Angewandte
Communications
Chemie
in combination with automated synthesis platforms was
reported just in a few publications,[13,14d,15] so far.
Grignard reagents are highly exothermic reactions. Due to the
involvement of solid species in the chemical reaction,
sampling and chemical analysis with NMR spectroscopy or
any other method is challenging. As the reaction is highly
exothermic, reaction monitoring is mandatory to shorten
reaction times while ensuring safety. The evolution of
Grignard reagents was already monitored via NMR[17] but
so far, the coupling of Grignard reagents to an aromatic
aldehyde was not investigated using online methods. Further
experimental details of the procedure are given in the
Supporting Information.
The Chemputer employed in this study was developed by
the Cronin Group, University of Glasgow[16a] and can transfer
liquids between modules with syringe pumps and six-way
selection valves actuated by the control software. Depending
on synthetic needs, different modules can be combined, in this
investigation that was a tempered reactor, liquid-liquid
separator, and rotary evaporator. The original platform can
be extended by analytical devices, in our case a compact NMR
instrument. In Figure 2, the design of the Chemputer and the
herein developed analytical module including the primary
communication infrastructure are shown.
Interchangeability of the analytical method and evalua-
tion software is ensured through a SQL database for
information transfer, only requiring a small interface for for
example, OPC UA based communication between SQL and
analytical software. The underlying concept is the separation
of the Chemputer and its respective controlling software (the
Chempiler and analytical labware python modules)[15] as an
executing platform with the ability to trigger analytical
measurements from the processing of analytical data. The
measurement results can be evaluated by any suitable
algorithm implementation or human input and the resulting
information, for example, species concentration or the binary
information about the presence of a distinct spectral feature
are fed back into the database. Decisions taken by the
Chemputer are solely based on traceable values in the
database, thereby achieving high transparency, modularity,
and simple integration. The application is demonstrated by
incorporation of compact NMR and PT100 thermocouple,
evaluated by a custom and a commercial algorithm. This
could be extended to simple pH measurements and basically
any standard analytical method. NMR spectroscopy was
chosen as applied PAT due to its matrix-independent linearity
between measured signal and species concentrations. This is
especially advantageous when several reaction species vary-
ing in concentration are present, causing time-consuming
calibration effort for all established PAT (e.g. UV/Vis,
Raman, and NIR), except for NMR, to obtain quantitative
results. Compact NMR instruments benefit from lower cost,
no need for cooling agents and portability compared to
common NMR instruments but suffer from peak broadening
and hence potential overlapping of single peaks due to lower
field strengths. Several methods exist which tackle this
common problem in (not only NMR) spectra evaluation, for
example, CRAFT algorithm,[18] quantum mechanical
approaches,[19] machine learning[20] or indirect hard modeling
(IHM).[21] As IHM has proven itself to be a robust method,[22]
The “Chemputer”[1,16] is the first universal automated
synthesis platform that was designed to implement the
abstraction of chemical reactions in a universal manner
according to their unit operations: additions, transfers, or
physical manipulations (heating, stirring) performed in mod-
ules, for example, a flask with a condenser. These unit
operations can be programmed, and the modularity of code
and hardware allows application to a broad range of reactions.
Thus far, the Chemputer relies upon precise instructions, yet
without PATand processing of chemical information, can only
follow well-defined synthesis routes programmed with fixed
quantities and reaction times. So far, analysis was performed
on reaction products, which is sufficient for optimized routes.
Integration of PAT might entail significant cost, but also be
helpful in cases like i) heterogeneous reactions with varying
time scales or reactions with initiation times, ii) thermal
runaway reactions, iii) unknown quantitative information
(concentration, purity, amount) on starting materials includ-
ing multi-stage synthesis with required adaption of subse-
quent stages, iv) reaction optimization and v) avoiding side-
reactions/overreaction. In the cases mentioned, online anal-
ysis offers insights into reaction progress, present species, and
concentrations, reveals optimization potential, and allows for
faster malfunction identification. Feedback control will in
turn lead to increased robustness, decreased chemical waste
generation and energy consumption by reducing waiting
times.
Herein we show the successful implementation of such
feedback control for different starting materials in the
Grignard reaction, selected to be the first benchmark
candidate to validate the concept of building a self-optimizing
Chemputer. Due to harsh measurement conditions caused by
evolution of solids, the reliable use of online NMR measure-
ment is one key aspect which will be addressed in the research
presented here.
The feasibility of the universal synthetic and analytical
approach presented was demonstrated for a prototypical
reaction class, the Grignard reaction (see Figure 1). This
reaction class was chosen because: (1) Grignard couplings are
applied in synthetic routes to pharmaceuticals and fine
chemicals; (2) The initiation time of the heterogeneous
Grignard formation may differ due to varying activation
states of the magnesium involved: (3) The preparation of
Figure 1. Examined reactions including Grignard reagent formation.
A: Synthesis of Diphenylmethanol; B: Synthesis of 1,2-Diphenyletha-
nol; C: Synthesis of 1,3-Diphenylpropan-1-ol.
2
ꢂ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 1 – 6
These are not the final page numbers!