The main part of this setup is the sample chamber, as shown in Fig. 1. The purpose of this custom-built part in the form of a hollow cylinder is to carry the liquid solution. It is made from casting silicone (SF45, Silikonfabrik; see Fig. 1b), which is highly chemically resistant against alkaline solutions and additionally facilitates proper sealing due to its high elasticity. When installed, the chamber is compressed by two round crystal quartz disks. In this way, both disks are in direct contact with the sample on one side each. The sample chamber was designed for small sample volumes of approximately 0.5 mL.

To inject the liquid, two syringes, one to inject the test fluid and one to simultaneously remove the air from the chamber, are used. The injection needles pierce small holes in the chamber wall when they are first inserted. Because of the high elasticity of the chamber the injection sites are self-sealing even when the injection needles are removed after sample injection.

Figure 2 shows different tested geometries of the sample chamber. For conductivity measurements, the chamber geometry in Fig. 2a is used due to its constant cross-sectional area. The geometry in Fig. 2b contains space for gas accumulation at the top. This turned out to be crucial for experiments at elevated temperatures, where gas bubbles develop more easily and have a negative impact on the measurement accuracy. The geometry in Fig. 2c shows an additional space to collect particles, which sediment at the bottom. Zeolites can reach sizes of several micrometers and, hence, are subject to sedimentation.

For all measurements, besides the conductivity measurements, the geometry according to Fig. 2c was used. To cast the sample chamber, a mold was printed, using a precision photopolymer 3D printer (Objet30 Pro, Stratasys) and photopolymer (VeroClear, Stratasys). To separate the mold from the silicone part after curing, the former was constructed from four screw-fastened parts, as shown in Fig. 3a.

The chamber housing contains the sample chamber, the quartz sensors and the auxiliary components, such as O-rings. The quartz disks were squeezed against the sample chamber by means of a 3D printed housing. In Fig. 4, the parts of this housing, including the sample chamber, are shown. Figure 4a, b and h were also 3D printed using the aforementioned system (Stratasys Objet30 Pro) and a high-temperature-resistant photopolymer (Stratasys RD525).

To contact the two quartz crystals, custom-made contact rings etched out of a nickel silver sheet, as similarly employed by Reichel et al. (2010), in combination with spring pins (811-S1-002-10-017101, Preci-Dip), were used (see Fig. 4b and f). The O-rings (Fig. 4d) facilitate proper bearing on the dry side.

The chamber housing is placed in an aluminum box, whose temperature is controlled by two Peltier elements (Nesarajah and Frey, 2016), two Pt100 temperature sensors and custom-made electronics. To ensure a failure-free operation and fast equilibration times, the cool side of the Peltier elements are equipped with ventilated heat sinks (Fig. 5). The fully assembled setup is shown in Fig. 6. This approach to control the temperature has been successfully implemented and tested for other devices (Ecker et al., 2021). The major aim was to control the temperature at a constant value; in the considered temperature range between 0 and 80 ∘C, changes in temperature set points led to equilibration times in the order of a few minutes.

The whole setup was operated with a Raspberry Pi 3B+ computer and a custom-made electronics board. In Fig. 7, this board is shown, including marked areas for its different sub-circuits. An analog–digital converter (AD7124-4 BRUZ, Analog Devices) was used for the readout of the Pt100 temperature sensor signal.

Controlling the current through the Peltier elements with pulse-wide modulated signals is facilitated by two H-bridges with high-current metal–oxide–semiconductor field-effect transistors (MOSFETs). The analog–digital converter in combination with an operational amplifier and a current measurement enables optional voltage or current-controlled output.

A dedicated readout unit (QCM50, Micro Resonant) in combination with a multiplexer (see Fig. 6c) and a Raspberry Pi embedded computer were used to capture the resonance frequencies and the quality factors of both quartz disks over time. The measurement data were collected using a custom-written Python code and were stored on the Raspberry Pi. In addition, a graphical user interface was programmed to adjust the set-point temperature of the setup.

To measure viscosity (η), two identical AT-cut quartz disks that operate in thickness-shear mode were used. To determine the viscosity–density product from the measured values of resonance frequency (fr) and quality factor (Q), the frequency-dependent acoustic impedance (Zfl) for Newtonian fluids (Follens et al., 2009; Reichel et al., 2014; Vogelhuber-Brunnmaier and Jakoby, 2012) in the three-parameter form was used: 1Zfl=jωρVfl+jωηρAfl+ηLfl, where Vfl represents the effective fluid volume, Afl is the effective shear-wave interaction surface, Lfl is the effective length of viscous damping, ρ is the density of the sample fluid, j is the imaginary unit and ω=2πfr is the angular frequency of the oscillation. With the assumption that the in-plane motion is dominant, the model is reduced to 2Zfl≈jωηρAfl. The unloaded resonator is characterized by an angular resonance frequency ω0, a mass coefficient m0, and a low intrinsic damping expressed as the unloaded quality factor Q0. The resonance frequency and quality factor in the loaded case are then expressed using the real ℜ and the imaginary ℑ parts of Eqs. (1) or (2) as follows: 31ω2=1ω021+ℑZflm0ω,41Q=1Q0+ℜZflm0ω0. The viscosity–density product according to Eq. (2) can be derived either from the resonance frequency or the quality factor. Two reference measurements in known fluids or gases are necessary to determine the unknown constants Afl and m0. Air, deionized water and NaOH with known concentrations are used for this purpose. Furthermore, to calculate the viscosity, the density of the sample is required. The density values from the literature at known temperatures are used, and no variation during the process is assumed.

Electrical conductivity is determined via electrochemical impedance spectroscopy (EIS). A potentiostat (Reference 600+, Gamry Instruments) performs the electrical impedance measurements between the two electrodes of the quartz sensors, which are in contact with the sample. Note that the present approach represents a combination of a dual QCMD setup with a potentiostat impedance measurement. Thus, while the required sensors are integrated in the devised measurement chamber, the readout electronics are still separate. For a commercial integrated EQCMD device, of course, the integration of the electronics into a single unit is practicable.

A first-order model for the impedance of the liquid in terms of a resistor in series with a capacitance is employed. The former models the Ohmic (ionic) resistance of the liquid sample, and the latter captures the effect of the electrochemical double layer at the electrode–liquid interface. For higher frequencies, the capacitive contribution is negligible, resulting in a phase shift of zero. Consequently, this impedance represents the Ohmic resistance of the sample and a constant resistance R0, representing the resistance of wires and connections. R0 is first determined with a reference measurement using a conductivity standard with conductivity σS. If the length l and the cross section A of the sample chamber as well as the electrical conductivity σ of the sample are known, R0 can be calculated using Pouillet's law and the measured resistance R (value of impedance measurement with the minimal phase shift) according to 5R0=R-lσS⋅A. After this calibration, for further measurements, the electrical conductivity σ of the sample can be calculated from Eq. (5) yielding 6σ=lR-R0⋅A.

Zeolite synthesis liquids (ZPLs) are prepared following a slightly altered methodology as explained by Van Tendeloo et al. (2015). Firstly, tetraethylorthosilicate (TEOS) was mixed with sodium hydroxide solution to achieve a molar ratio of 1 part TEOS (98 %, Acros Organics), 1 part 1 NaOH (>98 %, Sigma-Aldrich), and 25 parts H2O (Milli-Q), resulting in hydrolysis of TEOS. After 24 h of mechanical agitation, the mixture was rested, leading to phase separation into a top ethanolic water phase and a dense bottom phase, void of bulk water, containing small silicate oligomers and weakly hydrated sodium and hydroxide ions. After 3 d, this bottom phase, called hydrated silicate ionic liquid (HSIL), was collected and then mixed with sodium hydroxide, aluminum hydroxide (reagent grade, Sigma-Aldrich) and water to achieve a molar composition of 0.5 Si(OH)4 : 0.028 Al(OH)3 : 1 NaOH : 5 H2O. Homogenizing this mixture by mechanical agitation for 24 h finally yields the zeolite precursor liquid (ZPL), characterized by a single, transparent phase. Choosing a synthesis temperature of 60 ∘C, zeolite yield (white precipitate) is observable within several days, and synthesis can be considered complete after 1 week.

The setup is tested with viscosity measurements using NaOH solutions of various concentrations. The results are compared to data from the literature (Sipos et al., 2000). Testing the quartz disks with the planar surface on the measurement side (supplier and manufacture by LapTech Precision Inc.), the calibration with gases is problematic, as damping of the sample chamber and the O-rings causes large errors. Therefore, instead of calibration with air and water, the setup is calibrated with water and NaOH (mass concentrations 46 %) using the literature values from Sipos et al. (2000). The measurement results are shown in Fig. 8a.

The quartz disks from Taitien Electronics Co., Ltd. (supplier Digi-Key Electronics), which have a convex surface, were calibrated by measurements in air and water. As visible in Fig. 8, both measurements show that the measured values are in reasonable agreement with reference values. To get more accurate measurement data over this wide range of viscosities, calibration with viscosity standards on several different points would be necessary.

For the electric conductivity measurement, the setup is first calibrated with a conductivity standard (Consort B562 calibration solution 1 M KCl, σS=11.18 S m-1 at 25 ∘C). Using this known conductivity of the sample, the mechanical dimensions of the sample chamber, the measured value of R and Eq. (6), the constant is calculated as R0=7.050 Ω. In Fig. 9, the Bode diagram of this measurement is shown. It can be seen that our model fits adequately. For all conductivity measurements, it is very important that the chamber is absolutely free of air bubbles. Air bubbles locally decrease the cross section of the sample and, therefore, cause an increase in the measured solution resistance (i.e., a drop in conductivity). Furthermore, as temperature has a drastic effect on ionic conductivity, the sample was loaded 1 h prior to the measurement to ensure a uniform and steady temperature distribution.

To validate our setup, measurements with other standards (see Table 1) were performed. The calculated values for electric conductivity are presented in Table 1. As can be seen, the measurement values are in good agreement with the nominal conductivity values.

For monitoring the crystallization process, we put ZPL at 60 ∘C into the sample chamber. After the first measurements, we noticed that the gold electrodes of both quartz disks, which were in contact with the highly alkaline sample, partly degraded over the experimental period of about 7 d. An attempted solution to this issue was to cover the electrodes with thin layers of polymethyl methacrylate (PMMA) via spin coating. PMMA is highly resistant to alkaline solutions and easily deposited in micrometer-thick layers. With more massive layers, the measurement sensitivity of the QCMD would decrease due to the damping in the polymer layer. Additionally, as PMMA is a dielectric material, the electrical conductivity measurements become more difficult.

Figure 10 shows a measurement with and without the PMMA layer. As expected, the two measurement curves differ in sensitivity, which is apparent from the much lower quality factor of the coated quartz than the uncoated one. However, both measurements show a similar course. From the decreasing resonance frequency and quality factor over time, we assume the deposition of emerging particles on the quartz surfaces.

One intriguing effect observed is that bumps periodically appear in all curves (Fig. 10). To identify, whether this comes from the crystallization process itself or from spurious artifacts, we tested solutions of sodium and potassium hydroxide at various concentrations as well as water. The liquids, which had a similar alkalinity to the precursor solution, showed these periodic bumps too; however, we did not observe them in water. The initial suspicion that resonance effects are associated with spurious standing pressure waves, as discussed by Beigelbeck and Jakoby (2004), was ruled out, as the bumps should then appear in water as well. Therefore, we assume that the bumps were due to degradation effects caused by the high alkalinity. Images of the electrodes before and after confirm the appearance of severe corrosion (Fig. 11).

In the on-going investigation, coating materials with higher protective properties are tested to avoid electrode degradation.