characterization of the Dye Entrapped into a Poly(Cyanoacrylate) Matrix
Different authors have studied the effect of the polymer support on the photophysical properties of Ru(II) complexes entrapped into rigid matrices. Thus, it has been demonstrated the inhibition of the emission decay through the 3 MC pathway for photoexcited Ru(bpy) 3 2+ (and related complexes) immobilized in cellulose acetate [43], zeolites [49], or metal-organic frameworks (MOFs) [50,51], so that its small temperature dependence in the region of interest is then controlled by the additional (fourth) MLCT state. This effect is due to the anti-bonding nature of the excited state 3 MC, because its population is expected to lead to the elongation of the Ru-N bonds that is inhibited in the rigid medium.
We have monitored the changes of the luminescence spectrum of [Ru(phen) 2 (4-Clp)] 2+ dissolved in the cyanoacrylate monomer during its polymerization. The metal complex displayed a blue shift of its emission band as the monomer became harder (during the first three days, Figure 5). The emitting 3 MLCT cannot be stabilized by solvent reorganization in the solid matrix (rigidochromism), and, therefore, the radiative deactivation takes place from a more energetic excited state leading to the emission blue shift. The luminophore also undergoes a strong enhancement of its luminescence intensity ( Figure 5), largely due to the less accessibility of the oxygen to the indicator, which decreases the deactivation pathway that is caused by this effective quencher. The observed changes in the intensity and maximum emission wavelength of the dye dissolved in the monomer also allow for a continuous monitoring of the polymerization process to assess its progress. Table 4 lists the emission lifetimes obtained from the analysis of the luminescence decay kinetics of the [Ru(phen) 2 (4-Clp)]/PCA sensor tip. Under the two evaluated temperatures (0.0 and 40.0 • C), the emission decays could only be fitted correctly to 3-exponential functions. Several environments for the dye into the solid matrix are the origin of the different lifetimes found: the longest one is associated to very diluted ruthenium complexes entrapped in the rigid polymer medium, while the shortest one corresponds to aggregated dye molecules dwelling in other regions of the polymer monolith. The bright aggregates undergo nonetheless self-quenching of their emission lifetime, as it has been shown by luminescence lifetime imaging microscopy in O 2 -sensitive polymer layers that are also manufactured with Ru(II) complexes [52]. Actually, the situation is probably more complex and the number of luminophores in different microenvironments of the polymer matrix is expected to be higher than just three. Once the 3-exponential function provides an excellent fit, we do not try a larger number of exponential terms. Intermediate lifetimes correspond to populations of Ru(II) dye molecules separated by various distances or surrounded by different domains generated by the PCA chains.
Polymers 2018, 10, x FOR PEER REVIEW 11 of 17 also manufactured with Ru(II) complexes [52]. Actually, the situation is probably more complex and the number of luminophores in different microenvironments of the polymer matrix is expected to be higher than just three. Once the 3-exponential function provides an excellent fit, we do not try a larger number of exponential terms. Intermediate lifetimes correspond to populations of Ru(II) dye molecules separated by various distances or surrounded by different domains generated by the PCA chains.

Temperature Measurements Based on the Luminescence Lifetime of the Polymer Material
Macroscopic Temperature-Sensitive Optical Fiber Sensor. After evaluating the temperature optode response by steady-state emission measurements, the optical sensor was evaluated with a fiber-optic phase-sensitive luminometer. To that end, the sensor tips were coated with aluminum foil for protection against the income of external light to the detector, and placed at the end of a single optical fiber ( Figure S1). The sensitive terminal was introduced in a temperature-controlled water bath and the emission phase shift, as generated by the luminophore lifetime was continuously recorded ( Figure 6). This emission lifetime decreases linearly with temperature in the tested range (r = 0.99987), simplifying the optode calibration. A continuous monitoring of the emission intensity was simultaneously carried out thanks to the value of the amplitude of the modulated luminescence waveform provided by the instrument. Intensity values confirm the dye photostability during the measurements.
The optimal modulation frequency of the blue light source (78 kHz) was chosen to provide the maximum difference (14.3°) between the phase angles that were measured at 0 and 40 °C. This temperature range is the most useful for water courses monitoring, the sought application for our luminescent sensor. The apparent small discrepancy between the emission lifetimes of Ru(phen)2(4-

Temperature Measurements Based on the Luminescence Lifetime of the Polymer Material
Macroscopic Temperature-Sensitive Optical Fiber Sensor. After evaluating the temperature optode response by steady-state emission measurements, the optical sensor was evaluated with a fiber-optic phase-sensitive luminometer. To that end, the sensor tips were coated with aluminum foil for protection against the income of external light to the detector, and placed at the end of a single optical fiber ( Figure S1). The sensitive terminal was introduced in a temperature-controlled water bath and the emission phase shift, as generated by the luminophore lifetime was continuously recorded (Figure 6). This emission lifetime decreases linearly with temperature in the tested range (r = 0.99987), simplifying the optode calibration. A continuous monitoring of the emission intensity was simultaneously carried out thanks to the value of the amplitude of the modulated luminescence waveform provided by the instrument. Intensity values confirm the dye photostability during the measurements.
The optimal modulation frequency of the blue light source (78 kHz) was chosen to provide the maximum difference (14.3 • ) between the phase angles that were measured at 0 and 40 • C. This temperature range is the most useful for water courses monitoring, the sought application for our luminescent sensor. The apparent small discrepancy between the emission lifetimes of Ru(phen) 2 (4-Clp) 2+ embedded in PCA calculated from the observed phase shift (τ φ , Figure 6) and those calculated from the multiexponential decay of the luminescence (τ M , Table 4) is simply due to the different weight of the short and long components (τ i ) in the τ M calculation and in the overall phase shift. While the former depends on α i τ i , the latter rather depends on the modulation frequency, according to α i τ i / 1 + (2π f ) 2 τ 2 i [41] (pp. 191-192). Therefore, higher modulation frequencies yield shorter τ φ values due to the heavier contribution of the shorter-lived components.
Polymers 2018, 10, x FOR PEER REVIEW 12 of 17 Clp) 2+ embedded in PCA calculated from the observed phase shift (τφ, Figure 6) and those calculated from the multiexponential decay of the luminescence (τM, Table 4) is simply due to the different weight of the short and long components (τi) in the τM calculation and in the overall phase shift. While the former depends on i τ i , the latter rather depends on the modulation frequency, according to [1 + (2π ) 2 2 ] ⁄ [41] (pp. 191-192). Therefore, higher modulation frequencies yield shorter τφ values due to the heavier contribution of the shorter-lived components. The effect of the dye loading in the polymer was also investigated. We found a modest 7% increase of φ (0 °C) in going from an initial concentration of Ru(phen)2(4-Clp) 2+ in ethyl cyanoacrylate of 29 to 67 μM, accompanied by a 15% decrease of φ (40 °C); higher levels of doping (up to 114 μM) lead to negligible further increase of φ (0 °C), while significantly increasing φ (40 °C). Those variations translate into phase shift excursions in the 0-40 °C range of 14.3° at 29 μM, to 22.6° at 67 μM. More concentrated initial solutions actually yield smaller excursions of φ probably due to larger populations of aggregated dye species and the self-quenching of the photoexcited dye lifetime they undergo (see Section 3.4).
The response time of the temperature luminescent sensor is determined by the calibrated thermostatic circulator change rate. Nevertheless, instantaneous transfer of the fiber-optic sensor tip from an ice-water bath at 0 °C to the thermostatic bath at 40 °C led to response times on the order of 20 s (t90%). The optode response can be accelerated using thinner PCA/dye monoliths attached to the optical fiber at the expense of a slightly lower sensitivity in the investigated temperature range.
The optode resolution (R), calculated from the calibration slope (s) and the standard deviation,  σn-1, for the worst signal-to-noise ratio (measured at 40.0 °C), was found to be 0.05 °C under the optimized conditions (R = 3σn-1/s, n = 20). The sensor tip reproducibility and its response repeatability were better than 2% and 1%, respectively, also evaluated from the relative standard deviation of the temperature measurements. The effect of dissolved molecular oxygen, a well-known efficient quencher of Ru(II) luminophores (see above), on the response of the optode was also studied ( Figure S4, Supplementary Material). No significant changes were observed when dipping the sensor tip into air equilibrated or The effect of the dye loading in the polymer was also investigated. We found a modest 7% increase of φ (0 • C) in going from an initial concentration of Ru(phen) 2 (4-Clp) 2+ in ethyl cyanoacrylate of 29 to 67 µM, accompanied by a 15% decrease of φ (40 • C); higher levels of doping (up to 114 µM) lead to negligible further increase of φ (0 • C), while significantly increasing φ (40 • C). Those variations translate into phase shift excursions in the 0-40 • C range of 14.3 • at 29 µM, to 22.6 • at 67 µM. More concentrated initial solutions actually yield smaller excursions of φ probably due to larger populations of aggregated dye species and the self-quenching of the photoexcited dye lifetime they undergo (see Section 3.4).
The response time of the temperature luminescent sensor is determined by the calibrated thermostatic circulator change rate. Nevertheless, instantaneous transfer of the fiber-optic sensor tip from an ice-water bath at 0 • C to the thermostatic bath at 40 • C led to response times on the order of 20 s (t 90% ). The optode response can be accelerated using thinner PCA/dye monoliths attached to the optical fiber at the expense of a slightly lower sensitivity in the investigated temperature range.
The optode resolution (R), calculated from the calibration slope (s) and the standard deviation, σ n−1 , for the worst signal-to-noise ratio (measured at 40.0 • C), was found to be 0.05 • C under the optimized conditions (R = 3σ n−1 /s, n = 20). The sensor tip reproducibility and its response repeatability were better than 2% and 1%, respectively, also evaluated from the relative standard deviation of the temperature measurements. The effect of dissolved molecular oxygen, a well-known efficient quencher of Ru(II) luminophores (see above), on the response of the optode was also studied ( Figure S4, Supplementary Materials). No significant changes were observed when dipping the sensor tip into air equilibrated or fully deoxygenated solutions at identical temperature due to the low oxygen permeability of the selected polymer (PCA). Obviously, any influence of O 2 is suppressed for the aluminum foil-covered polymer monoliths.
Microscopic Temperature-Sensitive Films inside Organ-on-a-Chip Devices. Finally, the optimized sensor was tested for temperature monitoring into organ-on-a-chip devices. The dyed polymer was deposited as thin spots and the chips were subjected to temperature variations from ca. 27 to 52 • C (Figure 7). fully deoxygenated solutions at identical temperature due to the low oxygen permeability of the selected polymer (PCA). Obviously, any influence of O2 is suppressed for the aluminum foil-covered polymer monoliths. Microscopic Temperature-Sensitive Films inside Organ-on-a-Chip Devices. Finally, the optimized sensor was tested for temperature monitoring into organ-on-a-chip devices. The dyed polymer was deposited as thin spots and the chips were subjected to temperature variations from ca. 27 to 52 °C (Figure 7). The response is linear between 27 and 43 °C, a suitable range for monitoring cell cultures [47]. The micrometric thickness of the sensitive spots has positive and negative consequences. On the one hand, the response time (the delay between the thermistor and the optode signals) of the temperaturesensitive film is shorter than 5 s. On the other, the thin sensor is no longer insensitive to O2 (see Figure  S5, Supplementary Material). However, this interference is easily suppressed by measuring that species and temperature at the same time, something that can be readily performed in a multichannel instrument, like ours. Moreover, the interference of O2 can be eliminated by covering the dyed PCA spot with commercial aluminum foil which, besides decreasing external light interference blocks permeation of gases without lowering the sensor sensitivity to temperature.

Conclusions
Molecular engineering of Ru(II) polypyridyl complexes allows for the design and manufacture of luminescent temperature sensors that outperform those described up to date based on standard Ru(II) indicator dyes for oxygen monitoring. In this way, a novel heteroleptic complex with two 1,10phenanthroline and one 4-chloro-1,10-phenanthroline ligands embedded in poly(cyanoacrylate) provides optimum temperature response and photostability (at least) in the 0-50 °C range, using both emission intensity-or lifetime-based measurements with commercially-available instrumentation for optical O2 sensing. Moreover, this polymeric material provides affordable, easily made, and reproducible sensitive elements with excellent optical quality that can be tailored to the sought application.
The applications of polymer-based luminescent temperature sensors doped with Ru(II) complexes are countless and not only restricted to those illustrated here, namely temperature monitoring of surface waters in combination with other Ru(II)-polypyridyl-based optodes (O2, pH, CO2, NH3, Cu II , etc.) [10] plus a multichannel field-deployable optoelectronic unit, and cell culture monitoring with lab-on-a-chip systems. Other uses where such polymeric optical sensors may provide a definitive advantage are currently being investigated, e.g., temperature monitoring in The response is linear between 27 and 43 • C, a suitable range for monitoring cell cultures [47]. The micrometric thickness of the sensitive spots has positive and negative consequences. On the one hand, the response time (the delay between the thermistor and the optode signals) of the temperature-sensitive film is shorter than 5 s. On the other, the thin sensor is no longer insensitive to O 2 (see Figure S5, Supplementary Materials). However, this interference is easily suppressed by measuring that species and temperature at the same time, something that can be readily performed in a multichannel instrument, like ours. Moreover, the interference of O 2 can be eliminated by covering the dyed PCA spot with commercial aluminum foil which, besides decreasing external light interference blocks permeation of gases without lowering the sensor sensitivity to temperature.

Conclusions
Molecular engineering of Ru(II) polypyridyl complexes allows for the design and manufacture of luminescent temperature sensors that outperform those described up to date based on standard Ru(II) indicator dyes for oxygen monitoring. In this way, a novel heteroleptic complex with two 1,10-phenanthroline and one 4-chloro-1,10-phenanthroline ligands embedded in poly(cyanoacrylate) provides optimum temperature response and photostability (at least) in the 0-50 • C range, using both emission intensity-or lifetime-based measurements with commercially-available instrumentation for optical O 2 sensing. Moreover, this polymeric material provides affordable, easily made, and reproducible sensitive elements with excellent optical quality that can be tailored to the sought application.
The applications of polymer-based luminescent temperature sensors doped with Ru(II) complexes are countless and not only restricted to those illustrated here, namely temperature monitoring of surface waters in combination with other Ru(II)-polypyridyl-based optodes (O 2 , pH, CO 2 , NH 3 , Cu II , etc.) [10] plus a multichannel field-deployable optoelectronic unit, and cell culture monitoring with lab-on-a-chip systems. Other uses where such polymeric optical sensors may provide a definitive advantage are currently being investigated, e.g., temperature monitoring in bioreactors, aerospace applications, and temperature measurements within cells and tissues (the latter upon encapsulation of the indicator dye in polymer nanocapsules).
Supplementary Materials: The following items are available online at http://www.mdpi.com/2073-4360/10/ 3/234/s1, Figure S1: Picture of the temperature-sensitive optical fiber tips, Figure S2: Drawings/pictures of the 3-layer chip, temperature-sensitive spots into Micronit's organ-on-a-chip device, and of the Fluidic Connect PRO chip holder with fluidic connections and optical fibers, Figure S3: Variation with temperature of the luminescence intensity from the dye-doped polymer monoliths, Figure S4: Response to O 2 of the [Ru(phen) 2 (4-Clp)]/PCA temperature-sensitive optical fiber tip, Figure S5: Response to O 2 of the [Ru(phen) 2 (4-Clp)]/PCA temperature-sensitive spots in the organ-on-a-chip device.