Analysis of the grinding behaviour of
various Portland cement clinkers

1 Introduction
Chemical and physical properties of cement clinker are influenced by many factors including chemical, physical and mineralogical characteristics of the raw materials, the burning and cooling processes, the fuel used and the capacity of the kiln. Grindability and the energy consumption in clinker grinding are influenced by a number of factors as well. If it is defined as the rate of production of fine material for a given energy expenditure, then it is reasonably well accepted that grindability increases with porosity and alite content (3CaO · SiO₂ or C₃S) [1, 2] and decreases with belite (2CaO · SiO₂ or C₂S) [1, 3, 4] and aluminate content (3CaO · Al₂O₃ or C₃A) [2, 5].

There is controversy regarding the effect of a number of factors, since authors have reported different influences on clinker grindability. Examples are the alumina content, iron oxide content, aluminoferrite (4CaO · Al2O3 · Fe2O3 or C4AF) content, silica ratio, ratio of silicates and aluminates and the proportion of liquid phase [1, 2, 4–6].

Besides chemical and mineralogical composition, microstructure has been found to play an important role in clinker grindability. For instance, evidence exists that the belite content deteriorates the clinker grindability more significantly when crystals appear as clusters [3]. When they do not form clusters, belite crystals are simply liberated during grinding, so that grindability is comparatively easier [3]. There is also evidence that grindability reduces with the increase in the dust component of clinker [7, 8], the presence of microcracks within the constituent minerals [8, 9], the degree of crystallinity of the interstitial phase [2], the porosity and with the reduction in alite crystal sizes.

Some of the contradicting results presented in the literature may, at least in part, be explained by the fact that some of the well-known correlations were demonstrated by burning clinker from the same raw materials, by the same burning and cooling processes and are not often verified for clinkers produced under variable conditions [2]. Another reason is associated with the fact that not all factors affect grindability in the same way at different stages of grinding. Microstructure and composition affect grindability in different ways, sometimes making pre-grinding a relatively simple task, but final cement grinding a severe energy consumption task. This is particularly relevant since clinker nodules as coarse as 50 mm may be ground down to sizes finer than 20–50 μm in a single grinding mill in industry. Understanding how different components and microstructural features influence clinker grindability at different stages of cement grinding is relevant to improving the energy efficiency in the process.

This work takes advantage of single-particle breakage experiments and standard grindability tests to gain insights into the relationships between material composition and microstructure and the expected comminution behavior of cement clinkers during the different stages in grinding.

2    Experimental
Clinkers denoted by A–I have been selected from six industrial plants in Brazil. These are all produced in rotary kilns by the dry process, for ultimately producing ordinary Portland cement, except for clinker I, used in the production of white cement.

Characterization methods included helium pycnometry, X-ray fluorescence, X-ray diffraction (XRD) and optical microscopy under reflected light. Particle size analyses were determined by wet/dry sieving. Helium pycnometry was conducted on samples ground below 75 μm to measure their specific gravity.

Total porosity of each clinker sample was measured, for con­venience of handling, on particles contained in size range
5.6–4.75 mm by coating over 200 particles, one by one and by hand, with a thin layer of Teflon® tape and then measuring the specific gravity of the coated particles by pycnometry in water. The total porosity () was then calculated by

 = 100 1– ƒ coated – (1 – ƒ ) coated (1)
                     Teflon                             s
where s is the specific gravity of clinker, Teflon is the specific gravity of the Teflon® tape, ƒ is the weight fraction of Teflon® in the coated particles and coated their specific gravity. Measurements were run in triplicate.

Grinding in the first compartment of multi-compartment tube mills is often difficult to investigate in laboratory ball mills given the limitations in ball size and mill diameter. On the other hand, the comminution behavior of nodules of sizes ranging from a few to several millimeters can be appropriately investigated ­using drop-weight tests. Fragmentation response was analyzed from the relationship between the applied energy and the size distribution of progeny fragments from testing particles contained in size ranges from 22.4–19.0 mm to 11.2–9.5 mm. Results from these tests have been used to determine the ­amenability
of clinkers to fragmentation by impact, according to [10]

t10 = A[l – exp(-b Ecs)](2)

where t10 is the percentage of breakage product that passes 1/10th of the initial particle size; Ecs is the specific input energy (kWh/t) as calculated from the kinetic energy of the falling weight. The product A*b is used to compare the amenability of the different clinkers to fragmentation by impact. A high value of A*b means that the clinker has low resistance to impact breakage and vice versa.

The Bond ball mill grindability of the samples has been measured using the standard procedure and mill [11] and tests were conducted with the aim of determining the work index of the clinkers (Wi). In general, a fixed sieve size size (P1) that is similar to the cut size used in the industrial circuit is used to determine Wi. For an “ideal Bond material” Wi is a material property that should not vary with P1. However, for several materials, it changes depending on the fineness of grind [12], so that in the present work Wi was determined using the sieve sizes 300, 150 and 75 μm, which corresponded to P80 ­values of 165–245, 95–125, 50–60 μm and Blaine specific surface areas of about 100–150, 200–250 and 300–350 m2/kg, respectively.

3    Results and discussion
3.1 Chemical and physical characteristics
Table 1 shows a summary of the chemical composition of the clinkers. Table 2 shows that specific gravities varied within the narrow range of 3.20 to 3.25 g/cm3, except for clinker I, which presented a lower specific gravity, owing to its lower iron content (Table 1).

Size analyses of the clinker nodules are presented in Figure 1.
Variations in the sizes of nodules and particularly on the amount of fines can be associated to burning conditions inside the kiln [8]. Indeed, it has been suggested that grindability reduces with the increase in the proportion of fine material in the clinker, so that their presence is generally avoided by ensuring that strong nodules of clinker are formed by agglomeration in the kiln [13].

Analyses of the porosity of the clinker nodules and of the porous network were conducted and a summary is presented in Table 2.
Porosities varied significantly among the clinkers, from as low as 25  % (clinker G) to as high as about 35  % (clinker B). In spite of the comparatively high standard deviations of the measurements, they allowed discrimination between the ma­terials.

It has been suggested that porosity is directly related to the proportion of interstitial phase, the fineness of the raw meal, the amount of trace elements and firing temperature. García-
Márques et al. [14] demonstrated that porosity increases with the alumina content, owing to the contraction of C3A that ­occurs during cooling of the liquid phase during burning of clinker. Indeed, Figure 2 shows a good correlation between ­alumina content and total porosity.

3.2 Mineralogy
Quantitative mineralogy results from XRD and Rietveld ­analyses are summarized in Table 3, which demonstrate the ­significant variation in clinker composition. No C4AF has been identified in clinker I, since iron is almost absent in this clinker (Table 1), and only a small proportion of periclase is found in this clinker, which presented low MgO content. Qualitative mineralogical and textural analyzes from inspection of optical microscopy images made it possible to establish that alite crystals of all clinkers presented predominantly pseudohexagonal shapes
(Fig. 3) and mean grain sizes from 22 to 27 μm, being coarsest for clinkers E and F. Alite crystal sizes were smaller for clinkers C, D, G and I, suggesting that clinkering times were marginally short. Alite crystals were found to be extensively corroded for all samples (Fig. 3a), except for clinkers C, E and F (Fig. 3b). Such a degree of corrosion was particularly high in clinkers A and G, and alite (as well as belite) crystals were also found to be highly microcracked in these clinkers as well.

Belite crystals were generally found in clusters, which often presented regular shapes for all clinkers. In the case of clinkers C and I some compact zones are found, with little or no interstitial phase. In the case of clinkers G and I, clusters of irregular shapes are also present. The average size of belite clusters varied from 140 and 150 μm for clinkers G and H, respectively, to as high as 280 μm for clinkers C and I.

The interstitial phase was characterized as crystallized for ­clinkers H and I (Fig. 4a), crystallized to semi-crystallized for clinkers A, B, C and G, semi-crystallized for clinkers E
(Fig. 4b), and F and vitreous for clinker D, suggesting that secondary cooling was fastest for clinker D and slowest (perhaps too slow) for clinkers H and I. Except for clinker I, the interstitial phase for all other clinkers is composed by a greater proportion of C4AF in comparison to C3A.
Free CaO is distributed in regular zones for all clinkers, with sizes that varied, on average, from 60 and 80 μm for clinkers D and H, respectively, to as large as 180 and 210 μm for clinkers B and C, respectively. The coarse size of these zones and the reasonably large proportion found in clinkers B and C suggest that limestone grinding in the raw meal was probably insufficient in these clinkers. Finally, metallic iron grains were observed along the corroded borders of alites in clinkers C, F and I, suggesting the existence of a reducing environment in some parts of the kiln in the burning process. For all others, clinkering conditions are characterized as oxidizing.

3.3 Particle breakage and ball milling
Additional insights can be gained by analyzing the size distribution of progeny fragments from single-particle breakage of the clinkers at different impact energies. A summary from fitting these data to Eq. (2) is presented in Table 2. The product A*b may be used to characterize the initial grindability of clinker in a grinding mill and can also be shown to be proportional to porosity (Fig. 5), with a correlation coefficient of 0.58). This is consistent with observations by a number of researchers that showed that greater porosities result in lower strengths of a number of materials, including rocks and concrete [15].

Figure 6 summarizes data on the Bond ball mill work index. With the exception of clinkers A and G, the Wi for all others was found to increase with a decrease in sieve sizes. For such a porous material, an increase in Wi with a decrease in fineness is not surprising, since, as coarser particles are broken, the associated porosity decreases and individual mineral grains are liberated. The opposite behavior found for clinkers A and G may be explained in part by the low porosities of these clinkers (Table 2), which suggest poor initial grindabilities, and at the same time are clinkers that presented high degrees of corrosion and cracking of alite crystals, which makes finish grinding a relatively less demanding task.

Multiple regression analyses have been conducted with data from the Bond work index measured using the different sieve sizes and the effect of each chemical and mineralogical component investigated. Porosity and C3S content were the variables that more significantly influenced Wi at the sieve size of
300 μm. This is illustrated in Figure 7, which compares the data to the regression equation

Wi(300 µm) = 76.4 – 0.50 Porosity – 0.74 C3S(3)
where porosity and alite content are given in % and the work index is given in kWh/t. Eq. (3) shows that, at this sieve size, the work index decreases with an increase in both porosity and C3S content. All parameters in the equation are statistically significant at 95  % confidence. In addition, regression analyzes for Wi at the sieve size of 150 and 75 μm result in the expressions

Wi(150 µm) = 55.6 – 0.46 C33 – 1.39 C3A – 0.54 C4AF  (R² = 0.82)(4)

Wi(75 µm) = 42.8 – 2.79 C3A – 1.36 C4AF(5)

A comparison between data and Eq. (5) is shown in Figure 8.

These results show that as sieve size openings decrease, first porosity, then alite content no longer influence the results. In contrast to that, the influences of C3A and C4AF contents on Wi become more pronounced at finish grinding.

While the decreases in energy expenditure with the increase in porosity, C3S and C4AF contents observed are well supported by the literature [1, 2, 5, 7], its reduction with the increase in C3A observed (Eqs. 5 and 6) contradicts results from a number of authors [3, 5, 6].

This may be partially explained by the method used to estimate the mineralogical composition, which in the present work was carried out by XRD Rietveld whereas in the cited works it was estimated from the Bogue calculation, which is known to provide very different results [16].

Finally, no relationship was found between the degree of crystallinity of the interstitial phase or the size of alite crystals and grindability. Still, a good correlation was identified between the proportion of fine material (taken, for example, as the material passing the 1.18 mm sieve) in the as-received clinker (Fig. 1) and the Wi at a sieve opening of 75 μm. This result is consistent with the observation that clinkers containing large proportion of fines present more difficult grindability [13].

3.4 Grindability as a function of grinding stage
From the information provided by the different techniques, it is proposed that cement clinker grindability be characterized in different size scales. “Initial grindability” could be used to describe the response of clinker inside the first compartment of a tube mill or in a pregrinding stage, corresponding to particle sizes from almost a millimeter up to the coarsest nodule sizes of the clinker. “Intermediate grindability” could be used for the response of clinker up to the point that it reaches the final compartment of tube mills, corresponding to grinding down to a few hundred micrometers. “Finish grindability” would deal with the overall grinding response as the clinker approaches the final diaphragm of tube mills, corresponding to grinding down to particle sizes present in the mill discharge.

Within this framework, much of the research published in the literature on clinker grindability deals with “finish grindability”, with the exception of a few studies, such as that from Frigione et al. [5] which spans to “intermediate grindability”, with values of Blaine as low as 100 m2/kg. The greater attention generally devoted to finish grindability is quite understandable, since it is the stage responsible for the greatest part of the ­energy consumption in clinker grinding in industry. Comparatively limited work, however, has been conducted in elucidating the “initial grindability” of clinker [9], which is also of significance, particularly when defining grinding conditions in the first compartment of tube mills.

Relative grindabilities of the clinkers analyzed are summarized in Figure 9. Initial grindability (nodules coarser than a few millimeters) are characterized using the inverse of the parameter A*b from single-particle breakage (Table 2). Intermediate grindability is characterized from the Wi at a sieve size of
300 μm, whereas finish grindability is characterized from the Wi at a sieve size of 75 μm. It shows the significant variations in grindability of the samples. Whereas in finish grinding the most difficult clinker to grind took twice the amount of energy required than the easiest one, this ratio increases to five for initial grindability.
Figure 9 also shows that the relative grindability also varied significantly with the stage in grinding. Except for clinkers A and G, grindability of all clinkers is comparatively improved at coarser grinding.

4    Conclusions
Analyses of comminution behavior of cement clinkers at different size scales allowed the conclusion that:
–    Porosity is correlated with alumina content.
–    Single-particle breakage data show that porosity is the most significant factor that determines the initial grindability of cement clinker.
–    Data on the Bond work index at fine sieve sizes show that finish grindability is influenced significantly only by the contents of C3A and C4AF.
–    Intermediate grindability, determined from the Bond work index at large sieve sizes is progressively less influenced by porosity as size decreases and more influenced by C3A and C4AF
content, also being significantly affected by alite content.