J. Phys. Chem.

B 2000, 104, 1539-1545 1539

Atomic Force Microscopy under Defined Hydrodynamic Conditions: Three-Dimensional

Flow Calculations Applied to the Dissolution of Salicylic Acid

Shelley J. Wilkins, Marco F. Suárez, Qi Hong, Barry A. Coles, and Richard G. Compton*
Physical and Theoretical Chemistry Laboratory, Oxford UniVersity, South Parks Road,
Oxford OX1 3QZ, United Kingdom

George E. Tranter and David Firmin

Physical Sciences, GlaxoWellcome Medicines Research Centre, Gunnels Wood Road, SteVenage,
Hertfordshire SG1 2NY, United Kingdom
ReceiVed: August 12, 1999; In Final Form: October 27, 1999

The kinetics of the dissolution of salicylic acid in aqueous solutions is studied using in situ atomic force
microscopy, using a novel liquid flow cell for which the full three-dimensional flow pattern is known. This
allows the interpretation of dissolution rate in terms of an interfacial reaction mechanism, with excellent
agreement between the theoretically predicted and the experimental results. The use of a three-dimensional
simulation to obtain the flow velocities enables accurate prediction over a much wider range of flow rates
than is possible using a simpler two-dimensional model for the flow pattern. The dissolution of the (110)
face of salicylic acid in the presence of water and aqueous solutions containing sodium chloride has been
studied as a function of flow rate and is found to be consistent with a model combining a constant rate of
dissolution with some redeposition having a first-order dependence on the surface concentration [SA]0, with
the flux J ) kf - kb[SA]0. The parameters for kf are found to be 2.04 × 10-8, 1.65 × 10-8, and 8.85 × 10-9
mol cm-2 s-1 for dissolution in water and 0.1 M and 1 M sodium chloride, respectively, at a cell temperature
of 21 °C.

Introduction
The novel design of a flow cell which allows AFM to be
carried out under defined flow conditions has been described
in recent publications.1,2 This cell is based on the Topometrix3
liquid immersion cell, with the addition of an inlet tube shaped
to allow a jet of fluid to be applied directly to the sample surface,
flowing along the front of the cantilever support chip and over
the cantilever. The rate of interfacial reaction is measured
directly by averaging the absolute height of the imaged surface
as indicated by the z-piezo voltage and by following the changes
of this height obtained from successive scans. This cell
established that AFM could be used to make quantitative
determinations of heterogeneous reaction rates, discriminate
between proposed interfacial reaction mechanisms and provide
information about binding sites,4 analyze surface dissolution and
passivation,5 and identify the mechanisms of dissolution.6
The flow pattern in this cell is complex, and in the initial
work, this was solved using an approximate two-dimensional
simulation with a finite-element fluid dynamics program. Three-
dimensional simulations of the flow patterns have now been
completed, using the FIDAP 7.62 program7 on the Columbus
Superscalar service at the Rutherford Appleton Laboratory.
We have applied the velocity data from these simulations to
investigate the dissolution processes of the (110) face of salicylic
acid exposed to aqueous solutions of flows ranging from 2.5 to
17.4 µL s-1 corresponding to centerline jet velocities of 2.1-
14.8 cm s-1. The use of the 3D model is shown to permit a
precise description of the interfacial rate law.
* To whom all correspondence should be addressed. E-mail:
richard.compton@chemistry.oxford.ac.uk.
10.1021/jp992858k CCC: $19.00 © 2000 American Chemical Society
Published on Web 01/28/2000
Theory
Flow Simulation. FIDAP7 is a fluid dynamics program which
uses finite-element methods to solve flow problems; these may
be steady-state or transient and may include the effects of
temperature and chemical species. The flow region is divided
into elements, connected together at nodes, and appropriate
boundary conditions may be specified for those nodes which
are on an external surface. The equations of conservation of
momentum, mass, and energy are solved for each element, and
the values for velocities, pressure, temperature, and concentra-
tions at each node are obtained. The volume simulated is shown
in outline in Figure 1, with dimensions 0.414 cm × 0.09 cm ×
0.1188 cm, and contained 80 794 nodes. The sloping section at
the left represents the inside volume of the jet tube, which had
a square cross section; the bottom of the volume represents the
sample surface, and the AFM scanning cantilevers project into
the volume from the back.
Figure 1 shows an example of the particle path trajectories,
starting from a vertical set of coordinates inside the jet tube, as
the solution flows down the tube and then spreads out into the
cell, for a fairly low flow rate. At higher flow velocities the jet
of solution flows more closely along the sample surface with a
smaller degree of spreading.
For the simulation of chemical processes, it is not necessary
to consider velocities remote from the sample, so data files
have been prepared containing the x, y, and z components of
velocity at 35 769 points in the smaller volume which starts
from the mouth of the jet and encloses the scanning canti-
levers (Figure 2), for 14 flow rates encompassing the practicable
range of the cell. It should be noted that the crystal surface
1540 J. Phys. Chem. B, Vol. 104, No. 7, 2000 Wilkins et al.

BIFD calculations. The derivatives in eq 1 are approximated to

∂[C] [C]j,k - [C]j,k-1

) (2)
∂x ∆x
∂[C] [C]j+1,k - [C]j-1,k
) (3)
∂y 2∆y
∂2C [C]j+1,k - 2[C]j,k + [C]j-1,k
) (4)
∂y2 (∆y)2
Combining eqs 1-4, we obtain
Figure 1. Schematic view of the simulated portion of the AFM
cell, showing the jet tube, at the left, and particle path trajectories as Cj+1,k - 2Cj,k + 2Cj-1,k Cj,k - Cj,k-1
the solution spreads out into the cell. The sample surface is at the D - Vx -
(∆y) 2 ∆x
bottom.
Cj,k - Cj-1,k
Vy ) 0 (5)
should not normally extend back into the jet itself due to ∆y
the high velocities and gradients present. The surface should
be inert in this position by coating part of the crystal or The equation can be rewritten as
embedding the crystal in an inert matrix. These data files are
λVxCj,k-1 + (λy + λVy)Cj-1,k - (2λy + λVx + λVy)Cj,k +
now available and may be freely downloaded8 from the World
Wide Web. λyCj+1,k ) 0 (6)
Simulations of chemical processes in the cell were carried
where
out by finite difference methods in two dimensions and required
a much finer mesh than is normally used for flow simulation. D Vx Vy
To assist with this, two Fortran-77 programs were prepared. λy ) , λVx ) , and λvy )
(∆y) 2 ∆x ∆y
The first of these finds the solution flow path which passes over
the scanning tip position, and it generates a two-dimensional If a ) -(λy + λVy)/λVx, b ) (2λy + λVx + λVy)/λVx, and c )
grid file of velocities in a surface which follows the flow path. -λy/λVx, it follows that
The second program performs a bicubic spline interpolation from
this two-dimensional grid to provide the velocity components Cj,k ) aCj-1,k+1 + bCj,k+1 + cCj+1,k+1
in a user-specified grid which may contain any number of points. (7)
Further details are available in a paper9 published on the World Cj,k-1 ) aCj-1,k + bCj,k + cCj+1,k
Wide Web, which may be downloaded along with the programs
and the data files. Equation 7 is the final general equation for BIFD simulation.
Backward-Implicit Finite Difference Simulation. The mass The BIFD method only involves vector calculations. The vectors
transport from the crystal surface to the bulk solution is describe concentrations in the y direction for different values
described by the following time-dependent convective diffusion of x. The calculation proceeds downstream, each vector enabling
equation at steady state: the calculation of the next, starting from the vector defining
the boundary conditions specified upstream of the reacting
crystal of the concentration at the mouth of the inlet jet.
∂[SA] ∂2[SA] ∂[SA] ∂[SA] The boundary conditions used were as follows:
) DSA 2
- Vx(x,y) - Vy(x,y) )0
∂t ∂y ∂x ∂y
(1)
D (∂C∂y )
J,k
) 0, CNJ-1,k ) CNJ,k
where x is the direction of the flow over the crystal surface, y
is the direction perpendicular to the surface, DSA is the diffusion Thus,
coefficient of salicylic acid, and Vx and Vy represent the solution CNJ-1,k-1 ) aCNJ-2,k + bCNJ-1,k + cCNJ,k
velocities in the x and y directions, respectively, at the point
(x,y). We assumed that salicylic acid is in equilibrium with the CNJ-1,k-1 ) aCNJ-2,k + (b + c)CNJ-1,k
salicylate ion and H+. Experimental data were compared with
different theoretical models using rate constants as adjustable At the solid salicylic acid surface, assuming that salicylic acid
parameters. dissolves from the solid at a constant rate and partly precipitates
To apply the backward implicit finite difference (BIFD) on the surface,
method, the xy plane is covered with a two-dimensional finite
difference grid. Increments in the x direction are ∆x and in the
y direction are ∆y.
-D [∂C∂y ] j)0
) kf - kbC0,k

∆y kb∆y
yj ) j∆y j ) 0,1, . . . J where ∆y ) 2h/J C1,k - C0,k ) - k + C
D f D 0,k
xk ) k∆x k ) 0,1, . . . K where ∆x ) (xe - xs)/K DC1,k + kf∆y
C0,k )
D + kb∆y
where xs and xe are the start and finish coordinates of the BI
grid and 2h is the maximum distance from the surface for the where kf and kb are rate constants for dissolution and precipita-
Dissolution of Salicylic Acid J. Phys. Chem. B, Vol. 104, No. 7, 2000 1541

Figure 2. The solution volume included in the velocity data files.

tion, respectively.

C1,k-1 ) aC0,k + bC1,k + cC2,k

[ ]
akf∆y aD
) + + b C1,k + cC2,k
D + kb∆y D + kb∆y

( ) ( )( )
These simultaneous equations may be expressed as a (J -
1) × (J - 1) matrix equation:

{d} ) [T]{u}
d1 b(1) c(1) u1,k
d2 a2 b2 c2 u2,k Figure 3. Salicylic acid crystal habit after growth from a saturated
· · · · solution of salicylic acid in ethanol.
· · · ·
· · · ·
· · · · single crystal of salicylic acid was fixed in the flow cell exposing
) · ·
· · · · the (110) face for imaging. Solution was flowed through the
· · liquid cell using various flow rates ranging from 0.0025 to
dj a(j) b(j) c(j) uj,k
· · 0.0174 cm3 s-1. A series of images of the dissolving surfaces
· · scanning an area of 20 µm × 20 µm was then recorded
dNJ-1 · a· b(NJ-1) uNJ-1,k
(NJ-1) continuously at ca. 60 s intervals. In addition to conventional
where
topographical images, plots of the absolute z-piezo voltage were

[ ]
akf∆y
aD
d1 ) C1,k-1 ) + + b C1,k + cC2,k
D + kb∆y D + kb∆y
dj ) Cj,k-1 ) aCj-1,k + bCj,k + cCj+1,k
dNJ-1 ) CNJ-1,k-1 ) aCNJ-2,k + (b + c)CNJ-1,k
The tridiagonal form of the matrix enables the Thomas
algorithm10 to be used. The boundary condition Cj,0 ) Cinlet
supplies the vector {d}0 from which {u}1 is calculated. In the
absence of homogeneous chemical complications, {d}k ) {u}k,
so {u}k is calculated from {d}k-1 and so on until {u}K is
obtained. Thus, all the values Cj,k (j ) 1, 2, . . . J - 1, k ) 1,
2, . . . K) are evaluated.
Materials and Methods
A Topometrix TMX 2010 Discoverer atomic force micro-
scope operating in contact mode with SFM probes type 1520-
00 and a 75 µm scanner type 5590-00 was used to obtain the
images. The novel flow cell described earlier1,2 was used for in
situ AFM imaging. The rates of dissolution of the surfaces of
salicylic acid were measured using the following procedure. A
recorded. The change of the mean piezo voltage calculated from
each scan and therefore the change of the mean surface level
can be derived using the calibration factor 0.333 V µm-1. Note
that, although the thickness of the crystal is changing, the piezo
support moves to maintain the top surface of the crystal at a
constant height since the cantilever deflection is being main-
tained constant by the AFM feedback. The flow geometry and
velocities therefore do not change as dissolution proceeds. The
flow system was gravity fed with solution flowing from a
reservoir to the flow cell then to an outlet whose height was
adjusted relative to the reservoir. Measurements were taken at
a temperature of 21 °C.
Salicylic acid crystals were grown from a saturated solution
in ethanol by gradually decreasing the temperature from 30 to
25 °C over 200 h, yielding large numbers of good quality
crystals averaging 3 mm × 4 mm × 30 mm in size (Figure 3).
Salicylic acid crystals were indexed and the crystal faces
assigned using standard methods with a ENRAF NONIUS FR
590 diffractometer using Mo KR radiation.
The total solubilities of salicylic acid in water and solutions
containing 0.1 M and 1 M NaCl, respectively, were determined
by UV spectrophotometric analysis. An excess of salicylic acid
1542 J. Phys. Chem. B, Vol. 104, No. 7, 2000
Figure 4. In situ images of the (110) surface of salicylic acid during the dissolution process in 0.1 M NaCl flowing at a rate of 2.5 µL s-1.
Sequence after (a) 0, (b) 118, and (c) 236 s.
Figure 5. In situ images of the (110) surface of salicylic acid during the dissolution process in 0.1 M NaCl flowing at a rate of 12.5 µL s-1.
Sequence after (a) 0, (b) 59, and (c) 118 s.
was added to 10 mL flasks of each of the solutions. The flasks
were agitated for 24 h before removing a small aliquot from
each flask and centrifuging to separate the excess salicylic acid
from solution. The solutions were diluted 100 fold with
phosphate buffer of pH 7.429, and the solubilities were
calculated from the UV absorption spectra. The solubilities were
found to be 14.9 mM in water, 13.88 mM in 0.1 M NaCl, and
10.79 mM in 1 M NaCl. The diffusion coefficient for salicylic
acid was calculated to be DSA ) 7.1 × 10-6 cm2 s-1 using the
Wilke-Chang equation,11 and the dissociation constant of
salicylic acid is 1 × 10-3 M at 25 °C. The solubility values
used for the mathematical simulation were the concentration
of undissociated salicylic acid in a saturated solution, calculated
from the known solubility and the dissociation constant.
Solutions were made up using triply deionized water of
resistivity > 107 Ω cm (from an Elgastat system, High
Wycombe, UK). The chemicals used were salicylic acid 99+%
and sodium chloride 99.9% (BDH Ltd., Poole, Dorset). The
buffer solution for solubility measurements contained 0.0303
M Na2HPO4 and 0.0068 M KH2PO4 (BDH Ltd., Poole, Dorset).
Results and Discussion
Crystals of salicylic acid were imaged under a range of flow
conditions using the modified AFM liquid cell. AFM dissolution
experiments of the (110) plane were conducted using flow rates
of 2.5, 7.5, 12.5, and 17.4 µL s-1 for dissolution in pure water
and solutions of 0.1 M and 1 M NaCl. The rate of dissolution
was determined for the different flow rates by monitoring the
change in the mean z-piezo voltage from successive scans at
59 s intervals. During the course of each experiment some
erosion of the crystal occurred in the jet region. To prevent
interference of this erosion in the flow dynamics, the crystal
was shifted in the z direction between each sequence of images.
The (110) face was found to be flat with minor defects
typically less than 150 nm in height before exposure to the
aqueous solutions. Naturally grown faces were used; it was not
found possible to prepare faces by cleavage due to the fragile
nature of the crystals. Exposure to the three solutions at slow
flow rates was found to induce the formation of small pyramidal
growths on the surfaces, as demonstrated in Figure 4. Analysis
Wilkins et al.
of the z-piezo voltage plots showed the crystal height steadily
decreasing due to dissolution, with the small growths increasing
in height before falling again as dissolution proceeded. At slow
flow rates, these growths are observed where the salicylic acid
has dissolved in the region close to the jet then precipitated on
the surface downstream. The orientation and shape of these
growths are not clear from these images since the growths may
be sharper than the pyramidal AFM probes. However, the
growths typically form in areas of defect on the surface such
as steps and pits.
At higher flow rates, the small growths were absent and the
sequences of images clearly showed the formation of steps and
terraces on the crystal surface during dissolution, as shown in
Figure 5. The steps and pits grow in two directions, the most
rapid dissolution occurring in the (001) direction with steps also
dissolving perpendicular to the (001) direction but at a slower
rate. The height, size, and velocity of retreat varied considerably
between experiments; however, the overall dissolution of the
crystal was found to vary linearly with time. An example of
this is shown in Figures 6 and 7. In Figure 6, the major step
height in the first image is approximately 3200 nm; the step
height of the major step in Figure 7 is approximately 800 nm.
The larger step can be seen to retreat more slowly than the
smaller step. However, the overall flux remains linear since the
amount of salicylic acid dissolved from the surface is similar
in each case.
Experiments with the two NaCl solutions showed a suppres-
sion in the rate of dissolution compared with the dissolution in
water. The dissolution rates for the salicylic acid (110) plane
are shown as a function of the solution flow rate together with
the theoretical curves in Figure 8. The dependence of the flux
from the flow rate is a function of the heterogeneous rate
constants kf and kb. The relationship between solubility and the
rate constants is shown in eq 8.
kf
[SA]sol ) (8)
kb
Therefore, when the solubility is known, the relationship
between flux and flow rate is only dependent on one of the
heterogeneous rate constants and can be determined by the best
Dissolution of Salicylic Acid J. Phys. Chem. B, Vol. 104, No. 7, 2000 1543

Figure 6. In situ images of the (110) surface of salicylic acid during the dissolution process in 1 M NaCl flowing at a rate of 2.5 µL s-1. Sequence
after (a) 0, (b)118, (c) 236, and (d) 354 s.

Figure 7. In situ images of the (110) surface of salicylic acid during the dissolution process in 1 M NaCl flowing at a rate of 2.5 µL s-1. Sequence
after (a) 0, (b) 118, (c) 236, and (d) 354 s.

fit of the experimental curves by the model described. The and 8.85 × 10-9 mol cm-2 s-1 at 21 °C for dissolution of
parameters for kf were found to be 2.04 × 10-8, 1.65 × 10-8, salicylic acid in water, 0.1 M NaCl, and 1 M NaCl, respectively.
1544 J. Phys. Chem. B, Vol. 104, No. 7, 2000 Wilkins et al.

Figure 8. Plot of dissolution rate vs flow rate for the dissolution of the (110) face of salicylic acid in water (9), 0.1 M NaCl ([), and 1 M NaCl
(b) showing the theoretical curve using the 3D simulation compared to the experimental data.

Figure 9. Plot of dissolution rate vs flow rate for the dissolution of the (110) face of salicylic acid in water (9), 0.1 M NaCl ([), and 1 M NaCl
(b) showing the theoretical curve using the 2D simulation compared to the experimental data.

Figure 9 shows, for comparison, the theoretical curves using
a 2D simulation of the flow velocities. This gives less good
agreement to the experimental data at low flow rates and hence
demonstrates the importance of using a full 3D simulation for
the flow.
As the concentration of sodium chloride in the flow cell is
increased, there is a reduction in the rate of dissolution. The
3D theoretical model is in good agreement with the experimental
data.
Conclusion
It has been shown that an accurate calculation of the
dissolution flux of salicylic acid in the flow cell under different
flow conditions can be made using a 3D finite element fluid
dynamics simulation of the complex flow field. The use of the
novel flow cell reveals not only information about reaction rate
constants of solid/liquid reactions but the topographic images
also provide details of mechanisms of dissolution. This is of
considerable importance for example to pharmaceutical com-
panies in understanding the fundamental factors governing the
dissolution behavior of pharmaceutical drugs under different
conditions. The merits of hydrodynamic AFM for quantitative
studies of interfacial kinetics is evident; the only limitations of
the method arise since, as the upstream zone of the crystal
influences the kinetic behavior in the downstream imaged area
at low flow rates, it is necessary to assume that the interfacial
reaction mechanism remains unchanged over the full length of
the sample.
Acknowledgment. We thank GlaxoWellcome and BBSRC
for support for S.J.W. via a CASE award, COLCIENCIAS for
a scholarship for M.F.S., and EPSRC for the grant of super-
computer facilities (Grant GR/L34167).
Dissolution of Salicylic Acid J. Phys. Chem. B, Vol. 104, No. 7, 2000 1545

References and Notes
(1) Coles, B. A.; Compton, R. G.; Booth, J.; Hong, Q.; Sanders, G. H.
W. Chem. Comm. 1997, 619.
(2) Coles, B. A.; Compton, R. G.; Suárez, M.; Booth, J.; Hong, Q.;
Sanders, G. H. W. Langmuir 1998, 14, 218.
(3) Thermo Microscopes Ltd., Bicester, Oxon, UK (http://www.ther-
momicroscopes.com).
(4) Hong, Q.; Suárez, M.; Coles, B. A.; Compton, R. G. J. Phys. Chem.
B 1997, 101, 5557.
(5) Booth, J.; Compton, R. G.; Atherton, J. H. J. Phys. Chem. B 1998,
102, 3980.
(6) Suárez, M.; Compton, R. G. J. Phys. Chem. B 1998, 102, 7156.
(7) Fluent Europe Ltd., Sheffield, UK (http://www.fluent.com).
(8) ftp://joule.pcl.ox.ac.uk/pub/rgc/afmvmaps/.
(9) Coles, B. A.; Compton, R. G. http://physchem.ox.ac.uk:8000/
research/afm/afmflow.shtml.
(10) Lapidus, L.; Pinder, G. F. Numerical Solution of Partial Differential
Equations in Science and Engineering; Wiley: New York, 1982.
(11) Wilke, C. R.; Chang, P. AIChE J. 1955, 1, 264.