Introduction

The elementary steps underlying the reversible addition and removal of electrons from matter—metals, molecules or materials—are the basis of redox chemistry, electrocatalysis, and electrochemical energy storage. Since the electrochemical potentials of such reactions are only comparable within one solvent, the IUPAC suggests the use of the ferrocenium/ferrocene redox couple Fc+/Fc as a reference1,2. Due to the weak solvent dependence of its potential, other electrochemical potentials can be compared for various liquid media (Ferrocene Assumption)3.

By contrast to the multitude of reagents for electronation (addition of an e−) down to the potential level of solvated electrons, accessible and reliable reagents for deelectronation4 (removal of an e−) at potentials higher than about +0.5 to +1 V vs. Fc+/Fc are scarce, but would be useful5. Note that we use electronation4 and deelectronation in their strict sense, i.e., addition and removal of electrons in innocent reactions, respectively, hence without non-innocent complications such as complex formation, substitution or degradation reactions. For reasoning, see the reports by Radtke6 and Himmel7 et al. The typical systems available, for example, salts of the inorganic ions NO+ and Ag+, are applied in many fields of chemistry (organic synthesis8,9,10, nitrosonium complexes11) and the material sciences (batteries12, solar cells13, silver electrode14,15,16) to remove electrons from the system studied. Such commercial deelectronator salts include classical, but rather small weakly coordinating anions (WCAs), like [BF4]− or [MF6]− (M = P, As, Sb)17. Yet, solutions of such salts de+[WCA]− (de+ = Ag+, NO+) show a strong solvent (S) dependence5 of their potentials vs. Fc+/Fc, ranging for NO+ from +0.56 (S = DMF) to +1.00 V (S = CH2Cl2) and for Ag+ from −0.07 (S = MeCN) to +0.65 V (S = CH2Cl2)5,18,19,20,21. This potential dependence arises from partially strong interactions of the NO+ or Ag+ cations with the solvent molecules S. In addition, especially the systems with the highest de+ potentials—published in weakly basic and little interacting low-polarity organic media like CH2Cl2—suffer from the low solubility of the salts de+[WCA]−. This leads to reduced activities a(Ag+, S) or a(NO+, S) that may further be lowered by severe ion-pairing of de+[WCA]− in S. Equally important, the de+ ions form rather strong bonds to classical aromatic solvents19,22 that greatly lower the accessible potential. Moreover, solvated NO+ is only fleetingly stable in arenes, as the formed Wheland complex23,24 serves as an entry point to an electrophilic aromatic substitution reaction18,19.

Fundamentally, both problems—ion-pairing and solvent coordination, substitution, or degradation—may be overcome by using the salt de+[WCA]− in a combination of an almost non-interacting inert, but polar and redox-robust (aromatic) solvent with a redox-stable very good WCA. Despite being polar with a large dipole moment, the solvents have to act as weak bases (thermodynamics) and nucleophiles (kinetics). Together with large WCAs, such solvents enhance the solubility of de+[WCA]− by solvating de+ and [WCA]− and yield dissociated ions that have high activities a(de+, S) also in low dielectric media S (Fig. 1).

Fig. 1: Summary of the investigated solvents, cations and anions.
figure 1

Top: Structural formulae of the strategically substituted and dipole-maximized polar fluorobenzenes xFB with x = 1–5. Bottom left: Ball and Stick models of the Ag+ and NO+ cations. Bottom right: The WCAs [pf]− (= [Al(ORF)4]−) and [al-f-al]− (= [(FRO)3Al-F-Al(ORF)3]−; RF = C(CF3)3) in a representation with the electrostatic potential plotted on an isodensity surface (0.025 e− B−3) using like cut-off values and calculated at the (RI-)BP86/def2-TZVPP level of theory. The scale bar right indicates areas of more negative (red), neutral (green) and positive (blue) surface potentials.

Full size image

Our entry to the field was the preparation of the NO+ and Ag+ salts of the large WCAs [pf]− (= [Al(ORF)4]−)17,25 and [al-f-al]− (= [(FRO)3Al-F-Al(ORF)3]−; RF = C(CF3)3)17,26 (Fig. 1) that enable de+[WCA]− solubility in low-polarity solvents, i.e., CH2Cl2, 1FB and 5FB, but also the more polar 2FB, 3FB and 4FB (Fig. 1). We routinely use these WCA-reagents for syntheses with Ag+27,28,29 or NO+ ions30,31,32,33,34,35, for deelectronation34,35,36,37,38,39,40,41,42 or dehalogenation reactions43,44. More than 100 groups worldwide utilize the advantageous properties of [pf]− and closely related WCAs for applications in diverse fields, i.e., deelectronation45, supporting46,47,48 and battery49,50,51,52,53 electrolytes, organic solar cells54,55,56, photoacids57, catalysis58, etc.59,60,61.

This account reports how the above ion-pairing and solvent problems upon application of de+ salts are overcome by using strategically substituted and dipole-maximized polar fluorobenzenes62 xFB (x = 1–5) partnered with salts of the large63 redox-stable WCAs [pf]− and [al-f-al]− shown in Fig. 1. Especially the higher fluorinated 4FB and 5FB solvents are very weak ligands and induce, together with the aluminate WCAs, record high de+-potentials in xFB solution of up to +1.52 V for NO+(4FB)/NO and +1.50 V Ag+(5FB)/Ag, both vs. Fc+/Fc in S. In addition, problems with reported literature potentials in standard solvents were spotted19, addressed and corrected in this work5,18. Two innocent deelectronator reagents push the positive potential accessible in xFB solution to synthetically highly useful positive values of up to +1.89 V vs. Fc+/Fc. Both overcome the observed non-innocent reactivity of the Ag+ or NO+ ion33,64,65,66 and make use of the very large electrochemical xFB stability windows that exceeds 5 V at positive upper limits between +1.82 V (1FB) and +2.67 V (5FB) vs. Fc+/Fc.

Results

First, we describe the missing solvent properties of fluorobenzenes, before investigating the de+ potentials in standard and xFB solution with large WCAs, validate and relate the measured potentials to electrolyte species with crystal structures, IR spectra and high-level quantum chemical calculations. Finally, we utilized these strategies to even further push the positive potential limits with perfluorinated arenium cations.

Permittivity and principal solvent properties of xFBs

To understand the effect of the solvent, we determined fundamental xFB (x = 1–6) properties, such as dielectric constants, dipole moments, viscosities, and densities. The static dielectric constant εs is a measure for the dipole density and polarizability within a solvent S. Hence, it governs the stabilization of any ion in the solvent and is also used as a polarity descriptor for many quantum chemical solvation energy calculations. Dielectric relaxation spectroscopy was used to measure εs, the relaxation time, τ, and the infinite frequency permittivity, ε∞, of the neat xFB solvents as well as their binary 1:1 mixtures between 5 and 30 °C. From the dielectric relaxation strengths εs-ε∞, we derive67 the effective dipole moment, μeff, of xFB in the liquid phase (Supplementary Note 5). For completion, the in part unknown densities, kinematic/dynamic viscosities of all xFB solvents were measured with an Ubbelohde-viscometer at 23.1 ± 0.2 °C and were added to Table 1 (cf. Supplementary Note 9).

Table 1 Relaxation parameters of the neat solvents at 25 °C: static dielectric constant εs, infinite frequency permittivity ε∞, relaxation time τ, and the derived effective dipole moments μeff. Kinematic (ν) and dynamic (η) viscosities at 23.1 ± 0.2 °C as well as densities ρ of xFB at the indicated temperatures [T]
Full size table

Pure xFB solvents

The εs values of neat xFB at 25 °C in Table 1 agree, where known (1–2FB, 5–6FB)68,69, within ±0.5 units to the literature values measured between 22 and 25 °C. εs(xFB) increases with increasing fluorine content from 1FB to 3FB, but decreases upon further fluorination. As εs is largely determined by the alignment of molecular dipoles according to an external electric field, εs(xFB) mostly reflects the trends of the molecular dipole moments of xFB, and the μeff(xFB)70 values in Table 1 agree well with the values for isolated xFB molecules obtained from DFT calculations. They are only moderately affected by a slight preferential parallel alignment of xFB’s molecular dipoles in the liquid phase (Supplementary Note 9).

εs decreases with increasing temperature for all xFB solvents, consistent with thermal motion countering the correlations of the molecular dipoles according to the external field. The slopes of εs vs. temperature correlate with μeff and are steepest for the most polar 3FB, 2FB and 4FB (Supplementary Fig. 101). Cooling increases εs(3FB) considerably to 24.6 at 5 °C, which is 2.5 units higher than at 25 °C. This may be used in favor for synthetic purposes.

Binary 1:1 xFB mixtures

The trends for the binary 1:1 mixtures discussed in Supplementary Note 5 parallel those found for the neat solvents: With increasing fluorine content of one component mixed with another xFB, we find εs to peak for mixtures with 3FB, irrespective of the second component (Supplementary Figs. 105b–109b). Hence, binary mixtures of xFB’s yield solvents with a tunable permittivity between εs = 2–22 at 25 °C, which can be exploited in specific applications. The moderate parallel correlation of the dipoles observed for the neat xFBs are reduced in the mixtures, as apparent from the effective dipole moments μeff (Supplementary Fig. 110). All binary 1:1 mixtures also show a monotonic increase of εs with decreasing temperature (Supplementary Fig. 111).

Comparison of ε s(xFB) to standard solvents

The high static dielectric constant of 3FB of εs = 22.1 stems from the very high dipole moment of μeff = 4.7 D. It even exceeds the room temperature value of the prototypical polar, but coordinating solvent acetone (20.7) and by far exceeds that of the typically for synthesis used aprotic non-basic, mainly non-coordinating media like CH2Cl2 (8.9) or 1,2-Cl2C2H4 (10.8)68. In addition, also non-basic, but polar liquid sulfur dioxide (13.8 at 25 °C), e.g., used to stabilize carbocations and many other reactive cations, is surpassed71. We note that εs of 2FB (14.2) and 4FB (12.6) also exceed those of CH2Cl2 and 1,2-Cl2C2H4 and reach that of SO2. In addition, εs of 1FB and 5FB are similar in magnitude, and it appears that also the affordable 5FB is a hitherto underestimated non-basic solvent for reactions that need some polarity, i.e., reaching that of the frequently used, but toxic HCCl3 (4.8) or chlorobenzene (5.6). To fully understand the solvent’s effect on ionic solutes—besides considering ionic species being stabilized by the solvent as an effective medium with dielectric permittivity εs—also specific interactions between the solvent and the ionic species have to be considered. Such explicit xFB-solvation is discussed in the following sections.

Cyclovoltammetric measurements in xFB and standard solvents

Demand for reevaluation of de+ potentials in standard solvents

The published5 formal potentials of the Ag+(S)/Ag system in acetonitrile (S = AN) and dichloromethane (S = CH2Cl2) are by 0.61 V markedly different. In this light, the close similarity of the potentials of the NO+(S)/NO system in AN (+0.87 V) and CH2Cl2 (+1.00 V) was somewhat surprising5,18,19. Apparently, these opposing solvent effects are due to experimental challenges when measuring the potentials that are caused by the very low solubility of NO+[BF4]− in CH2Cl219. Therefore, Kochi et al.19 used NO(g) dissolved in CH2Cl2 for the measurements. Yet, in a footnote, the authors honestly stated that they were unable to accurately prepare solutions of gaseous NO with known concentrations19. This suggests this potential value to be considerably too low due to the very low activities a(NO+, CH2Cl2) accessible with [BF4]− counterion. In addition, several formal potential values of de+ ions included in the highly-cited 1996 Geiger-Conelly review5 are given as estimates. Similar to the NO+[BF4]− case above, some of these values may be drastically affected by ion-pairing. In the course of our investigations, we realized that xFB solutions of NO+[pf]− or NO+[al-f-al]−, although being colored due to the formation of Wheland intermediates (see below)37, are stable for many days at room temperature. Hence, a comprehensive (re-)investigation of the de+ potentials in standard as well as fluorobenzene solvents appeared timely.

Methodology for CV

With the help of a large WCA, we study the principal magnitudes decisive for any possible use of a solvent in electrochemistry. Thus, using 100 mm [NBu4]+[pf]− as supporting electrolyte in the solvent S, the negative potential limit Eneg, the positive potential limit Epos and the resulting widths of the potential stability windows (ECW) of xFB’s and several of the above stated standard solvents were determined. Note that other large WCA salts including [B(C6F5)4]− or [B(ArF)4]− (ArF = 3,5-(CF3)2C6H3) and that were recommended to be used as supporting electrolytes especially in low permittivity media72,73,74,75, reduce the ECW due to decomposition of the anions at higher potentials: Waldvogel et al. demonstrated that both WCAs decompose at anodic limit potentials76 of only +1.25 to 1.78 V vs. Fc+/Fc in AN in electrosynthetic yields with formation of the respective fluorinated biphenyls. Hence, the potentials of Ag+, NO+ and in addition that of the organic amine TBPA0 (TBPA0 = N(4-Br-C6H4)3)5, supposed to be less affected by medium effects, were investigated vs. Fc+/Fc. The general CV-measurement setup is described in Supplementary Note 3.1. Full results and all CV traces at scan rates between 20 and 200 mV s−1 are included in Supplementary Notes 3.2 to 3.7 of the Supplementary Information.

The values of the negative limit Eneg, the positive limit Epos and the resulting ECW-widths of xFB and standard solvents (Supplementary Note 3.3) are collected in Table 2, known literature values in Supplementary Table 6.

Table 2 Electrochemical properties of selected solvents S and redox active species therein
Full size table

Comparison to literature data for standard solvents

Our ECW-widths of the standard solvents measured with the supporting electrolyte [NBu4]+[pf]− in Table 2 are much larger than the corresponding values found in the literature (Supplementary Table 6). These differences are probably induced by the use of different supporting electrolytes (often [NBu4]+[ClO4]− or [NBu4]+[PF6]−) that influence the ECW. In part, a standard calomel electrode (SCE) was used as a reference electrode (RE) including a Liquid Junction Potential (LJP) between the aqueous RE and the non-aqueous solvent6,21. Further, in some cases, the supporting electrolyte used was not given and prevents a more accurate comparison, although measured with the same RE as in our work. Nevertheless, the literature values help put our ECWs into context. Among that, the used electrolyte salt is key for the ECW: Own experiments in CH2Cl2 and using either [NBu4]+[pf]− or [NBu4]+[PF6]− as supporting electrolyte show that the [pf]− salt gives an ECW which is by about 1 V wider than that of the [PF6]− salt (Table 2, Supplementary Fig. 26). The larger window, which is induced by the [NBu4]+[pf]− supporting electrolyte with reduced ion-pairing, is a favorable feature that can be exploited for further developments in the field of electrochemistry in general.

Performance of the xFB solvents

The Epos potential shows a progression from 1FB to 5FB and increases from +1.82 V (1FB) up to +2.67 V (5FB), consistent with the rising ionization energies (IEs) of these solvents with increasing fluorination (Table 2). The Eneg potentials of 1-4FB are rather similar and around −3.1 ± 0.1 V. Only the most fluorinated 5FB shows the highest Eneg potential at −2.37 V, as expected due to its pronounced77 C-H acidity. Altogether these potentials lead to ECWs exceeding 5 V that peak for 4FB at 5.51 V. They are compatible with a wide range of electrochemical syntheses. Hence, together with the poor capacity of the xFB solvent molecules to serve as ligands (see also the following sections), the basic electrochemical performance of xFB with [NBu4]+[pf]− as supporting electrolyte salt is very promising.

After having established the ECW-widths, we proceeded to determine the potentials of the electroactive salts de+[pf]− in these solvents and the effect of the—standard and xFB—solvent on their potentials. Our measurements of the redox potentials E of 10 mm solutions of the cations de+ = NO+ and Ag+ with [NBu4]+[pf]− as supporting electrolyte in the solvents xFB, CH2Cl2, 1,2-Cl2C2H4, dimethylformamide (DMF) and AN are collected in Table 2 and, for the standard solvents, compared to published data. Where appropriate, E1/2 values were converted into formal potentials E°´ = E1/2, i.e., for the Ag+/Ag and the Fc+/Fc systems. This is justified, if the diffusion coefficients of the electronated and deelectronated form of the redox system are identical. For the Fc+/Fc system, this was verified to be almost true in many solvents S by diffusion constant PFGSE-NMR measurements. Hence, from the measured diffusion constants, E°´ and E1/2 were calculated to be similar within about 10 mV in 1FB and 5FB, and less for the other solvents (Supplementary Note 3.2.1). However, for the NO+/NO system, where NO+ forms a Wheland complex [NO(xFB)]+ (see below), which diffuses much slower than a neutral NO molecule, we give E1/2(NO+(S)/NO). In addition, we measured the potentials of the deelectronation of 10 mm solutions of neutral TBPA0 giving the frequently used radical cation [TBPA]∙+, which is well-known from the magic blue salt (Supplementary Note 3.4.2)5,78. Using the neutral underlying the radical cation [TBPA]∙+ with the [pf]− counterion from the electrolyte solution prevents electrochemical side reactions of the [SbCl6]− anion, compared to using the magic blue salt directly. Exemplarily we display a stack plot of the measurements of Fc+[pf]− solutions in xFB and CH2Cl2 and with respect to the Ag+(S)/Ag redox system in Fig. 2.

Fig. 2: Formal potentials of the Ag+ cation in different xFB solvents.
figure 2

2nd cycles of the CVs (v = 100 mV s−1) of solutions of Fc+[pf]− (c = 10 mm) in the solvents xFB (x = 1–5) and CH2Cl2 with the conducting salt [NBu4]+[pf]− (c = 100 mm) corrected to the formal potential E°’ by −0.118 V (cf. Supplementary Note 3.2.1).

Full size image

Note that the systems are electrochemically quasi-reversible since the peak-to-peak separation of the waves in the CVs increases with the scan rate. Exemplarily, the E1/2 potential of a 10 mm solution of the salt NO+[al-f-al]− with the least coordinating, largest WCA was measured in the least polar and coordinating solvent 5FB. The potential influence of the counterion and possible ion-pairing effects are expected to be most prominent in this solvent. Yet, we find identical E1/2(NO+(5FB)/NO) values for both WCAs (+1.47 V), suggesting that the measured potential values are insensitive to the WCA counterion even in low-polarity 5FB.

Potentials of the de+/de systems in standard solvents S

The Ag+(CH2Cl2)/Ag and NO+(CH2Cl2)/NO potentials determined with [pf]− WCA in Table 2 differ from the published values by +0.23 V (Ag+ with [PF6]− counterion) and +0.40 V (NO+ with [BF4]− counterion). By contrast, the NO+(S)/NO potentials in S = DMF; AN and the Ag+(AN)/Ag potential agree within 0.04 V with literature values. Apparently, DMF and AN are polar and coordinating enough so that ion-pairing plays a minimal role, and the potentials are independent of the counterion. To prove this point, we measured the NO+(CH2Cl2)/NO potential using 10 mm NO+[PF6]− with 100 mm [NBu4]+[PF6]− as supporting electrolyte. This gave E1/2 as +0.95 V (Supplementary Note 3.6.1.3), close to the published value of +1.00 V19. But note, in this work19 CH2Cl2 was saturated with an unknown concentration of NO gas, which may account for some differences19. Therefore, it can be assumed that the [BF4]− as well as [PF6]− anions form a strongly bound contact-ion pair in solution. Due to the rather small amount of the solvated free NO+ cation, its activity and hence its potential can be markedly lowered. By contrast, the NO+ solution of the corresponding [pf]− salt in CH2Cl2 apparently provides activities unaffected by ion-pairing. Hence, the interaction with the [PF6]− anion appears to reduce the NO+-activity in solution by about 7 orders of magnitude corresponding to a potential drop of 7∙0.059 V = 0.413 V, assuming ideal Nernst’ian behavior. A similar consideration accounts for the 0.23 V difference between the Ag+ potentials in CH2Cl2: Again, the small [PF6]− anion induces strong ion-pairing and lowers the activity of Ag+ ions by four orders of magnitude, if compared to the corresponding [pf]− salt in CH2Cl2. Thus, the advantage of the [pf]− WCA as counterion to provide largely improved activities a(de+, S) of the electroactive ion in less polar and less coordinating solvents S like CH2Cl2 becomes very evident.

Potentials of the de+/de-systems in xFB solution (de = Ag, NO)

For E(de+/de, xFB), we find a clear increase in the potentials from 1FB to 4FB. From 4FB to 5FB, the NO+(xFB)/NO potential experiences a slight reduction, while that of Ag+(xFB)/Ag further increases. Very high potentials of up to +1.52 V (NO+(4FB)/NO) and +1.50 V (Ag+(5FB)/Ag) vs. E°´(Fc+/Fc, S) are reached. In addition, by appropriate choice of the solvent, the NO+(S)/NO potential can be tuned by 0.41 V and the Ag+(S)/Ag potential even by 0.75 V. Apparently, this variation is induced by the reduced interaction energies of the Wheland complexes [de(xFB)]+solv upon increasing the degree of fluorination of xFB. This complies with the IEs collected in Table 2 and the color of the NO+ solution (see below, also Supplementary Note 3.5.4).

Validation with Born-Fajans-Haber-Cycles

From the measured E1/2 potentials of 10 mm solutions of Fc+ and NO+ versus E(Ag+(10 mm, S)/Ag) and NO+ versus E(Fc+(10 mm)/Fc(10 mm), S) within the same solvent S, triangular Born-Fajans-Haber-Cycles (BFHC) can be constructed. Knowing two out of the three values in the BFHC, the third can be calculated (shown for 5FB and 1,2-Cl2C2H4 in the method section, and all data in the Supplementary Notes 3.5.3 and 3.7.3). With this relation, the internal consistency of all the measured data for nine solvents was checked by calculating mean |ΔE| and |ΔG| uncertainties that are lower than 0.05 V or 5 kJ mol−1 toward the absolute value. Hence, our measurements are consistent within this margin.

Interaction of de+[pf]− with the solvents xFBs, CH2Cl2 and 1,2-Cl2C2H4

The variation of the deelectronation potentials with solvent and the high values in xFBs suggest that despite their polarity, the direct coordinative interaction capacities of the xFB solvents with the electroactive de+ ions depend on the degree of fluorination of xFB and may be weak. To explore the coordination capacities of the solvents, we investigated their solvated single-crystal X-ray diffraction (scXRD) structures (mainly Ag+-complexes) and IR spectra (NO+-salts). Later this information is discussed in the context of their energetics.

Silver salts

The soft silver cation Ag+ is rather carbophilic40 and a large number of solvent complexes [Ag(arene)n]+ are published, including a very systematic investigation22 of the reference system [Ag(arene)3]+[B(C6F5)4]− with the very good WCA [B(C6F5)4]−. Yet, only a few [Ag(xFB)n]+ salts are known26,79, the majority of which were published by our group as [pf]− or [al-f-al]− salts. For S = CH2Cl2, the types [Ag(S)3]+ and [Ag(S)4]+ are known80,81.

We note that the structural type [Ag(S)1]+[pf]− with solvent-separated cations and anions is absent. Rather, tight ion-pairs {Ag[pf]}ip and {(S)Ag[pf]}ip are formed. The most relevant molecular (cation) structures of the structural types were collected and exemplarily compiled in Fig. 3.

Fig. 3: Coordination environments of the Ag+ cation in different xFB solvents.
figure 3

Exemplarily selected molecular (cation) structures of the most important structural types. For salt structures, the counterion [pf]− was omitted. In the tight ion-pair {Ag[pf]}ip, three further contacts to F-atoms of other [pf]-units in the unit cell are included. Thermal ellipsoids set at 50% probability. a: ref. 79 b: ref. 17 c: The structure of {Ag[pf]}ip crystallized from 5FB and 6FB is isotypic to that obtained earlier from isoperfluorohexane29.

Full size image

scXRD NO+ salts

We succeeded to crystallize [NO(2FB)]+[pf]−, thus establishing a η6-NO+-Wheland complex with a fluorinated arene (Fig. 4A). Hitherto, only NO+-Wheland complexes of electron-rich arenes (e.g., (alkyl-)benzenes) were crystallized. The electron-deficiency of 2FB lengthens the N…centroid distance in the [NO(2FB)]+ complex (2.255(9) Å) by more than 0.1 Å compared to the average N…centroid distances of known 1:1 adducts82,83,84. In addition, the C(F) carbon atoms have longer NO+ contacts of 2.759(9)–2.794(7) Å than their C(H) counterparts (2.49(1)–2.69(1) Å). The O-N…centroid angle, herein 164.8°, is highly flexible in NO+-Wheland complexes due to the π-type interaction from the highly delocalized HOMOs of the arene toward the π* LUMOs of the NO+. It is bent toward those C(H) atoms with large coefficients of the HOMO.

Fig. 4: Interaction of the nitrosonium cation with xFB solvents.
figure 4

A Molecular structure of the cation in solid [NO(2FB)]+[pf]−, color code: fluorine—green, oxygen—red, nitrogen—blue, carbon—gray, hydrogen—light gray, anion omitted for clarity, thermal ellipsoids set at 50% probability. B ATR-IR spectra of the solid residues obtained from evaporated solutions of NO+[pf]− in xFB (x = 1–6) between 1900 and 2400 cm−1. The spectra from 4FB to 6FB are multiplied by the factor 5, due to the low intensity of the NO band of non-coordinated NO+ (calc. at: 28 km mol−1) in comparison to the intensities of the NO bands in the Wheland complexes (calc. at: >1000 km mol−1). C The shape of the deformation density Δρ(1) corresponding to ΔEorb(1) (charge flow: red → blue) and the participating fragment orbitals of NO+ and 2FB at the BP86(D3BJ)/TZ2P level. The eigenvalue |v1 | gives the amount of the charge migration in e−. D EDA-NOCV results for [(xFB)NO]+ complexes using NO+ and xFB as interacting fragments at the BP86(D3BJ)/TZ2P level given in kJ mol−1. E Photo of the solutions of NO+[pf]− in xFB (x = 1–6) at a concentration of 1 mm. Note the decreasing color intensity with increasing degree of fluorination of xFB.

Full size image

Overall, the scXRD structures are indicative of weaker Ag+–xFB coordination for increasing x, and also, 2FB is weaker bound in the NO+-Wheland-complex than in non-fluorinated counterparts.

IR-data NO-stretches

The π-donor properties of different arenes in Wheland complexes with the nitrosonium cation were extensively studied in the 1990s by IR spectroscopy probing the redshift of the νNO vibration relative to that of free NO+ (2340 cm−1 in NO+[PF6]−)82. While νNO of NO+[PF6]− is shifted to 2075 cm−1 in benzene, the redshift is significantly larger in hexamethylbenzene with νNO at 1885 cm−1. Extending this series, we measured ATR-IR spectra of concentrated NO+[pf]− solutions in xFB and observed broad νNO bands that increase from 2007 to 2049 cm−1 in 1FB to 3FB (Supplementary Fig. 113). Unexpectedly, the bands of NO+[pf]− dissolved in xFB (x = 1–3) are redshifted, if compared to the NO+[PF6]− complex with benzene. This may result through interaction of the Wheland complex with the small [PF6]− counterion, i.e., the formation of the ion-pair {(C6H6)NO[PF6]}ip reducing the interaction strengths of NO+ with benzene and increasing νNO. In line with the strong decolorization of the NO+[pf]− solution in going from 3FB to 4FB (Fig. 4E), νNO was not visible in 4–6FB. Due to the low intensity of the band νNO in non-coordinated NO+, this stretching band was invisible in the solution. In addition, spectra of the evaporated residue of NO+[pf]− solutions in xFB were measured directly on the ATR-IR unit inside a glovebox to investigate the interactions of NO+ with xFB in the solid state. For 1–3FB, the spectra were similar to the spectra in solution, with νNO increasing from 1999 (1FB) to 2029 cm−1 (3FB). For residues stemming from 4 to 6FB solution, a weak νNO band was observed at 2338 cm−1 (Fig. 4B) and hence essentially at the same position as in neat NO+[pf]−. In agreement with this, we observed the solvent-free crystallization of NO+[WCA]− from 6FB (WCA = pf, al-f-al). Overall, we assume that the xFB…NO+ interaction for x = 4–6 is too weak to allow for co-crystallization as a Wheland complex.

Molecular structures as a function of solvent properties

Analysis of Figs. 3 and 4 gives some important clues on the species that exist in solution: Although εs(3FB) is the highest of all investigated solvents and one would therefore expect solvated ions to prevail (i.e., reduced ion-pairing), the scXRD structure corresponds to a tight ion-pair {(3FB)Ag[pf]}ip that saturates the Ag+ coordination sphere by contacts to the [pf]− ion and one 3FB solvent molecule. And although εs for 2FB and 4FB are virtually the same, only 2FB exhibits solvent-coordinated cations in the crystal structures [Ag(S)n]+[pf]− with n = 2 at room temperature and n = 3 for low-temperature crystallization. By contrast, 4FB forms a tight ion-pair {(4FB)Ag[pf]}ip similar to that found for 3FB. A similar consideration holds for the 1FB/5FB couple with comparable εs. Yet, the salt [Ag(1FB)3]+[pf]− forms in 1FB at all conditions tested, and from 5FB a tight ion-pair {Ag[pf]}ip without interaction to any solvent molecule is formed. Hence, it appears that with increasing fluorine content in xFB, the Ag+ ion complexation enthalpies ΔrHcomplex get sequentially weaker, so that starting with 3FB, the saturation of the Ag+ coordination sphere by the counterion [pf]− is preferred over the solvent. Apparently, ΔrHcomplex for 5FB is so low that even the weak contacts of the tight ion-pair {Ag[pf]}ip to fluorine atoms of other [pf]− ions in the lattice are more favorable than coordination of a 5FB solvent molecule. Interestingly, in CH2Cl2 with a εs value between 1FB/5FB and 2FB/4FB, clear salt structures [Ag(S)n]+[pf]− with n = 3 at room temperature and even n = 4 at low temperature are formed. In 1,2-Cl2C2H4, exclusively the salt type [Ag(S)3]+[pf]− with the more favorable 5-ring chelates is formed. Therefore, the interaction of CH2Cl2 or 1,2-Cl2C2H4 with the Ag+ ion must be at least similar in strength to 1FB, slightly stronger than Ag+ with 2FB, but much stronger than Ag+ with 3FB, 4FB and 5FB. This agrees with the formal potentials collected in Table 2.

Quantum chemical calculations

To quantify the extent of stabilization of the solvated species [de(S)n]+ (n = 1–4) and to further evaluate the presence of ion-pairs {Ag[pf]}ip,solv and {(S)Ag[pf]}ip,solv, the structures of all complexes [de(S)n]+ (de = Ag, NO), {Ag[pf]}ip,solv and {(S)Ag[pf]}ip,solv were optimized at the (RI-)BP86(D3BJ)/def2-TZVPP85 DFT86 level of theory, refined and extrapolated to the complete basis set limit (CBS) in a series of DLPNO-CCSD(T) single point calculations87,88,89. Solvation was considered by the COSMO-RS90,91,92 model and used to derive the gas-phase quantities ΔrH°(g) and ΔrG°(g) as well as ΔrH°(S) and ΔrG°(S) in S at standard conditions (g: 298 K, 1 bar; solv.: 298 K, a = 1 mol L−1). Full data is presented in the Supplementary Notes 4.1–4.4.

All calculated gas-phase quantities ΔrH°(g) and ΔrG°(g), as well as ΔrH°(S) and ΔrG°(S) in S at standard conditions, are collected in full detail in Supplementary Note 4.3. In Fig. 5, we only discuss the solution-values ΔrG°(S) relevant for the development of the formal potential of Ag+(S)/Ag as a function of S. Overall, the consecutive and total solvent complexation Gibbs energies and enthalpies of Ag+ greatly decrease with increasing fluorination of xFB, both in S and in the gas phase. Typically, two or three xFB molecules may be taken up in the solvates [Ag(S)n]+ (yellow, orange, red bars in Fig. 5). Yet, especially for the highly fluorinated xFB molecules with x = 4–5, the desolvation Gibbs energy ΔrG°(xFB) of 9 to 28 kJ mol−1 is very low and, hence, the solvated Ag+ ion is very reactive in these solvents. This agrees with the very high E°´(Ag+(S)/Ag) values of 1.47 (S = 4FB) and 1.50 V (S = 5FB) vs. E°´(Fc+/Fc, S) measured in S. In addition, for the three most fluorinated xFB solvents, the calculations predict that ion-pairs are more favorable than dissociated ions (cf. yellow bars to green/blue bars in Fig. 5). Pleasingly, the complexes or ion-pairs, calculated to be most favorable in solution, comply well with those found with scXRD analyses (Fig. 3).

Fig. 5: Solvation Gibbs energies of the silver cation.
figure 5

Calculated values of ΔrG°(S) for the formation of the silver ion solvates (Ag+ + nS → [Ag(S)n]+) and ion-pairs (Ag+ + [pf]− (+ S) → {Ag[pf]}ip / {(S)Ag[pf]}ip) at the DLPNO-CCSD(T)/CBS level with COSMO-RS solvation in S. The individual values are reported in full detail in Supplementary Note 4.3. The species with the largest negative value of ΔrG°(S) in the graph, is the most favorable one in solution in S at standard conditions. AN stands for H3CCN, DCE for 1,2-Cl2C2H4 and DCM for CH2Cl2.

Full size image

In addition, the complexation Gibbs energies of Ag+ with 2FB and CH2Cl2 are comparable as are their E°´(Ag+(S)/Ag)-values of +0.99 and +0.88 V vs. E°´(Fc+/Fc, S). Since ΔrG°(S) calculated for 1FB is more favorable than for 2FB and CH2Cl2, the E°´(Ag+(1FB)/Ag) potential is reduced to 0.74 V. The potential difference between the E°´(Ag+(S)/Ag)-values in 2FB and 1FB (0.99–0.74 = 0.25 V), also agrees with the difference of the sum of their 1st and 2nd complexation Gibbs energies in solution of 21.3 kJ mol−1; this corresponds with ΔG = −zFΔE to 0.22 V (cf. Supplementary Note 4.3). For xFB solutions with x > 2, the values are influenced by ion-pairing and cannot be used for such quantitative evaluations. By contrast, the complexation Gibbs energies of AN, which induce the lowest Ag+ potential of +0.02 V vs. Fc+/Fc, are also the largest of all the solvents assessed within Fig. 5. This also complies with the experiment. Hence, our results suggest that the E°´(Ag+(S)/Ag) values collected in Table 2 reflect the true reactivity of the electroactive Ag+ ions in S. They are induced by poor ligand capacities of xFB.

All calculated NO+ interaction energies are collected in Supplementary Note 4.4. Interestingly, while the formation of Wheland complexes is favored for all xFB molecules in the gas phase and, in part, even the uptake of a second xFB molecule is viable, the situation is very different in xFB-solution: Apparently, only 1FB—with the lowest IE of all xFB molecules (Table 2)—is electron-rich enough to slightly favor the complex [NO(1FB)]+ also in 1FB-solution.

By contrast, in all other cases, the calculations suggest the presence of uncomplexed NO+ in xFB (x = 2–5) solution, although this is on the edge for 2FB, which complies with the determined scXRD structure [NO(2FB)]+[pf]− (Fig. 4A). This strive toward the formation of free NO+ presumably results from the rather high calculated solvation Gibbs energies for the isolated small NO+ ion vs. the complexed [NO(xFB)]+ systems. Overall, from 1FB to 5FB the interaction energies gradually get lower, concomitant with the increase of the solvent IEs. This aligns with the potentials E1/2 collected in Table 2 that vary in xFB solution over +1.52 (4FB) – 1.11 (1FB) = 0.41 V. Also, the difference of ΔrG°(xFB) for the first arene complexation between 4FB and 1FB amounts to 32.1–(–6.8) = 38.9 kJ mol−1 or 0.40 V. Hence, the drastic increase of the NO+ potentials results from the inferior interaction of the cation with the higher fluorinated xFB solvents and the presence of free NO+ in the latter cases. In addition, the weak tendency to form Wheland complexes [NO(xFB)]+ aligns with the IEs of the xFB solvents of 9.20–9.63 eV in Table 2: They are close to (1FB), or even higher (xFB, x = 2–5) than that of NO of 9.26 eV. This can explain why solutions of NO+[pf]− may be stored for weeks in xFB solution without noticeable decomposition. Moreover, the visual inspection of these solutions indicates decreasing red coloration, the typical Wheland-complex color, with increasing fluorine content of xFB (Fig. 4E).

Additionally, we performed an energy decomposition analysis combined with the natural orbitals of chemical valence (EDA-NOCV). In line with the experimental observations and other DFT calculations, the total interaction energy (ΔEint) between the NO+ fragment and the different xFB fragments gradually decreases with increasing degree of fluorination (Fig. 4D). Interestingly, the decrease of the total interaction energy is not of a decrease of the orbital interaction energies (ΔEorb), but rather the electrostatic interaction (ΔEelstat). The entire EDA-NOCV results, including analyses for benzene and mesitylene, are included in Supplementary Table 40.

Extension of the positive potentials in xFB with innocent deelectronators

After having established that the combination of xFB solvents and [pf]− WCA can tune the potentials of the commonly used NO+ and Ag+ deelectronators to much higher values, we proceed to push these limits to even higher values by using different deelectronators. Compared to the positive solvent limits Epos collected in Table 2, the Ag+ and NO+ potentials do not stretch over the full Epos limits, which the xFB solvents can tolerate +1.82 to +2.67 V vs. E°’(Fc+/Fc, S). Moreover, we have noted numerous times in synthetic applications that both, Ag+ and NO+ deelectronator reagents react non-innocent with substrates in xFB solution—with complexation33,35,93,94,95, substitution30,32,34,37 or degradation33 being the key side—or even major reactions complicating or hindering the desired reaction outcome64. In addition, also the magic blue-like salts [TBPA]∙+[WCA]−, praised5 for their innocent behavior and even if partnered with the excellent counterion [pf]−, yield after deelectronations a neutral amine, that often does react non-innocent41.

Innocent deelectronators

To generate a desired deelectronated target system [M]+(S) (Fig. 6, Eq. 1a), the secondary system in Eq. 1b needs to have a suitably high deelectronation potential E°’(iD+/iN, S), to be non-reactive toward both, the solvent S and the generated reactive cation [M]+, and must be compatible with the counterion17, i.e., the very good WCAs [pf]− and [al-f-al]−25,26. Toward this goal, a series of innocent redox-couples iD+/iN was developed, from which we already published phenazineF41 and anthracenceHal33. Structural formulae of the neutral iN are shown in Fig. 6, together with their deelectronation potentials E°’(iD+/iN, S) vs. E°’(Fc+/Fc, S) in 2FB or 4FB solution. The iD+/iN-couples (iN = anthraceneF, phenantreneF) in Fig. 6 extend the positive deelectronation range in 4FB up to +1.89 V and considerably higher than the maximum of +1.52 V available with E1/2(NO+, 4FB) from Table 2. In addition, all deelectronated iD+ radical cations from Fig. 6 may be prepared as room temperature stable iD+[WCA]− salts in 74 to 78% yield (WCA = al-f-al). For the systems, also the CV traces at scan rates between 20 and 1000 mV s−1 and the scXRD structures of iD+[WCA]− are included in Fig. 6 and Supplementary Notes 3.4.3–3.4.4. Note that for phenazineF and anthraceneX (X = Hal, F) salts, their easiest preparation now follows the deelectronation of their iN neutrals in 4FB with the readily available NO+[WCA]− salts26,36, since their formal potentials E°’(iD+/iN, 4FB) of +1.29–1.47 V are below the +1.52 V potential of NO+[WCA]− in 4FB. In addition, we have shown that the silver salt Ag+[pf]− dissolved in 4FB and with a formal potential of +1.47 V is suited to quickly deelectronate anthraceneHal (E°’ = 1.42 V) and slower also anthraceneF (E°’ = 1.47 V) with formation of solid silver Ag0(s) in the reaction (Supplementary Notes 3.8). However, to ionize the phenantreneF neutral, one needs to use the silver salts Ag+[WCA]− in combination with 0.5 equivalents of I2. The concomitantly formed solid AgI provides the additional thermodynamic driving force to deelectronate the neutral iN. Note, that the Ag+[WCA]−/0.5 I2 system is only fully compatible with 4FB and higher fluorinated benzenes. We have estimated the limit for deelectronation with Ag+[WCA]−/0.5 I2 to roughly amount to +2.3–2.4 V vs. E°´(Fc+/Fc, S), since Epos(3FB) = +2.35 V (Table 2) and 3FB already reacts with the synergistic mixture Ag+[WCA]−/0.5 I2.

Fig. 6: Key data of the innocent deelectronators.
figure 6

Top: Requirements for innocent deelectronation described by Eq. 1a−b and structural formulae and formal potentials E°’(iD+, S) of the known33,41 innocent deelectronator/neutral couples. Bottom: Structural formulae and formal potentials E°’(iD+, S) of the innocent deelectronator/neutral couples with increased potential to be used for innocent deelectronation reactions. Middle: Chemically reversible electronation and deelectronation in cyclic voltammograms (2nd cycle) of anthraceneF and phenantreneF at different scan rates (black to red, 20–1000 mV s−1) of phenanthreneF (10 mm) in 4FB using [NBu4]+[Al(ORF)4]− (100 mm) as supporting electrolyte. Right: Molecular structures of deelectronated anthraceneF and phenantreneF as [al-f-al]− salts. Thermal ellipsoids set at 50% probability.

Full size image

Favorable for any application, all iD+[WCA]− salts—prepared from the iN molecules in Fig. 6—are long-term stable at room temperature (at least several months), given inert and anhydrous conditions. Hence, we expect innocent deelectronation with iD+[WCA]− salts in inert xFB solvents to largely extend the limits of currently possible deelectronation chemistry. All of this works using standard techniques and glassware with commercially available reagents. Hence, from this combination of reagents and solvents a wide variety of cationic chemical systems may be synthesized that potentially stretch from organic carbocations, over transition metal carbonyl and other organometallic cations, to even non-metal cations. Since the four iN in Fig. 6 are also reversibly addressable at an inert electrode, the pairs xFB / iN also do hold large promise as redox mediators, e.g., for electrosynthesis and -catalysis or maybe also as redox shuttle to mediate cut-off potentials in batteries96,97,98,99,100,101.

Discussion

Here we present the very favorable properties of the strategically and dipole-maximized substituted fluorobenzenes xFB as inert solvents that are tolerant to high deelectronation potentials. With increasing fluorination and if paired with the good WCAs [pf]− or [al-f-al]−, the xFB molecules turn into increasingly poor ligands and induce markedly increasing deelectronation potentials of the classical deelectronator ions Ag+ and NO+ that may reach values as high as +1.50/+1.52 V vs. E°´(Fc+/Fc, S). Concomitantly, this combination enables very good solubilities of the de+[WCA]− salts with high activities of the solvated de+ ions. The most relevant system properties for potential applications in many fields of chemistry or the material sciences are compiled in Table 3.

If these de+[WCA]− reagents (de = Ag, NO) do react non-innocently by coordination, substitution or degradation and despite the solvents being inert, one can switch to the room temperature stable innocent deelectronator salts iD+[WCA]− with [anthraceneF]+∙ and [phenantreneF]+∙ radical cations that further push the limits of reversibly addressable deelectronation chemistry up to +1.89 V vs. E°´(Fc+/Fc, 4FB). Since these potentials are also available at an inert electrode, we suggest that these combinations of innocent solvent and innocent deelectronator may also be advantageously used as redox mediators for electrosynthesis and catalysis or as redox shuttles in battery electrolytes. Favorably, the innocent solvent properties may be fine-tuned by mixing. For example, at room temperature static dielectric constants εr between 2 and 22 may be realized for binary xFB systems.

Methods

Syntheses

The used syntheses for the nitrosyl, silver and [NBu4]+ salt were already described in earlier publications of our group36,102,103. Nonetheless, it is worth mentioning again that the synthesis for the nitrosyl salt could be done in a at least 10 gram scale with regard to the lithium salt, using the commercial NO+[BF4]− salt and without lithium impurities in the product36,104. The synthesis of the solvent-free silver salt was a great progress to avoid potential harmful dichloromethane in further reactions25,29,102. Additionally, we could develop a new synthetic route for the known Fc+ salt using the nitrosyl, instead of the silver salt as oxidizing agent, to avoid colloidal silver which could be disturbing during further experiments. For a better allocation in the following NMR spectra, instead of the abbreviation [pf]− the detailed description [Al{OC(CF3)3}4]− will be used.

Nitrosonium tetrakis(nonafluorotertbutanolato)aluminate(III) (NO+[pf]−)

Caution! This procedure involves the work with liquified sulfur dioxide, which has a vapor pressure of ca. 4 bar at room temperature. Therefore, the synthesis requires trained personnel and proper equipment.

NO+[Al{OC(CF3)3}4]− was synthesized based on the procedure published in ref. 36: Li+[pf]− (9.74 g, 10 mmol, 1.0.eq) and NO+[BF4]− (1.75 g, 15 mmol, 1.5 eq.) were filled in one side of a double-bulb Schlenk vessel equipped with a G4 frit plate. Sulfur dioxide (10 mL) was condensed onto the mixture of the reagents at −78 °C. The vessel was brought to room temperature and stirred for 7 days. Afterward, the solution was filtered and the solvent was removed by vacuum. NO+[pf]− was obtained as a colorless solid (8.86 g, 89%).

Characterization

1H-NMR (300.18 MHz, CD2Cl2, calibration to CHDCl2 = 5.32 ppm105, 298 K): No signals observed.

7Li-NMR (116.66 MHz, CD2Cl2, 298 K): No signals observed.

11B-NMR (96.31 MHz, CD2Cl2, 298 K): No signals observed.

14N-NMR (21.69 MHz, CD2Cl2, 298 K): δ = 364.5 (s, NO+, 1 N) ppm.

19F-NMR (282.45 MHz, CD2Cl2, 298 K): δ = −75.7 (s, [Al{OC(CF3)3}4]−, 36 F) ppm.

27Al-NMR (78.22 MHz, CD2Cl2, 298 K): δ = 34.7 (s, [Al{OC(CF3)3}4]−, 1Al) ppm.

FTIR (ZnSe, ATR):ν/cm−1 = 2342 (vw), 1354 (vw), 1301 (m), 1248 (vs), 1209 (vs), 968 (vs), 863 (vw), 830 (vw), 757 (vw), 726 (vs), 642 (vw), 568 (vw), 560 (vw).

FT Raman (1000 scans, 250 m0W):ν/cm−1 = 2937 (vw), 2756 (vw), 2340 (m), 1355 (vw), 1315 (vw), 1275 (w), 1248 (vw), 1204 (vw), 1120 (vw), 975 (vw), 829 (vw), 815 (vw), 796 (vs), 746 (s), 571 (w), 562 (w), 537 (m), 366 (m), 325 (vs), 290 (m), 237 (m), 208 (w), 174 (w), 118 (m), 100 (m).

Silver tetrakis(nonafluorotertbutanolato)aluminate(III) (Ag+[pf]−)

Caution! This procedure involves the work with liquified sulfur dioxide, which has a vapor pressure of ca. 4 bar at room temperature. Therefore, the synthesis requires trained personnel and proper equipment.

Ag+[Al{OC(CF3)3}4]− was synthesized based on the procedure published in ref. 102 The reaction was performed analogously to the one yielding NO+[pf]−. Instead of NO+[BF4]−, AgF (1.91 g, 15 mmol, 1.5 eq.) was used. Additionally, the reaction was performed in the absence of light. Ag+[pf]− was obtained as a colorless solid (9.03 g, 84%).

Characterization

1H-NMR (400.17 MHz, CD2Cl2/Et2O, calibration to CHDCl2 = 5.32 ppm105, 298 K): 3.50 (q, 3JHH = 7.0 Hz, O(CH2CH3)2, 2H) and 1.20 (t, 3JHH = 7.0 Hz, O(CH2CH3)2, 3H) ppm.

19F-NMR (282.45 MHz, CD2Cl2/Et2O, 298 K): δ = −75.8 (s, [Al{OC(CF3)3}4]−, 36 F) ppm.

27Al-NMR (104.27 MHz, CD2Cl2/Et2O, 298 K): δ = 34.6 (s, [Al{OC(CF3)3}4]−, 1Al) ppm.

7Li-NMR (116.7 MHz, CD2Cl2/Et2O 2, 298 K): No signals observed.

Tetrabutylammonium tetrakis(nonafluorotertbutanolato)aluminate(III) ([NBu4]+[pf]−)

[NBu4]+[Al{OC(CF3)3}4]− was synthesized based on the procedure published in ref. 103: Li+[pf]− (19.5 g, 20 mmol, 1.0 eq.) and [NBu4]+Br− (6.45 g, 20 mmol, 1.0 eq.) were dissolved in a mixture of water and acetone (85:15 v/v, 150 mL) at room temperature. The solution was kept at a warm place/heated at 30 °C overnight, allowing the acetone in the solvent to evaporate, yielding a microcrystalline precipitate. The remaining solvent was removed by filtration and the residue was washed with water until all the bromide was removed (test with, e.g., silver nitrate). Afterward, the product was washed two times with hexane (2 × 100 mL). [NBu4]+[pf]− was obtained as a colorless powder (22.9 g, 94%).

Characterization

1H-NMR (300.18 MHz, CD2Cl2, calibration to CHDCl2 = 5.32 ppm105, 298 K): δ = 3.07 (m, [N(CH2CH2CH2CH3)4]+, 8H), 1.60 (m, [N(CH2CH2CH2CH3)4]+, 8H), 1.43 (m, [N(CH2CH2CH2CH3)4]+, 8H), 1.03 (t, 3JHH = 7.3 Hz [N(CH2CH2CH2CH3)4]+, 12H) ppm.

19F-NMR (282.45 MHz, CD2Cl2, 298 K): δ = −75.7 (s, [Al{OC(CF3)3}4]−, 36 F) ppm.

27Al-NMR (78.22 MHz, CD2Cl2, 298 K): δ = 34.6 (s, [Al{OC(CF3)3}4]−, 1Al) ppm.

7Li-NMR (116.7 MHz, CD2Cl2, 298 K): No signals observed.

14N-NMR (21.9 MHz, CD2Cl2, 298 K): No signals observed.

Bis(η 5-cyclopentadienyl)iron(III) tetrakis(nonafluorotertbutanolato)aluminate(III) (Fc+[pf]−)

NO+[pf]− (1.00 g, 1.01 mmol, 1.00 eq.) and Fc (0.23 g, 1.23 mmol, 1.22 eq.) were weighed, inside a glovebox, in one side of a double-Schlenk tube separated by a G3 or G4 frit and equipped with grease-free PTFE valves. Under reverse flow of Argon, 2FB (1,2-difluorobenzene, 10 mL) was added and led immediately to the formation of NO(g) and a dark blue solution. The solution was stirred at RT overnight and the solvent was removed under vacuo. To remove the excess of ferrocene, the residue was washed with n-hexane (5 mL). Therefore, the ferrocene solution in n-hexane was filtered through the frit and the solvent was condensed back to the side of the crude product, as many times, as the n-hexane solution was still colored yellowish before the filtration. Afterward, the crude product was dried under vacuo (10−3 mbar) to yield a blue powder of Fc+[Al{OC(CF3)3}4]− (0.99 g, 0.86 mmol, 85%).

Characterization

1H-NMR (300.18 MHz, 1,2-F2C6H4 (2FB), calibration to 1,2-F2C6H4 = 6.96 ppm against Si(CH3)4, 298 K): 33.87 (br. s., [Fe(C5H5)2]+, 10H) ppm.

19F-NMR (282.45 MHz, 2FB, 298 K): δ = −75.7 (s, [Al{OC(CF3)3}4]−, 36 F), −139.6 (s, 1,2-F2C6H4, 2 F) ppm.

27Al-NMR (78.22 MHz, 2FB, 298 K): δ = 34.7 (s, [Al{OC(CF3)3}4]−, 1Al) ppm.

FTIR (ZnSe, ATR):ν/cm−1 = 3126 (vw), 1423 (vw), 1352 (vw), 1299 (w), 1273 (m), 1266 (m), 1253 (m), 1239 (m), 1213 (vs), 1163 (w), 1064 (vw), 1014 (vw), 972 (vs), 856 (w), 832 (vw), 792 (vw), 756 (vw), 728 (vs), 571 (vw).

FT Raman (1000 scans, 200 mW):ν/cm−1 = 3133 (vw), 1425 (vw), 1363 (vw), 1304 (vw), 1273 (vw), 1113 (m), 1065 (vw), 851 (vw), 797 (vw), 746 (vw), 562 (vw), 538 (vw), 367 (vw), 321 (w), 299 (vs), 234 (vw), 170 (vw), 120 (vw), 82 (vw).

Nitrosonium bis{tris(nonafluorotertbutanolato)aluminum(III)}-(μ 2)-fluoride (NO+[al-f-al]−)

Caution! This procedure involves the work with liquified sulfur dioxide, which has a vapor pressure of ca. 4 bar at room temperature. Therefore, the synthesis requires trained personnel and proper equipment.

NO+[F{Al(OC(CF3)3)3}2]− was synthesized based on the procedure published in ref. 26. NO+[PF6]− (560 mg, 3.20 mmol, 1.0 eq.) and (H3C)3Si–F–Al(OC(CF3)3)3 (5.04 g, 6.1 mmol, 2.0 eq.) were filled in a Schlenk vessel inside a glovebox. Sulfur dioxide (10 mL) was condensed onto the mixture of the reagents at −78 °C. The vessel was equipped with a bubbler, brought to −35 °C and the temperature was held for 1 h. Subsequently, the reaction solution was slowly warmed to room temperature and the sulfur dioxide evaporated. The white powder was dried at 10−3 mbar for 2 h. [NO]+[al-f-al]− was obtained as a colorless powder (4.36 g, 90%).

Silver bis{tris(nonafluorotertbutanolato)aluminum(III)}-(μ 2)-fluoride (Ag+[al-f-al]−)

Caution! This procedure involves the work with liquified sulfur dioxide, which has a vapor pressure of ca. 4 bar at room temperature. Therefore, the synthesis requires trained personnel and proper equipment.

Ag+[F{Al(OC(CF3)3)3}2]− was synthesized based on the procedure published in ref. 26. The reaction was performed analogously to the one yielding NO+[al-f-al]−. Instead of NO+[PF6]−, Ag+[PF6]− (810 mg, 3.20 mmol, 1.0 eq.) was used. Additionally, the reaction was performed in the absence of light. Ag+[al-f-al]− was obtained as a colorless solid (4.68 g, 92%).

Decafluoroanthracene (anthraceneF)

AnthraceneF was synthesized based on the procedure published in ref. 106: 9,10-dichlorooctafluoroanthracene (2 g, 5.11 mmol, 1.0 eq.) and KF (0.98 g, 16.9 mmol, 3.3 eq.) were dissolved in a mixture of sulfolane (10 mL) and toluene (20 mL). The reaction mixture was heated for 2 h at 120 °C. Afterward, the toluene was removed from the reaction mixture. The reaction mixture was further heated for 4 h at 210 °C. At room temperature, water (50 mL) was added to the mixture. The mixture was filtrated and the brown residue was washed with water (3 × 20 mL). After drying at room temperature, the residue was taken up in dichloromethane (20 mL) and slowly cooled down to −40 °C. The solution was decanted and anthraceneF was obtained as a yellow to brown solid (0.63 g, 32%).

Decafluoroanthracenium bis{tris(nonafluorotertbutanolato)aluminum(III)}-(μ 2)-fluoride ([anthraceneF]+[al-f-al]−)

[NO]+[F{Al(OC(CF3)3)3}2]− (0.10 g, 1.0 eq, 66 μmol) and anthraceneF (28 mg, 1.2 eq., 78 μmol) were placed in a Schlenk flask and 1,2,3,4-tetrafluorobenzene (1 mL) was added. The solution instantaneously turned green-blue and a gas formation was observed. The solution was stirred 5 min at room temperature and was then layered with n-pentane (10 mL). Slow diffusion of the solvents over days led to the crystallization of [anthraceneF]+∙[F{Al(OC(CF3)3)3}2]− in blue plates, suitable for scXRD (90 mg, 49 μmol, 74%).

Table 3 Properties relevant to apply the combination xFB with deelectronators de[WCA] and [NBu4]+[pf]− as supporting electrolyte salt
Full size table

Characterization

FTIR (ZnSe, ATR):ν/cm−1 = 1600 (vw), 1517 (vw), 1491 (w), 1468 (vw), 1443 (vw), 1355 (w), 1300 (w), 1265 (s), 1243 (vs), 1211 (vs), 1177 (s), 1118 (w), 1038 (vw), 970 (vs), 957 (s), 863 (w), 810 (vw), 760 (vw), 726 (vs), 716 (m), 664 (w), 631 (w), 569 (w).

FT-Raman (1000 scans, 50 mW):ν/cm−1 = 1570 (s), 1548 (s), 1436 (vs), 1415 (s), 1395 (vs), 1294 (s).

Decafluorophenanthrenium bis{tris(nonafluorotertbutanolato)aluminum(III)}-(μ 2)-fluoride ([phenanthreneF]+[al-f-al]−)

Ag+[F{Al(OC(CF3)3)3}2]− (84 mg, 1.0 eq., 53 μmol), I2 (6.7 mg, 0.5 eq., 27 μmol) and phenanthreneF (19 mg, 1.0 eq., 53 μmol) were placed in a Schlenk flask and 1,2,3,4-tetrafluorobenzene (0.8 mL) was added to the solution. The solution was filtered and brownish crystals suitable for scXRD formed out of the dark solution upon slow removal of the solvent. The product was obtained as dark brown crystalline blocks (78 mg, 42 μmol, 78%).

Characterization

FTIR (ZnSe, ATR):ν/cm−1 = 1673 (vw), 1662 (vw), 1645 (vw), 1584 (vw), 1522 (w), 1495 (m), 1468 (vw), 1433 (vw), 1379 (w), 1354 (vw), 1301 (w), 1265 (m), 1242 (vs), 1213 (vs), 1180 (m), 1105 (vw), 1093 (w), 1069 (vw), 1011 (vw), 972 (vs), 866 (vw), 847 (m), 799 (vw), 727 (vs), 707 (m), 636 (w), 569 (w).

Dielectric spectroscopy

Complex permittivity spectra were recorded using an Anritsu MS4647A vector network analyzer connected to an open-ended coaxial probe at frequencies ranging from 1 to 70 GHz107. The reflectometer was calibrated using air, conductive silver paint, and N,N-dimethylacetamide108. Samples were placed into a double-walled sample holder connected to a Julabo-E12 thermostat to control the temperature at 5, 10, 18, 25, 33, or 40 °C. All complex permittivity spectra were modeled using a Debye relaxation70 to obtain εs, the low-frequency limit of the permittivity, ε∞, the high-frequency limit of the permittivity, and the dielectric relaxation time τ. Recorded spectra, together with the details of the analysis, are given in Supplementary Note 5.

Cyclovoltammetry, general procedures

All cyclic voltammograms were recorded in an argon-filled glovebox. A three-electrode arrangement was used with a 1 mm Pt disc working electrode (WE), a Pt mesh or wire as counter electrode (CE) and a Pt wire as pseudo reference electrode (RE) for NO+ (Supplementary Notes 3.1), ferrocene (Fc) and N(4-BrC6H4)3 or Ag wire for the Ag+ and NO+ evaluation (triangular) measurements in the Supplementary Notes 3.5.3 / 3.7.3. For all measurements [NBu4]+[pf]− (c = 100 mm) was used as supporting electrolyte. For each measured electroactive sample, c(de+) = 10 mm was used. The RE (Pt|Fc+, Fc or Ag+|Ag reference) was added in a glass compartment with a frit to allow for a direct measurement and was filled with solutions of [NBu4]+[pf]− (100 mm), Fc (10 mm) and Fc+[pf]− (10 mm) or [NBu4]+[pf]− (100 mm) and Ag+[pf]− (10 mm). For each solvent S, a potential stability window (ECW) was recorded in a solution of [NBu4]+[pf]− (100 mm) in S to identify any impurities and to determine the stability range of the pure solvent against the RE used. Scan rates were varied from 20 mV s−1, 50 mV s−1, 100 mV s−1 up to 200 mV s−1, and if not stated otherwise, the half-wave potentials E1/2 did not change with the rate (full details in Supplementary Note 3). Since several of the published potentials in the Geiger/Connelly Review5 may have been afflicted by ion-pairing and other effects, we also measured the Ag+ and NO+ potentials in the like setup, but the solvents CH2Cl2, 1,2-Cl2C2H4, dimethylformamide (DMF), acetonitrile (AN). In addition, the ECWs of nitromethane (MeNO2), propylene carbonate (PC) and tetrahydrofurane (THF) were determined with [NBu4]+[pf]− (100 mm) as supporting electrolyte salt for comparison.

CV-evaluation by triangular Born-Fajans-Haber-Cycles

This evaluation is exemplarily shown for the two solvents 5FB and 1,2-Cl2C2H4 in Fig. 7, all other triangular cycles are deposited in the Supplementary Notes 3.5.3 / 3.7.3.

Fig. 7: CV-evaluation by triangular Born-Fajans-Haber-Cycles.
figure 7

Measured half-wave potentials of [Fc]+ and [NO]+ versus Ag+(10 mm)/Ag and [NO]+ versus [Fc]+/Fc in exemplarily selected 5FB and DCE = 1,2-Cl2C2H4 solution at a scan rate of 100 mV s−1. Knowing two out of the three values of the measurements, the third can be calculated in a Born-Fajans-Haber-Cycle approach. Hence, |ΔE| and |ΔG| errors can be calculated by using the relation ΔG = −zFΔE, with z = number of electrons, ΔE = potential difference, F = Faraday constant = 96,485 C mol−1.68 The mean errors of the calculated potential difference (ΔE) / the corresponding Gibbs energy difference (ΔG) are given in red in the center of the triangle. Note, E1/2 potentials were rounded to two decimal places and rounding errors can occur.

Full size image

Quantum chemical calculations

An extended search of the potential energy surface was manually performed to find the lowest energy structures of all particles [de(S)n]+ (de = Ag, NO), {Ag[pf]}ip,solv and {(S)Ag[pf]}ip,solv at the dispersion109,110 corrected (RI-)BP86(D3BJ)/def2-TZVPP85 DFT86 level of theory. Corrections to statistical thermodynamics (ZPE, H° and S°) were taken from the frequency calculations111 at this level. The energies of the DFT structures were refined in a series of DLPNO-CCSD(T) single point calculations87,88,89 with Dunning’s112,113,114,115,116,117 basis sets cc-pVDZ, cc-pVTZ, cc-pVQZ and then extrapolated to the complete basis set limit (CBS). These accurate CCSD(T)/CBS values were used as electronic energies for the calculation of the thermodynamics of all particles, augmented by DFT-corrections to ZPE, H° and S°. Finally, contributions of solvation enthalpies and free energies in S were calculated with the COSMO-RS90,91,92 model at the BP86(D3)/def2-TZVPD//BP86(D3)/def-TZVP level, so that we overall derive the quantities ΔrH°(g) / ΔrG°(g) in the gas phase as well as ΔrH°(solv) / ΔrG°(solv) in solution in S at standard conditions (g: 298 K, 1 bar; solv.: 298 K, a = 1 mol L−1).