Experimental Investigation of Deposition of Silica Nanocolloids by Depressurizing Supercritical Water Vapor

Abstract

This article presents the results of an experimental investigation of silica deposition from depressurized supercritical steam. The case investigated is relevant for supercritical geothermal reservoirs with high temperature and pressure, where silica content is significant and deposition occurs rapidly upon depressurization. The purpose of the presented experiments is to accurately measure the deposited mass in two different areas in a flow tube and mathematically relate the measurement to particle formation behavior. In addition, SEM analysis permits valuable insight into the morphology of the scale formed under these conditions. The measured deposition is caused by silica solids formed when depressurizing supercritical fluids from around 350 bar and 500 °C by an isenthalpic valve to a state of superheated steam and pressures ranging from 60 to 150 bar. A test rig was designed, fabricated, and used for this purpose. The deposition mechanisms differ from silica particle formation in the water phase and the limited experimental research for the investigated conditions makes the gathered data highly interesting. The measured results are compared to validated models for deposition in straight pipes. The knowledge obtained on silica solidification and deposition can be used to optimize steam treatment of high-temperature pressurized geothermal sources for maximum power utilization by aiding in the development of advanced prediction tools for scaling and mineral extraction.

1. Introduction

Geothermal energy will be an important source of weather-independent base load and low-carbon emission energy in the near future. Geothermal energy from supercritical sources are superior due to their relatively low power cost if utilized efficiently [1,2,3,4]. Silica solubility in superheated steam increases with increasing pressure [5], and observations indicate that the kinetics of the precipitation process in pressurized steam will differ from the solidification of silica in liquid water [6]. Research, especially experimental data, on the precipitation of silica from superheated geothermal steam highly supersaturated with silica is scarce because the exploration of geothermal wells from supercritical waters, where the silica content in the steam phase is expected to be high, is a relatively new field. Knowledge of particle number density, size development, time scales of growth, and particle behavior in different depressurization scenarios is essential to efficiently handle deposits and minimize scaling in inconvenient parts of steam processing.

This paper considers production from hot reservoirs, inspired by deep wells near magmatic areas as attempted, among others, by the Iceland Deep Drilling Project (IDDP). The IDDP-1 well proved that production close to magma was possible. The record of the “hottest producing well on the planet” was achieved in 2010 [7,8]. If supercritical fluids with enthalpy around 3000 kJ/kg can be used directly in a steam turbine, the electric power per well can be close to 49 MW for a well with a flow rate of 50 kg/s, and the thermal efficiency, ${\eta }_{th}=W_{out}/Q_{in}$, corresponds to approximately 50% [9]. To compare, a conventional well tends to range between 8 and 20%. The direct utilization of high-enthalpy geothermal fluids represents a significant decrease in energy cost as power output per well is increased and a large portion of the investment cost relates to drilling. Conventional silica handling usually involves the quenching of the fluid. This may decrease the electrical power potential to 30 MW per well [9]. Improving methods for silica handling is therefore important to maximize power output, and gaining more knowledge about precipitation processes is an important aspect of this. Refs. [9,10] compare different technologies for the utilization of geothermal steam from supercritical fluid reservoirs with regard to power output. They consider the precipitation of silica and the formation of hydrochloric acid, respectively.

Geothermal water interacts with minerals in the rock reservoir. The silica content in a water reservoir is estimated based on the equilibrium concentration of quartz in the hot water. The dissolution of silica in the reservoir will follow the simplified reaction rate SiO₂ + 2H₂O = H₄SiO₄. Here, H₄SiO₄ is silicic acid dissolved in steam or liquid water. The precipitation of solid silica after pressure drops, as in the tests described herein, will follow the same reaction in the opposite direction. The crystalline forms of SiO₂(s), like quartz and cristobalite, are not common in geothermal systems as they have relatively slow crystallization rates. Amorphous silica, on the other hand, is a non-crystalline type of silica where the reaction rate is rapid enough to cause problematic precipitation in a geothermal plant. Changes in equilibrium conditions cause supersaturation, which is the chemical driving force behind silica precipitation. Supersaturation, S, is defined herein as the ratio of the actual concentration of silicic acid dissolved in the solution to the equilibrium concentration based on amorphous silica, $S={\displaystyle \frac{c_{act}}{c_{eq}}}$, where the concentrations are usually given in mg SiO₂ per kg water.

When the density of superheated geothermal steam is reduced due to pressure drops in the process plants, for instance, sudden and drastic changes in silica supersaturation are expected [11]. In this case, the dominant mechanism in the formation of a precipitate is bulk polymerization into nanocolloids, i.e., nano-sized particles. Polymerization occurs as monomers condense and bond, giving rise to nuclei and, further, nanocolloids. This is a form of homogeneous nucleation and the process is often associated with high deposition rates [12].

In Figure 1, precipitation through homogeneous nucleation, agglomeration, and deposition is illustrated. Downstream of the orifice (ii), nuclei and monomers have been generated in bulk fluid. The number of particles generated and the initial particle sizes will depend on supersaturation and the rate of change in supersaturation with time. The system moves towards equilibrium ((iii) in Figure 1). At this stage, the monomers may either (1) attach to nanocolloids (larger particles), (2) dissolve in the superheated water vapor, or (3) form new nanocolloids that further agglomerate into larger particles. The system strives towards equilibrium and fewer monomers will eventually become new stable nuclei. The larger particles will continue to grow until equilibrium is reached but at a slower phase (iv). From the various sources on silica kinetics in liquid water [13,14,15], the reaction rate of monomeric growth of the particles can be assumed to be relatively slow compared to agglomeration. Also, calculations by classical nucleation theory, as presented in [16], indicate that the polymerization of the majority of the supersaturation will occur within nanoseconds and that the remaining concentration available for monomeric growth becomes negligible.

The agglomeration rate is controlled by the particle transport mechanisms and particle–particle interaction forces. The small nucleus size particles in region (ii) are dominated by Brownian motion and the growth rate is relatively slow compared to the retention time in a high-velocity pipeline flow. As the particles grow in regions (iii) and (iv), the influence of turbulence becomes significant and the particle growth due to agglomeration accelerates exponentially [17]. Agglomeration is, however, also highly dependent on particle–particle interaction forces. Factors that influence the interaction forces include ionic strength of water, particle surface charge, pH, concentration, silica-water surface tension, and flow characteristics [12]. These, in turn, influence the collision efficiency.

Deposition onto surfaces will occur from the initial precipitation of particles (ii). From the literature, it is known that the deposition velocity will decrease with an increasing particle diameter in the diffusion-dominated regime [18]. Very high deposition rates are therefore expected in the vicinity of pressure drops, as a high number of fine particles are present there. The deposition can also be heterogeneous by nucleation directly on the wall or larger substrate. In the cases examined in this experiment, however, where the chemical driving forces are very high, forming primary nuclei in the steam is just as likely as nucleation directly on a surface. This is because the latter requires diffusion onto a surface, while the energy barrier for homogenous nucleation is low [12]. Down to a certain supersaturation and up to a certain particle concentration, the probability is thus higher for primary nucleation in the steam than onto a surface. The length scales of the particles generated are also very small compared to the radius of the pipe used in this experiment, which factors into the likelihood of diffusion onto the wall.

A calculation model that predicts the generation of solids through the numerical integration of classical nucleation theory and growth by agglomeration is presented by [16]. With initial supersaturation, pressure, temperature, and hydrodynamic conditions as the input, the model quantifies the total amount of deposited material in a specific region of a system. To complete the model, solid generation, growth, and deposition of particles onto a surface are calculated continuously. These three processes are separate, with co-dependences as illustrated by the reaction $H_{4}Si{O}_2\stackrel{Nuclation}{\to }Si{O}_{2(cn)}\stackrel{Agglomeration}{\to }Si{O}_{2(nano)}\stackrel{Deposition}{\to }Si{O}_{2(ppt)}$, where cn denotes critical nuclei from where these grow into nano, which are nano-sized agglomerates, and lastly, ppt, which denotes solid material deposited on a surface.

The agglomeration rate is affected by the concentration of solid particles in the solution. The size development of the particles is highly relevant to the rate of transport of solids to the surface, which is, in our case, the pipe wall. As solids are deposited and thereby lost from the solution, this rate will, in turn, determine the concentration of solid particles remaining. All size ranges of solid silica will deplete from deposition, including monomers, nuclei, nanocolloids, and agglomerates. As the deposition velocity and agglomeration rates vary in each size range, both processes will affect further agglomeration and deposition rates.

There is a general lack of research data concerning silica nanocolloid formation in steam, but a lot of experimental and theoretical research considering silica solidification and growth in liquid water can be found. Covering silica polymerization and growth into nanocolloids in pure water, there is [14,15,19], among others. In natural waters or simulated geothermal systems, we have [20,21,22,23,24], among others. Despite the vast amount of research performed on the kinetic processes of silica in liquid water, there is still strong disagreement both when it comes to reaction rates and the models used to describe silica precipitation kinetics [12].

Of the few experiments considering silica in the processing of superheated geothermal steam [25] is worth mentioning. He performed experiments on silica particle deposition and concentration in superheated steam using injected particles in the size range between 1 and 20 μm. These straight pipe experiments were, however, performed at moderate pressures (1.4 bar and 160 °C). The particles added to the steam were larger than would be expected after a significant pressure drop based on the classical nucleation theory calculation of minimum primary particle size followed by a limited period of growth by agglomeration.

A scaling experiment was performed by Trausti Hauksson Kemia and Sigurdur H. Markusson as part of the IDDP (Icelandic Deep Drilling Project) [8]. The fluids produced by the IDDP-1 well were exposed to a sequence of pressure drops induced by orifices in series with a pipe section in between. This is illustrated in Figure 2, where the measured deposited solids are given below the figure, and the pressures for each chamber are given above. A sharp increase in scaling was observed on the disks exposed to pressures below 75 bar. Uncontrolled parameters, such as the effect of other minerals, the amount of previously precipitated versus dissolved material in the fluid when entering the test rig, initial population distribution, temperatures, and exact turbulence and flow conditions in the chambers, limit the replicability of this experiment and make it difficult to pinpoint exactly what mechanisms are dominating.

Microstructural analysis showed the variable structure and cementation of the deposits. The characteristics of the scale changed from “flaky iron oxide” at 138 bar to “granular silica and iron oxide” at 95 bar and to “spherical and threadlike scale” at 34 bar [8]. In the photos with the finest resolution included in the report, 10–100 nm size particle shapes can vaguely be differentiated. Whether these larger structures were made up of smaller distinct units and what the number of newly generated nuclei with 1–10 nm diameter may be is not possible to determine at this resolution. Traces of iron and iron chloride were found in the deposits of the first orifices. Further downstream (from the sixth orifice after the impactor entrance), however, the deposit was pure SiO₂. Further on, threadlike structures of 100 nm diameter were observed, and, on the last orifice, thicker threads of approximately 1 micron diameter were also observed.

The silica concentration in this experiment was measured at 62 mg/kg upstream of the impactor. It is, however, likely that there is a significant amount of precipitated material in the IDDP steam when entering the system. Precipitation on a 10-micron filter was measured to 3.1 mg solids per kg steam. A fluid enthalpy of 3180 kJ/kg at a pressure of 75 bar will correspond to an equilibrium concentration of approximately 6.5 mg/kg silicic acid based on solubility data as discussed in [12]. Here, a linear correction based on actual density, compared to experimental density values and corresponding solubilities, is used. The process data compares well with the experiment discussed herein, and, as one of the few experiments regarding silica precipitation from a steam phase, aspects of the results are interesting in this context. The chemical composition of the fluid from IDDP-1 is, however, more complex and the hydrodynamic situation during deposition does not compare well. The hydrodynamic flow in the chamber of the experiment is not directly comparable to a straight pipe as each chamber between depressurizations will be the space between two circular plates. High velocities are expected at the orifice exit and there may also be regions with low velocity and recirculation zones. In a straight pipe, on the other hand, the velocity and boundary layer thickness are uniform. The measured deposition rates are therefore not expected to compare quantitatively with the experiment described herein.

The experimental results presented in this paper are used to map the particle formation behavior of silica precipitated from supercritical steam depressurized to superheated steam by measuring deposition at separate locations in a test rig. The aim of the experimental setup was to design a system with controlled hydrodynamic and chemical conditions that would be dynamically similar to the flow conditions expected in geothermal power plants handling fluids from deep hot wells, i.e., a turbulent flow of high-enthalpy, high-pressure water vapor.

The experimental rig used was designed and constructed to study the continuous deposition of generated and agglomerated solid submicron silica particles in depressurized supercritical steam. The generation and deposition of particles at different calculated supersaturation rates, S, was measured by quantifying the mass increase in two test sections, positioned at different locations downstream of a pressure drop, after a test period with steady flow. The experimental investigation thus aimed to validate calculated rates of precipitation, agglomeration, and deposition, as these three processes are related. The agglomeration and hence actual size range of particles is challenging to measure. A higher deposition rate in the first test section than in the one further downstream would indicate that the population balance has shifted towards a size distribution with larger particles (because deposition rate reduces for larger particles) and/or that the free concentration of solids is reduced (due to deposition between the test sections). For high velocities, however, the difference in dimensional deposition velocity between the two test sections was expected to be marginal. The generation of new solid particles, agglomeration, and deposition of particles are continuous processes between the two test sections, and they are treated accordingly.

2. Materials and Methods

The schematics of the test setup are shown in Figure 3. Purified water was pumped into an autoclave and heated to 500 °C and 350 bar pressure. Purified water with added silica was chosen as opposed to natural or geothermal waters to limit the uncertainties in the experiments. As there are knowledge gaps regarding the silica behavior in the supercritical steam and superheated steam phase, especially related to complex mixtures [12], the simplest chemical composition possible was chosen.

The autoclave was fed water through a pump containing a 250 mL reservoir of purified liquid water. The autoclave had a volume of 2 L and was prefilled with 20 g of 10 μm SiO₂ particles. The fluid in the autoclave was heated indirectly by a large electric air heater. During the heating process, the supercritical water was thus saturated with Si(OH)₄. Before entering the test rig, the supercritical water passed through a metallic filter with 0.5 μm pores to ensure that only dissolved silica (Si(OH)₄) entered the test rig. The silica particles in the bottom of the autoclave are reused for 5–10 tests. The original particles are large compared to the filter mesh and the overall mass of silica contained in the autoclave is large compared to what is dissolved. It is therefore safe to assume that only supercritical water saturated with silicic acid enters the test rig in this setup, even with an uneven dissolution of the original 10 μm solids in the autoclave.

During the experiments, the supercritical water flowed through the test section via the filter, a manual closing valve, and a pressure controller. The pressure controller maintained a constant pressure in the test section of the rig, and the superheated steam flowed through a test tube containing two 50 mm long removable sections where deposited solids could be measured after each test using a balance. Downstream of the test section, a manual control valve and an orifice setup were used to adjust the mass flow rate through the rig. A cooler condensed the steam to liquid water, which was collected. Complete control of all test variables proved challenging due to all the unknowns of operating under these very straining pressures and temperatures. Unpredicted material behavior, temperature-related wear of the equipment, and challenges in flow regulation caused several modifications and repairs on the test rig during the experiment. The results are divided into three separate test series, where each represents changes in procedure and equipment modifications to improve confidence in the results. Additional drawings and photos of the test rig and setup are given in Appendix C—Additional Photos and Drawings of the Test Equipment.

A manual closing valve separated the autoclave from the heated part of the test rig. This was only operated during the experiments. The pipeline, valves, test sections, and orifices were located inside a heated insulated box with a temperature approximately the same as the steam inside the pipe. Two heaters, with a 2 kW capacity each, were used to heat the box to the desired temperature. A thick layer of thermal insulation covered the outside. The primary purpose was to avoid temperature gradient biased results. Before each experiment, the rig was heated until a temperature sensor connected to the surface of the pressure regulator stabilized at the desired steam temperature of the test. In this way, we ensured stable thermal conditions inside the rig and minimized the effect of thermophoresis. The main pipe diameter, D = 1.38 [mm], was chosen to achieve sufficient turbulence with minimum flow rates in the test section. The Reynolds numbers, Re, generally ranged between 10³ and 10⁴.

A pressure controller regulated the pressure in the test sections. The controller consisted of a pressure regulator, “Equilibar Ultra High Temperature pressure regulator”, and a High-Pressure Dual Valve Vary-P Controller air supply, delivering a reference pressure to the regulator. In this setup, the Equilibar was used as a pressure-reducing regulator. This regulator is rated up to 413 bar and 500 °C. The controller was connected to a container with high-pressure inert gas (typically nitrogen), which needed to supply a higher pressure than the maximum process pressure to avoid damaging seals. As the maximum design pressure is 400 bar and the maximum operating pressure is 350 bar, a relatively large amount of 400 bar N₂ was required per test. The controller consisted of inlet and outlet valves, a proportional–integral–derivative (PID) controller, and a pressure sensor. Test section 1 was located approximately 52 mm downstream of the pressure controller. A second test section was located approximately 400 mm further downstream. At the test rig outlet, the water vapor was condensed by a small, water-cooled heat exchanger, and the condensate was collected. Figure 4 show pictures of the test rig and equipment used.

The manual control valves downstream of the test sections, used for flow regulation in the first test series, did not function as expected when heated to temperature. The flow was either too high, reducing pressure and temperature in the fluid reservoir, or too low, giving Reynolds numbers lower than what was the aim of the investigations. This instability during the tests is reflected in the relatively wide uncertainty bands in the results from the first tests.

Several redesigns were conducted, and a series of orifices were used to reduce pressure from the test pressure to ambient conditions in addition to one exit valve. At first, the orifices were mounted in a cylinder. This design had two major drawbacks; the orifices were impossible to retrieve after heating and pressurization of the cylinder and the very narrow holes (0.15–0.25 mm) in the orifices were vulnerable to larger particles in the steam, like loose corrosion products. Instead of inserting orifices into a cylinder, they were threaded to make it possible to screw them apart. In the third series, 3–4 of the orifices between 0.25 and 0.5 mm were used in addition to one control valve downstream of the orifices. Only the tests using the larger orifices were successful. This disqualified the higher test pressures planned for in the experiment, as obtaining stable flow in these tests without clogging the narrow orifice holes proved exceedingly difficult.

The main results of the experimental investigation were the increase in mass of each test section due to particle deposition. The effects of precipitation, agglomeration, and deposition were quantified by comparing the total mass of the test pieces before and after the test by use of a fine scale.

Based on the measurements, some calculations were performed to reduce the data to comparable measures. In addition, the contribution from thermophoresis had to be evaluated, as some temperature gradients were difficult to avoid in the test setup.

The wall particle flux, J = m/At [kg/m²s], was calculated based on the measured mass increase in the test section due to silica deposits, m, the test section surface area, A, and the duration of the experiment, t. The wall particle flux can be normalized by accounting for the estimated test segment inlet concentration of silica particles, c_SiO2 [kg SiO₂/kg water], and the friction velocity. The deposition can then be presented as dimensional deposition velocity, $V_d^{+}$, calculated as per Equation (1).

$$V_d^{+}={\displaystyle \frac{J}{c_{Si{O}_2}{u}_{\tau }}}={\displaystyle \frac{\dot{m}}{c_{Si{O}_2}{u}_{\tau }A}}$$

(1)

The friction velocity can be expressed as $u_{\tau }=\sqrt{{\displaystyle \frac{C_f}{2}}}u$, where u, is the bulk average fluid velocity. It was calculated based on steam flow rate measurements and a typical correlation for the friction factor for turbulent pipe flow, Fanning friction factor, $C_f={\displaystyle \frac{0.0791}{{Re}^{0.25}}}$. The concentration, c_SiO2, in a specific section of the pipeline downstream of the pressure drop was calculated by integrating the classical nucleation theory, particle growth, and deposition onto a surface numerically as presented by [16]. Likewise, this numerical calculation model was used to estimate particle size distributions since in situ measurements of the particle sizes in the steam were not possible during the experiments.

The thermophoretic force depends on the temperature gradient between the bulk fluid flow and the wall. The deposition of small particles is highly influenced by temperature gradients in the fluid. The effect of thermophoretic force can be significant for small particles even in moderate temperature gradients [18]. To quantify this effect in the experiments, the additional particle transport due to thermophoresis is estimated.

The correlation of Petukhov and Popov, as given in Equation (2) [26], can be used to determine the Nusselt number and the heat transfer coefficient $h_{th}$ by Equation (3). The friction, $f=(3.64\cdot \mathit{ln}(Re)-3.28{)}^{(-2)}$, is defined along with the correlation [26].

$$Nu={\displaystyle \frac{(f/2)RePr}{1+(13.6\ast f)+(11.7+1.8P{r}^{(-1/3)})\left(\right(f/2{)}^{(1/2)}\left)\right(\left(P{r}^{(3/2)}\right)-1\left)\right)}}$$

(2)

$$h_{th}={\displaystyle \frac{Nu\cdot k_{th}}{D}}$$

(3)

Following the recommendation of [27], the thermophoretic force is a corrected version of the expression derived by [28,29]. Compared with the expression derived by [30], the modified Cha–McCoy–Wood (Equation (5)) was shown to fit well for the whole range of Knudsen numbers. The dimensionless thermophoretic deposition velocity is calculated according to Equation (4), where m_p is the particle mass, $\tau ={\displaystyle \frac{{\rho }_p{d}_p^2}{18\mu }}\cdot C_c$, the particle relaxation time and $C_c=1+Kn\left(2.514+0.8e^{{\displaystyle \frac{-0.55}{Kn}}}\right)$, is the Cunningham slip correction factor as used by [18]. In this expression, $Kn={\displaystyle \frac{\lambda }{d_p}}$, is the Knudsen number and the mean free path between molecule collisions, $\lambda \approx {\displaystyle \frac{\mu }{p}}{\left({\displaystyle \frac{\pi R_{0}T}{2M_w}}\right)}^{1/2}$, is approximated according to the recommendations in [31], Chapter 5. In this expression, p is the gas pressure [bar], $\mu$ is the dynamic viscosity, T is the gas temperature [K], R₀ is the universal gas constant, and M_w is the molecular weight of the steam.

$$V_{thermophoresis}^{+}={\displaystyle \frac{F_{thermophoresis}\cdot \tau }{m_p\cdot u_{\tau }}}$$

(4)

The thermophoretic force acting on the particle is calculated from Equation (5) with $\alpha =0.22\sqrt{{\displaystyle \frac{\frac{\pi }{6}\varphi }{1+\frac{{\pi }_1}{2}Kn}}}$, ${\pi }_1=0.18{\displaystyle \frac{\frac{36}{\pi }}{\left(2-S_n+S_t\right)\frac{\pi }{4}+S_n}}, \mathrm{a}\mathrm{n}\mathrm{d} \varphi =0.25\cdot \left(9{\displaystyle \frac{c_p}{c_v}}-5\right)\cdot {\displaystyle \frac{c_v{M}_w}{R_0}}$. The momentum accommodation coefficients S_n and S_t are both assumed equal to unity. c_p and c_v are the specific heats of the gas at constant pressure and constant volume, respectively, M_w is the gas molecular mass, k_B is the Boltzmann constant, and d_m is the molecular diameter of the fluid.

$$F_{thermophoresis}=-1.15\cdot {\displaystyle \frac{Kn}{4\sqrt{2}\alpha \left(1+\frac{{\pi }_1}{2}Kn\right)}}\cdot \left(1-e^{-{\displaystyle \frac{\alpha }{Kn}}}\right)\cdot \sqrt{{\displaystyle \frac{4}{3\pi }}\varphi {\pi }_{1}Kn}\cdot {\displaystyle \frac{k_B}{d_m^2}}\nabla T{d}_p^2$$

(5)

An uncertainty analysis of the measured values was performed according to the method described below according to [32,33]. The estimated uncertainty in the experiments, $w_R$, is given according to Equation (6), where $w_n$, are the individual uncertainties.

$${\displaystyle \frac{w_R}{R}}={\left[{\left(a{\displaystyle \frac{w_1}{x_1}}\right)}^2+{\left(b{\displaystyle \frac{w_2}{x_2}}\right)}^2+\dots +{\left(N{\displaystyle \frac{w_n}{x_n}}\right)}^2\right]}^{{\displaystyle \frac{1}{2}}}$$

(6)

The measured variables in the experiment x₁, x₂, …, x_n were independent of each other. The result R only depended on the product of these so that $R=C{x}_1^a{x}_2^b, \dots , x_n^N$.

Each of the exponents and coefficients can be positive or negative. The results are often presented as R = B ± w_R, where B represents the calculated value of the result.

Table 1. Description of the main uncertainties.

Variable	Description	Wi
$L_{test section}$	Measured length of test section	$\pm$0.01 mm
$D_{pipe}$	Manufacturing tolerance in inner diameter of pipe	$\pm$0.14 mm
$\Delta m_{scale}$	Random uncertainty of the scale	$\pm$0.1 mg
$\Delta m_{increase}$	Weight increase due to other components than SiO2 like foreign materials, oxidation, or humidity. 1	+0.1%
$\Delta m_{decrease}$	An overall material loss was observed in each test. Control testing showed that the mass loss stems from mechanical material loss during connection and disconnection of fittings after exposure to high temperature. The test-specific material loss is obtained by control weighing and is subtracted from the deposition measure. Relatively high uncertainty is derived from the standard deviation and applied only where control weighing was not applied. 2	−0.5 mg
$c_{{SiOH}_4,quartz}$	Unstable pressure in the autoclave resulted in an uncertainty in the original concentration of dissolved silicic acid. There was also an uncertainty in the dissolution time at desired pressure and temperature. Higher than nominal flow rate out of the autoclave combined with low degree of mixing may have produced an undersaturated solution. 3	$\pm$20 ppm
$T_w$	The variable temperature in the heated chamber may increase or decrease deposition due to thermophoresis. This effect is significant when accounting for temperature fluctuation in only one direction. For some tests, the measured wall temperature fluctuates both on the positive and negative side, so that the effect could be assumed to cancel out. For most tests, however, there is an average lower or higher temperature measured on the pipe wall than in the bulk steam due to slow regulation. Thermophoresis is calculated and accounted for in the presented results based on the logged temperature in the autoclave and the pipe wall. 4	$\pm$1–5 K Measurement uncertainty, depending on the quality of temperature logging.
q	Variable steam volume flow and its possible effect on particle resuspension and validity of deposition model 5	$\pm$5 mL/min
$c_{{SiOH}_4,Amorphous}$	The test pressure in the test section varied, giving a variation in the equilibrium value of amorphous silica. In addition, data for determining the equilibrium solubility of amorphous silica in high-pressure steam is scarce, leading to unknown uncertainties in the actual concentration. This is in addition to the uncertainty regarding previously deposited material. 6	$\pm$10%
$r_p$	The mass average particle size is used in the calculation of dimensionless particle relaxation time and the graphical comparison of the results. While calculations indicate a relatively uniform distribution of particles, the SEM photo indicates a much wider range of particle sizes. Neither can be entirely accurate in representing the particle population balance in the steam. 7	Diameters ranging between 5 and 400 nm are observed on SEM.
t	Duration of experiment. The experiment duration was timed. This uncertainty accounts for slight logging errors, lag in valve response time, and variation in time required to establish stable flow conditions.	$\pm$1 min

¹ The EDS (Energy-Dispersive Spectrometer) analysis of the scale layer can be found in Appendix B. This showed traces of elements other than silica. The fraction was, however, very low. The contaminants stem mainly from corrosion products. During heating, an oxide layer formed on the outer surface of the test pieces. This process could be significant for new test pieces where no control weighing after scale removal was performed. The tests where control test pieces were used to verify the contribution of oxidation, however, showed no significant mass increase. These control test pieces were not connected to fittings and not exposed to steam, only temperature and air. The difference in dryness and possible effects of humidity in the surrounding air may affect the measured mass of a silica-covered sample. Although this was, at first, thought to be negligible, control weighing in some instances indicated a contribution to uncertainty. This is considered in the individual uncertainty analysis, where relevant. In these cases, the deviation between control measures is used as uncertainty. ² Only relevant for some tests. This potential error is corrected by control weighing the test sections after cleaning for most tests. In these instances, the uncertainty was reduced to 0.25 mg to account for humidity differences and/or incomplete removal of scale and other added masses only in the relevant direction. ³ This uncertainty is most relevant for the early tests (bulk 1), where the variation in the autoclave pressure was not logged and accounted for in the calculation. ⁴ The bulk steam temperature in the test section is calculated by considering the temperature probe inside the autoclave and assuming isenthalpic decompression across the regulation valve. Heat and friction losses are considered negligible. ⁵ This uncertainty may indirectly affect the results but is difficult to quantify in terms directly related to the deposition and was therefore not considered in the calculated uncertainty. ⁶ This has an indirect effect on the amount of solids in the solution by affecting the chemical driving force of phase transition. This effect is relevant in calculating a dimensionless deposition velocity and was not accounted for in the presentation of the absolute results. ⁷ The calculated average particle sizes are used when estimating the concentration development that is used in data reduction and thermophoresis estimation.

describes the uncertainties related to the measured variables of the experiments. The uncertainty values vary in each test as modifications were performed to limit uncertainty and both the measured parameters and results differ in each test. The values presented in Table 1 are examples and the specific calculated uncertainty for each test is given along with the results in Tables 2 and 3.

3. Results

3.1. Measured Deposition

The deposits accumulated on the test section surface areas during the experiments were primarily measured by weight differential in the test pieces before and after each experiment. The scale has an accuracy of $\pm$0.1 mg.

In

Table 2. Main parameters of the tests conducted.

Main Parameters
Test #	Autoklave Pressure [bar]	Autoklave Temperature [°C]	Test Section Pressure [bar]	Test Section Temperature [°C]	Reynolds Number [-]	Test Duration [min]
2	350	500	150	406	6001	36
11	342	500	84	357	4380	25
12	280	501	98	369	10,400	39
13	287	496	81	354	10,100	66
14	336	492	119	385	6610	134
15	341	498	140	399	3330	28
16	352	498	139	399	1170	60
18	338	497	105	374	3680	69
19	353	502	183	426	1310	60

, the main parameters of the tests presented are given. In

Table 3. Resulting deposition, uncertainty, and calculated effect of thermophoresis.

Results
Test #	Test Section 1 Measured Deposits [mg]	Deposition Rate Section 1 [mg/m2s]	Deposition Rate-Corrected for Thermophoresis [mg/m2s]	Uncertainty Test Section 1 [mg]	Uncertainty Test Section [%]	Test Section 2 Measured Deposits [mg]	Deposition Rate Section 2 [mg/m2s]	Deposition Rate-Corrected for Thermophoresis [mg/m2s]	Uncertainty Test Section 2 [mg]	Uncertainty Test Section 2 [%]
2	1.7	3.5596	4.1625	0.54	32	0.6	1.3023	1.5750	0.51	86
11	0.9	3.5577	3.1269	0.29	32	0.8	2.4121	3.2861	0.52	65
12	1	1.9328	1.3366	0.29	29	0.3	0.6011	0.3473	0.27	91
13	0.8	0.9137	2.8459	0.28	36	0.7	0.7413	1.8402	0.52	74
14	0.8	0.4500	1.3491	0.31	38	0.5	0.2813	1.0323	0.29	58
15	1.2	3.2305	3.1594	0.30	25	0.5	1.3461	1.6350	0.28	56
16	0.9	1.1307	1.1514	0.31	34	0.6	0.7538	0.7425	0.30	51
18	0.64	0.7026	0.4621	0.30	47	0.34	0.3732	0.1531	0.30	87
19	0.8	1.0051	0.9245	0.30	38	0.5	0.6282	0.7604	0.30	60

, measurements related to the experiments and respective uncertainties for each test are presented. The calculated contribution from thermophoresis based on the average wall temperature recorded is added to the results in a separate column. Some of the tests, especially in series 1, were susceptible to unstable flow in the flow section. This led to unstable pressures and temperatures in the autoclave. The instabilities were caused by problems with the pressure regulator and difficulty with the slow opening of the entrance valve in hot conditions. Test series 2 was prone to clogging in the orifice stacks used to regulate the flow and did not yield any tests with acceptable accuracy. The experience gained in the first test rounds led to higher overall confidence in the test results produced in series 3. In series 3, relatively stable flows for longer durations were achieved. Test pressures ranged from 70 to 140 bar. Continuous logging of the autoclave pressure and adjusting the expected initial concentration of silica accordingly also improved accuracy and confidence in the resulting data. The contribution from thermophoresis in each test is calculated by Equations (2)–(5). The dimensionless deposition velocity is converted to a deposition rate by Equation (1), using the estimated concentration calculated by the software provided by [16].

In Figure 5, the results from four tests are presented graphically, including uncertainty bands. The complete uncertainty analysis for each test is included in Appendix A. All these tests have negative thermophoresis, meaning that the average measured wall temperature, which is affected by the heated chamber surrounding the test rig, is higher than the steam temperature, causing a temperature gradient in the opposite direction to the particle transportation and thus slowing down deposition. The deposition rate termed “corrected for thermophoresis” adds the estimated contribution from thermophoresis to present the neutral value suited for comparison with general models. The effect of thermophoresis is noteworthy in most cases and is therefore also presented separately. The calculated contribution from thermophoresis in terms of dimensionless deposition velocity is expectedly lower in test section 2 than in test section 1 due to an increase in average particle size. For tests 14 and 19, ICP measurements, as presented in the next section, were also performed.

Other observations were also made during testing. The scale formed on the test specimen was not easily removed by isopropanol, acetone, or ultrasonic vibrations. NaOH was used to increase the pH to approximately 12 to efficiently dissolve the silica on the surfaces. Even then, some visible scale persisted. The mass of the test pieces that were used multiple times was reduced with an average of 0.7 mg during each test. A slight mass increase was expected for new test pieces due to the oxidation of exposed surfaces but control testing with unused test pieces, both connected and not connected, proved oxidation contribution to be negligible. The loss of mass observed was shown to be mechanical material loss during connection and disconnection after heating, as test pieces that were not exposed to the steam but connected to a dummy and heated had the same mass loss. The distinct characteristics observed on orifices as the pressure was reduced, further discussed in Section 3.3.2, show that the hydrodynamic and thermal aspects of particle behavior cannot be ignored when evaluating deposition rates.

3.2. Additional Verifications of Measurements

Additional methods to quantify deposited silica were tested to secure confidence in the results. Inductively coupled plasma optical emission spectroscopy (ICP-OES) was applied for silica quantification in some tests. This method generally has a higher accuracy than the weighing scale, but the uncertainty increases with low expected silica content, as is the case here. ICP is short for inductively coupled plasma. This is a method where samples are ionized by introducing them as a fine mist in a gaseous flow, typically ionized argon gas. Further, the samples are inductively heated, atomized, and ionized through high-temperature plasma (typically >3000 K). Lastly, the ions go through separation based on their mass-to-charge ratio. A mass spectrometer detects the ions, and the concentration of various species is determined based on signal intensities. These concentration measurements can have an accuracy down to parts-per-trillion (ppt).

Table 4. ICP results from test sections 1, 2, and at the exit after test 14 in measured mass of Si and calculated amount of SiO₂, based on the measured amount of Si and the molecular masses.

	Measured Si [mg]	Calculated SiO2 [mg]
Test piece 1	0.19	0.40
Test piece 2	0.12	0.26

shows the results from the ICP performed for the test pieces from test 14.

The amount of Si measured by ICP was used to calculate the complete mass of dry SiO₂ in the solution. The ICP results indicate a lower overall mass of deposited SiO₂ compared to the measured mass from the same experiment. Some Si may have been lost to the analysis due to the local formation of sodium silicate, but this species should, theoretically, not occur in the Si concentrations measured. The trapped remaining humidity, causing partly condensed silanol groups in the samples, may have increased the weighed results, but, as the test pieces were flushed with nitrogen when hot, the deposited material should be dry upon weighing and therefore comparable to the back-calculated SiO₂ from the ICP measurements.

The secondary verifications performed indicated that the deposition was less in the second than in the first test section. Although the absolute value of deposits was lower than the deposition measured by scale, the results correlate well with the anticipated decrease in deposition velocity between test sections 1 and 2. The difference between the weighed results and ICP results may be either an overprediction of the mass due to other contributors than silica or an underprediction of the ICP measurement due to an incomplete dissolution of the scale in the sample. Based on experience from cleaning the orifice disks, we know that some scales were difficult to remove even in a high pH solution. It is therefore possible that some Si remained trapped in the test piece and was not considered in the ICP.

3.3. SEM Analysis

The orifices placed directly downstream of the second test section had a flat surface that proved suitable for Scanning Electron Microscope (SEM) analysis. The conditions and particle size range at this point will be like that of test section 2, as no change in process conditions occurred and the orifices were located only a few centimeters downstream of the test section, giving little time to agglomerate or grow significantly in between. After each test, the test rig was flushed with nitrogen before cooling to dry out any moisture in the system. The components were then disassembled carefully, and the orifices were placed in a secure area before the samples were loaded into the SEM. Figure 6 shows some scaled surfaces of orifice plates after dismantling the test equipment. The SEM photographs presented in Figure 7 were taken directly from the orifice plate, so the deposited material was untouched and not in any way mechanically altered. As opposed to an in situ measurement, there is, of course, a change in temperature, change in the presence of steam, and time lapse, but drying and lowering the temperature should, theoretically, stop the growth process. Other effects on scale layer morphology and particle surfaces in the transition from hot steam to dry air at ambient conditions were not possible to observe in this experiment.

Energy-dispersive X-ray spectroscopy (EDS) showed that 99% of the surface consists of SiO₂. EDS was performed for a representative area on all samples to verify the composition of the deposit analyzed. An example of the results is given in Appendix B—EDS. A significant amount of copper was identified in the tests where copper paste was used as the sealant (tests 8 and 11). The sealant was only used between the orifice disks and cannot have affected deposition on the test pieces located upstream.

The quality of the test sample limits the resolution of the SEM photos. Generally, it is difficult to present clear images of structures of nanometer size. Figure 7a shows a uniform layer of silica scale with a porous, uneven surface texture at 1000× magnification. Increasing the resolution further to 10,000× magnification shows silica structures of various sizes and shapes in Figure 7b. In Figure 7c, one of the structures is further magnified to 50,000×. In this photo, you can see that many of the larger structures (200–500 nm), observed in Figure 7b, are loosely attached agglomerates of many much smaller particles (10–50 nm). One question that arises is whether these are agglomerated upon deposition or appear as agglomerates in the steam.

In the SEM photo shown in Figure 7d, the sample quality is good enough to make out some of the primary particles clogged together into a larger structure. This photo indicates that the particles are not necessarily spherical. Uneven asymmetrical round shapes and thread-like shapes are observed. This also strengthens the observation that the particle counts at 1 μm resolution is not necessarily representative of the particle size specter in the steam.

3.3.1. Particle Size Estimation Based on SEM

The photo presented in Figure 7b was used to perform a particle count and classification using the photo processing software Image-Pro v10.0.15. The minimum size counted in this analysis corresponds to one pixel and is hence limited to the resolution of the picture. The minimum size counted corresponds to 10 nm diameter particles for this photo. The real number of 10 nm particles must thus be assumed to be higher than counted and smaller particles may exist. Theoretically, the newly generated particles may be as low as 0.5–1 nm [16], depending on the chemical driving force for phase transition. The primary particles will thus not be visible in the size range obtained by this picture, but the contribution in terms of mass fraction is low. Figure 8 shows a partly processed photo in Image-Pro, where the counted deposits are marked with a blue color.

The average particle diameter in this count is 65 nm. Assuming the uniform density of the deposited material and evaluating the volume fraction of each particle size group, the relatively sizable number of particles in the range between 200 and 400 nm contribute significantly to the mass fraction of deposited material, giving a volume average diameter of 115 nm. As many of the larger structures comprise loosely formed smaller particles, assuming uniform density in both the larger and smaller particles may not be entirely correct. A few structures above 500 nm were observed, but most particles were below 200 nm.

Figure 9 shows the volume fraction distribution between the counted particles divided into seven groups. The three largest groups that appeared in the analysis are not accounted for as these had only one, three, and seven counted particles, with very uneven shapes. These are likely to be agglomerates with too low a contrast to be separated and are therefore excluded from the presentation.

The observations indicate a larger particle size range than can be explained by theoretical agglomeration from the critical nucleus size. Even though the water injected into the system is pure and the steam out of the autoclave contains a filter, we know there were some corrosion products in the steam. The larger particles may be due to solidification directly onto other solids in the steam. Both smaller and larger particles were expected, as the solution was not in equilibrium and new solid material was still generated.

3.3.2. Change in Hydrodynamic Behavior Through the Orifice Configuration

Figure 10 shows the surface of one of the orifice plates used in the second test series, where several orifices inside a cylindrical shell were used. All the holes are off-center in the plate and oriented 180 degrees offset compared to the previous orifice. This sample contains visible silica scale but also copper from the seals used in this test. There is a visible area with less scale where the jet from the upstream orifice impacts. This indicates a soft loosely attached scale, in agreement with the observations of silica scaling from depressurized superheated steam made previously [8]. Particle resuspension is probable in regions with high wall shear.

Some of the tests showed a distinct change in the hydrodynamic behavior of the silica scale between the orifices. The photos in Figure 11 are from the orifices retrieved from test 18. The pressure in this test was high and the deposition rate was low. We can see that the deposited material on the top orifice (a) is very thin. The second orifice (b), where pressure is reduced, had visible silica scaling and an area where deposited material is eroded due to the jet from the upstream orifice, as previously observed in Figure 10. In Figure 11c, on the other hand, where pressure is even further reduced, there is a material build-up that coincides with the hole in the upstream orifice. The hole was partly clogged in this last orifice, resulting in a gradual decrease in the flow rate during this test.

4. Discussion

Several trends can be identified in the gathered data. For instance, most tests show lower deposition in test section 2 than in test section 1, as expected. This tells us that silica deposition is severe directly downstream of a pressure drop and any scale mitigation should be placed as close as possible to the pressure drop. The observed clogging of exit orifices and visible deposition on orifice plates show that the concentration of solid silica in a straight pipe is still significant enough to cause severe process disruptions over 400 diameters downstream of the pressure drop with no mitigation.

The EDS (Energy-Dispersive Spectrometer) analysis performed proves that the amount of foreign species measured in the deposited layer is low. The test rig thus served well in producing relatively pure silica precipitate that deposited along the surfaces according to expectations. The tests were performed at various flow conditions. Using the same method of calculating dimensionless deposition velocity, we see some cases of decreasing deposition with an increasing Reynolds number. This could indicate that the sticking probability of the particles to the wall was less than 100% and decreased with increasing flow velocity. Increasing wall shear stress should, in theory, reduce the thickness of the inner boundary layer and therefore increase the rate of transport to the wall surface, but where larger particles are present, erosion or impact resuspension of the already deposited layer is common [34,35]. As observed in Figure 10 and Figure 11, the jet from the upstream orifice has blown clean the surface around the impact zone.

The uncertainty analysis elucidates a crucial aspect of the experiments. The short duration experiments that are only based on weighing have a relatively high uncertainty, as the measured deposited mass is close to the accuracy of the scale. When experiments were performed at higher test pressures, the equilibrium concentration of amorphous silica increased, and fewer particles were generated. With a lower concentration of solids in the steam, the deposition in the test sections was low compared to the possible error in the measurement. Therefore, the relative uncertainty would be higher for the high-pressure tests unless the experiment’s duration is increased accordingly.

The difficulty in keeping an exact constant temperature in the test rig wall contributed to uncertainty in some of the experiments. A temperature drop of 10 K across the boundary layer can increase the deposition velocity by 100% in the domain of some of the tests, and only a few degrees make a significant contribution to the deposition velocity. The effect of thermophoresis for silica particles in superheated steam is supported by the data gathered as the measured deposition fit better with the expected results when accounting for the measured temperature deviation. This information can be used to increase deposition in filters and areas with redundancy and decrease deposition in sensitive parts of the processing.

Particle size development is important when analyzing and using the data for validation. Figure 12a shows the calculated particle size distribution for three separate locations along the test rig, calculated by the continuous agglomeration of generated solids and size-dependent deposition onto the pipe walls, as described in [16], and by using parameters from the experiment. Figure 12b shows a comparison of the calculated distribution of particles in the steam at the orifice entrance and the particle count performed using Image-Pro from a sample of the deposited material taken at the orifice surface at approximately the same location for the same experiment. The calculations (green line) represent the distribution of particles in the steam. The sample count (pink line) is the result of accumulated deposits during the test. Although the effect of irregularities on the overall deposited mass is expected to be small, unavoidable unstable conditions during start-up and shut-down may have contributed to the distribution on the surface of the sample and therefore the particle count of deposited particles. In a slow startup condition, there is significantly more time to agglomerate larger particles. Particles large enough to be affected by inertia will easily deposit on the flat surface of the orifice. It is therefore not expected that the counts in Figure 9 and Figure 12b are perfect representations of the distribution of the steam at the location, nor necessarily along the walls of the pipe.

The overall reduction in concentration along the pipe in Figure 12a is due to deposition on the pipe wall. There is a slight reduction in the number of smaller particles moving along the pipe. Even as smaller particles are continuously generated in the calculation, the uniformity is predicted to persist both in test sections 1 and 2. In the diffusion-dominated domain, the smaller particles have a higher deposition probability than the larger particles. If the particles grow large enough for inertia to become dominant, on the other hand, deposition probability increases with increasing particle size. The average particle sizes measured by SEM after test 14 indicate that particles larger than estimated were present in the system. Particles displaying significant inertia effects cannot be ruled out based on the SEM photos, although it is theoretically unlikely in a steady-state situation with flow rates this high. The larger structures observed are therefore assumed to be caused by impurities or non-steady-state situations. The spectrum of observed particle sizes on the surface of the test samples investigated by SEM was wider than predicted theoretically, but differences between the distribution of particles deposited on the flat surface of the orifice and the calculated distribution in the steam are not surprising.

As the deposited material was relatively pure, larger particles of other species do not explain the larger structures observed. Some metal ions may, however, greatly affect the kinetics and growth of particles. It is possible that an enhancement of growth has occurred. Dissolved salts and the presence of fluoride ions in small amounts are known to influence nucleation in liquid water. Electrolytes promote faster equilibrium composition and fluoride acts as a catalyzer for silica polymerization [36]. As little as 1 ppm fluoride has a marked effect at low pH [19]. The EDS performed on samples from test 11 indicated traces of fluoride. An additional growth factor on the surface of the orifice may explain the larger particles observed on some of the samples from the orifice surfaces.

A main assumption in the calculation method and software presented by [16] is that when the supersaturation is high and the total concentration of Silicon in the solution is low, the monomeric deposition is insignificant, as the chemical driving force of nucleation is so high that solids will form in the fluid independently of surface contact. The remaining concentration of dissolved silicic acid thus rapidly decreases. This hypothesis was checked for all tests by using the concentration of remaining silicic acid along the timeline and calculating the transport rate to the wall based on the development of the critical radius and the correlation [37]. In Figure 13, the calculated contribution from monomeric deposition is presented along with the experimental results and the estimated overall deposition based on the calculation model discussed in [16], calculated for this specific test and corrected for thermophoresis. For the few tests carried out at higher pressures like test 19, however, the remaining concentration of dissolved silicic acid remains quite high through the test sections (approximately 90 mg/kg) while there is still a supersaturation. The calculated monomeric deposition is still negligible under these conditions. The contribution from monomeric deposition can be accounted for depending on the desired accuracy of results. The effect of thermophoresis, on the other hand, is significant in most tests, including test 14, as illustrated in Figure 13a.

The overall results compare relatively well with the estimations calculated from the theory described in [16]. The calculations related to deposition velocity and thermophoresis remain sensitive to parameters, like the mean free path between molecule collisions, that can be difficult to precisely determine [38].

The SEM photos provided in Figure 7 show structures that bear a resemblance to a jellification process observed in high ionic strength or high pH liquids [19,39]. The scala layer cannot be described as uniform or dense and the particles are almost exclusively non-spherical. These theories are further discussed in [12] and the effect of particle stability is further investigated in [40]. Whether the structures observed result from hydrodynamic aspects or chemical conditions is an interesting topic for further investigation and may be important in accurately determining the primary particle size and behavior.

5. Conclusions

The experimental rig designed to measure silica particle deposition from superheated steam in two straight pipes at various locations downstream of a pressure drop functioned well for the intended purpose, and the tests provided data with reasonable accuracy. The method provided relatively pure, newly precipitated material that deposited on the surfaces as very small particles, from a fluid that previous depressurization was without solids. This serves to prove that the generation of solids, growth, and deposition occur as interacting processes after a sudden shift in silica supersaturation in superheated steam and that the combined effect can be measured.

In addition to the data gathered, the documented learning on experimental control while measuring deposition from superheated pressurized steam in a flow-through system is valuable when planning similar experiments. The high relative uncertainty in some of the experiments limits the amount of data gathered, but the better part of the results compares reasonably well with theoretically calculated particle size distribution for the given conditions and dimensionless deposition rates according to acknowledged models. This indicates that with some tuning, it is possible to predict where in a system precipitation will occur and how much material will attach to the surfaces. This could be a major advantage when designing scale mitigation equipment, especially in the design and positioning of filter solutions.

As expected, the tests coherently show a lower deposition in test section 2 than in test section 1. This reduction in deposited material correlates well with increased particle size due to the growth and decrease in concentration due to previously deposited material, as predicted theoretically. Consequently, hydrodynamic and thermal conditions determining deposition velocity along the flow path are significant when evaluating silica scaling potential in supercritical high-enthalpy fluids and superheated steam systems. The observations discussed in Section 3.3.2 also highlight this.

The SEM analyses are of particular interest as the results prove that the deposited material is mainly silica, giving valuable information regarding the nature of the particles formed on a surface. Namely, deposited particles are non-spherical and have a wide size range.

The experimental results will be further used to validate and tune models for calculating silica deposition. Data confirming the particle size development and the effect of other chemical constituents on the processes would be useful to mathematically model the precipitation process in superheated geothermal steam derived from supercritical reservoirs high in silica. Supercritical geothermal steam is mineral-rich and methods of extracting solids from the steam may be interesting for the purpose of mineral extraction and scale mitigation in geothermal systems. In enabling the control of precipitation in the processing of supercritical geothermal waters, the energy efficiency of the power plant can be much improved, as discussed in the introduction to this research. This knowledge can in turn be used to enable better evaluation of various geothermal scale mitigation solutions engineered for maximum power utilization. The experiments show that to control deposition from a supercritical geothermal fluid, the significant effect of thermophoresis and particle charge can be used.