Comparing Positively and Negatively Charged Distonic Radical Ions in Phenylperoxyl Forming Reactions
In the gas phase, arylperoxyl forming reactions play a significant role in low-temperature combustion and atmospheric processing of volatile organic compounds. We have previously demonstrated the application of charge-tagged phenyl radicals to explore the outcomes of these reactions using ion trap mass spectrometry. Here, we present a side-by-side comparison of rates and product distributions from the reaction of positively and negatively charge tagged phenyl radicals with dioxygen. The negatively charged distonic radical ions are found to react with significantly greater efficiency than their positively charged analogues. The product distributions of the anion reactions favor products of phenylperoxyl radical decomposition (e.g., phenoxyl radicals and cyclopentadienone), while the comparable fixed-charge cations yield the stabilized phenylperoxyl radical. Electronic structure calculations rationalize these differences as arising from the influence of the charged moiety on the energetics of rate-determining transition states and reaction intermediates within the phenylperoxyl reaction manifold and predict that this influence could extend to intra-molecular charge-radical separations of up to 14.5 Å. Experimental observations of reactions of the novel 4-(1-carboxylatoadamantyl)phenyl radical anion confirm that the influence of the charge on both rate and product distribution can be modulated by increasing the rigidly imposed separation between charge and radical sites. These findings provide a generalizable framework for predicting the influence of charged groups on polarizable radicals in gas phase distonic radical ions.
KeywordsDistonic ions Phenyl radicals Peroxyl radicals Ion-molecule reactions Reaction kinetics Electronic structure calculations
Organic peroxyl radicals (ROO●) are an important class of reaction intermediate in the mechanisms of low-temperature combustion as well as in the oxidative processing of volatile organic compounds in the Earth’s lower atmosphere [1, 2]. Despite the importance of organic peroxyl radicals in these chemistries, the direct observation of peroxyl formation and decomposition reactions has remained a significant challenge for conventional experimental techniques. This is because, in the gas phase, peroxyl radicals are typically generated in very low concentrations and have a high propensity for subsequent reaction—including decomposition and self-reactions—resulting in relatively short lifetimes. Building on the pioneering work by Kenttämaa and co-workers [3, 4, 5, 6], a number of groups—including our own—have exploited the distonic radical ion approach for studying the chemistry of organic peroxyl radicals in the gas phase using ion cyclotron resonance and ion trap mass spectrometries [7, 8, 9, 10, 11, 12, 13, 14, 15, 16].
Distonic ions have radical centers that are spatially separated from the charge site [17, 18]. This class of radical ions has been much-less studied than conventional radical ions (where charge and radical are co-localized) since the latter are more commonly produced by traditional ionization techniques, namely, electron- and photo-ionization . Despite requiring greater regiochemical control in their synthesis, distonic ions can represent the global minimum on the radical ion surface and, moreover, provide an attractive model for studying radical reactivity by mass spectrometry. The benefits in deploying this strategy lie in the ability to generate high concentrations of reactive species without complicating self- or cross-reactions (owing to Coulombic repulsion of like charges), combined with the power to cleanly isolate and identify reaction products through changes in mass-to-charge ratio (m/z) [6, 20]. The role of the charge itself in modulating the reactivity of the remote radicals, however, has been the subject of recent interest, with computational studies indicating that polarizable peroxyl and nitroxyl radicals may be stabilized significantly by the presence of a negative charge in a distonic radical anion [21, 22]. Stabilization of the radical anion by up to 5 kcal mol−1 relative to neutral radical analogues has been reported, with smaller, but still significant effects, computed over intra-molecular distances > 10 Å. These findings suggest that radical reactivity may be modulated by the presence of a remotely charged moiety with both the charge polarity and charge-radical separation being key determinants of reaction outcomes.
3-Iodo-N,N,N-trimethylanilinium iodide and 4-iodo-N,N,N-trimethylanilinium iodide were previously synthesized according to literature methods [15, 24, 25]. 3-(4-Iodophenyl) adamantane-1-carboxylic acid was obtained from VBR Molecules (Hyderabad, India). Potassium 3-iodophenyltrifluoroborate (≥ 97%) and potassium 4-iodophenyltrifluoroborate (≥ 97%) were purchased from Advanced Molecular Technologies (Scoresby, Australia). 3-Iodobenzoic acid (98%), 4-iodobenzoic acid (98%), and dimethyl disulfide (98%) were purchased from Sigma-Aldrich (St. Louis, MO). Methanol (Optima LC/MS grade) was purchased from Thermo Fisher Scientific (Scoresby, Australia). Compressed helium (ultrahigh purity, 99.999%) was obtained from Coregas (Sydney, Australia). All commercial compounds were used as received without further purification.
Gas phase ion-molecule experiments were performed on a modified [25, 26] linear quadrupole ion trap mass spectrometer (LTQ XL, Thermo Fisher Scientific, San Jose, CA) operating Xcalibur version 3.0 software and equipped with a heated electrospray ionization (HESI) source. The HESI source was connected to both the instrument syringe pump and an HPLC system (Dionex UltiMate 3000 RSLC, Thermo Fisher Scientific, San Jose, CA) via a tee union. Methanol or acetonitrile solutions containing 5–10 μM of individual radical precursors were tee-infused at a rate of 5–10 μL min−1 into an HPLC flow of 100% methanol or acetonitrile (50–100 μL min−1) that eluted into the source. Typical source conditions for negative ion mode were capillary temperature 250 °C, source heater temperature 250 °C, sheath gas flow rate 25–30 arbitrary units, auxiliary and sweep gas flow rates 0–5 arbitrary units, spray voltage − 3.5–4 kV, capillary voltage − 35–40 V, and tube lens − 45–50 V. Typical source conditions for positive ion mode were capillary temperature 250 °C, source heater temperature 50 °C, sheath gas flow rate 25–30 arbitrary units, auxiliary and sweep gas flow rates 0–5 arbitrary units, spray voltage 4–4.5 kV, capillary voltage 10–20 V, and tube lens 35–45 V. For collision-induced dissociation experiments, ions were mass-selected with an isolation width of 1.5–2.0 Th, using an activation q parameter of 0.250. The normalized collision energy applied was typically 25–30 arbitrary units with an excitation time of 30 ms (unless otherwise specified).
Previous work by ourselves and others has indicated that synthesis of distonic radical ions by photodissociation can give greater regioselectivity than collision-induced dissociation [5, 27]. To facilitate photodissociation of trapped ions, instrument modifications that allow laser irradiation of ions confined within the linear ion trap are similar to those reported by Ly and Julian and have been described in detail elsewhere [12, 28]. Briefly, the posterior plate of the mass spectrometer was modified with a quartz window to transmit the fourth harmonic (266 nm) of a Nd:YAG laser (Continuum, Santa Clara, CA). The window was positioned on the backplate to direct laser access to the ions within the ion trap through the 2 mm aperture centered on the back lens. Laser pulses were aligned through the aperture in the back lens by two right-angle steering prisms, adjusted to optimize beam overlap with the ion cloud (see Figure S1, Supporting Information). Laser pulses were synchronized to the beginning of an MS2 ion activation step with a TTL trigger signal generated by the mass spectrometer to the laser via a digital delay generator . In these photodissociation experiments, the normalized collision energy was set to 0 (arbitrary units), such that all product ions arise from excitation by the laser pulse. Only a single laser pulse irradiates the target ions in each MS cycle.
Instrument modifications that allow for neutral reagent molecules to be seeded into the ion trap region of the instrument are similar to those previously described [15, 25]. Briefly, the native helium splitter was bypassed in order to directly connect an external reagent mixing manifold to the ion trap (see Figure S1, Supporting Information). Helium flow to the ion trap was controlled by a variable leak valve (Granville-Phillips Model 203, Boulder, CO) providing a pressure reading of 0.8 × 10−5 Torr on the instrument ion gauge. Neutral reagents, such as dimethyl disulfide, were introduced by placing a droplet of liquid into the end-cap of a stainless steel Swagelok cap and placing under vacuum before opening the line to the manifold. The flow of reagent was controlled by a PEEKsil restriction (100 mm length, internal diameter 25 μm). For observing reactions with dioxygen, no reagent was added and adventitious background air within the instrument provided a concentration of dioxygen that was sufficient for the experiments.
Pseudo-first-order rate constants, k1 (s−1), were obtained by recording a series of mass spectra as a function of storage time for the reaction between the radical ion and neutral reagent. Each spectrum recorded was an average of at least 60 individual spectra. The reaction time was defined as the interval between the isolation of the selected radical ion and ejection of all ions from the trap for detection. Reaction times of 50–10,000 ms were set using the activation time parameter within the instrument control software. A plot of the mean radical ion abundance against reaction time yielded a single exponential relationship. Fitting the data to Eq. (1) gave the pseudo-first-order rate constant for the reaction, k1 (s−1). Second-order rate constants, k2 (cm3 molecule−1 s−1), were obtained from k1 and the concentration of dioxygen ([O2]) present in the ion trap region of the mass spectrometer (Eq. (2)) , where [O2] was determined from the calibration reaction of 3-carboxylatoadamantyl radical anion with dioxygen under the same instrumental conditions . Reaction efficiencies, Φ (i.e., the percentage of ion-molecule collisions that resulted in product formation), were determined according to Eq. (3), where the collision rate constant, kcoll, is determined by average dipole orientation (ADO) theory at 307 K . The temperature of ions stored within a linear quadrupole ion trap has been estimated at 318 ± 23 K , consistent with an earlier estimate of 307 ± 1 K , which can be taken as the effective temperature for ion-molecule reactions observed herein [33, 34]. While the accuracy of the rate constant measurements is estimated to be ± 50%, the precision of the measurements is better than ± 10%.
The Gaussian 09 suite of programs was used for all computations . Electronic energies and thermally corrected (298 K) enthalpies for all ground-state species were computed using the hybrid meta-GAA M06-2X functional [36, 37] and the 6–311++G(d,p) basis set. DFT calculations for doublet states employed an unrestricted formalism. This computational strategy has previously been shown to be effective in modeling the thermochemical and spectroscopic properties of gas phase arylperoxyl radicals . All stationary points on the potential energy surface were verified to be either local minima (no imaginary frequencies) or transition states (one imaginary frequency) by computation of analytic vibrational frequencies. Transition states were connected to minima by calculation of the intrinsic reaction coordinate (IRC). Outputs from all calculations discussed in this study are summarized in the Supporting Information.
Results and Discussion
Formation of Charge-Tagged Phenyl Radicals in the Gas Phase
Aryl iodide precursor ions were generated by electrospray ionization (in either positive or negative ion mode), mass selected, and subsequently irradiated at 266 nm within the ion trap mass spectrometer to liberate the distonic radical ions 1–7 (Chart 1). Photodissociation of aryl iodides has been demonstrated as an effective means of generating aryl radicals with high regiospecificity, and the approach has been deployed to great effect by Julian and co-workers for molecular structure elucidation via radical-directed dissociation [28, 38]. In this study, the putative structure of the radical ions was confirmed for each instance by observing the gas phase reaction with dimethyl disulfide [11, 25, 39, 40, 41]. For all radical ions 1–7, the reaction resulted in thiomethyl abstraction consistent with the distonic radical ion structures shown in Chart 1 (see example spectrum in Figure S2, Supporting Information).
Charge Polarity Effect on Products of Phenyl Radical Reactions
Charge Polarity Effect on the Reaction Potential Energy Surface
The computational results, summarized in Figure 3, illustrate that addition of O2 to the neutral phenyl radical produces the phenylperoxyl radical (N2) with an exothermicity of 46.6 kcal mol−1. Homolytic cleavage of the O–O bond in N2 yields the phenoxyl radical plus atomic oxygen (3P) (N3) that is 7.9 kcal mol−1 below N1. Alternatively, the oxygen-centered radical of the peroxyl moiety in N2 can undergo addition to the ipso-carbon of the aromatic ring forming the dioxiranyl intermediate (N4) that is 26.3 kcal mol−1 below the entrance channel on the neutral surface. This isomerization occurs through the transition state TS N2 → N4 that represents a barrier of 31.3 kcal mol−1 with respect to the phenylperoxyl radical (N2) but is still some 15.3 kcal mol−1 below N1.
In the presence of the positively charged trimethylammonium moiety at the para-position, the addition of dioxygen to the phenyl radical cation (C2) is exothermic by 42.8 kcal mol−1 (Figure 3, red lines). Interestingly, on the cation surface, the minimum for the phenylperoxyl radical C2 lies in a well that is 3.8 kcal mol−1 shallower than the comparable neutral peroxyl radical (N2). The p-(N,N,N-trimethylammonium)phenoxyl radical (C3) formed upon homolysis of the O–O bond in C2 is 7.1 kcal mol−1 higher in energy than the neutral phenoxyl radical and lies just 0.8 kcal mol−1 below the entrance channel (C1). Similarly, the transition state TS C2 → C4 is predicted to lie 12.0 kcal mol−1 below C1, which is 3.8 kcal mol−1 above the comparable barrier on the neutral surface (TS N2 → N4). Furthermore, the dioxiranyl intermediate C4 is also predicted to be 6.2 kcal mol−1 higher in energy than the neutral archetype N4, lying 20.1 kcal mol−1 below C1. Interestingly, when the fixed positive charge-tag is substituted at the meta-position with respect to the radical site, the destabilizing effect of the charge-tag does not vary significantly from the para-substituted isomer (cf. values in parentheses in Figure 3). This finding indicates that the difference between the positively charged and neutral phenyl radical potential energy surfaces is not strongly influenced by formal charge or radical delocalization around the aromatic ring.
Considering the negatively charged analogue, addition of dioxygen to the p-trifluoroboratophenyl radical anion forms the p-trifluoroboratophenylperoxyl radical anion (A2) which resides in a minimum some 51.3 kcal mol−1 below the entrance channel (A1; Figure 3). When normalized to the energy of the entrance channel, the negatively charged peroxyl radical is 4.7 kcal mol−1 below its neutral counterpart N2. Comparison to the corresponding cation C2 reveals that the stabilization of the peroxyl radical by the proximate negative charge is larger in magnitude than the destabilization in the cation (cf. 3.8 kcal mol−1). The p-trifluoroboratophenoxyl radical anion (A3) formed upon cleavage of the O–O bond in A2 is predicted to be 9.6 kcal mol−1 lower in energy than N3, lying 17.5 kcal mol−1 below A1. The barrier to isomerization of A2 to form the dioxiranyl intermediate A4 is predicted to be 24.7 kcal mol−1 (TS A2 → A4) and is thus 4.7 kcal mol−1 lower in energy than the equivalent neutral transition state TS N2 → N4. A4 is computed to be 8.8 kcal mol−1 lower in energy than N4, lying 35.1 kcal mol−1 below the entrance channel. The net lowering of the potential energy surface relative to the entrance channel is also observed when the fixed negative charge-tag is in the meta-position with respect to the radical site (values in parentheses in Figure 3), indicating that resonance effects are only minor contributors to the overall stabilization afforded by the anion moiety.
The normalized reaction coordinate diagrams of the positively and negatively charge-tagged phenyl radicals in Figure 3 give some insight into the differences in experimentally observed reaction products for the two polarities (vide supra). For example, in the positively charge-tagged system, the barriers to direct decomposition or rearrangement of the intermediate peroxyl radicals 1-OO and 2-OO lie much closer to the entrance channel (cf. red lines in Figure 3). These findings suggest that the rate of decomposition will be significantly slower than for the corresponding anions or neutrals and affording a greater contribution from stabilization of the peroxyl adduct. This is consistent with the observation of phenylperoxyl radical formation (i.e., 1-OO and 2-OO) as the predominant reaction product with only trace amounts of the [M+O2-O]•+ and [M+O2-CHO]•+ decomposition products (see Figure 1). Conversely, for the negative ions, the transition states and intermediates in the decomposition pathways are lowered substantially relative to both the cation and neutral analogues (cf. blue lines in Figure 3). These calculations are thus consistent with higher rates for unimolecular transformation of the intermediate peroxyl radicals 3-OO and 4-OO to decomposition products. Such rapid transformations could outcompete stabilization of the adduct ions and explain why only trace amounts of the 3-OO and 4-OO anions are observed experimentally (Figure 2). The computational results illustrated in Figure 3 suggest that the distonic ion reactivity observed in this study is likely to bracket that of the neutral phenyl radicals. It would be interesting therefore to assess whether increasing the distance between charged and radical moieties might allow convergence (from above and below) on the energetics of the neutral phenyl radical archetype.
Charge-Radical Distance Effects on the Reaction Potential Energy Surface
To experimentally test the effect of charge-radical separation on the products of arylperoxyl forming reactions, the reaction of the 4-(1-carboxylatoadamantyl)phenyl radical anion (7) with dioxygen was investigated in the ion trap mass spectrometer and compared side-by-side with the reactions of the 4-carboxylatophenyl radical (5). Molecular orbital calculations estimate the distance between the radical site and the charged moiety in 7 is 9.1 Å, representing an increase in separation of 5.3 Å compared with 5 (data not shown). Figure 5a and b shows the mass spectra obtained after isolation of the radical anions 5 and 7, respectively, in the presence of dioxygen for 5 s. The spectrum obtained from 4-carboxylatophenyl radical anion (m/z 120) is dominated by the phenoxyl radical product ion at m/z 136 with the corresponding phenylperoxyl radical (5-OO) at m/z 152 only present at very low abundance (see ×40 magnification in Figure 5a). In contrast, analogous reaction of the 4-(1-carboxylatoadamantyl)phenyl radical anion (m/z 254) with O2 yields an abundant 7-OO product ion at m/z 286 (Figure 5b), with the phenoxyl radical plus atomic oxygen product channel (m/z 270) making a significantly diminished contribution (see ×40 magnification in Figure 5b). These observations are consistent with the computational prediction (vide supra) that increasing the charge-radical separation raises the barriers toward decomposition and isomerization of the phenylperoxyl radical, thus providing a greater opportunity for stabilization of the adduct ion.
Charge Polarity and Charge-Radical Distance Effects on the Rate of Phenyl Radical Reactions
Comparison of Second-Order Rate Constants and Reaction Efficiencies for the Ion-Molecule Reactions of Distonic Radical Ions 1–4 with O2
Through a combined experimental and computational examination, we have demonstrated that the polarity of the charge-tag significantly perturbs the potential energy surface of the Ph●+O2 reaction system, altering both the reaction kinetics as well as the reaction products observed. The presence of a fixed-positive charge was found to destabilize influential stationary points on the Ph●+O2 potential energy surface relative to the entrance channel. In contrast, a fixed-negative charge was found to lower the energy of the same stationary points on the corresponding anionic potential energy surface. The opposite but unequal perturbation arising from these charge-tags can account for the different product distributions of the Ph●+O2 reaction and also the measured differences in reaction efficiency. Computational exploration of these effects with increasing intra-molecular separation of charge and radical sites suggests that, in the gas phase, the charge can influence the energetics associated with the formation and fate of peroxyl radicals over surprisingly long distances (e.g., up to 14.5 Å). These predictions were supported by experimental examination of the reactivity of the 4-(1-carboxylatoadamantyl)phenyl radical anion (7), where the molecular structure imposes a charge-radical separation of 9.1 Å. Increasing the charge-radical separation in this system was shown to retard the reaction efficiency and significantly alter the product distribution in favor of the phenylperoxyl radicals. The demonstrated influence of remote charges in modulating peroxyl-forming reactions may have broader implications [21, 22]. In a biological context for example, the presence of remotely charged moieties may afford switching of peroxyl radical reactivity in enzymatic or free radical-induced oxidation chemistries. While the effective range of influence of charge-radical polarization would be diminished in polar solvents, in low-polarity regimes such as hydrophobic pockets of proteins or membrane bilayers, the effect could be comparable to that observed in the gas phase.
The data reported in this paper were obtained at the Central Analytical Research Facility (CARF) operated by the Institute for Future Environments at the Queensland University of Technology. Access to CARF is supported by generous funding from the Science and Engineering Faculty (QUT). A.J.T., B.L.J.P., and S.J.B. acknowledge financial support from the Australian Research Council (ARC) through the Discovery Project scheme (DP140101237 and DP170101596). The authors also acknowledge the generous allocation of computing resources by the NCI National Facility (Canberra, Australia) under Merit Allocation Scheme.
- 9.Harman, D.G., Blanksby, S.J.: Trapping of a tert-adamantyl peroxyl radical in the gas phase. Chem. Commun. 859–861 (2006)Google Scholar
- 10.Moore, B.N., Blanksby, S.J., Julian, R.R.: Ion–molecule reactions reveal facile radical migration in peptides. Chem. Commun. 5015–5017 (2009)Google Scholar
- 15.Kirk, B.B., Harman, D.G., Kenttamaa, H.I., Trevitt, A.J., Blanksby, S.J.: Isolation and characterization of charge-tagged phenylperoxyl radicals in the gas phase: direct evidence for products and pathways in low temperature benzene oxidation. Phys. Chem. Chem. Phys. 14, 16719–16730 (2012)CrossRefPubMedGoogle Scholar
- 26.Habicht, S.C., Vinueza, N.R., Archibold, E.F., Penggao, D., Kenttamaa, H.I.: Identification of the carboxylic acid functionality by using electrospray ionization and ion-molecule reactions in a modified linear quadrupole ion trap mass spectrometer. Anal. Chem. 80, 3416–3421 (2008)CrossRefPubMedGoogle Scholar
- 31.Langevin, P.: A fundamental formula of kinetic theory. Ann. Chem. Phys. 5, 245 (1905)Google Scholar
- 35.Frisch, M.J., Trucks, G.W., Schlegel, H.B., Scuseria, G.E., Robb, M.A., Cheeseman, J.R., Scalmani, G., Barone, V., Mennucci, B., Petersson, G.A., Nakatsuji, H., Caricato, M., Li, X., Hratchian, H.P., Izmaylov, A.F., Bloino, J., Zheng, G., Sonnenberg, J.L., Hada, M., Ehara, M., Toyota, K., Fukuda, R., Hasegawa, J., Ishida, M., Nakajima, T., Honda, Y., Kitao, O., Nakai, H., Vreven, T., Montgomery Jr., J.A., Peralta, J.E., Ogliaro, F., Bearpark, M.J., Heyd, J., Brothers, E.N., Kudin, K.N., Staroverov, V.N., Kobayashi, R., Normand, J., Raghavachari, K., Rendell, A.P., Burant, J.C., Iyengar, S.S., Tomasi, J., Cossi, M., Rega, N., Millam, N.J., Klene, M., Knox, J.E., Cross, J.B., Bakken, V., Adamo, C., Jaramillo, J., Gomperts, R., Stratmann, R.E., Yazyev, O., Austin, A.J., Cammi, R., Pomelli, C., Ochterski, J.W., Martin, R.L., Morokuma, K., Zakrzewski, V.G., Voth, G.A., Salvador, P., Dannenberg, J.J., Dapprich, S., Daniels, A.D., Farkas, Ö., Foresman, J.B., Ortiz, J.V., Cioslowski, J., Fox, D.J.: Gaussian 09, revision B.01 Gaussian, Inc., Wallingford, CT, USA (2009)Google Scholar
- 37.Zhao, Y., Truhlar, D.G.: The M06 suite of density functionals for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: two new functionals and systematic testing of four M06-class functionals and 12 other functionals. Theor. Chem. Accounts. 120, 215–241 (2008)CrossRefGoogle Scholar
- 43.Prendergast, M.B., Kirk, B.B., Savee, J.D., Osborn, D.L., Taatjes, C.A., Masters, K.-S., Blanksby, S.J., Da Silva, G., Trevitt, A.J.: Formation and stability of gas-phase o-benzoquinone from oxidation of ortho-hydroxyphenyl: a combined neutral and distonic radical study. Phys. Chem. Chem. Phys. 18, 4320–4332 (2016)CrossRefPubMedGoogle Scholar
- 44.Ormond, T.K., Scheer, A.M., Nimlos, M.R., Robichaud, D.J., Troy, T.P., Ahmed, M., Daily, J.W., Nguyen, T.L., Stanton, J.F., Ellison, G.B.: Pyrolysis of cyclopentadienone: mechanistic insights from a direct measurement of product branching ratios. J. Phys. Chem. A. 119, 7222–7234 (2015)CrossRefPubMedGoogle Scholar
- 53.Yu, T., Lin, M.C.: Kinetics of the C6H5 + O2 reaction at low temperatures. J. Am. Chem. Soc. 116, 9571–9576 (1994)Google Scholar
- 54.Tonokura, K., Norikane, T., Koshi, M., Nakano, Y., Nakamichi, S., Goto, M., Hashimoto, S., Kawasaki, M., Anderson, M.P.S., Hurley, M.D., Wallington, T.J.: Cavity ring-down study of the visible absorption Spectrum of the phenyl radical and kinetics of its reactions with cl, Br, Cl2, and O2. J. Phys. Chem. A. 106, 5908–5917 (2002)CrossRefGoogle Scholar