Hexachloroacetone as a Precursor for a Tetrachloro-substituted Oxyallyl Intermediate: [4+3] Cycloaddition to Cyclic 1,3-dienes

The enol phosphate 2,2-dichloro-1-(trichloromethyl)ethenyl diethyl phosphate (1), easily available by a Perkow reaction between hexachloroacetone and triethyl phosphite, reacts with sodium trifluoroethoxide/trifluoroethanol in the presence of cyclic 5-membered 1,3-dienes to furnish α,α,α',α'-tetrachloro-substituted[3.2.1]bicyclic ketones 2. A [4+3] cycloaddition of a tetrachloro-oxyallyl intermediate is postulated.


Introduction
A tetrachloro-substituted oxyallyl intermediate has been generated from pentachloroacetone (PCA) and utilized for many [4+3] cycloaddition reactions with 1,3-dienes and furans [1][2][3][4][5]. On preparation of PCA by chlorination of acetone, hexachloroacetone (HCA) is formed as a by-product in varying amounts [6]. Instead of disposing of the perchlorinated ketone, it seemed desirable to us to also utilize HCA for the generation of the tetrachloro-oxyallyl species. Moreover, HCA, a useful reagent, is also commercially available at present or may be prepared easily [7,8].
Standard metal reductions, e.g. with Zn/Cu couple, do not promise good selectivity in [4+3] cycloadditions because the tetrachloro-substituted cycloadducts formed are expected to be dehalogenated further.

Results and Discussion
Checking the literature, we became aware of the reaction between HCA and trialkyl phosphites, resulting in enol phosphates, e.g. 2,2-dichloro-1-(trichloromethyl)ethenyl diethyl phosphate (1) [9], which has been investigated because of its biological activity as a repellent against insects [10]. This dehalogenation, an example of the Perkow reaction [11], represents a selective reduction of HCA as well. The enol phosphate 1 has an allylic structure with an oxygen atom at the central carbon atom (C-2), and chloride leaving groups at one terminus. Therefore, we expected that ionization of this allyl chloride should be possible, leading eventually to a tetrachloro-oxyallyl intermediate. The ionization should be promoted by use of appropriate solvent systems, such as polyfluorinated alcohols [1,3].
The peak of the trifluoroethyl phosphate 3 was also found in the gas chromatograms of the reaction mixtures with furan (i.e. as by-product of 2a) and all the 1,3-dienes described below. Due to the polarity and volatility of the ester 3 [12] the cycloadducts 2 could be separated and isolated without any difficulties by ether extraction from the water/TFE phase, and work-up by distillation, crystallization or chromatography, if necessary. Scheme 1. Base-induced alcoholysis of 2,2-dichloro-1-(trichloromethyl)ethenyl diethyl phosphate (1), for R = CF 3 CH 2 .
From the Scheme it is evident that one equivalent of sodium trifluoroethoxide is needed for generating the oxyallyl intermediate. However, the consecutive side reactions would consume more base. Therefore we monitored the progress of the reaction by checking the pH with indicator paper. A slightly alkaline state of the reaction mixture (pH 7-8) was observed after 18 hours at ambient temperature.
This cyclocondensation protocol was also applied to cyclopentadiene and spiro [2.4]hepta-4,6diene, which furnished the carbobicycles 2b and 2c, respectively. The latter as yet unreported tricyclic spirocompound had been previously prepared by us in 45% yield from PCA using sodium 2,2,3,3tetrafluoropropoxide in 2,2,3,3-tetrafluoropropanol [3]. With triethylamine and lithium perchlorate/diethylether [2], the cyclocondensation between PCA and the spirodiene resulted in a very dark product mixture (see also the reaction with furan [3]). It was difficult to separate the tricycle 2c from the black, resinous side products that were formed in substantial amounts.
[a] This work Since TFE is an expensive solvent, we tried to replace it by ethanol, using sodium ethoxide as the base. However, not a trace of cycloadduct 2a was found upon reaction of 1 in the presence of furan. This further supports the unique properties of TFE.
A further approach towards ionization of 1 was made: a Lewis acid might assist in pulling off an 'allylic' chloro atom, possibly resulting in an O-phosphorylated oxyallyl intermediate. However, when 1 was allowed to react with boron trifluoride diethyl etherate and an excess of furan, no cycloadduct 2a could be detected. Other, more 'chlorophilic' Lewis acids were not investigated.

Conclusions
Our intention in these studies, utilization of hexachloroacetone as precursor for the generation and cycloaddition of oxyallyl intermediates, was verified in principle. It is true that the yields of the [4+3] cycloadducts are mediocre; but the preparations were not optimized. Possibly a more thorough investigation would result in better yields. The authors will not continue these investigations and would therefore welcome further research by other groups. Moreover, analogous derivatives of pentachloropropen-2-ol should be examined, e.g. pentachloro-2-(trimethylsiloxy)propene that can be prepared by the similar reaction of HCA with tris(dimethylamino)phosphane and chlorotrimethylsilane [13].