An efficient laboratory automation concept for process chemistry

Article abstract of DOI:10.1021/op900154c

This concept article presents and outlines a successful strategy to roll out an automation concept on a broad basis to AstraZeneca Process R&D laboratories. Examples of hardware and software are presented as well as a couple of examples related to process safety.

Full text of DOI:10.1021/op900154c

Organic Process Research & Development 2009, 13, 1059–1067
Concept Articles
An Efficient Laboratory Automation Concept for Process Chemistry
Christian Bernlind* and Csaba Urbaniczky
Global Process R&D, Process Chemistry, AstraZeneca, SE-151 85 So¨derta¨lje, SWEDEN
Abstract:
the industry to be able to smoothly pass the expected “paradigm
shift” within drug development.5
This concept article presents and outlines a successful strategy to
roll out an automation concept on a broad basis to AstraZeneca
Process R&D laboratories. Examples of hardware and software
are presented as well as a couple of examples related to process
safety.
Background
Around the start of the year 2000 the market for commercial
robotized systems experienced a boost. Similarly to many
companies, we walked down the aisle with a few robotized
systems and automated reactor platforms for parallel reactions
supplied by well-known manufacturers. The objective was to
increase efficiency both in order to save time and to increase
capacity to meet an increasing flow of candidate drugs from
our discovery units. The systems were widely employed in a
diversity of investigations during the first years. Although the
benefits of these systems, such as the ability to take many
samples and to perform unattended experiments remain unchal-
lenged, the overall impact on our development work was
limited. Compared to performing the same work manually, we
perceived that there were no clear gains in time or efficiency
when the full cycle time for an investigation was taken into
account. The automated systems still required a lot of previously
acquired knowledge about the chemistry and the physical
properties of the starting materials and reagents, (e.g., solubility,
stability, dissolution times, suitable sample concentrations as
well as robust and universal cleaning procedures for needles
and containers). Even an expert user would have to prepare
him or herself for a few wasted experiments before the
equipment could deliver useful results. This proved particularly
unfortunate for early screening work, where the supply of
starting material was minimal, as was the tolerance towards
failure. The overall conclusion was that these automated systems
could very well be used for homogeneous and diluted reagent
screens at moderate temperature ranges, but could generally not
be recommended for screens or optimisation studies where
factors such as concentration, stirring, prolonged addition times
or extreme temperatures were under investigation.
Introduction
Today, automated equipment plays an important role in
almost every process chemistry laboratory. The use of, for
example, HPLC autosamplers and diode-array detectors, NMR
spectroscopy, literature search tools and, to some extent, also
electronic lab notebooks have revolutionised planning, analysis
and documentation of chemical experiments in the pharmaceuti-
cal industry.1 Automation has also replaced some of the manual
work that is carried out in Process R&D laboratories. Automatic
weighing stations and liquid dispensers have the potential to
carry out tedious and repetitive work, leaving the chemists free
to perform more creative activities. Simple reaction blocks in
combination with stirring bars and test tubes are reliable
workhorses in the study of reactions, solubility and stability.
Automatic synthesizers and workstations for combinatorial
chemistry and high-throughput screening are today widely used
within medicinal chemistry departments to find new candidate
drugs.2 Automated systems for Process R&D work have also
been available on the market for many years,3,4 but since the
systems generally are very expensive and require dedicated
expert users, only a limited number of them are likely to have
found their way into the average Process R&D laboratory.
Hence, in order to provide lab automation to every single
Process R&D chemist, it is clear that an additional and
complementary approach is needed. A wide and universal use
of automation and logging of experimental data will be key to
In parallel to the evaluation of the commercial systems
another strategy was initiated at our department in 2001 with
the primary aim being to facilitate and integrate laboratory
automation in the preferred ways of working. The project was
* Author to whom correspondence may be sent. E-mail: christian.bernlind@
astrazeneca.com.
(1) Martin, V. Everything old is new again. Developing a new paradigm
for process research through application of new technologies. In
Proceedings of the Siegfried Symposium; Siegfried Symposium, Zurich,
CH, October 14, 2004.
(5) Federsel, H.-J. Drug News Perspect 2008, 21 (4), 193.
(6) Urbaniczky, C.; Bernlind, C. Oscar Wilde for the common chemist.
In Proceedings of the 8th International Conference & Exhibition:
Laboratory Automation. Faster Process DeVelopment - New Tools for
New Challenges; Scientific Update:Mayfield, U.K. 8th International
Conference & Exhibition: Laboratory Automation. Faster Process
Development - New Tools for New Challenges, London, U.K.,
September, 13-14, 2005.
(2) (a) Coates, W. J.; Hunter, D. J.; MacLachlan, W. S. Drug DiscoVery
Today 2000, 5 (11), 521. (b) Reader, J. C. Curr. Top. Med. Chem.
2004, 4 (7), 671.
(3) Rubin, A. E.; Tummala, S.; Both, D. A.; Wang, C.; Delaney, E. J.
Chem. ReV. 2006, 106 (7), 2794.
(4) For a comparison of three major brands see: Van Loo, M. E.;
Lengowski, P. E. Org. Process Res. DeV. 2002, 6 (6), 833.
10.1021/op900154c CCC: $40.75  2009 American Chemical Society
Published on Web 09/23/2009
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1059

Table 1. Example of RS232-controlled laboratory equipment
equipped with output channels that can be used to control
external relays and even generate programmable current and
voltage signals suitable for analog control of external devices.
In our view, the data logger equipped with a Pt100 temperature
sensor, together with a heater/chiller unit and a syringe pump,
forms a powerful base when it comes to logging and controlling
process parameters and being able to evaluate and understand
the ways our processes evolve.
An immediate issue was to ensure that several laboratory
devices could be conveniently connected to a single standard
computer. Consequently, a main priority became to increase
the number of serial communication ports. Most modern desktop
computers only have one or possibly two serial ports, and some
laptops even lack them completely. For this purpose, a range
of different USB-to-serial adapters was tested and evaluated in
terms of stability and robustness in order to increase the number
of comm ports up to a maximum of 16. Adapters that had
control chips and drivers manufactured by FTDI8 proved to
perform best at the time of testing. These chips power several
commercial USB-to-serial adapters from different manufactur-
ers. The USB-to-serial adapter was mounted inside the fume
hood and all laboratory devices were plugged into this adapter.
Only a single cable was then required between the fume hood
and the computer. For a photograph showing an example of a
typical Oscar Wilde arrangement for controlling and logging a
single reactor see Figure 1.
device type
model
manufacturer
data logger
Almemo 2390-5/8, 2590-2/3/4/9 Ahlborn Mess- and
Regelungstechnik
overhead stirrer IKA Power Control Visc
Heidolph RZR2051
IKA Werke
Heidolph Instruments
heater chiller
Julabo F, FS, FP, CF, Presto
Haake Phoenix II, DC50, F6
Huber ministat, unistat
Julabo Labortechnik
Thermo Fisher Scientific
Peter Huber
Ka¨ltemaschinenbau
syringe pump
Harvard pump 11
Harvard Apparatus
Metrohm
liquid dispenser Metrohm Dosimat
peristaltic pump Ismatec Reglo Digital
Ismatec SA Labortechnik
Dr. Ing. Herbert Knauer
KNF Neuberger
piston pump
Knauer Pump 100
diaphragm pump STEPDOS 03RC, 08RC
gear pump
mzr-7255
HNP Mikrosysteme
Ismatec mcp-z
Ismatec SA Labortechnik
laboratory
balance
Mettler AB, AX, PB, PG,
PR, SB, SR
Sartorius BP, ME, Combics3
Mettler-Toledo
Sartorius
heating plate
IKA RET Control Visc
IKA Werke
Sensirion
mass flow meter ASF1400
given the name Oscar Wilde, an acronym for Organic Synthetic
Chemist’s Automated Reactor-based Workstation with Integrated
Logging of Data from Experiments. The fundamental idea was
to use established, commercially available, over-the-counter
laboratory equipment, with the ability to communicate via an
RS232 interface, and to find a commercially available, easy-
to-use software package to control the equipment. An absolute
prerequisite was to incorporate only equipment that could
equally well be operated manually. This would allow the users
themselves to decide to which extent they would like to use
automation, and also remove the risk of down-time and wasted
experiments, a factor that is inherently associated with systems
that solely depend upon computer control. The investment
strategy was to gradually phase in new equipment as old units
needed to be replaced.
The concept could as well be applied to multiple reactor
systems. The picture to the left in Figure 2 shows an arrange-
ment of three 40-mL reactors (purchased from HEL9) with
gastight stirrer shaft connections and overhead agitation. In front
of the reactors is an orbital shaker designed for 4 mL disposable
vials. Three separate heater/chillers provide temperature control
to each reactor and to the vial blocks in the shaker. The right
picture in Figure 2 shows a permanent arrangement with four
identical reactors. The availability of starting material permitting,
the use of standard lab reactors in process validations, experi-
mental design or investigation of suitable workup procedures
offers excellent control over the process parameters.
Software
The quest for suitable software to control the equipment from
a computer proved to be a much less straightforward task than
was initially realised. Although most purchased equipment came
with a functional and fit-for-purpose control software, it was
not considered worthwhile to find a method to run the individual
pieces of software in parallel, and still be able to collect their
data using a single application.
For this purpose, less manufacturer-specific software plat-
forms had to be evaluated. Thus, LabView (National Instru-
ments), Camile (Argonaut Technologies), WinIso (HEL), Lara
(Radleys UK), VirtualLab and LabMax (Mettler-Toledo),
FlexyLab (Systag) and LabWorldSoft (IKA Werke) were
assessed. The LabWorldSoft software proved to be the most
suitable candidate in terms of cost, user-friendliness and
Hardware
Table 1 lists some of the devices that were either purchased
or were already available in-house during the progress of the
Oscar Wilde project. The listing does not, per se, imply any
endorsement of a particular vendor.
One of the key players in the range of Oscar Wilde devices
turned out to be the Almemo logger from Ahlborn.7 The device
is an affordable and robust state-of-the-art data logger which
can be equipped with a multitude of different sensors to
measure, for example, temperature, pressure, pH, resistance,
voltage, current, conductivity and humidity. All sensors have
built-in EEPROM chips which make them automatically
recognisable to the logger. Any calibration parameters will also
be conveniently stored onto the chips and follow the sensors
regardless of which logger is used. The loggers are also
(8) For information and download of latest drivers visit Future Technology
Devices International Ltd. http://www.ftdichip.com. Accessed May
2009.
(7) Ahlborn Messtechnik und Regelungstechnik GmbH: Holzkirchen, Ger.
http://www.ahlborn.com. Accessed May 2009.
(9) HEL Ltd. http://www.helgroup.com. Accessed May 2009.
1060
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Figure 1. Fume hood view of an Oscar Wilde arrangement with standard equipment to measure turbidity, energy and gas flow.
Figure 2. Examples of two multiple-reactor arrangements.
• be as easy to use as possible. The number of mouse-
clicks that is required to perform a certain task should
be kept to a minimum (in general this is an excellent
measure of software user-friendliness);
• record and document experiments in one single data
file in a universal format. For this purpose, the Microsoft
Excel spreadsheet format was chosen. The file would,
apart from the raw data and a chart, also contain other
important information about the experiment on work-
sheets within the file;
• have a robust and reliable error handling in case RS232
communication with the devices would fail;
• be retrieved from a central network disk available to
every chemist. This would allow rapid distribution of
new versions and bug fixes with minimal support and
effort.
manufacturer’s willingness to incorporate new devices and
develop the software. In close collaboration with the developers,
the software was modified and rolled out across the department
during 2002. The uptake was low, however, mainly due to the
fact that the user interface still was too demanding and time-
consuming to program. During the months that followed
dedicated resource was required to keep the software up and
running by, for example, delivering prewritten templates and
hands-on support, but unfortunately the degree of uptake
did not increase accordingly.
This made it necessary for us to start development of in-
house software applications. Recommended strategies to do this
have been presented previously by others.3,10,11 We had the
necessary knowledge of computer programming in-house and
had a clear list of requirements. Among the criteria for this
software were that it should
The first in-house developed software to be rolled out was
a simple data logging application which had been developed
in the powerful and, more importantly (for employees at a
company with a strong and restrictive IT department), readily
(10) Pollard, M. Org. Process Res. DeV. 2001, 5 (3), 272.
(11) Owen, M. R.; Dewitt, S. H. Laboratory Automation in Chemical
Development. In Process Chemistry in the Pharmaceutical Industry;
Gadamasetti, K. G., Ed.; Marcel Dekker: New York, 1999; p 429.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1061

Figure 3. Screen dump from an Oscar Wilde Light run.
available Visual Basic for Applications (VBA) using Microsoft
Excel as the development platform.12 The application contained
functionality allowing gas flow measurement and rough estima-
tion of energy release.
is beyond the scope of this article, but some of the features are
listed below:
• addition of a liquid reagent at such a rate as to maintain
a specified exotherm or a specified effect using a PID
control loop
• maintenance of a constant pH during the reaction
• ability to start equipment automatically (e.g., heater/
chillers) during out-of-office hours in order to decrease
the impact of lag time associated with temperature
stabilisation
• perform experiments that are completely controlled by
the software to reduce variations of parameters
• start an addition automatically via adaptive feedback,
e.g., when a temperature value is stable
• copy measured values to an electronic lab notebook
without the risk of transcription errors
This basic logger application was rapidly embraced by the
organisation, but it was soon realised that there was also a need
to be able to actively control and send set values to the devices
instead of just passively logging output data. For this purpose
the software Oscar Wilde Light was developed in-house and
rolled out in late 2005. Since its deployment Oscar Wilde Light
has been further refined in close collaboration with chemists
and applied on “live” process chemistry problems. Today, the
software is a pure Visual Basic application,13 and the number
of controllable laboratory devices and the level of uptake among
our chemists is steadily increasing. During the year 2008 nearly
3000 software-controlled experiments were performed at our
department, generating a total of 47000 h of data. A screen
shot of a running Oscar Wilde Light experiment is shown in
Figure 3. A more exhaustive description of Oscar Wilde Light
• automatic creation of suitable cooling profiles for heater/
chillers using best process understanding practice
(13) For Visual Basic 6, see e.g.: Norton, P.; Groh, M. Peter Norton’s
Guide to Visual Basic 6; SAMS:: Indianapolis, IN, 1998 orWong, W.
Visual Basic 6 for Dummies, 2nd ed.; IDG Books Worldwide Inc.:
Foster City, CA, 1998. For Visual Basic 2008 see e.g.: Halvorson,
M. Visual Basic 2008. Step by Step; Microsoft Press: Redmond, WA,
2008.
(12) A fully functional version of the data logger application named Dorian
Gray together with a manual and installation instructions are available
from the authors on request. The application requires an Almemo data
logger.
1062
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

component.16 In any case, the estimated figures for effect and
energy obtained by this method provide chemists with important
information about their process, such as suitable minimum
addition times for reagents and whether the reaction is ac-
cumulated. The effect trace for a given chemical reaction could
point out when the end of reaction has been reached and also,
under favorable circumstances, give information about the
overall reaction kinetics.
During an experiment, the power to the immersion heater
is either activated manually, or by software control. The
automatic option is performed by means of a 220 V relay switch
between the net adapter and the power plug which can be
conveniently operated via the Almemo logger. The standard
procedure is to first switch on the heater for a few minutes
before reaction, then the reaction is carried out and the heater
is started again once the reaction exotherm has declined.
In order to estimate the effect and energy for the reaction a
special procedure has been developed, which is performed by
an Excel macro.17 The mathematical algorithms in this macro
are based on a simplified formula for effect calculation18
(eq 1).
Figure 4. Immersion heater manufactured from a glass tube
and a coiled heater cable. A net adapter is shown in the
background.
• ability to freely connect and disconnect devices during
an active experiment in order to maximize their use of
operation
• perform ad hoc programming of experiments, i.e., add,
remove or change actions after the start of an experi-
ment
Today, Oscar Wilde systems have been installed for nearly
100 chemists at AstraZeneca. In addition, the Oscar Wilde Light
software is operating at several common resource laboratories
such as those used for small- and medium-scale parallel
equipment, the hydrogenation facility (<5 L), the scale-up
laboratories (5-10 L) and the continuous flow reactor system.
The ability to have the same automation software throughout
different stages of process development greatly facilitates
comparison of experimental data sets and enables “recycling”
of chemical recipes.
dTR
P ) UA × (TR - TJ) + m × cP ×
(1)
dt
where P is the effect, UA is the heat transfer coefficient U
multiplied by the wetted reactor area A, TR is the reactor
temperature, TJ the jacket temperature, m is the mass, and cP is
the specific heat for the reactor content.
The energy, Q, for a completed reaction is obtained by
integration of the effect.
Q )
P dt ) UA ×
(T - T ) dt
(2)
∫
∫
R
J
Effect and Energy Measurements
Although data for cP are readily available for a number of
pure solvents, it is not a trivial task to try to estimate cP for
complex reaction mixtures. Thus, in practice, eq 1 should be
considered to contain two unknown terms, namely UA and
m·cP.
Given the additional fact that the effect plateau during the
time the immersion heater is active should equal the nominal
effect value of the heater, there is also a second relationship
between the UA and m·cP terms which can be employed to
find a solution to eq 1.
An experiment is evaluated as follows: After collection of
the logged data from the experiment, the Excel macro men-
tioned above starts by assigning an arbitrary initial value to UA.
At the same time the factor m·cP is calculated from the fact
that the difference between the grouped values for the upper
and lower state of the effect curve should equal the effect of
the immersion heater. From eq 1 a preliminary effect trace is
printed out in the graph. Due to the presence of noise in the
input data, mainly due to slope calculations, it is necessary for
The ability to measure effect and energy for a chemical
reaction is a fundamental tool enabling rough estimation of the
thermal hazards for a given reaction.14,15 If this could be
performed in the standard equipment that is normally used by
process chemists, they could immediately get early information
about safety issues for a specific transformation. In order to
carry out such calorimetric measurements, the Oscar Wilde
concept uses immersion heaters, an established and largely
noncontroversial way to calibrate a reactor system before and/
or after reaction, since they simulate instantaneous chemical
reactions having known power outputs. The heater elements
could be constructed from heater cable or low-ohmic carbon
resistors connected to standard low-voltage net adapters as can
be seen in Figure 4. After manufacture, and at regular intervals,
each immersion heater should be calibrated in order to assign
the accurate effect value to it.
At this point it must be clearly stressed that the immersion
heater method described here does not replace a stringent
calorimetric measurement. Process safety is defined by many
factors of which thermal hazard assessment is only one
(16) Lees, F. P. Loss PreVention in the Process Industries; Hazard
Identification and Control, 2nd ed.; Butterworth Heinemann: Woburn,
MA, 1996; Vol. 1.
(17) More detailed information about the Microsoft Excel macro to calculate
effect and energy is available from the authors on request.
(18) Poling, B. E.; Thomson, G. H.; Friend, D. G.; Rowley, R. L.; Wilding,
W. V. Perry’s Chemical Engineers Handbook, 8th ed.; McGraw-Hill:
New York, 2008; Chapter 2, p 170.
(14) Zogg, A.; Stoessel, F.; Fischer, U.; Hungerbu¨hler, K. Thermochim.
Acta 2004, 419, 1.
(15) Fischer, U.; Hungerbu¨hler, K. Calorimetric Methods of Investigating
Organic Reactions. In The InVestigation of Organic Reactions and
their Mechanisms; Maskill, H., Ed.; Blackwell Publishing: Oxford,
U.K., 2006; Chapter 8, p 198.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1063

Figure 5. Screen shot from an iteration process to find the best square shape for the effect trace. The immersion heater was active
for 10 min. The effect trace was generated from the values for the reactor and jacket temperatures.
some human intervention to judge the appearance of the effect
curve. Using a scroll bar in the macro interface, the user can
assign new values to UA and then view the result of the new
UA and calculated m·cP. An example of the iteration process
is shown in Figure 5. The iteration process aims to produce an
effect trace shape that resembles a square (obtained in the middle
picture in Figure 5). Intuitively, this would be the expected shape
of an effect trace where an immersion heater is switched on
and off. Final values for UA and m·cP have then been assigned
and are used in the subsequent analysis of the chemical reaction.
The data in Figure 5 were obtained from a 250 mL reactor
that contained 150 mL of toluene and was equipped with an
8.1 W immersion heater. The observed m·cP value for the best
square-shaped effect curve (middle) was 248 J/K. For reference,
the expected cP for toluene at 25 °C is 1.79 J/K,18 and hence,
the theoretical value for m·cP would be 235 J/K. In practice, a
slightly higher value should be expected. This could be
attributed to the fact that the impeller, thermometer baffle and
the immersion heater themselves also contribute with their m·cP
terms.
It is important to activate the immersion heater both before
and after reaction. In cases where the jacket temperature is
changed or ramped between the start and the end of the reaction
this becomes especially important because the U term, for
example, is strongly dependent on the temperature. A linearity
algorithm in the Excel macro is used to balance the UA and
m·cP values before and after the reaction to be able to use them
in the calculation. After the macro has finished the calculations,
two additional data traces, effect and energy, have been added
to the graph. Some additional data of interest are also shown
in plain text, for example, the maximum effect during the
reaction, the theoretical adiabatic temperature increase (the
temperature that would be reached in a large-scale process if
the cooling should fail), and the degree of accumulation (energy
released after the end of charging).
Chart 1. Reaction and energy profile for hydrolysis of
acetic anhydride
investigated. Thus, effect and energy calculations do not require
an extra dedicated run, which is usually the case for other types
of calorimeters.
The hydrolysis reaction of acetic anhydride is well studied
and is frequently used to show the applicability of a certain
calorimetric method. When this reaction was studied using the
Oscar Wilde concept with an immersion heater in a standard
250 mL lab reactor, the heat of reaction was shown to be 57.4
kJ/mol, a value in very close accordance with the literature value
of 58 kJ/mol.19 Chart 1 shows the temperature profile and the
effect/energy traces for a reaction where 20 mL of acetic
anhydride was added to 150 mL of water over 10 min at 40
°C. The acetic anhydride addition started at 27 min. The
temperature increase from the 12.0 W immersion heater is seen
before and after reaction.
Chart 2 shows another effect and energy measurement that
was carried out in order to investigate suitable conditions for
an aqueous quench of methylmagnesium chloride. This example
clearly demonstrates the large amount of information about a
chemical process that can be extracted from a set of logged
experimental data for the reactor and jacket temperatures.
A standard 250 mL reactor was charged with 150 mL THF,
and the reactor inner temperature (green trace in Chart 2) was
stabilized at 25 °C. Starting at 25 min, 15 mL of methylmag-
nesium chloride (3 M in THF) was added from a syringe pump
over 5 min; to quench the Grignard reagent 1.62 mL of water
was added over 10 min starting at 57 min. Before, in between,
and after the additions, an immersion heater with a calibrated
effect of 10.45 W was activated. The effect and energy were
No additional data other than the temperature curves and
the effect value for the immersion heater are needed in order
to be able to estimate effect and energy for the chemical
reaction. The method works equally well for both small and
large reactors as long as the effect of the immersion heater is
chosen accordingly. Furthermore, the method provides excellent
results when the effect and energy are calculated for reactions
that take place during ramping of the jacket temperature. An
important advantage is also that the measurement can be carried
out in a standard lab reactor under normal experimental
conditions. It is simply an “add-on” to an otherwise standard
experiment where the usual process chemistry parameters are
(19) Smith, T. L. J. Phys. Chem. 1955, 59 (5), 385.
1064
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Chart 2. Effect and energy measurements for a water quench of methylmagnesium chloride
calculated by the procedures described above, and the resulting
data are included in Chart 2.
over it per time unit.22 Cuvette-based systems are also available,
in which the evolved gas repeatedly fills a container of a known
volume.23 The Oscar Wilde concept uses measurement of the
differential pressure drop that will build up over a constriction
pipesalso known as a Poiseuille flow metersin order to
estimate gas flow rate.24 A typical arrangement is seen in Figure
1 above.
The reactor is connected to a differential pressure probe,
preferentially through a piece of silicon tubing, needle and
septum or, as is the case in Figure 1, using a specially designed
screw cap with a glass connector pipe outlet on the side. The
pressure probe is, in turn, plugged into an Almemo data logger.
A disposable needle is pushed through a septum on the reactor
to act as a constriction pipe. During an experiment a small
overpressure will build up inside the reactor when gas evolution
occurs. The constriction needle will cause the pressure to reach
a steady-state value for a given gas flow. In a typical experiment
a 250 mL reactor equipped with a 1.2 × 50 mm disposable
needle would experience a pressure rise of 50-200 Pa
(0.0005-0.002 atm) for moderate gas flows, so the overpressure
in the reactor will be negligible.
The upper pressure limit for the Almemo differential pressure
probe is 1250 Pa (0.012 atm). In order to adjust the pressure
that is built up inside the reactor and to fit the method to different
reaction scales, the length and bore of the needle should be
adjusted accordingly. For larger reactors, other types of
constriction devices could be considered (see Figure 6).
The relationship between gas flow rate and the differential
pressure drop over a circular constriction pipe is described by
the Hagen-Poiseuille equation.24
Upon addition of MeMgCl there was an initial effect spike
that reached to a maximum of 17 W, which is likely to be
due to residual water in the THF. It is clear that there was
some accumulation of energy during the addition, apparent from
the effect trace (dotted line in Chart 2), since it did not drop
immediately at the end of the addition. The second charge
showed much less accumulation. Since an excess of water was
added (1.5 equiv), the effect dropped before the addition had
finished. The effect trace showed a spike at the end of the
addition, reaching to a maximum of 23 W. This additional
exotherm was attributed to precipitation of MgClOH and was
further corroborated by the turbidity trace (not shown in Chart
2). Some heat of mixing during the water addition was also
apparent from the effect trace. Not surprisingly, the quench
evolved methane gas. This is discussed in the Gas Flow
Measurements section below.
Today, the Oscar Wilde method is used routinely, in
conjunction with differential scanning calorimetry, DSC,20 as
the source of safety data required to take processes up to 100
L in the large-scale laboratories. The chemists who perform
these evaluations should be trained in the methodology and
equipment and should always consult experts within the process
safety group whenever there seems to be any potential safety
hazard with the process. For pilot-plant production (>100 L)
the Oscar Wilde method has not replaced more established
process safety assessment methods. Here, standardized equip-
ment (e.g., the Mettler RC1 calorimeter21 and the HEL Simular9)
operated and evaluated by process safety engineers is still the
preferred method of choice.
∆P × π × d4
128 × µ × L
Q )
(3)
Gas Flow Measurements
where Q is the volumetric flow rate, ∆P is the pressure drop,
d is the pipe diameter, µ is the dynamic viscosity and L the
Gas flow measurements for process chemistry can be carried
out using mass flow meters, where the reaction gas is allowed
to pass over a heated sensor. The degree of cooling depends
on the specific capacity of the gas and the mass that passes
(22) (a) For an overview of different techniques see e.g., http://en.
wikipedia.org/wiki/Flow_measurement or(b) Weisenburger, G. A.;
Barnhart, R. W.; Steeno, G. S. Org. Process Res. DeV. 2008, 12 (6),
1299.
(23) For information about the “U” tube, visit: Intertek ASG; http://
www.intertek-cb.com/asgmc/GasMeasurementApparatus.shtml. Ac-
essed May 2009.
(20) O’Neill, M. J. Anal. Chem. 1964, 36 (7), 1238.
(21) Mettler-Toledo Inc; http://www.mt.com. Accessed May 2009.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1065

Figure 6. Gas flow measurements on large and small scale. For the 10 L reactor (left) the constriction device was a 60 mm glass
pipe, i.d. ) 3.0 mm, mounted at the small red GL cap in the reactor lid. The black silicon tubing to the differential pressure probe
is visible below the pipe. For the 4 mL GC vial (right) the pressure difference was measured over a 60 cm coiled PEEK tubing
(inner diameter 0.25 mm).
Chart 3. Dynamic viscosities for some gases25,26
Chart 4. Gas flow measurement for a water quench of
methylmagnesium chloride
software has a built-in routine which integrates the pressure in
real-time so that the volume of gas is displayed on screen
directly, it is considered good practice to reintegrate the pressure
trace after the reaction. This would take care of any baseline
drift that can occasionally be observed for experiments which
stretch over several hours.
In order to perform a successful gas flow estimation there
are a few additional points that need to be made. Before the
calibration the pressure in the reactor should be allowed to
stabilize for a few minutes before addition of the calibration
gas. It is equally important to minimize the risk of solvent
vapours condensing inside the constriction needle. At elevated
temperatures the needle should preferably be placed above a
condenser, otherwise the needle should be positioned as far
away from the reaction mixture as possible to avoid splashes.
There are no imminent safety concerns about pressure building
length of the pipe. Chart 3 shows the dynamic viscosities for
some gases familiar to process chemists. Note the conversion
1 µP (micropoises) ) 0.1 µPa ·s.
It is wise to take a pragmatic approach to eq 3 since it
requires an otherwise hermetically gastight reactor in order for
the gas to exclusively pass through a constriction needle of exact
known inner diameter. Minor leaks through stirrer shaft con-
nections, ground glass joints and through the reactor lid gaskets
can never be ignored. Therefore, the recommended way to
quantify the volume of gas is to calibrate the flow by an injection
of a known gas volume before or after the presumed gas-
producing reaction. The gas can be injected using a standard
syringe, and the amount of gas should typically be 5-20% of
the nominal reactor volume. The integrated area of the pressure
increase during reaction can then be related to the known
integrated area of the calibration volume and thus quantified.
Alternatively, the gas can be injected using a syringe pump and
the steady-state value be used as a conversion factor between
pressure and gas flow rate. Although the Oscar Wilde Light
up in the reactor, should the needle become filled with solvent,
but the accuracy of the measurement will be significantly
reduced.
Chart 4 shows the gas flow profile for the water quench of
methylmagnesium chloride discussed in Chart 2 above. The
addition of methylmagnesium chloride at 25 min caused a small
initial gas evolution (residual water in reactor solvent). The
addition of water at 57 min caused extensive gas evolution.
The gas flow dropped rapidly after about half the addition of
water due to the excess of added water. In total, 0.8 L gas was
(24) Karger, B. L.; Snyder, L. R.; Horvath, C. An Introduction to Separation
Science; John Wiley & Sons: New York, 1973; Chapter 3, p 88.
(25) Weast, R. C., Ed. CRC Handbook of Chemistry & Physics, 64th ed.;
CRC Press: Roca Baton, FL, 1984; F42-F44.
(26) The viscosity values for isobutene were calculated using a Java
application for gas viscosities found at LMNO Engineering, Research
and Software Ltd. http://www.lmnoeng.com/Flow/GasViscosity.htm.
Accessed May 2009.
1066
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Chart 5. Gas flow calculation for addition of 0.6 M HCl
(aq) to 0.2 M NaHCO3 (aq) on a 4 mL scale
over 60 min. The estimated volume of evolved gas (orange line)
was 9.1 mL (theoretically 9.0 mL). Since an excess of acid
was added, the gas flow rate (blue trace) dropped after one
equivalent had been added. Most notably, the noise level during
the calibration phase was less than (2 µL/min!
Conclusions
A key to a smooth implementation of laboratory automation
is to find ways to lower the threshold between manual and
automated work. Easy-to-use software tools and laboratory
equipment that can also be operated manually can offer an ef-
ficient and cost-effective platform to allow every chemist to
enter the digital world in an incremental fashion. In parallel,
efforts must be made, via appropriate channels and examples,
to clearly show users that the ability to log a lot of different
data from each experiment will lead to a much deeper
understanding of the processes and add value to their work.
For example, the sensitivity of modern temperature probes can
easily measure even very small temperature changes of the order
of a few hundredths of a Kelvin. This can enable monitoring
of exotherms or endotherms that stem from crystallisations or
from the evolution of gas. Together with a sound risk assess-
ment, automation also enables fully or partly unattended
experiments to be performed during nights and weekends. In
addition, the accuracy and degree of control that computerized
systems provide, as well as the possibility to reuse the chemical
recipes and to perform repetitive runs, make them ideal
candidates for quality by design (QbD) work, which is an
emerging and challenging opportunity for pharma companies.28
released: the theoretical amount is 1.0 L (45 mmol) but some
methane is expected to dissolve in the THF.
For qualitative measurements it is sufficient to use air or
nitrogen as the calibration gas. If more accuracy is desired, the
values in Chart 3 together with eq 3 can be used to recalculate
the gas flow. When unknown gases or mixtures of gases are to
be measured (e.g., Boc-deprotection) the preferred alternative
is to perform the gas calibration after the gas-evolving reaction.
Then the atmosphere inside the reactor will consist of the gas
in question and injection of a small amount of air or nitrogen
will not disturb the system significantly. If desired, mathematical
methods to predict gas viscosities are also available.27
It is worth mentioning that the steady-state pressure value
for a given gas flow rate is independent of the headspace volume
in the reactor. Hence, no correction other than the displacement
volume from added material needs to be made for reactions
where a large volume of reagent is added during the gas
evolving process. Also, the gas does not physically pass through
the probe, a fact that will prolong the probe lifetime substantially
especially when corrosive gases are measured. Figure 6 shows
two examples of gas flow measurements on 10 L and 4 mL
scale, respectively, which used the same type of pressure probe
and method.
Acknowledgment
We thank Peter Jaksch for sharing his technical knowledge
and Jo¨rgen Blixt, Daniel Fahle´n and Patric Fransson, all in our
department, for valuable input and ideas. We also thank Dr.
Hans-Ju¨rgen Federsel and Dr. Adrian Clark for constructive
criticism on this manuscript.
The scalability of the Poiseuille flow meter is quite remark-
able. Chart 5 shows the gas flow measurement from the 4 mL
vial arrangement depicted in Figure 6. The vial was initially
filled with 2000 µL of 0.2 M NaHCO3 (aq) and was calibrated
by 500 µL of CO2 that was pumped into the vial using a syringe
pump over 10 min. Then 1000 µL 0.6 M HCl (aq) was added
Received for review June 16, 2009.
OP900154C
(28) (a) Pharmaceutical cGMPs for the 21st century, a risk-based approach;
Department of Health and Human Services, U.S. Food and Drug Administra-
tion: Washington, DC, September 2004. Final report. Available at: http://
www.fda.gov/cder/gmp/gmp2004/GMP_finalreport2004.htm. Accessed Jan
2009. (b) Seibert, K. D.; Sethuraman, S.; Mitchell, J. D.; Griffiths, K.;
McGarvey, B. J. Pharm. InnoVation 2008, 3, 105.
(27) (a) Yoon, P.; Thodos, G. AIChE J. 1970, 16, 300. (b) Reichenberg,
D. AIChE J. 1975, 21, 181.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1067