Adsorption properties of various forms of aluminium trifluoride investigated by PulseTA

Article abstract of DOI:10.1016/j.tca.2009.10.008

The current study employs the Pulse Thermal Analysis (PTA) method to investigate the adsorption of gaseous methanol onto various forms of aluminium trifluoride drawing special attention to the strongly adsorbing β-AlF₃ and the recently describe

Full text of DOI:10.1016/j.tca.2009.10.008

Thermochimica Acta 498 (2010) 100–105
Contents lists available at ScienceDirect
Thermochimica Acta
journal homepage: www.elsevier.com/locate/tca
Adsorption properties of various forms of aluminium trifluoride investigated by
PulseTA®
M. Feist, R. König, S. Bäßler, E. Kemnitz∗
Institute of Chemistry, Humboldt University, Brook-Taylor-Strasse 2, D-12489 Berlin, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 6 August 2009
Received in revised form 2 October 2009
Accepted 11 October 2009
Available online 20 October 2009
The current study employs the Pulse Thermal Analysis (PTA) method to investigate the adsorption of
gaseous methanol onto various forms of aluminium trifluoride drawing special attention to the strongly
adsorbing ␤-AlF3 and the recently described high-surface (HS) AlF3. Simultaneously or subsequently
occurring physisorption and chemisorption can be clearly distinguished and partly quantified. BET surface
and NH3-TPD data will be discussed together with the surface coverage by methanol determined by TA.
The extraordinary sorptive properties of HS-AlF3 are due to acidic sites at the remarkably large surface. The
crystalline modifications ␩-AlF3, ␪-AlF3, and ␬-AlF3 are less-important as they poorly adsorb exhibiting
thermogravimetric curves without clearly expressed mass changes. Their surface loading is comparably
poor as found for the stable modification ␣-AlF3.
Keywords:
Aluminium fluorides
Methanol adsorption
Pulse Thermal Analysis
© 2009 Elsevier B.V. All rights reserved.
1. Introduction
the strength of the Lewis acidic centres of HS-AlF3 and related
phases, including pyridine photoacoustic IR [12,13], NH3-TPD
During the last two decades, numerous studies were devoted
to aluminium fluoride in various crystalline and amorphous mod-
ifications as to aluminium-based hydroxyfluorides. The scientific
interest is primarily explained by the attempts to improve and
to understand the catalytic properties of these substances, but by
special applications as well, e.g. such as thin coatings. The stud-
ies led to the design of novel preparation routes, e.g. a sol–gel
process yielding nano-sized, mesoporous high-surface aluminium
fluoride (HS-AlF3) [1,2], microwave-assisted hydrothermal synthe-
sis [3,4], or mechanochemical activation by high-energy ball milling
[5,6]. Meanwhile, magnesium fluoride, being analogously sol–gel
prepared as HS-AlF3, can be deposited on glassy surfaces forming
stable transparent thin films exhibiting extraordinary optical and
protective properties [7].
HS-AlF3 and related phases were the subjects of a series of struc-
tural, chemical, and computational studies [8–10] which allowed
to conclude that the enormous Lewis acidity being comparable to
that of SbF5 [11] is responsible for the extraordinary properties.
Exemplarily, HS-AlF3 catalyzes the isomerisation of 1,2-
dibromohexafluoropropane to 2,2-dibromohexafluoropropane,
which is catalyzed only by the strongest Lewis acids known [12]. So
far, several techniques were applied to characterize and compare
[1,13] or IR spectroscopy of adsorbed CO molecules [8].
The present study will contribute to the knowledge of the sorp-
tion properties of the mentioned aluminium fluoride phases by
applying a promising tool in adsorption studies, i.e. the Pulse Ther-
mal Analysis® [14–16].
2. Experimental
2.1. Synthesis
Seven different modifications and specially prepared forms of
AlF3 have been synthesized (Table 1). The substances were pure and
did not contain amorphous parts as proved by their X-ray powder
diffractograms combined with a detailed spectroscopic study of the
phases by employing 27Al and 19F solid state MAS NMR, XPS and IR
spectroscopy. These results will be reported elsewhere [17].
␣-AlF3 was used as commercially available (Aldrich) or pre-
pared by thermal decomposition of (NH4)3[AlF6] in a nitrogen flow
at 700 ◦C. A typical experiment started from 3 g (NH4)3[AlF6]; (3 h
tempering, cooling down in nitrogen; XRD: PDF 44-231) [18].
␤-AlF3 was obtained by thermal decomposition of ␣-AlF3·3H2O
[(PDF: 43-436; to be carefully distinguished from ␤-AlF3·3H2O
(PDF: 35-872)]. A first reaction step consists in heating the educt in a
Schlenk tube in vacuum at 200 ◦C. The second step is re-heating the
intermediate tightly packed in an aluminium foil (“self-generated
atmosphere”) in a tube oven at the same 200 ◦C in a nitrogen flow,
followed by a tempering step at 400 ◦C (1 h). The sample is cooled
down under nitrogen [18,19].
∗
Corresponding author. Tel.: +49 30 2093 7555; fax: +49 30 2093 7277.
E-mail addresses: feistm@chemie.hu-berlin.de (M. Feist),
koenigre@chemie.hu-berlin.de (R. König), sigrid.baessler@chemie.hu-berlin.de
(S. Bäßler), erhard.kemnitz@chemie.hu-berlin.de (E. Kemnitz).
0040-6031/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.tca.2009.10.008

M. Feist et al. / Thermochimica Acta 498 (2010) 100–105
101
Table 1
Structural and surfacial characterization for various forms of AlF3.
a
b
c
Structure type
Ref.
Structural features
SBET [m2/g]
msample [mg]
nMeOH
nNH
3
[␮mol/g]
[mol%]
[␮mol/m2]
[␮mol/g]
[␮mol/m2]
␣-AlF3
␤-AlF3
␤-AlF3 gel
␩-AlF3
␬-AlF3
VF3 (LT)
HTB
HTB
Pyrochlore
TTB
Own type
Amorphous
[27]
[19]
[22]
[18,21]
[21]
3D network
3
31
43
2
19
64
250
31.48
43.81
14.04
36.68
17.64
26.44
12.80
0
107
221
8
91
0
0
0
–
380
130
160
–
–
Hexagonal channels
Hexagonal channels
Channels
Tetra/pentagonal channels
3D network
0.9
1.8
0.07
0.76
0
3.5
5.1
4.0
4.8
0
12.2
3.0
80
–
3.8
3.6
␪-AlF3
HS-AlF3
[20,29]
[1,2]
240
900
Mesoporous
1090
7.7
4.0
a
BET surface [31].
b
c
Surface loading with methanol from PTA experiments related to mS.
NH3-TPD (cf. [28] and Section 2).
␪-AlF3 is formed in a two-step decomposition reaction in vac-
2.4. Temperature-programmed ammonia desorption (NH3-TPD)
uum at 450 ◦C starting from [N(CH3)4]AlF4·H2O [20]. A typical
experimental setup is: 2 g educt, open corundum crucible in a
Quartz tube, evacuation with a membrane pump, tempering for
1 h; cooling down at normal air (XRD: PDF 47-1659).
␩-AlF3 has been characterized first by Herron et al. [21]. A com-
parably simple synthesis route has been described by Krahl: X-ray
amorphous aluminium chloride fluoride (ACF) is thermally decom-
posed in a dynamic vacuum at 450 ◦C [18].
The samples were pretreated at 250–300 ◦C for 1 h. Afterwards,
ammonia was adsorbed onto the surface at 393 K. Ammonia des-
orption was monitored upon heating (10 K/min up to 773 K) by FTIR
detection of the band at 930 cm−1 (FTIR system 2000, PerkinElmer).
The total amount of ammonia desorbed was determined volumet-
rically (excess of H2SO4, back-titration with NaOH solution [28].
␬-AlF3 was prepared according to Herron et al. [21] following
the three-step synthesis route passing PyHAlF4 and ␤-NH4AlF4 as
isolable intermediates.
3. Methodology and results
3.1. The PulseTA® method
The gel form of ␤-AlF3 has been synthesized by supercritical
drying at 300 ◦C in an autoclave starting from a {AlF3−x(OiPr)x/i-
PrOH} gel (cf. [1,2]) and MeOH. The autoclave was depressurized
at the same temperature. An extremely voluminous aerogel-like
material with a BET surface between 40 and 66 m2/g was obtained
[22].
The methanol used for the liquid injections was dried by com-
mon techniques, distilled, and stored over pre-dried molecular
sieve A4 under nitrogen.
The adsorption ability of solids can be studied by utilizing the
PulseTA® (PTA) technique. It is a promising tool in thermal analysis
[14–16] and has been successfully applied to the solid state chem-
istry of fluorides [23,24]. The PTA technique is normally used for the
quantitative interpretation of TA-MS or TA-FTIR curves based on a
preceding calibration of the ion current (IC) or IR signals [14,15].
The potential of PTA by quantitative signal evaluation is impress-
ing in the case of capillary-coupled TA-MS devices, but limited in
the case of skimmer systems which is due to its constructive design
[25,26]. A PTA apparatus can also be utilized in terms of a “catalytic
reactor” allowing for the injection of one or two gases onto a sample
subjected to a controlled heating program [23]. Low-boiling liquids
and even aqueous HF [24] can be injected as well (see Section 2).
A typical PTA experiment in the field of catalytically active oxides
or fluorides, which often contain surfacially adsorbed water and/or
OH groups, comprises the following sequence of steps performed
with one single sample:
2.2. Thermal analysis
The PulseThermalAnalysis® (PTA) experiments have been per-
formed using a NETZSCH thermoanalyzer STA 409 C Skimmer®
system, equipped with a BALZERS QMG 421, enabling on line-
coupled TA-MS measurements, and a commercial PTA box. The
self-made liquid injection unit comprises
a
heated (120 ◦C),
septum-tightened GC injector implemented in the heated stainless
steel tube of the gas supply system. It is placed between the PTA
box and the thermobalance. A microliter GC syringe is used. The
dead time between injection and detection amounted to 60–70 s
for the given flow conditions. The thermoanalytical curves (T, DTA,
TG and DTG) have been recorded together with the ionic current (IC)
curves in the multiple ion detection (MID) mode [14,15]. A DTA–TG
sample carrier system with platinum crucibles (baker, 0.8 ml) and
Pt/PtRh10 thermocouples was used. Samples of 25–40 mg each
were measured versus empty reference crucible. A constant purge
gas flow of 70 ml/min N2 5.0 (MESSER-GRIESHEIM), a constant
heating rate of 10 K/min or a isothermal regime with Tiso = 46–48 ◦C
was applied. The raw data have been evaluated utilizing the manu-
facturer’s software PROTEUS® (v. 4.3) and QUADSTAR® 422 (v. 6.02)
without further data treatment, e.g. such as smoothing.
(1) Pretreatment by heating in the appropriate carrier gas (prefer-
ably inert such as nitrogen) up to the chosen adsorption
temperature to be studied.
(2) Cooling down to room temperature.
(3) Starting the measuring run (i.e. constantly heating or setting an
appropriate isothermal temperature level).
(4) Injection of gas (typically 250–500 ␮L) or liquid pulses (1–5 ␮L)
during heating or at the chosen isothermal plateau and record-
ing of the TA-MS curves with appropriate mass numbers in the
MID mode.
(5) Desorption (i.e. further heating or re-heating after cooling down
to 25 ◦C).
Typical experimental curves according to steps (1), (4), and (5) are
presented in Figs. 5–7, respectively.
2.3. Surface determination
A Micromeritics ASAP 2010 apparatus has been used to record
the N2 adsorption isotherms at 77 K. Prior to each measurement,
the samples were degassed at 4 × 10−3 Torr and 200 ◦C for 10 h.
The surface area was calculated according to the BET method.
3.2. Blank experiments for studying the methanol adsorption
In order to design a blank experiment that should precede a
PTA measurement one has to consider three aspects: (1) a possible

102
M. Feist et al. / Thermochimica Acta 498 (2010) 100–105
of 0.6 ␮g that practically disappears over the experiment time of
120 min. All DTA signals exhibit the characteristic form of a sharp
exothermal effect indicating physisorption that is immediately fol-
lowed by the broader endothermal effect being due to desorption
by the continuous gas flow of the carrier gas (see below).
Again, the IC peak area of the first pulse is smaller than that of the
following ones; their areas are almost constant within the exper-
imental error. The area variation reflects the volume error that is
greater for the manual injection of liquids using a ␮L-syringe than
for the injection of permanent gases by an injection loop contained
in the commercially available PTA box.
3.3. The sorption behaviour of ˇ-AlF3
Fig. 1. TA-MS curves for preparing a PulseTA® experiment. Four injections of 2 ␮L
liquid methanol in the recipient, isothermally held at 46 ◦C, without a sample are
represented by the IC signals for the mass number m32 (CH3OH+) together with the
integral intensities (in A s). Neither a DTA effect nor a detectable mass gain can be
detected.
3.3.1. Adsorption of methanol
The findings reported before enabled us to perform analogous
sorption experiments with ␤-AlF3, a phase expected to adsorb
much stronger than ␣-AlF3. A sample of 44.29 mg has been pre-
treated by heating in vacuum for 2 h at 250 ◦C which lead to a water
release of 0.48 mg (1.08%). The remaining solid (43.81 mg), which
was then cooled down under nitrogen, provided a “fresh” surface
ready to be loaded with the adsorbate.
As shown in Fig. 3, four pulses of 3 ␮L liquid methanol are suf-
ficient for saturating the surface of ␤-AlF3. Only the TG step for
the first pulse indicates chemisorption alone (70 ␮g). The following
three pulses show the overlapping of chemisorption by an increas-
ing portion of physisorption. The initial mass gain at each injection
peak, caused by the methanol sorption, is afterwards reduced due
to the release of physisorbed methanol into the carrier gas stream.
The fifth injection peak already (not monitored) affects no mass
gain any more.
The remaining net mass gain for the second to fourth steps (50,
20 and 10 ␮g) represents the chemisorbed part of methanol. With
the total mass gain of 150 ␮g for the entire experiment one calcu-
lates a surface loading of 0.107 mmol/g (0.9 mol%) CH3OH.
It is noteworthy that both the qualitative interpretation of the
TG curve shape presented here and the quantitative determination
of the surface loading exactly correspond to the procedure pro-
posed for ZSM-5 zeolites [16]. Moreover, not only the TG curve, but
the DTA traces as well give a good insight in details of the sorption
behaviour. This is especially remarkable as, usually, the sensitivity
of TG data is considered to be almost one order of magnitude better
than that of DTA information. However, a careful inspection of the
DTA traces illustrates as well fine details of the sorption processes,
e.g. the different behaviour during the first and the following pulses
as well as differences in the heat exchange. So the first pulse in Fig. 3,
representing chemisorption alone, causes only a sharp exothermal
DTA signal. If desorption of physisorbed molecules begins to occur
Fig. 2. TA-MS curves for a PulseTA® experiment on pre-heated (250 ◦C in N2) ␣-
AlF3 (31.48 mg) in nitrogen with the IC curve for the mass number m31 (CH3O+)
representing the methanol injections.
enthalpic effect of injecting pulses of gaseous methanol into the car-
rier gas stream must be known; (2) the evolution of the peak areas
of the IC curves to be followed (m/z = 311 or 32 in this case) must
be known for the case of injecting into the apparatus with empty
crucibles; (3) it is further important to know the behaviour of an
inactive compound under the chosen experimental conditions. For
the present study of various modifications of aluminium fluorides,
a well-crystalline ␣-AlF3 seemed to be the typical representative of
an inactive compound as it is known not to act as Lewis acid catalyst
[13].
3.2.1. Injection into an empty apparatus
As can be seen in Fig. 1, no enthalpic effect can be detected that
could be attributed to a mixing effect of the evaporated methanol
in the carrier gas.
The IC area of the first pulse is somewhat smaller than the fol-
lowing ones. As no mass gain corresponds to the pulse, obviously
a certain fraction of the injected methanol is adsorbed at the inner
surface of the measuring cell. This is not surprising as the recipient
has been evacuated during the measurement preparation.
3.2.2. Injections onto ˛-AlF3
Fig. 2 demonstrates that the original assumption concerning
the inactivity of ␣-AlF3 cannot be confirmed. It shows, in contrast,
weak methanol adsorption predominantly as physisorption. Only
the first injection pulse (2 ␮L each) causes a very small mass gain
Fig. 3. TA-MS curves for a PulseTA® experiment on pretreated (2 h, 250 ◦C; vacuum)
␤-AlF3 (43.81 mg, ꢀm = +150 ␮g) in nitrogen with the IC curve for the mass number
m32 (CH3OH+).
1
The labelling of mass numbers, e.g. m/z = 31, will be shortened in the following
as m31.

M. Feist et al. / Thermochimica Acta 498 (2010) 100–105
103
Fig. 4. DTA curves for the first four injection pulses of methanol onto ␣- and ␤-AlF3
and without a sample (from Figs. 1–3). The peak areas (␮V s) are standardized to
equal injection volumina.
Fig. 5. TA-MS curves for the desorption of methanol from ␤-AlF3 loaded in a preced-
ing PTA experiment (cf. Fig. 3) after cooling down under nitrogen. The mass numbers
m18 (H2O+), m31 (CH3O+), and m32 (CH3OH+) are shown (if more than one mass
number is depicted, the scaling for the IC given in the plot corresponds to the most
intensive mass number, i.e. mostly that for water, m18).
(cf. the TG curve shape), the endothermal desorption affects the
broad signal that follows the strong exothermal peak. As expected,
the peak area for chemisorption is considerably greater than for
physisorption which can be deduced from the decreasing area
substituted by water (0.14 mg; 2.43 mol% H2O), which is released
between 50 and 200 ◦C with an IC maximum for m18 around 100 ◦C.
The observation that a certain part of the adsorbed methanol
is retained at the surface needs an interpretation. It cannot be
related to the channels existing in the HTB structure of ␤-AlF3 as
the methanol molecule is too large to get inside the channels.
2
values for the exothermal peaks for ␤-AlF3 as shown by the com-
parison in Fig. 4. By subtracting the area values for physisorption
alone from the sum value (cf. ␤-AlF3 in Fig. 4), one can separate the
net exothermicity for the chemisorption at the given surface (but
see below).
A further detail is noteworthy. For the case of pure physisorp-
tion, where no net mass gain persists (␣-AlF3 in Fig. 4), one could
have expected that the peak areas of exothermal physisorption and
endothermal desorption should be approximately equal as the pro-
cesses are reversible and no further enthalpic contribution should
appear. This is not the case, as clearly expressed for ␣-AlF3 (6–8 ␮V s
for the adsorption vs. 2–4 ␮V s for the desorption). Here, one should
bear in mind that both processes are competitive and partly overlap
each other with a differing time dependence. Hence, it is impossi-
ble to clearly separate the individual part of each contribution and
the peak areas are necessarily different. This is a limitation also
for the above-mentioned area subtraction when trying to separate
quantitatively the portion of chemisorption.
3.4. The sorption behaviour of HS-AlF3
The sequence of Figs. 6–8 represents the adsorption study of HS-
AlF3. The mass loss occurring during the thermal pretreatment is
greater (5.36%) than for the other AlF3 modifications investigated
To summarize, one should note that Fig. 4 illustrates qualita-
tively and quantitatively that ␣- and ␤-AlF3 behave differently.
Only the empty apparatus yields a real blank experiment whereas
␣-AlF3, unlike expected, is a weak adsorbent showing predomi-
nantly physisorption. ␤-AlF3, however, exhibits both chemi- and
physisorption and adsorbs strongly.
Fig. 6. PTA curves of pretreated (250 ◦C; N2) HS-AlF3 (12.8 mg, ꢀm = +0.41 mg).
3.3.2. Different desorption behaviour
The adsorption of methanol is not completely reversible. If the
loaded sample of ␤-AlF3 as presented in Fig. 3 is re-heated under
nitrogen in a subsequent TA run, 71% (0.10 mg) of the originally
adsorbed 0.15 mg methanol was released from the surface (Fig. 5).
The desorption starts at about 50 ◦C, i.e. only slightly higher than
the loading temperature 46 ◦C and is spread over a wide temper-
ature range (50–200 ◦C) with a desorption maximum at 110 ◦C.
As indicated by the IC curve for m18, a minor co-adsorption of
water during the preceding methanol injections cannot be fully
excluded.
Another result is obtained when a methanol-loaded ␤-AlF3
(0.12 mg on 25.86 mg AlF3; 1.23 mol% MeOH) is stored 24 h at
humid air. In this case, no desorption of methanol at all can be
detected under re-heating (curves not shown here). It is completely
Fig. 7. TA-MS curves for the desorption of methanol from HS-AlF3 (13.21 mg,
ꢀm = −0.48 mg) precedingly loaded with 0.41 mg MeOH (cf. Fig. 6). Note the coinci-
dence of the weakly expressed second IC maximum for m31 above 250 ◦C with the
IC maxima for m45(M−1)+ and m46(M+) caused by dimethylether.
2
Note that the area values for the desorption peaks (between −11 and −15 ␮V s
for ␤-AlF3 or −2 and −4 ␮V s for ␣-AlF3, cf. Fig. 4), have to be considered each as
practically constant within the experimental error.

104
M. Feist et al. / Thermochimica Acta 498 (2010) 100–105
qualitative character of exclusively chemisorption does not change.
The absolute methanol uptake, however, is lower (2.69% vs. 3.22%).
Here as well, the methanol adsorption is not finished with the 15th
pulse. The desorption after the second loading (not shown here) is
as complete as the first one (0.33 mg uptake vs. 0.31 mg loss).
If water is adsorbed, the behaviour is different. After pure
chemisorption at the beginning of the experiment, one observes
physisorption starting with the sixth injection pulse (of nine in
total) and the adsorption reaches saturation. The total mass gain
amounts to 1.63%, i.e. 3.8 mol% H2O. Even if this is only the half of
the methanol loading, a value of 3.8 mol%, nevertheless, indicates a
good sorption behaviour.
3.5. The adsorption behaviour of further modifications of AlF3
Fig. 8. Comparison of the PTA curves for the first and the second adsorption mea-
surements of pretreated (250 ◦C; N2) HS-AlF3.
Four further AlF3 phases have been investigated in the same
way: a gel-like form of ␤-AlF3, prepared via a special sol–gel synthe-
sis route [22] to give an especially large surface, and, finally, ␩-AlF3,
␪-AlF3, and ␬-AlF3, all representing crystalline modifications (cf.
Table 1).
As expected, the gel-like form of ␤-AlF3 showed a strong adsorp-
tion of methanol and the value of 0.2 mmol/g (1.6 mol%) surface
loading, being twice as much than for normal ␤-AlF3 (0.9 mol%), is
noteworthy. On the other hand, this is only 22% of the adsorption
capacity of HS-AlF3.
which is an expression of the strong sorption activity of HS-AlF3. As
the sample was glove box-stored and -weighed, the water uptake
occurs exclusively during the opening of the measuring cell and the
positioning of the TA crucible onto the sample holder (60–70 s on
humid air). Interestingly enough, this did not lead to an observable
pyrohydrolysis of the fluoride as proved by the course of the mass
numbers 19 and 20 (not shown here).
As can be seen in Fig. 6, the adsorption behaviour of HS-AlF3 is
qualitatively and quantitatively quite different from all the other
modifications studied here. Firstly, the absolute methanol uptake
is impressing: 1.19 mmol/g (7.7 mol%) which is more than 5 times
greater than for the comparably strongly adsorbing ␤-AlF3 gel
(Table 1). Moreover, the TG curve shape shows that the adsorp-
tion becomes weaker with the 13th to 15th injection pulses, but is
not yet definitely finished. Secondly, the exothermal DTA peaks for
the injections are not followed by the above-mentioned (Section
3.3.1) weak endothermal post-effects indicating the desorption of
physisorbed species in the gas flow. This means that exclusively
chemisorption occurs in the case of HS-AlF3! This is in line with
the TG curve shape, even if the individual steps are not as ideally
expressed as observed for ␤-AlF3 (cf. Fig. 3), where the distinction
between chemi- and physisorption is extraordinarily clear.
Not only the adsorption behaviour, but the desorption
behaviour of HS-AlF3 as well is different from all other modifica-
tions investigated. The IC curves in Fig. 7 to be compared with Fig. 5
demonstrate that the desorbed amount is within the experimental
error identical with the precedingly uptaken amount (0.41 mg vs.
0.48 mg). This is not generally the case for the other AlF3 modifi-
cations. Only methanol is released in the usual desorption range
(50–250 ◦C), a water release can be detected only above 250 ◦C
and is attributed to beginning pyrohydrolysis (see below). The des-
orption profile is spread over a broader temperature interval up
to 230 ◦C and the first maximum is shifted to higher tempera-
tures (130 ◦C). A weakly expressed second intensity maximum for
m31 appears at about 280 ◦C. Moreover, the m18 and m19 inten-
sities begin to increase. This can be explained by the formation of
dimethylether from methanol which is catalyzed by HS-AlF3 and
which produces water as the second product. As a consequence,
HF is released due to the beginning pyrohydrolysis (m19). This
behaviour is comparable to that observed for the precursors of
HS-AlF3 [12], where ether formation was established during the
post-fluorination yielding the final HS-AlF3 and where the di-i-
propylether originates from the i-propoxide groups contained in
the precursor phase.
␩-AlF3 exhibited a distinctly differing sorption behaviour.
Only the first pulse yields a clear picture indicating exclusively
chemisorption without any endothermicity after the exothermal
adsorption peak. Starting with the second injection already, an
unstable course of the TG curve was recorded. Adsorption and
desorption seem to alternate in an irregular manner. A net mass
gain of 10 ␮g persists which leads to a very low surface loading of
0.008 mmol/g (0.07 mol%). This is in the same order of magnitude
as observed for ␣-AlF3, where practically no persisting adsorption
could be established.
More or less identical findings have been obtained both for ␬-
AlF3 and ␪-AlF3. They show nearly the same sorption features, i.e. a
less-regular course of the TG curves: certain pulses do not effect a
detectable mass gain, certain others even show a slight mass loss.
Roughly, the three modifications ␩-AlF3, ␬-AlF3, and ␪-AlF3 behave
in the same way as does ␣-AlF3.
3.6. Discussion
A comprehensive overview about the sorption properties is
given in Table 1. The values for the surface loading are compared
with those for the NH3-TPD, directly representing the sum of Lewis
and Brønsted acid centres, and with the BET surface values. The
BET surfaces of four modifications are relatively close together,
except the three extreme values, i.e. the very small surface areas
of ␣- and ␩-AlF3 and the very high surface of HS-AlF3. As the NH3-
TPD data represent the ammonia amount thermally removed from
the covered surface, the values given in ␮mol NH3/m2 have to be
understood as ␮mol acid centres/m2.
Differently from NH3, which easily can act as Lewis or Brønsted
base, CH3OH should only behave as a Lewis base probing exclu-
sively the Lewis acidic centres.
However, since the fluorides are pretreated at 250 ◦C, the abso-
lute amount of Brønsted acidic centres determined by NH3-TPD is
negligible in this case. Based on the assumption that the values for
the amount of centres on the surface calculated by the methods
of TPD and of PTA represent the part of the interaction of Lewis
acidic centres and the probe molecules (TPD: NH3, PTA: CH3OH),
the values should be directly comparable.
The almost reversible adsorption of methanol which has been
deduced from the findings for the first loading described above
implicated the question of the repeatability of the surface load-
ing of HS-AlF3. As can be seen in Fig. 8, the methanol adsorption
is reversible, but in a limited manner. Interestingly enough, the
Unambiguously, the starting point for an interpretation of the
experimental and calculated data regrouped in Table 1 must be, on

M. Feist et al. / Thermochimica Acta 498 (2010) 100–105
105
the one hand, the interplay of acidic sites and porosity, and, on the
other hand, structural aspects such as the channel structure of cer-
tain AlF3 modifications together with the size and the coordinative
situation of surfacial AlF6 units. It can be stated that the evolution
of the values for SBET, nMeOH in ␮mol/g and in mol% is qualitatively
almost identical, except the non-adsorbing ␣- and ␪-AlF3. All these
data are mass-based. If the surface loading by methanol is related,
however, to the actual surface basing on SBET, the practical identity
of all calculated values given in ␮mol MeOH/m2 might surprise.
This is especially true if they are compared with the surface-related
values (in ␮mol NH3/m2) obtained with the NH3-TPD method.
Reconsidering, however, the different size of the probe molecules
(NH3 would fit into the channels of ␤- or ␩-AlF3, whereas CH3OH
is too large), the values obtained by NH3-TPD are possibly overesti-
mated. An uptake of small molecules into the channels as recently
observed for water and HTB tungstates [30] seems to play a role
also for the adsorption of NH3 onto ␤- and ␩-AlF3 resulting in
higher surface-related values for these phases (Table 1, last col-
umn). Taking further into account that MeOH, unlike NH3, interacts
exclusively with Lewis acidic centres, the analytical information
obtainable by MeOH-PTA seems to be more specific.
Concerning the values obtained with the MeOH-PTA experi-
ments, one has further to bear in mind, that the geometric and
coordinative situation of the AlF6 octahedra, being the basic struc-
tural unit of all AlF3 modifications, is not so strongly different from
each other, even if they adopt different crystal structures or amor-
phous forms. Interpreting the calculated values for the fluorides,
the surface area and the morphology of the investigated phase
is clearly one of the biggest influencing variables governing the
adsorption properties and therewith the amount of offered acidic
centres. Additionally, the particular Lewis acidic behaviour of ␤-
AlF3 and HS-AlF3, one of the strongest solid Lewis acids known, can
be proved by the extraordinary adsorption behaviour of MeOH.
The variations of distortions and linking of the AlF6 octahe-
dra, which make the main differences (among them the channel
feature) between the different crystalline modifications, seem to
exhibit only a minor effect. Nevertheless, the structural variations
which, on the other hand, are clearly to be identified by XRD or other
structural methods such as NMR of various nuclei [17] are at least
responsible for subtle differences, e.g. as shown for the characteris-
tics of ␤-AlF3, which shows Lewis acidic behaviour in comparison to
the other crystalline fluorides, such as ␬- or ␩-AlF3 both exhibiting
channel structures.
channel structures, but exclusively by surface effects, i.e. primarily
by Lewis acid sites and the size and the morphology of the phase
particles.
It could be demonstrated that PTA, indeed, is a valuable tool
for investigating the sorption properties of solids which allows
to follow and to visualize fine details of the interaction, i.e. the
energetic balance and the mass changes as well. But these quan-
titative aspects (enthalpy and mass) are, at least for this group of
AlF3 polymorphs, better understood than the interplay of acid sites
and porosity, on the one hand, and structural aspects such as the
channel structure of certain AlF3 modifications on the other hand.
Comparing the information that could be obtained here from NH3-
TPD with that deduced from PTA experiments, the latter seems to
differentiate remarkably better in the case of the investigated AlF3
modifications.
Acknowledgements
Dr. Ekkehard Füglein (Netzsch Gerätebau GmbH, Selb, Germany)
and Dr. Jan Hanss (University of Augsburg, Germany) contributed
to the presented work by supporting us for the area integrations of
the DTA curves in Fig. 4 (E.F.) and by helpful trilateral discussions
of experimental details. This is gratefully acknowledged.
References
[1] E. Kemnitz, U. Groß, St. Rüdiger, Ch. Shekar, Angew. Chem. Int. Ed. 42 (2003)
4251–4254.
[2] St. Rüdiger, E. Kemnitz, Dalton Trans. (2008) 1117–1127.
[3] D. Dambournet, A. Demourgues, C. Martineau, S. Pechev, J. Lhoste, J. Majimel,
A. Vimont, J.-C. Lavalley, C. Legein, J.-Y. Buzaré, F. Fayon, A. Tressaud, Chem.
Mater. 20 (2008) 1459–1469.
[4] M. Gaudon, J. Majimel, J.-M. Heintz, M. Feist, D. Dambournet, A. Tressaud, J.
Fluorine Chem. 129 (2008) 1173–1179.
[5] G. Scholz, M. Feist, E. Kemnitz, J. Solid State Sci. 10 (2008) 1640–1650.
[6] G. Scholz, R. König, J. Petersen, B. Angelow, I. Dörfel, E. Kemnitz, Chem. Mater.
20 (2008) 5406–5413.
[7] H. Krüger, E. Kemnitz, A. Hertwig, U. Beck, Thin Solid Films 516 (2008)
4175–4177.
[8] Th. Krahl, A. Vimont, G. Eltanany, M. Daturi, E. Kemnitz, J. Phys. Chem. C 111
(2007) 18317–18325.
[9] E.K.L.Y. Hajime, J.L. Delattre, A.M. Stacy, Chem. Mater. 19 (2007) 894.
[10] A. Wander, C.L. Bailey, S. Mukhopadhyay, B.G. Searle, N.M. Harrison, J. Mater.
Chem. 16 (2006) 1906.
[11] K.O. Christe, D.A. Dixon, D. McLemore, W.W. Wilson, J.A. Sheey, J.A. Boatz, J.
Fluorine Chem. 101 (2000) 151.
[12] S. Rüdiger, U. Groß, M. Feist, H.A. Prescott, C.S. Shekar, S.I. Troyanov, E. Kemnitz,
J. Mater. Chem. 15 (2005) 588–597.
[13] A. Hess, E. Kemnitz, J. Catal. 149 (1994) 449–457.
[14] M. Maciejewski, C.A. Müller, R. Tschan, W.-D. Emmerich, A. Baiker, Ther-
mochim. Acta 295 (1997) 167.
[15] M. Maciejewski, A. Baiker, Thermochim. Acta 295 (1997) 95.
[16] F. Eigenmann, M. Maciejewski, A. Baiker, Thermochim. Acta 6276 (2000) 1.
[17] R. König, G. Scholz, K. Scheurell, D. Heidemann, I. Buchem, W.E.S. Unger, E.
Kemnitz, J. Fluorine Chem. (in press).
[18] T. Krahl, Dissertation, Humboldt-Universität zu Berlin, 2005.
[19] D.L. Xu, Y.Q. Luo, Y. Li, L.B. Zhou, W.K. Gai, Thermochim. Acta 352 (2000) 47–52.
[20] A. Le Bail, J.L. Fourquet, U. Bentrup, J. Solid State Chem. 100 (1992) 151–159.
[21] N. Herron, D.L. Thorn, R.L. Harlow, G.A. Jones, J.B. Parise, J.A. Fernandezbaca, T.
Vogt, Chem. Mater. 7 (1995) 75–83.
This might explain that the undoubtedly extraordinary position
of HS-AlF3 in the sequence of phases investigated is not expressed
in a comparably extraordinary value for the methanol loading given
in ␮mol/m2. If, however, the loading is regarded mass-based, the
situation is unambiguous. This can be regarded as a hint for the
great potential of the PTA method and would be a further demon-
stration of the deep insight in fine details of the sorption process
itself, which is obtainable by employing PTA. NH3-TPD, being well-
established in catalysis research, yields a quantitative information
about the surface coverage and is completed by the qualitative
description deducable from the desorption profile. PTA, on the
other hand, not only yields quantitative data via enthalpy and mass
changes (the latter often being underestimated in catalysis papers),
but also allows to distinguish between different kinds of adsorption,
and even to inject pulses of various liquids or gases onto a solid in
one single carefully designed experiment. This might be considered
as a “titration” of the surfaces.
[22] EU Research Project “Funfluos”, NMP3-CT-2004-505575, Activity report, Octo-
ber 31, 2005, p. 18.
[23] M. Feist, I.K. Murwani, E. Kemnitz, J. Therm. Anal. Calorim. 72 (2003) 75.
[24] M. Feist, E. Kemnitz, Thermochim. Acta 446 (2006) 84.
[25] J. Hanss, A. Kalytta, A. Reller, Hyphenated Techniques in Thermal Analysis,
in: Proc. 5th Selber Kopplungstage, E. Kapsch, M. Hollering (eds.), Bad Orb,
Germany, May 25–28, 2003, pp. 151–63.
[26] E. Post, et al., Coupling techniques in thermal analysis, in: Proc. 4th Selber
Kopplungstage, Selb, Germany, June 24–26, 2001.
[27] J. Ravez, A. Mogus-Milankovic, J.-P. Chaminade, P. Hagenmuller, Mater. Res.
Bull. 19 (1984) 1311.
[28] V. Quaschning, J. Deutsch, P. Druska, H.J. Niclas, E. Kemnitz, J. Catal. 177 (1998)
164.
[29] R.B. King (Ed.), Encyclopedia of Inorganic Chemistry, vol. III, 2nd edn., Wiley &
Sons, Chichester, 2005, p. 1540.
4. Conclusions
[30] V. Luca, C.S. Griffith, J.V. Hanna, Inorg. Chem. 48 (2009) 5663–5676.
[31] A.J. Lecloux, Texture of catalysts, in: J.R. Anderson, M. Boudart (Eds.),
Catalysis—Science and Technology, vol. 2, Akademie-Verlag, Berlin, 1983, pp.
171–230.
Summarizing the reported observations it is concluded that the
methanol adsorption behaviour of the various forms of AlF3 is not
governed by an influence of the structure, e.g. by the existence of