Isolation , Structure and M O Calculational Investigations of a Highly Stable , Hydrogen-bonded Primary Amine-Phosphine Oxide Adduct , 2-Aminobenzothiazole-HMPA , c-) NH 2 * O = P ( NMe 2 ) 3 ; a Possible Model to Explain the Carcinogenicity of HMPA ( HMPA = Hexamethylphosphoramide )

Colourless needles of the 1 : 1 primary amine-phosphine oxide adduct CsH4SC(=N)NH2.0=P(NMe2)3 are obtained in high yield simply by chilling a toluene solution of the two components; the solid-state structure of the highly stable adduct (which can be prepared in water, and which sublimes easily) consists essentially of dimers held by (amine) N-H...O and (amine)N-H...(heterocyclic) N hydrogen bonds.

We report the isolation and structure of a 1 : 1 adduct between 2-aminobenzothiazole m N ) N H 2 l a and hexamethylphosphoramide [HMPA, O=P(NMe&].The solid adduct consists essentially of dimers, (la.HMPA)2, with N-H.e.0-P and N-H-..(heterocyclic) N hydrogen bonds.It is unaffected by air or water, and it sublimes readily.Extensive ab initio M O calculations have been used to explore the bonding in the adduct since, to our knowledge, la.HMPA is the first amine (primary or otherwise)-phosphine oxide (HMPA or otherwise) complex to be structurally characterised.This uniqueness, plus the stability of the adduct, suggest it as a model to explain the carcinogenicity of HMPA.
was isolated subsequently simply by dissolving the components in toluene at SO "C, then cooling.It can indeed also be similarly prepared using water as the solvent.The adduct is also thermally stable; it sublimes intact (140 "C, 0.1 mmHg) and the monomeric molecular ion (mlz 330, for la.HMPA.H+) is intense in its FAB mass spectrum.
The crystal structure$ of the adduct consists essentially of dimers, (la.HMPA)2, in which one amino-H (H2) of each  In benzene solutions, cryoscopic measurements show that (la-HMPA), exists mainly as a monomer ( n = 1) (though slight further dissociation into the separate components is apparent).? Such monomer formation reflects in part the relative weakness of N(2)H(1) -.-N(1a) interactions in the solid.However, other features, and estimates of the energies involved, have come from ab initio M O calculationrs on models for a la.HMPA monomer and for a dimer.The models n for a monomer included adducts of Hc=CHSC(=&)NH* l a ' with O=P (NH& and oi l a itself with O=PH3.Structures A with a double hydrogen bridge were examined first [cf.Fig. l(a); cleavage of H(l)...N(la), then both H(2) and H(1) interacting with O( l)].However, during attempted optimisation of dibridged forms, the phosphine oxide moves, possibly owing to N...O repulsions or possibily to introduce secondary interactions (see below).A single N-H.-.O interaction results, allowing two isomers for each model; the NH.e.0 unit can be cis to the amine ring N or cis to the ring S. Typical results are shown in Fig. 2 for ( a ) la.0=PH3.cis to N, and ( h ) la.0=PH3, cis to S. The calculated adduct formation energies (AE, the decrease in energy on bringing together the two components) are 55.3 and 42.3 kJ mol-1 for the cis-N and cis-S specier, respectively.The difference is due to the cis-N isomer having, over and above the (N)H...O-P interaction, a secondary interaction between the P atom (charge, +0.85) of O=PH3 and H

H -N:
;O=P-A 8 A h initio MO calculations: 6-31G basis set with d orbitals on P and S atoms,7 using the program GAMESS.8 Geometries were freely optimised except that, since for l a ' adducts a plane of symmetry resulted from optimisation even when using no constraints, a plane of symmetry was assumed for l a adducts.The total energies (in a.u.) calculated for the optimised structures mentioned in the text are: O=PH3 -417.284250;O=P(NH2 tion energy for cis-N monomers, since NH...S interactions should be weak.This is confirmed by further MO calculations § on dimers of l a ' [(la')2 with NH-a-N attachments and with NH--.S ones] and on dimers of la'.0=PH3 adducts {(la'-O'=PH3)2 with NH...N links so NH...O is cis to S [Fig.

2(c)
] and with NH-..S ones so NH..-O is cis to N}.For (la')2, the dimerisation energies are 58.2 kJ mol-1 for the NH...N species, but a mere 3.3 kJ mol-1 for the NH...S one.Even more strikingly, whilst the cis-N isomer of monomeric la'.0=PH3 is favoured over the cis-S one (AE = 55.6 and 40.6 kJ 11101-1, respectively), this preference is reversed on forming the dimers, (la'.0=PH3)2: the dimerisation energies are 2.9 kJ mol-1 for the cis-N, NH...S bonded dimer and 45.0 kJ mol-1 for the cis-S, NH...N bonded one.The calculations thus rationalise the structural features of solid (la.HMPA)2, in that, overall, the cis-S, NH.-.N bonded dimer is favoured by 12.1 kJ mol-1.They predict that, on forming the solution monomer, the amine would rotate so that H(l) (cis to N), not H(2) (cis to S) [Fig. l(a)], is involved in NH.-.OP interactions.
They also estimate the energies of the two sorts of strong hydrogen bonding found in the solid: 42.3 kJ mol-' for the NH-m-OP interaction ( A E for the cis-S form of la.0=PH3, assuming no secondary interactions) and 22.5 kJ mol-1 for the NH...N bond [half the dimerisation energy for ( la'.0=PH3)2, formed from cis-S monomers].
The structure and stability (especially towards water) of the la-HMPA adduct suggest a model to help explain the carcinogenicity of HMPA.Amine l a has similarities to the natural bases along DNA: A , G and C have exo-NH2 groups and ring N and/or NH units, whilst T has ring NH units.Hydrogen bonds between base pairs (A-T, C-G), individually weak, but numerous, impart the high stability essential for preservation of genetic information.Clearly, strong and competitive complexation of HMPA to these amines would severely hinder replication of such information.More generally, the formation of such a strong adduct as la-HMPA reflects the dipolar (ylidic) nature of HMPA, whose 0-P bond is better described by 0--P+ than by O=P.We are investigating other ylide adducts with P+-E-...H-X interactions (E-= CH2-, NH-, 0-, S-and X = N, 0, S etc.).
We thank the SERC and the Associated Octel Co. Ltd (CASE award to M. G. D.), Gonville and Caius College, Cambridge (Research Fellowship for D. S. W.) and the DAAD, NATO Scholarship (D. S.) for financial support.the heterocyclic N (charge, -0.69; cf. that on S +0.25 in the cis-S isomer) of l a .To introduce this further bonding, the H...O-P angle is only 123.0"for the cis-N isomer (cf.161.7"for the cis-S one).For la'.0=P(NH2)3, more pronounced secondary interactions occur, and for both isomers (though more so for the cis-N one).For the czs-N species, AE = 72.4kJ mol-l, LH...O-P = 132.6",and the interaction involves one H (charge +0.48) of O=P(NH2)3 and the heterocyclic N (charge -0.67) of la'.For the cis-S species, AE = 53.6 kJ mol-l, H...O-P = 159.3",and the interaction occurs between one N (charge -0.97) of O=P(NH2)3 and the heterocyclic S (charge +0.22) of la'.
Seemingly in conflict with the above results, in solid (la.HMPA)* the NH..-O link is cis to S. However, this then leaves the ring N "(I), Fig. l (a)] free to engage in hydrogen bonding to give the dimer.One expects only a low dimerisa- Received, 23rd September 1991;Com. 1104901A Fig. I( a )  Basic dimeric \tructure of the 1: 1 adduct m ) N H I .O = P ( N M e l ) i , la-HMPA; ( b ) further intermolecular association of (IaeHMPA)?