Functionally graded metal syntactic foam: Fabrication and mechanical properties

Accepted Manuscript Functionally graded metal syntactic foam: Fabrication and mechanical properties

Nima Movahedi, Graeme E. Murch, Irina V. Belova, Thomas Fiedler PII: DOI: Article Number: Reference:

S0264-1275(19)30089-9 https://doi.org/10.1016/j.matdes.2019.107652 107652 JMADE 107652

To appear in:

Materials & Design

Received date: Revised date: Accepted date:

19 November 2018 7 February 2019 9 February 2019

Please cite this article as: N. Movahedi, G.E. Murch, I.V. Belova, et al., Functionally graded metal syntactic foam: Fabrication and mechanical properties, Materials & Design, https://doi.org/10.1016/j.matdes.2019.107652

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Functionally Graded Metal Syntactic Foam: Fabrication and Mechanical Properties Nima Movahedi, Graeme E. Murch, Irina V. Belova, Thomas Fiedler Centre for Mass and Thermal Transport in Engineering Materials School of Engineering The University of Newcastle

PT

Callaghan, NSW, Australia

Abstract

RI

In this research study a novel functionally graded metal syntactic foam (FG-MSF) was

SC

manufactured using expanded perlite and activated carbon particles. A tailored arrangement of these fillers was infiltrated with ZA27 alloy in a single-step process. The structure of the FG-MSF contained two individual layers: ZA27/Expanded Perlite (EP-MSF) and

NU

ZA27/Activated Carbon (AC-MSF) syntactic foam. The density of these FG-MSFs varied between 2.11- 2.15 g.cm-3. Microstructural studies confirmed that no relevant chemical

MA

reaction occurred within the foam, in particular in the vicinity of the particle-matrix interfaces. The mechanical properties of the produced FG-MSF were evaluated using quasistatic compression testing. The results showed that the deformation mechanism of the FG-

D

MSF is a mixed mode and varies between the two different filler layers. The energy

PT E

absorption of the FG-MSF sample was increased compared to uniform syntactic foams containing only a single particle filler.

CE

Keywords: Functionally graded metal syntactic foam; Expanded perlite; Activated carbon;

AC

Deformation mechanism; Energy absorption.

1

ACCEPTED MANUSCRIPT

Nomenclature Activated carbon metal syntactic foam

𝜌B

Bulk density

EP-MSF

Expanded perlite metal syntactic foam

𝜌P

Envelop density

FG-MSF

Functionally graded metal syntactic foam

𝜌s

Solid density

FGM

Functionally graded materials

𝜌𝑀

𝑚SF

Mass of syntactic foam

𝐷0

Volume fraction of matrix

𝜙P

Volume fraction of particles Volume fraction of voids

Initial

diameter

𝐷𝑚

Barrelling diameter

ℎ/𝑑

Aspect ratio

𝜎𝑝𝑙

Plateau stress

EDS

of

Energy dispersive spectroscopy

PT E

1. Introduction

D

MA

𝜙V

Matrix density

RI

𝜙M

NU

Volume of syntactic foam

SC

sample

𝑉SF

PT

AC-MSF

Functionally graded materials (FGM) are advanced engineering materials that exhibit a gradual and controlled positional change of at least one property. This can be achieved by

CE

changing the volume fraction of constituents, microstructure or material type from one location to another. Designing functionally graded materials with tailored properties

AC

(mechanical or physical) has the potential to grant considerable benefits [1]. Static and dynamic behaviour of functionally graded plates were analysed numerically [2, 3]. Quan et al studied the influence of parameters such as mechanical loads on nonlinear vibration of functionally graded shallow shells [4, 5]. Metal foams are a class of advanced materials with distinctive properties such as lightness and high-energy absorption capabilities. These materials have attracted attention as a viable solution for energy absorbers in the automotive industry [6, 7]. The mechanical properties and deformation mechanism of these foams depend strongly on their density [8]. Therefore, it

2

ACCEPTED MANUSCRIPT is of interest to introduce a density gradient along one direction of its structure and thus produce a functionally graded metal foam. In the literature, various approaches have been described to produce and evaluate functionally graded foams using different production techniques. He et al. [9] produced uniform and functionally graded (FG) closed-cell aluminium foam using a specially modified cast technique. Quasi-static compression tests indicated noticeable strain hardening. Dynamic

PT

loading of uniform samples revealed a high peak stress at low strain followed by a rapid decline in the plateau region. In contrast, dynamic loading of the FG foams indicates a lower

RI

peak stress and a more extended plateau region. This behaviour is beneficial as a more constant deformation resistance translates into near constant accelerations of an impacting

SC

mass. Hangaiet al. [10] fabricated FG foam using a sintering-dissolution technique. During the dissolution step, the removal of space holders (NaCl) was partially stopped. The resulting

NU

functionally graded material was a combination of open-cell Al foam and Al/NaCl composite. In another study [11] this research group prepared a functionally graded closed-

MA

cell aluminium foam using the friction stir processing (FSP) technique. They used different amounts of foaming agent in the precursor. The produced samples consisted of two layers with different porosities. By tailoring the high and low porosity layers, it was possible to

D

control the deformation mechanism of the composite FG foam. Hassani et al. [12] fabricated

PT E

a functionally graded open cell aluminium foam using a powder metallurgical method. They used carbamide particles as space holders and changed the particle size in the longitudinal direction of the foam.

CE

Metal matrix syntactic foams are a special group of metal foams, which consist of a metal matrix with embedded filler particles [13]. Different types of fillers have been successfully

AC

used in previous studies. These range from fabricated hollow spheres to naturally occurring porous particles [14, 15]. In most cases, metal matrix syntactic foams have been produced by the homogeneous distribution of fillers inside the matrix. As a result, metal syntactic foams with near-uniform physical properties have been achieved. To gain the benefits of functionally graded materials, it is of interest to introduce a property gradient into syntactic foams by altering the size or volume fraction of particles and/or by employing tailored distributions of two or more different particle types. The former method has been investigated in recent years. Ferreira et al. [16] produced a functionally graded aluminium syntactic foam using a centrifugal casting technique. In their study, a changing volume fraction of microballoons was distributed along the radial direction of the cast samples. In a related 3

ACCEPTED MANUSCRIPT work, random distributions of two different filler particles in an aluminium matrix syntactic foam (hybrid metal syntactic foams) was studied by Májlinger and Orbulov [17]. They investigated random mixtures of hollow ceramic and hollow iron spheres embedded inside an aluminium alloy. Following the above review, the aim of this research is to study the controlled combination of different filler particles within a metal matrix syntactic foam to create a functionally graded

PT

structure. The available research [16] on FG metal syntactic foams focused on the volume fractional change of one particle type in the structure of the material. To the author’s

RI

knowledge, no research studies on the usage of two or more different particle types in a tailored arrangement within FG metal syntactic foams have been published to date and the

SC

paper therefore presents novel findings. This research study shows how filler particles with different properties (bulk density and strength) can be used to create FG-MSF. Furthermore,

NU

the mechanical properties and deformation mechanisms of FG-MSFs and uniform syntactic foams under quasi-static loading conditions are investigated. To this end, a tailored

MA

distribution of two different types of particles is infiltrated with a zinc-aluminium alloy. In addition, uniform foams that are identical to each section of the functionally graded foam are produced. Quasi-static compressive testing is conducted to determine and evaluate the

PT E

D

mechanical properties of these foams.

CE

2. Experimental procedures

2.1.Sample Fabrication

AC

2.1.1. Fabrication of uniform ZA27 syntactic foams Due to its excellent castability, ZA27 alloy was used as the matrix material. The chemical composition of this alloy is 25-28wt% aluminium and 2-2.5wt% copper as the major alloying elements according to ASTM B86-13 [18].Expanded perlite and activated carbon particles with 2-2.8 mm size were used as the fillers (supplied by Australian Perlite Pty and Seachem ® respectively). Uniform syntactic foams were produced using either expanded perlite or activated carbon particles. The established counter gravity infiltration casting technique was used to manufacture the samples. This technique was successfully employed to manufacture the aluminium alloy syntactic foam using expanded perlite as fillers [15]. A schematic of the

4

ACCEPTED MANUSCRIPT setup is shown in Fig. 1. In the first step, a graphite mould was filled using tapping and vibration technique in four individual steps. This technique is used to minimize the density gradient within the structure. Next, a stainless steel mesh was located at the open end of the graphite mould to keep the particles in place. A ZA27 solid block was inserted into a graphite crucible. The filled mould was inserted upside down in the same crucible leaving only a tight gap to permit relative motion. The assembly was then heated inside a resistance furnace at

PT

535oC and kept at this temperature for 30 minutes in an argon atmosphere to minimize oxidation. In the next step, the assembly was removed from the furnace and casting was initiated by pressing the mould into the crucible with a force of 9.8 N. The mould was slowly

RI

cooled in atmospheric conditions before the sample was removed for machining of the flat

SC

surfaces to remove the metallic meshes. In the final step, heat treatment was performed on all syntactic foams. The samples underwent a solution treatment at 365oC for 1 hr followed by

NU

water quenching. Finally, samples were artificially aged at 140oC for 24 hrs and cooled to room temperature at atmospheric conditions [19].

MA

2.1.2. Fabrication of functionally graded ZA27 syntactic foams For fabrication of the FG-MSFs, modifications are restricted to the filling of the mould. The

D

graphite mould was divided into two equal volumes in its longitudinal direction (see Fig.1).Filling the mould was performed in two sequential steps. First, the activated carbon

PT E

particles were filled into the mould using the packing and vibration technique to achieve a uniform particle packing density. A thin piece of a paper was placed on top of the particle column to avoid subsequent particle mixing. Second, the remaining space was filled with

CE

expanded perlite particles using the same procedure. Henceforth, the manufacturing of the

AC

FG-MSFs was completed similar to uniform syntactic foams (see section 2.1.1).

5

SC

RI

PT

ACCEPTED MANUSCRIPT

2.2 Characterization

MA

2.2.1 Characterization of the filler particles

NU

Fig.1. Schematic of the production process.

The bulk density of particles was determined by measuring their weight and the macroscopic volume that these particles occupy within a volumetric measurement cylinder. In addition,

D

their envelope density (𝜌𝑃 )was obtained using the flour method [20]. Particle porosity was

s   p s

(1)

CE

Particle porosity =

PT E

calculated according to Equation (1):

AC

where 𝜌s is the solid density of the cell wall material. In order to investigate the behaviour of particles under compression, the individual particles (either expanded perlite or activated carbon particle)were tested in a 5 KN SHIMADZU compression testing machine with 0.2 mm.min-1 crosshead speed. The obtained force displacement data were used for this purpose.

2.2.2. Physical properties of the uniform syntactic foams

6

ACCEPTED MANUSCRIPT The density of uniform ZA27 syntactic foam samples was calculated after measuring their mass mSF (on a precision scale with accuracy of 0.01 g calliper with accuracy of ± 0.01 mm), i.e. SF 

and volume VSF (using a digital

mSF . VSF

The volume fractions of the matrix (M ) particles (P ) and voids (V )

in the uniform

PT

samples were determined according to the following equations [21]:

RI

mSF  mp

M

M 

(2)

SC

VSF

B P

MA

P 

(3)

NU

mP  B VSF 

D

(4)

PT E

V  1  M  P (5)

carbon),  M

CE

Where mP is the combined mass of the filler particles (either expanded perlite or activated is the solid density of the metal matrix and  B

and  P

are the bulk and

AC

envelope density of the embedded filler particles. 2.2.3. Physical properties of the FG-MSFs The overall density of the produced FG-MSF samples was measured using their initial weight and macroscopic dimensions. The volume fraction of each layer i was calculated according to Equation (6):

i 

Vi  Vj

(6)

7

ACCEPTED MANUSCRIPT To determine the volume of each layer (𝑉i ), the cross sectional area and the height of each layer were measured.The particle volume fraction for each layer (𝜙P,i )was assumed to be the same as in uniform foams (i.e. according to Equation (4)) and the total particle volume fraction (𝜙P ) of FG-MSF was obtained as Equation (7): P  P,i  i

(7)

PT

The combined particle mass (𝑚P ) is calculated by combining Equations (3) and (4) for each layer:

RI

mP   (Vi  P,i  P,i )

(8)

SC

where 𝜌P,i is the envelope density of the particles within the respective layer. The overall void fraction of FG-MSF was estimated according to Equation 9: mSF  mP VSF  M

NU

V  1  P 

(9)

MA

where, 𝑚SF and𝑉SFare the total mass and volume of the FG-MSF respectively. Then, the volume fraction of the matrix within the layered foam is measured using Equations (9) and

D

(7):

(10)

PT E

M  1  P  V

CE

2.2.4. Microstructural analysis of the FG-MSFs The microstructure of FG-MSF was studied using a Zeiss Sigma VP FE-SEM Scanning

AC

Electron Microscope equipped with an energy dispersive spectroscopy (EDS) sensor. A particular focus was given to the interface region of the two particle types. To this end the FG-MSF sample was truncated in the longitudinal direction and polished using silicon carbide grinding papers with different grit sizes. An elemental analysis was performed to identify the distribution of chemical elements and probe for possible chemical reactions. 2.2.5. Mechanical properties of the produced syntactic foams The mechanical properties of all samples were determined using quasi-static compression tests. A 50 kN SHIMADZU uni-axial testing machine with a crosshead positional accuracy of ± 0.01 mm, crosshead speed accuracy of ± 0.1% and load cell accuracy of ± 0.5% of the 8

ACCEPTED MANUSCRIPT indicated force was used to compress the samples. A constant cross head speed of 1 mm.min 1

was used in all tests. The load and cross head displacement were captured using the

Trapezium 2 software.

The recorded force-displacement data were converted into

engineering stress and strain using the initial sample cross sectional area and height. An analysis of experimental uncertainty is included in Appendix 1. Prior to testing, both ends of the cylindrical samples were lubricated in order to minimize friction between samples and

PT

compression platens. The lowest density sample in each group was tested first. Its plateau stress was used to determine the unloading cycle for the compression tests of the remaining samples in the corresponding group [22]. The main mechanical properties of the syntactic

RI

foams were calculated from these stress strain curves according to ISO 13314 [22]. 1% Proof

SC

stress, plateau stress between 20% and 40% compressive strain, plateau end strain, energy absorption and energy absorption efficiency were considered. The energy absorption 𝑊 and

NU

energy absorption efficiency 𝜂 were calculated using Equations (11) and (12): 0.5

W    d



MA

0

W 0.5   max

(11)

(12)

D

A digital camera was used to visually capture the deformation sequence of the syntactic foam

PT E

samples during quasi-static compression.

2.2.6. Mechanical properties of the solid ZA27 matrix material

CE

To investigate the mechanical properties and the deformation behaviour of the matrix material, two solid ZA27 alloy cylinders were compressed using identical testing conditions

AC

to the syntactic foam samples. To ensure that these solid ZA27 samples exhibit a similar micro-structure as the matrix material in the syntactic foam samples, the solid samples have been machined out of the excess melt from foam casting and subjected to the same thermal treatment. Due to a 50 kN load constraint of the uni-axial testing machine, a smaller solid sample diameter of 10 mm was required. The height of the solid samples h  16 mm was then selected to match the aspect ratio to the syntactic foam samples.

3. Results and Discussion

3.1 Properties of the constituents 3.1.1.

Filler particles 9

ACCEPTED MANUSCRIPT In Fig.2, light photography of the two different particle types are shown. The expanded perlite and activated carbon fillers exhibit a polygonal and spherical shape, respectively. The particle size of expanded perlite is varied between 2-2.8 mm and the mean diameter of the activated carbon particles is 2.39 mm (according to the supplier data sheet).Table1 shows the physical properties of these two particle types. The bulk and envelope densities of the activated carbon particles are higher than for expanded perlite. This is caused predominantly

PT

by a lower volume of pores in the structure of the activated carbon particles. For the same reason, these particles exhibit a higher strength under compressive loads. The particle size is known to influence the mechanical properties of MSF. This may occur by altering the micro-

RI

structure of the metallic struts [20] or due to changes in particle properties [23]. In order to

PT E

D

MA

NU

SC

eliminate this effect, the same particle size range was selected for all foams of this study.

10mm

(b)

(a)

AC

CE

Fig.2. Typical morphology of (a) activated carbon and (b) expanded perlite particles.

Particle

Table 1. Physical properties of the particles

Bulk density (𝜌B ) (g.cm-3)

Particle density (𝜌P ) (g.cm-3)

Expanded perlite 0.09

0.16

Activated carbon0.49

0.83

[24]

Packing density(𝜙P )

Porosity

0.56

94%

0.59

63%

Fig.3 shows the compressive load-displacement curves for the compression of single particles. The activated carbon particles have dentate load-displacement curves. At the early stages of deformation, a high peak force is reached that is followed by a rapid decline before the load increases again. The cycle repeats several times resulting in the serrated morphology 10

ACCEPTED MANUSCRIPT of the load displacement curve. This behaviour is caused by brittle cracking and fracture of activated carbon particles under compressive loads. In contrast, the expanded perlite particles show a gradual load increase over displacement. The deformation force of expanded perlite is considerably lower than that for the activated carbon fillers. This deviation in strength is mostly related to the high porosity of the expanded perlite particles [15].According to Equation (1) the average porosities of expanded perlite and activated carbon particles were

PT E

D

MA

NU

SC

RI

PT

determined to be 94% and 63% respectively.

Fig.3. Compressive load-displacement curves of individual particles.

CE

3.1.2. ZA27 matrix alloy

Fig.4 shows the compressive stress strain curves of two heat-treated solid ZA27 samples

AC

taken from a previous study [25]. Compression tests were performed on two heat-treated ZA27 solid samples. It can be seen in Fig.4 that the resulting stress- strain curves are similar indicative of consistent base material properties. The curves exhibit strain hardening, i.e. the stress increases gradually with strain. It was shown in [25] that the heat treatment changes the deformation mechanism of the ZA27 alloy from brittle to ductile mode. During heattreatment, the dendritic microstructure of the as-cast ZA27 alloy is transformed to a more spheroidized morphology resulting in a considerable improvement in ductility [19, 25].According to [26], the density of the ZA27 matrix is (𝜌𝑀 ) considered to be 5.00 g.cm-3.

11

RI

PT

ACCEPTED MANUSCRIPT

SC

Fig.4. Compressive stress-strain curves for heat treated solid ZA27 samples [25].

NU

3.2. Properties of the Syntactic Foams 3.2.1Physical properties

MA

The physical properties of the uniform syntactic foams are shown in Table 2.Unlike fly ash microballoons with a single large pore, damaged expanded perlite and activated carbon particles cannot be completely filled during casting [27]. As a result, the density of the

D

samples in this study is distinctly below the values reported in the literature for ZA22

PT E

alloy/Ni-coated fly ash microballoon syntactic foams (3.3-5.1 g.cm-3) [27]. Despite using two types of particles with distinctly different bulk densities (see Table 1), the densities of the uniform ZA27 syntactic foams fall within a relatively narrow band from 1.92 to 2.15 g.cm-3.

CE

One explanation is the higher particle volume fraction in AC-MSF that decreases the available volume for the higher density ZA27 matrix. In addition, AC-MSF exhibits a higher

AC

void volume fraction of 7.85-10.04% compared to 4.54- 6.46% for EP-MSF (see Table 2.).The difference in the void volume fraction is likely attributed to differences in the wettability of particle surfaces. The higher particle and void fractions in AC-MSF thus compensate the increased AC particle density. Table.2. Physical properties of uniform syntactic foams.

Sample No

Density (g.cm-3)

ΦM (%)

ΦP (%)

ΦV (%)

EP-MSF-1

1.92

36.66

56.88

6.46

EP-MSF-2

1.99

38.04

56.88

5.08

EP-MSF-3

2.02

38.58

56.88

4.54

AC-MSF-1

2.13

32.48

59.03

8.49

12

ACCEPTED MANUSCRIPT AC-MSF-2

2.04

30.93

59.03

10.04

AC-MSF-3

2.15

33.12

59.03

7.85

PT

Fig.5 shows a truncated FG-MSF sample (this sample was not considered in compression testing). Functionally graded properties are introduced into the syntactic foam by changing

RI

the filler type in longitudinal direction. The physical properties of the FG-MSFs are shown in

SC

Table 3.As expected, the overall volume fractions of matrix and particles in FG-MSFs fall between the values of two uniform syntactic foams. The FG-MSF fractions are estimated

PT E

D

MA

value of the uniform foam (see Table 2).

NU

based on the assumption that the particle volume fractions in each layer corresponds to the

CE

10mm

AC

Fig.5.Truncated FG-MSF sample showing particle layers.

Table.3. Physical properties of FG-MSF samples. Sample No

Density (g.cm-3)

ΦM (%)

ΦP (%)

Φv (%)

ΦP, AC (%)

ΦP, EP (%)

FG-MSF-1

2.11

36.41

57.92

5.67

59.03

56.88

FG-MSF-2

2.15

37.30

57.96

4.74

59.03

56.88

FG-MSF-3

2.15

38.12

57.88

4.00

59.03

56.88

FG-MSF-4

2.14

37.26

57.94

4.80

59.03

56.88

13

ACCEPTED MANUSCRIPT

3.2.2Microstructural analysis of the FG-MSFs A SEM image of the interface between two layers in the FG-MSF is shown in Fig.6a. The elemental distribution is obtained by an EDS line scan along the arrow in Fig.6a. The results

PT

of this analysis are shown in Fig.6b. In the segment crossing the expanded perlite particle, silicon and oxygen are the dominant elements. This can be explained by chemical

RI

composition of expanded perlite, which is mainly composed of SiO2 [15].In the interface between the two particles (i.e. within the matrix), the highest intensities are related to zinc

SC

and aluminium. The higher intensity of aluminium in this region is most likely related to the presence of an aluminium-rich phase (such as α dendrites [28]) in this region. Within the

NU

region of the activated carbon particle, the carbon intensity increases rapidly. The analysis of elemental distribution suggests that no chemical reactions occurred near the interface of the

MA

two particle types in the FG-MSF sample. The visible carbon and silicon peaks towards the centre of the matrix are likely caused by the propagation of particle fragments due to sample

CE

PT E

D

polishing.

Fig. 6(a) SEM image of the interface between expanded perlite and activated carbon in FG-MSF structure

AC

(b)EDS line scan.

3.2.3Mechanical properties Uniform syntactic foam Fig.7a shows the engineering stress-strain curves of EP-MSF syntactic foams. In a previous study [25], the effect of heat treatment on the compressive behaviour of EP-MSF syntactic foams was investigated. It was shown that the ductility of the ZA27 struts will increase due to microstructural changes. As a result, heat treated syntactic foams exhibit a smooth stress 14

ACCEPTED MANUSCRIPT plateau with gradually increasing stress. This is caused by ongoing plastic deformation and strain hardening of the foam struts. Light photography of the syntactic foam deformation is shown in Fig.7b.The heat-treated EP-MSF syntactic foam deforms in a layer-by-layer mode that is perpendicular to the loading direction. In addition, barrelling of the sample without the formation of shear bands is observed during deformation of EP-MSF. The compression behaviour of EP-MSF syntactic foam resembles the solid ZA27 samples (see Fig.4), albeit at

PT

much lower stresses. It can be inferred that the deformation of EP-MSF syntactic foam is mainly controlled by its metallic matrix. The expanded perlite particles introduce porosity into the syntactic foam; however, their low crushing strength (see Fig.3) prevents them from

AC

CE

PT E

D

MA

NU

SC

RI

significantly contributing to the strength of the syntactic foam [15].

15

RI

PT

ACCEPTED MANUSCRIPT

SC

Fig.7 (a). Compressive stress strain curves EP-MSF samples,

NU

(b) Compressive deformation behaviour of EP-MSF with density of 1.99 g.cm-3 [25].

In Fig.8, the compressive stress-strain curves of the AC-MSF syntactic foams are shown. The

MA

stress-strain curves contain three distinct regions. An initial elastic region leads to a stress maximum and is followed by a stress drop towards the plateau region. The plateau region exhibits a serrated morphology up to the densification of the syntactic foam. In spite of using

D

the same matrix material (heat-treated ZA27 alloy), the quasi-static compressive stress-strain

PT E

data of the uniform syntactic foams in Figs. 7a and 8a differ significantly. Unlike EP-MSF syntactic foam, the stress drops inFig.8a indicate the formation of shear bands during early deformation. A comparison between the compression of a single activated carbon particle

CE

(see Fig.3) and AC-MSF (seeFig.8) indicates the contribution of the activated carbon fillers to the deformation behaviour of the syntactic foam as both exhibit similar plateau stress

AC

oscillations. The higher strength of the activated carbon particles also results in distinctly higher stresses of AC-MSF at low strain.

16

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

Fig.8 (a) Compressive stress- strain curve of AC-MSF samples. (b)Compressive deformation sequence of AC-MSF with density of 2.04 g.cm-3.

The deformation sequence of the AC-MSF is shown in Fig.8b. During early deformation 𝜖 ≈ 0.05, a shear plane emerges with an orientation of approximately 45o relative to the loading direction (see dashed lines in Fig.8b). With increasing strain, this shear bands grows and at least one additional shear band (no images of the back of the sample were captured) forms in the upper part of the sample (see Fig.8b at ε=0.1). The formation of these shear bands coincides with the plateau stress fluctuations in Fig.8a and is likely assisted by the fracture of activated carbon particles. Brittle deformation of metal syntactic foams with

17

ACCEPTED MANUSCRIPT formation of shear bands was also observed in aluminium syntactic foams using pumice and expanded perlite fillers [29, 30]. FG-MSF The compressive stress strain curves of the functionally graded ZA27 syntactic foams are presented in Fig. 9a. The morphology of the stress strain curves is similar for all tested samples. Minor deviations in the compressive stress strain curves are most likely related to

PT

the density difference between samples and are similar to the scattering of the uniform syntactic foams.Fig.9b shows the deformation of the FG-MSF sample with 𝜌 = 2.15g.cm-3

RI

(see arrow inFig.9a). During early deformation, the expanded perlite layer starts to deform.

SC

This coincides with previous studies of functionally graded metal foam where the deformation originates within the weakest section [31, 32].Upon further compression, the

NU

perlite section gradually deforms with the previously observed layer-by-layer deformation mechanism (similar to EP-MSF, see Fig. 7a).This process continues until the onset of densification of this layer at 𝜀 ≈ 0.25.Similar to uniform EP-MSF, strain hardening is

MA

observed in conjunction with layer-by-layer collapse. It is important to note that the bottom layer (equivalent to AC-MSF) does not undergo considerable macroscopic deformation at low strains. However, at 𝜀 ≈ 0.3 deformation of the activated carbon layer is observed. This

D

coincides with a stress drop due to the formation of shear bands in the activated carbon layer

PT E

of the FG-MSF sample (see Fig.9).At 𝜀 ≈ 0.4-0.65, minor stress fluctuation appear due to the ongoing shear deformation of the AC layer. Macroscopic foam densification commences at

AC

CE

𝜀 ≈ 0.65.

18

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig. 9 (a) Compressive stress strain curves of FG-MSF samples. (b) Compressive deformation of FG-MSF

AC

sample with density of 2.15 g.cm-3.

3.3Analysis and Discussion In Fig.10a the compressive stress-strain curves of the uniform and functionally graded syntactic foams are plotted together for better comparison. The morphology of these curves is different for each foam type. Importantly, the stress-strain curves of FG-MSF sample contain two distinct plateau regions. Each of these plateaus shares characteristics of the stress-strain curves of one of the uniform syntactic foams. The ascending behaviour of the FG-MSF stress strain curve in the first plateau is followed by a stress drop in towards the second plateau. 19

AC

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig.10 (a) Compressive stress strain curves of uniform and FG-MSF samples. (b) Modified compressive stress strain curves.

FG-MSF shows a strong concentration of initial deformation (see Fig. 10a): at low strains, deformation is restricted to the EP layer whereas at higher strains deformation occurs predominantly within the AC layer. As a result, each layer undergoes a higher localised strain and Fig. 10.a can be modified for a better comparison. The height of each filler layer in FGMSF is about half of the overall sample height. Therefore, the strain values of the uniform 20

ACCEPTED MANUSCRIPT syntactic foams were multiplied by the constant factor of 0.5 to compensate for this deviation. In addition, the stress-strain curves of AC-MSF where shifted to the onset of the non-linear AC layer deformation in FG-MSF. The resulting stress strain curves are shown in Fig. 10.b and form the basis of the subsequent discussion. Deformation stage 1 The first deformation stage (blue background) of the FG-MSF curve is similar to the

PT

modified stress strain curve of EP-MSF. In addition, the deformation mode of the EP-MSF

RI

layer in FG-MSF is similar to EP-MSF (compareFigs.7b and 9b).

In both cases, EP-MSF deforms in a layer-by layer mode with some barrelling. This results

SC

in strain hardening visible in the stress-strain curves (see Fig.10b). In an attempt to quantify and better compare the degree of barrelling, an optical analysis of the photography has been

NU

conducted. To this end, the diameter of the undeformed sample (D0) and its maximum barrelling diameter (Dm) at different stages (points A-D indicated in Fig.10b) were considered

MA

to calculate the diameter increase (Di-D0).The average results for each group of samples are shown in Fig. 11b. The findings indicate that the diameter of EP-MSF (with a higher aspect ratio) increases faster and more than the EP layer in FG-MSF (see Fig.11a). In the case of

D

FG-MSF, the onset of significant barrelling is delayed until after point B. This is in

PT E

agreement with the findings of Altinbalik and Can [33]. They investigated the influence of aspect ratio on deformation and barrelling of a solid aluminium alloy during compression and observed the samples with smaller aspect ratio showed smaller diameter increment during

CE

compression. Due to the smaller aspect ratio of the EP layer, the functionally graded syntactic foam is compressed with more stability (i.e. undergoes less barrelling) which likely explains

AC

its higher strength at low strains compared to EP-MSF.

21

AC

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig. 11(a) Schematic deformation of studied samples.(b) Diameter increment in EP-MSF and EP layer in FGMSF samples.

Deformation stage 2 The second plateau of the FG-MSF sample (orange background) shares characteristics with the stress-strain data of AC-MSF. Similar to the uniform foam, the stress of FG-MSF decreases; however (i) at a lower rate and (ii) without significant stress oscillations. These deviations can be explained by differing deformation mechanisms caused by a variation of the aspect ratio. The aspect ratio of AC-MSF is ℎ/𝑑 = 1.5, whereas the aspect ratio of the 22

ACCEPTED MANUSCRIPT AC layer in FG-MSF is only ℎAC /𝑑 = 0.75. Majilinger and Orbulov [17, 34] showed that changing the aspect ratio of uniform aluminium alloy syntactic foams alters their deformation mechanism during quasi-static compression, in particular the formation of shear bands. In AC-MSF with ℎ/𝑑 = 1.5, multiple shear bands form at an angle of about 45o relative to the loading direction in different parts of the sample (see Fig. 8).The AC-MSF sample then fractures along these shear bands and is divided in two halves. Due to the high aspect ratio,

PT

these halves are able to slide along each other whilst most parts of the sample remain undeformed prior to densification (see Fig. 11a).In contrast, the AC layers of FG-MSF with ℎAC

= 0.75 develop multiple shear bands that tend to intersect (e.g. see Figs.9 and 11,𝜀 ≈

RI

𝑑

0.4).This kind of deformation was previously observed in syntactic foam samples with small

SC

aspect ratiosℎ/𝑑 ≤ 1 [17, 34].Due to the lower aspect ratio of the AC layers, the shear bands are unable to bisect the entire sample (see red lines in Fig. 11b). As a result, relative sliding

NU

of FG-MSF sample fragments along a shear plane is impossible and the material is instead compressed between the pressure platens. This triggers the formation of additional shear

MA

bands. It further permits to explain the slower stress decline and decreased stress fluctuations

D

in the second deformation stage of FG-MSF compared to AC-MSF.

PT E

3.3.1 Effective Mechanical properties

The evaluated mechanical properties of the samples are plotted in Fig.12 against the density. The density difference between uniform and functionally graded samples is relatively small

CE

(the maximum density difference is 11%). In contrast, the mechanical properties show a remarkable variation.

AC

The 1% proof stress of the uniform AC-MSF is distinctly higher compared to EP-MSF samples. As described in Fig.3the compressive strength of the activated carbon filler is considerably higher than that of expanded perlite. Non-surprisingly, their corresponding uniform syntactic foams follow a similar trend. Comparison between FG-MSFs and EP-MSF shows that the 1% proof stress of the functionally graded structure is only slightly higher than the uniform EP sample. The likely explanation is that the deformation of FG-MSF originates within the expanded perlite section (weaker layer) and thus exhibits a similar proof stress value as EG-MSF.

23

ACCEPTED MANUSCRIPT The plateau stress (𝜎𝑝𝑙 )depicted in Fig. 12 is one of the important characteristics of metallic foams and determines the resistance of the material during compression. According to ISO 13314 [22], 𝜎𝑝𝑙 is the arithmetical means of stress between 20% and 40% of compressive strain. The plateau stress of the functionally graded samples clearly exceeds both uniform syntactic foams. The average plateau stress for FG-MSF is 88.14 MPa ± 0.45, compared to 56.17 ± 0.28 MPa and 45.94 ± 0.23 MPa for EP-MSF and AC-MSF, respectively. The higher

PT

plateau stress of FG-MSF can be explained by the macroscopic deformation mechanisms described in the previous Section: FG-MSF does not undergo catastrophic shear failure

RI

resulting in a decreased stress drop and increased plateau stress. The higher plateau stress further results in an increased volumetric energy absorption of FG-MSF. The average

SC

volumetric energy absorption of FG-MSF is 35.74 MJ.m-3 compared to 27.22 MJ.m-3 for EPMSF and 26.84 MJ.m-3 for AC-MSF.

NU

The plateau end strain corresponds to 1.3 times the plateau stress. This value is known to be sensitive to the deformation mechanism and should be interpreted with care. Layer-by-layer

MA

collapse and work hardening results in a progressive plateau stress increase of EP-MSF. As a result, these foams exhibit a relatively low plateau end strain. The plateau end strain of most AC-MSF samples is relatively low due to plateau stress oscillation. The highest values are

D

found for FG-MSF which exhibits a near constant plateau stress up to high strains. This

AC

CE

PT E

further results in the maximum energy absorption efficiency of this study.

24

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

Fig.12. Main mechanical properties of the produced uniform and FG-MSFs.

4. Conclusions A novel functionally-graded metallic syntactic foam (FG-MSF) was produced using two different types of filler particles, i.e. expanded perlite (EP) and activated carbon (AC). These particles were vertically separated to form two distinct layers of equal volume containing only one particle type. For comparison, uniform MSF samples with either activated carbon (AC-MSF) or expanded perlite (EP-MSF) were manufactured and tested. The density of the 25

ACCEPTED MANUSCRIPT FG-MSF is 2.11 to 2.15 g.cm-3which lies between the density of uniform EP-MSF (2.02 g.cm-3) and AC-MSF (2.15 g.cm-3). FG-MSF was subjected to compression testing and results were compared to uniform foams. It was found that initial deformation of FG-MSF is concentrated within the weaker EP layer. At low strains, the FG-MSF deformation closely resembled the compression behaviour of uniform EP-MSF. However, differences were found at higher strains during the subsequent

PT

deformation of the AC layer. In FG-MSF, catastrophic shear failure was suppressed due to a smaller aspect ratio. This resulted in superior FG-MSF performance, quantified by an

SC

RI

increased volumetric energy absorption and energy absorption efficiency.

Data availability

MA

the data also forms part of an ongoing study.

NU

The raw/processed data required to reproduce these findings cannot be shared at this time as

Appendix. 1: Analysis of Experimental Uncertainty

D

Measured properties

PT E

Mass was measured using a AD GX-6100 precision scale with an uncertainty um  0.01 g Based on the lowest sample mass, the highest fractional uncertainty is therefore 0.01 g  0.053% (1) 19.04 g

CE

m 

AC

Length was measured using a calliper with an uncertainty 𝑢𝑙 of ± 0.01 mm. The average of three independent measurements at different locations was used to determine sample diameter 𝑑 and height ℎ. Based on the samples’ dimensions the fractional uncertainties are

d 

0.01 mm   0.050% and 20 mm

h 

 0.01 mm  0.033% 30 mm (3)

(2)

26

ACCEPTED MANUSCRIPT Displacement was measured using the testing machine cross-head displacement. The specified uncertainty is ± 0.01 mm. Elastic deflection of the cross-head and compression platens was disregarded due to the relatively low compression loads. At 50% macroscopic compressive strain, the fractional uncertainty for the displacement measurement is thus DP 

0.01 mm  0.067% . 0.5  30 mm

(4)

PT

Force was measured using a 50 kN Shimadzu load cell with a specified fractional uncertainty of F  0.5%

RI

Derived properties

SC

d2 Volume is obtained using V     h . (5) 4

NU

Based on the experimental uncertainty of the length measurements for 𝑑 and ℎ (both ± 0.01 mm) the fractional uncertainty for the volume is thus [35] (6)

m . (7) v

PT E

D

Density is obtained using  

MA

V  2d2  h2  0.078%

According to the fractional uncertainties of mass and volume, the fractional uncertainty of

CE

the density is

ρ  V2  m2  0.094%

(8)

AC

This uncertainty value is too small and not considered.

Strain is calculated using  

l . (9) h

Based on the experimental uncertainty of the length measurement and machine crosshead displacement, the uncertainty for the strain is: 2   2h  DP  0.075%

This uncertainty value is too small and not considered.

Stress is calculated using  

F . A

(11) 27

(10)

ACCEPTED MANUSCRIPT The uncertainty of stress is measured according to:

σ  F2 2d2  0.505%

(12)

The numerical measurement data alongside uncertainty intervals are given in the Table below:

Density

Proof Stress (MPa)

Plateau stress (MPa)

Plateau end strain (-)

33.66 ± 0.17 39.84 ± 0.20 43.19 ± 0.22 74.03 ± 0.37 67.10 ± 0.34 74.55 ± 0.38 49.78 ± 0.25 49.95 ± 0.25 51.76 ± 0.26 50.56 ± 0.26

51.31 ± 0.26 57.17 ± 0.29 60.04 ± 0.30 66.63 ± 0.34 31.92 ± 0.16 39.28 ± 0.20 87.35 ± 0.44 85.95 ± 0.43 90.30 ± 0.46 88.95 ± 0.45

0.43 (± 0.00) 0.42 0.43 0.67 0.41 0.48 0.73 0.71 0.72 0.71

SC

(g.cm ) 1.92 (± 0.00) 1.99 2.02 2.13 2.04 2.15 2.11 2.15 2.15 2.14

D

MA

NU

code EP-MSF-1 EP-MSF-2 EP-MSF-3 AC-MSF-1 AC-MSF-2 AC-MSF-3 FG-MSF-1 FG-MSF-2 FG-MSF-3 FG-MSF-4

RI

-3

PT

Sample

PT E

References

[1] M. Naebe, K. Shirvanimoghaddam, Functionally graded materials: A review of fabrication and properties, Appl. Mater. Today. 5 (2016) 223-245.

CE

https://doi.org/10.1016/j.apmt.2016.10.001

AC

[2] N.D. Duc, Nonlinear Static and Dynamic Stability of Functionally Graded Plates and Shells, first ed., Vietnam National University Press, Hanoi, 2014. [3] N.D. Duc, Nonlinear dynamic response of imperfect eccentrically stiffened FGM double curved shallow shells on elastic foundation, Compos. Struct. 99 (2013) 88-96. https://doi.org/10.1016/j.compstruct.2012.11.017

[4] T.Q. Quan, P. Tran, N.D. Tuan, N.D. Duc, Nonlinear dynamic analysis and vibration of shear deformable eccentrically stiffened S-FGM cylindrical panels with metal-ceramic-metal layers resting on elastic foundations, Compos. Struct. 126 (2015) 16-33. https://doi.org/10.1016/j.compstruct.2015.02.056

[5] T.Q. Quan, N.D. Duc, Nonlinear vibration and dynamic response of shear deformable imperfect functionally graded double-curved shallow shells resting on elastic foundations in thermal environments, J. Therm. Stresses. 39 (2016) 437-459. https://doi.org/10.1080/01495739.2016.1158601 28

ACCEPTED MANUSCRIPT [6] M.F. Ashby, A. Evans, N.A. Fleck, L.J. Gibson, J.W. Hutchinson, H.N.G. Wadley, F. Delale, Metal Foams: a Design Guide, first ed., Butterworth-Heinemann, Oxford, 2000. [7] X. Yang, W. Wang, L. Yan, Q.C. Zhang, T.J. Lu, Effect of pore morphology on crossproperty link for close-celled metallic foams, J. Phys. D. 49 (2016) 1-6. https://doi:10.1088/0022-3727/49/50/505301

[8] H.P. Degischer, B. Kriszt, Handbook of Cellular Metals Production, Processing, Application, first ed., Wiley-VCH Verlag GmbH, Germany, 2002.

PT

[9] S.Y. He, Y. Zhang, G. Dai, J. Q. Jiang, Preparation of density-graded aluminium foam, Mater. Sci. Eng. A. 618 (2014) 496-499. https://doi.org/10.1016/j.msea.2014.08.087

SC

RI

[10] Y. Hangai, K. Zushida, O. Kuwazuru, N. Yoshikawa, Functionally graded Al foam fabricated by sintering and dissolution Process with remaining spacers, Mater. Trans. 57(5) (2016) 748-750. https://doi.org/10.2320/matertrans.M2015404

https://doi.org/10.1016/j.msea.2011.11.100

NU

[11] Y. Hangai, K. Takahashi, T. Utsunomiya, S. Kitahara, O. Kuwazuru, N. Yoshikawa, Fabrication of functionally graded aluminium foam using aluminium alloy die castings by friction stir processing, Mater. Sci. Eng. A. 534 (2012) 716-719.

MA

[12] A. Hassani, A. Habibolahzadeh, H. Bafti, Production of graded aluminium foams via powder space holder technique, Mater. Des. 40 (2012) 510-515.

D

https://doi.org/10.1016/j.matdes.2012.04.024

PT E

[13] N. Gupta, P.K. Rohatgi, Metal Matrix Syntactic Foams, Processing, Microstructure, Properties and Applications, first ed., DEStech Publication, Pennsylvania, 2015.

CE

[14] A. Szlancsik, B. Katona, K. Bobor, K. Májlinger, I. N. Orbulov, Compressive behaviour of aluminium matrix syntactic foams reinforced by iron hollow spheres, Mater. Des. 83 (2015) 230-237. https://doi.org/10.1016/j.matdes.2015.06.011

AC

[15] M. Taherishargh, I.V. Belova, G.E. Murch, T. Fiedler, Low-density expanded perlite aluminium syntactic foam, Mater. Sci. Eng. A. 604 (2014) 127-134. https://doi.org/10.1016/j.msea.2014.03.003

[16] S.C. Ferreira, A. Velhinho, R.J.C. Silva, L.A. Rocha, Corrosion behaviour of aluminium syntactic functionally graded composites, Int. J. Mater. Prod. Technol. 39 (1-2) (2010) 122135. https://doi.org/10.1504/IJMPT.2010.034265 [17] K. Májlinger, I. N. Orbulov, Characteristic compressive properties of hybrid metal matrix syntactic foams, Mater. Sci. Eng. A. 606 (2014) 248-256. https://doi.org/10.1016/j.msea.2014.03.100

29

ACCEPTED MANUSCRIPT [18] ASTM, Standard Specification for Zinc and Zinc-Aluminium (ZA) Alloy Foundry and Die Castings, B86-13, 2013. [19] Y. Liu, H.Y. LI, H. F. Jiang, X. C. Lu, Effects of heat treatment on microstructure and mechanical properties of ZA27 alloy, Trans. Nonferrous Met. Soc. China. 23 (3) (2013) 642649. https://doi.org/10.1016/S1003-6326(13)62511-X

PT

[20] K. Al-Sahlani, S. Broxtermann, D. Lell, T. Fiedler, Effects of particle size on the microstructure and mechanical properties of expanded glass-metal syntactic foams, Mater. Sci. Eng. A. 728 (2018) 80-87. https://doi.org/10.1016/j.msea.2018.04.103

SC

RI

[21] K. Al-Sahlani, M. Taherishargh, E. Kisi and T. Fiedler, Controlled shrinkage of expanded glass particles in metal syntactic foams, Materials. 10 (9) (2017)114. https://doi.org/10.3390/ma10091073

NU

[22] ISO, Mechanical Testing of Metals- Ductility Testing- Compression Test for Porous and Cellular Metals, 13314, 2011 (Switzerland). [23] M. Taherishargh, M.A. Sulong, I.V. Belova, G.E. Murch, T. Fiedler, On the particle size effect in expanded perlite aluminium syntactic foam, Mater. Des. 66 (2015) 294-303.

MA

https://doi.org/10.1016/j.matdes.2014.10.073

[24] M. Arifuzzaman, H. S. Kim, Novel mechanical behaviour of perlite/sodium silicate composites, Constr. Build. Mater. 93 (2015) 230-240.

D

https://doi.org/10.1016/j.conbuildmat.2015.05.118

PT E

[25] N. Movahedi, G.E. Murch, I.V. Belova, T. Fiedler, Effect of heat treatment on the compressive behaviour of zinc alloy ZA-27 syntactic foam. Submitted to the Journal of Materials.

CE

[26] Zinc aluminium alloys-ZA27. https://www.azom.com/article.aspx?ArticleID=9286

AC

[27] A. Daoud, Synthesis and characterization of novel ZnAl22 syntactic foam composites via casting, Mater. Sci. Eng. A. 488 (1-2) (2008) 281-295. https://doi.org/10.1016/j.msea.2007.11.020

[28] G. Chen, M. Tong, Z. Zhu, Study on the macrosegregation of aluminium in centrifugalcast ZA27 alloy, Mater. Sci. Eng. A. 265 (1-2) (1999) 306-309. https://doi.org/10.1016/S09215093(98)01087-9

[29] M. Taherishargh, I.V. Belova, G.E. Murch, T. Fiedler, Pumice/aluminium syntactic foam, Mater. Sci. Eng. A. 635 (2015) 102-108. https://doi.org/10.1016/j.msea.2015.03.061 [30] M. Taherishargh, I.V. Belova, G.E. Murch, T. Fiedler, On the mechanical properties of heat-treated expanded perlite–aluminium syntactic foam, Mater. Des. 63 (2014) 375-383. 30

ACCEPTED MANUSCRIPT https://doi.org/10.1016/j.matdes.2014.06.019

[31] A. H. Brothers, D. C. Dunand, Mechanical properties of a density-graded replicated aluminium foam, Mater. Sci. Eng. A. 489 (1-2) (2008) 439-443. https://doi.org/10.1016/j.msea.2007.11.076

PT

[32] Y. Hangai, T. Utsunomiya, O. Kuwazuru , S. Kitahara, N. Yoshikawa, Deformation and plateau region of functionally graded aluminium foam by amount combinations of added blowing agent, Materials. 8 (10) 7161-7168. https://doi.org/10.3390/ma8105366.

RI

[33] T. Altinbalik, Y. Can, An upper bound analysis and determination of the barrelling profile in upsetting, Indian. J. Eng. Mater. S. 18 (2011) 416-424.

SC

[34] I.N. Orbulov, Compressive properties of aluminium matrix syntactic foams, Mater. Sci. Eng. A. 555 (2012) 52-56. https://doi.org/10.1016/j.msea.2012.06.032

AC

CE

PT E

D

MA

NU

[35] P. J. Pritchard, Fox and Mc Donalds Introduction to Fluid Mechanics, eighth ed., John Wiley & Sons, Inc. USA, 2011.

31

ACCEPTED MANUSCRIPT Author contribution

AC

CE

PT E

D

MA

NU

SC

RI

PT

Mr Mohavedi manufactured all samples, conducted the compression tests and wrote the initial draft of the paper. Prof. Murch co-wrote and edited the final draft of the manuscript. Prof. Belova assisted with the experimental design and data evaluation. A/Prof. Fiedler supervised the project, contributed to the initial draft and co-wrote the final draft of this paper.

32

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

Graphical Abstract

33

ACCEPTED MANUSCRIPT Highlights A novel functionally graded metal matrix syntactic foam (FG-MSF) was manufactured.



Two different particles were used in a tailored arrangement to produce FG-MSF.



The density of the FG-MSFs is between the densities of uniform MSF.



The mechanical properties of the FG-MSF are governed by the contained layers.



FG-MSF showed a higher energy absorption in comparison to uniform MSFs.

AC

CE

PT E

D

MA

NU

SC

RI

PT



34