Solution and Solid-State Effects on NMR Chemical Shifts in

Dec 6, 2011 - Solution and Solid-State Effects on NMR Chemical Shifts in Sesquiterpene Lactones: NMR, X-ray, and Theoretical Methods. Martin Dračíns...
1 downloads 9 Views 3MB Size
ARTICLE pubs.acs.org/JPCA

Solution and Solid-State Effects on NMR Chemical Shifts in Sesquiterpene Lactones: NMR, X-ray, and Theoretical Methods Martin Dracínsky ,† Milos Budesínsky,*,† Beata War_zajtis,‡ and Urszula Rychlewska‡ † ‡

Institute of Organic Chemistry and Biochemistry, Flemingovo nam. 2, 16610 Prague, Czech Republic Faculty of Chemistry, Adam Mickiewicz University, 60-780 Poznan, Poland

bS Supporting Information ABSTRACT: Selected guaianolide type sesquiterpene lactones were studied combining solution and solid-state NMR spectroscopy with theoretical calculations of the chemical shifts in both environments and with the X-ray data. The experimental 1 H and 13C chemical shifts in solution were successfully reproduced by theoretical calculations (with the GIAO method and DFT B3LYP 6-31++G**) after geometry optimization (DFT B3LYP 6-31 G**) in vacuum. The GIPAW method was used for calculations of solid-state 13C chemical shifts. The studied cases involved two polymorphs of helenalin, two pseudopolymorphs of 6α-hydroxydihydro-aromaticin and two cases of multiple asymmetric units in crystals: one in which the symmetry-independent molecules were connected by a series of hydrogen bonds (geigerinin) and the other in which the symmetry-independent molecules, deprived of any specific intermolecular interactions, differed in the conformation of the side chain (badkhysin). Geometrically different molecules present in the crystal lattices could be easily distinguished in the solid-state NMR spectra. Moreover, the experimental differences in the 13C chemical shifts corresponding to nuclei in different polymorphs or in geometrically different molecules were nicely reproduced with the GIPAW calculations.

’ INTRODUCTION Sesquiterpene lactones are a large group of natural compounds isolated mostly from various plant materials. They are interesting for their various biological activities and also as markers in chemotaxonomy (for a review, see ref.1). The large number of skeletal types and the presence of chiral centers as well as variable substituents make sesquiterpene lactones attractive for structural studies. The structures of sesquiterpene lactones were established mainly using a detailed analysis of the experimental NMR spectra in solution (1H and 13C chemical shifts, proton coupling constants and NOE) or, less frequently, by an X-ray structure analysis of their crystals. A review of the 13C NMR data of sesquiterpene lactones was published several years ago.2 More recently, progress in quantum-chemical calculations has opened new possibilities for theoretical calculations of conformational isomers and NMR parameters, and in favorable cases, the predicted models agree extremely well with the existing experimental data (e.g. refs 35). The dependence of chemical shifts on the solvents is well-known. The solvent effect is caused by the interactions of the solvent molecules with the solute, for example, through hydrogen bonds, van der Waals forces, or other nonbonded interactions. For NMR experiments, accompanying ab initio calculations have become standard for isolated molecules6,7 and are also becoming increasingly popular for the solid state8 as well as for liquids and solutions. We have previously explored different approaches for modeling the solvent r 2011 American Chemical Society

effects in the calculation of NMR parameters. We have shown that averaging a large number of molecular dynamic geometries is essential for reasonable accuracy of the calculations. Implicit solvent models (polarizable continuum models, PCM) completely fail in the calculation of solvent effects on NMR parameters.9,10 In the solid phase, bulk chemical shifts reflect a change in the NMR resonance owing to the compact arrangement of the molecules, in full analogy with the solvent effects. Since the packing forces are more intense in the solid phase, the spectroscopic signatures are usually stronger than in solutions. Several modeling and simulation techniques have been proposed to describe the influence of intermolecular interactions on chemical shifts. The cluster approach and finite-charge models are commonly used.11,12 In the cluster model, neighboring molecules or fragments are considered explicitly during the chemical shielding calculations. Modeling a solid as a “large molecule” or cluster has many difficulties. While NMR spectroscopy is generally sensitive mostly to the local environment, there are cases where the longrange effects are significant, for example, electrostatic effects and ring currents. The choice of the cluster, in particular its termination, is critical, as the calculations must be maintained at a manageable size.13 It is thus more efficient to exploit the translational repetition in crystals. In the past decade, the gauge-including Received: September 29, 2011 Revised: November 24, 2011 Published: December 06, 2011 680

dx.doi.org/10.1021/jp209408b | J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

Figure 1. Stereochemical formulas of sesquiterpene lactones discussed in the paper and the atom numbering scheme.

projector-augmented wave (GIPAW) procedure has been developed for the prediction of the magnetic resonance parameters.14 In contrast to the traditional molecule-based methods, plane waves are used to form a basis set. The translational symmetry of the system is ensured, but pseudopotentials must be used to represent the interactions around the nuclei. The pseudopotentials permit both the removal of the core electrons from the calculation and the smoothing of the valence wave functions near the nuclei. The power of the GIPAW approach for calculating NMR properties for fully periodic crystal structures, specifically in this context of organic solids, has been well documented.12,1523 It is quite common for crystalline solids to display the ability to exist in more than one crystalline arrangement. This behavior is called polymorphism if the different crystalline forms have the same chemical and molecular composition, or pseudopolymorphism, if the different crystalline forms include solvates or hydrates of the material. The latter term, although commonly used, has been questioned by some members of the crystallographic community.24 Another nonstandard option that the organic substance can exercise during the crystallization process is the inclusion of more than one molecule in the asymmetric part of the unit cell (Z0 > 1).25,26 Previous studies have shown that chiral molecules, in particular natural products, to which the investigated group of compounds belongs, are more likely to form high Z0 structures.27 Apart from diffraction methods, the 13 C cross-polarization (CP) magic-angle spinning (MAS) NMR is also an efficient method for determining Z0 by means of comparing the number of observed resonances with the number of nonequivalent carbon atoms present in the molecule.28 For our study, we have chosen five sesquiterpene lactones from the guaianolide family, 15 (Figure 1), that represent different packing modes in crystals. Mexicanin (1) was the standard case with just one symmetrically independent molecule in the unit cell (Z0 = 1). Helenalin (2) was studied in its two polymorphic forms (monoclinic and orthorhombic) while 6αhydroxydihydroaromaticin (3) was investigated in the hydrated and nonhydrated crystalline form. The remaining two compounds, geigerinin (4) and badkhysin (5), represent the case of Z0 = 2 with the two independent molecules differing either in their involvement in the hydrogen-bond formation or in molecular geometry. In this paper, we shall demonstrate that, unlike the NMR spectra in solution, which lead to the observation of only one set of signals, the solid state NMR spectra of crystalline material clearly reflect the fine structural differences described above and, moreover, that quantum-chemical calculations allow the reproduction of observed NMR spectra in both solution and the solid state.

20 years ago in the Department of Medical Plants, Karol Marcinkowski University of Medical Sciences, Pozna n, Poland, from the plants of Telekia speciosa (3) and Helenium aromaticum (4). The crystals of 3 3 H2O (mp 116119 °C) were grown from butanol while those of 4 (mp 202203 °C) were grown from propanol. The solid-state 13C NMR spectrum of one of the five samples of 3 that were at our disposal clearly showed the doubling of signals for each carbon atom obviously owing to the presence of two different crystal forms of this compound in the 1:1 ratio, one of which corresponded to the crystal structure obtained from butanol. Close examination under the optical microscope of this sample revealed two distinct crystal morphologies, which confirmed the preliminary NMR findings. X-ray measurements were then performed for both types of crystals. At the structure refinement stage, it became apparent that one set of crystals represented the hydrated (mp 116119 °C) and the other nonhydrated (mp 150 °C) crystals of 3. The diffraction data for both crystal forms of 3 as well as for 4 were collected at room temperature on a SuperNova diffractometer equipped with a Cu microfocus source (λ = 1.5405 Å) and 135 mm Atlas CCD detector. The data reduction and analysis were carried out with the CrysAlisPro program.29 The intensity data were corrected for the Lorentz and polarization effects. The multiscan correction for absorption was also applied.30 The crystal data and experimental details are summarized in Table S1. The structures were solved by direct methods and refined on F2 using the full-matrix leastsquares method. The anisotropic displacement parameters were refined for all of the non-hydrogen atom positions. The hydrogen atoms attached to the carbon atoms were placed in calculated positions (methyl CH = 0.96, methylene CH = 0.97, and methine CH = 0.98 Å) while those attached to hydroxyl and water oxygen atoms were located on difference Fourier maps and included in the refinement with the standardized oxygen hydrogen distances of 0.85 Å. During the refinement, isotropic displacement parameters for H-atoms were used which were fixed at 1.5  Ueq (O) and 1.2  Ueq (C). All of the H-atoms were refined as riding. The absolute structure was unequivocally determined using the Flack parameter.31 NMR Spectroscopic Details. The liquid-state one-dimensional 1H and 13C spectra, two-dimensional PFG (pulsed-field gradient) 1H1H COSY, 1H13C HSQC (heteronuclear singlequantum correlation), and HMBC (heteronuclear multiplebond correlation) NMR spectra were recorded in CDCl3, CD3OD, and CD3COCD3 at 27 °C with a Bruker Avance II 600 spectrometer. This instrument was equipped with a 5-mm diameter cryoprobe operating at 600.1 MHz in the 1H and 150.9 MHz in the 13C. The following referencing of spectra was used: in CDCl3 (1H to TMS; 13C to CDCl3 at 77.0 ppm), in CD3OD (1H and 13C to the solvent peak at 3.31 ppm and 49.0 ppm, respectively), and in CD3COCD3 (1H and 13C to solvent peak at 2.05 ppm and 29.84 ppm, respectively).

’ METHODS X-ray Diffraction Analysis. The samples for X-ray analysis have been isolated by the late El_zbieta Bzoszyk approximately 681

dx.doi.org/10.1021/jp209408b |J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

Table 1. Average Ranges of the 1H and 13C Chemical Shifts (Δav) of a Given Atom of Lactones 15 in the Three Solvents Used

The solid-state NMR spectra were recorded with a Bruker Avance II 500 spectrometer operating at 499.8 MHz in the 1H and 125.7 MHz in the 13C experiments by using 3.2 mm rotors. The chemical shifts were referenced to crystalline α-glycine as a secondary reference (δst = 176 ppm for carbonyl carbon). The ramped amplitude shape pulse was used during the crosspolarization. The contact time in the basic 13C CPMAS experiments was 2 ms, the relaxation delay was 4 s, and the spinning rate was 12 kHz. In the nonquaternary suppression (NQS)32 experiment, the dephasing delay was 48 μs. In the short CP (cross-polarization) experiments, the contact time was 40 μs. Computational Details. The 1H and 13C chemical shifts were calculated with the DFT B3LYP 6-31++G** method after geometry optimization in vacuum (unconstrained optimization of the starting X-ray geometry using DFT B3LYP 6-31G**). The chemical shifts were determined using the shielding constants of TMS (1H 31.69 and 13C 192.76) calculated at the same level of theory as the sesquiterpene lactones. The 13C chemical shifts of the crystalline samples of sesquiterpene lactones were calculated using the following procedures. The atomic coordinates for mexicanin I (1), two polymorphs of helenalin (2), and badkhysin (5) were derived from the Cambridge Crystallographic Database33 (refcodes SEZLUE, FEGWAP, FEGWAP02, and KOBYUV, respectively) while those for 6α-hydroxyaromaticin (3) and geigerinin (4) were obtained from the newly determined crystal structures. The shielding values of the infinite crystals were calculated by using the CASTEP program,34 which is a DFT-based code. The calculations have been performed for an (1) X-ray structure without any optimization, (2) X-ray structure where the positions of the hydrogen atoms were adjusted in the process of constrained geometry optimization, (3) X-ray structure where the positions of all the atoms were adjusted in the process of full geometry optimization. The electron-correlation effects were modeled using the generalized gradient approximation of Perdew, Burke, and Ernzerhof.35 For the geometry optimization, we employed “ultrasoft” pseudopotentials,36 a planewave cutoff energy of 340 eV with integrals taken over the Brillouin zone using a Monkhorst Pack37 grid of a minimum sample spacing of 0.07 Å1. The NMR spectroscopic calculations were performed using the gaugeincluding projector-augmented wave approach (GIPAW)14 at the larger cutoff of 550 eV, also using CASTEP. 38 The lattice volume was fixed to the experimental value, so that the intermolecular interactions were effectively taken into account. The calculated chemical shift values were obtained from the calculated shieldings using the following relation: δcalc = (σcalc  172.84)/1.0736.

1

2

3

4

5

Δav ( H)

0.10

0.08

0.08

0.08

0.08

Δav (13C)

1.52

1.58

1.55

1.27

1.41

1

for example, the carbonyl carbons in methanol were less shielded than in chloroform and in acetone more shielded than in chloroform. The average ranges of 1H and 13C chemical shifts in three used solvents (Δav) are given in Table 1. The values Δav for the individual lactones 15 are very similar (below 0.1 ppm for 1H and about 1.5 ppm for 13C) and smaller than the estimated accuracy of the chemical shifts calculated by modern quantumchemical methods.39 The solvent effect on the observed proton coupling constants (not shown) is in general very small (the largest differences are smaller than 1 Hz but typically less than 0.3 Hz), indicating that minimal conformational changes have been induced by the solvent. The simulation of the solvent effect in the quantum-chemical calculation of the NMR parameters is not yet sufficiently solved, and it is known that simple methods with nonexplicit solvent models (like PCM) do not provide realistic data. Considering also the small experimental differences in the chemical shifts of compounds 15 in the various solvents, we have decided to disregard the solvent effect completely and to probe the utilization of chemical shifts calculated in vacuum (see Supporting Information, Tables S4S8). The chemical shifts calculated in vacuum were then compared with the experimental data obtained for three different solvents, i.e., CDCl3, CD3OD, and CD3COCD3. The agreement of the calculated chemical shifts with the experimental ones was generally very good. The characteristic parameters of the correlation of the calculated and observed chemical shifts (resulting from a linear regression analysis) are given in Table S9 in Supporting Information. The correlation of the calculated 1H and 13C chemical shifts with the experimental data in the most commonly used solvent CDCl3 for 15 is shown in Figure 2. The application of linear regression yields parameters that can be used for the correction of the calculated chemical shifts using the following equations: δH ðcorrÞ ¼ 0:933  δH ðcalcÞ þ 0:209½ppm

ð1Þ

δC ðcorrÞ ¼ 1:042  δC ðcalcÞ  4:756½ppm

ð2Þ

The calculated corrected values δH(corr) and δC(corr) provide somewhat lower mean absolute error (MAE) values (0.090.14 for 1H and about 1.72.7 for 13C atoms) for 15 (see Supporting Information, Table S9). Solid-State Structures and 13C NMR Spectra of 15. Experimental 13C CP-MAS spectra are shown in the Supporting Information. The structural assignment of the signals (Table 2) was done with the help of an NQS (nonquaternary suppression) experiment and a complementary CPMAS experiment with a short contact time (40 μs) where the signals of the quaternary carbon atoms were suppressed. A comparison with the complete structural assignment in solution (Supporting Information, Tables S4S8) was also helpful. The small differences between the carbon chemical shifts in different polymorphs or between two symmetry nonequivalent

’ RESULTS AND DISCUSSION Experimental and Calculated 1H and 13C Chemical Shifts in Solution. The NMR spectra of lactones 15 were recorded in

three different solvents [low-polar CDCl3 and two polar solvents capable of participating in H-bonding as either proton donor (CD3OD) or proton acceptor (CD3COCD3)] to check the variations of the chemical shifts with the type of solvent. The observed chemical shift data are summarized in Supporting Information (Tables S4S8). The ranges of chemical shifts observed for given hydrogen or carbon atom in three used solvents (Δ values) were generally rather small. The largest Δ values were observed, as expected, for the hydrogen and carbon atoms proximal to the polar groups in the molecules. Typically, 682

dx.doi.org/10.1021/jp209408b |J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

Figure 2. Correlation of the calculated and observed 1H and 13C chemical shifts for 15 in CDCl3.

Table 2. Experimental 13C Solid-State NMR Chemical Shifts of Lactones 15 2 carbon

a

1

orthorhombic

3

4

monoclinic

hydrate

without solvent

molecule A

5 molecule B

molecule A

molecule B

1

53.73

48.65

52.33

45.51

46.77

a

a

132.72

131.39

2

166.24

164.96

169.11

24.65

24.06

34.00

32.62

196.05

196.05

3 4

131.15 214.76

131.66 211.52

128.91 217.93

40.63 224.84

40.57 221.22

73.29 86.31

75.07 91.00

134.98 173.26

132.72 171.80 48.86

5

57.40

57.58

55.76

57.55

57.58

a

a

49.23

6

65.58

67.62

68.42

75.61

76.98

37.89

38.17

79.44

79.90

7

52.70

54.35

54.00

52.56

52.58

a

a

46.29

46.29

8

76.09

79.40

78.71

77.07

77.58

81.90

80.64

67.80

68.20

9

44.52

37.75

37.13

45.00

44.31

a

a

42.45

41.78

10

27.22

27.80

26.66

29.29

30.14

30.29

30.29

141.96

143.53

11 12

137.34 172.78

137.32 172.11

137.02 169.89

140.20 170.00

138.09 170.54

142.26 170.28

142.92 170.73

35.93 179.87

35.93 179.47

13

121.85

122.18

121.06

121.44

126.53

118.92

118.92

13.35

13.77

14

19.53

21.24

19.43

19.49

20.65

19.83

20.61

20.70

20.70

15

19.53

18.14

18.74

20.23

20.65

19.52

18.85

20.70

20.70

16

167.30

169.39

17

126.80

126.80

18

141.08

134.98

19 20

15.56 20.70

15.56 20.70

The signals of the carbon atoms C1, C5, C7, and C9 of compound 4 appear in a multiplet 43.0545.82 ppm.

atoms after geometry optimization was carried out for proton-only and all-atom optimizations by using a maximum plane-wave cutoff energy of 340 eV. When constrained, the heavy atoms were still affected by average forces (given as Cartesian components) of 0.7 eV/Å for carbon as compared to 0.008 eV/Å for the protons, indicating that further relaxation of the heavy atoms was necessary. Where all of the atoms were relaxed, the average forces were determined to be of a similar magnitude for all of the atomic species, i.e., 0.009 (C), 0.012 (O), and 0.011 (H) eV/Å.

molecules suggest that the accuracy of the calculations needed for the spectral assignment is higher than that obtained in the calculations of the chemical shifts of isolated molecules. We, therefore, used the CASTEP program, which exploits the translational symmetry in crystals by using plane waves as a basis set, and we used the GIPAW method for the chemical shift calculations within CASTEP again. The geometry optimization was performed by starting with the crystal structure determined by X-ray diffraction. A comparison of the typical average forces remaining on the 683

dx.doi.org/10.1021/jp209408b |J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

Mexicanin I (1)40. Only one geometrically independent mo-

then recalculated. The differences between the carbon shielding constants calculated with and without geometry optimization were up to 23 ppm. The differences between the calculated data with and without H optimization are given by the fact that X-ray crystallography cannot position hydrogen atoms correctly. Finally, we have calculated the carbon chemical shifts after the optimization of the positions of all the atoms. The chemical shielding calculated on the fully optimized structure provided the best correlation with the experimental chemical shifts (see Figure 4). Helenalin (2)41. Helenalin has been reported to exist in three polymorphic forms, deposited in the CSD as FEGWAP,42 FEGWAP01,40 and FEGWAP02,31 two of them crystallizing in the monoclinic (FEGWAP and FEGWAP01) and one in the orthorhombic (FEGWAP02) space groups. The molecules in all three polymorphs exhibit nearly the same molecular conformation. The three crystalline modifications differ mainly in the packing mode and in the hydrogen bonding pattern. In the orthorhombic form of 2, like in 1, the molecules related by a single translation and connected by C6—αOH 3 3 3 OdC12 hydrogen bonds arrange into linear chains while in both monoclinic forms the chains are formed by molecules related by the screw axis and joined together either by C6—αOH 3 3 3 OdC4 (FEGWAP, the structure discussed in this paper) or by C6— αOH 3 3 3 OdC12 hydrogen bonds. Hence, in the crystal structures of the three polymorphs of 2, we witness competition between two strong hydrogen-bond acceptors [carbonyl oxygen constituting the exomethylene-γ-lactone moiety (O12) vs α, β-unsaturated ketone (C4dO4) for a single C6—αOH donor group (Figure 5)]. The 13C NMR spectra measured for the available samples of FEGWAP and FEGWAP02 nicely mirror this different hydrogen-bonding pattern in the two types of crystals (Figure 6). The very small difference in the carbon chemical shifts for C6 is in accordance with a single-protondonor function of the C6 hydroxyl in both crystal forms, whereas large differences for C4 and C12 carbonyl carbons acting as hydrogen-bond acceptors in FEGWAP and FEGWAP02, respectively, reflect the different involvement of the two carbonyl groups in hydrogen bonding. Moreover, the involvement of the C4 carbonyl group in hydrogen bonding also affects the chemical shifts of the carbon atoms constituting the whole α, β-unsaturated ketone fragment. The possibility to distinguish two crystalline polymorphs with the calculation of solid-state 13C chemical shifts has been tested on 2. We calculated the chemical shift differences between the monoclinic and orthorhombic forms of 2 and compared them

lecule of 1 appears in the unit cell. The molecules related by the unit translation along the b-direction are connected by C6— βOH 3 3 3 OdC12 hydrogen bonds into linear chains (Figure 3). Only one of the two potential hydrogen-bond acceptor atoms present in the molecule (O4 and O12) is utilized in the crystal structure. To our knowledge, no other polymorphic form of this compound has so far been reported. Experimental 13C CP-MAS spectrum of mexicanin I is shown in Figure S1. We calculated the chemical shifts of solid 1 using the GIPAW method. When the X-ray crystal structure was used as the input geometry, agreement with the experimental chemical shift differences was rather modest. Subsequently, the positions of the hydrogen atoms were optimized, with the heavy atoms fixed to their original positions, and the shielding constants were

Figure 3. Arrangement of the molecules of 1 in the crystal. View along the a-direction. The dashed lines indicate H-bonds.

Figure 4. Correlation of the calculated chemical shieldings and experimental chemical shifts for solid mexicanin I (1).

Figure 5. Hydrogen-bonding pattern in the two polymorphic forms of helenalin (2) discussed in this paper: orthorhombic (FEGWAP02) (left) and monoclinic (FEGWAP) (right). The hydrogen bonding is shown with dashed lines. 684

dx.doi.org/10.1021/jp209408b |J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

with the experimental differences (see Figure 6). The experimental data were reproduced very well with the GIPAW calculations after the optimization of the hydrogen positions as well as after full geometry optimization. For all of the carbon atoms, the calculated chemical shift differences (Δδ) had the same sign as the experimental Δδ and the absolute values of Δδ were in most cases similar. The correlation of the calculated Δδ on the nonoptimized structure with the experimental data was rather poor (data not shown). For comparison, the Δδ values calculated for isolated molecules (X-ray structures without optimization) using the Gaussian program are also plotted in Figure 6. These values do not agree with the experimental Δδ values. From this comparison, it follows that the chemical shift differences between the polymorphs of 2 arise from their different crystal packing, which is not reflected by the calculations of isolated molecules, and hence, calculations with periodic conditions (such as GIPAW) are necessary. 6α-Hydroxydihydroaromaticin (3). This compound forms two distinct crystalline phases, hydrated (1:1 ratio) and nonhydrated. Both crystal structures are reported in this paper. In the nonhydrated crystals, the O6H donor group forms a threecenter hydrogen bond of which one constituent (to O4) is intramolecular and the other (to O12 at x + 3/2, y + 2, z + 1/2) is intermolecular, the intramolecular hydrogen bond being relatively weak (see Table S2 of Supporting Information). This molecular arrangement creates geometrical conditions favorable

for the formation of an additional, albeit weak, C13—H13a 3 3 3 O4 (at x + 3/2, y + 2, z  1/2) hydrogen bond involving the C13 exomethylene group and construction of the supramolecular R22(7) ring motif (Figure 7; for a definition of the graph-set notation, see ref 43). One consequence of the involvement of the α-exomethylene-γ-lactone fragment in cyclic hydrogen-bond interactions is the apparent flattening of the C13C11 C12O12 fragment (see Table S3). As expected, the packing and hydrogen-bond interactions in 3 3 H2O crystals are entirely different. The solvent water molecule acts as a double hydrogen-bond donor (to O4 at x + 1, y  1/2, z + 3/2 and O6 at x + 2, y  1/2, z + 3/2) and a single acceptor from the O6H hydroxy group. As a result, the hydrogen-bonded molecules form layers parallel to the (001) lattice plane. Interestingly, in the hydrated form neither the lactone carbonyl (C12dO) nor the C11dC13 methylene participate in the hydrogen-bond interactions and, presumably because of that, the α-exomethylene-γ-lactone fragment is more puckered than in the nonhydrated crystal (Table S3). The polymorphism of 2 and pseudopolymorphism of 3 both arise as a result of the competition between the two strong hydrogenbond acceptors for a single hydrogen-bond donor. From the above description, it follows that the major 13C NMR chemical shift differences should appear at C4 as well as at C11, C12 and, particularly, C13 carbons, constituents of the α-exomethyleneγ-lactone ring. That this indeed is the case can be seen in Figure 8.

Figure 6. Calculated 13C chemical shift differences between the monoclinic (FEGWAP) and orthorhombic (FEGWAP02) forms of helenalin (2) compared with the experimental differences.

Figure 8. Calculated 13C chemical shift differences between the two crystal forms of 6α-hydroxydihydroaromaticin (3) compared with the experimental differences.

Figure 7. 6α-Hydroxydihydroaromaticin (3): the hydrogen-bonding scheme in the hydrated phase (left) and in the nonhydrated phase (right). Strong hydrogen bonds are indicated with bold dashes and weak with regular dashes. Note the involvement of the C13 hydrogen atom in the formation of the R22(7) hydrogen-bond ring motif (right). 685

dx.doi.org/10.1021/jp209408b |J. Phys. Chem. A 2012, 116, 680–688

The Journal of Physical Chemistry A

ARTICLE

Figure 9. Two symmetry-independent but geometrically slightly different molecules of geigerinin (4), distinguished by color, are connected by OH 3 3 3 OH hydrogen bonds that extend in the form of ribbon throughout the crystal. Note the different involvement in the hydrogen bonds of the analogous hydroxy groups in the two independent molecules. For clarity, the neighboring molecules are represented solely by hydroxy groups.

Figure 11. Badkhysin (5): two geometrically independent molecules present in the crystal lattice differ mainly in the conformation of the angeloyloxy group (compare the corresponding torsion angles O8C16C17C18 = 150.5° and O80 C160 C170 C180 = 51.5°). The numbering scheme differs from the one used in the original crystallographic publication45 and deposited in the CSD.

the number of independent molecules in the crystal. Of the four hydroxy groups (from two different molecules), two are symmetrical donoracceptor groups (O3 and O40 ), one is a single donor and a double acceptor (O4), and one acts exclusively as a single donor (O30 ). The experimental and calculated chemical shift differences between the two different molecules of 4 are plotted in Figure 10. Owing to the different involvement in hydrogen bonding of the corresponding hydroxy groups, the main difference in the 13C NMR chemical shifts for the two independent molecules is seen for the C4 carbons while the shifts at the C12 nuclei remain nearly unaffected (Figure 10). Badkhysin (5)45. Figure 11 shows the shapes of two symmetryindependent molecules of 5. The molecules are held together in crystal mainly by van der Waals forces. While the conformation of the tricyclic skeleton in the two molecules is very similar, the main structural difference is seen in the ester side chain, where the angeloyloxy (2Z-2-methylbut-2-enoate) groups differ in the conformation around the C16C17 bonds. Moreover, in the original crystallographic publication45 it has been mentioned that one of the two symmetry independent molecules showed signs of positional disorder affecting the angeloyloxy substituent. Since no account was taken of this disorder during the structure refinement process, the accuracy of atomic coordinates is low and clearly affects the chemical shift values calculated with H only optimization. The experimental and calculated chemical shift differences between the two different molecules of 5 are plotted in Figure 12.

Figure 10. Experimental and calculated chemical shift differences between the two independent molecules in the solid 4. The molecules mostly differ in the involvement of their C3OH and C4OH groups in hydrogen bond formation. The experimental values for carbon 1, 5, 7, and 9 have not been determined.

Very good agreement between the experimental and calculated Δδ values was obtained for the two crystal forms of 3 (see Figure 8). Geigerinin (4). Although reported so by us in 1989,44 the crystal structure of geigerinin has not been deposited in the CSD. Therefore, we have decided to collect a new data set for these crystals and deposit the newly determined crystal structure. In the crystal structure of 4, there are two symmetrically independent molecules in the unit cell (primed and unprimed) which differ only slightly in the conformation of the tricyclic skeleton, mainly in the γ-lactone region (the differences of endocyclic torsion angles are