Basicity-Tuned Reactivity: diaza-[1,2]-Wittig versus diaza-[1,3]-Wittig Rearrangements of 3,4-Dihydro-2H-1,2,3-benzothiadiazine 1,1-Dioxides

The base-induced (t-BuOK) rearrangement reactions of 3,4-dihydro-2H-1,2,3-benzothiadiazine 1,1-dioxides result in a ring opening along the N–N bond, followed by ring closure with the formation of new C–N bonds. The position of the newly formed C–N bond can selectively be tuned by the amount of the base, providing access to new, pharmacologically interesting ring systems with high yield. While with 2 equiv of t-BuOK 1,2-benzisothiazoles can be obtained in a diaza-[1,2]-Wittig reaction, with 6 equiv of the base 1,2-benzothiazine 1,1-dioxides can be prepared in most cases as the main product, in a diaza-[1,3]-Wittig reaction. DFT calculations and detailed NMR studies clarified the mechanism, with a mono- or dianionic key intermediate, depending on the amount of the reactant base. Also, the role of an enamide intermediate formed during the workup of the highly basic (6 equiv of base) reaction was clarified. The substrate scope of the reaction was also explored in detail.


Data Collection
A colorless prism crystal of C 11 H 14 Cl 2 N 2 O 4 S having approximate dimensions of 0.40 x 0.04 x 0.02 mm was mounted on a cactus needle. All measurements were made on a Rigaku RAXIS RAPID imaging plate area detector with graphite monochromated Cu-Kα radiation.
Indexing was performed from 4 oscillations that were exposed for 900 seconds. The crystal-todetector distance was 127.40 mm.
Cell constants and an orientation matrix for data collection corresponded to a primitive tetragonal cell (laue class: 4/mmm) with dimensions: a = 20.0729(4) Å c = 7.2451(2) Å V = 2919.21(12) Å 3 For Z = 8 and F.W. = 341.21, the calculated density is 1.553 g/cm 3 . Based on the systematic absences of: 0kl: k ± 2n hhl: l ± 2n packing considerations, a statistical analysis of intensity distribution, and the successful solution and refinement of the structure, the space group was determined to be: Readout was performed in the 0.100 mm pixel mode.

Data Reduction
Of the 32203 reflections that were collected, 2821 were unique (R int = 0.086).
The linear absorption coefficient, µ, for Cu-Kα radiation is 54.845 cm -1 . An empirical absorption correction was applied which resulted in transmission factors ranging from 0.640 to 0.877. The data were corrected for Lorentz and polarization effects.

Structure Solution and Refinement
The structure was solved by direct methods 1 and expanded using Fourier techniques 2 . The nonhydrogen atoms were refined anisotropically. Some hydrogen atoms were refined isotropically and the rest were refined using the riding model. The final cycle of full-matrix least-squares refinement 3 on F was based on 20447 observed reflections (I > 2.00σ(I)) and 207 variable parameters and converged (largest parameter shift was 0.00 times its esd) with unweighted and weighted agreement factors of: The standard deviation of an observation of unit weight 4 was 3.13. Unit weights were used. Plots of Σ w (|Fo| -|Fc|) 2 versus |Fo|, reflection order in data collection, sin θ/λ and various classes of indices showed no unusual trends. The maximum and minimum peaks on the final difference Fourier map corresponded to 14.10 and -31.00 e -/Å 3 , respectively.