Optimisation of a Side Inlet for H 2 Entry into an Ultrasonic Spray Pyrolysis Device

: An ultrasonic spray pyrolysis (USP) device consists of an evaporation and two reaction zones of equal length, into which an aerosol with a precursor compound enters, and where nanoparticles are formed in the ﬁnal stage. As part of this research, we simulated the geometry of a side inlet, where the reaction gas (H 2 ) enters into the reaction tube of the device by using numerical methods. Mixing with the carrier gas (N 2 ) occurs at the entry of the H 2 . In the initial part, we performed a theoretical calculation with a numerical simulation using ANSYS CFX, while the geometries of the basic and studied models were prepared with Solidworks. The inlet geometry of the H 2 included a study of the position and radius of the inlet with respect to the reaction tube of the USP device, as well as a study of the angle and diameter of the inlet. In the simulation, we chose the typical ﬂows of both gases (N 2 , H 2 ) in the range of 5 L/min to 15 L/min. The results show that the best geometry is with the H 2 side inlet at the bottom, which the existing USP device does not allow for. Subsequently, temperature was included in the numerical simulation of the basic geometry with selected gas ﬂows; 150 ◦ C was considered in the evaporation zone and 400 ◦ C was considered in the other two zones—as is the case for Au nanoparticle synthesis. In the ﬁnal part, we performed an experiment on a USP device by selecting for the input parameters those that, theoretically, were the most appropriate—a constant ﬂow of H 2 5 L/min and three different N 2 ﬂows (5 L/min, 10 L/min, and 15 L/min). The results of this study show that numerical simulations are a suitable tool for studying the H 2 ﬂow in a UPS device, as the obtained results are comparable to the results of experimental tests that showed that an increased ﬂow of N 2 can prevent the backﬂow of H 2 effectively, and that a redesign of the inlet geometry is needed to ensure proper mixing. Thus, numerical simulations using the ANSYS CFX package can be used to evaluate the optimal geometry for an H 2 side inlet properly, so as to reconstruct the current and improve future USP devices.


Introduction
Nanoparticles range in size from 0.1 nm to 100 nm, and gold nanoparticles have recently been at the forefront of research. This is due to their different properties [1] compared to bulk gold, which are known at the macro level. Gold nanoparticles have a large surface area in terms of volume, so they are useful for many new applications. They have functional properties, among which biocompatibility and catalytic activity make them different from other metal nanoparticles [2,3].
Ultrasonic spray pyrolysis (USP) is a method of nanoparticle synthesis. Compared to other methods (laser ablation [4], nanoemulsion [5], reduction in liquid [3]) it enables the continuous synthesis of nanoparticles and represents a promising method for their industrial production [6]. The existing USP device ( Figure 1) [7], which has three temperature zones, allows for the controlled synthesis of metal nanoparticles on a laboratory scale. They are synthesised by the transformation of the pre-prepared solution (the so-called precursor) by ultrasound into an aerosol, which is transported with the help of a carrier gas within three zones of the heated reaction tube. In the first zone, the solvent evaporates. In the second, the thermal decomposition of the precursor compound and the reaction with H 2 gas occur, and the nanoparticles are densified in the third zone. Thus, the formed nanoparticles can be trapped in liquid, where a suitable stabiliser is dissolved, which prevents their possible agglomeration [1]. The H 2 gas enters a USP device in the second zone via a side inlet at a fixed inlet angle with respect to the reaction tube. The inlet angle and geometry must be chosen to ensure the proper mixing of the carrier gas, reaction gas, and precursor compound, with the aim of producing as many controlled nanoparticles as possible, defined by the target final size and shape. Effective mixing of all the components is key to ensuring a fast and reliable synthesis. Common issues that can be attributed to bad mixing include unreacted residue, which can exacerbate the agglomeration of the suspended nanoparticles, and material losses due to the backflow of reduction gas into the evaporation chamber, causing the premature formation of nanoparticles that are deposited on the walls of the reaction tube.
The numerical models used by Wang [8] and Ternik [9] have been shown to be able to simulate the behaviour of aerosolised droplets in a USP device effectively. Therefore, computer fluid dynamics (CFD) seemed to be an appropriate path for the purpose of evaluating the most optimal geometry and dimensions of the side inlet for H 2 gas into a three-zone USP device. To achieve this, we used a (CFD) programme, which is, nowadays, an essential tool for fluid dynamics.
With the ANSYS CFX programme, we can model and prepare geometries for simulations, and the programme also supports various types of files from other modelling programmes [10]. All computer fluid dynamics programmes follow basic conservation laws. As part of our scientific work, we performed the law of conservation of mass, the law of conservation of momentum, and the law of conservation of energy when performing all simulations. Detailed records of all laws are presented in references [11,12]. The "thermal energy" model was used to calculate the heat transfer. The Reynolds-averaged Navier-Stokes (RANS) approach was used for modelling the turbulent flow, with which one-equation and two-equation turbulent models can be used to calculate fluctuating current field quantities. In the presented research work, we used a two-equation k-ε turbulent model to inventory the turbulent flow [13].
As industrial applications of USP are still relatively rare [14,15], significant optimisation steps must be taken in order to improve the method's viability and efficiency within an industrial setting [16]. In order to obtain sustainable mixing and prevent the agglomeration of nanoparticles, which is common in the current design of the USP device, the optimisation of the reaction gas inlet was performed in the presented study. The obtained results indicate that the proper design of the reaction gas inlet can eliminate backflow (leakage). This results in the production of nanoparticles with different sizes and characteristics for a broader range of operating regimes, and it brings the usability of USP devices closer to the industrial scale. The presented study also indicates that numerical simulations are an appropriate tool for studying flow in a USP device. The obtained results give significant information about gas flow in the device and enable the study of possible future designs of USP devices.

USP Device
The USP device is shown schematically in Figure 1. The ultrasonic generator has a diameter of 20 mm, operates at a frequency of 1.6 MHz, produces aerosol droplets of 3 µm, and is able to convert 500 mL/h of liquid into an aerosol [7]. The furnace has a metal housing and fireclay brick insulation inside. Rock wool is used around the reaction tube so that it does not rest on the brick, and the insulation is tight. The furnace has a temperature regulator on the side for each zone separately. All glass tubes in the UPS device are made of quartz glass [17,18].
The USP device has 3 zones. The first is an evaporation zone in which the solvent evaporates, and it operates at temperatures of 100 to 300 • C. The second is the reaction zone into which the H 2 [19][20][21][22][23][24], enters and where the reaction with the precursor compound takes place. It has temperatures from 400 • C to 1000 • C. The last zone can be a reaction or cooling zone, so temperatures range from 20 • C to 1000 • C. Temperature setting and monitoring were performed with a temperature controller located directly on the furnace (Pixsys, model ATR244). N 2 was used as the carrier gas [19,21,25,26]. Gas flow control instruments were from the manufacturer Messer, with a maximum flow of 16 L/min or a minimum flow of 1 L/min at 1 bar, and with a possible step-scale of 1 L/min.
The photo in the process of the experiment was taken with a Nikon D7500 camera and a Nikkor AF-S f2.8 24-70 mm lens.
In the case of the presented research, where the optimisation of the H 2 side inlet was performed, we had to consider the following limitations and assumptions: (i) experimental testing was performed on the basic geometry of the USP device; (ii) the research work focused only on the flow in the reaction tube, and not on the entire USP device; (iii) the experimental testing was performed with coloured water vapour (green pigment), which was not taken into account in the simulations; (iv) the numerical simulation was performed on at least three different geometries; and (v) nanoparticles were not considered in the simulations and experimental testing.

Numerical
Model-Description of the Geometry 2.2.1. Description of the Geometry All geometries were formed from a reaction tube, with a length of 1780 mm, an inner diameter of 35 mm, and an outer diameter of 40 mm, and a side inlet, which has a centreline at 655 mm from the entrance of the reaction tube due to the shape of the furnace of the USP device. The side inlet has an inner diameter of 6 mm and an outer diameter of 10 mm. In the basic geometry, the side inlet enters the main tube at an angle of 90 • to the reaction tube, as shown in Figure 2.

New Geometries-A Case Study
As part of the research, we prepared new side-feed geometries, which are shown in Table 1. The following references [27][28][29][30][31] were helpful in selecting the different geometries. The geometries differed in the inlet angle of the side inlet for H 2 . The side inlet had an inside diameter of 6 mm in all cases except for the last geometry, where we considered an inside diameter of 10 mm. For the mesh of the basic model we chose CFD for "physics preference", CFX for "solver preference", and 10 mm for the size of the elements. The reaction tube and side feed were then crosslinked separately. For the reaction tube (Figure 3a), the size of the mesh elements was chosen to be 2.5 mm, and an "inflation layer" was added, which is a layer on the tube walls that makes the mesh denser and allows for a more accurate calculation of the boundary layer. This layer had a thickness of 2 mm and was distributed across 10 layers of elements. Based on the thickness and the growth factor, which was 1.2, the programme itself determined the thickness of the layers. For the side inlet (Figure 3b,c), we selected the size of the grid elements as 1 mm, to which we also added the "inflation layer". This layer had a thickness of 1 mm, 8 layers, and a layer growth factor of 1.2.
With this method, we got a computational mesh with 1.24 million elements. The references [32][33][34] were helpful when preparing the computational mesh. The same settings were used in the preparation of the computational meshes of all the new geometries.

Conditions
To determine the H 2 flow, it was necessary to make certain calculations, which showed that the theoretical minimum required flow was 3.3 L/min. Therefore, a minimum flow rate of 5 L/min was chosen for the study. A gas temperature of 20 • C was used for the simulations. The gravitational force (9.81 m/s 2 ) was considered in the simulations. It was necessary to determine the inlet, outlet, and walls on the numerical model. The inlet to the reaction tube to which N 2 was fed is marked as "IN1", the inlet to the side feed to which H 2 was fed is marked as "IN2", and the outlet from the reaction tube is "OUT" (Figure 4).

Conditions for Basic Geometry Disregarding Temperature
For the basic geometry, we decided to perform simulations at five different boundary conditions. We chose different combinations of H 2 and N 2 flows. With the selected combinations, we wanted to gain theoretical insight into the events in the evaporation and reaction zones of the considered UPS device, and, thus, indirectly, the influence of the flow of an individual gas on the events in the USP process. The selected flows are shown in Table 2.

Conditions for Basic Geometry Taking into Account Temperature
Subsequently, we also performed a simulation on the basic geometry considering the temperatures in all three zones of the furnace. The first zone is the evaporation zone, for which a temperature of 150 • C was selected, the second is the reaction zone, in which the temperature was 400 • C, and the third, the reaction or cooling zone, also had a temperature of 400 • C. For the flow boundary conditions, we used the flows of the third simulation of the basic geometry, with a flow rate of H 2 5 L/min and N 2 10 L/min, as it was necessary to ensure a sufficient flow rate of N 2 for the reaction to take place in the reaction zones. In the temperature-based simulations, we had to adjust for the carrier and reaction gases that the density of the gases varied with the temperature. We did this by writing the equations in the CEL module of the CFX programme. We also marked the points on the model where we monitored the gas temperatures ( Figure 5). These points included Point 1-beginning of the evaporation zone; Point 2-end of the evaporation zone; Point 3-end of the first reaction zone; Point 4-end of the second reaction zone.

New Geometries
For all new geometries, we had a flow of both gases of 5 L/min, as we wanted to determine a geometry in which H 2 would not leak back through the tube, even at the lowest flows. Simulations for all geometries were performed without considering the wall temperatures.

Experimental Setup
The experimental test in the USP device was performed by staining the water with a water-soluble organic dye powder of green colour (E142). When the ultrasonic generator was started, an aerosol was formed from the water, which was transported to the reaction tube with the help of N 2 , and H 2 was fed to the side inlet. The reaction tube was eventually opened without a collection system because we did not produce nanoparticles that would need to be collected. The experiments were performed at three different N 2 flows and a constant H 2 flow, as shown in Table 3. The flows were selected based on those used later in the simulations. Table 3. Gas flows in the experiment.

5
To facilitate the monitoring of the aerosol transport process in the USP device reaction tube and to observe the mixing of N 2 and H 2 , the experiment was performed by lifting the device cover as well as the quartz tube. For better visibility, we used an additional light source (a lamp) to be able to capture the key moments of the mixing of all the components (gases and aerosol) in the reaction tube with the camera. The results of the experimental testing were used further for the validation of the numerical results on the basic geometry.

Results of the Numerical Simulations
All simulations were performed as time dependent. The total time of each simulation was 10 s. The selected time step was ∆t = 0.01 s. We set the value of 10-4 s for the convergence criterion. The average time of one simulation was 8 h. The presented results show the conditions at 10 s of flow in the model. In all presentations of the results, the H 2 profile is coloured blue and the N 2 profile is grey.

Results of Simulations on the Basic Geometry
First, simulations were performed with different flows on the basic geometry. All the results shown in Table 4, show the state in the reaction tube after 10 s of current.
We found that in the simulations of the basic geometry, the N 2 flow had the greatest influence on the H 2 leakage back into the reaction tube. The leakage profile H 2 is wedgeshaped. The calculations show that at a flow of N 2 and H 2 5 L/min, the wedge length was approximately 250 mm. At flows of N 2 = 5 L/min and H 2 = 10 L/min, it was approximately 300 mm, and at flows of N 2 = 10 L/min and H 2 = 5 L/min, it was approximately 25 mm. For other flows, the leakage was negligible, or no H 2 leakage was obtained. Table 4. Simulation results of basic geometry (blue-H 2 profile, grey-N 2 profile).

Selected Flow
Image of the Simulation 1. Flow H 2 = 5 L/min Flow N 2 = 5 L/min For the temperature-based simulation, flow rates of H 2 = 5 L/min and N 2 = 10 L/min were selected. These flows were chosen because they proved to be the most optimal if we did not consider other external conditions. The inlet temperature of the gases was 20 • C. The graphical results of the numerical simulation are shown in Figure 6. Prior to the simulation, we marked the points on the model at which we measured the gas temperatures (5). At Point 1, the constant temperature was 20 • C. The highest calculated simulation temperatures at the other points were as follows: Point 2-39.8 • C, Point 3-164.6 • C, Point 4-238.4 • C. In this simulation, it was shown that H 2 was not leaking back, but sufficient temperatures were not reached (Figure 7).

Results of Simulations on the New Geometries
The results of simulations on the new geometries are shown in Table 5. Of all the selected different geometries, geometry No. 2, where the side feed was at the bottom of the reaction tube, the backflow of H 2 was the lowest. For other geometries, H 2 leaked back down the reaction tube to at least one-third of its length, its leakage profile being wedgeshaped. For this purpose, we also performed a simulation considering the temperature on geometry No. 2. Table 5. Simulation results of different geometries (blue-H 2 profile, grey-N 2 profile). For the simulation of the new geometry, taking into account the temperature, we determined that the flow of both gases was 5 L/min, the temperature of the evaporation zone was 150 • C, and the temperature of both reaction zones was 400 • C. The inlet temperature of the gases was 20 • C. The simulation shown in Figure 8 shows that H 2 would still leak back, but much less so than in the basic geometry simulation with the same flows.  Figure 9 shows how H 2 cooled N 2 at the entrance to the reaction tube since the N 2 had a higher temperature when leaving the evaporation zone than when entering the first reaction zone. The temperature measurement points during the simulation were the same as in the simulation with the basic geometry ( Figure 5). The calculated simulation temperatures at the points were as follows: Point 2-79.5 • C, Point 3-261.9 • C, Point 4-326.5 • C. The differences in the calculated temperatures between the basic and the new geometries were due to the lower N 2 flow. The calculations also show that the temperatures stabilised later than in the basic geometry simulation (Figure 10). The results of our simulations were also compared with other studies [34][35][36][37][38][39][40][41][42] and were consistent.

Results of Experimental Tests
During the experiment, it was observed that the ultrasonic generator generated a sufficiently large amount of aerosol, which the N 2 transported successfully to the evaporation zone of the reaction tube. During this aerosol transport, the pigment was deposited on the walls of the reaction tube of the USP device, as shown in Figure 11. Accurate observations and monitoring of the process show that before the side inlet, that is, before the entry of H 2 , the flow of aerosol and carrier gas was laminar ( Figure 12). In the part of the side feed where the H 2 entered, it was possible to see how it entered, the collision of the H 2 molecules with the wall of the reaction tube, and the process of displacing the H 2 with N 2 molecules, which are generally 14× heavier than H 2 . From this point on, before the first reaction zone, it was seen that the flow of the mixture of both gases and vapour was turbulent ( Figure 12). Based on the obtained images, we determined that the turbulent flow zone started at 15 mm from the side inlet, centrally in front of the first reaction zone (Figure 12). At the maximum N 2 flow rate (15 L/min), the experiment did not show that H 2 would leak back into the evaporation zone, as shown in Figure 13.
In the experiment, when the N 2 and H 2 flows were 5 L/min, it was possible to observe the leakage of H 2 back into the evaporation zone ( Figure 14). The leakage profile is wedgeshaped, with a measured length of 155 mm into the evaporation zone. The Video S1: H 2 -backflow is available online.   Among the results of the performed simulations and experiments, the following consistencies were confirmed. Changes in the flow at the entry of the H 2 were laminar to turbulent. H 2 displaced N 2 at lower flows in the upper part of the reaction tube, causing H 2 leakage back into the evaporation zone. The H 2 leakage profile into the evaporation zone was wedge-shaped; with the numerical simulation, the length of the H 2 leakage wedge was calculated to be approximately 250 mm, while the experiment showed a wedge length of approximately 155 mm. The reasons for the deviation in length are more than an incomplete consideration of the actual situation and include other restrictions, such as the accuracy of the gas flow regulators, and so forth. A high N 2 flow (>10 L/min) prevented H 2 leakage back into the evaporation zone, which reduced the process of displacing H 2 with N 2 .
The general conclusion of the comparison of the results is that the numerical simulations predicted a longer wedge range, where the return flow (leakage) of H 2 occurred. Thus, the results were not quantitatively comparable to the experimental tests. However, numerical simulations can predict the trend of H 2 behaviour of the flow in the reaction tube, so they can be used to study the process of optimising the design of individual elements of the USP device.

Discussion
Limitations and understanding of the USP process were the basis for conducting a theoretical study with numerical methods. We first performed simulations with minimum and maximum flow values (H 2 flow of 5-10 L/min and N 2 flow of 5-15 L/min) so that we could gain insight into the processes and events in the selected zones. The research work was related to limitations, among which the key one was not allowing for the change of the basic geometry of the USP device, which has constrained reaction tube dimensions-length = 1780 mm, inner φ = 35 mm, and outer φ = 40 mm; the side feed is located at 655 mm from the entrance of the reaction tube and has an inner φ of 6 mm and an outer φ of 10 mm. In this way, we obtained information for the possible optimisation of the structural parts of the USP device in the future. Despite the above limitation, it was possible to change the gas flows in the existing USP device, as enabled by the flow regulator for H 2 or N 2 . As it is difficult to regulate and measure the gas flows precisely, they were chosen so that they would not deviate too much and, thus, influence the course of the experiment (5, 10, 15 L/min). On the USP device, it was also possible to change the temperatures of individual zones. During the setting, the selected operating temperature rose evenly in all three zones. The experiment was performed by first mixing the pigment in deionised water, following the references of other authors [43], stating that the aerosol retains colour. The USP system was operated by pumping coloured water from a reservoir that was filled to the top before start-up into an ultrasonic generator. This was followed by the activation of the ultrasonic generator and, immediately afterwards, the opening of the N 2 carrier gas supply valve, which transported the aerosol into the reaction tube. During the transport process, aerosol droplets in the carrier gas stream were deposited on the walls of the feed tube (before entering the reaction pipe, where they condensed and collected on the inner wall of the tube). This was followed by the opening of the valve on the side inlet, allowing H 2 to enter the reaction tube. During this process, the formation of a wedge-shaped H 2 profile was observed, which did not separate the boundary sharply from the rest of the mixture (Figure 14). Before entering the heating part of the reaction tube, aerosol retention was observed at the bottom of the tube, which was seen by the fact that there was greater transparency in the upper half of the reaction tube than in the lower half ( Figure 14). This suggests that the mixed pigment increased the density of the aerosol, which was higher than the density of the N 2 . Based on the perceived effect of raising the carrier gas above the aerosol, it is necessary to point out that when preparing the precursor solution and the formation of the aerosol, it is necessary to know the density of the formed droplets. In this way, a suitable inert carrier gas can be selected, which must be denser than the aerosol to ensure a sufficient flow of the precursor compound to the reaction part of the USP device where the nanoparticles are formed. When monitoring the experiment, we could not provide the optimal viewing angle due to the placement of the USP device, so the images are not the best in terms of resolution. For further research and to achieve better resolution on the USP process monitoring images, it would be necessary to provide a more appropriate background and illumination of the reaction tube to obtain an even more accurate H 2 wedge profile. We found that the wedge formed by the displacement of H 2 was technologically problematic, as H 2 escaping through the tube back into the evaporation zone can cause too rapid a reaction with the precursor compound at a lower temperature, leading to irregular nanostructures with inadequate chemical composition. Such nanoparticles are unusable for many intended applications, as they do not achieve the required properties related to their size and structure [3][4][5][6][7].

Conclusions
The study came to the following scientific conclusions:

1.
The results of numerical simulations, in which the H 2 inlets above and below were compared, show that placing the inlet on the bottom is the best proposed design solution for the USP device.

2.
H 2 , which entered from the side inlet into the reaction tube of the USP device, escaped back into the evaporation zone because N 2 displaced it at lower flows (5 L/min). The leakage profile H 2 was wedge-shaped.

3.
To prevent H 2 leakage back into the evaporation zone, it was necessary to increase the N 2 carrier gas flow to at least 10 L/min.

4.
The experiment showed that the turbulence zone occurred approximately 15 mm from the side inlet in the reaction tube after H 2 entered at a flow rate of 5 L/min and a N 2 flow rate of 10 L/min.

5.
Due to the high flow of N 2 , cooling occurred in the reaction tube of the USP device, so it was necessary to ensure higher temperatures in all three zones of the USP device with temperature control, or to extend the lengths of the zones in the USP device. 6.
The results of the validation with the experiment showed that the numerical simulations give results with reasonable accuracy and are able to predict the flow in the glass tube of the USP device qualitatively. The numerical approach provides results with sufficient accuracy and can provide trends for the future optimisation of USP devices.

Conflicts of Interest:
The authors declare no conflict of interest.

CFD
Computer fluid dynamics RANS Reynolds-averaged Navier-Stokes USP Ultrasonic spray pyrolysis