Crystal structures of four thioglycosides involving carbamimidothioate groups

The structures of the four thioglycosides, all Z-configured across the C=N(CN) moiety, differ in many important torsion angles. The C—N bond lengths at the central carbon atom of the carbamimidothioate group are almost equal. Three of the four structures form layers by hydrogen bonding.

The compounds 2 0 ,3 0 ,4 0 ,6 0 -tetra-O-acetyl-�-d-glucopyranosyl N 0 -cyano-Nphenylcarbamimidothioate (C 22 H 25 N 3 O 9 S, 5a), 2 0 ,3 0 ,4 0 ,6 0 -tetra-O-acetyl-�-dgalactopyranosyl N 0 -cyano-N-phenylcarbamimidothioate, (C 22 H 25 N 3 O 9 S, 5b), 2 0 ,3 0 ,4 0 ,6 0 -tetra-O-acetyl-�-d-galactopyranosyl N 0 -cyano-N-methylcarbamimidothioate (C 17 H 23 N 3 O 9 S, 5c), and 2 0 ,3 0 ,4 0 ,6 0 -tetra-O-acetyl-�-d-galactopyranosyl N 0 -cyano-N-p-tolylcarbamimidothioate (C 23 H 27 N 3 O 9 S, 5d) all crystallize in P2 1 2 1 2 1 with Z = 4.For all four structures, the configuration across the central (formal) C N(CN) double bond of the carbamimidothioate group is Z.The torsion angles C5-O1-C1-S (standard sugar numbering) are all close to 180 � , confirming the � position of the substituent.Compound 5b involves an intramolecular hydrogen bond N-H� � �O1; in 5c this contact is the weaker branch of a three-centre interaction, whereas in 5a and 5d the H� � �O distances are much longer and do not represent significant interactions.The C-N bond lengths at the central carbon atom of the carbamimidothioate group are almost equal.All C-O-C O torsion angles of the acetyl groups correspond to a synperiplanar geometry, but otherwise all four molecules display a high degree of conformational flexibility, with many widely differing torsion angles for equivalent groups.In the crystal packing, 5a, 5c and 5d form layer structures involving the classical hydrogen bond N-H� � �N cyano and a variety of 'weak' hydrogen bonds C-H� � �O or C-H� � �S.The packing of 5b is almost featureless and involves a large number of borderline 'weak' hydrogen bonds.In an appendix, a potted history of wavelength preferences for structure determination is presented and it is recommended that, even for small organic crystals in non-centrosymmetric space groups, the use of Mo radiation should be considered.

Chemical context
Many synthetic nitrogen heterocycles are utilized in medicinal chemistry (Azzam et al., 2023;Elboshi et al., 2024).Several of these have played an important role in the search for potent antiviral drugs (Santos et al., 2021).As part of our program aimed at developing new, effective and straightforward procedures for the synthesis of antimetabolites (Elgemeie et al., 1998a(Elgemeie et al., ,b, 2015(Elgemeie et al., , 2022;;Mohamed-Ezzat et al., 2024), we have described several effective syntheses of folic acid, pyrimidine nucleoside and mercaptopurine analogues.One of these (Elgemeie et al., 2015) presented the synthesis and structure of a carbamimidothioate, namely methyl N 0 -cyano-N-(1,5-dimethyl-3-oxo-2-phenyl-2,3-dihydro-1H-pyrazol-4-yl)carbamimidothioate.Recently, the synthesis of nucleoside analogues and their integration into DNA sequences for the investigation of ligand-DNA and protein-DNA interactions has attracted increased attention (Dantsu et al., 2021).Numerous nucleoside derivatives that involve an alteration or removal of the functional groups of heterocyclic bases have been synthesized (Hammad et al., 2018;Masoud et al., 2017).The synthesis of oligodeoxynucleotides with a single functional group at a preselected position, involving various novel thioglycosides that demonstrate antagonistic activity, is made possible by such analogues (Pe ´rez-Rentero et al., 2012;Warren et al., 1998).The use of dihydropyridine thioglycosides as substrates or inhibitors of glycosylation of proteins was reported (Scala et al., 1997).These results have made the synthesis of modified and acyclic thioglycosides relevant in the quest for more potent agents (Elgemeie et al., 2017;Galmarini et al., 2003).
This work reports the one-pot synthesis of glycosyl isothiourea derivatives as a new class of acyclic thioglycosides, the N 0 -cyano-N-(alkyl or aryl)carbamimidothioates 5a-d.The potassium 1-cyano-isothiourea salts 3a-c were chosen as the key reagents.The sequences of reactions are summarized in Fig. 1.Cyanamide 1 was heated with substituted isothiocyanate derivatives 2a-c in KOH/EtOH to give the corresponding stable potassium N-substituted carbamimidothioates 3a-c.These salts reacted with 2,3,4,6-tetra-O-acetyl-�-dgluco-or galactopyranosyl bromides 4a,b in DMF at room temperature to afford the corresponding S-glycosides 5a-d in high yield.The structures of the compounds 5a-d were established by their elemental analyses and spectroscopic data (see Synthesis and crystallization).For example, the 1 H NMR spectra of 5a showed the anomeric proton as a doublet at � 5.82 ppm; the other six glucose protons resonated at � 4.05-5.39ppm and the four acetyl groups appeared as four singlets at � 1.92-2.05ppm.The structures of compounds 5a-d were unambiguously confirmed by single-crystal X-ray structure determinations, which are reported here.

Structural commentary
All four compounds 5a-d crystallize solvent-free in space group P2 1 2 1 2 1 with Z = 4.The molecular structures are shown in Figs.2-5, with selected molecular dimensions in Tables 1-4.For all four structures, the configuration across the formal double bond C15 N2 is Z, with the cyano group and the sulfur atom mutually cis, which avoids a steric 'collision' between the cyano and aryl groups (where present).The absolute configurations at C1-C5 are SRSSR for 5a and SRSRR for the other structures, the designations at C4 corresponding to the change of sugar from glucose in 5a to galactose in 5b-d.The torsion angles C5-O1-C1-S1 are all close to 180 � , confirming the � (equatorial) positions of the substituent at the sugar ring.The C1-S1 bond lengths are consistently longer than C15-S1, corresponding to the different hybridizations of the carbon atoms.Compound 5b involves an intramolecular hydrogen bond from the NH group to the sugar oxygen atom O1, with H01� � �O1 = 2.11 (2) A ˚; in 5c the longer H01� � �O1 distance, 2.52 (3) A ˚, represents the weaker branch of a three-centre interaction, whereas in 5a and 5d the H01� � �O1 distances are even longer at 2.84 (2) and 2.90 (2) A ˚, respectively, and the NH group is thus effectively only involved in intermolecular hydrogen bonds (see Supramolecular features).
The carbamimidothioate groups are consistently numbered as S1-C15 (-N1-Cxx) N2-C16 N3, where xx is 21 for the aryl substituents but 17 for the methyl substituent.The six atoms (excluding Cxx) are approximately coplanar, with r.m.s.deviations of 0.05, 0.03, 0.009 and 0.005 A ˚for 5a-d in that order; the interplanar angles to the aryl group are 20.04 (1), 47.89 (2) and 48.05 (4) � , respectively, for 5a, 5b and 5d.The small interplanar angle for 5a is associated with a short intramolecular H22� � �N2 contact of 2.34 A ˚.The sugar atom C1 lies 1.529 (1) A ˚out of the carbamimidothioate plane for 5a, 0.541 (1) A ˚for 5b, 0.148 (2) A ˚for 5c and 0.321 (2) A ˚for Figure 1 The reaction scheme for the syntheses of compounds 5a-d.5d, corresponding to a wide range of C1-S1-C15-N1 torsion angles.The bond lengths are broadly as expected, in accordance with the different hybridizations of Cxx.The near equality of bond lengths for C15-N1 (the aryl-or methylsubstituted nitrogen) and the formal double bond C15 N2 (the cyano-substituted nitrogen) indicate a considerable degree of delocalization in this region, as do the angle sums at N1 (359-360 � ) and the sp 2 angles at C15, N1 and N2.The C15-N1 bond is slightly longer (by ca 0.02 A ˚) than C15-N2 except for the N-methyl derivative 5c, where it is 0.07 A shorter.The glucose derivative 5a, however, shows some appreciable differences; thus the angle S1-C15-N1, 115.89 (4) � , is narrow, while C15-N1-C21 is very wide at 128.26 (5) � .Also, the angle at sulfur is appreciably narrower for 5a, 100.24 (3) � compared to a mean value of 104.7 � for 5bd.These differences, even between 5a and 5b, which have the same phenyl substituent at N1, can scarcely be attributed directly to the change of sugar, because the relevant atom C4 is quite remote from the affected carbamimidothioate group.Similarly, the strong intramolecular hydrogen bond in 5b is absent in 5a, but it is difficult to see how this would directly cause the observed differences.
A further explanation might be sought based on the torsion angles of the carbamimidothioate groups.For all four compounds, the torsion angles S1-C15-N1-Cxx correspond to an antiperiplanar geometry, and S1-C15-N2-C16 to synperiplanar.For the three galactose derivatives 5b-d, the torsion angles O1-C1-S1-C15 are roughly constant at Selected geometric parameters (A ˚, � ) for 5a.

Figure 2
The molecule of compound 5a in the crystal.Ellipsoids represent 50% probability levels.Only the major site of the disordered acetyl group at O3 (atoms C9, C10, O8) is shown.

Figure 3
The molecule of compound 5b in the crystal.Ellipsoids represent 50% probability levels.The dashed line indicates an intramolecular hydrogen bond.
about À 70 � , whereas for 5a the value is À 101.53 (4) � ; the groupings C1-S1-C15-N1 for 5b-d are roughly synperiplanar (torsion angles 7-17 � ), but the value for 5a is 57.56 (5) � .Finally, the torsion angles C15-N1-C21-C22 are widely different for 5a, 5c and 5d, corresponding to different rotations of the aryl ring.Despite the many degrees of torsional freedom, it seems that the unusual values for some bond angles of 5a may tentatively be connected with its lack of synperiplanarity for the region C1-S1-C15-N1.Without detailed calculations, however, this is difficult to prove (and takes no account of packing effects, see following section).
In the acetylated sugar moieties, all the C-O-C O torsion angles of the acetyl groups correspond to a synperiplanar geometry, but otherwise these too display a high degree of conformational flexibility.For the galactoses, the torsion angles C3-C4-O4-C11, C1-C2-O2-C7 and (to a lesser extent) C2-C3-O3-C9 remain reasonably constant, but there are large differences in C4-C5-C6-O6, which is À 167.08 (6) � for 5b, where the extended configuration is clear in Fig. 3, but À 63.41 (17) � for 5c and À 61.75 (13) � for 5d.The glucose derivative 5a necessarily deviates appreciably from 5b-d in torsion angles involving the region at C4, where the configuration is reversed.

Figure 5
The molecule of compound 5d in the crystal.Ellipsoids represent 50% probability levels.

Table 3
Figure 4 The molecule of compound 5c in the crystal.Ellipsoids represent 50% probability levels.The intramolecular contact H01� � �O1, not drawn explicitly, is the weaker branch of a three-centre hydrogen bond (see Supramolecular features).

Supramolecular features
All the compounds involve several potential hydrogen-bond acceptors A (nine oxygens, two nitrogens and the sulfur; the nitrile nitrogen atom is the acceptor for all three intermolecular hydrogen bonds, see below) but only one classical hydrogen-bond donor (the NH group).This means that several C-H� � �A 'weak' hydrogen bonds might be expected, and this is indeed the case.For completeness, the hydrogenbond tables (Tables 5-8) contain a number of borderline cases, not all of which are discussed.The packing diagrams are drawn to include only the shortest contacts, for the sake of clarity.The space group P2 1 2 1 2 1 is well known for its propensity to provide complex three-dimensional packing patterns if the interactions involve more than one 2 1 axis.
In compound 5a, the classical hydrogen bond N1-H01� � �N3 and the 'weak' but short interaction C1-H1� � �O10 combine via the 2 1 axis parallel to b to form a layer structure parallel to the ab plane (Fig. 6).Because the layer is quite thick, a side view in projection parallel to the a axis is shown as Fig. 7 as an aid to interpretation.
The packing of compound 5b involves no strikingly short (< 2.5 A ˚) H-A contacts; the NH group is involved in an intramolecular hydrogen bond (see above), and its intermolecular contact to O10, at 2.68 (2) A ˚and with an angle of only 101 (1) � at H01, can probably be neglected.We were unable to construct a clearly assimilable packing diagram, but Fig. 8 shows the pattern generated by the contacts H25� � �S1, H10C� � �O7, H3� � �N3 and H23� � �N3.
In compound 5c, the classical hydrogen bond H01� � �N3 combines with the four shortest 'weak' contacts to form a layer research communications Acta Cryst. (2024)

Figure 7
The layer from Fig. 6 is shown here in projection parallel to the a axis.
structure parallel to the ab plane (Fig. 9), and the same is true for compound 5d (Fig. 11).Again, side views of the layers, in projection, are shown as Figs. 10 and 12, respectively.

Database survey
The searches employed the routine ConQuest (Bruno et al., 2002), part of Version 2024.1.0 of the Cambridge Structural Database (Groom et al., 2016).A search for the acyclic residue   The layer from Fig. 9 is shown here in projection parallel to the a axis.

Figure 12
The layer from Fig. 11 is shown here in projection parallel to the a axis..311-1.349, av. 1.327 (13).This corresponds reasonably well to our values of 1. 7593-1.7759, av. 1.7646; 1.3084-1.323, av. 1.3305; and 1.316-1.3382, av. 1.3155A ˚.The six atoms of the carbamimidothioate group are essentially coplanar in all these structures (maximum mean deviation 0.039 A ˚), but in two cases (OWAHOK and XAZKIT) the methyl group at sulfur is rotated out of the plane (by 0.84 A ˚).

General procedure for the synthesis of 5a-d
The reaction scheme is given in Fig. 1.Cyanamide 1 (0.42 g, 0.01 mol) was added to a cold solution of potassium hydroxide (0.56 g, 0.01 mol) in absolute ethanol (20 mL) and the mixture was stirred for 10 min.The appropriate substituted isothiocyanate derivative (2a, 2b or 2c; 0.01 mol), was then added gradually over a period of 15 min and the mixture was stirred at room temperature for 4 h, after which the reaction was complete.The solvent was evaporated under reduced pressure, and the residue was dissolved in DMF (15 mL).A solution of 2,3,4,6-tetra-O-acetyl-�-d-gluco-or galacto-pyranosyl bromide 4a or 4b (4.2 g, 0.01 mol) in DMF was then added dropwise over 30 min.Stirring was continued at room temperature until the reaction was judged complete by thinlayer chromatography (6-8 h).The mixture was poured into ice-water, and the resulting precipitate was collected by filtration, dried, and crystallized from ethanol to give compounds 5a-d.

Refinement
Crystal data, data collection and structure refinement details are summarized in Table 9.
The hydrogen atoms of the NH groups were refined freely.The methyl groups were included as idealized rigid groups allowed to rotate but not tip (command 'AFIX 137'), with C-H = 0.99 A ˚and H-C-H = 109.5� .Other hydrogen atoms were included using a riding model starting from calculated positions (C-H methylene = 0.99, C-H methine = 1.00,C-H arom = 0.95 A ˚).The U(H) values were fixed at 1.5 � U eq of the parent carbon atoms for the methyl group and 1.2 � U eq for other hydrogens.
For compound 5a, the acetyl group at O3 (atoms C9, C10, O8) was disordered over two positions.The occupation factor of the minor component refined to 0.085 (2).Appropriate restraints were employed to improve refinement stability, but the dimensions of disordered groups (and particularly the minor components) should always be interpreted with caution.In the discussion sections above, the minor component is not considered.

Appendix: The choice of radiation type for X-ray measurements
The large and well-formed crystals of compounds 5a and 5b were clearly suitable for measurements using Mo K� radiation; 5c consisted of smaller crystals, and one of these was measured with Cu radiation.For compound 5d, we originally recorded a dataset using Cu K� radiation; data for this are given in Table 9 in the final column '5d(Cu)'.The reasons for this, and for preferring the Mo dataset (measured later with the same crystal), will be discussed here (together with a potted history of the fashions in X-ray wavelengths over the last 50 years, as experienced by PGJ) as they may be of general interest, in particular to younger crystallographers.One central criterion is the ability to determine absolute configuration (more generally 'absolute structure', see below) and  Orthorhombic, P2 1 2 1 2 1  Orthorhombic, P2 1 2 1 2 1  Orthorhombic, P2 1 2 1 2 1  Orthorhombic, P2 1 2 1 2 1  Orthorhombic, P2 1 2 1 2   Computer programs: CrysAlis PRO (Rigaku OD, 2023), SHELXT (Sheldrick, 2015a), SHELXL2019/3 (Sheldrick, 2015b) and XP, (Bruker, 1998).
the other, connected with this, is the need to collect data of adequate intensity.
The first automated four-circle diffractometers were introduced in the early 1970s, and the institute where I was working acquired such a diffractometer (shared by both chemistry departments) in 1974; the tubes were changed from copper (henceforth Cu) to molybdenum (henceforth Mo) radiation or vice versa every six months.These were the only types of X-ray tube generally available at the time.As an X-ray beginner, I was told that Cu radiation was used for organic structures and Mo radiation for inorganic structures.As I soon realised, this is an oversimplification; a more accurate formulation would be that Cu radiation is used for crystals that diffract less strongly and Mo for those that diffract more strongly, because Cu radiation has a higher intrinsic intensity (main beam intensities were much weaker then; the nominal diameter of the main beam for Mo measurements was ca 0.7 mm, and the chosen crystals were often of this size, unless they were highly absorbing materials, in order to maximize the measured intensities).This was brought home to me by a typical beginner's mistake, my unwise measurement of a large crystal of the 'organic' compound sodium acetylphosphonate acetic acid solvate using Cu radiation; the diffraction was extremely strong, but the absorption and (particularly) extinction effects were so pronounced that the structure was unusable, and had to be repeated using Mo radiation (Jones & Kennard, 1978), which, with its shorter wavelength, is absorbed less strongly.Like many oversimplifications, the assumption that inorganic materials (including metal complexes) diffract more strongly than organic materials has some validity; the scattering power of a crystal of a given size will depend on the number of electrons in the crystal, which in turn depends on the density, and densities are generally greater for inorganic materials (a second-order effect is that their U values tend to be lower).
One important reason for using Mo radiation is to reduce absorption effects, because absorption corrections at the time largely relied on face-indexing the crystals, a procedure that was often difficult or impossible.The quality of datasets increased significantly when the first generalized absorption correction methods became available; this major step was provided by Flack (1974), who introduced the -scan method to diffractometry.Nowadays, the highly redundant datasets (see below) have made the 'multi-scan' (Blessing, 1995) the method of choice.The use of Cu radiation decreased drastically during and after the 1970s, which is partly attributable to the more serious absorption effects, but also to the limited amount of data that could be measured; a complete sphere of Cu reflections was inaccessible because of the restricted geometry of bulky diffractometer components (typically, measurements above 2� = 120 � were difficult to obtain) and even a complete sphere of Cu data to the theoretical limit of 180 � would only correspond to 2� = 55 � using Mo radiation.Routine measurements were usually conducted at room temperature with Mo radiation and 2� max = 50 � ; above this value, significant intensity was difficult to detect for many organic and organometallic samples.
The other connection between Cu radiation and organic crystals is associated with enantiomerically pure materials, of which many natural products constitute an important subclass; the determination of absolute configuration, often of great importance for these materials, relies on the measurement of generally small intensity differences between Friedel opposite reflections hkl and À h, À k, À l, caused by the phenomenon 'anomalous dispersion' (perhaps not a wellchosen name, because physicists tell us that there is nothing anomalous about it); the longer the wavelength, the more pronounced are these effects.The correct absolute configuration should then give a better R value than the incorrect, inverted, structure.Measurable differences at the time could only be detected in the presence of heavy atoms (as a rule of thumb, elements of the fourth or higher periods), and the effects were generally unobservable for light-atom structures, so that it was often necessary to synthesize heavy-atom derivatives of natural products in order to determine their absolute configuration.It should also not be forgotten that direct methods in the early 1970s were still in their infancy and often unreliable (especially for weak datasets), and a heavy-atom derivative was often needed to solve the structure in the first place (by the Patterson method).
As an example of an old structure determination of a natural product, the peptide l-serylglycine (Jones et al., 1978) shows the standard practices of the time.The structure was measured using Cu radiation (for a week) to a 2� max value of only 116 � .A total of 1186 intensities were measured, giving 713 unique data > 4�(F); weak reflections were omitted from the measurement.Friedel opposites were presumably not merged, but the data are no longer directly available.The number of parameters is not given, but must have been roughly 125.There is no mention of the absolute configuration.From a modern viewpoint, albeit nearly 50 years on and with hindsight, this seems embarrassingly lackadaisical.
In general, any non-centrosymmetric structure must be compared with the corresponding inverted structure to ensure that the structure is correctly refined; the procedure is not confined to the space groups adopted by enantiomerically pure materials (the 'Sohncke' space groups such as P2 1 2 1 2 1 ; these were often informally and incorrectly called the 'chiral' space groups, but this name should now be used for the space groups that occur in pairs with opposite sense of the screw axis, such as P3 1 and P3 2 -formerly known as 'enantiomorphic' space groups).The general procedure is now normally referred to as the 'determination of absolute structure', as suggested by Jones (1984a), although the use of the word 'absolute' has correctly been questioned by Glazer & Stadnicka (1989).
For datasets where the determination of absolute structure was not expected to succeed, some bad habits were common (Jones, 1984b(Jones, , 1986)); either the Friedel opposite reflections À h, À k, À l were not measured at all (because the space group, often determined only after data collection, was assumed to be centrosymmetric, or to save diffractometer time at a time when measurements were very slow by today's standards), or the Friedel pairs hkl and À h, À k, À l were considered exactly equivalent and were merged (in SHELX using the command 'MERG 3', which is now effectively banned).
The determination of absolute structure/configuration relies on the existence of a method to test which configuration gives a significantly better fit.The 'Hamilton R method' (Hamilton, 1965) was the first statistical test to be generally used, but the results were capable of misinterpretation.The first significant improvement was made by Rogers (1981; see also Jones, 1984a), by refining a factor � that multiplied the anomalous scattering parameters f".The correct structure should then give an � value of +1 and the incorrect (inverted) structure a value of À 1.This method gave a standard deviation for �, so that the reliability of the determination could also be judged.The next improvement was introduced by Flack (1983; reviewed by Watkin & Cooper, 2020), using the parameter x to estimate the extent of inversion twinning, whereby both the parent structure and the inverted structure are present in the same crystal.This had mathematical advantages over the � method and became the accepted method of determining absolute configuration; a value of x = 0 indicated the correct structure and 1 the incorrect (inverted) structure.The latest improvement was provided by Parsons et al. (2013; see also Parsons, 2017), who used the quotients [(I + ) À (I À )]/[(I + ) + (I À )] (where I + is the intensity of hkl and I À the intensity of À h, À k, À l) to improve the sensitivity with which the x parameter could be determined.The excellent review article by Linden (2017) on the determination of absolute structure was published just before the Parsons method became generally known.
In the 1990s, the use of area detectors increased the speed and precision of intensity measurements.Whereas the older 'serial' diffractometers measured one reflection at a time, and a dataset, even if consisting of only the independent data, took days or weeks to record, it was now possible to measure tens or hundreds of reflections per exposure ('frame').Datasets typically consist of hundreds of frames, whereby each reflection is measured many times; the redundancy leads to a statistical improvement in data precision (by merging many equivalents of each reflection) and enables the 'multi-scan' absorption correction.Continuous improvements in detector sensitivity and source intensity have brought measurement times down to hours rather than days.
The second, vital, development in the 1990s was the development of routine measurements at low temperature (without the restriction of severe icing problems).This is largely attributable to the efforts of Stalke.The advantages are well-known: the most important is the reduction of thermal motion, which in turn reduces the U values and thereby leads to an increase in the number and intensity of reflections that are available at higher angle (see e.g.Kottke & Stalke, 1993, and references therein).In our opinion, any X-ray structure determination at room temperature (in the absence of extenuating circumstances such as phase changes at low temperature) represents a missed opportunity to collect good data.A recent issue of Acta Cryst.E contained 21 lowtemperature and 16 room-temperature structures.
The third major change was the use of refinements based on F 2 rather than F, and using all data, including the weak reflections; this was introduced into the SHELX program system in the 1990s (Sheldrick, 2008(Sheldrick, , 2015a,b),b).It is fitting to pay tribute here to George Sheldrick, who has developed and maintained SHELX for some 50 years, and has always been quick to incorporate the newest developments (e.g. the Parsons method).
In the 2000s, the introduction of microsources, with typical beam diameters of 0.1-0.2mm, for Cu radiation appreciably increased the available intensity.This had two important consequences.First, structures from weakly diffracting organic crystals with average dimensions as small as 10-50 mm, previously considered unmeasurable, could now be successfully determined (e.g.Abu-Zaied et al., 2024); secondly, the anomalous scattering of oxygen atoms, previously regarded as negligible, was now often sufficient to determine the absolute structure reliably.The less bulky detectors and a favourable modified kappa geometry also meant that data could be collected to much higher angles (currently 2� max ' 160 � ).This led to a renaissance in the use of Cu radiation.
It was first recognized by Escudero-Ada ´n et al. (2014) that the absolute configuration of light-atom structures could be determined reliably even using Mo radiation, if high-energy sources were used and the datasets were recorded at low temperature to higher (by the standards of the time) diffraction angles 2�, typically 55 � .The reason is that the anomalous scattering is approximately independent of 2�, whereas the normal scattering decreases with increasing 2�, so that the contribution of the anomalous scattering becomes more pronounced at high angles.The problem was that few lightatom structures diffracted to sufficiently high angles, but the Parsons method has made matters easier in this respect; thus we found that a steroid derivative, containing four oxygen atoms as anomalous scatterers, measured by us using a standard Mo source to 2� = 61 � (cholest-5-en-3-yl 3-formylphenyl carbonate, C 35 H 50 O 4 ; refcode LUCVOX; Jones & Kus ´, 2020), gave a correct (known) absolute configuration with x = 0.15 (16), whereas the x value without the Parsons modification had been indeterminate.The development of Mo microsources has helped further; with these, light-atom crystals can diffract significantly to 80 � or more, and the absolute structure can then often be determined successfully with Mo radiation even for light-atom structures.The improvements in detector sensitivity have also made an important contribution.
Returning finally to the two datasets measured for 5d, we originally thought that the very small crystals would need to be measured using Cu radiation.Although the absolute configuration of this galactose derivative is known, a confirmation using X-ray methods is always welcome.The crystal diffracted so strongly with Cu radiation, however, that we decided to re-measure the same crystal using Mo radiation.For practical purposes, both measurements were designed to run until the following day; the measurement times were ca 6 h for the Cu dataset and 22 h for the Mo dataset.Both datasets could certainly have been measured significantly faster, had it been necessary.The diffraction pattern for Mo extended to (at least) 2� = 72 � (in our experience, the intensity statistics of the data reduction often indicate that significant intensity is still present at angles where no maxima can be recognised in the frames), and the Flack x parameter is unambiguous at 0.005 ( 18).[Of course, the presence of sulfur, which nowadays counts as a 'heavy' atom, greatly facilitates the determination of the x parameter; this is usually no problem with elements of the third period.A good example of a structure with no atom heavier than oxygen is l-arabinose, which we measured to 2� = 157 � using Mo radiation, and which gave an x value of 0.03 (11) (refcode ABINOS04; Jones, 2023)].The number of independent intensities is doubled compared to the Cu data (11008, cf.Cu 5508), so that the s.u.'s of molecular dimensions are somewhat lower (by a factor of approximately ffi ffi ffi 2 p , as would be expected if other things are equal).Furthermore, despite the generally effective absorption corrections that are now employed, it should not be forgotten that absorption effects are lower with Mo radiation.We therefore prefer the Mo dataset for 5d, and would indeed recommend that, even for small organic crystals, the use of Mo radiation should not be dismissed out of hand.Both datasets are included here and are thus available to the interested reader.

Special details
Geometry.The symmetry employed for this shelxl refinement is uniquely defined by the following loop, which should always be used as a source of symmetry information in preference to the above space-group names.They are only intended as comments.Refinement.The symmetry employed for this shelxl refinement is uniquely defined by the following loop, which should always be used as a source of symmetry information in preference to the above space-group names.They are only intended as comments.

Special details
Geometry.The symmetry employed for this shelxl refinement is uniquely defined by the following loop, which should always be used as a source of symmetry information in preference to the above space-group names.They are only intended as comments.

Special details
Geometry.The symmetry employed for this shelxl refinement is uniquely defined by the following loop, which should always be used as a source of symmetry information in preference to the above space-group names.They are only intended as comments.

Special details
Geometry.All esds (except the esd in the dihedral angle between two l.s.planes) are estimated using the full covariance matrix.The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry.An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s.planes.

Figure 6
Figure 6 Packing diagram of compound 5a viewed parallel to the c axis, showing one layer in the region z ' 0.25.Dashed lines indicate the hydrogen bonds H01� � �N3 (thick) and H1� � �O10 (thin).Hydrogen atoms not involved in hydrogen bonding are omitted for clarity.Two atoms are labelled to indicate the asymmetric unit.

CFigure 8
Figure 8 Packing diagram of compound 5b viewed parallel to the a axis, showing the region x ' 0.75.Thin dashed lines indicate 'weak' hydrogen bonds.Hydrogen atoms not involved in hydrogen bonding are omitted for clarity.Two atoms are labelled to indicate the asymmetric unit.

Figure 9
Figure 9 Packing diagram of compound 5c viewed parallel to the c axis, showing one layer in the region z ' 0.25.Dashed lines indicate the hydrogen bonds H01� � �N3 (thick) and four H� � �O (thin).The longer contact H1� � �O10 is also present in this layer, but is omitted for clarity, as are hydrogen atoms not involved in hydrogen bonding.Two atoms are labelled to indicate the asymmetric unit.

Figure 11
Figure 11 Packing diagram of compound 5d viewed parallel to the c axis, showing one layer in the region z ' 0.75.Dashed lines indicate the hydrogen bonds H01� � �N3 (thick) and four H� � �O (thin).Hydrogen atoms not involved in hydrogen bonding are omitted for clarity.Two atoms are labelled to indicate the asymmetric unit.

Table 9
Experimental details.