Content uploaded by Jean-Pierre Bucher
Author content
All content in this area was uploaded by Jean-Pierre Bucher on Nov 02, 2017
Content may be subject to copyright.
1
Hierarchy of Chemical Bonding in the Synthesis of
Fe-Phthalocyanine on Metal Surfaces: a Local
Spectroscopy Approach
Shawulienu Kezilebieke1, Anis Amokrane1, Mathieu Abel2, Jean-Pierre Bucher1*
1Université de Strasbourg, Institut de Physique et Chimie des Matériaux de Strasbourg, CNRS
UMR 7504, F-67034 Strasbourg, France and 2Université Aix-Marseille, IM2NP, CNRS UMR
7334, Campus de Saint-Jérôme, Case 142, F-13397 Marseille Cedex 20, France.
AUTHOR INFORMATION
Corresponding Author
* To whom correspondence should be addressed. E-mail: jean-pierre.bucher@ipcms.unistra.fr
2
ABSTRACT Scanning tunneling spectroscopy (STS) has become a key tool for accessing
properties of organometallic molecules adsorbed on surfaces. However, the rich variety of
signatures makes it sometimes a difficult task to find out which feature is intrinsic to the
molecule i.e. relevant for a metal-ligand interaction or related to the interaction of the molecule
with the substrate. Here we study the prototype covalent self-assembly of FePc and probe how
electronic/magnetic properties at the local scale change as a function of temperature-induced
step-by-step assembly, starting from TCNB (1,2,4,5- Tetracyanobenzene) molecular and Fe
atomic precursors. Intermediate complexes with tetra-coordinated Fe atoms are then used both,
as synthons for the FePc and as identifiers of specific features of the STS. As observed by STS
and confirmed by spin-polarized DFT calculations, the occupied dπ states of Fe are present in
both, the FePc and Fe(TCNB)2 on Au(111). The main difference appears in the dz2 states, which
play a key role in magnetism as confirmed by the presence/absence of the Kondo resonance. A
comprehensive picture is obtained by following with STS the hybridization of the dz2 orbital of
Fe to various substrates (Cu, Au and Co). Finally it is demonstrated that FePc units can be
created by on-surface polymerization from the Fe(TCNB)2 network upon thermal annealing.
TOC GRAPHICS
3
KEYWORDS. STM, Scanning tunneling spectroscopy, molecular states, covalent self-
assembly, molecular magnetism, nanochemistry, molecule-substrate interaction.
Metallophthalocyanines (MPc) are technologically relevant molecules for organic electronic
devices. They have been extensively studied for their chemical and optoelectronic properties and
can be employed as building blocks for a wide range of applications such as gas sensors, field
effect transistors, organic light emitting diodes, or data storage devices [1]. They have become a
dedicated system for local scanning tunneling microscopy and spectroscopy (STM/STS) studies
due to their advantageous geometry, leading to face-on adsorption and providing easy access of
metal atom and ligands to the STM probe tip [2-16]. Additionally, the physical and chemical
properties of the MPc molecules have also been studied by a number of other techniques such as
XPS [17-19], XRAS [19,20], NEXAFS [21] and XMCD [22] providing complementary
information on metal-ligand and molecule-surface interaction. The local magnetism of MPc on
metal surfaces was studied both, by STS from the behavior of the Kondo resonance [4-10] and in
favorable cases by spin polarized (SP)-STM [11,12]. However, the incidence of specific bonds
on the local magnetism remains a delicate issue [9,10], mainly because the substrate alters such
properties.
It is the purpose of this work to understand by means of STM/STS to which extent the property
of the metal atoms are modified by their chemical environment. To this end, the on-surface
synthesis of MPc by covalent self-assembly is addressed, using organic ligands and metal atoms
as precursors. Beyond the tremendous challenge of the direct on-surface synthesis for 2D
molecular electronics, the thermally induced, stepwise transformation of organic and inorganic
species into organometallic compounds then offers a way to study metal atoms involved in
4
either, coordination or covalent bonds and their interaction with a substrate. Prior to the MPc
synthesis, the obliged pathway usually leads to the formation of 2D metal-organic coordination
networks [23]. It was found by X-ray magnetic circular dichroism (XMCD) measurements [24-
27] that under certain circumstances, such networks of transition metals with appropriate ligands
e.g. TCNQ (7,7,8,8- Tetracyanoquinodimethane) and TCNB (1,2,4,5- Tetracyanobenzene) [26-
27] possess interesting magnetic properties on their own. On the other hand, the formation of
robust MPc molecules or 2D polymers as anticipated earlier [28] is expected only in specific
molecule-substrate systems upon high temperature annealing. While the feasibility of such a
polymeric form has been demonstrated recently for MnPc [29], a detailed step-by-step STS study
towards the chemical bond identification is still missing. In this work, Fe atoms and TCNB
molecules are used as metal and organic precursors. The 2D polymerization reaction is confined
on atomically flat crystal templates. To keep the interaction with the environment to a minimum,
the Au(111) surface is primarily used as a substrate but Cu(111) and Co islands are used
occasionally for comparison.
Before the Fe-TCNB assembly is addressed, a thorough characterization of FePc is performed
in the following. Although the synthesis leads to FePc(CN)8 complexes, we find it more relevant
to work on freshly deposited FePc since it allows a better, noise-free, STS characterization, as
will be shown latter on, and an easy comparison with the existing literature. The molecular
structure of Fe-phthalocyanine (FePc) consists of a central metal ion surrounded by a macrocycle
with alternating carbon and nitrogen atoms. Among the eight nitrogen atoms, four are in the aza-
position while the others belong to the pyrrole rings. The Fe atom in the FePc is involved in
covalent bonds with the nitrogen atoms from the pyrrole rings thus giving rise to the fourfold
5
symmetry observed by STM. The Fe is in the +2 oxidation state and shows a paramagnetic
behavior both, in the gas phase and when adsorbed on Au(111) [2,3]. It has a d6 electron
configuration and bears two unpaired electrons leading to S=1 [5,7].
The measurements were performed in a low–temperature STM operating at temperature of 4.6
K and under ultrahigh vacuum (UHV) condition. The FePc molecules were deposited onto the
Au(111) surfaces by means of a home-made evaporator. Prior to deposition, the crucible with the
FePc was heated overnight just below the evaporation temperatures for purification. The
Au(111) surface was kept at room temperature during the deposition. As shown in the STM
images of Figure 1, the FePc molecules appear with their characteristic planar adsorption (face
on), four-lobe structure and bright center on the Au(111) surface. Three representative
orientations of the molecules are identified in the enlarged images of Figure 1b-1g. Furthermore,
as reported earlier, the molecules adopt two types of adsorption configurations on the Au(111)
surface, i.e. the on-top and bridge configurations [6,7]. In the on-top (bridge) configuration, the
Fe atom is located in the on top (bridge) site with the molecular plane parallel to the surface. The
spectral features of both configurations have been analyzed in great details close to the Fermi
energy (± 5 meV), principally in relation with the Kondo behavior [6,7], however, no detailed
study has been published yet on the full spectroscopic analysis over several eV.
6
Figure 1. (a) STM image (63 × 63 nm2) of FePc molecules on Au(111), I= 0.2 nA, V= -0.34 V.
(b-g) Topography and schematics of the three different orientations of the FePc molecule (dotted
black arrow) making angles of -5°, 28° and 58° with respect to the [11-2] crystallographic
direction of gold (dotted red arrow). The frequencies of occurrence of the molecular orientation
show that configuration 2 (43%) slightly dominates over configuration 1 (30%) and
-1,0 -0,5 0,0 0,5 1,0
7
6
4
3
1
2
dI/dV(a.u)
Sample bias (eV)
Ligand
Fe
5
(a) (b) (c)
(d) (e)
(f) (g)
(h)
7
configuration 3 (27%). (h) dI/dV spectra above Fe (red) and above the ligand (black) under the
same condition.
The local electronic structure was investigated by low-temperature high resolution STS on
individual FePc molecules with different orientations and on different locations above the
molecules. Changes between different adsorption geometries are found to be marginal. Two
dI/dV spectra taken above the Fe atom (red) and on the benzene rings (black) of a FePc molecule
are presented in Figure 1h. There are seven main features at sample bias -0.9 V, -0.79 V, -0.72
V, -0.37 V, -0.15 V, +0.5 V, and 0 V, respectively. Since none of the above features were
observed on a pristine Au(111) surface, the peaks must necessarily be attribute to the molecular
structure of FePc adsorbed on Au(111). The features at positive bias voltages correspond to the
unoccupied states of the FePc molecules on the surfaces. In this energy domain, the spectra are
featureless except for a broad signature at +0.5 V (No. 7). The fact that the peak at +0.5 V only
appears above the ligand, and not above the Fe atoms, strongly speaks in favor of a LUMO
contribution of the molecule on Au(111) surface. For negative sample voltages, there is a peak at
-0.72 V (No. 3) on the ligand, that we attributed to the HOMO, i.e. p-orbital contribution, in
good agreement with previous valence-band photoemission spectra studies on FePc/Au(111)
[19]. Peaks recorded above the Fe atom are essentially determined by dz2 , dxz and dyz orbitals,
since in STM significant overlap between tip states and molecular orbitals are only expected for
those orbitals (contributions from dxy and dx2-y2 states are usually marginal). A sharp peak denoted
as No 1 and located at -0.9 V is then the lower lying occupied molecular orbital (HOMO-1)
while the shoulder at -0.79 V (No. 2) has been attributed to the dxz/yz orbitals of iron [6]. The
relatively intense resonance at -0.38 V, noted No. 4 resembles the one observed by Gao et al [6]
most probably arising from the hybridization of the Fe atom of the FePc molecule and the
8
Au(111) electronic surface state. Similar surface induced states, labeled SI in the following, have
been observed for FePc on other surfaces and are quite common for atoms on noble metal
surfaces [9, 10].
Figure 2. The constant height dI/dV maps over the FePc (2 × 2 nm2). The feedback loop was
opened over the Au(111) surface at 0.01 nA and at the bias voltage indicated in each map. The
asymmetry in contrast is due to the surface reconstruction of Au(111).
Additional information on the spatial distribution of the molecular orbitals is obtained from the
differential conductance recorded above the molecule. Figure 2 shows constant-height dI/dV
maps at different bias voltages. These maps were recorded by exploring the x-y space at a
constant tip-height above a single FePc molecule, which provides a better signal over noise ratio
compared to the constant current dI/dV maps, chiefly because in this mode, the feedback loop
remains open during the data acquisition [13]. As can be seen from Figure 2, dominant
9
conduction channels, corresponding to Fe d-orbital contributions are found over the Fe atom at
negative bias at -0.80 V, -0.20 V, and -5 mV, confirming the spectral features of Figure 1. The
peak close to the Fermi level (EF = 0 V), noted No. 6 in Figure 1, is a well known Kondo
signature observed previously in FePc on a Au (111) [6, 7]. Furthermore, dI/dV maps acquired at
-0.9 V and -0.7 V (Figure 2) show a dominant contribution from the ligands of the molecule as
expected for the HOMO-1 and HOMO states observed in the dI/dV spectra of Figure 1. At
positive bias, a contrast (Figure 2) is again observed on the metal and on the ligands at +0.5 V
corresponding to the LUMO orbital (peak No. 7 in Figure 1). Note the different geometry of the
HOMO and LUMO orbitals, eight lobes are observed for the low-lying orbitals (-0.8 V and -0.6
V) while only four lobes are seen on the LUMO (+0.5 V). Similar situations have been reported
before on other MPc systems [9] while O2-functinalized tips show different contrasts [14].
The peak located at -0.17 V, just below the Kondo peak (No. 5 in Figure 1), was not reported
in previous studies of FePc on Au(111). However, a similar feature has been observed for FePc
on Ag(100) [10] and Ag(111) [15] surfaces. According to its localized contrast associated with
peak No. 5 (see Figure 2), the origin of this feature is attributed to the out of plane dz2 state of the
Fe atom. To gain more insight into the hybridization of the Fe dz2 state with the substrate, we
carried out STS measurements above the FePc adsorbed on various surfaces. Figure 3 shows the
dI/dV spectra of FePc absorbed on the Au(111), Cu(111), and a cobalt nano-island. The
corresponding dI/dV spectra, acquired above the center of the FePc molecule, then exhibit an
increasing shift of the Fe d-state towards to the Fermi level EF. The positions of the Fe d-states
are summarized in the Table 1 for different surfaces.
10
Figure 3. The position of the dz2 orbital of FePc changes with different substrates. Feedback
loop opened at 0.4 nA and -0.3 V. The spectra are vertically shifted for clarity. Vertical doted
lines show the position of the dz2 states while the doted green line shows the position of the
Kondo resonance.
The shifting of the d-states of Fe on different surfaces, can be rationalized by introducing the
so-called d-band model [30, 31] that accounts for the interaction of adsorbate electrons with the
valence states of surface atoms. In the metal, the highly itinerant s and p-band electrons form a
broad band of states whereas the d-band electrons form a relatively narrow band owing to the
small coupling matrix element between the more localized d orbitals. The coupling mechanism
between the adsorbate and the surface can be viewed as a two-step process. First, the adsorbate
couples to the broad metal sp-band, leading to a shift and a broadening of the adsorbate states.
Second, the renormalized adsorbate states couple to the metal d-states. Since d-bands are narrow,
-0 ,3 -0 ,2 -0 ,1 0,0 0,1 0,2 0,3
dI/dV (a.u)
Sample bias (V)
Co
Cu
Au
Fe dz2 states Kon do
(a)
11
the interaction of an adsorbate state with the d electrons of a surface often gives rise to bonding
and antibonding states just as in a simple two-state problem. If the first step, coupling to the
metal sp-band, can be considered to be essentially similar for the transition and noble metals,
then, the interaction energy between an adsorbate and a metal surface are governed by the second
step, which is the coupling to the metal d-band. The key parameter in the d-band model is the
position of the d-band center with respect to the Fermi energy, Ed-Ef. The noble metals exhibit a
low-lying d-band center, because the d-band is filled with electrons. The d-state shift observed
experimentally for different metal surfaces obeys the same order Ag<Au<Cu<Co as the d-band
filling and d-band center.
Table 1. Position of the dz2 resonance of Fe for the FePc molecule on different metal surfaces,
values on Ag(100) and Ag(111) are from ref. [10] and [15].
Ag(111)
Ag(100)
Au(111)
Cu(111)
Co island
Peak position (mV)
-250
-250
-170
-50
-10
The data of Table 1 can be rationalized in the following way. In the case of Ag, where the d-
band center is lowest in energy the interaction with virtually any adsorbate gives rise to a large
fraction of antibonding states, which results in a weaker binding energy of the adsorbate to the
surface. On the contrary, for the Co surface which has a d-band center close to the Fermi level, a
large fraction of the antibonding states between the adsorbate and the surface will be pushed
above the Fermi level, resulting in a strong bonding character with the surface. This might then
explain the strong shift of d states of Fe in FePc on the cobalt nanoislands. Interestingly, when
looking at the relation between the center of d-band [32] and the shift of the d states of Fe in
12
FePc on different surfaces, we observe that the d-state shift shows the same trend as the d band
shift as we move from Ag to Co surfaces.
One of the key issues of this work is to show to which extent a stepwise identification of the
reaction path towards the final FePc product can be used to gather information on orbital
occupation and spin state of the species. In particular this approach may help us identify which
STS features present in the FePc are already present in the intermediate compounds. Since
TCNB radicals are intermediate species in the formation of phthalocyanines they represent a
natural choice for the surface confined self-assembly with Fe atoms [33]. The weak interaction
of the TCNB layer with the Au(111) surface, as quantified by STS [34], is an additional
advantage for the present study. When TCNB molecules and Fe atoms are vapor deposited on
Au(111), the temperature induced step-by-step self-assembly leads to tetra-coordinated
Fe(TCNB)4 and Fe(TCNB)2 precursor. The first complex was studied elsewhere [34] and does
not show relevant enough features for this work. Only the later complex has the appropriate
stoichiometry for the synthesis of FePc and will be studied here for the purpose of comparison. A
simple calculation shows that one Fe atom for two TCNB molecules correspond to the exact
number of atoms necessary to synthesize a 2D FePc polymer.
The deposition of TCNB molecules and iron atoms on Au(111) in a proportion of 2:1 and post-
annealing up to 450 K leads to a fully reticulated Fe(TCNB)2 network as shown in Figure 4a and
4b. The network is composed of Fe atoms interconnected with TCNB molecules. Large
homochiral and mirror symmetric Fe(TCNB)2 domains with a lateral extension up to 50 nm are
present on the surface (see Figure 4a). It is worth mentioning that the same result is obtained on
Ag(100), indicating that the coordination network is relatively insensitive to the surface template
13
i.e. to the crystallography and to some extent chemical nature of the underlying substrate.
Furthermore, our results also allow the correct interpretation of networks observed earlier under
similar conditions [28], that must likely be attributed to Fe(TCNB)2. The control of the
stoichiometry however is crucial, since an excess of metal atoms or molecules can hamper the
formation of well-ordered networks and lead to disordered phases. A careful analysis of the STM
data shows that the Fe(TCNB)2 network has a square structure with a measured periodicity of
1.15 ±0.1 nm in both orthogonal directions as shown in Figure 4b. The unit cell of the
Fe(TCNB)2 network (Figure 4b and 4c) contains eight nitrogen atoms among which four have a
coordination bond with one single Fe atom. Additional information concerning the chirality of
the Fe(TCNB)2 domain and its relationship to the underlying Au(111) substrate can be found in
the Supporting Information.
Figure 4. STM topography images. (a, b) Fe(TCNB)2 network on Au(111) acquired at I = 0.2 nA
and V = -0.9 V. (e, f) Units of FePc(CN)8 on Au(111), obtained upon annealing the previous
14
Fe(TCNB)2 structure to 550 K, acquired at I = 1.8 nA and V = -0.9 V, (c, g) Corresponding
schematics. Notice the simple assembling of TCNB molecules with Fe atoms in (c), while the
covalent bonding involves a full rearrangement of atoms in (g).
Further annealing to 550 K for 3 minutes induces the initial stage of polymerization. As shown
in Figures 4e,f, units of octacyano-FePc [FePc(CN)8] involving covalent bonds have formed
from the previous Fe(TCNB)2 phase. In Figure 4f, the Fe atoms appear as bright protrusions,
confirming the presence of dz2 orbitals of Fe while the organic ligands appear as four symmetric
lobes, forming together a cross-like molecule. In the middle of Figure 4f, a dimmer protrusion
may tentatively be ascribed to a fourfold coordinated Fe atom. Since no clear spectroscopic
identification could be done for the moment this atom is not represented in Figure 4g. For more
details see the SI. Full transformation into covalent bonds is further ascertained from the absence
of chirality on the ligands. It is worth noticing that during the temperature increment, a
significant amount of TCNB molecules originally coordinated to the Fe atoms desorb from the
surface (compare Figure 4a and 4e) and are not available anymore to pursue the polymerization
process. A spectrum of FePc(CN)8 is shown in Figure S3 in comparison with FePc. Although it
was difficult to obtain good spectra after the polymerization process, the Kondo resonance and
the dz2 feature just below Ef (both relevant for magnetism) are clearly visible, showing again their
close interrelation and ultimately the similar status of the Fe atoms in both systems.
STS measurements of the Fe(TCNB)2 phase on Au(111) were performed and compared with
those of FePc on Au(111). Although each Fe atom in the Fe(TCNB)2 phase has 4 nitrogen
neighbors as in the FePc molecule, the spectra show significant differences but also remarkable
similarities. The differences mainly come from the fact that Fe in Fe(TCNB)2 is forming
15
coordination bonds with the TCNB moieties while the Fe atom in the FePc is also involved in
covalent bonds with the nitrogen atoms of the pyrrole rings. Figure 5 shows dI/dV spectra for
both Fe(TCNB)2 and FePc with the STM tip above the Fe atom (Figure 5a) and above the ligand
(Figure 5b). The two systems show three similar features, two of them on the Fe atom and one on
the ligand. In particular, the feature at -0.79 V on the Fe, attributed to the dπ (dxz and dyz) orbitals,
is similar in both systems, indicating that the Fe feels a similar environment from the four TCNB
ligands and from the phthalocyanine in the FePc molecules. It is worth noting that the feature at -
0.79 V also resembles the one observed on iron clusters on the Au(111) surface surrounded by
the 3,4,9,10-perylene-tetracarboxylic-dianhydride (PTCDA) molecules [35]. The second
remarkable feature, is the surface induced state (No. 4 in Figure 5a) found above the Fe atom at -
0.38 V on both, the FePc and the Fe(TCNB)2 leading to the conclusion that in both cases, the Fe
atom produces the same effect on the Au(111) electronic s, p surface state.
Figure 5. Comparison of dI/dV spectra taken on the Fe(TCNB)2 network and on the FePc
molecule on Au(111) surface, under the same condition (feedback loop opened at V = -0.7 V, I =
-1,0 -0,5 0,0 0,5 1,0
dI/dV(a.u)
Sample bias(V)
Fe in Fe(TCNB)2
(a)
2
4
5
1
Fe in FePc
6
Kondo
dz2
SI
dxz/yz
5.0Å
-1,0 -0,5 0,0 0,5 1,0
dI/dV(a.u)
Sample bias(V)
Ligand in
Fe(TCNB)2
(b)
Ligand
in FePc
LUMO
SI
HOMO
5.0Å
16
0.2 nA). (a) dI/dV spectra taken above the Fe in FePc and in Fe(TCNB)2 respectively (b) dI/dV
taken above the ligands of Fe(TCNB)2 and above the lobes of FePc, respectively.
Finally, the ligand spectra of Figure 5b show that the LUMO orbital of the network is located
close to the LUMO orbital of FePc (+0.50 V). On the other hand, the HOMO orbital is located at
-0.76 V, again close to the one of FePc (-0.72 V). This leads to a HOMO-LUMO gap of about
1.2 V in both systems. Features that are similar in both, Fe(TCNB)2 and FePc are summarized in
the Table 2.
Table 2. Summary of similar peak positions for FePc and Fe(TCNB)2 on Au(111) given in eV.
No.
FePc
Fe(TCNB)2
dxz/yz
2
-0.79
-0.79
HOMO
3
-0.72
-0.76
SI
4
-0.38
-0.38
dz2
5
-0.17
-
Kondo
6
yes
no
LUMO
7
+0.5
+0.5
The appearance of a Kondo resonance for FePc on Au(111) (peak No. 6 in Figure 5a) with a
corresponding Kondo temperature of about 200 K deserves a special attention, particularly in
respect to its absence in Fe(TCNB)2. In FePc, the valence of Fe can be described as Fe2+ (S=1)
with the following electronic configuration: (dxy)2, (dz2)1, (dπ)3, similar to the solid and gas phase
FePc. The fact that the Kondo resonance of FePc/Au(111) is located at the metal ion (see Figure
2) strongly speaks in favor of a screened Kondo spin originating from the dz2 orbital (local
17
moment Kondo system). This result is far from being trivial since on similar MPc systems
(adsorbed on noble metals), the maximum intensity Kondo resonance has been found at the
ligands arising from unpaired spins in the dπ orbital [10]. This is also in contrast with the fully
delocalized Kondo resonance observed over the Co-porphyrin molecules on Cu(111) [36]. Our
result is in agreement with Minamitani et al. [7] who found that for the on-top position of Fe
above the Au(111), the strong coupling of the dz2 orbital overcomes the zero field splitting,
providing a temperature window where the Kondo screening becomes dominant (effective
Hamiltonian with S=1/2). The partial screening of the S=1 spin of FePc is also coherent with the
interpretation of Stepanow et al. [22].
The features observed by STS can be discussed in more detail in terms of bonds that determine
the shape and positions of the experimental resonance most sensitive to the STS measurement.
The overlap of occupied valence Fe states and empty s orbitals of N in TCNB drives the
formation of a metal-ligand bonds, which may involve a charge transfer from the Fe to the N
atoms. Only the out of plane π-orbitals due to the interaction of dxz/yz orbitals with the pz orbitals
of the N atoms are observed in STS (see Figure 5) while the main σ interaction between the Fe
and N atoms involving the dx2−y2 orbital of Fe and the s, p orbitals of the N, are usually not seen.
To rationalize the experimental results, density functional theory (DFT) calculations were
performed for FePc molecules and compared with Fe(TCNB)2 complexes. In order to come
closer to the situation of a molecule interacting with the noble metal surface, calculation was also
performed for the [FePc]- ion similar to Ref. [10]. Experimental results are then compared to the
calculated projected density of state (PDOS) onto different orbital symmetries, allowing the
assignment of dI/dV peaks to specific orbitals (see Figure 6). The spin-polarized (SP)-PDOS and
the calculated spin value of S=1 for the free FePc molecule of Figure 6a, are in good agreement
18
with previous studies [7, 10, 18]. The d-orbital PDOS shows a dxz/dyz contributions at -0.25 eV
and -1.6 eV, while the dz2 peak is located at +0.3 eV. The effect of adsorption is depicted in
Figure 6b, for the case of [FePc]-, the dz2 resonance shifts downwards and is now spanning the
Fermi energy in good agreement with the experimental peak No. 5. Simultaneously the dxz/yz
resonances shift upwards, thus a dxz/yz resonance is found at -1.2 eV after adsorption, only slightly
below the peak measured by STS (No. 2 in Figure 5). It must be noticed that hybridization with
the underlying substrate occurs primarily with orbitals that have strong z-component, while dxy
orbitals undergo negligible shifts in the vicinity of the Fermi level upon adsorption.
19
Figure 6. Spin-resolved projected density of state (PDOS) of (a) FePc, (b) (FePc)- and (c)
Fe(TCNB)2 onto different d-orbitals. The insets show spin-density isosurfaces for an electron
density of ±0.004 e/ Å3.
E-E
f
(eV)
!" !# $ # "
%&'()*+,
"
-% & . / 0!
1
1
.234'567,
189"
18:9;<9
18:<
18:"!<"
%&./
1
1
1
1
=!=>1'&? ,
1
1
1
20
As detailed in the Supporting Information, the PDOS on both, FePc and FePc(CN)8 are the
same due to the fact that the orbital configuration of the Fe atom is not modified by the
peripheral functionalization (CN instead of H), and thus the magnetic moment and spin state are
the same in both systems. Figure 6c shows the SP-PDOS of the unsupported Fe(TCNB)2
complex. This calculation is sufficient for comparison, owing to the comparatively weak
interaction of the tetracoordinated Fe atom with the Au surface [34]. It shows that the dxz/yz peak
at -0.85 eV reproduces quite well the experimental resonance No. 2 at -0.79 eV. On the other
hand, the dz2 peak located above the Fermi level is reminiscent of the broad peak centered on
+0.5 eV in the STS spectrum of Figure 5a and corresponds to an empty dz2 state in contrast with
what is observed in FePc, that shows an occupied dz2 resonance just below the Fermi level. The
difference may explain by itself that the Kondo channel is opened solely to screen the Fe spin on
the FePc molecule (Peak No. 6 in the STS data of Figure 5). The lack of Kondo signature for
Fe(TCNB)2 cannot be ascribed to a different Fe-substrate distance and is most probably due to
the intrinsic differences mentioned above. Corresponding spin densities are shown in the insets
of Figure 6. As expected, FePc and [FePc]- clearly show a localized spin density on the Fe with
S=1, the opposite spin density being only marginal on N. The Fe(TCNB)2 on the contrary shows
a weekly delocalized spin density on the ligand, the Fe bearing only part (S=0.68) of the total
spin.
In conclusion, the controlled synthesis of FePc on Au(111) from the first principles (Fe atoms
and TCNB molecules) provides exceptional insight into the evolution of the chemical/electronic
structure along the reaction path. Information gathered from high-resolution dI/dV spectroscopic
labeling shows that the dπ features appear already on the intermediate Fe(TCNB)2 complex as
well as a remarkable surface induced state. However, the appearance of the dz2 peak close to the
21
Fermi energy, as well as the related Kondo effect is definitely a signature of the final FePc,
clearly evidencing the contribution of the covalent character. The lack of Kondo resonance in
Fe(TCNB)2 is related to the absence of the dz2 feature just below Ef and appears to be intrinsic to
the complex. While the on surface synthesis of FePc by covalent self-assembly of TCNB and Fe
at 550 K is successfully demonstrated, the approach presented here is promising as well for the
interpretation of dI/dV spectra of other systems, similarly assembled on surfaces. Furthermore,
the ability to control the covalent bond formation on metal surfaces allows extending these
concepts to polymerization of single poly-MPc sheets as shown recently on a smaller scale [29]
thus providing an interesting subject, not only from the electronic point of view (2D delocalized
π-conjugated system) but also from the magnetic point of view [37], since the regular array of
spin centers can serve as model systems for fundamental studies of spin-spin interactions.
EXPERIMENTAL METHODS
The measurements were performed in a low–temperature STM (CreaTec LT-STM) operating
at a base pressure of 10-11 mbar at 4.6 K. The Au(111) surface was cleaned by repeated cycles of
Ne ion sputtering and annealing at 700 K. First TCNB molecules (97% Alpha Aesar) and then Fe
atoms were vapor deposited to the sample using a resistively heated alumina crucible and
electron beam evaporators, respectively. The sample was held at room temperature (RT) during
the deposition steps and then annealed for 15 min at about 450 K to form Fe(TCNB)2 networks
on Au(111) surface. Commercially available FePc (Aldrich, 98%) was thermally evaporated at
540 K onto Au(111) surface held at room temperature (RT). The STM measurements were
performed at 4.6 K with an electrochemically etched tungsten tip. All tips were treated in vacuo
by sputter/annealing cycles and then by soft indentation into the clean gold surface. The well
22
known steplike onset of the Au(111) Shockley surface state then appears as a sharp and
reproducible feature [34]. All given voltages are applied to the sample. Spectra of the differential
conductance (dV/dI) were acquired with a lock-in technique using a modulation amplitude of 5
mVrms and frequency of 600 Hz.!!
Electronic structure calculations were performed within the framework of density functional
theory (DFT) using SIESTA package. The wave function of the valence electrons is expanded on
a double-zeta plus polarization basis set for each atom [38, 39]. The core electrons are treated
within the frozen core approximation in which norm-conserving Troullier-Martins
pseudopotentials were used [40]. The exchange-correlation energy is treated within the
generalized gradient approximation (GGA) using parametrization proposed by Perdew-Burke-
Ernzerhof [41]. In this study, spin polarized calculations are done on molecules in a gas phase
configurations with a fixed total spin of S=1 that corresponds to the most stable configuration. A
k grid (4 × 4 × 1) was used in the case of Fe(TCNB)2 and (1 × 1 × 1) for the calculations of
isolated FePc molecules.
ASSOCIATED CONTENT
Supporting Information. Chirality of the Fe(TCNB)2 phase and its relationship to the substrate,
enantiomers domains (Figure S1); arrays of FePc(CN)8 on Au(111) (Figure S2); dI/dV of the
FePc(CN)8 complex (Figure S3); comparison between spin-resolved density of states of FePc
and FePc(CN)8 projected onto different d-orbitals of Fe (Figure S4). This material is available
free of charge via the Internet at http://pubs.acs.org.”
AUTHOR INFORMATION
23
Corresponding Author
*Email: jean-pierre.bucher@ipcms.unistra.fr
Notes
The authors declare no competing financial interests.
ACKNOWLEDGMENT
This work was supported by “UNION” (ANR grant 10-EQP-52), the International Center for
Frontier Research in Chemistry (Grant FRC-2010-JBu-0001), and the Institut Universitaire de
France (IUF).
REFERENCES
(1) Bekaroglu, O. et al. Functional Phthalocyanine Molecular Materials (Springer, Berlin,
2010).
(2) Lu, X.; Hipps, K. W.; Wang, X. D.; Mazur, U. Scanning Tunneling Microscopy of
Metal Phthalocyanines:! d7 and d9 Cases. J. Am. Chem. Soc. 1996, 118, 7197–7202.
(3) Lu, X.; Hipps, K. W. Scanning Tunneling Microscopy of Metal Phthalocyanines:! d6 and
d8 Cases. J. Phys. Chem. B 1997, 101, 5391–5396.
(4) Zhao, A.; Li, Q.; Chen, L.; Xiang, H.; Wang, W.; Pan, S.; Wang, B.; Xiao, X.; Yang, J.;
Hou, J.G.; et al. Controlling the Kondo Effect of an Adsorbed Magnetic Ion through its
Chemical Bonding. Science 2005, 309, 1542-1544.
(5) Tsukahara, N.; Noto, K.-I.; Ohara, M.; Shiraki, S.; Takagi, N.; Shin, S.; Kawai, M.
Evolution of Kondo Resonance from a Single Impurity Molecule to the Two
Dimensional Lattice. Phys. Rev. Lett. 2009, 102, 167203.
(6) Gao, L.; Ji, W.; Hu, Y. B.; Cheng, Z. H.; Deng, Z. T.; Liu, Q.; Jiang, N.; Lin, X.; Guo,
W.; Du, S.; et al. Site-specific Kondo Effect at Ambient Temperatures in Iron-Based
Molecules. Phys. Rev. Lett. 2007, 99, 106402.
24
(7) Minamitani, E.; Tsukahara, N.; Matsunaka, D.; Kim, Y.; Takagi, N.; Kawai, M.
Symmetry-Driven novel Kondo Effect in a Molecule. Phys. Rev. Lett. 2012, 109,
086602.
(8) Katoh, K.; Yoshida, Y.; Yamashita, M.; Miyasaka, H.; Breedlove, B.K. Kajiwara, T.;
Takaishi, S.; Ishikawa, N.; Isshiki, H.; Feng, Y.; et al. Direct Observation of
Lanthanide(III)-Phthalocyanine Molecules on Au(111) by Using Scanning Tunneling
Microscopy and Scanning Tunneling Spectroscopy and Thin-Film Field-Effect
Transistor Properties of Tb(III)- and Dy(III)-Phthalocyanine Molecules. J. Am. Chem.
Soc. 2009, 131, 9967–9976.
(9) Li, Z.; Li, B.; Yang, J.; Hou, J.G. Single Molecule Chemistry of Metal Phthalocyanine
on Noble Metal Surfaces. Acc. Chem. Res. 2010, 43, 954-962.
(10) Mugarza, A.; Robles, R.; Krull, C.; Korytár, R.; Lorente, N.; Gambardella, P. Electronic
and Magnetic Properties of Molecule-Metal Interfaces: Transition-Metal
Phthalocyanines Adsorbed on Ag(100). Phys. Rev. B 2012, 85, 155437.
(11) Iacovita, C.; Rastei, M.V.; Heinrich, B.W.; Brumme, T.; Kortus, J.; Limot, L.; Bucher,
J.P. Visualizing the Spin of Individual Cobalt-Phthalocyanine Molecules. Phys. Rev.
Lett. 2008, 101, 116602.
(12) Brede J.; Atodiresei, N.; Kuck, S.; Lazić, P.; Caciuc, V.; Morikawa, Y.; Hoffmann, G.;
Blügel, S.; Wiesendanger, R. Spin- and Energy-Dependent Tunneling through a Single
Molecule with Intramolecular Spatial Resolution. Phys. Rev. Lett. 2010, 105, 047204.
(13) Heinrich, B.W. Iacovita, C.; Brumme, T.; Choi, D.J.; Limot, L.; Rastei, M.V.; Hofer,
W.A.; Kortus, J.; Bucher, J.P. Direct Observation of the Tunneling Channels of a
Chemisorbed Molecule. J. Phys. Chem. Lett., 2010, 1, 1517–1523.
(14) Cheng, Z.; Du, S.; Guo, W.; Gao, L.; Deng, Z.; Jiang, N.; Guo, H.; Tang, H.; Gao, H.J.
Direct Imaging of Molecular Orbitals of Metal Phthalocyanines on Metal Surfaces with
an O2-Functionalized Tip of a Scanning Tunneling Microscope. Nano Res. 2011, 4,
523–530.
25
(15) Gopakumar, T.G.; Brumme, T.; Kröger, J.; Toher, C.; Cunibertti, G.; Berndt, R.
Coverage-Driven Electronic Decoupling of Fe-Phthalocyanine from a Ag(111)
Substrate. J. Phys. Chem. C, 2011, 115, 12173-12179.
(16) Sperl, A.; Köger, J.; Berndt, R. Demetalation of an Organometallic Complex. J. Am.
Chem. Soc. 2011, 133, 1107.
(17) Ahlund, J.; Nilson, K.; Schiessling, J;. Kjeldgaard, L.; Berner, S.; Mårtensson, N.;
Puglia, C.; Brena, B.; Nyberg, M.; Luo, Y. The Electronic Structure of Iron
Phthalocyanine Probed by Photoelectron and X-Ray Absorption Spectroscopies and
Density Functional Theory Calculations. J. Chem. Phys. 2006, 125, 34709.
(18) Isvoranu, C.; Wang, B.; Schulte, K.; Ataman, E.; Knudsen, J.; Andersen, J.N.; Bocquet,
M.L.; Schnad, J. Tuning the Spin State of Iron Phthalocyanine by Ligand Adsorption. J.
Phys.: Condens. Matter 2010, 22, 472002.
(19) Ahmadi, S.; Shariati, M. N.; Göthelid, M. Molecular Layers of ZnPc and FePc on
Au(111) Surface: Charge Transfer and Chemical Interaction. J. Chem. Phys. 2012, 137,
084705.
(20) Annese, E.; Casolari, F.; Fujii, J.; Rossi, G. Interface Magnetic Coupling of Fe-
Phthalocyanine Layers on a Ferromagnetic Surface. Phys. Rev. B 2013, 87, 054420.
(21) Gargiani, P.; Angelucci, M.; Mariani, C.; Betti, M.G. Localized and Dispersive
Electronic States at Ordered FePc and CoPc Chains on Au(110). Phys. Rev. B 2010, 81,
085412.
(22) Stepanow, S.; Miedema, P.S.; Mugarza, A.; Ceballos, G.; Moras, P.; Cezar, J.C.;
Carbone, C.; de Groot, F.M.F.; Gambardella, P. Mixed-Valence Behavior and Strong
Correlation Effects of Metal Phthalocyanines Adsorbed on Metals. Phys. Rev. B, 2011,
83, 220401 (R).
(23) Barth, J.V. Molecular Architectonic on Metal Surfaces. Annu. Rev. Phys. Chem. 2007,
58, 375-407
26
(24) Gambardella, P.; Stepanow, S.; Dmitriev, A.; Honolka, J.; de Groot, F.M.F.;
Lingenfelder, M.; Gupta, S.S.; Sarma, D.D.; Bencok, P.; Stanescu, S.; et al.
Supramolecular Control of the Magnetic Anisotropy in Two-Dimensional High-Spin Fe
Arrays at a Metal Interface. Nature Mater. 2009, 8, 189-193.
(25) Umbach, T.R.; Bernien, M.; Hermanns, C.F.; Krüger, A.; Fernández-Torrente, I.; Stoll,
P.; Pascual, J.I.; Franke, K.J.; Kuch, W. &Ferromagnetic Coupling of Mononuclear Fe
Centers in a Self-Assembled Metal Organic Network on Au(111). &Phys. Rev. Lett. 109,
267207 (2012).
(26) Abdurakhmanova, N. Tseng, T.-C.; Langner, A. ; Kley, C.S.; Sessi, V.; Stepanow, S.;
Kern, K. Superexchange-Mediated Ferromagnetic Coupling in Two-Dimensional Ni-
TCNQ Networks on Metal Surfaces. Phys. Rev. Lett., 2013, 110, 027202.
(27) Giovanelli, L. ; Savoyant, A. ; Abel, M. ; Maccherozzi, F. ; Ksari, Y. ; Koudia, M. ;
Hayn, R. ; Choueikani, F.; Otero, E.; Ohresser, P.; et al. Magnetic Coupling and Single-
Ion Anisotropy in Surface-Supported Mn-Based Metal–Organic Networks. J. Phys.
Chem. C, 2014, 118, 11738−11744.
(28) Abel, M.; Clair, S.; Ourdjini, O.; Mossoyan, M.; Porte, L. Single Layer of Polymeric
Fe-Phthalocyanine: An Organometallic Sheet on Metal and Thin Insulating Film. J. Am.
Chem. Soc., 2011, 133, 1203-1205.
(29) Koudia, M.; Abel, M. Step-by-Step On-Surface Synthesis : from Manganese
Phthalocyanines to their Polymeric Form. Chem. Comm. 2014, 50, 8565-8567.
(30) Norskov, J. K. Covalent Effect in Effective-Medium Theory of Chemical Binding.
Phys. Rev. B 1982, 26, 2875-2885.
(31) Hammer, B.; Morikawa, Y.; Norskov, J. K. CO Chemisorption at Metal Surfaces and
Overlayers. Phys. Rev. Lett. 1996, 76, 2141.
(32) Hammer, B.; Norskov, J. K. Theoretical Surface Science and Catalysis-Calculations and
Concepts. Advances in Catalysis, 2000, 45, 71-129.
27
(33) Yang, C.H.; Lin, S.F.; Chen, H.L.; Chang, C.T. Electrosynthesis of the Metal
Phthalocynine Complexes. Inorganic Chemistry, 1980, 19, 3541-3543.
(34) Kezilebieke, S.; Amokrane, A.; Boero, M.; Clair, S.; Abel, M.; Bucher, J.P. Steric and
Electronic Selectivity in the Synthesis of Fe-TCNB Complexes on Au(111): from
Topological Confinement to Bond Formation. Nano Res. 2014, 7, 888-897.
(35) Mendez, X.; Caillard, R.; Otero, G.; Nicoara, N.; Martin-Gago, J.A. Nanostructured
Organic Material: From Molecular Chains to Organic Nanodots. Adv. Mater. 2006, 18,
2048–2052.
(36) Perrera, U.G.E.; Kulik, H.J.; Iancu, V.; Dias da Silva, L.G.G.V.; Ulloa, S.E.; Marzari,
N.; Hla, S.W. Spatially Extended Kondo State in Magnetic Molecules Induced by
Interfacial Charge Transfer. Phys. Rev. Lett. 2010, 105, 106601.
(37) Zhou, J.; Sun, Q. Magnetism of Phthalocyanine-based Organometallic Single Porous
Sheet. J. Am. Chem. Soc., 2011, 133, 15113–15119.
(38) Sanchez-Portal, D.; Ordejon, P.; Artacho, E.; Soler, J.M. Density-Functional Method
for Very Large Systems with LCAO Basis Sets. Int. J. Quantum Chem., 1997, 65, 453-
461.
(39) Soler, J.M.; Artacho, E.; Gale, J.D.; Garcia, A. ; Junquera, J.; Ordejon, P.; Sanchez-
Portal, D. The SIESTA Method for ab initio Order-N Materials Simulation. J. Phys.:
Condens. Matter. 2002, 14, 2745-2779.
(40) Troullier, N.; Martins, J.L. Efficient Pseudopotentials for Plane-Wave Calculations.
Phys. Rev. B, 1991, 43, 1993-2006.
(41) Perdew, J.P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made
Simple. Phys. Rev. Lett. 1996, 77, 3865-3868.
