Novel synthetic pathway for the production of phosgene

A new approach for the production of the industrially important intermediate phosgene based on chloride catalysis was developed.


INTRODUCTION
Since its discovery in 1812 by Davy (1), phosgene [C(O)Cl 2 ] has evolved as one of the most important industrial chemicals along with, sulfuric acid, ammonia, ethylene, and chlorine. As an "intermediate" chemical, it serves as starting material for polymers, agrochemicals, and pharmaceuticals to mention only a few (2). Currently, 12 million metric tons are produced per year mainly for the synthesis of polyurethanes and polycarbonates, and it is estimated that the production of phosgene will rise to 18.6 million metric tons/year until 2030 (2). Because of its high toxicity, phosgene is only manufactured by a few specialized companies, typically on multi-ton scale requiring a multilevel safety concept. It is obtained by gas phase reaction of carbon monoxide and chlorine at elevated temperature using activated carbon as a catalyst (Eq. 1) (2) CO (g) + Cl 2(g) cat.
⇌ C(O ) Cl 2(g) (1) Although the exact mechanism of the phosgene formation is still under debate, it is widely accepted that the reaction is initiated by activation of the Cl─Cl bond. As proposed by Lennon and co-workers (3), the first step is a dissociative chemisorption of Cl 2 on the carbon surface. The adsorbed chlorine atoms react with gaseous CO (Eley-Rideal mechanism) to form a surface-bound acyl chloride entity [C(O)Cl (ad) ], which further reacts with a surface-bound chlorine atom, leading to phosgene (Langmuir-Hinshelwood mechanism). In contrast, Lercher and co-workers (4) assume a two-step Eley-Rideal mechanism, when C 60 fullerene is used as model system, while the catalytically active species is the surface-bound triplet diradical [C 60 ···Cl 2 ] •• . Further studies on nitrogen-modified carbon materials propose a polarization of the Cl─Cl bond by interaction with electropositive carbon sites of the material (Lewis acid catalysis). Reaction with CO leads to an acyl chloride cation [C(O)Cl + ] and a weakly bound Cl − , which react with each other to form phosgene (5).
However, the high exothermicity of the reaction (H = −107.6 kJ mol −1 ) and subsequent dissipation of process heat is more problematic as the temperature in the iron tube reactors can rise up to 550°C at hotspot reaction sites (2). Because of the high temperatures, the catalysts also slowly degrade by attack of Cl 2 and Cl • atoms on carbon defects, leading to the corrosive formation of HCl and CCl 4 , which leads to shorter maintenance cycles of the reactor (7,9).
At the outset of this work, we anticipated that the reaction of CO and Cl 2 to C(O)Cl 2 could be catalyzed by activation of Cl 2 using a weakly coordinated chloride anion. In the reaction of Cl − and Cl 2 , polychlorides are formed, which show a broad structural diversity and promising applications (10).
The bonding properties of various trichlorides, the simplest polychlorides, were recently analyzed by experimental and computed electron density studies. Accordingly, there seems to be a smooth transition from an asymmetric [Cl … Cl─Cl] − unit to a symmetric [Cl─Cl─Cl] − anion with two equal Cl─Cl bonds in a crystalline environment. These different bonding types of the [Cl 3 ] − anion are crucial for its chemical reactivity (11).
Depending on the cation, the trichloride [Cl 3 ] − is a yellowish salt or a room temperature ionic liquid (RT-IL). Alkylammonium salts such as triethylmethylammonium chloride, [NEt 3 Me]Cl, are considered to be potential materials for the efficient and convenient storage of elemental chlorine as an RT-IL at atmospheric pressure (12). This could enable a more flexible chlorine production that can be adapted to the availability of (renewable) electrical energy and thus represents a secondary energy storage system. Here, we report a preparation of phosgene from carbon monoxide and elemental chlorine that proceeds in a homogeneous reaction at room temperature and atmospheric pressure using a [NEt 3  ] where x depends on the partial pressure of chlorine and the size of x has a great influence on the properties of the system (see the Supplementary Materials). When x < 0.8, the system exists as a yellow solid, while an increase of the chlorine concentration (x > 0.8) results in the formation of an RT-IL (see Fig. 1 and movie S1).
Initially, the stoichiometric reaction of CO with liquid [NEt 3 Me] [Cl(Cl 2 ) x ] (x = 1.1) to C(O)Cl 2 at room temperature was investigated by gas-phase infrared (IR) spectroscopy, indicating rapid formation of phosgene. If by consumption of Cl 2 x becomes smaller than 0. To improve the contact time between liquid and gaseous reactants, we designed a flow setup, which consists of a gas-washing bottle, filled with a dispersion of [NEt 3 Me][Cl(Cl 2 ) x ] in oDCB, and a peristaltic pump for continuously circulating gaseous CO and already formed C(O)Cl 2 in the system (see figs. S1 and S2). This setup allows a spectroscopic in situ monitoring of the reaction progress by passing the gas mixture through IR and ultraviolet/ visible (UV/Vis) cells. The consumption of CO (IR spectrum; Fig. 2A) and Cl 2 (UV/Vis spectrum; Fig. 2B) as well as a simultaneous formation of C(O)Cl 2 were recorded, indicating an immediate start of the reaction.
Control experiments with Cl 2 and CO in our setup have shown that both the beam of the UV/Vis spectrometer and visible light slightly contribute to the formation of C(O)Cl 2 by photolytic activation of Cl 2 . Therefore, all experiments were conducted in the dark to avoid light-induced radical formation and to emulate industrial processing of phosgene in stainless steel tubes. To demonstrate that the phosgene formation requires no photoactivation, we conducted a series of control experiments. CO was treated with Cl 2 without a chloride salt (a), in the presence of catalytic amounts of solid [NEt 3    was observed in the dark, but in experiments b and c, C(O)Cl 2 was formed (see fig. S22 3 ], was applied as another catalyst to investigate whether a solid/gas reaction could be used for the production of phosgene (d). Using this catalyst, formation of C(O)Cl 2 was substantially slower compared to reaction b (see the Supplementary Materials). Notably, no attack of the cation [NEt 3 Me] + in [NEt 3 Me][Cl(Cl 2 ) x ] (x = 1.6) was observed, when stored under an atmosphere of chlorine gas at 1 bar at room temperature for years, as shown by Raman spectroscopy.
To rule out a radical mechanism, we treated the ionic liquid [NEt 3 Me][Cl 3 ] with methane in the dark. As neither chlorinated methane nor HCl was observed IR spectroscopically (for details, see the Supplementary Materials), a radical-based mechanism for the formation of phosgene under these conditions was rejected.
To achieve further mechanistic understanding, we carried out quantum-chemical CCSD(T)-  Fig. 4), which spontaneously dissociates into C(O)Cl 2 and Cl − . In summary, the reaction of CO and [Cl 3 ] − can be regarded as the insertion of CO into an activated Cl─Cl bond, leading to C(O)Cl 2 and Cl − by releasing about 60 kJ mol −1 .
However, the computational estimation of the reaction barrier depends on cation-anion interactions. When the influence of only one cation is taken into account, the free energy of the transition state is relatively high (77.6 kJ mol −1 ). In contrast, when no cation effects but only solvent effects are considered, the transition state is calculated to have a substantially lower energy (56.9 kJ mol −1 ). As the real system involves the interaction of multiple cations and solvent molecules, the real free energy of the transition state can be expected to be in the same range between 56.9 and 77.6 kJ mol −1 . For the uncatalyzed reaction of CO and Cl 2 to phosgene, a free energy activation barrier of about 230 kJ mol −1 was calculated. This indicates that the activation barrier is tremendously reduced by our chloride catalyst comparable to the activated carbon-catalyzed process.
On the basis of these results, a catalytic reaction scheme can be proposed in which phosgene is prepared at room temperature by chloride catalysis. The prepared phosgene can be used for further processes, most importantly the synthesis of isocyanates to produce polyurethanes. This was demonstrated for the synthesis of phenyl isocyanate by adding aniline to the generated phosgene solution (see the Supplementary Materials). In the reaction of amines with phosgene to isocyanates, HCl is released, which, in an industrial process, is to some extent typically electrolyzed to regenerate elemental chlorine (Fig. 5).
Concluding  both as a convenient chlorine storage medium and as an efficient catalyst for the production of phosgene opens up new industrial options.

Apparatus and materials
All substances sensitive to water and oxygen were handled under an argon atmosphere using standard Schlenk techniques and oil pump vacuum up to 10 −3 mbar. Dry oDCB was obtained after storage over activated 3-Å molecular sieves. Commercially available triethylmethylammonium chloride (TCI) and tetraethylammonium chloride (TCI) were dried in vacuo at 150°C for 1 hour before use. Aniline (Acros), chlorine (5.0, Linde), and carbon monoxide (2.0, Linde) were used without further purification. Raman spectra were recorded on a Bruker (Karlsruhe, Germany) MultiRAM II equipped with a low-temperature Ge detector (1064 nm, 100 to 180 mW, resolution of 4 cm −1 ). IR spectra were recorded on a Nicolet iS5 Fourier transform IR (FTIR) spectrometer (gas IR cell: 10 cm, KBr windows) or a Bruker Vector 22 FTIR spectrometer (gas IR cell: 10 or 20 cm, silicon windows). UV/Vis spectra were recorded on a PerkinElmer Lambda 465

General description of the flow setup To investigate the kinetics of the reaction between [NEt 3 Me][Cl 3 ]
and CO, a glass vacuum line was connected to a gas-washing bottle (reactor) via a peristaltic pump, which successively circulates gaseous reactants through the reactor, a UV/Vis, and an IR flow cell, to monitor the formation of C(O)Cl 2 and the consumption of Cl 2 and CO as well (see figs. S1 and S2). To ensure proper mixing of all reactants, the liquid and solid reactants were filled into the gaswashing bottle, and all gaseous reactants are pumped through the system. All connections were made using either perfluoroalkoxy alkane or C-Flex Ultra (Cole-Parmer) tubing. This setup was used for experiments using stoichiometric and varying catalytic amounts of [NEt 3 Me]Cl for the reaction of CO + Cl 2 . In addition, blind experiments for the noncatalyzed reaction between Cl 2 and CO were performed, which showed that the beam of the UV/Vis spectrometer and visible light can induce the formation of phosgene from Cl 2 and CO. Therefore, all following experiments have been performed in the dark (see fig. S3). Also, the reaction between [NEt 3 Me]Cl, Cl 2 , and CH 4 was studied to verify a nonradical mechanism. The progress of the reaction was monitored by integrated absorbances of the IR bands in the spectral regions between 1995 and 2250 cm −1 (CO) and 1760 and 1885 cm [NEt 3 Me]Cl (0.460 g, 3.033 mmol) and 20 ml of oDCB were filled into a gas-washing bottle, which was connected to a glass vacuum line and an IR flow cell. The reaction mixture was degassed, and the system was filled with a mixture of CO and Cl 2 (CO: 1000 mbar, 22.32 mmol, 553 ml; Cl 2 : 208 mbar, 4.65 mmol, 553 ml). The gaseous reactants were pumped through the system using a peristaltic pump for 226 min. IR spectra were recorded after 0, 6, 11, 16, 21, 26,  31, 36, 41, 46, 51, 56, 61, 66, 71, 76, 81, 86, 91, 96, 101, 106, 111, 116,  121, 126, 131, 136, 141, 184, 191, 196, 201, 206, 211 ) and 20 ml of oDCB were filled into a gas-washing bottle, which was connected to a glass vacuum line and an IR flow cell. The reaction mixture was degassed, and the system was filled with CO (1000 mbar, 553 ml, 22.32 mmol). The gaseous reactants were pumped through the system using a peristaltic pump for 153 min. IR spectra were recorded after 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 98, 103, 108, 113, 123, 128, 133, 138, 144,  154, 164, 174, 184, 194, 204, 214, and 224 min (figs. S12 to S14).
[ )] and 20 ml of oDCB were filled into a gas-washing bottle, which was connected to a glass vacuum line, an IR flow cell, and a UV/Vis flow cell. The reaction mixture was degassed, and the system was filled with CO (1000 mbar, 800 ml, ca. 32 mmol). The gaseous reactants were pumped through the system using a peristaltic pump for 420 min. IR and UV spectra were recorded every 5 min [NEt 4 ]Cl (0.230 g, 1.393 mmol) was filled into a gas-washing bottle, which was connected to a glass vacuum line and an IR flow cell. The system was evacuated and filled with a mixture of CO and Cl 2 (CO: 850 mbar, 20.20 mmol, 588 ml; Cl 2 : 150 mbar, 3.530 mmol, 588 ml). The gaseous reactants were pumped through the system using a peristaltic pump for 628 min. IR spectra were recorded after 0, 1, 6,11,21,27,32,37,42,47,52,65,77,137,197,257,317,377,437,497,558, and 628 min (figs. S15 to S17) Cl 2 + CO → C(O ) Cl 2 (flow) oDCB (20 ml) was filled into a gas-washing bottle, which was connected to a glass vacuum line and an IR flow cell. The oDCB was degassed, and the system was filled with a mixture of CO and Cl 2 (CO: 850 mbar, 20.20 mmol, 588 ml; Cl 2 : 150 mbar, 3.53 mmol, 588 ml). The gaseous reactants were pumped through the system using a peristaltic pump for 811 min. IR spectra were recorded after 0, 4,9,14,18,19,34,49,64,79,94,109,124,139,154,261,262,323,384,445,506,567,628,689,750, and 811 min (figs. S18 to S20). Repetition of the experiment without exclusion of light yields the formation of phosgene, which was shown by gas-phase IR spectroscopy ( fig. S21 [NEt 3 Me]Cl (0.499 g, 3.290 mmol) and 20 ml of oDCB were filled into a gas-washing bottle, which was connected to a glass vacuum line and an IR flow cell. The reaction mixture was degassed, and the system was filled with a mixture of argon, CH 4 , and Cl 2 (argon: 705 mbar, 588 ml; CH 4 : 95 mbar, 588 ml, 2.25 mmol; Cl 2 : 200 mbar, 4.74 mmol, 588 ml). The gaseous reactants were pumped through the system using a peristaltic pump for 242 min. IR spectra were recorded 0, 3,6,9,12,15,19,22,25,28,39,40,100,161,221,282,342, 403, 363, and 524 min ( fig. S23).
Proof for the formation of phenyl isocyanate [NEt 3 Me]Cl (0.367 g, 2.420 mmol, 0.31 equiv.) was loaded into a 500-ml two-neck Schlenk flask, dried in vacuo at 150°C for 1 hour, and suspended in 20 ml of oDCB. The solution was degassed, and chlorine was added until the system retained a pressure of 200 mbar (0.560 g, 7.898 mmol, 1 equiv.). CO (800 mbar) (ca. 16 mmol, 2.3 equiv.) was added to the flask, and the reaction mixture was stirred for 3 days in the dark. The reaction mixture was cooled to −196°C and degassed. Then, the system was filled with dry argon gas and connected to a dropping funnel and a condenser that was cooled by using −15°C cold ethanol. The condenser was opened to the fume hood via a gas bubbler and a series of four gas-washing bottles, two of which are filled with a KOH solution and an NH 4 OH solution, respectively, each followed by an empty bottle. A solution of 0.5 ml of aniline (0.51 g, 5.476 mmol, 0.69 equiv.) in 5 ml of oDCB was added slowly via the dropping funnel to the reaction mixture held at −15°C, and the reaction mixture was then heated to 100°C for 8 hours. After that, excess of phosgene was removed in vacuo, and the analysis of the reaction products using gas-phase IR spectroscopy revealed the characteristic NCO stretching band at 2273 cm −1 of the reaction product ( fig. S24).

298.15K
) and entropic contributions (−TΔS) were obtained from harmonic normal mode analyses. To incorporate higher-order electron correlation effects, a correction term Δ CCSD(T) − F12 was calculated from the difference of the SCS-MP2 and CCSD(T)-F12 energies of the anionic systems in vacuum and without any counterion. Correction terms to the Gibbs free energies of reactions in solution under standard conditions (ΔG COSMO − RS ;298.15 K, 0.1 MPa) were obtained using the COSMO-RS solvation model (40)(41)(42)(43). To this end, additional single-point calculations at the vacuum and polarizable continuum model (PCM) optimized structures were carried out using the TURBOMOLE program version 7.5.0 (44)(45)(46). These calculations were performed at the DFT-BP86 (47,48) level of theory in conjunction with def2-TZVPD (49) basis sets for all atoms, the multipole-accelerated resolution-of-identity approximation, and the refined COSMO cavity construction algorithm (keyword $cosmo_isorad) (50)(51)(52)(53)(54). Subsequent COSMO-RS computations used the COSMOtherm program version C30_1201 and a BP-TZVPD-FINE level parameterization (BP_TZVPD_FINE_ HB2012_C30_1201). For an in-depth analysis of the reaction profile in terms of its thermochemistry, different thermochemical quantities were evaluated: the pure electronic energy ΔE 0K , the enthalpy at 298.15 K with zero-point energy corrections added, ΔH 298.15K , the Gibbs free energy in vacuum, ΔG 298.15K , and the Gibbs free energy in oDCB solvent, ΔG 298.15K, oDCB . All consecutive steps of the reaction of [Cl 3 ] − with CO were modeled with and without inclusion of the counter-cation [NEt 3 Me] + .