CHAPTER 1: Fundamentals of Quenched Phosphorescence O₂ Sensing and Rational Design of Sensor Materials

In the last decades, optical oxygen sensors (oxygen optodes) became indispensable analytical tools, which are nowadays widely applied in academia and industry. Their popularity is explained by the numerous advantages offered by the optical detection method such as the absence of electromagnetic interferences, minimally invasive character (measurement though a transparent wall of a reactor), versatility of sensor formats varying from planar foils and fiber-optic sensors to nanoparticles, suitability for imaging etc. Optical oxygen sensors do not consume the analyte, which favourably distinguishes them from electrochemical sensors such as Clark electrode or galvanic cells. Optodes allow for oxygen measurement in gases and solutions with dynamic ranges, which can be adjusted over many orders of magnitude. Finally, optical oxygen sensors are very useful for measurement of air pressure on surfaces (pressure sensitive paints) or as transducers for enzymatic sensors making use of oxygen consumption such as glucose or lactate sensors.

Oxygen is one of the most powerful luminescence quenchers. Quenching of fluorescent dyes (excited singlet state, S₁) and phosphorescent dyes (excited triplet state, T₁) is spin-allowed. Moreover, the energies of excited states of oxygen (¹∑_g⁺ and ¹Δ_g) are lower than the energies of the excited states of most organic dyes and metal complexes (Figure 1.1), which makes quenching via energy transfer favourable.

The mechanism of oxygen quenching is rather complex and the exact pathways and formed products depend on many factors.¹  Electron-exchange Dexter-type energy transfer is the predominant mechanism of oxygen quenching. Quenching of fluorescent dyes (D) can result in the formation of the dye in the triplet excited state or in the ground state:¹ 

The triplet state of the dye is deactivated to the ground state:

For the quenching of phosphorescence, the dye is deactivated into the ground state and singlet oxygen is formed:

Depending on the triplet energy of the dye, formation of singlet oxygen either only in the ¹Δ_g state (e.g. for PtTPTNP, Figure 1.1) or in both ¹Δ_g and ¹∑_g⁺ states (e.g. for PtOEP) is possible. Notably, O₂(¹∑_g⁺) deactivates very fast into O₂(¹Δ_g) state.

Apart from the energy transfer, electron transfer leading to superoxide is also possible:

This process can play a significant role for metal complexes with strong reducing properties (particularly in the excited state), for instance Ir(iii) cyclometalated complexes.²  Rapid back electron transfer can result in the formation of singlet oxygen and the sensitizer in the ground state.

Importantly for all these processes, singlet oxygen represents one of the main products. Since its deactivation to the triplet state regenerates the analyte, optical oxygen sensors do not consume the analyte in theory. However, the lifetime of singlet oxygen in polymers can be much longer compared to that in the aqueous phase (∼3 µs), which can be sufficient for it to react with the sensor components (dye or polymer), see Chapter 1.7.

Independent of the quenching mechanism, the quenching behavior for dissolved dyes is described by the Stern–Volmer equation:

where I₀(τ₀) and I(τ) are the luminescence intensity (decay time) in the absence and in the presence of oxygen, respectively, k_q is the bimolecular quenching constant and K_SV is the Stern–Volmer constant.

From eqn (1.1) it is evident that the efficiency of quenching depends on both the bimolecular quenching constant and the decay time of the luminophore τ₀. The k_q constant is determined mostly by oxygen diffusion since the diffusion of the much larger dye is significantly slower already in solution and is virtually non-existant for immobilized dyes. The k_q constant often approaches the diffusion-controlled limit k_diff for quenching of fluorescence¹  but is lower for quenching of phosphorescence. For common phosphorescent indicators such as Pt(ii) porphyrins or Ru(ii) polypyridyl complexes, it is usually close to 1/9 of k_diff,²  where 1/9 is the spin statistical factor accounting for the formation of both products in the singlet state. However, k_q is sometimes higher than this value even in the case of purely energy transfer-based quenching¹  and may be even higher if electron transfer is involved.² 

Clearly, the τ₀ has a much stronger influence on the Stern–Volmer constant than k_q. In fact, assuming a k_q = k_diff = 2.1 × 10¹⁰ M⁻¹ s⁻¹ for an air-saturated toluene solution (C(O₂) ≈ 1.8 mM) of a typical fluorescent dye with τ₀ of 4 ns, the I₀/I value calculated with eqn (1.1) is only 1.15. On the other hand, for a phosphorescent indicator such as PtOEP (τ₀ = 85 µs) even with much lower k_q 2.4 × 10⁹ M⁻¹ s⁻¹ (1/9k_diff) the luminescence intensity and decay time decrease 367-fold in the same conditions. Since the diffusion of oxygen is significantly slower in the polymers compared to the solution, it is evident that only phosphorescent indicators will provide the required resolution when embedded in common polymeric matrices. Additionally, whereas the tunability of fluorescence decay times is usually limited by 1–2 orders of magnitude, the phosphorescence decay time can vary from several microseconds to hundreds of milliseconds. This provides virtually unlimited flexibility in designing oxygen-sensing materials for very different applications.

In order to navigate among hundreds of reported oxygen indicators it is useful to define the important parameters, which should be considered when the indicators are selected. These include:

• Compatibility of the indicator with the light sources, photodetectors and other optical components should also be considered. Whereas a wide range of light sources is available for the whole spectral range, the detectors are mainly limited to avalanche photodiodes, CCD-arrays and photomultipliers. Although the sensitivity of PMTs is generally very high, it deteriorates fast in the NIR part of the spectrum. In the case of fiber-optic sensors, the quartz glass fibers are compatible with all oxygen indicators. In contrast, much cheaper plastic fibers show strong absorption in the NIR part of the spectrum, which limits the practically useful length to 1–2 meters.

• (ii) Brightness of luminescence. Brightness can be defined as the product of molar absorption coefficient ε and luminescence quantum yield ϕ. Clearly, in case of phosphorescent indicators high efficiency of inter-system crossing (ISC, S₁ → T₁ transition, Figure 1.1) is one of the prerequisites for bright phosphorescence. Bright indicators allow for thinner sensing layers and therefore for faster response times. In the case of very bright but less photostable indicators the operational lifetime of the sensor can be extended by using low intensities of the excitation light nevertheless allowing for acceptable signal to noise ratio.

• (iii) Luminescence decay times. The luminescence decay times along with permeability of the polymer govern the sensitivity of the sensor. In order to obtain oxygen sensors for physiologically relevant conditions on the basis of polymeric matrices such as polystyrene (moderate oxygen permeability) the optimal decay times are about 20–100 µs, the requirement met by most Pt(ii) porphyrins. Indicators with shorter decay times would deliver lower resolution, whereas the sensors based on the indicators with much longer lifetimes (e.g. Pd(ii) porphyrins) will only be suitable for trace sensing. On the other hand, if trace sensing applications are intended, the long luminescence lifetimes (≫1 ms) are highly desirable.

• In respect to photobleaching, two aspects should be mentioned. First, photobleaching rates can be significantly different in the absence of oxygen and under air saturation. In the second case, oxidation of the dye by photosynthesized singlet oxygen may be the predominant mechanism of photodegradation. Second, drift of the sensor properties can be observed not only due to oxidation of the dye but also due to degradation of the matrix polymer resulting in accumulation of the moieties acting as quenchers of phosphorescence of the indicator. In contrast to the first mechanism (photobleaching of the dye itself) characterized by degradation of the luminescence intensity but relative stability of the decay time, in the second mechanism both parameters will be affected to approximately the same degree. Although this effect can be overcome in theory by using an oxidation-resistant polymer, the choice of such polymers is very narrow and the compatibility of the indicator and the polymer can represent a serious problem.

• Only a few indicator classes show the properties which fulfil most of the requirements. Dyes with the best combination of properties will be good indicators for general use covering the majority of applications. However, for some applications one or the other requirement can be a cut-off criterion whereas other properties are of less importance. Here a specialized tailor-made indicator dye may be a much better choice.

An impressive number of phosphorescent dyes were reported in the last decades and most of them qualify as oxygen indicators.^4,5  However, there is a significant gap between chemists who design the sensing materials and end-users such as biologists, oceanographers, geologists etc. Therefore, the most widely used indicators are either commercially available or can be easily prepared in the lab. Fortunately, some comparably new indicators such as benzoporphyrin complexes have been also commercialized recently.

The structures of the most common oxygen indicators are shown in Figure 1.2 and their photophysical properties are summarized in Table 1.1 Ruthenium tris(4,7-diphenyl-1,10-phenanthroline) (Ru-dpp) is the brightest among the polypyridyl complexes and was applied in many sensing materials. This dye can be synthesized in only one step. The limitations include moderate luminescence brightness and high temperature dependency of the luminescence decay time, which is also comparably short so that immobilization in highly oxygen-permeable matrices is essential to achieve optimal sensitivity. Platinum(ii) and palladium(ii) complexes with porphyrins and their derivatives are extremely popular. The main reasons are simple synthesis and often commercial availability, acceptable photophysical properties and versatility of the porphyrin structure which allows for numerous modifications. Pt(ii) and Pd(ii) complexes with octaethylporphyrin (OEP) possess good brightness upon excitation in the Soret band (UV range). Molar absorption coefficients for excitation with green light are much lower (Table 1.1). The photostability of OEP complexes is, however, moderate. Therefore, for most applications they have been substituted by the complexes of the same metals with meso-(pentafluorophenyl)porphyrin (TFPP) which are known to be very photostable. As a trade-off, the brightness of these dyes is about 2-fold lower than for the OEP complexes.

As was mentioned above, NIR dyes are advantageous for several reasons. Platinum(ii) complex with ocaethylporphyrin-ketone¹¹  (PtOEPK) enjoyed great popularity as a NIR emitting oxygen indicator. Spectral properties of the Pt(ii) complex with tetra(pentafluorophenyl)porpholactone¹⁵  are very similar to that of the PtOEPK. Although these dyes feature absorption and emission spectra, which are bathochromically shifted compared to the parent porphyrins, the position of the Q-band is not fully adequate for in vivo measurements. Moreover, the brightness of the dyes is moderate. Interestingly, the Pd(ii) complexes with porphyrin-ketones and porphyrin lactones show very weak phosphorescence, which is different from the general trend shown by the porphyrins (∼2–3 fold lower brightness of the Pd(ii) complexes, Table 1.1).

Although NIR-emitting metal complexes of benzoporphyrins were reported more than two decades ago,¹⁶  only in the last decade have they become popular indicator dyes. This is explained by the development of convenient synthetic methods leading to a great variety of structures (pyrrole condensation under Lindsey conditions with oxidative aromatization elaborated by Vinogradov and co-workers¹⁷  and more recently modified template method¹⁸ ) and also of increasing interest to researchers in in vivo applications. Indeed, the absorption of the Q-band for the Pt(ii) and Pd(ii) complexes of tetraphenyltetrabenzoporphyrin (TPTBP) peaks at 614 and 628 nm, respectively, which is very close to so called “NIR optical window”. The molar absorption coefficients are exceptionally high, which in combination with good quantum yields, results in excellent brightness of these compounds. It should be noted that some recent investigations¹⁹  indicate that the quantum yields are not as high as previously reported by various groups, but the dyes nevertheless remain the brightest in this spectral range. The group of Vinogradov designed various benzoporphyrin-based dendrimeric probes for extracellular measurements and in vivo applications^20–22  some of them also suitable for two-photon excitation. In an alternative approach, lipophilic complexes of TPTBP can be embedded into water-dispersible nanoparticles with or without cell penetration capabilities.^23–25  The absorption and emission spectra of tetranaphthoporphyrins such as TPTNP (Figure 1.2) are even more bathochromically shifted and fully match the NIR optical window. Unfortunately, the luminescence quantum yields and photostability are significantly lower than those of the benzoporphyrins. Additionally to these main classes several molecular hybrids were reported. Hybrids of benzo- and naphthoporphyrins²⁶  possess spectral properties tuneable over a broad range resulting in tailor-made dyes, matching different available LED sources. The hybrids of benzoporphyrins and phthalocyanines (aza-benzoporphyrins)²⁷  share very high photostability and comparably narrow absorption bands of phthalocyanines and possess very large Stokes shift. A large palette of NIR oxygen indicators is very useful when designing sensor arrays for in vivo applications, such as dual sensors for simultaneous subcutaneous monitoring of glucose (via an oxygen transducer) and oxygen.²⁸ 

The indicators described above can be viewed as general-purpose indicators. However, for some specific oxygen sensing applications more exotic dyes can be a better choice. These include, for instance, cyclometalated iridium(iii) coumarin complexes featuring ultrabright emission,²⁹  visible light-excitable europium(iii) complexes which possess typical narrow band emission of Eu(iii),³⁰  or luminescent chelates of Gd(iii),³⁰  Al(iii) and BF₂ ³¹  which possess long phosphorescence decay times for the design of trace oxygen sensors.

Two main groups of the oxygen-sensing materials can be distinguished: (i) molecular probes including water-soluble dyes, dendrimers, dye-protein conjugates etc. and (ii) materials based on oxygen indicators immobilized in some matrix (mostly polymers or sol–gels) or adsorbed on/covalently coupled to the surface of a (non)porous material.

The materials of the first group are not reusable so they are predominantly used for microscopic applications and for assays in samples of small volumes such as microwells in plate readers. In order to be useful, such probes must possess excellent solubility in water. Although this criterion is met by water-soluble phosphorescent dyes, they are poorly suited to practical applications for several reasons: (i) their phosphorescence is typically weak due to efficient radiationless deactivation in water; (ii) the response to oxygen is highly influenced by the environment and is likely to vary significantly upon interaction of the probe with biological components such as proteins; (iii) such probes are more prone to cross-talk to ionic species which can act as quenches; (iv) they can show high cytotoxicity due to close proximity to cells. Therefore, in order to overcome these limitations, a phosphorescent indicator is usually protected by creating a polymer shell (dendrimers),²²  immobilization into nanoparticles or covalent coupling to a protein³²  or a peptide.³³  A polymeric shell or a protein not only protects the dye from undesired interference but also tunes the oxygen permeability of the environment thus helping to design a probe with an optimal dynamic range.

The second group of materials is used for the preparation of reusable sensors. Here, the indicator is immobilized into a polymeric or sol–gel matrix, which acts a solvent and a support for the dye and as a permeation-selective barrier. Due to variety of the available polymers, polymer-immobilized dyes are more common than those based on silica or Ormosils (organically modified silica).³⁴  Among the vast group of polymers, hydrophobic ones are preferable since potentially interfering species (ions, proteins, water) are prevented from interaction with the indicator.

The following aspects should be considered in respect to the polymer choice:

• (i) Oxygen permeability of the polymer. This parameter along with the phosphorescence decay time of the dye is essential for the design of the sensors with the desired dynamic range (see below);

• (ii) Compatibility of the dye and the polymer. Compatibility can be low if both components differ significantly in polarity. For instance, most dyes will aggregate readily in silicone rubber or in amorphous perfluorinated polymers (such as Teflon AF) so that modification of the dyes which improves their solubility in such matrices cannot be avoided.

• (iii) Physical properties and chemical stability of the polymer. Whereas many polymers will form a mechanically stable layer, the sensors prepared from some crystalline polymers will tend to build cracks and detach from the support. Cracking of the sensing layers is also common for sol–gel materials. The glass-transition point (T_g) is another important parameter. One can expect a significant change in the sensing properties of the material if the T_g point is located within the measurement temperature range or the temperature corresponding to the T_g point is reached during sterilization by autoclaving. Finally, stability of the polymers towards oxidation via photosensitized singlet oxygen can be very important in some cases and particularly for trace sensors. Unfortunately, the number of truly oxidation-resistant polymers is very limited and most of them do not fulfil the requirements (i) and (ii).

One should mention that the materials of the second group often contain additional components. For instance, scattering substances such as titanium dioxide can be added into the sensing layer (or in the next layer above the sensing layer) to increase the brightness. A layer containing carbon black can be added to provide optical isolation i.e. to avoid saturation of the detector and photobleaching of the dye due to ambient light and to eliminate the disturbance of the excitation light in case of photosynthetic systems.

The design of oxygen sensitive nanoparticles is very similar to the probes of the second group i.e. they are based on phosphorescent indicators immobilized into polymer or a sol–gel. Often functional groups on the surface of the nanoparticle are introduced to improve dispersibility in water and provide additional properties, e.g. cell penetration ability. On the other side, the application area of such nanosensors is identical to that of the materials of the first group: intra- and extracellular imaging, assays in respiration vials and microplates.

Sensitivity and dynamic range of optical oxygen sensors is determined by two main factors: the phosphorescence decay time of the indicator τ₀ (Figure 1.3a) and the oxygen permeability of the polymer (Figure 1.3b). This, in turn, can be viewed as a product of oxygen solubility in the polymer and the diffusion coefficient. In other words, the properties of the indicator and the polymer are equally important for design of the sensing material. For instance, Figure 1.3 shows that the sensitivity of the sensor based on an Ir(iii) coumarin complex (Ir(Cs)₂acac) immobilized in highly oxygen-permeable ethylcellulose is similar to that of the material based on Pt(ii) tetraphenyltetrabenzoporphyrin (τ₀ much longer than for the Ir(iii) complex) embedded into less oxygen-permeable polystyrene.

Among the polymeric matrices poly(trimethylsilyl propyne) shows the highest oxygen permeability⁵  but tends to significantly change the properties over time. Silicon rubber and perfluorinated amorphous Teflon AF belong to stable polymers with very high oxygen permeability. Oxygen permeability of ethylcellulose, some Ormosils and Hyflon AD is moderately high. Polymers with moderate oxygen permeability such as polystyrene, polysulfone and poly(methyl methacrylate) are the most commonly used matrices due to their good chemical stability, low cost, good solubility in various organic solvents and optimal sensitivity at ambient conditions in combination with Pt(ii) porphyrins. Polymers with low oxygen permeability such as poly(vinylidene chloride) and poly(acrylonitrile) are not useful matrices but some of them are excellent support materials (poly(ethylene terephthalate)).

As a rule of thumb, one can say that the dynamic range of optical oxygen sensors based on bright indicators is about three orders of magnitude in pO₂/concentration of dissolved oxygen. The lower level is determined by the slope of the curve (Stern–Volmer constant) and noise of the read-out instrument. At the upper limit, decrease of the phosphorescence intensity due to quenching is too high to deliver an acceptable S/N ratio. Sensors based on less bright indicators will have a shorter dynamic range in the same conditions. As a consequence, it is impossible to design an oxygen sensor matching all possible applications. As was mentioned above, the combination of Pt(ii) porphyrins with polymers of moderate oxygen permeability results in sensors resolving from about 0.1 to 100 kPa pO₂. The sensitivity can be improved by roughly 1 order of magnitude via substitution of the Pt(ii) porphyrin via a Pd(ii) analogue. However, even this sensor is not suitable for some applications, for example for the quantification of oxygen in oxygen minimum zones³⁵  (nM concentration range). Trace sensors for these applications can be prepared via combining Pd(ii) (benzo)porphyrins and highly oxygen-permeable polymers such as Teflon AF or Hyflon AD.^36,37  For instance, an oxygen sensor based on PdTFPP embedded into Hyflon AD 60 (“LUMOS sensor”)³⁶  showed the LOD of about 0.5 nM. In terms of detection limit, resolution, sampling rate and simplicity of manufacture this optode favourably compares with the most sensitive electrochemical Clark-type sensor reported to date (STOX sensor),³⁸  Figure 1.4. This direct comparison undoubtedly illustrates the high potential of the optode technology for oxygen sensing.

As can be observed from Figure 1.3, quenching behaviour in polymers is more complex compared to the solutions where strictly linear Stern–Volmer plots are always observed. In contrast, in most polymers Stern–Volmer plots are not linear which can be explained by localization of the indicator in microenvironments of different oxygen permeability. The so called “two site model”³⁹  assumes the existence of only two microenvironments:

f₁ and f₂ - fractions of the total emission for each component (f₁ + f₂ = 1); K_SV¹ and K_SV² - the Stern–Volmer constants for each component. Eqn (1.2) is known to fit both the intensity and the lifetime Stern–Volmer plots very well, despite that, in the second case, the fit is not physically meaningful.

Luminescence intensity is notoriously known to be a rather ambiguous parameter, which depends on many factors including intensity of the excitation light, sensitivity of photodetector, scattering in the sample etc. Therefore, referencing in sensing and imaging is essential for precise quantification of pO₂. Fortunately, phosphorescence decay time represents a self-referenced parameter, which is not affected by the above-mentioned factors. Phosphorescence decay time can be conveniently measured with low cost compact phase fluorometers. Several major suppliers (PreSens, Pyro Science, Ocean Optics etc.) offer optical oxygen sensors in combination with the dedicated read-out devices and these products are nowadays widely applied in industry and academia. For imaging of oxygen distribution with optical sensors both the frequency domain (FLIM camera from PCO) and the time domain (FLIM/PLIM from Becker and Hickl) are useful. Many set-ups, however, are not equipped with detectors for lifetime imaging but rather with several photodetectors, which enable simultaneous intensity measurement in different spectral ranges. In this case, referenced measurement of luminescence intensity is very useful. To obtain a suitable material a fluorescent dye is immobilized along with the oxygen indicator. Clearly, both dyes should be excitable at the same wavelength but emit in different spectral ranges. This requirement is easily met since the Stokes shift is much smaller for the fluorescent dyes compared to phosphorescent oxygen indicators. In the case of light harvesting systems, the excitation light is absorbed by the added fluorophore and the energy is transferred to the oxygen indicator via Förster Resonance Energy Transfer (FRET) mechanism. This approach not only yields ratiometric intensity measurement (utilizing residual fluorescence of the fluorophore) but also allows for significant increase of fluorescence brightness due to high molar absorption coefficient of the added dye.⁴⁰  It should be kept in mind that the ratiometric referencing scheme does not compensate for photobleaching or leaching of the oxygen indicator or the reference. The reference dye might show much higher photobleaching rates in the presence of oxygen indicator due to the formation of highly reactive singlet oxygen. Leaching can be critical in the case of nanoparticles due to their high surface to volume ratio and therefore much more efficient interaction with the environment. Here, covalent immobilization of indicators is strongly preferable. Polymeric materials based on conjugated polymers (combining dual function of the matrix and fluorophore), where an oxygen indicator is covalently grafted into the polymer and acts as a FRET acceptor, were also shown to be highly promising for imaging applications.^24,41  Importantly, ratiometric character is also preserved under two-photon excitation since the conjugated polymer acts as an excellent two-photon antenna.²⁴ 

Although optical oxygen sensors belong to very robust analytical tools, precise oxygen determination can be compromised by a number of errors, cross-sensitivities and artefacts. As for all chemical sensors, oxygen optodes are cross-sensitive to temperature which affects the phosphorescence decay time of the indicator (τ₀ decreases due to thermal quenching) and the Stern–Volmer constant (increases in most cases due to faster diffusion of oxygen). These effects must be compensated for in an additional temperature measurement. Calibration might be another critical issue. When the exact form of the Stern–Volmer plot is known, two-point recalibration is almost always sufficient. Whereas anoxic conditions can be easily obtained using high purity gases or a solution of an oxygen scavenger (sodium sulfite, sodium dithionite, glucose solution with added glucose oxidase), the calibration in air is more critical. If the sensor is calibrated in the gas phase, temperature, atmospheric pressure and relative humidity should be measured. Although the last parameter is constant in water, it must be ensured that 100% air saturation at the given temperature is obtained. Colder or warmer solutions brought to the calibration temperature might be over- or undersaturated, respectively.

Evidently, the above errors can be avoided by careful calibration; however, the measurements can still be compromised by several artefacts that originate from the nature of the sensing material itself. As was discussed above, optical oxygen sensors, in theory, do no consume oxygen, which is highly advantageous over other analytical methods. However, this assumption might not be true in reality. Photosynthesized singlet oxygen belongs to highly reactive species and may react with the polymer before it deactivates back to the triplet state. As a consequence, not only is the matrix chemically modified (which may be accompanied e.g. by the change in oxygen permeability or accumulation of luminescence quenchers) but the analyte is also consumed. In fact, as was demonstrated for materials bearing an additional gas-blocking layer, most of the commonly employed polymers do react with singlet oxygen.⁴²  However, due to abundance of the analyte at air saturated conditions oxygen consumption might not be detectable at all for comparably fast responding sensors. On the other hand, the situation is radically different in the case of trace sensors. Here, oxygen consumption can dramatically affect the results.⁴³  For instance, the sensor based on a phosphorescent BF₂-chelate immobilized in polystyrene and equilibrated with 0.05% vol. O₂ showed a more than 10-fold decrease in quenching upon increasing the intensity of the excitation light by about 20-fold. This effect was even more pronounced at lower pO₂ so that the Stern–Volmer plot showed a characteristic upward curvature. Notably, the same matrix is fully adequate for measurements at physiological conditions even at comparably high light intensities typical e.g. for fiber-optic microsensors.

Evidently, oxygen consumption can be eliminated by making use of oxidatively robust matrices. Amorphous perfluorinated polymers such as Teflon AF, Hyflon AD and Cytop represent excellent candidates for this purpose. Unfortunately, conventional phosphorescent indicators are poorly soluble in these matrices and tend to aggregate even at low concentrations. Modification of indicators with perfluoroalkyl chains^31,37  renders them compatible with the perfluorinated polymers but the synthesis and purification of the dyes are not straightforward.

Trace, and particularly ultratrace, oxygen sensors also show several effects, which alter the form of the response curve.⁴³  First, the very long emission lifetime of the oxygen indicator can result in accumulation of the dye in the triplet state and therefore depopulation of the ground state. This effect affects the luminescence intensity plots but not the decay time plots. It can be minimized by using lower light intensities. Second, since dye concentration in polymers is typically high (0.01 mol l⁻¹) accumulation of the indicator in the triplet state can result in triplet–triplet annihilation, i.e. a process generating one dye molecule in the S₁ state and one molecule in the S₀ state upon interaction of two molecules in the T₁ state. Triplet–triplet annihilation results in shorter lifetimes and lower luminescence intensities. Since the efficiency of this process is proportional to the concentration of the dye in the triplet state, the negative effect of triplet–triplet annihilation can be minimized by using lower intensities of excitation light and by decreasing dye concentration in the polymer. Finally, the lifetime of singlet oxygen in some polymers can be rather long (>1 ms in perfluorinated polymers)⁴³  which can convert part of the analyte into the form not available for luminescence quenching. This effect can become significant at high light intensity (=high concentration of triplets) in combination with low oxygen concentration since one oxygen molecule has to quench several dye molecules within their lifetime. This effect can be minimized by using lower dye concentration resulting in higher oxygen to dye ratio. It can be concluded that such undesired effects can be minimized by using lower dye concentrations in the matrix and lower excitation intensities (the latter also helping to minimize oxygen consumption), but the trade-off is the lower signal-to-noise ratio. Here dyes with the highest phosphorescence quantum yields and optimized set-up, which minimizes the light losses, are beneficial. Fortunately, the above effects are marginal under “normal conditions” i.e. high pO₂ values and comparably low light intensities.

It can be concluded that rational design of optical oxygen sensing materials is necessary for them to show the required performance in envisaged applications. Although it is not feasible to create an oxygen optode suitable for all potential applications, a great variety of oxygen indicators and polymeric matrices allow for virtually unlimited possibilities providing great flexibility in respect to optical properties and dynamic range of the sensors. Whereas many existing sensors already enable reliable measurements in many fields of science and technology, new materials for some demanding and emerging applications have yet to be developed. These include, for instance, a sensor operating in organic solvents with performance comparable to that of the state-of-the-art sensors for aqueous phase or a fast responding microoptode showing no drift upon prolonged usage at very high measurement rates (5–10 Hz).