On the biodegradation performance of an Mg–Y–RE alloy with various surface conditions in simulated body fluid

On the biodegradation performance of an Mg–Y–RE alloywith various surface conditions in simulated body fluid

Anja C. Hanzi, Petra Gunde, Michael Schinhammer, Peter J. Uggowitzer *

Laboratory of Metals Physics and Technology, Department of Materials, ETH Zurich, 8093 Zurich, Switzerland

Received 27 March 2008; received in revised form 25 June 2008; accepted 31 July 2008Available online 14 August 2008

Abstract

This study documents the influence of di!erent surface conditions produced by various heat treatments on the in vitro degradationperformance of an Mg–Y–RE alloy (WE43) investigated by immersion in simulated body fluid. WE43 samples were, respectively (i)annealed at 525 !C (plus artificial aging at 250 !C in one case) and afterwards polished; and (ii) polished, annealed at 500 !C in airand subsequently investigated in the oxidized state. Thermogravimetric analysis (TGA) indicates a mass gain during oxidation in air,following a square-root law over time. X-ray di!raction spectra imply a growing Y2O3 layer upon oxidation, and Auger electron spec-troscopy depth profiles show an increased oxide layer thickness which develops according to the behavior observed by TGA. Macroscop-ically, the degradation performance of the di!erently heat-treated samples can be divided into two groups. Annealed and polishedsamples show a fast and homogeneous degradation which slows with time. Their degradation behavior is approximated by a paraboliclaw. Oxidized samples exhibit a slow initial degradation rate which increases when the protection of the oxide layer is reduced. Overall,they reveal a sigmoidal degradation behavior. Here the di!ering degradation performances of the annealed–polished and the oxidizedsamples are related to the di!erent surface conditions and explained on the basis of a depletion hypothesis." 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Magnesium biodegradation; Simulated body fluid; Surface conditions; Oxide layer; Yttrium depletion

1. Introduction

In the field of bioabsorbable implant materials, magne-sium has become an interesting candidate as a bioactivesubstance which naturally degrades within the body. Ingeneral, bioabsorbable implants o!er the advantage thatthey need not be removed in a second surgical interventionafter serving their purpose, which makes them both conve-nient for the patient and more economical. Magnesium hasgreat potential as a bioabsorbable material: it is an essen-tial element in the human body, it degrades in aqueoussolutions and its mechanical properties are more advanta-geous than those of degradable polymeric materials. How-ever, to be suitable bioabsorbable implant material, Mgalloys must possess certain properties: (i) appropriate

mechanical characteristics at room temperature, in orderto o!er adequate mechanical support to the injured tissue;(ii) a moderate degradation performance able to assist thehealing process e!ectively and to prevent from early desta-bilization; and (iii) ions released during degradation needto be biocompatible.

At pH values below 11.5 magnesium corrodes in aque-ous solutions [1]. Its degradation can then be describedby the following electrochemical equation:

Mg! 2H2O ! Mg2! ! 2OH" !H2 #1$

The formation of OH" ions and gaseous hydrogen uponthe degradation of Mg, however, may harm injured tissue,as the pH value increases locally and gas bubbles build up.Therefore a low initial degradation rate is desirable toavoid further deterioration of the adjacent tissue. Extensiveresearch has been performed on enhancing the corrosionresistance of Mg alloys by modifying their microstructure

1742-7061/$ - see front matter " 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.actbio.2008.07.034

* Corresponding author. Tel.: +41 44 632 2554; fax: +41 44 633 1421.E-mail address: [email protected] (P.J. Uggowitzer).

Available online at www.sciencedirect.com

Acta Biomaterialia 5 (2009) 162–171

www.elsevier.com/locate/actabiomat

or their surface, including the application of conversioncoatings [2,3]. These methods, however, are often rathersophisticated and may include the release of harmful sub-stances. Heat treatments o!er a simple method for improv-ing corrosion resistance, by altering the microstructureand/or the surface condition of the alloy without addingundesired substances.

The goals of this study are twofold: (i) to investigate thepotential of di!erent heat treatments on the decrease of theinitial degradation rate of an Mg–Y–RE alloy (WE43); and(ii) to analyze the influence of surface condition on the deg-radation performance. We report on the e!ect of thermaloxidation treatments of WE43 on the formation of Y-con-taining surface films, and the corresponding change in thein vitro degradation behavior in physiological media. Thealloying system was selected because yttrium is known tobe beneficial in enhancing the corrosion resistance of Mg-alloys [4–6], and WE43 has been shown in recent studiesto meet many requirements for implant applications [7–9].

2. Materials and methods

2.1. Sample preparation and denotations

Disks of 22 mm in diameter and 3.6 mm thick were cutfrom an extruded bar. Material specifications are given inTable 1, according to ASTM designations. The whole disksurface was ground and polished (to 0.25 lm diamond pol-ish) either before or after the oxidation heat treatments,and cleaned in isopropanol in an ultrasonic bath. All heattreatments were performed in an air-circulating oven(Heraeus 170/2). Samples of one series were first heat-trea-ted at 525 !C for 6 h (solution heat-treated condition, SHT)and then water-quenched. Some of these samples wereadditionally artificially aged at 250 !C for 16 h (conditionT6). Following the heat treatments all samples of the firstseries were ground and polished. Samples of the second ser-ies were first polished, then annealed at 500 !C and left inthe ‘‘thermally oxidized” state, being covered by an oxidelayer due to the heat treatment (oxidized condition). Theoxidation in air was carried out for 1, 8, 24, 48 and 168 h(1 week), respectively, and the samples were air-cooledafter annealing. Table 2 gives an overview on the heattreatments performed and sample denotations used in thisstudy. In order to minimize the contact area between spec-

imen and holder, the samples to be oxidized were placed ona tripod of alumina sticks during heat treatments to facili-tate homogeneous oxidation of the entire sample surface.

2.2. Microstructure and characterization

The microstructures of all samples were investigatedusing optical microscopy. The oxidized samples were addi-tionally investigated by thermogravimetric analysis (TGA),X-ray di!raction (XRD) and Auger electron spectroscopy(AES). Isothermal TGA measurements (Netzsch STA449C) were carried out at 500 !C in air for 120 min andthe mass gain over time was recorded. The TGA specimensmeasured 5 mm % 5 mm % 9 mm, and were ground withSiC abrasive paper up to a grit of 4000 and degreased inisopropanol directly before measurement. For phase iden-tification of the oxidized samples XRD measurements(PANalytical X’Pert PRO-MPD) were conducted using aCu Ka1 (k = 0.15406 nm) source operated at 38 kV and45 mA. AES (Perkin-Elmer PHI 4300 SAM) was per-formed at 5 keV using an LaB6 filament. Depth profileswere recorded using an Ar ion gun operated at 4 keV.The sputter rate was 25 nm min"1 and calibrated forTa2O5. Taking into account the thicker surface layer onthe samples oxidized for 48 h and 1 week at 500 !C, thesputter rate for these samples was increased to 37.5 and75 nm min"1, respectively.

2.3. Immersion testing

The degradation performance was evaluated by immers-ing the heat-treated samples in simulated body fluid (SBF).According to Eq. (1) every dissolved mole of magnesiumproduces 1 mol of hydrogen, which makes it possible toderive the corrosion rate from the amount of hydrogenevolved. Song et al. [10] developed the hydrogen evolutionmethod, which is based on the collection of hydrogen gasduring degradation of magnesium in aqueous solution.The experimental set-up is simple: a magnesium sample isplaced in a beaker containing the immersion liquid. A mea-suring cylinder filled with the liquid is placed over the sam-ple in order to collect the hydrogen formed during thesample’s corrosion. The amount of hydrogen collected inthe measuring cylinder over time reveals the degradation

Table 1Chemical composition of WE43

Element Mass fraction (wt.%)

Y 3.7–4.3REa 2.4–4.4Zr >0.4Mg Balance

Ni <5 ppmFe <0 ppmCu <20 ppma RE consists of Nd and heavy RE (Yb, Er, Dy, Gd).

Table 2Heat treatments, sample denotations and sample series

Temperature (!C) Annealing time (h) Label Sample series

525 6 SHT pa

525 + 250 6 + 16 T6 p

500 1 ox1 oxb

500 8 ox8 ox500 24 ox24 ox500 48 ox48 ox500 168 ox168 oxa Polished.b Oxidized.

A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171 163

rate of the Mg sample. Immersion testing was performed inSBF at 37 ± 2 !C in ambient air in an incubator (WTB Bin-der). SBF is an aqueous solution which simulates the ionconcentration of human blood but does not contain largerparticles such as proteins, lipids or blood cells. The SBFused in this study was chosen because it reflects the ion con-tent of human blood very precisely, as indicated in Table 3[11]. For immersion testing, the samples were placed onPOM MultiClips (Struers). For each sample 300 ml ofSBF was used, and three samples were investigated pertreatment. The SBF was changed regularly such that thepH value did not exceed 8, in order to keep the conditionsas constant as possible.

3. Results

3.1. Microstructure

Slight di!erences in microstructural features wereobserved between the oxidized samples. The grain sizereached approximately 100 ± 50 lm after 1 h at 500 !Cand increased to values of 200–250 lm after longer anneal-ing.Mainly Nd-rich particles were identified, indicating thata significant amount of Y was dissolved in the Mg matrixupon heat treatment [12]. There are also slight microstruc-tural di!erences between the specimens of the polished sam-ple series. While solution heat-treated samples exhibitmicrostructural features similar to those observed for theoxidized samples, T6 samples reveal Y-rich particles, andthus less homogeneous distribution and lower solid solutioncontents of Y can be assumed compared to SHT or oxidizedspecimens [13]. Their grain size, however, is similar to thoseof SHT specimens. In a recent study it has been shown thatthe grain size of an Mg alloy in near-equilibrium state(wrought condition) does not significantly influence the deg-radation behavior in SBF [14]. Therefore the influence of theslightly varying grain size on the degradation performance ofthe samples investigated in this study is discounted.

3.2. Characterization of oxidized samples

Isothermal TGA measurements at 500 !C indicated thatthe mass gain due to oxidation approximately follows a

square-root law over time (data not shown here). Very sim-ilar behavior was also observed by Wang et al. [15], whoinvestigated the early oxidation behavior of Mg–Y alloysat high temperatures. XRD measurements, represented asnormalized intensity over 2h in Fig. 1, indicate Mg andY2O3 as crystalline phases on the oxidized specimens.Intermetallic phases in the bulk or other oxide phases onthe surface were either below the detection limit of theXRD device or did not occur in crystalline form. The inten-sity of the Y2O3 signal increases with increasing oxidationtime, implying a growing oxide layer on the surface. AESdepth profiles emphasize the XRD and TGA findings(Fig. 2). In the polished condition only a thin (&10 nm)MgO layer is found on the surface, which grows naturallyand is most likely rather amorphous. After thermal oxida-tion the surface layers consist mainly of O, Y and a littleMg. Other alloying elements (Nd, Zr and REs) were notdetected in the oxide layer and are not displayed inFig. 2. In contrast to the XRD measurements, AES depthprofiles also display the presence of MgO, which is foundmainly in the outermost part of the oxide layer (100–200nm) but also to a marginal extent inside the oxide layer.However, the MgO contribution to the surface layer isminor. Further, it is emphasized that the oxide layer growswith increasing oxidation duration. After 1 h at 500 !C alayer thickness of approximately 500 nm is achieved. Dur-ing further annealing at 500 !C the layer grows to a thick-ness of approximately 800 nm after 8 h, 1200 nm after 24 h,1900 nm after 48 h, and 2700 nm after 168 h, respectively.Again, a square-root law of oxide growth over oxidationtime is seen, as already observed for the initial oxidationduring TGA measurements. Since the calibration for thethickness calculation from the sputter rate was conductedwith Ta2O5, the absolute values given here may in truthvary slightly due to di!erent yields and sensitivity factors.However, this does not influence the general conclusion

Table 3Ionic content (in mmol l"1) of human blood and the SBF used in thisstudy [11]

Ions Blood SBF

Na+ 142.0 142.0K+ 3.6–5.5 5.0Mg2+ 1.0 1.0Ca2+ 2.1–2.6 2.5Cl" 95.0–107.0 109.0HCO3

" 27.0 27.0HPO4

2" 0.65–1.45 1.0SO4

2" 1.0 1.0

pH 7.3–7.4 7.3–7.4Fig. 1. XRD spectra of WE43 after oxidation in air at 500 !C for 1, 8, 24,48, respectively, and in the polished condition (F).

164 A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171

of a growing oxide layer which exhibits square-root behav-ior upon oxidation in air.

In Fig. 2, it is important to note the Y content beneath theoxide layer. Even though the sensitivity of AES for low ele-mental amounts is limited, a reduced Y content in the oxi-dized samples is recognizable, particularly for the ox1samples.

3.3. Immersion testing

The degradation performance in SBF of the di!erentlyheat-treated samples is given in Fig. 3. One curve of each

condition represents the degradation behavior of the threesamples with the same heat treatment. Generally, thereproducibility of hydrogen evolution at a certain degrada-tion time is moderate. This is mainly ascribed to the highreactivity of Mg in aqueous solutions in the neutral pHregime [16], which reduces the accurate predictability ofcorrosion attack. The relative errors of the data presentedin Fig. 3 were less than 10%. To allow a clearer picture,however, error bars were not included in the plot. Immer-sion testing in SBF revealed di!erent trends for the twosample series. Annealed and polished samples have a highinitial degradation rate, which slows with time. The hydro-gen evolution of these samples can be approximated by aparabolic law. The degradation can be characterized asrapid at the beginning (undesired initial ‘‘hydrogen burst”),but relatively homogeneous (Fig. 4a and b). The T6 sam-ples corroded slightly faster at the beginning than samplesin the SHT condition. After immersion the surface of thepolished samples was evenly rough, with a deposition ofcorrosion products [17]. The degradation performance ofthe oxidized samples indicates certain di!erences betweenthe briefly oxidized samples compared to samples oxidizedfor a longer time, but essentially displays the same charac-teristics. Brief oxidation induces a low initial degradationrate, which at some point rapidly increases with increasingimmersion time. The degradation becomes fast and ratherinhomogeneous as soon as the protective oxide layer is pen-etrated or removed (Fig. 4c–e). Pronounced localized cor-rosion sites are observed in the areas of dismantled oxidelayer, where the formation of a corrosion deposit occurssimultaneously. After long oxidation a very slow and

a

d e f

b c

Fig. 2. (a–f) AES depth profiles of WE43 (a) in the polished condition, and after oxidation in air at 500 !C for (b) 1 h, (c) 8 h, (d) 24 h, (e) 48 h and (f) 168h, respectively.

Fig. 3. Hydrogen evolution of WE43 samples in SBF at 37 !C aftervarious heat treatments.

A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171 165

homogeneous initial degradation is observed. It remainsslow as long as the oxide layer still covers a reasonableamount of the surface. Once the protective surface layeris penetrated or removed (detailed investigation of the pen-etration/degradation mechanisms by the authors is in pro-gress and is discussed elsewhere), the degradation rateincreases and sites of localized corrosion grow, but not asfast as observed in the briefly oxidized samples (Fig. 4f–h). When larger areas of uncovered metallic substrate areexposed to the immersion liquid the degradation behaviorbecomes similar to that of the briefly oxidized samples.The degradation rate is nearly constant until the surfacelayer is removed and degradation is slowed. This takeslonger than for the brief oxidation condition, however,since the oxide layer is thicker.

It is important to note that, once the oxide layer is moreor less completely removed, the overall hydrogen evolution

of the oxidized samples is greater than that of the polishedcounterparts.

4. Discussion

WE43 has been shown to be a suitable candidate forimplant application material [7–9]. This is essentially dueto its excellent corrosion resistance in aqueous solutions,which allows relatively slow and homogeneous degrada-tion. In addition, its good electrochemical properties areaccompanied by considerable mechanical properties,including an elastic modulus similar to that of human bone[8]. The good corrosion resistance of WE43 is ascribedmainly to the beneficial influence of the alloying elementY. The incorporation of oxidized Y in the surface filmwas found to be the reason for the enhanced degradationresistance of Mg–Y alloys [4,5]. In addition, Davenport

Fig. 4. (a–h) Photographs of immersed samples after di!erent immersion periods in SBF at 37 !C. (a) SHT, 1 day; (b) SHT, 7 days; (c) ox1, 1 day; (d) ox1,7 days; (e) ox1, 11 days; (f) ox24, 1 day; (g) ox24, 7 days; (h) ox24, 11 days.

166 A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171

et al. [6] reported on the advantageous influence of Y-richregions in the Mg matrix, which slow the propagation ofcorrosion. They also showed that the corrosion resistanceis enhanced by homogenizing the distribution of Y (andof other alloying elements), which could be achieved byappropriate heat treatments.

We performed thermodynamic calculations to simulatethe content of the alloying elements dissolved in the Mgmatrix (a-phase) in dependence on the temperature, usingthe Pandat software package [18]. The thermodynamic sim-ulations were conducted assuming equilibrium conditions:they indicated the highest amounts of Y in solid solutionfor temperatures above 300 !C. Below 300 !C, the contentof Y in the Mg matrix decreases significantly due to the for-mation of Y-containing precipitates [13]. According to thethermodynamic calculations, the consequences of the heattreatments performed in this study are as follows: (i) thehighest amount of Y dissolved in the Mg matrix may beexpected for the SHT and the oxidized samples; and (ii)the T6 samples feature less Y in solid solution due to arti-ficial aging at 250 !C after the heat treatment at 525 !C.

Macroscopically, the two sample series lead to two basi-cally di!erent types of degradation performance; these areschematically represented in Fig. 5a. Since degradation ofsuch Mg samples is ascribed to a statistical process, H2 evo-lution regimes rather than defined curves are used to qual-ify the degradation performance after a certain pre-treatment. The hydrogen evolution of the annealed andpolished samples can be described by a parabolic law.The characteristics are a high initial degradation rate dueto the large reactive metallic surface, which is initiallyexposed to the immersion medium (Fig. 5b). Over timethe degradation slows due to the formation of corrosionproducts at the surface and their barrier action, whichimpedes further degradation [17]. Corrosion accompaniedby surface layer formation due to depositions commonly

follows a parabolic law over time since the reaction rateis controlled by the di!usion of reaction species throughthe surface layer [19]. The higher initial degradation rateof the T6 samples compared to the SHT samples is ascribedto microstructural di!erences. Since the artificial aging partof the T6 heat treatment enhances the formation of Y-con-taining precipitates, the solid solution content of Y in theMg matrix, and thereby its protective influence on degrada-tion resistance, is diminished. The more heterogeneous dis-tribution of Y, in turn, may lead to the formation of localelements. The second-phase particles act as cathodes andinduce the active dissolution of the Mg matrix. Since theSHT samples also lack a single-phase structure, this phe-nomenon also takes place here, though it is less pro-nounced than observed for the T6 samples.

The degradation performance of the oxidized samplescan first be approximated by a sigmoidal law. Initially, avery low degradation rate is observed due to the protectiveinfluence of the oxide layer, which covers the whole surfaceof the samples (Fig. 5c). As soon as the oxide layer is pen-etrated, however, the reactive surface area is increased(Fig. 5d) and degradation accelerates. With time, the deg-radation is impeded at these sites due to the barrier actionof deposited corrosion products, while at the same time atother sites ‘‘fresh” reactive surface is exposed to the liquidfeaturing accelerated degradation. This condition persistsuntil most of the oxide layer is removed and corrosionproducts cover the degraded surface. Now degradationslows, similar to the annealed–polished samples, and tendstowards parabolic behavior.

The oxidation duration causes a delay in the onset ofaccelerated degradation with increasing oxidation time,plus a later turn-in to the parabolic behavior, but doesnot change the basic type of degradation performance.The H2 evolution regimes are also assumed to becomebroader with increasing oxidation time, due to the statistical

a b

c

d

Fig. 5. (a) Schematic degradation performances of WE43 samples after di!erent heat treatments; (b) initial surface condition of annealed and polishedsamples featuring a large reactive surface; (c) initial surface condition of oxidized samples carrying an oxide layer; (d) penetrated oxide layer upondegradation. Xi represents the reactive area fraction.

A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171 167

process of the degradation (Fig. 5a). A thick oxide layer isassumed to contain more inhomogeneities than a thin layerbecause of the increasing influence of the di!erence in ther-mal expansion coe"cients between the oxide and the metal.This in turn renders more unpredictable just when andwhere the first penetration of the surface layer will occurwhich finally generates the onset of increased degradation.

In Fig. 6a, the development of the H2 evolution curveshape of oxidized samples is illustrated. Once the oxide layerhas been removed the oxidized samples basically degradelocally via the samemechanism as observed for the annealedand polished samples. Therefore one curve represents theannealed and polished condition indicated with a delayedonset, which is ascribed to the protection by the oxide layerat the beginning of immersion. Upon penetration of the sur-face layer, the actual reactive surface area increases, and isdenoted by the reactive area fraction Xi. The reactive areafraction also significantly influences the curve developmentof the oxidized samples’ hydrogen evolution. Xi follows acharacteristic S-shaped curve from 0 to 1 and can be deter-mined by evaluating the immersed samples after variousimmersion periods. The H2 evolution curve shape of the oxi-dized samples is then the result of multiplying the annealed–polished by the Xi curve. It is characterized by a steepincrease at the beginning when the reactive area fraction isstill increasing considerably. Later the degradation mecha-nism of the annealed–polished samples becomes more pro-nounced and mainly determines the curve evolution.

Fig. 6b presents the maximal H2 evolution rate per reac-tive surface area for SHT, ox1 and ox24 samples. The errorbars indicate the scattering of the three samples per condi-tion. In the SHT samples the maximal degradation rate isobserved at the beginning of immersion, when the reactivesurface area is maximal. The slope of the curve indicatesthe degradation rate. The reactive surface area of the oxi-dized samples was determined by evaluating photographsof the immersed samples taken during every SBF change(i.e. at least every second day). The reactive surface was iden-

tified as the areas where the oxide layer was removed or obvi-ously damaged. Hence, the maximal degradation rate perreactive surface area of the oxidized samples is found at theonset of the accelerated degradation regime where the H2

evolution is increasing but only a very small reactive surfacearea exists. The onset was determined as the intersection ofthe tangents to the H2 evolution curves of the initial andthe increased degradation regimes, respectively. The indica-tion is that the maximal H2 evolution rate per reactive sur-face area is similar for the SHT and ox24 samples butsignificantly higher for the ox1 samples. Taking into accountthe small reactive surface area at the beginning of theincreased degradation regime, the amount of H2 evolvedfrom the ox1 samples is astonishingly large. This phenome-non is emphasized when the H2 evolution in the onset regionof the ox1 and the ox24 samples is observed in more detail(see small inset in Fig. 3). Compared to the ox24 samples,the ox1 samples evolve much more H2 within the sameimmersion period. The explanation for this phenomenonwill be based on a depletion hypothesis.

Fig. 7a schematically presents the Mg surface with oxidelayer in air. The Y concentration in solid solution is givenas a function of distance from the surface. The Y concen-tration in the bulk, C0, is defined. Directly underneaththe oxide layer a reduced Y concentration, Cd, is assumed,and will be reconsidered later. Ion currents and di!usioncoe"cients of the oxide and the metal are indicated inFig. 7b. Since very little data is available on di!usion coef-ficients or activation energies of Y in Mg, or Y3+ and O2"

in Y2O3, the following assumptions have been made: (i) dif-fusion is hindered more in ionic materials than in metalsdue to the lower mobility and defect density at tempera-tures around 500 !C; (ii) for the continuous formation ofY2O3, O2" di!uses through the (already formed) oxidelayer to the metal–oxide interface rather than Y3+ di!usingto the surface in contact with air. The latter assumptiondoes not influence the general conclusion of the depletionhypothesis but arranges the e!ect more clearly. It follows

sht ox1 ox240.00

0.05

0.10

0.15

0.20

H .xaM

2 rat

e mc·h/l

m[ aera evitcaer rep2 ]H2 evolution

Immersion time

Reactive area

fraction Xi

1

0

reactive areas

annealed-polished

oxidized

Fig. 6. (a) Development of hydrogen evolution curve shape of oxidized samples; (b) maximal H2 evolution rate per reactive surface area of SHT, ox1 andox24 samples.

168 A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171

that the di!usion of O2" through the oxide layer is the rate-determining step. This seems reasonable as di!usion of ionspecies in oxides is commonly determined by rather highactivation energies and low di!usion coe"cients, especiallyat the low homologous temperatures used in this study.Since the formation enthalpy of Y2O3 is approximatelythree times higher than that of MgO, a much higher oxygena"nity of Y compared to Mg results [20,21]. Hence, Y willreadily di!use towards the alloy’s surface to form Y2O3,leaving behind a Y-depleted zone in surface-near regions(see Dc in Fig. 7a). For the formation of an Y2O3 layeron the surface, the following simplified condition must befulfilled:

3JO ' 2JY #2$where JO and JY are the di!usion currents of O2" in Y2O3

and of Y in Mg, respectively. According to Fig. 7b, the dif-fusion currents can also be expressed by the followingequations according to Fick’s law:

Jo ' DO ( b ( dldx

' DO ( b ( Dld

; JY ' DY ( dcdx

' DY ( Dcn#3$

where DO and DY are the di!usion coe"cients of O2" inY2O3 and of Y in Mg, respectively, b is a dimensional con-stant, Dl is the chemical potential di!erence between theinner and outer sides of the oxide layer, d is the thicknessof the oxide layer, Dc is the concentration di!erence nearthe surface and n is the depth of the concentration gradient.Including Eq. (3) in Eq. (2), it follows that:

2DY ( Dcn

' 3Do ( b ( Dld

orDcn

' Const: ( 1d

#4$

Hence, brief oxidation of the samples generates a signif-icant Y-depletion in surface-near regions due to the consid-erable mobility of Y in Mg and the still rather shortdi!usion paths of O2" through the oxide layer (Fig. 7d).The Y approaching the metal–oxide interface is rapidly

consumed by the formation of the oxide layer: the concen-tration gradient increases Dc ' C0 " C1. Upon longer oxi-dation, however, the concentration gradient levels outDc ' C0 " Ch since now the rate-determining influence ofthe O2" di!usion through the oxide becomes more domi-nant (Fig. 7e). The di!usion paths through the oxide layerincrease, the O2" supply at the metal–oxide interfacedecreases and the concentration gradient is weakened dueto the still-high mobility of Y in Mg. For brief oxidationC0 is assumed to be lower than observed for long oxidationdue to less time being taken for the Y-rich particles to dis-solve [12] (not indicated in Fig. 7).

The AES results illustrated in Fig. 2 reveal a reduced Ycontent beneath the oxide layer and thus support the viabil-ity of the proposed depletion hypothesis. The consequenceof the Y-depletion and depletion hypothesis on the degra-dation performance of the oxidized samples is evident. Assoon as the protective oxide layer is removed, the reactivesurface is exposed to the immersion liquid. In the case ofbrief oxidation, the considerable depletion of Y in thenear-surface regions and the amount of still undissolvedY-rich intermetallics generate an enhancement of the sur-face reactivity, which is displayed in the faster accelerationof degradation in the briefly oxidized samples (small insertsin Fig. 3). This e!ect is not as pronounced for the samplesoxidized for a long time because they feature less pro-nounced Y-depletion and a higher solid solution contentof Y. Hence, the increase in degradation is slower andthe turn-in to the high corrosion rate regime takes longer.However, due to the Y-depletion, the overall hydrogenevolution of the oxidized samples is higher than that ofthe annealed and polished samples where possible deple-tion was removed by polishing the surface after heat treat-ment (Fig. 3). After very long oxidation there is onlyminimal Y-depletion underneath the oxide layer and themaximal H2 evolution rate per reactive surface approachesthe values found for the SHT samples (Fig. 6b).

c d e

a b

Fig. 7. (a) Schematic illustration of the sample surface with oxide layer and Y gradient due to depletion; (b) ion currents and di!usion coe"cients near theoxide layer; (c) Y concentration profile in the bulk; (d) Y concentration profile with high Y-depletion after brief oxidation; (e) Y concentration profile withlow Y-depletion after long oxidation.

A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171 169

Basically the surface reactivity discussed above alsoinfluences the development of the H2 evolution curve-shapeof the oxidized samples. Due to the higher Y-depletion andthe amount of Y-rich intermetallics, the surface reactivityof the briefly oxidized samples is higher than that of sam-ples oxidized for a long time. This would imply a modifica-tion in Fig. 6a that generates a steeper ascent of H2

evolution in the accelerated degradation regime of thebriefly oxidized samples. This detail was not included inFig. 6a, however, because it a!ects only a small part ofthe curve-shape and does not change the general degrada-tion behavior.

A simple calculation illustrates the depth of the depletedzone. In Y2O3 about 78 wt.% Y is present. Assuming a layerthickness of approximately 1 lm, a linear concentration gra-dient as illustrated in Fig. 7a and a depletion C0 " Cd of 2wt.%Y, the depth n is calculated to be approximately 80 lm.

It is worth mentioning that, once the corrosion modechanges from homogeneous to heterogeneous degradationupon sample immersion, prediction of the degradation per-formance is no longer possible. Severe localized corrosionprobably initiates at a certain point in immersion time onevery specimen, particularly on the polished samples,which o!er a large reactive surface area. Then the degrada-tion rate significantly increases and can no longer beapproximated by a parabolic law. Optical inspection ofthe briefly oxidized samples immersed for longer time peri-ods disclosed sites of localized attacks, which to a certainextent also explain the large amount of H2 evolved duringimmersion. However, the proposed models allow a goodprediction of the degradation performance of such or sim-ilarly treated WE43 samples, or in general of comparableMg–Y alloys, as long as there is a certain continuity andconsistency in the degradation.

It is debatable whether the di!erent modes of coolingfollowing heat treatment (water quenching for the solu-tion-annealed and polished samples in contrast to air cool-ing of the oxidized samples) further influences thedegradation behavior of the two sample series. It is notexpected, however, that a large di!erence in Y dissolvedin the Mg matrix will arise via this procedure, becauseMg–Y–RE alloys are not particularly quench-sensitive[22]. For the oxidized samples air cooling was chosen inorder to reduce the proneness to defect formation in theoxide layer due to the di!erence in the coe"cient of ther-mal expansion between metal and oxide. Further, the dif-ference in annealing temperature (525 !C for polishedsamples, 500 !C for oxidized samples) is not expected tosubstantially influence the results of this study. 525 !Cwas chosen for annealing the polished samples as this tem-perature represents a standard for the heat treatment (espe-cially T6) of WE43 [22]. For oxidation, a more universaltemperature was selected, one less close to the solidus tem-perature of the various alloys currently being investigatedby the authors.

It was initially speculated that the increase in the pHvalueto 8 upon sample degradation would reduce the corrosion

tendency so much that self-passivation would initiate,decreasing corrosion rates with increasing immersion timesas indicated by the parabolic behavior of annealed–polishedsamples. This, however, was not so: samples immersed in atitrated solution with a constant pH value featured the samedegradation behavior as the samples investigated in thisstudy. Due to the constantly lower pH value they onlydegraded a little faster but showed essentially the same per-formance (parabolic law) as described here.

Further discussion addresses the issue of whether anincreased degradation resistance accompanies increasingoxidation duration. Immersion testing with samples oxi-dized for 1 week at 500 !C indicated that these samples aremore prone to degradation than those oxidized for 24 h or48 h. Thicker oxide layers are assumed tobemore susceptibleto crack formation because the di!erence in thermal contrac-tion upon cooling has more significant consequences. Thisimplies that samples oxidized for a very long time carrymorepotential sites where corrosion can initiate, even if they nowpossess a more homogeneous distribution of Y in the Mgmatrix and a thicker oxide layer, i.e. less depletion beneaththe oxide layer. Hence, in general, decreased degradationresistance is expected after very long oxidation at high tem-peratures. This has not been taken into account in Fig. 5,however, since it is di"cult to predict the onset of theincreased degradation regime in such samples.

Thermal oxidation has proved to be an e"cient and sim-ple method for increasing the initial degradation resistanceof WE43 samples in SBF. In the context of potentialimplant applications, however, possible plastic deforma-tion during the surgical intervention also has to be takeninto account. Deformation upon placement of the implantmay harm the oxide layer and decrease its protective influ-ence. The authors address this issue in a further study [23].

Similar thermal oxidation treatments at lower tempera-tures are also expected to improve the initial degradationresistance, without altering the microstructure to the sameextent. Similar high protection by the oxide layer, however,cannot be expected because the layer will be less thick dueto the restricted mobility of the di!usion species at lowertemperatures. AES depth profiles on samples oxidized at325 and 425 !C have indicated much thinner oxide layers,as found for samples oxidized at 500 !C. Alterations indensity must also be taken into account.

The results of immersion testing emphasize that heattreatment followed by removal of the protective oxide layeris unsuitable for achieving a reduced initial degradationrate, as sought in this study. The presence of further sec-ond-phase particles and the reduced protective influenceof Y in solid solution in the Mg matrix are, in particular,believed to be the cause of the lower initial degradationresistance of T6 samples as compared to solution heat-trea-ted samples. In contrast, oxidation of the samples generatesreasonable improvement of the degradation resistance,especially at the crucial early stage of degradation whenslow degradation is most important for implantapplications.

170 A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171

5. Conclusions

The degradation performance of WE43 in SBF afterdi!erent types of heat treatment was investigated andthe potential of these heat treatments for enhancing theinitial corrosion resistance was analyzed. Annealing fol-lowed by polishing and thereby removing the surface’sthermally formed oxide layer turned out to be an unsuit-able approach, because the samples in question su!er anincreased initial degradation rate. The corrosion perfor-mance of annealed and polished WE43 exhibits parabolicbehavior, with decreased degradation over time due to thebarrier action of corrosion products deposited at the sur-face. Samples left in the thermally oxidized state, how-ever, feature enhanced corrosion resistance at thebeginning of immersion. XRD, TGA and AES measure-ments indicated the presence of MgO and crystallineY2O3 in the surface film, which grows with increasing oxi-dation duration. The decreased initial degradation rate ofoxidized WE43 is ascribed to the protective e!ect of thisoxide layer. Once the layer is penetrated or removed, deg-radation accelerates until the deposition of corrosionproducts slows further degradation. Thermal oxidationproved to be a simple method for increasing the initialdegradation resistance of WE43 in SBF. The varying deg-radation performance of the variously treated samples isconnected with the di!erent surface conditions and isexplained on the basis of a depletion hypothesis. Solutionheat-treated WE43 contains a relatively high amount ofdissolved Y in the Mg matrix whose beneficial influenceon degradation resistance is mirrored in comparably lowmaximal degradation rates. In contrast, brief oxidationgenerates reasonable Y-depletion underneath the oxidelayer, which causes significant acceleration of degradationonce the surface film has been removed. The diminishedbeneficial influence of Y in the Mg matrix is identifiedas the cause of this. Long oxidation results in lower Y-depletion, and the samples involved exhibit a maximaldegradation rate similar to that observed in theannealed–polished samples. The models proposed hereallow a good prediction of the degradation performanceof such or similarly treated WE43 samples, or of compa-rable Mg–Y alloys in general, as long as there is a certaincontinuity and consistency in the degradation.

Acknowledgements

The authors thank Pierre Elser, B.Sc, for performingimmersion testing experiments. Financial support by theAustrian Forschungsforderungsgesellschaft (FFG), thestate of Upper Austria and the Austrian Research CentersGmbH (ARC) within the frame of the Kplus research pro-grams, and by the Staub/Kaiser Foundation, Switzerland,is gratefully acknowledged.

References

[1] Pourbaix M. Atlas of electrochemical equilibria in aqueous solu-tions. Houston, TX: National Association of Corrosion Engineers;1974.

[2] Song G, Atrens A. Understanding magnesium corrosion – aframework for improved alloy performance. Adv Eng Mater2003;5:837–58.

[3] Shaw BA, Wolfe RC. Corrosion of magnesium and magnesium-basealloys. In: Cramer SD, Covino BS, editors. ASM handbook.MaterialsPark, OH: ASM International; 2005. p. 205–27.

[4] Miller PL, Shaw BA, Wendt RG, Moshier WC. Assessing thecorrosion-resistance of nonequilibrium magnesium–yttrium alloys.Corrosion 1995;51:922–31.

[5] Yao HB, Li Y, Wee ATS. Passivity behavior of melt-spun Mg–Yalloys. Electrochim Acta 2003;48:4197–204.

[6] Davenport AJ, Padovani C, Connolly BJ, Stevens NPC, Beale TAW,Groso A, Stampanoni M. Synchrotron X-ray microtomographystudy of the role of Y in corrosion of magnesium alloy WE43.Electrochem Solid State Lett 2007;10:C5–8.

[7] Mani G, Feldman MD, Patel D, Agrawal CM. Coronary stents: amaterials perspective. Biomaterials 2007;28:1689–710.

[8] Staiger MP, Pietak AM, Huadmai J, Dias G. Magnesium and itsalloys as orthopedic biomaterials: a review. Biomaterials2006;27:1728–34.

[9] Witte F, Kaese V, Haferkamp H, Switzer E, Meyer-Lindenberg A,Wirth CJ, Windhagen H. In vivo corrosion of four magnesium alloysand the associated bone response. Biomaterials 2005;26:3557–63.

[10] Song G, Atrens A, St John D. A hydrogen evolution method for theestimation of the corrosion rate of magnesium alloys. MagnesiumTechnol 2001:255–62.

[11] Muller L, Muller FA. Preparation of SBF with di!erent HCO3"

content and its influence on the composition of biomimetic apatites.Acta Biomaterialia 2006;2:181–9.

[12] Mengucci P, Barucca G, Riontino G, Lussana D, Massazza M,Ferragut R, Hassan Aly E. Structure evolution of a WE43 Mg alloysubmitted to di!erent thermal treatments. Mater Sci Eng A2008;479:37–44.

[13] Nie JF, Muddle BC. Characterisation of strengthening precipitatephases in an Mg–Y–Nd alloy. Acta Materialia 2000;48:1691–703.

[14] Wang H, Estrin Y, Zuberova Z. Biocorrosion of a magnesium alloywith di!erent processing histories. Mater Lett 2008;62:2476–9.

[15] Wang XM, Zeng XQ, Zhou Y, Wu GS, Yao SS, Lai YJ. Earlyoxidation behaviors of Mg–Y alloys at high temperatures. J AlloysCompd 2008;460:368–74.

[16] Song G, Atrens A. Corrosion mechanisms of magnesium alloys. AdvEng Mater 1999;1:11–33.

[17] Rettig R, Virtanen S. Composition of corrosion layers on amagnesium rare-earth alloy in simulated body fluids. J BiomedMater Res A, in press, doi:10.1002/jbm.a.31887.

[18] Available from: http://www.computherm.com/pandat.html.[19] Paarek VK, Ramanarayanan TA, Mumford JD, Ozekcin A, Scanlon

JC. The role of morphology and structure in the kinetic evolution ofiron-sulfide films on Fe-base alloys. Oxidation of Metals1994;41:323–41.

[20] Lavut EG, Chelovskaya NV. Enthalpy of formation of diyttriumtrioxide. J Chem Thermodyn 1990;22:817–20.

[21] Lide DR. CRC Handbook of chemistry and physics. 88th ed. BocaRaton, FL: CRC Press; 2007–2008.

[22] Elektron WE43 Wrought Alloy. Manchester: Magnesium Elektron.Report no. 478. 4 p.

[23] Gunde P, Furrer A, Hanzi AC, Schmutz P, Uggowitzer PJ. Theinfluence of heat treatments and plastic deformation on the biodeg-radation of an Mg–Y–RE alloy. J Biomed Mater Res A, accepted forpublication.

A.C. Hanzi et al. / Acta Biomaterialia 5 (2009) 162–171 171