Equipment

The experiments were performed in Chalmers’ pilot evaporator, with Figure 1 showing the most important equipment and sensors relevant to this study. The evaporator itself was of the vertical falling film type, with the liquor on the outer side of a vertical tube. It was designed to be large enough to produce results relevant to industrial conditions and simultaneously sufficiently flexible to be able to simulate a high variety of operational conditions. The evaporator tube was 4.5 m long with an outer diameter of 60 mm ( d o ), which gave it an active outer heat transfer area of 0.85 m 2 .

Figure 1

Simplified flowsheet of the pilot falling film evaporator used in the experiments. Sensors: CLD = chord length distribution, D = density, F = flow rate, L = level, P = pressure, R = refractive index, T = temperature and V = viscosity.

The salt solution or black liquor was fed into the top of the evaporator using a circulation pump: an overflow distributor integrated in the top of the tube ensured that a circumferentially uniform distribution of liquid, with a variety of flow rates, was obtained. The vapour and the concentrated liquid flowed from the evaporator tube into a tank serving both as a buffer volume (about 50 l) and vapour-liquid separator. The bottom of the tank was then connected to the circulation pump as shown in Figure 1; the vapour flowed into a condenser controlling the pressure in the system.

The pilot evaporator was equipped with sensors to monitor the operation and acquire data: the most important sensors are shown in Figure 1. The flow rate and density were measured using Endress and Hauser Promass Coriolis meters with a stated accuracy of 0.15 % for flow rate and 0.5 kg / m 3 for density. PT100 meters were used to measure temperature with a stated accuracy of 0.1 K. Yokogawa EJA meters were used to measure pressure with a stated accuracy of 0.20 %. The refractive index was measured at 589 nm using K-Patent PR-23-SD with a stated accuracy of 0.0002. The inner side of the tube was heated with condensing saturated steam. Further details of the pilot evaporator can be found elsewhere (Gourdon 2009, Karlsson 2017, Åkesjö 2018).

Scaling measurements

Scaling was measured in three different ways:

1. Continuous measurement of the global heat transfer coefficient from the condensate flow rate

2. Continuous measurement of local heat transfer coefficient in 7 different vertical positions from thermocouple pairs installed in the tube wall

3. Local thickness in 1 or 2 positions from mechanical measurements made at the end of the experiment (see Figure 1)

where d i is the inner diameter of the tube and k tubewall is the conductivity in the tube wall according to Johansson (2008). The different heat fluxes (both q global and q local ) are thereafter used to calculate the overall heat transfer coefficients (U):

which, in turn, is used to calculate the heat transfer coefficients of the black liquor:

A i and A o are the inner and outer heat transfer areas of the tube, respectively. h cond is the heat transfer coefficient of the condensing steam, which was estimated using a correlation for condensation on vertical surfaces by Numrich and Müller (2010). A m is the logarithmic mean area:

The measured heat transfer coefficients for black liquor decreased during the experiment for two reasons: scaling and increased viscosity (as the DS increased). The measured values were therefore adjusted for the effect of viscosity to isolate the effect of scaling using the following expression:

Here, μ is the measured viscosity and the parameter −0.41 is as suggested by Johansson et al. (2007). The validity of this formula was also checked by diluting the black liquor at the end of the experiment (see Procedure).

Finally, h adj was recalculated to an apparent scale thickness, δ, to make it comparable with the local mechanical thickness measurements:

Here, k scale is the thermal conductivity of the scales, which was estimated by Smith (2000) to 1.73 W/mK (1.5–2 W/mK) from literature data. The average values of h adj just prior to nucleation and at the end of the experiment were used to find h clean and h scaled .

Materials

The chemical composition of each black liquor and salt solution used in the experiments is presented in Table 1. The parameter w denotes the content of salt (i. e. sodium carbonate and sodium sulphate) per dry mass black liquor. The liquors used were sourced from two Swedish mills that process softwood black liquor: three liquors were received from Mill 1 (Liquors 1.1, 1.2 and 1.3, sampled in May 2018, November 2018 and February 2019) and one liquor from Mill 2 (Liquor 2.1, sampled in June 2019). Three different batches of salt solution were prepared by dissolving anhydrous sodium carbonate and sodium sulphate in deionized water. The pH levels of the liquors or solutions were not measured or adjusted since they were taken directly from the evaporation plant at the mills.

Soap and different fractions of tall oil brine from Mill 1 were used as additives in the experiments (sampled in May 2018). The mill uses a continuous tall oil process with a centrifugal separator (PINOLA TOPP from HEAD Engineering AB), from which the following streams or fractions were sampled:

1. Soap: the tall oil soap skimmed from the black liquor and fed to the tail oil plant.

2. Acid: the brine from the centrifuge, excluding residues from cleaning.

3. Sludge: the liquid produced when the centrifuge has been washed out with hot water.

4. Mix: the tall oil sent to the evaporator plant, including cleaning residues. It was sampled after being alkalised with sodium hydroxide to pH 12.9.

Figure 2

Appearance of the four additives used in the experiments after 2 months of sedimentation.

The chemical composition of the additives is shown in Table 2. The Sludge had significantly higher contents of magnesium and calcium. The Mix corresponds to a combination of Acid and Sludge, but it also contained some additional cleaning liquids from the tall oil plant (in addition to wash water from the centrifuge). The chemical composition here is comparable to the brine analysed by Lindgren et al. (2018), even though they used another tall oil process (HDS).

As can be seen in Table 2, the Acid, Sludge and Mix are comprised of components that are already present in large amounts in the black liquor, i. e. water, sodium, carbonate and sulphate. The table also includes the total amounts of other, more interesting, components contained in the additives to provide a more comprehensive comparison.

The Acid and Sludge were also analysed for organic composition using pyrolysis gas chromatography mass spectrometry (Py-GC/MS). The liquid water phase was found to be comprised mainly of dissolved lignin, while the sedimented solid phase (seen in Figure 2) was constituted of fatty acids and resin acids. Although this was true for both the Acid and Sludge, the latter contained a larger proportion of solid phase and its water phase also contained fatty acids and resin acids. This means that most of the unseparated tall oil (or extractives) are found in the Sludge.

Procedure

The study started with 13 pre-experiments with the aim of developing the experimental method and discovering whether or not a change in scaling due to the addition of brine could be detected. Liquor 1.1 was used in the pre-experiments. The main experiments were comprised of 104 experiments using Liquors 1.2, 1.3, 2.1 and three batches of salt solution of two different compositions (see Table 1). Additional sodium carbonate and/or sodium sulphate were added to the black liquor in the main experiments to be able to change chemical composition and crystallizing species: the salts were first dissolved in deionised water and added to the liquor before the experiment began.

Fresh liquor was used in each pre-experiment, which limited the number of experiments that could be performed. In the main experiments, however, the same liquor was used several times when appropriate (being discarded after the addition of an additive) to allow more experiments to be undertaken. The fresh liquor had relatively low DS, < 40 % (see Table 1), and was carefully mixed before pumped into the system to minimize inhomogeneity due to precipitated salt. The only inhomogeneity that could be visually observed was some soap floating on the surface in the liquor storage container, but it disappeared after the mixing. Each experiment used around 60 kg of black liquor, which was first heated to the operational temperature of 120 °C for at least 1 h to dissolve any precipitated salt. Evaporation was initiated and the condensate was returned to keep a constant DS: when stable operation was reached, the condensate was removed to increase the DS and the experiment was started. The temperature difference between the steam and liquor was kept at 15 °C, which is higher than industrial conditions in order to maintain a similar heat flux since the tube in the pilot is thicker (5 mm instead of 1.2 mm). The circulating flow rate was 500 l/(h·tube), which corresponds to a wetting rate of 0.8 l/(m·s).

At 40–45 % DS, 1.2 kg (in the base case) of the additive to be tested was fed into the system using a dosage pump. Nucleation normally occurred at 52–54 % DS and evaporation continued to 70–75 % DS. When the experiment had ended, the thickness of the scale was measured at 1.8 (for some experiments) and 3.8 m from the inlet (see Figure 1). The liquor was then diluted with condensate to around 55 % DS while the evaporation was still ongoing: when the heat transfer remained lower than the same DS before nucleation, scaling was confirmed. Samples of scales were collected from the evaporator tube at 1.9 m (for some experiments) and 3.9 m from the inlet. Finally, the black liquor was ejected before the system was washed at 120 °C with condensate overnight. When the same liquor was to be reused in the following experiment, it was diluted to around 40 % DS and then recirculated at 120 °C overnight to dissolve all crystalline material.

Scale samples were taken at the end of an experiment and some of them were analysed using the following methods:

1. Qualitative chemical composition of dried samples using attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR).

2. Quantitative chemical composition of dissolved samples using the same methods as for black liquors (Table 1).

3. Physical structure and elemental composition using a scanning electron microscope (SEM) with energy-dispersive X-ray spectroscopy (EDX).

Pre-experiments

The results from the 13 pre-experiments with Liquor 1.1 are presented in Figure 3. Scaling was detected in several experiment and all the three different scale thickness measurements showed the same trend. The mechanical thickness and local heat transfer measurements were generally higher and, as they were obtained mainly from the lower part of the tube, therefore indicated a conical scale distribution with thicker scales in the bottom. This was also confirmed visually and represents the normal situation as reported by Gourdon et al. (2007). These local measurements also showed larger variations as they become more sensitive for local variations in the thickness of the scales.

Figure 3

Experimental results obtained for Liquor 1.1 during the pre-experiments with and without additives. The thickness of the scales from the three methods are shown for all experiments except Exps. 1.1 and 1.2, in which the local heat transfer measurements were unsuccessful.

The first two reference (Ref.) experiments (Exps. 1.1 and 1.2) showed significant scaling (Figure 3). Interestingly enough, scaling was fully reduced in Exps. 1.3 and 1.4 when Mix was added, meaning that the experimental method seems to be able to detect scale inhibition caused by the addition of tall oil brine. Comparing Exps. 1.5 and 1.6, it seems that Soap had no effect on scaling, while Sludge reduced it in Exp. 1.7. It is, however, surprising to see that Acid increased scaling in Exps. 1.8, 1.11 and 1.13. Unfortunately, the reference experiments were not repeatable as the scaling decreased in Exp. 1.5 to none in Exp. 1.9, thus making it difficult for any definite conclusions to be drawn from the pre-experiments. As the scaling was strong in Exp. 1.8, the absence of scaling in Exp. 1.9 is probably not related to insufficient cleaning of the equipment, and all experiments were run within a few weeks so aging effects of the black liquor are also unlikely. No additional experiments could be performed with Liquor 1.1 since it was all consumed during the pre-experiments.

Conditions with predictable scaling in the reference case were necessary in order to improve the repeatability of the main experiments. The solution, as shown in the following section, was to add extra salt to the liquors to increase scaling. Otherwise the experimental method was found successful: scale inhibition could be detected and no additional changes were needed. However, to avoid potential inhomogeneity of the liquor to cause unrepeatable behaviour, extra care was put into ensuring sufficient mixing of the liquor before feeding the evaporator in the main experiments.

Main experiments

The first part of this section shows the results obtained from the main experiments without any additive. The second and third parts report the results obtained with additives, where the latter also includes the results from the pre-experiments.

The influence of carbonate and sulphate contents

Liquors 1.2, 1.3 and 2.1 did not scale at all during the reference experiments but, when additional salt was added, scaling was initiated during nucleation. The weight fraction of salt, w, at which scaling started is summarized in Table 3: it differed for the various liquors and was dependent on ϕ in the liquor. The underlying experimental points are shown in Figure 4. During carbonate-rich conditions ( ϕ > 0.68), i. e. in the dicarbonate and sodium carbonate sulphate regions, scaling started at w ≈ 23 % for Liquor 1.2, at w ≈ 21 % for Liquor 1.3 and at w ≥ 28 % for Liquor 2.1. The results were similar in both the dicarbonate and sodium carbonate sulphate regions, suggesting that there is no difference in scaling behaviour between the two regions.

Table 3

Salt content required to achieve repeatable scaling depending on the mole fraction of carbonate, ϕ, for the investigated liquors. Liquor 1.1 is included even though it could not be examined by addition of salt.

	Burkeite	Sodium carbonate sulphate	Dicarbonate
	ϕ < 0.68	0.68 < ϕ < 0.82	ϕ > 0.82
Liquor 1.1		> 15.6 %
Liquor 1.2	19.5 %	23 %	23 %
Liquor 1.3	17 %	21 %	21 %
Liquor 2.1	23–24 %	≥ 28 %

Figure 4

Experimental results from the main experiment without additives, but with sodium carbonate and sodium sulphate added to increase scaling. The colour of the points represents the amount of scaling. The horizontal lines represents the divisions of crystalizing species according to Frederick et al. (2004) and DeMartini and Frederick (2008).

During sulphate-rich conditions, ϕ < 0.68, i. e. the burkeite region, scaling started at lower contents of salt: Liquor 1.2 scaled at w ≈ 19.5 %, Liquor 1.3 at w ≈ 17 % and Liquor 2.1 at w ≈ 23 % (Table 3). Since all liquors scaled easier during sulphate-rich conditions, they contradict previous reports, i. e. that the dicarbonate region scales more than the burkeite region. Liquor 2.1 required very high levels of salt to form scales, higher than what is normal in industry. Since all liquors required different levels of salt, it seems that the scaling behaviour cannot be described by ratio and content of carbonate and sulphate alone.

The influence of additives

Experimental results for black liquors with and without additives are shown in Figure 5, where the experiments without additives are the same as those presented in Figure 4. Under sulphate-rich conditions, ϕ < 0.68, (Figure 5 left), scaling could be inhibited by adding Mix, Sludge or Soap to both Liquors 1.2 and 1.3 if w ≤ 20 %: higher salt contents could not be inhibited. The same behaviour at high salt contents was shown by Liquor 2.1, which could not be inhibited at w ≈ 23 %. One explanation might be that more additive is needed for inhibition at high salt contents: high salt contents ( w ≈ 24.5 %) and three times more Sludge than standard (Sludge high) were therefore tested for Liquor 2.1 and resulted in a clear inhibition of scaling. However, high sludge addition to Liquor 1.3 ( w ≈ 24 %) gave no clear inhibition even though the scaling was lower than without additive at the same w.

Figure 5

Experimental results for black liquors, with and without additives, at sulphate rich conditions (left), where ϕ < 0.68, and carbonate rich conditions (right), where ϕ > 0.68. The points that are labelled have the additive stated. The colour of the point is related to ϕ.

At carbonate rich conditions, ϕ > 0.68, (Figure 5 right), scaling was clearly inhibited by the addition of Mix and Soap in the experiments with Liquor 1.3 at w ≈ 22 %. For Liquor 1.2, neither Mix nor Sludge additions inhibited scaling but, as described previously (Table 3), it required a higher content of salt ( w ≥ 23 %) than Liquor 1.3 to scale. The point with the addition of Sludge at w ≈ 20 % for the same liquor is below this limit, so the low scaling rate is therefore not caused by the additive. Adding Acid to pure Liquor 1.2 was also tested, but it did not increase the scaling as for Liquor 1.1 in the pre-experiments. Addition of Mix to pure Liquor 2.1 gave no effect either. High Sludge addition was tested for Liquor 2.1 at very high salt content ( w ≈ 28 %) using similar procedure as during sulphate rich conditions: no inhibition, instead the scaling slightly increased.

Soap is often considered to cause scaling, and therefore two levels of Soap addition were tested for pure Liquor 1.2 (Soap and Soap high in Figure 5 right at w ≈ 14 %, corresponding to an addition of 47 and 94 g raw soap/kg DM), but no increase in scaling was detected. Uloth and Wong (1986) reported that sodium salt scaling increased at higher soap contents (up to 46 g/kg black liquor solids were tested, corresponding to unskimmed liquor): this was detected after 8 h of continuous operation in a rising film pilot evaporator, however, so the experiments might not be fully comparable (e. g. falling film evaporators are known to be less sensitive to scaling in comparison to rising film evaporators). Instead, Soap addition showed clear inhibition of scaling at w ≈ 20 % for Liquor 1.3 (58 g raw soap/kg DM). On the other hand, Liquor 1.2 just showed minor reduction at w ≈ 20 % (54 g raw soap/kg DM added) and in the pre-experiments the effect of Soap was unclear: it seems that Soap can inhibit scaling, but possibly less efficient than brine. Two experiments with Soap were also repeated after dilution to DS ≈ 40 % and recirculation overnight to dissolve all crystals: these points, denoted “dissolved” in Figure 5, showed less inhibition of scaling for some reason.

The results of the three batches of salt solution with and without additives are presented in Figure 6. The scale thickness without an additive was in the same range as, or lower than, that of black liquor under scaling conditions. The scale thickness in Batch 3 in particular was very low, which was caused by the scales falling off, and Batch 2 was a milder case where some scales fell off. As with black liquor, scaling was not significantly stronger in the dicarbonate region (Batch 1, ϕ = 0.85), even though this was observed by Gourdon et al. (2010a). It is interesting to note that additions of both Mix and Sludge increased scaling, which appeared to be connected to a lower tendency for the scales to fall off. The points denoted “dissolved” were repetitions of a previous experiment with an additive in the same manner as for Soap in Figure 5 (i. e. dilution and dissolution overnight to dissolve all crystals): then the scaling increased even more, and it also started at a lower concentration. Again, it is clear that also other factors than the crystallizing species can have a strong influence on the scaling.

Figure 6

Experimental results of the three batches of salt solutions with and without additives.

The influence of the content and dissolution of the brine

The brine contains large amounts of water, sodium, carbonate and sulphate, all of which cannot be active in scale inhibition since they are already present in large amounts in the black liquor. However, when evaluating the brine addition, only components that are potentially active in the inhibition is of interest, which can be quantified as the addition of interesting components or as organics according to the recalculation in Table 2. Furthermore, as the salt content in the liquor seems to be important, the amount of additive can be calculated per total mass of salt in the system. This way of treating the data also takes into account the fact that the content of interesting components varies between the different additives (Table 2): Acid naturally contains more spent acid (not active) whereas Sludge contains more of other, potentially active, components. The results can be seen in Figure 7, which shows all experiments with additives (except Soap) performed under scaling conditions, with the scaling rate as a function of both added interesting component (A) and organics (B). Both figures are very similar: the only clear difference is that the three experiments with Liquor 1.1 and lowest amount of addition (the pre-experiments with Acid) show lower addition relatively the other points in Figure 7B.

Figure 7

Scaling as a function of the amount of additive (excluding Soap) for all experiments under scaling conditions, i.e. w according to Table 3 or higher. The amount of Acid, Sludge or Mix added were recalculated (according to Table 2) to the addition of interesting components (A and C) and organics (B and D) per total mass of salt in the system. The pre-experiments (Liquor 1.1) and experiments with salt solution are included as well. C and D are the same as A and B, but experiments that were repeated after dilution and dissolving overnight (Dissolved) and experiments where new additive was added the day after overnight dissolution (Reused) are also included.

It appears to be some relationship between scaling and the amount of additive used: bellow around 7 g organics per kg salt, inhibition is unsuccessful (Figure 7B). This means that the reason for the unsuccessful inhibition in the salt experiments, as well as with the addition of Acid in the pre-experiments (Liquor 1.1), can be interpreted as the addition of active component was not enough. From 8 g of organics per kg salt and above, several experiments show a trend of decreased scaling with the addition of more additive. Liquor 1.1 shows a fully consistent trend. Liquors 1.2 and 2.1 show consistent inhibition in the burkeite region, while no clear inhibition in the carbonate rich regions. Liquor 1.3 shows a strange behaviour with more inhibition at lower amounts of additives.

Figures 7C and D also includes the “Dissolved” salt experiments previously seen Figure 6, along with some liquor experiments which were repeated after dilution to < 40 % DS and dissolution overnight (same procedure as the dissolved Soap experiments shown in Figure 5). Only two of these liquor experiments that had previously showed clear inhibition, did so the day after, which is in agreement with the results for Soap in Figure 5. This behaviour was investigated further: two of these “Dissolved” experiments, one with inhibition and one without, were diluted and dissolved overnight again with fresh Mix then added according to the normal method, shown as “Reused” in Figures 7C and D. In this case, the dissolved experiment with the higher content of additive and successful inhibition showed no inhibition when it was “reused”, while the opposite applied to the other reused experiment. In addition, one of the “Dissolved” experiments with Liquor 1.3 showed increased scaling. It is clear that the inhibition effect decreases (and, in some cases, even increases) after dissolution and kept hot overnight (120 °C according to the normal procedure). The form or state of the active component appears to be important, or it can degrade in some way. The behaviour is probably the same effect as seen with to low amount of addition ( < 7 g organics per kg salt).

Figure 8

The pilot evaporator tube after experiments with Liquor 1.2 (A before and B after taking scale samples) and salt solution Batch 3, ϕ = 0.75, (C and D). C shows an experiment when scales fell off, giving an uneven scaling layer. D shows the result after the addition of Sludge, where thick and even scales were formed (some had been removed on the right-hand side during sampling).

In the experiments with Soap addition (Figure 5), substantially higher amounts of interesting components and organics were added per mass salt, corresponding to 140–150 g/kg (using the value by Gullichsen and Fogelholm (1999) that half of the Soap is black liquor, implying the other half to be interesting components): the active components are likely to be the same, but the additives might be too different to allow this kind of comparison. The higher amounts can also be interpreted as Soap is less efficient, and as it is also a valuable by-product, Soap might be less attractive as an agent to prevent scaling.

Appearance and behaviour of the scales

Typical appearances of scales, once experiments had ended, are shown in Figure 8. The appearance of the scales was the same, regardless of crystallizing species: they were soft and clay-like for all the liquors tested and easily removed by scraping the surface, so perhaps fouling is a more appropriate term. The salt solutions produced mineral-like scales (i. e. hard and brittle) that had to be cracked during sampling. Mineral-like scales have also been found for black liquor, e. g. by Gourdon et al. (2008) and Karlsson et al. (2017) using the same pilot evaporator.

As mentioned previously, scales fell off at varying degrees for the reference experiments with salt solution. When Mix and Sludge were added to the salt solutions, however, this behaviour ceased: the scales appeared to be stronger and the overall degree of scaling become more severe. It was common that scales also fell off during the black liquor experiments, and especially so immediately after being formed during the primary nucleation phase. This was detected by the local heat transfer measurements, which first decreased strongly and then suddenly returned to the nonscaled value. If the scaling was strong, new scales were formed and the heat transfer decreased again, sometimes causing an uneven distribution. This is exemplified in Figures 8A (black liquor) and 8C (salt solution), where parts of the tube are free from scales or have significantly thinner scales. Liquor 1.1, however, showed low tendency to fall off, so this was not the reason for the low repeatability in the pre-experiments.

In the case of the pure black liquors, no scaling was formed at all (except for Liquor 1.1, which had the highest content of natural salt). However, at intermediate levels of w (still below the values required to achieve repeatable scaling, as shown in Table 3) scales were actually formed, although most either fell off directly after formation or were just formed locally and were therefore not detected in the final scale thickness measurements. These results imply that, as the salt concentration increased, the system became more prone to form scales and, as they also attach more strongly, the behaviour changed quite drastically from non-scaling to scaling conditions.

Composition and structure of the scales

A total of ten scale samples were analysed: six from black liquor experiments and four from salt solution experiments. They were first analysed by ATR-FTIR, which confirmed that they were comprised mainly of sodium carbonate and sodium sulphate. Two of the four scales from the salt solution experiments also contained traces of sodium bicarbonate ( NaHCO 3 ). All of the black liquor scales also contained additional components, but they were more difficult to identify: probably lignin and potentially also other organics were present, which is not surprising considering their physical appearance. The samples taken from Liquor 1.2 seemed to contain more lignin than those from the other liquors. The scales from experiments with additives did not show any notable difference.

The chemical composition of these ten samples of scales were also analysed quantitatively; the results are reported in Table 4 together with general experimental conditions. Although the level of calcium was significantly higher in the samples from experiments with additives (Samples 1, 5 and 10), it was still not more than by a factor of 2–3. The amounts of sodium carbonate and sodium sulphate (w) were above 94 % in the scales from salt solution (Samples 7–10) while around 60–70 % in those from black liquor (Samples 1–6): this relates to the mineral-like verses clay-like appearance discussed in the previous section. In addition to the results in Table 4, the ash content of another sample of scales from black liquor was measured to 80.8 %, confirming the presence of significant amounts of organics.

For the salt solution experiments, ϕ in the scales and solution generally agrees with the literature: a solution with ϕ = 0.85 (Sample 7) gave dicarbonate ( φ = 0.62 – 0.75) and ϕ = 0.75 (Samples 8 and 9) gave sodium carbonate sulphate ( φ ≈ 0.60). The latter crystal species, however, has only been reported previously for black liquor and not for salt solutions. An interesting observation that can be made here is that the experiment with an addition of sludge (Sample 10), which hade stronger scaling, had scales with a lower φ (0.53). In the case of black liquor scales, the relationship between ϕ (solution) and φ (scales) showed less agreement with the literature but, since the scales were not pure crystals, such a comparison is less relevant.

In the case of Batch 3 (Sample 9) in the salt solution experiments, most of the scales fell off when there was no additive, as mentioned previously. The data in Table 4, however, shows no significant difference to Batch 2 (Sample 7), where only minor parts of the scales fell off, that can explain the difference in behaviour. It is not known if the higher calcium content or lower ϕ in Sample 10 is connected to the strong scaling of Batch 3 with addition of Sludge.

More insight into the structural differences between scales from black liquor and salt solution was obtained using SEM (Scanning Electron Microscopy). Figure 9 shows the different appearances at 100 and 1000× magnification: the black liquor scales (Sample 2) were crumbly and fell apart easily, whereas the salt solution scales (Sample 10) were hard and dense. The large voids in the black liquor scale were probably created when it was fragmented during sample preparation and does not reflect its true structure.

Figure 9

SEM secondary electrons images at 100 and 1000× magnification, of a fragment of black liquor scale from Sample 2 (A and C) and a fracture surface of salt solution scale from Sample 10 (B and D). The large voids in the black liquor scale fragment were probably formed when the fragment was broken.

Figure 10 shows the SEM images of the black liquor scales (Sample 3). As for Sample 2 (Figure 9), the scales seem to be comprised of particles in the micrometre range, partly attached to each other in some form of aggregates. The images from the backscattered electrons (Figures 10B and D), in combination with EDX (Energy-Dispersive X-ray spectroscopy), show that the sample is not homogenous: brighter areas contained mainly sodium, sulphur and oxygen (i. e. Na 2 SO 4 ), while the darker areas also had high concentrations of carbon, some potassium and traces of chlorine (probably a higher content of black liquor residues). Although the composition was similar to Sample 2, the EDX spectra showed a lower content of carbon and higher content of sulphur. This agrees with the chemical composition reported in Table 4; the same agreement was also seen for salt solution scales.

Figure 10

SEM images of black liquor scale from Sample 3 at 1000 and 3000× magnification. A and C show the surface structure by detecting the secondary electrons, while B and D show the contrast in atomic number (larger atoms are brighter) by detecting backscattered electrons.

Figure 11

SEM secondary electrons images of salt solution scale Sample 10, showing the surface facing towards the evaporator tube at 50× (left) and 3000× (right) magnification.

The structure of the salt solution scales was dependent on the surface being investigated. The fracture surface perpendicular to the evaporator tube had a layered structure (Figure 9); the surface towards the evaporator tube, Figure 11 (left), was smooth but had a few cavities (especially in the left and right part of the left image). At 3000× magnification (right image), however, the surface is no longer smooth and appears instead to consist of agglomerated particles in the range of 1 μm upwards. This is likely related to how the crystals adhere to the tube steel surface and, during sampling, it was clear that this adhesion was weaker than the strength of the scale itself.

Calcium carbonate deposits

Insoluble scales built up slowly on the evaporator tube during the course of the experiments. Figure 12 shows the appearance after the 13 pre-experiments: the tube was dotted with fur-like deposits (needle crystals) that XRD analysis confirmed to be calcium carbonate (calcite) scales. This type of deposit has not been observed before on the pilot evaporator tube and must be caused by the high calcium content of the tall oil brine. Calcium carbonate scales were also formed during the main experiments: however, they did not seem to affect the scaling behaviour of the sodium salts. The tube was still cleaned three times, using sulphamic acid solution, over the course of the experimental campaign to avoid interference. The acid washing effectively removed all calcium carbonate scales.

Figure 12

Traces of calcium carbonate scaling found on the evaporator tube after the pre-experiments. The scales were subsequently dissolved and removed using a sulphamic acid solution.

The risk of calcium carbonate scaling occurring when tall oil brine is introduced is already known (e. g. as reported by Gullichsen and Fogelholm 1999), but whether or not it leads to actual problems probably depends on the situation at each mill. The two mills that were contacted in connection to this study did not reported any particular problems with calcium carbonate scaling caused by tall oil brine. It should be mentioned that the type of crystal species affects the scaling behaviour of calcium carbonate, and it has been shown that lignin can have a pronounced inhibiting effect on the calcium carbonate crystallisation (Wikander et al. 2006).

Formation of scales

The results obtained for both black liquor and salt solutions are interesting in that they show a behaviour that differs to what has been reported previously, i. e. that dicarbonate scales much more than burkeite. Euhus et al. (2002) detected strong scaling of dicarbonate at around 70 % DS (burkeite started to crystallize at around 55 % DS, leading to the depletion of sulphate and a shift in ϕ to the dicarbonate region), but no such second nucleation event was observed in the present study. Gourdon et al. (2008) reported strong dicarbonate scaling with black liquor at 55 % DS but they used additional salt, giving ϕ = 0.85 and w = 28 – 29 % (burkeite was not tested), which is an even higher w than that used in the present study. In another study by the same authors (Gourdon et al. 2010a), dicarbonate was shown to give stronger scaling than burkeite with salt solutions. Strong dicarbonate scaling has also been reported in mill studies, for example by Verrill et al. (2004). In the present study, it is shown that the amount of salt is important, and can be an alternative explanation for the strong scaling reported by Gourdon et al. (2008). It cannot, however, explain the results of Euhus et al. (2002) and Gourdon et al. (2010a). There seems to be additional factors affecting scaling that are not yet understood, requiring further research to be undertaken.

The reason for strong scaling in the dicarbonate region given by Frederick et al. (2004) is that finer crystals are produced there leading to agglomerate on the heat transfer surface, whereas burkeite gives rise to larger crystals that remain in the bulk phase. It was also shown by Karlsson et al. (2017) that larger crystals (>40 μm for salt solution) were necessary to avoid strong scaling. All black liquor scales in the current study had the same appearance; the SEM images show that they were loose aggregates with crystals sizes down to a mere few micrometres, but other components from the black liquor were also incorporated. This means that adhesion by weak intermolecular forces (in contrast to the strong ionic bonds in mineral-like scales) was the mechanism of scale formation for both dicarbonate and burkeite in this study.

The solubility is important for when crystallization starts, but it is not known if it also influences the scaling behaviour (i. e. crystallization kinetics). The additional sodium carbonate and sodium sulphate added in the experiments in this study influenced the solubility, which is also connected to the residual alkali (the concentration of sodium ions) due to the common ion effect. It is, however, unlikely that these changes are the reason for the deviation between the results here and previous studies.

The present study shows that scales can actually fall off the surface of the tube: this can be considered a new finding, as it does not seem to have been reported previously. In the case of black liquor, there was some kind of transition region where lower additions of salt triggered the initiation of scaling, but scales fell off the tube due to weak adhesion. When the salt content was increased even further, scaling was stronger and became permanent. Although a tempting explanation for this behaviour is the soft and loose appearance of the black liquor scales, the scales from salt solutions also fell off despite their being hard and mineral-like. Again, there must be an additional unknown factor that affects the crystallization process and alters the strength of the scales. It is also unknown why the black liquor scales in this study were clay-like whilst they were mineral-like in other experiments performed in the same pilot evaporator. Scales in industry, however, are often more clay-like (Lagerberg-Nilsson 2012).

When the metastable concentration limit is passed, crystallization is initiated due to primary nucleation. The liquor is then supersaturated, and the scaling is usually the fastest (Karlsson et al. 2013). This means that the scaling rate is not constant, which is especially true during batch evaporation as in the current study, and, in addition, if scales are also falling of the rate will fluctuate considerably. These are the reasons for why the scaling is presented as the accumulated thickness after the whole experiment in the result section: it was found more robust than a rate (e. g. mm/h) often used in previous studies. The experimental runtime under crystallizing conditions was 30–60 min without scaling and 70–100 min with scaling (more scales gave lower evaporation rate), but was also dependent on the solubility limit (i. e. at which DS scaling started). This means that the avarage scaling rate in this study for the experiments with high scaling are was in the order of 1 mm/h, which can be used for comparisions with other studies. Under industrial conditions, the startup procedure after washing implies operation at increasing DS with potentially no seed crystals and resembles the operation procedure in the current study. However, when operational steady state conditions have been resched, the scaling rate can be expected to be lower.

Inhibition of scaling

It is clear in the present study that tall oil brine and soap has the potential of inhibiting sodium salt scaling, although it seems as a minimum concentration is necessary in relation to the amount of salt crystalizing out. The inhibition is more efficient at the lowest ϕ, but there are still two experiments in the burkeite region without inhibition despite high amount of additive. The inhibiting effect also varies between different liquors, meaning that there are also other factors affecting the behaviour. As the mechanism of scale formation seems to be aggregation, the active component(s) in the tall oil brine or soap probably cause a weakening of the bonds between the crystals or prevent aggregation completely. It seems important that the active component(s) has the correct form as shown by the fact that the inhibition effect decreases when brine is added to the black liquor in advance: the active component(s) probably then dissolve or change form in another way. Taking all the results into consideration, the most likely active component appears to be one or several fatty acids or resin acids, and they should maybe not be dissolved if they are to have the ability to inhibit scaling. Sodium soaps are more soluble than calcium and magnesium soaps (which is a common problem when using hand soap in hard water) and, as tall oil brine generally contains less magnesium, calcium soaps are the most likely to precipitate. These results might seem, at first, to contradict Green and Hough (1992), who believed that the precipitates in brine increase scaling, but they discussed mainly calcium scaling (and fibres), which is actually in agreement with the present study.

An alternative active component that should be mentioned is calcium ions: calcium is abundant in tall oil brine and soap, but its solubility is complex as the ions interact with organic material. Shi et al. (2003) showed that calcium can inhibit the crystallization of burkeite, and Euhus et al. (2003) found that the addition of lime mud (calcium carbonate) reduced scaling. Further research is, however, needed to isolate the active component(s) and to gain an understanding of the mechanisms which, in turn, will provide a great opportunity for finding new, effective measures for controlling scaling.

Industrial significance

This study has shown that the amount of salt in the liquor (w) seems to be more important than the ratio between carbonate and sulphate (ϕ). In industry, it is common that ash from the recovery boiler, which consists mainly of the same salts but with a low fraction of sodium carbonate ( ϕ ≈ 0.3), is added to the evaporator by first mixing it with black liquor. If the concentration of the liquor is below the solubility limit of these salts, which is common, they will dissolve and increase the concentrations of sodium carbonate and sodium sulphate in the liquid phase. The ash typically accounts for at least 5 % of the black liquor dry solids, but can vary from 4 to 15 % according to Guimaraes et al. (2014), which is in the same range as the amount of salt added here (0–14 %). An important consequence of this is, therefore, that adding and dissolving ash can cause a shift from non-scaling to scaling conditions, even if a shift is made simultaneously from the dicarbonate to the burkeite region. From this perspective, it is much more beneficial that the ash is added to the liquor well above the solubility limit (e. g. to the firing liquor).

The amount of additive used in the experiments corresponds to the amount of brine added to the evaporator in Mill 1; depending on the salt content of the liquor, the amount of ash added and the crystallizing species, it might then be close to the limit required to achieve inhibition. The scaling in Effect 1C in this mill, operating at around 60 % DS, was investigated by Karlsson (2020): scaling generally disappeared when brine was added, but there were also periods when there was no inhibiting effect. There seems to be a potential for improving the situation in the mill by increasing the addition of brine and reducing the amount of salt. This is, nevertheless, not an easy task: the production of brine is dependent on the amount of soap, and the amount of salt in the liquor is controlled by other parts of the mill. Positive action that can be taken is to have a high degree of soap separation as possible (i. e. have a high production of tall oil) and operate the tall oil plant as smoothly as possible. Another, more unconventional idea, would be to add soap to Effect 1C. Adding ash above the solubility limit can also be considered, as mentioned previously.

The results also show that the way in which the brine is added is important if the inhibiting effect is to be retained: it should be added as close as possible to the scaling evaporator. This has also been confirmed in industry by personnel in several mills, interviewed by the authors of this study, who reported a clear decrease in scaling when the addition of brine was changed from the thin liquor to Effect 1. There is also a potential for reducing problems of calcium carbonate scaling when the brine is added to Effect 1: fewer effects are exposed to the high levels of calcium, and because Effect 1 is often less sensitive to scaling.