Nuclear Instruments and Methods in Physics Research B68 (1992) 227-230 North-Holland
Beam Interactions with ilatedals"Atoms
Characterization by Rutherford backscattering, elastic recoil and nuclear reaction analysis of near-surface modifications of glasses submitted to a DC potential b and C .A . Achete a, C .M . Lepienski Departamento de Fisica, Pontifcia Universidade Cat6lica do Rio de Janeiro, Cx. Postal 38071, Rio de Janeiro 22453 RJ, Brazil n CEPEL - Centro de Pesquisas de Energia Elétrica, Cx. Postal 2754, Rio de Janeiro 20001 RJ, Brazil Programa de Engenharia Metalkrgica e de Materiais, COPPE, UFRJ, Cx. Postal 68505, Rio de Janeiro 21910 RJ, Brazil
F .L . Freire Jr.
The near-surface modifications of glasses with different compositions were studied by the combination of three nuclear techniques for surface analysis : the resonant nuclear reaction 23Na(p, a)2° Ne at Er =591 .6 keV, 2 .2 MeV 4 He elastic recoil detection analysis and Rutherford backscattering spectrometry . Glass samples were submitted to a DC potential during different time intervals and temperatures. -The influence of the electrodes on the surface modifications was studied . The modified layer thickness was related to the total charge which passed through the sample during the treatment. Continuous recording of the current flowing through the glass shows that, for a fixed voltage, there is a general decrease in the current with time. This behaviour was related to the surface composition modifications determined by the nuclear techniques. 1. Introduction It is well known that alkali based glasses show an ionic migration when polarized with a DC potential  . Alkali ions drift under the influence of an electric field leaving a depleted layer close to the anode and creating an enriched ionic region near the cathode , where dendrites are formed . Such dendrites growing towards the anode can produce a conductive channel provoking the dielectric breakdown . From the technological point of view, a better understanding of the ionic migration in glasses is very important for the study of dielectric breakdown of the glasses employed as insulators in high-voltage DC transmission lines  . Besides this technological interest, some fundamental aspects of the ionic migration in glasses are still not completely understood, such as the determination of the charge carrier in the alkali-depleted layer near the anode, justifying thus new studies on this subject [2,41. In this work, we carried out nuclear surface analyses on glasses previously submitted to a DC potential with the aim of characterizing the surface composition modifications when different kinds of electrodes were used . We also establish correlations between surface modifications and the time evolution of the current which passes through the polarized sample . The laser induced pressure pulse technique (LIPP) was also used to measure the space charge distribution in order to provide complementary information . A detailed study on the LIPP characterization has been performed by the authors [5,6] . In these previous papers the space
charge and electric field distribution inside the sample were investigated . The present paper is focused on the surface characterization of the polarized glasses. 2. Experimental procedure The composition of the glasses used in the experiments is reported in table 1. Glass A is the one used as insulator in high-voltage DC transmission lines and glass B is a commercial soda-lime microscope slide . The samples were 1.2 mm thick and the electrode area was 2.0 cm 2. Two kinds of electrode were used: the first one was a vacuum deposited aluminiumcopper bilayer, each layer 200 ram thick . The copper overlayer prevents hydrogen to diffuse from the atmosphere into the sample and also preserves the alu-
Table 1 Glass composition [wt .%] BaO Fe 20 3 CaO K 20 Si02 A1 203 MgO Na 20
0168-583X/92/$05 .00 0 1992 - Elsevier Science Publishers B.V. All rights reserved
2.0 0.4 5.9 10.5 68.7 4.0 3.5 5.0
0.4 8.5 0.6 73.2 1 .3 3.7 12.3 V. BACKSCATTERING
F.L. Freire Jr. et al. / Near-surface modifications ofglasses
minium film from degradation. The second type of electrode were slightly pressed aluminium plates . Before being treated the glasses were etched in 4% HF during 4 min to remove the hydrated surface layer. Typically, a 2kV DC voltage was applied on the glass samples. The samples were kept in an oven at a temperature of about 150'C and the current flowing through the glass was continuously recorded . Before the nuclear analysis the electrodes were removed in order to obtain better depth resolution and sensitivity. To avoid surface charging, which could be detrimental to the analysis, the samples were coated with a 15 nm thick carbon film. Sodium depth profiles were obtained by means of the 'Na(p, a)`GNe nuclear reaction, which has a resonance at 591.6 keVwith a width of 0.6 keV. We used a proton beam provided by the 4 MV Van de Graaff accelerator at the Physics Department, PUC, Rio de Janeiro. The depth resolution on silicates was - 15 nm and the concentration profile was determined by increasing the proton energy by steps of 2 keV, corresponding to, - 30 nm. At each step the a-particles emitted with a dose of 5 WC of protons were counted. A NaCl crystal used as standard allowed both the absolute sodium concentration to be calculated and the position of the surface to be accurately determined. The maximum depth where sodium can be measured is 600 nm . Hydrogen depth profiling was measured using the elastic recoil detection (ERD) technique, with a probe beam of 2.2 MeV 'He' incident on the target at an angle of 75 ° with respect to the surface normal . Detection of the recoiled hydrogen particles was made with a surface barrier detector at a forward scattering angle of 30°'. A 10 pm mylar film covered the detector to prevent the detection of the scattered 4He particles . The depth resolution at the surface was of the order of 50 nm and the maximum depth where hydrogen could be measured was 700 nm. Calibration was obtained by using an a-Si : H (H atomic concentration of 5.6 x 10 21 cm -3 ) as a standard . Acareful control of the hydrogen loss induced by the beam during analysis was mandatory. In all cases the hydrogen signal was very stable. Rutherford backscattering spectrometry (RBS) was used for measuring. the depth profiles of the heavy elements. In this case the beam (2.2 MeV 'He') impinges the sample with a direction parallel to the normal surface. The backscattered ions were detected at an angle of 165° with the incident beam direction.
TEYPERATURE'150°C VOLTAGE .2kV THICKNESS 12-
-. . . .. . . .. 1. i .
Fig. I shows the time dependence of the current through the glasses for both kinds of electrodes. Be-
Fig. 1 . Current through the glass as a function of time . Full line: aluminium-copper bilayer electrodes . Dotted line : alu minium plates used as electrodes. For the composition of glasses A andB see table 1 . sides the large difference of resistivity between the glasses A and B, it is clear from the figure that for a fixed voltage, there is a general decrease in the current with time . In order to monitor the surface composition modifications caused by the treatment, several samples of glasses B were polarized with 16 kV cm - 1 at 150° C, for different time intervals. Figs . 2a-2c show the calcium, sodium and hydrogen depth profiles, in the region of the anode, for metallized samples through which total electrical charges of 16, 58 and 480 mC were passed, corresponding to treatment times of 120, 540 and 7200 s, respectively. All samples present a 0)
5 4 Z 3 O 2 F
W U Z O
4 Na . . .. ... H 3 2
5 --- ca Z
200 400 600 800 DEPTH( ..)
Fig. 2. Concentration depth profiles of hydrogen, sodium and calcium beneath the anode of soda-lime glasses (Al-Cu electrodes) which had passed charges of 8 (a), 29 (b) and 240 mC/cm2 (c). Samples treated during 120, 540 and 7200 s, respectively (2 kV DC, temperature 150'C, thickness 1.2 mm).
F.L. Freire Jr. et al. / Near-surface modifications ofglasses
ENERGY (keV )
Fig. 3 . ERD spectra from soda-lime glasses treated for 2 h, polarized with 16 kV /cm at 150°C .The spectra were taken on the anode face of the samples. The hydrogen depth profile obtained from the spectrum of the nonmetallized sample indicates a 5 x 1021 H/cm3 constant concentration . The 600 keV peak corresponds to the carbon film. sodium depleted layer whose thickness increases with time . Whenever this thickness can be measured by the nuclear reaction, there is a good agreement between the total charge transported through the glass and the equivalent charge of sodium (Na') originally present in the depleted layer. A calcium depletion is also observed near the surface, but this is more evident for longer treatments than the one necessary for theobservation of sodium depletion . Furthermore, calcium piles up in the region between where it is depleted and where the sodium reaches its bulk concentration. It is also observed that there is a hydrogen accumulation at the interface between the region where the calcium piles up and the unperturbed glass . The hydrogen probably comes from the residual water present on the surface before the electrodes have been deposited . When the aluminium plates were used as electrodes the RBS results do not show any modification of the
Fig. 4. Sodium depth profiles beneath anode for glasses polarized with 25 kV/cm for 2 h at a temperature of 200 ° C . calcium profiles when compared with the untreated sample (not shown in fig. 2). As in the previous case, the thickness of the sodium depleted layer is in good agreement with the total charge that was passed through the sample . The only significative difference was found for the hydrogen profiles, where the ERD spectra (see fig . 3) indicate that a 5 x 10 2) H/cm 3 constant concentration was found throughout the sodium depleted layer. The hydrogen concentration is nearly equal to the sodium bulk concentration. Fig. 4 presents the sodium profiles obtained from the nuclear reaction analysis. It is evident from this figure that the thickness of the sodium depleted layer depends on the electrode used. For the nonmetallized sample the sodium depleted layer is 180 nm thick and when aluminium-copper electrodes are used its thickness is 300 nm . RBS and ERD spectra from glass A (3 kV DC, 200°C for 2 h) are shown in fig. 5. Although the RBS spectrum for the nonmetallized sample seems to be quite similar to the one obtained for an untreated
-NON-METALLIZED GLASS -AL- ELECTRODES
Q9 0.8 GLASS A 0.3
n 0 C W
Fig. 5. (a) RBS spectra from glasses (composition A) polarized with 25 kV/cm during 2 h at a temperature of 200 ° C. Thick line : aluminium plates used as electrodes. Thin line: AI-Cu electrodes . The arrows indicate the surface position of the elements ; (b) ERD spectra from the same samples. The 600 keV peak corresponds to the carbon film . The spectra were taken from the anode side of the samples. V. BACKSCATTERING
F.L. Freire Jr. et al /Near-surface modificationsofglasses
sample, a potassium depleted layer 150 nm thick can be determined. As for the soda-lime glass, the ERD spectrum from these samples show a hydrated layer,
alkali depleted layer. The LIPP results do not have sufficient depth resolution to allow the exact determination of the depths of these charged layers. However,
The total charge that flowed through the sample, 27 mC, was three times greater than the equivalent of
bilayer electrodes are used, a double layer was formed by uncompensated negative charge (nonbridging oxy-
170 nm thick, with a concentration of 5 x 10 2 ' H/cm3, which is nearly egi!ai to the alkali b ,.,iic rorccnt-ation
sodium ions presented initially in the depleted layer. Nevertheless, it agrees well with the total alkali (K+ plus Na') charge contained in that layer, as measured by nuclear, techniques.
The ERD spectrum from the metallized sample of glass A presents the same feature previously observed for the soda-lime glass. In this spectrum, besides the hydrogen peak at 600 keV due to the carbon film, there is anotherone at arr energy which corresponds to the depth position of the interface between the sodium depleted layer and the unperturbed glass. For the same sample, the RBS spectrum indicates the movement of different species. In particular, it shows that the barium piles up at a depth of 200 nm. An unambiguous determination of the calcium and potassium profiles from the IRS spectrum is difficult . However, it is clear that there is a calcium and potassium depletion at the near-surface layer. In fact, we can assign the peak at 1350 keV in the RBS spectrum as being the calcium accumulation at a depth of 200 run, leaving a depleted layer of 170 nm thick. If potassium ions drift without pileup at the interface between the depleted layer and the unmodified one, as we observed for the sodium, the peak at 1200 keV would be
attributed to the
modifications on the height of RBS signal due to the reduction on the stopping power caused by the composition changes of the surface layer  . A potassium depleted layer with a thickness between 250 and 300 nm could be assigned. Charge measurements support this interpretation.
4. Discussions and conclusions Besides the nuclear techniques the samples were also characterized by the LIPP technique [5,6]. Those results do not depend on the glass composition but on the type of electrodes used . In fact, they show an electric field in the alkali depleted layer with opposite
signs for metallized and nonmetallized samples. The LIPP results for the metallized samples are attributed to the existence of a double charged layer beneath the
anode, the deeper one being positive. The nonmetallized samples presented a thin positive surface layer followed by a negative charge distributed along the
these results can be explained with the aid of those obtained usingthe nuclear techniques. So, when Al-Cu
gens) and by the calcium ions (or calcium and barium in the case of glass A) rich layer . For the nonmetallized
samples the hydrogen penetration does not exactly counterbalance the nonbridging oxygens, producing a negatively charged layer in the alkali depleted zone. Summarizing, in this work we have studied the
surface composition modifications in glasses polarized with a DC potential using the combination of three
nuclear techniques: RBS, ERD and nuclear reaction analysis. The results show that the alkali ions are the most mobile charge carriers followed by hydrogen species, most probably H+. The Cat' and Bat+ ions
are also mobile. These differences in mobility are responsible for the observed decrease in the current with time when AI-Cu bilayer electrodes are used . These electrodes block the indiffusion of hydrogen from the atmosphere during the treatment. Furthermore, through the observed surface modifications we were
able to explain the results obtained with the LIPP technique.
Acknowledgements This work was supported in part by Secretaria de Ciência e Tecnologia/PR, FAPERJ and CNPq .
References [l] D.E . Carlson, K.W . Hang and G.F. Stockdale, J. Amer. Ceram. Soc. 55 (1972) 337, and references therein .  U.K. Krieger and W.A. Lanford, J. Non-Cryst. Solids 102 (1988) 50.  C.A .O . Peixoto, G. Marrone, L. Pargamin and G. Carrara, IEEE Trans . on Power Delivery 3 (1988) 776.  A. Doi, Y. Meniou, T. lshikawa and Y. Abel, J. Appl . Phys. 67 (1990) 691, and references therein .  C.M . Lepienski, G.F . Leal Ferreira, J.A . Giacometti, R.M . Faria, C.A . Achele and F.L. Freire Jr., presented at 7th Int . Symp . on Etectrets, Berlin, September 1991 (unpublished) .  C.M. Lepienski, J.A. Giacometti and C.A . Achele, Solid State Commun ., in press.  W.K. Chu, J.W. Mayer and M.A. Nicolet, Rutherford Backscattering Spectrometry (Academic Press, New York, 1978).