www.sciencemag.org/cgi/content/full/323/5920/1455/DC1
Supporting Online Material for
The Initial Stages of Template-Controlled CaCO3 Formation
Revealed by Cryo-TEM
Emilie M. Pouget, Paul H. H. Bomans, Jeroen A. C. M. Goos, Peter M. Frederik,
Gijsbertus de With, Nico A. J. M. Sommerdijk*
*To whom correspondence should be addressed. E-mail: [email protected]
Published 13 March 2009, Science 323, 1455 (2009)
DOI: 10.1126/science.1169434
This PDF file includes:
Materials and Methods
SOM Text
Figs. S1 to S8
Tables S1 and S2
References
2
Supporting Online Material 1
Materials and Methods
Cryogenic Transmission electron microscopy (cryo-TEM) and cryo-electron tomography
(cryoET)
The sample vitrification procedure was carried out using an automated vitrification robot (FEI
Vitrobot™ Mark III) equipped with humidity and temperature controlled glove box. CryoTEM
grids, R2/2 Quantifoil Jena grids, were purchased from Quantifoil Micro Tools GmbH. Prior to
the vitrification procedure the grids were surface plasma treated using a Cressington 208 carbon
coater. The cryoTEM experiments were performed on a FEI Technai 20 (type Sphera) TEM or on
the TU/e CryoTitan (FEI) (www.cryotem.nl). A Gatan cryo-holder operating at ~ -170 °C was
used. The Technai 20 is equipped with a LaB6 filament operating at 200 kV and the images were
recorded using a 1k x 1k Gatan CCD camera. The TU/e CryoTitan is equipped with a field
emission gun (FEG) operating at 300 kV. Images were recorded using a 2k x 2k Gatan CCD
camera equipped with a post column Gatan Energy Filter (GIF).
The 3-dimensional (3D) reconstructions were performed with the software Inspect 3D v.3.0 (FEI
Company). For the segmentation and visualization of the 3D volume, Amira 4.1.0 (Mercury
Computer Systems) was used. The ACT-4 (Automated Crystallography for the TEM), TLS
EDAX 2003 software was used as an aid to analyze the diffraction patterns. The TEM image in
figure 1 has been enhanced using the software Matlab 2007b. The filters used are an anisotropic
linear filter with 10 iterations and a coefficient of 0.25 and subsequently a sharpening filter with a
coefficient of 0.5.
Experimental design
As a model system, we used monolayers of stearic acid (CH3(CH2)16COOH) as a template,
deposited on a supersaturated solution of Ca(HCO3)2 (S1,S2). The mineralization reaction then
proceeds by the out-gassing of CO2 from the solution which shifts the equilibrium below to the
right, resulting in the template controlled precipitation of CaCO3.
Ca(HCO3)2
CaCO3 + CO2 + H2O
3
The reaction was performed directly on the cryoTEM grid, in a glove box in which the
temperature and the humidity were controlled at 22 oC and 100% relative humidity. At the
beginning of the reaction the cryoTEM grids were present in the solution, on a metallic support.
The cryoTEM grids are 200 mesh copper grids, covered by a carbon film that contained a regular
pattern of 2 μm holes in the carbon layer (Figure S1, Step 1). By draining the solution, the
monolayer could be loaded onto these cryoTEM grids with minimal disturbance of the system
(Step 2). The CO2 evaporation from the supersaturated solution induced the formation of CaCO3
particles (Step 3). After a controlled reaction time (the formation of the CaCO3 particles starts at
Step 1 and the time is estimated from Step 1 up to the moment of vitrification, Step 5), the grid
was transferred to a fully automated vitrification robot (Vitrobot) (S3). The environmental
conditions were kept constant at 22 °C and 100% relative humidity. A suspension of hydrophobic
CdSe nanorods in chloroform was added on the hydrophobic side of the monolayer (Step 4) to
mark the monolayer in the cryoTEM experiments (see Supplementary Information 6). Finally, the
excess of liquid was removed by automatic blotting to produce a thin liquid layer (around 100 nm
thick) containing the monolayer and the calcium carbonate particles, and the reaction was
quenched by plunging the grid into melting ethane (Step 5).
For all crystallization experiments, a 9 mM supersaturated solution of Ca(HCO3)2, prepared
following the Kitano method (S2), was used. The required calcium carbonate was purchased from
Merck, purity ≥ 99%, the stearic acid from Fluka, purity ≥ 99.5%, and the chloroform from
Biosolve, purity > 99.9%. Hydrophobic CdSe nanorods were provided by Patrick Chin (TU/e,
Eindhoven) following a slightly modified procedure from Bunge et al. (S4).
The monolayers were prepared by spreading a chloroform solution of stearic acid (0.5 mg/ml).
The amount of stearic acid has been calculated to cover the surface with a close-packed
organization (S1).
4
Figure S1: Left- Vitrification robot coupled with the glove-box positioned for the transfer of the
grid prior to vitrification. Right- Scheme of the reaction on the cryoTEM grid. Step 1: cryoTEM
grid submerged in a supersaturated Ca(HCO3)2 solution with a stearic acid monolayer spread at
the air/water interface. Step 2: draining of the liquid placing the monolayer on the grid. Step 3:
CO2 out-gassing and CaCO3 formation. Step 4: transfer of grid to the vitrification robot after
different reaction times and addition of hydrophobic nanorods to mark the monolayer. Step 5:
blotting and vitrification by plunge freezing. During the whole process, the temperature is 22 °C
and the relative humidity >97%.
5
Supporting Online Material 2.
Analytical Untracentrifugation.
Materials
Analytical ultracentrifugation measurements were performed using a XLI Ultracentrifuge
(Beckman Coulter, Palo Alto CA), equipped with Rayleigh interference optics at 60000 rpm.
Data were evaluated with SEDFIT evaluation software (version 11.3b P. Schuck 2008).
Analytical Ultracentrifugation
A 9 mM supersaturated Ca(HCO3)2 solution was prepared (see supporting online material 1) and
during out-gassing of CO2 from the solution the pH evolution was monitored. Figure S2 shows
the change in pH during the early stages of the mineralization process. When nucleation occurs, a
significant drop in pH can be observed. Due to larger volume to surface ratios of the solutions
used for the pH measurements, compared to the on-the-grid preparation of the cryoTEM samples,
the outgassing rate of the solution and the associated reaction kinetics are slower, causing
nucleation to occur only after ~90 min.
Figure S2- Change in pH during the crystallization process. The arrows indicate moments in time
at which the measurements were done.
6
Several samples were taken from the crystallization solution at different time points (Figure S2,
red arrows). Within these samples the sedimentation coefficient of the different species present
was estimated by analytical ultracentrifugation. Assuming the Ca(HCO3)2 solution is a mixture of
non-interacting ideally sedimenting molecules, a diffusion corrected continuous distribution
function c(s) can be calculated for all samples using the Lamm equation model (Figure S3) (S5).
Sedimentation coefficient s is defined as the velocity νt of the particle per unit gravitational
2
acceleration (centrifugal acceleration : ω r , where ω = angular velocity and r = radial distance to
rotation axis)
s=
vt
(equation S1)
ω 2r
Large and dense particles sediment faster than the smaller or less dense ones, which induces a
higher sedimentation coefficient value.
1.2
0 min (pH=6.06)
10 min (pH=6.20)
30 min (pH=6.48, pH=6.43)
66 min (pH=7.00)
90 min (pH=7.44)
1.0
repetition
c(s)
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
5
6
7
s (Sv)
Figure S3: Diffusion corrected sedimentation coefficient distribution c(s). The peaks in the
region between 0 – 3 Svedberg, Sv (Sv = 10-13 s) were normalized. A typical uncertainty in
sedimentation coefficients is 0.3 Sv.
7
Three noteworthy regions can be observed in Figure S3. The first region (s = 0-0.6 Sv) indicates
the presence of ions. The second region (s = 1.5 - 3 Sv) indicates the presence of small particles
which are slightly bigger than ions. These particles, which already are present in the Ca(HCO3)2
solution after 10 minutes (before nucleation), can be considered as the nanometer sized prenucleation clusters (S6). After approximately 30 minutes, also larger particles appear in the
solution (≥ 4.5 Sv). The presence of these particles can be explained by the formation of
aggregates between pre-nucleation clusters. This observation confirms the presence of the
nucleation clusters (0.6-1.1 nm) and their aggregates (2-4 nm) as was also shown by high
resolution cryoTEM (see manuscript text). After nucleation the aggregates were no longer
visible, suggesting they grew to transform into the precipitating amorphous calcium carbonate
(ACC) nanoparticles. Nevertheless, at this stage the pre-nucleation clusters were still present.
8
Supporting Online Material 3
Cryoelectron tomography settings
The reconstructions shown in Figures 2E-G were obtained from tilt series recorded using the
following settings:
Early stages (fig 2E)
Tilt range: -65° to +64°, using a Saxton scheme.
Magnification 19000x.
Defocus -8 μm
Total dose = 43 e-.Å-2
Intermediate stages (fig 2F)
Tilt range: -60° to +68°, using a Saxton scheme.
Magnification 19000x.
Defocus -6 μm
Total dose = 51 e-.Å-2
Mature stages (fig 2G)
Tilt range: From -70° to +70°, using a Saxton scheme.
Magnification 11500x.
Defocus -10 μm
Total dose = 70 e-.Å-2
9
Supporting Online Material 4
Comparison of electron diffraction patterns in figure 1.
We compared the electron diffraction patterns of the ACC particles in Figure 1-A (See
manuscript) with those of vitrified films of pure water, aqueous 10 mM CaCl2 solutions and
fresh 9 mM Ca(HCO3)2 solutions. To this end the position of the first ring of the diffraction
patterns of water, CaCl2 (10 mM), Ca(HCO3)2 solution (9 mM) and ACC was measured and
averaged.
Table S1. Comparison of the radial integration of the diffraction patterns of pure water,
aqueous 10 mM CaCl2 solutions and fresh 9 mM Ca(HCO3)2 and ACC.
<X> (Å)
σ(<X>)
n
Water
3.66
0.06
8
CaCl2
3.71
0.05
16
Ca(HCO3)2
3.80
0.05
12
ACC
3.81
0.03
12
<X> is the average, σ(<X>) is the standard deviation calculated as σ =
2
∑ (x − x)
(
(n − 1)
2
) and
n is the number of measurements.
Welch-Aspin generalized T-test
To test whether the values are different the Welch-Aspin generalized T-test (S7) was
used with the data for the water, CaCl2 solution, Ca(HCO3)2 solution and ACC. For the
combination water-CaCl2 solution a T-value of 2.17 resulted, to be compared with T = 3.01
and T = 2.17 for a 1% and 5% two-sided reliability interval, respectively. From these data we
conclude that the averages are not significantly different. For the combination waterCa(HCO3)2 solution T = 5.73, to be compared with T = 2.98 and T = 2.15 for a 1% and 5%
two-sided reliability interval, respectively. From this comparison we conclude that these
average values are significantly different. Similarly for the ACC- Ca(HCO3)2 solution T =
0.60, to be compared with T = 3.20 and T = 2.10 for a 1% and 5% two-sided reliability
interval, respectively. From this comparison we conclude again that these average values are
not significantly different. Finally, for the ACC and the water, T = 6.97 to be compared with
10
T = 3.20 and T = 2.24 for a 1% and 5% two-sided reliability interval, respectively. From this
we conclude that, as expected, these values are statistically different.
Table S2. Comparison of the T-values of the combination of Water-CaCl2, WaterCa(HCO3)2, CaCl2- Ca(HCO3)2, Water-ACC and Ca(HCO3)2-ACC and the significance
levels for 1% and 5% two-sided reliability intervals.
T
T1%
T5%
Water-CaCl2
2.17
3.01
2.17
Water- Ca(HCO3)2
5.73
2.98
2.15
CaCl2- Ca(HCO3)2
4.82
2.81
2.06
Water-ACC
6.97
3.20
2.24
Ca(HCO3)2-ACC
0.60
2.89
2.10
Conclusion
Overall we conclude that the d-spacings associated with the first diffraction ring for
ACC particles of different size (3.81±0.03 Å) show small but statistically significant
differences as compared to that of vitrified water (3.66±0.06 Å). In contrast, the values for
the ACC particles were indistinguishable from those determined for a freshly prepared
Ca(HCO3)2 mineralization solution (3.80±0.05 Å). It was also demonstrated that the dspacings determined from patterns recorded for vitrified 10 mM CaCl2 solutions (3.71±0.05
Å), in agreement with literature reports (S8), were statistically indistinguishable from those
of water. This shows that at the concentrations used the longer d-spacing observed for the 9
mM Ca(HCO3)2 solutions cannot simply be attributed to the presence of inorganic ions in
solution. Moreover the finding of d-spacings indistinguishable to those found for ACC
particles suggests that already in a freshly prepared Ca(HCO3)2 solution amorphous prenucleation clusters are present.
11
Supporting Online Material 5
Detailed indexation of the diffraction patterns of particles 4, 5 and 6 shown in Figure 2.
Figure S4- Detailed indexation of the diffraction patterns shown in Figure 2. Scale bar = 5 nm-1.
A) intermediate crystal, particle 4. Some weak spots can be indexed as vaterite (0 0.1): 1 = 1 .0,
2 = 1 0.0 and 3 = 0.0. The spots marked by yellow arrows could not be indexed as they belong
to different crystalline domains, however, all d-spacings correspond to vaterite. B) intermediate
crystal, particle 5. Vaterite (0 0.1). 1 = 1.0, 2 = 0.0, 3 = 1.0. The spots marked by yellow
arrows are not indexed but the d-spacings correspond to vaterite. C) mature crystal, particle 6.
Vaterite (0 0.1). 1 = 1 0.0, 2 = 0 1.0, 3 = 1 1.0.
12
Supporting Online Material 6
Marking the monolayer with hydrophobic nanorods, purpose and consequences
In low dose cryoTEM it is difficult if not impossible to distinguish the position of the monolayer
which has a low and homogeneous density. To mark the interface for tomographic analysis,
hydrophobic nanorods were added to the hydrophobic side of the monolayer at the end of the
sample preparation process. As the hydrophobic markers were confined to the only hydrophobic
zone of the sample, the hydrophobic tails of the monolayer, their high contrast allowed
positioning of the interfacial monolayer both in 2D cryoTEM images as well as in the
tomographic reconstructions.
However, these nanorods diffract and their near-random orientation gives rise to a nonhomogeneous ring (Figure S5, A) close to the diffuse first ring of ACC. This thin ring interferes
with the interpretation of the diffraction patterns of CaCO3 (Figure S5, B). By tilting the sample
45° it was possible to record a low dose selected area electron diffraction (LDSAED) pattern of
the particle without the diffraction contribution of the nanorods and only the diffuse ring of ACC
and the spots arising from crystalline CaCO3 are observed (Figure S5, C).
A
B
C
tilt 45°
Figure S5: Diffraction patterns of A- the hydrophobic particles added on the monolayer, B- the
particle 2 of the intermediate stage and C- the same particle after a tilt of the grid of 45°. The ring
observed on the pattern A due to the nanorods markers disappears when the sample is tilted, i.e.
when the selected area is focused only on the particle.
13
Supporting Online Material 7
Crystallization in a 9 mM Ca(HCO3)2 solution in the absence of a monolayer.
The crystals observed are randomly oriented calcite and vaterite.
Figure S6: CryoTEM images (top) and low dose electron diffraction patters of CaCO3 crystals
obtained in a 9 mM Ca(HCO3)2 solution after 40 min of reaction, without the presence of the
monolayer. Scale bars = 500 nm and 5 nm-1 A) Calcite, orientation (2 2.3). 1=0 1. ; 2= 1.0: 3=
2. . The spot marked with the white arrow shows a d-spacing corresponding to calcite but is
not indexed. B) Calcite, orientation (0 0.1). 1=1 0.0; 2=0 1.0; 3=1 1.0. C) Polycrystal. The dspacings correspond to vaterite but the pattern is not indexable. D) The spots marked with green
arrows can be indexed with vaterite oriented (0 0.1). The ones marked with blue arrows can be
indexed with vaterite ( 2.-1) and the ones marked by yellow arrows are not indexed but the dspacings correspond to vaterite. E) Vaterite, orientation (6 5. ). 1= 0 . ; 2= 0. ; 3= . .
14
Supporting Online Material 8
The templating effect of stearic acid monolayers
The results presented in this report complete and connect the results from previous studies on this
stearic acid (SA) monolayer / Ca(HCO3)2 solution system. Two decades ago, it has been shown
that depending of the concentration of the Ca(HCO3)2 solution, calcite (1 .0) or vaterite (00.1)
can be observed at the interface with a SA monolayer covering a solution contained in a
crystallizing dish (S1,S9). In 2003, it was demonstrated in two different ways that the selection of
the metastable vaterite over the thermodynamically calcite is not an effect of the organization of
the functional groups in the monolayer but of the CO2 evaporation rate (S10, S11) A rapid CO2
evaporation favors the kinetic product, vaterite. Moreover, synchrotron X ray scattering
experiments showed randomly oriented crystals, even challenging the ability of SA to exert a
templating effect (S11).
In the present study we confirmed that (1 .0) oriented calcite was formed under monolayers of
SA deposited on 25 mL of a 9 mM Ca(HCO3)2 solution held in a crystallization dish, i.e. using a
low CO2 evaporation rate (data not shown), while vaterite (00.1) forms when the reaction is
performed on a TEM grid, hence at a much faster evaporation rate. These results imply that the
selection of vaterite over calcite is indeed due to the kinetics of the reaction but the systematic
orientation of the crystals with respect to the monolayer confirms the templating effect. The
finding of polycrystals with different crystallographic orientations of vaterite in the first stages of
crystallization is in line with the observations of DiMasi et al. (S11) and may explain the
discrepancy between the results from the different laboratories (S1, S9, S10, S11).
Figure S7: SEM image of the calcite (1 .0) crystals formed in crystallization dishes.
15
Supporting Online Material 9
Summary of particle size distributions from TEM images.
Prenucleation clusters:
Most of the pre-nucleation clusters show a diameter between 0.6 and 1.1 nm but already in the
very early stages of the reaction a small number of bigger clusters with diameters up to 4 nm are
also present.
ACC particles
The majority of the ACC particles have diameters between 20 and 70 nm with a maximum in the
distribution around 30 nm. However, a small fraction has diameters as large as 120 nm. The size
distribution of the ACC particles does not change significantly during the experiment,
maintaining a distribution maximum at 30 nm, both for the particles in solution as well as for the
particles attached to bigger ones.
25
12
20
10
8
15
6
10
4
5
2
0
0
20
30
40
50
60
70
80
d(nm)
90 100 110 120
20
30
40
50
60
70
80
90 100 110 120
d(nm)
Figure S8. ACC particle size distribution during the (left) early stages of the reaction (0-15
min) and (right) advanced stages of the reaction (25-45 min).
Crystals
Polycrystals are generally observed with a size between 70 and 300 nm and a small amount
shows bigger diameters (< 500 nm). The single crystals observed had diameters starting at 300
nm.
Critical particle sizes
The observed distributions suggest that the pre-nucleation clusters and the ACC nanoparticles
have an optimum volume-to-surface area distribution at diameters of 0.7 and 30 nm, respectively.
16
Similarly, the observations that most ACC particles have a size < 70 nm while a minimal particle
diameter of ~70 nm exists for the polycrystalline CaCO3 implies that this size is critical for the
development of crystalline domains inside an amorphous matrix. The transition from polycrystals
to single crystals of vaterite is a more gradual process and takes place at particle diameters
between 300 and 500 nm.
References:
S1.
S. Mann, B. R. Heywood, S. Rajam, J. D. Birchall, Nature, 334, 692 (1988)
S2.
Y. Kitano, K. Park, D. W. Hood, J. Geophys. Res, 67, 4873 (1962)
S3.
M. R. Vos, P. H. H. Bomans, P. M. Frederik, N. A. J. M. Sommerdijk,
Ultramicroscopy, 108, 1478 (2008)
S4.
S. D. Bunge et al., J. Mater. Chem., 13, 1705 (2003)
S5.
P. Schuck, Biophys. J., 78, 1606 (2000)
S6.
D. Gebauer, A. Völkel, H. Cölfen, Science, 322, 1819 (2008)
S7.
J. R. Green and D. Margerison, Stastistical treatment of experimental data,
Elsevier,Amsterdam, 1978, 1st rev. reprint
S8.
R. Mancinelli, A. Botti, F. Bruni, M. A. Ricci, A. K. Soper, Phys. Chem. Chem. Phys.,
9, 2959 (2007)
S9.
S. Rajam, et al. J. Chem. Soc. Faraday Trans., 87, 727 (1991)
S10.
E. Loste, E. Diaz-Marti, A. Zarbakhsh, F. C. Meldrum, Langmuir, 19, 2830 (2003)
S11.
E. DiMasi, M. J. Olszta, V. M. Patel, L. B. Gower, CrystEngComm, 5(61), 346 (2003)