yttrium segregation and intergranular defects in alumina
TRANSCRIPT
This article was downloaded by: [Thammasat University Libraries]On: 08 October 2014, At: 23:07Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Philosophical MagazinePublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/tphm20
Yttrium segregation and intergranulardefects in aluminaDaniele Bouchet a , Sylvie Lartigue-Korinek b , Regine Molins c &Jany Thibault da Laboratoire de Physique des Solides UA CNRS 8502, Bât 510 ,Université Paris-Sud , 91405 – Orsay, Franceb CECM-CNRS , UPR 2801, 94000 Vitry sur Seine, Francec Centre des Matériaux , Ecole des Mines , BP 87, 91003 – Evry,Franced Laboratoire TECSEN , UMR 6122 CNRS, Case 262–Université Aix-Marseille III , Paul Cézanne–Faculté des Sciences et Techniques deSt Jérôme–13397 Marseille, FrancePublished online: 21 Aug 2006.
To cite this article: Daniele Bouchet , Sylvie Lartigue-Korinek , Regine Molins & Jany Thibault(2006) Yttrium segregation and intergranular defects in alumina, Philosophical Magazine, 86:10,1401-1413, DOI: 10.1080/14786430500313804
To link to this article: http://dx.doi.org/10.1080/14786430500313804
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
Philosophical Magazine,Vol. 86, No. 10, 1 April 2006, 1401–1413
Yttrium segregation and intergranular defects in alumina
DANIELE BOUCHET*y, SYLVIE LARTIGUE-KORINEKz,REGINE MOLINS} and JANY THIBAULT�
yLaboratoire de Physique des Solides UA CNRS 8502, Bat 510,Universite Paris-Sud, 91405 – Orsay – France
zCECM-CNRS, UPR 2801, 94000 Vitry sur Seine – France}Centre des Materiaux, Ecole des Mines, BP 87, 91003 – Evry – France
�Laboratoire TECSEN, UMR 6122 CNRS,Case 262 – Universite Aix-Marseille III, Paul Cezanne – Faculte des Sciences et
Techniques de St Jerome – 13397 Marseille – France
(Received 29 April 2005; in final form 19 August 2005)
Intergranular segregation of yttrium is investigated in a rhombohedral twin grainboundary of alumina. The deviation from the twin orientation is compensatedby a periodic arrangement of intergranular dislocations. TEM tools (CTEM andHREM, EDXS, EFTEM and EELS/ELNES) have been used to characterizechanges in chemical and electronic environments along this twin. Withinthe experimental limits, no Y is detected in the perfect twin parts. On theother hand, Y segregation occurs very locally in the dislocation cores on aboutfour atomic planes perpendicularly to the GB, which in fact correspondsto the step height associated with the interfacial dislocations. By comparisonwith an undoped bicrystal, both the dislocation distribution and theY segregation localization confirm the influence of Y on the dislocationmobility. Furthermore, in the Y-rich defects, high spatial resolution analysis ofthe energy loss near edge structures (ELNES) of the Al-L23 absorption edgebrings some enlightenments about the Al3þ cation environment, provided theeffects of local radiation damage are fully considered and under control.
1. Introduction
The utmost importance of grain boundaries and intergranular segregations in thevarious macroscopic properties of polycrystalline materials is undoubtedlyrecognized but not fully understood from the origin point of view. The effect ofsegregation, whatever the property, is mainly conditioned by the inducedchanges of bonding in the structural units of the grain boundaries. Nevertheless,up until now, very few papers have been devoted to the electronic and atomiccharacterisation of intergranular dislocations [1].
Our study is motivated by the predominant role of yttrium on severalproperties of alumina. In particular, the creep rate is strongly reduced with Y doping
*Corresponding author. Email: [email protected]
Philosophical Magazine
ISSN 1478–6435 print/ISSN 1478–6443 online # 2006 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/14786430500313804
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
[2, 3]. The rather low solubility limit of Y in bulk alumina [4] leads to segregation ofthis element in the GBs [5–7]. This large sized cation Y3þ is substituted to Al cationand induces local distortion of the oxygen ion lattice [8]. Besides, the Y segregationdecreases the GB diffusivity and thus impedes the intergranular dislocation climbmotion. In polycrystals, the intergranular Y content has been found to depend on theorientation of the GB plane [9].
With this in mind, the purpose of the present work is to deepen our knowledgeof the atomic and chemical arrangement of an yttria-doped �-alumina bicrystal.Up until now, the structure of the rhombohedral undoped-twin boundary, eitherfabricated by diffusion bonding procedure or issued from mechanical twinning, hasbeen widely investigated by HRTEM [11–14]. It is described by a dense commonð01�112Þ plane, and a common ½0�1111� direction in the GB plane which is based oncoincident empty octahedral aluminium sites. On this basis, four different structuralmodels characterized by different translation states have been proposed. They pro-vide identical simulated images, which describe perfectly the experimental HRTEMone. Among these four models, it has been found that the predicted structure pre-sents a 2-fold screw symmetry axis including a glide component and leads to the leastdistorted Al–O bonds at the interface [10, 12]. On the basis of the coupled analysisof HRTEM image and O-K ELNES, the simulations using local density functionaltheory (LDFT) [13] and lattice static calculations [14] confirm the lowest energyof this screw twin configuration which corresponds to the S(V) structure inreference [15].
Taking into account an experimental sensitivity limit of 0.2 atom/nm2, until nowneither yttrium [6] nor other impurity (Ca, Si) [16] segregation has been detected bySTEM-EDXS analyses in such a ð01�112Þ facet of the rhombohedral twin. Anyway, abinitio calculations of segregation revealed a propensity for yttrium to segregate on Alsites along this twin [17]; this propensity varies with the different modelledstructures [18]. The apparent contradiction between experiment and calculation isattributed for kinetic reasons, by the authors, as the rhombohedral twin GB struc-ture is very close to the bulk structure.
In this paper, we present the analysis of a rhombohedral twin slightlydeviated from the exact coincidence, thus containing intergranular defects. We willconcentrate mainly on the Y-doped bicrystal in order to localise the dopant andto emphasize the atomic environment.
2. Materials and experiments
The alumina bicrystals are prepared by diffusion bonding and yttrium is laser-deposited before bonding [19]. The bonding temperature is 1700�C for 30min. Thespecimen is first thinned by the mechanical tripod technique, then made electron-transparent by a soft argon ion milling (DUO-MILL Gatan at 3 kV with an Arþ
0.5mA current). A protective amorphous carbon layer is evaporated on both faces ofthe foil, which is finally baked under high vacuum (10�6 Torr) in order to preventfurther contamination in the microscopes.
Four different transmission electron microscopy (TEM) techniques have beenused on four different microscopes dedicated to each technique: conventional TEM
1402 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
on a JEOL 2000EX, HRTEM (High Resolution Transmission Electron Microscopy)on a Topcon ABT running at 200 kV, EDXS (Energy Dispersive X-raySpectroscopy) on a TECNAI F20/ST in STEM mode, EFTEM (Energy FilteredTEM) on a JEOL 3010-LaB6 equipped with a GIF (Gatan Imaging Filter) andEELS/ELNES (Electron Energy Loss Spectroscopy/Energy Loss Near EdgeStructure) on a STEM-VG HB501/FEG. Each of them give complimentaryand essential information such as the geometry of the GB, the description ofdislocations, intergranular dopant content and extent, localization of dopant andatomic environment in the different parts of the studied GB. The four-implementedTEM methods are used on the same thin foil. Careful interest and attention are givento the various radiation damage problems occurring especially in the case of fineand intense probes.
3. Results and discussion
3.1. GB structure and intergranular defects
The structure of the studied Y-doped GB is considered by reference to the structureof the undoped bicrystal [12]. Both undoped and doped GBs present a similar tiltdeviation from the rhombohedral twin orientation. This deviation is accommodatedby a periodic array of two types of dislocations alternately distributed and quasi-parallel to the ½2�11�110�1 axis (figure 1). Their Burgers vectors and associated stepheights are determined by the King and Smith method on HRTEM images [20].
The two Burgers vectors of the coupled dislocations are ~bbx ¼ ~bb1 þ ~bb2 and ~bby ¼ ~bb3(figure 1a), knowing that ~bb1, ~bb2, ~bb3 are translation vectors of the �7 GB [21]. Thedistance between ~bbx and ~bby is 26 nm in agreement with the 0.8� angular deviationfrom the coincidence orientation. The ~bb2 and ~bb3 Burgers vectors have a same
a
b
Figure 1. CTEM images of the rhombohedral twin GB in alumina: (a) the undoped bicrystalcontains a periodical array of parallel coupled dislocations. (b) Only some parts of theY-doped bicrystal (white curly brackets) contain arrays similar to the one in (a). The otherparts of the GB contain extrinsic dislocations with different Burgers vectors (one of themis white arrowed).
Yttrium segregation and intergranular defects in alumina 1403
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
component but with alternate sign along the ½2�11�110�1 axis such that ~bb2 þ ~bb3 is normalto the GB plane; this results in the long-range stress field cancellation. The vectors ~bbxand ~bby are associated with geometrically necessary GB steps [20] whose heightsare 1.75 and 1.39 nm respectively (i.e. 5 d and 4 d perpendicularly to the GB).The ~bb1 Burgers vector is contained in the GB plane, so the ~bb1 distribution accountsfor the vicinality of the original bounded surfaces resulting in a non-symmetricalGB plane.
The main modification of the GB due to the Y-doping (figure 1b) is the presenceof numerous dislocations with several Burgers vectors that lead to a non-equilibriumconfiguration. In the Y-doped bicrystal, the dislocations distribution is non strictlyperiodic and some areas exhibit extrinsic dislocations whose Burgers vectorsare various combinations of ~bb1, ~bb2 and ~bb3 with different associated step heights.As a result, a long-range stress field may be present.
In the next part, the Y segregation is investigated by EDXS along the GB withspecial focus on the specific dislocation array found in the Y-doped bicrystal.
3.2. Yttrium segregation at intergranular dislocations by EDXS
For the EDXS, EFTEM and EELS analyses and unlike for the CTEM observationsof figure 1, the GB has to be oriented parallel to the electron beam. The normal ofthe foil being ½2�11�110�1, the dislocations are also seen almost parallel to the beam. Theoverall elemental analysis and Y distribution in the GB is performed with the follow-ing procedure. First, spectra taken from rectangular scanned area (box size of200 nm by 5 nm) in the grains and along the grain boundary confirm the purityof the material: no other significant minor element is detected, except copper comingfrom the grid, argon coming from ionic thinning and carbon coming from accidentalcontamination. The Y-K� peak is only detected when the box is centred on the grainboundary, given that no yttrium enrichment is observed in the adjacent aluminamatrix. Secondly, in order to locate yttrium in the GB, a 1 nm probe is scannedwith 1 nm steps along and across the GB around a defect. The segregation does notlie continuously in the GB: within an experimental detection limit of 0.2 at.%Y, no Yis detected in the perfect twin parts, while strong Y enrichment is clearly evidencedonly at isolated dislocations (figure 2a). These two line scans – parallel and perpen-dicularly to the GB – allow to localize the distribution of the Y at the defect(figure 2b). The measured dimensions of the Y segregated area, defined as theFWHM of the obtained intensity Y-K� profiles adjusted for probe size and beambroadening, is about 4 nm along the GB and 2.5 nm perpendicularly to it. Theapparent 4 nm length comes from the slight inclination of the dislocation aboutthe projection axis. The relative elemental concentrations are extracted from thespectra using the Tecnai software (Tecnai Imaging and Analysis). Considering thatthe foil thickness is about or less than 50 nm, no absorption correction is made.Given the used probe size in this experiment, the Y result is slightly underestimateddue to the small tilt of the dislocation. A concentration value is thus estimated in therange of 0.70 at % Y in the core of the defect. To be compared to the results foundin the literature, this value has to be expressed in Y density per GB surface unity,which assumes that Y lies in one intergranular plane. This will be experimentallyconfirmed in the section 3.3. Then, a rough estimation of the areal Y density
1404 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
Inte
nsity
(a.
u.)
Energy (keV)
0 155 20
Y-Kα
Cu-Kα
Dislocation
Twin
0
0,1
0,2
0,3
0,4
0,5
0,6
0 5 10 15 20
Y a
t %
2.5 nm
b
a
0
0,1
0,2
0,3
0,4
0,5
0,6
0 2 4 6 8 10 12
Y a
t%
Distance (nm)
4.5 nm
c
Distance (nm)
Figure 2. EDXS analysis: (a) spectra showing the presence of Y segregation in theintergranular defect and not in the rhombohedral twin parts of the GB. Y profiles across(b) and along (c) the GB allow the measurement of the Y extent in the defect (4 nm� 2.5 nm).
Yttrium segregation and intergranular defects in alumina 1405
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
calculation [6] needs a convolution of the Y content result in the probed defect by thesite density of Al in alumina and the width of the analyzed area (determined mainlyby beam size). The final value is 1.70� 1 at/nm2. By comparison with previous resultsfrom other authors, our value is closed to the mean value of Y segregation in generalGBs in polycrystals, which is found to be 1.75 atom/nm2 whereas the largest valuemeasured is 3.20 atom/nm2 [6]. This last value corresponds to the coverage of 1/4monolayer.
3.3. Yttrium location in the dislocation core by EFTEM
It has to be noticed that, as mentioned above, the intergranular dislocation (figure 3a)is linked to a large step and its associated line is slightly inclined with the ½2�11�110�1projection axis. As a consequence, any corresponding EELS signal is very low andvery difficult to detect, even considering the very sharp Y-L23 white lines. Moreover,the high energy loss of these Y-L23 white lines (�2100 eV) implies very long record-ing times of the image series which can promote radiation damage and other shiftproblems. Consequently, the Y-N23 edge which lies at much lower electron energyloss (37 eV) is used [22]. One series of energy-filtered images is recorded made offifteen filtered images 256 pixels� 256 pixels in size, every 2 eV from 15 eV to 43 eVaround Y-N23 edge with a slit of 2 eV. The exposure time is 1 s per image. UsingDigital Micrograph (Gatan Inc., USA), the images of the series are shown on figure3(a) after alignment. Two line spectra are extracted from the EFTEM series: figure3(b) corresponds to the line spectrum extracted across the perfect GB interface; inthis case, each spectrum corresponds to a rectangular averaging area (box size:30� 3pixels i.e. 3.4 nm� 0.34 nm) and the step between two spectra is 0.34 nm.Figure 3(c) shows the line spectrum extracted across the dislocation core; in thiscase, the sampling size and step are different: each spectrum corresponds to anaveraging area, i.e. (1.1 nm� 0.11 nm), the step between two spectra being 1 pixel.Figure 3(d) exhibits two spectra: one spectrum is recorded in the matrix (full line)and the other one results from the sum of 3 spectra recorded at 3 different spots onthe dislocation core projection (dashed line) to enhance the signal to noise ratioaround Y-N23 edge. On all the spectra (figure 3b, c, d), the large peak at 25 eVcorresponds to the bulk plasmon of �-Al2O3. The Y-N23 edge at 37 eV is onlydetectable at the dislocation (figure 3c). The intensity of the Y-N23 edge is insufficientto perform the most accurate localization. A better Y-N23 edge contributionis obtained by summing spectra recorded at different spots along the defectprojection (figure 3d). The number of spectra affected by this feature on figure 3(c)is about 10, i.e. 1.1 nm corresponding to the dislocation core localisation. Theapparent length parallel to the GB comes from the dislocation tilting with respectto ½2�11�110�1 axis.
Additional information can be drawn from the broad shoulder whose onset isat 32 eV and which is characteristic of the �-Al2O3 low loss area. It has to bepointed out that this feature is slightly less intense where Y is present, i.e. in thedislocation core. This might be due to an irradiation effect which decreases thethickness by a very local abrasion under the beam. However, such a local effect israther unlikely to occur under the parallel beam used in EFTEM experiment.On the other hand, could this 32 eV shoulder be characteristic of the defect’s
1406 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
structure? To answer this question, figure 4 shows some low loss spectra whichhave been previously recorded on the STEM VG microscope. These are referencespectra or ‘‘fingerprints’’ obtained from several alumina phases at the samethickness of 50 nm: �-Al2O3, amorphous a-Al2O3 and �-Al2O3. The 32 eV shoulderclearly appears only for the �-Al2O3 phase but never for the other alumina phasessuch as the amorphous a-Al2O3 and �-Al2O3 and more generally when there isless than six O anions around Al cation (figure 4). Therefore, the observationof a flattening of the 32 eV shoulder on the defect introduces the hypothesisof a decrease of the Al coordination in the defect as compared to theadjacent grains and this will be carefully analyzed on the Al-L23 edge in thenext part.
0.348 nm
25 eV
box (1.1 x 0.11 nm)
15 20 25 30 35 40 45
E (eV)
Inte
nsit
y [a
rb. u
nits
]
- - - - - - dislocation core______ matrix
a
b c dY-N23
Figure 3. EFTEM experiments: (a) Image extracted from a filtered series recorded at anenergy loss of 25 eV with a 2 eV slit. It has to be pointed out the excellent contrast preservationof the HREM image even in these conditions. (b) and (c) show line spectra between 15 eV and43 eV extracted from the EFTEM series. (b) is across the perfect GB interface and (c) isacross the dislocation; about 10 spectra exhibit the Y-N23 edge at 37 eV. The ‘‘width’’ of thedislocation core corresponds to 1 nm which is in fact of the order of the step height associatedwith the dislocation core (see a)). (d) The comparison between a spectrum recorded in thematrix (full line) and one resulting from the sum of 3 spectra recorded at 3 different spotson the dislocation core projection (dotted line) enhances the signal to noise ratio aroundY-N23 edge.
Yttrium segregation and intergranular defects in alumina 1407
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
3.4. Al 3Q environment in the yttrium segregated dislocation core by ELNES
The atomic environment change of Al3þ due to Y segregation in the GB is investi-gated by EELS/ELNES. Few studies have been devoted to isolated dislocation
by these techniques [1]. In our case, the ELNES analysis of Y and O in such sub-
nanometric areas is very tricky because of the weakness of these signals. For differentreasons, it is not possible to extract positive information from the ELNES of any of
the three Y edges (Y-M, Y-N, Y-L). As for oxygen environment, we believe thatreliable variations of O-K ELNES are very difficult to obtain with a good signal to
noise ratio compatible with the absence of any radiation damage.About ten intergranular dislocations have been analyzed. For most of them, the
extraction of the dislocation spectrum by processing data does not lead to any
specific ELNES. In some other cases, a specific defect spectrum can be obtainedwhich is not valid because of a suspicion of radiation damage. Finally, for a
few dislocations, the Al-L23 edge is not the same as the one in the bulk and in the
twin part. We present and discuss these results.On the HAADF (high angle annular dark field) image of the GB, the dislocations
appear as white dots (figure 5a) and the perfect twin parts have no contrast at all.
These bright patches may be attributed to the Z-contrast intensity of Y, consistentwith the segregation of this element, although the 50mrad inner ADF angles
used here cannot completely exclude a diffraction/strain contrast contribution.The apparent size of a dislocation is 4.5 nm� 2.5 nm, similarly to EDXS results.
In a first step, over the energy range containing the Al-L23 edge, an image-
spectrum [23] (dprobe¼�1 nm, 16 pixels� 16 pixels in size, 1 pixel¼ 0.5 nm, recordingtime¼ 50ms/spectrum, 0.2 eV/channel) is recorded on the intergranular defect
0 10 20 30 40 50 60 70
Inte
nsity
(a.
u.)
Energy loss (eV)
alpha
gamma/amorphous
Figure 4. Low loss reference spectra of several alumina phases recorded at the samethickness (t¼ 50 nm) on a STEM-VG. Only one spectrum is shown for the �-Al2O3 and�-Al2O3 spectra which are similar.
1408 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
shown on figure 5(a). A chemical map of the Al atoms, extracted from this image-spectrum, does not detect any Al content variation between the grains and the twinpart of the GB but clearly evidences an Al deficiency. The area of this deficiencymatches well the HAADF image of the defect (see figure 5a). Although this Aldeficiency is not possible to quantify owing to diffraction contribution, this resultis coherent with the hypothesis of the substitution of some Al3þ cations by Y3þ in thedislocation core.
26 nm
Image-
HAADF Al map
8 nm
a
b
Figure 5. EELS experiments: (a) The main high angle annular dark field image shows therhombohedral twin and the defects separated by 26 nm (see figure 1). The black square definesthe area of the image-spectrum around a defect (16� 16 pixels, 1 pixel¼ 0.5 nm, 50ms/spectrum). The Al map indicates a decrease of this element in the Y rich defect. (b) Bulkand dislocation Al-L23 absorption edges extracted from a line-spectrum across the intergra-nular defect (32 spectra, t¼ 200ms, step¼ 0.28 nm). (c) Processing data by both MSAand spatial difference allow to extract the defect spectrum (the combination of the energyvalues of the ELNES indicates a spinel type structure) and the absence of any radiationdamage during the recording of the line-spectrum (i.e. the difference spectrum between thelast and the first spectrum is flat).
Yttrium segregation and intergranular defects in alumina 1409
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
In a second step, Al bonding variations between bulk, perfect twin parts andintergranular defects are studied by monitoring the spatially resolved ELNESchanges on the Al-L23 edge starting from the onset at 76 eV and extending overthe next 20 eV. This provides information about the local symmetry predominatedby oxygen coordination around the central Al atom. The signal to noise ratios ofthe mentioned above spectra obtained with a recording time of 50ms is good enoughto estimate a chemical composition variation in an image-spectrum, but unfortu-nately not to analyze the ELNES. In this study, we are particularly careful about thespecific radiation damage problems which are known to occur preferentially at GBsand defect areas [24]. With this aim in mind, the recording time is increased upto 200ms which is a good compromise given that a longer time of 500ms is toomuch damaging. Furthermore, we have not used the image-spectrum tool whichinduces very strong cumulative charging effect in insulators [24]. A less damagingline-spectrum (32 spectra, t¼ 200ms, step¼ 0.28 nm) is recorded across the defectinstead.
According to the intensity profile, the defect is ‘‘localized’’ on four pixels perpen-icularly to the GB; this corresponds to �1.2 nm in size close to the EFTEM result.Two well-known data processing techniques are used to extract the weak and specificdefect spectrum: spatial difference [25] and multivariate statistical analysis (MSA)[26]. The two methods give the same result which is the dislocation spectrum dis-played on figure 5(c). It is characterized by the ELNES features located at 77, 78.3,79.5, 84 and 99 eV. In the absence of any reliable knowledge and useful simulation ofthis complex Al-L23 edge, the only way to outline an interpretation of this spectrum isto compare it to Al-L23 fingerprints. Such fingerprints have been established forseveral alumina phases and different Al coordinations at the same energy resolution[24] and for various Y-Al-O compounds [27]. First of all, the Al–L23 ELNES obtainedfor the defect are completely different from any of the Al–Y–O known compounds.The careful comparison between the dislocation spectrum (figure 5c) and availableAl-L23 ELNES energy values reported on table 1 from [24] brings out the followingimportant points: firstly, the defect structure is not amorphous but well ordered andsecondly, Al cation is no more 6-fold coordinated by oxygen atoms (Alocta): we can
c
Figure 5. Continued.
1410 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
assert that it is neither purely 4-fold coordinated (Altetra) nor 5-fold coordinated(Alpenta). The amorphous structure would be recognized mainly by the absence ofany peak at about 84 eV and by the presence of a peak at 80 eV and not at 79.5 eV.Similarly, a pure 4-fold Al coordination would be characterized by a strong peak at86 eV and a sharp peak at 80 eV, which is not the case here. Thus, the dislocationspectrum is not a simple linear convolution of Altetra-L23 and Alocta-L23 spectra. Onthe other hand, the dislocation spectrum fits well with the fingerprints spectra ofAl-L23 in either �-Al2O3 or �-Al2O3, thus indicating a combination of Altetra andAlocta in a spinel type arrangement. The substitution of Y3þ cations to Al3þ ones inthe defect induces a change in the atomic environment and part of the Al cations maybecome under coordinated. At the moment, we are not able to distinguish between thefingerprints of the two spinels �-Al2O3 and �-Al2O3 which differ by the ratio Alocta/Altetra. In the literature, several electronic structure calculations on special GB [28, 29]and EELS experiments on general GBs conclude also to a partial 4-fold Al coordina-tion [30]. Our result emphasizes the wealthy and complex information delivered byELNES showing that the atom coordination is only one of the numerous parameterssuch as charge transfers, bond lengths, electronic configuration and so on, that areneeded for their interpretation.
Furthermore, given that the segregated cation Y3þ is oversized (nearly twice theradius of Al3þ), bondings between Al and O may be distorted as found by calcula-tions [18]. Is this distortion detectable on the shape of the Al-L23 edge? We haveshown before [24] that the small bump preceding the threshold of the edge andlocated at 77 eV is not related to 4-fold coordination as often mentioned in theliterature [25] but undoubtedly due to an amorphization or to some slight intrinsicdisordering of the analyzed area. Besides, this 77 eV ELNES often appears in thecase of damaged GB structures by a local and surreptitious radiation effect.Moreover, as far as radiation damage is concerned, an ELNES experiment is reliableonly if one has checked that the bulk spectrum is strictly identical at the beginningand at the end of the line-spectrum, because of the above mentioned charge effects.All these considerations support the idea that no ELNES variations across the twin�7 part of the GB is detected, indicating Al3þ in octahedral sites, as shown before[6]. Nevertheless, such results obtained on a twin with an electron probe larger than
Table 1. Energy values over the first 11 eV above threshold of the ELNES features of Al-L23
fingerprints for six different atomic environment of Al atoms [24]. The results of this studyabout the ELNES of the intergranular defect is added showing clearly the similarity with
the two spinel aluminas.
Phases Al-coordination
Energy values (eV) of the ELNES features (�0.3 eV)
a b c d e f g h i j
AlPO4 4 78 80 86Al2Ge2O7 5 77.5 79.5 86�-Al2O3 6 79 80 83 85a-Al2O3 Mixed 77 80�-Al2O3 (0.3) 4þ (0.7) 6 78 79.5 84�-Al2O3 (0.5) 4þ (0.5) 6 78 79.5 84
Dislocation 77 78 79.5 84
Yttrium segregation and intergranular defects in alumina 1411
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
the GB plane are to be treated with caution. Unfortunately, increasing the spatialresolution by oversampling is too much damaging in these GBs.
4. Summary and conclusions
A combination of TEM techniques has shown that within the detection limits, Y issegregated in the dislocation cores of the rhombohedral twin �7 in alumina slightlyaway from the exact coincidence and that noY is detected along the perfect parts of theGB between the dislocations. This is in agreement with the fact that the twin GB isof low energy and with experiments already performed on intergranular defects inother system such as Ni-doped molybdenum bicrystals [31] where segregation hasbeen found to occur only in dislocation cores or along high energy GBs. In the presentstudy, the expansion of the perfect twin expected from simulations [18] and due to Ysegregation has not been reliably measured partly since no large perfect twin area canbe found due to dislocation distribution. Besides, within the detection limit, neither Y,nor any change in Al environment have been detected between dislocations. However,the localisation of Y is precisely enough determined to say that Y appears all along thedislocation line. From our experiments, Y segregation in the defect core/step is alsoaccompanied by a change in the Al cation electronic environment and noticeablyELNES indicate that the core remains locally ordered even though a small bondingdistortion is due to the strain induced by the large substituted Y cations. Finally, animportant point is that ELNES recorded on the dislocation core allow us to assert thatAl3þ environment is similar to that found in a spinel. At present, simulations of theelectronic structure of Al engaged in the alumina structures are not sufficiently refinedand thus do not permit to fully interpret this result.
Nevertheless, these electronic changes and strains sustain the basic idea thatsegregation influences the dislocation mobility as illustrated by the non equilibriumdislocation distribution in the doped bicrystals.
Acknowledgements
E.A. Stepantsov and A.L. Vasiliev from the Institute of Crystallography of Moscoware gratefully acknowledged for the preparation of the bicrystals. We are pleasedto thank C. Colliex for careful and rewarding reading of the manuscript.
References
[1] Z. Zhang, W. Sigle and M. Ruhle, Phys. Rev. B 66 1 (2002).[2] J. Cho, M.P. Harmer, H. Chan, et al., J. Am. Ceram. Soc. 80 1013 (1997).[3] S. Lartigue-Korinek, C. Carry and L. Priester, J. Europ. Ceram. Soc. 22 1525 (2002).[4] J.D. Cawley and J.W. Halloran, J. Am. Ceram. Soc. 69 C-195 (1986).[5] P. Gruffel and C. Carry, J. Eur. Ceram. Soc. 11 189 (1993).[6] M.A. Gulgun, V. Putlayev and M. Ruhle, J. Am. Ceram. Soc. 82 1849 (1999).[7] C.M. Wang, G.S. Cargill, H.M. Chan, et al., Acta Mater. 48 2579 (2000).[8] M.K. Loudjani and R. Cortes, J. Eur. Ceram. Soc. 77 349 (1994).[9] D. Bouchet, F. Dupau and S. Lartigue-Korinek, Microsc. Microanal. Microstruct. 4 561
(1993).
1412 D. Bouchet et al.
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014
[10] T. Geipel, K.P.D. Lagerloff, P. Pirouz, et al., Acta Mater. 42 1367 (1994).[11] F.R. Chen, C.C. Chu, J.Y. Wang, et al., Philos. Mag. A 72 529 (1995).[12] S. Lartigue-Korinek and S. Hagege, in Intergranular and Interphase Boundaries in
Materials (Trans Tech Publishing Ltd., Prague, Czech Republic, 1998), p. 281.[13] S. Nufer, A.G. Marinopoulos, T. Gemming, et al., Phys. Rev. Lett. 86 5066 (2001).[14] H. Nishimura, K. Matsunaga, T. Saito, et al., J. Am. Ceram. Soc. 86 574 (2003).[15] A.G. Marinopoulos and C. Elsasser, Acta Mater. 48 4375 (2000).[16] T. Gemming, S. Nufer, W. Kurtz, et al., J. Am. Ceram. Soc. 86 581 (2003).[17] S. Fabris and C. Elsasser, Acta Mater. 51 71 (2003).[18] C. Elsasser and A.G. Marinopoulos, Acta Mater. 49 2951 (2001).[19] A.L. Vasiliev, E.A. Stepantsov, Z.G. Ivanov, et al., Interface Science 5 223 (1997).[20] A.H. King and D.A. Smith, Acta Cryst. A 36(3) 335 (1980).[21] H. Grimmer, R. Bonnet, S. Lartigue, et al., Philos. Mag. A 61 494 (1990).[22] Ch. Grigis and S. Schamm, Ultramic. 74 159 (1998).[23] C. Jeanguillaume and C. Colliex, Ultramic. 28 252 (1989).[24] D. Bouchet and C. Colliex, Ultramic. 96 139 (2003).[25] J. Bruley, Microsc. Microanal. Microstruct. 4 23 (1993).[26] N. Bonnet, N. Brun and C. Colliex, Ultramic. 77 97 (1999).[27] M.A. Gulgun, W.J. Ching and M. Ruhle, Materials Sciences Forum 294–296 289 (1999).[28] P.R. Kenway, J. Am. Ceram. Soc. 77 349 (1994).[29] S.D. Mo, W.Y. Ching and R.H. French, J. Am. Ceram. Soc. 79 627 (1996).[30] K. Kaneko, T. Gemming, I. Tanaka, et al., Philos. Mag. A 77 1255 (1998).[31] J.M. Penisson and T. Vystavel, Acta Mater. 48 3303 (2000).
Yttrium segregation and intergranular defects in alumina 1413
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
23:
07 0
8 O
ctob
er 2
014