CMAC P O ST E R CO L L EC T I O N

POSTER 39 POSTER 40

Reactive Crystallisation of
Real-Time X-ray Imaging of
Benzaldehyde Sodium Bisulfite
Christopher McArdle Crystallisation Processes
Project Aim Lab Work Oliver V. Towns1*, Sven L. M. Schroeder1, Joanna Leng1, Tariq Mahmud1, Maximilian Besenhard1, Helen Wheatcroft2,
Anna Jawor-Baczynska2
The aim of this project is to use in-situ measurement techniques for real-time 1 Chemical and Process Engineering, University of Leeds, Leeds, UK
monitoring of reactive crystallization systems using bisulfite addition to
Synthesis of Benzaldehyde sodium 2 Chemical Development, Pharmaceutical Technology & Development, Operations, AstraZeneca, Macclesfield, UK
benzaldehyde as a model system. The objectives are therefore as follows:
bisulfite
Benzaldehyde (0.09 mol, 9.6 mL) was *cm15ovt@leeds.ac.uk
• To evaluate spectroscopic and particle characterization techniques for measuring
mixed with 83 mL of ethanol, and sodium

Introduction
the required process and product attributes.

• Investigate the impact of experimental parameters on the product attributes and

bisulfite powder (0.9 mol, 9.4 g) was
dissolved in 35 mL of water. The Experimental Setup
benzaldehyde/ethanol mixture was placed
determine kinetic information.
in a beaker and a pH and temperature
probe were placed in it. The sodium
Crystallisation is a hugely important process,
• Devise a workflow for the development of reactive crystallization systems and
bisulfite powder was then added to the involved in the manufacture and purification of many
demonstrate its application in an industrial case study with GSK.
mixture resulting in the benzaldehyde pharmaceuticals. Having better process control may
sodium bisulfite “crashing” out. The pH
and temperature changes were then achieve considerable efficiency gains, both in terms
Background recorded, the precipitate filtered, washed of time and physical resource savings. Having more
with a small amount of ethanol, dried in a
Formation of the bisulfite adduct (product) occurs via nucleophilic addition of the vacuum oven, and then weighed again.
Figure 2: Experimental Set-up accurate models that predict particle size and shape
bisulfite ion to the aldehyde or ketone carbonyl carbon as shown in figure 1 [1]. This is therefore key. The goal of the project is to create
reaction has multiple uses that make it particularly useful: Table 2: Results From Initial Experiments
Parameter Value
an online imaging technique that can glean 3D
• Purification of aldehydes: the resulting bisulfite adduct can undergo further information of crystal growth, that can better
Conversion • First reaction: 92.35%
reaction to regenerate the aldehyde [2]. • Second reaction: 96.08% influence industrial population balance models.
• Third reaction: 92.29%
• Storage of aldehydes: aldehydes typically suffer from stability issues (i.e. Temperature • Starting point: 20.4 ºC
Change • End point: 33.9 ºC
degradation or oxidation) [3]. Reacting the aldehyde with bisulfite will produce a

X-ray Phase Contrast Imaging

pH Changes • Starting point: 5.34
crystalline solid which has a better stability. • Reaction begins: >8
• Decreases back to 5.89
• Drug delivery: the bisulfite adduct is soluble in water, which means it can be • Slowly increases to an end point of

delivered to patients by methods other than orally (i.e. intravenously) [1].

6.09 Figure 3: Benzaldehyde Sodium
Bisulfite
Many pharmaceutical crystals are poor absorbers of
Preliminary Solubility Data
The bisulfite addition reaction is a complex equilibrium scheme that is dependent on The solubility of the product must be determined prior to determining
X-rays. Phase contrast imaging allows them to be
the pH of the overall system. This is due to the multiple protonation and solubility curves (see future work). A known mass of benzaldehyde visually captured. Taking advantage of coherent
deprotonation reactions that can occur. sodium bisulfite was placed inside a beaker with a magnetic stirrer and synchrotron X-rays, at a phase boundary such as the Figure 3. Experimental set-up plan.
set on top of a hotplate set to 1000 rpm. The solvent was added step-
interface between a crystal and liquid, phase shifting
0\ sulfurous acid
wise and left for 1 minute between each addition to allow for
dissociation to occur. When the solution has fully clear, the mass of occurs which causes interference. This presents itself
The crystallisation vessel will be cooled to induce

u addition

sulfurous acid
elimination
solvent was measured and recorded. The results of this are shown in
table 3. Table 3: Preliminary Solubility Data
Solvent Benzaldehyde Sodium Sodium Bisulfite
as enhanced contrast at the edge of a crystal.
Phase shift
nucleation and growth. The peristaltic pump will
circulate the crystal suspension so they flow through
the X-ray beam.
Composition Bisulfite (g/kg) (g/kg)
(Water:Ethanol w/w)

l 100 % Water 19.79 (22.1 oC) 436.00 (19 oC) Coherent waves

Algorithm
pK, = 1.77 - 1.84
65/35 5.00 (25.4 oC) 135.24 (20.7 oC)
Crystal
0\ bisulfite
HOi~~O
_o
50/50
35/65
3.73 (23.7 oC)
1.84 (23.1 oC)
34.24 (21.3 oC)
9.37 (22.7 oC)
Object detection algorithms have been created that
u addition

bisulfite
elimination
u
Figure 1. Crystal causes initially coherent x-rays to become phase shifted
Images can be taken at intervals to create dynamic
videos of processes or the sample can be rotated with
can automatically separate crystals from the
background.
pH 3- 7
Future Work images taken in steps which can then be Canny Threshold Label
Gaussian
reconstructed as 3D tomography. Image
l pK, = 6.83 - 7.20 ll pK, = 9.16 - 10.68 • Validation of final product: using NMR to determine what is in the
filtrate and the precipitate
Blur
edge
detection
and area
filter image

_o Figure 4. Current work-flow of the object detection algorithm

0\ • Amorphous or crystalline?: using PXRD to check the solid form of
Paganin Filter
u
sulfite
addition

sulfite
elimination 6''
Figure 1: Bisulfite Equilibrium Scheme [4]
pH 10 -1 2
the sample.

• Unreacted sodium bisulfite: PXRD will be used to determine where

unreacted sodium bisulfite ended up, i.e., in the filtrate or the
precipitate.
Paganin et al. derived the below equation which allows
thickness to be extracted from phase contrast images
The algorithm currently relies on simple edge
detection. This is due to XPCI having low contrast
apart from at the edges of boundaries.

The equilibria of this reaction is well-characterized in the literature (see table 1), • Solubility data: using the preliminary solubility data as a basis, (1).
with papers as far back as the 1930s, however, the nucleation and crystal growth of solubility curves will be constructed using the Crystal-16
these compounds have not been discussed. This research aims to understand the equipment via measuring the clear point temperature.
crystallisation kinetics and use it as a basis for gaining a greater understanding of
reactive crystallisation systems in general. • Applying PAT to the system: after gaining a better understanding of
the crystallisation kinetics, PAT will be used to monitor and control
Table 1: Equilibrium and Reaction Rate Constants of the pH-dependent
Reaction Scheme from the Literature the product attributes. When used, 2D information can be obtained like in a
Reaction Equilibrium Constant (M-1) Reaction Rate
normal video, but also 3D information can be extracted
(M-1 s-1) which can be used to influenced more accurate models.
Sulfurous
Acid Not found 0.097 (21oC) [6]
Acknowledgments Figure 5. Left: Canny edge detection of crystals. Right: Labelled output image
Addition
6.4 x 103 (𝜇𝜇 = 1.0 M, pH 3.55-5.27, 294 K, 5 v/v% EtOH in H2O) [5]
Supervisors: Alison Nordon and Jan Sefcik (University of Strathclyde); Joe The algorithm has real-time controls such as
Harris, Ali Saleemi and Andrew Dominey (GSK).
Bisulfite
1.4 x 104 (no pH given, 286 K, pure H2O) [6]
4.7 x 103 (no pH given, 296 K, pure H2O) [6] 0.43 (21oC) [4]
minimum area of detection. In the left image there
Addition 1.9 x 103 (no pH given, 306 K, pure H2O) [6]
1.7 x 104 (no pH given, 273 K, pure H2O) [7]
254 (13oC) [4]
0.71 (𝜇𝜇 = 1.0 M, 25oC) [6]
Funding: CAMS, CMAC, CPACT, and the University of Strathclyde. are small white areas that aren’t labelled in the right
6.57 x 102 (no pH given, 293 K) [7]
4.37 x 102 (no pH given, 303 K) [7] Affiliations: CMAC Hub and the Department of Pure and Applied Chemistry
image due to insufficient area.
Sulfite 1.25 x 104 (21oC) [4]
Addition 9.12 x 10-1 (𝜇𝜇 = 1.0 M, pH 8-12.6, 294 K, 5 v/v% EtOH in H2O) [5] 2.15 x 104 (𝜇𝜇 = 1.0 M, 25oC)
[6]
References The algorithm can extract size information of crystals
[1] - J. Clayden, N. Greeves, S. Warrens and P. Wothers in Organic Chemistry, New York,, 2001, pp. 148-149. such as height, width and area and also shape
[2] - D. Blackadder and C. Hinshelwood, Journal of the Chemical Society, pp. 2720-2727, 1958.
[3] - M. Kissane, S. Frank, G. Rener, C. Ley, C. Alt, P. Stroud, R. Vaid, S. Boini, L. McKee, J. Vicenzi and G. Figure 2. Left: Original phase contrast image. Right: Paganin filter applied information such as geometry and convexity.
.
Stephenson, Tetrahedron Letters, vol. 54, no. 48, pp. 6587-6591, 2013.
[4] - L. H. Donnally and T. D. Stewart, Journal of the American Chemical Society, vol. 54, pp. 3559-3569, 1932.
[5] - F. C. Kokesh and R. E. Hall, The Journal of Organic Chemistry, vol. 40, no. 11, pp. 1632-1636, 1974.
[6] - T. M. Olson, S. D. Boyce and M. R. Hoffman, The Journal of Physical Chemistry, vol. 90, pp. 2482-2488, 1986. 1. Paganin, D., Mayo, S.C., Gureyev, T.E., Miller, P.R. and Wilkins, S.W. Simultaneous phase and amplitude extraction from a single defocused image of a homogeneous
[7] - Y. Ogato and A. Kawasaki in The Carbonyl Group, John Wiley & Sons Ltd., 1970, pp. 33-37.
object. Journal of Microscopy. 2002, 206: 33-40.

40 W W W. C M A C . A C . U K 41