Boronate Covalent and Hybrid Organic Frameworks Featuring P^III and P=O Lewis Base Sites

Abstract

Two covalent organic frameworks comprising Lewis basic P^III centers and Lewis acidic boron atoms were prepared by poly‐condensation reactions of newly obtained tris(4‐diisopropoxyborylphenyl)phosphine with 2,3,6,7,10,11‐hexahydroxytriphenylene and 2,3,6,7‐tetrahydroxy‐9,10‐dimethylanthracene. Obtained materials exhibit significant sorption of dihydrogen (100 cm³ g⁻¹ at 1 bar at 77 K), methane (20 cm³ g⁻¹ at 1 bar at 273 K) and carbon dioxide (50 cm³ g⁻¹ at 1 bar at 273 K). They were exploited as solid‐state ligands for coordination of Pd⁰ centers. Alternatively, in a bottom‐up approach, boronated phosphine was treated with Pd₂dba₃ and poly‐condensated, yielding hybrid materials where the polymer networks are formed by means of covalent boronate linkages and coordination P−Pd bonds. In addition, the analogous materials based on phosphine oxide were synthesized. The DFT calculations on framework–guest interactions revealed that the behavior of adjacent boron and phosphorus/phosphine oxide centers is reminiscent of that found in Frustrated Lewis Pairs and may improve sorption of selected molecules.

The first boron‐phosphine covalent organic framework (BP‐COF) is reported, which can be considered as a new group of functional porous materials joining characteristics of frustrated Lewis pairs (FLPs) with enhanced sorption selectivity and solid‐state ligands for the coordination of transition‐metal atoms.

Introduction

There is a continuous interest in covalent organic frameworks (COFs)—a class of porous organic materials composed of light elements such as carbon, boron, oxygen, nitrogen, silicon, and sulfur connected by strong covalent bonds.1, 2, 3, 4, 5, 6 Since the first report by Yaghi and co‐workers in 2005,7 numerous COF architectures were designed. From the structural topology point of view they can be divided in two major groups. Two‐dimensional (2D) COFs are obtained from building blocks possessing planar structures.8 In contrast, three‐dimensional (3D) COFs are formed starting with precursors featuring a general 3D (e.g., tetrahedral) topology which imposes a structure of a resulting polymer network. Within this group, classical examples include highly porous materials such as COF‐102 and COF‐103 based on tetraboronic acids C[p‐C₆H₄B(OH)₂]₄ and Si[p‐C₆H₄B(OH)₂]₄, respectively.9

The use of organophosphine linkers for the preparation of various multifunctional porous coordination polymers dates back to 2008. Since then, a number of materials were designed, especially by Humphrey and co‐workers (phosphine coordination materials, PCMs).10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 It should be noted they generally comprise carboxylic groups coordinating metal secondary building units (SBUs). The pseudo‐tetrahedral geometry around phosphorus centers favors the formation of 3D frameworks. The presence of lone electron pair at the P^III centers provides a possibility for pre‐ or post‐synthetic functionalization through alkylation,16, 26 arylation27 or chalcogenation.28 Such materials are also considered as solid‐state ligands (SSLs) for coordination of transition metals.12, 29, 30, 31, 32, 33, 34, 35 Regarding the latter ability, the most common strategy involves the formation of robust coordination polymers with metals (Ca, Zr, Sc) weakly coordinated to phosphorus and hard‐ligand donors (carboxylates, imidazolates, alkoxylates). In such porous networks, the softer phosphorus function allows the facile coordination of precious metal atoms including Pd⁰, Rh^I, Ir^I, Au^I, Ag^I, as well as low‐valent Cu^I, Co^II metals. Importantly, their catalytic performance is usually retained from corresponding single‐molecule phosphine‐based catalysts. Thus, such materials offer a very attractive way to transfer catalytic processes from homogenous to heterogeneous environments. This concept can also be used to achieve stronger and more selective binding of small‐molecule substances, which can further be exploited for storage and separation processes. In addition to the P^III materials, the analogous MOFs constituting phosphine oxides functions are obtained and used as a solid‐state platforms for the lanthanide‐metal coordination (La^III, Dy^III).17, 19

Although the chemistry of PCM‐MOFs is already well developed, the analogous COFs containing phosphine or phosphine oxide fragments have not been considered so far. Limited progress in this field was due to the lack of procedures allowing for the facile synthesis of phosphine‐based COF precursors. In particular, boronated COFs are especially interesting due to the presence of weakly acidic boron centers along with stronger Lewis base phosphorus sites. In our efforts for the synthesis of functional porous materials36, 37 we have developed a new approach to synthesize boron‐phosphine COFs (BP‐COFs) (Scheme 1). Their non‐planar tripodal topology is unique for all COF precursors. Thus, the prepared materials cannot be regarded as typical 3D COFs, which are based on nodes featuring the 43m point‐group symmetry. The presence of Lewis base P^III or P=O oxygen atoms and weakly acidic boron centers may significantly enhance the sorption properties of such materials. Furthermore, they can serve as a robust platform for the coordination of transition metals. Hence, in the current contribution we present pre‐ and post‐synthetic modifications of COFs with palladium(0) using labile Pd₂dba₃ complex as a metal source.

Scheme 1.

New class of COF materials based on boronated phosphines.

Results and Discussion

Tris(4‐bromophenyl)phosphine 1, a starting material for the preparation of 2, was obtained on a large scale (0.5 mol of the starting 1,4‐dibromobenzene) from the metathesis of PCl₃ with 4‐bromophenyllithium according to the published procedure.38 It was subjected to triple bromine–lithium exchange followed by a boronation with trialkyl borate reagent. After careful optimization, the approach based on the in situ addition of tBuLi (6 equiv) to the mixture of 1 and B(OiPr)₃ in THF at −78 °C was found to give the best results (Figure 1 a). In our first attempts the reaction mixture was hydrolyzed in order to isolate tris(4‐dihydroxyborylphenyl)phosphine 2 a as a precursor of COFs. However, ¹H and ³¹P NMR analyses revealed that 2 a exists in equilibrium with its hydrated zwitterionic form (2 a’) featuring protonated phosphorus atom and the anionic boronate group (Scheme S1, Supporting Information). Furthermore, we have observed its oxidation under air to corresponding phosphine oxide (4 a), which is accompanied by a slow partial deboronation. Therefore, we decided to isolate compound 2 by treating its tris(boronic) ate complex precursor with Me₃SiCl followed by removal of volatiles and extraction with heptane. Compound 2 was obtained in a pure form as colorless crystals in multigram quantities and reasonable yield (50 %). The structure of 2 was confirmed by multinuclear NMR spectroscopy and X‐ray diffraction (Figure 1 b). Notably, the crystal structure lacks intermolecular interactions between Lewis acidic boron atoms and Lewis base phosphorus centers.

Figure 1.

(a) Synthesis of triboronated phosphine 2 and phosphine oxide 4. (b) The molecular structure of 2 (ellipsoids drawn at the 50 % probability level, H‐atoms and disordered sites omitted for clarity). LEP=Lone Electron Pair.

The synthesis of boronated triphenylphosphine oxide precursor 4 involved the in situ Br/Li exchange/boronation of brominated precursor 3 39 using the protocol analogous to that described for 2. Compound 3 was isolated as a yellow solid and characterized by ¹H, ¹¹B and ³¹P NMR spectroscopy. Unlike 2, it is only sparingly soluble in CDCl₃, presumably due to extensive aggregation in solid state through the P=O→B dative bonds.

As shown in Figure 2 a the syntheses of materials BP1–2 were performed by polycondensation reactions of 2 with 2,3,6,7,10,11‐hexahydroxytriphenylene (HHTP, 1 equiv), or 2,3,6,7‐tetrahydroxy‐9,10‐dimethylanthracene (THDMA, 1.5 equiv), respectively. A general protocol similar to those developed for the preparation of other boronate ester COFs was employed.1, 2, 6, 7, 8, 9 Thus, a stoichiometric mixture of 2 and a polyol was stirred in a 1:1 (v/v) mesitylene −1,4‐dioxane mixture at 85 °C followed by repeated washing of a crude product with anhydrous THF and final drying under high vacuum at 150 °C. The prepared materials are grayish‐green (BP1) or olive‐green (BP2) powders. Their hydrolytic degradation in wet [D₆]DMSO (with added D₂O) and ¹H NMR analyses of resulting samples showed that their composition was in agreement with the theoretical stoichiometry. Since the precursor 2 is not the boronic acid but the respective ester, its poly‐condensation reactions with HHTP and THDMA can be classified as transesterification‐based processes. We suggest they should proceed more easily and faster than dehydrative poly‐condensation reactions which occur during formation of COFs from free boronic acids and are typically conducted solvothermally although some examples involving mild conditions were reported.40 Thus, the synthesis of BP1–2 was repeated at room temperature and both materials were obtained in good yields. Furthermore, their composition and appearance is very similar to the samples obtained with the heating applied. On the other hand, high reaction rates favor amorphization. Following the same protocol the synthesis of phosphine oxide‐based materials (BPO1–2) was performed by poly‐condensation reactions of 4 with HHTP (1 equiv) or THDMA (1.5 equiv), respectively.

Figure 2.

(a) Synthesis of BP1–2 and BPO1–2 porous materials. (b) SEM images of BP1–2. (c) ¹¹B and ³¹P MAS NMR spectra of BP1 and BPO1 recorded at the spinning rate of 10 kHz. Asterisks (*) indicate peaks arising from spinning side bands.

In the next step, hybrid materials containing Pd⁰ metal centers were prepared utilizing two general strategies: (a) bottom‐up approach involving Pd⁰ coordination to the precursor 1, followed by polycondensation with HHTP/THDMA (BP1‐Pd(BU), BP2‐Pd(BU)), (b) post‐synthetic modification of previously obtained BP1–2 COFs (BP1‐Pd(PS), BP2‐Pd(PS)). It should be noted that the modification of organoboron polymers through metal complexation is still rather unexplored.41 The synthesis of BP1‐Pd(BU) and BP2‐Pd(BU) was performed in two steps as shown on Figure 3 a. At first, compound 2 was treated with Pd₂(dba)₃ at low temperature (ca. −50 °C) in DCM. The amounts of starting materials were taken to achieve the Pd:P ratio of 1:4.

Figure 3.

(a) Synthesis of Pd‐functionalized COF materials using bottom‐up approach. (b) ¹¹B and ³¹P MAS NMR spectra of BP1‐Pd(BU) recorded at the spinning rate of 10 kHz. Asterisks (*) indicate peaks arising from spinning side bands.

Initially, the reaction mixture had a purple color due to the presence of dissolved Pd₂(dba)₃. However, a solution gradually turned olive‐brown on warming to ambient temperature indicating the progress of ligand exchange at the Pd atom and formation of a species of assumed stoichiometry Pd[2]₄ although it may equilibrate with Pd[2]₂ and/or Pd[2]₃ as well as free ligand 2. The ³¹P NMR spectrum of this species showed a few resonances in the range of 24–26 ppm which confirms the formation of Pd‐P dative bonds.42 In the second step, the obtained complex was treated with a THF solution of a polyhydroxy linker, i.e., HHTP or THDMA, in a stoichiometric proportion with respect to 2. The gradual precipitation of a voluminous brownish or olive‐green precipitate was observed during several hours. The slurry was stirred for 3 days at room temperature and filtered under argon followed by thorough washing of a solid with several portions of anhydrous DCM. Drying under high vacuum at 80 °C resulted in final materials as free‐flowing dark‐colored powders. Analysis of a yellow filtrate revealed the presence of free ligand (dba) in amounts slightly less (ca. 15 %) than those initially added with the used Pd precursor. It was not significantly contaminated by other compounds including boronated triphenylphosphine derivatives and polyhydroxy linkers, which indicates their quantitative incorporation in the bulk materials.

In the second approach, BP1–2 materials were immersed in Pd₂dba₃/THF solution during 3 days. We assumed that the labile character of the complex should result in the attachment of Pd⁰ to P centers43, 44 on the surface of pores having size sufficient for penetration with Pd₂dba₃. In both cases, the purple solution containing the unreacted complex was removed and the resulting materials were washed thoroughly with DCM and dried under vacuum (10⁻³ torr, 50 °C) to give the final products BP1‐Pd(PS) and BP1‐Pd(PS) as dark powders.

The thermal stability of BP1–2 (dried at 150 °C prior to analysis) was investigated under N₂ atmosphere by TGA technique (Figure S1). A minor mass loss (up to 5 %) is observed up to 350 °C. It can be associated with the removal of adsorbed gases, and some volatile impurities (e.g., traces of solvents or by‐products resulting from completion of condensation reactions). A gradual decomposition starts as the total mass loss at 600 °C reaches ca. 40 % for both COF materials. To summarize, obtained boronate ester COFs exhibits comparably high thermal resistance consistent with the exclusive presence of strong covalent bonds. The presence of Pd nodes did not affect the thermal stability of BP2‐Pd(BU) as the decomposition started also at ca. 350–400 °C (Figure S2). In turn, BP1‐Pd(BU) material losses about 15 % of its mass when heating up to 200 °C, which is probably associated with the presence of significant amount of solvent and dba molecules resided in the porous framework.

¹¹B MAS NMR spectrum of BP1 (Figure 2 c) shows the resonance at ca. 24 ppm, i.e., in the range characteristic for trigonal boron atoms in arylboronic acids and esters.45 The ³¹P MAS NMR spectrum of BP1 shows a sharp resonance at ca. 0 ppm, i.e., shifted downfield by ca. 6 ppm with respect to PPh₃.46 Very similar ¹¹B and ³¹P MAS NMR spectra were recorded for BP2 (see Supporting Information, Figures S23, S24). Based on these results one can conclude that the plausible structure stabilization of BP1–2 due to formation of strong dative interactions between boron and phosphorus atoms should be excluded. However, a slight deshielding (relative to PPh₃) of ³¹P MAS NMR signals for both materials might indicate occurrence of weak communication between boron and phosphorus centers.47 The ¹¹B MAS NMR spectrum of BPO1 (Figure 2 c) shows the resonances at 8.1, 13.7, 23.1 ppm. The latter can be assigned to the trigonal boron atom featuring CBO₂ environment whereas two remaining ones may indicate that some boron atoms are coordinated with oxygen atoms of P=O groups and/or THF guest molecules. The ³¹P MAS NMR spectrum of BPO1 is characterized by persistence of intense spinning sidebands indicating that the chemical shift anisotropy frequency range is larger than the MAS rate. The spectrum has one isotropic chemical shift at 34.6 ppm consistent with the presence of triphenylphosphine oxide motifs. Very similar ¹¹B and ³¹P MAS NMR spectra were recorded for BPO2 (Figures S25, S26), thus confirming that the change of the linker topology does not significantly influence the ³¹P NMR chemical shift anisotropy characteristics of the studied nuclei. ¹¹B MAS NMR spectra of the hybrid BP1‐Pd(BU) and BP2‐Pd(BU) materials (Figure 3 b) are in general analogous to those recorded for BPO1–2 as they point to dative interactions between some boron atoms and oxygen‐based donors including remaining THF or dba molecules. The ³¹P MAS NMR spectrum of BP1‐Pd(BU) features a set of spinning sidebands with isotropic chemical shifts at −2.1 and 34.5 ppm (similar values of −0.7 and 36.3 ppm were recorded for BP2‐Pd(BU)). The former one can be assigned to free triphenylphosphine moieties which indicates that P^III nodes characteristic for BP1 are also present in BP1‐Pd(BU) to a significant extent. However, the latter more intense (ca. two‐fold) signal can be assigned to triarylphosphine P atoms engaged in Pd−P dative bonds.48 It should be noted that the presence of P−Pd coordination is also supported by XPS data which are discussed below.

Scanning electron microscopy (SEM) indicates mesoporous morphology of BP1–2 (Figure 2 b). This material forms sponge‐like agglomerates of irregular shape and dimensions in the range of 0.5–20 μm. The size of particles (presumably nanocrystallites) forming these structures are very small and ranges from 50–200 nm. The impregnation of such a COF structure with palladium centers does not change the general morphology of the structure (Figure S41).

XPS analysis was employed to evaluate the amount of Pd centers adsorbed on the surfaces of COF materials obtained using bottom‐up and post‐synthetic methodologies. The survey of XPS spectra is shown in Figures S29–S32 in Supporting Information. Pd 3d and P 2p high‐resolution XPS spectra of BP1‐Pd(BU) and BP1‐Pd(PS) COFs are shown in Figure 4, while the C 1s, B 1s, and O 1s HR‐XPS spectra regions and analogous spectra for BP2‐derived hybrid materials are placed in the SI (Figures S33–S36).49 As indicated by the analysis of Pd 3d region, two different Pd species can be distinguished. The doublet at binding energies (B.E) of 341.5 eV (Pd 3d_3/2) and 336.3 eV (Pd 3d_5/2) can be assigned to bulk palladium nanoparticles—(Pd(bulk), Table 1).50 The bigger doublet is shifted towards higher binding energies (B.E=343.2 eV, 337.9 eV) and is associated with palladium centers coordinated to phosphorus atoms (Ar₃P‐Pd).51 The appearance of metallic palladium on the COF surface results from the relatively low stability of Pd₂(dba)₃ precursor and may indicate that its decomposition followed by formation of Pd clusters is catalyzed by the COF surface. The deconvolution of XPS curves clearly shows that total amount of Pd atoms adsorbed in the BP1‐Pd(BU) and BP2‐Pd(BU) materials is much lower with respect to post‐synthetically impregnated COFs, which is consistent with stoichiometry of P/Pd centers in precursors. Furthermore, the Pd‐P/Pd(bulk) molar ratio is much lower for latter systems, specifically in case of BP2‐Pd(PS) Pd bulk atoms outnumber those bound to P atoms. The deconvolution of P 2p core level spectra for BP1‐Pd(BU) gives two doublets representing two set of P 2p_1/2 and P 2p_3/2 peaks. The peaks at 131.0 eV and 131.9 eV correspond to P atoms bound to three aromatic rings of the COF framework (Ar₃ P). The adjacent peaks at higher binding energies (B.E=133.0 eV and 133.8 eV) can be attributed to the P centers coordinating palladium atoms (Ar₃ P‐Pd). The comparison of Pd‐containing materials obtained using either bottom‐up or post‐synthetic approaches shows that proportion of free and bound P atoms is different in both material classes. In the former materials, the molar ratio of palladated vs. free phosphine (Ar₃ P‐Pd/Ar₃ P) is close to 1.5 indicating that more than half of P atoms is bound to Pd, while in the case of post‐synthetically modified COFs, this ratio is close to 1. Finally, as suggested by the molar ratio of coordinated P and Pd sites (Ar₃ P‐Pd/Pd‐P), in the case of BP1‐Pd(BU) and BP2‐Pd(BU) each Pd center is bound to 3 or 4 P atoms. Thus, these COFs preserve the stoichiometry of its precursor Pd[2]₄ and can be treated as mixed‐hybrid networks composed of covalently linked organoboron network interconnected by tetra‐coordinated palladium centers. In turn, for BP1‐Pd(PS) and BP2‐Pd(PS) the molar ratio Ar₃ P‐Pd/Ar₃ P is close to 1, which indicates that the structure of these materials is mostly preserved from initial BP1–2 COFs. Overall, it can be concluded that the post‐synthetically Pd‐functionalized COFs are more preferred for catalytic applications, where the density of the catalytic centers plays the major role, while the palladium centers would be more dispersed and electronically active in materials obtained from a bottom‐up approach.

Figure 4.

High‐resolution XPS spectra for BP1‐Pd(BU) (a, c) and BP1‐Pd(PS) (b, d): Pd 3d (a, b) and P 2p (c, d).

Table 1.

XPS‐derived phosphorus and palladium atom amounts located on the surface of corresponding COF materials. Pd(bulk) and Pd‐P denote bulk and bonded palladium atoms, Ar₃ P and Ar₃ P‐Pd correspond to three‐coordinate and Pd‐bonded P atoms, respectively. Values are provided in atomic percent. The respective molar ratios are given below.

	BP1‐Pd(BU)	BP2‐Pd(BU)	BP1‐Pd(PS)	BP2‐Pd(PS)
Pd(bulk)	0.05	0.08	0.23	0.75
Pd‐P	0.32	0.25	0.62	0.63
Ar3 P	0.64	0.66	0.72	0.83
Ar3 P‐Pd	1.03	0.98	0.75	0.71
Pd‐P/Pd(bulk)	6.40	3.13	2.70	0.84
Ar3 P‐Pd/Ar3 P	1.61	1.48	1.04	0.85
Ar3 P‐Pd/Pd‐P	3.22	3.9	1.21	1.13

The porosity of all obtained materials was initially evaluated using N₂ gas adsorption at 77 K. Prior to the measurements, samples were activated by heating at 150 °C under high vacuum (10⁻³ torr) for 24 hours to remove any possible guest molecules. All recorded isotherms show a sharp increase of N₂ uptake at low relative pressures (below 0.02 P/P ₀) which is common for microporous materials (Figure 5 a, b) including classical boron COFs. All studied materials exhibit similar type‐II sorption isotherms with relatively slow and almost constant increase of sorption in the 0.05–0.8 P/P ₀ pressure range. BP1 shows the highest N₂ uptake reaching ca. 650 cm³ g⁻¹ STP. The N₂ sorption is significantly decreased (to ca. 200 cm³ g⁻¹ STP at P=P ₀) for the BP1‐Pd(BU) hybrid material. On the other hand, both materials based on THDMA have similar N₂ sorption properties (up to ca. 300 cm³ g⁻¹ at P=P ₀) which suggests that the porosity is not significantly affected by introduction of Pd coordination centers. In addition, desorption measurements revealed that the isotherms are almost reversible for these two materials. For BP1 and BP1‐Pd(BU), a more distinctive hysteresis loop was observed which may indicate the occurrence of capillary condensation within mesopores, due to the strength of adsorbate‐adsorbent and adsorbate‐adsorbate interactions.52

Figure 5.

(a, b) N₂ sorption isotherms (@77 K) for studied porous materials. (c) H₂ (@77 K), (d) CO₂ and CH₄ (@273 K) sorption isotherms for BP1–2.

The Brunauer–Emmett–Teller model was used to calculate the relevant parameters including apparent surface area S _BET based on the Rouquerol's consistency criteria,53 and the pore volume V _p at P/P ₀=0.98 (Table 2). The S _BET value for BP1 was significantly higher (669 m² g⁻¹) than the remaining materials (72–532 m² g⁻¹). Obtained results are comparable with those reported for related boron 2D COFs such as COF‐1, COF‐5,6 and BTP‐COF.54

Table 2.

Summary of surface areas, S _BET, gas uptakes and total pore volumes, V _p for all studied COF materials.

	BP1 [a]	BP2 [a]	BPO1	BPO2	BP1‐Pd(PS)	BP2‐Pd(PS)	BP1‐Pd(BU)	BP2‐Pd(BU)
S BET [m2 g−1]	669(703)	532(480)	235	245	485	71	310	383
V p [cm3 g−1]	0.95(0.48)	0.56(0.55)	0.17	0.22	0.76	0.30	0.34	0.49
N2 @77 K, P/P 0=0.98	615.1(310.6)	362.0(358.1)	112.2	145.4	489.2	184.2	220.0	316.5
H2@77 K, P=850 torr	98.7(111.5)	88.5(68.6)	64.3	77.9	86.3	57.5	68.6	77.9
CO2@273 K	48.7(54.7)	38.7(33.5)	32.0	33.2	36.9	21.6	32.6	33.7
CH4@273 K	15.0(16.7)	12.1(9.8)	9.8	10.5	14.7	7.8	10.1	10.9

[a] data for materials prepared at room temperature are given in parentheses.

We have also studied H₂, CO₂ and CH₄ sorption for BP1–2 (Figure 5 c,d) as we hypothesized that the presence of Lewis acidic (B atoms) and Lewis basic (P atom) centers may be beneficial for selectivity of obtained materials due to synergistic effects. Specifically, highly inhomogeneous electric field could enhance interactions with molecules featuring polar bonds such as CO₂, but may also induce bond polarization in homoatomic molecules such H₂. Both materials exhibit type‐I sorption of H₂@77 K which is not saturated at P=P ₀. The H₂ uptake for BP1 is only slightly higher than for BP2 (98.5 vs. 88.5 cm³ g⁻¹ STP, P=850 torr) which implies a much higher H₂ versus N₂ sorption selectivity in favor of the latter material (Table 2). In case of BP1‐Pd(PS) and BP2‐Pd(PS) the N₂ uptake was decreased. This effect was stronger for the latter material (485 cm³ g⁻¹ STP for BP1‐Pd(PS) vs. 71 cm³ g⁻¹ STP for BP2‐Pd(PS)). The sorption of H₂ was also lowered (86.3 and 57.5 cm³ g⁻¹ STP for BP1‐Pd(PS) and BP2‐Pd(PS), respectively, P=850 torr) but to a lesser extent. This suggests the microporous structure was not strongly deteriorated by an introduction of Pd⁰ centers on the surface of original materials BP1–2.

Nitrogen sorption of hybrid materials BP1‐Pd(BU) and BP2‐Pd(BU) was much lower (ca. two‐fold) compared to their non‐palladium counterparts BP1–2 (Table 2). However, H₂ uptakes for BP2 and BP2‐Pd(BU) are in fact comparable (88.5 vs. 77.9 cm³ g⁻¹ STP, respectively, P=850 torr) which shows that introduction of Pd centers seems to be beneficial in terms of H₂/N₂ sorption selectivity.

Investigation of the sorption properties of BPO1‐2 revealed that their N₂ uptake at 77 K amounts to ca. 100 cm³ g⁻¹ STP at P/P ₀=1, i.e., it was significantly lower compared to their analogues with P^III centers. The sorption of H₂@77 K was also lower relative to BP1–2 but the difference was much smaller by the factors of ca. 1.5 and 1.1 for BPO1–2, respectively (Figure S51). This is consistent with an increased H₂/N₂ uptake ratio for BPO1–2 relative to BP1–2. This can be attributed to structural peculiarity due to the presence (and perhaps cooperation) of donor (P=O motif) and acceptor (B atom) sites.

The sorption of CO₂ and CH₄ was investigated at 273 and 293 K. For BP1–2, the sorption of CO₂ is at a moderate level (ca. 50 and 30 cm³ g⁻¹ STP @273 K at P=P ₀) and is much higher than for CH₄ and N₂ (Figure 5 d). The isosteric heats of adsorption of CO₂ and CH₄ for higher degrees (>0.3) of surface coverage equal to ca. 27 and 20 kJ mol⁻¹, respectively, and thus they are comparable to values found for most COFs. The sorption of CO₂ and CH₄ were also studied with Pd‐functionalized COFs as well as BPO1–2. In all cases gas uptake was lower which can be ascribed to reduced porosity. Furthermore, coordination of Pd⁰ centers or oxidation of phosphorus centers did not result in a significant change of CO₂/CH₄ selectivity relative to BP1–2.

Laboratory PXRD analyses of all obtained materials indicate their crystallinity is rather low. Nevertheless, we succeeded in performing such analyses for BP1 COF using the synchrotron radiation (λ=0.178 Å). The PXRD pattern of BP1 displays several broad peaks at lower 2θ angle (Figure 6). We proposed several structural models and compared the generated PXRD patterns to the experimental one. The best fit was obtained with a 3D structure resembling the topology of COF‐105 and COF‐108 obtained from condensation of tetrahedral precursors M(4‐B(OH)₂Ph)₄ (M=C, Si) and HHTP as a trigonal linker.9 However, in BP1 $1/4$ of connections are replaced by the lone electron pair of the phosphorus atom. This partially resembles the situation observed in COF materials reported by Dichtel and Bunck by co‐condensation of tetrahedral C(4‐B(OH)₂Ph)₄ and truncated CR(4‐B(OH)₂Ph)₃ (R=n‐C₁₂H₂₆, allyl).55 In our case, the structure is even more labile and supposedly vulnerable for catenation. In our model we proposed a two‐fold interpenetration level, where two neighboring networks are related by the inversion center. However, a higher degree of catenation can also be considered. The PXRD pattern of BP2 also shows some degree of structural ordering (Figure S37) which indicates that both materials can be categorized as COFs. We note, however, the proposed model should properly describe only the short‐range order in this material. It seems the structures BP1 and BP2 are not strictly defined due to variable orientation of phosphorus centers leading to significant disorder and lack of well‐defined long‐range order.

Figure 6.

(a) Synchrotron‐measured (black solid line) and simulated (blue dashed line) PXRD pattern of BP1. (b) Proposed structural model for the BP1 material. (c) Connolly surface generated in the Mercury program; (d) two neighbored polymeric fragments are related by the symmetry center; (e) rotation around C−P bond as a source of structural disorder.

In order to study the specificity of host‐guest interactions in obtained materials, we have performed series of quantum chemical calculations at M062X/cc‐pVTZ level of theory.56, 57 To simplify the calculations, the networks of BP1 and BPO1 were reduced to two closest fragments comprising organoboron and organophosphorus monomers with the initial distance between boron and P or P=O oxygen atoms set to 2.4 Å (BP1’ and BPO1’). In the case of BP1’ the geometry optimization led to the increase of the B…P distance to about 3.5 Å, meaning that moderate compatibility of phosphorus LP and boron free p orbital is insufficient to overcome the steric hindrance. In turn, in BPO1’ the P=O…B distance was shortened to 2.11 Å resulting in appearance of dative P=O→B bond and slight pyramidalization of geometry around boron atom. The calculated interaction energy of 15.5 kJ mol⁻¹ suggests that BPOs would exhibit tendency for additional interconnections through the dative P=O→B interactions, but this effect seems to be too weak to organize the material into the well‐defined crystal samples. Conversely, such a labile bond would rather decrease the ordering level.

Since polymer networks of BP1 and BPO1 are considered labile, it can be expected that Lewis base and acid centers would appear in the appropriate distance to invoke the local electric field effect, which could then enhance the interactions with guest molecules. The electrostatic potential maps plotted on electron density isosurfaces show strong electronegative regions around the phosphorus (BP1’) or P=O oxygen (BPO1’) atoms, while the slightly positive regions appear above the planes of boronate groups. (Figure S86). Due to rather low Lewis acidity of the boron atom, the expected binding effect would be rather weak and unsymmetrical with respect to classical frustrated Lewis pairs (FLPs).58 It would be mostly dominated by the interaction with P^III/P=O donor, and only supported by the interaction with boron atom. Indeed, the performed calculations on the host‐guest systems confirms that the presence of boron and phosphorus centers supports binding of H₂, CH₄ and CO₂ molecules by the framework, but this effect is insufficient to promote heterolytic bond splitting. The computed interaction energies clearly show that BPO1’ displays higher affinity toward H₂ and CH₄ molecules, while BP1’ strongly interacts (ΔE=−33.4 kJ mol⁻¹) with CO₂ molecule. In line with these observations, the P/O⋅⋅⋅H (H₂, CH₄) distances are shorter in BPO1’ and P=O⋅⋅⋅CO₂ distance is longer with respect to P⋅⋅⋅CO₂ in BP1’ (Table 3). This can be simply understand by the higher basicity of oxygen atom, and stronger nucleophilic character of phosphorus atom.

Table 3.

Characterization of host‐guest interaction energies. ΔE denotes the interaction energy between host and guest molecule, d is interatomic distance between P/O, B and guest molecule XY (X,Y=H for H₂, CH₄; X=C and Y=O for CO₂), ρ and ${\nabla }^2$ ρ are electron density and its Laplacian at P/O…X (BPC1) and Y…B (BCP2) bond critical points.

Host	Guest	ΔE [kJ mol−1]	d P/O⋅⋅⋅X [Å]	d B⋅⋅⋅Y [Å]	d P/O⋅⋅⋅B [Å]	ρ(BCP1) [eÅ−3]	▿2 ρ(BCP1) [eÅ−5]	ρ(BCP2) [eÅ−3]	▿2 ρ(BCP2) [eÅ−5]
BP1	H2	−1.1	2.503	2.273	4.850	0.12	1.02	0.09	0.94
	CO2	−33.4	2.393	3.155	6.014	0.41	0.80	0.03	0.46
	CH4	−0.8	2.835	2.426	6.483	0.07	0.57	0.07	0.69
BPO1	H2	−5.7	1.963	2.101	4.458	0.16	2.34	0.11	1.09
	CO2	−6.3	2.437	2.526	5.186	0.14	2.23	0.11	1.19
	CH4	−4.2	2.556	2.933	6.306	0.05	0.56	0.03[a]	0.33[a]

[a] BCP2 was found between C−H hydrogen and boronate ester oxygen atom.

The deeper insight into the host‐guest binding mechanism is provided by the topological analysis of electron density in the framework of Bader's quantum theory of atoms in molecules (QTAIM).59 Table 3 gathers the most important geometrical and electron density topological features of host‐guest interactions. Molecular graphs showing the formation of bond paths and bond critical points (BCP) in the binding pocket are presented on Figure 7, full molecular graphs are given in SI (Figures S87–S92). The topological analyses of electron density recognized bond paths and bond critical points (BCPs) between all donor centers and H (H₂, CH₄) or C (CO₂) atoms from guest molecules. The electron density values follows the general trend observed for interatomic distances. Specifically, for BPO1’‐H₂ adduct, where donor…H distance is the shortest from studied series (d _O….H=1.963 Å), the electron density at P⋅⋅⋅H bond critical point reaches the value of 0.16 eÅ⁻³ with negative Laplacian of 2.34 eÅ⁻⁵ and host‐guest interaction energy equals to −5.7 kJ mol⁻¹ indicating that this interaction can be classified as weak hydrogen bond. As indicated by longer P⋅⋅⋅H distance (2.503 Å) and lower ρ(BCP1) value of 0.12 eÅ⁻³, the binding of dihydrogen by BP1’ is weaker (ΔE=−1.1 kJ mol⁻¹), but still satisfies the criteria of very weak intermolecular hydrogen bond. In both cases, the donation from donor to H₂ molecule results in elongation of H−H bond from 0.744 Å (equilibrium distance in free H₂ calculated at the same level of theory) to 0.761 Å in BP1’ and 0.780 Å in BPO1’, which is accompanied by the reduction of electron density at H‐H BCP from 1.84 eÅ⁻³ to 1.81 eÅ⁻³ (BP1’) and 1.68 eÅ⁻³ (BPO1’). Our observations are somewhat consistent with theoretical studies conducted by Pinter et al.60 on FLPs featuring low‐energy dihydrogen activation transition states termed “early” (such as tBu₃P⋅B(C₆F₅)₃). In contrast to so‐called “late” FLPs, such “early” complexes are characterized by relatively short H‐H distances (0.79–0.80 Å), slightly decreased electron density at H₂ BCP (1.6 eÅ⁻³), long P⋅⋅⋅H distances and electron density values at BCPs of P⋅⋅⋅H bond in the range of 0.2–0.3 eÅ⁻³. Nonetheless, “early” FLPs systems are active hydrogenation catalysts. The comparison between FLPs and our systems leads to the conclusion that in the latter the donor binding effect is weaker presumably due to the lack of strong support from acceptor side. Although, topological analysis of electron density revealed the formation of H…B bond path in both models BP1’ and BPO1’, the value of electron density at BCP2 oscillates near 0.1 eÅ⁻³ and it is more than two times smaller with respect to “early” tBu₃P⋅H₂⋅B(C₆F₅)₃ complex. On the other hand, the lower activity of BP1’ and BPO1’ is compensated by the higher stability of formed host‐guest adducts.

Figure 7.

(a,c) Modelling of host‐guest interactions in BP1 and BPO1 materials. (b,d) Fragment of molecular graphs showing bond paths (black dashed lines) and bond critical points (small blue spheres) in host‐guest interaction region.

An interesting observation can be made by the comparison of interatomic distances and electron density features of B⋅⋅⋅H versus donor P^III/P=O⋅⋅⋅H interactions with dihydrogen and methane. Despite the C−H bond polarization, the donor⋅⋅⋅H interactions are weaker with methane than dihydrogen. It should be also noted that in case of the BPO1’‐CH₄ bond path was recognized between C−H hydrogen atom and oxygen atom from boronate ester group with electron density of only 0.03 eÅ⁻³ at BCP2. Thus the contribution from boron atom is questionable. Regardless of binding mechanism, the C−H bonds from donor and acceptor sites are elongated and electron density depleted at BCP with respect to free CH₄ molecule. Furthermore, the calculations show that both H₂ and CH₄ adducts are unstable in the absence of boron counterparts.

The BP1’‐CO₂ is the most distinct adduct from the studied series. The exceptionally high value of electron density at P⋅⋅⋅CO₂ BCP of 0.46 eÅ⁻³ confirms the strong character of this interaction and points to the chemical nature of CO₂ adsorption. This also leads to the significant bending of CO₂ molecule (α_O‐C‐O=153°), elongation of C−O bond (from 1.155 Å to 1.178 Å), and reduction of electron density at corresponding BCP (from to 3.158 eÅ⁻³ to 3.038 eÅ⁻³) with respect to free CO₂ molecule. On the other hand, the boron atom is barely involved. Turning to less nucleophilic phosphine oxide, the interaction energy with CO₂ (ΔE=−6.3 kJ mol⁻¹) and the amount of electron density localized at BCP1 (0.14 eÅ⁻³) are comparable to BPO1’‐H₂ adduct. Accordingly, the geometry of CO₂ molecule approach to linear with the O‐C‐O angle of 171°. Interestingly, the formation of moderate P=O⋅⋅⋅CO₂ interaction is accompanied by the increased contribution from boron atom reflected by relatively short B⋅⋅⋅O contact (2.526 Å) and electron density value of 0.11 eÅ⁻³ at BCP2.

The analysis of Hirshfeld61 and Bader59 atomic charges revealed that the charge is transferred from donor to guest molecule, however, it is only partially transferred further to acceptor unit (Tables S2 and S3). For most systems, the population at the donor atom (P^III/P=O) drops by 0.02–0.04 e (Hirshfeld charges), while the magnitude of charge donation from the guest molecule to the boron atom is of about 0.005–0.025 e. In line with previous findings, the magnitude of charge transfer from BP1’ donor to CO₂ molecule is larger. The charge at P atom is depleted by ca. 0.1 e, while guest molecule and boron atom increase electron population by 0.162 e and 0.025 e, respectively. It is also noticeable that charge is distributed unsymmetrically within the guest molecule, wherein it is mostly shifted toward the acceptor hydrogen (H₂, CH₄) or carbon (CO₂) atoms. This again emphasizes the dominant character of P^III/P=O donor site in the binding processes.

Conclusions

We have demonstrated that the general approach to phosphine coordination materials mostly developed for MOF family can be successfully extended to covalent organic frameworks. The transesterification of triboronated triphenylphosphine 2 and its oxide 4 esters with polyhydroxy HHTP and THDMA linkers is a facile route toward the formation of boron‐phosphine COFs that are characterized by higher selectivity toward H₂, CH₄ and CO₂. The tripodal topology of the employed boronic linkers stems from the presence of the phosphorus(III) center. Obtained materials display lower crystallinity with respect to related boron COF materials resulting from fast transesterification rates and statistically random orientation of phosphorus center with respect to three aryl substituents and lone electron pair. However, synchrotron radiation experiments revealed some short‐range structure ordering. The proposed BP1 structural model assumes two‐fold interpenetration level and general topology preserved from COF‐105, but it is characterized by higher flexibility and ordering discontinuity.

The BET surfaces derived from nitrogen sorption isotherm are moderate. However, obtained materials exhibit uptakes of hydrogen, CO₂ and methane at a level similar to those found for boron COFs featuring higher N₂‐based S _BET values. In other words, the obtained porous materials have much higher relative affinity with respect to H₂, CO₂ and CH₄ than standard boron COFs. We suppose that this effect can be ascribed to the presence of boron Lewis acidic and phosphorus Lewis basic centers which is beneficial for binding of more polarizable molecules due to generation of local electric field gradients. Theses suppositions has been confirmed by theoretical calculations. The interaction with gaseous molecules has some common features with “early” FLP systems reflected by slightly decreased distance and electron density at BCP of H−H bond, elongated donor⋅⋅⋅guest distances with electron density localized at corresponding BPC ranging from 0.2–0.3 eÅ⁻³ and 0.02–0.04 e charge transferred from basic center to guest molecule. On the other hand the contribution of Lewis acidic boron atom is less pronounced. Thus, an overall effect is beneficial for sorption effectiveness but it is insufficient to promote stronger host‐guest interactions or bond splitting. In contrast, theoretical calculations suggest that CO₂ is bound by BP1 more effectively due to stronger nucleophilic character of P^III atom and intrinsic polarity of C=O bonds.

Post‐synthetic modification of the materials was performed by impregnation with Pd₂(dba)₃ in DCM. XPS analyses demonstrated that the presence of phosphorus donor centers results in a strong affinity to Pd⁰ leading to the high Pd/P ratio of ca. 0.9. Incorporation of the metal resulted in significant decrease of the BET surface. However, there is still space left available for uptake of guest molecules. In a different bottom‐up approach, the precursor 1 was first reacted with Pd₂(dba)₃ followed by polycondensation with HHTP or THDMA. Thus the resulting materials possess the hybrid character due to presence of Pd−P dative bonds as well as boronate ester moieties typical of boron COFs. The sorption properties of these amorphous networks are slightly worse than those found for BP1–2. According to XPS analyses, the obtained materials exhibit much lower surface Pd/P ratio of 0.26–0.31 consistent with the assumed stoichiometry of the Pd complex used as a precursor. All obtained Pd‐containing porous materials could potentially serve as heterogeneous catalysts, e.g., for hydrogenation and cross‐coupling reactions. We will test such applications in our future approaches.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

P. Pacholak, K. Gontarczyk, R. Kamiński, K. Durka, S. Luliński, Chem. Eur. J. 2020, 26, 12758.