The Effects of Water Recycling on Flotation at a North American Concentrator—Part 1

Water chemistry and its impact on mineral processing operations are not well understood and often not adequately monitored. CanmetMINING, as part of its water management research program, has been involved in a project initiated to identify opportunities for improving water recovery, water treatment, and recycling in the mining and mineral processing operations. One of the main objectives of this work is to evaluate and assess water chemistry and identify factors that impact mineral recovery, concentrate grade, and metal extraction efficiencies in order to understand and mitigate negative impacts of water recycling and improve process efficiency. In col-laboration with a North American concentrator, CanmetMINING has been involved in assessing the water chemistry in the mill and evaluating water recycling options for select process streams to reduce fresh water intake and maximize recycling. The overall goal of the project is to investigate options for water recycling (increase the thickener overflow recirculation from thickener overflow tank) without affecting nickel and copper metallurgy. The results of the sampling campaigns showed that the water chemistry of the streams was fairly consistent throughout the year with no significant seasonal variations. The laboratory tests illustrated that when higher quantities of thickener overflow from thickener overflow were used, the nickel + copper grade versus nickel recovery curves shifted towards lower values. These observations were observed for the plant water samples obtained in April, June and August 2019.


Introduction
Water is a critical component as both a transport and a reaction medium in mineral processing and hydrometallurgical operations. Mining companies primarily focus on the environmental impact aspect of water quality issues and not on process efficiency. The importance of water quality on minerals and metals recovery is grossly underestimated and not fully understood despite the fact that most of the water in mining operations is used in mineral processing plants and flotation circuits. Understanding the effects of variations of water chemistry and quality on the mineral processing circuits' operational efficiency is critical in guiding decisions with respect to the water make-up and recycling [1] [2].
The mining industry is being compelled to reduce fresh water consumption due to various reasons. Some of these reasons are environmental, lack of fresh water, and government regulations [3] [4]. These operations may recycle water from tailings dams, thickener overflow stream, dewatering, filter products, industrial effluents and treated sewage [2] [4] [5]. Water recirculation is advantageous because less fresh water is used, however, these streams may have a high level of impurities, which can affect flotation performance.
Mineral separation uses a large quantity of water, which represents 80% to 90% of the pulp in flotation [2]. Water in flotation circuits is used for transporting solids and is a critical part in the processing of minerals [4], therefore, understanding its effects on flotation is critical. However, mill personnel do not often have sufficient understanding of the effects of recycled water on flotation and do not have the time and resources to carry out a detailed investigation to take remediation actions. Unfortunately, there may be significant economic losses mill personnel may not be able to estimate.
Flotation performance depends on water quality. The use of seawater results in lower molybdenum recovery due to calcium and magnesium hydroxyl complexes [6]. Studies on the impact of sewage effluent and other waste streams with high organic concentrations have shown that dissolved and suspended organics lower flotation rates and cause frothing problems in different mineral processing operations such as copper-molybdenum flotation [7]. Therefore, before water streams are recirculated, a holistic approach has to be taken by mining companies. Firstly, all the water streams should be assayed for dissolved species such as calcium, thiosalts, sulphate, TDS, total organic carbon, total inorganic carbon, and microbial load. The sampling of the concentrator should be done over the course of at least one year to determine whether there are any seasonal trends.
Secondly, a detailed flotation program is required at the laboratory scale. The results of this test work will provide mill personnel with the effects of all the streams on flotation. Once this information becomes available, various ratios of the streams can be tested on flotation performance. If the streams are highly concentrated with species (in solutions and precipitates), various water treatment technologies may have to be considered.
Water treatment technologies should also be considered in the overall assess- thorough testing program has to be performed to determine the proportion of the stream and which streams have to be treated to maximize paymetal recovery.
Afterward a detailed economic analysis is recommended to justify the utilization of such technologies.
This paper consists of two sections. In the first section, the characterization of the process water, flotation feeds (the circuit has two parallel roughers-scavenger banks-side A and side B, which will be described later) and the thickener overflow, from thickener overflow tank, streams are discussed. The measurements such as pH, temperature and ORP of the streams are reported. In the second section, the results of flotation testing using process water (treated water that can be released to environment), thickener overflow and a combination of the two streams are discussed. The probable causes of the effects, pyrrhotite and gangue recoveries, of using thickener overflow on flotation are discussed.   Water samples for microbial analysis were decanted to separate the water and solids, then filtered through a Pall Supor 0.2 micron sterile filter unit. The filters were then processed using the Qiagen DNEasy Water extraction kit. Extracted DNA was quantified using a Qubit3 fluorometer with high sensitivity assay.

Laboratory Testing
The nickel-copper ore was crushed to −2 mm (−10 mesh), blended and split into 1 kg charges. Table 1 shows the external reference distribution for the ore used in this test work. All the relative standard distributions (RSD) were less than 5%, which means that the sample was well blended.
The charges were ground to 56% passing 75 micrometres using a laboratory rod mill. The grinding media used was a combination of mild steel and stainless steel rods. The percent solids used in grinding was 60%.
The laboratory rougher-scavenger flotation tests were done using a Denver flotation machine and a 1 litre cell. The reagents used were potassium isobutyl xanthate (PIBX) from Prospec Chemicals and Polyfroth W31 from Quadra Chemicals. The pHs for flotation were 9.2 and 8.0; these were adjusted using lime and sulphuric acid both from Fisher Scientific, respectively. Process water and thickener overflow obtained from the concentrator were used for the flotation tests ( Table 2).
The process water and thickener overflow samples were received from the concentrator on a monthly basis (some months sampling was not done due to mill shutdown). Only the results of the flotation tests using the water types obtained in April 2019, June 2019 and August 2019 will be presented in this publication. All plant water samples were taken on the same day. Table 2   was compared to the probability > F statistic (Pr > F which is equivalent to p) in the SAS output. If 0.05/m was less than Pr > F (SAS output), then the factor in question was considered to be significant at the 95% level. The methodology is described in [8].

Statistical Analysis
There are three major prerequisites for MANOVA to be applicable; the data have to be independent and normal and the variance has to be constant. These conditions were verified prior to performing the MANOVA analysis and were found to meet the requirements for all the tests discussed. The following are the null hypothesis tests used in the analysis of the test work results.

Manova
For the notation in this section, PW stands for process water and TK O/F stands for thickener overflow. Also, μ stands for the mean of the variable in question.
For example, μ Ni rec signifies the mean of the nickel recovery.
Null hypothesis

Confidence Intervals for Flotation Tests
The method described by Napier-Munn (2012) [9] was used to establish the confidence intervals (Cl) for the flotation tests.
where t = 2.92 for 90% confidence, s is the standard deviation and n is the number of replicates (n = 3 for the flotation testing in this work). Note that a 90% confidence level was used to create error bars for the flotation test results.

Results
The results of the streams that can impact flotation will be discussed. These are the process water, side A flotation feed, side B flotation feed, and thickener overflow streams. The sampling period was from November 2017 to June 25, 2019.
The process water comes from the tailings treatment system (this is clean water) and the thickener overflow is recirculated back to the grinding circuit. A portion of the thickener overflow is recirculated.
A. Di Feo et al. Table 3 shows the sampling dates for the process water, A side flotation feed, B side flotation feed, and thickener overflow. Certain streams were not sampled during this period.     The process water came from the tailings treatment system (this is clean water); therefore, the pH of this stream tended to be between 6 and 9 with one sur- Therefore, when the overflows of the copper and nickel concentrate thickeners and fines thickener were combined, the feed to the thickener overflow became alkaline. The process water and thickener overflow generally had lower temperature relative to the flotation feeds. This is to be expected because the water and ore were heated during grinding prior to flotation. Generally, the flotation feeds and the thickener overflow tended to have reducing pulp potentials. The low ORP may be caused by oxidation processes in the water and pulp [6]. Also, changes in the ore treated (different ratios of pentlandite, chalcopyrite and pyrrhotite in the ore) and water chemistry can cause variations in the ORP.

Stream Characterization
The filtrates for sulphate analysis of the sampled streams were frozen until assaying. When these were thawed (prior to analysis or assaying), white precipitates formed. To determine what these precipitates were, A side flotation feed and thickener feed #3 samples were filtered then the solids were dried and weighed followed by XRD analysis. Figure 5 shows the XRD (Rigaku, model Dmax 2500) analysis of the precipitates for the flotation feed side A only. The XRD analysis showed that the precipitates were gypsum. The results of the XRD analysis of the precipitates for the thickener #3 feed also confirmed that the  Table 3.
Generally, for all the streams and species, there are no trends between the seasons. However, in some campaigns, there were peaks in concentration of species such as in Figure 14 and Figure 15. These peaks in total organic carbon and total carbon may have been caused by higher collector and frother dosages in the flotation circuit. The thickener overflow that the concentrator would like to increase the amount recirculated, had higher concentrations of total inorganic    carbon (TIC), total organic carbon (TOC), total carbon TC and total dissolved solids (TDS) than those of the process water. The higher TC and TDS can have significant impacts on flotation [1]. For example, accidental activation of undesired minerals, increased gangue recovery through entrainment, slime coating, passivation of valuable minerals, etc [1]. Therefore, water treatment may have to A. Di Feo et al.  be employed to reduce the TC and TDS prior to recirculation. This is beyond the scope of this publication.
The average %solids of the campaigns for the process water, thickener overflow, side A flotation feed and side B flotation feed were 0.10%, 0.11%, 48.11% and 46.56%, respectively. The suspended solids in the thickener overflow were assayed. Figure 18 shows the assays of the suspended solids in the thickener overflow stream for the dates indicated in Table 3         Another observation worth noting is the concentration of calcium in the solids. For some campaigns, the % calcium in the solids was almost as high as 20%.
Scaling is already an issue at the concentrator and if the quantity of thickener overflow is increased, scaling most likely will become worse. The implication is more frequent shutdowns to clean or replace the piping, which would have a significant financial impact. The recommended next step is to carry out water treatment tests in the future to determine the conditions that would allow the reduction or elimination of calcium precipitate. Figure 19 shows the particle size distribution of the solids suspended in the thickener overflow. The 80% passing of the suspended solids was approximately 75 micrometers. These solids should be liberated because the feed size to flotation is 56% passing 75 micrometers. If these solids are ground further, fines will be produced and most likely will be lost in the final tailings. Therefore, filtering this stream and sending the solids directly to copper-nickel separation in the flotation circuit will increase the copper and nickel recoveries to final copper and final nickel concentrates, respectively.

Flotation Testing
In the plant, the rougher circuit consists of the primary rougher, secondary rougher and scavenger. The primary rougher concentrate goes directly to copper-nickel separation. Therefore, the nickel + copper grade of this concentrate is important because it influences the final copper and final nickel concentrate grades. If the primary rougher concentrate grade is too low, then the final copper and final nickel concentrates grade specifications may not be met.  Table 2. The water samples were stored in a refrigerator until further use [10]. Also, a weighted average was computed for the metallurgical results. Figure 21 shows the nickel + copper grade versus nickel recovery for the water types. The nickel + copper grade versus nickel recovery decreased as the amount of thickener overflow used increased. The reasons will be discussed shortly.    Figure 24 and Figure 25 show the pyrrhotite recovery versus water recovery and gangue recovery versus water recovery, respectively. For pyrrhotite, the relationship is not a straight line, which means entrainment was not the main recovery mechanism. The gangue recovery versus water recovery for the water types was fairly linear, which means that the main recovery mechanism for gangue was by entrainment. When thickener overflow was used, the gangue recovery by entrainment was higher relative to the other water types, which is most likely one of the reasons for higher gangue recovery observed in Figure 23. However, accidental activation may be another cause for higher gangue recovery; this will be discussed later in the publication. impact on these parameters. The MANOVA analysis was performed using the parameters days (blocks) and water ratio. Since two parameters were used, an α of 0.025 (α = 0.05/2 = 0.025 − Bonferroni adjustment) was used for the significance testing for both the days (blocks) and water ratio. Table 4 shows the results of the MANOVA analysis for the days (blocks). For the day or block parameter, all the probabilities for the four different test statistics applied were greater than 0.025 so we cannot reject the null hypothesis that the nickel and copper recoveries between the days (blocks) were the same. This means that the nickel and copper recoveries were not affected by the water chemistry differences, if any, between the days the flotation tests were done. Figure 26 shows the chemistry of the water types for the three days of testing. For all the species, the concentrations for the same water types between the days were approximately the same, thus, this is consistent with the conclusion that the days (blocks) were not significant or did not affect the metallurgy. Table 5 illustrates the results of the MANOVA analysis of the effect of water ratio on nickel and copper recoveries. All the probabilities for the test statistics were less than 0.025 with the exception of the Hotelling-Lawley Trace test. Since three out of the four test statistics were significant, we can reject the null hypothesis that the nickel and copper recoveries were the same. This implies that at least one of the water types had a significant effect (at 95% level) on nickel    In order to determine which water type had a significant impact on nickel and/or copper recovery, the contrast method in MANOVA analysis was used. Tables 6-8 show the results of this analysis. Since three comparisons are made (3 water types), the significance level α is divided by three to make the Bonferroni adjustment for the significance testing. Hence, the significance levels for the test statistics will be compared to a significance level α = 0.05/3 = 0.017. The only significant comparison was process water versus thickener overflow (p < 0.017 for all the test statistics). Hence, we can reject the null hypothesis that the nickel and/or copper recoveries were the same. In order to determine whether nickel recovery and/or copper recovery were significantly different between those obtained with process water and thickener overflow, the ANOVA portion of the output of the analysis was used (the MANOVA analysis also produces an ANOVA output to determine which recovery difference was significant). The ANOVA output will be discussed shortly. Table 9 shows the ANOVA portion of the MANOVA analysis for nickel and A. Di Feo et al. less than or equal to 0.017. Thus, we can reject the null hypothesis that the copper recoveries between the water types were the same and the differences in copper recoveries between the water types were significant. Figure 27 shows the profile plot for the nickel and copper recoveries for the flotation tests conducted using the April 2019 plant water. The copper recoveries tended to be higher than those obtained for nickel. The highest nickel and copper recoveries were obtained with the tests conducted using thickener overflow and they were significant at 95% confidence level relative to the recoveries obtained with process water. Figure 27. Profile Plot for nickel and copper recoveries for flotation tests completed using spring 2019 plant water (PrWater-Th: 50% process water and 50% thickener overflow, PrWater: process water, and ThOF: thickener overflow). Figure 28 illustrates the nickel + copper recovery versus nickel recovery for the process water and thickener overflow samples taken in June 2019. The flotation tests conducted with thickener overflow only resulted in a lower nickel + copper grade versus nickel recovery curves. This decrease was significant at the 90% confidence level. Figure 29 and Figure 30 show the pyrrhotite recovery versus pentlandite recovery and gangue recovery versus pentlandite recovery, respectively. The selectivity between pentlandite and pyrrhotite minerals decreased for thickener overflow only, which explains the lower nickel + copper grade versus nickel recovery for the tests conducted with the thickener overflow. The selectivity between gangue and pentlandite did not vary between the water types. Therefore, gangue in this case did not lower the nickel + copper grade in the primary and secondary rougher concentrates. Figure 31 and Figure 32 show the pyrrhotite recovery versus water recovery and gangue recovery versus water recovery for the tests done using the plant water obtained in June 2019, respectively. The pyrrhotite recovery versus water recovery for the water types was not a linear relationship, which implies that the main recovery mechanism for pyrrhotite was by true flotation. The gangue recovery versus water recovery for the water types was almost a straight line; thus, the main recovery mechanism for gangue was by entrainment for all the water types.     Table 10 shows the results of the MANOVA analysis for the days (blocks). Since we are using blocks for days and water ratio in the MANOVA analysis, the number of comparisons is 2 (m = 2), thus, the significance level α = 0.05/2 = 0.025 (Bonferroni adjustment) was used for the days (blocks) and water ratio analysis. For the day (block) parameter, all the probabilities for the four test statistics were greater than 0.025, so the null hypothesis for the nickel and copper recoveries between the days (blocks) cannot be rejected or we do not have enough evidence to reject it. This means that the nickel and copper recoveries were not affected by the water chemistry differences, if any, between the days the flotation tests were conducted. Figure 33 shows the chemistry of the water types for the three days of testing.

Plant Water-June 2019
For all the species, the concentrations for the same water type between the days were approximately the same, which is consistent with the conclusion that the days (blocks) were not significant or did not have any impact on metallurgy. Table 11 illustrates the results of the MANOVA analysis of the effect of water ratio on nickel and copper recoveries. All the probabilities for the four test statistics were greater than 0.025 with the exception of the Roy's Greatest Root test. Figure 33. Species concentration in process water, 50% process water/50% thickener overflow and thickener overflow-June 2019. Since three out of the four tests were not significant, there was not enough evidence to reject the null hypothesis that the nickel and copper recoveries between the water types were the same. This implies that the water types did not have a significant effect at 95% level on nickel and/or copper recovery. Figure 34 shows the nickel + copper grade versus nickel recovery for the plant water obtained in August 2019. The results were consistent with those obtained in April 2019 and June 2019. When thickener overflow was used, the nickel + copper grade versus nickel recovery decreased. We are 90% confident that the differences for the nickel + copper grade between process water and those obtained with 50% process water/50% thickener overflow and thickener overflow

Plant Water-August 2019
were not due to experimental error or chance (error bars do not overlap).   to decrease for the thickener overflow (50% and 100%). The gangue recovery curves were similar and the error bars overlapped. Therefore, gangue was not responsible for the lower nickel + copper grade for the thickener overflow. Figure 37 and Figure 38 illustrate the pyrrhotite recovery versus water recovery and gangue recovery versus water recovery, respectively. The pyrrhotite versus water recovery was not a straight line; thus, the recovery mechanism of pyrrhotite was mainly by true flotation. The gangue recovery as a function of water recovery was almost linear; thus, the recovery was mainly by entrainment for all the water types.
A statistical analysis, MANOVA, was also done for the flotation tests conducted using the plant water obtained in August 2019. Table 12 shows the re-      between the water types were the same. This implies that the water types did not have a significant effect on nickel and copper recoveries.

Discussion
There are no seasonal variations for the pH, ORP, calcium, sodium, sulfide, sulfate, TDS, thiosalts, total organic carbon, total inorganic carbon and total carbon for the process water, flotation feed side A water, flotation feed side B water and thickener overflow. This implies that it will be easier to implement a water treatment strategy, to treat the thickener overflow, for the mill in question without any significant issues related to seasonal changes that might exist.
Microbial cells can induce significant changes in the surrounding water chemistry, particularly with respect to pH and ORP [11] [12], and can metabolize certain flotation additives [11] [13] [14] [15], all of which could have serious effects on mineral recovery and flotation selectivity. Laboratory studies have also shown that microbial cells and cellular constituents (e.g., proteins, lipids, DNA) can directly affect mineral flotation due to adsorption onto mineral surfaces [16] A. Di Feo et al.
[17] [18]. Consequently, monitoring of the microbial loading within a flotation system is an important consideration if water is to be recycled, thereby greatly extending its residence time within the milling circuit and possibly allowing microbial numbers to increase to deleterious levels. In this study, the only samples which consistently yielded detectable concentrations of extractable microbial DNA were the process water samples. These were drawn directly from holding ponds that are exposed to the natural environment. Consequently, these samples were more susceptible to seasonal variations in temperature than were the other points sampled within the mill. Warmer summer temperatures may have led to increased microbial growth in the holding ponds, which was reflected by the spike in process water DNA yields during both summer sampling campaigns. It is recommended to sequence the extracted DNA to identify the microbial communities and determine whether they could have negative impacts on water quality and flotation as a result of their metabolic activity. However, the fact that no similar seasonal trends were observed in the chemical parameters our measurements suggest that increased microbial loading and activity in the process water was not having any subsequent detectable effect on water quality.
The fact that no extractable DNA could be detected in the side A and side B flotation feed and thickener overflow samples suggests one of two possibilities.
The first is that microbes could have been associated preferentially with the suspended solids (i.e., were attached to the mineral surfaces) which were removed from the water sample prior to DNA extraction. However, we have run separate, solid-phase DNA extractions on selected solids and were unable to detect any DNA in association with the solid phase (data not shown). The second possibility is that the microbial load was simply too low to yield detectable quantities of extractable DNA from the 1 L samples that were collected. In such cases filtering larger volumes could result in higher yields, but the funding and logistics involved in collecting, freezing, transporting and processing such volumes precluded that possibility at the time. In any case, we believe that the microbial load was so low within the milling circuit that it was unlikely to have any significant effect on flotation. That being said, prolonged recycling of the water within the milling circuit could eventually lead to a buildup of microbial load over time that might adversely affect water quality and flotation. For this reason, continued monitoring of the microbial load and activity throughout the milling circuit is recommended.
Laboratory flotation tests were done using process water and thickener over-  When thickener overflow was used, nickel + copper grade versus nickel recovery curves tended to decrease due to higher pyrrhotite (April, June and August 2019) and gangue recoveries (gangue recovery was higher only for April 2019). Figure   the thickener overflow could have caused higher pyrrhotite recovery through accidental activation. For example, the accumulation of the decomposition products of xanthate due to water reuse caused non-selective adsorption in copper-lead-zinc flotation [19]. Therefore, in the results presented in this publication, higher pyrrhotite recovery obtained with thickener overflow could have been caused by the higher TOC and TC concentrations.
For the flotation tests performed with thickener tank overflow stream in April 2019, the higher gangue recovery may have resulted by the higher concentrations of TOC and TC as well as higher recovery by entrainment as discussed earlier. From Figure 41, it can be observed that the TDS of thickener overflow was much higher than that observed for the process water. The TDS in water generally causes the air bubbles to be smaller, higher bubble surface area flux, causing a higher particle-bubble collision probability [1]. The higher bubble surface area flux caused by the thickener overflow may provide more bubble surface area for the pyrrhotite to float. Also, when the TDS is higher in the water, some species can cause accidental of unwanted minerals [1]. This may be another cause for higher pyrrhotite recovery when the thickener overflow is used. A surface analysis study is beyond the scope of this publication. However, a surface analysis investigation will be performed in the future to establish the species, if any, responsible for pyrrhotite and gangue flotation.   The MANOVA analysis for the April 2019 test work showed that the nickel and copper recoveries were significantly (95% confidence) higher for thickener overflow compared to those for process water. However, for the nickel and copper recoveries obtained with thickener overflow for the months of June 2019 and August 2019 were not significantly higher compared to those for process water.
When the null hypothesis cannot be rejected, it implies that we do not have enough evidence to reject it (nickel and copper recoveries between the water types are the same). This means that more tests are required and that the recoveries between water types may actually be different. However, it is important to point out that when the thickener overflow was used, there was no loss in nickel and copper recoveries.
It is recommended to investigate the impact of complete thickener overflow recirculation as a follow-up of this study. Also, various water treatment technologies on thickener overflow should be tested. The water treatment technology that will be chosen, and its effect on the percentage of the thickener overflow that has to be treated, will play an important role in the economics.

Conclusions
There were no seasonal or other trends (between the seasons) during the sampling period for the chemistry (species in solution such as calcium, thiosulphate etc.) of the process water, side A flotation feed water, side B flotation feed water and thickener overflow streams. Although process water samples displayed clear seasonal trends with respect to microbial loading, the increased summer loads did not appear to have adversely affected the water quality and flotation parameters measured. The nickel + copper grade versus nickel recovery decreased when higher amounts of the thickener overflow stream (50% and 100%) were used. The lower nickel + copper grade versus nickel recovery in the presence of the thickener overflow stream were due to higher pyrrhotite and gangue recoveries (non-sulphide gangue recovery was higher for April 2019 only).
For nickel recovery, the tests conducted using thickener overflow stream (100% thickener overflow) obtained in the April 2019 showed an improvement relative to that obtained with process water. For copper recovery, the tests conducted with 50% and 100% thickener overflow (April 2019) resulted in an improvement relative to those obtained with process water.
Most likely the higher pyrrhotite recovery in the presence of the thickener overflow stream was due to the higher TDS, TOC and TC. The higher TDS in the water will most likely create smaller bubbles and higher bubble surface area flux, which can increase particle-bubble collision efficiency resulting in higher recovery. Whereas, the TOC and TC may cause inadvertent activation of pyrrhotite and gangue.