Self-Assembly of Covalently Linked Porphyrin Dimers at the Solid–Liquid Interface

The synthesis and surface self-assembly behavior of two types of metal-porphyrin dimers is described. The first dimer type consists of two porphyrins linked via a rigid conjugated spacer, and the second type has an alkyne linker, which allows rotation of the porphyrin moieties with respect to each other. The conjugated dimers were equipped with two copper or two manganese centers, while the flexible dimers allowed a modular built-up that also made the incorporation of two different metal centers possible. The self-assembly of the new porphyrin dimers at a solid–liquid interface was investigated at the single-molecule scale using scanning tunneling microscopy (STM). All dimers formed monolayers, of which the stability and the internal degree of ordering of the molecules depended on the metal centers in the porphyrins. While in all monolayers the dimers were oriented coplanar with respect to the underlying surface (‘face-on’), the flexible dimer containing a manganese and a copper center could be induced, via the application of a voltage pulse in the STM setup, to self-assemble into monolayers in which the porphyrin dimers adopted a non-common perpendicular (‘edge-on’) geometry with respect to the surface.


Introduction
Metal-porphyrins occur ubiquitously in nature, e.g., as magnesium porphyrin pigments in the light harvesting system [1,2] and as iron porphyrins in the blood protein hemoglobin [3] and the oxidation enzyme cytochrome P450 [4,5]. In biomimetic chemistry, many artificial porphyrin systems have been developed with the aim to approach the efficient functional properties of their natural counterparts, for instance in areas of molecular electronics [6][7][8] and catalysis [9][10][11]. In the past decades, our group has extensively investigated the catalytic properties of Mn(III) porphyrins in the epoxidation of alkenes [12][13][14][15][16]. The majority of these studies have been carried out using supramolecular systems in solution, and in order to obtain more detailed information about the reaction mechanisms during catalysis, we have also investigated the behavior of single porphyrin molecules adsorbed at a surface. To this end, we constructed self-assembled monolayers of metal-porphyrins at a solid-liquid interface and investigated their properties with scanning tunneling microscopy (STM) [17][18][19][20]. This technique does not only enable the probing of structural properties of single molecules but also provides electronic information, e.g., about changes in the redox state of the porphyrin metal center during a chemical reaction. Furthermore, a solid-liquid interface environment closely resembles the reaction conditions encountered in laboratory-scale chemical reactions: Room temperature, atmospheric pressure, and the presence of a liquid to which reagents can easily be added and products withdrawn. For these reasons,

Synthesis
Conjugated porphyrin dimer H41 was synthesized in five steps, starting from 4hydroxybenzaldehyde, which was alkylated with 1-bromooctadecane to give 3 in 78% yield (Scheme 1). This compound was subsequently condensed with pyrrole to give porphyrin 4 in 38% yield. Using a modified previously reported protocol [40], 4 was regioselectively dihydroxylated at one of its βpyrrole moieties using osmium tetraoxide to give diol 5 in 41% yield. This compound was oxidized with 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) to diketone 6 in 38% yield. Finally, two equivalents of 6 were condensed with one equivalent of 1,2,4,5-tetraaminobenzene to give the desired bis-porphyrin H41 in 38% yield. Compound H41 could be readily converted into Cu21 by reacting it with copper(II) acetate, and into Mn22 by reacting it with manganese(II) acetate tetrahydrate followed by aerobic oxidation in brine (both in nearly quantitative yields). The synthesis of acetylene-linked porphyrin dimer H42 and its metal complexes Cu22 and MnCu2 has been reported by us previously [39]. Bis-manganese dimer Mn22 was synthesized in nearly quantitative yield from H42 following the same procedure as for the synthesis of Mn21. The progress of all-metal insertion reactions was monitored by UV-vis spectroscopy, in which chloroform solutions of the reaction mixtures were acidified with concentrated aqueous HCl; as soon as absorptions correlated to protonated free base porphyrins remained absent, the metallation was considered complete and the reaction stopped (see Supplementary Materials for spectral data of all new compounds).

Synthesis
Conjugated porphyrin dimer H 4 1 was synthesized in five steps, starting from 4-hydroxybenzaldehyde, which was alkylated with 1-bromooctadecane to give 3 in 78% yield (Scheme 1). This compound was subsequently condensed with pyrrole to give porphyrin 4 in 38% yield. Using a modified previously reported protocol [40], 4 was regioselectively dihydroxylated at one of its β-pyrrole moieties using osmium tetraoxide to give diol 5 in 41% yield. This compound was oxidized with 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) to diketone 6 in 38% yield. Finally, two equivalents of 6 were condensed with one equivalent of 1,2,4,5-tetraaminobenzene to give the desired bis-porphyrin H 4 1 in 38% yield. Compound H 4 1 could be readily converted into Cu 2 1 by reacting it with copper(II) acetate, and into Mn 2 2 by reacting it with manganese(II) acetate tetrahydrate followed by aerobic oxidation in brine (both in nearly quantitative yields). The synthesis of acetylene-linked porphyrin dimer H 4 2 and its metal complexes Cu 2 2 and MnCu2 has been reported by us previously [39]. Bis-manganese dimer Mn 2 2 was synthesized in nearly quantitative yield from H 4 2 following the same procedure as for the synthesis of Mn 2 1. The progress of all-metal insertion reactions was monitored by UV-vis spectroscopy, in which chloroform solutions of the reaction mixtures were acidified with concentrated aqueous HCl; as soon as absorptions correlated to protonated free base porphyrins remained absent, the metallation was considered complete and the reaction stopped (see Supplementary Materials for spectral data of all new compounds).

Self-Assembly of Conjugated Porphyrin Dimers Cu21 and Mn21 at a Solid-Liquid Interface
In Figure 2, STM images of self-assembled monolayers of Cu21 and Mn21 at the highly oriented pyrolytic graphite (HOPG)/1-octanoic acid interface are shown. Droplets of 3-5 μL with concentrations of ~10 −5 M of the compounds in 1-octanoic acid were applied to a freshly cleaved HOPG surface. Only in approximately 20% of the attempts, monolayers could be successfully imaged at (sub)molecular resolution. In all other cases, typically poorly resolved structures were found revealing at best the presence of vague lamellar arrays. This is probably due to incomplete assembly of the molecules in a fully packed layer, or to fast adsorption/desorption kinetics [41][42][43] of the molecules at the solid-liquid interface, preventing the formation of a layer that is immobilized enough to be resolvable by STM.
The orientation of the dimers of Cu21 within the self-assembled monolayer can be derived from the image in Figure 2A. Based on their shape and size, and taking into account the arrangement of the side-groups, each of the bright-appearing (i.e., elevated) spots is assigned to a porphyrin moiety that is oriented with its aromatic plane in a largely coplanar geometry with respect to the surface. Consequently, two of such signatures may constitute one porphyrin dimer molecule. The dimers are aligned with their long axis along the lamellae, as is shown in the proposed structural model in Figure  2E. This organization was concluded by considering the distances between the porphyrin moieties along and perpendicular to the lamella, respectively, especially the observation of alternating small and large spacings. The unit cell of the monolayer (drawn in Figure 2A) is defined by vectors a = 3.5 ± 0.2 and b = 3.9 ± 0.2 nm, with an angle of 79 ± 8° between them. Based on the distance between the lamellae, the C18-alkyl chains of adjacent porphyrin dimers have to be interdigitated. In some parts of the image, these chains are discernable in the darker regions between the bright signatures (inset in Figure 2A), and careful analysis revealed that they are fully adsorbed on the HOPG surface in an interdigitated geometry. They are, however, not always ordered alternately, i.e., one chain originating from a dimer in a lamellar array is not necessarily followed on the surface by a chain of a dimer from an adjacent lamella. This irregular distribution is also shown in the model in Figure 2E. The chains are organized in pairs as a result of the limited spacing between the porphyrin moieties within one dimer, which does not allow interdigitation of a chain from a dimer in an adjacent lamella. Scheme 1. Synthesis of porphyrin dimers with a rigid conjugated linker, and their metal derivatives.

Self-Assembly of Conjugated Porphyrin Dimers Cu 2 1 and Mn 2 1 at a Solid-Liquid Interface
In Figure 2, STM images of self-assembled monolayers of Cu 2 1 and Mn 2 1 at the highly oriented pyrolytic graphite (HOPG)/1-octanoic acid interface are shown. Droplets of 3-5 µL with concentrations of~10 −5 M of the compounds in 1-octanoic acid were applied to a freshly cleaved HOPG surface. Only in approximately 20% of the attempts, monolayers could be successfully imaged at (sub)molecular resolution. In all other cases, typically poorly resolved structures were found revealing at best the presence of vague lamellar arrays. This is probably due to incomplete assembly of the molecules in a fully packed layer, or to fast adsorption/desorption kinetics [41][42][43] of the molecules at the solid-liquid interface, preventing the formation of a layer that is immobilized enough to be resolvable by STM.
The orientation of the dimers of Cu 2 1 within the self-assembled monolayer can be derived from the image in Figure 2A. Based on their shape and size, and taking into account the arrangement of the side-groups, each of the bright-appearing (i.e., elevated) spots is assigned to a porphyrin moiety that is oriented with its aromatic plane in a largely coplanar geometry with respect to the surface. Consequently, two of such signatures may constitute one porphyrin dimer molecule. The dimers are aligned with their long axis along the lamellae, as is shown in the proposed structural model in Figure 2E. This organization was concluded by considering the distances between the porphyrin moieties along and perpendicular to the lamella, respectively, especially the observation of alternating small and large spacings. The unit cell of the monolayer (drawn in Figure 2A) is defined by vectors a = 3.5 ± 0.2 and b = 3.9 ± 0.2 nm, with an angle of 79 ± 8 • between them. Based on the distance between the lamellae, the C 18 -alkyl chains of adjacent porphyrin dimers have to be interdigitated. In some parts of the image, these chains are discernable in the darker regions between the bright signatures (inset in Figure 2A), and careful analysis revealed that they are fully adsorbed on the HOPG surface in an interdigitated geometry. They are, however, not always ordered alternately, i.e., one chain originating from a dimer in a lamellar array is not necessarily followed on the surface by a chain of a dimer from an adjacent lamella. This irregular distribution is also shown in the model in Figure 2E. The chains are organized in pairs as a result of the limited spacing between the porphyrin moieties within one dimer, which does not allow interdigitation of a chain from a dimer in an adjacent lamella. Based on the orientation of the dimers, one would expect to observe an even number of bright signatures along each lamella within a monolayer domain. Overall, this is indeed observed in the monolayer of Cu 2 1, for example when the lamella along line 1 in Figure 2A is considered. However, the lamella along line 2 reveals an odd number of bright signatures. The cross-sections of both lines are shown in Figure 2C. Cross-section 1 shows the apparent height profile of six dimers, which show up as twelve protrusions, each corresponding to a porphyrin moiety within the dimers. The gray boxes indicate the separate dimers. The cross-section furthermore shows that the depression in apparent height between the porphyrin moieties in a dimer is less deep than the depression between the porphyrin moieties of two adjacent dimers, which is the basis of our assignment. If we now consider cross-section 2, dimers can also be assigned based on the apparent height of the porphyrin moieties. Again the gray boxes indicate which of the protrusions form a dimer. Based on this analysis, the 5th protrusion from the left, with a width of about 1.8 nm, is not a fragment of a dimer. We, therefore, assign this feature to a vacant site, although the small protrusion may be caused by one or more trapped and poorly resolved solvent molecules [44,45]. The model in Figure 2E also shows such a vacancy. Since the alkyl chains are not ordered in an alternating fashion, the next bright signature available is at a distance half a dimer distance further, at 1 2 a = 1.75 ± 0.1 nm, which is in agreement with the width of the described vacancy feature. Based on the orientation of the dimers, one would expect to observe an even number of bright signatures along each lamella within a monolayer domain. Overall, this is indeed observed in the monolayer of Cu21, for example when the lamella along line 1 in Figure 2A is considered. However, the lamella along line 2 reveals an odd number of bright signatures. The cross-sections of both lines are shown in Figure 2C. Cross-section 1 shows the apparent height profile of six dimers, which show up as twelve protrusions, each corresponding to a porphyrin moiety within the dimers. The gray boxes indicate the separate dimers. The cross-section furthermore shows that the depression in apparent height between the porphyrin moieties in a dimer is less deep than the depression between the porphyrin moieties of two adjacent dimers, which is the basis of our assignment. If we now consider cross-section 2, dimers can also be assigned based on the apparent height of the porphyrin moieties. Again the gray boxes indicate which of the protrusions form a dimer. Based on this analysis, the 5th protrusion from the left, with a width of about 1.8 nm, is not a fragment of a dimer. We, therefore, assign this feature to a vacant site, although the small protrusion may be caused by one or   Figure 2B shows an STM image of a self-assembled monolayer of Mn 2 1 at the HOPG/1-octanoic acid interface. Although the resolution is not as good as that of the monolayer of Cu 2 1, it is clear that also Mn 2 1 self-assembles into a lamellar structure, with unit cell vectors a = 3.4 ± 0.4 and b = 3.9 ± 0.3 nm, under an angle of 80 • ± 10 • . Within experimental error, these values are the same as those of the unit cell of Cu 2 1, indicating that the substitution of the metal centers does not change the spatial ordering of the dimers on the HOPG surface. The difference in resolution between the monolayers of Cu 2 1 and Mn 2 1 may be explained by the presence of the axial chloride ligand at the manganese centers of Mn 2 1. This ligand may hinder the adsorption of the dimers at the surface, or alter the tunneling pathway between the surface and the tip. The self-assembled monolayers of Mn 2 1 appeared to be not very stable. They only persisted when the tunnel current was maintained below 10 pA, and even then they typically disappeared after about 30 min of scanning. Although the resolution is not optimal, the signatures of some of the manganese porphyrin moieties appear to differ in brightness. The cross-section in Figure 2D shows an apparent height difference of 0.1 nm between bright and less bright appearing porphyrins. The brighter signatures have a radius of 1.2 nm, and thus span only one manganese porphyrin, not the whole dimer structure. Due to the lack of sub-molecular resolution, it is not possible to determine whether more than two different signatures are present in the monolayer. So far, the origin of the different signatures is unclear. They might be the result of redox reactions occurring at the manganese centers. Unlike the Cu(II) centers of Cu 2 1, the Mn(III) centers of Mn 2 1 may undergo reduction to Mn(II), or oxidation to Mn(IV) upon reaction with oxygen, as has been demonstrated in our previous STM work on monomeric manganese porphyrins [25,26]. Such reactions would most certainly lead to a difference in the appearance of the porphyrins in the STM images. An alternative explanation would be that the two porphyrins within an Mn 2 1 dimer adsorb in a different geometry to the underlying surface. Since the manganese centers have an axially coordinated chloride ligand, Mn 2 1 can exist in two isomeric forms, at least in principle. Assuming kinetically stable coordination of the inner shell chloride ligand to the metal center, a molecule of Mn 2 1 can have both chloride ligands coordinated at the same (syn-isomer) or at the opposite face (anti-isomer) of a porphyrin dimer ( Figure 2F). Due to the paramagnetic nature of the manganese centers, the existence or abundance of these isomers could not be investigated with the help of NMR spectroscopy. The rigid, conjugated linker between the porphyrin moieties furthermore inhibits isomerization, via rotation, from the synto the anti-isomer and vice versa. Assuming the existence of both isomers, one can envisage that they will adsorb in different geometries at the solid-liquid interface. While the conjugated porphyrin plane of the syn-isomer can adsorb in a cofacial geometry with respect to the surface, this is impossible for the anti-isomer since one axial chloride ligand will always inhibit this geometry. In the latter case, the dimer may be adsorbed in a geometry in which it is lifted on the side where the chloride ligand points in the direction of the surface. As a result of this tilted geometry, the amount of π−π stacking and van der Waals interactions between the surface and the molecules is not ideal, which may be an explanation for the limited stability of the monolayer structure in time.

Self-Assembly of the Non-Conjugated Porphyrin Dimers Cu 2 2, Mn 2 2, and MnCu2 at a Solid-Liquid Interface
In a previous paper [39] we reported the self-assembly properties of Cu 2 2 and MnCu2 at the solid-liquid interface. Both compounds formed stable and highly organized monolayers, and in some cases, bilayer formation was observed for MnCu2, in which the dimers were stacked in two 2D layers on top of each other. The top 2D layer was found to be able to laterally move stepwise with respect to the bottom 2D layer, which was strongly immobilized to the HOPG surface. The fact that the molecules of MnCu2 formed stable and extended uniform 2D layers of molecules that were oriented with their porphyrin planes coplanar with the underlying HOPG substrate, indicates that the single axial chloride ligand at the manganese centers of these molecules does not hinder such an adsorption geometry. It was therefore of interest to investigate the assembly behavior of Mn 2 2, which has two of these chloride ligands.
It turned out to be very difficult to obtain assemblies of Mn 2 2 at the HOPG/1-octanoic acid interface that were stable enough to be imaged with STM. In most cases, only vague and very dynamic structures were observed and no reliable unit cell could be determined. Only sparingly, small patches (<30 × 30 nm 2 ) emerged, which were not as close-packed as those of Cu 2 2 or MnCu2 (Figure 3). The layers were even more unstable than those of the rigid conjugated dimer Mn 2 1 and typically disappeared within several minutes. Like Mn 2 1, also Mn 2 2 can exist as isomers in which the two chloride ligands are oriented in different directions with respect to each other. However, unlike in the case of Mn 2 1, the isomers of Mn 2 2 can interconvert since the two porphyrin moieties can freely rotate with respect to each other about the alkyne linker. The fact that MnCu2 forms extended and stable assemblies and Mn 2 2 does not, and the fact that also layers of Mn 2 1 are unstable, suggests that the presence of two axial chloride ligands within a given manganese porphyrin dimer inhibits the formation of stable layers at a surface. This may be the result of a higher stability of dimers with an anti-configuration of the two ligands compared to those with a syn-configuration, which in turn may be governed by a lower dipole moment in the former isomers.
Molecules 2019, 24, x 7 of 15 geometry. It was therefore of interest to investigate the assembly behavior of Mn22, which has two of these chloride ligands.
It turned out to be very difficult to obtain assemblies of Mn22 at the HOPG/1-octanoic acid interface that were stable enough to be imaged with STM. In most cases, only vague and very dynamic structures were observed and no reliable unit cell could be determined. Only sparingly, small patches (<30 × 30 nm 2 ) emerged, which were not as close-packed as those of Cu22 or MnCu2 (Figure 3). The layers were even more unstable than those of the rigid conjugated dimer Mn21 and typically disappeared within several minutes. Like Mn21, also Mn22 can exist as isomers in which the two chloride ligands are oriented in different directions with respect to each other. However, unlike in the case of Mn21, the isomers of Mn22 can interconvert since the two porphyrin moieties can freely rotate with respect to each other about the alkyne linker. The fact that MnCu2 forms extended and stable assemblies and Mn22 does not, and the fact that also layers of Mn21 are unstable, suggests that the presence of two axial chloride ligands within a given manganese porphyrin dimer inhibits the formation of stable layers at a surface. This may be the result of a higher stability of dimers with an anti-configuration of the two ligands compared to those with a syn-configuration, which in turn may be governed by a lower dipole moment in the former isomers.

2D Polymorphism of Porphyrin Dimer MnCu2
During the STM measurements of porphyrin dimer MnCu2, we observed an interesting case of 2D polymorphism. In less than 10% of the attempts to image a self-assembled structure of this compound, no structures were observed at all at the HOPG surface. It is then common practice to induce monolayer formation by electric polarization, applying one or more mild voltage pulses to the tunnel junction, in this case between +1 and +2 V with a duration of 900 μs. In about 5% of these attempts, within 5 min a monolayer structure as shown in the STM images in Figures 4A-B grew over the whole scan area. After the other attempts, no structure emerged or the STM tip became too blunt to be used for further imaging. Although the STM images are inconclusive about the precise orientation of the dimers, it is plausible that they are oriented with their porphyrin planes perpendicular ('edge-on') to the surface, as is drawn in the schematic representation in Figure 4C, in which each rectangle corresponds to one porphyrin moiety within a dimer. 2D polymorphic changes upon the application of a voltage pulse have been reported previously for other types of disk-shaped molecules adsorbed at a solid-liquid interface [46,47].
Similar as was observed in the other STM images of the porphyrin dimers, we propose that the bright lamellae correspond to the aromatic parts of MnCu2 and the darker regions to the alkyl chains. Similar `edge-on' orientation geometries at the solid-liquid interface have previously been observed for porphyrin oligomers containing six [17] and twelve porphyrin moieties [18,19], phthalocyanine derivatives [48], and other planar molecules with extended aromatic surfaces [49][50][51][52][53]. The formation

2D Polymorphism of Porphyrin Dimer MnCu2
During the STM measurements of porphyrin dimer MnCu2, we observed an interesting case of 2D polymorphism. In less than 10% of the attempts to image a self-assembled structure of this compound, no structures were observed at all at the HOPG surface. It is then common practice to induce monolayer formation by electric polarization, applying one or more mild voltage pulses to the tunnel junction, in this case between +1 and +2 V with a duration of 900 µs. In about 5% of these attempts, within 5 min a monolayer structure as shown in the STM images in Figure 4A-B grew over the whole scan area. After the other attempts, no structure emerged or the STM tip became too blunt to be used for further imaging. Although the STM images are inconclusive about the precise orientation of the dimers, it is plausible that they are oriented with their porphyrin planes perpendicular ('edge-on') to the surface, as is drawn in the schematic representation in Figure 4C, in which each rectangle corresponds to one porphyrin moiety within a dimer. 2D polymorphic changes upon the application of a voltage pulse have been reported previously for other types of disk-shaped molecules adsorbed at a solid-liquid interface [46,47].
Similar as was observed in the other STM images of the porphyrin dimers, we propose that the bright lamellae correspond to the aromatic parts of MnCu2 and the darker regions to the alkyl chains. Similar 'edge-on' orientation geometries at the solid-liquid interface have previously been observed for porphyrin oligomers containing six [17] and twelve porphyrin moieties [18,19], phthalocyanine derivatives [48], and other planar molecules with extended aromatic surfaces [49][50][51][52][53]. The formation of the 'edge-on' assemblies of MnCu2 is most likely governed by stabilizing π-stacking between the aromatic porphyrin planes of the dimers. Despite the fact that individual dimers can be hardly discerned in Figure 4A-B, in the darker parts between the lamellae signatures are visible, which we propose to be the alkyl chains of the dimer molecules. Based on this assumption, a unit cell was determined with vectors c = 6.2 ± 0.3 and d = 0:53 ± 0.2 nm, under an angle of 48 • ± 5 • . These unit cell dimensions give no clue as to whether the alkyl chains of dimers in adjacent lamella are interdigitated or not. The unit cell parameters correspond to an orientation of the dimers as drawn in Figure 4C, i.e., in which the porphyrin rings are oriented in an off-set geometry, which is typical for π−π stacking of small molecules. From the parameters, a porphyrin stacking distance of about 4 Å can be calculated, which is a realistic value for stacked aromatic surfaces. Remarkably, the bright parts of the lamellae all show a brighter side and a less bright side, which can also be discerned in the cross-section drawn in Figure 4A,D. These differences in brightness may be explained by the presence of different metal centers in the two halves of the porphyrin dimers, and a unidirectional orientation within the lamellae. In the schematic drawing in Figure 4E, the squares representing the porphyrin moieties are shown either filled or empty, corresponding to porphyrin moieties with a manganese or a copper center, respectively. A unidirectional orientation would mean that in a given stack of porphyrin dimers the same metal centers are always aligned behind each other, resulting in an apparent height difference between the two halves of the lamella. Such an ordering might be governed by favorable dipole-dipole interactions between the Mn-Cl metal-ligand pairs of adjacent dimers. However, similar internal differences in lamellar brightness have previously been observed in STM images of porphyrin hexamers [17] and dodecamers [18,19] which contained no or identical metal centers. Another, and probably more likely explanation for the contrast differences in the lamellae of MnCu2 may, therefore, be variations in the orientation of the stacks of the dimers with respect to the atomic ordering of the underlying solid surface. The ordering of the stacks may not be fully commensurate with the ordering of the underlying graphite, leading to the observed differences in the sub-lamellar signature. of the `edge-on' assemblies of MnCu2 is most likely governed by stabilizing π-stacking between the aromatic porphyrin planes of the dimers. Despite the fact that individual dimers can be hardly discerned in Figures 4A-B, in the darker parts between the lamellae signatures are visible, which we propose to be the alkyl chains of the dimer molecules. Based on this assumption, a unit cell was determined with vectors c = 6.2 ± 0.3 and d = 0:53 ± 0.2 nm, under an angle of 48° ± 5°. These unit cell dimensions give no clue as to whether the alkyl chains of dimers in adjacent lamella are interdigitated or not. The unit cell parameters correspond to an orientation of the dimers as drawn in Figure 4C, i.e., in which the porphyrin rings are oriented in an off-set geometry, which is typical for ππ stacking of small molecules. From the parameters, a porphyrin stacking distance of about 4 Å can be calculated, which is a realistic value for stacked aromatic surfaces. Remarkably, the bright parts of the lamellae all show a brighter side and a less bright side, which can also be discerned in the crosssection drawn in Figures 4A,D. These differences in brightness may be explained by the presence of different metal centers in the two halves of the porphyrin dimers, and a unidirectional orientation within the lamellae. In the schematic drawing in Figure 4E, the squares representing the porphyrin moieties are shown either filled or empty, corresponding to porphyrin moieties with a manganese or a copper center, respectively. A unidirectional orientation would mean that in a given stack of porphyrin dimers the same metal centers are always aligned behind each other, resulting in an apparent height difference between the two halves of the lamella. Such an ordering might be governed by favorable dipole-dipole interactions between the Mn-Cl metal-ligand pairs of adjacent dimers. However, similar internal differences in lamellar brightness have previously been observed in STM images of porphyrin hexamers [17] and dodecamers [18,19] which contained no or identical metal centers. Another, and probably more likely explanation for the contrast differences in the lamellae of MnCu2 may, therefore, be variations in the orientation of the stacks of the dimers with respect to the atomic ordering of the underlying solid surface. The ordering of the stacks may not be fully commensurate with the ordering of the underlying graphite, leading to the observed differences in the sub-lamellar signature.  The assemblies of 'edge-on'-orientated dimers of MnCu2 were exclusively observed after the application of a voltage pulse to the tunnel junction, which strongly indicates that the pulse plays a decisive role in their formation. A first explanation for this may be that the pulse removes the axial chloride ligands from the manganese porphyrin moieties, as has been proposed before for the dissociation of the same ligand from a single manganese porphyrin upon application of a voltage pulse to the interface of HOPG and a 1-octanoic acid solution of that compound [27]. Alternatively, or perhaps also concomitantly, the strong electric field involved with the pulse might align the dimers already in solution and induce the deposition a small seed of 'edge-on' oriented dimers at the solid-liquid interface. Subsequently, other porphyrin dimers can attach to this seed resulting in the growth of extended 'edge-on' patterns over the whole scanning area. The 'edge-on' patterns remained stable over time and no porphyrin dimers that were adsorbed coplanar ('face-on') to the surface were observed.
Remarkably, the scanning by the STM tip seemed to play a role in the growth of the edge-on domains of MnCu2. In the STM image in Figure 5A an area with edge-on oriented assemblies of MnCu2 is shown, directly after scanning it after the application of a voltage pulse. The next STM image ( Figure 5B) is a zoom-out in which the scanning area was enlarged from 150 × 150 to 300 × 300 nm 2 . The square drawn in this image corresponds with the scanning area of Figure 5A. It can be clearly seen that the edge-on assemblies are predominantly present in the confined area of the previous scan. The STM image in Figure 5C shows the same area as in Figure 5B, but now after 20 min of continuous scanning. A large part of the scanning area is now covered with the 'edge-on' assemblies, which indeed indicates that the STM tip plays an active role in their growth and that the domains are formed from a seed of 'edge-on' oriented molecules, as described above. The precise mechanism by which the tip assists in the domain growth is as yet unknown. With the applied tunneling parameters (V bias = −450 mV, I t = 5 pA) the tip is relatively far removed from the surface, minimizing direct physical contact with the monolayer. The porphyrins are not very polar, but to some extent polarizable along their planar axes. The electric field in the tip-sample junction is rather high (in the regime of 10 9 V/m). Polar and polarizable species may form bridges in electric fields, which can act as nuclei for further growth with a molecular orientation of the most polarizable direction along the field direction, i.e., in this case, perpendicular to the surface. This effect is best known for water molecules [54], which are polar, however, a static polarization during the pulse duration may suffice to initiate growth nuclei of polarizable species of such an 'upright' configuration. Furthermore, the movement of the tip may perturb the supernatant solution containing molecules or assemblies of MnCu2 and thereby provide them with energy to self-assemble into ordered arrays at the solid-liquid interface. The assemblies of `edge-on'-orientated dimers of MnCu2 were exclusively observed after the application of a voltage pulse to the tunnel junction, which strongly indicates that the pulse plays a decisive role in their formation. A first explanation for this may be that the pulse removes the axial chloride ligands from the manganese porphyrin moieties, as has been proposed before for the dissociation of the same ligand from a single manganese porphyrin upon application of a voltage pulse to the interface of HOPG and a 1-octanoic acid solution of that compound [27]. Alternatively, or perhaps also concomitantly, the strong electric field involved with the pulse might align the dimers already in solution and induce the deposition a small seed of `edge-on' oriented dimers at the solidliquid interface. Subsequently, other porphyrin dimers can attach to this seed resulting in the growth of extended `edge-on' patterns over the whole scanning area. The 'edge-on' patterns remained stable over time and no porphyrin dimers that were adsorbed coplanar ('face-on') to the surface were observed.
Remarkably, the scanning by the STM tip seemed to play a role in the growth of the edge-on domains of MnCu2. In the STM image in Figure 5A an area with edge-on oriented assemblies of MnCu2 is shown, directly after scanning it after the application of a voltage pulse. The next STM image ( Figure 5B) is a zoom-out in which the scanning area was enlarged from 150 × 150 to 300 × 300 nm 2 . The square drawn in this image corresponds with the scanning area of Figure 5A. It can be clearly seen that the edge-on assemblies are predominantly present in the confined area of the previous scan. The STM image in Figure 5C shows the same area as in Figure 5B, but now after 20 min of continuous scanning. A large part of the scanning area is now covered with the 'edge-on' assemblies, which indeed indicates that the STM tip plays an active role in their growth and that the domains are formed from a seed of 'edge-on' oriented molecules, as described above. The precise mechanism by which the tip assists in the domain growth is as yet unknown. With the applied tunneling parameters (Vbias = −450 mV, It = 5 pA) the tip is relatively far removed from the surface, minimizing direct physical contact with the monolayer. The porphyrins are not very polar, but to some extent polarizable along their planar axes. The electric field in the tip-sample junction is rather high (in the regime of 10 9 V/m). Polar and polarizable species may form bridges in electric fields, which can act as nuclei for further growth with a molecular orientation of the most polarizable direction along the field direction, i.e., in this case, perpendicular to the surface. This effect is best known for water molecules [54], which are polar, however, a static polarization during the pulse duration may suffice to initiate growth nuclei of polarizable species of such an 'upright' configuration. Furthermore, the movement of the tip may perturb the supernatant solution containing molecules or assemblies of MnCu2 and thereby provide them with energy to selfassemble into ordered arrays at the solid-liquid interface.

General Materials and Methods
All commercially obtained chemicals were used without further purification unless stated otherwise. CHCl 3

STM Experiments
Approximately 10 µL droplet of a solution of a porphyrin dimer (concentration~10 −4 M) was applied to a freshly cleaved HOPG substrate (10 × 10 mm 2 , NT-MDT, ZYB) which was mounted into a home-built liquid-STM setup [25]. The STM-tip (mechanically cut from Pt 0.8 Ir 0.2 wire, diameter 0.5 mm) was immersed in the droplet and the surface was scanned at a rate of 300 nm/s. To minimize thermal drift effects, the STM setup was mounted and left to equilibrate for several hours before the solution containing the compound was added. The STM measurements were performed in constant-current mode using an Omicron Scala SPM controller. The images shown were only corrected for background and no additional filters were applied to the raw data. All STM measurements were performed in the thermostatted environment (21.5 ± 0.5 • C) of the NanoLab Nijmegen.

Conclusions
A porphyrin dimer with a rigid aromatic linker was readily synthesized in five steps, and its bis-copper and bis-manganese complexes were obtained nearly quantitatively from the free base precursor. The bis-copper dimer self-assembled into stable monolayers at a solid-liquid interface. The bis-manganese dimer formed monolayers with a similar structure, but these were much less stable than the layers of the bis-copper dimers. Similar unstable layers were observed for bis-manganese porphyrin dimers in which the porphyrins are linked by a less-rigid alkyne spacer. The instability of both layers is attributed to the presence of inner shell axial chloride ligands at the manganese centers, which are believed to hinder a cofacial orientation of the porphyrin planes with respect to the surface.
The application of a mild voltage pulse induced the emergence of an adlayer of a copper-manganese porphyrin dimer with an alkyne spacer, in which the dimer molecules were exclusively oriented with their porphyrin planes perpendicular to the surface. The formation of this surface polymorph was attributed to (i) dissociation of the axial chloride ligand from the manganese center allowing cofacial intermolecular stacking, and/or (ii) electric field-induced preorganization of small stacks in solution which, after adsorption to the surface, serve as seeds for further growth of the edge-on polymorph.
Future research will be directed to investigating the (cooperative) reactivity of the bis-manganese dimers towards molecular oxygen. In addition, these dimers will be equipped with non-coordinating anions (e.g., PF 6 − or BPh 4 − ) via ion-exchange reactions, to improve their adsorption at a solid-liquid interface into more stable adlayers.

Conflicts of Interest:
The authors declare no conflict of interest.