**About the Special Issue Editors**

**Shahrooz Nafisi**, Dr., is an adjunct Professor at the University of Alberta, Canada. He received his B.S. and M.S. in Metallurgical Engineering from the Iran University of Science and Technology and his Ph.D. from the University of Quebec. He has co-authored two books, "A New Approach to Gating ´ Systems" (1st edition, 1997, 2nd edition, 2001) and "Semi-Solid Processing of Aluminum Alloys", ISBN 978-3-319-40333-5, Springer, Sep 2016 (republished in China, Jan 2020, ISBN: 978-7-122-34281-2), and more than 70 journal articles and conference papers. He was the 2017 profile in achievement from Professional Engineers of Canada (APEGS); the Vanadium award recipient in 2014 (Institute of Materials, Minerals and Mining "IOM3"); the 2013 best paper award of the International Metallographic Society and Metallography, Microstructures, and Analysis; and the recipient of the 2012 Association of Iron and Steel Technology (AIST) Hunt-Kelly Outstanding Paper Award.

**Reza Ghomashchi**, Prof., graduated from the Iran University of Science and Technology (B.Eng. Metallurgical Engineering,1978), and Cambridge (M.Phil. Materials Technology, 1979) and Sheffield (Ph.D. Metallurgy, 1983) Universities in the U.K. After a few years working at universities in the U.K., he migrated to Australia to work for BHP Steel in 1988. In early 1990, he joined the University of South Australia and worked as a Lecturer and then Senior Lecturer until 2001 when he was offered a Natural Science and Engineering Research Council of Canada (NSERC) Industrial Research Chair-Tier I (in collaboration with ALCAN, now Rio-Tinto ALCAN) and Professor Position at the University of Quebec in Canada. He has also been a visiting Professor at MIT, 1994; IUST, 1999; and ´ an adjunct Professor at the School of Mechanical Engineering at the University of Adelaide, 2007. In early 2008, he returned to Australia to take up the position of Manager, Materials Research and Development, at Sunday Solar Technologies, a start-up R&D company in Sydney and, in August 2010, he accepted his current position at the school of Mechanical Engineering of the University of Adelaide, Australia.

### *Editorial* **Semi-Solid Processing of Alloys and Composites**

#### **Shahrooz Nafisi 1,\* and Reza Ghomashchi 2,\***


Received: 29 April 2019; Accepted: 7 May 2019; Published: 8 May 2019

A quick look through the past two centuries tells us that we may be in our third industrial revolution. The first industrial revolution (19th century) was mainly due to the introduction of steam energy, while the second (20th century) was mainly due to inexpensive oil and gas—which, by the way, brought us some unwelcome consequences; the so-called greenhouse effect and subsequent global warming and unpredictable weather patterns. We are now embracing a third industrial revolution, which could be termed the green energy and communication era. As a result, our manufacturing technologies should also follow a similar pattern: "Green manufacturing" with less energy consumption. Semi-solid metal (SSM) processing may be branded as a step forward towards green manufacturing, as it consumes less energy than its conventional counterparts. However, in spite of the many advantages of SSM processing and its viable manufacturing route, including a reduction in energy consumption, its implementation in the metal industry has been very sluggish. As strong advocates within the SSM processing community, we believe such a delay in recognizing the benefits of SSM casting of light alloys is predominantly due to the lack of proper communication between research and development (R&D) investigators and industry leaders. The Editors have tried to close the communication gap through publication of a new book [1] and the introduction of the current special issue of the Metals Journal on SSMs as an extra effort to the biannual S2P conference. We hoped an invitation of key players to highlight the latest advancements in the field would contribute towards better usage of SSM processes in industrial applications.

This special issue is focused on the recent research and findings in the field, with the aim of filling the gap between industry and academia, and to shed light on some of the fundamentals of science and technology of semi-solid processing.

This special issue provides new researches on the two main routes of semi-solid metal processing; Rheo and Thixo - casting. In addition, a variety of alloying systems and composite materials are covered in this special issue, including interesting information on welding, tribology and corrosion of SSM-processed alloys. Rheology and the correlation between structure and properties have been covered in two outstanding review articles. We would like to thank all the authors for their contribution and consideration of the reviewers' comments. Additionally, the continuous assistance of the Metals editorial staff is gratefully acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Reference**

1. Nafisi, S.; Ghomashchi, R. *Semi Solid Processing of Aluminum Alloys*; Springer International Publishing: Basel, Switzerland, 2016.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Rheological Characterization of Semi-Solid Metals: A Review**

#### **Michael Modigell 1,\*, Annalisa Pola <sup>2</sup> ID and Marialaura Tocci <sup>2</sup>**


Received: 18 February 2018; Accepted: 4 April 2018; Published: 7 April 2018

**Abstract:** In the present review, the main findings on the rheological characterization of semi-solid metals (SSM) are presented. Experimental results are a fundamental basis for the development of comprehensive and accurate mathematics used to design the process effectively. For this reason, the main experimental procedures for the rheological characterization of SSM are given, together with the models most widely used to fit experimental data. Subsequently, the material behavior under steady state condition is summarized. Also, non-viscous properties and transient conditions are discussed since they are especially relevant for the industrial semi-solid processing.

**Keywords:** rheology; semi-solid alloys; thixotropy; rheometer; compression test; viscosity

#### **1. Introduction**

In the early 1970s Flemings and coworkers [1] discovered metallic alloys in the semi-solid state with non-dendritic structure to have special rheological properties which can be exploited for a new, attractive forming process. The non-dendritic structure can be easily achieved by stirring the alloy while cooling it from the liquid state down into the semi-solid temperature range. This results in a suspension consisting of a liquid metallic phase and primary solid particles with globular or rosette-type shape [2,3], such as that shown in Figure 1, in comparison with a typical dendritic microstructure.

**Figure 1.** Typical microstructures of an A356 component obtained by (**a**) thixocasting (billet preheated to a solid fraction of 0.52–0.54) and (**b**) conventional casting (pouring temperature = 680 ◦C).

The rheological properties of this special type of slurry give the advantages of the semi-solid metals (SSM)-processing. In detail, the slurry can either flow like a liquid—but with non-constant viscosity—or it can behave like a solid. This is typical for suspensions with high solid fraction (*Fs*) [4], whereas fully liquid metals show Newtonian flow behavior, which is water-like [5]. The rheological properties are responsible for the die-filling behavior of SSM, which is different from fully liquid (or fully solid) materials [6,7] and which results in specific advantages in the quality of the product (low gas porosity, less shrinking, higher mechanical properties, etc.), besides those related to technological aspects (longer tool life in comparison with conventional casting processes due to the lower metal temperature, etc.) [2]. To obtain these advantages, it is necessary to fully understand the rheology of the material. This enables understanding of flow-specific phenomena, such as instabilities or segregation, and allows optimization of the process. Carefully performed experiments with respect to the mechanical, fluid dynamical, and thermal conditions lead to the development of comprehensive and accurate mathematical models, which picture the physics properly and which are used in computer simulation to design and optimize the process effectively.

Few publications provide a comprehensive discussion on the rheological behavior of SSM [8,9]. In addition, in this regard, it should be mentioned that still some aspects about the flow of semi-solid metals are not clear or contradict data are available in the literature [8]. For this reason, it appears useful to provide an overview of the current knowledge about this topic, with particular attention to the scientific innovation that took place in the past 10–15 years, while details about semi-solid metal processing can be found elsewhere [2,7,10].

In Section 2, the rheological classification of SSM is explained. In Section 3, the principles of the most frequently used devices for investing rheological properties of SSM are presented: in Section 3.1 the rotational rheometer and in Section 3.2 the compression test. Section 4 gives a summary of the applied rheological models and the corresponding constitutive equations. An overview of recent published results of the equilibrium viscosity of Al-alloys is given in Section 5. Section 6 deals with special rheological phenomena of SSM: yield stress and thixotropy. Section 7 explains the influence of Ostwald ripening in rheological experiments, which is important to consider for long-term experiments.

Before analyzing the literature, it is useful to clarify some rheological terms that are related with the rheological properties of SSM and which are frequently used improperly in the literature of SSM processing.

#### **2. Rheological Classification of SSM**

The rheological behavior of any material is found to be between two limiting, ideal cases: the ideal solid body (Hookean body), which shows deformation proportional to the stress, and the ideal viscous material (Newtonian body), which shows rate of deformation proportional to the stress. Within the viscous materials, besides the Newtonian fluids (with constant viscosity, only depending on temperature and pressure), we find the Non-Newtonian fluids. Among these, we have the non-linear pure viscous fluids, whose viscosity depends additionally on the stress and which exhibit shear thinning or shear thickening behavior. Another class of the non-linear materials are the plastic ones, which show solid behavior below a certain threshold of stress (yield stress), while they are characterized by linear (Bingham body) or non-linear behavior above the yield stress. Another class of materials shows time-dependent properties, whose rheological properties do not change immediately after change in strain but follow a specific kinetics. Viscoelastic materials have simultaneously elastic and viscous properties. Thixotropic materials show a gradual decrease of viscosity under constant stress and a recovery of the viscosity when the stress is removed; in particular, the viscosity of the initial condition will be recovered totally. As shown in Figure 2, due to thixotropy, at the beginning of shearing or after a rapid change in shear rate, the instantaneous viscosity is different from the steady state values, and it takes time for the viscosity to reach a constant value, which reflects the equilibrium condition of the structure. The opposite behavior, an increase of viscosity with time, is called rheopexy.

**Figure 2.** Schematic diagram showing the change in viscosity with time following changes in shear rate to illustrate the thixotropic behavior of semi-solid slurries [11].

Regarding the rheological classification of SSM, it is generally accepted that they are shear thinning fluids, that means that viscosity will drop with increasing shear rate (Figure 3) [12]. Additionally, they are thixotropic materials.

**Figure 3.** Example of plot of viscosity measurements in isothermal conditions from shear rate experiments for A356 alloy at various solid fractions showing the typical shear thinning behavior of SSM [12].

Finally, SSM with high solid fraction exhibit yield stress, which is nicely demonstrated with the picture of a block of Al alloy in the semi-solid state able to wear its own weight and which can be cut with an ordinary knife. A representative image of this phenomenon is shown in Figure 4; analogous pictures are available in the scientific literature, as for instance in [1].

The specific rheological properties are finally nothing more than the consequences of the changes in the internal structure of the slurry due to external forces. A simple physical model can be set up, in agreement with the general understanding of the kinematics of the SSM. It can be assumed that cohesive forces are acting between the particles of the SSM, resulting in the formation of agglomerates [13]. The particles of such agglomerates can be connected temporarily by formation of welded necks or e.g., by capillary forces. Within the agglomerates, a certain amount of liquid is immobilized, which leads to a higher apparent solid fraction [14]. Under the influence of shear forces, the agglomerates will be partially or totally disintegrated whereby the liquid phase is released and the apparent solid fraction will approach the true fraction. With this model, the steady state behavior of the material, i.e., shear thinning, can be explained since the viscosity decreases with decreasing solid fraction.

**Figure 4.** Billet of a semi-solid metal cut with an ordinary knife.

By reducing the shear rate, particles that meet in the shear field have the chance to agglomerate, resulting in an increase in the apparent solid fraction.

It follows that the structural change is reversible, which explains one feature of thixotropy. The deagglomeration and agglomeration processes do not happen instantaneously, but they take some time. The agglomeration is diffusion controlled and, therefore, it is much slower than the deagglomeration phenomenon.

Similar to other suspensions, the most important parameter influencing the rheological properties is the solid fraction, which depends on the temperature [2]. Experimental measurements show that an increase in solid fraction results in an increase of viscosity. In addition, also the yield stress increases with higher solid fraction, as demonstrated in scientific literature [15–18], together with the presence of the thixotropic effects. Other factors, such as particle diameter, or diameter distribution, and the shape of the particles are of minor importance. Both these parameters can be combined with the specific surface area of solid and liquid phase. The dependency on the viscosity of the liquid phase is not very strong because the liquid phase viscosity is orders of magnitude lower than of the SSM.

This rather simple structural model has been accepted widely—although there is no clear experimental evidence. Indeed, all metallographic micrographs have been produced from ordinary solidified samples with low cooling rates. It has been demonstrated [19,20] that, with cooling rates smaller than −10 K/s, diffusional processes will significantly change the size of the particles and, therefore, the appearance of the structure. To investigate the structure is difficult. There are only a few publications [20–24] that are dealing with structural experiments with X-ray tomography for SSM in rest and compression. However, up to now, no work has been published for investigations of SSM under pure shear. The evaluation of the X-ray images shows that with increasing shear rate the distribution of the particles will become more homogenous over the volume, which confirms qualitatively the physical model [24]. The compression experiment under the X-ray beam shows that for high solid fraction (0.70) the material behaves as a saturated sponge and the liquid phase is pressed out of the skeleton with the consequence that the solid fraction will change locally. This has been previously found, as well by analyses of the solid distribution in a billet after compression [25].

#### **3. Experimental Methods for the Measurements of Rheological Properties**

#### *3.1. Shear Experiments in Rotational Rheometers*

The most widely used shear rheometers for the study of semi-solid metals are the rotational rheometers with concentric arranged cylinders. The outer cylinder is a cup that contains the SSM material and in which the inner cylinder, the bob, is inserted. In the Couette-type rheometer the cup is rotating and the bob is fixed, whereas in the Searle rheometer the bob is rotating while the cup is fixed. Because of the relative movement of cup and bob, the material is sheared in the gap between them. The shear stress at the wall is related to the torque, which is measured, and the shear rate is related to the rotational speed and to the geometry. Due to inertia forces, the Searle system is sensitive for secondary flows, the Taylor vortices, which dissipate energy and cause an increase in the measured torque [26]. Depending on the geometry of the system and the properties of the sample, the vortices can occur already at rather low rotational speed. Simple criteria are available to calculate the onset of the vortices [26], which can be applied for all viscous fluids. Another effect that falsifies the measurements consists in turbulent vortices that occur for both rheometers at higher rotational speeds, which is defined by a critical Re number [27].

For Newtonian fluids, the evaluation of the viscosity from the torque and the rotational speed is rather simple. For non-Newtonian fluids, it is complicated when the rheological nature of the fluid is unknown. In this case, frequently the way of evaluation valid for Newtonian fluids is applied. This results in an apparent viscosity value and in an apparent flow curve, which does not reflect properly the physical properties of the material. For purely viscous materials, the approach of the representative shear location should be applied [28], which results in physical correct values. Alexandrou et al. [29] have shown recently that this method does work for viscoplastic materials in special cases only. In general, the processing of data collected from rotational rheometer should be evaluated with the help of computational rheology [30].

Wall slip is another phenomenon that affects rheological measurements in suspensions with any kind of shear device. The slip is caused by segregation of a thin layer of the liquid phase adjacent to the wall. This thin layer has the effect of a lubricant that reduces the friction and, consequently, the torque measured by rotational rheometers, resulting in apparently lower viscosity values. In the literature, some different geometries for the bob have been proposed to avoid slip. For instance, Modigell et al. [31] have shown that vane-type bobs are not suitable because they lead to secondary flows that influence torque measurements, whereas a grooved bob prevents slip without affecting the torque significantly. Another way to treat slip is to apply the Kiljanski method for Searle or Couette rheometers [32] (or the Mooney method for capillary systems). The idea of both methods is to evaluate the slip velocity with the help of two different geometries. Harboe et al. [33] could show that the application of the Kiljanski method results in the same flow curve as the application of a grooved rod, but it requires significantly more experimental effort.

For SSM with low solid fraction—and low corresponding viscosity—the influence of the surface tension on the experimental result must be considered. Tocci et al. [34] could demonstrate that small deviation of the symmetry of the measuring system leads to secondary forces caused by the surface tension, which is dominant for small shear rates. Consequently, the material appears to be strongly shear thinning, although it is almost Newtonian.

At the beginning of the development of SSM processes, most of the rheological investigations have been performed with low melting Sn-Pb alloys because of the lack of high temperature rheometers. Nowadays, commercial instruments for testing Al alloys are available, while, to our knowledge, only one commercial instrument is available on the market for studying steels [35]. Yekta et al. [36] and Modigell et al. [37] have used own developed instruments.

Typically, for experiments with rotational rheometers, a first important part of the procedure is the preparation of the semi-solid material by shearing it for a certain time during cooling to the desired temperature, according to the solid fraction. A proper material preparation is fundamental, especially under consideration of the Ostwald ripening (see Section 7), since the flow behavior of semi-solid metals is strongly related to the microstructure [2]. An example of the evolution of viscosity during the material preparation is presented in Figure 5 for an Al-Si alloy for a constant shear rate of 100 s−<sup>1</sup> [12]. First, the material is sheared in the fully liquid state (630 ◦C) to ensure the homogeneity of the material. When the temperature decreases, a severe increase in viscosity takes place, mainly due to the formation

of solid particles. Finally, when the temperature reaches the value corresponding to the desired solid fraction (0.35 at 583 ◦C), a first steep decrease in viscosity is observed due to the change of dendrites into globular particles because of the application of shear forces. The following less steep decrease of the viscosity is due to Ostwald ripening.

Values of viscosity for the evaluation of flow curve in steady state condition are calculated from experimental data at different shear rates.

**Figure 5.** Plot of viscosity versus time at constant shear rate (100 s<sup>−</sup>1) during cooling from liquid state (630 ◦C) to semi-solid condition (solid fraction of 0.35 at 583 ◦C) for an Al-Si alloy measured by means of a Searle rheometer [12].

#### *3.2. Compression Tests*

The compression test is a conventional testing method to acquire strain-stress curves by squeezing a sample either under a constant load between two parallel plates or with a constant speed of displacement of the plates [38]. Compression experiments are usually performed with materials characterized by a solid fraction higher than 0.5.

Various experimental configurations are possible according to the used device, an example is shown in Figure 6. Usually, the sample is first heated to the required temperature in a separate furnace or directly in the testing chamber, while, after compression, it can be quenched in water for further study of the microstructure. The applied force and the obtained displacement is monitored by a proper load cell.

**Figure 6.** Scheme of parallel plate deformation set up [39].

From the stress-strain curve, it is possible to calculate rheological parameters and obtain a flow curve in terms of viscosity as a function of the shear rate, as illustrated by Laxmanan et al. [40]. It is important to mention that, with this method, the viscosity at a given shear rate is calculated under the assumption of Newtonian behavior (comparable to the simple approach with rotational rheometers). Consequently, the calculated viscosities are apparent values only. Nevertheless, the evidence of the shear thinning behavior of SSM is obtained when the values of apparent viscosity calculated at different shear rates are compared.

Another drawback is the flow condition at the surfaces of the plates. For pure shear flow, the material should adhere to the plates. Slipping conditions result in elongation flow, which must be evaluated in a different way. In practice, none of the two conditions are completely fulfilled and the flow is of mixed mode. Additionally, the flow is non-stationary and at least 2-dimensional. Since usually the ram speed is high in order to simulate forging conditions, high accelerations are achieved, and the evaluation of thixotropic effects is difficult (as it is discussed in Section 6). At this regard, Hu et al. [40] performed compression tests with Al alloy under conditions close to forging processes and applied ram speeds up to 1000 mm/s, which results in experimental times of approximately 0.01 s. Similar values were achieved by Becker et al. [41,42] during experiments with steel.

Two practical problems arise with SSM during compression. Even at low ram speeds liquid phase is squeezed out of the sample when solid fraction is less than 0.80. This results in an inhomogeneous composition of the sample, which additionally is changing by time, although the experiments have been performed isothermally. Moreover, another problem is the cracking of the free surface with increasing deformation [43].

Temperature and compression rate can be varied to reproduce real cavity die-filling conditions, which is one advantage of this technique in comparison with shear experiments. On the other hand, the possible experimental procedures are wider for shear experiments, allowing to completely characterize the rheological behavior of the material.

Particularly, the compression rate is a key parameter since a slow compression can provide information not adequate for the understanding of the actual industrial process, which is known to take place in less than 1 s. For this reason, rapid compression tests were carried out to investigate the transient behavior of SSM [44–46]. A schematic representation of the load-displacement curve for these kinds of experiments is shown in Figure 7.

**Figure 7.** (**a**) Typical signal response to rapid compression of semi-solid A356 alloy (ram speed: 500 mm/s; soak time: 0 min; and temperature 575 ◦C) [44]; (**b**) Schematic appearance of a load vs. displacement response. Four distinct regions are present: (1) zero load prior to reaching the die; (2) an initial breakdown, or peak load (stress); (3) a relatively constant, or plateau, load after the initial breakdown; and (4) a rapid increase in the load as the die approaches complete filling [45].

In addition, the test temperature and the holding time in isothermal condition should be carefully chosen when the main aim of the experiments is to provide information adequate for the industrial process.

Finally, this experimental procedure can be applied using a close die to investigate the liquid flow during compression. This is helpful in predicting the formation of liquid segregation and surface cracks

in the products obtained by SSM processing [47]. Also, drained compression tests can be performed to investigate the compressibility of the solid phase in isothermal condition [10].

#### **4. Modelling of Rheological Properties**

The evaluation of experimental rheological investigations should result in mathematical equations, called constitutive models, which should reflect the physics of the flow and the deformation process. Surely, it is difficult to include all phenomena in one model and it is generally accepted to make simplifications according to the application of the model. The simplest models are the one phase, equilibrium models, which assume the SSM to be a homogenous fluid without time-dependent properties. Under these assumptions, the Ostwald-de-Waele model—or power law—is the simplest one, assuming viscous properties only. The relationship between shear stress and shear rate can be expressed by the following equation:

$$
\pi = m \cdot \dot{\gamma}^n \tag{1}
$$

And the apparent viscosity is given by:

$$
\eta = \pi/\dot{\gamma} = m\dot{\gamma}^{n-1} \tag{2}
$$

where for *n* < l, the fluid exhibits shear thinning properties; *n* = l, the fluid shows Newtonian behavior; *n* > l, the fluid shows shear-thickening behavior.

The terms *m* and *n* are two empirical parameters, the flow index and shear exponent, respectively.

Because of the simplicity of this equation, it is widely applied to process data from rheological experiments. Besides the first studies on the characterization of the rheological behavior of semi-solid slurries [38,48], mainly focused on SnPb15 alloy, also more recent papers used the Ostwald-de-Waele relationship to express the viscosity as a function of shear rate for various Al alloys and steels [49,50].

The simplicity relates to a couple of disadvantages. First, the flow index m does not have a fixed dimension because this depends on the power index. More serious is the fact that for small and large shear rates the equation results in physically non-correct values. This gives problems in its application in numerical simulation. Frequently Ostwald de Waele parameters are presented with shear exponents less than zero. This will result in physically nonsensical results, as e.g., a positive pressure gradient in a simple tube flow.

The Herschel-Bulkley equation is typically used to describe the flow of viscoplastic fluids [26]. It is a generalization of the Bingham plastic model to consider the case of a non-linear relationship between shear stress and shear rate [51].

$$
\pi = \pi\_{\mathcal{Y}} + \ m \,\dot{\gamma}^{n} \tag{3}
$$

With *τ<sup>y</sup>* the yield stress and m and *n* the flow index and the shear exponent.

It is believed *τ<sup>y</sup>* to be a fundamental parameter for the modeling of semi-solid metals behavior [52]. For this reason, a Herschel-Bulkley model was applied to the numerical simulation of the semi-solid processing [30,53,54], fitting experimental results for Sn-Pb15 alloy. More recently, the same model was modified to better describe the phenomena taking place in the early stages of deformation of the solid structure [30].

To model thixotropy, three main approaches are applied. One is to describe the change in the structure with the change of the viscosity, which needs to define a rate equation for the temporal development of the viscosity [55]. Another approach is to calculate the number of existing and broken connections between the particles, which depend on the local shear rate [56].

The most promising model is the one that was originally worked out by Moore [57]. He defined a structural or coherency parameter *λ*, which is defined to be 1 in a fully saturated state of the structure and to be 0 when all particle bonds are broken. A rate equation for the structural parameter is set up to consider the creation and the destruction of bonds. It is frequently assumed that all parameters of the Herschel-Bulkley equation are depending on *λ* [58].

A more detailed model has been set up by Petera et al. [59]. The SSM is modelled as a two-phase system with a semi-fluid approach for the solid phase. A kinetical equation is introduced reflecting the change in structure. The model has been successfully applied for the simulation of die-filling experiments where the flow front was videotaped. Good agreement was achieved between simulation and experiment for the development of the flow front, the transient pressure drop in the die and the final distribution in the solid phase due to segregation [60].

The Cross model considers that at extreme boundary condition, i.e., at very low or very high shear rate, thixotropic fluids assume a Newtonian viscosity [61]. This is expressed by the following equation:

$$
\eta = \eta\_{\infty} + \frac{\eta\_0 - \eta\_{\infty}}{1 + k \cdot \dot{\gamma}^n} \tag{4}
$$

with *η*<sup>0</sup> the viscosity for zero shear rate and *η*<sup>∞</sup> the viscosity for high shear rates and *k* and *n* parameters as in the Ostwald de Waele equation.

This model has been applied to fit the experimental results from various researches on SnPb15 alloy in a satisfactory way [62], even though consistent data about the extreme conditions are hardly available in literature and, therefore, the reliability of the model cannot be stated [7]. From the practical point of view, it does not provide an advantage compared with Ostwald de Waele model since the Cross model reduces to the Ostwald de Waele one if the extreme condition viscosities are not determined.

The above-mentioned approaches are applicable if the solid fraction is below approximately 0.65, which corresponds to the maximum packing of the solid particles. Above this value, the SSM can be treated as a "porous solid body" and the approaches of the continuum mechanics must be applied to model the relation between stress and deformation.

#### **5. Steady State Condition: Time-Independent Properties**

As aforementioned, it is fundamental to distinguish between the properties of SSM in steady state and transient conditions. In this paragraph, the main findings related to time-independent behavior will be reviewed according to the experimental procedure applied.

Comparison of data available in the literature have been done in the past for A356 and A357 [44]. It was found that the flow curves for both alloys, expressed using a power law relationship, were characterized by a slope of approximately −1, corresponding to the shear exponent. The comparison was carried out among results obtained by means of various techniques and conditions (Figure 8), which is expected to lead to discrepancies in the flow curves, even when studying the same alloy.

As additional evidence of this, Lashkari et al. [63] and Blanco et al. [49] studied a similar Al alloy containing approximately 4.5% Cu using respectively compression tests and shear rate jump experiments with a Searle rheometer. In this case, the difference is also in the range of shear rate investigated since compression tests results correspond to very low shear rates (in the order of 10<sup>−</sup>3–10−<sup>2</sup> s<sup>−</sup>1), while in the other study a very different range was investigated (60–260 s−1).

Regarding shear experiments, more recently Das et al. [64] performed various experiments on A356 alloy with a Searle-type rheometer applying the power law model to evaluate their results. Furthermore, they compared the obtained m and n parameters with the findings of previous researches. The difference between the values was mainly due to the different ranges of shear rates considered for the fitting of the flow curve since Das et al. [64] performed rheological measurements up to 1500 s−<sup>1</sup> as shear rate values, while the other authors considered a narrower range (approximately up to 200 s<sup>−</sup>1).

To update the available information, in the present review the attention was focused on the studies from 2003 up to now. Values of flow index and shear exponent for Al alloys were collected from various scientific publications, when available, and they are shown in Tables 1 and 2. It was chosen to organize the data according to the experimental methods used, i.e., shear (Table 1) or compression experiments (Table 2) in order to better represent the rheological behavior of these materials. At this regard, it is important to mention that experiments with rotational rheometers are useful for the investigation of the rheological behavior of semi-solid slurry with a solid fraction of 0.2–0.5, while compression experiments can provide information for materials characterized by a solid fraction higher than 0.5.

**Figure 8.** Comparison of apparent viscosities obtained by various experimental techniques and conditions [44].

It appears that the shear exponent assumes values between −1.5 and −1.2 for shear rates up to 300 s−1, while it decreases if a wider shear rate range is applied. Values very close to 0 are found for investigations carried out at low solid fraction. This is reasonable if it is considered that liquid metals exhibit Newtonian behavior. Consequently, it is expected that the flow of semi-solid slurries at low solid fraction would shift towards a Newtonian-like behavior.


**Table 1.** Power law parameters from experiments with rheometer for different alloys.

This is particularly evident if the flow curves corresponding to the parameters listed in Table 1 are plotted in a viscosity vs shear rate graph, as in Figure 9. It clearly appears that the slope of the flow curves obtained for solid fraction above 0.3 are comparable, while for lower values of solid fraction the viscosity is less dependent on the shear rate, reflecting the increasing contribution of the Newtonian liquid phase. Furthermore, also the influence of solid fraction is visible if the results for the same A356 alloys are considered. On the other hand, a scattering in the viscosity values of semi-solid AlSi alloys is observed, which underlines how these kinds of measurements can be affected by various parameters, as such as material preparation, holding time, shear history of the samples, etc. Despite the large number of experiments available in scientific literature, it is still difficult to define viscosity values for SSM in an unambiguous and systematic way.

**Figure 9.** Comparison of flow curves according to parameters reported in the previous Table 1.

It is more complex to perform the same analysis for flow curves obtained from compression experiments, since less abundant data are available in scientific literature, as shown in Table 2.


**Table 2.** Power law parameters from compression experiments for different alloys.

It is important to mention that most of the shear exponents are negative values. As mentioned above, this gives strange physical results. It must be assumed that the experiments were dominated by secondary effects that influence the experimental results, e.g., wall slippage. The application of these results seems to be doubtful. This is even more evident for the results of the compression tests, which indicate the difficulty to extract reasonable results for the shear behavior from compression tests.

On the other hand, compression experiments can provide important information about the transient and time-dependent properties of SSM close to practical application [68] even when their use for simulation might be doubtful.

#### **6. Non-Viscous Properties**

#### *6.1. Yield Stress*

It is obvious from the previous explanations that the rheological properties of SSM strongly depend on time—which is frequently forgotten. This is insofar important as any die-filling process, either casting or forging, is a non-stationary process by nature with temporally changing border conditions for the deformation of the material.

An important question is how the typical time scales of the process and the material behavior compare when discussing an appropriate approach to model the material behavior properly [51,69]. This matter becomes evident when investigating the phenomenon of yield stress in SSM. The existence of yield stress is subject of basic discussions, which seem to have philosophical character [70,71]. There is no doubt that the experimental detection of yield stress depends on the quality of the experiment. However, this holds for any mechanical property as well. To identify e.g., thixotropic properties it is necessary to have an instrument with a certain minimal temporal resolution.

In general, it is accepted that the introduction of yield stress for materials with viscoplastic behavior is at least a reasonable engineering approach to model adequately the flow behavior [11,52,72]. As with other materials, the yield stress in SSM is the consequence of the formation of a network of the particles, which is more or less mechanically stable. The direct way to investigate yield stress phenomena is the application of stress or strain ramps and the observation of the temporal behavior of strain or stress respectively [69]. Experiments [6,15] show that the yield stress depends on the solid fraction, the particle size and shape and on the resting time. With increasing resting time, the yield stress increases nearly exponentially, which leads to the concept of three different values of yield stress [15,16]:


The isostructural value cannot be measured directly and must be found, for instance, by extrapolating the dynamic yield stress to resting zero time. For two Al alloys, it was found that the isostructural yield stress is less than 1/10 of the static value, which was for both alloys about 400 Pa. Comparable experiments have been performed by Solek [73] with high carbon steel, which confirms the above-mentioned findings.

Application of oscillatory shearing and comparison of the temporal development during rest of the yield stress and the loss- and storage-moduli show that the ratio of the both moduli decreases while the yield stress increases [74]. Due to the formation of a stable structure in the material the rheological nature of the material changes from viscous to elastic—fluid to solid—as indicated by the loss angle. The temporal increase of the storage modulus compares with the increase of the yield stress.

The other way to determine yield stress is an indirect one by evaluating the flow curve on the base of an appropriate visco-plastic model. A Herschel-Bulkley model was found to fit in a proper way the experimental results for semi-solid Sn-Pb15 alloy at different solid fractions [60]. The alloy was considered as a homogeneous material with thixotropic properties and it was tested under isothermal conditions. Among the parameters obtained from the application of the model to experimental results, it was possible to calculate finite values of yield stress as a function of the solid fraction. A Herschel–Bulkley approach was applied to the numerical simulation of die filling and it was validated with the results from die-filling experiments [75] using a Sn-Pb alloy. In these experiments, the evolution of the flow front as well as the pressure drop were observed during the die filling and the process was performed for different flow conditions. Similarly, a model corresponding to the Herschel-Bulkley was developed to represent the behavior of semi-solid metallic suspensions in fast transient conditions [76]. The validation of the model was carried out with short time measurements of Sn-Pb15 alloy under rapid shear rate changes, already published by other authors [77]. The model can show the increase of the shear rate in a shear rate jump and the gradual decrease after reaching a maximum. For the yield stress, the authors found interestingly a constant value of 100 Pa, independent of the time the material was in rest (which was between 0 and 5 h).

The evaluation of yield behavior of Al-Si alloys was carried out also using different experimental procedures, as such as the compression and the cone penetration method [72]. Yield stress was measured as a function of temperature, i.e., solid fraction, taking also into account different billet processing methods, as such as the addition of grain refiner and the application of magnetohydrodynamic stirring (Figure 10).

**Figure 10.** A comparison between measured average yield stresses of grain refined (GR) A356 and magneto-hydrodynamically stirred (MHD) A356 alloys with yield stress values predicted by the Loue–Sigworth (LS) model [72].

It was found that not only the solid fraction is affecting the yield behavior, but that also the entrapped liquid and the morphology of the solid globules can play an important role in determining the deformation resistance. A high amount of entrapped liquid can lead to an increase in yield stress because the "effective" liquid fraction is decreased, while the presence of more rounded solid particles can result in a smaller deformation resistance and, consequently, in a decrease in yield stress.

The viscoplastic behavior of A356 alloy in semi-solid state was expressed by the Herschel-Bulkley model by Simlandi et al. [78]. They also considered a time-dependent structural parameter to comprehensively describe the material behavior at different solid fractions.

An enhanced model based on the Herschel-Bulkley equation was proposed also for other metallic alloys, as such as steels [79] and Mg alloys [80]. Pouyafar et al. [79] calculated the yield stress of M2 steel from the flow curve obtained from experiments in steady-state flow by interpolation of the experimental results to evaluate the stress at zero shear rate. The enhanced model was compared to the classical one and it was found that the new model could predict in a more precise way the behavior of the material, especially at high shear rate, than the conventional one. This was confirmed also by the application of the model to the simulation of the rheometer flow. Moreover, the same model was used to simulate the material behavior during compression test [36].

#### *6.2. Transient Behavior: Time-Dependent Properties*

In the introduction, the main thixotropic properties of SSM were briefly discussed. It was mentioned that thixotropic materials show a decrease of viscosity under constant stress and a recovery of the viscosity when the loading stops. This behavior is linked to microstructural evolution of SSM, in particular to the agglomeration and deagglomeration phenomena.

Various experiments can be performed to better investigate these time-dependent properties, which focus on the response of the SSM to changes in the applied deformation rate or shear stress, as such as shear rate jump or hysteresis loop experiments [60,65,68] or rapid compression tests [45,81].

Modigell et al. [60] investigated thixotropic behavior under isothermal condition on a Sn-Pb15 alloy. Hysteresis loops experiments are particularly useful to provide qualitative information on the degree of thixotropy of the material. It was found that the faster the ramp is performed, the higher the hysteresis area and, therefore, the more evident the thixotropic behavior (Figure 11). On the other hand, Brabazon et al. [65] tested an AlSi4 alloy in a similar way but found that with slower ramp the peak viscosity increases corresponding to an increase in thixotropy, not considering the change around the hysteresis loop. They also evaluated the effect of rest time on the material thixotropy and measured an increase in the viscosity with increasing rest time.

**Figure 11.** Hysteresis experiments at different sweep times (*T* = 198 ◦C, *Fs* = 0.45) [60].

It is reported [60] that, immediately after an increase in shear rate, the measured viscosity makes a sudden jump and, subsequently, a decrease follows according to the thixotropic nature of the material. The opposite happens when the shear rate is decreased. This phenomenon is said to be "isostructure behavior". It is assumed that the structure of the alloy—corresponding to the initial shear rate—will not change instantaneously to the structure corresponding to the new shear rate. Evaluation of special shear rate jump experiments in terms of an "isostructural flow curve" shows that the SSM immediately after a shear rate jump reacts in a shear thickening way [17]. Recent analyses of the literature [10] regarding this effect confirm the findings of [60]. They showed that it needs a solid fraction larger than 0.36 to be observed. It has been clarified that for shear rate jumps from shear rate zero (sample at rest) this shear thickening effect is masked by yield stress effects.

As above-mentioned, it is important to compare the time scale of the time-dependent rheological properties and the typical process times. Typical process times for die casting or forging are in the order of 1/10 s [40,82], as also confirmed by other authors [83,84] for the thixoforging of steel components. This is interestingly in the order of the response time of modern rotational rheometer. Detailed investigations about kinetics of the thixotropy in SSM are rare. The first well-known investigations of this phenomenon were done by Quaak and Peng et al. [85,86]. Quaak assumed that the process of deagglomeration and agglomeration after a change in shear rate is composed of two steps: a fast process to destroy or create agglomerates and a slow one of coarsening and sintering. This has been confirmed for Sn-Pb alloy by Koke [18], who could model the transient process assuming two different

kinetics. The time scale for the fast process was in the order of 0.5 s—and seemed to be independent of shear rate—whereas the scale of the slow process was 100 times slower. Liu at al. [77] studied the effect of shear jumps on the temporal development of the shear stress for Sn-Pb15 as well (Figure 12). They only investigated the short-term behavior and the time scale they found for the thixotropic effects is comparable with the results of Quaak [85] and Koke.

**Figure 12.** Shear rate jumps (1–200 s<sup>−</sup>1) for Sn15Pb alloy for two different solid fractions (*Fs* = 0.5 and 0.36) in isothermal condition [77].

Rapid compression tests on AlSi alloys have been performed by Hogg et al. [45], to investigate the rheological properties under the conditions of the real manufacturing process. Time scale of the test has been around 0.065 s, so much shorter than the scale of rotational rheometer. The experiments provided data also about the effect of the holding time and the reheating temperatures. The viscosity estimated from their experiments were by around a factor of ten higher then viscosity values evaluated from shear experiments. Assuming a time scale for the thixotropy of some 1/10 s it can be concluded that in these experiments the material was not in the final state of equilibrium. Based on these studies, also numerical models were developed to describe accurately the transient behavior [30,40,46,76].

Another interesting result regarding thixotropy can be extracted from the work of Bührig-Polaczek et al. [87] who performed experiments with Al 356 with a capillary rheometer. The device, which was developed in house, was equipped with four pressure sensors along the capillary. This allowed evaluating the pressure drop along the path of flow and, consequently, it was possible to analyze the thixotropic reaction of the material. The shear rates they could realize in the capillary were up to 3200 s−<sup>1</sup> and the resident time in the capillary was in the range of 1/10 s, which compares to typical process times in casting. They found shear thinning behavior in the shear rate range from 1000 to 3200 s−1, whereby the viscosity for 1000 s−<sup>1</sup> was approximately ten times higher than the viscosity measured in a rotational rheometer at shear rate equal to 250 s−1, which compares with the findings of Hogg [45]. Evaluation of the pressure drop along the capillary showed that the viscosity close to the entrance was around 2.5 Pa·s, while it dropped towards the end of the capillary to 0.6 Pa·s (both for a shear rate of 3200 s<sup>−</sup>1). The drop was almost linear and showed no tendency to become constant. This indicates that, within the residence time of 0.15 s, the thixotropic reaction is not finalized.

The consequences of this are that obviously the short term and transient behavior of the SSM is of much more interest for the typical casting and forging processes than the equilibrium flow curves. This is in accordance with numerical analyses of the flow of a viscoplastic material in the gap of a rheometer under conditions relevant for technical processes performed by Alexandrou et al. [30]. They composed the stress function by two contributions: the "steady state" stress, evaluated from long shearing time experiments and assumed to follow a Herschel-Bulkley model, and the stress due to slurry strength, which depends on a coherency parameter, depending on time. The simulation showed that for short times the latter contribution is dominant and defines the propagation of the shear in the gap. The relevant time frame is in the order of less than 0.1 s. Evaluation of experiments in this time scale need the application of computational rheology rather than the classical evaluation of the experimental data.

#### **7. Ostwald Ripening**

Another important aspect to consider to completely describe the rheological behavior of SSM is the growth and coarsening of the globular solid particles under rest or shear condition by recrystallization. This will happen under isothermal conditions. The phenomenon is usually indicated as Ostwald ripening and it takes place in different materials, as such as emulsion systems, etc. During this process, small particles dissolve, while the larger ones coarsen with the consequence that the surface energy is minimized [88] so that in total the mean particle diameter in the material will increase.

This phenomenon was investigated in the past by several researches [89–92] mainly by means of the observation of 2D sections of samples quenched from the semi-solid state, which represented a serious limitation in the investigation of the real phenomenon. More recently, agglomeration and growth of necks between solid particles, which are different from Ostwald ripening, and the later were investigated by X-ray in situ tomography [93,94].

In systems in rest, Ostwald ripening is a diffusion-controlled process that is rather slow and has less technical importance in the frame of the subjects discussed here. The growth of the particle as a function of time can be calculated on the base of the LSW-theory (Lifshitz, Sloyozov [95] and Wagner [96]) which gives a linear relationship between the volume of the particle and the time. In agitated systems, the growth of the particle is strongly enhanced by convective transport, which results in a growth of particles in a time frame which is at least relevant for a couple of experimental investigations.

The rheological effect of the Ostwald ripening is a decrease of the viscosity because of the increase of the mean particle diameter in the sample—although solid fraction will not change because of isothermal condition. This effect is demonstrated in Figure 13.

The Sn-Pb melt, initially fully liquid, is cooled down into the two-phase condition. First, the viscosity will rise, subsequently it will reach a maximum and will start to drop. The first decrease after the maximum is due to the formation of non-dendritic particles. At the first small peak, this process is more or less finalized. The further decrease of the viscosity is mainly due to convective Ostwald ripening. The increase in particle size is clearly demonstrated in the metallographic pictures. The time scale of this process is in the order of 60 min, depending on the shear rate. The higher the shear rate, the faster the growth.

**Figure 13.** Viscosity vs. time at constant shear rate (110 s−1). Initial cooling rate is 1 ◦C/min. Steady-state temperature is 198 ◦C, corresponding to a solid fraction *Fs* = 0.45. The metallographic pictures show samples quenched after 1 and 6 h of constant shearing (Sn–15% Pb) [17].

It is obvious that the Ostwald ripening falsifies the experimental results when performing long-term investigations, such as, for example, making shear rate jump experiments. For instance, in the case of A356 alloy tested in a rotational rheometer at 0.40 solid fraction, it can be seen in Figure 14 that at the end of the experiment the viscosity for a shear rate of 80 s−<sup>1</sup> is significantly lower than at the beginning. Evaluation of these data will result in an overestimation of the shear thinning effect. Modigell et al. [97] have developed a simple model considering convective transport in the system, which allows correcting the data for any shear, as shown in Figure 14. For sufficient large shear rates, the temporal change of the diameter is linear with time.

**Figure 14.** Shear rate experiment results with A356 under solid fraction of 0.40 [97].

The effect of isothermal and non-isothermal stirring on particle size was also evaluated for Al-Si alloys [98–101] and Mg alloys [102]. Sukumaran et al. [99] investigated the evolution of particle diameter with shearing time for an Al-Si alloy with and without the addition of grain refiner (Figure 15). In this case, it was found that first the particle size decreases due to the fragmentation of the dendritic structure. Subsequently, particle size reaches a minimum value before starting to increase due to the discussed Ostwald ripening mechanism, which is considered the main mechanism for coarsening of solid globules, while agglomeration phenomena is believed to contribute in a minor way. Due to the presence of the grain refiner, the growth of dendrites during semi-solid processing is inhibited, promoting the formation of an equiaxed microstructure, in comparison with the base material, and accelerating the formation of a globular microstructure, as well as the coarsening and ripening phenomena. Therefore, the coarsening of particles takes place earlier for the alloy with grain refiner in comparison with the base alloy, as visible in Figure 15.

**Figure 15.** Plot of nominal particle diameter vs. time of isothermal stirring (shear rate 210 s<sup>−</sup>1) at 615 ◦C [99].

The same mechanism is illustrated also by Chen et al. [102], who provided the following schematic diagram (Figure 16) and correlated the evolution of the particle size to the apparent viscosity for a Mg alloy.

**Figure 16.** Schematic diagram of the variation of microstructure for: (**a**) decreasing apparent viscosity and (**b**) increasing apparent viscosity [102].

#### **8. Summary**

In the present review, a summary of the basic aspects of the rheology of semi-solid metals are presented. These materials exhibit a complex behavior and several parameters must be considered to correctly characterize it. For this reason, the distinction between steady state and time-dependent properties is fundamental. Furthermore, additional important aspects as the evaluation of yield stress and Ostwald ripening mechanism are discussed.

In general, the analyses of the recent literature show that a lot of experimental and theoretical work has been done and that the initial findings of the early researches are confirmed. A database has been provided for higher melting alloys due to technical development in the field of rheometers. Even for steels, basic data are now available provided by shear experiments.

Most of the data give information about rheological behavior under equilibrium conditions. In technical practice, the processes are far from equilibrium due to the high process speeds. For better understanding of the process phenomena and for reliable simulation of the forming process, the short-term behavior of the SSM must be investigated more deeply than it has been done up to now. This means that the thixotropic properties, as well as the yield phenomena, must be studied to clearly understand the process phenomena under technical conditions. Additionally, there is a lack of knowledge in the relationship between the composition of the alloy and the rheological properties.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Microstructure and Properties of Semi-Solid Aluminum Alloys: A Literature Review**

#### **Annalisa Pola 1,\* ID , Marialaura Tocci <sup>1</sup> and Plato Kapranos <sup>2</sup>**


Received: 4 February 2018; Accepted: 6 March 2018; Published: 13 March 2018

**Abstract:** Semi-solid processing of aluminum alloys is a well-known manufacturing technique able to combine high production rates with parts quality, resulting in high performance and reasonable component costs. The advantages offered by semi-solid processing are due to the shear thinning behavior of the thixotropic slurries during the mold filling. This is related to the microstructure of these slurries consisting of solid, nondendritic, near-globular primary particles surrounded by a liquid matrix. This paper presents a review on the formation of this nondendritic microstructure, reports on the different proposed mechanisms that might be responsible, and illustrates the relationship between microstructure and properties, in particular, tensility, fatigue, wear, and corrosion resistance.

**Keywords:** semi-solid; microstructure; mechanical properties; wear; corrosion

#### **1. Introduction**

Semi-solid metal (SSM) processing is a manufacturing technique where an alloy, in the form of a slurry of near-globular primary particles in a liquid matrix, is injected into a die, allowing the production of near-net-shape components. The main advantage of this technology is related to the flow properties of the metal in the form of a slurry, which, in the semi-solid state, is non-Newtonian and exhibits shear thinning behavior [1,2]. The viscosity of the SSM slurry is higher than when fully liquid, reducing the risk of turbulent [3] or spray flow, which is more typical of conventional pressure die-casting [3,4]. However, thanks to the shear thinning behavior of the slurry, under the influence of a shear force acting on it when it flows into the die, the viscosity decreases and the metal slurry is able to fill the cavity completely in a nonturbulent manner. As a consequence, semi-solid cast parts are almost free of gas porosity.

Injecting a partially solidified alloy slurry has the added benefit that shrinkage porosity is virtually absent [5]. The low or even absent porosity allows the production of structural parts with good mechanical properties that can also undergo subsequent heat treatments or welding operations.

Semi-solid processing guarantees higher performance than die-casting, while maintaining a number of the advantages of die-casting, such as good dimensional tolerances, high production rates, high surface quality, complex near-net-shape parts, and thin sections with very limited need of any finishing operations [6]. In addition, when compared to conventional die-casting, SSM processing increases die life because of the lower stress associated with the lower injection temperatures and speeds (i.e., lower mold attack and erosion, lower thermal shock), reduced cycle times and risk of hot tearing due to the lower temperature of the metal slurry (no over-heating), and associated lower energy consumption [4,6,7].

On the other hand, SSM manufacturing requires specialized equipment for alloy preparation, combined with strict control of process parameters, particularly the alloy temperature, i.e., the solid/liquid

fraction. Unfortunately, all these tend to increase the production costs [8], even though recent investigations have shown that SSM processing is only slightly more expensive than conventional die-casting and cheaper than some other competing foundry processes [9].

Since its development in the early 1970s at MIT [10], much research has been performed worldwide, aimed mainly at developing new and alternative routes of feedstock production for different alloys suitable for SSM processing. At present, there are a number of different techniques used to produce semi-solid castings, differentiated by the percentage of liquid/solid fraction they employ and the way they produce the alloy in the semi-solid state. SSM methods can be divided in two main categories according to their processing route, known as rheocasting and thixocasting [11]. In rheocasting, the semi-solid slurry is prepared in situ from the liquid state down to a certain percentage of solid fraction (usually between 10 and 30% [12]), and then directly transferred into the shot sleeve for being injected into the die. In thixocasting, a billet, characterized by an almost globular or rosette-like microstructure developed through some specific route, is reheated in the mushy zone (semi-solid region) to an appropriately chosen solid fraction (usually between 50% and 60% [13]), placed in a modified shot sleeve and finally injected into the die [14]. Another classification methodology distinguishes the SSM methods according to the initial step used for obtaining the semi-solid feedstock, i.e., from the liquid state, by controlled solidification or from the solid state, via heavy plastic deformation and recrystallization [11].

Some of the most common routes for feedstock material preparation are: mechanical stirring, such as the SSR [15,16] or GISS [17] processes, electromagnetic stirring (EMS or MHD) [18,19], ultrasonic stirring (UTS) [20,21], New Rheocasting (NRC or UBE) [22,23], cooling slope [24], twin screw [25], Rheometal [26], liquid mixing method [27], SEED [28], thermomechanical [29], and SIMA [14,30].

A number of interesting reviews about the different technologies available for obtaining nondendritic slurries can be found in the literature [11,13,14,31–33]. Nevertheless, independently from the chosen technique, the fundamental concept is based in developing feedstock with a microstructure of a solid, near-globular phase surrounded by a liquid matrix when in the semi-solid state. As already mentioned, when a shear stress is applied, the near-globular solid particles move easily between and over each other, reducing the viscosity and making the material behave like a liquid. On the contrary, when a shear stress is applied on a dendritic microstructure, the liquid remains entrapped between dendrite arms and prevents them from moving freely, thus increasing the viscosity of the alloy [32].

This paper presents a review on the formation of nondendritic microstructures, microstructures that have a key role in semi-solid processing, and discusses the different proposed mechanisms together with ways to analyze SSM microstructures. In addition, the review provides information on the variation of mechanical properties and corrosion behavior through modification of microstructures that are typical of SSM processing.

#### **2. Formation of Nondendritic Microstructures in SSM Processing**

During the early experiments performed by Spencer et al. on a Sn–Pb15 alloy [10], it was found that the microstructure of the material was strongly affected by the constant shearing of the alloy when in semi-solid state. Particularly, it was shown that shearing action causes the formation of a nondendritic grain structure, which is the distinctive characteristic of semi-solid alloys. Moreover, with further shearing during cooling, it is also possible to obtain spheroidal particles, typically with some entrapped liquid [4]. The authors also reported that high shear rates and slow cooling rates can promote the formation of spherical particles instead of rosette-like ones [10].

The steps for the formation of nondendritic microstructures have been extensively studied over the years and one of the first proposed mechanisms is shown in the schematic illustration of Figure 1.

**Figure 1.** Globule formation during stirring in the semi-solid range: (**a**) initial dendritic fragment, (**b**) dendritic growth, (**c**) rosette, (**d**) ripened rosette, and (**e**) spheroid [4].

According to Flemings and co-workers [4], in the early stages of solidification, as it happens for all metallic materials, dendrites form in the liquid. However, unlike conventional solidification, the shearing action affects the dendritic morphology, which changes into that of a "rosette" due to different phenomena. Various explanations about the conversion mechanisms from dendritic to globular morphology can be found in the literature like ripening, shear, bending and abrasion with other growing crystals, dendrite fragmentation, remelting of dendrite arms, and growth control mechanisms [2]. Vogel et al. [34], for instance, proposed that under a shear force the dendrite arms bend plastically, thus introducing large misorientations into the arms and forming dislocations. At high temperatures these dislocations rearrange themselves inducing, under specific conditions, the detachment of dendrite arms [35] as shown in Figure 2. These dendrite fragments act as nuclei, coarsening and leading to the presence of globules of the primary phase [4,35,36].

**Figure 2.** Schematic model of fragmentation mechanism: (**a**) undeformed dendrite, (**b**) after bending, (**c**) dislocation rearrangement to give grain boundary, and (**d**) grain boundary wetting [35].

In contrast, Molennar et al. [37] proposed that rosette-like particles are the result of cellular growth. Mullis [38] reported that bending could bring about rosette formation without any need of mechanical effects due to shearing. According to Hellawell [39], the secondary dendrite arms can separate at their roots because of solute enrichment and thermosolutal convection that determines their remelting rather than breaking off for simple mechanical interactions (Figure 3). He suggested that in the solidification range, the solid is completely ductile and dendrites can be bent but not broken. Hence, the detachment of the secondary arms can be explained by a local remelting phenomenon. In particular, the remelting can occur either by recalescence of the whole system or by local recalescence due to fluctuations caused by convective phenomena or stirring [14,40].

**Figure 3.** Schematic model of dendrite arms remelting [4].

The effect of the fluid flow characteristics on the morphology of solidification structures was also studied, by means of Monte Carlo simulation, by Das et al. [41]. They found that a rotational motion under laminar flow promotes rosette-like morphology due to a periodic stabilizing and destabilizing of the solid–liquid interface, while a turbulent flow hinders dendritic growth resulting in a compact morphology due to a stable solid–liquid interface. The presence of a concurrent mechanism was also proposed.

An interesting review of the various proposed mechanisms is reported in [2].

Figure 4 shows a typical microstructure of semi-solid castings, which consists of rosette-like or even globular grains (Figure 4a) and a dendritic structure typical of conventional casting processes (Figure 4b) [42].

**Figure 4.** Typical microstructures of AlSi7 component (**a**) semi-solid cast and (**b**) conventionally cast.

#### **3. Microstructural Analysis of Semi-Solid Alloys**

In the analysis of microstructures of SSM components, dendrite arm spacing cannot be measured as in the case of conventional castings [43]; instead, the microstructural parameters taken into account are the size of the globules, their shape factor, i.e., their roundness, and the amount of entrapped liquid [44].

The globule size should be large enough to build an almost rigid solid phase network and, at the same time, small enough that the slurry can flow into the die cavity similarly to a liquid. The dimensions of the globules are usually defined as their mean diameters. It is typically assumed that the minimum thickness that can be filled by SSM should not be lower than 20–30 times the grain radius [11]. In particular, according to some findings, the optimum primary particle size for SSM alloys is lower than 100 μm [45]. However, the grain size distribution measured by 2D analysis is difficult to determine and sometimes an extensive analysis of serial sections is needed to guarantee reliable data. A more accurate and consistent analysis of the grain size distribution and evolution can be performed via 3D examination methods, like X-ray microtomography [46–50], as shown in Figure 5.

**Figure 5.** (**a**) Schematic representation of 2D analysis [50], (**b**) 3D image processing with 2D slice, and (**c**) volume representation of the acquired tomography data [51].

The shape factor is known to strongly affect the slurry viscosity. For laminar die filling in particular, the solid particles should preferably be round and separated from each other [11]. The shape factor (*F*) is defined as:

$$F = 4\pi A/P^2$$

where *A* is the area and *P* is the perimeter of the particles.

A shape factor equal to 1 represents the case of a perfect circle and it reduces to zero with an increasing amount of irregularity, i.e., highly branched or elongated microstructures. In semi-solid processing, a shape factor above 0.6 is considered as appropriate [44]. Because the SSM particles can have complex morphologies, care should be taken when interpreting metallographic sections, as errors can arise if isolated secondary branches are taken into account as real single particles [52]. Thus, in order to obtain more reliable results, a high number of particles have to be measured.

The entrapped liquid is a distinct feature of the thixocasting route [11,53]. This liquid, as shown in Figure 6, does not contribute to the sliding of the globules during processing. Therefore, it follows that the liquid fraction is lower than the theoretical one and that the viscosity increases during die filling due to a sponge effect that can be induced [11]. The amount of entrapped liquid can be estimated by image analysis of 2D polished section areas, however, as in the case of the shape factor and globule size, errors can be introduced when some of the entrapped liquid islands appear to be isolated even though they are connected to the liquid phase at deeper levels. A more thorough investigation by means of 3D analysis allows more reliable data to be obtained.

**Figure 6.** Thixocasting microstructure with entrapped liquid in the globules.

Clearly, different production routes of semi-solid parts can result in different microstructures, from rosette-like to near globular and details about them can be easily found in the literature [2,14].

#### **4. Performance of Semi-Solid Aluminum Alloys**

In recent years, several studies have been focused on the mechanical properties of parts fabricated by semi-solid/thixoforming processes, often in comparison to conventional routes.

#### *4.1. Mechanical Properties*

Al alloys are widely used in semi-solid processing as discussed in the previous section. The proven better quality of components obtained by SSM processes, and associated better properties, are often cited in textbooks about the topic [11,13], in particular highlighting the possibility to perform T6 heat-treatment to further increase their characteristics. There are also many scientific studies on the mechanical properties of Al SSM alloys (both casting and wrought ones), even though significantly less than those on microstructural modification in comparison with conventional casting alloys.

#### 4.1.1. Tensile Behavior

Since the very first production attempts, it was evident that semi-solid parts have high mechanical properties comparable to those of the forged material and better than permanent mold castings [4]. Over the years, these findings have been confirmed by many authors. In fact, the enhancement in performance of parts manufactured by SSM processing compared to traditionally cast parts is supported by various studies mainly on Al–Si [54–59], Al–Cu [60,61], and Al–Zn alloy families [62,63].

Bergsma et al. [64], for instance, reported that the tensile strength of 357 and modified 319 semi-solid formed aluminum alloys are superior to conventionally cast alloys due to the reduction in porosity and the spherical microstructure, when an effective optimization of heat treatment parameters is achieved. Cerri et al. [65] showed excellent ultimate tensile strength and yield strength of 319 alloy after appropriate heat treatment, in the order of 350 MPa and 280 MPa, respectively, thus almost 100 MPa higher than the conventionally cast counterparts. Similarly, Zhu et al. [66] analyzed different casting and forging alloys used industrially for the production of compressor wheels, finding that strength and ductility approach those of the forged components after T61 heat treatment, as shown in Figure 7. Nevertheless, in their work, the influence of microstructural features is not thoroughly investigated.

**Figure 7.** SSM 319 vs. cast C355, 354.0, 319 alloys and forged 2618 alloy after T61 heat treatment [66].

On the contrary, Haga et al. [67] discussed the influence of the size of primary globules on tensile properties of semi-solid castings. They underlined that the remarkable tensile properties are due not only to the nondendritic microstructure, but also to the small size of the primary α-Al, which is especially effective in enhancing the elongation to fracture. In particular, for the A356, the elongation can reach up to 18% by using the cooling slope method.

Another study on an Al–Si–Mg–Fe alloy demonstrated the influence of the shape of primary α-Al globules on ultimate tensile strength and elongation [68], a fact that is shown in Figure 8, i.e., when the shape factor increases (the more rounded the primary globules), tensile and elongation properties also increase.

Lü et al., investigating the behavior of rheocast 5052 alloy in comparison with gravity (GC) and high pressure (HPDC) die casting [69], noticed that fine and uniform microstructure throughout the entire SSM sample would effectively reduce stress concentrations at the grain boundaries under applied stress. They concluded that the globular shape is effective in enhancing the tensile strength and ductility, as detectable by fractographic analyses that show an almost ductile fracture mode for the rheocast alloy instead of the mixed ductile brittle fracture experienced by the gravity cast samples and with smaller dimples than those on the conventional casting sample (Figure 9).

**Figure 8.** Effect of shape factor of primary α-Al phase on ultimate tensile strength and elongation of semi-solid slurry for Al–Si–Mg–Fe alloy in as-cast condition [68].

**Figure 9.** SEM micrographs of tensile fracture for samples produced under different processing conditions: (**a**) Rheo-HPDC; (**b**) conventional HPDC, and (**c**) conventional GC [69].

In Figure 10a, the comparison between the properties of semi-solid, casting, and forging Al–Si alloys summarized by Brochu et al. [70] is shown, further supporting the above-mentioned advantages. Clearly, different SSM process routes can result in different mechanical properties, as shown in Figure 10b [71]. However, the improved trend, as compared to traditional casting process, appears to hold in all cases as a consequence of the higher soundness of the components and the enhanced microstructure of semi-solid alloys, as discussed above.

**Figure 10.** (**a**) Mechanical properties of SSM, cast and forged aluminum alloys [70] and (**b**) comparison of mechanical properties of A357 alloy under T5 condition obtained from different processing technologies [71].

Apart from the influence of primary α-Al grains, the effect of different alloying elements, as well as the influence of secondary phases in SSM alloys, were examined. For example, the addition of Si and Fe in a 206 aluminum alloy for an automotive application was investigated by Lemieux et al. [72], who found remarkable performance of the tested rheocast components.

The influence of a significantly higher Si content (approximately 20 wt. %) was examined in some studies, like that on a rheocast Al–Si–Cu alloy [73]. Recently, hypereutectic alloys processed by SSM methods have attracted the interest of researchers because of their heat-resistant properties. In hypereutectic Al–Si alloys, primary Si grains solidify as coarse plate-like particles, which can be refined by SSM processing, thus improving tensile properties, as demonstrated by Zheng et al. [74] studying the properties of AlSi30 rheo-diecast (RDC) compared to conventional die-casting. They pointed out that the UTS, elongation, and hardness of the SSM samples are approximately 57.9%, 42.9%, and 20.6% higher than those of the die-casting ones, respectively. They attributed this to the finer compact primary Si grains, which can reduce or even eliminate crack initiation, combined with reduced porosity. The development of a series of hypereutectic alloys based on the A390 composition (17% Si, 5% Cu, 0.5% Mg) and their thixoforming, have already been described by Kapranos et al. [75], together with their resulting microstructures and mechanical properties. Again, the main advantages of thixoforming were related to the improved size and morphology of brittle Si particles in comparison with conventionally cast parts, in addition to the expected spheroidisation of the Al matrix. The successful thixoforming of an automotive brake drum was reported and represented an interesting example of substitution of a conventional cast iron part with an aluminum one.

Concerning the effect of secondary phases (i.e., iron intermetallic compounds), interesting analyses can be found in the work of Shabestari et al. [56]. This is particularly interesting since numerous studies have reported the correlation between amount and morphology of secondary phases, for instance, intermetallic ones containing Fe, Mn, Cr, Ni, and mechanical properties [76–80] for casting Al–Si alloys. In particular, these authors found that the peculiar microstructural characteristics of the thixoformed alloy, such as the extremely low porosity, fine and equiaxed morphology of the α-Al grains and uniform distribution of intermetallic compounds fragmented by the process route, enhance strength and elongation, in comparison with the as-cast condition. A similar topic was also discussed by Möller et al. [81] in a study of the microstructural and tensile properties of semi-solid metal high pressure die cast F357 alloy with various additions of Fe, Ni, and Cr. In this case, the formation of intermetallic phases containing Fe and Ni lead to a decrease in strength and ductility due to their microcracking during tensile tests.

Nowadays, the attention of the researchers is focused on the investigation of the properties of less conventional alloys for SSM processing, like AlSi8 [82], Al5Fe4Cu [83], Al–Zn–Mg–Cu [84], AlZnMg alloys with Sc addition [85], etc., in order to evaluate the advantages in their application as semi-solid manufactured products.

#### 4.1.2. Fatigue Behavior

Different researchers have compared high cycle fatigue resistance of SSM alloys (mainly based on Al–Si) with that of conventional casting ones [64,70,86–97]. There is a general consensus in these papers that the SSM samples show higher fatigue resistance, even though the data about the high cycles fatigue strength results is quite scattered (ranging from 60 to 180 MPa) depending on the process route used, casting and heat treatment parameters, as well as the possible presence of defects. The authors agree that the superior fatigue performance of SSM parts is related to the fact that they are almost free from defects like gas and shrinkage porosity as well as oxide inclusions. Thus, in defect-free parts, the microstructural constituents play a key role in fatigue crack nucleation and propagation [98]. As reported in [70,99], for instance, a high volume fraction of primary α-phase significantly increases the resistance to crack initiation. In particular, Park et al. found that fatigue cracks propagate mainly by cutting through the primary α-phase when a low volume fraction of α-phase is present; on the contrary, when the volume fraction is high, they mainly bypass the primary α-phase following the phase boundaries [87].

Other reasons for fatigue improvements seem to be the smaller globule size as well as the finer size and distribution of eutectic Si particles of SSM castings [100]. Relating to the former, it is reported that fatigue strength increases with decreasing primary α-phase size and that also, the globular morphology plays a positive role [70,86]. Additionally, the level of α globules agglomeration, which determines the size and distribution of Al–Si eutectic regions, influences the fatigue crack threshold. Concerning the Si particles, their interface with the α-phase in the eutectic can act as nucleation point of fatigue cracks because of the mismatch of plastic deformation between each other that causes fracture and/or decohesion of Si lamellae. It follows that their more uniform distribution and fine size improve fatigue crack initiation resistance, as documented by different authors [70,87].

Ragab et al. [89] showed that grain boundaries in SSM microstructures act as a barrier to propagation of short cracks. Additionally, they provide evidence that fatigue failure is also associated with the presence of oxides as well as slip bands and intermetallic phases. Clearly, as in the case of conventional casting, with the presence of platelet-like and needle-like shapes, Fe-rich intermetallic compounds reduce the fatigue properties, as their morphology is conducive to high stress concentrations, thus making them a source of cracks able to cause failure [95]. Figure 11 shows an example of a fatigue fracture surface [95]. A mixed fracture mode can be clearly seen with both cleavage cracks, which induce brittle fracture, and dimples in the α-phase, which denote a ductile fracture.

**Figure 11.** Scanning electron microscope analysis of the fatigue fracture surface of SSM A357 samples after T6 treatment [95].

It is known that iron up to 1.20% is needed in conventional die-casting to prevent alloy sticking onto die surfaces at high temperatures due to chemical, metallurgical, and mechanical interactions [101]. Interestingly, Al alloys for SSM contain lower amounts of Fe than conventional alloys for die-casting due to the reduced risk of die soldering related to the lower injection temperatures and speeds used during the process.

#### *4.2. Wear Resistance*

Regarding wear resistance, the dry sliding behavior of some SSM alloys has been evaluated in comparison with conventionally cast parts. Dey et al. show that the globular microstructures of A356 alloy castings appeared beneficial for dry wear resistance, as compared to dendritic ones [102]. In particular, the results of friction coefficients of SSM samples were lower than that of conventional cast specimens for all loads. The same holds for the wear loss.

A remarkable improvement in wear resistance of semi-solid specimens in comparison with conventional castings is also reported by Bayoumi et al. [103], who measured a lower wear rate for semi-solid processed A356 alloy. On the other hand, a comparable erosion mechanism was observed. Concurring results were found by other authors. The results from another study on the tribological properties of A356 alloy [104] showed that, for all applied loads, heat treatment enhanced the alloy behavior, while improvements caused by thixocasting were not systematic, i.e., a lower friction coefficient was noticed just for lower specific loads. Overall better wear resistance of thixocast materials, as compared to the original alloy, was attributed mainly to the improved distribution and smaller size of Si particles.

Several hypereutectic Al–Si compositions have also been investigated. It is a known fact that the presence of hard Si particles distributed in the matrix induces an outstanding wear resistance. However, the presence of casting defects like porosity, typical of traditional foundry processes, reduces their performance. In this regard, SSM processes make these alloys promising candidates for heavy wear applications. Very interesting results were obtained with A390 SSM alloy [105].

Birol et al. stated that the enhanced wear performances of hypereutectic Al–Si alloys are linked to the uniform distribution of fine primary Si particles [106]. Thus, the combination of a favorable silicon dispersion and the better soundness induced by the semi-solid processing gives superior wear performance.

Likewise, other chemical compositions showed comparable results in terms of improved wear behavior [63,107]. Recent studies on A319 confirm that the uniform distribution of Si, the reduction in porosity level, and the different morphology, size, and distribution of intermetallic phases obtained by SSM processing are responsible for better wear behavior than conventionally cast samples, even though the predominant wear mechanism remains the same for both the alloys [108,109].

Other damaging mechanisms of SSM components have been investigated over the years, such as cavitation resistance [110,111]. It was reported that the globular microstructure, as obtained by ultrasound treatment methods (UTS), increases the cavitation erosion resistance of the alloy because of the higher chemical and microstructural homogeneity, the morphology of the primary particles, and the refined structure of the eutectic due to the treatment itself. A comparison of the eroded area, at macroscopic scale, of conventionally cast (NUST) and SSM (UST) Al–Si7 samples is shown in Figure 12. It can be clearly seen that the highest damage was experienced by the conventionally cast alloy in the as-cast condition (Figure 12a), whereas the highest erosion resistance was exhibited by the heat-treated semi-solid sample (Figure 12d).

**Figure 12.** Surface topographies of the damaged areas of conventionally as-cast (**a**) before and (**b**) after T6, semi-solid in (**c**) as-cast and (**d**) after T6 [111].

#### *4.3. Corrosion Resistance*

Corrosion behavior of SSM aluminum alloys has not been explored to the same extent as mechanical performance and investigations on this property are more recent. For instance, Bastidas et al. [112] studied the pitting corrosion of A357 rheocast alloy, showing that it preferentially moves through the eutectic regions due to the Si particles that play a remarkable role in the corrosion process and to the cathodic properties of the intermetallic compounds.

A comparison between SSM and permanent mold cast A357 alloy was studied by Yu et al., who showed that both the resistance to corrosion and stress corrosion cracking is higher for semi-solid microstructures [97]. Similar results were obtained by other authors when comparing A356 semi-solid and low-pressure die-cast components. They highlighted that the number of pits and the degree of corrosion is higher for the conventionally cast products than in rheocast ones [99]. Similarly, Arrabal et al. [113] provide evidence about the better resistance of SSM as compared to gravity cast A356 alloy, as detectable by its lower cathodic current densities (Figure 13).

**Figure 13.** Cathodic polarisation curves of (**a**) gravity cast (GC) and (**b**) SSM (RC) A356 alloy after 1 h immersion in 3.5 wt. % NaCl naturally-aerated solution [113].

The improved corrosion resistance of the SSM Al–Si alloys is related to the reduced area ratio between Si particles and α-phase lamellae in the eutectic compared to that of their traditionally cast counterparts [99,114,115]. In particular, the differences in Si particle size and shape and, consequently, in the area ratio between silicon and α-phase in the eutectic, are related to the different solidification rates combined with the different applied pressures during the casting processes [114].

Analyzing the damage by scanning electron microscopy, it is noted that the α globules appear almost not to have been attacked, while the Si particles and the intermetallic compounds in the eutectic have a fundamental role in the corrosion damage [113,116], acting as local cathodes (Figure 14).

**Figure 14.** Cathodic effect of silicon and intermetallic particles [116].

Interestingly, some authors have also investigated the effect of surface eutectic segregation on corrosion resistance [117,118]. This thin layer of liquid phase can be found in semi-solid products due to the "sponge" effect [11], when the slurry is injected under pressure into the die, and it contains a higher amount of alloying elements than the bulk. Because, as reported above, the pitting corrosion in Al–Si alloys occurs preferentially in the eutectic area, the semi-solid parts characterized by the presence of this segregation layer are more prone to the phenomenon. Recently, 3D SEM tomography showed that corrosion takes place more at the interface between α-phase and Fe-rich intermetallics than at the eutectic Si ones [119].

Some investigations are also available in the literature about other Cu-containing aluminum alloys for SSM processing. Again, the intermetallic compounds were shown to exhibit a cathodic behavior as compared to the α-phase [120].

#### **5. Summary**

In the present review paper, the microstructural characteristics of various semi-solid Al alloys are thoroughly summarized, together with the description of the evolution of their typical nondendritic microstructure during solidification. Furthermore, the influence of microstructural features on mechanical properties is systematically analyzed. This is fundamental in order to understand the different performance of SSM parts in comparison with components obtained by conventional production routes. Apart from tensile properties, other important characteristics are also discussed in order to provide a complete overview of the performance of semi-solid Al alloys, such as fatigue behavior, wear, and corrosion resistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **A Comparison between Anodizing and EBSD Techniques for Primary Particle Size Measurement**

**Shahrooz Nafisi 1,2,3,\*, Anthony Roccisano <sup>1</sup> , Reza Ghomashchi <sup>1</sup> and George Vander Voort <sup>4</sup>**


Received: 26 March 2019; Accepted: 24 April 2019; Published: 27 April 2019

**Abstract:** Proper understanding and knowledge of primary particle or grain size is of paramount importance in manufacturing processes as it directly affects various properties including mechanical behavior. Application of optical microscopy coupled with etching techniques has been used conventionally and in conjunction with color metallography (polarized microscopy) has been the preferred method for grain size measurement. An advanced technique as an alternative to light microscopy is using electron backscatter diffraction (EBSD). A comparison is made between these two techniques using Al-7Si alloy produced with various casting techniques to highlight the cost and time of the sample preparation and analysis for both techniques. Results showed that color metallography is certainly a faster technique with great accuracy and a much cheaper alternative in comparison with EBSD.

**Keywords:** polarized light microscopy; anodic etching; EBSD; grain; globule; Al-Si alloy; semi-solid metal processing; EMS; thixocasting

#### **1. Introduction**

Grain size plays an important role on the properties of metallic materials. Normally, the structure is observed on the plane of the polished surface using optical microscopy which is a two-dimensional (2D) measurement. The 2D analysis is not a complete representative of the 3D structure and in some instances may lead to a biased conclusion. However, it is the current practice for microstructural characterization and therefore there is an urge to ensure the procedure is as effective and accurate as possible. An example is shown in Figure 1 highlighting the deficiency of 2D analysis. The secondary dendritic branches are treated as individual and isolated particles during automatic image analysis processing [1].

**Figure 1.** Branches of dendrites, A356 alloy, quenched from 598 ◦C [1]: (**a**) bright field, and (**b**) polarized light image with sensitive tint plate, polished and anodized with Barker's reagent.

Various techniques have been used to overcome this inadequacy including advanced serial sectioning [2,3], X-ray tomography [4,5], and analysis of crystallographic orientations of the particles by electron backscatter diffraction (EBSD) technique (either 2D or 3D) [6,7]. The competing issues in using any of the aforementioned methods are the required equipment, the operator's skill and the analysis time which are eventually interpreted in terms of the cost of these methods. This is the theme of the current report; the accuracy, the cost (equipment), and time of two widely used methods of "anodic etching-polarized light microscopy" and EBSD in the analysis of semi solid metal (SSM) processed Al-Si alloy structure. The distinction between the individual grains and particles is important for SSM processes as it not only verifies the effectiveness of SSM process but also greatly affects the finished SSM parts mechanical properties.

Anodic oxidation, or anodizing, is an electrolytic process for depositing an oxide film on the metal surface epitaxial to the underlying grain structure. It is similar in nature to heat tinting or tint etching where an interference film is produced on the surface of metals. In this method, the sample is placed in the anodizing solution as the anode connected to a stainless steel or graphite plate or bar acting as the cathode. The resulting interference film colors when viewed under a polarizing light are a function of the anodic film thickness. The thickness, however, depends on the anodizing voltage, the anodizing solution composition, and the composition and/or structures of the phases present in the specimens and anodizing time. There are plenty of reports in open literature noting that certain etchants for Body Centered Cubic (BCC) and Face Centered Cubic (FCC) metals produce an interference film on the grain which results in grain contrast effect in bright field light microscopy. When viewed in polarized light, it yields color images that can be further enhanced by adding a sensitive tint plate [8,9].

EBSD is a technique that provides crystallographic information by analyzing crystalline samples in the scanning electron microscope (SEM). In EBSD, a stationary electron beam strikes a 70 degrees tilted sample and the diffracted electrons form a pattern on a fluorescent screen. The diffraction pattern is unique to the crystal structure and crystal orientation of the sample region from which it was generated. Diffraction patterns are used to measure the crystal orientation, grain boundary misorientations, discriminate between different materials, and to provide useful information about local crystalline perfection. By scanning in a grid across a polycrystalline sample and measuring the crystal orientation at each point, the resulting map will reveal the constituent grain morphology, orientations, and boundaries. In addition, the data shows the preferred crystal orientations (texture) present within the material [10,11].

Color metallography techniques have been developed for many metals and alloys, but they have not been utilized widely by metallographers, despite their obvious benefits. Color etchants are usually phase-specific and they will fully reveal the grain structure or specific second-phase constituents. Some, like Barker's anodizing solution, produce results that can only be observed using cross polarized light. Coloration in all cases, if weak, can be enhance by adding a sensitive tint filter (sometimes called a lambda plate). The coloration is due to variations in the crystallographic orientation of the grains or particles. Black and white etchants cannot reveal such differences. EBSD is commonly used to reveal crystallographic orientations between grains and phases, but the process requires an SEM equipped with an EBSD system. Coarser step sizes between diffraction patterns reduce the precision of the method, although it shortens the scan time. Specimen preparation for EBSD is much more challenging for most laboratories, and more time consuming. All preparation-induced damage must be removed to get high-quality EBSD grain maps, and the specimen surfaces must remain perfectly flat [12,13]. So, the use of color metallography by anodizing Al and its alloys, followed by examination using polarized light and a sensitive tint filter, is much faster, easier, and less expensive than EBSD.

This article attempts to highlight the effectiveness of employing polarized light microscopy in conjunction with anodizing, in revealing the individual grains or primary particles in SSM processed Al-Si alloy. The distribution of grains and particles are then compared with the grain size obtained from EBSD technique to emphasize the simplicity and low cost of color metallography. As a result of such practice, it was managed to identify the critical degrees of misorientation as the required criterion to differentiate between grains and subgrains in SSM processed alloys when using EBSD, which in its own right is an important finding by itself.

#### **2. Materials and Methods**

Binary Al-7% Si alloys (6.7–6.9% Si, 0.8–0.81% Fe) were prepared in an electric resistance furnace. Two different molds were used; for higher cooling rate, a copper mold with a water-cooled jacket was used and for the lower cooling rate, a CO2 bonded silica-sand mold was employed. Ingots were 76-mm in diameter and 300-mm long. The entire configuration was placed in an electromagnetic stirring machine (EMS). The frequency was set to 50 Hz and the current was 100 and 30 A for copper and sand molds, respectively (the application of magnetic field was stopped when the alloy temperature in the molds reached 400 ◦C on cooling). The application of different currents was intended to ensure that the applied magnetic force to stir the molten alloy is almost the same for both the sand and copper molds. Experimental details are explained in further detail elsewhere [14]. Pouring temperature was changed between 630 and 690 ◦C. The cooling rate in the copper and sand molds for the conventional ingot (with no stirring) was about 4.8 and 3.3 ◦C/s, respectively (the cooling rates were calculated in the liquid state above the liquidus temperature).

For thixocast trials (reheating to semi-solid region), samples were cut from the transverse sections (200 mm from the bottom of EMS billets), in areas between the billet center and wall, and were reheated in a single coil 5 kW induction furnace operating at 80 kHz. Samples were placed vertically on an insulator plate. Temperature variation during the tests was monitored by attaching thermocouples to both the billet center and the wall. The induction furnace was controlled by the central thermocouple and the wall thermocouple was used to establish if there is any transverse temperature gradient within the billet. The reheating cycle included 2–3 min of heating up to 583 ± 3 ◦C and 10 min holding time at this temperature before water quenching (at the selected holding temperature of 583 ± 3 ◦C, there is about 38–40% fraction of solid alloy according to ThermoCalc calculations). For EMS cast samples, the metallographic specimens were cut transversely at 200-mm from the bottom of the billets, mounted in Bakelite, ground, and polished conventionally down to 0.05 μm colloidal silica to develop high quality EBSD (Struers, Willich, Germany) maps. Thixocast samples were prepared simply by cutting a transverse section of the quenched samples.

EBSD analysis was undertaken with an FEI Inspect F50 Field Emission Gun (FEG) Scanning Electron Microscope (SEM) with an EDAX Hikari detector. In each specimen at least four regions were located using a microhardness indent and were scanned through EBSD. At least four regions per specimen were scanned at 200X magnification, with a scan frame of 800 <sup>×</sup> 800 <sup>μ</sup>m2 and a step size of 1.5 μm taking approximately 45 min to scan. In addition, each specimen had a region scanned at 100× magnification with a scan frame of 1600 <sup>×</sup> 1600 <sup>μ</sup>m2 and a step size of 2 <sup>μ</sup>m2 taking approximately 90 min to scan. After the EBSD analysis, the samples were anodized. Samples were immersed in a solution of Barkers reagent (1.8% fluoroboric acid in water) and anodized until a color shift was observed under polarized light microscopy (20 V direct current, 4 min). Polarized light microscopy was undertaken on a Zeiss AXIO Imager.M2m (Carl Zeiss Microscopy, Jena, Germany) equipped with reflected light polarizer and, rotatable analyzer with lambda plate. Regions that were scanned with EBSD were subsequently viewed under polarized light microscopy and the grain boundary properties determined under each were compared.

The acquisition time for color metallography and EBSD is provided in Table 1 indicating the approximate time spent acquiring grain boundary data. Sample preparation time is typically 30 min longer for EBSD samples due to the increased colloidal silica polishing time required to produce strong patterns and the stricter cleaning regime required to produce good scans. The EBSD acquisition time is heavily dependent upon the available equipment and the resolution of the scan with the newer detectors offering shorter scan times for the same resolution, however, there are extended waiting periods in all models from pumping and venting the chamber as well as locating the scan area. In addition to the above time allocation to acquire the desired outcome, it is important to point out that the level of skill to run the EBSD is far more demanding than the one for anodizing and optical microscopy. On top of that, the capital investment is nowhere comparable.

**Table 1.** Time consumption acquiring grain boundary data through color metallography and electron backscatter diffraction (EBSD).


#### **3. Results and Discussions**

Generally, aluminum hypoeutectic alloys consist of two main constituents, primary α-Al particles and eutectic mixture of α-Al and Si (in this specific alloy, due to the extra addition of iron, some β-iron intermetallics were formed). Effects of different cooling rates on the silicon and iron flake sizes and distribution in the as-cast EMS samples are shown in Figure 2. Lower cooling rate in the sand mold shows larger dendrites comparing to that of the copper mold specimens. Dendrite size had a great difference since the cooling rate is higher in the copper mold (Figure 2c,d). At higher magnifications, a great difference could be seen in the size of the silicon eutectic flakes and β–iron intermetallics. By pouring the alloy in the copper mold, i.e., with a higher cooling rate, silicon and β–iron intermetallics got thinner and shorter (compare Figure 2b with Figure 2d).

**Figure 2.** Optical micrographs showing the effect of cooling rate on the microstructure of as-cast magnetic stirring machine (EMS) billets: (**a**,**b**) sand mold and (**c**,**d**) copper mold, pouring temperature 690 ◦C, etched with 0.5% Hydrofluoric acid (HF) (arrows show some of the iron-intermetallics). Scale bar is 400 μm for (**a**) and (**c**), 50 μm for (**b**) and (**d**).

During isothermal holding, the eutectic mixture was re-melted while the primary α-Al phase was coarsened. There was also a driving force towards reduction of interfacial area which results in globularization of the primary α-Al particles (Figure 3).

In order to discuss the correlation between EBSD and anodizing, it is better to examine the thixocast structure first which, due to the better clarity of individual particles, makes comparison much easier and convenient than for the EMS which the particles sometimes are intertwined.

Figure 4 shows an example of color metallography and EBSD of the same area of the sample cast in copper mold at 630 ◦C and isothermally held at ~583 ◦C for 10 min. As discussed earlier, by isothermal holding, primary α–Al particles became spherical and some of the primary particles possibly sintered together. The colors in the polarized light micrograph, Figure 4a, were indications of different orientations of the primary particles while a similar color indicates the same nucleation and growth pattern. In this specific example, particles K–L, J–M, M–N, O–P, T–X, X–Y, and Z–O possibly sintered, i.e., two or more isolated solid particles joined and formed a pseudo cluster that is associated mainly to the stabilized contact during post heat treatment processing [1]. In the semi-solid science, this phenomenon is defined as "agglomeration" [15–17] where primary particles come into contact and possibly sinter together to form agglomerates.

**Figure 3.** Optical micrographs showing the effect of isothermal holding on the microstructure of thixocast samples: (**a**,**b**) sand mold and (**c**,**d**) copper mold, pouring temperature 690 ◦C, etched with 0.5% HF (arrows show some of the iron-intermetallics). Scale bar is 400 μm for (**a**,**c**), 50 μm for (**b**,**d**).

**Figure 4.** Thixocast sample, copper mold, poured at 630 ◦C; (**a**) polarized light image with sensitive tint plate, polished and anodized with Barker's reagent, (**b**) inverse pole figure (IPF) map, (**c**) 5◦ grain map, (**d**) 10◦ grain map, (**e**) 15◦ grain map.

Table 2 presents information such as misorientation between particles, and whether they are detectable as individual grains when the grain boundary misorientation criteria were set at five, 10, and 15 degrees. For example, particles K and L had a misorientation of 55.6 degree specifying that these two are indeed two separate grains. Changing the critical misorientation angle from five to 10 and then 15 (Figure 4c–e) did not change the grain map and proved that these two particles are two individual grains and polarized image confirmed the fact. The investigated areas in all samples were analyzed using the same methodology.


**Table 2.** EBSD data for thixocast sample, copper mold, poured at 630 ◦C (Y, Yes; N, No).

Figure 5 presents another example of color metallography and EBSD of the sample cast in copper mold at 690 ◦C and isothermally held at ~583 ◦C for 10 min. In comparison to Figure 4a, the α-Al particles in Figure 5a are larger and present fewer colors, meaning that fewer individual particles of varying crystal orientation formed within this process. This is due to the fact that the higher pouring temperature resulted in fewer nucleation sites and more dendritic structure. This subject was explained in another publication [14].

Some of the particles in Figure 5 belong to one primary dendrite as they show the same color contrast. This is regardless of changing the misorientation angle criteria. An example is particles 1, 2, 3, and 4. By increasing the grain misorientation criterion from 5 to 10 or even 15-degrees, particles 1-4 still have the same color contrast. However, by increasing the misorientation to 10 degrees, particles E and E1 showed similar color to particles 1–4 (Figure 5d,e). This indicates similar origin and disintegration mechanism for all of the adjacent particles of similar color contrast. The same rationale is applicable for particles C1–C3 meaning that they are originated from the same dendrite. C1–C3 appear as three distinct particles when the misorientation is five degrees. This may indicate these sub grain particles are separated from each other by five degree or less. When the misorientation criterion increases to 10 or 15 degrees, they show the same color which means they were originated from the same dendrite (considering that the sample was electromagnetically stirred and dendrites cramped and/or broken during this process, however they all have their origin in one dendrite). In other words, what it is being said here is the fact that although the particles are detected as individual particles with clear grain boundaries, they may be connected from underneath the plane of polish or broken down by stirring action during semi-solid processing. Nonetheless, the most important finding here is the similarity of polarized light microscopy of anodized specimen with those of EBSD characterization and grain mapping, particularly when a critical misorientation angle of 10 degrees is implemented. Table 3 shows additional EBSD information on some specific particles and whether they are detectable as individual grains in color metallography.

(**a**) (**b**)

**Figure 5.** Thixocast sample, copper mold, poured at 690 ◦C; (**a**) polarized light image with sensitive tint plate, polished and anodized with Barker's reagent, (**b**) IPF map, (**c**) 5◦ grain map, (**d**) 10◦ grain map, (**e**) 15◦ grain map.


**Table 3.** EBSD data for thixocast sample, copper mold, poured at 690 ◦C (Y, Yes; N, No).

Representatives of as-cast EMS sand billets are shown in Figures 6 and 7. The application of electromagnetic stirring results in forced convection of the bulk liquid which encourages dendrites fragmentation as well as dendrite arm root re-melting due to the thermal and solutal convection [1]. It is also important to consider that in contrast to the conventional casting, electromagnetic stirring of the melts with higher superheat results in more re-melting of the nuclei [18].

**Figure 6.** As-cast EMS billets, sand mold, poured at 630 ◦C; (**a**) polarized light image with sensitive tint plate, polished and anodized with Barker's reagent, (**b**) IPF map, (**c**) 5◦ grain map, (**d**) 10◦ grain map, (**e**) 15◦ grain map.

(**c**) (**d**) (**e**)

**Figure 7.** As-cast EMS billets, sand mold, poured at 690 ◦C; (**a**) polarized light image with sensitive tint plate, polished and anodized with Barker's reagent, (**b**) IPF map, (**c**) 5◦ grain map, (**d**) 10◦ grain map, (**e**) 15◦ grain map.

Quality of the colored images depends not only to the anodizing technique (e.g., solution, voltage, time, temperature which all were constant for these samples), but also to the alloying elements and manufacturing process. By isothermally holding the samples, the image quality and color differentiation improves (compare Figures 4–7).

As a result, detection of the individual grains in color metallography is a challenging task. For instance, in Figure 6, particles 3 and 4 are two different grains according to EBSD grain maps; however, polarized light microscopy is unable to detect the differences between the two grains (Figure 6a). By increasing the critical misorientation angle to 10 degrees, some particles such as 4–15, 15–16, 18–28, and 16–25 present similar color in EBSD grain mapping. Table 4 provides additional information on the area presented in Figure 6.


**Table 4.** EBSD data for as-cast EMS billets, sand mold, poured at 630 ◦C (Y, Yes; N, No).

The same is applicable in Figure 7 for as-cast EMS billets, sand mold, poured at 690 ◦C. Particles A and E, A and D, K and Q are different grains according to EBSD; however, polarized light microscopy is unable to detect the grains. By increasing the critical grain misorientation to 10 degrees, particles B–F, D–E, I–M, and M–L exhibit similar color (see Table 5), once more confirming that a 10-degree misorientation is an acceptable grain detection limit for EBSD technique (see Figure 6).


**Table 5.** EBSD data for as-cast magnetic stirring machine (EMS) billets, sand mold, poured at 690 ◦C (Y, Yes; N, No).

Various areas were analyzed in these four samples and Figures 8 and 9 presents the comparison graphs. In these graphs, misorientations were categorized in different bins; each bin spans over 5-degree misorientation. To develop the graphs in Figures 8a and 9a, the total number of boundaries identified through each technique were counted and separated into their respective misorientation intervals. Figures 8b and 9b represent the number fraction of the boundaries in each interval and was determined by dividing the number of boundaries in each interval found in Figures 8a and 9a by the total number of boundaries identified. For the thixocast samples, the number and fraction of grain boundaries using both color metallography and EBSD follow each other closely and have the same trend. In general, the EBSD technique detects more boundaries than color metallography. This is evident in the graph for the number of identified boundaries (Figures 8a and 9a). However, the fraction graphs support the idea that we could definitely rely on the results of the color metallography as the trend is similar.

**Figure 8.** Number of identified boundaries (**a**) and fraction of boundaries (**b**) based on the misorientation angles, thixocast sample.

**Figure 9.** Number of identified boundaries (**a**) and fraction of boundaries (**b**) based on the misorientation angles, as-cast EMS billets, sand mold billets.

In the case of as-cast EMS sand billets, the number and fraction of grain boundaries in both color metallography and EBSD follow each other closely and have the same trend. However, comparing Figures 8 and 9, the difference between color metallography and EBSD results in as-cast EMS samples are more than the thixocast samples. This is due to the color distinction in these samples. EBSD is more accurate for samples having complications during color metallography and EBSD detects far more boundaries than were detected using polarized light and sensitive tint after anodizing. Below five degrees, color microscopy was unable to detect the difference between any boundaries as the difference in color was not significant enough to detect. Whilst EBSD did not detect any boundaries below five degrees either, this was due to the critical misorientation angle being set to five degrees and not an inherent limitation of EBSD.

When reviewing the results of grain boundary identification through color metallography and EBSD, whilst the end results were comparable, the time spent acquiring the data was significantly different. As explored in Table 1, the total time to analyze a single region with color metallography was 100 min faster than through EBSD analysis. The higher testing time associated with EBSD can be linked to the longer sample preparation time required to develop high quality scans as well as the upwards of 75 min required to produce a scan. Samples can be anodized and then analyzed through optical microscopy in as little as 20 min hence, for the time it takes to scan and analyze a single EBSD sample, four samples could be anodized and analyzed through color metallography. When comparing the cost of carrying out each technique, polishing costs are higher for EBSD than for color metallography due to the longer polishing times. Whilst this higher polishing cost is somewhat offset by the requirement of an anodizing solution, the capital expenditure and booking fees are significantly higher for an SEM than an optical microscope.

#### **4. Conclusions**

In order to study the capability of color metallography as a reliable microstructural characterization technique in distinguishing the primary particles, anodizing-polarized light microscopy was compared with the EBSD technique for Al–Si samples produced with different casting procedures;


**Author Contributions:** Conceptualization, S.N.; methodology, S.N., validation, S.N. and G.V.V.; formal analysis, S.N. and A.R.; investigation, A.R.; resources, S.N. and R.G.; data curation, A.R.; writing—original draft preparation, S.N.; writing—review and editing, S.N., A.R., R.G. and G.V.V.; visualization, S.N. and A.R.; supervision, S.N.; project administration, S.N.; funding acquisition, R.G.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors acknowledge the facilities, and the scientific and technical assistance, of the Australian Microscopy and Microanalysis Research Facility at Flinders University. This work was performed in part at the OptoFab node of the Australian National Fabrication Facility utilizing Commonwealth and SA State Government funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
