Materials Science in Semiconductor Processing 63 (2017) 184–189

Contents lists available at ScienceDirect

Materials Science in Semiconductor Processing

journal homepage: www.elsevier.com/locate/mssp

Band gap measurement of Bi2MoxW1-xO6 by low loss electron energy loss MARK
spectroscopy
⁎
Damasio Morales-Cruza,b, , Francisco Paraguay-Delgadoa,e, Raúl Borja-Urbyc, Sofía Basurto-
Cerecedaa, Guillermo Herrera-Péreza, Paolo Longod, Marek Malace
a
Centro de Investigación en Materiales Avanzados (CIMAV) S. C., Miguel de Cervantes 120, Chihuahua 31136, Chihuahua, México
b
Escuela Superior de Ingeniería Mecánica y Eléctrica. Instituto Politécnico Nacional, Unidad Prof. Adolfo López Mateos, Zacatenco, Delegación Gustavo A.
Madero, 07738 México City, México
c
Centro de Nanociencias Micro y Nanotecnología. Instituto Politécnico Nacional, Unidad Prof. Adolfo López Mateos, Zacatenco, Delegación Gustavo A.
Madero, 07738 México City, México
d
Gatan, Inc., 5794, W Las Positas BLVD Pleasanton, CA 94588, USA
e
National Institute for Nanotechnology, 11421 Saskatchewan Dr., Edmonton, Canada T6G 2M9

A R T I C L E I N F O A BS T RAC T

Keywords: This work shows the comparison of high-resolution electron energy loss spectra (HR-EELS) in the low loss
Electron energy loss spectroscopy (EELS) region (0−15 eV) to investigate the electronic structure from koechilinite Bi2MoO6 to rusellite Bi2WO6 varying
Transmission electron microscopy the stoichiometric relation Bi2MoxW1−xO6. The effect of the Mo to W ratio on the bandgap energy was evaluated
Band gap on individual particles. Two approximations were considered in order to determine the band gap energy value,
Energy loss function
the first one was a linear fit and the second one was a mathematical fit. Both analyses are in agreement with
Bi2MoxW1−xO6
those ones collected and analyzed by UV–Vis characterization. Our results suggest a direct electronic transition
that increases from about 2.53 eV to about 3 eV as the W content increase from 0% to 100% wt. X-ray
diffraction was used to corroborate the crystal structure and crystal size; transmission electron microscopy was
used to monitor the morphology evolution and UV–Vis spectroscopy in diffuse reflectance mode to determine
the Eg. These techniques complement the characterization of these materials.

1. Introduction yielding the structure, bonding and electronic properties of the

materials at nanoscale [6]. The 1−50 eV loss region of the energy-loss
Nanostructured tungsten oxides have attracted interest arising from spectrum can be utilized to retrieve the electronics and optical proper-
properties that can lead to new potential applications, including ties of a material [7,8]. While the spatial resolution of a STEM can be
photoluminescence, optical fibers, scintillator materials, and photo- below 1 nm the sample damage by the electron beam often leads to
catalysts [1]. The development of visible light driven photocatalyst has choice of moderate (few nm) probe size [9].
attracted much attention as a possible approach to increase efficiency The present study reports systematic investigation about electronic
of solar energy utilization [2]. Heterogeneous semiconductor photo- bandgap evolution (Eg) for individual nanoparticles of Bi2MoxW1−xO6
catalysts including koechilinite (Bi2WO6) and rusellite (Bi2MoO6) have (X values from 0 to 1) [3,4] observed at high spatial resolution using
sparked worldwide interest in converting solar energy to chemical EELS in a STEM in the low energy loss region 1 – 15 eV. The
energy and decomposing organic contaminants [3,4]. The photocata- estimation of Eg is carried out by Kramers-Kronig (K-K) analysis of
lytic activity depends on numerous factors, especially the morphology the energy loss function (ELF). The comparison between two approx-
and size of the tungsten oxide nanoparticles (NP). Typically, optical imations (linear and parabolic fit) was performed on the ELF in order
band gap is measured by optical methods which have high energy to determine the band gap energy Eg evolution on these materials. UV–
resolution (~2 meV), but the spatial resolution is not adequate for Vis spectroscopy in diffuse reflectance mode was performed in order to
nanostructured photocatalysts (NPs with size below 200 nm) [5]. corroborate the bandgap evolution on the bulk of these materials. The
Electron energy loss spectroscopy (EELS) in a (scanning) transmis- Structural evolution was study by X-ray diffraction and Raman
sion electron microscopy (STEM) mode provides an alternative char- spectroscopy, the morphology evolution was known by transmission
acterization that allows to fully utilizing analytical capabilities of a TEM electron microscopy which complements the characterization of these

⁎
Corresponding author at: Centro de Investigación en Materiales Avanzados (CIMAV) S. C., Miguel de Cervantes 120, Chihuahua 31136, Chihuahua, México.
E-mail addresses: [email protected], [email protected] (D. Morales-Cruz).

http://dx.doi.org/10.1016/j.mssp.2017.02.016
Received 23 August 2016; Received in revised form 5 December 2016
1369-8001/ © 2017 Elsevier Ltd. All rights reserved.
D. Morales-Cruz et al. Materials Science in Semiconductor Processing 63 (2017) 184–189

materials.

2. Experimental procedure

The Bi2MoxW1−xO6 samples were synthesized from bismuth nitrate

pentahydrate [Bi(NO3)3 5H2O (1 g)], ammonium metatungstate hy-
drate [(NH4)6H2W12O40 H2O], ammonium heptamolybdate tetra hy-
drate [(NH4)6Mo7O24 4H2O], hydrogen peroxide [H2O2 (6 ml) 30 wt%]
and nitric acid [HNO3 (0.5 ml) 70 wt%] purchased from Sigma-
Aldrich. The atomic solution concentrations of Mo in relation with W
atoms were “y %” (y=0, 3, 5, 10, 15, 25, 50, 70, 90 and 100 of Mo, then
(100-y) % of W). The procedure for synthesizing the samples started by
dissolving 1 g of Bi (NO3)3 5H2O in 20 ml of tri-distilled water. Then
0.5 ml of HNO3 was added to this solution referred to as SBi. Then the
(100-y) % of ammonium metatungstate hydrate and y % of ammonium
heptamolybdate tetrahydrate were separately dissolved each one in
10 ml of tri-distilled water. The solutions are referred to as SW100−y
and SMoy, respectively. To get the final solution for synthesis
Bi2MoxW1−xO6 the procedure was as follows: solution SW100−y was
added to SBi and then SMoy, then 6 ml of H2O2 was added to the last
solution. The resulting solution was transferred to a quartz vial for heat
treatment in a Multiwave PRO microwave (Anton Paar, Rotor type:
8NXQ80). The synthesis was performed at 200 °C under pressure of 35
bars for 25 min and to 800 W FR power.
The EELS experiments were carried out in a JEOL JEM-
ARM200CF transmission electron microscope equipped with a Gatan
image filter 965-ER Quantum operated at 80 kV. The low loss spectra
were acquired in STEM mode. The electron emitter of microscope is Fig. 1. (a - h) X-ray diffraction patterns evolution from bismuth molybdate to bismuth
equipped with a cold field emission source. The transmission electron tungstate.
microscopy (TEM) images were acquired using a JEOL JEM-7200FS,
in the STEM mode. observed in the patterns b, c, d, e, f, and g that the Bi2MoxW1−xO6
In order to complement the structural and Mo solid solution in the sample also exhibits an anisotropic growth [15]. These results can be
Bi2WO6 was studied by X-ray diffraction (XRD) patterns using a attributed to the square-plate morphology of the crystallites as can be
PANalytical XPert´PRO with X´Celerator detector diffractometer with observed in the micrographs collected by TEM and discussed in the
λ (CuKα) =1.5406 Å monochromatic radiation. The crystallite size of next section.
the samples was estimated from the Full Width at Half-Maximum One can observe three interesting aspects, the first one is that the
(FWHM). We consider the (020*) and (131*) reflections in the incorporation of the Mo, the intensity of the (020) reflection increases.
Scherrer equation, D = Kλ/β cos θ, where K is the shape factor with The second one the reflections located at 2θ, 32.87°, 47.16°, and 55.93°
a typical value of 0.89, λ is the X-ray wavelength as we describe above, are splitting gradually, indicating the solid solution formation, due to
β is the line broadening at FWHM in radians (in which the instru- the systematical change with the Mo content and the third one, the
mental resolution is consider), and θ is the Bragg angle. The instru- (131) reflections sharpen, indicating an increase of the crystallite size.
mental resolution function was experimentally determined by using of In fact, the crystallite size was determined from the FWHM of (131)
a LaB6 standard. The Raman spectra were acquired by Micro Raman reflection by using the Scherrer equation. Our results suggest a gradual
Lab Ram HR Vis 633 instrument, Horiba with He-Ne light at 632.8 nm increase of crystal size from 35 nm for Bi2WO6, 51 nm for
wavelength laser. The UV-Visible absorbance (UV–Vis) by diffuse Bi2Mo0.5W0.5O6 to 62 nm for Bi2MoO6. These results are in agreement
reflectance was obtained with a Lambda 9 Perkin Elmer UV–Vis with those one reported previously by L. Zhou et al. [3].
spectrometer in diffuse reflectance mode, using Labsphere reflectance The results concerning to the gradual evolution of the normal
patterns in the range of 1000–250 nm at a scan speed of 900 nm min−1 phonon modes determined by Raman spectroscopy for Bi2WO6,
and lamp replacement (visible-ultraviolet) at 350 nm. Bi2WxMo1−xO6 solid solution up to pure Bi2MoO6 are shown in
Fig. 2. As the first step, the vibrational bands for the russellite were
3. Results and discussion identified at 831 cm−1, 816 cm−1, 797 cm−1, 724 cm−1, 714 cm−1,
698 cm−1, 594 cm−1, 515 cm−1, 438 cm−1, 417 cm−1, 324 cm−1,
Fig. 1(a-h) shows the X-ray diffraction pattern evolution from pure 303 cm−1, 281 cm−1, 258 cm−1, 227 cm−1, 218 cm−1, 205 cm−1. One
bismuth tungstate to pure bismuth molybdate solid solutions. These can observe in this figure that all Raman modes are in agreement with
results suggest that the samples contain a unique well-crystallized those ones reported in the literature [16].
phase with an orthorhombic structure with Pca21 space group [10,11]. The Raman spectra could be divided in two main regions. The first
The diffraction pattern for Bi2MoO6 (pattern h) can be indexed as one is in the range of 200–550 cm−1. This range is characterized by the
pure koechilinite phase consistent with PDF No. 76–2388 [12]. The bending modes of MO6 octahedra (where M is occupy by W cation and
diffraction pattern intensities for Bi2WO6 (pattern a) can be indexed as then is gradually substituted by Mo cation) coupled with stretching and
pure rusellite phase consistent with PDF No. 39–0256 [13]. One can bending modes of bismuth-oxygen polyhedra [12]. It can be seen in the
observe in the XRD patterns that there are systematic differences in the evolution of the spectra the gradual change of Raman shifts for the
relative intensities for the (020), (131) (060) (200) and (002) reflec- bands located at 417 cm−1, 324 cm−1, and 303 cm−1 accompanied of
tions for all the samples, suggesting the possibility of different gradual change in the Raman band intensity increasing Mo concentra-
preferential orientation growth. In addition, the ratio of intensity tion. In particular, the intensity of Raman band at 324 cm−1 increases
between I (131) / I (002) for Bi2WO6 is smaller than 2, implying that gradually. This result could be attributed to the deformation degree of
the nanoparticles have anisotropic growth [14]. Meanwhile, it was

185
D. Morales-Cruz et al. Materials Science in Semiconductor Processing 63 (2017) 184–189

Fig. 2. Raman spectra of Bi2MoxW1−xO6.

MO6 octahedra by the substitution of W cation for Mo cation.

Additionally this result reflect the gradually variation in the M-O bond
lengths in tungstate species [17] by the rearrangement of the crystal
structure. The second region to be analyzed is founded in the range
between 550 cm−1 up to 1000 cm−1 where the Raman bands are Fig. 3. STEM a) BF image for Bi2WO6 and (b) HAADF image for Bi2Mo0.5W0.5O6.

associated to the stretches of the W-O or Mo-O bands. This region is

particularly interesting because one can notice differences in the The Fig. 3(a) shows the bright field (BF) STEM micrograph for
stretching vibrational bands that provide information on the structural rusellite Bi2WO6. The morphology of this material shows sheet-like
variations. One can appreciate these variations in the Raman spectra as shape in the nanometer range. Fig. 3(b) shows high angle annular dark
changes in the width and the relative intensity. One can observed that field (HAADF) STEM micrograph for Bi2Mo0.5W0.5O6. The micrograph
the triplet conformed by the Raman bands centered at 698 cm−1, reveals an agglomeration of plate-like nanoparticles. Both morpholo-
714 cm−1 and 724 cm−1 shows a gradual variation in the width and in gies indicate preferential growth of the crystal; it means that an
the intensity due to Mo incorporation as solid solution. This variation important influence in the crystal growth with a decrease in the crystal
was also observed for the doublet conformed by the Raman bands size as can be observe in Fig. 3(b).
located at 816 cm−1 and 831 cm−1. A little apparent change of Raman The UV–vis in diffuse reflectance (DRS) mode can be used to
shift is observed for the Raman band centered at 797 cm−1. The bands measure volume averaged band gap energy (Eg) values. For homo-
at 797 cm−1 and 816 cm−1 are attributed to antisymmetric and sym- genous semiconducting powder materials, DRS is the method of choice.
metric Ag modes of the WO6 octahedra [15]. It is important to notice The band gap values determined using this method for Bi2MoxW1−xO6
the shift position of Raman band located at 816 cm−1 to high samples are shown in the first column of Table 1. The EELS data
wavenumbers as Mo content increasing systematically. This shift provide alternative means for bandgap measurement that allows attain
suggests the possible rearrangement of the structure associated with high spatial resolution. Fig. 4 shows high energy-resolution EEL
the gradual substitution of W cation by Mo cation. Additionally, this spectra of Bi2MoxW1−xO6 in the low-loss region. One can observe the
result is associated to the different metal-oxygen bond length within zero loss peak center at 0 eV. A sharp onset, interpreted as the edge of
the WO6 and MoO6 octahedra. The corresponding metal-oxygen bond the band gap can be observed between 2 and 3 eV. Gu et al. [5]
length in the russellite and koechilinite, were determined according to reported previously that various spectra processing techniques are used
the following expression for the band gap energy measurements.
The first step of data analysis on the energy-loss spectrum consists
RM − O=0. 48239ln (32, 895/ υ), (1) in fitting the zero-loss peak by a Gaussian for the zero energy
calibration. Subsequently, the spectrum is corrected for multiple
where υ is the Raman stretching frequency in wavenumbers and R is scattering events by Fourier-Log (FLOG) deconvolution based on the
the metal-oxygen bond length in angstroms [18]. Our results shows a Fourier transform [6]. This procedure results in the energy loss
little difference for W-O =1.793 Å and for Mo-O =1.794 Å. According to function (ELF). Fig. 5 shows deconvoluted Bi2MoxW1−xO6 spectra.
Zhang et al. [15] the gradual increasing Mo concentration has an effect One can observe in these spectra that the zero loss peaks and the
on the Raman band at 797 cm−1 that is shifted to higher wavenumbers, multiple scattering were removed.
for this reason one can expected the small value for W-O bond length in In the next step is the Kramers-Kronig (K-K) analysis was used to
comparison with Mo-O bond length.

186
D. Morales-Cruz et al. Materials Science in Semiconductor Processing 63 (2017) 184–189

Table 1
Comparison of band gap (Eg) values obtained with JEM-ARM200F spectrometers for
Bi2MoxW1−xO6.

Sample UV–VIS DRS Linear fit Parabolic fit Reference [3]

Eg (eV) Eg (eV) Eg (eV) Eg (eV)

Bi2WO6 2.99 2.95 2.99 2.94

Bi2Mo0.03W0.97O6 – 2.73 2.74 –
Bi2Mo0.05W0.95O6 2.90 2.90 2.94 –
Bi2Mo0.1W0.9 O6 – 2.54 2.55 –
Bi2Mo0.15W0.85O6 2.78 2.77 2.79 –
Bi2Mo0.25W0.75O6 2.82 – – 2.69
Bi2Mo0.5W0.5O6 2.61 2.60 2.60 2.70
Bi2Mo0.7W0.3O6 2.66 – – –
Bi2Mo0.9W0.1O6 2.56 2.56 2.59 –
Bi2MoO6 2.67 2.66 2.69 2.72

Fig. 6. Energy-loss function (Im (−1/ε )) spectra for Bi2MoxW1−xO6 in the range de 0–
80 eV.

can observe in Fig. 7 that the band gap energy estimated by linear fit
method (dotted line) results to be approximately at 2.54–2.95 eV.
An alternative approximation is based on a previous work reported
by Rafferty and Brown [21]. This one consider a mathematical fit of the
form a(E-Eg) b, where a is scaling parameter, E is the energy, Eg is the
band gap energy, and b takes values of 0.5 or 1.5 depending whether is
direct transition or indirect transition [21–25]. An estimate of the band
gap using a (E-Eg) 0.5 fit suggests a direct electronic transition. Fig. 7
show a parabolic fit (dashed line) that estimate band gap energy at
2.55–2.99 eV.
All spectra in Fig. 7 were obtained at 80 keV, except of spectrum
Bi2Mo0.5W0.5O6 (Fig. 7(f)), which was obtained at 200 keV. Comparing
of the whole spectra we can observe a systematic behavior among
Fig. 4. High energy resolution valence EEL spectra for Bi2MoxW1−xO6. spectra. Therefore, our fit of all spectra at low energies is very
appropriate.
All spectra were analyzed, i. e. the energy band gap was obtained
using FLOG deconvolution method and fit to the energy loss function
by K-K analysis. The result of the both the linear and parabolic fit is
shown in Table 1. This one summarizes the results of the measured
band gap energies. All results are consistently reproduced on measure-
ments made on a series of different samples deposited under the same
conditions.
The theoretical volume plasmon energy ℏωP can be calculated for
the free electron gas model using the equation
ℏωP=ℏ(ne 2 / m ϵ 0)1/2 , (2)
where ℏ is Planck´s constant divided by 2π , n is the valence electron
density (valence electrons/m3), ϵ 0 is the dielectric permittivity of
vacuum, e is the electron charge, and m is the rest mass of the electron
[6]. The theoretical plasmon energy EP for Bi2Mo0.15W0.85O6 is 20.8 eV
which agrees reasonably well with the plasmon energy measured by
EELS (20.5 eV), for an electron density n =3×1029 e/m3.
Fig. 5. Zero-loss spectra deconvoluted showing the band-gap onset.
In our study we report the determination of the energy loss function
(Im (−1/ε )) and consecutive analysis to determine the energy band gap.
retrieve the ELF, which is related the reciprocal dielectric function
A problem with the EELS technique (as well as any other technique) is
(ELF = Im (−1/ε (E ))). Thus, K-K analysis was performed on single
the possible influence of surface contaminations [9]. As a result of the
scattering distribution obtained from the low loss spectra using FLOG
inevitable presence of noise, EELS spectra collected under the same
deconvolution implemented in the Gatan Digital Micrograph™ soft-
conditions will differ from experiment to experiment.
ware. Fig. 6 shows typical energy loss function spectra, which have
Our derived values for band gap energies are consistent when
been appropriately deconvolved (Fig. 5). We using the ELF for
comparing the applied method here. Our measured band gaps values
determination of the band gap energy. The band gap energy (Eg) can
are found to be in good agreement with the experimental values
be estimated from a linear fit to the onset in the ELF spectrum
reported in the literature [3]. In the case of linear fit, this model
according to Garvie et al. [19] and Park et al. [20]. Fig. 7 shows the
depends much of the spectrum shape of the energy loss function.
band gap energy determination from eight sample; these images show
Therefore, is inappropriate to make a fit in EELS. The parabolic
the band gap region in detail. Finally, we applied the linear fit method
approximation suggests that this mathematical fit (a (E-Eg) b) is the
proposed by Garvie and Park in order to determine the band gap. One
most appropriate method for approach the band gap energy. One can

187
D. Morales-Cruz et al. Materials Science in Semiconductor Processing 63 (2017) 184–189

Fig. 7. Energy-loss function spectra for Bi2MoxW1−xO6 retrieved using linear fit and parabolic fit, in the range 0–16 eV.

be observe that in Table 1 the differences between the Eg evolutions 4. Conclusions

determined in the individual nanoparticles respect to the bulk results is
small. This work shows the analysis of the low loss energy region using

188
D. Morales-Cruz et al. Materials Science in Semiconductor Processing 63 (2017) 184–189

HR-EELS in a TEM which is important to determine the band gap Eg [2] J. Tang, Z. Zou, J. Ye, Catal. Lett. 92 (2004) 53–56.
[3] L. Zhou, M.M. Yu, J. Yang, Y. Wang, C. Yu, J. Phys. Chem. C 114 (2010)
energy of individual nanoparticles. The analysis of ELF spectra was 18812–18818.
performed for individual particles of Bi2MoxW1−xO6. XRD patterns and [4] A. Etogo, R. Liu, J. Ren, L. Qi, C. Zheng, J. Ning, Y. Zhong, Y. Hu, J. Mater. Chem.
Raman spectra confirm a gradual increase of Mo as a solid solution and A 4 (2016) 13242–13250.
[5] L. Gu, V. Srot, W. Sigle, C. Koch, P. van Aken, F. Scholz, S.B. Thapa, C. Kirchner,
crystal size change from 35 nm for rusellite to 62 nm for koechilinite. M. Jetter, M. Rühle, Phys. Rev. B 75 (2007) 195214.
The comparison between two approximations linear and mathematical [6] R.F. Egerton, Electron Energy Loss Spectroscopy in the Electron Microscope, 2nd
fit was performed in the ELF which was determined by the K-K edition, Plenum Press, New York, 1996.
[7] T. Kiguchi, N. Wakiya, K. Shinozaki, T.J. Konno, Mater. Sci. Eng. B 161 (2009)
analysis. The fitting suggests a direct electronic transition for all 160–165.
samples. The parabolic approximation suggests that is the most [8] O. Prytz, O.M. Lovvik, J. Tafto, Phys. Rev. B 74 (2006) 245109.
appropriate method for determining the band gap energy. In order to [9] R.F. Egerton, P. Li, M. Malac, Micron 35 (2004) 399–409.
[10] K.S. Knight, Mineral. Mag. 56 (1992) 399–409.
confirm these results, UV–Vis in reflectance diffuse mode was per-
[11] N.A. McDowell, K.S. Knight, P. Lightfoot, Chem. -Eur. J. 12 (2006) 1493–1499.
formed in the bulk. The UV–Vis analysis of each spectrum support the [12] L. Zhang, T. Xu, X. Zhao, Y. Zhu, Appl. Catal. B Environ. 98 (2010) 138–146.
direct electronic transition obtaining similar Eg values. [13] T. Saison, P. Gras, N. Chemin, C. Chanéac, O. Durupthy, V. Brezová, C. Colbeau-
Justin, J.P. Jolivet, J. Phys. Chem. C 117 (2013) 22656–22666.
[14] C. Zhang, Y.F. Zhu, Chem. Mater. 17 (2005) 3537–3545.
Acknowledgments [15] L. Zhang, Y. Man, Y. Zhu, ACS Catal. 1 (2011) 841–848.
[16] P.R. Graves, G. Hua, S. Myhra, J.G. Thompson, J. Solid State Chem. 114 (1995)
D. Morales thanks to CONACyT postdoctoral scholarship. F 112–122.
[17] F.D. Hardcastle, I.E. Wachs, J. Raman Spectrosc. 26 (1995) 397–405.
Paraguay-Delgado and M. Malac acknowledge the support from [18] F.D. Hardcastle, I.E. Wachs, J. Phys. Chem. 95 (1991) 10763–10772.
NINT/NRC. Thanks to the CEMIE-Sol Project P25 from Sener- [19] L.A.J. Garvie, P. Rez, J.R. Alvarez, P.R. Buseck, Solid State Commun. 106 (1998)
CONACyT project. Thanks to W. Antunez, C. Ornelas, E. Lestarjette 303–307.
[20] J. Park, S. Heo, J.-G. Chung, H. Kim, H. Lee, K. Kim, G.-S. Park, Ultramicroscopy
for their technical help at Nanotech-Cimav-Mexico. G. Herrera-Pérez 109 (2009) 1183–1188.
would like to thank CONACyT, México, through the SNI 1; Basic [21] B. Rafferty, L.M. Brown, Phys. Rev. B 58 (1998) 10326.
Research No. 253605; Catedra Research Fellow No. 530 and Catedra [22] R. Erni, N.D. Browning, Ultramicroscopy 104 (2005) 176–192.
[23] S. Lazar, G.A. Botton, M.Y. Wu, F.D. Tichelaar, H.W. Zandbergen, Ultramicroscopy
Research No. 2563.
96 (2003) 535–546.
[24] I. Alexandrou, A.J. Papworth, C.J. Kiely, G.A.J. Amaratunga, L.M. Brown, Diam.
References Relat. Mater. 13 (2004) 1408–1411.
[25] S. Schamm, G. Zanchi, Ultramicroscopy 88 (2001) 211–217.

[1] X. Zhao, T. Xu, W. Yao, C. Zhang, Y. Zhu, Appl. Catal. B Environ. 72 (2007) 92–97.

189