Deliquescence and Efflorescence of Hygroscopic Salt Particles in Dust Cakes on Surface Filters

2.1.1 Materials and Methods

Sodium chloride (NaCl) was used as hygroscopic salt, since it has advantageous properties for the experiments. Unlike many other salts, the deliquescence humidity of NaCl is relatively independent of the temperature. At 0 °C, it is at a value of 75.9 % and drops at 25 °C by only 0.5 % to a value of 75.4 % ⁵, ⁸. This facilitates the comparison of the investigations in the ESEM carried out at 1 °C and those at the filter test rig performed at 20–25 °C. Furthermore, the deliquescence humidity of around 75 % is not that high that other effects such as capillary condensation in the dust cake could overlap with the deliquescence effects. In addition, air borne NaCl crystals are cubic, which simplifies their localization and the determination of their size in the microscopic images. As a non-hygroscopic bulk material mainly finely ground limestone test dust (Ulmer Weiß XMF, x_50,3 = 3 μm) was used. To clarify certain processes, glass spheres (SiLibeads Glass beads Type S) with a significantly larger size were employed additionally.

The microscopical investigations were performed with a Quanta 450 FEG ESEM (FEI, Hillsboro, OR, USA) equipped with a Peltier cooling stage. In this case, the relative ambient humidity closely around the sample in the sample chamber is defined as a function of the controlled temperature of the cooling stage and the chamber pressure, which is generated by using pure water vapor. The microscope was operated at an acceleration voltage of 10 kV using backscattered electrons (BSE) for imaging with a gaseous analytical detector (GAD). The BSE signals enable a clearer distinction between different substances in a sample. Due to the high material contrast, e.g., between water and NaCl, liquid accumulations are clearly visible. This is due to the fact that the yield of BSE is proportional to the average atomic number of the element, which is hit by the electron beam.

High pressures have a negative impact on the image quality, as the electron beam is spread and weakened by the gas in the chamber, resulting in a significantly lower resolution of the images obtained. To overcome this problem, the samples were cooled to 1 °C so that the working pressure in the chamber could be limited up to 660 Pa (RH(1 °C, 656 Pa) = 100 %). In this pressure range, the deliquescence as well as the further absorption of the water by the liquid bridges could be observed.

The characterized samples were prepared in two different ways on sample holders suitable for the microscopical investigations in an ESEM. First, powder mixtures of hygroscopic salt particles and non-hygroscopic particles were created from bulk materials. The created powder mixtures were then used to generate packed beds on a carrying plate or on a fine wire cloth, which were mounted on a sample holder for the cooling stage in the sample chamber. In a second method, airborne non-hygroscopic particles and salt aerosol particles were subsequently separated on a substrate (like fine wire cloths) by filtration, so that a dust cake was formed in the end. A dust disperser with rotating brush helped to generate airborne non-hygroscopic particles from bulk materials. To generate hygroscopic salt aerosols from salt solutions with different concentrations, a TSI Constant Output Atomizer (Model 3076) in combination with a TSI Diffusion Dryer (Model 3062) was applied.

2.1.2 Characterization of Salt Particles and Particle Contacts in Dust Cakes

Previous studies on the deliquescence and efflorescence of hygroscopic salt particles in particle-wall and particle-particle contacts ⁶ demonstrated that investigations on a loose powder or packed beds with particles on a sub-micrometer or micrometer scale with scanning electron microscopy (SEM)/ESEM/energy-dispersive X-ray spectroscopy (EDX) techniques must be handled with great care and prudence. This results in the following approach to investigate phenomena with the prepared samples. After preparation, the samples were transferred to the sample chamber of the ESEM, which was then evacuated. Subsequently, the sample chamber was purged repeatedly with water vapor at moderate pressures (10–200 Pa) so that ideally only water molecules remain in the atmosphere of the chamber. The cooling stage was then set to a temperature of 1 °C. Thereafter, the chamber pressure was increased with a small difference between the setpoint and the actual value (maximum 15 Pa), up to a pressure of 400 Pa (RH < 65 %). After an equilibration period of at least half an hour, all relevant imaging settings were made.

Once all imaging settings have been completed, the corresponding parameters for the subsequent examinations with the sample are no longer changed. These can only be made meaningful while the sample is being scanned by the electron beam. However, this would in turn lead to an additional energy input. A very short dwell time of the electron beam per image/frame was chosen, i.e., recording times of less than 19 s. At the expense of image quality, the energy input is greatly reduced in this way. In order to achieve a further reduction of the energy input, the electron beam was blanked out between the individual image recordings. Then, only the pressure was subsequently varied in steps of 10 Pa. After reaching the desired pressure, an equilibration time of at least 5 min was maintained. Thereafter, the electron beam was switched on for the duration of a single complete scan of the sample before it was blanked out again.

2.2.1 Particle Movements

Fig. 1 presents the BSE images of glass beads in contact with ground NaCl on a fine wire cloth. The first image (Fig. 1a) shows the initial situation. Upon reaching the deliquescence humidity, liquid bridges form from salt solution drops. Due to the forces acting on the contacts, smaller glass spheres are moved towards a larger sphere out of their original position (Fig. 1b). After the subsequent efflorescence, the smaller spheres are moved further towards the larger one (Fig. 1c).

This type of particle movement, the so-called self-organization, has already been examined by Schröter and also made visible with the aid of an ESEM ⁹. In contrast to the experiments presented here, however, condensing water vapor was investigated there to form liquid bridges. The fluid bridges have a concave shape, similar to the liquid bridges that have been shown in the previous studies on the deliquescence of salt particles in particle-wall and particle-particle contacts ⁶. The amount of fluid that arises here during the spontaneous, complete deliquescence is significantly larger. Therefore, less of a bridge than much more a drop, in which the glass ball dips, forms. Unlike the liquid bridges, this drop is convex, which in principle indicates an overpressure within the liquid, which in turn would tend to exert a repulsive force on the particles.

On closer examination of the contact lines of saline solution and glass sphere, it can be seen that a contact angle of less than 90° is present. Therefore, it can be assumed, however, that a concave bridge forms at the moment when the drop touches the glass sphere. This exerts an attractive force on the sphere, causing the movement to be triggered. Since the drop is convex in shape, the movement of the particles ends as soon as there is a balance between the attractive and repulsive forces. At this equilibrium, the spheres, apparently depending on their diameter, are submerged to some degree. Smaller spheres seem to be further submerged than large ones. This is due to the fact that they displace a smaller volume at the same immersion depth.

On closer inspection, a concave fluid bridge between two spheres in the upper right edge of the image can be seen in the second image (Fig. 1b). These two spheres align according to the forces acting on them. A more detailed description of the processes can be found in the studies of Schröter ⁹.

As the relative humidity is reduced, the various fluid formations undergo efflorescence at different relative ambient humidity. As already described in the previous studies ⁶, the influence of the electron beam on this process cannot be completely neglected. Therefore, these observations should be taken with caution. Nevertheless, it becomes clear during the experiment that the concave liquid bridge crystallizes at significantly lower ambient humidity. Since both deliquescence and efflorescence are processes that are based on the partial pressure difference as a driving force, this observation is also plausible. The salt crystals that form, thereby strive for their natural cubic structure. The glass beads are partially enclosed in these crystals.

2.2.2 Caking

The following pictures (Fig. 2) illustrate a caking process of finely ground limestone and NaCl particles. For this experiment, NaCl particles were deposited on a packed bed of limestone powder as a salt aerosol stream overflowed the bed for a certain duration. In the first image (Fig. 2a), NaCl particles are clearly visible on the limestone particles. The salt crystals stand out visually, both because of their brighter appearance, as well as due to their cubic structure, which clearly differs from the broken structure of the limestone particles. Upon reaching the deliquescence humidity, liquid droplets are formed (Fig. 2b), as in the previous examination. Limestone particles are partially pushed aside and appear, as before the glass spheres, to penetrate into the drops to a certain equilibrium depth.

The biggest difference to the first shown experiment is displayed in the last picture (Fig. 2c). After the solution drops have undergone efflorescence, the limestone particles are not only visibly agglomerated, but apparently also coated with a crust of NaCl. It is obvious that the cohesion of the particles was very likely increased significantly by this procedure. Cubic crystal structures of NaCl are no longer recognizable here, which is influenced by the numerous contact areas to limestone particles, the irregular surface structure, and the wettability of surfaces of these particles.

2.2.3 Restructuring

The microscopic processes with a view to the deliquescence and efflorescence of salt particles in a dust cake were examined with a suitably prepared sample. In this case, airborne limestone particles were generated and then separated on a fine wire mesh by filtration with an air-to-cloth ratio of 180 m h⁻¹, until a closed and uniform layer of limestone dust particles with a particle mass per filter area of about 5–7 g m⁻² was generated. Subsequently, salt aerosol particles were separated on the previous dust cake by filtration with the same flow condition, i.e., deposited mass per filter area < 1 g m⁻². The as-prepared samples were mounted on a sample holder and characterized by the ESEM.

When evacuating the sample chamber, the previously formed dust layer on the sample holder is always partly destroyed by flow forces, so that the middle area of a sample is no longer available for targeted investigations. The following results were obtained along the edges of a sample where the original dust cake has been largely preserved.

With increasing ambient humidity, movements of the particles in the dust cake can already be detected below the deliquescence humidity. These movements occur only slightly in this first phase. In a second phase, when exceeding the deliquescence humidity sudden, much more intense movements of the particles in the dust cake were observed. A further increase in humidity then only leads to further lighter movements in a third phase.

It is presumed that local deliquescence of the salt particles occurs at the contact faces of salt particles with other particles in the first phase. The formed liquid bridges and the accompanied attractive forces, as already seen in accumulations of pure salt particles in the previous studies ⁶, ensure that the particles start to move until equilibrium arises. In the second phase, in which the deliquescence humidity is reached, the solid components of the salt dissolve, which leads to a considerable increment in the amount of liquid in the dust cake. The previously formed balance is abruptly disturbed significantly, resulting in a sudden restructuring of the particles in the dust cake. In the third phase, further absorption of water vapor by the saline solution from the ambient atmosphere leads to another disbalance and a subsequent further self-organization of the particles.

The subsequent decrease in pressure in the chamber results in further movements due to the evaporation of water from the saline solution and the concomitant change in the liquid bridges. The shrinking liquid bridges change the balance of forces, which act on the dust particles of the dust cake. This restructuring process in the dust cake is ongoing until the efflorescence of salts has been carried out throughout the dust cake. Fig. 3 presents two images of a dust cake recorded before (Fig. 3a) and after (Fig. 3b) the deliquescence and efflorescence of the salt particles.

There is a contraction of the particle layer visible. The particles that lay on the metal wires were pulled away laterally. The previous homogeneous, finely pored particle layer gains significantly enlarged pores between the particle dendrites, which itself become considerably more compact than the previous homogeneous particle structure. It can be assumed that such a change in the structure has a significant influence on the flow resistance of the particle layer.

3.2.1 Pressure Losses

Fig. 5 displays the pressure loss curve of a test on a standard test filter. The main dust cake was built up to a pressure drop of 900 Pa. Subsequently, a further pressure loss of 100 Pa was caused by the deposition of fine salt particles (x_50,3 = 1.2 μm) in the dust cake. The salt aerosol was generated by atomization of a 20 g L⁻¹ NaCl solution, and the raw gas with dry salt aerosol during the filtration has an RH of < 20 %.

During adjusting the RH of the air flow to humidify the dust cake, the gas extraction through the dust cake was continued, so that it was exposed to the ambient humidity (RH < 40 %). This resulted in a surprising jump-like increase in pressure loss of about 70 Pa. During the subsequent humidification (RH upstream slightly over the deliquescence humidity of the salt), an exponential decrease in the pressure loss results, which causes a pressure drop reduction of about 125 Pa. In the following drying process, first a sudden drop in pressure loss by about 45 Pa occurs, which is followed by a slight, softer increase of about 20 Pa. This characteristic course was found in all investigations. The extent of the pressure changes depends on the amount of the added salt particles – the more salt particles added, the greater the pressure changes.

At present, it is not fully understood why the sudden increase in pressure loss occurs when the dust cake is exposed to air with fairly low relative humidity (RH < 40 %). It is presumed that, with the relative humidity below 40 %, sorption layers due to water vapor absorption are formed on particle surfaces. Even though the absorbed amount of liquid is relatively small, it causes an apparent, measurable increase in the specific cake resistance in a finely pored dust cake. This is supported above all by the change in pore size of such pores on a micrometer und sub-micrometer scale. To clarify this phenomenon, comparable dust cakes made exclusively of limestone test dust were slightly moistened (40 % RH) and dried cyclically. When the moist air hits the dust cake, a sudden increase in pressure loss could also be recorded here. The additional pressure loss is significantly lower in the investigations without salt compared with the change in the tests with salt particles, which can be explained by the fact that subsequently added fine salt particles considerably reduce the pore size in the dust cake. During subsequent drying, this additional pressure loss was degraded again.

To highlight the characteristic course of humidification/dehumidification of the dust cake with a fraction of salt particles, the relative humidity of the air upstream and downstream the filter during this process is considered in addition to the pressure drop curve (Fig. 6). When the humid air stream (about 80 % RH) hits the dust cake, the pressure drop decreases exponentially, as already described. At the same time, the relative humidity of the clean gas (Fig. 6, RH downstream) initially rises abruptly, but then gradually adapts to the raw gas humidity (Fig. 6, RH upstream). When a constant pressure loss is reached, the clean gas humidity becomes equal to the raw gas humidity. In the subsequent drying process, the clean gas humidity drops more slowly than the raw gas humidity. Again, an adjustment of the humidity can only be determined when a constant pressure loss is reached. The difference between the raw and clean gas humidity during humidification and dehumidification of the dust cake is due to the fact that the salt particles absorb or release moisture and, thus, the moisture is removed from or added to the gas as it flows through the filter cake.

3.2.2 Regeneration

In respect of regeneration of the filters, the dust cake discharge is the main focus of the investigation. For this purpose, screening tests were carried out with rescaled test filters (exposed diameter = 40 mm, four pieces per test). Comparable dust cakes with the same final pressure loss under identical filtration conditions were generated by varying the amount of the salt particles added subsequently in the main dust cake. Before the regeneration was initiated, the dust cakes underwent a standard humidification/dehumidification process as described above. The tank overpressure for the regeneration was set relatively low (1 bar tank overpressure and a valve opening time of 80 ms, cf. ⁷) and uniform for all tests, so that some cake patches remain on the filter area.

After regeneration, the total filter area was inspected by using a digital incident light microscope; a following image analysis then yields an area-related regeneration result. Fig. 7 illustrates the single steps for the evaluation method. First, single shots of a filter at a specified magnification (20× magnified) are assembled to recover the entire filter area (Fig. 7a). Remaining cake patches are then marked with polygons and measured (Fig. 7b). A post-edited image is generated for further use (Fig. 7c).

Tab. 1 indicates the four test series for the investigation on the regeneration with varied parameters and evaluation results. Fig. 8 summarizes typical regeneration results of the four test series presented as post-edited images of the filters after regeneration. It can be clearly seen that the regeneration results differ significantly from each other.

All generated dust cakes have the same final pressure loss of 1000 Pa at an air-to-cloth ratio of 180 m h⁻¹, whereby in the test series a) and b) only limestone test dust was used to prepare the dust cake, and in the test series c) and d) different amounts of salt aerosol particles were deposited in main dust cakes prepared with appropriate pressure loss. A standard humidification/dehumidification process was carried out for the test series b), c), and d) before regeneration, so that test series a) serves as a reference.

A comparison between the test series a) and b) proves that if a dust cake is only conditioned with water vapor, particle cohesion in the cake as well as adhesion between the cake and the filter medium are enhanced simultaneously and to a comparable degree, so that no clearly better regeneration results are obtained. With a focus on the size of cake fragments formed at regeneration or cake patches remaining on the filter area, considerably larger cake fragments are found in both test series c) and d) compared with the test series without salt particles, which is often advantageous for the dust cake discharge. At a closer look into the cleaned regions on the filter area, virtually no small remaining cake patches are found in both test series c) and d), which is also a sign of significantly increased particle cohesion in the dust cake.

Since the remaining cake patches in cases of dust cake with salt particles are found primarily on the border to the frame of the filter holder, it can be assumed that solid bridges are also formed at contacts of dust particles to the frame of the filter holder, which is one reason for the poor regeneration results with significantly higher standard deviation (STD) in the test series d) compared with other tests.