Peculiarity of the Cathodoluminescence of Αlpha- Alumina Prepared by Calcination of Gibbsite Powder or Generated by Oxidation of a Metallic FeCrAl Alloy

Most of metallic materials functioning at high temperature need to have oxidation resistance. This resistance can be achieved when the material develops, through oxidation, an oxide film which acts as a diffusion barrier while keeping a good adherence. In this respect, alpha alumina clearly acts as such. The oxides of aluminum have been the subject of many investigations because of their commercial importance and scientific interest. The thermal stability and optical properties of pure nanometer-sized alumina powder have received much attention because of their intrinsic interest and commercial value. Nanometer-sized alumina powders are widely applied today. One of its applications is in fluorescent lamps due to the absorption of ultraviolet light. In fact, it can also emit the light under excitation with a suitable wavelength. It is important to note that there are many works about alpha alumina using X-ray diffraction, but there is a need for a more detailed structural analysis. To achieve this more exhaustive structural characterization we have used the Rietveld refinement method and cathodoluminescence (CL) measurements.

In the presence of lattice defects, extra luminescence emissions can be observed in the ultraviolet (UV) region upon highly energetic excitation. The main intrinsic defects in thealumina crystals are oxygen vacancies in different charge states: a neutral vacancy, a vacancy capturing one electron (a F + -center), and a vacancy capturing two electrons (a Fcenter) (Kislov et al., 2004;Michizono et al., 2007;Yu et al., 2002). The observed UV spectrum in -alumina can be deconvoluted into two distinct sub-band components: an F +center band, located at around 3.8 eV, and a less intense F-center band, located at around 3.0 eV (Brewer et al., 1980;. Depending on the defect introduction method one can create also F 2 -centers, 2 F  -centers and 2 2 F  -centers (double oxygen vacancy with four, three and two trapped electrons respectively). -Al 2 O 3 crystals with defects in the oxygen sublattice are actively studied as promising storage materials (Kortov & Milman, 1996). In this connection, it is interesting to study luminescence properties of the nanostructured aluminium oxide and compare them with analogous properties of crystalline samples.
In this chapter, we present X-ray diffraction (Rietveld analysis) and CL measurements ofalumina powders formed by calcination of gibbsite or generated by oxidation of a metallic FeCrAl alloy. The peculiarity of the cathodoluminescence under comparable conditions of -alumina is discussed.
The present work was performed on ferritic ODS commercial FeCrAlY alloy PM2000 (20 wt% Cr, 5.8 wt% Al, 0.5 wt% Ti, 0.5 wt% Y 2 O 3 ). This alloy is an alumina-forming alloy. The specimens, with dimensions 25×25×5 mm 3 , are cut from a rolled plate. In the case of the PM2000, sample was oxidised in air at 1223 K for 72h. The thermogravimetric method (TG) and differential thermal analysis (DTA) data were recorded under a dry air flow with a heating rate of 10 K/min in a SETARAM TGDTA92-16.18 thermal analyser. TG measurements were corrected for temperature-dependent buoyancy by subtracting the data of a measurement carried out on an inert sample. The crystalline structure of the sample was investigated by XRD using a PANalytical X'Pert Pro MRD diffractometer configured as follows: Cu tube operating at 40 kV and 30 mA (λ(Kα 1 ) = 0.15406 nm, λ(Kα 2 ) = 0.15444 nm). The scan rate (2θ) was 1°/min at a step size of 0.025°. The data were processed to realize the conditions of the software program Fullprof Suite for the structure refinement.
The FTIR technique was used in the absorbance mode in the 200-4000 cm -1 range. For oxides all bands have characteristic frequencies between 200 and 1000 cm -1 . For the FTIR measurements, samples were prepared by grinding the oxide films scraped from the substrate (PM2000). After calcination, 10 to 100 µg of the powder was drawn, then grinded with 23 ±2 mg of CsBr in order to obtain, a pellet of 200-250 µm in thickness. After grinding, the powder was placed in a mould (5 mm diameter) and a cold isostatic pressure (CIP) of 150 MPa was applied for 5 min.
The FTIR spectra are obtained using a Perkin-Elmer spectrometer at resolution of 8 cm -1 . For each sample, 120 scans were used. The apparatus is equipped with a system allowing the reduction of the optical course in air in order to minimize the perturbations associated with ambient air (water vapour and CO 2 ). The uncertainty on the position of the various peaks is equal to ± 2 cm -1 .
The emitted light under electron beam excitation in a UHV system was analyzed through a quartz window with a Jobin Yvon CP 2000 spectrograph and a CCD detector. The wavelength range 200-1000 nm was investigated.

Results and discussions
3.1 Differential thermal analysis and thermal gravimetry analysis (DTA-TG) of gibbsite Fig. 1 shows the typical TG-DTA curves of the gibbsite. TG and DTA curves are indicated as dotted and solid lines, respectively. Concerning the dehydration-dehydroxylation process of gibbsite, the dehydration appears to occur in two steps (around 598 and 803 K respectively) at higher temperature. The expected theoretical loss due to dehydration is 34.6%, the experimental loss is 34.3% a little lower. This difference 0.3% is a bit larger than the experimental uncertainty 0.1%, the starting gibbsite may be slightly dehydrated. The last step (803 to 1273 K) gives no thermal event but appears as a continuous mass loss (about 2%) which corresponds to the elimination of residual hydroxyls. The formation ofalumina occurs between 1473 and 1533 K (MacKenzie et al., 1999). Finally, the structural transformations to well-crystallised −alumina are described by nucleation (T<1470 K) and growth mechanisms (T>1470 K).

Gibbsite
Gibbsite ( -Al(OH) 3 ) has monoclinic symmetry (a=0.8684, b=0.5078, c=0.9736 nm, =94.54°) with the space group P21/n, and the unit cell contains eight Al(OH) 3 units (Saalfeld & Wedde, 1974). Gibbsite is characterized by the stacking of two-layer units (AA or BB) of hydroxyl sheets with the sequence ABBAABBA... where hydroxyl sheets of the adjacent Gibbsite layers face the c direction (Kogure, 1999). In Fig. 2a, the XRD pattern obtained on the as received gibbsite powder shows a good agreement with the reference XRD pattern (33-0018 JCPDS file).

Phase transitions induced by heat treatment of gibbsite
When heating up fine-grained gibbsite, most OH groups are eliminated, and various forms of alumina are formed with the sequence: gibbsite →χ-→κ-→ -Al 2 O 3 when temperature increases. In order to study these phases, we performed XRD measurements on four samples prepared from gibbsite calcined for 24h at 773, 1073 (62h), 1173 and 1573K respectively.
According to Fig. 2b, at 773K ,the χ phase is expected. In spite of many investigations since the1950's (Brindley & Choe, 1961;Kogure, 1999;Saalfeld, 1960;Stumpf et al., 1950;Yu et al., 2002), the crystal structure of χ-alumina is still uncertain. Stumpth et al. (Stumpf et al., 1950) assumed a cubic (not spinel) unit cell of lattice parameter a=0.795 nm (04-0880 JCPDS file). On the other hand, two hexagonal structures have been suggested, either with the parameters a=0.556 nm and c=1.344 nm (Saalfeld, 1960) or with a=0.557 nm and c=0.864 nm (Brindley & Choe, 1961) (13-0373 JCPDS file). The two previous hexagonal unit cells may be described respectively as a stacking of 6 and 4 close-packed oxygen layers, of approximately the same thickness (0.224 and 0.216 nm) as the Al-OH layers in gibbsite (0.212 nm). More recently, Kogure (Kogure, 1999) proposed a hexagonal lattice with a=0.49 nm and an undefined c length indicating that χ-alumina structure can be regarded as random close packing of gibbsite-like layers.
For samples prepared at 1173 K, the κ phase is expected (see Fig. 2d). Contrary to the χ phase, the crystal structure of κ-alumina is well known (see for example Ref. (Ollivier et al., 1997) and references therein). κ-alumina is orthorhombic with the space group pna2 1 and results in ten independent atoms positions (four Al and six O). The experimental XRD pattern at 1173K is specific of a pure κ-alumina (Fig. 2d). Nevertheless, the presence of remnant χ-phase cannot be excluded as all the χ peaks also appear in κ structure. Note that the experimental XRD patterns show well crystallized phases, in contrast with the χ phase. When heating up at temperature above 1573K for 24h, gibbsite transforms into -alumina, the stable structure. The XRD pattern of the 1573K sample ( Fig. 6) shows that only -Al 2 O 3 is present when compared with 42-1468 JCPDS file. In this structure, tetrahedral Al 3+ ions are no longer present and only AlO 6 octahedron remain.

General properties of alumina (Al 2 O 3 )
Aluminum oxide, commonly referred to as alumina, possesses strong ionic interatomic bonding giving rise to it's desirable material characteristics. It can exist in several crystalline phases which all revert to the most stable hexagonal alpha phase at elevated temperatures. This is the phase of particular interest for structural applications and the material available from Accuratus.
The exceptional properties of alumina (Al 2 O 3 ), such as great hardness, high thermal and chemical stability, and high melting temperature, make it a very attractive material. The crystalline -Al 2 O 3 phase (corundum or sapphire) is the single stable modification of alumina. The crystalline -Al 2 O 3 has the band gap Eg≈8.5 eV and is widely used in optical devices. Sapphire doped with chrome (ruby) or titanium is applied as an active medium in laser systems. In microelectronics sapphire is used as a substrate for growing silicon and gallium nitride (GaN). Alumina is a highly radiation-resistant material and is used as a sensitive element in detectors when measuring the ionizing radiation parameters.
High purity alumina is usable in both oxidizing and reducing atmospheres to 1925°C. Weight loss in vacuum ranges from 10 -7 to 10 -6 g/cm 2 .sec over a temperature range of 1700° to 2000°C. It resists attack by all gases except wet fluorine and is resistant to all common reagents except hydrofluoric acid and phosphoric acid. Elevated temperature attack occurs in the presence of alkali metal vapors particularly at lower purity levels.
The composition of the ceramic body can be changed to enhance particular desirable material characteristics. An example would be additions of chrome oxide or manganese oxide to improve hardness and change color. Other additions can be made to improve the ease and consistency of metal films fired to the ceramic for subsequent brazed and soldered assembly.
Mechanical, thermal and electrical properties of Al 2 O 3 are summarized in table 1.

Oxidation of PM2000
The oxidation resistance of high-temperature alloys and metallic coatings is dependent on the formation of a protective surface oxide. In an ideal case, the oxide layer should be highly stable, continuous, slow growing, free from cracks or pores, adherent and coherent. -Al 2 O 3 is an oxide which comes close to satisfying these requirements; the slow growth rate is related to its highly stoichiometric structure and its large band gap which makes electronic conduction difficult.
One of the most crucial factors in the oxidation of alumina-formers is the temperature, which must be high enough to promote the formation of -Al 2 O 3 in preference to the less protective transition alumina. Another critical factor is the aluminum content which must be sufficiently high to develop and maintain an alumina layer and prevent subsequent breakaway oxidation.
The addition of chromium to Fe-Al alloys promotes the formation and maintenance of a complete layer of -Al 2 O 3 by acting as a getter and preventing internal oxidation of the aluminum (Wood, 1970).
Iron and chromium are the major impurities present in Al 2 O 3 scales formed on PM2000.
Primarily their oxides formed during the transient stage and were incorporated into the -Al 2 O 3 scale. Fe segregated to some -Al 2 O 3 grain boundaries, but not Cr. The Al 2 O 3 scale became progressively purer with oxidation time. It is possible that the Fe in the Al 2 O 3 scale increases the scaling rate and, in particular, enhances lateral growth that causes scale convolution.
The growth of -Al 2 O 3 scales that form on FeCrAl alloys during high temperature oxidation is generally considered to be controlled by oxygen inward diffusion through oxide grain boundaries (Mennicke et al., 1998;Quaddakkers et al., 1991). Aluminum also diffuses out, which can cause growth within the scale (Golightly et al., 1979). The degree of Al outward transport can be significantly reduced by the presence of reactive elements, such as Y, Hf or Zr (Mennicke et al., 1998;Quaddakkers et al., 1991), which segregate to Al 2 O 3 grain boundaries (Przybylski et al., 1987). However, the extent of outward growth seems to differ appreciably among several reactive-element doped Fe based alloys.
The EDS results of the average Fe and Cr concentrations in the scale as a function of scale thickness from different transmission electron microscopy (TEM) specimens are summarized in Table 2.  EDX analysis (Fig. 5) of the sample oxidized at 1123 K indicates that the film mainly consists in aluminium and oxygen elements and very small amount of the substrate constituting elements are observed. As the film thickness decreases, like sample oxidized at 1023 K, the signature of the substrate increases due to interactions of the electrons with the underlying substrate mater as the film is thinner than that formed at 1123 K.

Rietveld refinement of the structures
The Rietveld refinement of the structures was performed using the WinPlotr/FullProf suite package (Rodríguez-Carvajal, 1993). The peak shape was described by a pseudo-Voigt function, and the background level was modeled using a polynomial function. The profiles were refined using the space group and structure models of Al 2 O 3 ( 3 RC , JCPDS 46-1212) (in this structure cations occupy the 12c sites and oxygen ions the 18e sites).
The XRD data for the 2θ regions between 20° and 80° was used for the refinement. The observed, Rietveld refined and difference patterns are shown in Figs. 6 and 7.
It was obvious that the agreement between the experimental data and the simulations was excellent since the R wp (weighted residual error) factor was small (≤ 14.1%). The Rietveld results (cell parameters, atom position, reliability factors and crystallite size (D)) are given in Table 3 and 4. The size of -alumina crystallites (D=36 nm) obtained using PM2000 alloy was smaller than that (D=43 nm) obtained using gibbsite precursor. Furthermore, the -Al 2 O 3 from PM2000 has a greater lattice parameter (a=0.4763nm, c=1.3047nm) than that of JCPDS file 46-1212 (a=0.4758 nm, c=1.2991 nm, c/a=2.730) and a greater c/a ratio (c/a =2.739) than that of JCPDS file 46-1212. However, the -Al 2 O 3 from gibbsite has a smaller lattice parameter (a=0.4752nm, c=1.2980nm) and a similar c/a ratio (c/a =2.731) than that of JCPDS file 46-1212. Furthermore, the -Al 2 O 3 from gibbsite has a greater calculated density (3.996 g/cm 3 ) than that of -Al 2 O 3 from PM2000 (3.961 g/cm 3 ). Phase transformations are frequently accompanied by microstructural changes. This fact could explain the crystallographic parameters differences between a two -alumina.

Fourier transform infrared (FTIR) analysis
Fig. 8 compares the FTIR absorbance spectra of -Al 2 O 3 obtained after calcination of gibbsite and oxidation of PM2000. For the samples calcined at 1573 K, significant spectroscopic bands at ~640, ~594, ~447 cm -1 and ~386 cm -1 appear which are identified to be the characteristic absorption bands of -Al 2 O 3 (Barker, 1963). This is in good agreement with XRD observations. Common bands exist in all cases, such as the broad OH band centered around 3420 cm −1 , and the 1640 cm −1 H 2 O vibration band (Ma et al., 2008). The very high surface area of these materials results in rapid adsorption of water from the atmosphere because the FTIR samples were kept and grinded in air. Three peaks of very weak intensities at 2850 cm -1 , 2920 cm -1 and 2960 cm -1 are observed which are due to C-H stretching vibrations of alkane groups. The absorption in ~2356 cm −1 is due to CO 2 molecular presence in air. Fig. 8. FTIR absorption spectra of -alumina powder prepared by calcination of gibbsite (black line) or generated by oxidation of PM2000 (red line).

Cathodoluminescence
CL is the phenomenon of light emission from specimens as a result of interaction with an electron beam. In insulating crystals, the origin of the luminescence arises from impurity atoms (e.g. transition metals or rare earths) in the crystal lattice. Using an electron microscope to produce the electron beam, the spatial distribution of luminescent sites can be observed with submicron spatial resolution, and correlated with features of the specimen morphology or microstructure.
The mechanisms for CL are similar to those for photoluminescence, but the energy input or excitation source is that of an electron beam rather than a visible or ultraviolet light beam. When an energetic (keV range) electron beam propagates within a semiconductor or insulator, the primary electrons lose energy by the creation of electron-hole pairs. These www.intechopen.com electron-hole pairs then recombine via radiative and non-radiative processes. Only the radiative recombination process which leads to the creation of a photon is viewed with CL. Radiative recombination may be intrinsic (arising from electronic states of the perfect crystal) or extrinsic (arising from electronic states that are localized at defects or impurities in the crystal). Extrinsic luminescence thus provides information about defects and impurities in the crystal lattice. For conciseness, the defects and impurities that give rise to extrinsic luminescence are often denoted the luminescence centers. Each type of luminescence center in a particular crystal has a characteristic emission spectrum. The spectrum may contain both narrow lines and broad bands, depending on the energy level structure of the luminescence center and the coupling of the center to the host lattice. CL is advantageous compared to photoluminescence. CL can potentially give additional information about local positions in the sample because the electron beam can be focused on several nanometers. In addition, the CL system operates under UHV conditions of less than 10 −9 Torr. Hence, CL measurements can be performed in a contamination-free environment, which is very effective in detecting weak luminescence. Moreover, the depth dependent emission profiles can be examined in CL by controlling the accelerating voltage.

Oxygen vacancy in Al 2 O 3
CL signal is a good signature of the material qualities and is used in this study to characterize the point defects associated to oxygen vacancies in -Al 2 O 3 . The CL spectra of -Al 2 O 3 formed from gibbsite and PM2000 are given in figure 9 and 12.
In figure 9, the CL spectra shows that wide band over the interval of (200-600 nm) consists of a series of overlapping bands. The main emission bands located at about 250 nm (4.96 eV), 281 nm (4.41 eV), 325 nm (3.81eV), 373 nm (3.32 eV) and 487 nm (2.54 eV) occur in alpha alumina powder and also in alpha alumina films generated by oxidation of a metallic FeCrAl alloy. We believe that the observed CL peak at 4.96 eV is related to the interband transitions or to defect that is different in origin to the F or F + centers in -alumina. The luminescence band at 4.41 eV is detected only if the excitation density is high and was previously observed in -alumina by Kortov et al. (Kortov et al., 2008).
In -Al 2 O 3 (corundum structure) each O atom is surrounded by four Al atoms forming two kinds of Al-O bonds of length 0.186 and 0.197 nm. This is why in corundum F-type centers have low C 2 symmetry. Besides, an O vacancy has two nearest neighbor O atoms, forming the basic O triangle with O-O band length of 0.249 nm in perfect corundum. Thus, the Ftype centers are surrounded by six nearest atoms which determine mainly their optical properties.
Defects induced in Al 2 O 3 may be of various kinds: F centers (oxygen vacancy with two electron), F + centers (oxygen vacancy with one electron), F 2 centers (two oxygen vacancies with four trapped electrons), 2 F  centers (two oxygen vacancies with three electrons) and 2 2 F  centers (two oxygen vacancies with two electrons) (Ghamdi & Townsend, 1990).
In the case of -alumina and sapphire, there have been reported a number of F-type centers including the F + and F centers (Evans, 1995). As for the luminescence of irradiation defects, it is known that the luminescence of the F + center is observed at the UV region around 3.8 eV (325 nm) while that of the F centers is observed at a lower photon energy region around 3.0 eV (410 nm). Thus the presently observed luminescence from -alumina and sapphire which is centered at 330 and 420 nm can be attributed to the F + and F centers. As for the others, it is noted that the luminescence at 250 and 290 nm is also observed in the case ofalumina. The luminescence intensity at these bands was found to be sensitive to thermal annealing at higher temperatures, and then might be attributed to the effect of some impurities such as OH.
In Ref. (Oster & Weise, 1994), the absorption bands of 220 and 260 nm in pure -Al 2 O 3 crystals have been attributed to F+ absorption bands. Due to the presence of C 2 -symmetry in the F + -center (oxygen vacancy occupied by a single electron) in pure -Al 2 O 3 crystals, the excited state is split into three levels, 1 B, 2 A, and 2 B, according to the theory of La et al. (La et al., 1973), giving three polarized optical absorption bands located at 255, 229 and 200 nm.
The optical properties of these luminescent centres are well known. They possess absorption and emission bands which are produced in the gap as summarised in Fig. 10. F+ centre is characterised by three absorption bands at 6.3, 5.4 and 4.8 eV and emits at 3.8 eV (330 nm). F centre absorbs at 6 eV and emits at 3 eV (415 nm).  In Ref. (Evans, 1994), synthetic sapphire single crystals grown by four different techniques all showed an anisotropic 5.4 eV (230 nm) absorption broad band. There was not any other absorption band presented in his cases. As for the nanometer-sized Al 2 O 3 powder apart from normal lattice vacancy-type defects such as the F-, F + -and F 2 -type centers it also has the surface defect because the nanometer powder has a larger specific surface area and there exists a lot of dangling bonds in the surface.
Experimental measurements of absorption and luminescence energies for single-vacancy and for dimer centers in Al 2 O 3 are collected in Table 5.  Table 5. Experimental measurements of optical properties for single-vacancy and for dimer centers in Al 2 O 3 , energies are in eV (Crawford, 1984;Evans et al., 1994).
The band of F-centers is absent in both samples due to the existence of impurity Cr (Aoki, 1996). For instance, the CL intensity at 3.81 eV (F + -center) of the -Al 2 O 3 formed from gibbsite is approximately 5 times higher than that of the -Al 2 O 3 formed from PM2000 measured under the same excitation conditions. The main specific feature of the CL spectra of the -Al 2 O 3 is the presence of a new emission band with the maximum at 3.32 eV. It is a possible that the new emission band is related to surface F s -centers concentrating on nanoparticle boundaries (Evans, 1995).
Luminescence spectra obtained for anion-deficient aluminum corundum exposed to different types of excitation and stimulation exhibit luminescence with nanosecond (F +centers), microsecond (Ti 3+ and Al + i ) and millisecond (F and Cr 3+ ) decay times (Surdo et al., 2005). Significantly, the aforementioned centers, which actively participate in relaxation processes, have considerably different decay times t and emission band maxima h (  Table 6. Basic parameters of the emission of most active centers (Surdo et al., 2001;Springis & Valbis, 1984) The band 487 nm (2.54 eV) can be related to the aggregate F 2 -centers produced by doubleoxygen vacancies and the centers formed by interstitial aluminum ions. It is known that these centers are responsible for the green luminescence in highly disordered crystals of aluminum oxide (Tale et al., 1996;Springis & Valbis, 1984).

Cr 3+ in Al 2 O 3
A classic example of the isolated luminescent centre is Cr 3+ in Al 2 O 3 (ruby) when the excited electronic energy levels of the host are at much higher energy than those of the dopant ion. The dopant ion colours the colourless host lattice red. If the concentration of the dopant ion is low, the interaction between the dopant ions can be neglected. This is what we consider here as an isolated luminescent centre.
Different impurities in corundum ( -Al 2 O 3 ) produce different color varieties. All colors of corundum are referred to as sapphire, except for the red color, which is known as ruby.
Corundum has a trigonal lattice D 6 3d structure. The crystals have an approximately hexagonal closed packing structure of oxygen and metal atoms. The six oxygen ions are octahedrally coordinated cations; and only two-thirds of the octahedral sites are filled. If corundum has more than 1000 ppm Cr 3+ ions as impurities, it is referred to as a ruby. Rubies can be used in solid-state lasers (Soukieh et al., 2004), and they fetch high prices in gem markets. Chromium can be substituted for the aluminum in corundum and is present as chromium oxide. The Cr 3+ ion is slightly larger than Al 3+ ; therefore, it naturally enters easily into the corundum structure. As a result, Cr 3+ ions form 3d 3 , with only three unpaired electrons in the 3d orbitals. If the Cr 3+ ion is located in the Al 3+ site in corundum, it coordinates the six oxygens into a distorted octahedral configuration (Nassau, 1983).
For Cr 3+ in Al 2 O 3 crystal, Cr 3+ substitutes for some of Al 3+ , and adopts octahedral ligand coordination. The 3d levels are extremely host sensitive. The strong crystal field in Al 2 O 3 leads to the splitting of 3d electron orbits of Cr 3+ and produces the ground level: 4 A 2 , and the excited states: 2 E, 4 T 2 , and 4 T 1 , etc. the transitions from 4 A 2 to 4 T 2 , and 4 T 1 are spin-allowed, so these energy levels act as broad pumping levels. The 2 E is the narrow lowest excited band, acting as emitting level. The unusual magnitude of this crystal field splitting extends the lowest 2 E state 14400 cm -1 (694 nm) above the ground state. Thus the 2 E-4 A 2 transition of Cr 3+ : Al 2 O 3 crystal lies in visible spectral region. Exciting any of the pumping bands of 4 T 2 , and 4 T 1 results in fast relaxation to lowest 2 E excited state. At room temperature, the fluorescence emitting from 2 E state appears as a sharp band with a peak at 694 nm corresponding to the transition to the 2 E terminal state. The Cr 3+ ion has two strong absorption bands in the visible part of the spectrum, which explain the red color, i.e., 2.2 eV light can be absorbed to raise the chromium from the 4 A 2 ground level to the 4 T 2 excited level as absorption in the yellow-green, and 3.0 eV light raises it to the 4 T 1 level as violet absorption. In addition, the absorption decreases to zero in the red region below 2.0 eV. Therefore, rubies have a red color with a slight purple overtone.
Chromium impurity in -Al 2 O 3 lattice is characterised by two bands of absorption (3.1 and 2.2 eV) and one fine emission structure peaked at 1.8 eV (693 nm) as summarised in Fig. 11. In oxide insulators, a number of transition metal and rare earth impurities act as luminescence centers. The trivalent chromium ion (Cr 3+ ), with electronic configuration 3d 3 , is an efficient luminescence center in many light-metal oxides, including Al 2 O 3 and MgO. The trivalent chromium ion enters substitutionally and is surrounded by an octahedron of oxygen ions. In aluminum oxide, the surroundings of the chromium ion are not quite cubic, as the oxygen octahedron is stretched along its trigonal symmetry axis C 3 .
Chromium in -Al 2 O 3 lattice gives a luminescence in the visible domain. In CL, the narrow band at 693 nm is attributed to chromium impurity (Ghamnia et al., 2003). In Al 2 O 3 :Cr 3+ (ruby) the apparent lifetime of the R-line emission may increase from the intrinsic value of 3.8 ms up to 12 ms (Auzel & Baldacchini, 2007).
The typical CL spectra of -Al 2 O 3 , formed by calcination of gibbsite powder at 1573 K for 24 h (black line) and by oxidation of PM2000 at 1223K for 72h (red line), obtained in the region of 600 to 800 nm at room temperature are shown in Fig. 12 (Jovanic, 1997;Geschwind et al., 1962;Crozier, 1965). Fig. 12. CL spectra of -Al 2 O 3 , formed by calcination of gibbsite powder at 1573 K for 24 h (black line) and by oxidation of PM2000 at 1223K for 72h (red line), obtained in the region of 600 to 800 nm at room temperature.
A similar emission was recently reported in -Al 2 O 3 microcones (Li et al., 2010). Thus, it can be concluded that the incorporation of Mn 4+ ions in -Al 2 O 3 observed in the experiment is irreversible and occurs during its formation from the Mn 4+ ions dissolved quite uniformly in the bulk of the low-temperature polymorphic modifications of alumina (gibbsite →χ-→κ-→ -Al 2 O 3 ). The Mn 4+ impurity emission at 677 nm is absent in -alumina films obtained by oxidation of a metallic FeCrAl alloy. Taking into account high sensitivity of the method, this indicates very low concentration of such ions.

Conclusion
-Al 2 O 3 was prepared either by calcination of gibbsite and also generated by oxidation of a metallic FeCrAl alloy. The Mn 4+ impurity emission at 1.83 eV is absent in -alumina thin films obtained by oxidation of a metallic FeCrAl alloy. The band of F-centers is absent in both samples due to the existence of impurity Cr. The difference in oxygen vacancies (F +centers) amount between -Al 2 O 3 from gibbsite and from PM2000 was confirmed by CL spectra.