Michel
Meyer
a,
Laurent
Frémond
a,
Alain
Tabard
a,
Enrique
Espinosa
a,
Guy Yves
Vollmer
a,
Roger
Guilard
*a and
Yves
Dory
b
aLaboratoire d’Ingénierie Moléculaire pour la Séparation et les Applications des Gaz (LIMSAG, CNRS UMR 5633), Université de Bourgogne, Faculté des Sciences, 6 boulevard Gabriel, 21100, Dijon, France. E-mail: roger.guilard@u-bourgogne.fr; Fax: +33 3 80 39 61 17; Tel: +33 3 80 39 61 11
bLaboratoire de Synthèse Supramoléculaire, Département de Chimie, Université de Sherbrooke, 2500 boulevard de l‘Université, Sherbrooke, PQ J1K 2R1, Canada
First published on 14th December 2004
The 14-membered cyclic diamide 1,4,8,11-tetraazacyclotetradecane-5,12-dione (5,12-dioxocyclam) can be considered as a trans-autodiprotected tetraazamacrocycle and provides a convenient starting material for the preparation of macrobicyclic receptors. As an example, the secondary amine nitrogen atoms located at the 1 and 8 positions were cross-bridged with a 1,3-pyridyl strap, affording the constrained ansa-dioxocyclam ligand 1,9,12,18,22-pentaazatricyclo[7.6.6.13,7]docosa-3,5,7(22)-triene-13,19-dione (L1). The proton binding properties of this cage-type compound, which possesses a hemispherical cavity, were fully investigated by spectroscopic (IR, NMR, UV, MALDI-TOF MS), quantum chemical, and potentiometric methods. While both bridgehead tertiary amines have their free lone pairs oriented inside the cavity, intramolecular hydrogen bonding was found to play a key role in determining the structural features of the free base and its protonated forms. L1 behaves as a diprotic base in water with log K011 = 8.94(1) and log K012 = 2.32(9), but most interestingly shows slow proton-transfer rates on the NMR timescale.
Reinforced28,29 and cross-bridged17,30–32 cyclic tetraamines such as 1,4,7,10-tetraazacyclododecane (cyclen) and 1,4,8,11-tetraazacyclotetradecane (cyclam), are among the most common and readily accessible polynitrogenated bicyclic architectures.33 They are obtained by linking either two adjacent or two nonadjacent secondary amines of the parent macrocycle. The latter possibility has given rise to a new family of macrobicyclic ansa-ligands, the so-called bowl adamanzanes,34 with a number of these also displaying “proton sponge” properties and very slow protonation kinetics.35,36
In contrast, amide-containing azamacrobicyclic ligands have attracted much less attention,37 although they provide a convenient access to strapped cyclen or cyclam derivatives upon reduction of the protecting carbonyl groups.8,9,38–41 From a synthetic point of view, trans-dioxocyclam (5,12-dioxocyclam or 1,4,8,11-tetraazacyclotetradecane-5,12-dione, compound 1 in Scheme 1)42,43 is an attractive starting material since two out of the four secondary amines of the 14-membered ring, namely those located in positions 4 and 11, take part in two peptidic bonds and thus are masked. Optically active C-functionalized analogs, obtained by photolysis of chromium carbene complexes in the presence of suitable imidazolines, have been described more recently by Hegedus and coworkers.44–47 Provided these trans “autodiprotected” synthons are reacted with the appropriate biselectrophile, hemispherical cage-type molecules are isolated in a single step and in high yield.8,9,38,39,46,48–56 Their dual personality stems from their intermediate structure as they possess all the features of the corresponding saturated polyamines and those of oligocyclopeptides.57,58 However, the physico-chemical consequences of their interplay in a constrained environment remain largely unknown today. Recently, we reported on the unusual properties displayed by the copper(II) complexes formed with ligand L1.1 Depending upon the medium’s acidity, either a red tetra- or a blue pentacoordinated species was isolated, the former possessing one and the latter two N-deprotonated amide groups. The fast pH-triggered interconversion was shown to be accompanied by metal translocation and a simultaneous molecular reorganization of the ligand scaffold reminiscent of a glider motion.
Scheme 1 |
Although a few o-xylyl,48 α,α′-pyridyl,52,53,55 or polyethylene glycol capped trans-dioxocyclams51 have been structurally characterized, to the best of our knowledge no report describing the solution structure and the acid-base properties of this class of compounds has appeared so far. The structural data indicate that the strap spanning across the dioxocyclam fragment brings the bridgehead nitrogen atoms in close vicinity and enforces an endo orientation of both lone pairs, most probably freezing out the Walden inversion. The resulting ring conformation contrasts markedly with the extended layout adopted by 1 in either its free base or diprotonated form, suggesting that both amines behave independently of each other.43 Indeed, solution studies revealed negligible Coulombic interactions between both basic sites for trans-dioxocyclam, since their protonation gives rise to a weak cooperative effect. Hereafter, the thermodynamic and kinetic proton-uptake properties of an ansa-dioxocyclam macrobicycle are discussed for the first time in light of NMR, potentiometric, spectrophotometric, and quantum-chemical studies of the pyridine derivative L1 (Scheme 1).
The steric constraints imposed by the rigid pyridyl handle prevents the chiral tertiary amines from undergoing the Walden inversion at room temperature. Thus, reaction of the dihalogenated biselectrophile 3 with the C2h-symmetric dioxocyclam unit affords a racemic mixture of (R,R) and (S,S) diastereoisomers.52 The analytical purity was checked by elemental CHN analysis, a negative silver nitrate assay, and by MALDI-TOF mass spectroscopy. The mass spectrum of the free base evidenced a featureless baseline with a single peak at m/z = 331.1, showing the predicted isotopic line pattern for the radical-cation [L1]+˙ (calcd m/z = 331.2). The one-mass-unit shifted signal found for the monohydrobromic acid salt crystallized in acetonitrile further confirmed the compound’s identity. It may be noted that none of the possible pyridinyl-bridged bismacrocyclic [2 + 1] or macrotricyclic [2 + 2] cyclization by-products is formed in appreciable amounts if the reaction is carried out at the centimolar concentration level, thus avoiding high-dilution conditions.
In order to ascertain the formation of intramolecular N–H⋯N hydrogen bonds, spectra were recorded in CDCl3 between 1500 and 3500 cm−1. As far as no shift in energy for any band was detected upon increasing the concentration of L1, it is concluded that the amidic protons are effectively engaged in such interactions both in the solid state and in solution. In addition, solubilization of amorphous L1 in chloroform perturbs only very slightly the vibrational energy levels as evidenced by comparing the KBr pellet and solution spectra. While the carbonyl stretch remains unaffected, the frequency difference for the νNH doublet and the δCNH mode does not exceed 17 cm−1. These results further support the intramolecular nature of the hydrogen bonds and suggest also a close conformational similarity of the molecule in both states. Another piece of structural information is provided by the absence of a lower-energy satellite associated to the strong C–D stretching band arising from the solvent at 2253 cm−1. The appearance of such a band near 2170 cm−1, as observed when increasing amounts of triethylamine are mixed with CDCl3, would be indicative of C–D⋯X (X = N or O) hydrogen bond formation between the solvent and the solute. This criterion can be used to probe the relative orientation of the bridgehead tertiary amine’s lone pairs pointing either inside (in configuration) or outside (out configuration) the cavity.63 Since hydrogen bond formation is only possible when the doublet is exposed outwards the cavity to the bulk, the bridgeheads of L1 adopt an in-in configuration.
The upper panel in Fig. 1 shows the aliphatic region of the high-resolution 1H NMR spectrum of L1 recorded at room temperature. Like the 13C NMR spectrum, it displays only a single set of signals corresponding to one-half of the number of atoms, indicating that the molecular cage assumes a time-averaged C2-symmetric structure on the NMR timescale. However, all methylenic protons are diastereotopic and give rise to complex geminal and vicinal coupling patterns because of the magnetic anisotropy induced by the pyridyl moiety.
Fig. 1 Experimental (upper panel) and gNMR66 simulated (lower panel) aliphatic part of the 1H NMR (500 MHz) spectrum of L1. Solvent: CDCl3; T = 20 °C. |
The methylenic hydrogen atoms in the α position to the aromatic nucleus appear as two doublets, typical of an AB spin system (2J = 16.6 Hz) at 3.78 and 3.97 ppm. Interestingly, L1 gives rise to a more pronounced anisochrony of these two pairs of protons (Δν = 95 Hz) compared to L22 (Δν = 45 ppm).60 Spectroscopic results obtained in solution together with crystallographic data for the latter diazacryptand argue for a more rigid framework where the m-xylyl linker is maintained through a three-center hydrogen bond almost perpendicular to the dioxocyclam subunit. Therefore, the higher Δν value found for L1 might reflect a more tilted orientation of the pyridine versus the benzene ring, as expected if the electrostatic repulsion between the three nitrogen lone pairs converging in the center of the cavity is taken into account. The dynamics of the back-and-forth flipping of the 2,6-pyridinedimethylene fragment has been investigated by variable temperature experiments. Cooling a sample down to 213 K in CDCl3 did not freeze out the rocking motion, which would have doubled the total number of resonances. Hence, the activation barrier for the interchange between the various conformational arrangements adopted by the macrocyclic scaffold is well below 40 kJ mol−1 at temperatures as low as −50 °C. Similarly, the lineshape remained unchanged upon heating up to 100 °C.
Two-dimensional homonuclear (1H–1H COSY) and heteronuclear (1H–13C HETCOR) experiments allowed the unequivocal assignment of all proton and carbon resonances (see ESI), although their relative exo or endo orientation with respect to the aromatic cross-bridge could not be determined by these techniques. The most up-field shifted triplet in the 1H NMR spectrum arises from the diastereotopic methylenic protons HA and HA′ located in α positions with respect to the carbonyl group.43 The spectral pattern can be analyzed in terms of coincidentally isochronous nuclei coupled with the neighboring pair of HB (2.84 ppm) and HB′ (3.03 ppm) protons, since they appear as a two resolved multiplets in D2O. The fact that two distinct peaks are also observed in the case of ligand L22 further supports this view.60 The splitting into five lines of the signal assigned to proton HB′ suggests that both vicinal 3J scalar coupling constants take almost identical values and that these should be very close to one-half of |2JB,B′|. Likewise, similar values for 3JB,A and 3JB,A′ (3JB,A = 3JB,A′ < 1/2 × |2JB,B′|) are expected in order to explain the doublet of triplets attributed to HB. The set of ethylenediamine proton resonances is readily identified by a COSY experiment. While HC and HC′ resonate respectively at 2.52 and 2.89 ppm as doublets of doublets of doublets giving rise to eight-line patterns, protons labelled HD and HD′ appear as complex 12-line manifolds indicating additional coupling with the adjacent amide proton (broad singlet at 9.72 ppm). Scalar coupling between the HD and N–H protons is further confirmed by a cross peak in the COSY correlation. Assignment of the well-resolved experimental 1H NMR spectrum was confirmed by numerical simulation using the gNMR software (lower panel in Fig. 1).66 Iterative refinement of the entire spectrum by full lineshape analysis further enabled an accurate determination of the chemical shifts of L1, as well as those of the geminal and vicinal coupling constants, which are reported in detail in the Experimental section.
(1) |
(2) |
(3) |
L1 was titrated from low to high p[H] (p[H] = −log[H3O+]) in the range 2.1–11.8 and vice versa. The neutralization curve of an acidic solution shows a first and sharp inflection point around p[H] 6 after the addition of one equivalent of base, followed by a buffering region that extends until the second equivalent is reached. Thus, L1 possesses at least two well-separated protonation constants, one close to 9 and one below 3. Nonlinear least squares refinement of four independent data sets with HYPERQUAD67 indicated that only two out of the three protonation constants can be measured by glass-electrode potentiometry. The average values are log K011 = 8.94(1) and log K012 = 2.32(9), the numbers in parenthesis corresponding to the standard deviations of four replicates expressed as the last significant digit.
Compared to 5,12-dioxocyclam [log K011 = 7.60(1) and log K012 = 7.44(3)],43 the cross-bridging of the secondary amines profoundly affects the proton affinity since the value of the first protonation constant of L1 is about two orders of magnitude higher. Assuming an in,in orientation of both electron doublets belonging to the bridgehead nitrogen atoms, as suggested by the spectroscopic properties and the DFT calculations, the basicity increase encountered by L1 indicates that the first added proton is encrypted and tightly bound inside the macrocyclic cavity. The more acidic nature of 2,6-dimethylpyridine (log K011 = 6.73; I = 0; T = 25 °C)68 also corroborates this assumption and clearly evidences a so-called cryptand effect.63,69
Replacement of the dioxocyclam fragment by a cyclam unit further increases the ligand’s basicity. Indeed, the first protonation constant of the corresponding macrobicyclic cage was found to be too high to be measured by glass-electrode potentiometry and thus the ligand was considered as a proton sponge.39 The electron-withdrawing properties of amide groups might explain the weaker hydrogen ion affinity of L1 compared to its cyclam-based analog. In both cases, however, the stabilization brought by the bridging group strongly suggests that all three donor atoms are involved in the proton binding, probably via intramolecular hydrogen bonds, although the precise localization from a microscopic point-of-view cannot be inferred at this stage.
The presence of a cross-linker also exerts a strong influence on the second protonation constant. Whereas both secondary amines of 5,12-dioxocyclam behave independently of each other, minimizing the repulsive Coulombic interactions between positive charges,43 the addition of a second proton to L1 is accompanied by a high energetic cost as evidenced by the very large gap between K011 and K012 (6.6 log units). The confinement inside the cavity provided by a rather rigid cage might explain this difference in binding strength, which is much larger than that expected from a purely statistical point of view.70 Thus, factors such as hydrogen bond breakup, conformational reorganization or charge accumulation, which increase the repulsion energy, might be invoked to explain the low value of K012. This difficulty to protonate a second site also suggests an inside location of the additional hydrogen atom.20,21,71,72
In order to get a more precise picture of the acid-base properties displayed by L1, the UV absorption changes taking place upon protonation of the bridgehead tertiary amines and/or pyridine core have been investigated in the 230–300 nm range. Since nitrate salts strongly absorb in this region, spectroscopically “silent” inert electrolyte and acid solutions such as KCl and HCl had to be used. The electronic spectrum of the free base tails out below 250 nm and shows the so-called pyridine B band (π → π*) centered at 263.5 nm with an extinction coefficient (εmax = 2930 M−1 cm−1) that is close to the value found for pyridine in water (εmax = 2750 M−1 cm−1 at λmax = 257 nm) but slightly lower to that of 2-methylpyridine (εmax = 3560 M−1 cm−1 at λmax = 262 nm).59 Acidification of the medium down to p[H] ∼ 7 (Fig. 2) induces severe perturbations of the spectra: the broad B band undergoes a slight bathochromic shift (λmax = 267 nm), accompanied by a strong hyperchromic effect resulting in the doubling of the apparent extinction coefficient. Concomitantly, the intensity of the high-energy absorption tail decreases as a consequence of a blue shift of the corresponding absorption maximum. Further addition of acid causes, between p[H] 4 and 2.5, a strong decrease in intensity of the major π → π* transition, which resolves into two peaks (λmax = 262 and 268 nm) and shows a low-energy shoulder near 257 nm (Fig. 2). Lowering the p[H] below 2.5 did not result in additional spectral changes.
Fig. 2 Spectrophotometric titration of L1 by HCl in water. [L1]tot = 1.95 mM; I = 0.1 (KCl); T = 25 °C; l = 0.1 cm. Spectra 1–36: p[H] = 2.52, 2.54, 2.56, 2.59, 2.62, 2.65, 2.68, 2.71, 2.75, 2.78, 2.83, 2.87, 2.93, 2.99, 3.05, 3.13, 3.23, 3.35, 3.51, 3.76, 4.38, 7.40, 8.15, 8.46, 8.68, 8.87, 9.04, 9.21, 9.40, 9.61, 9.85, 10.11, 10.33, 10.49, 10.62, 10.73. |
Factor analysis based on singular-value decomposition of the entire multiwavelength data set, recorded between p[H] 2.52–10.73 and comprising 36 spectra with 448 evenly distributed wavelengths ranging between 230 and 300 nm (Fig. 2), indicated that three eigenvectors are sufficient to describe the observed features above the noise level. Subsequent nonlinear least-squares treatment of the experimental absorbance values with the program SPECFIT73–75 confirmed the hypothesized chemical model, comprising only the three absorbing species previously detected by potentiometry. The best fit converged to the final values log K011 = 8.98(2) and log K012 = 2.40(8) with an overall residual standard deviation of 1.25 × 10−4 absorbance unit. Identical results within experimental error were also obtained by HYPERQUAD when the potentiometric and spectrophotometric data were refined together. Moreover, both protonation constants and their associated errors are in good agreement with those determined by glass-electrode measurements alone [log K011 = 8.94(1) and log K012 = 2.32(9)].
Fig. 3(a) displays the calculated absorption spectra of L1 (λmax = 263.5 nm; εmax = 2700 M−1 cm−1), [H(L1)]+ (λmax = 267 nm; εmax = 5725 M−1 cm−1), and [H2(L1)]2+ (λmax = 262 nm; εmax = 3350 M−1 cm−1 and λmax = 268 nm; εmax = 2657 M−1 cm−1), which are in excellent agreement with the spectra recorded after dissolution of L1 in basic or strongly acidic aqueous media. It is well-known that the formation of a pyridinium cation induces a strong hyperchromic effect on the π → π* transition, while protonation of an amine engages the lone pair of the nitrogen atom and therefore moves the n(N) → π* charge-transfer (CT) transition to higher energy.59,76 Hence, the spectral changes observed upon addition of the first equivalent of hydrochloric acid are consistent with the protonation of the pyridine nitrogen atom. The close resemblance of the limiting high- and low-pH absorption spectra of 2,6-bis(bromomethyl)pyridine with those exhibited by L1 and [H(L1)]+ further support this assignment.
Fig. 3 (a) Calculated electronic absorption spectra of L1 and its protonated forms. (b) Comparison of the absorption spectra of [H2(L1)]2+ and of the monoprotonated form of N,N′-dimethyl(pyridine-2-yl)methylamine. I = 0.1 (KCl); T = 25 °C. |
The remarkably strong hypochromic effect observed upon formation of the diprotonated species is associated to the disappearance of the absorption due to the pyridinum chromophore and suggests a redistribution of the added hydrogen ions between both types of binding sites. Thus, regioselective protonation is expected to occur at both bridgehead tertiary amines, together with a concomitant deprotonation of the pyridine nitrogen atom (Scheme 2).
Scheme 2 |
Additional stabilization of the [H2(L1)]2+ species might be provided by hydrogen bonding if the ligand takes up a conformation in which the lone pair of the pyridine nitrogen atom interacts with one or possibly both ammonium protons, hence reducing the Coulombic repulsion between positive charges in the center of the cavity. The rearrangement of the hydrogen bonding pattern induced by the second protonation is also consistent with the spectral similarity of [H2(L1)]2+ and the monoprotonated form of N,N′-dimethyl(pyridine-2-yl)methylamine [Fig. 3(b)]. The slight shift of ca. 5 nm between both spectra is ascribed to the different substitution level of the aromatic ring for each substance.
1H NMR | 13C NMR | |||||
---|---|---|---|---|---|---|
Signal | L1 | [D(L1)]+ | [D2(L1)]2+ | L1 | [D(L1)]+ | [D2(L1)]2+ |
A | 2.24 | 2.29 | 2.91 | 33.8 | 35.3 | 31.2 |
2.33 | 2.55 | |||||
B | 2.84 | 2.78 | 3.69 | 50.7 | 51.9 | 51.5 |
2.92 | 3.04 | 4.13 | ||||
C | 2.45 | 2.64 | 3.40 | 54.1 | 53.2 | 59.9 |
2.84 | 2.96 | 3.91 | ||||
D | 3.01 | 3.30 | 3.23 | 38.7 | 36.7 | 36.0 |
3.22 | 3.62 | 4.32 | ||||
E | 3.87 | 4.14 | 4.83 | 58.3 | 53.7 | 57.6 |
F | 159.6 | 153.1 | 148.6 | |||
G | 7.13 | 7.79 | 7.62 | 120.7 | 124.8 | 124.2 |
H | 7.66 | 8.41 | 8.13 | 138.7 | 147.0 | 142.1 |
CO | 176.0 | 175.3 | 174.9 |
The NMR data confirm the protonation scheme deduced from the spectrophotometric UV titration. Formation of a pyridinium cation between p[D] 11.8 and 5.4 is evidenced by the following observations: (i) the aromatic protons HG (Δδ = 0.66 ppm) and HH (Δδ = 0.75 ppm) are strongly deshielded, (ii) in CDCl3 containing traces of acid, a broad lowfield-shifted resonance appears at 13.55 ppm in the 1H NMR spectrum, which is characteristic for a hydrogen-bonded pyridinium proton.78,79 As the [D(L1)]+ species is converted into the [D2(L1)]2+ one, relocation of the deuteron from the pyridinum to the amine sites causes an upfield shift of the aromatic signals (Δδ = 0.17 and 0.28 ppm for the HG doublet and HH triplet, respectively), while the methylenic HE protons from the handle resonate at much lower field (Δδ = 0.69 ppm). Moreover, the occurrence of an off-diagonal cross-peak that correlates the HB and HC signals in the COSY spectrum of [D2(L1)]2+ implies a four-bond coupling through the bridgehead nitrogen atoms, which further argues in favor of their deuterated state. Deuteration of both tertiary amines is also corroborated by the upfield shift of the resonances assigned to the carbon atoms CA and CD located in β positions with respect to the bridgehead nitrogen atom (Table 1).80,81
Fig. 4 shows a stack plot of pertinent 300 MHz 1H NMR spectra recorded over the p[D] range 1.7–11.8. The most striking features are the progressive disappearance of the free base resonances upon acidification and the simultaneous build-up of the signals due to the monodeuterated species with a concomitant line broadening observed as long as some free base is present (p[D] > 7). Conversion below p[D] 5 of [D(L1)]+ to [D2(L1)]2+ occurs in a similar way. Identical spectral changes could be observed when the titration was followed by 13C NMR spectroscopy. This unexpected result is a clear indication that the deuteration kinetics of L1 is slow on the NMR timescale, whereas potentiometric studies revealed a fast protonation process in the second time range.
Fig. 4 Stack plot of representative 1H NMR spectra of L1 recorded at 300 MHz in D2O as a function of p[D]; T = 25 °C. |
Using the mass-balance eqn. (4) and the corresponding expressions of the successive equilibrium constants given by relation (2), the theoretical fractional concentration of each species in solution is described by eqns. (5) to (7), where KD01i represents the ith deuteration constant. Experimental values of the relative concentration for a given species were readily obtained by integration of the 1H NMR signals corresponding to the aromatic HG and HH protons (Fig. 5). Deuteration constants were adjusted by nonlinear minimization of the sum of the squared deviations between the theoretical and observed concentration ratios using a Microsoft Excel spreadsheet. The solid lines in Fig. 5 correspond to the calculated species distribution curves drawn with the optimized values of log KD011 = 9.36 and log KD012 = 3.10. Uncertainties are not reported since a single titration was performed. Estimates of the protonation constants in water, corrected for the deuterium effect according to the empirical correlation proposed by Delgado et al.82 (log KH011 = 8.66 and log KH012 = 2.66), are in fair agreement with those determined by potentiometric and spectrophotometric experiments, considering the error introduced in the absence of supporting electrolyte by the evolving ionic strength during the NMR titration.
[L1]tot = [L1](1 + KD011[D+] + KD011KD012[D+]2) | (4) |
(5) |
(6) |
(7) |
There are only a very few chemical systems that undergo slow protonation or deprotonation, the dynamics of such reactions being diffusion limited in the majority of cases. Typical examples are found among proton sponges.25 For the compounds belonging to the prototypical class of N-alkylated 1,8-diaminonaphthalene, the slow proton-transfer rates might be ascribed to the strong intramolecular N–H⋯N hydrogen bonding interaction.83–85 Tricyclic tetraamines, also called adamanzanes, exhibit analogous properties and were found to be very inert with respect to the proton entrapped in the cavity.34–36 Of particular relevance with respect to the present work are the macrobicyclic diazacycloalkanes20–22,26 and cryptands,63,71,86 in which the proton is tightly concealed inside an often rigid cavity. Bridged bicyclic structures of this type contain medium to large rings (at least 7-membered) in which the lone pairs on bridgehead nitrogen atoms can point either inward or outward, thus allowing inside or outside protonation.87 The slow insertion rate of a proton into the cage of the extraordinarily strong base 1,6-diazabicyclo[4.4.4]tetradecane (pKa ∼ 25) is believed to be the consequence of an indirect process involving a 1,2- or 1,5-transfer of an α-methylenic hydrogen atom.23,24 For larger representatives, internal protonation was shown to occur according to a sequence of concerted prototropic reactions and nitrogen atom inversion starting from the rapidly formed exo-protonated state.20–22 Similarly, the endo-endo conformation is the thermodynamically most stable form for the most constrained [1.1.1] cryptand and its protonated species. The slow protonation rates result from the high activation energy barrier (ca. 110 kJ mol−1) for the incoming proton, whereas nitrogen inversion followed by outside protonation is a much faster process.71 Conversion of the exo to the endo orientation of the ammonium group constitutes the rate-limiting step and most probably occurs via a nitrogen inversion pathway, which implies a deprotonation-inversion-reprotonation sequence. Because the associated rates in water are slow at room temperature, the reaction was followed as a function of time by integrating the respective signals assigned to both states of the molecule.
Fig. 5 Relative concentrations of L1 (black dots), [D(L1)]+ (blue circles), and [D2(L1)]2+ (red squares) determined by integration of the 1H NMR spectra. The nonlinear least squares distribution curves calculated according to eqns. (5)–(7) are represented as solid lines. Solvent: D2O; T = 25 °C. |
This is in contrast with the present investigation as no time dependence of the recorded spectra could be detected over a period extending up to one month. Internal protonation of L1 occurs obviously much faster since any exo conformer could never be observed by NMR spectroscopy. Assistance of the pyridine moiety might be invoked to explain the accelerated and direct entry of the protons in the cavity of L1. Assuming that the internal hydrogen-bonding network, which stabilizes both [H(L1)]+ and [H2(L1)]2+ species, is preserved in water, the pyridyl nitrogen site can act as a proton relay and slow reorganization on the NMR timescale of the three-center hydrogen bond takes place upon proton binding.
As expected, the lowest energy structure computed for L1 [Fig. 6(a)] shows an inward orientation of both tertiary amine lone pairs with a transannular N⋯N separation of 4.42 Å, which is in good agreement with the distance measured on the crystal structure of L22 (4.64 Å).60 The slightly bent dioxocyclam ring has both amidic hydrogen atoms directed towards the center of the cavity, one interacting with the nitrogen atom belonging to the tilted pyridine ring (N⋯Npy = 2.96 Å, N–H⋯Npy = 157°) and the other with the neighboring amine located 2.26 Å away (N⋯N = 3.05 Å, N–H⋯N = 134°). Hence, the molecular modelling calculations carried out on L1 reproduce all the key structural features displayed by closely related pyridine-capped C-substituted dioxocyclams.55
Fig. 6 DFT-optimized molecular structures of (a) L1 and (b) [H(L1)]+. Possible hydrogen bonds are visualized with dashes. For the sake of clarity, only the pertinent hydrogen atoms are represented. |
The reliability of the minimization procedure was checked by examining the planarity of both amide groups, which is conveniently expressed by the OC–N–H torsion angles, which are 166 and 169°. Further validation was provided by computing estimates of the NMR 3JαNH coupling constants corresponding to the H–N–C–H dihedral angles (ϕ) using a revised Karplus equation.89 For the axial methylenic proton adjacent to the amide group (ϕ = −76 and −83.3°), 3JαNH averages 6.7(5) Hz, which is in fair agreement with the value derived from the full lineshape analysis of the C2-symmetric time-averaged 1H NMR spectrum recorded at room temperature in CDCl3. The magnitude of the vicinal coupling constant with the equatorial hydrogen atom (ϕ = 167.9 and 161.2°) is expected to be smaller [3JαNH = 2.5(5) Hz] as experimentally observed (3JαNH = 3.9 Hz).
It is therefore clear that the DFT calculations are fully supported by a vast collection of experimental and spectroscopic data in this particular case. This reliable tool can now be used efficiently to describe the effect of protonation on the geometry of L1. Thus, from a structural point of view, formation of the monoprotonated species breaks the intramolecular hydrogen-bond network and results in a ca. 90° rotation of both almost planar amide groups (OC–N–H = −176 and −179°), affording a cone-shaped molecular cavity delimited by both carbonyl groups oriented on the same side as the aromatic ring [Fig. 6(b)]. In turn, protonation has only a limited influence on the distance between the bridgehead nitrogen atoms, which shrinks by only 0.07 Å, but enforces a less tilted orientation of the capping pyridine moiety. The pyridinium proton points in-between both amines in an almost perpendicular fashion with respect to the dioxocyclam skeleton, forming a three-center hydrogen bond with them (NpyH⋯N = 2.22 and 2.32 Å, sum of NHN angles = 359.6°) as already found in the free-base structure of L22.60 Furthermore, molecular modelling of the diprotonated species confirmed that the in,in-diammonium salt was the most stable conformer (see ESI).
Titrations were performed using a T110 (Schott) piston burette equipped with a calibrated TA10 interchangeable unit of 10 mL and a TR250 (Schott) data acquisition unit, both instruments being controlled by TR600 software (release 5.2) running on an IBM-compatible PC computer. The volume of the burette-delivered aliquots was corrected according to a linear calibration function obtained by weighing known quantities of water and by taking into account the buoyancy effect.91 Magnetically stirred solutions were maintained under an argon atmosphere and at constant temperature [25.0(1) °C] by using a water-jacketed titration vessel fitted to a Lauda RE 106 water circulator.
Protonation constants were measured by titrating ca. 0.1 mmol of ligand dissolved in 25 mL of supporting electrolyte (0.1 M KNO3). The electrochemical cell was allowed to equilibrate for at least 30 s after each addition of a 0.03 mL increment of titrating solution. The collected potential readings were converted into p[H] values, spanning the range 2.1–11.8, with the help of an Microsoft Excel spreadsheet by iterative solving of the rearranged Nernst equation: p[H] = (E0 − Emes + Ja[H3O+] + JbKw[H3O+]−1)/S. The titration data were analyzed by the weighted nonlinear least-squares procedure implemented in the HYPERQUAD 2000 program.67 The selected weighing scheme, derived from the error propagation law,94 seeks to reduce the weight of the less accurate measurements (i.e., points located in the steep region of a titration curve). On the basis of the instrument’s calibration, the uncertainties associated with the experimental p[H] values (σpH) and the volumes delivered by the piston burette (σv) were estimated as 0.003 pH unit (or ∼0.1 mV) and 0.005 mL, respectively. In the final refinement step, the total amounts of titrated ligand and initially added base or acid were also allowed to vary. The fit between calculated and experimental emf data was evaluated through the squared sum of residuals (1.0 < σ < 1.8) and an approximately normal distribution of the residues.94 The thermodynamic reversibility was checked by cycling from low to high p[H] and vice versa. Data from forward and backward titrations afforded statistically identical values of the adjusted thermodynamic parameters, pointing out the chemical stability of the ligand over the explored pH range. For each data set, the standard deviations of the log K01h values were derived from the full variance/covariance matrix of the refined global constants (log β01h).95 The extradiagonal terms, which correspond to the correlation coefficients between the stepwise protonation constants,95 were close to 0.5. The final values are reported as the arithmetic mean of four independent measurements together with their standard deviations, which were systematically higher compared to those derived for a single experiment from the full variance/covariance matrix.
The multiwavelength data sets were decomposed into their principal components by factor analysis with the SPECFIT program.73–75 The protonation constants and the extinction coefficients corresponding to each species were subsequently adjusted to the reduced data sets by the Marquardt nonlinear least-squares algorithm. Furthermore, the data from each titration were also processed by the program HYPERQUAD 2000, which allows the simultaneous treatment of potentiometric and spectrophotometric data.67 While SPECFIT uses the measured p[H] values to calculate directly the free H3O+ concentration, the mass-balance equations for all components, including the hydronium ion, are solved at each titration point by the program HYPERQUAD. Since the precision of spectrophotometric and potentiometric measurements is different, it is of prime importance to define a proper weighing scheme relying on the error propagation rule, which gives to each type of observation an approximately equal contribution to the residual sum of squares. Realistic parameter estimates of the quadratic absorbance error function (σA = a + bA + cA2) were evaluated by computing, for each recorded wavelength of 50 replicates of the ligand spectrum, the average absorbance and its associated standard deviation.67,94
Footnotes |
† For the previous paper in this series see ref. 1. |
‡ Electronic supplementary information (ESI) available: 1H–1H NMR COSY and 1H–13C NMR HETCOR charts for L1 (Figs. S1–S3); DFT-calculated energies of L1, [H(L1)]+, and [H2(L1)]2+ in different geometries (Tables S1–S3); Cartesian x, y, z coordinates of the DFT-minimized lowest-energy conformers of L1, [H(L1)]+, and [H2(L1)]2+ structures (Tables S4–S6). See http://www.rsc.org/suppdata/nj/b4/b410260f/ |
§ This paper is dedicated to Dr Jean-Pierre Sauvage on the occasion of his 60th birthday. |
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2005 |