Noncovalent Chalcogen Bonds and Disulfide Conformational Change in the Cystamine-Based Hybrid Perovskite [H3N(CH2)2SS(CH2)2NH3]PbIII4

DOI:10.1002/ejic.201301017

Noncovalent Chalcogen Bonds and Disulfide

Conformational Change in the Cystamine-Based Hybrid Perovskite [H

N(CH

)

SS(CH

)

NH

]Pb

I

Nicolas Louvain,*

Gilles Frison,

Jens Dittmer,

Christophe Legein,

and Nicolas Mercier*

Keywords:Halogenometallates / Perovskite phases / Noncovalent interactions / Chalcogens

The cystamine-based hybrid perovskite, α-[NH₃(CH2)2S–S- (CH2)2NH3]PbI4(1a), can be transformed into its polymorph, β-[NH₃(CH2)2S–S(CH2)2NH3]PbI4(1b), by heat activation (T

= 150 °C). The crystal structures have been characterised by single-crystal X-ray diffraction, whereas the phase transition was followed by both solid-state ¹H,¹³C cross-polarisation magic-angle spinning (CPMAS) NMR spectroscopy and ther- modiffractometry techniques. At 150 °C, compound 1a is transformed into 1b, and, remarkably, theβphase (1b) can be nearly retained down to room temperature, which means that both polymorphs1aand1bcan coexist over a large tem- perature range. The structure of1bhas been solved, and it was found that cystamine molecules are disordered over two positions: the two related components with opposite helical conformations. Solid-state ¹H,¹³C CPMAS NMR spectro-

Introduction

Noncovalent intermolecular interactions that involve chalcogen atoms are well known in that they can be respon- sible for controlling the conformation of large molecules from biological to synthetic architectures.^[1–9]In particular, they play noninnocent roles in determining protein struc- tures and their folding pathways.^[1] Different studies have underlined experimental features specific to S···X (X being any chalcogen or a pnictogen) nonbonded interactions:

electrophiles tend to approach Y–S–Z (Y, Z being any atom except hydrogen) groups along a direction perpendicular to [a] Institut des Sciences et Technologies Moléculaires d’Angers, MOLTECH ANJOU, CNRS UMR 6200, Université d’Angers, 2 Bd. Lavoisier, 49045 Angers, France

E-mail: nicolas.mercier@univ-angers.fr http://moltech-anjou.univ-angers.fr

[b] Laboratoire des Mécanismes Réactionnels, Department of Chemistry, Ecole Polytechnique and CNRS,

91128 Palaiseau CEDEX, France

[c] LUNAM Université du Maine, CNRS UMR 6283, Institut des Molécules et des Matériaux du Mans,

Avenue Olivier Messiaen, 72085 Le Mans CEDEX 9, France [‡] Current address: Institut Charles Gerhardt UMR CNRS 5253

(AIME), Université Montpellier 2, CC1502, Place E. Bataillon, 34095 Montpellier CEDEX 5, France

E-mail: nicolas.louvain@univ-montp2.fr

Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/ejic.201301017.

scopic measurements show a significant broadening of the NMR spectroscopic line associated with two disordered car- bon atoms when cooling1b from 160 to 50 °C, thereby re- vealing the presence of exchange between these related atoms, and this favours a molecular dynamical disorder. Di- sulfide bridges of cystamine molecules are engaged in weak interactions with neighbours, either another cystamine mole- cule in1a(SS···SS interactions), or iodine atoms in1b(SS···I interactions). To evaluate the donating and accepting abili- ties of the disulfide bridge, and their impact on such weak interactions, a detailed partition of the interaction energy of ten dimer models has been calculated and revealed that the main contribution to the intermolecular bonding comes from the dispersion forces.

Y–S or Z–S bonds (Y–S···X or Z–S···X ≈ 90°; type I), whereas nucleophiles would interact with sulfur atoms pref- erentially along the extension of one of those covalent bonds (Y–S···X or Z–S···X ≈ 180°; type II; Figure 1, a).[1,3,6,10–14]These geometrical features have been interpre- ted in terms of donating or withdrawing ability of the in- volved sulfur atom as an orbital-type np–σ* formalism, in which a donating lone pair interacts with an accepting anti- bonding σ* orbital (Figure 1, b).^[15] Disulfide S–S func- tional groups are post-translational modifications that con- trol the ternary and quaternary structures of proteins, such as human insulin protein.^[16,17]Organic molecules with di- sulfide functional groups (R–S–S–R⬘) are inclined to adopt two different screwed structures in solution as well as in crystalline states to form a pair of chiral enantiomers, which are described asP- andM-helical forms.^[18–22]The racemi- sation is fairly rapid in solution on account of the relatively low barrier of rotation of S–S bonds in the case of nonbulky R and R⬘organic groups.^[23,24]In materials science, organic disulfides can be used as moderate donors towards soft metal ions, as well as flexible ligands in the fields of coordi- nation polymers or supramolecular chemistry.^[25–32]Experi- mental studies devoted to the interactions of the disulfide bridge with its environment, either in organic and inorganic crystals^[1,6,10] or in protein structures,^[11,33] show geometri-

cal trends that could be explained by the donating–ac- cepting formalism previously exposed, and similar conclu- sions on the basis of experimental evidence have been speci- fically made with regard to S–S···S–S, thereby placing the emphasis on the dispersion forces as well as on a np–σ*

orbital character to explain their propensity to inter- act.^[10,11,34]

Figure 1. (a) Geometrical features of S···X interactions and (b) np–

σ* interactions between chalcogen centres (X being any chalcogen or pnictogen).

The diprotonated cystamine (cys) molecule [H₃N–(CH₂)₂– SS–(CH₂)₂–NH₃]²⁺, here denoted by H₂cys, is an organic disulfide with two ethylammonium functions at both ends.

Until recently, this molecule was not often incorporated as a counterion in ionic-like compounds, even though it a pri- ori fulfils the requirements (i.e., nonbulkiness and the pres- ence of primary ammonium groups) for the self-assembly of layered hybrid perovskites.^[35] Such hybrids can be con- sidered multifunctional materials that combine properties of both organic and inorganic components.^[36–39] For in- stance, as organic compounds, they can be easily made into crystalline thin films,^[40] whereas their interesting electronic properties often come from the inorganic framework.^[37–42]

In the course of our investigations on perovskite-like com- pounds, we recently focused on the synthesis of halogen- ometallates of Bi^III and Pb^II hybrid materials that contain the H2cys dication. The diprotonated cystamine could af- ford an unprecedented series of iodoplumbate salts based on PbnI_{4n + 2}^{(2n+ 2)–} ribbons, namely, the “H₂cys series”

{(H2cys)[(2n+ 2 –u)/2]PbnI4n+ 2·(uC⁺, vG); with C⁺ and G being any monocation and neutral guest molecule, respec- tively, incorporated in the structure}.

Up to now, the “H₂cys series” was composed of (H₂cys)₂PbI₆·2H₃O (n= 1, being composed of isolated PbI₆ octahedra),^[43](H2cys)2Pb2I10·2H3O [n= 2, the formula be- ing reduced to (H₂cys)PbI₅·H₃O],^[44](H₂cys)₄Pb₃I₁₄·I₂(n= 3) and (H2cys)6Pb5I22·4H2O (n= 5).^[45]All inorganic anions for each member of the series can be regarded as a dimen- sional reduction of 2D hybrid perovskite layers, thus the dual nature of the H2cys cation is highlighted: on the one hand, it is able to stabilise networks of corner-sharing PbI₆ octahedra as is expected for nonbulky primary ammonium cations, and on the other hand, it might prevent the forma- tion of hybrid perovskite layers owing to the strong capa- bility of trapping guest molecules by means of noncovalent

interactions between the disulfide bridge and its neighbour- ing environment. The most interesting feature concerns the helical conformational change of disulfide components in the solid state that can be observed for then= 1 andn= 2 compounds, the latter of which results in an exceptional solid-state conglomerate α-(H2cys)PbI5·H3O to true race- mateβ-(H₂cys)PbI₅·H₃O reversible transition.^[44]The struc- tural transformation of the room-temperature acentric α phase to centric βphase is reversible and was followed by variable-temperature second-harmonic-generation (SHG) measurements, as only the acentric salt is optically active.

As a hysteresis was observed in the SHG = f(T) curve, both phases can coexist at a given temperature, thereby making such materials good candidates for SHG switches. Never- theless, the temperature range of phase coexistence is small, that is, a range of 20 °C for (H2cys)PbI5·H3O. One strategy that was envisaged as a plausible way to increase the tem- perature range of coexistence was to rigidify the inorganic scaffold and thus obtain a 2D hybrid perovskite (the “n =

⬁” member of the “H₂cys series”), in which the rotation of H2cys cations would be restrained (Scheme 1).

Scheme 1. Schematic representations of the (a)n= 2 and (b)n=

⬁members of the H₂cys series showing the potential directions of expansion of each system with green arrows; for then=⬁member (b), the expansion is restrained in the direction parallel to the inor- ganic layers owing to the rigidity of the polymeric anions.

Herein we report the synthesis, solid-state NMR spectra, X-ray structural characterisation and quantum chemical studies of the hybrid perovskite α-(H2cys)PbI4(1a,n =⬁), which undergoes a reversible structural transformation in the solid state to form the polymorphβ-(H₂cys)PbI₄(1b,n

= ⬁) at 150 °C. In the first part, the centric crystal struc- tures of 1a and 1b are described. We will show that great structural changes occur through the transition. In particu- lar, the disulfide bridges of cystamine molecules are ordered and interact together through SS···SS noncovalent contacts in1a, whereas in1bthey are disordered over two positions and are approximately turned perpendicular to the inor- ganic layers, thus interacting with iodine atoms. In the sec- ond part, we report on the variable-temperature measure- ments of the X-ray powder diffraction (XRPD) and solid- state¹³C NMR spectroscopic experiments, which show that 1bis retained down to 40 °C. This indicates that both poly- morphs of1a and1b can coexist over a large temperature range (40–150 °C). Moreover, the results of the ¹³C NMR spectroscopic study converge to assign a dynamical mode to the molecular disorder observed in 1b. Finally, in the

third part, we give the results of our theoretical investi- gations of SS···SS and SS···I interactions, which are system- atically observed in cystamine-based iodometallates, includ- ing the title hybrid compounds, and are suspected of play- ing a non-innocent role in the helical conformational mo- lecular change and/or in the molecular disorder phenomena in the solid state.

Results and Discussion

Foreword on the Synthetic Conditions

As mentioned previously, then=⬁member of the H₂cys series was, until recently,^[46]unreported because of the syn- thetic difficulty to prevent the incorporation of guest mole- cules through nonbonded interactions with the disulfide bridge of the H2cys cation. We speculated that this phenom- enon disturbs the self-assembly of the organic cations in solution during the nucleation process by keeping them from forming a suitable array of positive charges that would allow the formation of anionic 2D hybrid perovskite layers, and it mainly leads to 0D and 1D hybrid iodometallates.

To achieve a higher inorganic dimensionality by using H2cys, it is necessary to be able to control the interactions of the disulfide bridge in solution prior to the crystallisa- tion. In the “H2cys series”, it is interesting to note that all compounds have been crystallised using acetonitrile as a solvent, and all compounds, fromn= 1 ton= 5, incorpo- rate guest molecules (H3O⁺, H2O or even I2).^[43–45]We sup- posed that treating PbI₂with H₂cys and HI in a mixture of acetonitrile and a polar protic solvent should give interest- ing results owing to the fact that the created ionic species such as H₃O⁺should be more strongly attracted by the pro- tic solvent (i.e., solvated) than by the sulfur atoms of the disulfide group of H₂cys dications. Indeed, the preparation of1a, the hybrid perovskiteα-(H2cys)PbI4, was achieved by heating at 85 °C equimolar amounts of PbI₂ and H₂cys in an acidified 1:1 acetonitrile/ethanol solution. The resulting compound is guest-free, thus corroborating our first as- sumptions. Interestingly, this compound has been recently reported, serendipitously obtained from a mixture of PbI2

and the monocation HS–(CH₂)₂–NH₃⁺in a highly concen- trated acidic solution that was expected to give the [HS–

(CH2)2–NH3]2PbI4hybrid perovskite.^[46]The in situ forma- tion of the H₂cys dication by the oxidation of the thiol group of the thioethylammonium probably occurred after the self-assembly of the organic monocations in solution, therefore leading to the preformation of the 2D hybrid lay- ers. Interestingly, the authors stated that their crystals of1a are always obtained with another phase, the latter being a 1D-based hybrid material, which nicely emphasises the advantage of our rational synthetic approach.

Crystal Structures of 1a and 1b

The asymmetric unit of 1a, which crystallises in the P2₁/nspace group (Table 1), consists of two Pb atoms, eight

I atoms and two crystallographic independent H2cys cat- ions (Figure 2). Compound 1a is a hybrid perovskite that consists of PbI₄^2– perovskite layers that are separated by organic cations (Figure 3, a). In the organic layers, both en- antiomeric P and M forms of disulfide molecules are en- countered in the same layer, related to each other by then glide plane. The –CH₂–NH₃⁺ fragments of cystamine are located at the boundary of the inorganic sheets, thanks to intramolecular hydrogen bonds established between the –NH₃⁺ group and the neighbouring sulfur atom of each cystamine (Figure 2), thus reducing the typical hydrogen bonding between the ammonium of the organic layers and the iodide of the perovskite sheets. This particular feature of ammonium cations bearing an acceptor of hydrogen bonds in theβposition [X–(CH₂)₂–NH₃⁺, X = Cl, Br, OH]

limits the tilt of the PbI6 octahedra related to the mean plane of the layers, that is to say, the lead and equatorial iodide atoms approximately lie in thea,bplane, as already emphasised for similar compounds.^[47,48]Last but not least, the main attribute of 1a is the nonbonded interactions of the cystamine cations through the sulfur atoms of their di- sulfide bonds: d[S(2)···S(3)] = 3.53 Å and d[S(4)···S(4)] =

Table 1. Crystallographic data for1aand1b.

α-(H₂cys)PbI₄ β-(H₂cys)PbI₄

(1a) (1b)

Crystal system monoclinic monoclinic

Space group P2₁/n P2₁/a

a[Å] 17.7855(10) 9.1524(10)

b[Å] 8.5500(4) 8.2510(10)

c[Å] 23.270(2) 11.686(2)

α[°] 90 90

β[°] 98.780(10) 104.870(10)

γ[°] 90 90

V[ų] 3497.2(4) 852.9(2)

ρ_calcd.[mg m^–3] 3.301 3.384

2θmax. 25.92 30.05

T[K] 293(2) 313(2)

Reflections collected 26470 (6756) 17320 (2493) (unique)

Data (parameters) 6756 (235) 2493 (80) R₁[I⬎2σ(I)] [wR₂(all data)] 0.0477 (0.1270) 0.0559 (0.1692)

Figure 2. The conformations of the cystamine cations in (a)1aand (b)1b. The dashed bonds emphasise the intramolecular NH₃···S short contacts [1a, d(S1···H1B) = 2.9283(4) Å, d(S2···H2B) = 2.7233(4) Å,d(S3···H3B) = 2.7810(4) Å,d(S4···H4B) = 2.7941(4) Å;

1b,d(S1B···H0B) = 2.9899(9) Å]; some atoms are disordered over two positions (see the Exp. Section). Torsion angles: 1a, C2–S1–

S2–C3 76.7(7)°, C6–S3–S4–C7 –74.6(8)°;1b, C2A–S1A–S1B–C2B 82(2)°.

3.71 Å [Figure 3; these values are inferior, or reasonably close, to twice the van der Waals radius of a sulfur atom;

2·r_vdw(S) = 3.66 Å].^[49]This leads to define supramolecular tetramers, whereas we also observe that all disulfide bridges of cystamine are approximately parallel to the inorganic layers (Figure 3, a). These contacts, far from being well understood, are not unusual (see below).

Figure 3. Quite different molecular positions in both polymorphs (a)1aand (b)1bviewed along thebaxis.

Upon heating at 150 °C, compound 1a is transformed into its polymorph1bthrough a reversible solid-state reac- tion. During the1ato1btransformation, most crystals are broken into tiny fragments, and the transition could not be

Figure 4. The layout of perovskite layers in structures of (a)1aand (b)1bviewed along thecaxis (only one component of the disordered molecule is shown in1b).

followed in operando on the same single crystal. As shown by powder thermodiffractometry (see below), the structure of the high-temperature phase (1b) is retained down to 40 °C. This allowed us to collect data at T = 313 K of a single crystal of 1b. The structure of 1b crystallises in the P21/a space group, and its unit cell is four times smaller than that of1a(a_β≈a_α/2,b_β≈b_α,c_β≈c_α/2). The asymmet- ric unit consists of one Pb, two I atoms and one-half of the H₂cys cations, which is disordered over two positions, thus describing bothM- andP-helical conformations for one or- ganic moiety (Figure 2). This also unambiguously proves that a helical conformational change has occurred during the transformation from1ato1b.

The overall structure underwent major structural changes through the transition and the most spectacular ones concern the cystamine cations. Firstly, in contrast with 1a, the –CH₂–NH₃⁺fragments in1bare almost perpendicu- lar to the inorganic layer, with only the ammonium heads being located in the cavities of the perovskite sheets (Fig- ure 3, b). Indeed, whereas the torsion angle S1B–C2B–C1–

N is equal to 55.1(4)°, the torsion angle of the other disor- dered component of the cystamine cation, S1A–C2A–C1–

N, is equal to –178.5(2)°, thereby preventing the presence of the intramolecular hydrogen bond between the sulfur atom S1A and the ammonium group and increasing the interactions between the latter and the perovskite layers.

This effectively influences the geometry of the inorganic sheets by tilting the PbI6 octahedra out of the a,b plane.

Secondly, whereas disulfide bridges are connected together through the a priori weak SS···SS interactions and lie in a plane parallel to the inorganic layers in1a, they are roughly perpendicular to the perovskite layers in1b (viewed along the b axis, Figure 3, b) and interact unusually with apical iodides of the inorganic part:d(S···I) = 3.31 and 3.53 Å (the sum of van der Waals radii of S and I atoms is 3.83 Å).^[49]

Finally, the inorganic framework also undergoes a remark- able change from1ato1b, as can be seen in Figure 4. The layout of PbI6 polyhedra that belong to a single layer is similar in both structures (the Pb–I–Pb angles in both 1a and 1bare close to 160°) but the relative positions of two consecutive perovskite layers differ from each other. That means a concerted rotation of polyhedra that belong to the same layer (motion of equatorial iodides) occurred through the transition.

Thermodiffractometry and Solid-State NMR Spectroscopic Studies

The phase transformation from 1a to 1b has been fol- lowed by XRPD techniques (Figure 5). A slight shift of the (002) diffraction peak to lower angle is observed along with an increase in temperature (20 to 130 °C), which reveals that the gap between perovskite layers is increasing. Then, as shown by the clear shift to low angle of the (002) reflec- tion at 150 °C, the supplementary space between two inor- ganic layers might be provided to allow organic ligands to adjust their conformation to form1b. Finally, it seems that the structure of 1a breathes before yielding 1b. Above 160 °C and up to 190 °C, an unknownγphase is obtained, and upon cooling, the β phase is recovered rapidly (T = 130 °C). When the temperature reaches 190 °C, theβphase (1b) is transformed into an unknown γ phase, the XRPD pattern of which could be roughly indexed by the mono- clinic unit cell: a_γ= 11.703 Å, b_γ= 8.756 Å,c_γ= 9.023 Å, β_γ = 101.2°,V_γ = 907.3 ų(please refer to the Supporting Information). The main noticeable change concerns the cell parameters that lie in the perovskite layers, which are of 8.251 and 9.152 Å for theβphase (T= 40 °C,1b) and 8.756 and 9.023 Å for theγphase (T= 190 °C).

Figure 5. Thermodiffractometry study of 1a. Details of the pro- cedure: Different XRPD patterns of an initial sample of1ahave been collected between room temperature and 190 °C, then cooled to 40 °C. Compound1a(α) is transformed into1b(β) at 150 °C, which becomes theγphase at 190 °C; the upperβ(40) pattern was collected after the sample had been kept at 40 °C for two weeks.

On the contrary, theα-phase1ais recovered below 40 °C, which is much lower than the temperature of the initial phase transformation (T= 150 °C). Moreover, theβphase 1b is stable (at least) for two weeks at 50 °C after being heated at 160 °C, and no α phase is detected during this period (Figure 5). The results indicate that both poly- morphs can then coexist over the broad temperature range of 50–140 °C.

1H,¹³C cross-polarisation magic-angle spinning (CPMAS) solid-state NMR spectroscopic experiments un- der variable temperature were carried out on 1a to study the phase transition that leads to1b. Figure 6 compares the spectra of (H2cys)PbI4 between room temperature and

164 °C on the way up (Figure 6, spectra a–f) as well as on the way down (Figure 6, spectra f–l). The spectrum of1aat 27 °C shows one broad and low resolved peak (Figure 6).

This spectrum was reconstructed with six contributions with relative intensities (two of them with larger line widths and double intensities) that are in agreement with the exis- tence of two crystallographic independent cystamine dicat- ions (i.e., eight different C atoms with the same multiplicity) for1a(see Figure S5a and Table S5a in the Supporting In- formation). According to the expected¹³C isotropic chemi- cal shift (δ_iso) values of the C(S) and C(NH3) atoms of the cystamine dication [estimated as δ_isoC(S) = 32.6 ppm and δisoC(NH3) = 41.1 ppm], the three lowerδisovalues may be assigned to C(S) (i.e., C2, C3, C6 and C7) and the three higher δ_iso values to C(NH₃) (i.e., C1, C4, C5 and C8).

When the temperature increases from 27 to 107 °C, the sig- nal of1adoes not evolve significantly (Figure 6, spectra a–

c). In contrast, it collapses at 122 °C, with the emergence of two new lines (Figure 6, spectrum d), whereas it completely disappears above 137 °C (Figure 6, spectra e and f). The two emerging lines are probably the signatures of the β phase, which indicates that both αandβphases would co- exist between 122 and 137 °C, whereas beyond, only theβ phase would be present. At this stage, we could note a slight discrepancy with the XRPD study, as theαphase was ob- served at 137 °C in the NMR spectra, and both α and β phases were still detected at 150 °C by diffractometry (Fig- ure 5). Taking into consideration that these techniques are sensitive to different physical variations within the solid state, dissimilarities might be due to both inherent tempera- ture errors and temperature gradient of samples, especially for the thermodiffractometry technique (see the Exp. Sec- tion). Another plausible hypothesis might be a change of molecular behaviour within the α phase before the struc- tural transformation from theαtoβphase occurs. In other words, one might postulate that NMR spectroscopy detects a modification in the cystamine conformation at a tempera- ture at which XRPD only identifies an increasing layer-to- layer distance [shifting to lower angles of the (002) diffrac- tion peak] in theαphase (Figure 5). The spectrum recorded at 164 °C that can be unambiguously assigned to 1b(Fig- ure 6, spectrum f) was reconstructed with two contributions at δ_iso = 36.0 ppm and δ_iso = 42.5 ppm (Figure S5b and Table S5b in the Supporting Information).¹³C δ_isoestima- tions and the similar intensities of these two lines allow the assignment of the first one to C2A and C2B and the second to C1. Thus, C2A and C2B either have the sameδ_isovalue or are in the fast intermediate exchange regime (i.e., they exchange their positions with a rate constant larger than half the magnitude of the chemical-shift difference fre- quency).^[50] A decrease in the temperature from 164 to 50 °C (Figure 6, spectra f–k) leads to a gradual broadening of the two NMR spectroscopic lines of1b, especially those assigned to C2A and C2B, as shown by the reconstruction (Figure S5c in the Supporting Information). From 164 to 50 °C, the width of the line assigned to C1 increases by a factor of 2.6 and the width of the line assigned to C2A and C2B increases by a factor of 3.9 (Table S5c in the Support-

ing Information). Whereas one expects broadening due to lower mobility, the strong increase of the width of the line assigned to C2A and C2B indicates the presence of ex- change (or dynamical disorder from a crystallographic point of view) between these two atoms. As the exchange rate decreases along with the temperature, the line broad- ens, and at 50 °C it is already close to the coalescence point.

Figure 6.¹H,¹³C CPMAS NMR spectra of1aand1b; the spectra were acquired in the order a-l.

Whereas upon an increase in temperature the phase tran- sition takes place at 122 °C, upon a decrease in temperature, the recovery of1atakes place below 50 °C. This is consis- tent with the hysteresis revealed by the temperature-con- trolled XRPD study. Moreover, the evolution of the cyst- amine conformation in the solid state as studied by NMR spectroscopy is highly compatible with the results of the single-crystal analysis, and therefore the presence of short nonbonded SS···SS contacts in theαphase and SS···I con- tacts in theβphase might play a role in the establishment of such a large hysteresis.

Figure 7. (a) Polar scattergram plotting the SS···S interaction angles (ω₃and ω₄; see Figure 8 for their definition) against the SS···S interaction distance. (b) Histogram showing the SS···S interaction angles over the compounds of the CSD.

Structural and Theoretical Investigations of SS···SS and SS···I Noncovalent Interactions

An exhaustive review of the Cambridge Structural Data- base (CSD)^[51] confirmed the findings previously high- lighted. Indeed, among the 637 structures that contain a C(sp³)–S–S–C(sp³) fragment, 91 showed SS···SS intermo- lecular contacts shorter than 3.7 Å, approximately twice the van der Waals radius of sulfur.^[49] The presence of these close contacts suggests that these might be due to true inter- molecular interactions rather than artefacts of crystal pack- ing. As can be seen from Figure 7, the S–S···S angles (ω₃ andω₄; see their definition on Figure 8) can be roughly be separated into two groups that correspond to either type I (S–S···S ≈90°) or type II (S–S···S ≈180°) interactions.^[3,6]

In the case of 1a, let us notice again that SS···SS contacts are of 3.533(7) Å (S2···S3) and 3.715(7) Å (S4···S4; Fig- ure 9), whereas the SS···S angles are in the range 131–168°

[S1–S2···S3, 168.6(4)°; S4–S3···S2, 142.0(1)°; S3–S4···S4, 131.3(1)°]; these results fall into the trend of the distribution of the experimental values derived from the CSD. Similar results can also be extracted for the S–S···I interactions (ple- ase refer to the Supporting Information).

Figure 8. Definition of the five important geometrical parameters, the distance r(X1–X3), and the orientation angles ω₁(y,[X1X3]), ω₂(z,[X3X4]), ω₃([X1X3],[X3X4]), ω₄([X2X1],[X1X3]) and ω₅([Y1X1],[X1X3]).

Figure 9. The intermolecular SS···SS and SS···I interactions at the organic–inorganic interface in the structure of1a.

These geometrical features can be interpreted in terms of the donating or withdrawing ability of the disulfide bridge.

As already emphasised, this interpretation arises from ex- perimental investigations on chalcogen com- pounds.[3,10,34,52–56] Thus, when the S–S···S angle ap- proaches 180°, the disulfide entity could be an acceptor of electronic density through theσ*_S–Sorbital of the S–S cova- lent bond. In the other situation, in which the S–S···S angle equals 90°, two different cases should be considered: on the one hand, if C–S···S≈ 0°, the disulfide bridge can behave as an accepting unit through its C–S bond (i.e., its σ*C–S

antibonding orbital), and on the other hand, if C–S···S ≈ 90°, it could donate electronic density thanks to one of its lone pairs.

To shed light on the origin of those interactions, their nature and their role in the conformation change, a quan- tum chemistry study was undertaken. We based our meth- odology on the work of Gleiter et al. in which chalcogen–

chalcogen interactions between homo- and heteronuclear species (CH3)2E¹···CH3E²Z (E¹, E²= O, S, Se, Te; Z = CH3, C₂H, CN) have been investigated by quantum chemistry methods.^[8,9]Their strategy was to study the influence of the inductive ability of substituents attached to the chalcogen atom on the force of the interaction. The contact between two molecules in the solid state was modelled by a supra- molecular dimer, with one monomer being considered an electron-accepting unit (CH₃E²Z) and the other one being an electron-donating unit [(CH3)2E¹], as defined by the or- bital-type np–σ* formalism (Figure 1). By varying the Z substituent of the CH3E²Z monomer from Z = CH3

(methyl, electron-donating) through Z = C₂H (alkynyl, in- termediate) to Z = CN (nitrile, electron-withdrawing), it was possible to enhance the chalcogen–chalcogen interac- tion. The electron- withdrawing effect of the nitrile group amplifies the electrophilicity of the E² atom and therefore increases the force of the E¹···E²np–σ* contact while short- ening the E¹···E²distances (shorter than the sum of the van der Waals radii of E¹and E²species). A similar methodol- ogy has therefore been followed to investigate the disulfide–

disulfide contacts encountered in1aand the disulfide–iod- ide contacts of1b, and the model dimersA–H(Figure 10) were designed for that purpose. Models A–F were con- structed to unravel the donating and/or accepting ability of

the R–S–S–R group, whereas models G and Hwere built to closely characterise the SS···SS interaction in the solid state.

Figure 10. (a) Model systems A–H. (b) Details of the modelH, H1–2andH2–3.

All the model systems, excluding G and H, were sub- jected to full geometry optimisation at the MP2 level of theory as described in the Exp. Section. The main resulting parameters are summarised in Table 2. The equilibrium dis- tances of the optimised dimers are in relatively good ac- cordance with twice the van der Waals radius of sulfur as demonstrated by Gleiter et al. with the MP2 method.^[8,9]

On the other hand, DFT^[57]/B3LYP^[58–61]optimisations led to inconsistent geometries with sulfur–sulfur intermolecular distances between 3.8 and 7 Å (see the Supporting Infor- mation).

Table 2. Calculated interaction energy E_int [kcal mol^–1], intermo- lecular equilibrium distancer[Å]^[a]and orientation anglesω_1–5[°].

Models Eint r ω₁ ω₂ ω₃ ω₄ ω₅

A –2.792 4.03 113.9 19.7 175.8 74.7 74.7

B –3.853 3.49 97.3 12.6 174.7 85.3 85.3

C –3.123 3.69 104.5 25.7 166.9 81.5 79.7

D –2.893 4.63 124.4 75.2 – 76.7 60.4

E –3.702 3.57 99.3 21.6 167.8 85.4 82.7

F –3.179 3.89 109.3 60.6 135.8 77.6 78.4

G^[a] –2.351 3.53 46.8 62.7 168.6 142.0 89.5 H1–2^[a] –2.264 3.53 46.8 62.7 168.6 142.0 89.5 H2–3^[a] –0.534 3.71 43.7 90.0 131.3 131.3 102.2 [a] For nonoptimised model systems, the geometrical features are given as the experimental ones.

Within the donor–acceptor hypothesis, the ideal geome- try to obtain a maximal np–σ* interaction would be reached forω1≈ 90° andω2 ≈0°. Examination of our re- sults (Table 2) indicates it is rather reasonable to rationalise the present disulfide–disulfide interactions in terms of an np–σ* perturbative interaction, except for D (Figure 1). In- deed, theω₁angle is approximately equal to 90° for all the optimised models (i.e.,A–F), and it decreases when the elec- tron-withdrawing effect of the X₄substituent becomes more important (going fromAtoB, or fromDtoE). When con- sidering the A, Band C models, one can notice from the interaction distances and energies listed in Table 2 that the Me–S–S–Me entity is a moderate accepting unit with an intermediate S···S distance relative to Me–S–Me and Me–

S–CN (3.69 versus 4.03 and 3.49 Å, respectively). The cal- culations onD,EandFare based on the Me–S–S–Me as a donating unit. Firstly, the S···S intermolecular distances are greater than those of A,BandC, which indicates that the interactions are weaker and that the Me–S–S–Me monomer is a poor donating unit. With a weak accepting unit (i.e., in theDmodel), the geometry of the supramolec- ular assembly cannot be described in terms of the np–σ*

orbital interaction because the geometry does not show sig- nificant S···S interactions and mainly reveals hydrogen bonds between sulfur atoms and methyl groups (Figure 11).

Models EandF, with Me–S–CN and Me–S–S–Me as ac- cepting units, deviate from the ideal geometry of an orbital- type interaction because of the presence of the second sul- fur atom on the donating monomer (Figure 11). This is indicated by an increasing ω2 angle from Eto F(21.6 to 60.6°). Finally, it is clear that the alignment of disulfide bridges, seen as a fingerprint of the SS···SS interaction in the solid state, does not remain still during the optimisation because of the moderate electron-accepting and poor elec- tron-donating nature of the Me–S–S–Me moiety. Therefore, other forces also influence the structure.

The diagram in Figure 12a depicts the main results that come from the natural bond orbital (NBO) calculations, achieved to give us a qualitative representation of the rela- tive strength of sulfide–sulfide noncovalent interaction ver- sus hydrogen bonding. This was done by interpreting the sums of the second-order interactions terms of the NBO program in terms of hydrogen and sulfide–sulfide bonding

Figure 11. Equilibrium geometry for the model systemsC, D, E and F (from left to right). Black dashes: S···S interactions, light grey: CH···S hydrogen bonds.

(as shown in Figure 1; see the Exp. Section). This result supports the analysis of the geometries optimised at the MP2 level: the more accepting one monomer is, the more important the contribution from the S···S interaction rela- tive to hydrogen bonding. If we consider A,Band C, the nature of the accepting monomer has a direct influence on the origin of the interaction. ForA, the optimised geometry is governed by hydrogen bonding, whereas in B the S···S interaction is the main term. Results forC show the inter- mediate nature of the disulfide bridge as an accepting unit.

This statement is in good accordance with the conclusions from Gleiter et al. about sulfide–sulfide interactions; sulfur is a pivotal element in the chalcogen group.^[8,9]Indeed, oxy-

Figure 12. (a) Relative contribution of the S···S and hydrogen bonding obtained from the NBO analysis for models A–H. (b) Contribution of the electrostatic (Eelst), induction (Eind), dispersion (E_disp) and exchange-correlation (E_exch) energies to the interaction energyEint,SAPTas derived from the SAPT analysis (see the Exp.

Section).

gen–oxygen interactions are mainly of an electrostatic na- ture, whereas selenium–selenium or even tellurium–tel- lurium interactions are dominated by the np–σ* per- turbative interaction. Therefore, the sulfide–sulfide interac- tions represent a transition between the two previous de- scriptions, with the main contribution depending on the force of the accepting unit. The results are somehow similar forD,EandF, and the contribution of the sulfide–sulfide interaction increases with the force of the accepting unit.

Nevertheless, the E_S–Xterm is strictly nil for theDmodel, and very small for F, thus emphasising the poor electron- donating ability of the disulfide entity. On the contrary, the sulfide–sulfide interaction is preponderant for the experi- mental geometriesG,H1–2andH2–3. From the symmetry- adapted perturbation theory (SAPT) analysis, it is empha- sised that the dispersion forces are the main bonding contri- bution to the intermolecular interaction in every model di- mer studied (Table 3; Figure 12, b). The SAPT formalism has been shown to allow for a fine decomposition of non- covalent interaction energies into “chemically understand- able” values such as electrostatic, induction and dispersion forces, plus a term that comes from the electron exchange, namely,E_elst,E_ind,E_dispandE_exch.^[62]As a reminder, electro- static forces describe the ionic character of an interaction, the induction forces represent the distortion of the electron cloud of a neutral specie by the permanent dipolar moment of an adjacent specie, dispersion forces (instantaneous di- pole-induced dipole interactions) are commonly called van der Waals forces and the exchange term comes from the antisymmetrisation of the wave function to take into ac- count the Pauli principle.^[63]

Table 3. Partition of the energies [kcal mol^–1] derived from SAPT calculations in electrostatic (E_elst), inductive (E_ind), dispersive (Edisp), and exchange (Eexch) energies for model systems A–G as defined by Equation (4) to Equation (8) in the Exp. Section.^[a]

Model E_elst E_ind E_disp E_exch E_int,SAPT^6-311G**

A 0.606 –0.334 –2.304 0.377 –1.951

B 1.038 –0.543 –2.740 0.633 –2.217

C 1.004 –0.356 –2.721 0.544 –1.931

D 0.904 –0.314 –2.639 0.492 –1.906

E 1.130 –0.407 –2.839 0.710 –1.852

F 1.209 –0.238 –3.317 0.720 –1.932

G 0.928 –0.248 –2.021 0.339 –1.283

[a] The last column collects the sum of the four contributions plus δ_HF. For details, see the Supporting Information.

Along with the dispersion forces, the induction forces are also bonding but on a smaller scale. The absolute values of these forces (Eindand E_disp) increase with the electron- accepting nature of the monomer, fromA toBor fromD toE, which is consistent with the polarisability of the sulfur atom. The electrostatic forces are repulsive in all systems, which is as expected due to the nature of sulfur atoms. For instance, in analogous oxygen-containing models, the elec- trostatic forces were found to be bonding. When consider- ing A, B and C, dispersion and induction forces increase fromAtoBand are intermediate forC, which emphasises

the weak accepting nature of Me–S–S–Me. ForD,EandF, changing the accepting group has a different effect. Interest- ingly, the E_disp absolute value for model F is the highest among our systems, along with a smallerE_indvalue. Those findings highlight that the intermolecular interaction is strong although not primarily being of orbital origin, al- though a longer interaction distance is obtained relative to C(3.89 versus 3.69 Å) and this might be due to the presence of hydrogen bonds on the donating monomer (Figure 11).

Finally, the G model has been calculated from the experi- mental geometry. The results show small E_ind and E_disp, which could be explained if we state that the polarisability of the sulfur atom in the experimental system is weaker than expected.

Conclusion

The cystamine ligand is of interest considering its disul- fide bridge as well as its preferential chiral conformations.

It is remarkable that five halogenoplumbate salts based on it were achieved, and that all the compounds define an un- precedented series of structures based on PbnI_{4n+ 2}^{(2n+ 2)–}

networks. However, the main interesting features, which oc- cur in then = 1,n= 2 andn=⬁salts, are the solid-state phase transformations that involve a helical conformational change of disulfide components on one hand, and the coex- istence of the two related phases over a temperature range on the other hand. Although this temperature range was restricted to about 20 °C inn= 1 and 2, it is quite fortunate that the coexistence temperature range in (H2cys)PbI4(1a/

1b, n=⬁) is 100 °C. The coexistence of both polymorphs might be of great interest if each compound can be associ- ated with a different response to an external stimulus as in n= 1 and 2, theαphases (room temperature) of which are optically active (SHG), whereas their β phases (high tem- perature) are not. The most likely reason for such a large temperature range of coexistence is that the 2D inorganic frameworks are more rigid than the low-dimensional ones, which makes it more difficult for the organic cations to ad- just their conformations in the solid state, especially when the sample is cooled to low temperature.

The results of our theoretical calculations show that the SS···SS and SS···I interactions are weak, with the dispersion forces being the main bonding contribution to the intermo- lecular interaction, which most certainly indicates that such interactions have a minor role in the reversible 1a-to-1b phase transition. Calculations also show that the short ex- perimental S···I distance (3.31 Å) that is present for both components of the disordered cystamine in theβphase (1b) corresponds to a positive interaction energy. This could ex- plain the disorder phenomenon as a result of an oscillation of cystamine molecules between two unfavourable posi- tions. This hypothesis of a dynamic disorder is also sup- ported by the study of¹H,¹³C CPMAS NMR spectroscopy versus temperature. It will be valuable in the future to see if the application of an electric field is able to induce a pref- erential helical chiral conformation of cystamine molecules in motion, thus leading to polar and SHG active materials.

Experimental Section

Synthesis and Characterisation:(H₂cys)PbI₄(1a) was prepared un- der solvothermal conditions in a 25 mL Teflon^®-lined PARR auto- clave. Hydriodic acid (1 mL,M_r= 80.92,d= 1.49, 48 %) was added to an acetonitrile/ethanol solution in a 1:1 ratio (8 mL) that con- tained PbI₂ (146.30 mg, 0.317 mmol) and H₂N–(CH₂)₂–SS–

(CH₂)₂–NH₂·2HCl (71.41 mg, 0.317 mmol). After the heating/cool- ing steps (2 h from room temperature to 80 °C, 3 h at 80 °C and 50 h from 80 °C to room temperature), nice orange block crystals of1awere obtained. Crystals were filtered, then washed with cold ethyl acetate and dried at 40 °C to result in a yield of 49.7 % (137 mg based on PbI₂). The XRPD pattern of the sample showed that all reflections could be indexed in the monoclinic cell of the hybrid perovskite of 1a. Depending on the experiments, a small amount of yellow crystals of (H₂cys)₃Pb₅I₁₆ could be crystallised as impurities. This compound can be obtained as a pure phase in ethanol under the same conditions.^[64]Differential scanning calo- rimetry (DSC) and thermogravimetry (TG) measurements were performed with DSC-2010 and TGA-2050 TA Instruments systems in the temperature range of 20–350 °C and 20–900 °C, respectively (given in the Supporting Information). The first endothermic peak at 125 °C in the DSC curve corresponds well to the phase transfor- mation from1ato1b, whereas another endothermic peak atT= 169 °C is assigned to another solid-state reaction that leads to an unidentified phase (see above). The resulting phase was then stable up to 250 °C and then decomposed in two steps up to 600 °C.

X-ray Crystallography: X-ray diffraction data of selected single crystals were collected with a Bruker-Nonius KAPPA-CCD dif- fractometer equipped with a graphite-monochromated Mo-K_αra- diation (λ= 0.71073 Å) at room temperature for1aand at 40 °C for1b. A summary of crystallographic data and refinement results for1aand1bis listed in Table 1. Structures were solved using Sir92 program^[65] and refined using the SHELXL97 package^[66] im- plemented in the WinGX software suite.^[67,68]Heavy atoms (Pb, I, S) were first located using direct methods, with C and N atoms being then localised from the analysis of the Fourier-difference maps. Positions and agitation parameters were refined by full-ma- trix least-square routines against F². All hydrogen atoms were treated with a riding model in each structure. There are two lead atoms and iodine atoms in the asymmetric unit of1a, which are balanced by two crystallographic independent H₂cys cations; how- ever, there are one lead and two iodine atoms in the asymmetric unit of1b, which are balanced by half of a disordered H₂cys cation.

In fact, the C–S fragment is disordered over two positions (C2A/

C2B and S1A/S1B) in such a way that only one kind of cation that comprises both components is then defined with the site-occupa- tion factor of each component equal to 0.5 (Figure 2). Two C–C bond lengths that involve C1 are longer or shorter than the ex- pected values (1.50 Å), which correlates well with the quite high anisotropic agitation parameters of C1 (U_iso= 0.090⫻10^–3Ų). Fi- nally, refinements of positions and anisotropic displacements pa- rameters of all non-hydrogen atoms lead toR1 = 0.048 (1a) and R1 = 0.055 (1b). A complete list of crystallographic data, along with the atomic coordinates, the anisotropic displacement param- eters and bond lengths and angles for each compound are provided in the Supporting Information.

CCDC-724583 (for1a), -724584 (for1b) and -724585 (for2) con- tain the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallo- graphic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

Powder Thermodiffractometry:X-ray powder diffraction measure- ments were carried out on a D8 Advance Bruker diffractometer

using CuK_αradiation, equipped with a linear Vantec Super Speed detector and a TTK-450 temperature chamber. A set of 15 min patterns in the 6–40° 2θrange were collected at different tempera- tures. Each pattern was collected 10 min after the expected tem- perature was reached, whereas the heating rate of the sample was 2 °C min^–1to prevent any potential overshooting. Because the sam- ple is heated mainly from below by a copper sample holder, with the thermocouple being placed with heat-conducting grease within the holder a few millimetres below the sample, the temperature of the sample is subject to gradient and the temperature reading of the XRPD might suffer from it. Several pre-studies were carried out to refine both the heating rate and stabilisation time to obtain coherent temperature-controlled diffractograms.

Solid-State NMR Spectroscopy: NMR spectroscopic experiments have been performed with a Bruker Avance III spectrometer with nominal 300 MHz proton frequency, equipped with a double-reso- nance WVT-type magic-angle spinning (MAS) probe head for 4 mm rotors. The temperature was calibrated by means of ²⁰⁷Pb spectra of PbNO₃spun at 5 kHz, the same spinning frequency as in the main¹³C experiments. In addition, the temperature gradient over the dimension of the rotor was assessed by these experiments, which ranged from about 1 °C at 27 °C over 3 °C at 77 °C up to 5 °C at 164 °C. The temperatures given are therefore average tem- peratures with the above variation. The main experiments were

1H,¹³C cross-polarisation (CP) experiments with a contact time of 1 ms. For the temperature series, the spectra were acquired in the order 27, 77, 107, 122, 137, 164, 137, 122, 107, 77, 50, 27 °C. The sample was kept at each temperature for 50 min for the accumu- lation of 1024 transients, followed by temperature change and sta- bilisation for about 30 min. Stabilisation was verified by short test spectra. The number of transients for the spectrum of1ashown in Figure 6 was 2048. The spectra were referenced to TMS. NMR spectra were handled and reconstructed using the DMFIT soft- ware.^[69]

Analysis of the Cambridge Structural Database: The Cambridge Structural Database (CSD) is the well-known database provided by the Cambridge Crystallographic Data Centre (CCDC).^[51]In Octo- ber 2010, it contained no fewer than 525093 structures. We per- formed a statistical analysis on version 5.31 (November 2009 and August 2010 updates) of the CSD by using the search program ConQuest (version 1.12),^[70] the visualisation program Mer- cury^[70–73] and the analysing program VISTA.^[74]The search was carried out by defining two fragments as C(sp³)–S–S–C(sp³) for the SS···SS search; one of these fragments was replaced by one iodine atom for the SS···I search. We requested a sulfur–sulfur interaction distance smaller than the sum of the van der Waals radius of sulfur [r_vdw(S) = 1.83 Å]^[49] plus 0.04 Å and a sulfur–iodine distance smaller than sum of the van der Waals radius of sulfur and the effective ionic radius of iodide anion [r_ionic(I^–) = 2.20 Å]^[75] plus 0.04 Å. To broaden our findings, another series of searches was carried out by scrutinising the nonbonded contacts for the SS···SS and SS···I systems by defining the fragments as R–S–S–R (with R being any atom except hydrogen) in close contact with a single sulfur or iodine atom (see the Supporting Information). Additional structural manipulations, and exhaustive checking of the results provided by VISTA, have been achieved by using the software Dia- mond (Crystal Impact, version 3.2f).^[76]

Theoretical Calculations:The methods and basis sets used to study the model dimers (Figure 9) were chosen according to a previous study.^[8,9]The basis sets were obtained by means of the Basis Set Exchange (BSE) software and the EMSL Basis Set Library.^[77]All MP2 calculations were performed with the Gaussian 03 package