Chemical Reduction of Ni^II Cyclam and Characterization of Isolated Ni^I Cyclam with Cryogenic Vibrational Spectroscopy and Inert-Gas-Mediated High-Resolution Mass Spectrometry

Abstract

Ni^II cyclam (cyclam = 1,4,8,11-tetraazacyclotetradecane) is an efficient catalyst for the selective reduction of CO₂ to CO. A crucial elementary step in the proposed catalytic cycle is the coordination of CO₂ to a Ni^I cyclam intermediate. Isolation and spectroscopic characterization of this labile Ni^I species without solvent has proven to be challenging, however, and only partial IR spectra have previously been reported using multiple photon fragmentation of ions generated by gas phase electron transfer to the Ni^II cyclam dication at 300 K. Here, we report a chemical reduction method that efficiently prepares Ni^I cyclam in solution. This enables the Ni^I complex to be transferred into a cryogenic photofragmentation mass spectrometer using inert gas-mediated electrospray ionization. The vibrational spectra of the 30 K ion using both H₂ and N₂ messenger tagging over the range 800 – 4000 cm⁻¹ were then measured. The resulting spectra were analyzed with the aid of electronic structure calculations, which show strong method dependence in predicted band positions and small molecule activation. The conformational changes of the cyclam ligand induced by binding of the open shell Ni^I cation were compared with those caused by the spherical, closed-shell Li^I cation, which has a similar ionic radius. We also report the vibrational spectrum of a Ni^I cyclam complex with a strongly bound O₂ ligand. The cyclam ligand supporting this species exhibits a large conformational change compared to the complexes with weakly bound N₂ and H₂, which is likely due to significant charge transfer from Ni to the coordinated O₂.

Graphical Abstract

I. Introduction

A major challenge in chemistry is to study the structure and reactivity of intermediates formed in catalysis^1–7. These species often have short lifetimes and are too unstable to analyze using traditional techniques, such as X-ray diffraction or NMR spectroscopy. For example, the coordination and activation of CO₂ by reactive transition metal complexes is proposed to be a crucial elementary step in the catalytic conversion of CO₂ to a range of more valuable chemicals including CO and methanol^8–13. The present literature, however, rarely explores the binding of CO₂ to the highly reactive and unstable complexes that are proposed as intermediates in catalysis and typically focusses on more stable model complexes^14–17. Here, we investigate Ni complexes supported by the planar macrocyclic cyclam (cyclam = 1,4,8,11- tetraazacyclotetradecane, hereafter denoted “L1”) ligand (Figure 1). Complexes of the this type are efficient and selective electrocatalysts for the reduction of CO₂ to CO¹⁸ with the Ni^I-L1 intermediate hypothesized to play a key role in the high selectivity^19–21, but experimental information about the structure of Ni^I-L1 is limited²².

Figure 1.

Schematic structure of the trans-III isomer of Ni^I-L1.

A recently developed method for studying the activation of small molecules in compounds such as Ni^I-L1 involves isolating the singly charged molecular cations in the gas phase with open coordination sites on the metal center and then attaching substrates in a temperature-controlled ion trap^23–28. This approach follows earlier methods for the study of intermediates that employed an infrared-ultraviolet double-resonance scheme for acquiring the infrared spectra of cold ions²⁹. Although the present method was successful for studying CO₂ binding to a Ni^I complex with a specifically designed, non-catalytic ligand system^30–31, it has been difficult to efficiently prepare the key Ni^I-L1 complex in a manner compatible with cryogenic ion spectroscopy. The only reported characterization of the Ni^I-L1 cation in isolation involved a scheme that employed exothermic gas-phase electron transfer to the stable Ni^II-L1 complex^32–35 followed by buffer gas cooling to 300 K and did not enable a full IR spectrum to be measured.

Here, we describe an alternative method involving chemical reduction of the stable Ni^II-L1 compound with decamethylcobaltocene ((C₅(CH₃)₅)₂Co^II), hereafter denoted (Cp*)₂Co) to generate Ni^I-L1 cations in solution. These ions are extracted into the gas phase using electrospray ionization (ESI) and characterized at low temperature via vibrational spectroscopy. Spectra are recorded using the messenger tagging approach, where weakly bound H₂ and N₂ complexes prepared in a cryogenic (30 K) radiofrequency ion trap undergo infrared photodissociation (IRPD). The resulting band patterns in the 800 – 4000 cm⁻¹ range are compared with predictions from a variety of computational methods. Surprisingly, the computational predictions of small-molecule activation by Ni^I-L1 show strong method dependence, which indicates there are challenges associated with calculating the structure of these complexes. We compared the experimental structure of Ni^I-L1 with Li^I-L1, which contains a closed-shell metal center. This enabled us to determine the impact of the Ni d electrons on the binding of the L1 ligand and determine that cyclam binding is primarily governed by electrostatic factors. Finally, using O₂ as a substrate, we demonstrate that it is possible to bind a more reactive molecule than N₂ or H₂ to the Ni^I center by carrying out a three-body attachment of O₂ in an ion guide. The resulting complex was characterized using cryogenic vibrational spectroscopy.

II. Experimental Methods

A. Oxidation state control.

The stable Ni^II-L1 PF₆ salt was reduced with a slight excess of (Cp*)₂Co in dry and deoxygenated acetonitrile to generate Ni^I-L1. The reduction was performed in a glove box and the products were transferred to a gas-tight syringe (Hamilton Company) for transport and injection into the ion source. This method generates a solution containing the Ni^I-L1 cation that is stable for several hours under the oxygen- and water-free conditions maintained by the syringe. Samples were analyzed with the Yale hybrid photofragmentation mass spectrometer, which has been described previously²⁶. In short, the instrument combines a commercial ThermoFisher Orbitrap Velos Pro mass spectrometer (hereafter denoted “Orbitrap”) with a custom-built, double-focusing tandem time-of-flight photofragmentation mass spectrometer. Electrospray ionization of the Ni^I-L1 solution described above carried out with the Orbitrap’s ion source using N₂ sheath gas yields the mass spectrum (hereafter denoted “MS”) in Figure 2b. This is compared with the MS from the Ni^II-L1 sample solution prior to reduction with (Cp*)₂Co in panel Figure 2a. These results highlight the utility of the Orbitrap’s high resolution capability by confirming successful synthesis of the labile Ni^I intermediate before transfer to the photofragmentation spectrometer. Specifically, this capability allowed unambiguous identification of a minor deprotonated Ni^II-L1 contaminant, which is formed during generation of Ni^I-L1 (Figure 2b).

Figure 2.

High resolution mass spectrometric preparation and isolation of the Ni^I-L1 cation. a) Mass spectrum of unreduced Ni^II-L1 and b) Ni^I-L1 generated by reduction with (Cp*)₂Co in solution. Isotopic pattern simulations are color-coded to the inset chemical structures and inverted underneath the experimental data. The isotopic pattern of the reduced sample b) is in excellent agreement with the expected combined mass spectrum for Ni^I-L1 and a lower abundance of the deprotonated Ni^II-L1 complex. The Ni^II species is expected to contribute less than 5% to the peak around m/z 258 that was isolated for spectroscopic characterization. The green-shaded inset in panel b) expands the m/z axis scale to demonstrate the Orbitrap’s capability to resolve the features of Ni^I-L1 and deprotonated Ni^II-L1, which are separated by less than 0.05 m/z.

III. Results and Discussion

A. Mass isolation of Ni^I complexes and spectroscopic verification of product oxidation state

To verify that the present solution-phase preparation yields the same structures as the gas-phase approaches employed in our previous work^30–31, we applied the synthesis described above to another Ni^I complex whose IRPD spectrum we earlier acquired following gas-phase synthesis: Ni^I-L-N₄Me₂ (L-N₄Me₂ = N,N’-dimethyl-2,11-diaza[3,3](2,6)pyridinophane, hereafter denoted “L2”). Those experiments generated the coordinatively unsaturated Ni^I-L2 center by collisional dissociation of a stable Ni^I precursor ion in the gas phase^30–31. The solution chemistry applied to the generation of Ni^I-L1 indeed provides an alternative route to this Ni^I-L2 species: comparison of Ni^I-L2 infrared spectra taken as part of the present study and those reported previously (Figure S1) confirms that both solution- and gas-phase methods produce the same forms of the cation. We note that isolating the desired Ni^I-L2 (m/z = 326.104) cation from this reaction mixture with ESI is challenging because the spray yields an intense peak at m/z = 329 due to the oxidized reducing agent, [(Cp*)₂Co^III]⁺ (329.168 m/z), which saturates the ion trap required for the spectroscopy experiments, thus suppressing the yield of the target Ni^I-L2 cation. This problem was overcome using the Orbitrap’s mass isolation function to select m/z = 326 prior to injection into the photofragmentation mass spectrometer, as illustrated by the mass spectra before and after isolation in Figure S2.

B. Vibrational spectrum of Ni^I-L1 at 30 K using H₂ tagging

The ability to generate robust yields of the Ni-L1 in the critical Ni^I oxidation state enables the acquisition of its vibrational spectrum in a linear action regime using cryogenic IR photodissociation of complexes with weakly bound “mass tags”. This represents a significant advance because the previous spectroscopic study of this system³² was limited to the 800–1600 cm⁻¹ range, and was obtained using 300 K infrared multiphoton dissociation (IRMPD) of the bare ion. Here, we report the spectrum of the Ni^I-L1-2H₂ cluster ion, where the H₂ tags are expected to bind to the Ni^I metal center, over the range 800–4100 cm⁻¹ (Figures 3a and 4b). This is compared with the predicted harmonic spectrum (Figure 3b) and previously acquired IRMPD spectrum (Figure 3c) of bare Ni^I-L1, whose minimum energy structure is displayed in the Figure 3 inset. The doubly-tagged (2H₂) complex was chosen to avoid mass overlap between the ⁶⁰Ni^I-L1 isotopologue of the bare cation and ⁵⁸Ni^I-L1-H₂. The H₂ molecules were observed to attach at approximately 40 K, which is typical for operation in the tagging regime where the spectra of the H₂ complexes are observed to be very close to those of the isolated, cold parent ions³⁶. Further evidence of the weak interaction between Ni^I-L1 and H₂ is the fact that the H₂ stretching fundamental is weak and observed at 4137 cm⁻¹ (Figure 4b, blue line), very close to the value (4161 cm⁻¹, grey line in Figure 4) in isolated H₂³⁷. The H₂ molecule is predicted, at the M06-2X/SDD level of theory^38–41, to attach by primarily electrostatic interaction with the metal center in an “end-on” arrangement as indicated in the structure displayed in Figure S3. Activation of the IR-forbidden diatomic stretching transitions is attributable to polarization of, and charge transfer to, the diatomic tag by the molecular ion. This behavior and the expected relationship between transition strength and red shift (transition strength and red-shift increase together) has been treated in previous work and agrees with the trends observed here^42–47. Red shifts of the H₂ stretching transitions observed in this work are summarized in Figures S5 and S6.

Figure 3.

Comparison between the vibrational spectra of the Ni^I-L1 cation obtained by a) vibrational predissociation of the Ni^I-L1–2H₂ cluster at 30 K and c) 300 K IRMPD of the bare ion reproduced from an earlier report³². The calculated spectrum for the bare Ni^I-L1 cation (M06–2X/SDD) is displayed in b) along with qualitative descriptions of the modes involved in the various fundamentals. The corresponding minimum-energy structure is shown as an inset. The asterisk marks a feature which we do not assign. The observed tag-dependent multiplet structure in the region expected for the NH stretching modes (see Figure S4) is not predicted at the harmonic level.

Figure 4.

CIVP spectra of a) Ni^I-L1-N₂ cation, b) Ni^I-L1–2H₂ cation, c) Li^I-L1-N₂ cation, d) Li^I-L1-H₂ cation, and e) Ni^I-L1-O₂N₂ cation. The structure of Ni^I-L1-O₂ predicted at the M06–2X/SDD level of theory is included in f).

The spectrum of the Ni^I-L1-2H₂ ion at 30 K in Figure 3a generally displays features in the same locations as those observed with IRMPD of room temperature ions (Figure 3c). As expected, the relative intensities are somewhat different – and the features better resolved – in the linear spectrum of the cold ion. However, this technique also introduces the possibility that the tag molecules perturb the intrinsic structure and vibrational band pattern of the isolated Ni^I-L1 cation. To address this issue, we noted that N₂ attachment to form the Ni^I-L1-N₂ cluster ion occurred at a similarly low temperature, suggesting that N₂ is also very weakly bound^46–47 to the molecular cation. The observed IRPD spectra of the H₂- and N₂-tagged ions are compared in Figure 4. These spectra are essentially identical with the exception of tag-dependent multiplet structure (labeled *) in the NH stretching region (Figure S4), which is evident in the features near 3300 cm⁻¹ in Figure 4a and 4b. This overall similarity, combined with the fact that the N₂ stretch is extremely weak and occurs at 2328 cm⁻¹, very close to the 2330 cm⁻¹ value of the N-N stretching vibration⁴⁸ in free N₂ (red vertical line in Figure 4), lends strong support to the conclusion that the tag molecules do not significantly affect the band pattern, and hence structure, of the isolated ion.

The greatest advantage of the tagging method for acquiring vibrational spectra is that the band patterns can be directly compared with (scaled) harmonic predictions of the absorption spectra predicted for candidate minimum energy structures. In the case of Ni^I-L1, there are two dominant isomeric forms²⁰ that differ according to the orientation of the NH groups (the so-called trans-I and trans-III isomers, shown in Figure S7). The spectrum calculated (M06-2X/SDD) for the trans-III isomer, which is predicted to lie ~3.5 kJ/mol lower in energy than the trans-I form, is presented in Figure 3b and is in excellent agreement with the observed pattern. A comparison of the observed spectrum with those calculated for the trans-I and trans-III isomers is presented in Figure S7. The agreement in the lower energy region near 1000 cm⁻¹ for the trans-III isomer is particularly compelling. Specifically, the strong narrow peak corresponding to the ring deformation in the trans-I spectrum is predicted to degrade into a plethora of bands in the trans-III spectrum. We note that the trans-III form was identified as the isomer generated in the gas phase in the earlier report³². Finally, it is valuable to acknowledge how strongly the calculated spectra depend on the choice of DFT functionals and basis sets. Figure S8 illustrates the surprisingly large variations in the harmonic spectra of Ni^I-L1-N₂ calculated using 28 different combinations of functionals and basis sets. Remarkably, the harmonic predictions for N₂ stretching frequencies span a range of ~1200 cm⁻¹. This work does not seek to identify the cause of this variation, but rather to report the linear experimental spectrum and thus provide an experimental benchmark to guide future theoretical efforts. It is useful to note, however, that the large variation only occurs in the frequency of the N₂ stretch; the general pattern of bands from scaffold are largely consistent across the methods surveyed.

C. Investigation of metal-ion-dependent scaffold deformation

The large variation in theoretical predictions for the interactions between Ni^I-L1 and N₂ highlights the challenges for widely accessible electronic structure theory to accurately predict the behavior of the open shell (nominally d⁹) Ni^I ion. To empirically explore the extent to which the response of the cyclam ligand is primarily electrostatic in character, we recorded the vibrational spectra of the Li^I analogues Li^I-L1-H₂, and Li^I-L1-N₂ (Figures 5a, 4c and 4d), which are compared to the Ni^I-L1-2H₂ and Ni^I-L1-N₂ spectra in Figure 4a–d, respectively. The bands associated with the complexes tagged by either H₂ and N₂ are once again essentially identical: the positions of all bands remain within the ~3 cm⁻¹ bandwidth of our laser on moving from the H₂ to N₂ tags. A survey of predicted Li^I-L1-N₂ spectra (Figure S9) using the same methods applied to the Ni^I-L1-N₂ complex (Figure S8) shows that the calculated N₂ stretching transitions are much more consistent in intensity and somewhat more accurate in terms of frequency, spanning a ~700 cm⁻¹ range. The experimental spectrum for Li^I-L1-H₂ is compared with the harmonic prediction in Figure 5 to illustrate the performance of the M06-2X/SDD level of theory. The predicted bands in the CH region are very accurate, and the CH/NH bending and ring deformation features in the fingerprint region are also satisfactory. Weaker bands in the CH rock/NH bend region involving collective motions are in worse agreement, however, as are the NH stretches near 3300 cm⁻¹. Thus, the poor agreement with the NH stretching bands mentioned above cannot be solely attributed to complications arising from the open shell d⁹ configuration of the Ni^I cation.

Figure 5.

Comparison of the a) experimental and b) predicted (M06–2X/SDD, scaled by 0.95) vibrational spectra of Li^I-L1-H₂. The predicted minimum energy structure is shown in c).

The most important result from this study is that we have established a robust method to generate catalytically relevant Ni^I-L1. In principle, this enables the investigation of the reactivity of Ni^I-L1 with small molecules, which has previously been challenging. As a demonstration of this capability, we attached O₂ to Ni^I-L1 by introducing O₂ into an ion guide that transfers the cation to the cryogenic ion trap at ambient temperature. The fact that O₂ binds at such high temperature confirms that the O₂ is very strongly bound to the metal center. Since the resulting Ni^I-L1-O₂ complex is too tightly bound to allow the collection of photofragmentation spectra below ~2500 cm⁻¹ in a linear regime, the O₂ complex was tagged with N₂ in the cryogenic ion trap at 40 K. The vibrational spectrum of Ni^I-L1-O₂-N₂ ion is displayed in Figure 4e. As evidenced by the dramatic changes in the fingerprint-region band pattern relative to those of the isolated ion Figure 4a–d, attachment of O₂ indeed induces substantial rearrangement of the ligand. The calculated structure of the O₂ complex is displayed in Figure 4f and features substantial charge transfer to the O₂. These large changes in the ligand bands are easily rationalized by considering the substantial charge redistribution between the Ni^I-L1 complex and the O₂ substrate. Calculations at the same level of theory applied to the bare cation indicate about 0.5 e⁻ is transferred from Ni to the O₂ moiety. The band associated with the O₂ stretch, which is predicted to be weak and fall in the congested fingerprint region at around 1100 cm⁻¹, was not readily identified by inspection⁴⁹.

Unfortunately, our attempts to isolate CO₂ bound to Ni^I-L1 using this approach were not successful, indicating that this interaction is weak compared to the behavior of the nominally similar Ni^I-L2 system that readily attaches CO₂ at elevated temperature³⁰. This difference raises the issue of how the local coordination environment around the Ni^I ion drives chemical interactions with small molecules. This aspect of Ni^I coordination chemistry can be addressed by introducing a solvent molecule to one of the two open sites on the metal center in the planar cyclam scaffold before condensing CO₂ onto the remaining open site. Modifications to the instrument are currently underway to explore how substrate activation responds to such manipulations of the coordination environment around the Ni^I metal center.

IV. Conclusion

We have demonstrated an efficient method to isolate Ni^I-cyclam with the metal ion in the labile Ni^I oxidation state. This approach enables the characterization of Ni^I-cyclam with high-resolution mass spectrometry and cryogenic vibrational spectroscopy. The Ni^I cyclam cation was generated by chemical reduction of the stable Ni^II cyclam dication in solution, then isolated by extraction into a photofragmentation mass spectrometer. Isolated Ni^I-cyclam displayed relatively inert chemical behavior toward N₂ and H₂ and appears similar in terms of both structure and chemical activity to Li^I-cyclam, which was chosen to preserve the charge and size of the central ion, while eliminating contributions from d-orbitals. The isolated Ni^I-cyclam cation binds O₂ very strongly, and its vibrational spectrum reveals substantial distortion of the cyclam ligand, consistent with partial charge transfer to the O₂ substrate. The methods reported here provide a useful protocol for investigating the intrinsic interactions of this system with catalytically relevant substrates, and further work is ongoing to explore the role of solvent coordination in the activation of small molecule substrates, such as CO₂.