Lanthanide Luminescence to Mimic Molecular Logic and Computing through Physical Inputs

The remarkable advances in molecular logic reported in the last decade demonstrate the potential of luminescent molecules for logical operations, a paradigm‐changing concerning silicon‐based electronics. Trivalent lanthanide (Ln3+) ions, with their characteristic narrow line emissions, long‐lived excited states, and photostability under illumination, may improve the state‐of‐the‐art molecular logical devices. Here, the use of monolithic silicon‐based structures incorporating Ln3+ complexes for performing logical functions is reported. Elementary logic gates (AND, INH, and DEMUX), sequential logic (KEYPAD LOCK), and arithmetic operations (HALF ADDER and HALF SUBTRACTOR) exhibiting a switching ratio >60% are demonstrated for the first time using nonwet conditions. Additionally, this is the first report showing sequential logic and arithmetic operations combining molecular Ln3+ complexes and physical inputs. Contrary to chemical inputs, physical inputs may enable the future concatenation of distinct logical functions and reuse of the logical devices, a clear step forward toward input–output homogeneity that is precluding the integration of nowadays molecular logic devices.


Introduction
In a period of a few decades, the computers and related systems become smaller and smaller due to the substantial scaling shrinkage of silicon-based technology components. [1][2][3][4][5] Whereas the first field-effect transistor has the dimensions of functions were performed. Later on, it was recognized that the molecules can additionally perform logical operations of higher complexity, such as molecular multiplexers and demultiplexers or molecular binary adders and subtractors [18,[20][21][22][23][24][25] All the above-mentioned molecular devices operate in wet conditions, so its immediate use was in biomedicine. Molecular and biomolecular computing/logic systems, [26][27][28] including supramolecules, [16] proteins, [14] DNA, [29] DNA/conjugated polymer systems, [17,21] and enzyme based-systems, [30] were used to process chemical input signals mimicking most of the logical operations of semiconductor-based electronic devices.
As the signal processing is performed by molecules, parallel computing can be implemented, allowing an exponential gain on computing power in comparison with nowadays electronic devices. [31] However, the characteristic broadband emission and fast decay times of organic compounds, conjugated to the well-known photobleaching, precludes multiplexing using light as a logical input or output, making difficult its implementation in applications other than proof-of-concept molecular logic devices.
The use of trivalent lanthanide ions (Ln 3+ ) as luminescent optical active centers in molecular logic is particularly attractive as these ions typically emit in a wide wavelength range covering the UV-vis-NIR spectral regions, with characteristic line-like emission bands (<10 nm) and long-lived excited state lifetimes (>1 µs). Despite it is nowadays recognized that Ln 3+ ions can improve the state-of-the-art molecular logical devices, only a handful of papers have been published so far. [11,[32][33][34][35][36] Up to now, all the logical gates based on Ln 3+ ions respond to chemical inputs and thus operate exclusively in wet conditions, except the Eu 3+ /Tb 3+ -based self-assembled polymer monolayer functionalized Si surface proposed by some of us in 2016. [36] It is recognized in the literature that if physical inputs (viz., light, pressure, temperature) are used as inputs and outputs of a molecular logical device it will render advantages in the input-output homogeneity when the integration of distinct logical functions, no-contamination, and reuse of the device are demanded. [37] It also permits higher integration of distinct logical functions because it does not require physical access, and monolithic structures may contain thousands of individual logic elements. For example, the optical inputs signals can be remotely delivered and read from the molecular devices as time-gated pulses (covering time ranges until the picosecond, [37] faster cycling even in comparison to the electronic counterparts). All these potential benefits motivate the transfer of molecular logic devices from wet (chemical or physical inputs and physical outputs) to dry (exclusively physical inputs and outputs) operation conditions. Up to now all the works performing molecular logic and computing using exclusively physical inputs (involving reversible phototransformations of photochromic molecules) [37][38][39][40][41] operate, however, in wet conditions.
Here, we report a noteworthy step forward relative to the state-of-the-art examples presenting high-sophistication-level logical operations based exclusively on physical inputs and outputs and operating in nonwet conditions. By exploiting an Eu 3+ /Tb 3+ -functionalized Si platform that produces narrow and stable emission lines and that is fully compatible with existent Si-based electronic devices, we fabricate molecular logical arrays performing basic (AND, INH, DEMUX), sequential (KEYPAD LOCK), and arithmetic (HALF ADDER and HALF SUBTRACTOR) logic operations.

Fabrication and Structural Characterization of the Logical Arrays
The formation and the structure of the Eu 3+ /Tb 3+ -based selfassembled polymer monolayer functionalized Si surface are depicted in Figure 1a,b, respectively. Original monocrystalline Si wafers were covered with a 1-µm-thick layer of SiO 2 and the surface was activated by amination with 3-aminopropyl triethoxysilane (APTES). The functionalization with the Eu 3+ and Tb 3+ complexes was realized by Michael's addition to the surface amine groups of an acrylate poly(ethylene glycol) (PEG) polymer spacer terminated in a β-ketoester ligand ( Figure S1 and Table S1, Supporting Information). This group fixes the Eu 3+ and Tb 3+ ions to the surface allowing further complexation of these ions with light-harvesting 2,6-piridindicarboxilic acid (DPA) ligands. Different types of coatings were prepared that differ in the length of the PEG polymer spacer ( Figure 1c, Table 1). Samples with and without the addition of the Ln 3+ (Ln = Eu, Tb) ions and DPA ligands and using two different reaction times for Michael addition of acrylate to surface amines were prepared (Figures S2-S4 and Tables S2-S4, Supporting Information).
The extension of the reaction time from 24 h (samples S1B-S6B) to 48 h (samples S1D-S6D) did not result in a significant change in the composition and structure of the samples, and therefore samples of the series B and D are equivalent from the luminescence and structural point of view. The same applies to the samples without Eu 3+ and Tb 3+ ions prepared with a reaction time of 24 h (samples S1A-S6A) and 48 h (samples S1C-S6C).
The samples were analyzed by X-ray photoelectron spectroscopy (XPS) before and after the formation of the Eu 3+ and Tb 3+ complexes, confirming the functionalization of the Si surface and the incorporation of the Ln 3+ ions (Figure 2).
Field emission scanning electron microscopy analysis of the samples shows a very smooth surface and a uniform chemical composition for the samples with medium and small polymer chain lengths, whereas some degree of granulation and chemical heterogeneity is discerned for the sample with the longest polymer length. Atomic force microscopy analysis ( Figures S5-S8, Supporting Information) shows a clear increase of the surface roughness on the Ln 3+ -functionalized samples concerning amine-modified samples. This loss of functionalization effectiveness in long polymeric chain lengths is also observed in the Ln 3+ -doped samples (S1B to S6B), and it is reflected in a lower concentration of Eu 3+ and Tb 3+ on the surface (Tables S2 and S3, Supporting Information).

Photoluminescence
The photoluminescence of the samples prepared with distinct acacPEGA chain lengths or with different functionalization routes does not show notable differences (Figures S10-S17, www.advopticalmat.de Supporting Information) and, thus, we selected for further investigation three representative samples treated at 48 h with the longest (S1D), middle (S3D), and shortest (S6D) polymer chain length.
The room temperature excitation spectra of the three samples, monitoring the 5 D 4 → 7 F 5 (Tb 3+ ) and 5 D 0 → 7 F 2 (Eu 3+ ) transitions, are similar and dominated by broadband in the 250-290 nm range associated to the DPA ligand excited states ( Figure S10, Supporting Information). [42] Increasing the temperature, the integrated areas of the 5 D 4 → 7 F 5 (Tb 3+ , I 1 ) and 5 D 0 → 7 F 2 (Eu 3+ , I 2 ) transitions decrease by 70% and 30%, respectively (Figure 3 and Figure S12, Supporting Information), meaning that the I 1 /I 2 ratio is temperature-sensitive, as reported for other Eu 3+ /Tb 3+ -bearing materials. [43][44][45][46][47] Figure S18 (Supporting Information) presents the temperature evolution of the I 1 /I 2 ratio in a single heating-cooling cycle for S1D, S3D, and S6D. During the heating from 295 to 369 K, the intensity ratio decreases linearly, probably due to the activation of nonradiative decay pathways involving the polymeric chains that depopulate the 5 D 4 and 5 D 0 emitting levels. The high energy of the DPA ligand triplet state, ≈27 050 cm −1 , [46,48] prevents the 5 D 4 -to-ligand and 5 D 0 -to-ligand energy back-transfer, discarding, then, any role of the ion-to-ligand back transfer in the decrease of the I 1 /I 2 ratio as temperature increases. Furthermore, Figure 1. a) Schematic diagram of Si surface functionalization with the self-assembled monolayers incorporating the Ln 3+ ions showing the activation and functionalization steps. b) The structural formula of monolayers functionalized with the Ln 3+ ions. c) Schematic representation of the S1D, S3D, and S6D samples. The details of the logic arrays structure and fabrication are summarized in Table 1. 3000 S1A S1C S1B S1D Adv. Optical Mater. 2020, 8,2000312 www.advopticalmat.de thermally activated Tb 3+ -to-Eu 3+ energy transfer is also settled out as no evidence of that transfer is found in the excitation spectra and lifetime measurements (in accord with the relatively low amount of the two ions, making the energy transfer extremely improbable). Moreover, the I 1 /I 2 ratio shows a hysteretic behavior in a heating/cooling cycle ( Figure S18, Supporting Information). This is not related to the optical bistability of Ln 3+ -bearing crystals excited with high power densities, [49] but to changes in the conformation of the acacPEGA polymer chains, as reported previously by some of us in an analogous polymer. [36,50] On cooling, two linear regimes are distinctly discerned, whose intersection occurs at the same temperature for all the samples, (T c = 313 ± 1 K) ( Figure S18, Supporting Information). This indicates that the two regimes are independent of the polymeric chain length and different functionalization routes used (Figures S13 and S14, Supporting Information), reaffirming the crucial role of the temperature dependence of the polymer conformation on the temperature dependence of the I 1 /I 2 ratio. The hysteretic response of the samples will be used in Section 3.2 to demonstrate sequential logical operations.
The thermometric response of these systems was accessed and summarized in Table S6 and Figures S18-S20 (Supporting Information). The maximum relative thermal sensitivities are 2.84, 2.23, and 2.27% K −1 for S1D, S3D, and S6D, respectively. These values are high enough to consider the application of these samples as luminescent molecular temperature probes ( Table 2), particularly in the physiological range of temperatures interesting for biological applications. However, the focus of this manuscript is to use Ln 3+ luminescence, including its temperature dependence, to mimic molecular logic and computing through physical inputs.

Combinational Logic Functions
The Si arrays functionalized with Eu 3+ and Tb 3+ ions emitting centers can be employed for the construction of optically active molecular logic gates, using exclusively physical inputs and outputs. For the following discussion, we use S1D as an illustrative Figure 2. a,b) XPS spectra of the C 1s region for short (samples S6A, S6B), medium (S3A, S3B), and long (S1A, S1B) acacPEGA chains in samples A series and B series ( Table 1 of the main text). The carbon oxidation states in acacPEGA are CC (C 1s 1 and C 1s 2), CO (C 1s 3), OCO/CO (C 1s 4), and OCO (C 1s 5). Remarkable differences are found in the CO signal depending on the length of the acacPEGA chain. Series B and D, containing Eu 3+ and Tb 3+ ions, show the same behavior. c) XPS spectra of S3A (before anchoring the complexes) and S3B (after anchoring the complexes). The Tb 3d and Eu 3d regions are highlighted to show the differences resulting from the anchoring of the Ln 3+ complexes. d) Evolution of Eu 3d core level with the length of the acacPEGA organic chain, for short (S6B), medium (S3B), and large (S1B) chains.
www.advopticalmat.de example. We tackle the design of molecular logics exploiting the heat and the excitation wavelength as logical inputs and several emission intensity ratios as the logical outputs (Figure 3d,e). The excitation light is always assumed as an input in the reports on all-optical or all-photonic logic gates, [37,[61][62][63][64] however it is the first time that temperature increase (or heat) is being used as an input for molecular logical gates. Using simply the dependence of the emission spectra upon 271 nm excitation as the temperature is being raised (Figure 4a), we clearly observe changes in the emission profile that can be quantified using three ratiometric outputs: i) I 1 /I 2 , defined above, ii) I 2A /I 2B , involving the intensity peak ratio of two Stark components of the 5 D 0 → 7 F 2 transition (Figure 4b), and iii) I 3A /I 3B , an intensity peak ratio of two Stark components of the 5 D 0 → 7 F 4 transition (Figure 4c). The temperature evolution of these intensity ratios is plotted in Figure 4d,e. Whereas I 1 /I 2 and I 2A /I 2B decrease 66% and 50%, respectively, as temperature increases by 70 degrees, for I 3A /I 3B this same increment results in a 55% enhancement, relatively to the value at 300 K. These dissimilar evolutions upon heating induced the development of one inhibit (INH) and one AND logical gate.
Using the I 1 /I 2 ratio (or I 2A /I 2B ) as logical output, we set a threshold value of 0.06 to define the logical output TRUE (1, when I 1 /I 2 > 0.06) or FALSE (0, the opposite case). The logical inputs are the excitation at 271 nm (denoted by UV) and a temperature increase of 70 degrees (denoted by ΔT). This logical gate does not produce any output without UV and thus the logical output is 0. Only in the presence of excitation at 271 nm and in the absence of ΔT the logical output is 1 and, thus, this is the transfer function characterizing the INH logic gate (Figure 4a,c,e). We notice the high obtained switching ratio (66% of the initial value), defined as the relative change of the output signal when the logical output signals change from "1" and "0." The I 3A /I 3B ratio is used as logical output setting a threshold value of 0.8 to define the logical output TRUE (1, when I 3A /I 3B > 0.8) or FALSE (0, the opposite case). The logical inputs are the same defined for the INH logic gate. Again, this logical gate does not produce any output without 271 nm excitation and thus the logical output is 0. Only when the system is excited at 271 nm and heated by ΔT the logical output is 1 and, thus, this is the transfer function characterizing the AND logic gate (Figure 4b,d,f). This gate operates based on the distinct thermal The Tb 3+ and Eu 3+ transitions are signed in green and red, respectively. The asterisk marks the spectral region of overlap between the 5 D 0 → 7 F 1 (Eu 3+ ) and the 5 D 4 → 7 F 4 (Tb 3+ ) transitions. The peak intensities I 1 , I 2 , and I 3 are also identified. b,c) Magnification of 5 D 0 → 7 F 2 (displaying the I 2A and I 2B components) and 5 D 0 → 7 F 4 (displaying the I 3A and I 3B components) transitions. The blue and red spectra correspond to the spectra measured at the lowest (295 K) and the highest (369 K) temperatures. d,e) Temperature dependence of the I 1 /I 2 intensity ratio and I 2A /I 2B and I 3A /I 3B intensity ratios. The threshold values (above which the logical value of the ratio is "1" and below is "0") are identified by the horizontal lines.
www.advopticalmat.de population ratio of two 7 F 4 Stark components. We observe that on one hand, the integrated intensity of the high-energy 7 F 4 Stark component is almost temperature independent whereas that of the remaining ones displays a strong temperature dependence ( Figure S17, Supporting Information) evidencing the hypersensitive nature of the 5 D 0 → 7 F 4 Eu 3+ transition. [65] The switching ratio of this gate is about 60%, a similar value to that calculated from the data in ref. [66] (67%).
Observing the inputs homogeneity on the two logical gates described above, we design a port constituted by the integration of the AND and the INH logic gates. The resulting logical gate, presented in Figure 4g,h, is a 1:2 demultiplexer (DEMUX). This is a clear step forward relative to our previous work [36] (the first example using Eu 3+ and Tb 3+ emissions and temperature for logic applications), which reported the same DEMUX using a much complicated and time-consuming strategy.

Sequential Logical Functions
Contrarily to the combinational logic functions described in the previous section, in the sequential logic functions the output signal switches ON when a specific inputs order is set, [67][68][69] implying a memory function of the device. The KEYPAD LOCK function has been reported previously employing both chemical [68,[70][71][72] and physical [37,71,[73][74][75][76][77][78] inputs. Here, a KEYPAD LOCK was implemented using the temperature increment ΔT and the characteristic temperature T c as physical inputs (Figure 5a). We set the logical value 0 for the ΔT input representing the switching off of the temperature controller and the T c input in the T < T c range. The logical value 1 is set for the switching on of the temperature controller and T = T c , respectively. Each input might be 0 or 1, leading to eight possible two-digit order combinations. From this state, when both physical inputs present the logical value 0, the output is strictly (I 1 /I 2 ) 0 . Regardless of the order in which the temperature controller is switched on (ΔT > 0) while the second input is T < T c , the output will be (I 1 /I 2 ) c < (I 1 /I 2 ) < (I 1 /I 2 ) 0 . On the other hand, when the temperature is fixed to T c and no heating source is acting (and vice versa), the output will be (I 1 /I 2 ) c . When the heating is on, and then T = T c , the output yields (I 1 /I 2 ) c ≤ I 1 /I 2 < (I 1 /I 2 ) 0 .
As the initial state, we consider the sample a room temperature (denoted by (I 1 /I 2 ) 0 , Figure 5a,c). The only possibility to open the KEYPAD LOCK is to have first the system at T = T c and then impose a ΔT such that I 1 /I 2 < (I 1 /I 2 ) c .

Arithmetic Operations
A key requirement of digital computers is the ability to use logical functions to perform arithmetic operations. Traditionally, the operation is implemented using an exclusively OR (XOR for the sum digit) and an AND logical gate (for the carry digit). We notice that using the UV and ΔT inputs it is not possible to recover the XOR transfer function, so we implement the HALF-ADDER operation using the temperature increase ΔT = T H −T C (Figure 5a,b, logical value 1 means heating and 0 means no temperature increase) and the characteristic temperature value T = T c as inputs, and combining the I 1 /I 2 output and the relative thermal sensitivity, defined by Equation (1).
Defining the logical value "1" for I 1 /I 2 ratio lower than the threshold value (I 1 /I 2 ) H , we recall the AND logic gate (for S1D (I 1 /I 2 ) H = 0.025). The (I 1 /I 2 ) H is defined as the ratio of integrated intensities on cooling corresponding to the temperature T C . The temperature T H is defined as the temperature value for the heating regime for which the (I 1 /I 2 ) H is obtained (Figure 5a). For the XOR logic gate, we use the S r output defining the logical value "1" when S r is in the range S C < S r < S H corresponding to the S r values at room temperature and at T H . The resulting HALF-ADDER logical gates arrangement is presented in Figure 5e and the corresponding truth table in Figure 5f.

Conclusions
In summary, monolithic structures consisting of Si surfaces functionalized with acacPEGA polymeric chains of controllable length and functionalized with Eu 3+ and Tb 3+ ions are reported for applications in molecular logic and computing using exclusively physical inputs. Basic (AND, INH, and DEMUX), sequential (KEYPAD LOCK), and arithmetic (HALF ADDER and HALF SUBTRACTOR) logic operations are demonstrated using the excitation light and the temperature as inputs and ratios of emission intensities and the relative thermal sensitivity as outputs. The system is fully reversible containing ≈10 6 emitting centers µm −2 . Comparing the integration density of the Eu 3+ /Tb 3+ based self-assembled polymer monolayer functionalized Si surface with that of the modern Si-based chips (≈10 2 ), [6] we realize that the former presents 10 4 times higher value. Concerning the molecular logical devices reported so far, this work demonstrates clear advantages on the integration of distinct logical functions and reuse of the device for several cycles. In this work, there is no physical contact of the inputs with samples (like in those operating in wet conditions) that result eventually in the device contamination. The implementation of the reported logical functions on a functionalized Si-chip also permits better integration with current electronic devices and the use of Ln 3+ -based emitting centers is quite suitable for emission and excitation by commercial detectors and LED sources. Although a molecular computer able of substituting current conventional computers is still just an ambitious future vision, [24] judiciously designing molecular systems that perform logic functions is paving the way for developing useful molecular logic applications. Most of the drawbacks of the devices reported so far can be mitigated if solid samples interrogated by physical inputs are used, preferentially giving narrow emitting lines. In this context, Ln 3+ -based emitting centers will certainly play a central role, since the same ions are widely spread in optoelectronic devices, smoothing the interface between novel molecular and conventional electronic devices.

www.advopticalmat.de
and 3000 Da is the M w molecular weight), N,N-dimethylaniline (DMA), and acryloyl chloride (97%) were purchased from Aldrich and used as received.
Synthesis of the PEG Polymer Chains: The PEG chains were prepared as described in the Supporting Information.
Preparation of SiO 2 -Coated Wafers: p-type monocrystalline Si wafers (Okmetic) were coated with a SiO 2 layer of 1083.0 ± 11.3 nm thickness by plasma-enhanced chemical vapor deposition of tetraethylorthosilicate (TEOS) at 380 °C and a pressure of 1 atm.
Surface Activation: The silicon wafers were cut in 1 × 1 cm 2 dices, washed with ethanol and acetone, and dried under nitrogen. Surface hydroxylation was performed by O 3 treatment for 10 min using a UVOcleaner Model 342 (Jelight Company Inc., USA).
Amine Surface Functionalization: Functionalization with amine groups was achieved by sol-gel reaction with APTES. The APTES coating was carried out by vapor deposition, as described in a previous publication. [36] Briefly, the surfaces together with a vial containing 1 mL of APTES were placed inside a desiccator previously thermostated at 80 °C and connected to a vacuum. The desiccator was pumped down and sealed off. The samples were left in contact with the APTES vapor for 3 h. Afterward, the samples were washed with acetone, dried under nitrogen, and cured in an oven at 50-60 °C for 20 min.
Grafting PEG Polymer Chains to the Silicon Substrates: The PEG polymer chains grafting were achieved by Michael's reaction of acrylate acacPEGA terminal groups with surface amine groups. After coating with APTES, the silicon surfaces were immersed in acacPEGA solution Figure 5. a,b) An illustrative plot of the I 1 /I 2 intensity ratio and S r of S1D. The vertical pointy lines mark the characteristic temperatures T c and T H , whereas the horizontal ones indicate the corresponding values of I 1 /I 2 (a) and S r (b). The other lines in (a) are the best fit to experimental data. In (a) and (b), the shadowed regions mark the ranges in which the logical output is set as the logical value "1." c) Bar plot of KEYPAD LOCK using the heating (ΔT) and T c as inputs. d) The ON state below the threshold (interrupted line) is only accessible when the inputs are applied in the correct sequence (first T c and then ΔT) leading to the lock opening. e,g) Logic gates of arithmetic operations for sum or subtract two binary numbers. f,h) The corresponding truth tables.

www.advopticalmat.de
(samples 1b-6b, 8 × 10 −3 m, Supporting Information) in water at pH = 8.5 and allowed to react at 60 °C, washed with distilled water, and dried under nitrogen flux. The reaction was carried out individually for each wafer in an Eppendorf containing 2 mL of an 8.9 × 10 −3 m aqueous solution of the polymer. Two series of silicon substrates with grafted polymer chains were prepared, labeled as S1A-S6A for a Michael reaction time of 24 h and as S1C-S6C, for a Michael reaction time of 48 h ( Table 1).
Anchoring of Lanthanide Coordination Compounds: Europium and terbium ions in a 1:3 molar ratio were attached to the polymer selfassembled monolayer by coordination to the β-ketoester polymer terminal groups in a 1:1 polymer/Ln 3+ ratio. The polymer-coated wafers were placed in a vial with 2 mL of water, to which 15.8 µL of 1 m NaOH, 100 µL of a solution of EuCl 3 (H 2 O) 6 (1.44% wt), and 100 µL of a solution of TbCl 3 (H 2 O) 6 (4.42% wt) were added, and the pH was adjusted to 6.7. After 3 h, 340 µL of a solution of DPA (15.75 mg mL −1 , pH 6.76) was added. The complex was formed overnight. The surfaces were then thoroughly washed with distilled water and dried under nitrogen. Two series of samples were obtained, samples S1B-S6B (deriving from samples S1A-S6A), and samples S1D-S6D (deriving from samples S1C-S6C) ( Table 1).
Surface Characterization by XPS: The XPS analysis was carried out using a Kratos Axis Ultra spectrometer employing a monochromatic Al Kα (1486.6 eV, 10 mA, 15 kV) X-ray source and a power of 150 W. All samples were introduced in the analysis chamber simultaneously and were analyzed in the same experimental conditions. To avoid any induced effect in the chemistry of the samples, they were analyzed as received without any previous etching. Differential surface charging was minimized using a charge neutralizer system (flood gun). Survey spectra were recorded using an analyzer pass energy of 160 eV and 1.0 eV energy step. High-resolution spectra of C 1s, O 1s, N 1s, Si 2p, Eu 3d, and Tb 3d regions were collected using pass energy of 20 eV and 0.1 eV energy step. The spectra were analyzed using CasaXPS software. The background for all spectra was subtracted using a Shirley baseline. Due to the use of charge neutralizer, spectra need to be calibrated. Binding energies (BE) were referenced to the C 1s (CC) peak at 285.0 eV.
Photoluminescence Characterization: The emission spectra were recorded with a modular double grating excitation spectrofluorimeter with a TRIAX 320 emission monochromator (Fluorolog-3, Horiba Scientific) coupled to an R928 Hamamatsu photomultiplier, using a front face configuration. A 450 W Xe arc lamp was used as the excitation source. Both recorded emission and excitation spectra were corrected with the spectrofluorimeter optical spectral response and the spectral distribution of the lamp intensity using a photodiode reference detector, respectively. On the other hand, the temperature was controlled using an IES-RD31 controller and a Kapton thermofoil heater from Minco mounted on a copper holder and monitored using a thermocouple thermometer Barnant 100 (model 600-2820) with an accuracy of 0.1 K, accordingly to the manufacturer. The temperature ramp in the heatingcooling cycles is about 0.25 K min −1 .
Thermometric Performance: The relative thermal sensitivity (S r ) was calculated using and the temperature uncertainty using being δΔ/Δ the relative uncertainty in Δ calculated using where δI i /I i (i = 1.2) is calculated dividing the readout fluctuations of the baseline by the maximum value of each intensity, i.e., I 1 and I 2 . As the integrated areas are calculated from the same emission spectra δI 1 = δI 2 = δI. The experimental Δ = I Tb /I Eu values were fitted to straight lines (Table S6, Supporting Information) and the resulting parameters were used to calculate S r and δT.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.