Applied Physics A

, Volume 115, Issue 2, pp 651–660

Aluminum integral foams with tailored density profile by adapted blowing agents


    • Department of Materials Science and EngineeringUniversity of Erlangen-Nuremberg
  • Tobias Fiegl
    • Department of Materials Science and EngineeringUniversity of Erlangen-Nuremberg
  • Carolin Körner
    • Department of Materials Science and EngineeringUniversity of Erlangen-Nuremberg

DOI: 10.1007/s00339-014-8377-4

Cite this article as:
Hartmann, J., Fiegl, T. & Körner, C. Appl. Phys. A (2014) 115: 651. doi:10.1007/s00339-014-8377-4


The goal of the present work is the variation of the structure of aluminum integral foams regarding the thickness of the integral solid skin as well as the density profile. A modified die casting process, namely integral foam molding, is used in which an aluminum melt and blowing agent particles (magnesium hydride MgH2) are injected in a permanent steel mold. The high solidification rates at the cooled walls of the mold lead to the formation of a solid skin. In the inner region, hydrogen is released by thermal decomposition of MgH2 particles. Thus, the pore formation takes place parallel to the continuing solidification of the melt. The thickness of the solid skin and the density profile of the core strongly depend on the interplay between solidification velocity and kinetics of hydrogen release. By varying the melt and blowing agent properties, the structure of integral foams can be systematically changed to meet the requirements of the desired field of application of the produced component.

1 Introduction

Aluminum integral foams are monolithic structures possessing a foamed cellular core, a compact, dense skin and a smooth transition region in between. At a constant relative density level, material distribution over the cross section of the produced component can differ. This can be due to a variation of the skin thickness, to a differing transition region or to a combination of both. This variation of the integral foam structure represents a decisive factor for the adaption of such components to the desired application such as flexural load, crash absorption or damping [1].

Thus, in order to improve the mechanical properties for employment as load bearing structures, an increase in skin thickness is favorable [2, 3]. For crash absorption, the compressive strength is an important factor which is at least partly dependent on the distribution of porosity within the foam. In this case, the density distribution should show a pronounced plateau. To benefit from the excellent damping properties of aluminum foams, the compact layer has to be as thin as possible since only porous material regions contribute strongly to damping [4].

There are two different existing process modifications in integral foam molding: the high pressure (HP-IFM) and the low pressure (LP-IFM) variant [5]. The process steps are similar to those of the foam injection molding of polymers on the basis of which the steps are developed. A detailed description of the HP-IFM can be found in [6], the LP-IFM is discussed in detail in [7] (see also Fig. 3). In both cases, a certain amount of blowing agent, typically magnesium hydride (MgH2), is entrained into the cavity by the aluminum melt during mold filling at high velocity by a moving piston, where it decomposes thermally activated into magnesium (Mg) and hydrogen (H2). The released gas leads to the pore formation. In case of the low pressure process, mold filling and pore formation take place simultaneously as the cavity is underfilled with melt. At the cold mold wall, the melt solidifies to a compact skin due to a very high solidification rate, whereas the inner core region is foamed. In the case of HP-IFM, mold filling and foam formation are separate process steps. As the mold filling is completed, the volume of the cavity is increased by a core puller system after a certain delay time which initiates pore formation in regions where the alloy has not yet solidified. In both cases, there are two competing physical processes which by interaction define the integral foam structure: the decomposition of the blowing agent which is very sensitive to the local melt temperature [5, 811] and the velocity of the moving solidification front. In case of parts produced by HP-IFM, this interplay results in U-shaped density profiles whereas parts produced by LP-IFM show a more V-shaped profile.

There are three possibilities to increase the thickness of the solid skin: increasing the delay time between piston stop and core puller movement [6], reducing the melt temperature or lowering the cooling temperature of the die [2, 4]. But neither of those alterations of the process is able to change the basic distribution of porosity in the core region or to decrease skin thickness, which would be desirable for damping applications. Furthermore, changing the thermal conditions toward faster solidification not only leads to a thicker compact skin but also in most cases to a decrease in foam structure quality [4, 6] or even delaminations of the skin layer [6]. Figure 1 shows cross sections of LP-IFM parts where the thickness of the compact skin is increased either by reducing the melt temperature (θmelt) or by decreasing the cooling channel temperature (θcc) of the die. A lower θmelt is reached by simply delaying the moment the piston movement starts after dosing the melt into the shot chamber, from immediate start to a 2 s delay. The skin thickness is clearly increased as already shown by Trepper [4] but at the expense of surface and pore structure quality. By reducing the heat content of the melt, not only the decomposition kinetics of the blowing agent is lowered but also the flowability of the melt so that pronounced cold flow occurs. Furthermore, the higher solid phase fraction in the moment of pore formation leads to disrupted pore structures.
Fig. 1

Standard process options for increasing the compact skin thickness of aluminum integral foams. Left lowering the temperature of the injected melt (θmelt). Right reducing the cooling channel temperature (θcc) of the die. Both options lead to cold flow and poor pore structure quality

Because of these disadvantages as well as the fact that the thickness is only variable within narrow limits as already shown by Wiehler [2], other approaches are necessary. That is why the goal of the current paper is to present on the one hand a way to increase skin thickness without any negative influence on surface quality and pore structure. And on the other hand, we also show the way how the skin thickness can be reduced. This as well as a complete modification of the porosity profiles of the foams is implemented by manipulation of the decomposition kinetics of the MgH2 powder. Furthermore, two different aluminum alloys are used, showing a strong difference in the solidification temperature influencing the decomposition kinetics of the entrained blowing agent [5].

2 Experimental

2.1 Blowing agent modification: milling and oxidizing

The commercial blowing agent powder MgH2 (Tego Magnan, Evonik Goldschmidt) is treated in different ways. A powder fraction with a d50-value of 55 μm (d10 = 30 μm, d90 = 94 μm) is milled in batches of 40 g in a centrifugal ball mill (S1, Retsch; grinding balls to powder weight ratio of 10:1) for 1, 2, 3, 5 and 10 min at 370 rpm.

Powder as supplied (d10 = 25 μm, d50 = 50 μm, d90 = 92 μm) as well as milled powder (3 min at 320 rpm; d10 = 2 μm, d50 = 19 μm, d90 = 47 μm) is oxidized in a furnace at 270 °C and at 40 °C in a drying oven in a humid atmosphere (so-called “aging” [12]) for up to 23 h. A similar powder modification was already performed by Matijasevic-Lux et al. [13, 14] for titanium hydride (TiH2) to retard hydrogen release during the heating-up process and to improve the foaming properties of powder metallurgical aluminum foams. In our case, 270 °C is chosen to accelerate the oxidation process without reaching the theoretical decomposition temperature of MgH2 of about 280 °C. In addition, for the aging process under humidity, a much lower temperature is applied, but still slightly above room temperature to ensure a certain evaporation.

2.2 Analysis of particle size and powder reactivity

To quantify the properties of the so fabricated powders, particle size and decomposition temperature (θdec) are measured. The particle size distributions of the different powders are measured by laser diffractometry (Mastersizer 2000, Malvern Instruments), and pre-dispersed by a sonotrode in isopropanol. The resulting particle sizes (Fraunhofer model) are volume weighted and correspond to the diameter of spheres with equivalent cross-sectional areas as the real non-spherical particles. The mean particle size d50 is used as a representative value for the respective powder.

The decomposition properties of the prepared powders are measured by thermogravimetrical analysis (STA 409 CD, Netzsch). An amount of m0 = 25 mg is, therefore, weighed out into an alumina (Al2O3) crucible and heated under argon flow (6 l/h) at a heating rate of 25 K/min above θdec where the mass loss due to hydrogen release is detected. The high heating rate is used to see clearly the influence of powder modification (milling and oxidizing) on the kinetics of hydrogen release. In that way, a signal of mass m/m0 (starting at 100 %) as a function of the temperature is recorded as shown in Fig. 2 exemplarily for three different powders. The so-called onset temperature (equivalent to θdec) is determined to quantify the beginning of the decomposition. θdec is defined as the intersection of the base line at 100 % mass and the tangent through the point of inflection of the descending decomposition section (Fig. 2, gray lines and dots).
Fig. 2

Thermogravimetrical analysis of different blowing agent modifications at a constant heating rate of 25 K/min. The mass loss due to hydrogen desorption is given as a function of the temperature. The different decomposition temperatures (θdec) are determined by intersection of the base line and the tangent through the point of inflection

2.3 Characterization of the alloys by solidification simulation

The two different aluminum die casting alloys Al9Si3Cu(Fe) (226D) and Al5Mg2SiMn (Magsimal-59, Rheinfelden) are used to analyze the influence of the melting range on the integral foam structure of LP-IFM parts. Solidification simulations (Thermo-Calc, Access) according to Scheil–Gulliver are performed to gain information about the temperature as a function of solid phase fraction during cooling. Especially the temperature range around the liquidus temperature (θliq) is of interest as it represents the decisive value for the temperature of the melt surrounding the blowing agent particles during entrainment.

In addition, computational fluid dynamics simulation (CFD; Flow-3D, Flow Science) of particle entrainment and cavity filling is carried out to obtain further information on the temperature fields the blowing agent particles are exposed to. Altogether, this allows an estimation of the decomposition kinetics which helps to explain integral foam structures.

2.4 Foam fabrication by LP-IFM

A cold-chamber pressure die casting machine (DAK450-54, Frech) is used to produce the parts by LP-IFM. An amount of approximately 1,450 g of aluminum melt (Al9Si3Cu(Fe), Al5Mg2SiMn; 740 °C) is dosed into the shot chamber resulting in a part weight of approximately 600 g. As a complete filling of the cavity would correspond to a part weight of 1,000 g, the produced parts have a global porosity of 40 %. The part has side lengths of 178 mm and thicknesses of 10, 12 and 14 mm in three regions (Fig. 3, bottom right).
Fig. 3

LP-IFM process steps visualized by simulation (Flow-3D, Flow Science). t1 blowing agent powder placed in the runner system. t2 capture of the powder by flowing melt. t3 turbulent entrainment of powder into the melt. t4 solidification of melt at the surface and mixing of blowing agent in the inner region of the component. t5 homogeneous distribution of particles by melt flow and pore formation due to blowing agent decomposition (not implemented into simulation). Bottomright geometry of the produced part

Powders with five different decomposition temperatures are used (Table 1) and placed in the runner system before closing the mold (Fig. 3, t1). The standard powder as supplied and the milled powder are used for both alloys so that the impact of the θliq on the component structure can be investigated.
Table 1

Overview of the powders (particle size, θdec and oxidation treatment) and alloys used for the fabrication of integral foam parts


Powder size

θdec (°C)




Time (h:min)


As supplied


Al9Si3Cu(Fe), Al5Mg2SiMn




Al9Si3Cu(Fe), Al5Mg2SiMn


As supplied


dry, 270 °C






dry, 270 °C




As supplied


wet, 40 °C






dry, 270 °C



As the piston advances, the melt entrains the powder in a turbulent way into the mold where solidification and gas release take place simultaneously (Fig. 3, t2t5). The melt wets the surface where it solidifies instantaneously whereby the resulting core density profile depends on the heat flow and decomposition kinetics in the inner region.

2.5 Foam characterization

The characterization of the produced parts comprises the preparation of cross sections from the 12 mm height sector to get an impression of the foam structure quality as well as of the skin thickness.

In addition, cylindrical samples with a diameter of 13 mm are prepared from the same sector (see Fig. 3, bottom right) for micro-computed tomography (μCT 40, Scanco Medical). With an X-ray acceleration voltage set to 60 kV, an initial current of 133 μA and an integration time of 300 ms, the samples are analyzed three dimensionally with a voxel resolution of 15 μm3. In that way, the relative density is recorded as a function of the distance to the surface of the samples which allows gaining information about the porosity distribution within the sample volume.

3 Results and discussion

3.1 Modification of the blowing agent reactivity

Figure 4 depicts exemplarily the decrease in particle size with increasing milling time. The initial d50 value of 55 μm decreases exponentially and reaches 8 μm after 10 min. θdec (heating rate 25 K/min) increases slightly after 1 min milling time before it shows a similar behavior as the mean particle size. The temperature plateau at short milling time may be explained by the fact that the powder is heated up due to friction so that oxidation of the powder is facilitated lowering the decomposition kinetics. Simultaneously, the natural oxide layer (MgO) of the powder particles due to the fabrication process (so-called phlegmatic MgH2, Evonik Goldschmidt) retarding decomposition [10] is broken up.
Fig. 4

Left decrease in mean particle size (d50) as a function of the milling time (370 rpm). θdec (25 K/min) is lowered in the same way except for a slight increase for very short milling times. Right linear correlation of θdec and particle size

Hydrogen release takes place either by diffusion through the oxide layer and recombination on the surface [11, 1518] or by breaking up the layer due to the formation of inner hydrogen gas pressure within the particle [19, 20] which occurs especially at very high heating rates [18]. Thus, cracking the oxide layer accelerates desorption [16]. In addition, the reduction of particle size increases the total particle surface and accelerates the hydrogen release by shorter diffusion paths [21, 22]. Not only the particle size changes during milling but also the grain size is reduced which also correlates with faster desorption [19]. Finally, also incorporated impurities by abrasion processes during milling like iron (Fe) or iron oxide (Fe3O4) can act as catalysts for hydrogen desorption [23, 24]. On the whole, the combination of microstructure change, catalyst incorporation [21], particle size reduction [25] and breaking up of the hindering oxide layer [19] during milling leads to faster decomposition kinetics. There is a linear relation between d50 and θdec (Fig. 4, right). Thus, not only the desired particle size can be adjusted by milling but also the reactivity of the powder changes in a defined way.

To investigate the possibilities of retarding decomposition of MgH2 powders, two aging environments (dry at 270 °C and wet at 40 °C) as well as two powders (as supplied and milled) are studied. Figure 5 gives an overview of the shift of onset temperature of decomposition as a function of oxidation time. Although starting from a lower θdec, the milled powder “ages” faster than the powder as supplied in both atmospheres. Varin et al. [12] already showed that milled powders are prone to aging in a humid atmosphere. This can be due to numerous reasons: first of all as the oxide passivation layers are broken up during the milling process, the milled particles only exhibit a thin oxide layer [26] making new oxidation easier as the diffusion paths are shorter. But even at equivalent θdec of milled and as supplied powder (e. g. after 1.5 h of oxidation at 270 °C), the increase in θdec is faster, probably due to the fact that the particle surface is significantly larger for milled powders. The velocity of the θdec—increase slows down with increasing oxide layer thickness resulting in a root-like curve progression.
Fig. 5

Aging of MgH2. Left oxidation of powder as supplied and milled powder (3 min at 320 rpm; d50 = 19 μm, θdec = 422 °C) at 270 °C. Right oxidation of powder as supplied and milled powder at 40 °C under humid atmosphere. Milling and humid atmosphere accelerate the aging process

Aging under wet atmosphere leads to a much faster shift of θdec to higher values, especially for milled particles. The humidity causes a hydrolysis reaction [27]:
$$ {\text{MgH}}_{ 2} + {\text{ 2 H}}_{ 2} {\text{O }} \to {\text{ Mg}}\left( {\text{OH}} \right)_{ 2} + {\text{ 2 H}}_{ 2} $$

Magnesium hydroxide [Mg(OH)2] forms on the surface [12] and the hydrogen content decreases. This Mg(OH)2 layer starts to decompose to MgO and H2O above 350 °C [28, 29]. Thus, it probably acts in the same way as the oxide layer but grows faster explaining the accelerated aging compared to dry oxidation.

These aging curves are used as master curves which help to exactly time the duration of oxidation of blowing agents to obtain powders with a defined decomposition temperature.

3.2 Solidification properties of the alloys Al9Si3Cu(Fe) and Al5Mg2SiMn

Figure 6 shows the results of Thermo-Calc calculations of the two alloys which are consistent with results of Otarawanna et al. [30] who analyzed Al5Mg2SiMn as well as of Fragner et al. [31] and Bünck et al. [32] who studied Al9Si3Cu(Fe). The calculation shows that there is a temperature difference of about 40 K between the Al9Si3Cu(Fe)-alloy and the Al5Mg2SiMn-alloy up to a solid phase fraction of about 50 % which represents the relevant range for blowing agent decomposition and pore growth [2]. That is, the temperature during mold filling and foam formation is expected to be about 40 K higher for Al5Mg2SiMn compared to Al9Si3Cu(Fe).
Fig. 6

Melt temperature of the two aluminum alloys as a function of solid fraction for a solidification according to Scheil–Gulliver (Thermo-Calc)

3.3 Modification of compact skin thickness and relative density profiles

Figure 7 depicts cross sections of integral foams produced with different reactive blowing agent powders and different alloys, with the process parameters remaining constant. Using reactive milled powders, a clear decrease in compact skin thickness can be observed due to the faster decomposition kinetics. Pores are also formed near the surface resulting in a clearly decreased skin thickness. This is further pronounced using the Al5Mg2SiMn-alloy with the higher θliq. On the other hand, aged powders (as supplied) lead to an increase in the compact skin thickness with increasing θdec. This behavior is independent of aging condition—dry (445 °C) or wet (456 °C). The delayed hydrogen release prolongs the time delay before pore formation in the melt takes place. It is important to state that no adverse influence, neither on the surface quality (Fig. 7, bottom, for θdec = 456 °C) of the samples nor on the pore structure, can be detected.
Fig. 7

Modification of skin thickness of integral foams (indicated by white arrows). The white frames indicate the position of detailed views. Lowering θdec (III → II) as well as an increase in θliq (III → I) leads to a clear decrease in skin thickness. Augmenting θdec (III → IV/V) leads to a stepwise increase in skin thickness. Pore structure and surface quality (bottomleft for θdec = 456 °C) are not affected in a negative way

Figure 8 shows characteristic density profiles of LP-IFM foams. The profiles demonstrate clearly the dependence on θdec and the aluminum alloy. The foam produced with Al9Si3Cu(Fe) and the supplied blowing agent exhibits a typical V-profile for LP-IFM foams (Fig. 8, middle left). The profile resulting from milled powder shows a more U-shaped profile with a fast increase in porosity below the solid skin so that a plateau with a more homogeneous distribution in the core develops (Fig. 8, top left). Aging of the milled powder leads to a thicker skin with a V-profile (Fig. 8, bottom left). Using Al5Mg2SiMn instead of Al9Si3Cu(Fe) leads to a very thin skin layer with a steep transition to a quite homogeneous plateau of high porosity (Fig. 8, middle right).
Fig. 8

Density profiles (from μCT-measurements) of the LP-IFM foams as a function of blowing agent properties and alloy. Left a higher reactivity leads to a U-shaped profile whereas a lower reactivity amplifies the V profile. Right using Al5Mg2SiMn results in a more U-like density profile with a quite homogeneous plateau of high porosity. The higher the powder reactivity, the thinner the skin and the more pronounced are additional density peaks between transition region and foamed core

Very often we find two local density maxima within the foamed core (Fig. 9, top right). In case of milled particles, the skin thickness is not only decreased but the density maxima also appear to be more pronounced (Fig. 8, top right). Aging of the powder leads again to a thicker skin and a more V-shaped density profile with no density maxima (Fig. 8, bottom right). This phenomenon is not a stochastic fluctuation but reflects the filling history of the mold. At the position from which the samples were taken (Fig. 3, bottom right), the skin and the core were formed with a time lag so that two melt fronts stuck together leading to a characteristic material accumulation (indicated in Fig. 9, right). Figure 9 shows cross sections of the plate from the CFD simulation during the cavity filling process. The position of the μCT sample is marked by the gray/dotted frame. It can be clearly seen that the surface of the mold is already wetted by the arriving melt on its way up. During this filling period, the solid wall forms. The internal section forms about 20 ms later during backfilling.
Fig. 9

Left CFD simulation of particle entrainment; cross sections perpendicular to the gating system with marked withdrawal position of μCT sample. Time measurement starts at first melt/particle interaction in the runner system. Right cross section of the Al5Mg2SiMn sample reconstructed from μCT images and corresponding density profile. Marked local density maxima within the foamed core

During this time delay, foaming already starts in the outer region. This can be understood by calculating the decomposition curve of standard MgH2 at 618 °C (Fig. 10, continuous line) according to the Johnson–Mehl–Avrami theory based on experimental thermogravimetrical curves [5, 11]. The activation energy (EA)of 195 kJ/mol is determined with an Arrhenius type plot as described in [33] based on own experimental data. Thus, between the first contact of the melt with the wall (~40 ms after first melt/particle-contact) and the arrival of the backflow 20 ms later, a high amount of hydrogen is released. This leads to foam formation in the outer region before filling of the inner region by an also bubbling melt.
Fig. 10

Decomposition curves of MgH2 calculated from thermogravimetrical analysis at different heating rates according to the Johnson–Mehl–Avrami approach [5] for different process conditions: Entrainment of as supplied powder (θdec = 432 °C) by Al5Mg2SiMn (θmelt = 618 °C) or Al9Si3Cu(Fe) (θmelt = 580 °C)

Furthermore, due to the fact that strong hydrogen release already occurs before piston stops, the porosity level and pore size in the whole core region are homogeneous. The pores develop within the first 100 ms when the melt temperature over the cross section is still nearly constant. In contrast, in case of using the Al9Si3Cu(Fe)-alloy, the local pore size strongly depends on the distance to the mold wall as pores start to expand later when mold filling is already finished and a temperature gradient along the cross section is established (Fig. 10, dash-dotted line). As the MgH2 is very sensitive to the local surrounding temperature, the velocity of hydrogen release strongly depends on the distance from the center resulting in a graded pore size distribution (Fig. 8, middle left). A retarded decomposition by powder aging has a similar effect on the density profile of Al5Mg2SiMn foams as shown in Fig. 8, bottom right.

4 Conclusions

The structural design of aluminum integral foams is successfully implemented by alloy and blowing agent modification. The thickness of the solid skin—decisive for the respective field of application—can be varied without any intervention regarding the heat content of the melt. Instead of reducing the melt temperature or accelerating the solidification rate, which has negative effects on the surface and pore quality of the produced parts, the skin thickness is increased by slowing down the decomposition kinetics of the powder. It is possible to adjust the desired reactivity of MgH2 by either milling (higher decomposition rates) or oxidizing milled or non-milled powder (lower decomposition rates). Another parameter represents the alloy’s liquidus temperature which has direct influence on the hydrogen release velocity with the powder being very sensitive to the ambient conditions [5, 10]. By timing the decomposition and, therefore, the pore evolution within the process window, the density profiles over the component’s cross section can be varied according to the desired specifications. By combining the temperature fields from CFD with the decomposition kinetics of the powder the differences in the density profiles can be explained. This opens up new possibilities in the design of integral foam structures tailored to the needs for new components.


The authors gratefully acknowledge the financial support of the Deutsche Forschungsgemeinschaft (DFG), grant no. KO 1984/5-2. The authors also thank Dr. Andreas Borgschulte, EMPA, for his help and expertise concerning the decomposition behavior of magnesium hydride and Dr. Ralf Rettig for providing the Thermo-Calc calculations.

Copyright information

© Springer-Verlag Berlin Heidelberg 2014