A versatile iron(II)-based colorimetric sensor for the vapor-phase detection of alcohols and toxic gases

Yunnan Guo , Shufang Xue , Marinela M. Dîrtu and Yann Garcia *
Institute of Condensed Matter and Nanosciences, Molecules, Solids and Reactivity (IMCN/MOST), Université catholique de Louvain, Place L. Pasteur 1, Louvain-la-Neuve 1348, Belgium. E-mail: yann.garcia@uclouvain.be

Received 23rd January 2018 , Accepted 1st March 2018

First published on 21st March 2018

A mononuclear iron(II) neutral complex (1) was screened for colorimetric sensing abilities for a wide spectrum of vapor-phase analytes including toxic gases. Although 1 possesses a single sensing unit without any other cross-reactive elements, it can capture information about analyte molecules in a distributed fashion that is encoded sufficiently to allow discrimination from other closely related chemical structures. Clear color differentiation among four different toxic industrial chemicals (TICs) and alcohols was demonstrated. Different TICs and alcohols were readily identified using a standard chemometric approach (hierarchical clustering analysis), with no misclassifications over 18 trials. 57Fe Mössbauer spectroscopy was applied to investigate the driving spin state change of the analytes in this Fe-azole sensor material. This approach opens up perspectives for the further development of iron materials as potential optical sensor arrays through colorimetric techniques.


There is currently a huge appeal for “chemosensors” based on colorimetric sensor array (CSA) technology for the sensitive and selective detection of gas- and vapor-phase analytes,1,2 developing a wide range of applications including chemical threat alerts,3 medical diagnostics,4 and environmental monitoring.5 These sensor materials always incorporate a targeted set of chemically responsive colorants. Therefore, dyes of coordination complexes are developing an indispensable role in the colorimetric detection of metal-binding species, especially, toxic small molecules such as amines, covalent hydrides and alcohols.6 They mainly benefit from two contributions of metalo-colorants: (a) in well documented examples, analytes can directly coordinate to the metal center,7 causing significant changes to the d–d absorptions; (b) metal centers can respond to redox reactions.

In order to enhance the discrimination ability of these sensor devices, a large number of chemically responsive sensing elements are necessary, which are usually obtained from instable and expensive reagents.8 Therefore, the currently tedious and taxing sensing probe synthesis requires the development of new materials for simple, cost-effective, and rapid, yet sensitive measurements for practical applications.9 One possible way is to develop an elegant dye, which can capture information about the analyte molecule in a distributed fashion that is encoded sufficiently to allow discrimination from other closely related chemical structures.10 Good candidates for exploiting these effects are spin crossover (SCO) materials.11,12 Beyond those benefits of coordination complexes mentioned above, in fact, SCO materials present a unique ability to switch between high-spin (HS) and low-spin (LS) states in response to external stimuli, such as temperature or pressure variations,13 and light irradiation, which have been intensely studied for use in “physical-sensors”.14,15 As a chemosensor, SCO materials show an intrinsic nature wherein the metal center is very sensitive to subtle changes driven by guest intercalation, thus resulting in alterations of the optical outputs (Fig. 1a).16,17 In particular, some systems are known to be very sensitive to solvent exchange through a molecular network. For example, recently, Aromí et al. reported on an FeII mononuclear complex that showed vapochromism with MeOH and acetone as a result of the SCO mechanism.18 Using Hoffmann Clathrate-type systems, indirect analyte–metal interactions can also afford a spin-state change for Fe(II) centers.19–21

image file: c8tc00375k-f1.tif
Fig. 1 (a) Electronic diagram of the HS and LS states for an Fe(II) ion in an octahedral ligand field; (b) molecular structure of our colorimetric sensor 1;27 (c) colorimetric sensor 1 printed directly on a plexiglas capsule.

Obviously, the main bottleneck of SCO materials is how to detect and discriminate among the wide range of analytes showing an olfactory-like response, rather one or two specific guests. Toward this aim, we have been engaged in the design of new SCO sensors through the synthesis and use of 1,2,4-triazole derivatives.22,23 The Fe-azole system gives rise to a family of coordination complexes, many exhibiting SCO behavior, which are often extremely sensitive to their hydration degree.24–26 Therefore, a mononuclear iron(II) neutral complex (1), [Fe(trz-tet)2(H2O)4]·2H2O with trz-tetH = 5-(4H-1,2,4-triazol-yl)-2H-tetrazole (trz-tetH), was developed as a colorimetric sensor, which is specific to MeOH sensing among an alcohol series (Fig. 1b).27 The mechanism is based on molecular sieving and a subsequent spin-state change of iron centers via a crystal-to-crystal transformation. The sensing process is visually detectable, fatigue-resistant, highly selective, and reusable. We were thus intrigued by the ability of the molecular sieving effect in 1. In the present work, we applied 57Fe Mössbauer spectroscopy to investigate the driving spin state change of analytes in this Fe-azole sensor material (1) and established it as a reversible and bidirectional sensor for the detection of guest MeOH and EtOH molecules with a memory effect. Furthermore, 1 was developed as a simple colorimetric sensor that can differentiate a wide range of volatile analytes, which was then applied to the detection of toxic gases with the help of a hierarchical cluster analysis (HCA).

Results and discussion

Selective and reusable sensor for alcohols

When the native hydrate sensor pellet 1 (Fig. 1c) is in contact with liquid methanol [MeOH(l)] (Fig. S1, ESI) at room temperature, it instantly turns to pale pink. The sensor begins to turn pink within five min, but it takes hours to transform into dark pink when the material is saturated with MeOH(g) to give 1-MeOH (Fig. 2). What is more interesting is that 1-MeOH maintained its pink color (the LS state, shown below) after removing MeOH by drying under vacuum at room temperature, denoted 1-off-MeOH (Fig. S2, ESI). Finally, 1-off-MeOH reversibly turns back to colorless (1) when exposed to water vapor/washed with water for regeneration. 1 thus detects MeOH with a memory effect by visual, optical, and magnetic feedback.
image file: c8tc00375k-f2.tif
Fig. 2 The process of selective and reusable colorimetric sensing of alcohols using sensor 1.

The process of reversible and bidirectional colorimetric sensing can be quantitatively analyzed by 57Fe Mössbauer spectroscopy. As shown in Fig. 3a, the sensor pellet of 1 at 298 K shows a unique doublet with an isomer shift δ = 1.17(2) mm s−1 and a quadrupole splitting ΔEQ = 3.18(3) mm s−1, which corresponds to the value of the crystalline sample [Fe(trz-tet)2(H2O)4]·2H2O.271-MeOH reveals a much more complex spectrum with four signals: a new quadrupole doublet in the center [δ = 0.37(1) mm s−1; ΔEQ = 0.20(2) mm s−1, 19%], characteristic of a LS FeN6 core with azole ligands;27 two new HS FeII contributions: HS-1 [δ = 1.14(2) mm s−1; ΔEQ = 2.56(5) mm s−1; 33%] and HS-2 [δ = 0.83(3) mm s−1; ΔEQ = 2.40(5) mm s−1; 18%], corresponding to the FeN4O2 octahedra, which are more distorted than the remaining contribution of 1, which is also detected with δ = 1.17(1) mm s−1 and ΔEQ = 3.17(1) mm s−1. This MeOH intrusion is presumably part of an allosteric effect, wherein an uncoordinated tetrazole moiety of the neighboring mononuclear unit is displaced from its mean position to be coordinated to iron, thus leading to a LS FeN6 core.27 In contrast to the crystalline state situation,27 30% of the HS FeN4O2 octahedra are not modified by the MeOH intrusion within the sensor pellet.

image file: c8tc00375k-f3.tif
Fig. 3 Spin-state tracking by 57Fe Mossbauer spectroscopy at 298 K of (a) 1, (b) 1-MeOH, and (c) 1-off-MeOH, which shows the memory effect and (d) 1-Regeneration from 1-off-MeOH after treatment with water, which shows full reversibility.

For classic alcohol sensors, control over the memory effect is unlikely because of the fading from analytes leaving. Indeed, after experiencing a color change due to guest molecule detection, the color is lost upon guest release. This situation holds for various SCO materials reported up to now.28,29 Our sensor, however, keeps track of the color change due to alcohol detection, even after its release (see Fig. 3c). The LS state is unaffected upon removal of the MeOH guest, as well as other sites. This indicates that MeOH is the driving force to produce the FeN6 core, but it cannot recover the FeN2O4 sites after the removal of MeOH because of the lack of H2O compensation. The reversibility of the process was confirmed by Mössbauer spectra, which show a complete disappearance of the LS signal in 1-Regeneration from 1-off-MeOH after treatment in a water vapor atmosphere (Fig. 3d and Table S1, ESI).

The selectivity of detection decreases in the order MeOH > EtOH > i-PrOH > … with increasing alcohol molecule size.27 A weaker vapochromic behavior was indeed noticed when a pellet of 1 was exposed to EtOH vapors, as shown in Fig. 2 and Fig. S3 (ESI). In contrast to the MeOH case, the response time lasted up to two days. Such a color change was only observed as a slight effect on the crystal surface of 1-EtOH, due to the lower available surface area.27 A similar memory effect to that found for 1-off-MeOH was observed for 1-off-EtOH, with the presence of only one modified FeN4O2 site, indicating fewer modifications in the crystal lattice. Higher analogues in the alcohol series were exposed to 1 for several days (i-PrOH, BuOH, pentan-1-ol) and none showed any color change.

It is known that the disadvantage of limited reversibility for most CSAs is that it necessitates disposable colorimetric sensors.1 The valuable point of 1 lies in the reusable sensing process for environmental concerns. Obviously, such a material should be inherently sensitive to host–guest interactions and constitutes a playground for molecular recognition by switching iron centers, thanks to the non coordinated tetrazole moiety.

Colorimetric sensor for toxic gases

Detection and discrimination among the wide range of high-priority TICs remain a major challenge and the topic of substantial recent research, especially for closely related chemical structures. For example, recently, Suslick et al. reported on TIC identification using a CSA consisting of 36 different chemically responsive pigments with the ability to differentiate 19 TICs.30 In view of the ability to selectively sense the MeOH molecule by molecular sieving in 1, we also undertook a further examination of small molecular TICs, choosing four representative examples (HCl, HBr, NH3 and N2H4) from lists generated by the International Task Force 25 report.31 Excitingly, 1 exhibits a vapochromic response upon incorporating these four TIC guests by changing into specific and distinguishable colors (shifts of optical adsorption band as shown in Fig. S4, ESI). In contrast, without statistical analysis, the four TICs could be determined from these unique colors even by the eye (Fig. 4).
image file: c8tc00375k-f4.tif
Fig. 4 Images depicting vapochromic transformation upon the sensing of four TICs by 1.

The mechanism of colorimetric sensing was also revealed quantitatively by 57Fe Mössbauer spectroscopy (Fig. 5). The spectra of 1-HCl and 1-HBr were fitted best with one similar quadrupole doublet centered at δ = 1.06(1) and 1.16(1) mm s−1 with ΔEQ = 3.26(1) and 3.14(1) mm s−1, respectively. These parameters are assigned to HS FeII without any sign of oxidation or Fe-related phases (e.g., FeCl2 and FeBr2 salts). This indicates that the Fe atoms are still coordinated by water molecules leading to a FeN2O4 core and that the Cl or Br anions do not enter into the coordination sphere. The Mössbauer spectrum of 1-NH3 is, however, different. It shows a new quadrupole doublet with δ = 0.36(3) mm s−1EQ = 0.65(5) mm s−1), typical of LS FeII, indicating that 1 absorbs the NH3 molecule to form a LS FeN6 center. Indeed, the isomer shift fits well with the one of the LS site of 1-MeOH, the increase of ΔELSQ being attributed to an increase of the lattice contribution to the electric field gradient in 1-NH3. No thermally induced spin state crossover was observed on cooling to 78 K (Table S2, ESI). The situation differs with vapors of hydrazine (N2H4), with the emergence of a HS FeII center with δ = 1.03(2) mm s−1 and ΔEQ = 2.61(8) mm s−1. The spectrochemical position of hydrazine is higher than that of ammonia, i.e. FeN6 of hydrazine is more inclined to be LS than ammonia.32 Instead, however, the new spin state in 1-N2H4 signifies FeN4O2 octahedra, as found in 1-MeOH, which are indicative of only relevant distortion from hydrazine. Indeed, taking into account that the molecular sizes of N2H4 and MeOH are exactly the same, a molecular sieving effect may presumably dominate this colorimetric process, rather than spectrochemical considerations. More quantitative investigation of the apparently complicated analytes driving the spin state change was beyond the scope of this work, but will be pursued in further studies.

image file: c8tc00375k-f5.tif
Fig. 5 Spin-state tracking by 57Fe Mossbauer spectroscopy at 298 K of 1-HCl, 1-HBr, 1-NH3 and 1-N2H4.

Standard chemometric methods were employed in order to examine the potential of the colorimetric sensor in the recognition of toxic industrial chemicals. The printed sensor was imaged digitally with a smart camera after exposure to each analyte. A six-dimensional vector (that is, three in RGB values (red, green and blue) and three in HSV values (Hue, Saturation and Value) for the after-exposure image) was chosen for the quantitative comparison of the color changes of the sensor. This way, it was not necessary to retain the full digital images: each analyte was represented digitally by the 6-dimensional vector and could be compared by standard chemometric techniques. Simple comparisons of the spatial distances in their full original vector space between an observed color and a precollected library were extremely rapid. The raw, digital color vectors from a total of 18 experimental trials are given in Table S3 (ESI).

The color change provoked by HCl and NH3 can be identified from the sensor pellet in a matter of a few minutes. However, N2H4 and HBr color changes are observed in less than one hour, as shown in Fig. 6. These observations indicate that under identical conditions, the vapor pressure of different TIC members decides the extension of the solvent vapor exposure period for colorimetric sensing (PHCl > PHBr; PNH3 > PN2H4).

image file: c8tc00375k-f6.tif
Fig. 6 Total Euclidean distance vs. response time for the four TICs and MeOH under vapor diffusion (the solid line is a guide to the eye). The Euclidean distance corresponds to the total length of the 6-dimensional color vector shown in Table S4 (ESI).

The process of image digitization is easily quantified by efficient pattern recognition techniques. The HCA provides a straightforward method for quantitatively displaying analyte dissimilarity. This method relies on a classification scheme based on the spatial distances of data points in their full dimensionality.30 Each experimental trial was defined as a 6-dimensional vector that consists of three RGB values and three HSV values in our sensor. HCA was performed using the minimum variance (Ward's) method for different analytes. The HCA dendrogram for the responses to saturated analyte vapors is shown in Fig. 7, demonstrating a 100% correct classification (in triplicate trials including a humidity test). It is well known that control over the humidity is extremely problematic for a given electronic sensor. However, 1 was cultivated in H2O solution so that changes in the humidity would present no difficulty. As shown in Fig. 4, the color map is unaffected by relative humidity under water vapor, which has also been proven by the reusable sensing for alcohols.

image file: c8tc00375k-f7.tif
Fig. 7 HCA dendrogram with Ward linkage for 4 TICs, MeOH and one humidity test at full vapor pressure at room temperature. The Euclidean distances are defined by the color vectors of each experimental trial in the full 6-dimensional space made up of 3 RGB values and 3 HSV values. All experiments were performed in triplicate: no confusion could be observed in the classification for 18 trials.

In the field of CSAs, the collection of sensors always contains as much chemical diversity as possible, so that the array responds to the largest possible cross-section of analytes to reduce the error rate.8 Undoubtedly, sensor 1 can be considered as an excellent candidate as a crucial element for designing new CSAs to allow discrimination from closely related chemical structures.


In summary, an Fe-azole complex (1) was employed as a dye in the development of a colorimetric sensor for the vapor-phase detection and discrimination of alcohols (MeOH and EtOH) and four TICs, including HCl, HBr, NH3 and N2H4. These small molecules of volatile analytes lead to the intrusion of 1 by a molecular sieving effect. Hierarchical cluster analysis reveals that 1 can capture information about these analyte molecules in a distributed fashion that is encoded sufficiently to allow discrimination from other closely related chemical structures (i.e. MeOH and EtOH; HCl and HBr; NH3 and N2H4). The main advantage of this chromogenic system is the striking signal transduction and operation at ambient temperatures without the need for cryogenic facilities or temperature pretreatment, which is very common, e.g. for metal–organic frameworks (MOFs).33 Thus, in the present investigation, we have demonstrated that an FeII material can be used as a colorimetric sensor based on the effect of analytes driving a spin state change. Interestingly, our sensor retains a memory effect after alcohol detection in contrast to most sensors based on spin crossover complexes. This opens up possibilities for developing such materials as potential optical sensor arrays through colorimetric techniques.



All chemicals purchased were of reagent grade and used without further purification. A crystalline sample of 1 was prepared according to a literature procedure.27

Sensor pellet preparation

The crystalline sample was ground to reduce the particle size by the following procedure: (i) add a spatula full of 1 into an agate mortar and grind it to a fine powder until crystallites can no longer be seen and it becomes somewhat “pasty” and sticks to the mortar. (ii) Insert the powder into a Qwik Handi-Press and press the powder twice to form a 7 mm sensor pellet in a square collar, as shown in Fig. S1 (ESI). Disassemble the press and take out the collar. (iii) Insert the collar together with the pellet onto the sensor holder.

Physical measurements

The 57Fe Mössbauer spectra were recorded in transmission geometry with a constant acceleration mode conventional spectrometer equipped with a 50 mCi 57Co(Rh) source and a Reuter Stokes proportional counter. The sensor pellet was sealed in aluminum foil and spectra were recorded at 298 K. The spectra were fitted using Recoil 1.05 Mössbauer Analysis software (K. Lagarec, D. G. Rancourt, Recoil, Mössbauer Spectral Analysis software for Windows 1.0, Department of Physics, University of Ottawa, 1998). The isomer shift values are given with respect to α-Fe at room temperature. The diffuse reflectance spectra were obtained with a Perkin Elmer Lambda 9 UV/vis/NIR spectrophotometer equipped with a 60 mm integrating sphere and converted into absorption spectra by using the Kubelka–Munk function, using BaSO4 as a reference. Real time imaging of the vapochromic behavior of the sensor kept under analyte vapors was performed with a Sony 16 MP IMX298 stacked image sensor system (HUAWEI).

Data analysis

The color patterns were obtained by taking the RGB values (red, green and blue) and HSV values (Hue, Saturation and Value) from the center of the sensor pellet (5 mm) images. Each image yields a color vector of six dimensions. Therefore, the eight-bit color imaging was created by each dimension of RGB value ranging from 0 to 255, Hue from 0 to 360 and Saturation and Value from 0 to 100. The vectors are provided in Tables S3 and S4 (ESI). Measurements can be performed using the GNU Image Manipulation Program. All experiments were run in triplicate, and chemometric analyses were carried out on the difference vectors using the SPSS Statistics software Version 22.

Conflicts of interest

There are no conflicts to declare.


We thank the Fond National de la Recherche Scientifique-FNRS for a PDR (T.0102.15) and the fellowships of chargé de Recherches allocated to Y. GUO (28109061), S. XUE (24918505) and M. M. Dîrtu (16511866). We thank Prof. J. Wouters for the courteous use of the diffuse reflectance spectrometer. We also thank the COST actions CM1305 and CA15128.


  1. J. R. Askim, M. Mahmoudi and K. S. Suslick, Chem. Soc. Rev., 2013, 42, 8649–8682 RSC.
  2. M.-C. Tu, J. A. Cheema, U. H. Yildiz, A. Palaniappan and B. Liedberg, J. Mater. Chem. C, 2017, 5, 1803–1809 RSC.
  3. K. Urmila, H. Li, Q. Chen, Z. Hui and J. Zhao, Anal. Methods, 2015, 7, 5682–5688 RSC.
  4. P. J. Mazzone, X.-F. Wang, Y. Xu, T. Mekhail, M. C. Beukemann, J. Na, J. W. Kemling, K. S. Suslick and M. Sasidhar, J. Thorac. Oncol., 2012, 7, 137–142 CrossRef PubMed.
  5. L. Feng, C. J. Musto, J. W. Kemling, S. H. Lim, W. Zhong and K. S. Suslick, Anal. Chem., 2010, 82, 9433–9440 CrossRef CAS PubMed.
  6. N. A. Rakow and K. S. Suslick, Nature, 2000, 406, 710–713 CrossRef CAS PubMed.
  7. M. Albrecht, M. Lutz, A. L. Spek and G. van Koten, Nature, 2000, 406, 970 CrossRef CAS PubMed.
  8. K. J. Albert, N. S. Lewis, C. L. Schauer, G. A. Sotzing, S. E. Stitzel, T. P. Vaid and D. R. Walt, Chem. Rev., 2000, 100, 2595–2626 CrossRef CAS PubMed.
  9. N. Fahimi-Kashani and M. R. Hormozi-Nezhad, Anal. Chem., 2016, 88, 8099–8106 CrossRef CAS PubMed.
  10. T. A. Dickinson, J. White, J. S. Kauer and D. R. Walt, Nature, 1996, 382, 697–700 CrossRef CAS PubMed.
  11. A. Lapresta-Fernandez, M. P. Cuellar, J. M. Herrera, A. Salinas-Castillo, M. d. C. Pegalajar, S. Titos-Padilla, E. Colacio and L. F. Capitan-Vallvey, J. Mater. Chem. C, 2014, 2, 7292–7303 RSC.
  12. M. C. Young, E. Liew and R. J. Hooley, Chem. Commun., 2014, 50, 5043–5045 RSC.
  13. P. Gütlich, A. B. Gaspar and Y. Garcia, Beilstein J. Org. Chem., 2013, 9, 342 CrossRef PubMed.
  14. J. Linares, E. Codjovi and Y. Garcia, Sensors, 2012, 12, 4479 CrossRef CAS PubMed.
  15. M. Khusniyarov, Chem. – Eur. J., 2016, 22, 15178 CrossRef CAS PubMed.
  16. X. Zhang, H. Xie, M. Ballesteros-Rivas, Z.-X. Wang and K. R. Dunbar, J. Mater. Chem. C, 2015, 3, 9292–9298 RSC.
  17. D. Gentili, N. Demitri, B. Schafer, F. Liscio, I. Bergenti, G. Ruani, M. Ruben and M. Cavallini, J. Mater. Chem. C, 2015, 3, 7836–7844 RSC.
  18. J. S. Costa, S. Rodríguez-Jiménez, G. A. Craig, B. Barth, C. M. Beavers, S. J. Teat and G. Aromí, J. Am. Chem. Soc., 2014, 136, 3869–3874 CrossRef CAS PubMed.
  19. A. Mishima, T. Koshiyama, J. A. Real and M. Ohba, J. Mater. Chem. C, 2017, 5, 3706–3713 RSC.
  20. I. y. A. Gural'skiy, S. I. Shylin, B. O. Golub, V. Ksenofontov, I. O. Fritsky and W. Tremel, New J. Chem., 2016, 40, 9012–9016 RSC.
  21. J.-Y. Li, C.-T. He, Y.-C. Chen, Z.-M. Zhang, W. Liu, Z.-P. Ni and M.-L. Tong, J. Mater. Chem. C, 2015, 3, 7830–7835 RSC.
  22. Y. Garcia, F. Robert, A. D. Naik, G. Zhou, B. Tinant, K. Robeyns, S. Michotte and L. Piraux, J. Am. Chem. Soc., 2011, 133, 15850–15853 CrossRef CAS PubMed.
  23. M. M. Dîrtu, A. D. Naik, A. Rotaru, L. Spinu, D. Poelman and Y. Garcia, Inorg. Chem., 2016, 55, 4278–4295 CrossRef PubMed.
  24. J. M. Herrera, S. Titos-Padilla, S. J. A. Pope, I. Berlanga, F. Zamora, J. J. Delgado, K. V. Kamenev, X. Wang, A. Prescimone, E. K. Brechin and E. Colacio, J. Mater. Chem. C, 2015, 3, 7819–7829 RSC.
  25. M. Gimenez-Marques, M. L. Garcia-Sanz de Larrea and E. Coronado, J. Mater. Chem. C, 2015, 3, 7946–7953 RSC.
  26. T. Zhao, L. Cuignet, M. M. Dirtu, M. Wolff, V. Spasojevic, I. Boldog, A. Rotaru, Y. Garcia and C. Janiak, J. Mater. Chem. C, 2015, 3, 7802–7812 RSC.
  27. A. D. Naik, K. Robeyns, C. F. Meunier, A. F. Léonard, A. Rotaru, B. Tinant, Y. Filinchuk, B. L. Su and Y. Garcia, Inorg. Chem., 2014, 53, 1263–1265 CrossRef CAS PubMed.
  28. (a) Y. Garcia, P. J. van Koningsbruggen, E. Codjovi, R. Lapouyade, O. Kahn and L. Rabardel, J. Mater. Chem., 1997, 7, 857–858 RSC; (b) P. J. van Koningsbruggen, Y. Garcia, E. Codjovi, R. Lapouyade, O. Kahn, L. Fournès and L. Rabardel, J. Mater. Chem., 1997, 7, 2069–2075 RSC.
  29. S. Rodriguez-Jimenes, H. L. C. Felthon and S. Brooker, Angew. Chem., Int. Ed., 2016, 55, 15067–15071 CrossRef PubMed.
  30. S. H. Lim, L. Feng, J. W. Kemling, C. J. Musto and K. S. Suslick, Nat. Chem., 2009, 1, 562–567 CrossRef CAS PubMed.
  31. A. K. Steumpfle, D. J. Howells, S. J. Armour and C. A. Boulet, Final report of international task force-25; hazard from toxic industrial chemicals, 1996.
  32. C. Furlani, G. Mattogno, A. Monaci and F. Tarli, Inorg. Chim. Acta, 1970, 4, 187–191 CrossRef CAS.
  33. O. S. Wenger, Chem. Rev., 2013, 113, 3686–3733 CrossRef CAS PubMed.


Electronic supplementary information (ESI) available: Additional data regarding IR spectroscopy, gas adsorption measurements, reflectance-on-temperature dependence. See DOI: 10.1039/c8tc00375k

This journal is © The Royal Society of Chemistry 2018