Submolecular imaging of chloronitrobenzene isomers on Cu(111) Abstract

Transcription

1 Submolecular imaging of chloronitrobenzene isomers on Cu(111) Eeva Niemi, 1, Violeta Simic-Milosevic, 2 Karina Morgenstern, 3 Antti Korventausta, 1 Sami Paavilainen, 1 and Jouko Nieminen 1 1 Institute of Physics, Tampere University of Technology, P.O. Box 692, FIN Tampere, Finland 2 Institut für Experimentalphysik, FB Physik, FU Berlin, Arnimallee 14, D Berlin, Germany 3 Institut für Festkörperphysik, Universität Hannover, Appelstr. 2, D Hannover, Germany (Dated: Version of July 14, 2006) Abstract We compare computer simulations to experimental STM images of chloronitrobenzene molecules on Cu(111) surface. The experiments show that adsorption induced isomerisation of the molecules takes place on the surface. Furthermore, not only the submolecular features can be seen in the STM images, but also different isomers can be recognized. Todorov Pendry approach to tunneling produces simulated STM images which are in good accordance with the experiments. Alongside with STM simulations in tight binding basis, ab initio calculations are performed in order to analyze symmetry of relevant molecular orbitals, and to consider the nature of tunneling channels. Our calculations show, that while the orbitals delocalized to the phenyl ring create a relatively transparent tunneling channel, they also almost isolate the orbitals of the substitute groups at energies which are relevant in STM experiments. These features of the electronic structure are the key ingredients of the accurate submolecular observations. PACS numbers: Ef, Fg, d Electronic address: eeva.niemi@tut.fi 1

2 I. INTRODUCTION Scanning tunneling microscopy [1] provides surface scientists the most accurate information available on local molecular or atomic scale. However, the interpretation of molecular STM images is not straightforward, since the internal structure of an STM image does not necessarily reflect the positions of atomic nuclei directly. Rather, the STM images reflect the electronic structure of the combined adsorbate-substrate system, sometimes also influenced by the electronic structure of the microscope tip. The adsorption site also affects the STM image, since mixing of the electronic states of surface and adsorbate depends on it. Although this dependence of the STM image on the electronic structure makes the interpretation of images difficult, it may, on the other hand, be used to obtain more information than the mere geometry. Local electronic structure can be probed using scanning tunneling spectroscopy [2], and changes in intramolecular bonding are revealed by inelastic electronic tunneling spectroscopy (IETS) [3]. In the first approximation, STM measures the electronic structure of the sample near the Fermi level [4]. In a more explanatory analysis, STM images result from interference of tunneling through different tunneling channels [5 7]. In this framework, good qualitative accordance between experimental and simulated images can be found in relatively simple calculations in tight binding (TB) basis. This approach is very useful in controlling how geometrical or chemical configuration affects the STM image of an adsorbate molecule. This article presents both experimental and computational results on substituted benzene molecules on Cu(111) surface. In the adsorption of meta-chloronitrobenzene molecules on copper, a spontaneous isomerisation of the molecule is observed, and three different isomers of chloronitrobenzene (ClNB) are recognized and identified from the topographic STM images. These ClNB molecules provide an interesting research subject for computational STM, as the variations in the images can be explained in detail by different mechanisms in the formation of the tunneling current. With the Todorov Pendry approach (TP) [8, 9] a Green s function method with a strong explanatory power we find a good accordance to the experiments. The images are analyzed and interpreted by taking a closer look at molecular orbitals and their configurations with, in addition to TB method, a DFT method. A special feature of ClNB seems to be the wide gap between the orbitals delocalized to the 2

3 phenyl. On one hand this makes the phenyl itself rather transparent and, on the other hand, this allows no indirect mixing between the strongly localized resonant path through the nitro group and the chlorine channel. This is the key point in explaining why different isomers of ClNB can be recognized from topographic images. Only in the case, where the orbitals of the substitutional groups directly overlap, the tunneling channels are more involved. II. EXPERIMENTAL A. Sample preparation Ortho- (o-clnb), meta- (m-clnb) and para-chloronitrobenzene (p-clnb) molecules (C 6 H 4 ClNO 2 ) adsorbed on Cu(111) have been studied at 5 K with a ultrahigh-vacuum low-temperature scanning tunnelling microscope [10]. The Cu(111) sample was first cleaned by repetitive cycles of Ne+ sputtering (1.3 kv, 2.2 µa, 45 min) and annealing (870 K for 10 min followed by 890 K for 2 min) and then cooled on the manipulator to 19 K. Commercial meta-chloronitrobenzene (ClNB) (100 % purity in chromatogram) was further cleaned by several pump cycles. After each pump cycle mass spectra were recorded. The mass spectra showed the presence of not only ClNB but also of dissociation products induced by spectrometer hot filaments. The cleaning procedure was repeated until the mass spectra between two subsequent pump cycles did not change. The deposition was performed through a leak valve with the aid of a stainless steel tube for 30 s directly onto the sample held at 19 K. In order to avoid dissociation of the molecules during deposition, all hot filaments in the chamber were switched off. The sample was then transferred into the custom-built STM, which operates at 5 K. For the described preparation, a molecule density of less than 0.1 molecules per nm 2 was measured by STM. All STM images presented were recorded at a bias voltage of 100 mv and a tunneling current of 50 pa. B. Results The substitution of two hydrogen atoms within the basic benzene structure C 6 H 6 by a single nitro group (NO 2 ) and a chlorine atom (Cl) offers the possibility to create three different chloronitrobenzene isomers. The isomers that differ in the arrangement of the two attached substituents are ortho-clnb, meta-clnb, and para-clnb. Confinement to a 3

4 surface increases the number of possible isomers to 5, because for both ortho- and meta- ClNB two mirror images are possible. Indeed, after adsorption to the surface five different images of molecules are observed. Fig 1 shows submolecular resolution STM images of these five configurations. (Note that the configuration in Fig. 1(c) has been only observed once within several hundred molecules and might thus be stabilized by a surface defect.) Different parts in STM images can be identified based on previous adsorption experiments of nitrobenzene molecules (C 6 H 5 NO 2 ) on Cu(111) combined with tight-binding calculations [11]. We know that the phenyl ring is imaged as a dark depression while a nitro group appears as an ellipsoidal approximately 80 pm high protrusion. In the STM images of ClNB molecules we observe an additional circular protrusion of smaller apparent height (30pm) at different locations on the phenyl ring with respect to the nitro group. In the following we will show that this additional protrusion represents the chlorine atom, and thus, the differently appearing molecules represent the different isomers of chloronitrobenzene. Thus, the adsorption of the molecule even at 19 K leads to isomerisation. III. THEORY AND CALCULATIONS There are a couple of purposes for the theoretical calculations in this work. First, the simulations of the STM images show, whether the proposed assignments of the observed images to the different isomers are correct. Second, we analyze the formation of different tunneling channels, and discuss their origin and nature. Third, and this is the most practical issue, we explain what factors in the electronic structure make it possible to distinguish between different isomers of ClNB. A. Modeling the electronic and geometric structure We utilize Green s function in tight-binding (TB) basis to analyze the relation of the electronic structure to tunneling channels, and the subsequent formation of STM images. Following the method in Ref. 11, the Green s function for the substrate and the tip is calculated utilizing the recursion method described by Horsfield et al. [12], with first order terms for off-diagonal elements of Green s function. To make formation of tunneling channels more tractable, only s-orbitals of the substrate atoms are taken into account. 4

5 The Green s function of the adsorbate is formed exactly in TB basis starting from the Green s function of independent atomic orbitals G 0 αα(e) = 1 E E α + iη, where E α refers to the onsite energy of the orbital α, and η is the convergence parameter. Dyson s equation G = G 0 + GV G 0, is applied to the TB Hamiltonian in a way described in Ref. 13, 14. The basis consists of s-orbitals for hydrogen atoms, and (s, p x, p y, p z ) for carbon, chlorine, oxygen and nitrogen. The angular scaling and distance dependence of the TB hopping integrals follow Ref. 11. Concerning the geometry of the substrate we use exactly the same configuration as in Ref. 11, where a slab of 224 Cu atoms in 4 layers and a tetrahedral tip at the bottom of the slab is utilized. In STM calculations, Lees-Edwards boundary conditions [15] are applied to the simulation cell in order to simulate scanning the tip across the substrate surface. The constant current mode is simulated by varying the height of the simulation cell at each step of calculation [13]. B. Simulation of tunneling current The analysis of the experimental STM images is done in two stages. First, we simulate STM images using the TP-approach [8, 9]. Then, we qualitatively analyze the dependence between open tunneling channels and geometrical configuration. The Green s function of the adsorbate-substrate system appears in two roles in STM simulations. First, the Green s function gives the density matrix ρ 0 of non-interacting system for the substrate and the tip. Second, the transition matrix T between the tip and the substrate, can be written in terms of the Green s function of the interacting system and the Hamiltonian matrix elements between the tip and the surface orbitals. The basic formula for TP-method gives current as I = 2πe h EF +ev T r[ρ 0 ττ (E ev )T τ E σ (E) (1) F ρ 0 σ σ(e)t στ (E)]dE, 5

6 where ρ 0 is the density of states matrix for the uncoupled system. It is useful to write the transition matrix in the form T = V + V G + V, where V is the Hamiltonian matrix including only off-diagonal elements, coupling the different subsystems, and G + is the Green s function of the coupled system. The T-matrix can be rewritten in terms of different tunneling channels: T στ = V στ + V σµ G + µσ V σ τ + V σµ G + µνv ντ, where σ, τ refer to the s-orbitals of the substrate and the tip, and µ, ν refer to atomic orbitals of the adsorbate molecule. This form shows a clear division between different tunneling paths: tunneling through vacuum, T 1 = V στ, the scattering path T 2 = V σµ G + µσ V σ τ and the path through the adsorbate, T 3 = V σµ G + µνv ντ, with µ, ν running over all atomic orbitals participating in tunneling. This decomposition provides a fairly straightforward way to define tunneling channels and paths [11, 16]. Magoga and Joachim [7] make a distinction between terms tunneling path and tunneling channel. The former term describes the different paths in space through which electrons may tunnel, e.g., the through vacuum and through molecule -routes according to Ref. 6, are different tunneling paths in this terminology. In the case of complicated molecules, tunneling paths may open through different parts of the molecule or functional groups. Tunneling routes in terms of atomic or molecular orbitals are called tunneling channels. If the states are sufficiently localized into certain atoms or groups of atoms in a molecule, it may be possible to subdivide a tunneling path into different tunneling channels. The significance of the transition matrix depends essentially on, whether the tunneling through the structure is resonant or non-resonant [7]. For resonant tunneling, the density of states near tunneling energies is very high, meaning a large value of the imaginary part of G, which creates an easy tunneling route for electrons. Non-resonant tunneling takes place through the gap between occupied and unoccupied levels. Hence, the tunneling wave function is decaying, but even more importantly the tunneling route may exhibit a phase change in the wave function. In the formalism of the present study, this behaviour is seen in the matrix element of G corresponding to the nonresonant channel. This matrix element of Green s function has only the real part, leading to only real elements in the transition matrix. Subsequently, non-resonant channels may be 6

7 in the same phase or in antiphase with respect to each other. These kinds of interference effects have been presented widely in our previous works [16, 17]. Surprisingly, resonant states of the adsorbate in the vicinity of Fermi energy makes the channel analysis fairly complicated. The imaginary part of the Green s function is clearly large, and the tunneling current is dominated by the through adsorbate path T 3. channels T 1 and T 2 can be attributed to direct through vacuum tunneling by using the fact that ρ αβ = η π γ G + αγg γβ. [16] By applying the definitions of T 1 and T 2 and Dyson s equation, the integrand of Eq. (1) can be rewritten as The T r[ρ σ σ(e)v στ ρ 0 ττ (E ev )V τ σ ], (2) where the tunneling current is proportional to the LDOS of the substrate in the presence of an adsorbate. Thus, the effect of T 2 in the STM image can be interpreted as the change in substrate LDOS induced by the adsorbate. In resonant tunneling the orbitals of the adsorbate may have a strong overlap with the substrate orbitals near Fermi level. This may alter the substrate orbitals significantly at those energies where tunneling takes place. Thus, in terms of tunneling channels, T 2 is crucial. If the LDOS of the substrate decreases, the tunneling amplitude may locally diminish, despite a resonant level for tunneling electrons [13]. Consequently, the strength of the tunneling current cannot be reliably predicted without calculating the effect of substrate LDOS. This effect has also be seen very recently in DFT calculations combined to non-equilibrium Green s function techniques [18]. In the case of non-resonant tunneling, the electronic structure of the substrate at the tunneling energies may remain more intact, and thus serve as a good starting point for investigating the tunneling channels through free molecules. We have demonstrated in Refs. 14, 16 that the phase differences, and thus interference, between different tunneling channels determine whether the total tunneling current is high or low. Then, of course, the interpretation is refined by taking into account the coupling between the electronic structure of the substrate and the adsorbate. 7

8 C. Comparison of TB and DFT methods: Molecular orbitals Although the molecular orbitals mix with the substrate orbitals in an adsorbate-substrate system, it is useful to analyze the real space images of the orbitals of free molecules. This gives an important insight to contribution of those orbitals to tunneling channels, since the symmetry of the orbital determines, whether the overlap between a particular molecular orbital with the tip and/or the substrate is significant. This, on the other hand, is a fundamental factor in opening a tunneling channel through an orbital. First principles density-functional calculations are carried out using the generalized gradient approximation (GGA) with Perdew-Wang exhange-correlation functional [19]. We use DM ol3 code [20] with all-electron atomic orbital basis set. In order to assess the validity of the TB method, these calculations are carried out also in the TB basis. The coefficients between atomic and molecular orbitals are obtained from the secular equation. The density of states for molecular orbitals are projected to real space as a linear combination of atomic orbitals using the analytic form of Slater atomic orbitals[21]: ψ nlml (r, θ, φ) = Nr neff 1 e Z effr/a 0 n eff Y lml (θ, φ) (3) where N is a normalization constant, Y lml is a spherical harmonic, a 0 is Bohr radius, and nlm l are the quantum numbers of the state. The effective (n eff ) and real principal quantum number (n) are identical for the studied atomic species while values for effective atomic number (Z eff ) are obtained from Ref. 22. Using atomic wavefunctions instead of Slater orbitals does not affect the LDOS isosurfaces significantly. IV. THEORETICAL RESULTS AND ANALYSIS The simulations have been performed for several different chloronitrobenzene isomers on Cu(111). For the adsorption geometry we assume the geometry of nitrobenzene molecule from Ref. 11, where the phenyl ring is adsorbed horizontally on the surface with the carbon atoms at twofold bridge sites around the center copper atom in so called top0-geometry; nitrogen atom on top site next to the phenyl ring, and the oxygen atoms on bridge sites and relaxed towards surface, see Fig. 2, left side. Instead of the 1.8 Å used in Ref. 11, the molecule is brought slightly closer to the surface, with the distance of phenyl ring from 8

9 copper atoms approximately at 1.55 Å. This distance was chosen as the best fit with the experimental images, and it also corresponds to the phenyl - Pt(111) distance in Ref. 23. For each different isomer, a single hydrogen atom has been replaced by a chlorine atom, with the adsorption site proposed by the experimental results, the chlorine sitting on a bridge site with a bond length of about 2.5 Å. The calculations have been done for o-clnb, m-clnb and p-clnb molecules. The constant current mode is simulated starting with tip height at about 5.9 Å above the bare surface, and the bias voltage is held at 100 mv. A. STM images The computational results for each configuration are presented in Fig. 3. The phenyl rest is seen as a flat depression, with a corrugation of about 0.6Å. Next to the depression, the nitro group is seen in each case as an ellipsoidal protrusion of 0.4Å, and the chlorine tends to cause a slightly smaller protrusion, approximately 0.3 Å. As the chlorine atom is brought closer to the nitro group, first from para-configuration to meta-configuration and again to ortho-configuration, the symmetry of the nitro group image is greatly affected. Actually, in o-clnb, the image shows two rather round protrusions, Fig. 3(c), proposing a loss of a channel either through the chlorine or through the oxygen atom in the middle. The comparison to the experimental images of Fig. 1 is presented in Fig. 3(d), where the black ellipses and circles correspond to the shapes of the protrusions in the experimental images and red contours to the simulations. The simulated images are in good accordance with the experiments, even though the depression caused by the phenyl ring makes the comparison of linescans difficult. The shape of the protrusion caused by ortho-isomer differs slightly from the experiments. In our simulations, the image of the chlorine is clearly fused with the nearest oxygen, but in experiments, the chlorine forms a more separate feature. However, in the experiments, this feature is somewhat dependent on the tip condition, and thus the resolution of the experiment. Certain tip conditions give experimental images that are qualitatively rather similar to our simulations. Apart from the tip condition, we discuss other possible factors in the subsequent analysis section. The approximations, i.e. in the Green s function for the substrate, naturally affect the results of the simulations. Keeping this in mind, the corrugations cannot be interpreted as 9

10 absolute values and as such comparable to the experimental STM, but in a more qualitative sense. B. Features of the STM images and density of states As discussed above, the features of topographic STM images can be attributed to variations of LDOS of the sample or amplitudes and phases of tunneling channels generated by the adsorbate molecule. First, we consider what the dominant features of LDOS imply for the contrast of the STM image of different parts of the adsorbate molecule. Particularly, we analyze how the LDOS projected to the substitute groups vary between different isomers. As discussed in Ref. 11, the apperance of the phenyl ring as a dark protrusion seems to originate from the adsorbate induced change of the DOS of the copper atom beneath the adsorbate. This factor is especially important, since the different tunneling channels through the phenyl ring essentially cancel each other, and would thus make the molecule transparent. A clear difference between the DOS of a clean surface copper and the copper below a phenyl ring is observed, see Fig. 4. The density of states near the Fermi energy is reduced to about half of its value compared to the bare substrate, and this change results in the STM images to a flat depression centered to the copper atom in the middle of the phenyl rest. The elliptical protrusions in the STM images of p- and m-clnb, Fig. 3(a-b), result from a common channel of the whole nitro group, and are centered around the two oxygens, while the smaller protrusion is centered on the chlorine atom. Separate to this, the protrusions of the substitute groups of o-clnb, Fig. 3(c), are more blended into each other. Plotting the DOS for each atomic orbital shows, for each configuration, clear resonant channels through both oxygens and chlorine, whereas the carbon atoms and the nitrogen seem to have a weaker contribution near the tunneling energies. Since the phenyl ring disconnects the orbitals of Cl and NO 2, their contribution to the STM image can be separated. Only, if there is a direct overlap, their orbitals are mixed. Thus both of the groups are seen independent in the STM image in the decoupled cases, m- and p-clnb. However, in the case of ortho-configuration, the direct overlap has visible consequences. The symmetry of the two oxygens in the NO 2 group is broken and, eventually, a corresponding change is observed in the DOS of the chlorine atom. This can be seen clearly 10

11 in Fig. 5, where the combined DOS of s and p-orbitals has been calculated for both oxygen atoms and the chlorine atom in three different configurations of ClNB adsorbed on Cu(111). For meta- and para-configurations, the DOS of both oxygens are clearly symmetric, Fig. 5(a-b), and the chlorine has a resonant state separate from the nitro group, Fig. 5(c). As the chlorine is brought closer to the ortho-configuration, the states of chlorine and the nearest oxygen atom, Fig. 5(b-c), are mixed, and their resonant channels blend in to one common channel. C. Analysis of the tunneling current in terms of channels To be able to track down the relevant tunneling channels, we need to study the molecular orbital density of states in addition to the calculations in atomic orbital basis. Fig. 6 shows the real space LDOS isosurfaces and energy configuration of the molecular orbitals for each isomer, computed with DFT-based DM ol3-code. As is shown in our former study [11], reduction of the symmetry group D 6h of benzene to lower symmetry of the substituted molecules has some significance to the nature of tunneling channels. In the case of benzene, there is a wide gap between the doubly degenerate frontier orbitals, and it can be shown that these orbitals form non-resonant channels, whose mutual interference leads to transparency of the molecule in the STM image. A reconstructed nitro group reduces the symmetry to C s, which forms two new frontier orbitals with a negligible gap. These orbitals are localized to the nitro group, and a resonant channel opens, resulting in a bright ellipse in STM images. The reconstruction seems to be essential, since it makes nitro goup orbitals remain more detached from the delocalized π-orbitals of the phenyl ring. The appearance of the chlorine in STM images, however, does not depend on the change of the symmetry group of the molecule. In the case of the para-isomer, the symmetry group remains C s, but the meta-isomer further reduces the symmetry to C 1. In both the experimental and simulated images the three parts of these two molecules have the same nature: phenyl is seen as a dark depression, nitro group as a bright ellipse, and chlorine as bright sphere. Obviously, the key point is that the bonding between the phenyl ring and the chlorine is related to orbitals which are not anywhere near energies relevant to tunneling. Rather, orbitals localized to chlorine are formed, which open isolated non-resonant channels having a constructive interference with the through-vacuum channel. Consequently, the two 11

12 relevant orbitals of the nitro group remain intact, Fig. 6(a-b), and that is why the ellipse seen in case of NB remains in the topographic images in Fig. 3(a-b). Let us next consider the orbitals, which are relevant to the appearance of chlorine in the STM images. In the case of para- and meta-isomers, two interesting orbitals localized to chlorine are formed: the HOMO-1 and HOMO-2 states are now composed entirely of the chlorine p-orbitals. One of these orbitals has a strong overlap with the microscope tip as well as with the surface, since it has strongly localized p z type behaviour at Cl. The DOS of the atomic orbitals of the adsorbed chlorine in Fig. 5(c) shows, that a resonant state can be found and thus the protrusions in Fig. 3(a-b) result also from resonant tunneling. The ortho-isomer is different from other isomers, since the orbitals of the chlorine directly overlap with the nitro group. The strong overlap, as indicated also in the bottom row in Fig. 5, affects greatly the frontier orbitals. Two π-type orbitals are now observed near the Fermi level, Fig. 6(c), and the current is dominated wholly by these near resonant states. For comparison, the molecular orbitals have been calculated besides the DFT method, with the same TB based method as used for STM simulation. Fig. 7 presents LDOS isosurfaces of the two closest orbitals below and above the Fermi level of meta-clnb calculated with the TB method. These match very well for the shape and size of real space presentation of the orbitals, and reasonably well by their energies with the orbitals calculated with DFT presented in the middle column of Fig. 6, even though the energy gap is clearly wider. A slight difference between TB and DFT calculations appears when looking at the HOMO orbitals, which lie within a narrow energy range. According to DFT, two of the orbitals are quite strictly localized to the chlorine, but TB predicts some mixing with the phenyl orbitals. Due to this overlap, those levels shift somewhat lower in energy and, consequently, the order of the three orbitals differs between the two computational methods. Of course, these orbitals of free molecules do not fully explain the origin of the tunneling current in the case of adsorbed molecules, since the states may shift and broaden in the adsorption. The difference in the shape of the protrusion in the experimental and simulated STM images of ortho-isomer (Figs. 1(a,b) and 3(c)) may, in part, be due to the description of mixing between orbitals of chlorine, oxygen and the copper below. In such a situation, the details of the outcoming STM image may be sensitive to modeling of the substrate Green s function. On the other hand, another kind of model for the tip might affect the resolution, and detect the channels through the chlorine as separate features in STM image. 12

13 V. CONCLUSIONS For meta-chloronitrobenzene molecules adsorbed on Cu(111) surface, five specific STM images are observed. The difference in the STM images is explained by different molecular configurations, and the adsorbates are identified as para-, meta- and ortho-clnb, of which the two latter are observed in different chiralities. This suggests, that the adsorption of the molecule leads to a spontaneous isomerisation, even at 19 K temperature. Our simulated STM images are in a good qualitative accordance with the experimental images. To explain the detailed features of the images, we scrutinize relevant molecular orbitals of different isomers and the corresponding tunneling channels. In addition, a good congruence between the TB and DFT calculations of molecular orbitals adds solidity and reliability to the theoretical analysis. There are essentially three factors of the electronic structure to consider in explaining the characteristics of the STM images: the change in the LDOS of the substrate on the adsorption, the strength of the overlap between different subsystems, and resonant tunneling channels through localized adsorbate states. Our analysis of LDOS and tunneling channels indicate that the electronic structure of the phenyl ring is very essential to the submolecular resolution observed in STM study of the present molecules. First, the frontier orbitals of benzene open such tunneling channels that make the phenyl ring itself rather transparent. Second, the gap between these orbitals is rather wide, which disconnects the orbitals of the substitute groups from each other at the energies near Fermi level, unless there is a direct overlap between the orbitals. Hence, the submolecular resolution of STM is a consequence of certain independence of the electronic structure of different parts of the molecule. Considering the factors of submolecular resolution, a direct overlap between substitutional groups makes the electronic structure more complex, which is readily seen when comparing the simulated and experimental images of the o-clnb. Most interestingly, this configuration seems to be somewhat dependent on, e.g., tip shape in the experiments. This problematics would motivate another interesting study of sensitivity of the images and tunneling channels to, e.g., modeling the underlying substrate or the condition of the microscope tip. 13

14 VI. ACKNOWLEDGEMENTS We acknowledge financial support from the VolkswagenStiftung. The DFT calculations in this article were performed with the help of resources of CSC, the Finnish IT center for science. [1] G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel, Appl. Phys. Lett. 40, 178(1982); G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel, Phys. Rev. Lett. 49, 57(1982). [2] R. J. Hamers, R. M. Tromp, and J. E. Demuth, Phys. Rev. Lett. 56, 1972(1986). [3] B. C. Stipe, M. A. Rezaei and W. Ho, Science 280, 1732(1998). [4] J. Tersoff and D. R. Hamann, Phys. Rev. B. 31, 805(1985). [5] J. Nieminen and S. Paavilainen, Phys. Rev. B. 60, 2921(1999). [6] P. Sautet, Surf. Sci. 374, 406(1997). [7] M. Magoga and C. Joachim, Pys. Rev. B, 59, 16011(1999), [8] T.N. Todorov, G.A.D. Briggs and A.P. Sutton, J.Phys.: Condens. Matter 5, 2389(1993). [9] J.B. Pendry, A.B. Prêtre and B.C.H. Krutzen, J.Phys.: Condens. Matter 3, 4313(1991). [10] M. Mehlhorn, H. Gawronski, L. Nedelmann, K. Morgenstern, submitted. [11] J. Nieminen, E. Niemi, V. Simic-Milosevic and K. Morgenstern, Phys.Rev. B. 72, (2005). [12] A.P. Horsfield, A.M. Bratkovsky, D.G. Pettifor, and M. Aoki, Phys. Rev. B 53, 1656 (1996). [13] J. Nieminen, S. Lahti, S. Paavilainen, and K. Morgenstern, Phys. Rev. B. 66, (2002). [14] J. Nieminen, E. Niemi, and K.-H. Rieder, Surf. Sci. 552, L47(2004). [15] A.W. Lees and S.F. Edwards, J. Phys. C: Solid St. Phys. 5, 1921(1972). [16] E. Niemi and J. Nieminen, Surf. Sci. 600, 2548(2006). [17] E. Niemi and J. Nieminen, Chem. Phys. Lett. 397, 200(2004). [18] J.M. Blanco, C. Gonzáles, P. Jelínek, J. Ortega, F. Flores, R. Prez, M. Rose, M. Salmeron, J. Wintterlin, and G. Ertl, Phys. Rev. B (2005). [19] J.P. Perdew, Electronic structure of Solids 91, Akademie Verlag, Berlin [20] B. Delley, J. Chem. Phys. 113, 7756(2000). [21] P. W.Atkins and R. S.Friedman Molecular Quantum Mechanics, 3rd ed., Oxford Univ. Press, 14

15 New York (1997). [22] E. Clementi and D. L. Raimondi, J. Chem. Phys. 38, 2686 (1963). [23] C.F. Tirendi, G.A. Mills and C. Dybowski, J. Phys. Chem. 96, 5045 (1992). FIG. 1: STM images of five different configurations after adsorption of meta-chloronitrobenzene on Cu(111) identified as (a, b) ortho-chloronitrobenzene (o-clnb) (c,d) meta-chloronitrobenzene (m-clnb), (e) para-chloronitrobenzene (p-clnb); l and r denote that chlorine is adsorbed to the right-hand side and left-hand side of the nitro group, respectively, reflecting adsorption induced chirality. (f), line scans above isomers taken along dashed lines marked in (a), (c), and (e). The experiments were performed at 5 K at a bias voltage of 100 mv and a tunneling current of 50 pa. FIG. 2: Assumed adsorption geometry for chloronitrobenzene molecules on Cu(111). (a) para- ClNB with chlorine sitting on bridge site, (b) meta-clnb and (c) ortho-clnb. On the right hand side, side view showing that the reconstruction tilts the hydrogens and the substitutes slightly away from the surface. 15

16 FIG. 3: Simulated STM-images. (a) para-clnb, (b) meta-clnb and (c) ortho-clnb. The comparison between simulated and measured STM images is presented in (d), where the model of the m-clnb is taken from the theoretical simulations. The red ellipses and circles are obtained by drawing the lines over the nitro group and the chlorine in calculated STM images, while the black contours are the outer shapes of the molecules as they appear in measured STM images. FIG. 4: The DOS of the s-orbital of the Cu atom diminishes in the adsorption; on dash-dotted line, the DOS of a surface Cu atom far from the adsorbate, and on solid line, the DOS of the Cu atom under the phenyl ring. 16

17 FIG. 5: Combined DOS of s- and p-orbitals of (a,b) oxygen atoms and (c) chlorine atom of ClNB adsorbed on Cu(111) in three different configurations. The symmetry between the two oxygens is broken, when the chlorine is brought to the ortho-configuration, and a strong mixing of states can be observed between the chlorine and the other oxygen. 17

18 FIG. 6: Molecular orbitals for (a) para-clnb, (b) meta-clnb and (c) ortho-clnb computed with DM ol3-code. In para- and meta-configurations, the HOMO is localized in the nitro group, and the HOMO-1 and HOMO-2 consist solely of chlorine orbitals, whereas in the ortho-configuration they are blended together. 18

19 FIG. 7: Frontier orbitals of a free meta-chloronitrobenzene on the reconstructed adsorption geometry computed with TB method. 19