DOI: 10.1021/la8043347 7353Langmuir 2009, 25(13), 7353–7358 Published on Web 05/26/2009
pubs.acs.org/Langmuir
© 2009 American Chemical Society
c(4
2) Structures of Alkanethiol Monolayers on Au (111) Compatible with
the Constraint of Dense Packing
Oleksandr Voznyy* and Jan J . Dubowski
Department of Electrical and Computer Engineering, Centre of Excellence for Information Engineering
(CEGI), Universit
e de Sherbrooke, Sherbrooke, Qu
ebec J1K 2R1, Canada
Received December 31, 2008. Revised Manuscript Received April 26, 2009
Using alkanethiol dense packing as a starting point, we have found six prototypical packing structures commensurate
with the (3
2
√
3) supercell of the Au (111) surface. Five of the six structures are not compatible with the flat surface
conditions but can be fitted to a reconstructed surface. Combined with density functional theory calculations and
simulations of grazing incidenceX-ray diffractionmaps and of scanning tunnelingmicroscopy images, this allowed us to
refine and assess the recently proposedmodels of the c(4
2) self-assembledmonolayers involving thiolate-adatom and
thiolate-adatom-thiolate species and to propose a new model with four gold adatoms per unit cell.
1. Introduction
Self-assembled monolayers (SAMs) of organic molecules on
solid substrates attract a lot of interest for their potential
applications in nanofabrication, molecular electronics, bio- and
chemical sensing, surface protection, and so forth.1-3 Resolving
the atomistic structure of the SAM is an important step toward
understanding the process of its formation and creating mono-
layers with desired functionality. Alkanethiol SAMs on Au (111)
became a prototypical system for studying self-assembly because
of their stability and ease of preparation, and have been intensely
studied for the last two decades.4 Nevertheless, the exact structure
of these SAMs in the “standing-up” phase remains highly
debatable. Experimental determination of the structure is proble-
matic because of the impossibility of direct imaging of the inter-
face between the monolayer and the surface. Early theoretical
simulations, assuming an atomically flat surface, proposed sev-
eralmodels for the c(4
2) superstructuremarginallymore stable
than the (
√
3

√
3)R30
phase (hereafter,
√
3)5-10 following the
experimental suggestions.4,11 However, most of them were
derived for short chain thiols and were not compared to experi-
mental structural data to convincingly select the preferred model.
Only recently have several experiments explicitly shown that
thiol SAMs onAu (111) contain gold adatoms.12-14 The presence
of surface reconstruction introduces additional degrees of free-
dom for modeling and further complicates determination of the
exact structure of the monolayer. Several models involving
thiolate-adatom,15 thiolate-adatom-thiolate,16-19 and thio-
late-vacancy16-18 species were proposed recently, based on the
analysis of different data sets, each having its own drawbacks.
Particularly, density functional theory (DFT) is known to have
deficiencies in describing van derWaals interactions, and most of
the theoretical studies had concentrated on short-chain thiols to
avoid this problem.9,10,17,20,21 However, the c(4
2) structure of
interest was never observed experimentally for methyl- and
ethylthiolate SAMs, making the short-chain thiols an unreliable
model system for resolving the c(4
2) structure. On the other
hand, models obtained by fitting to experimental grazing inci-
dence X-ray diffraction (GIXRD) data16,22 were performed
independently of theoretical calculations, sacrificing the thermo-
dynamic stability of the structures for the quality of the fit.
In this work we derive several c(4
2) models using thiol
densest packing as a primary constraint, which concurrently has
the capability to suggest a surface reconstruction rather than
adapt to a predefined one. Comparison to a wide set of experi-
mental data allowed us to exclude some of the previously
proposed models and reduce the discussion to two structures.
2. Methodology
We adapt the approach previously proposed by us for
thiol SAMs on GaAs (001).23 Infrared reflection spectroscopy
(IRS) data24,25 and tilt angles obtained from near-edge X-ray
*Corresponding author. E-mail: o.voznyy@usherbrooke.ca.
(1) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M.
Chem. Rev. 2005, 105, 1103.
(2) Schreiber, F. J. Phys.: Condens. Matter 2004, 16, R881.
(3) Seker, F.; Meeker, K.; Kuech, T. F.; Ellis, A. B. Chem. Rev. 2000, 100, 2505.
(4) Schreiber, F. Prog. Surf. Sci. 2000, 65, 151.
(5) Fischer, D.; Curioni, A.; Andreoni, W. Langmuir 2003, 19, 3567.
(6) Gerdy, J. J.; Goddard, W. A. J. Am. Chem. Soc. 1996, 118, 3233.
(7) Li, T. W.; Chao, I.; Tao, Y. T. J. Phys. Chem. B 1998, 102, 2935.
(8) Pertsin, A. J.; Grunze, M. Langmuir 1994, 10, 3668.
(9) Vargas, M. C.; Giannozzi, P.; Selloni, A.; Scoles, G. J. Phys. Chem. B 2001,
105, 9509.
(10) Yourdshahyan, Y.; Rappe, A. M. J. Chem. Phys. 2002, 117, 825.
(11) Vericat, C.; Vela,M. E.; Salvarezza, R. C.Phys. Chem. Chem. Phys. 2005, 7,
3258.
(12) Kautz, N. A.; Kandel, S. A. J. Am. Chem. Soc. 2008, 130, 6908.
(13) Maksymovych, P.; Sorescu, D. C.; Yates, J. T. Phys. Rev. Lett. 2006, 97.
(14) Maksymovych, P.; Yates, J. T. J. Am. Chem. Soc. 2008, 130, 7518.
(15) Yu,M.; Bovet, N.; Satterley, C. J.; Bengio, S.; Lovelock, K. R. J.; Milligan,
P. K.; Jones, R. G.; Woodruff, D. P.; Dhanak, V. Phys. Rev. Lett. 2006, 97.
(16) Cossaro, A.; Mazzarello, R.; Rousseau, R.; Casalis, L.; Verdini, A.;
Kohlmeyer, A.; Floreano, L.; Scandolo, S.; Morgante, A.; Klein, M. L.; Scoles,
G. Science 2008, 321, 943.
(17) Gronbeck, H.; Hakkinen, H.; Whetten, R. L. J. Phys. Chem. C 2008, 112,
15940.
(18) Wang, J. G.; Selloni, A. J. Phys. Chem. C 2007, 111, 12149.
(19) Mazzarello, R.; Cossaro, A.; Verdini, A.; Rousseau, R.; Casalis, L.;
Danisman, M. F.; Floreano, L.; Scandolo, S.; Morgante, A.; Scoles, G. Phys.
Rev. Lett. 2007, 98.
(20) Gronbeck, H.; Hakkinen, H. J. Phys. Chem. B 2007, 111, 3325.
(21) Molina, L. M.; Hammer, B. Chem. Phys. Lett. 2002, 360, 264.
(22) Torrelles, X.; Barrena, E.; Munuera, C.; Rius, J.; Ferrer, S.; Ocal, C.
Langmuir 2004, 20, 9396.
(23) Voznyy, O.; Dubowski, J. J. Langmuir 2008, 24, 13299.
(24) Laibinis, P. E.; Whitesides, G. M.; Allara, D. L.; Tao, Y. T.; Parikh, A. N.;
Nuzzo, R. G. J. Am. Chem. Soc. 1991, 113, 7152.
(25) Nuzzo, R. G.; Korenic, E. M.; Dubois, L. H. J. Chem. Phys. 1990, 93, 767.

7354 DOI: 10.1021/la8043347 Langmuir 2009, 25(13), 7353–7358
Article Voznyy and Dubowski
absorption fine structure (NEXAFS)26,27 and GIXRD measure-
ments28 indicate that thiols in the SAMs on Au (111) possess the
densest possible packing, comparable to that of crystalline
alkanes.29,30 In our approach, thiol dense packing structures are
obtained from molecular mechanics (MM) simulations in the
absence of surface and then checked for commensurability with
the known size of the surface unit cell. No attention to adsorption
sites of the molecules is given at this step. Further fitting
of crystalline structures of the surface and the monolayer is
performed by introducing surface reconstructions (adatoms or
vacancies), taking into account the known orbital shapes, bond
lengths, and steric limits of sulfur and gold atoms, and, conse-
quently, optimizing the geometry with DFT. The proposed
method is simpler to implement and requires less computational
effort than quantummechanics/molecular mechanics (QM/MM)
calculations,5,16 since DFT and MM are never used simulta-
neously. Independently of the level of elaboration, current
theoretical approaches inevitably conserve the amount of atoms
in the unit cell, limiting in such a way possible outcomes of the
simulation. Our approach offers a powerful complementary
tool for a systematic search of initial guesses, helping to find
reconstructions that optimize both thiol packing and thiol-surface
interactions. Full details of the implementation of the method
can be found in the Supporting Information and in our
previous work.23
Togenerate possible dense packing structures of thiols, we used
the DREIDING force field31 as implemented in the Accelrys
Discovery Studio package. The accuracy ofMMcalculations was
tested on the most stable packing structure of polyethy-
lene (orthorhombic unit cell). Obtained distances and angles
between the alkane chains23 are within 1% error from the most
recent experiments.32 The DFT geometry optimization and total
energy calculations were performed for butanethiol and penta-
nethiol SAMs (hereafter, C4 and C5) using the SIESTA code33
within the generalized gradient approximation (GGA) with
Perdew-Burke-Ernzerhof (PBE) functional.34 Optimized bases
forAu,35C,Hand S36were used. Surface slabsweremodeledwith
fouror five fullAu layers and a vacuumregionof∼30 A˚.A (5
4)
Monkhorst-Pack k-grid for the (3
2
√
3) unit cell was used.
Forces on atoms were converged to 30meV/A˚. In order to reduce
the basis set superposition error, “ghost” Au atoms (i.e., basis
orbitals without actual atom) were placed in ideal bulk positions
around adatoms for calculation of adatom formation energy and
total energies of the “atop-adatom” models. Obtained lattice
constant, bulk modulus, surface energy, and adatom formation
energywerewithin the ranges of values reportedpreviously.10,18,21
Constant-current scanning tunneling microscopy (STM) images
were simulated within the Tersoff-Hamann approach37 for a
range of biases from -3 to 3 eV. In-plane diffraction intensity
maps were simulated using the ANA-ROD program,38 taking
into account the mirror and the 3-fold rotational symmetries of
the substrate, following the procedure and notation conventions
described in ref 22.
3. Results
3.1. Prototypical Packing Structures. Figure 1 shows four
of the six prototypical packing structures obtained by applying
the proposed approach. The two structures not shown are the
simple monoclinic, widely used in previous works to describe the
√
3 phase,5-8 and a “3 + 1” structure similar to that shown in
Figure 1d but with precession angle of 60
from the next nearest
neighbor (NNN) thiol direction, and thus incompatible with
GIXRD data.28 For all of the structures the chain centers form
ideal hexagons; however, sulfur positions deviate from this ideal
pattern as a result of the a zigzag shape of thiol chains.
It should benoted, that all of the structures have somedegree of
misfit with the surface (red arrows with white outline in Figure 1).
The tilt angle of ∼35
needed to achieve full interlocking
of chains39 (also common to phases of Langmuir layers on
liquids29,30) is larger than the 31.5
allowed by the area available
Figure 1. Prototypical alkanethiol packing structures commensu-
rate with the (3
2
√
3) supercell of the Au (111) surface: panels a,
b, c, andd showtheviewalong thiol chains (i.e.,∼35
off-normal to
the surface, thiol tilt direction is indicated by bold dashed arrows)
with thiols unit cell indicated. Red arrows indicate the compressive
strain required to be imposed on relaxed packing structures to
achieve full commensurability with the surface (the length of the
arrows is proportional to the amount of strain, being 7%in the case
of ortho2a structure). The difference between ortho2a and ortho2b
packing structures and the incommensurability of the latterwithan
atomically flat surface is shown in panel e. Same numbering of
chains as in panels b and c is used. Light blue hydrogen atoms are
guides to the eye, darker and lighter shades of gray are used to
distinguish the chains with different twists, and sulfur atoms are
yellow.
(26) Hahner, G.; Woll, C.; Buck, M.; Grunze, M. Langmuir 1993, 9, 1955.
(27) McGuiness, C. L.; Shaporenko, A.;Mars, C. K.; Uppili, S.; Zharnikov,M.;
Allara, D. L. J. Am. Chem. Soc. 2006, 128, 5231.
(28) Fenter, P.; Eisenberger, P.; Liang, K. S. Phys. Rev. Lett. 1993, 70, 2447.
(29) Kaganer, V. M.; Mohwald, H.; Dutta, P. Rev. Mod. Phys. 1999, 71, 779.
(30) Kitaigorodskii, A. I. Organic Chemistry Crystallography; Consultants
Bureau: New York, 1961.
(31) Mayo, S. L.; Olafson, B. D.; Goddard,W. A. J. Phys. Chem. 1990, 94, 8897.
(32) Takahashi, Y.; Kumano, T. J. Polym. Sci. B 2004, 42, 3836.
(33) Soler, J. M.; Artacho, E.; Gale, J. D.; Garcia, A.; Junquera, J.; Ordejon, P.;
Sanchez-Portal, D. J. Phys.: Condens. Matter 2002, 14, 2745.
(34) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865.
(35) Junquera, J.; Paz, O.; Sanchez-Portal, D.; Artacho, E. Phys. Rev. B 2001,
64.
(36) Voznyy, O.; Dubowski, J. J. J. Phys. Chem. B 2006, 110, 23619.
(37) Tersoff, J.; Hamann, D. R. Phys. Rev. Lett. 1983, 50, 1998.
(38) Vlieg, E. J. Appl. Crystallogr. 2000, 33, 401. (39) Ulman, A.; Eilers, J. E.; Tillman, N. Langmuir 1989, 5, 1147.

DOI: 10.1021/la8043347 7355Langmuir 2009, 25(13), 7353–7358
Voznyy and Dubowski Article
per thiol on Au(111) surface.4,30 Such a misfit of the SAM and
surface unit cells along the tilt direction can be easily avoided by
reducing the tilt angle at the expense of reduced chain interlock-
ing.39 Themisfit in the perpendicular direction, however, can only
be avoided by changing the twist of the chains, in agreement with
the requirement of rotational distortions implied by IRS data.25
Depending on the value of themisfit, this may result in significant
disordering of the SAM.
The structure in Figure 1a (hereafter “ortho”), based on the
orthorhombic packing observed experimentally for bulk
alkanes,29,30,32 cannot be fitted on an atomically flat surface.
Dense packing restricts the positions of sulfur atoms so that not
all of them can occupy the energetically favorable bridge face-
centered cubic (fcc) site5,9,10,18,21 (seeFigure S2c for example). The
DFT geometry optimization of such structure results in sulfur
atoms falling down into bridge-fcc sites, destroying the dense
packing. However, surface reconstructions with 1 or 2 adatoms
per thiol are found to be compatible with dense packing. These
structures are similar to inverted honeycomb (IHC) and honey-
comb (HC) structures, proposed previously for monoclinic pack-
ing,21 although, here the adatoms occupy both fcc and hexagonal
close-packed (hcp) sites (see Figure 2a and Figure S3). The total
energy differences for the structures discussed here are compiled
in Table 1.
Positions of adatoms identical to those in “ortho-atop” struc-
ture (Figure 2a) were proposed for the “polymer” model20 with
sulfur atoms bridging the fcc (red) and hcp (blue) adatoms. This
resulted in energetic gain compared to the flat-surface
√
3
structure due to reduced steric repulsion from the surface while
still maintaining two bonds to the surface for sulfur. In the
original “polymer” model, the two methylthiolates in the unit
cell were tilted along the
√
3 direction (perpendicular to the one in
our ortho-atop model) while in opposite directions to each
other.20 In our calculations we imposed the same tilt direction
for all butanethiol chains, which resulted in a noticeable steric
hindrance between thiols and made this structure isoenergetic to
Figure 2. (a)Topand (b) sideviewsof theDFT-relaxed structures.Red spheres:Auadatoms in fccholloworbridgeadsorptionsites: blue:Au
adatoms in hcp hollow sites. Hydrogen atoms in panel a are omitted for clarity. For ortho2a-2, the bottommostH atoms are shown to reveal
the absence of steric repulsion between themand adatoms.The (3
2
√
3) cell is shownwith solid line. (c) Simulated diffraction intensitymaps
for C10 thiols. Dotted lines indicate the
√
3 peaks, dashed line indicates the reciprocal (3
2
√
3) unit cell. (d) Simulated STM images for C4
thiols produced for 0.8 eV bias (empty states) and 10-9 a.u.-3 LDOS isosurface (located at 2.6 A˚ above themolecular apex). The shape of the
bright protrusions in STM reproduces the positions of the two topmost hydrogen atoms (small circles) for chains with an even amount of
carbon atoms, and reduces to a smaller circular shape for chains with odd amount of carbons, i.e., with only one topmost H.

7356 DOI: 10.1021/la8043347 Langmuir 2009, 25(13), 7353–7358
Article Voznyy and Dubowski
the ortho-atop model (see Table 1). The optimized tilt angle for
this model appeared to be only 10
(see the three-dimensional
(3D) structure in the Supporting Information), while the tilt
direction of the original polymer model is clearly incompatible
with experimental data and our prototypical packing structures.
Packing structures shown in Figure 1b,c (hereafter “ortho2a”
and “ortho2b”, respectively) also exhibit orthorhombic symme-
try. They differ from the ortho packing (Figure 1a) by the pattern
of equivalent molecules within the unit cell, while the difference
between the ortho2a and ortho2b themselves concerns the clock-
wise and counterclockwise twists with respect to the original
monoclinic phase (see the StructuresGeneration Procedure in the
Supporting Information). Different twist direction results in a
better interlocking of the chains in ortho2b structure (see
Figure 1e), and a smaller misfit with Au (111) surface along the
NNN Au direction (Figure 1b,c). Similarly to ortho packing,
molecules in ortho2a structure on an atomically flat surface
cannot simultaneously accommodate bridge-fcc adsorption sites
favored at low coverage. Nevertheless, ortho2a structure can
easily adopt the atop-adatom geometry, requiring adatoms in a
mixture of fcc and hcp sites (Figure 2a). In contrast to IHC
structure, these adatoms are paired, insignificantly enhancing the
stability of the structure (see Table 1).
The ortho2b (Figure 1c) and 3 + 1 (Figure 1d) structures are
clearly not compatible with an atomically flat surface, since two
and one chains in them, respectively, are shifted vertically, so that
sulfur atoms are situated at different heights above the surface
(Figure 1e). Such height modulation of S atoms was suggested by
normal incidence X-ray standing wave (NIXSW) experiments.40
The resulting end group height modulation might be also respon-
sible for the observed STM contrasts.41-44
3.2. Atop-Adatom Model. On the basis of the obtained
prototypical packing patterns, the exact structure of the recently
proposed qualitative c(4
2) model involving thiolate-adatom
species (four adatoms per unit cell)15 can be resolved. Among our
six prototypical packing structures, three can be adsorbed atop an
adatom (namely, mono, ortho, and ortho2a) but only ortho
and ortho2a require adatoms in a mixture of fcc and hcp
sites (Figure 2a) as NIXSW data suggest.15 Since the atomic
scattering factor of Au is much higher than those of C and S, the
resulting GIXRD maps would be dominated by adatoms (in our
simulations forC10 thiols, addition of adatoms increases the peak
intensities ∼100 times). This, in turn, implies that it is enough to
impose the centered rectangular symmetry45 only on adatoms,
while the chain twists can be arbitrary. Indeed, the absence of
modulation in superlattice rods (and its presence in
√
3 rods)
observed experimentally suggests that the superlattice peaks are
not primarily the result of an ordering (twist) of the hydrocarbon
chains, but rather that of S or Au atoms.45 Among the three
possible atop-adatom models, only ortho2a-atop structure has
the required symmetry of adatoms (Figure 2a), resulting in a
qualitatively correct diffraction map (i.e., has the missing (0;1),
(2;0), (2;2) spots), as shown in Figure 2c (note that “mono-atop”
not presented in the figure exhibits only
√
3 peaks). Nevertheless,
all the three structureshavepractically identical energies (Table 1),
providing no apparent reason for ortho2a-atop to be preferred.
Since in the atop-adatom structures the adsorption geometries
and chain twist angles of all the thiols are practically identical, no
end group height modulation is observed in the simulated STM
images (Figure 2d) in contrast to experimental observations.41-44
Such a structure looks essentially like the
√
3 phase in STM,
despite the underlying c(4
2) symmetry of the adsorption sites.
Moreover, tilt of thiol chains causes a tilt of S-Au bonds as well
(Figure 2b). Combined with a downward relaxation of adatoms,
this reduces the distance from S to the nearest bulk Au plane to
1.9 A˚ (Figure S4), in contrast to 2.5 A˚ observed by NIXSW
experiments.15,46
3.3. Alternative Model with Four Adatoms. Earlier work
proposed a significantly different interpretation of NIXSW
data,40 requiring a ∼0.8 A˚ height modulation of sulfur atoms,
with the lower S atom situated atop Au, and the higher S close to
the fcc site. The originally proposed “sulfur pairing model”
compatible with this interpretation40 was inconsistent with the
more recent experimental data and thus was abandoned.4,22 In
contrast, our ortho2b and 3 + 1 structures have an intrinsic
heightmodulation of∼1 A˚ of sulfur atoms. Figure 2a,b shows the
structure based on ortho2b packing with four adatoms per unit
cell (hereafter ortho2b-4) obtained by fitting it with the surface.
It is similar to the previously reported c(4
2) structure on an
atomically flat surface, derived using QM/MM calculations.5
Addition of adatoms allowed for the improvement of packing
density, while the formation of two bonds to Au for each
S significantly enhanced the binding energy compared to the
atop-adatom geometry. This resulted in ortho2b-4 structure
being 0.82 eV per unit cell more stable (Table 1). In contrast to
Table 1. Relative Stability of the Structures with Respect to Adsorption on an Atomically Flat Surfacea
total energy difference, eV
structure Au atoms difference this work
Molina, Hammerb
ref 21.
Gronbeck, Hakkinen
ref 20.
Gronbeck, Hakkinen,
Whetten ref 17.
flat surface, mono (
√
3) 0 0 0 0 0
thiolate-vacancy, mono (HC) -4 -0.55 -0.68 -0.53
thiolate-vacancy, ortho -4 -0.35
-Au-S- polymer +4 1.30 -0.56c -0.19c
atop adatom, mono (IHC) +4 1.29 1.24 0.94 1.28
atop adatom, ortho +4 1.30
atop adatom, ortho2a +4 1.22
ortho2b-4 +4 0.4
thiolate-adatom-thiolate, ortho2a-2 +2 -0.54 -0.46c -0.84
aEnergies per unit cell (four thiolates) are presented. Lower total energy corresponds to a more stable structure. Formation of adatoms (vacancies) is
taken into account assuming exchange with the infinite bulk reservoir. Values from previous works are for methylthiolates, in current work they are for
butylthiolates. bResults obtained using PW91 functional. cExact geometries differ from the ones used in this work. See text for details.
(40) Fenter, P.; Schreiber, F.; Berman, L.; Scoles, G.; Eisenberger, P.; Bedzyk,
M. J. Surf. Sci. 1998, 413, 213.
(41) Lussem, B.; Muller-Meskamp, L.; Karthauser, S.; Waser, R. Langmuir
2005, 21, 5256.
(42) Riposan, A.; Liu, G. Y. J. Phys. Chem. B 2006, 110, 23926.
(43) Zeng, C. G.; Li, B.;Wang, B.;Wang, H. Q.;Wang, K. D.; Yang, J. L.; Hou,
J. G.; Zhu, Q. S. J. Chem. Phys. 2002, 117, 851.
(44) Zhang, J. D.; Chi, Q. J.; Ulstrup, J. Langmuir 2006, 22, 6203.
(45) Fenter, P.; Eberhardt, A.; Eisenberger, P. Science 1994, 266, 1216.
(46) Roper, M. G.; Skegg, M. P.; Fisher, C. J.; Lee, J. J.; Dhanak, V. R.;
Woodruff, D. P.; Jones, R. G. Chem. Phys. Lett. 2004, 389, 87.

DOI: 10.1021/la8043347 7357Langmuir 2009, 25(13), 7353–7358
Voznyy and Dubowski Article
the QM/MM-derived flat-surface structure,5 sulfur lateral and
vertical positions in our model are compatible with NIXSW
data.40 Figure 2c,d shows the simulated GIXRDmap (calculated
for C10 thiols and averaged over three rotational domains to be
directly comparable to experimental data45) and STM image for
C4 thiols. Since adatoms possess the required centered rectangu-
lar symmetry, the diffraction intensity map is in qualitative
agreement with experimental data (correct missing spots and
relative intensities of the peaks).45 The
√
3 peaks are stronger in
the experimental map, as expected for
√
3 and c(4
2) phases
coexisting on the surface.22 Modulation of the headgroup and
terminal group heights ∼1 A˚, combined with a zigzag pattern
common to ortho2 structures, produces an STM image similar to
those observed experimentally.4,42,43
3.4. Thiolate-Adatom-Thiolate Model. Adsorption of
thiolate with S atom situated at the same level above the surface
as aAuadatom, as observed for half of the chains in the ortho2b-4
geometry (see left chain in ortho2b-4 in Figure 2b), maximizes the
binding energy of thiolate. Two bonds to the surface are formed,
while adsorption atop Au allows avoiding steric repulsion, pre-
sent in the bridge-fcc adsorption geometry on a flat surface.21
Adsorption of all thiols in such atop geometry allows for the
reduction in the amount of adatoms, resulting in the formation of
thiolate-adatom-thiolate complexes, also observed experimen-
tally in the low-density “striped” phase,13 and compatible with
NIXSW15,40,46 and photoelectron diffraction (PED)47 data. Pre-
vious works inspired by this geometry could not, however, find an
appropriate structure of the SAM. Addition of the adatom
resulted in the reduction of tilt from the correct 31
to ∼20

and in the loss of dense packing due to disturbance of steric limits
of the nearby CH2 units.
18 Introduction of the ortho2a packing
allows us to solve this problem. The resulting structure based on
ortho2a packing and two adatoms per unit cell (hereafter
ortho2a-2) is shown in Figure 2a,b. The opposite orientation of
CH2 units allows steric repulsion from adatoms to be avoided
without affecting the dense packing. This structure is identical to
the recently proposed c(4
2) model for methylthiolate,17
however, deduced from completely different considerations.
The simulated diffraction intensity map of the proposed
structure for C10 thiols is similar to that of the ortho2b-4 model
(Figure 2c) and compares well with experimental data.45 Despite
equivalent adsorption sites, the simulated STM image in
Figure 2d of the DFT-optimized structure exhibits a slight height
modulation of thiol end groups. This modulation is due to the
presence of a 7% misfit between the SAM and the surface,
resulting in a slight deviation from ideal packing structure.
The obtained zigzag pattern is similar to that typically observed
experimentally; however, the ∼0.1 A˚ height modulation in
this model is significantly smaller than 0.4-0.7 A˚ observed
experimentally.41,42
4. Discussion
All our models by construction possess a tilt angle compatible
with IRS24,25 and GIXRD28 data. They also exhibit in one form
or another the atop adsorption sites suggested by PED47 and
NIXSW.15,46 We could not, however, find a suitable surface
reconstruction involving thiolate-vacancy coexisting with thio-
late-adatom complexes16,18 compatible with any of the obtained
prototypical packing structures. Discriminating between the
proposed models based on available NIXSW or GIXRD data
remains a complicated task because of controversies in the
interpretation15,40 or unavailability of the full experimental data
sets to analyze the numerical values of the fit quality factors. Thus,
we try to analyze the models from other points of view.
Our theoretical findings regarding the stability and structural
properties of the atop-adatom model present arguments against
its feasibility for explanation of the c(4
2) structure. Compar-
ison of the stability of the models with different amount of
adatoms, however, requires more attention. In previous works
and in Table 1, stability of the new structures versus the flat-
surface
√
3 phase was assessed assuming that adatoms come from
an infinite bulk reservoir.17,18,21 However, since there is no
external source of gold atoms (as it could be in molecular beam
epitaxy system), adatoms can only be pulled out from the surface
itself,16 lifting the herringbone reconstruction13,48 or leaving a
vacancy behind.48,49 This leads to a 2-fold increase of the forma-
tion energy, from 0.57 eV for adatom coming from infinite bulk,
to 1.1 eV for adatom-vacancy pair. Considering the increased
adatom formation energies, all our structures, including the
thiolate-adatom-thiolate model, become unstable compared
to the flat-surface
√
3 model, in contrast to results from previous
works.17,20,21 Our calculated adatom formation energy reduces
to∼0.8 eV at surface defects or step edges, a value comparable to
that previously reported.50 This value brings the thiolate-
adatom-thiolate model∼0.1 eV lower in energy than flat-surface
structure, while HC model remains clearly unfavorable (0.37 eV)
since it requires the formation of 2 times the amount of defects.
The amount of adatoms present in the thiolate-adatom-
thiolate structure corresponds to 0.17 monolayers of gold, com-
patible with recent experiments on SAM desorption using hydro-
gen flux.12 The assumption that the creation of adatoms is
accompanied by the creation of vacancies may also explain the
presence of vacancy islands (etch pits) observed after SAM
formation.4,44,49 Taking into account that some initial amount
of adatoms forms by lifting the herringbone reconstruction of the
surface13,48 and that, at step edges, formation of adatom does not
leave a vacancy behind,44 the amount of adatoms in the thiolate-
adatom-thiolate model compares well to the 0.04 monolayers of
etch pits reported previously and 0.07-0.1 monolayers estimated
by us on the basis of the analysis of STM images from other
works.4,44,49
Using similar reasoning, adatoms that are already present on
the surface cannot escape into bulk12 and can disappear only by
attaching to a step edge. Thus, the
√
3 phase should contains the
same amount of adatoms as c(4
2), since these phases were
observed to interchange easily.11 The atop-adatom IHC model21
is a good candidate for a
√
3 counterpart of the c(4
2) phase
presentedby the ortho2b-4model.However, the energydifference
in this pair is too high (0.21 eVper thiol) to explain the coexistence
of both phases on the surface.11,51 On the other hand, it is
impossible to form a
√
3 phase from thiolate-adatom-thiolate
complexes. Recently it was proposed that disorder in their
positions can explain the
√
3 GIXRD pattern,19 although, no
exact structure was proposed to test its appearance in STM.
In our simulations of STM images the thiolate-adatom-
thiolate model looks very similar to the
√
3 phase. The ex-
change-correlation functional and the geometry relaxation pro-
cedure used in ourDFTcalculations forC4 thiols are not sensitive
enough to the strain in the ortho2a-2 SAM to stipulate any
(47) Kondoh, H.; Iwasaki, M.; Shimada, T.; Amemiya, K.; Yokoyama, T.;
Ohta, T.; Shimomura, M.; Kono, S. Phys. Rev. Lett. 2003, 90.
(48) Dishner, M. H.; Hemminger, J. C.; Feher, F. J. Langmuir 1997, 13, 2318.
(49) Poirier, G. E.; Pylant, E. D. Science 1996, 272, 1145.
(50) Esplandiu, M. J.; Carot, M. L.; Cometto, F. P.; Macagno, V. A.; Patrito, E.
M. Surf. Sci. 2006, 600, 155.
(51) Pflaum, J.; Bracco, G.; Schreiber, F.; Colorado, R.; Shmakova, O. E.; Lee,
T. R.; Scoles, G.; Kahn, A. Surf. Sci. 2002, 498, 89.

7358 DOI: 10.1021/la8043347 Langmuir 2009, 25(13), 7353–7358
Article Voznyy and Dubowski
noticeable deviation from the prototypical packing geometry.
However, our MM simulations for longer chains have shown
that a 7%shrinking of the prototypical ortho2a packing structure
to the size of the unit cell imposed by Au (111) surface would
result in rearrangement of chains, transforming the ortho2a
structure into one of the less strained prototypical packings.
Enforcement of the thiolate-adatom-thiolate binding geometry
gives rise todisruptionof ideal packing at the bottomof the chains
and a gradual change of chain twists to accommodate the
monoclinic, ortho2b, or 3 + 1 structure near the chain termini
(Figure S5). These changes may be potentially responsible for
the
√
3,11,52 2 + 2 (zigzag), and 3 + 141,42,51 phases, respectively,
observed by STM. Gauche defects at the bottom of the chains
could also help to achieve ortho2b packing in the unit cell with
two adatoms, although they are unlikely for short chain thiols
where reduction of packing density would be less expensive
energetically.
Variation of bias from -3 to 3 eV and the choice of local
density of states (LDOS) isosurfaces used in our STMsimulations
did not result in any qualitative change of the obtained STM
patterns (see also Figure S6), in contrast to experimental observa-
tions.41,42,51 Introduction of vacancies on the surface and even
detachment of the SAM from the surface did not affect the STM
images, suggesting that topographic effects are more important
than changes in LDOS (at least in the Tersoff-Hamann
approach), in agreement with previous findings.43We expect that
phenomena other than LDOS should be invoked to explain the
change of STM contrast upon changing the bias, e.g., structural
changes in the monolayer induced by the tip, as described above
for the ortho2a-2 structure or as proposed previously.44
Amongour prototypical structures, only ortho2b (Figure 2d) is
compatible with the zigzag pattern observed experimentally.41-43
Similar patterns can be obtained for monoclinic, ortho, and
ortho2a structures by cutting the chains at different heights, i.e.,
introducing sp- and sp3-like bonding configurations of sulfur.
However, our DFT calculations suggest that sp-like geometry is
much less stable than the sp3-like one. High-resolution STM or
constant-height imaging can potentially resolve the sp- versus
sp3-like bonding43,51 by distinguishing CH-up versus CH2-up
orientation of the chain termini (Figure 2b,d, and Figure S7),
and discriminate between the proposed packing structures.
GIXRD remains the most useful tool to assess the amount of
adatoms at the interface. It can potentially verify whether a zigzag
pattern observed in STM is the result of disruption of the ideal
packing within the thiolate-adatom-thiolate model or is a
consequence of the structure with more adatoms per unit cell
and native ortho2b packing. It should be noted, however, that
strong deviations from bulk positions of gold atoms in the top
surface layers, reported to be up to 0.5 A˚,16,22 can induce a signal
stronger than that from thiol chains, thus providing extra freedom
to fit experimental data even with incorrect chain geometries.
Another variation of GIXRD analysis based on direct compar-
ison of simulated and experimental electron densities and capable
of suggesting the positions of adatoms38 can be particularly useful
to overcome this drawback.
5. Conclusions
We have adapted our previously proposed approach for
searching the SAM structure on surfaces with unknown recon-
struction to alkanethiol SAMs on Au (111). Using the known
dimensions and symmetry of the c(4
2) unit cell, we obtained six
prototypical structures compatible with the constraint of thiol
dense packing. On the basis of these prototypical packings, the
exact structure for the atop-adatommodel was deduced, and two
new structures with two and four adatoms per unit cell were
proposed. No definitive conclusion regarding the preferredmodel
can yet be made. Available theoretical and experimental data
favor the thiolate-adatom-thiolate bonding geometry. How-
ever, in order to be compatiblewith STMdata, some disruption of
the ideal packing in this model is required. Additional experi-
ments using our prototypical packings as a starting point can help
to resolve the longstanding controversy about the structure of
the c(4
2) phase. Application of our approach to other
SAM-substrate systems is envisioned.
Acknowledgment. We thank Xavier Torrelles for preliminary
fits of our structures to experimental GIXRD data, and Peter
Maksymovych for fruitful discussions. The funding for this
research has been provided by theNatural Sciences andEngineer-
ingResearchCouncil ofCanada (STPGP350501-07) andCanada
Research Chair in Quantum Semiconductors Program (J.J.D.).
Computational resources were provided by the R
eseau qu
eb
ecois
de calcul de haute performance (RQCHP).
Supporting Information Available: Description of the
procedure for systematic generation of dense packing struc-
tures and their verification for commensurability with the
surface, 3D structures of the discussed SAM models, and
simulated STM images for different tunneling conditions
and packing configurations. This material is available free of
charge via the Internet at http://pubs.acs.org.
(52) Torrelles, X.; Vericat, C.; Vela, M. E.; Fonticelli, M. H.;Millone, M. A. D.;
Felici, R.; Lee, T. L.; Zegenhagen, J.; Munoz, G.; Martin-Gago, J. A.; Salvarezza,
R. C. J. Phys. Chem. B 2006, 110, 5586.