Article pubs.acs.org/JPCC
Thermodynamics of the Segregation of a Kinetically Trapped TwoDimensional Amorphous Metal−Organic Network Sören Krotzky,†,⊥ Claudius Morchutt,†,∥,⊥ Vijay S. Vyas,† Bettina V. Lotsch,†,‡,§ Rico Gutzler,*,† and Klaus Kern†,∥ †
Max Planck Institute for Solid State Research, Heisenbergstraße 1, 70569 Stuttgart, Germany Department of Chemistry, Ludwig Maximilian University, Butenandtstraße 5-13, 81377 Munich, Germany § Nanosystems Initiative Munich (NIM) and Center for Nanoscience, Schellingstraße 4, 80799 Munich, Germany ∥ Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland ‡
S Supporting Information *
ABSTRACT: Surface-confined self-assembled molecular networks are commonly described as crystalline structures with short- and long-range order. Few exceptions to this trend are found in the literature reporting on amorphous and glassy networks and discussing the origins of their disordered nature at the atomic scale. Here, we show that an amorphous twocomponent metal−organic network can be synthesized on a crystalline metal surface at room temperature by vapor deposition of organic molecules and metal atoms. Scanning tunneling microscopy reveals the transformation of a crystalline closepacked organic layer into a disordered porous layer formed through the self-assembly of an organic semiconductor functionalized with two cyano groups upon introduction of iron atoms on Ag(111). In contrast to the commonly observed preference for trigonal crystal symmetry in 2D systems that form glassy networks, our metal−organic network favors rhombic or square tessellation of the surface. Statistical analysis of the network’s morphology reveals that entropy plays a critical part in its stabilization. Phase segregation of the binary mixture of monodisperse metal atoms and organic molecules into spatially separate homogeneously crystalline domains of molecules and metal atoms upon thermal annealing and subsequent cooling is observed. This suggests that the amorphous network is kinetically trapped at room temperature during the preparation process.
■
INTRODUCTION Two-dimensional (2D) amorphous materials have experienced increased attention in recent years, in particular by employing real-space imaging techniques that circumnavigate the shortcomings of diffraction techniques.1 The atomic structure of 2D silica glass films has been elucidated by scanning tunneling microscopy (STM)2 and (scanning) transmission electron microscopy (S)TEM.3 The solid-state transition from crystalline to amorphous structure by e-beam-induced conversion of graphene into a 2D carbon glass was analyzed by TEM, thus yielding valuable insight into the transitional states.4 STM especially offers the salient capability to elucidate the structure of noncrystalline materials confined to surfaces with atomic resolution, as shown on a copper oxide surface5 and for monolayers of different molecules.6−8 The characterization of two-dimensional glasses and amorphous networks selfassembled from organic molecules on surfaces thus offers fundamental insight into the origin of their disordered appearance. Studies of disordered molecular networks describe systems synthesized in ultrahigh vacuum9−13 as well as at the liquid/solid interface,14−16 and offer various explanations for © 2016 American Chemical Society
the observed disorder. A random cytosine network adsorbed on Au(111) was shown to be the consequence of the assembly of cytosine into several distinct elementary structural motifs stable under the experimental conditions, which subsequently arrange into the observed disordered structure. Crystallization is precluded by a large energetic barrier, kinetically trapping the random network.7 Entropy as the driving force behind random networks was reported for monocomponent6 and bicomponent15 networks at the liquid/solid interface. Another random metal−organic string network assembles on Ag(111) as a result of small energy differences between 3- and 4-fold coordination of two stereoconformers of a ditopic organic molecule.8,17 Conformational flexibility of an organic linker can likewise result in disordered metal−organic networks in which the molecule is either randomly distorted or exhibits random orientations.9 The ability to resolve the atomic and molecular structure of 2D amorphous solids and glasses thus leads to inReceived: December 18, 2015 Revised: January 26, 2016 Published: February 3, 2016 4403
DOI: 10.1021/acs.jpcc.5b12406 J. Phys. Chem. C 2016, 120, 4403−4409
Article
The Journal of Physical Chemistry C depth understanding of the physical concepts that underlie the formation of disordered networks and why these are favored over crystalline structures under given experimental conditions.18 Evaluation of kinetic and thermodynamic properties19−21 provides crucial insights for predicting the outcome of self-assembly processes22 in molecular and biological structures.23,24 The relevance of these studies extends beyond 2D systems to amorphous metal−organic frameworks,25,26 coordination polymers,27 and other such 3D materials.28 Herein, we investigate the self-assembly of a derivative of [1]benzothieno[3,2-b]benzothiophene (BTBT), which is an important organic semiconductor.29 Our recent success in functionalization of BTBT into a ditopic prochiral dibromo[1]benzothieno[3,2-b]benzothiophene30 encouraged us to further functionalize this semiconductor with N-coordinating cyano groups. The resulting 2,7-dicyano[1]benzothieno[3,2-b]benzothiophene (cBTBT) was deposited on a Ag(111) surface at room temperature under ultrahigh vacuum condition leading to a self-assembled crystalline network. A subsequent deposition of iron atoms on the surface containing the selfassembled molecular layer resulted in the formation of a disordered metal−organic network. Annealing leads to the segregation of metal clusters and the organic molecule and the reformation of the organic network. A theoretical model based on Helmholtz’ free energy is proposed that helps to explain the relative stability of the amorphous state as compared to the crystalline state, and why the segregated structure is thermodynamically more stable.
Figure 1. (a) Molecular structure of prochiral cBTBT and its two enantiomers. (b) Scheme of experimental sequence: self-assembly of cBTBT, formation of amorphous network through addition of Fe, and its segregation upon annealing. (c) Self-assembled structures of cBTBT on Ag(111) formed at room-temperature deposition (U = −842 mV, I = 0.12 nA). Unit cell is depicted in blue. The right inset shows a zoom of the self-assembled structure (unit cell in light blue). Carbon atoms are shown in gray, nitrogen in blue, sulfur in yellow, and hydrogen in white.
■
EXPERIMENTAL METHODS 2,7-Dicyano[1]benzothieno[3,2-b]benzothiophene was synthesized from 2,7-dibromo[1]benzothieno[3,2-b]benzothiophene30 as reported in the Supporting Information. STM measurements were performed in an ultrahigh vacuum (UHV, p < 10−9 mbar) with a home-built STM. The Ag(111) crystal was cleaned by sputtering with Ar+ ions and thermal annealing at 540 °C. The molecule was evaporated from a Knudsen cell at ∼200 °C, while iron was deposited from a standard e-beam evaporator (Omicron GmbH). The flux was calibrated using a combination of STM and a quartz crystal microbalance. All STM data were acquired with etched tungsten tips; bias is reported with respect to the sample.
the formation of disordered metal−organic networks (Figure 2a and b). Previous studies of linear cyano-functionalized molecules with cobalt report the formation of ordered networks with hexagonal unit cells where each Co atom coordinates to three molecules that form an angle of 120° between each other.31,32 Irregularities in the networks can be enforced through the deposition of more molecules than required for a saturated honeycomb mesh, resulting in 4-, 5-, and 6-fold coordination of the Co atoms and leading to pores of various shapes.32 In our STM images, we observe an apparent random tessellation of the surface with differently shaped polygons. For further insight and comparability with other amorphous networks, a statistical analysis of over 1000 pores within the network was performed. The mean value of the number of edges is 4.01. The distribution of triangles, quadrilaterals, pentagons, etc., is depicted in Figure 2c as a histogram. The edge number does not follow a normal distribution around the mean value but rather shows a positive skewness. The second moment (variance; μ2) and third moment (skewness; μ3) of the distribution give values of μ2 = 0.74 and μ3 = 0.61, respectively (calculated with 4 as mean value). As compared to other 2D amorphous materials,1 the variance as a measure of spread of our distribution falls in the middle of the two extremes Cu2O5 (μ2 = 0.42) and amorphous SiO21 (μ2 = 1.06) and is close to amorphous graphene33 (μ2 = 0.78). The skewness as a measure of the asymmetry of the distribution about its mean is large as compared to the other amorphous materials, the largest value being in amorphous SiO2 (μ3 = 0.67) and the smallest in Cu2O
■
RESULTS AND DISCUSSION The ditopic prochiral molecule cBTBT contains two fused thiophene rings and is functionalized with two cyano groups (Figure 1a) suitable for the coordination to single metal atoms. The two-step synthesis procedure of the amorphous network is depicted in Figure 1b, along with its segregation in the final annealing step. First, evaporation of cBTBT onto the clean Ag(111) surface held at room temperature under ultrahigh vacuum condition leads to its self-assembly into a crystalline network. The chevron-like structure contains both enantiomers of cBTBT (Figure 1c). The structure is stabilized by intermolecular hydrogen bonding between terminal nitrogen atoms in the cyano group as well as the electron-rich sulfur atoms and hydrogen atoms of the aryl groups of adjacent molecules. The unit cell is shown in blue and measures a = 1.6 nm, b = 1.1 nm, and γ = 96°. Ball-and-stick models are depicted as an overlay on top of the STM image (Figure 1c). In the second step, the introduction of iron atoms on the surface containing the self-assembled molecular layer results in 4404
DOI: 10.1021/acs.jpcc.5b12406 J. Phys. Chem. C 2016, 120, 4403−4409
Article
The Journal of Physical Chemistry C
numbers is shown in Figure 2d, from which a mean value of 3.66 is calculated. Variance and skewness are 0.42 and −0.37, respectively. The occurrence of a smaller experimental mean value as compared to the theoretically predicted value of 4 is largely influenced by the Fe centers found at the boundary of the islands, which is commonly made up of metal centers to which three molecules are coordinated, two making up the border and one pointing inward into the metal−organic structure. A large number of 3-fold coordinated metal centers as compared to 4-fold centers within the metal−organic network excludes the shortage of metal atoms as a possible cause of the disordered structure of the network. Additional iron clusters found on the surface alongside the amorphous network additionally support the hypothesis that sufficient iron atoms are always available for the formation of crystalline structures. The noncrystalline nature of the network is rather a consequence of similar coordination energies for 3-fold and 4fold coordination of the Fe atoms, which is apparent from the mere 2-fold increased occurrence of 4-fold nodes as compared to the 3-fold nodes. The relative energetic stability of the 4-fold node can be calculated assuming a Boltzmann distribution for the coordination number, yielding larger stability of the 4-fold coordination by only 13 meV. For cobalt atoms coordinated to cyano-functionalized polyphenylene molecules, the 3-fold coordination bond is stronger by 90 meV than the 4-fold coordination motif.31 This energy difference is sufficient to significantly favor one binding motif (3-fold) over the other (4fold), leading to ordered hexagonal networks. Another reason for the amorphous structure is the large geometric flexibility of the coordination nodes, which is manifested in angles different from 90° between two molecules in 4-fold coordinated centers, and angles different from 120° in 3-fold coordinated centers. The interplay between the two contributions, variable coordination number and structural flexibility within the node, results in the noncrystalline nature of the network that can be described by a pair correlation function (PCF, Figure 2e). The normalized PCF shows only two broad peaks, one centered at 1.4 nm, the expected Fe−Fe distance between two coordination nodes connected by one molecule, and the other at 2.8 nm, corresponding to second-next-nearest neighbors. Because of the randomness of the network in polygon size and angle between molecules forming the same coordination node, no additional clear peaks are visible in the PCF except for some minor modulation for Fe−Fe distances larger than 4 nm. The absence of the next-nearest neighbor peak expected at √2 × 1.4 nm = 2.0 nm in a square grid is particularly surprising and stems from the angular flexibility. Even minor deviations from the optimal 90° angle in a 4-fold coordination node alter the diagonal distance within a square to sufficiently eliminate its contribution to the PCF. The absence of long-range order as observed here is typical for glassy systems and can be interpreted as a sign of a kinetically trapped structure. This poses the question whether a thermodynamically more stable crystalline structure can be created by applying thermal energy to overcome kinetic barriers. Interestingly, annealing the sample does not lead to a more ordered crystalline metal−organic network or the reorganization in an amorphous state, but the segregation of the two components into pure domains. Annealing at temperature higher than 370 K results in the rearrangement of Fe atoms in the network into small all-metal clusters surrounded by molecules (Figure 3a) and larger clusters (Figure 3b, black
Figure 2. (a) Overview STM image of amorphous metal−organic network of cBTBT and Fe (U = −1 V, I = 0.1 nA). (b) Highresolution STM image of the black rectangle in (a). cBTBT molecules (white bars) and Fe atoms (white circles) can be observed forming polygons of different size (U = −1 V, I = 0.1 nA). (c) Polygon edge number distribution with a mean value of 4.01. (d) Distribution of the number of cBTBT molecules coordinating to a Fe atom. The mean coordination number is 3.66. (e) Normalized pair correlation function for the metal centers in the coordination nodes.
close to 0. The origin of the large skewness in our sample is due to the small mean value of 4, which results in the triangle being the only possible smaller polygon. The distribution of polygons is skewed to the right due to the presence of pentagons, hexagons, and larger polygons, which outweigh the rather large number of triangles. The different 2D amorphous materials characterized in ref 1 have a mean value of 6 for the polygon edge number, with at least two possible polygons with smaller edge number (pentagons and quadrilaterals) that can make the distribution more symmetric than in our case. The maximum in the histogram is found for the four-sided polygon, which together with the mean value of 4.01 implies a preference for rhombi or square tessellation of the surface. This remarkably singularizes our network from the other glassy networks, in which 3-fold coordination is the most reported motif. We found 4-fold tessellation is possible if the coordination number of each Fe atom within the network is also equal to 4. Note that also metal dimers,34 trimers,35 and possibly larger clusters could constitute the coordination nodes, and that the coordination number could depend on cluster size. However, there is little experimental evidence to support this hypothesis, and literature reports unanimously agree that cyano ligands coordinate to only one metal center.8,31,32,36 The distribution of coordination 4405
DOI: 10.1021/acs.jpcc.5b12406 J. Phys. Chem. C 2016, 120, 4403−4409
Article
The Journal of Physical Chemistry C
layer that are stabilized and separated from the surrounding by cBTBT molecules. The small number of metal atoms renders 2D cluster growth energetically favorable over island growth into the third dimension, which dominates upon adding additional Fe atoms. Metal−organic networks with “metal−cyano” coordination motifs are usually imaged at low temperature due to their limited thermal stability31,32,38 with some notable exceptions.39,40 More stable networks are based on metal− carboxylate coordination bonds and can readily be studied at room temperature (RT)34,41,42 and even higher temperatures.43 In addition, the dissolution of Co clusters upon reacting with carboxylated molecules can be followed in real time at RT.44 Thermal annealing steps can be required to force the networks into their thermodynamic most stable state.45 It is thus clear that an intricate interplay between molecular functionalization and metal center within the network governs thermal stability and the formation of a given crystal structure. Although the phase segregation, to the best of our knowledge, is not addressed in the literature on surface-supported metal−organic networks, the organic phase is commonly deposited before metal sublimation to ensure the growth of networks. The deposition of the metal prior to the molecules in a first instance leads to metal clusters, which often cannot be broken up through thermal annealing, and thus no metal−organic phase can be formed. In the following, we rationalize this observation on the basis of thermodynamic arguments. To understand the disordered morphology and the segregation of the metal−organic phase, entropic and enthalpic contributions in both amorphous and phase-segregated structures can be estimated. The segregated structure appears to be thermodynamically favored over the amorphous one as it prevails at room temperature after thermal annealing. Entropy enters the thermodynamic description as the configurational entropy of the amorphous structure, which reflects the number of ways molecules and metal atoms can arrange on the surface. For a given and fixed probability distribution pn (for example, pn from Figure 2d), where n is the number of coordinated molecules, any possible geometric realization of 2-, 3-, or 4-fold and higher coordination sites on the surface has the same internal energy and entropy. The configurational entropy can be written as Sconf = −kB∑pn ln pn, where kB is Boltzmann’s constant. Translational and rotational entropies are zero for immobile species on the surface, and intramolecular vibrational entropy of the molecule is assumed to be not affected by selfassembly.46,47 These three contributions are hence neglected in the following discussion. However, vibrational entropy also acquires contributions from vibrations of the molecule perpendicular to the surface,48 vibrations in the coordination node,49 and intermolecular modes in the self-assembled structure,50 which are discussed further below. The Helmholtz free energy of the network can be expressed as A = −TSconf + U = kBT∑pn ln pn − ∑pnεn, where εn is the energy of one coordination node with n ligands, T is the temperature, and U is the internal energy. The first term in the equation corresponds to the entropic contribution to the Helmholtz free energy, and the second represents the average internal energy per coordination node. For the probability distribution pn and at a given value ε4, corresponding to the most abundant coordination motif n = 4 and thus the largest stabilizing contribution, the internal energy is expressed as U = −(ε4 − 0.0081 eV) (see the Supporting Information for details). The amorphous state is hence always accompanied by a non-
arrows). Supramolecular networks of cBTBT self-assemble in the remaining parts of the surface.
Figure 3. (a) Annealing at 420 K results in the segregation of the network into the regions of pure metal islands (indicated by blue dots) surrounded by cBTBT in a radial geometry (black bars). The selfassembled structure of cBTBT can be observed in the upper left part (U = 0.99 V, I = 0.1 nA). (b) Self-assembled molecular domains and large Fe clusters (white protrusions pointed out by black arrows) are spatially separated after annealing at 370 K.
The metal clusters formed after the separation of metal and organics in Figure 3a measure a few nanometers in diameter and are surrounded by cBTBT. The molecules bind almost perpendicular to the tangent of the metal clusters, allowing as many cyano−metal contacts as sterically possible, thus stabilizing the energetically unfavorable metal atoms at the periphery of a cluster. The second cyano group of thus coordinated cBTBT molecules is likely binding to the substrate.37 The metal clusters in Figure 3b extend spatially into the third dimension away from the surface as indicated by their large contrast in the STM image. Contrary, the smaller clusters in Figure 3a radially surrounded by molecules appear flat and are probably built up from a single layer of Fe atoms. These clusters are likely intermediate structures in the transition from metal−organic network to the phase-segregated 4406
DOI: 10.1021/acs.jpcc.5b12406 J. Phys. Chem. C 2016, 120, 4403−4409
Article
The Journal of Physical Chemistry C favorable internal energy as compared to the crystalline state with 4-fold coordination nodes. The configurational entropy Sconf of the amorphous network equals 7.1 × 10−5 eV K−1 and for temperatures greater than 115 K compensates the reduced internal energy rendering the amorphous network more stable. At room temperature, TSconf = 0.021 eV, which is roughly 2.5 times larger than the internal energy penalty caused by disorder. The disordered network is stabilized by a small but relevant entropic contribution. An additional entropic stabilization from the angular flexibility within the coordination nodes further favors the disorder over crystallinity, which was not taken into account here.18 The vibrational entropy Svib has contributions from molecular vibrations normal to the surface with an vibrational energy of the order of 10 meV48 and contributions from vibrations in the coordination node ranging from 25 to 75 meV49 (see also the Supporting Information). The vibrational entropy can be calculated using the Einstein equation46 and is of the same order as the configurational entropy. However, these contributions are the same in the amorphous and hypothetical crystalline network as the number of molecules is the same in both cases. This leaves the number of vertical modes unchanged, and because the mean coordination numbers in the amorphous structure as well as the crystalline network are almost the same (3.66 and 4, respectively), intracoordination vibrational modes are also unaltered. In both of the networks, TSvib ≈ 4 × 0.05 eV, dominated by the soft vertical modes, and thus giving no preference to either ordered or disordered network structure. For the segregation to be feasible, the Helmholtz free energy of the phase-segregated state needs to be lower than that in the disordered state. In the crystalline separated state with hypothetically defect-free structures, the entropy is zero and A = Umolecules + Umetal, where U is the internal energy in the molecular layer and the metal clusters. Typical coordination bond energies of ε are 0.3−0.6 eV,51−53 while energies for hydrogen bonds range from 0.05 to 0.7 eV in self-assembled molecular monolayers, and from 1 to 5 eV for metal−metal bonds.52,54 Iron within a metal cluster has a binding energy of about 4.4 eV,54 which clearly suggests that the segregation is strongly enthalpic in origin, driven by the very large binding energy within a metal cluster. Removing one iron center from a coordination node (energy penalty
