328 Laurens, Ichharam, and Modro
˚
which was expected to yield the thio-analogue of the
noncyclic triamidate B, failed to give satisfactory re-
sults. Under mild conditions (four mol equivalents
of PhNH2, CH2Cl2, מ
40ЊC–room temperature) no
significant reaction progress was observed. Change
of the reagents and conditions (CH2Cl2/reflux, two
mol equivalents of Et3N/THF, aq NaOH/CCl4, BuLi/
THF/מ
78ЊC) resulted in some reactions, but the
crude product always consisted of a complex mix-
ture of phosphorus-containing compounds (31P
NMR). The attempt to substitute oxygen for sulfur
by treating 1a (1, R ס
Ph) with P(S)Cl3/DMF [3] re-
sulted also in a mixture of products from which 2a
could not be isolated. A solution to the problem was
derived from our most recent results [4], according
to which the previously difficult to prepare bis(2-ar-
ylaminoethyl)amines (3) were obtained by exhaus-
tive hydrolysis of triamidates 1. The triamine 3a (3,
R ס
Ph), obtained in that way, when treated with
P(S)Cl3 in the presence of base, was successfully con-
verted to the desired bicyclic product 2a. The con-
version 1a r 2a is presented in Scheme 2. Pure 2a
was obtained as a colorless, highly crystalline ma-
terial, suitable for X-ray analysis; the structure of 2a
is shown in Figure 1. Since the X-ray structure of 1a
had been determined before [1], it was now possible
to compare the molecular parameters of those two,
closely related bicyclic structures. Selected bond dis-
tances and angles for both compounds are listed in
Table 1. It is clear that all parameters that describe
the geometry of the 2,5,8-triaza-1k5-phosphabicy-
clo[3.3.0]octane system are in both cases approxi-
mately the same. The bonding of the thioamidate
function does not show any unusual features, with
A) is also very typical; according to the Cambridge
Structural Data Base [6], the average value of the
Pס
S bond distance in 74 listed compounds of the
˚
(N,NЈ,NЉ)Pס
S type is 1.929 A. Compounds 1a and
2a represent therefore similar geometry, and any dif-
ferences in chemical reactivity should therefore re-
flect the difference between the phosphoryl and thio-
phosphoryl center, and not the different geometries
of the bicyclic skeleton.
The first reaction studied for 2a was the acid-
catalyzed solvolysis of the amide bond. For 1a, the
reaction with ROH/H (R ס
Me, Et) led to a selective
cleavage of the P–N (bridgehead) bond giving the
eight-membered cyclic diamidoester 4a as the ki-
netic product, which underwent spontaneous rear-
rangement to the isomeric 1,3,2-diazaphospholidine
5a (thermodynamic product) [7]. The same se-
quence of reactions was therefore expected to occur
for 2a (Scheme 3). The results obtained for 2a as a
substrate were, however, very different for those re-
ported earlier for 1a [7], and, until now, we were not
able to prepare either of the corresponding thio de-
rivatives 4b or 5b. The conversion of 2a was com-
plete (31P NMR), and a single phosphorus-containing
product (dP ס
76, as opposed to dP ס
81 for 2a) was
formed. The attempts to isolate the product and to
assign its structure were unsuccessful since it de-
composed into a mixture of phosphorus-containing
compounds. Gas chromatography–mass spectrome-
try (GC-MS) analysis of the crude reaction product
gave, however, interesting results, worth reporting.
Chromatographic separation yielded two compo-
nents in an approximately 1:1 ratio. The second frac-
tion (retention time 48.5 minutes) was unambi-
gously identified as the phospholidine derivative 5b,
presumably formed via the solvolysis of 2a followed
by a rearrangement. The MS of that fraction gave the
˚
the average P–N bond distance of 1.674 A, a typical
value reported [5] for the amides of phosphoric acid.
The length of the thiophosphoryl bond (Pס
S, 1.926
expected peak of m/z ס
347 (M , 51%), together with
SCHEME 2
FIGURE 1 ORTEP drawing of 2a.