Nb ratio in LiNbO3 crystals. Effect of polarity and mechanical processing on LiNbO3 surface chemical composition

Applied Surface Science 389 (2016) 387–394

Contents lists available at ScienceDirect

Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc

XPS study of Li/Nb ratio in LiNbO3 crystals. Effect of polarity and mechanical processing on LiNbO3 surface chemical composition E.A. Skryleva ∗ , I.V. Kubasov, Ph.V. Kiryukhantsev-Korneev, B.R. Senatulin, R.N. Zhukov, K.V. Zakutailov, M.D. Malinkovich, Yu.N. Parkhomenko National University of Science and Technology MISiS, Leninsky prosp., 4, Moscow, 119049, Russia

a r t i c l e

i n f o

Article history: Received 19 April 2016 Received in revised form 14 July 2016 Accepted 17 July 2016 Available online 19 July 2016 Keywords: Lithium niobate XPS Li/Nb atomic ratio Polarity Mechanical processing Adsorption

a b s t r a c t Different sections of congruent lithium niobate (CLN) crystals have been studied using X-ray photoelectron spectroscopy (XPS). We have developed a method for measuring the lithium-to-niobium atomic ratio Li/Nb from the ratio of the Li1s and Nb4s spectral integral intensities with an overall error of within 8 %. Polarity and mechanical processing affect the Li/Nb ratio on CLN crystal surfaces. The Li/Nb ratio is within the tolerance (0.946 ± 0.074) on the negative cleave surface Z, and there is excess lithium (Li/Nb = 1.25 ± 0.10) on the positive surface. The positive surfaces of the 128◦ Y cut plates after long exposure to air exhibit LiOH formation indications (obvious lithium excess, higher Li1s spectral binding energy and a wide additional peak in the O1s spectrum produced by nonstructural oxygen). XPS and glow discharge optical electron spectroscopy showed that mechanical processing of differently oriented crystals (X, Z and 128◦ Y) and different polarities dramatically reduces the Li/Nb ratio. In situ fluorine adsorption experiments revealed the following regularities: fluorine adsorption only occurred on crystal cleaves and was not observed for mechanically processed specimens. Positive cleave surfaces have substantially higher fluorine adsorption capacity compared to negative ones. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Ferroelectric lithium niobate crystals LiNbO3 (LN) are widely used in functional devices of acoustic and optical electronics (modulators, gates, surface acoustic wave filters, piezoelectric ultrasonic emitters, piezoelectric actuators, various components of integral optics devices etc.) due to their unique electrooptical, piezoelectric and nonlinear properties. It is well known that LiNbO3 exists in a wide range of solid solutions, i.e. from 44.5 to 50.5 mol % Li2 O. Most of commercially grown crystals have chemical compositions near the congruent melting point, i.e. 48.3–48.6 mol % Li2 O [1]. Congruent crystals may have spatial composition inhomogeneity expressed as fluctuations of cations atomic ratio, Li/Nb.

∗ Corresponding author. E-mail addresses: [email protected] (E.A. Skryleva), [email protected] (I.V. Kubasov), [email protected] (Ph.V. Kiryukhantsev-Korneev), [email protected] (B.R. Senatulin), rom [email protected] (R.N. Zhukov), [email protected] (K.V. Zakutailov), [email protected] (M.D. Malinkovich), [email protected] (Yu.N. Parkhomenko). http://dx.doi.org/10.1016/j.apsusc.2016.07.108 0169-4332/© 2016 Elsevier B.V. All rights reserved.

Even greater Li/Nb ratio deviations are typical of LiNbO3 coatings used as electrooptical and ferroelectric layers in various devices. Lithium deficiency in these films is detected by X-ray diffraction as the presence of the LiNb3 O8 phase [2,3]. Synthesis of layers with the optimum composition requires Li/Nb ratio control. Li/Nb ratio is of special importance for surface chemical composition analysis of LiNbO3 coatings and crystals. Polar and nonpolar LiNbO3 cut surfaces have become independent objects of analysis in theoretical works [4–8]. The calculations for the cut surfaces using a first-principles density functional theory demonstrated that the surface stoichiometry differs from that for a bulk-terminated face. For the LN Z-cut, a strong influence of the spontaneous polarization of the material on the surface structure was found. The stoichiometry of the positive and negative surfaces under the same experimental conditions should be quite different: the −Nb-O3Li2 and the O-Li- terminations for the positive and negative faces respectively [4,7]. For the nonpolar LN X-cut, the dominant −Li12 termination was predicted by density functional theory and atomic force microscopy supported [7,8]. Analysis of the structural models of the most stable LN surface terminations [4,7,8] was made in the assumption of the absence of foreign adsorbates. The effect of adsorbed molecules on the LN Z-cut surface under ambient conditions is investigated elsewhere

388

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

[6]. The adsorption of the main air components at the LN surface was simulated from first-principles, and striking differences in the adsorption behavior at oppositely polarized surfaces were found. A theoretical study of surface reconstruction and surface charge at different temperatures [5] has shown that all the thermodynamically stable surfaces formed at different temperatures reduce their surface charge, suggesting that the compensation of polarization charge is a driving force for the observed structural modifications. Various analytical tools are used for LiNbO3 crystal surface investigation [8–11], e.g. atomic force microscopy, coaxial-impact collision ion scattering spectroscopy (CAICISS), ion scattering spectroscopy (ISS), temperature programmed desorption (TPD), reflection high energy electron diffraction (RHEED), low energy electron diffraction (LEED), X-ray and ultraviolet photoelectron spectroscopy (XPS and UPS). However, those works do not provide Li/Nb ratio data although XPS offered that possibility because fundamentals of quantitative analysis were set forth long ago by P. Steiner and H. Hochst [12]. It seems that the absence of Li/Nb ratio data is caused by the very low intensity of the Li1s photoelectron line. The ISS method does not detect lithium because of its low atomic weight. Raman spectroscopy is often used for assessing LiNbO3 stoichiometry. It was shown [13] that the widths of the E(TO) mode of the 150 cm−1 line and the A1 mode of the 872 cm−1 line depend on stoichiometry. Transition from stoichiometric to congruent composition leads to an increase in the width of the E(TO) line from 6.75 cm−1 to 10.97 cm−1 and the an increase in the width of the A1 mode from 20.1 cm−1 to 30.35 cm−1 . Taking into account the error of the measurements, the sensitivity of the methods could be assessed to be 0.06 mol % Li2 O. However, this is only an indirect estimate as the shape and position of the lines also depend on structural defects that are not always avoidable. In this work we used X-ray photoelectron spectroscopy (XPS) as the main analytical tool in spite of the difficulties caused by the very low intensity of the Li1s line. Glow discharge optical electron spectroscopy (GDOES) was used as an additional tool for it is widely used for rapid product control in metallurgy and can be adapted to the analysis of oxide crystals and coatings [14]. The initial aim of this work was to develop an XPS Li/Nb ratio assessment method and to estimate the potential error. Congruent lithium niobate (CLN) crystals were used as reference specimens. Test specimens were cut and cleaved out of differently shaped crystals with various crystallographic orientations; in addition, commercial treated plates were analyzed. The Li/Nb ratio fluctuations proved to be high for different surfaces. Lithium and niobium peak intensity ratios were reproducible and close to the theoretical ones [12,15,16] for non-polar cleave specimens, and we therefore used these specimens for testing our Li/Nb ratio measurement technique. The new quantitative analytical tool was used for controlling Li/Nb ratio deviations observed on the surfaces of polar cleaves and for analyzing the dependence of these deviations on charge sign and mechanical processing, and this became the final aim of this work.

Fig. 1. Photo of the specimen obtained from rectangular batch C crystal by cutting with a diamond ring in the peripheral area and subsequent breaking in the central part ex-situ before loading into the spectrometer: (1) central part − Z cleave and (2) peripheral area − Z cut.

used. The crystals differed in shapes, sizes and surface treatment, see Table 1. Cleaved specimens were fractured ex-situ before loading into the spectrometer. The cleave surfaces were not smooth but had small sections that were either parallel to the Z axis, or perpendicular to the Z axis or perpendicular to the X axis with a satisfactory accuracy. The rectangular batch C and batch D crystals were first notched with a diamond ring in the peripheral areas so a 2–3 mm diam. central part remained intact, and before loading into the spectrometer the specimens were broken. As a result two samples of different polarity were obtained and each individual specimen had two sections: the central one with a cleaved surface and the peripheral one with a cut surface (Fig. 1). 2.2. Polarity control Crystal surface polarity was controlled using a rapid method for detecting the sign of the charge induced on the crystal edges during its straining in the direction perpendicular to the surface; a storage oscilloscope was used. A slight impact of the oscilloscope probe on the test surface produced elastic compression strain in the crystal, and excess charge accumulated on the crystal edges located perpendicularly to the strain direction. A negative charge was induced on the surface located at the end of the dipole moment vector, and a positive charge was induced on the surface located at the beginning of the dipole moment vector. The surface electric potential was measured with the oscilloscope which displayed a wave-shaped extinction signal. If the first half-wave had a negative sign, then a negative charge was induced on the respective surface and the respective surface had the “+” sign, and on the contrary, if the first half-wave had a positive sign, then that surface had the “–” sign. The other half-waves of the signal are associated with crystal restoration to the initial unstrained state and are therefore of no interest for polar surface sign determination. 2.3. XPS

2. Material and methods 2.1. CLN single crystals Single crystal specimens were made from commercially available congruent Li0,946 NbO2,973 single crystals (ELAN Ltd, SaintPetersburg, Bogoroditsk Plant of Technochemical Products) that were Cz-grown from the melt containing 48.6 mol % Li2 O and 51.4 mol % Nb2 O5 and polarized along the [0001] crystallographic direction, the Z axis. Polar Z- and Y 128◦ -cut and X-cut crystals were

XPS studies were carried out on two instruments, PHI 5500 ESCA and PHI 5000 VersaProbe II. In the PHI 5500 ESCA, photoemission was excited by standard MgK␣ radiation (h␯ = 1253.6 eV) with a power of 250 W. The analysis area diameter was 1100 ␮m. In the VersaProbe II spectrometer we used monochromatic Al k␣ radiation with a power of 25 or 50 W and analysis area diameter of 100 or 200 ␮m, respectively. The residual gas pressure in the chamber was within 1 × 10−7 Pa. The full elemental composition was determined from the survey spectra in the 0–1100 eV range recorded at the analyzer pass energy

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

389

Table 1 CLN crystal parameters and specimen notations. Batch

Description

Analysis Area

Surface Charge

Specimen Notation

A

0.5 mm thick plates, 128◦ Y −cut

B C

1.75 mm thick plates, Z-cut Rectangular section 6.6 × 4.3 mm2 10 mm long crystals, Z-cut

D

Rectangular section 5 × 5 mm2 10 mm long crystals, X-cut

Polished surface Ground surface Cleaved parallel to the Z axis Cleaved perpendicular to the Z axis Cut perpendicular to the Z axis Cleaved perpendicular to the Z axis Cut perpendicular to the Z axis Cleaved perpendicular to the X axis Cut perpendicular to the X axis

negative positive – positive positive negative negative – –

A “–” A “+” Z cleave II Z cleave “+” Z cut “+” Z cleave “–” Z cut “–” X cleave X cut

2.4. GDOES O1s

Nb4s Li1s

70

65

60

55

Nb3d

50

Nb3p

O KLL Nb3s

C1s Nb4p

1000

800

600

400

200

0

Binding energy (eV) Fig. 2. Survey spectrum of the X cleave specimen; Inset: 50–70 eV spectral range.

Epass = 93.9 eV and a data density of 0.8 eV/step. The high resolution C1s, Nb3d and O1s spectra were recorded at Epass = 11.75 eV and a 0.1 eV/step. The binding energy scale was calibrated by the Nb3d5/2 peak, 207.1 eV. The very low intensity of the Li1s line (the photoabsorption cross-section of the Li1s level is almost 100 times as low as that for the Nb3d line [15]) and its small width preclude the use of survey spectra for assessing lithium concentration using the elemental relative sensitivity factors. To assess the lithium and niobium atomic ratio we recorded a short range of the spectrum, 50–70 eV, which contained the photoelectron lines Li1s and Nb4s (Fig. 2). The I(Li1s)/I(Nb4s) intensity ratio does not depend on the thickness of surface hydrocarbon contamination layers because the photoelectrons emitted from the Li1s and Nb4s levels have nearly the same kinetic energies. Therefore analysis can be carried out on the initial surface with any adsorbed impurities, and no ion etching is required that can change the CLN composition [17]. If Mg K␣ radiation is used (h␯ = 1253.6 eV), the inelastic mean free path ␭ in this spectral range is 2.3 nm; for Al k␣ radiation (h␯ = 1486.6 eV) it is 2.7 nm [18]. The Li1s and Nb4s peak intensities were measured after nonlinear background subtraction (Iterated Shirley) and 50–70 eV spectrum range approximation using the nonlinear least squares method. The narrowness of this spectral range allows its recording with a sufficiently high resolution, Epass = 23.5–29.35 eV, and a data density of 0.125–0.25 eV/step. These recording parameters were chosen so to obtain the maximum signal-to-noise ratios (100 + ) with the smallest Li1s and Nb4s line overlap. The spectra were approximated using the nonlinear least squares method; the Gaussian was used for the Li1s peak, and asymmetric Gaussian-Lorentzian shape was used for the Nb4s peak (Fig. 3 left panel).

Glow discharge optical electron spectroscopy is based on recording the intensities of element lines at a selected wavelength simultaneously with sample etching in an electric discharge. The measurements were carried out on a Profiler 2 Horiba Jobin Yvon instrument. Niobium and oxygen were detected by the second order Nb line (316.340 nm) and the first order O line (130.217 nm) with a polychromator. Lithium was detected with a monochromator set to 460.289 nm (the first order Li line). For checking the Li and Nb sensitivity of the instrument we used Horiba Jobin Yvon’s references and Al-Mg-Ca-Li alloy. GDOES studies of LN specimen bear methodical difficulties that are primarily caused by the low electric and thermal conductivity of the specimens and their low resistivity to ion etching thermal impact. The etching mode was as follows: HF pulse frequency 3 kHz, power 25 W, output power 3 W, pressure 850 Pa, duration 30 s. After etching the crater depth was measured with a Tencor AlphaStep IQ profilometer. The line intensities were averaged over the whole etching depth which was about 100 nm. 3. Results 3.1. Non-polar CLN crystal cleaves and development of Li/Nb measurement technique The non-polar cleave specimen were the batch D crystal cleave perpendicular to the X axis (X cleave) and the batch B cleave perpendicular to the Z axis (Z cleave II). Column 4 inTable 2 shows the I(Li1s)/I(Nb4s) integral intensity ratios for non-polar X cleave and Z cleave II specimens (column 1), calculated from the Li1s and Nb4s XPS spectra (Fig. 3 left panel). The measurements were carried out on two spectrometers (column 2) using two radiation sources: Mg k␣ and Al k␣ (column 3). For comparison column 5 shows the I(Li1s)/I(Nb4s) ratios calculated for the Li0,946 NbO2,973 composition using Eq. (1). In both cases electrons are emitted from the S shell (Li1s and Nb4s), and hence the asymmetry parameter ␤ for the angular dependence of the photoabsorption cross-section is the same. As inelastic mean free path ␭ is also the same, the I(Li1s)/I(Nb4s) integral intensity ratios are determined by the following equation: I(Li1s)/I(Nb4s) = 0, 946(Li1s)/(Nb4s)

(1)

where ␴(Li1s) and ␴(Nb4s) are the full photoabsorption crosssections of the Li1s and Nb4s levels for free atoms calculated by Scofield J.H. for Mg k␣ and Al k␣ [15]: ␴(Li1s) = 0.0593, ␴(Nb4s) = 0.333 for Mg k␣ and ␴(Li1s) = 0.0568, ␴(Nb4s) = 0.402 for Al k␣. Columns 6 and 7 shows theoretical and experimental I(Li1s)/I(Nb4s) obtained by P. Steiner and H. Hochst [12] for Al k␣. If I(Li1s)/I(Nb4s) for Li0,946 NbO2,973 is calculated using the relative sensitivity factors of Li1s and Nb4s for an averaged matrix (AMRSF)

390

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

Nb4s

O1s Li1s

a)

a)

Nb4s

O1s Li1s

b)

b)

Nb4s

O1s

c)

70

c)

Li1s

65

60

55

50

538

536

534

532

530

528

526

Binding energy (eV)

Binding energy (eV)

Fig. 3. (Color online) Li1s and Nb4s (left panel) and O1s (right panel) spectra of the (a) Z cleave “+”, (b) X cleave and (c) Z cut “+” specimens. The experimental spectra (continuous line) for the left panel were fit using two bands (dashed line): the Gaussian shape Li1s peak at 55,0 eV and the asymmetric Gaussian-Lorentzian shape Nb4s peak at 60,2 eV. The O1s experimental spectra (continuous line) for the right panel were fit using two Gaussian-Lorentzian shape bands (dashed line): the peak at 530.2 eV produced by structural oxygen and the peak in the 531.2–532.1 eV range produced by adsorbed oxygen.

Table 2 I(Li1s)/I(Nb4s) ratios for non-polar CLN specimen cleaves. Specimen

Instrument

Radiation

1 Li0,946 NbO2,973

2

X cleave

PHI 5500 ESCA PHI 5000 VersaProbe II PHI 5000 VersaProbe II

3 Mg k␣ Al k␣ Mg k␣ Al k␣ Al k␣

Z cleave II

I(Li1s)/I(Nb4s) Experiment

Calculation using Eq. (1)

Calculation [12]

Experiment [12]

Calculation [16]

4

5 0.168 0.134

6 – 0.137

7 – 0.133

8 0.168 0.134

0.164 ± 0.010 0.135 ± 0.007 0.136 ± 0.008

[16], the results are identical to those obtained using Eq. (1), see column 8. The experimental I(Li1s)/I(Nb4s) were obtained in this work by averaging spectral data taken at different specimen points (from 5 points for the X cleave specimen to 20 points for the X cleave II specimen). Furthermore, we varied recording parameters (Xray beam diameter and power, analyzer pass energy and energy step). The overall error of the measurement was estimated for a confidential probability of 0.95. The experimental I(Li1s)/I(Nb4s) data (0.164 ± 0.010 for Mg k␣ radiation and 0.136 ± 0.008 for Al k␣ radiation) agreed well with the calculated values within the confidential intervals, and we therefore considered it safe to accept these values as nominal ones, i.e. describing the bulk ionic composition of the CLN crystals. The Li/Nb ratio for the test specimens can be determined using the following equation: Li/Nb = 0.946Rm/Rnom,

(2)

where Rm is the measured I(Li1s)/I(Nb4s) intensity ratio for the test specimen and Rnom is the nominal I(Li1s)/I(Nb4s) intensity ratio which is accepted to be 0.164 and 0.136 for Mg k␣ and Al k␣ radiations, respectively. The overall Li/Nb U95 error for the above mentioned Rnom and Rm determined using Eq. (2) was approx. 8 % for both spectrometers. This error precludes the use of this method for detecting minor composition deviations, e.g. congruent and stoichiometric com-

positions. If Li/Nb deviates beyond the tolerance limit 0.87–1.02, composition changes can be detected by XPS, and the detection probability is the higher the greater is the deviation. The experimental spectral binding energies (Eb ) Li1s, Nb4s and O1s at Eb (Nb3d5/2) = 207.1 eV for non-polar cleave specimens were 55.0 ± 0.1 eV, 60.2 ± 0.1 eV and 530.2 ± 0.1 eV, respectively. 3.2. Polar cleaves of Z-cut specimen Polar cleaves were studied for two crystals with different polarities made from one rectangular crystal of batch C that contained a cleave area and a cut area (Fig. 1). Both “+” and “−” specimens were loaded into the VersaProbe II instrument on the same holder. The Li/Nb ratios were determined for the cleave areas using Eq. (2). For the Z cleave “–” specimen Li/Nb was 0.89–1.05 which differs only slightly from the nominal bulk value tolerance range 0.87–1.02, and for the Z cleave “+” specimen this ratio was noticeably higher, 1.15–1.35. This however proved to be not the only difference between different sign cleave surfaces. One more difference, not detectable at an early stage, could only be seen after a certain exposure in the chamber. Achieving the desired measurement accuracy was a timeconsuming task and required more than one working day. After the expiry of that period the survey spectra of the specimens indicated

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

0,6

Z «+»

F/Nb

0,5 0,4 0,3 0,2

Z «–»

0,1 0,0 0

1

2

3 4 Time (day)

5

6

2

688

684

ber atmosphere which was detected by the lithium niobate polar cleaves, the sensitivity of the “+” cleave being far higher. Analysis of the high-resolution F1s spectra revealed two peaks: the main peak at 684.7 eV and the additional one at 686.4–686.7 eV with an intensity of 13–19 %, see Fig. 4. The main peak is in the metal fluoride region and it is therefore reasonable to attribute it to LiF for which the reference peak position is 684.9 eV [19]. LiF formation indicates chemical interaction of fluorine with the crystal. The origin of the additional peak is not clear; we can assume it to be Li2 SiF6 because literary data for NbF5 peak position suggest a greater energy, 687.5 eV [20]. Survey spectra were also recorded throughout the experiment for the cut areas of the specimens. Fluorine adsorption was not observed on the cuts. 3.3. Mechanical processing

1

692

391

680

Binding energy (eV) Fig. 4. F/Nb atomic ratio for the Z “+” and Z “–” specimens as a function of analytical chamber exposure time (upper panel) and F1s spectrum of the Z “+” specimen (lower panel). The growth rate of F/Nb is far greater for Z “+”, the saturation limit being approx. 0.53 compared to only 0.13 for Z “–”. The F1s experimental spectrum (continuous line) was fit using two Gaussian-Lorentzian shape bands (dashed line): peak 1 at 684.7 eV produced by LiF and peak 2 at 686.5 eV with an intensity of 15 %.

the presence of a new element, fluorine. More detailed analysis of the survey spectra also showed silicon peaks that were less intense. This result was surprising taking into account the low overall residual gas pressure in the chamber, i.e. 3 × 10−8 Pa. Fluorine could be brought to the analytical chamber during an earlier ion etching study of a fluorine containing specimen. When the first indications of fluorine were observed it became clear that its concentration is higher on the positive surface. Further observations confirmed this conclusion. Fig. 4 shows fluorine concentration growth in two specimens observed during a 6-day long study. The growth rate of the fluorine to niobium atomic ratio F/Nb as determined from the F1s to Nb3d line intensity ratio divided by the elemental sensitivity factors was far greater in the “+” specimen, the saturation limit being approx. 0.53 compared to only 0.13 for the “–” specimen. Fluorine concentration growth was accompanied by an increase in silicon concentration; a rough F/Si estimate based on the survey spectra was 5–6 for the “+” specimen. This is greater than 4 (SiF4 ) but as there are no other possible variants one can assume that silicon tetrafluoride was the component in the residual cham-

Studies of the peripheral areas of the cuts (diamond ring, water) of the X cleave and the Z cleave specimens showed that the Li/Nb ratio in these areas was lower than the nominal values for the bulk. From comparison of the Li1s and Nb4s spectra a) and c) in Fig. 3 from areas 1 and 2 of the same crystal (Fig. 1) is clearly seen that the I(Li1s)/I(Nb4s) intensity ratio for Z cut “+” is less than that for Z cleave “+”. Li/Nb atomic ratio obtained by Eq. (2) varied on the cuts from 0.40 to 0.83. Comparison of the Z cut “+” and Z cut “–” specimens did not reveal any difference, the values varied over a relatively wide range, indicating lithium deficiency. Similarly, there was no difference between the Z cuts and X cuts. In the survey spectra of these areas the relative intensity of the C1s peak was several times as high as for the cleaves having the same notations. The oxygen spectra contained the structural oxygen peak at 530.2 eV and a relatively intense and wide peak in the 531.2–532.1 range associated with adsorbed oxygen (Fig. 3 right panel). As mentioned above, fluorine adsorption was not observed on these cut areas. To understand the processes accounting for lithium depletion we refer to the results of another dry mechanical processing experiment. The specimens were dry ground by butts of crystals that were similar to the specimens; this was possible for batch C and batch D specimens having narrow rectangular shapes. The resultant surface had remaining fine particles, and there Li/Nb reached 5, the Li1s binding energy grew to 55.4 eV, the intensity of the additional O1s peak at 532.1 eV was almost 70 % and the C1s spectrum had a new peak at 289.5 eV, i.e. the fine particles were Li2 CO3 . These results were obtained for the Z and X cleaves. There is selective lithium precipitation producing a separate compound; similar processes can occur during wet mechanical processing (LiOH, Li2 CO3 ) but the fine particles are washed away and the surface finally becomes depleted of lithium. 3.4. Polar surfaces of 128◦ Y cut plates The polar surfaces of batch A 128◦ Y cut plates were studied without preliminary cleaning: we analyzed the surface chemical composition of plates that were stored for several years in the air. The opposite surfaces of the plates had different surface charge signs and roughness: “–” for polished surfaces and “+” for ground surfaces. The Li/Nb atomic ratios calculated using Eq. (2) from the Li1s and Nb4s spectra shown in Fir. 5 (left panel) are shown in Table 3 (column 4). The Li/Nb difference for the opposite surfaces was visible even despite the relatively large scatter of the ratios for different points on the surfaces of each surface (at least 5 measurement points). On the ground “+” surface, Li/Nb (1.7–2.3 for Mg k␣ and 1.7–2.1 for Al k␣) was far above the nominal values, while on the polished

392

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

O1s Nb4s

Li1s 1 2

O1s

Nb4s 1 Li1s Na2s 2

70

65 60 55 Binding energy (eV)

50 538

534 530 Binding energy (eV)

526

Fig. 5. (Color online) Li1s, Nb4s (left panel) and O1s (right panel) spectra of the A “+” (upper spectra) and A “–” (lower spectra) specimens. The Li1s and Nb4s experimental spectra (continuous line) were fit using two or three bands (dashed line): the Gaussian shape Li1s peak, the asymmetric Gaussian-Lorentzian shape Nb4s peak and the Gaussian shape Na2 s peak (for the A “+” specimen only). Eb (Li1s) is 55.5 eV and 55.0 eV for the A “+” and A “–” specimens respectively. The O1s experimental spectra (continuous line) were fit using two Gaussian-Lorentzian shape bands (dashed line): peak 1 at 530.2 eV produced by structural oxygen and peak 2 in the 531.8–532.3 range produced by LiOH. The fraction of O1s peak 2 is 62 % and 38 % for the A “+” and A “–” specimens respectively.

Table 3 Li/Nb and I(O2)/I(O) ratios for polar surfaces of batch A CLN crystals. Specimen

Instrument

Radiation

Li/Nb

I(O2)/I(O), %

1 A “–” A “+” A “–” A “+”

2 PHI 5500 ESCA PHI 5500 ESCA VersaProbe II VersaProbe II

3 Mg k␣ Mg k␣ Al k␣ Al k␣

4 0.6–1.0 1.7–2.3 0.7–0.8 1.7–2.1

5 34 62 38 64

The XPS data on different Li/Nb for polar surfaces of 128◦ Y cut plates refer to the superficial layers with a thickness of not greater than 3␭, i.e. 8 nm. The difference in Li/Nb for opposite surfaces of batch A plates was also confirmed by GDOES. The lithium to niobium peak intensity ratio for the “+” surface proved to be 1.1 times lower that for the “–” surface. 3.5. Effect of mechanical processing on GDOES Li/Nb data

“–” surface Li/Nb (0.6–1.0 for Mg k␣ and 0.7–0.8 for Al k␣) proved to be slightly below the nominal range 0.87–1.02. Analysis of the oxygen spectra (Fig. 5 right panel) and the binding energies of the Li1s spectra also revealed differences. For the cleaved specimens Eb (Li1s) was 55.0 ± 0.1 eV and the O1s spectrum contained in fact only one peak with Eb = 530.2 eV with the fraction of second peak (531.4–531.7 eV) being within 10 %. For the mechanically treated surfaces of the batch A crystals the O1s spectrum contained two peaks: narrow peak 1 of structural oxygen at 530.2 eV and wide peak 2 in the 531.8–532.3 range. The fraction of peak 2, I(O2)/I(O), for the A “+” specimen was 62–64 % and for the A “–” specimen its fraction was 34–38 %. These data are provided in column 5 of Table 3. Eb (Li1s) for the A “+” specimen was 55.20–55.5 eV and for the A “–” specimen, 54.9–55.0 eV. These two phenomena, i.e. the high content of surface nonstructural oxygen and the positive shift of the Li1s line for the “+” surface in comparison with the “–” surface, can be accounted for by the formation of lithium hydroxide LiOH. According to literary data [21,22], Eb (O1s) is 531.0–531.5 eV in LiOH and 531.9–532.2 in Li2 CO3 . This very wide peak can cover both these states along with the adsorbed oxygen and water molecules. However, the impossibility to clearly divide the peaks for nonstructural oxygen of different chemical nature in the wide high-energy region of the O1s spectrum does not allow one to draw any definitive conclusion. Li2 CO3 can be excluded from consideration because there is no respective peak in the C1s spectrum near 290.0 eV.

To confirm lithium concentration reduction due to mechanical processing we carried out a special experiment for batch B plate specimens that were preliminarily annealed with a temperature gradient and therefore had a polydomain structure and uncharged surfaces, as well as for the A “+” and A “–” specimens. We measured lithium to niobium peak intensity ratios ILi /INb for specimens before and after mechanical processing. The specimens were mechanically treated with 120–130 ␮m grain emery paper for 1 min. Specimen surfaces before and after grinding were washed with isopropyl alcohol. The roughness factors Ra measured for the same plate before and after mechanical processing were 0.11–0.14 ␮m and 1.0–1.5 ␮m, respectively. The measurement results are illustrated by the ILi /INb diagram for mechanically processed specimens normalized to ILi /INb for the initial specimens (Fig. 6). It can be well seen that mechanical processing of all the specimens caused a distinct reduction of Li/Nb. In the non-polar specimen B the ILi /INb intensity ratio after mechanical processing was on the average 0.65 of the initial value and in the A “+” and A “–” specimens it was even lower, 0.4 of the initial value. Surface charge sign was not found to affect this reduction. 4. Discussion Analysis of the XPS Li/Nb data for the surfaces of different test specimens (Fig. 7) shows that they are beyond the tolerance limit

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

ILi/INb

0,6

0,4

0,2

0 A "–"

A "+"

B

Fig. 6. GDOES data − ILi /INb optical line intensity ratio for mechanically treated specimens normalized to ILi /INb for initial specimens. In the non-polar specimen B the ILi /INb intensity ratio after mechanical processing was 0.65 of the initial value and in the A “+” and A “–” specimens it was 0.4 of the initial value.

393

only shows itself if the surface does not undergo any treatment, i.e. on cleaved surfaces, this regularity cannot be ignored. The effect of polarity was most clearly seen in a fluorine adsorption experiment with exposure in the analytical chamber. The “+” cleave surface proved to adsorb fluorine more efficiently: the F/Nb concentration ratio in the saturation area was 5 times that for the “–” surface. Analysis of the F1s spectra showed the presence of lithium fluoride, and therefore these processes can be attributed to chemisorption. Adjacent specimen areas that were treated by cutting did not adsorb fluorine. In a special earlier experiment [11] for BF3 adsorption on polar (0001) surfaces at 125 K, the F/Nb curves were similar regardless of polarity. Our data suggest that the cause of the similar adsorption behavior of the specimens was their long history before the experiment. The plate surfaces were first mechanically processed then exposed to long-term annealing at 900 ◦ C in air and finally oxygen plasma treated at 300 ◦ C in vacuum. The charge passivation mechanisms typical of these treatments most likely completed there and hence the initial surface state was modified. Reduction of the Li/Nb ratio as a result of mechanical processing was confirmed by two methods, XPS and GDOES. Dry grinding and chemical composition analysis of the fine particles allow one to simulate the processes occurring during mechanical processing with water (or with isopropyl alcohol as in the GDOES experiment), i.e. lithium precipitation in the form of Li2 CO3 or LiOH particles followed by particle removal by the process liquid. No effect of crystal polarity on Li/Nb ratio reduction due to mechanical processing was observed.

5. Summary Fig. 7. Li/Nb ranges as determined by XPS for CLN crystal specimen surfaces, dashed lines − 95 % tolerance limit obtained for nominal values at non-polar cleaved surfaces (X cleave and Z cleave II). The Li/Nb ranges for the Z “–” cleave are within the permissible range. The Li/Nb ranges for the Z “+” cleave are above the upper limit and for A “ + ” they are much higher than the upper limit. The Li/Nb ranges for the mechanically treated specimens (Z cut, X cut, X polish and A “–”) are below the lower level.

0.87–1.02 obtained for non-polar cleaved surfaces (X cleave and Z cleave II). Different Li/Nb values were obtained for polar Z cleaves: at “+” they were higher than the nominal ones (1.15–1.35), while at “−” they were within the acceptable range. The excess of lithium at the plus-cut could be accounted for by different stoichiometries of the polar surfaces predicted in theoretical works [4,7]. Unfortunately, a number of reasons preclude comparison of our experimental data with theory. The models of theoretical works are based on the perfect LiNbO3 structures of stoichiometric compositions in the absence of surface contact with the ambience. Our experimental data were obtained for crystals of a congruent composition for which an increase in temperature may already change the composition at moderate temperatures, ∼600 K [5]. Cleaved specimens were fractured ex-situ, and adsorption of hydrocarbons was observed; the C1s peak is present in all the survey spectra (Fig. 2) indicating the participation of adsorbed molecules in relaxation processes. The higher Li/Nb at the plus than at the minus was also observed for specimens of processed surfaces of A 128◦ Y −cut series, and there the difference was even greater: 1.7–2.1 for the A”+” specimens and 0.7–0.8 for the A”−“ specimens. Analysis of the O1s spectra and the binding energies of the Li1s spectra in these specimens gave us quite a reasonable explanation for this difference. At the positive surface, long-term storage in air caused LiOH formation, i.e. the chemical reactivity of A”+” is higher than that of A”−“. Though, the pure effect of polarity on surface chemical composition

We show that XPS can be used for quantitative analysis of the Li/Nb ratio in LiNbO3 specimens with a total U95 error of within 8 %. On the surfaces of the CLN crystals the Li/Nb ratio may vary over a range exceeding the tolerance limit 0.946 ± 0.074 depending on surface polarity and mechanical processing. Li/Nb for the negative Z cleave surface is within the tolerance limit, and on the positive Z cleave surface Li/Nb is 1.25 ± 0.10. The obvious lighium excess at the positive side of the “old“128◦ Y cut specimens is caused by LiOH formation. Study of the mechanically processed polar surfaces of the batch A 128◦ Y cut plate specimens revealed an even greater lithium excess (Li/Nb reached 2) on the positive surface with an increase in the content of nonstructural oxygen on the surface (to 65 %) and a positive shift of the Li1s line (55.20–55.5 eV) on the positive surface compared to the negative surface and to the bulk (54.9–55.0 eV). We attribute these results to LiOH formation on the positive surfaces of the 128◦ Y cut plates. XPS and GDOES data showed that mechanical processing of crystals with different crystallographic orientations (X, Z and 128◦ Y) and different polarities reduces Li/Nb to 0.4–0.9. We assume that lithium depletion mechanism is refinement-enhanced lithium interaction with the ambient atmosphere, formation of Li2 CO3 or LiOH particles and further removal of these particles by the liquid. Study of fluorine adsorption on Z cleaves of different polarities within a single experiment showed that opposite surfaces have different fluorine adsorption capacities. Positive cleave surfaces adsorb fluorine more efficiently, the F/Nb concentration ratio in the saturation area being 5 times that for the negative surface. Fluorine was not observed in the cut specimen areas throughout the entire experiment. Thus the cuts did not adsorb fluorine which suggests a drastic reduction in the adsorption capacity of polar cleave surfaces after mechanical processing.

394

E.A. Skryleva et al. / Applied Surface Science 389 (2016) 387–394

Acknowledgments The authors thank Vladimir Antipov for his assistance in carrying out this work. The study was supported by the Ministry of Education and Science of the Russian Federation (Federal Targeted Program for Research and Development in Priority Areas of Development of the Russian Scientific and Technological Complex for 2014–2020) (Project ID RFMEFI57815 × 0102) References [1] P.F. Bordui, R.U. Norwood, C.D. Bird, G.D. Calve, Compositional uniformity in growth and poling of large-diameter lithium niobate crystals, J. Cryst. Growth 113 (1991) 61–68. [2] H.J. Kim, D.Y. Kim, J.Y. Ha, C.Y. Kang, M.Y. Sung, B.H. Cho, S.J. Yoon, H.J. Kim, Deposition of LiNbO3 thin films for selective etching, J. Electroceram 17 (2006) 933–935. [3] D.A. Kiselev, R.N. Zhukov, A.S. Bykov, M.I. Voronova, K.D. Shcherbachev, M.D. Malinkovich, Yu.N. Parkhomenko, Effect of annealing on the structure and phase composition of thin electro-optical lithium niobate films, Inorg. Mater. 50 (2014) 419–422. [4] S.V. Levchenko, A.M. Rappe, Influence of ferroelectric polarization on the equilibrium stoichiometry of lithium niobate (0001) surfaces, Phys. Rev. Lett. 100 (2008) 256101. [5] S. Sanna, R. Hölscher, W.G. Schmidt, Temperature dependent LiNbO3 (0001): surface reconstruction and surface charge, Appl. Surf. Sci. 301 (2014) 70–78. [6] R. Hölscher, W.G. Schmidt, S. Sanna, Modeling LiNbO3 surfaces at ambient conditions, J. Phys. Chem. C 118 (2014) 10213–10220. [7] Simone Sanna, Wolf Gero Schmidt, Lithium niobate X-cut, Y-cut, and Z-cut surfaces from ab initio theory, Phys. Rev. B 81 (2010) 214116. [8] S. Sanna, W.G. Schmidt, S. Rode, S. Klassen, A. Kühnle, Unraveling the LiNbO3 X-cut surface by atomic force microscopy and density functional theory, Phys. Rev. B 89 (2014) 075403. [9] Y. Yun, M. Li, D. Liao, L. Kampschulte, E.I. Altman, Geometric and electronic structure of positively and negatively poled LiNbO3 (0001) surfaces, Surf. Sci. 601 (2007) 4636–4647.

[10] T. Yamada, T. Choso, K. Tabata, E. Suzuki, Study of the topmost surface structure of a Y-Cut LiNbO3 single crystal with coaxial impact collision ion scattering spectroscopy (CAICISS), Appl. Surf. Sci. 171 (2001) 106–112. [11] M.W. Herdiech, H. Mönig, E.I. Altman, An X-ray photoelectron spectroscopy study of BF3 adsorption on positively and negatively poled LiNbO3 (0001), Surf. Sci. 626 (2014) 53–60. [12] P. Steiner, H. Hochst, X-ray excited photoelectron spectra of LiNbO3 , Z. Phys. B 35 (1979) 51–59. [13] P. Galinetto, M. Marinone, D. Grando, G. Samoggia, F. Caccavale, A. Morbiato, M. Musolino, Micro-Raman analysis on LiNbO3 substrates and surfaces: compositional homogeneity and effects of etching and polishing processes on structural properties, Opt. Laser. Eng. 45 (2007) 380–384. [14] P.V. Kiryukhantsev-Korneev, Elemental analysis of coatings by high-frequency glow discharge optical emission spectroscopy, Prot. Met. Phys. Chem. Surf. 48 (2012) 585–590. [15] J.H. Scofield, Hartree-Slater subshell photoionization cross-sections at 1254 and 1487 eV, J, Electron Spectrosc. 8 (1976) 129–137. [16] XPS and AES Average Matrix Relative Sensitivity Factors, Materials of the Centre of National Physical Laboratory Great Britain http://www.npl.co.uk/science-technology/surface-andnanoanalysis/services/xps-and-aes-average-matrix-relative-sensitivityfactors (accessed 15.04.16). [17] R. Courths, P. Steiner, H. Hochst, S. Hufner, Photoelectron-spectroscopy investigation and electronic properties of LiNbO3 crystal surfaces, Appl. Phys. 21 (1980) 345–352. [18] QUASES-IMFP-TPP2M software http://www.quases.com/products/quases-imfp-tpp2m/(accessed 15.04.16). [19] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, Handbook of X-Ray Photoelectron Spectroscopy, 1st ed., Perkin-Elmer Corp., Physical Electronics Division, Eden Prairie, MN, 1992. [20] Y. Luo, P. Wang, L.P. Ma, H.M. Cheng, Hydrogen sorption kinetics of MgH2 catalyzed with NbF5 , J. Alloys Compd. 453 (2008) 138–142. [21] A.C. Kozen, A.J. Pearse, C.F. Lin, M.A. Schroeder, M. Noked, S.B. Lee, G.W. Rubloff, Atomic layer deposition and in situ characterization of ultraclean lithium oxide and lithium hydroxide, J. Phys. Chem. C 118 (2014) 27749–27753. [22] K.P.C. Yao, D.G. Kwabi, R.A. Quinlan, A.N. Mansour, A. Grimaud, Y.L. Lee, Y.C. Lu, Y. Shao-Horn, Thermal stability of Li2 O2 and Li2 O for Li-Air batteries: In situ XRD and XPS studies, J. Electrochem. Soc. 160 (2013) A824–A831.