Surface modification of fluorinated aluminas: Application of solid state NMR spectroscopy to the study of acidity and surface structure

Article abstract of DOI:10.1016/j.jcat.2004.02.013

Solid-state NMR methods have been used to determine the mechanism for the fluorination reaction of CHClF₂ on alumina. The initial stages of fluorination lead to the formation of terminal F-Al groups (i.e., fluorine bound through a single bond t

Full text of DOI:10.1016/j.jcat.2004.02.013

Journal of Catalysis 224 (2004) 69–79
www.elsevier.com/locate/jcat
Surface modification of fluorinated aluminas: Application of solid state
NMR spectroscopy to the study of acidity and surface structure
Peter J. Chupas 1 and Clare P. Grey ∗
Department of Chemistry, State University of New York at Stony Brook, Stony Brook, NY 11794-3400, USA
Received 2 December 2003; revised 6 February 2004; accepted 11 February 2004
Abstract
Solid-state NMR methods have been used to determine the mechanism for the fluorination reaction of CHClF on alumina. The initial
2
stages of fluorination lead to the formation of terminal F–Al groups (i.e., fluorine bound through a single bond to an aluminum atom),
which is consistent with a recent XPS report [O. Boese, W.E.S. Unger, E. Kemnitz, S.L.M. Schroeder, Phys. Chem. Chem. Phys. 4 (2002)
27
2824]. Al NMR indicates that five-coordinate aluminum sites are initially consumed during the fluorination reaction, suggesting that the
five-coordinate aluminum species are predominately localized near the surface of the high surface area (∼ 3–5 nm) alumina particles. The
31
sorption of basic phosphines as probe molecules for acid sites, coupled with P NMR, was used to follow the changes in acidity after
fluorination of the surface. The untreated alumina samples do not contain Brønsted acid sites, partial fluorination of the surface leading
to their creation. In contrast, the Lewis acid sites, initially present on the untreated alumina, decrease in concentration on fluorination.
Continued fluorination of the surface eventually leads to the destruction of the Brønsted acid sites, some Lewis acid sites remaining but in
lower concentrations than on the initial untreated alumina.
2004 Elsevier Inc. All rights reserved.
Keywords: NMR; Fluorinated; Alumina; Acid
1. Introduction
The surface structure of catalytic aluminas should first
be considered to monitor further chemical modification of
surface structure. The common catalytic aluminas, γ -Al2O3
and η-Al2O3, are nonstoichiometric in that they contain sig-
nificant concentrations of protons. The surfaces of these
catalytic transition aluminas are covered with hydroxyl
groups bound to sites of various geometries. Knozinger and
Ratnasamy’s review on the surface structure of aluminas de-
scribed five types of surface sites [8]. The µ1-OH groups are
referred to as type Ia and Ib and are bound to single tetra-
hedral and octahedral aluminum atoms, respectively. Two
types of µ2-OH groups, one bound to a tetrahedral and
octahedral aluminum (IIa) and the other to two octahedral
aluminum atoms (IIb) and a µ3-OH group bound to three
octahedral aluminum atoms (IIIb) were also discussed [8].
Other types of surface hydroxyls may also result from the
presence of five-coordinate aluminum species, such as a µ2-
OH group bound to a five- and six-coordinate aluminum site.
The different surface hydroxyls have inherently differ-
ent acidic properties. For example, µ2-OH species show
Brønsted acid properties while µ1-OH species do not [8].
Fluorination can significantly modify the acid properties of
Surface modification is often used to tune reactivity in
the field of heterogeneous catalysis, and it is common for
solid catalysts to be pretreated or chemically modified before
the desired reactivity is observed. For example, the addi-
tion of chlorine and/or fluorine to the surface of alumina has
been shown to modify acidity and, subsequently, activity for
acid-catalyzed F/Cl halogen exchange [1,2], alkylation [3,4],
isomerization [5,6], and cracking reactions [7]. Recently,
halogenated aluminas, both chlorinated and fluorinated, have
attracted considerable attention for use as solid acid catalysts
[1–7]. One potential drawback to these catalysts is the ob-
served decrease in activity with time on stream. The drastic
changes in activity that result from fluorine addition are be-
lieved to be associated with distinct structural changes, and
it is these changes that have yet to be fully elucidated.
*
Corresponding author. Fax: (631)632-5731.
E-mail address: cgrey@mail.chem.sunysb.edu (C.P. Grey).
Current address: Materials Science Division, Argonne National Labo-
1
ratory, Argonne, IL 60439, USA.
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2004.02.013
2004 Elsevier Inc. All rights reserved.

70
P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
oxide catalysts and, for alumina, can lead to a strengthening
of the acid sites [1–7]. Several techniques have been em-
ployed to study the effects of the fluorination of aluminum
oxide, including NMR [9–12], XANES (X-ray absorption
near-edge spectroscopy) [13], IR [14–17], and X-ray diffrac-
tion [9,10,19–21]. There has been considerable discussion
in the literature as to the basis of the modification of both
Brønsted and Lewis acidity on the surface of alumina. Two
conclusions seem to be universal: (1) that stronger Lewis
acid sites form as a result of fluorination and (2) that fluo-
rination leads to the formation of Brønsted acid sites on the
surface of partially fluorinated alumina.
by the chlorofluorocarbon CHClF2. The basic probe mole-
cule (phenyl dimethylphosphine) is used in conjunction with
31P NMR to quantify the number and types of acid sites.
We show that fluorination of the surface leads to the forma-
tion of Brønsted acid sites with the concurrent consumption
of Lewis acid sites and provide a structural model for the
change in the surface structure that is also able to rationalize
the observed changes in acidity.
2. Experimental
Previous studies of fluorinated aluminas have employed
fluorinating agents such as NH4F [9,11,12,14,17,18], HF
[23], F2 [21], and CHClF2 [10,13,22]. Our work has,
however, shown that the fluorination mechanism depends
strongly on the fluorinating agent employed [13]. Acidic
protons on the fluorinating agent, and the mode of oxygen
removal as either H2O or CO, for example, may alter the sur-
face structure and acidic properties of the intermediates that
are formed. Changing the fluorinating agent may also affect
the surface sites that are attacked first during the initial fluo-
rination reactions. This becomes particularly relevant when
halocarbons are used as fluorinating agents. For example,
CHClF2 and CHF3 would be expected to attack different
surface sites, CHClF2 chemisorbing as a carbocation and
CHF3 adding as a carbanion.
Studies investigating the acidic properties of fluorinated
aluminas have coupled the use of probe molecules with IR
spectroscopy, allowing the changes to acid properties to be
quantified. For example, Peri has demonstrated an increase
in the strength of the Lewis acid sites on aluminas fluori-
nated with NH4F [14]. In line with this, Scokart et al. have
applied IR methods and have concluded that strong Lewis
acid sites and stronger Brønsted acid sites form as a result
of fluorination of the oxide surface, in aluminas fluorinated
with aqueous solutions of NH4F [17].
A recent report by Boese et al. has investigated the ini-
tial stages of the fluorination of γ -Al2O3 with CHClF2 us-
ing XAES (X-ray excited Auger electron spectroscopy) and
XANES [13]. They demonstrated that the surface of the par-
ticles is first fluorinated with the formation of singly bound
fluorine species (F–Al). This is followed by destruction of
the oxide lattice and concurrent formation of strong Lewis
acid sites. Lastly, an AlF3 phase with a high defect concen-
tration is formed. Our previous work, which utilized in situ
X-ray diffraction, has also shown that further fluorination of
the oxide leads to the formation of an AlF3 phase that re-
sembles α-AlF3 but contains significant concentrations of
defects [10]. While these analytical methods are able to track
the specific structural changes that occur, they do not allow
these changes to be directly correlated with the changing na-
ture of the acid sites present on the surface.
2.1. Sample preparation
Aluminum-sec-tributoxide (99%) and chlorodifluoro-
methane (HCFC-22) were obtained from Aldrich. The alu-
minas were prepared as follows: Aluminum-sec-tributoxide
(ASB) was dissolved in 2-butanol (molar ratio 1:16). While
stirring vigorously, 3.5 eq of water per ASB molecule were
added dropwise at room temperature. The solution was
stirred vigorously for 1 h at room temperature and was then
aged for 24 h at 95 ◦C in a covered glass dish. The resulting
gel was subsequently dried at 95 ◦C for 24 h. The dried ma-
terial was calcined under nitrogen as follows: 10 ◦C/min to
400 ◦C, 400 ◦C hold for 30 min, followed by rapid cooling to
room temperature. This sample, referred to as untreated alu-
mina, is labeled U/Al2O3. A dehydrated sample, D/Al2O3,
was prepared by drying a sample under vacuum to 400 ◦C
at a rate of 2 ◦C/min and holding at 400 ◦C for an addi-
tional 4 h.
Surface fluorination of the samples was performed in a
controlled manner using a vacuum line for low fluorination
levels and a reactor tube for higher fluorination levels. All
samples were heated and dried under vacuum, from room
temperature to 400 ◦C at a rate of 2 ◦C/min. 100 Torr of
CHClF2 was added to a 162-mL volume vacuum line to
which a tube containing a shallow bed of approximately
100 mg of alumina (described above) was attached. After the
sample was heated for 2 h at 400 ◦C, the sample was evac-
uated for 30 min. This procedure was repeated once for the
sample labeled F1/Al2O3 and twice for sample F2/Al2O3.
Sample F1/Al2O3 was heated an additional 2.5 h at 400 ◦C
so that both samples received the same overall heat treat-
ment. Sample F3/Al2O3 was prepared in a 1/4-in. Inconel
reactor, by flowing CHClF2 over the alumina sample at a
rate of 40 cc/min. The sample was initially heated to 280 ◦C
under a flow of dry N2. The flow of CHClF2 was then initi-
ated and continued for a total of 75 min. Gas flow was then
switched back to dry nitrogen and the sample was heated to
400 ◦C for an additional 2 h. Sample surface areas were de-
termined by N2 physisorption according to the BET method
using a Micromeritics ASAP 2010 volumetric sorption ana-
lyzer, and the results are given in Table 1. A Leo 1550 field
emission gun scanning electron microscope was used to con-
duct EDS (energy-dispersive X-ray spectroscopy) analysis,
Here, we report a single- and double-resonance solid-
state NMR study of the structural changes that are associ-
ated with the fluorination of the surface of alumina xerogels

P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
71
Table 1
Sample composition and surface properties
Sample
Surface area
% F by mass
# fluorine per
surface Al
Lewis acid
sites
(mmol/m )
Lewis acid
sites
(#/surface Al)
Brønsted
acid sites
(mmol/m )
Brønsted
acid sites
(#/surface Al)
2
(m /g)
2
2
D/Al O
406
392
372
345
0
0
0.001253
0.000416
0.000147
0.000529
0.1258
0.0417
0.0148
0.0531
0
0
2
3
a
F1/Al O
6.8
10.7
20.8
0.91(1.82)
1.52(3.04)
3.19(6.38)
0.000505
0.001099
0.000333
0.0507
0.1104
0.0334
2
3
3
3
F2/Al O
2
F3/Al O
2
a
Calculated by assuming Al–F terminal groups and (Al–F–Al bridging F atoms).
to determine the fluorine and chlorine contents of the treated
samples.
as described above. The typical contact time was approxi-
mately 100 µs for MAS frequencies of 35 kHz. Repetition
times of 4 s were used.
1H-decoupled 31P MAS NMR experiments were per-
formed with a double-tuned 4.0-mm Chemagnetics probe
on a CMX-360 spectrometer at operating frequencies of
360.03 for 1H and 145.45 MHz for 31P. Chemical shifts are
referenced to tetramethylsilane and 85% H3PO4 as exter-
nal standards at 0.0 ppm. Zirconia rotors capable of spin-
ning 18 kHz were used for all experiments. Samples were
moved directly from a dry atmosphere glove box to the NMR
probe, where dry nitrogen was used for the purge and spin-
ning gases. 19F/31P REDOR (Rotational Echo in DOuble
Resonance) experiments were performed using a double-
tuned 3.2-mm Chemagnetics probe. REDOR experiments
were performed at −120 ◦C at a spinning rate of 20 kHz,
by applying π pulses to the 31P channel using XY-8 phase
cycling [26].
2.2. Phosphine loading
Dimethyl phenylphosphine (Aldrich) was used as the
probe molecule for the study of the acidity of the fluorinated
alumina samples. The samples were handled under dry N2
for all manipulations performed after the fluorination proce-
dure. Phosphine loadings were made directly on the samples
in the glove box by using a µL syringe with 0.05-µL ac-
curacy. The loading level of phosphine was approximately
0.0018 mmol/m2 for each sample. The samples were im-
mediately sealed in a 4-mm rotor with specially designed
airtight spacers. The samples were allowed to equilibrate for
1 h before spectra were acquired.
2.3. NMR
19F MAS NMR experiments were performed with a
custom-built 2.0-mm probe, capable of spinning to speeds of
50 kHz, on a Chemagnetics (8.45 T) spectrometer. Fluorine
contents of the samples were determined by 19F spin count-
ing using a known mass of aluminum fluoride trihydrate,
AlF3·3H2O, as a standard (Table 1). The Hartmann–Hahn
condition for the 19F/27Al CP (cross polarization) MAS ex-
periments was determined with AlF3·3H2O and optimized
for the so-called fast spinning regime [24], using a radio fre-
quency (rf) field strength for 27Al of approximately 18 kHz.
A rotor synchronized contact time of 95 µs was used for
MAS frequencies of 42 kHz, with repetition times of 2 s.
19F/27Al TRAPDOR (TRAnsfer of Populations in DOuble
Resonance) [27] NMR experiments were performed using
an 27Al rf field strength of 90 kHz. Chemical shifts are
referenced to 1.0 M aqueous aluminum sulfate and hexaflu-
orobenzene as external standards at 0.0 and −163.0 ppm,
respectively.
3. Results
3.1. Surface areas
The method of fluorination was chosen so as to maintain
high surface areas of the fluorinated materials. As the fluo-
rine content increases from 0 to 21% by mass, the surface
area decreases from 406 to 345 m2/g (Table 1). Although
this is a measurable decrease in surface area, the overall sur-
face areas remain high, which makes the systems ideal for
study by NMR methods.
3.2. Determination of chlorine content
EDS analysis was performed on samples D/Al2O3, F1/
Al2O3, F2/Al2O3, and F3/Al2O3, to determine if chlorine
was present in the treated samples. No chlorine was ob-
served in the samples, while the fluorine contents obtained
were similar to those determined from the NMR experi-
ments reported in the next section. The fluorinated samples
used in this study were prepared under low-pressure condi-
tions and were heated under vacuum to remove any adsorbed
molecular species before examination by NMR. This may
result in the complete removal of HCl as a gas-phase by-
product of the fluorination. These results are in agreement
27Al MAS NMR and 19F/27Al HETCOR (HETeronuclear
CORrelation) NMR experiments were performed with a
double-tuned 2.5-mm Bruker probe, on a Bruker Avance 700
(17.1 T) spectrometer. Small 27Al flip angles were used
(< 15◦) to ensure uniform excitation of all spins. CP was
used to transfer magnetization from 19F to 27Al for the
19F/27Al HETCOR experiments. The Hartmann–Hahn con-
dition for the 19F/27Al HETCOR experiments was deter-
mined with anhydrous aluminum fluoride and optimized

72
P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
tween −170 and −220 ppm. As the fluorination level in-
creases, the chemical shifts of both groups of resonances
shift to more negative frequencies and the relative inten-
sity of the second group of resonances decreases. The group
centered between −110 and −165 ppm has chemical shifts
consistent with aluminum oxyfluoride environments (i.e.,
AlO6−xFx, 1 < x < 6) [27]. Chemical shifts of −180 to
−200 ppm have been observed for alkali metal aluminum
fluoride salts [28], the fluorine atoms bonded to a single
aluminum (F–Al species) showing chemical shifts in the
−200 ppm region. Chemical shifts of −180 to −200 ppm
have also been observed for tetrahedral aluminum fluoride
species in organic salts and as isolated species in zeolites
[29,30].
The surface fluorine concentration was determined by
spin counting and is given in Table 1 in units of weight
percent and fluorine atoms per surface aluminum atom. The
latter values were estimated by making use of the numbers
previously reported for the concentrations of tetrahedral and
octahedral Al atoms on the surface of nano-sized alumina
particles, of 2.8 and 4.0 atoms/nm2, respectively [8]. The
numbers in Table 1 were determined by assuming, first, that
each F atom is coordinated to a single Al atom (forming a
terminal Al–F species) and, second, that each F atom is coor-
dinated to two Al atoms in bridging Al–F–Al linkages. The
second value is given in parentheses in Table 1, these two
assumptions representing the two extreme situations for the
maximum and minimum numbers of F coordinated to the
surface Al atoms. At the highest fluorine loadings, a Al:F
ratio greater than 1:6 assuming Al–F–Al bridging groups is
consistent with the formation of AlF6 octahedral environ-
ments and the fluorination of more than the first layer of Al
surface atoms.
19
Fig. 1. F MAS NMR collected at a spinning speed of 42 kHz for sam-
ples (a) F1/Al O , (b) F2/Al O , and (c) F3/Al O . Guidelines are shown
2
3
2
3
2 3
19
at the F chemical shift positions of bulk anhydrous aluminum fluoride
at −172 ppm (dashed line) and at −200 ppm (solid line; terminal Al–F
groups).
with the results of a detailed XPS study by Boese et al., who
found no significant difference between surface fluorination
with CHClF2 versus CHF3 [13]. Although they found evi-
dence for very small amounts of chlorine, <1% (mole), in
their samples, fluorination was the dominant reaction, with
final F contents of between 5 and 14% (mole) being ob-
served [13]. In addition, in situ diffraction studies of the
fluorination of alumina with CHClF2 have shown the forma-
tion of aluminum fluoride phases [10]. Clearly, fluorination
of the surface is a primary result of treatment with CHClF2,
which is reasonable considering that the formation of alu-
minum fluoride is thermodynamically more favorable than
aluminum chloride [2].
The 27Al MAS NMR spectra of the samples U/Al2O3,
D/Al2O3, F1/Al2O3, F2/Al2O3, and F3/Al2O3 are shown
in Fig. 2a, 2b, 2c, 2d, and 2e, respectively. The spectra
each contain three major groups of resonances, centered
at 70, 35, and 10 ppm, which correspond to four-, five-,
and six-coordinate aluminum species, respectively. Signifi-
cant changes are apparent as the fluorination level increases.
The most apparent change is in the relative number of five-
coordinate aluminum sites, the number of these sites de-
creasing with fluorination level. This is best illustrated in
Fig. 2f, which shows the overlaid 27Al NMR spectra of sam-
ples D/Al2O3 and F1/Al2O3. The six-coordinate resonance
shifts to more negative values on fluorination. This, based
on our earlier work [32], indicates an increase in the num-
ber of octahedral aluminum oxyfluoride groups, AlO6−xFx
and in the value of x. Consistent with this, the resonance
at −16 ppm in Fig. 2e is assigned to an AlF6 group [32].
Thus, the major process that appears to occur during flu-
orination is the reaction of the coordinatively unsaturated
aluminum sites (i.e., five-coordinate sites) to form octahe-
dral aluminum oxy-fluoride species. The resonance due to
the four- coordinate Al atoms sharpens on fluorination and
3.3. 19F and 27Al MAS NMR
The 19F MAS NMR spectra of the fluorinated samples
F1/Al2O3, F2/Al2O3, and F3/Al2O3 are shown in Fig. 1a,
1b, and 1c, respectively. The application of very fast spin-
ning MAS (in excess of 40 kHz) at a field of 8.45 T al-
lows the full chemical shift range, which spans approxi-
mately 100 ppm, to be collected without any overlap of the
isotropic resonances with the sidebands of different reso-
nances. As the fluorination level increases from F1/Al2O3
to F3/Al2O3, the chemical shifts and relative abundance
of the different sites change with a clear trend. For the
lowest level fluorinated sample, F1/Al2O3, two groups of
peaks are observed, one between −110 and −165 ppm and
the other, which consists of several overlapping peaks, be-

P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
73
19 27
Fig. 3. F/ Al CP MAS collected at a spinning speed of 42 kHz, for
19
sample F2/Al O . The spectra were collected with the F offset located at
2
3
19
(a) −110 ppm and (b) −220 ppm and a weak F rf field strength of 38 kHz.
method to determine whether the majority of fluorine chem-
ical shifts in the range of −180 to −200 ppm are associated
with octahedral, five-coordinate, or tetrahedral aluminum
species.
The 19F/27Al CP NMR spectra shown in Fig. 3a and 3b
were collected with fluorine offsets corresponding to 19F
shifts of −110 and −220 ppm, respectively. The only signal
observed in both CP spectra has a chemical shift of approx-
imately 0 ppm, which corresponds to octahedral aluminum
atoms. There is no indication that either group of 19F res-
onances are associated with five-coordinate or tetrahedral
aluminum species.
3.5. 19F/27Al TRAPDOR NMR
27
Fig. 2. Al MAS NMR collected at a spinning speed of 35 kHz, for samples
(a) U/Al O , (b) D/Al O , (c) F1/Al O , (d) F2/Al O , and (e) F3/Al O .
A 19F/27Al TRAPDOR NMR experiment was performed
on F1/Al2O3 (Fig. 4) to explore differences in the F–Al dipo-
lar couplings, and thus connectivities, of the different fluo-
rine sites. The experiment was performed using fast spin-
ning speeds to obtain the highest resolution possible; the
TRAPDOR experiment is still feasible under these condi-
tions due to the small quadrupole coupling constants of the
Al atoms [25]. It is clear from the difference spectra that all
the fluorine sites are located in close proximity (i.e., directly
bound) to aluminum atoms. The resonances between −180
and −220 ppm show a noticeably smaller TRAPDOR frac-
tion than the other resonances at higher frequencies. This is
consistent with their assignment to terminal fluorine species
bound to a single aluminum atom since they experience an
overall smaller F–Al dipolar coupling than fluorine bound to
two aluminum atoms in Al–F–Al bridges.
2
3
2
3
2
3
2
3
2 3
(f) Overlaid spectra of D/Al O (heavy line) and F2/Al O (thin line). The
2
3
2 3
guideline is shown at 10 ppm, the chemical shift of a six-coordinate AlO
6
species in γ -Al O ; the assignment of the resonances to different Al coor-
2
3
dination numbers are given in (a).
decreases slightly in intensity as some of these sites are also
consumed.
3.4. 19F/27Al CP NMR spectroscopy
The cross-polarization (CP) experiment relies on the po-
larization transfer of net magnetization on one set of nuclei
to others in close proximity. The large 19F chemical shift
range in the fluorinated samples can be exploited to examine
the type of aluminum species associated with the different
groups of 19F resonances, as the CP efficiency is very sen-
sitive to the offset frequency of the 19F radio frequency (rf)
field, the polarization transfer rate decreasing by an amount
v12/(v12 + δo2ff), as the offset frequency (δoff) of the applied
rf field strength (v1) is increased [31,32]. Here, we use this
3.6. 19F/27Al HETCOR NMR spectroscopy
The 19F/27Al HETCOR spectrum for sample F2/Al2O3
is shown in Fig. 5, and provides a second method for iden-

74
P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
extracted. As the 19F chemical shift moves from −120 to
−170 ppm, the chemical shifts of the bonded 27Al species
move from 0 to −10 ppm. This shift correlates to an in-
creasing fluorination level (i.e., x increasing in AlO6−xFx),
again consistent with our earlier work [32]. The 19F sig-
nal at −200 ppm (not shown) overlaps with the sideband
of the resonance centered at −150 ppm, and it is therefore
impossible to say with certainty which aluminum species it
is connected to.
3.7. 27Al MQ-MAS NMR spectroscopy
The 27Al MQ-MAS spectra collected at 16.3 T, for the
D/Al2O3 and all the fluorinated samples, are shown in Fig. 6.
There are two major clusters of resonances. The isotropic
chemical shift of the four-coordinate tetrahedral Al site is
centered at 70 ppm, while that for the six-coordinate oc-
tahedral site is centered near 10 ppm, in the anisotropic
dimension. As the fluorination process proceeds, an upfield
shoulder of the six-coordinate peak, due to the oxyfluoride
sites, emerges. These sites are particularly noticeable in 6c
and the AlF63− resonance is clearly seen in 6d. Two different
four-coordinate sites are also resolved in the spectra shown
in Fig. 6a, 6b, and 6c, at approximately 70 and 75 ppm (in
the anisotropic dimension of the MQ-MAS spectra). The
75 ppm resonance is no longer present in the MQ-MAS
spectrum of the most highly fluorinated sample and only a
single, broad four-coordinate site is visible at lower frequen-
cies. The four-coordinate sites most affected by fluorination
are likely either localized near the surface and, therefore,
consumed first, or are connected to a five-coordinate site
and thus could be indirectly perturbed by the fluorination of
that species. The five-coordinate sites, just visible above the
noise in Fig. 6a, are consumed during fluorination. This is
seen more clearly in higher signal-to-noise MQ-MAS spec-
tra acquired with a larger volume 4-mm probe.
19 27
Fig. 4. F/ Al TRAPDOR NMR of sample F3/Al O ; (a) control,
2
3
(b) double resonance, and (c) difference, obtained by subtracting the spec-
trum shown in (b) from that in (a). A spinning speed of 40 kHz was used
with 8 rotor periods of irradiation on the Al channel.
27
3.8. 31P MAS NMR
The variable temperature 1H decoupled 31P MAS NMR
spectra of the alumina and fluorinated alumina samples
loaded with dimethyl phenylphosphine (DMPP) are shown
in Fig. 7. Three discrete sites are observed for the spec-
tra acquired at −150 ◦C, which are assigned to physisorbed
phosphine molecules (approximately −45 ppm), phosphines
bound to Lewis acid sites (i.e., coordinatively-unsaturated
aluminum; −35 ppm) and protonated phosphine ions bound
directly to the conjugate bases of the Brønsted acid (surface)
sites (approximately 0 ppm) [23,27]. At room temperature,
molecular exchange of the phosphine molecules occurs, par-
ticularly between the Lewis acid site-bound phosphines and
the physisorbed molecules, but this exchange can be frozen
out by cooling to −150 ◦C. This exchange results in inac-
curate measurements of the 31P chemical shift and relative
number of sites at room temperature and the concentrations
19 27
Fig. 5. F/ Al HETCOR NMR two-dimensional spectra for sample
27
Al O /F2, collected at a spinning speed of 35 kHz. The Al irradiation
offset was placed at a shift of 30 ppm.
2
3
tifying the Al atoms in close proximity to fluorine. A short
contact time ensures that the major peaks in the spectrum
correspond to directly bonded Al–F species. The utility of
this type of experiment is apparent when a comparison is
made between the 27Al dimension and the 27Al one-pulse
spectrum of the same sample shown in Fig. 2. The sig-
nal from aluminum atoms in close proximity (i.e., directly
bonded) to fluorine is selected in the 27Al dimension. The
projection of the 27Al dimension is rather featureless, indi-
cating a distribution of sites. The resonances due to these
sites are more dispersed in the two-dimensional spectra, and
the 19F chemical shift due to each site can now be directly

P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
75
27
Fig. 6. Al MQ-MAS NMR collected at a spinning speed of 35 kHz on a 16.3 T spectrometer, for samples (a) D/Al O , (b) F1/Al O , (c) F2/Al O , and
2
3
2
3
2 3
3−
(d) F3/Al O . The assignments of the resonances to different Al coordination numbers are shown in (a) and the AlF
6
species is marked in (d).
2
3
of the different sites given in Table 1 were obtained from the
low-temperature NMR data.
shifts is apparent on sorbing the phosphine (Fig. 1a), the two
major groups of resonances shifting to lower frequencies to
−120 to −160 and −160 to −200 ppm. The larger change
in chemical shift seen for the terminal Al–F groups (−160
to −200 ppm) is reasonable since these species might be
expected to interact more strongly with the phosphine mole-
cules than the more buried Al–F–Al species. This change in
chemical shift has also been observed for samples that have
been allowed to hydrate and might explain why samples flu-
orinated under hydrous environments by other workers have
shown very different 19F NMR spectra [11,12].
Differences are apparent between the untreated alumina
and fluorinated samples. The untreated alumina does not
contain any strong Brønsted acid sites, as no resonance is
observed in the 0 ppm region. As the surface fluorination
proceeds, the concentration of Lewis acid sites decreases
from the level originally observed on the untreated sample.
In contrast, the concentration of Brønsted acid sites grows,
from F1/Al2O3 to F2/Al2O3, but drops noticeably for the
highest F content sample.
Two discrete groups of fluorine resonances from fluorine
atoms near phosphorus, one centered at −135 ppm and the
other at −180 ppm, are seen in the difference spectrum. The
relative concentration of the terminal F atoms (−180 ppm) is
increased in the difference spectrum relative to the control,
again consistent with the presence of these groups on the
surface. Sample F1/Al2O3 contains two types of acid sites,
3.9. 19F/31P REDOR NMR
We have used the REDOR NMR experiment (Fig. 8) to
probe the proximity of fluorine on the catalyst surface to the
phosphorus atom on the adsorbed phosphine for F1/Al2O3
loaded with DMPP. A noticeable change in the 19F chemical

76
P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
suggesting that these two F sites may both be associated with
the Brønsted acid site.
4. Discussion
Sol–gel-derived aluminas were used in this study for sev-
eral reasons. First, the synthesis of high surface area mate-
rials is relatively straightforward with alkoxide precursors.
High surface areas (>300 m2/g) are clearly advantageous in
the application of NMR, an inherently insensitive technique,
to the study of surface structure. Second, these aluminas con-
tain few chemical impurities relative to commercial alumina
catalyst supports. Since we are studying subtle structural fea-
tures on the surface of the aluminas, we wanted to minimize
the potential of contaminate species.
The phase stability of aluminas prepared using the alu-
minum alkoxide route has been examined by Wang et al.,
as a function of temperature [33]. Their work has shown
that samples calcined at 400 ◦C predominately contain the
γ -Al2O3 phase [33]. The 27Al MAS NMR spectrum of
D/Al2O3 shows an intensity ratio of approximately 1 to 2
for the tetrahedral to octahedral aluminum sites, in line with
the ratios expected for γ -Al2O3 [10]. All our samples were
dehydrated to 400 ◦C prior to fluorination; thus, the obser-
vation of a phase with regions that are locally similar to
γ -Al2O3 is not too surprising. Note that the presence of
four-coordinate aluminum species is strong evidence for the
presence of a spinel-related phase (i.e., a γ or γ ꢀ alumina):
All the other transition aluminas such as boehmite contain
six-coordinated aluminum sites only. The small differences
between the 27Al MAS NMR of U/Al2O3 and D/Al2O3 pri-
marily arise from the loss of some of the residual hydroxyl
groups following dehydration under vacuum, which will re-
sult in an increase in the number of coordinatively unsatu-
rated aluminum atoms on the surface. The presence of high
concentrations of five-coordinate aluminum is a direct con-
sequence of the increased number of defects caused by the
high surface area of the materials used in this study and from
the disorder that results from the synthesis method used to
prepare the materials.
1
31
Fig. 7. Variable temperature H decoupled P NMR spectra of dimethyl-
phenylphosphine-loaded (a) D/Al O , (b) F1/Al O , (c) F2/Al O , and
2
3
2
3
2 3
31
(d) F3/Al O . P resonances corresponding to coordination to the Brøn-
sted (B) and Lewis (L) and the physisorbed molecules are marked in the
low-temperature spectrum of (b).
2
3
There have been many detailed studies of fluorinated alu-
minas, which have used different fluorinating agents and/or
conditions; thus, the local structures are unlikely to be iden-
tical for all procedures. In particular, the mechanism of flu-
orination of the surface of aluminas is dependent in part on
the fluorinating agent and conditions used [2–14]. This is
particularly evident when we consider the use of NH4F as
a fluorinating agent. Here, several reports indicate NH4AlF4
phase formation [9], which obviously has no role in the flu-
orination mechanism when halocarbons are used. Thus, in
the subsequent discussion of the mechanism of fluorination,
we pay particular attention to those studies using halocar-
bons.
31 19
Fig. 8.
P/ F REDOR NMR of sample F2/Al O loaded with
2 3
dimethylphenylphosphine; (a) control, (b) double resonance, and (c) dif-
ference (×2). A 34 rotor period dephasing time was used with XY-8 phase
cycling.
which might be expected to be located near different fluorine
species. However, Brønsted acid sites are created on fluo-
rination, while the Lewis acid sites are partially destroyed,
The very fast spinning 19F MAS NMR spectra clearly
indicate the presence of two major groups of resonances.

P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
77
The resonances in the range from −110 to −165 ppm are
consistent with octahedral aluminum oxyfluoride environ-
ments [32]. The observation of these resonances and their
assignments are consistent with other NMR studies. An ad-
ditional group of resonances between −180 and −220 ppm
has not previously been observed on fluorinated aluminas.
We assign the majority of these peaks to terminal fluorine
species (i.e., fluorine bound to a single aluminum atom)
bonded to an octahedral aluminum atom. The chemical shift
range of these species, however, is consistent with both tetra-
hedral and octahedral aluminum fluoride species, with termi-
nal F–Al linkages. Terminal AlIV–F species are not consis-
tent with the 19F/27Al CP results. The presence of terminal
fluorine species has been clearly observed in fluorination
procedures using CHClF2 by Boese et al. using XAES and
XANES [13], but these methods were not able to distin-
guish between four- and six-coordinate species. There are
several potential explanations as to why these types of ter-
minal groups have not been previously observed with NMR
spectroscopy. Most earlier NMR studies have utilized NH4F
in aqueous solutions to fluorinate the surface and the pres-
ence of moisture and the use of fluoride salts may not lead to
stable terminal F–Al species. Additionally, overlap with the
spinning sidebands of the more intense resonance at lower
frequencies may have concealed the resonances of the ter-
minal fluorine species in these earlier studies.
Fig. 9. Diagram of proposed reaction pathway for the first states of fluori-
nation.
The 27Al MAS and MQMAS NMR spectra provide in-
sight as to which aluminum species are consumed dur-
ing the initial fluorination reaction. It is apparent from the
27Al MAS NMR spectra that the number of five-coordinate
Al decreases during the fluorination process. As these five-
coordinate and (higher frequency) four-coordinate species
(seen by MQMAS) are consumed, octahedral aluminum
oxyfluoride species are formed. The preference for the at-
tack of the five-coordinate sites during the initial stages
of fluorination is most likely thermodynamic in nature.
Five-coordinate aluminum species are predominately ob-
served in aluminas that have been synthesized through
low-temperature procedures, which are capable of trap-
ping metastable five-coordinate Al species. Since the five-
coordinate sites are consumed during the first stages of fluo-
rination, this suggests that these species are localized on the
surface of the nano-sized (∼5 nm) particles. The 27Al NMR
of the dehydrated as-prepared material (D-Al2O3) suggests
that the γ -Al2O3 structure, which is composed of a cubic
closed packed array of oxygen atoms with aluminum occu-
pying 8/9 of the octahedral sites and 4/9 of the tetrahedral
sites, is a reasonable model of the structure of the particles.
The five-coordinate aluminum atoms are most likely located
near the surface of such small particles due to surface re-
laxation and termination effects and are thus more readily
fluorinated.
ous studies have utilized phosphines as probe molecules,
though several have not paid particular attention to the pos-
sibility of site exchange at ambient temperatures. In these
systems, the motion is not frozen out of the NMR time scale
until approximately −80 ◦C, allowing accurate concentra-
tions of the number of acid sites to be determined. From
the quantification of the −150 ◦C, 31P MAS NMR spectra
it is clear that the number of Lewis acid sites decreases on
fluorination (Table 1). This is consistent with the increase in
the average coordination number of aluminum on the surface
as fluorine replaces oxygen. The measured Lewis acidity in-
creases slightly from D/Al2O3 to F3/Al2O3 as evidenced by
the slight downfield shift in 31P chemical shift from −39
to −35 ppm. This may be due to one of several reasons:
first, nucleation of a new aluminum fluoride/hydroxyfluoride
phase or reaction with a tetrahedral Al atom may open up ad-
ditional coordinatively unsaturated aluminum centers. Sec-
ond, fluorination of nearby Al may increase acidity due to
inductive effects. The 31P MAS NMR also clearly indi-
cated the formation of Brønsted acid sites from D/Al2O to
F2/Al2O3, but these sites are consumed as the fluorine to
surface aluminum ratio approaches 3.0 in sample F3/Al2O3.
Based on the results of this NMR study, we can propose a
mechanism for the first stages of fluorination of the surface
of alumina with CHClF2 (Fig. 9). The first step involves the
chemisorption of a CHClF2 molecule to the surface. Since
the relative bond strengths are C–Cl < C–H < C–F, this
is likely to occur via C–Cl bond breaking and carbocation
formation [34]. The terminal µ1-OH (Al–OH) groups are
stronger nucleophiles than µ2-OH (Al–O(H)–Al) groups,
The variable temperature 31P NMR of the adsorbed phos-
phines clearly indicates the importance of making quanti-
tative measurements at low temperatures to freeze out any
molecular exchange processes on the surface. Many previ-

78
P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
for example, and can attack the CHClF2 molecules forming
bound methoxy species, F2HC–O–Al (step 1 in the scheme
shown in Fig. 9). The released chloride ions will combine
with the proton of the µ1-OH group, forming HCl, which
will be removed as a gaseous by-product. The formation of
these types of methoxy species formed through C–Cl bond
cleavage has been directly observed for the adsorption of
CH3Cl to alumina, where the presence of hydroxyl groups
was found to be essential for the disassociative chemisorp-
tion of chloroalkanes [35].
Following chemisorption and formation of the methoxy
species, the subsequent steps most likely occur through in-
termediates with only a transient existence. The fluorine and
hydrogen atoms from the methoxy (–O–CHF2) group mole-
cule must clearly react with the surface, as HF will not be
desorbed from the surface at the temperature of the fluorina-
tion reactions. Five-coordinate, unsaturated sites nearby the
methoxy group will be readily fluorinated to form octahedral
environments (step 2), leading to an increase in the average
coordination environment on the surface, and the conversion
of five-coordinate to octahedral environments.
minum hydroxyfluoride environments [9] and Fischer et al.
have used 19F NMR to distinguish different types of surface
Al sites, based on their fluoride substitution levels [11]. In
both studies there was no evidence for terminal Al–F groups,
suggesting that the use of aqueous methods can modify the
fluorination pathway, and either do not lead to the forma-
tion of the more labile terminal Al–F species, or if they
are formed, they then react further under these conditions to
form Al–OH or Al–F–Al groups. DeCanio et al. have shown
that high levels of NH4F also lead to the dissolution of alu-
minum species and the formation of NH4AlF4 salts on the
surface, which decompose into β-AlF3 [9]. This is in marked
contrast to the gas-phase reaction of halocarbons with the
surface of aluminas, where terminal Al–F groups are clearly
observed. Further evidence for the formation of terminal
Al–F species in nonaqueous fluorinations was provided by
Krahl et al., who have used halocarbons in the fluorination
of AlCl3 and have identified terminal Al–F species in the
products [36]. Further fluorination with HCF2Cl leads to the
condensation of the terminal groups, and eventually the for-
mation of α-AlF3, rather than β-AlF3, albeit with a high
concentration of defects [10].
The carbon from the originally chemisorbed CHClF2
molecule is removed from the surface as a carbon monox-
ide molecule (step 3), the oxygen coming from the initial
methoxy bridge. Desorption of CO versus CO2 was observed
in our earlier GC studies [10] and is consistent with the flu-
orination agent used in this study. CHClF2 would not be
expected to undergo any oxidation or reduction reactions
following reaction with the alumina surface and must, there-
fore, be desorbed as CO. CO2 can, however, be formed as the
product of the disproportionation of CO. The desorption of
the CO molecule must occur with the simultaneous attack of
the Al atom by fluorine so that no unstable three-coordinate
aluminum site are formed. The formation of a terminal F–Al
group is consistent with the work of Boese et al. who inves-
tigated fluorination with the same halocarbon, CHClF2 [13].
Charge balance is maintained, following CO loss, by the
protonation of a bridging oxygen species to a form Brønsted
acid sites. This process was followed in the 31P NMR spec-
tra, the untreated alumina showing no evidence of the Brøn-
sted acidity increasing steadily on fluorination as the Lewis
acidity decreased and the lower coordinate sites were con-
sumed. Reaction of a terminal Al–F group with the nearby
Al atom results in the formation of a Al–F–Al linkage, a new
µ1-OH group, and the consumption of the Brønsted acid
site. The new µ1-OH group can then react further with an-
other CHClF2 molecule. Once all the unsaturated sites have
been fluorinated, the reaction must then proceed by cleavage
of Al–O–Al bonds connecting nearby surface sites (reac-
tions 5 and 6) or an Al–O bond of the µ1-OH group. These
reactions no longer produce Brønsted acid sites.
5. Conclusion
The combination of single- and double-resonance solid-
state NMR and the use of basic probe molecules have been
used to examine the mechanism of fluorination of alumina
using CHClF2 as the fluorinating agent. Our study clearly
supports a mechanism that occurs in the following steps:
(1) the CHClF2 molecule reacts with a µ1-OH group with
the formation of HCl,
(2) the five-coordinate aluminum sites and Lewis acid sites
are consumed, resulting in the formation of Brønsted
acid sites and terminal Al–F groups,
(3) condensation of Al–F groups, which results in Al–F–Al
linkages, the consumption of the Brønsted acid sites, and
the formation of additional µ1-OH groups, and
(4) the reaction of the newly formed µ1-OH groups with
more CHClF2 molecules and a repeat of steps (2)–(3).
Once the unsaturated Al sites are fluorinated, no new Brøn-
sted acid sites can be created, but fluorination can still con-
tinue by cleaving Al–O–Al bonds. The model is consistent
with the 27Al and 19F NMR studies of the changing sur-
face structure and the 31P NMR spectra of the sorbed basic
phosphines. Whereas no Brønsted acidity is observed in the
untreated alumina, strong Brønsted acid sites form during
the initial stages of fluorination, which then disappear as the
surface is fully fluorinated and the five-coordinated Al sites
are consumed.
It is useful to consider the differences between fluorina-
tion with halocarbons and aqueous methods. Several work-
ers have investigated the structural changes associated with
aqueous impregnation of aluminas with NH4F. For example,
DeCanio et al. have used 27Al NMR to identify different alu-
Controlling the acid properties of alumina requires the de-
velopment of mechanisms for the modification of its surface.
The fluorination mechanism proposed in this work repre-

P.J. Chupas, C.P. Grey / Journal of Catalysis 224 (2004) 69–79
79
sents the first step toward the rational control of the acidity
of fluorinated aluminas. An understanding of and ability to
control acidity may eventually allow acid properties to be
tailored to catalyze specific reactions.
[13] O. Boese, W.E.S. Unger, E. Kemnitz, S.L.M. Schroeder, Phys. Chem.
Chem. Phys. 4 (2002) 2824.
[14] J.B. Peri, J. Phys. Chem. 72 (1968) 2917.
[15] Q. Dai, G.N. Robinson, A. Freedman, J. Phys. Chem. B 101 (1997)
4940.
[16] S.S. Deshmukh, V.I. Kovalchuk, V.Y. Borovkov, J.L. d’Itri, J. Phys.
Chem. B 104 (2000) 1277.
[17] P.O. Scokart, S.A. Selim, J.P. Damon, P.G. Rouxhet, J. Colloid Inter-
face Sci. 70 (1979) 209.
Acknowledgments
[18] E.R.A. Matulewicz, F.P.J.M. Kerkoff, J.A. Moulin, H.J. Reitsma,
J. Colloid Interface Sci. 77 (1980) 110.
[19] A. Hess, E. Kemnitz, A. Lippitz, W.E.S. Unge, D.H. Menz, J. Ca-
tal. 148 (1994) 270.
[20] A. Hess, E. Kemnitz, J. Catal. 149 (1994) 449.
[21] J.M. Saniger, N.A. Sanchez, J.O. Flores, J. Fluorine Chem. 88 (1998)
117.
Martine Ziliox is thanked for her assistance in collection
of the high-field NMR data. Financial support from the Basic
Energy Sciences program of the Department of Energy is
gratefully acknowledged (DEFG0296ER14681).
[22] R.I. Hedge, M.A. Barteau, J. Catal. 120 (1989) 387.
[23] J.H. Lunsford, W.P. Rothwell, W. Shen, J. Am. Chem. Soc. 107 (1985)
1540.
References
[1] L.E. Manzer, Science 249 (1990) 31;
[24] A.J. Vega, Solid State Nucl. Magn. Res. 1 (1992) 17.
[25] C.P. Grey, A.J. Vega, J. Am. Chem. Soc. 117 (1995) 8232.
[26] T. Gullion, D.B. Baker, M.S. Conradi, J. Magn. Res. 89 (1990) 479.
[27] P.J. Chupas, D.R. Corbin, V.N.M. Rao, J.C. Hanson, C.P. Grey, J. Phys.
Chem. B 107 (2003) 8327.
L.E. Manzer, V.N.M. Rao, Adv. Catal. 39 (1993) 329.
[2] E. Kemnitz, D.-H. Menz, Prog. Solid State Chem. 26 (1998) 97.
[3] L.M. Rodriguez, J. Alcaraz, M. Hernandez, Y. Ben Taarit, M. Vrinat,
Appl. Catal. A 169 (1998) 15.
[4] L.M. Rodriguez, J. Alcaraz, M. Hernandez, M. Dufaux, Y. Ben Taarit,
M. Vrinat, Appl. Catal. A 189 (1998) 53.
[28] L.S. Du, A. Samoson, T. Tuherm, C.P. Grey, Chem. Mater. 12 (2000)
3611.
[5] Z. Cheng, V. Ponec, J. Catal. 148 (1994) 607.
[29] C.P. Grey, D.R. Corbin, J. Phys. Chem. 99 (1995) 16821.
[30] Y. Xiao, M.S. thesis, SUNY-Stony Brook, 2001.
[31] K.H. Lim, C.P. Grey, Chem. Commun. 20 (1998) 2257.
[32] K.H. Lim, F. Jousse, S.M. Auerbach, C.P. Grey, J. Phys. Chem. B 105
[6] R. Covini, V. Fattore, N. Giordano, J. Catal. 7 (1967) 126.
[7] G.B. McVicker, C.J. Kim, J.J. Eggert, J. Catal. 80 (1983) 315.
[8] H. Knözinger, P. Ratnasamy, Catal. Rev.-Sci. Eng. 17 (1978) 31.
[9] E.C. DeCanio, J.W. Bruno, V.P. Nero, J.C. Edwards, J. Catal. 140
(1993) 84.
[10] P.J. Chupas, M.F. Ciraolo, J.C. Hanson, C.P. Grey, J. Am. Chem.
Soc. 123 (2001) 1694.
[11] L. Fischer, V. Harle, S. Kastelan, J.B. d’Espinose de la Caillerie, Solid
State NMR 16 (2000) 85.
(2001) 9918.
[33] J.A. Wang, X. Bokhimi, A. Morales, O. Novaro, T. Lopez, R. Gomez,
J. Phys. Chem. B 103 (1999) 299.
[34] D.F. McMillan, D.M. Golden, Ann. Rev. Phys. Chem. 33 (1982) 493.
[35] T.P. Beebe, J.E. Crowell, J.T. Yates, J. Phys. Chem. 92 (1988) 1296.
[36] T. Krahl, R. Strosser, E. Kemnitz, G. Scholz, M. Feist, G. Silly, J.-Y.
Buzare, Inorg. Chem. 42 (2003) 6474.
[12] W. Zhang, M. Sun, R. Prins, J. Phys. Chem. B 106 (2002) 11805.