Unmasking melon by a complementary approach employing electron diffraction, solid-state NMR spectroscopy, and theoretical calculations - Structural characterization of a carbon nitride polymer

Article abstract of DOI:10.1002/chem.200601759

Poly(aminoimino)heptazine, otherwise known as Liebig's melon, whose composition and structure has been subject to multitudinous speculations, was synthesized from melamine at 630°C under the pressure of ammonia. Electron diffraction, solid-state NMR spectroscopy, and theoretical calculations revealed that the nanocrystalline material exhibits domains well-ordered in two dimensions, thereby allowing the structure solution in projection by electron diffraction. Melon ([C₆N₇(NH₂)(NH)] _n, plane group p2gg, a = 16.7, b = 12.4 A, γ = 90°, Z=4), is composed of layers made up from infinite 1D chains of NH-bridged melem (C₆N₇(NH₂)₃) monomers. The strands adopt a zigzag-type geometry and are tightly linked by hydrogen bonds to give a 2D planar array. The inter-layer distance was determined to be 3.2 A from X-ray powder diffraction. The presence of heptazine building blocks, as well as NH and NH₂ groups was confirmed by ¹³C and ¹⁵N solid-state NMR spectroscopy using ¹⁵N-labeled melon. The degree of condensation of the heptazine core was further substantiated by a ¹⁵N direct excitation measurement. Magnetization exchange observed between all ¹⁵N nuclei using a fp-RFDR experiment, together with the CP-MAS data and elemental analysis, suggests that the sample is mainly homogeneous in terms of its basic composition and molecular building blocks. Semiempirical, force field, and DFT/ plane wave calculations under periodic boundary conditions corroborate the structure model obtained by electron diffraction. The overall planarity of the layers is confirmed and a good agreement is obtained between the experimental and calculated NMR chemical shift parameters. The polymeric character and thermal stability of melon might render this polymer a pre-stage of g-C₃N₄ and portend its use as a promising inert material for a variety of applications in materials and surface science.

Full text of DOI:10.1002/chem.200601759

FULL PAPER
DOI: 10.1002/chem.200601759
Unmasking Melon by a Complementary Approach Employing Electron
Diffraction, Solid-State NMR Spectroscopy, and Theoretical Calculations—
Structural Characterization of a Carbon Nitride Polymer
Bettina V. Lotsch,[a] Markus Dçblinger,[a] Jan Sehnert,[b] Lena Seyfarth,[b]
Jürgen Senker,*[b] Oliver Oeckler,[a] and Wolfgang Schnick*[a]
Dedicated to Professor Wolfgang Beck on the occasion of his 75th birthday
Abstract: Poly(aminoimino)heptazine,
otherwise known as Liebigꢀs melon,
whose composition and structure has
been subject to multitudinous specula-
tions, was synthesized from melamine
at 6308C under the pressure of ammo-
nia. Electron diffraction, solid-state
NMR spectroscopy, and theoretical cal-
culations revealed that the nanocrystal-
line material exhibits domains well-or-
dered in two dimensions, thereby al-
lowing the structure solution in projec-
tion by electron diffraction. Melon
give a 2D planar array. The inter-layer
distance was determined to be 3.2 Š
from X-ray powder diffraction. The
presence of heptazine building blocks,
as well as NH and NH2 groups was
confirmed by 13C and 15N solid-state
NMR spectroscopy using 15N-labeled
melon. The degree of condensation of
the heptazine core was further substan-
tiated by a 15N direct excitation mea-
surement. Magnetization exchange ob-
served between all 15N nuclei using a
fp-RFDR experiment, together with
the CP-MAS data and elemental analy-
sis, suggests that the sample is mainly
homogeneous in terms of its basic com-
position and molecular building blocks.
Semiempirical, force field, and DFT/
plane wave calculations under periodic
boundary conditions corroborate the
structure model obtained by electron
diffraction. The overall planarity of the
layers is confirmed and a good agree-
ment is obtained between the experi-
mental and calculated NMR chemical
shift parameters. The polymeric charac-
ter and thermal stability of melon
might render this polymer a pre-stage
of g-C3N4 and portend its use as a
promising inert material for a variety
of applications in materials and surface
science.
([C6N7(NH2)(NH)]n, plane group p2gg,
A
a=16.7, b=12.4 Š, g=908, Z=4), is
composed of layers made up from in-
finite 1D chains of NH-bridged melem
Keywords: ab initio calculations ·
carbon nitrides · electron diffrac-
tion · solid-state NMR spectrosco-
py · solid-state structures
(C6N7(NH2)3) monomers. The strands
G
adopt a zigzag-type geometry and are
tightly linked by hydrogen bonds to
Introduction
The quest for binary carbon nitride C3N4, whose postulated
material properties are believed to push the borders of
ultra-hard materials, has provided a tight link between Lie-
bigꢀs historic work and modern materials chemistry.[1] While
being re-invented in a new context in the 1980s, carbon ni-
tride had been discussed as the ultimate de-ammonation
product of a series of “ammonocarbonic acids” such as cyan-
amide or melamine as early as 1834.[2] All attempts to syn-
thesize pure carbon nitride were invariably spoiled by the
presence of hydrogen, yielding an amorphous, infusible ma-
terial which was first obtained by Berzelius and given the
name “melon” by Liebig.[2,3] As reviewed by Franklin,[3] the
latter was obtained by heating to redness the yellow precipi-
[a] Dr. B. V. Lotsch, Dr. M. Dçblinger, Dr. O. Oeckler,
Prof. Dr. W. Schnick
Department Chemie und Biochemie
Ludwig-Maximilians-Universität
Butenandtstrasse 5–13 (D), 81377 München (Germany)
Fax : (+49)89-2180-77440
E-mail: wolfgang.schnick@uni-muenchen.de
[b] J. Sehnert, L. Seyfarth, Prof. Dr. J. Senker
Anorganische Chemie I
Universität Bayreuth
Universitätsstrasse 30, 95447 Bayreuth (Germany)
Fax : (+49)921-55-2788
E-mail: juergen.senker@uni-bayreuth.de
Supporting information for this article is available on the WWW
under http://www.chemeurj.org/ or from the author.
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4969

tate formed by the action of chlorine on a solution of potas-
sium or sodium thiocyanate, by heating ammonium or mer-
cury thiocyanate, and as the final de-ammonation product
on heating mixed “aquo-ammonocarbonic acids” such as
ammeline or urea.[4,5] Owing to its unusual thermal stability,
some doubt arose whether melon should be classified as an
organic or an inorganic compound. Despite the lack of
sharply defined composition, melon was assigned the empir-
ical formula H3C6N9, which led Redemann and Lucas to
draw two alternative planar structure models based on the
heptazine (cyameluric) core C6N7 (Scheme 1, 1a + 1b). Al-
ternatively, a symmetric triangular form was proposed,
whose empirical formula approaches asymptotically the
limit C3N4 if condensations extend indefinitely (Scheme 1,
1c).[4]
most stable form of carbon nitride, and owing to its analogy
with graphite was attributed the triazine core C3N3 as ele-
mentary building block (Scheme 1, 2a).[1,7] Since then, a ple-
thora of experimental and theoretical efforts have been
made to elucidate the structure of g-C3N4, which failed due
to the ill-defined nature of the obtained polymers.[7,8]
The search for graphitic carbon nitride typically emanates
from derivatives or precursors of triazine, whose structural
integrity is believed to be preserved upon thermal treatment
of the starting materials. Thus, a considerable number of
possible triazine-based structures—akin to graphite—has
been devised and their topologies as well as their relative
stabilities computed.[1d,8,10,11] For example, Kawaguchi et al.
synthesized
a
material with the presumed formula
ꢀ
[(C3N3)2(NH)3]n ( C6N9H3) by polycondensation of cyanuric
chloride and melamine, whose electron diffraction pattern
was indexed on a hexagonal cell with a=8.2 Š and d100
=
7.1 Š.[7b] Demazeau et al. reported on the solvothermal syn-
thesis of g-C3N4 by reacting melamine with hydrazine as a
nitriding agent under supercritical conditions.[7a,g,l,o,11]
Only recently, an alternative structure model based on the
heptazine nucleus emerged (2b, Scheme 1), reminiscent of
the extensive work on melon, which still represents a highly
controversial chapter in the history of C/N chemistry.[8a,12–15]
Although the heptazine-based nitrogen compounds such as
melem or cyameluric acid had been known for decades, the
significantly less stable unsubstituted nucleus, tri-s-triazine,
was synthesized and structurally characterized only in the
1980s.[16] Komatsu proposed polymers of melem to form
upon polycondensation/pyrolysis of NH4SCN and various
heptazine derivatives.[9b,c] The as-obtained graphite-like
“pseudo carbon nitrides” were classified according to their
composition as “symmetric triangular forms” (cf. model 1c,
Scheme 1) or as partially condensed, irregular forms of hep-
tazine-based polymers. In addition, the synthesis of a “cya-
meluric high polymer” by thermal de-ammonation of melon
via a melem-dimer was postulated, affording a linear poly-
mer composed of 42 monomers.[9d] In most cases, the X-ray
powder patterns were indexed on orthorhombic unit cells
(a=7.104, b=16.190, c=12.893 Š;[9b] a=7.229, b=21.512,
c=13.589 Š).[9d] In contrast, the attempt to prepare carbon
nitrides by polycondensation of melamine with the Lewis
acid ZnCl2 afforded a polymer with similar orthorhombic
metrics, which was assigned a vacancy-network structure
consisting of triazine cores.[9a]
As outlined above, the synthetic strategy that appears to
be most promising to obtain reasonably well-defined carbon
nitride materials includes the controlled pyrolysis of CNxHy
precursors.[15] However, the advantage of comparatively high
crystallinity usually accompanies the presence of hydrogen
in the products, which at the same time seems to stabilize
the structure and to function as defect site. On the other
end of the experimental spectrum, high-energy techniques
(ion beam sputtering, CVD, laser- or shock-wave tech-
niques) almost exclusively afford amorphous materials with
substoichiometric nitrogen content with respect to the ideal-
ized formula C3N4.
Scheme 1. Structure models proposed for melon (1a, 1b) and cutouts of
hypothetical structure models of graphitic carbon nitride based on tri-
azine (2a) and heptazine building blocks (2b). The melon-derived struc-
ture “C36N52H12” (1c) has an empirical formula intermediate between
melon and g-C3N4.
Other structure models, including triazine-based variants,
have been put forward for melon owing to the lack of de-
tailed knowledge on the structure of its presumed monomer,
melem.[6] Franklin found that the empirical composition of
melon varied with the method of preparation, as he ob-
tained samples with a hydrogen content between 1.1 and
2.0 wt%. Products with a hydrogen content of only 0.6 wt%
were rationalized by assuming a mixture of several “com-
pact” condensation products of triangular shape as for ex-
ample model 1c (Scheme 1).[3,4a] On this background, it was
conjectured that “it is probably incorrect to assign any one
structure to melon, for it is more than likely a mixture of
molecules of different sizes and shapes. This gives rise to its
amorphous character”.[4a]
The gap between historic and modern carbon nitride
chemistry is bridged by the hypothetical graphitic modifica-
tion of C3N4 (g-C3N4), which has been calculated to be the
4970
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980

Polymeric Carbon Nitride Materials
FULL PAPER
In this work we present the 2D structure of a material,
subject to numerous speculation, that has commonly been
identified with graphitic carbon nitride. We thus provide the
first structural characterization of a polymeric carbon(IV)
nitride based on electron diffraction and solid-state NMR
spectroscopy, which at the same time resolves the controver-
sial identity of melon.
The FTIR spectrum of C/N/H-graphite is displayed in
Figure 1. Owing to obvious analogies with the vibrational
spectra of melon already present in the literature,[9d,19] we
will only focus on the most important results here. The well-
Results and Discussion
Synthesis and characterization: A brownish polymer of the
approximate composition C3N4.4H1.8, (N: 61.2; C: 36.0; H:
1.8) determined from elemental analysis and with an aver-
age C/N ratio of 0.68 (theor. for C3N4: 0.75) was obtained by
heating melamine (triamino-s-triazine) in sealed silica glass
ampoules at 6308C. Whereas at temperatures below 6208C,
melem (C6N7(NH2)3) was found in the products, thermolysis
A
Figure 1. FTIR spectrum of melon recorded as a KBr pellet between 400
above 6408C leads to rapidly increasing carbonization. It
should be pointed out that heating melamine at 5008C in an
open system for several hours affords polymeric materials
with different colors ranging from light yellow to dark beige.
Whereas under these conditions, largely amorphous com-
pounds are obtained, the crystallinity is significantly en-
hanced by pyrolyzing melamine in a closed system under an
autogenous pressure of ammonia that arises from the con-
densation reactions. The brownish color of the as-obtained
products suggests that crystallization accompanies the onset
of carbonization of the sample. Small amounts of oxygen
(up to 2 wt%; typically 0.5–0.7 wt%) detected in the bulk
suggest that the material is prone to water absorption, as
was pointed out in the literature.[17] In addition, melamine
crystals were detected as impurities, which likely result from
depolymerization of the product induced by ammonolysis at
elevated temperatures and pressure of ammonia.[18] Possible
depolymerization products other than melamine, such as di-
cyandiamide, were not observed as crystalline side phases;
however, the presence of very small amounts of amorphous
by-products such as (poly)imides below the detection level
of solid-state NMR spectroscopy (ꢁ5% for CP measure-
ments) may not entirely be excluded. Also, the comparative-
ly harsh synthetic conditions applied in the present context
likely induce heterogeneous crystallization and probably
yield a broad spectrum of differently ordered domains, rang-
ing from amorphous to nanocrystalline. Presumably, a mix-
ture of polymers with different chain lengths is obtained,
which can be considered a feature intrinsic to thermally in-
duced polymerization processes. In this sense, the as-synthe-
sized material, which will be referred to as “C/N/H-graph-
ite” owing to its commonly inferred graphite-like charac-
ter,[7] cannot be considered an overall homogeneous phase.
In situ temperature-programmed X-ray powder diffraction
shows that the material is stable up to about 7708C without
passing through phase transitions prior to its decomposition.
The observed thermal stability of this lightweight material is
therefore comparable to or somewhat higher than that of ar-
omatic polyamides and -imides.[9d]
and 4000 cmꢂ1
.
resolved bands indicate a fairly high degree of ordering.
Whereas the sharp band at about 810 cmꢂ1 can be attributed
to the ring-sextant out-of-plane bending vibration character-
istic of both triazine or heptazine ring systems,[13,15,20] the
prominent absorption bands at 1206 and 1235 cmꢂ1 and pos-
sibly 1316 cmꢂ1 have been shown to be characteristic of the
[19,21,22]
ꢂ
ꢂ
C NH C unit in melam.
Therefore, a similar structur-
ꢂ
ꢂ
al motif, corresponding to either trigonal C N(-C) C (full
ꢂ
ꢂ
condensation) or bridging C NH C units (partial condensa-
tion), can be inferred for the polymer. Absorption found in
ꢂ1
ꢂ
the N H stretching region between 3250 and 3070 cm
proves the presence of NH and/or NH2 groups, which are
most likely integral parts of the structure.
Scanning electron microscopy images demonstrate the
micro- and nanocrystalline character of the CNxHy material
as shown in Figure 2. The morphology of the material re-
sembles that of microcapsules, the hollow tubes and spheres
with a diameter of several micrometers containing nanocrys-
tallites. While the crystalline particles appear to be platelike,
the spheres presumably result from the isotropic bulging of
the particles during gas evolution in the course of condensa-
tion. Owing to the small crystallite size, structure elucidation
by X-ray methods does not seem to be particularly promis-
ing.
In accordance with literature data, the X-ray diffraction
pattern is indicative of a layered substance with an interlay-
er spacing of 3.19 Š.[7,8,14,15,24] The powder pattern resembles
that of graphite, which has an interlayer spacing slightly
larger than that of the carbon nitride material (3.33 Š) as
shown in Figure 3. Remarkably, the strong reflection, which
is commonly indexed as 002 by analogy with graphite, is
sharp, whereas all hk0 and h0l reflections—if present at
all—are weak and broadened. In addition, asymmetric
shapes of the reflections (“tailing” towards higher 2q values,
as seen for the low-angle reflection at 2qꢃ12.68), are visible.
These features could result from “streaking” along c* due to
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
4971

W. Schnick, J. Senker et al.
sion) experiment was carried
out. By evaluating the charac-
teristic time dependence of the
polarization inversion dynam-
ics of the different 15N nuclei,
the number of protons cova-
lently bonded to the latter can
be ascertained. As outlined in
Figure 5, three types of signals
can be distinguished: The sig-
nals between d=ꢂ177 and
ꢂ195 ppm, as well as the reso-
nance at d=ꢂ225 ppm exhibit
a moderate intensity loss, the
continuous, slow decrease of
the polarization being charac-
teristic of tertiary nitrogen
atoms. In contrast, the polari-
zation of the signals at d=
ꢂ245 and ꢂ265 ppm follows a
two-step process induced by
the covalently bonded protons.
The intensity at the cross-over
Figure 2. Scanning electron microscopy images of C/N/H-graphite, taken from various sample regions. The
images reveal crystallite sizes on the nanometer- and micrometer-length scales, as well as the “porous” mor-
phology of the sample.
Figure 3. X-ray powder pattern (CuKa1 radiation) of C/N/H-graphite (con-
tinuous line) and simulation for graphite (bars). The interlayer distance
(strong reflection) amounts to 3.19 Š.
2D planar disorder. However, the sharp main reflection and
absence of splitting of the latter points to the layers being
planar with a sharply defined interlayer spacing.
Solid-state NMR spectroscopy: Solid-state NMR spectrosco-
py is a valuable tool for probing the structure of semicrystal-
line or amorphous materials on local and intermediate
length scales, as it is not dependent on the long-range order
in contrast to diffraction techniques. Owing to the low sensi-
tivity of the 15N nucleus, a 15N-enriched C/N/H-graphite
(degree of enrichment of ꢃ25%) was synthesized starting
from 15N-labeled melamine.
The 15N CP-MAS spectrum of the C/N/H-graphite is dis-
Figure 4. 15N (top) and 13C (bottom) CP-MAS solid-state NMR spectra of
the C/N/H-graphite. The signal assignments according to ab initio calcu-
lations of the chemical shift values for the DFT-optimized cell (see inset)
are indicated on top of the experimental spectra. The respective carbon
sites in the structure are indicated by numbers (black: C, gray: N).
played in Figure 4 (top). The comparatively high resolution
is diagnostic of a semicrystalline rather than an amorphous
material. For a reliable signal assignment (as indicated in
Figure 4), a CPPI (cross-polarization with polarization inver-
4972
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980

Polymeric Carbon Nitride Materials
FULL PAPER
Figure 6. 15N direct excitation spectrum using a recycle delay of 8 h as es-
timated from a 15N T1 measurement to ensure a total recovery of the
magnetization. Number of scans: 16; spinning frequency: 9 kHz.
Figure 5. Evolution of the signal intensities (cf. Figure 4) during the
course of a 15N CPPI experiment. The polarization of the 15N nuclei is
given as a function of the inversion time ti.
nuclei is circumvented, thereby rendering the signal intensi-
ties independent of the number of protons in spatial proxim-
ity. Owing to the very long relaxation time of the 15N nuclei,
a recycle delay between successive scans of 8 h had to be ap-
plied to ensure total magnetization recovery before each
scan. Comparatively long relaxation times are diagnostic of
well-ordered materials. Owing to the isotopic enrichment of
melon, 16 scans were sufficient to acquire a spectrum with a
reasonable signal-to-noise ratio. As depicted in Figure 6, the
signal intensities show significant variations as compared to
the CP-MAS spectrum in Figure 4. As expected, the tertiary
nitrogen resonances gained intensity at the expense of the
protonated nitrogen signals. A quantitative fit of the relative
intensities indicates a ratio of the four signal groups
between the dipolar and spin-diffusion regime is given by [2/
ACHTUNG-
T
between NH (crossover at 0) and NH2 (crossover at ꢂ1/3)
nitrogen nuclei.[13,25]
Therefore, the assignment of all signals according to the
proton environment of the 15N nuclei is feasible. Notably,
apart from tertiary nitrogen atoms expected for a fully con-
densed network, NH and NH2 groups are present, which
suggest the formation of an only partially condensed C/N/H
network.
Information on the nature of the building blocks can be
obtained from the isolated signal at d=ꢂ225 ppm in the 15N
spectrum. This tertiary nitrogen resonance is strongly remi-
niscent of the central nitrogen atom (“Nc”) of the heptazine
core, since it exhibits an up-field shift compared to the outer
nitrogen nuclei of the ring in essentially all heptazine-based
compounds studied so far.[13,26] For example, this nitrogen
resonance is observed at d=ꢂ234.2 ppm for melem.[13,24]
Thus, these findings furnish the first significant indication of
the heptazine-based nature of the graphitic C/N/H material.
The 13C CP-MAS spectrum of the product shows remark-
able similarity with that of melem. In detail, two signal
groups are found with peak maxima at dꢃ164 and 157 ppm.
In melem, two groups of carbon resonances are observed at
d=164.3/166.4 and 155.1/156.0 ppm, respectively, where the
low-field signals (d=164–166 ppm) were assigned to the
carbon positions adjacent to the amino groups.[13,24] Based
on DFT calculations, an analogous assignment of the low-
N
tert:Nc:NH:NH2 roughly corresponding to 7:1:1:1. In partic-
ular, the equal intensities of Nc, NH, and NH2 suggests a
degree of condensation approximately corresponding to that
expected for melon (6:1:1:1, cf. 1a + 1b, Scheme 1). To
sum up, evidence of only one single type of building block—
heptazine—is provided by the CP-MAS and direct excita-
tion spectra. Since atom connectivities and, hence, local
structural motifs dominate the chemical shift of the nuclei,
this missing spread in chemical shift values renders the pres-
ence of triazine ring systems highly unlikely. In addition,
DFT calculations of the chemical shift parameters of fully
condensed triazine- and heptazine-based C3N4 systems sug-
gest that the chemical shift range expected for triazine-
based structures would spread out way downfield as com-
pared to the observed data, for both planar and corrugated
structures.[27] This adds to the above evidence that the C/N/
H-graphite is composed of heptazine rather than triazine
building blocks. Accordingly, the somewhat higher intensity
observed for Ntert may result from the admixture of minor,
heptazine-based side phases exhibiting a slightly higher
degree of condensation (cf. for instance structure 1c,
Scheme 1). The line width observed in the NMR spectra is
slightly increased (by a factor of about 2) as compared to
typical line widths of crystalline molecular compounds. This
is indicative of varying chemical environments of the hepta-
zine nuclei, which depend on the orientation of the adjacent
layers as determined by the local stacking sequence. Similar
field signals (d=163–165 ppm) to CN2(NHx) and the high-
A
field signal (d=157 ppm) to CN3 moieties was accomplished
for the C/N/H-graphite as will be discussed below. The use
of different CP contact times tc corroborates this assign-
ment.
To quantify the relative 15N signal intensities and to draw
conclusions on the degree of condensation of the presumed
heptazine core, an experiment using direct excitation of the
15N nuclei was carried out (Figure 6). In this experiment, the
magnetization transfer from the abundant protons to the 15N
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
4973

W. Schnick, J. Senker et al.
effects may be induced by different polymer lengths, which
give rise to slightly altered local magnetic fields at the sites
of nuclei in the vicinity of chain or layer terminations. In
contrast, for a completely amorphous material the line
widths should be increased by a factor of 10 to 20.
A fp-RFDR experiment was carried out to probe the ho-
monuclear, through-space magnetization transfer between
the 15N nuclei. It can therefore be considered as a sensor for
spatial proximities of the nuclei in the sample. By selectively
exciting the NH2 signal, the time dependence of the magnet-
ization transfer to the surrounding 15N nuclei (other than
NH2) was monitored by successively varying the mixing
time in small intervals. Figure 7 (top spectra) outlines the
will be detailed below (see also Figure 4). This exchange be-
havior suggests homogeneity of the sample in a regime up
to 30 Š. Note that the coexistence of different domains is
thus limited to different stacking variants and possibly hep-
tazine-based structural isomers, which do not give rise to ad-
ditional signals in the 1D CP-MAS spectrum as outlined
above. Summing up, this experiment suggests that all 15N
nuclei are in spatial neighborhoods and no structurally dis-
tinct phases are detectable by 15N NMR spectroscopy.
Electron diffraction: Electron diffraction (ED) can provide
structural insights into nanometer-sized materials. The di-
mensions of the C/N/H-graphite particles can be estimated
to amount to 50–200 nm by transmission electron microsco-
py (TEM, Figure 8). In the diffraction mode, the presence of
domains (Figure 9) of varying crystallinity is evident, as can
Figure 8. TEM images of the C/N/H-graphite on a carbon coated copper
grid, showing the platelike nano- and microcrystals.
Figure 7. 15N CP-MAS spectrum of the C/N/H-graphite and the corre-
sponding overall fit (gray line; 2nd spectrum from bottom). Individual
resonances of the fits (cf. Figure 8) (bottom), and 1D-fpRFDR spectra
for mixing times tm =1, 25, 205 ms.
be seen by the asymmetric peak shapes and diffuse scatter-
ing, the latter being either intrinsic or due to amorphous
cover layers. In agreement with the platelike character of
the crystallites, the preferred orientation is such that most
diffraction patterns are taken along the [001] zone axis,
other zone axes are hardly accessible. As indicated by the
X-ray powder patterns, varying degrees of “streaking” along
c* are observable (Figure 10, right), resulting from planar
defects or possibly turbostratic stacking disorder. Neverthe-
less, SAED patterns of crystalline domains with perfect pe-
1D-RFDR spectra obtained for three mixing times (tm =1,
25, and 205 ms) corresponding to short, intermediate and
long periods of magnetization transfer. Whereas for tm =
1 ms no magnetization transfer to the other nitrogen nuclei
is visible, exchange commences at tm =25 ms for the high-
field part of the tertiary nitrogen signals, which is visible by
the growing intensity of the latter. Very weak signals can
also be distinguished for the Nc and NH resonances. For
long mixing times, all initial resonances are observable,
which indicates a dipolar exchange of magnetization among
the NH2 nitrogen and all other 15N nuclei in the sample.
This observation is confirmed by the build-up curves (see
Figure S4 in the Supporting Information), which trace the
increase of magnetization at the tertiary and NH nitrogen
nuclei and the concomitant loss of signal intensity at the se-
lectively excited NH2 nitrogen atoms. Build-up is fastest for
the tertiary 15N nuclei from d=ꢂ195 to ꢂ187 ppm, which
therefore represent the closest neighbors of the NH2 group.
This finding is in agreement with the ab initio 15N chemical
shift parameters calculated for the DFT-optimized cell
based on the structure solution from electron diffraction, as
Figure 9. Selected area electron diffraction (SAED) patterns of the hk0
plane (zone axis [001]) of the C/N/H-graphite. All patterns exhibit differ-
ent degrees of diffuse scattering and are indicative of partial disorder
and/or thick sample sizes.
4974
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980

Polymeric Carbon Nitride Materials
FULL PAPER
diffraction data.[28] After refinement, the difference Fourier
synthesis has no significant maxima, which can be demon-
strated by arbitrarily removing one atom from the structure
model. The removed atom then yields by far the highest dif-
ference Fourier peak. The bond length precision after re-
finement is 0.045 Š, within this experimental error (neglect-
ing systematic errors), the distances and angles calculated
are reasonable within a 2s interval. Crystallographic data
are summarized in Table 1.[31]
Figure 10. Left: Electron diffraction pattern of the hk0 plane (zone axis
[001]). The pseudo-hexagonal intensity distribution is indicated by white
circles. Merging the reflection groups contained in the circles (shown ex-
emplarily; the same applies for all reflection groups) into one single spot
would generate a diffraction image similar to that of graphite. Right:
SAED pattern viewed perpendicular c*. Whereas the 00l reflections are
clearly visible, “diffuse bars” are observed parallel to c* with a non-dis-
crete intensity distribution (“streaking”).
Table 1. Crystallographic data of the structure solution and refinement
of melon based on ED data.
formula
C6N9H3
201
Mw [gmolꢂ1
]
plane group
p2gg
instrument type
instrument details
JEOL 2011 TEM
single tilt holder, TVIPS CCD camera
(F114)
incident radiation energy [kV] 200
dmin [Š]
a [Š]
b [Š]
0.65
16.7[a]
12.4[a]
3.2[b]
riodicity at least in two dimensions can be obtained, which
are unaffected by the stacking disorder (Figure 10, left).
The commonly claimed graphite-like character of this ma-
terial, which has led to the assumption that it is related to
graphitic carbon nitride, is associated with the pseudo-hex-
agonal intensity distribution in the hk0 plane. Groups of
strong reflections can be correlated with reflections of
graphite with respect to their symmetry as well as the en-
coded distance information (Figure 10, left). Indexing the
hk0 patterns yields a rectangular mesh with cell parameters
a=16.7, b=12.4 Š. These values are similar to those deter-
mined by Komatsu from X-ray powder patterns (a=7.104
layer distance along c [Š]
symmetry-independent reflec- 208
tions
number of parameters
46
R1
Rrim[29]
26.44%
40.2%
[a] The error for determination of the lattice parameters from ED is esti-
mated to be ꢃ5%. [b] Determined from X-ray powder diffraction.
To countercheck the consistency of the structure model,
kinematical SAED diffraction patterns based on the struc-
tural parameters were calculated. The simulation of the hk0
plane is shown in Figure 11 (right), together with experi-
ꢀ
ꢀ
ꢀ
(
2c), b=16.190 ( a), c=12.893 Š ( b)) given the layer
spacing of 3.2 Š.[9b] Taking advantage of a number of benefi-
cial factors, such as the planarity of the layers, the light
atom structure, and large unit cell (small structure factors,
minimum multiple scattering), the structure was solved in
projection based on the electron diffraction data. Owing to
the small sample thickness, the kinematical approximation
2
Ihkl /jFhkl
j
could be used. Evaluation of the observed ab-
sences in the base plane (h0: h=2n+1; 0k: k=2n+1)
allows for the plane group p2gg. Strong reflections remain
sufficiently strong to find phase relationships using direct
methods,[28a] and, thus, a figure of merit of 20.06% was ob-
tained by employing SIR-97[29] for structure solution. Reduc-
ing the symmetry to p2 and introducing the mirror planes as
a twin law yields better residuals. However, the data/param-
eter ratio obtained is very low. Therefore, the improvement
of the residuals is not significant and is accompanied by an
unreasonable distortion of the heptazine rings. Upon regula-
rizing the heptazine units, the residuals again increase, so
that the twin approach does not lead to a real improvement.
All carbon and nitrogen atoms could be located with rea-
sonable bond angles and distances. Upon refinement (208
reflections, 46 parameters), R1 was 26.44%. As reflection
intensities are affected by dynamical diffraction, better re-
siduals have not been obtained; however, they lie in the
range usually reported for refinements based on electron
Figure 11. Experimental (left, middle) hk0 diffraction patterns, from
which the structure solution in projection was accomplished. The left dif-
fraction pattern was recorded under “static” conditions and its intensities
merged with a second pattern (not shown); the middle pattern was re-
corded by using the precession technique.[27g,30] Right: Simulation of an
SAED pattern of the 0th layer based on the proposed structural model,
assuming ideal kinematical scattering of electrons.
mental patterns used for structure solution. The patterns
agree well, although the experimental intensity distribution
deviates slightly from the mm2 symmetry of the plane
group. Note however, that the symmetry of the precession
patterns is significantly less violated.
The 2D projection of the crystal structure is displayed in
Figure 12. The layers comprise infinite chains of “melem-
monomers” condensed via N(H) bridges, thereby forming a
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
4975

W. Schnick, J. Senker et al.
narity of heptazine cores and amide/imide moieties[7a–c,8,14]
indicate overall planarity of the layers, the maximum sym-
metry of a 3D structure in the case of eclipsed layers is
given by space group Pbam. However, theoretical calcula-
tions (see below) suggest a lateral displacement of adjacent
layers to be far more likely. These displacements become fa-
vorable, since p-stacking interactions usually require a slight
displacement of adjacent layers to prevent repulsion of the
negatively polarized p orbitals.[32] The above requirements
can be met by considering layer-offsets in which p2gg sym-
metry is retained in the projections. In other words, it is rea-
sonable to assume monoclinic 3D symmetry, which implies
that the layers are laterally shifted along either a or b with a
monoclinic angle ¼ 908 and an AAA… type of stacking.
When viewed along [001], the associated diffraction patterns
are very similar to those obtained for an orthorhombic set-
ting. Therefore, the structure can alternatively be solved in
the P21/a space group (cell setting a=12.4, b=16.7 Š, b be-
tween 92 and 1158, layers shifted along the a axis). Note
that a layer offset along the b axis has the same negligible
effect on the structure solution. Apart from slight distortions
of the heptazine building blocks for b¼908, the structure
projection of eclipsed layers using orthorhombic and mono-
clinic variants, respectively, are hardly distinguishable. Other
ordered models, such as those with an ABA… layer se-
quence, are incompatible with the experimental data; how-
ever, alternative explanations might be considered.
It seems unlikely that the thickness of the sample corre-
sponds to only a few layers, which would yield 2D scattering
and lead to an interpretation of the diffraction pattern in
terms of a section through diffuse streaks along c*. Howev-
er, in the case of statistically shifted (but not rotated) layers,
without any periodicity of the stacking sequence (cf. powder
diagram, Figure 3), the z component of Patterson vectors is
lost, and the hk0 section would be likewise representative of
interatomic distances only perpendicular to c. In other
words, a structure determination would yield the structure
of a single layer. Diffraction patterns like that shown in
Figure 10 (right) might corroborate this situation;[33] howev-
er, it is impossible to ascertain the exact zone axis orienta-
tion. Therefore, we can conclude that the true situation is to
be found between the borderline cases outlined here.
Figure 12. Projection of the structure of melon. Hydrogen atoms and mo-
lecular fragments from adjacent strands are omitted for clarity; black: C,
gray: N.
closely packed two-dimensional array. The heptazine strands
are arranged in a zigzag-type fashion, allowing close N···N
contacts of 3.1–3.4 Š between adjacent strands, which are
bridged by medium strong hydrogen bonds between the
ring-nitrogen atoms and the NH and NH2 groups, respec-
tively. Covalent interactions within the heptazine backbone
and formation of a delocalized p system seem to play a piv-
otal role in fixing the planar geometry of the strands. Van
der Waals-type interactions between the layers are likely to
contribute significantly to the overall stabilization of the
system; however, they appear to be largely unselective
toward a particular kind of stacking sequence.
As the structure is built up from infinite 1D chains instead
of 2D atomic arrays it does not represent any structure
model postulated for hypothetical g-C3N4. In contrast, it cor-
responds to the “polymer” structure model proposed for
melon (1b, Scheme 1),[3,4,9b] whose existence has therefore
been substantiated. Detailed statements on the structure of
the melon strands are now possible, as for instance the
zigzag-type arrangement of the heptazine motifs as well as
the close packing of the chains leading to a tightly hydro-
gen-bonded 2D array. Furthermore, the alternative structure
model postulated for melon, which is based on a triangular-
shaped trimer of melem (1a, Scheme 1), can be discarded at
least in the present case. However, this does not in principle
disavow the existence of this structural isomer of melon.
The structure solution presented above allows the unique
determination of the 2D structure of melon, and is in full
agreement with the spectroscopic data presented above.
However, the stacking disorder of the bulk sample seems to
impede a comprehensive 3D approach, as due to the disor-
der it may not seem straightforward that the projection
yields the structure of a single layer. Nevertheless, possible
interpretations of the structure projection will be delineated
in the following.
Theoretical calculations: Theoretical calculations were used
as an independent verification of the structure model ob-
tained from electron diffraction. To this end, complementary
calculations were carried out based on three different com-
putational methods. Both a single chain comprising nine
heptazine molecules, as well as an array of six chains con-
taining six heptazine monomers each were optimized by
using the PM3 method.[34,35]
Energy minimization reveals that the single zigzag chain
as found by ED already possesses a stable molecular ar-
rangement in the gas phase. Its geometry after optimization
is outlined in Figure 13 (right). As an alternative, the classi-
cally proposed straight melon chain bends upon energy min-
imization as demonstrated in Figure 13 (left). The tendency
As a starting basis, we assume that the 2D projection de-
lineated above is only compatible with some sort of AAA…
type stacking. Given the fact that X-ray powder patterns,
theoretical calculations (see below), and the expected pla-
4976
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980

Polymeric Carbon Nitride Materials
FULL PAPER
Table 2. Experimental and calculated lattice parameters for the structure
of melon obtained from ED and different theoretical approaches.
Method
a [Š]
b [Š]
Interlayer distance [Š]
(4.43)[a]
CASTEP[a]
17.1
16.6
12.8
12.7
12.8
12.4[c]
ACHTREUNG
PM3[b]
DREIDING
TEM/XRD
16.7–16.8
3.2–3.4
3.2[d]
16.7[c]
[a] Constrained during geometry optimization. [b] Cluster of 36 hepata-
zine units arranged in six planar zigzag chains. [c] Estimated error ꢃ5%.
[d] X-ray powder diffraction.
Figure 13. PM3-optimized polymeric models of melon. Left: “Classical”
model of an initially linear (starting geometry) melon chain with identical
alignment of the heptazine monomers. The linear chain bends upon
energy minimization. Right: Zigzag chain of melon as obtained from ED.
Gray: N; black: C; gray: H; the positions of the latter are empirically
fitted.
layer spacing of 3.2 Š does not lead to a minimum on the
energy surface. This renders orthorhombic metrics associat-
ed with an AAA… stacking extremely unlikely.
An overall description including dispersion interactions
between individual layers is possible based on force field
methods.[37,38] The a and b axes are in excellent agreement
with the values calculated by using PM3 and CASTEP
(Table 2). For an eclipsed layer arrangement, which is con-
sidered to be energetically unfavorable in terms of interlay-
er van der Waals interactions, the repulsive inter-planar
forces are maximized, and a layer distance of 3.4 Š is ob-
tained. In contrast, for layer arrangements exhibiting an
offset stacking—not complying with orthorhombic metrics—
cells with interlayer distances of 3.2 Š were found, as ob-
served in the experiment.
of a pair of NH-bridged heptazine rings to bend apart is
also visible in the linear zigzag arrangement, whereby the al-
ternating curvatures about every NH bridge compensate,
yielding an overall linear array. When assembling six zigzag
chains into a 2D arrangement according to the structure
proposed by ED, a planar oligomer is obtained upon optimi-
zation, which is highly stabilized by a tight hydrogen-bond-
ing network (compare Figure 14).
For the DFT-optimized unit mesh, ab initio 13C and 15N
NMR chemical shift parameters were calculated. As demon-
strated in Figure 4, the calculated 13C and 15N chemical shifts
are in good agreement with those in experimental spectra
and confirm the overall signal assignments extracted from
the CPPI experiment. The 13C chemical shift range is repro-
duced by the calculations; in the theoretical 15N spectrum,
the resonance of the central nitrogen atom is slightly shifted
towards higher field; the same tendency is observed for the
NH and NH2 groups. The calculated chemical shifts also
allow one to distinguish between tertiary nitrogen atoms
bonded to different NHx groups, since they are directly af-
fected by the surrounding NH (low-field shift) and NH2
moieties (high-field shift), respectively. These observations
match well with the results from the 1D fp-RFDR NMR ex-
periment and confirm the structure model derived from the
ED experiments.
Figure 14. Planar cluster of melon after geometry optimization under pe-
riodic boundary conditions using CASTEP. The hydrogen positions are
empirically fitted and the hydrogen-bonding network is indicated by
black dots. Gray: N; black: C; gray: H.
To calculate the NMR chemical shift parameters for a
single layer of melon, the latter was further optimized by
using DFT (Figure 14).[36] The periodicity of the PM3-opti-
mized core region of the above planar oligomer allowed the
extraction of the a/b cell parameters and the establishment
of a rectangular unit mesh. Although symmetry restrictions
were not applied during structure optimization, a planar
structure with a rectangular mesh was retained. Whereas the
resulting b axis matches the value found by calculations at
the PM3 level, the a axis is slightly elongated as outlined in
Table 2. The overestimation of the a axis can be rationalized
by the poor description of dispersion forces in hydrogen-
bonding networks obtained by using DFT. Furthermore, at-
tempts to optimize an orthorhombic unit cell with an inter-
Conclusion
The above results shed light on the long-standing debate
concerning the identity of Liebigꢀs inert compound “melon”,
whose structure has now been proven for the first time. The
2D structure of melon reveals the heptazine molecular
building blocks and polymeric nature of this important CNx
precursor. Owing to the weak attractive interlayer forces,
the “bulk” 3D structure is intrinsically affected by inherent
planar defects associated with translations of the layers. Dif-
ferent energetically similar stacking modes with various
translational layer offsets are likely to contribute to the
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
4977

W. Schnick, J. Senker et al.
overall 3D structure.[39] This feature adds to the special im-
portance of understanding the 2D arrangement of the mo-
lecular building blocks.
which was sealed off under an atmosphere of argon and heated to tem-
peratures around 6008C (1 Kminꢂ1) for at least 12 h. Alternatively, ad-
mixtures of the starting material melamine were removed by sublimation
using a cold finger integrated into a glass tube. Upon purification, the
brownish material lightened up, thereby changing its color into reddish-
brown.
Considering the fact that melon is composed of planar
layers built up by a carbon(IV) nitride core (heptazine
units), this compound may be considered as a “defective g-
C3N4” material, whose graphite-like topology may in fact
result from the presence of network terminations in the
form of NH and NH2 groups. These function as triggers for
strain release in such a way that buckling of the sheets is
prevented. As has recently been noted,[14b] melon (in that
work called g-C3N4) may therefore be considered a promis-
ing functional material in catalysis or surface science, and at
the same time represent a pre-stage of graphitic carbon ni-
tride. Along the same lines, the recently synthesized sp3-hy-
bridized carbon nitride imide C2N2(NH), which represents a
formal pre-stage of 3D C3N4, can be considered the high-
pressure analogue of melon.[40]
However, the existence of melon highlights characteristic
differences between graphite and “graphitic” carbon nitride
materials, and in doing so, demonstrates that the analogy be-
tween the two systems is clearly limited both with respect to
chemical and structural aspects. Given the fact that essen-
tially all attempts to prepare g-C3N4 from hydrogen-contain-
ing precursors yield materials with similar structural features
and hydrogen contents, we hypothesize that the products of
the frequently claimed syntheses of graphitic carbon nitride
have in fact rather been polymers such as melon or related
compounds.[7,14,11,9a–c] We can thus add the first direct evi-
dence to the ongoing debate on the structure of the hypo-
thetical g-C3N4, that triazine-based models, which are still
predominantly being discussed in the literature, should
henceforward be assessed very critically. Therefore, the pres-
ent work may stimulate a careful re-evaluation of commonly
accepted paradigms about the existence and structure of
graphitic carbon nitride.
15N-enriched melamine was prepared according to the procedure intro-
duced by Jürgens et al.[13, 24] Sodium tricyanomelaminate Na3
[C6N9] was
R
prepared by heating sodium dicyanamide Na[N(CN)2] (2 g; ꢄ96%,
Fluka) to 5008C (5 Kminꢂ1) in a quartz tube under pressure equalization
and subsequently reacting the as-obtained material (742.5 mg, 2.78
10ꢂ3 mol) with 15NH4Cl (183 mg, 3.36”10ꢂ3 mol, ꢄ98%, Cambridge Iso-
topes) in a Duran tube (length: 160 mm, 1ext. : 26 mm, 1int. : 24 mm) at
4708C (1 Kminꢂ1) for 12 h.[13,24] The raw material was further purified by
sublimation (1 Pa, 2208C).
General techniques: Elemental analyses were performed by using a com-
mercial C, H, N elemental analyzer system Vario EL (Elementar Analy-
sensysteme GmbH).
XRD measurements were performed on a Stoe-Stadi P diffractometer.
High-temperature in situ X-ray diffraction was carried out on a STOE
Stadi P powder diffractometer (Ge(111)-monochromated MoKa1 radia-
G
tion, l=70.093 pm) with an integrated furnace using unsealed quartz ca-
pillaries (1 0.5 mm) as sample containers.
FTIR measurements were carried out on a Bruker IFS 66v/S spectrome-
ter. Spectra of the samples were recorded by utilizing KBr pellets (1 mg
sample, 500 mg KBr, hand press with press capacity 10 kN) in an evacuat-
ed cell equipped with a DLATGS detector at ambient conditions be-
tween 400 and 4000 cmꢂ1
.
Scanning electron microscopy was performed on a JEOL JSM-6500F
equipped with a field emission gun at an acceleration voltage of 10 kV.
Samples were prepared by putting the powder specimen on adhesive con-
ductive pads and subsequently coating them with a thin conductive
carbon film.
Solid-state NMR spectroscopy: 13C and 15N CP-MAS solid-state NMR
spectra were recorded at ambient temperature on the conventional im-
pulse spectrometers DSX Avance 500 (Bruker) and DSX Avance 400
(Bruker) operating at a proton resonance frequency of 500 MHz and
400 MHz, respectively. The samples were contained in 4-mm ZrO2 rotors,
which were mounted in standard double-resonance MAS probes
(Bruker). The 13C and 15N signals were referenced with respect to TMS
and nitromethane, respectively. Data collection of all experiments was
performed applying broadband proton decoupling using a TPPM se-
quence.[41] For the CP MAS spectra of both nuclei a ramped cross-polari-
zation sequence was employed where the 1H pulse amplitude was de-
creased linearly by 50%. Contact times between 10 ms and 20 ms were
used and the recycle delay was chosen to allow a nearly complete mag-
netization recovery optimized via 1H spin-lattice relaxation experiments.
The spinning frequencies nrot varied between 7 and 12 kHz. To determine
the number of covalently bonded protons to the nitrogen atoms, a 15N
CPPI[25] (cross-polarization combined with polarization inversion) experi-
ment was performed (nrot =6 kHz). The polarization inversion behavior
was probed by varying the inversion time from 0.2 to 800 ms (28 spectra),
with an initial contact time of 2 ms. 15N direct excitation spectra were ac-
quired with three back-to-back 908 pulses to eliminate unwanted contri-
butions from the probe and ringing effects.[42] The nutation frequency and
the recycle delay were adjusted to 75 kHz and 28800 s (=8 h) (estimated
from a 15N T1 measurement) to ensure a total recovery of the magnetiza-
tion. A total amount of 16 scans was collected and the spinning frequency
was set to 9 kHz. To probe homonuclear 15N connectivities and distances,
selective excitation fp-RFDR experiments were performed by using a
XY-16 phase cycle.[42] For the fp-RFDR mixing block active rotor syn-
chronization was applied with nrot =15 kHz. The length of the soft 1808
pulse in the middle of each rotor period was adjusted according to
Experimental Section
Synthesis of melon: Pyrolyses of melamine were carried out in sealed
silica glass ampoules under vacuum at temperatures between 560 and
6008C, or predominantly under dry argon at temperatures between 620
and 6408C. Typically, melamine (230 mg, 1.83”10ꢂ3 mol; ꢄ 99%, Fluka)
was transferred into a thick-walled silica glass tube (1ext. 15 mm, 1int.
11 mm), from which water had been removed by thorough heating and
evacuation. The tube was sealed by using a hydrogen burner at a length
of about 120 mm. The ampoule was then placed in a vertical tube furnace
and heated (1 Kminꢂ1) to 6308C, at which temperature the sample was
held for 12 to 36 h. The cooling rates did not notably affect the crystallin-
ity of the products, so that the furnace was typically switched off and the
sample was allowed to cool down to room temperature over several
hours. The sample was then isolated by carefully breaking the ampoule,
upon which a significant amount of HCN and ammonia was released.
The yield of brownish residue typically amounted to 38–57%; approxi-
mately 8–10% were recovered from the top of the ampoule in the form
of a brown sublimate mixed together with long, needle-shaped melamine
crystals. To purify the residue and eliminate melamine and—present to
only a small extent—melem crystals, the product (ꢃ80–100 mg) was
transferred into a Duran ampoule (pre-dried in vacuo under heating),
p(1808)=0.3·1/nrot corresponding to a nutation frequency of 25 kHz. To
E
ensure a selective excitation of the 15N resonance of the NH2 groups a
CP sequence with a short CP contact time (90 ms) was followed by a
comb of two 908 pulses (3.4 ms) on-resonant on the NH2-signal with an in-
terpulse spacing
t between 280 ms and 340 ms corresponding to t=
4978
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980

Polymeric Carbon Nitride Materials
FULL PAPER
1= (n(NH)ꢂn
C
sion pattern, especially I. Daberkow and J. Portillo. We also thank Prof.
E. Rçssler (University of Bayreuth) for making available his NMR
equipment, and Mrs. C. Buhtz for her support with the data analysis.
4
Electron diffraction/transmission electron microscopy: ED and TEM
measurements were carried out on a JEOL 2011 instrument equipped
with a tungsten cathode operating at 200 kV. The images were recorded
using a TVIPS CCD camera (F114). The sample was finely dispersed by
sonication in ethyl alcohol suspension for 30 minutes, and a small amount
of the suspension was subsequently dispersed on a copper grid coated
with holey carbon film. The grids were mounted on a single tilt holder
with a maximum tilt angle of 308 and subsequently transferred to the mi-
croscope. Suitable crystallites were singled out among those yielding dif-
fraction patterns of main poles, typically with the zone axis [001] aligned
along the electron beam. The selected-area aperture was adapted in each
case to the size of the selected thin crystalline domains. Precession ex-
periments were conducted by using a FEI Tecnai 12 transmission electron
microscope with a LaB6 cathode, operating at 120 kV and equipped with
a Spinning Star precession interface (NanoMEGAS). A precession angle
of approximately 1.58 was applied. The images were recorded on a
TVIPS 2k CCD camera (F224HD) with a dynamic range exceeding
25000:1. Reflection intensities were extracted by using the ELD program
package;[44] for simulation of the kinematical diffraction patterns the pro-
gram JSV1.08 Lite[45] was employed. Calculation of the electron diffrac-
tion patterns was done using the programs VEC[46] and JEMS.[47] In prin-
ciple, it is possible to solve a structure provided the strong reflections
remain sufficiently strong to find phase relationships using direct method-
s.[28a] To quantify all strong hk0 reflections, intensities of two hk0 diffrac-
tion patterns were merged, yielding a dataset of 208 independent reflec-
tions. Evaluation of the observed absences in the base plane (h0: h=
2n+1; 0k: k=2n+1) indicates the presence of the plane group p2gg.
Owing to the lack of detailed three-dimensional information, the struc-
ture was solved by using the space groups P21212 or Pbam, which corre-
spond to the plane group p2gg in (001) projection. The most probable so-
lution as found by SIR-97[29a] had a figure of merit of 20.06%.[29b] The
heptazine molecular unit was obtained without prior fixation of parame-
ters. Refinement of the ED data for calculation of Fourier maps was
done with the program SHELX-97,[48] using the electron form factors as
given by Doyle and Turner.[49]
Calculations: In the cluster approach the semiempirical PM3 method[34]
was used for structure optimization with the Gaussian03 program pack-
age.[35] The input structures were created from the ED structure solution
with hydrogen added to the NH and NH2 groups. DFT calculations
under periodic boundary conditions were performed with the MS Model-
ling 4.0 package by Accelrys. The input cell was created in orthorhombic
symmetry from a cutout in the core region of the PM3-optimized struc-
ture. For the CASTEP[36] calculations the PBE functional and ultrasoft
pseudopotentials were taken with sampling over 6 k-points. In the struc-
ture optimization of the input cell an energy cutoff of 280 eV was used.
To ensure that the dispersion interaction between neighboring layers
become negligible, a constrained slab of 4.43 Š along the c axis was intro-
duced. NMR parameters were calculated with a cutoff of 350 eV for the
optimized cell. For a core cutout of the PM3-optimized cluster the
Hirschfeld charges were determined with the DMol3 program, the PBE
functional and the DNP basis set.[38] These partial charges were used in
flexible body structure optimizations of the input cell with the Dreiding
force field.[37] By default, 1,4-intramolecular electrostatic interactions
were excluded from the energy evaluation.
[1] a) M. L. Cohen, Phys. Rev. B 1985, 32, 7988; b) A. Y. Liu, M. L.
Cohen, Science 1989, 245, 841; c) C.-M. Sung, M. Sung, Mater.
Chem. Phys. 1996, 43, 1; d) D. M. Teter, R. J. Hemley, Science 1996,
271, 53.
[2] J. Liebig, Ann. Pharm. 1834, 10, 10.
[3] E. C. Franklin, J. Am. Chem. Soc. 1922, 44, 486.
[4] a) C. E. Redemann, H. J. Lucas, J. Am. Chem. Soc. 1940, 62, 842;
b) L. Pauling, J. H. Sturdivant, Proc. Natl. Acad. Sci. USA 1937, 23,
615.
[5] B. Bann, S. A. Miller, Chem. Rev. 1958, 58, 131.
[6] a) L. Costa, G. Camino, G. Martinasso, Polym. Prepr. Am. Chem.
Soc. Div. Polym. Chem. 1989, 30, 531; b) H. May, J. Appl. Chem.
1959, 9, 340.
[7] a) I. Alves, G. Demazeau, B. Tanguy, F. Weill, Solid State Commun.
1999, 109, 697; b) Kawaguchi, K. Nozaki, Chem. Mater. 1995, 7, 257;
c) E. G. Gillan, Chem. Mater. 2000, 12, 3906; d) Q. Guo, Y. Xie, X.
Wang, S. Lv, T. Hou, X. Liu, Chem. Phys. Lett. 2003, 380, 84; e) J. L.
Zimmermann, R. Williams, N. Khabashesku, J. L. Margrave, Nano
Lett. 2001, 1, 731; f) M. Todd, J. Kouvetakis, T. L. Groy, D. Chandra-
sekhar, D. J. Smith, P. W. Deal, Chem. Mater. 1995, 7, 1422; g) H.
Montigaud, B. Tanguy, G. Demazeau, I. Alves, S. Courjault, J.
Mater. Sci. 2000, 35, 2547; h) V. N. Khabashesku, J. L. Zimmerman,
J. L. Margrave, Chem. Mater. 2000, 12, 3264; i) J. Kouvetakis, A.
Bandari, M. Todd, B. Wilkens, Chem. Mater. 1994, 6, 811; j) D. R.
Miller, D. C. Swenson, E. G. Gillan, J. Am. Chem. Soc. 2004, 126,
5372; k) D. R. Miller, J. Wang, E. G. Gillan, J. Mater. Chem. 2002,
12, 2463; l) S. Courjault, B. Tanguy, G. Demazeau, C. R. Acad. Sci.
Ser. IIc: Chim. 1999, 2, 487; m) Z. H. Zhang, K. Leinenweber, M.
Bauer, L. A. J. Garvie, P. F. McMillan, G. H. Wolf, J. Am. Chem.
Soc. 2001, 123, 7788; n) Q. Guo, Y. Xie, X. J. Wang, S. Zhang, T.
Hou, S. Lv, Chem. Commun. 2004, 1, 26; o) H. Montigaud, B.
Tanguy, G. Demazeau, I. Alves, M. Birot, J. Dunogues, Diamond
Relat. Mater. 1999, 8, 1707.
[8] a) E. Kroke, M. Schwarz, Coord. Chem. Rev. 2004, 248, 493; b) S.
Muhl, J. M. Mendez, Diamond Relat. Mater. 1999, 8, 1809; c) T.
Malkow, Mater. Sci. Eng. A 2001, 302, 311; d) M. C. dos Santos, F.
Alvarez, Phys. Rev. B 1998, 58, 13918.
[9] a) T. Komatsu, J. Mater. Chem. 2001, 11, 799; b) T. Komatsu, J.
Mater. Chem. 2001, 11, 802; c) T. Komatsu, T. Nakamura, J. Mater.
Chem. 2001, 11, 474; d) T. Komatsu, Macromol. Chem. Phys. 2001,
202, 19.
[10] D. T. Vodak, K. Kim, L. Iordanidis, P. G. Rasmussen, A. J. Matzger,
O. M. Yaghi, Chem. Eur. J. 2003, 9, 4197.
[11] G. Demazeau, J. Mater. Chem. 1999, 9, 15.
[12] E. Kroke, M. Schwarz, E. Horath-Bordon, P. Kroll, B. Noll, A. D.
Norman, New J. Chem. 2002, 26, 508.
[13] B. Jürgens, E. Irran, J. Senker, P. Kroll, H. Müller, W. Schnick, J.
Am. Chem. Soc. 2003, 125, 10288.
[14] a) M. Groenewolt, M. Antonietti, Adv. Mater. 2005, 17, 1789; b) F.
Goettmann, A. Fischer, M. Antonietti, A. Thomas, Angew. Chem.
2006, 118, 4579; Angew. Chem. Int. Ed. 2006, 45, 4467.
[15] a) B. V. Lotsch, W. Schnick, Chem. Mater. 2005, 17, 3976; b) B. V.
Lotsch, W. Schnick, Chem. Mater. 2006, 18, 1891.
[16] a) R. S. Hosmane, M. A. Rossman, N. J. Leonard, J. Am. Chem. Soc.
1982, 104, 5497; b) A. M. Halpern, M. A. Rossman, R. S. Hosmane,
N. J. Leonard, J. Phys. Chem. 1984, 88, 4324; c) M. Shahbaz, S.
Urano, P. R. LeBreton, M. A. Rossman, R. S. Hosmane, N. J. Leo-
nard, J. Am. Chem. Soc. 1984, 106, 2805.
Acknowledgements
We gratefully acknowledge financial support that was granted from the
Deutsche Forschungsgemeinschaft (DFG) (projects SCHN 377/12–1 and
SE 1417/2–1), Fonds der Chemischen Industrie (FCI), the BMBF, and
the Studienstiftung des Deutschen Volkes (scholarships for B. V. Lotsch).
We gratefully acknowledge NanoMEGAS and TVIPS for providing us
with their equipment and for experimental support to obtain the preces-
[17] D. D. Cubicciotti, W. M. Latimer, J. Am. Chem. Soc. 1948, 70, 3509.
[18] C. Grundmann, A. Kreutzberger, J. Am. Chem. Soc. 1955, 77, 6559.
[19] a) A. I. Finkelꢀshtein, N. V. Spiridova, Russ. Chem. Rev. 1964, 33,
400; b) A. I. Finkelꢀshtein, Opt. i Spektr. 1959, 6, 17.
[20] a) M. K. Marchewka, Bull. Korean Chem. Soc. 2004, 25, 466;
b) M. K. Marchewka, J. Chem. Res. Synop. 2003, 518; c) P. J. Larkin,
Chem. Eur. J. 2007, 13, 4969 – 4980
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
4979

W. Schnick, J. Senker et al.
M. P. Makowski, N. B. Colthoup, Spectrochim. Acta 1999, 55, 1011;
d) W. Jeremy-Jones, W. J. Orville-Thomas, Trans. Faraday Soc. 1959,
55, 193; e) E. N. Boitsov, A. I. Finkelꢀshtein, Russ. Chem. Rev. 1962,
31, 712.
y, Jr., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyen-
gar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N.
Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K.
Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda,
O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian,
J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E.
Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W.
Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J.
Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C.
Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari,
J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cio-
slowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaro-
mi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng,
A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W.
Chen, M. W. Wong, C. Gonzalez, and J. A. Pople, Gaussian, Inc.,
Wallingford CT, 2004.
[21] a) M. Takimoto, Kogyo Kagaku Zasshi 1961, 64, 1452; b) M. Taki-
moto, Kogyo Kagaku Zasshi 1964, 85, 168.
[22] B. V. Lotsch, W. Schnick, Chem. Eur. J. DOI: 10.1002/
chem.200601291.
[23] V. V. Khorosheva, A. I. Finkelꢀshtein, Zh. Fiz. Khim. 1962, 36, 1055.
[24] B. Jürgens, Ph. D. thesis, University of Munich (Germany), Shaker,
Aachen, 2004.
[25] C. Gervais, F. Babonneau, J. Maquet, C. Bonhomme, D. Massiot, E.
Framery, M. Vaultier, Magn. Reson. Chem. 1998, 36, 407.
[26] a) A. Sattler, Diploma thesis, University of Munich (Germany),
2005; b) A. Sattler, L. Seyfarth, J. Senker, W. Schnick, Z. Anorg.
Allg. Chem. 2005, 631, 2545.
[27] J. Sehnert, K. Bärwinkel, J. Senker, J. Phys. Chem. B, 2007, submit-
ted.
[36] M. D. Segall, P. J. D. Lindan, M. J. Probert, C. J. Pickard, P. J.
Hasnip, S. J. Clark, M. C. Payne, J. Phys. Condens. Matter 2002, 14,
2717.
[37] S. L. Mayo, B. D. Olafson, W. A. Goddard III, J. Phys. Chem.1990,
94, 8897.
[38] a) B. Delley, J. Chem. Phys. 1990, 92, 508; b) B. Delley, J. Chem.
Phys. 2000, 113, 7756.
[39] J. E. Lowther, Phys. Rev. B, 1999, 59, 11683.
[40] E. Horvath-Bordon, R. Riedel, P. F. McMillan, P. Kroll, G. Miehe,
P. A. van Aken, A. Zerr, P. Hoppe, O. Shebanova , I. McLaren, S.
Lauterbach, E. Kroke, R. Boehler, Angew. Chem. 2007, 119, 1498;
Angew. Chem. Int. Ed. 2007, 46, 1476 .
[41] A. E. Bennett, C. Rienstra, M. Auger, K. V. Lakshmi, R. G. Griffin,
J. Chem. Phys. 1995, 103, 6951.
[28] a) T. E. Weirich, X. Zou, R. Ramlau, A. Simon, G. L. Cascarano, C.
Giacovazzo, S. Hovmçller, Acta Crystallogr. Sect. A 2000, 56, 29;
b) T. E. Weirich, R. Ramlau, A. Simon, S. Hovmçller, X. Zou,
Nature 1996, 382, 144; c) U. Kolb, G. N. Matveeva, Z. Kristallogr.
2003, 218, 259; d) T. E. Weirich, J. Portillo, G. Cox, H. Hibst, S. Nic-
olopoulos, Ultramicroscopy 2006, 106, 164; e) D. L. Dorset, C. J. Gil-
more, Acta Crystallogr. Sect. A 2000, 56, 62; f) D. L. Dorset, Z. Kris-
tallogr. 2003, 218, 458; g) R. Vincent, P. A. Midgeley, Ultramicrosco-
py 1994, 53, 271.
[29] a) A. Altomare, M. C. Burla, M. Camalli, G. L. Cascarano, C. Giaco-
vazzo, A. Guagliardi, A. G. G. Moliterni, G. Polidori, R. Spagna, J.
Appl. Crystallogr. 1999, 32, 115; b) A. Altomare, G. L. Cascarano, C.
Giacovazzo, A. Guagliardi, J. Appl. Crystallogr. 1993, 26, 343.
[30] M. S. Weiss, J. Appl. Crystallogr. 2001, 34, 130.
[42] S. Zhang , X. Wu , M. Mehring, Chem. Phys. Lett. 1990, 173, 481.
[43] Y. Ishii, J. Chem. Phys. 2001, 114, 8473.
[31] The structure solution was verified by using precession intensity
data. Owing to their robustness with respect to sample misalignment
and less pronounced dynamical effects, the data quality obtained
under otherwise identical experimental conditions can be improved.
Accordingly, Rrim is 20% and the straightforward structure solution
has a figure of merit of 19.5%. The structure refinement (184 inde-
pendent reflections, no. of reflections/parameter ꢃ6, dmin. 0.78 Š)
was less successful due to the large sample thickness, which gives
rise to stronger dynamical effects (cf. Figure 11, middle).
[32] a) C. Janiak, J. Chem. Soc. Dalton Trans. 2000, 3885; b) C. A.
Hunter, J. M. Sanders, J. Am. Chem. Soc. 1990, 112, 5525.
[33] a) J. M. Cowley, Acta Crystallogr. 1961, 14, 920; b) J. M. Cowley, A.
Goswami, Acta Crystallogr. 1961, 14, 1071.
[44] a) X. D. Zou, Y. Sukharev, S. Hovmçller, Ultramicroscopy 1993, 49,
147; b) X. D. Zou, Y. Sukharev, S. Hovmçller, Ultramicroscopy 1993,
52, 436.
[45] S. Weber, Java Structure Viewer, v1.08, 1999.
[46] Z. H. Wan, Y. D. Liu, Z. Q. Fu, Y. Li, T. Z. Cheng, F. H. Li, H. F.
Fan, Z. Kristallogr. 2003, 218, 308.
[47] P. A. Stadelmann, Ultramicroscopy 1987, 21, 129.
[48] G. M. Sheldrick, SHELX-97, Programs for the solution and the re-
finement of crystal structures, University of Gçttingen, Gçttingen
(Germany), 1997.
[49] P. A. Doyle, P. S. Turner, Acta Crystallogr. A 1968, 24, 390.
[34] J. J. P. Stewart, J. Comput. Chem. 1989, 10, 221.
[35] Gaussian 03, Revision C.02, M. J. Frisch, G. W. Trucks, H. B. Schle-
gel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. A. Montgomer-
Received: December 7, 2006
Published online: April 5, 2007
4980
www.chemeurj.org
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 4969 – 4980