Manuel C.
Maier
ab,
Michael
Leitner
ab,
C. Oliver
Kappe
bc and
Heidrun
Gruber-Woelfler
*ab
aInstitute of Process and Particle Engineering, Graz University of Technology, Graz, Austria. E-mail: woelfler@tugraz.at
bCenter for Continuous Flow Synthesis and Processing (CC FLOW), Research Center Pharmaceutical Engineering GmbH (RCPE), Graz, Austria
cInstitute of Chemistry, University of Graz, Graz, Austria
First published on 27th May 2020
Utilization of highly reactive compounds in novel flow syntheses requires new tools for process development. This work presents such a tool in the form of a modular calorimeter designed for direct heat flux measurements in continuous flow applications. The calorimeter consists mainly of 3D printed parts, which can be adapted and reassembled easily to meet user-defined applications. By utilizing selective laser melting (SLM) of stainless steel and digital light processing (DLP) of a UV curable resin, a device is produced to meet the requirements of handling highly reactive organic compounds. Calorimeter segments are temperature-regulated independently of each other by a microcontroller, allowing isothermal operation conditions. Direct heat flux measurements are possible in the device through Seebeck elements which are calibrated internally at prevailing process conditions with the aid of heating foils. Functionality of the designed calorimeter is shown by good agreement of conducted heat flux measurements with literature.
Reaction calorimetrical investigations with their different modes of operation7 play a crucial role to provide fundamental data like enthalpy of reaction, activation energy, heat capacity of a reaction mixture as well as reaction rate. A key parameter for reactor design and safety evaluation is the reaction enthalpy. Standard equipment for reaction enthalpy measurements like batch calorimeters, e.g. the RC1 from Mettler Toledo,8 can only be used to a certain extent for novel reactions under extreme reaction conditions as seen today in flow chemistry. While there exist batch calorimeters with relatively small volumes,9–11 their mode of operation is different compared to flow chemical setups, i.e. they cannot provide the conditions required to gather meaningful thermodynamic and kinetic data for continuous flow applications in harsh reaction environments.
Modifying batch calorimeters to meet flow applications is possible through standard HPLC equipment,12 nevertheless, this approach has some drawbacks. It does not provide optimal connectivity of parts to the calorimeter, leads very often to poor mixing performance caused by standard laboratory tubing and connectors, and requires existing software to be adaptable. Another possibility to use existing technology for heat flux measurements is infrared thermography.13 Disadvantages of this method are a required optical access to the ongoing reaction as well as an expensive camera providing necessary resolution of the small channels.
Flow calorimeters recently developed use thermoelectric principles to directly detect heat fluxes through the Seebeck effect.14–19 Their general setup consists of a chip-like reactor separated from a heat sink by Seebeck elements which generate voltage signals proportional to the transferred heat flux. These elements can be manufactured directly on the reactor chip14,15 or bigger and cheaper commercially available Seebeck elements can be used.18 By miniaturization of Seebeck elements, spatially resolved measurements are shown and additional information about reaction time scales are obtained by some designs.17,18 Calibration of the Seebeck elements can be done externally or internally by integrated heating in the assembled device, which accounts for heat losses of the system.17,19
All of the mentioned designs lack the possibility to react locally to changing temperature caused by a reaction in the corresponding section of the reactor. Heat fluxes from each section are transferred through a Seebeck element uniformly to one heat sink, with its temperature being regulated by a thermostat set to the desired reaction temperature. Sufficient heat transfer from heat sink to reactor cannot always be ensured in such a setup due to the poor heat transfer characteristics of the utilized Seebeck elements. As a result, isothermal conditions cannot be achieved if highly reactive compounds are used and formation of local hot spots within the reactor is likely. To overcome the poor thermal conductivity of Seebeck elements, it is desired to control locally the temperature within a reactor by specifically applying increased heat fluxes.
The designed calorimeter described within this work is guided by the abovementioned devices but utilizing the advantages of 3D printing as well as already existing knowledge of reactor design20,21 and microcontroller-based regulation. We present here a novel isothermal heat flow calorimeter which can withstand harsh reaction conditions commonly found in flow chemistry by a combination of two different additive manufacturing techniques and cheap, commercially-available electronics. In addition to its modular and extendable design, the calorimeter's segments are regulated independently from each other and allow to react locally to ongoing reactions. Applicability of the flow calorimeter is shown by studying thermodynamic properties such as reaction enthalpy, heat capacity and molar excess enthalpy, and thus providing necessary data for reactor scale-out as well as safety aspects.
Two additive manufacturing techniques were used to manufacture the device, selective laser melting (SLM) and digital light processing (DLP). Reactor plate and cooling blocks were manufactured by SLM of stainless steel to provide high chemical and mechanical stability as well as excellent heat transfer rates, allowing the usage of highly reactive compounds and organic solvents at elevated pressures. The casing of the calorimeter was manufactured by DLP of a cheap UV-curable resin. This material in combination with an internal support structure showed good thermal insulation against the environment while providing necessary strength to fixate the whole device. Both manufacturing techniques allow reactor elements to be produced, which can be easily modified to meet a desired configuration of the device in terms of number of reaction segments with different size and mixing geometries.
The calorimeter measures direct heat fluxes using the thermoelectric Seebeck effect with the aid of Seebeck elements contacting the reactor plate. A Seebeck element consists of a series of thermocouples, which themselves are consisting of two dissimilar electrical conductors connected at one end. By applying different temperatures at each end of a thermocouple, it generates an electrical voltage proportional to the temperature difference. By arranging thermocouples in series across two surface areas as within Seebeck elements, it is possible to measure the transferred heat by means of an electrical voltage. Conversely, when a voltage is applied to such a device, a hot and cold side is formed, which allows it to transfer heat. This configuration is described by the Peltier effect and the same electronic components are referred to as Peltier elements within this work.
The generated voltages from the Seebeck elements are measured by a self-made electrical circuit utilizing a microcontroller which is programmed with the Arduino integrated development environment (IDE). To obtain the actual heat flux, a calibration with integrated heating foils has to be made. With these heaters, an exothermic reaction can be simulated and by applying a defined power input, a calibration at prevailing process conditions can be obtained. This calibration already accounts for heat losses of the device operated at a defined temperature. Influences of the heating foils during an actual calorimetric measurement were not investigated and neglected as they are assumed to be very small. Furthermore, ideal heat input of the heating foils was assumed within this work.
In contrast to other designs,17,18 this device features a microcontroller-based temperature control of each calorimeter segment. This allows to ensure an operation of each segment independently as close as possible to the desired isothermal set point. The reactor segment's temperature is measured and transmitted to a microcontroller which adjusts the heat flux of a Peltier element through a PID based control strategy, see Fig. 2.
The control strategy presented above only allows the investigation of exothermic measurements. Endothermic measurements can be recognized by the Seebeck elements, since the thermoelectric voltage can change signs. However, the electrical contacts of the Peltier elements would have to be switched to enable a temperature control. This would result in switched cold and hot sides to provide the necessary heating of endothermic events. In addition, no internal calibration for endothermic events is possible with the implemented heating foils.
The reactor plate was designed with a precooling section and two reaction sections utilizing a split and recombine structure, Fig. 2. The internal diameter was set to 0.8 mm, which gives a combined reaction volume of 220 μL for both reaction segments. Due to the modular design, these reaction sections can be duplicated and changed depending on the desired task to reach a defined residence time as well as mixing properties. Reactor sections were separated from each other to reduce heat conduction between them. The additional material mentioned above was removed by laser cutting to separate the sections. Connections of the reactor plate to peripheral tubing can be made with standard flat bottom HPLC fittings. Maximum operation pressure of the reactor is therefore limited only by the specifications of the used fittings, in this case 100 bar.
The designed cooling block for heat removal of the Peltier elements needed an internal support structure before printing (see ESI† Fig. S1). Besides being necessary for fabrication, this structure increases the internal contact area for the heat transfer of the coolant to the metal. If needed, a connector port for temperature measurements close to the Peltier contact area can be implemented as well. This temperature measurement is indicated in Fig. 1, but was not used within this work.
Additive manufacturing was then carried out by an SLM system from EOS which utilizes an ytterbium fibre laser with 400 Watt maximum power input, scanning through each of the 40 μm high sliced layers. A 316L stainless steel powder bed with a mean particle size of 35.9 μm was used for the calorimeter elements. Post-processing of the stainless steel parts included several cleaning procedures with compressed air, treatments within an ultrasonic bath to free entrapped particles, sandblasting of the outer surfaces, CNC milling and laser cutting to achieve the planned geometries. Accurate contact of the reactor plate and cooling block to the Seebeck and Peltier elements was ensured by the CNC milled surfaces, which have the same roughness properties as standard milled stainless steel parts. Besides removing of excess particles, no further surface treatment of the internal channels was carried out.
Casing elements of the calorimeter were designed to be easily 3D printable without any additional support structure, see Fig. S2 in ESI.† All elements feature prism like internal structures, which connect each layer, give the parts necessary mechanical strength, reduce the amount of material needed during production and provide necessary thermal insulation against the environment. All casing elements were designed to be printed by DLP of a UV curable resin. The elements were produced with the Photon from Anycubic and a standard black resin from the same manufacturer. Based on preliminary knowledge of the printer, a layer height of 50 μm with a normal exposure time of 14 seconds was chosen to produce the casing elements. Due to the small printing area of the device, several printing jobs had to be carried out but the whole production time did not exceed two days. By using this approach of 3D printed casing, future adaption and extension of the calorimeter can be easily carried out. Post-processing of these parts included only the removal of the parts from the build platform and cleaning with ethanol. The parts could be used directly after printing but a final curing in sunlight until the next day was envisioned.
(1) |
rx = coΔhRX | (2) |
(3) |
(4) |
(5) |
tr = tr,pre + tr,r1 + tr,r2 | (6) |
tr,i = fcalibr (USE,i) | (7) |
In a steady state operation, the temporal change of energy equals zero in eqn (1) and the reaction enthalpy can be calculated as shown in eqn (8) whereby convective heat fluxes of inlet streams need to be subtracted and the heat fluxes of the outlet have to be added.
(8) |
Because of these limitations, the diazo coupling published by Bourne et al.22 was chosen for mixing evaluations of the designed reactor plate, see Scheme S1 in the ESI.† In a first step, diazo coupling of 1-naphthol (A) and diazotised sulfanilic acid (B) takes place and gives the monoazo isomers p-R and o-R. Poor mixing promotes the secondary coupling of p-R and o-R with B giving the bisazo dye S. Better mixing is indicated by less formation of the secondary coupling product S.
(9) |
(10) |
(11) |
(12) |
Solutions were pumped with equal flow rates through the reactor plate at total flow rates of 0.2, 0.5, 1, 2, 4, 6, 8 and 10 ml min−1. Samples were collected after approximately three residence times and stored in a dark container. Before analysis, samples needed to be diluted (1:8) with the 444.4 mM buffer to account for the long path length of 10 mm of the available UV-vis flow cell (Flow Cell-Z-10, Avantes). Spectral data was produced by passing the light from a UV light source (AvaLight-DS-DUV) through the flow cell and to a detector (AvaSpec-ULS2048) with an integration time of 1.05 ms, averaging of 100 samples and saving the obtained data from an interval of 390 to 700 nm in 10 nm steps. Concentrations were calculated by a multi-parameter-linear-regression of the absorption spectra as described in a previous work.21
For the calibration of each reactor segment, a known and steady electrical heat flux was supplied to the respective segment simultaneously. This heat flux was delivered by integrated heating foils (TSC0400040gR7.91, pelonistechnologies.com) for each reactor segment separately and recognized by the respective Seebeck element opposite the reactor plate as a thermoelectric voltage USE,i. Heating foils were connected in parallel to a power supply (Manson NRP-3630) which provided a known electrical power input. Each foil's electrical resistance had to be measured to calculate the true applied electrical heat flux for the respective foil. Because of the inaccurate voltage and current display of the power supply, an exact voltage and current measurement was installed with digital multimeters for exact power input characterization. Changes in the electrical resistance of all elements (ESI† Fig. S4) were seen by varying temperatures with a changing overall resistance of the system. This changing resistance was accounted in the calibration experiments. Calibration data was obtained at steady state conditions with 31 points between zero and 7 W applied to each reactor segment for an operation temperature of 25 °C.
CH3COOH + NaOH → H2O + CH3COONa | (13) |
To demonstrate an additional application of the designed calorimeter, measurements of excess molar enthalpy of methanol (CHROMASOLV™ ≥99.9%, Honeywell) and deionized water at 25 °C were carried out. Within the experiments a constant total flow rate of 4 ml min−1 was used while different flow rates of the respective pumps were set to obtain mixture data throughout the binary system.
Fig. 5 Evaluation of the reactor plate's mixing performance. Lower yield of the bisazo dye S at higher flow rates indicates increasing mixing performance. The evaluated reactor plate performed well compared to commercial mixers evaluated within literature.30 It is compared to a standard T-mixer (Upchurch Scientific), an X-mixer (Little Things Factory GmbH, type X) and the Slit interdigital micromixer SIMM-V2-ss (Institut für Mikrotechnik Mainz GmbH). |
A comparison of reactor performance with literature is possible with the Bourne reaction since it was carried out at the proposed standard conditions.27 Compared to commercial equipment,30 the designed reactor plate performed very well. It is compared in Fig. 5 with a standard T-mixer (Upchurch Scientific), a X-mixer (Little Things Factory GmbH, type X) and the Slit interdigital micromixer SIMM-V2-ss (Institut für Mikrotechnik Mainz GmbH). This comparison is of course still dependent on the chosen pumps; however, both evaluations utilized high-accuracy syringe pumps and similar pumping performance can be assumed.
The calibration at 25 °C for each reactor segment is shown in Fig. 6. As reported in literature,18 a correlation of heat flux and thermoelectric voltage USE,i of the respective Seebeck element was found in the form of a polynomial of second order. This polynomial matches to the theoretical Joule heating by electrical power input P expressed with electrical potential U, when assuming an ideal resistor R, and expressing electrical current with Ohm's law as shown in eqn (14).
(14) |
Experimental data of this experiment can be seen in the ESI† in Fig. S5 and S6. Within this experiment, the necessary heat flux to cool down both feeds was calculated and compared to the heat capacity of water 75.34 J mol−1 K−1.31 A heat capacity of 73.36 J mol−1 K−1 with a variation of ±2.62 J mol−1 K−1 was obtained from the measurements. Increasing accuracy of the measurement was seen at higher flow rates above 2 ml min−1 total flow rate. At these high flow rates, temperature measurements of in- and outlets can be expected to better represent the true fluid temperature in and out of the device because of heat losses between the sensors and the device. Excluding flow rates below 2 ml min−1 leads to a value of 75.74 J mol−1 K−1 with a variation of ±1.18 J mol−1 K−1 for the remaining experiments. An accurate function of the device even at changing process conditions could be shown with a deviation of 0.53% of the mean value from the literature value.
Fig. 7 Neutralization of AcOH with NaOH with a comparison to the theoretical heat of neutralization for 1 mol of water.32 The grey area represents the directly measured heat flux with indication of the energy balance shown by an extending frame. |
From the obtained experimental data, (see ESI† Fig. S7–S9 for a comparison between experiments and Fig. S10–S17 for time resolved data of the respective measurement) it can be seen that at lower flow rates the reaction finished in the first reactor segment and only remaining heat gets transferred through convection to the second segment. Above flow rates of 6 ml min−1 a certain increase of detectable heat flux can be seen on the second segment. Most probably the reaction was shifted downstream at increasing flow rates. The pulsating nature of the HPLC pumps at higher flow rates is likely to be the main factor influencing this behaviour. Probably higher and lower concentrated plugs travel through the calorimeter and may experience back mixing to a uniform concentration field after a certain amount of mixing elements. Nevertheless, a constant operation for each measurement point was obtained after approximately 3 minutes.
Measured heat fluxes with contribution of the heat balance were compared to the expected neutralization enthalpy of −57.4 kJ mol−1 for the neutralization of AcOH with NaOH.32 Within all evaluations, heat capacity of pure water was used for the calculations. All measurements obtained a reaction heat of −52.95 kJ mol−1 with a variation of ±6.68 kJ mol−1. As in the previous experiment, low total flow rates of 1 and 2 ml min−1 gave dramatically bigger errors compared to higher flow rates. Without these low flow rates, a reaction heat of −57.00 kJ mol−1 with a variation of ±1.08 kJ mol−1 can be obtained from the measurements. The higher error occurring at lower flow rates was assumed to be caused by ineffective temperature measurements of the inlet stream due to heat losses. In this case, the heat balance was most probably not closed correctly.
In Fig. 8, the obtained measurements at 25 °C were compared to the literature values presented by Piñeiro et al.10 A slight offset to the reported values from literature can be detected but the general shape of the curve was perfectly mimicked by the obtained measurement data. Apart from the qualitative comparison of the two curves, no further evaluations were carried out with the data at current state.
Fig. 8 Measured molar excess enthalpy HE at 298.15 K for water (1) + methanol (2). With a slight offset, the obtained data nicely resembles measurements from literature.10 |
From all measurements a heat capacity of 80.45 J mol−1 K−1 with a variation of ±7.51 J mol−1 K−1 was obtained. In contrast to the other measurements, discarding low flow rates did not shift the obtained value closer to the literature value of 75.34 J mol−1 K−1 (ref. 31) but the variance of measurement did improve. Without low flow rates, a heat capacity of 84.01 J mol−1 K−1 with a variation of ±2.70 J mol−1 K−1 was measured.
As seen from the experimental data during measurements (see ESI† Fig. S21 and S22), this temperature jump did not produce a high change in recognizable heat flux due to the low temperature difference. Increasing measurement error can be attributed to the small detectable heat flux where reduced measurement efficiency can be expected. This problem could be solved by applying greater temperature jumps between the segments. But accurate heat capacity measurements are generally shown with an experimental setup as reported before within the functionality test with warm water.
The designed calorimeter was validated with a series of experiments which produce a well-known heat flux. Through these experiments, the calorimeter design was proven to be applicable to measure reaction heats of fast reactions, important material properties like specific heat capacities, and system-specific properties as molar excess enthalpies.
Improvements of mixing performance for lower flow rates should still be considered for future models of the reactor plate. Adaptations of the existing calorimeter can be easily made due to its modular design. The reactor plate used can be exchanged to meet the requirements of different flow syntheses for higher or lower flow rates to provide additional mixing performance if required. In addition, additional connector ports can be added to the design if mixing of multiple streams is desired.
Future work will focus on extending the reactor plate to increase internal volume of the device as well as the addition of smaller Seebeck elements for a more locally resolved measurement. Further improvements of the setup will be made by coupling a controllable heat exchanger to both inlets for a precise preconditioning of the streams. Additionally, it is planned to investigate hazardous chemical syntheses with the current device as well as with advanced versions of it.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0re00122h |
This journal is © The Royal Society of Chemistry 2020 |