Effects of excess oxygen on the 4.5–6.3 eV absorption spectra of oxygen-rich high purity silica

Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

Contents lists available at ScienceDirect

Nuclear Instruments and Methods in Physics Research B journal homepage: www.elsevier.com/locate/nimb

Effects of excess oxygen on the 4.5–6.3 eV absorption spectra of oxygen-rich high purity silica R.H. Magruder III ⇑, S.J. Robinson Belmont University, Department of Chemistry and Physics, 1900 Belmont Blvd., Nashville, TN 37212, United States

a r t i c l e

i n f o

Article history: Received 17 November 2015 Received in revised form 14 March 2016 Accepted 18 March 2016

Keywords: Ion implantation Point defects Silica Spectroscopy

a b s t r a c t Type III silica samples were implanted with O using a multi-energy process that produced a layer of constant concentration to within ±5% beginning
80 nm from the surface and extending to
640 nm below the surfaces of the samples. The concentrations of excess oxygen in the layer ranged from 0.035 to
2.1at. %. In these samples we show that E0 centers and NBOHCs, as well as the normal cadre of ODC (II) centers, were suppressed, and the optical absorption from 4.7 to 6.4 eV was primarily due to oxygen excess defects. Using Gaussian fitting techniques to examine the optical difference spectra, we have been able to identify four defect centers that are related to excess oxygen defect bands at 4.76 eV, 5.42 eV, 5.75 eV and 6.25 eV. Ó 2016 Published by Elsevier B.V.

1. Introduction Understanding the formation of oxygen related defect bands in a-SiO2 is of importance because of its use as a dielectric layer as well as in fiber optics [1–3]. Radiation induced displacements of oxygen ions can result in the diffusion of these ions, resulting in various defects that are detrimental to device as well as fiber optic performance [1–6]. At the same time, optical absorption (OA) in the range between 4 and 7.6 eV in high-purity silica is still not fully understood due to the overlap of broad bands in this spectral region [1–5,7,8]. Skuja et al. point out that the non-bridging oxygen hole centers (NBOHCs) have optical absorption (OA) bands ranging from
2 eV to 7.8 eV with large full widths at half maximum (FWHMs) [1–3,9–11]. Other established bands are the E0 center at 5.8 eV with a FWHM of 0.9 eV, the ODC(I) at 7.6 eV with a FWHM of
0.65 eV, the D center (which is generally associated with ion implantation and thought to be an ODC) at 7.15 eV with a FWHM of
0.8 eV and the ODC(II) bands (or B bands, B2a and B2b with FWHMs of 0.35 eV and 0.5 eV respectively) which have strong OA bands in the 5.0–5.2 eV region [9–11]. These bands complicate separating out the various components to the optical absorption from other sources [1–4,6–9]. Only by suppressing these primary bands, most of which are oxygen deficient bands, can oxygen excess bands be studied [2,5,6]. These ODC bands and their OA energies, photoluminescence (PL) and electron para-

⇑ Corresponding author. E-mail address: [email protected] (R.H. Magruder III). http://dx.doi.org/10.1016/j.nimb.2016.03.036 0168-583X/Ó 2016 Published by Elsevier B.V.

magnetic resonance (EPR) properties are reviewed in Ref. [3,5,9,10–12]. The formation of excess oxygen rich silica and measuring the amount of excess oxygen has also been problematic until more recently. Skuja et al. [13,14] have recently made a strong case for PL at 1232 nm for monitoring the interstitial O2 molecules and their role in defect formation as well as to confirm the amount and presence of interstitial O2 molecules. Interstitial oxygen molecules, O2, have been reported as one of largest forms of excess oxygen in silica and associated with the Schuman Runge bands for OA for energies >5 eV and PL at 1232 nm. However, these interstitial oxygen molecules have been shown to react in many cases with the silica matrix to form oxygen excess defects [2,5,15]. The O3 interstitial molecule has been suggested to exist in oxygen-rich silica with an absorption at
4.8 eV and a full width at half maximum (FWHM) of
0.9 eV [4]. This band overlaps the NBOHC band at 4.8 eV with a FWHM of
1.05 eV and could be easily obscured by the NBOHC presence. The O0, O2 and O3 molecules have also been studied to ascertain their role in the diffusion of oxygen in silica because of its importance in defects that affect the use of silica in electronic and fiber optic applications [1–3,5]. To form excess oxygen layers, we have used ion implantation to implant constant concentration layers of O by using multi-energy implants to create these excess oxygen concentrations [6–8,16]. While implanting excess oxygen does create excess oxygen stoichiometry in silica, it also is expected to help suppress or alter the concentrations of the oxygen deficient defect centers (ODCs) like the E0 center and the B2a and B2b bands [6–8,15,17]. A result of the bleaching of ODCs is the suppression of the optical

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

absorption bands associated with these defects, which provides an opportunity to study oxygen excess defects if they are formed. Szymanski et al. [18] has recently shown that the strain induced and remaining in the silica matrix depends on the formation process and the quenching rate. Galeener and others [19–21] have shown that the fictive temperature is important in determining the structure of the silica that results from different fictive temperatures [11,19–21] and the strain in the silica as well as the state of the disorder. These results suggest that the mode of formation is important for determining the types and numbers of defects formed. With the high fictive temperatures and fast cooling rates associated with the implantation process [22] it may be possible that some of these more subtle oxygen excess defects, if they exist, can be preserved in the bulk that may not be as readily formed by other methods [6,7,11,21,22]. The explicit purpose of implanting excess O is to suppress the formation of ODCs and increase the possibility of creating excess oxygen defects using the fast heating and quenching rate of the implantation process to form excess oxygen structures in regions of known excess oxygen concentration [6–8,10,11,15,22]. We recently reported the observation of a 5.32 eV band believed to be composed of two or more oxygen excess defects as well as a possible band at
5.5 to 5.8 eV in oxygen-rich silica but did not provide possible structures [6]. Here, we attempt to provide more details about these bands and the deconvolution of any overlapping bands due to oxygen excess defects that may be present. We have employed Gaussian fitting analysis of the optical difference spectra with increasing oxygen concentration in an attempt to separate the bands responsible for the absorption due to oxygen excess defects that are normally obscured in the optical absorption 4.0–6.4 eV region due to the predominance of ODCs and NBOHCs.

2. Methods The use of multi-energy implant techniques to produce a layer of constant concentration of an implanted ion has been previously described [6,7,16]. Samples of type III silica [Corning 7940] were implanted with multi-energy oxygen ions to form a layer in each sample
600 nm thick [7,8,16]. The computer programs Profile [23] and SRIM [24] were used to calculate the energies and doses necessary to form an approximate constant concentration layer. One face of a 2-cm diameter by 1-mm thick disc was implanted uniformly. The atomic percent of oxygen implanted in the layers was 0.035, 0.07, 0.35, 0.7 and 2.1 [7,8,16]. Full experimental details have been previously reported [6,7,16]. The ion energies varied from 35 to 250 keV, and the resulting depth profiles were constant to within ±5% beginning
80 nm from the surface and going to
640 nm. This provides an implanted layer that is approximately 7 times that of the ionization layer, minimizing the ionization effects in our analysis [6–8]. The optical absorption was measured at wavelength intervals of 1 nm from 1.8 eV to 6.5 eV using a dual-beam (Cary 5) spectrometer with an unimplanted sample in the reference beam and a beam spot size of 0.15 cm2. Hence, all absorption measurements are the difference between the absorption of implanted samples and that of an unimplanted sample and are reported in units of optical density. The absorption was measured at five different positions on the samples and the scatter in the absorption due to position was found to be less than ±5%. The noise in the data was less than 7% for the three highest concentrations and <20% for the lowest two concentrations. To test stability of the samples with time, they were measured a number of times after being made. They were first measured

41

within 48 h of being finished and were measured subsequently a number of times up to 24 months after being finished. No changes in absorption were observed outside of error for the 24-month period for any of the samples. Infrared reflectance measurements were made from 4000 cm1 to 400 cm1 using a Thermo Scientific Nicolet 6700 FTIR Spectrometer on all implanted samples and an unimplanted blank sample. Each spectrum was measured by averaging 100 scans using a 2 cm1 resolution. Five measurements were made on each sample at various positions. The scatter in the measurements due to position was less than ±5%. The reflectance spectra were measured at an angle of incidence of 10°. All measurements were referenced to a silver mirror to provide consistent normalization. The error in the peak position is ±2 cm1. 3. Results The Savitzky-Golay smoothing technique [25] in the commercial program IGOR by Wavemetrics [26] was used to reduce the noise in the optical spectra as the absorption is low in these samples, in particular for the two lowest oxygen concentrations. This technique was applied to data in the construction of Figs. 1 and 2 to better display possible peak positions and trends in the spectra. The smoothing technique was not used in the Gaussian fitting of data described in the following. Fig. 1 shows the smoothed optical spectra for the O implanted samples for all oxygen concentrations [6]. As the oxygen concentration increases, there are noticeable changes in the shape and magnitude of the spectra. The two lowest concentrations of 0.035 and 0.07 at.% show almost identical behavior as a function of oxygen concentration in both shape and magnitude with a peak at
5.06 ± 0.05 eV and increasing absorption for higher energies. Though much lower in magnitude, these optical spectra are very similar in shape to that of the Si and Ar implanted silica with similar energies and concentrations [6–8]. As the oxygen concentration increases to 0.35 at.%, a prominent shift occurs in the 5.06 eV peak from 5.06 eV to 5.21 eV and then to
5.32 eV in the higher oxygen concentrations. This peak shift was not observed in Si, N, B and Ar implanted samples [6–8] and suggests a fundamental change in the defects responsible for the optical absorption in the oxygen implanted samples as the oxygen concentration increases. Concurrent with this shift in peak position are changes

Fig. 1. Optical absorption as a function of photon energy for the different atomic percent of oxygen concentrations for multi-energy implanted samples.

42

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

Fig. 2. Difference in optical absorption for the three highest oxygen concentration implanted samples and the 0.07at.% oxygen implanted sample as a function of photon energy.

in the amplitude of the inflection in the 5.4 to 5.7 eV region over that observed in the lower oxygen concentration samples. The peak at
5.32 eV becomes the dominant feature in the spectra—also unlike that observed in Si, N, B and Ar implanted silica [6–8,27,28], indicating a different origin for the absorption. As we are interested in oxygen related bands and how increasing oxygen affects the spectra, Fig. 2 shows the difference spectra of the data obtained by subtracting 0.07at.% oxygen absorption from the higher oxygen concentration spectra. The 0.07at.% sample was chosen to subtract from the higher concentrations because the lower concentration sample showed similar behavior and the concentrations greater than 0.07at.% are where the spectra showed changes in peak position and shape. The difference spectra show the effects of increasing oxygen concentration. The growth of the peak at
5.32 eV is clearly seen as well as the shoulders at
4.7,
5.5 and
6.2 eV, particularly in the two highest oxygen concentration samples. In the difference spectrum of the 0.35–0.07 at.% sample as shown in Fig. 2, an increase in absorption is observed from 4.0 to 6.5 eV. Several maxima with a separation of 0.6 eV at energies <4 eV are observed, while peaks at energies >4 eV, which are separated by different energies are also observed. Each sample’s spectrum reveals a low-amplitude (i.e., 5–10% of the largest peaks) oscillation that permeates the data (especially noticeable in the 0.35–0.07 at.% sample in Fig. 2). With increasing O implantation, these effects are reduced in relative size due to an overall increase in absorption, giving rise to a better signal-to-noise ratio. Its ubiquity at all wavelengths and the fact that our spectrometer was being used at its lowest limits of measurement lead us to believe that it is an instrument source or detector artifact. It was taken into account as seen in Fig. 3 with the addition of regularly-spaced, small-amplitude peaks. All other absorption peaks are independent of and above this background noise, so we believe they accurately represent true sample peaks. By fitting the difference spectra data, we have generally minimized contributions from ionization and noise sources [7,8,16]. Part of these oscillations may also be attributed to the low signal-to-noise ratio in the lower concentration implanted samples due to the small absorption; this was discussed in previous work [6,7,16] and will not be discussed further. As the primary goal is to better understand the origin of the bands responsible for this new broad peak at 5.32 eV as well as

the shoulders on this peak between 4.6 and 6.5 eV (and assuming the bands are Gaussian in shape), we fit the difference data in Fig. 2 to a series of Gaussian bands. For fitting purposes the raw difference data (i.e. unsmoothed) was used. We also used the difference data to minimize any effects from the interference in the absorption due to the implanted region as well as any effects due to ionization. Based on prior research most of the ionization effects are saturated in the lowest implanted concentration samples and with additional higher oxygen concentrations ionization effects are expected to be similar in magnitude [6–8]. By using the difference data these effects are expected to be minimized. We used the commercial Gaussian fitting program PeakFit [29] by SeaSolve. So as to avoid any preconceived location of bands, we assumed no number of peaks nor their positions nor full widths at half maximum (FWHM). We chose to fit the raw difference data for the minimum number of peaks that would fit the data, assuming only that the peaks are Gaussian in shape. Four bands at 4.76, 5.42, 5.75 and 6.25 eV with FWHMs of 0.98 eV, 0.78 eV, 0.33 eV and 0.44 eV respectively were identified as needed to fit the difference spectra data as shown in Fig. 3 for the 2.1–0.07 at.% concentration sample. The optical density 90% prediction interval is
±7  104 (3% of the dominant peak); this prediction interval changed negligibly when the peak positions were adjusted within ±0.05 eV. The error in the FWHM is considered to be less than 0.1 eV. As these bands change with oxygen concentration and—as discussed below—ODCs and NBOHCs have been suppressed, we assume that these identified peaks are related to oxygen excess defects. Table 1 gives the magnitude for the four bands found in the difference spectra required to fit the data in terms of optical density as a function of increasing oxygen concentration. We have previously reported the EPR data on these samples and the error analysis [7,8] but include the results here in Table 2 for the convenience of the reader to compare EPR data with the optical difference data in Table 1. From EPR measurements in Table 2, the E0 center concentration drops by a factor of 60 [7,8]. From Table 2 we also note the decrease in NBOHCs concentrations from very small concentrations in the two lowest excess oxygen concentrations to below levels that are detectable with our spectrometers in the largest three oxygen concentration samples. These results are indicative of the suppression of NBOHCs by increasing oxygen concentration. Peroxy radical (POR) concentrations change less than 15% for the two lowest excess oxygen concentrations. The POR concentration increases by a factor of
2 for the next highest concentration and then remains constant within error for the three highest oxygen concentration samples. We infer from these results that the POR concentration saturates for concentrations greater than 0.07at.% excess oxygen in these samples. We have previously shown that, as identified by EPR measurements, a new oxygen excess center (the OS center) is formed in the sample with 0.35at.% excess oxygen implanted silica, increases by a factor of
3 then decreases by a factor of
2 for the highest excess oxygen concentration. As the OS center was not seen in the Si, B, N and Ar multi-implanted silica samples [27,28] and along with the g-values obtained for it, we previously suggested it is related to the excess implanted oxygen defect center. Fig. 4 shows the infrared reflectance (IRR) spectra for the implanted samples and an unimplanted sample. The IRR spectra has been shown to be sensitive to the intermediate range order (IRO) of the silica with implantation [11,19–21,30]. We define the intermediate range order (IRO) as the range from 2–5 nm in accordance with that given by Galeener [19]. This intermediate range order can affect the average ring size of the silica network, providing a significant change in the silica structure while maintaining the chemical order of the silica host [31]. The main TO

43

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

Fig. 3. Sum of contributing optical density peaks to the 2.1–0.07at.% oxygen implanted sample as a function of photon energy.

Table 1 Optical density peak amplitudes (10–3) and FWHMs (in eV) with varying oxygen concentrations. The 0.07at.% dataset has been subtracted from each dataset. Atomic % O

4.76 eV band

0.35 0.7 2.1

5.42 eV band

5.75 eV band

6.25 eV band

Amp.

FWHM

Amp.

FWHM

Amp.

FWHM

Amp.

FWHM

3.93 7.44 8.99

0.848 0.946 0.976

4.51 11.9 18.1

0.592 0.786 0.781

2.95 4.23 3.60

0.277 0.447 0.325

6.36 4.85 7.07

0.680 0.354 0.437

Table 2 Overview of the densities of the observed EPR-active defects as a function of the O implantation dose. Quoted values of O implantation and defect density are in units of atomic percent and 1018 cm3, respectively. All data is taken from Ref. [7].

1

O (at.%)

[E0 ]

[NBOHC]

[POR]

[OS]

0.035 0.07 0.35 0.7 2.1

1.7 ± 0.5 0.7 ± 0.2 0.5 ± 0.1 0.08 ± 0.02 0.028 ± 0.008

6±2 4±1 N.D.1 N.D.1 N.D.1

18 ± 5 17 ± 5 32 ± 9 36 ± 10 27 ± 8

N.D.1 N.D.1 13 ± 4 42 ± 13 18 ± 5

Not detected; estimated below <4  1018 cm3.

mode at 1122 cm1 shifts from 1122 cm1 in the unimplanted sample to
1108 cm1 with implantation of oxygen. This shift is indicative of the compaction of the network [11,15,19,21,22] as well as a change in the fictive temperature of the implanted layer compared to the unimplanted silica. However, there are no changes outside error with increasing oxygen concentration beyond the first implanted concentration, suggesting a saturation of compaction effects and change in the fictive temperature. Effects due to radiation damage have been shown to be maximal after relatively low doses on the order of 1014 ions/cm2 [9,11,19–21,30]. For all of our samples, the highest energy dose will give the maximum radiation damage throughout the implanted layer. This will result in a small compaction effect and, hence, small changes in the density of the implanted regions. However, we expect any changes will be constant for all samples as the highest dose in each sample is greater than 1014 ions/cm2. By using the difference spectra, we anticipate elimination of ionization

Fig. 4. Percent reflectance for the different atomic percent of oxygen concentrations for multi-energy implanted samples as a function of wavenumber.

effects as well as the minimization of any compaction or radiation damage effects that would affect the optical absorption. Sputter effects are expected to occur. Two things were done to minimize their effect on the optical spectra. First, we used sputtering coefficients in the calculations for dose and energy needed to build our constant profiles in SRIM and Profile calculations. Second,

44

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

this is also why we used the difference spectra. While the effects of sputtering would be slightly different for each dose, the use of difference spectra to study the effects of excess oxygen will account for some of the effect. One of the reasons for building multi energy implanted samples was to construct an implanted layer where the size of the layer would exceed the ionization layer by a significant factor. So this layer, while present and different in size from one sample to the next, will only contribute a small amount to the total absorption. Coupled with the difference spectra, its effect would be minimized—though still present in a small amount estimated to be less than 7%.

4. Discussion The multi-energy process we have used provides the capability of creating non-equilibrium concentrations of excess oxygen in regions of known and constant oxygen excess in a rectangular shaped distribution instead of the typical Gaussian or log normal shaped distribution generally associated with ion implantation [6–8,10,11,21,22]. A minor effect due to a defect distribution is possible and the distribution may have a small difference in typology in depth [6–8,10,11]. Though ionization effects can occur through the entire implanted region, there are two regions in the samples in which all of the initial Si and O ions may not be displaced. One of these is at the implanted surface where the implanted ions are scattered inelastically. While there may also be some elastic scattering causing displacements, this region will be small compared to the implanted region and is expected to have a minimal effect on changes in the total optical absorption. We assume that the radiation damage in this region is primarily ionization events [9–11]. The other region in which there may be primarily ionization events is the end of the range of the highest energy implanted ion. When the energy of the implanted ion becomes less than the displacement energy of the ions with which it collides, the remainder of its energy is expended in ionization events. We estimate based on Profile calculations [23] and SRIM [24] calculations that the thickness of these layers is less than 60 nm total with the size of the constant concentration region at
560 nm [7,8,10,12], resulting in the optical absorption of the implanted layer being much larger than that of the ionization layers; therefore these layers are expected to contribute no more than
10% to the absorption. Moreover, by using the difference spectra, these effects will be minimized by the subtraction process and are estimated to account for less than 5% of the optical absorption in the difference spectra. For the ion implantation process, several different approaches have shown that the fictive temperature of the ion implanted layer is between 2500 and 3500 K [10,11,21,22,32]. The fictive temperature can be quite high and achieved very rapidly compared to silica heated by conventional means [10,11,15,21,22,32]. The importance of the fictive temperature is that changes in it can result in a variety of different formations of O excess related defects not readily observed in conventional type III silica as well as the compaction of the silica matrix in the implanted region [10,11,19,21,22,30]. With fast cooling rates associated with ion implantation, nonequilibrium states of silica can be frozen in the implanted layer of the silica substrate. The chemical interaction (or lack thereof) of the implanted ion with the host silica can modify the simple compaction process and bring its own set of changes to defect concentrations [9,11,19,21,31,33]. Infrared reflectance (IRR) spectroscopy has been shown to be sensitive to changes in the IRO which are affected by the fictive temperature of the silica as seen in the IRR spectra with the position and shift of the TO mode at 1122 cm1 [11,19–22]. Other IRR modes have been used to characterize the effects of the fictive temperature and the disorder in the

silica matrix [10,11,15,19] Kajihara et al. [1] have shown that the efficiency of the Frenkel process due to ionization is influenced by the degree of disorder of the a-SiO2; hence the fictive temperature may be important in determining the defects produced and retained [3,5]. Fig. 4 shows the IRR spectra for the samples compared with an unimplanted sample. After the initial change with the lowest implanted oxygen concentration, there are no additional changes in the shifted peak position of the TO peak from 1122 to 1108 cm1 with increasing oxygen concentration. Coupled with a lack of change in the LO mode at 1250 cm1 and the mode at
800 cm1, we conclude that the fictive temperature and IRO are essentially the same for all implanted samples [10,11,15,19,21,33]. In addition, the lack of change in the 1105 cm1 mode, associated with non-bridging oxygens [11,19,21,22], indicates that there is no change in their concentration after the first implantation. These results suggest that the disorder is, within error, the same for all samples and the maximum compaction effect of the silica matrix is achieved with the lowest oxygen concentration implanted. Therefore because of the lack of change in the IRR spectra after the lowest concentration implanted sample then by using the difference spectra we expect to minimize any effects on the difference optical spectra for our fitting procedures due to disorder and or compaction which is a well know effect in ion implanted silica [9–11,19–22]. We conclude that all the optical changes observed must then be related to point defects as a function of the implanted oxygen concentration and their interaction with the SiO2 matrix [21,30]. Oxygen implantation can give rise to Schottky or Frenkel defects depending on the reaction of the ions [1,11,12]. In the case of O implantation, a somewhat more complex series of interactions can occur as the implanted ions and the ions most likely to be displaced by the implantation process are O ions [9–11,15,17]. The separation of any optical absorption bands due to excess oxygen defects is complicated by the fact that these bands can overlap in energy and their oscillator strengths are usually small (less than 103 [9,10,17]). This makes them difficult to detect, as the oscillator strengths for ODC sites are reported to be, in general, one to two orders of magnitude higher [4,7,9,10]. Therefore, the suppression of ODCs like the E0 center and the B2 bands as well as the NBOHCs (which have multiple absorption bands [1–5]) are essential to ascertain the effects of excess oxygen on other bands in the optical absorption. Skuja et al. [3] point out that the almost universal presence of the E0 center has limited the ability to study defects in the 5.5 to 6.5 eV region of the optical absorption spectra. The NBOHC also presents somewhat the same problem because of the large number of optical absorption bands associated with its presence from 4 to 7 eV [3–5]. The E0 center and NBOHC almost always occur together, as they are one of the main products that occur from disruption of the omnipresent strained bond precursor [3]. The ability to separate the contribution from various bands to the optical absorption in the 5.2 to 6.0 eV region is therefore always difficult unless the E0 and NBOHC concentrations can be suppressed. The presence of NBOHCs is usually established by EPR measurements, photoluminescence (PL) measurements at
1.9 eV and a strong optical absorption band at 4.8 eV as well as other NBOHC optical absorption bands that contribute to the optical absorption from 4.8 eV to
7.5 eV [3,5]. We infer, based on the lack of PL and EPR signals as well as optical absorption spectra associated with the NBOHCs, that we have effectively suppressed the presence of the NBOHCs for oxygen concentrations greater than 0.07at.% [6–8]. The suppression of the NBOHCs is also supported by the fact that the absorption spectra we observe in our oxygen implanted samples and the spectra suggested by Skuja et al. [3] to be almost pure NBOHCs are very dissimilar (Figs. 1 and 4 in Ref. [3]). Based on our results, NBOHCs, including even the EPR silent NBOHCs as defined by Skuja et al. [3–5], are either not

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

present or in very small concentrations in the three highest concentration oxygen excess samples. If any do remain we can expect their contribution in the optical difference spectra to be minimal and even smaller in the difference spectra. Coupled with the drop in E0 center concentration by a factor of 60 and the loss of optical absorption for the 5.0–5.2 eV region for oxygen concentrations greater than 0.07at.%, we conclude that the normally dominant ODC bands in the 4.7 to 6.0 eV region have been significantly suppressed in our suite of samples. We infer from these results that the origin of the four bands we see in the difference-fitted spectra are due to other sources than the NBOHC, E0 , B2 bands and other ODC bands. While the E band at 7.6 eV [9] is observed in oxygen implanted silica [15,34], based on its high energy and relatively small FWHM, we expect it to have a minimal impact in the optical absorption difference spectra from 4.0 to 6.4 eV. However, its presence and the effect on its concentration by increasing oxygen concentration may provide indications of possible mechanisms of formation for various defects as discussed below. Hosono and Matsunami [15] have suggested that with oxygen implantation of a single energy similar to those used in our samples, PORs and O2 molecules are the major defects. The concentration of the POR center is inferred from the attribution of part of the optical absorption at 7.5–7.6 eV to the presence of POR centers. The attribution of some of the absorption at
7.6 eV to the presence of large numbers of PORs resulted in their conclusion that the concentration of Si–Si or E band defects in oxygen implanted silica was approximately an order of magnitude lower in concentration than in other types of ion-implanted silica. If the POR does not have an absorption at 7.6 eV, this would suggest that the concentration of E band defects will be significantly greater. Recent calculations and other results have called into question whether the POR center has an absorption at 7.5–7.6 eV, indicating that there are more E band defects present than previously thought to exist before in oxygen implanted silica [1–4,9,15,35]. We attribute the bands fitted to the difference spectra at 4.76 eV, 5.42 eV, 5.75 eV and 6.25 eV to be associated with O excess defect structures. Using quantum mechanical calculations as guides and our results compared with previous EPR results on this suite of samples, we provide the following assignments of these four bands. The band at 4.8 eV would generally be attributed to POR centers [4,9,10] or NBOHCs. As discussed above, we expect only a minimal—if any—contribution from the NBOHCs. For these oxygen implanted samples, while the POR is detected in all samples, the EPR concentration of the PORs saturates within error for the three samples with the highest oxygen concentrations. Based on these results, we expect at most only a minimal change in the OD difference spectra due to the POR centers if their absorption is at
4.8 eV as has been suggested [4,9,10]. These can explain changes as large as those seen in Table 1 for the 4.76 eV band and the 5.42 eV band. The 4.76 eV band shows a continuous increase in the difference spectra magnitude by a factor of 2 from the 0.035–0.07 at.% difference to the 2.1–0.07 at.% difference samples (Table 1). We have previously suggested that an additional diamagnetic center exists at
4.6–4.8 eV, but here the evidence for the center is much stronger, as neither the POR nor the NBOHC can account for the optical difference change [6–8]. As the POR concentration saturates and the NBOHC is not present, the difference spectra presented here provides more definitive evidence that an additional oxygen excess defect must be present in these samples. Recent ab initio quantum mechanical calculations by Sousa, et al. [35], have shown that the lowest energy of absorption of the POR should be
5.5 eV, which agrees with the assignment of surface PORs at 5.4 eV [3,5]. Their calculations indicate a strong absorption for POR centers at
5.5 eV and a weaker absorption at

45


6.5 eV. They rule out optical absorptions for the POR at 2, 4.8 and 7.6 eV that have been previously suggested [4,9,10]. So while we cannot definitively rule out the attribution of absorption at
4.8 eV to PORs, we can say that the existence of an additional structure of at least one other center absorbing at this energy is necessary. While interstitial O and O2 molecules could be possible candidates for the absorption at 4.76 eV, recent work by Skuja et al. [3–5] and Kajihara [1,2] has shown these interstitials interact, forming peroxy linkage (POL) and POR type structures. Another possible candidate for absorption at 4.76 eV is the O3 interstitial molecule, which is recognized to have an absorption band at
4.8 eV [4,38] and would be expected to grow with increasing oxygen concentration as seen in Table 1 for the OD difference for the 4.76 eV band. In the following we suggest several mechanisms. The mechanisms proposed are not the only ones that are possible; however the mechanisms operative in these samples must meet certain criteria based on the results and calculations discussed above. We expect that multiple processes are occurring simultaneously. First, any mechanism must be compatible with the formation of the POL and take up a large number of interstitial oxygens as indicated in recent work [1–3,5]. Secondly, the mechanism(s) must suppress the formation and bleach the E0 center and the NBOHC as our results are indicative of this. Third, the mechanism(s) must not result just in the formation of an increasing number of PORs as, while they are observed, the POR cannot alone account for the total number of defects and the different peak energies observed. Fourth, one of the mechanisms must result in the formation of the E center as it is observed in oxygen implanted silica [10,15]. Fifth, the calculations on the energy of formation must be taken into account to suggest the mechanisms [37]. These mechanisms meet the criteria of the list discussed above and are perhaps the simplest. There are always more complex mechanisms, but the end results need to be similar to these suggested. While we list several possible mechanisms in the following, it should be noted that, for the level of excess oxygen used here, we anticipate that several mechanisms will be operative. All would be driven by the spike in local temperature on the intermediate range order and short range order [19] in the silica due to the ion implantation process. The fictive temperature with the implantation process itself is estimated to be quite high (2500–3000 K), certainly enough to allow for local rearrangement of the local ions and form the complexes suggested. The difference spectra allows the sputtering, radiation damage effects, and other effects to be minimized and allows focus on the effects on the optical spectra of excess oxygen in the samples. A simple mechanism would be implanted oxygens and/or displaced O ions forming the ozone interstitials, such as

2Oþ þ 2e þ BSiAOASiB ! BSiASiB þ 2O0 þ O0 ! BSiASiB þ O3 where Si represents a Si bonded to three oxygens. In this mechanism, two of the oxygens are coming from implanted ions and one from a displaced O ion to form the O3 interstitial. This mechanism was chosen as possible as an E band defect (ODC I) or a B2a band would also be formed with the implantation process [15,17,34], assuming one of the oxygens comes from the matrix. Based on the bleaching of optical absorption in the 5.0 eV region, we discount the formation of the B2a centers. However, this mechanism could account for the relatively high content of E band defects in O implanted silica that have previously been attributed to the formation of PORs [15] and that we have observed in our preliminary VUV measurements [34]. In theory, one might expect the excess oxygen to bleach the E center, but differing rates

46

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

of formation and bleaching could account for increasing E center concentrations. The plausibility of this mechanism is evidenced by the presence of a strong absorption in the 7.6 eV region with ion implantation of various ions including O [10,15,17,34]. Electrons to neutralize the implanted O+ in this formation mechanism would be furnished by the metal backing of the substrate and metal prongs attaching the substrate to the metal block holding the substrate during implantation. This neutralization process is expected to occur rapidly [10,11]. The FWHM for the O3 is expected to be 0.9 eV, which is close to the value of 0.98 eV we find for the 4.76 eV peak in fitting the difference spectra. Based on these and prior results, we suggest that the band at 4.76 eV is diamagnetic and is due to the O3 structure. A simple mechanism that fits our criteria outlined above is þ



O þ O2 þ e ! O3 where the O2 and O3 are interstitial and O+ comes from implantation. For the 5.42 eV band, calculations of optical absorption and the observation of surface defect structures for the optical absorption of the POR for a band at 5.42 eV, we suggest that there are two defect states and that the POR and POL oxygen excess structures are plausible candidates. From EPR measurements, the POR concentration is saturated in these samples and the optical difference spectra increases by a factor of 6; thus, we can say it does not contribute significantly to the optical difference measurements. These results indicate that even if calculations are correct and the POR contributes at
5.42 eV, there also has to be a second center close in energy to the POR to account for the continuing increase in optical absorption with oxygen concentration. The second candidate for optical absorption at 5.42 eV is the POL. Based on calculations, the POR and POL are expected to have absorption very close in value—to the point that their absorption may be difficult to separate [35]. Calculations by Stefanov and Raghavachan indicate the POL would have an optical absorption at
5.5 eV [36], which would mean the POR and POL bands would overlap; however the POR is expected to be EPR active while the POL will be diamagnetic. Considering the energy of formation of defects, Richard et al. [37] have shown that the dominant defects are oxygen interstitials which, based on their calculations, form charged and neutral peroxide bridges and edge-sharing tetrahedra (Table 2 in Ref. [34]). These calculations indicate that the probability of forming an interstitial O–2 similar to the alkali halide case is much smaller than the probability of forming the peroxy structures. These interstitial oxygen defects are also shown to be more energetically favored over other types of Si and O related defects. These calculations are consistent with earlier work by Skuja et al. [38] as well as calculations by Hamann [39]. Based on these results and calculations, we infer that the neutral and charged POL are plausible large sources of oxygen excess defects in oxygen excess implanted silica. Weeks et al. have previously reported the growth of a 5.4 eV band which was diamagnetic in type III silica with 5 eV radiation from a KrF laser [40]. The 5.4 eV band was suggested to be an oxygen related structure though no specific model was proposed. Based on the present work, this band could be explained if it is the POL band. The formation of POL structures would result from ionization processes forming atomic oxygen and then reacting as suggested by Kajihara et al. [1,2]. The interstitial O0 reacts in various ways to form the POL structure and could account for the 5.4 eV absorption observed in KrF irradiated samples of Type III silica [40]. This is consistent with earlier work [38] and more recent work [1–3,5] suggesting this POL center as a possible structure for large concentrations of oxygen interstitials. Calculations by Hamann [39] and Richard et al. [37] on energies of formation as well as calculations of optical absorption energies of Stefanov

and Raghavachan [36] are consistent with this assignment. Stefanov and Raghavachan’s calculations indicate that this would be a weak band due to its low oscillator strength, which is consistent with our observations of its magnitude in this suite of samples. This would in part explain its apparent lack of presence in samples where the oxygen excess is present but where the ODCs and NBOHCs have not been sufficiently suppressed and would dominate the spectra. We suggest that the implanted O results in the formation of interstitial O2, forming POL structures [2,4,9] as the oxygen reacts with the silica matrix. A possible mechanism for the POL formation by ion implantation that would result in relatively large numbers of POL centers and the rapid growth of the 5.42 eV band is

Oþ þ e þ BSiAOASiB ! BSiAOAOASiB where the excess O comes from the implantation process. The bleaching effect of the B2a (E00 ) could also form POL centers through the incorporation of either implanted O ions or O2 molecules as suggested in prior work [2].

2Oþ þ 2e þ BSiASiB ! BSiAOAOASiB or

O2 þ BSiASiB ! BSiAOAOASiB The bleaching of the E0 center along with the interaction of the NBOHC could be accomplished with implanted O, with the driving force for this reaction being furnished by the heat of the implantation process

Oþ þ e ! O0 

O0 þ BSi þ e þ  OASiB ! BSiAOAOASiB where BSi⁄ is the E0 center and OASiB is the NBOHC center. These mechanisms could account for the large increase in the 5.42 eV band, the complex behavior of the defects in the VUV measurements of oxygen implanted silica [15,17,34] and the bleaching of the E0 centers and the NBOHCs. Kajihara et al. have shown [2] that atomic oxygen would interact with the network to form POLs and that O2 exists mostly as the POL. Hamann [39] has suggested that O2 diffuses by forming POLs. The center would be diamagnetic and, as expected, shows no correlation with any of the observed EPR data in Table 2. Based on the large increase in the 5.42 eV OA band and the calculations suggesting the POL band would have optical absorption at
5.5 eV, we conclude that a significant fraction of displaced and/or implanted O in these materials is in the form of molecular oxygen bridges or POL structures and is the source of the 5.42 eV optical absorption. Optical absorption at
5.8 eV is normally associated with the E0 center band as the dominant absorption source. Here, we have decreased the concentration of the E0 centers as measured by EPR by a factor of
60 which, coupled with the increase observed in this region in the difference spectra, indicates that there is another band present at
5.75 eV [4,7,8,17,34,41]. A possible candidate center for the absorption at 5.75 eV, based on calculations and observation from surface centers, would be the oxygen double bond @Si@O, the silanone group (SG) center [3]. It is expected to have an optical absorption at 5.65 eV for surface centers and typical bulk values are expected to be within
0.3 eV [3,5] which would include the 5.75 eV region. This defect center could arise from implantation events driving the formation and the rapid cooling freezing them in. We found that the FWHM of the 5.75 band is 0.325 eV, which is not consistent with the 0.9 eV suggested value for FWHM for the SG center. Moreover, the SG center is expected to have a PL band at 2.27 eV with excitation at

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

5.5 eV. We specifically looked for the PL at 2.27 eV but were not able to observe any with our spectrometers. The lack of evidence for a PL band at 2.27 eV in our samples and the much smaller FWHM argues against this being the structure responsible for absorption at the 5.75 eV. While we have suggested the center responsible to be diamagnetic in previous work [7,8,12] by fitting to the optical difference spectra and by minimizing contributions from other defects, we have a better understanding of the characteristics of this band. From Tables I and II, there is a qualitative correlation of the 5.75 eV absorption band with the OS center, a paramagnetic center. This qualitative correlation is based on the increase, then decrease, of the optical absorption with increasing oxygen concentration as observed in EPR measurements of the OS centers. While this correlation may be fortuitous, we can say with certainty that this band is necessary to provide a good fit to the difference data, and its presence is observed only because we were able to suppress the NBOHC and the E0 centers. It is also consistent with previous work suggesting a second defect at
5.7 to 5.8 eV [4,40,41]. The positively charged peroxy bridge is paramagnetic and has been suggested as a candidate responsible for the OS signal [7,8]. Its energy of formation is larger than that of the neutral peroxy bridge [37]; however, the energy of formation could easily be furnished by the implantation process. Once formed by implantation, the fast cooling rates associated with the implantation process could result in the freezing in of this structure. Based on the similarity to the neutral POL and POR, it is not unreasonable to expect that its absorption would be close to the POL and POR which have been calculated to be at
5.5 eV. The OS center previously identified in these samples has g-values similar to charged oxygen complexes [7,8]. Both the EPR concentration and the optical absorption have a maximum in the difference spectra for the 0.7–0.07 at.% difference and decrease at the maximum oxygen concentration implanted. Its presence would be masked in most samples due to the E0 center presence; hence it is difficult to ascertain its relationship with EPR and optical absorption if many E0 centers exist. Based on calculations and our results, we tentatively attribute the 5.75 eV band to a charged variation of the POR as a positively charged peroxy bridge and believe that it is related to the OS center. The 6.25 eV band is more difficult to analyze as our data goes only to 6.5 eV in these measurements, and bands beyond this value could easily influence this region of the spectra. Guzzi et al. [41] have previously reported a 6.2 eV band in fused quartz which they attribute to a diamagnetic intrinsic center. Preliminary VUV data on these samples indicate a center at
6.2 to 6.5 eV [34]. The 6.25 eV band amplitude in the difference spectra decreases, then increases, exhibiting the opposite behavior to the EPR concentration of the OS centers and does not correlate with any of the defect states measured by EPR; this suggests it is probably diamagnetic as suggested by Guzzi et al. [41] for the band they reported at 6.2 eV. At this point we suggest that is related to an excess oxygen defect and diamagnetic. However, further experiments will need to be done as well as accounting for influences in this region from defects at higher energies than 6.4 eV before a more definitive model can be suggested [34]. 5. Conclusions We used Gaussian fitting techniques to examine the optical difference spectra of a suite of oxygen excess type III silica samples where the E0 centers and NBOHCs as well as the normal cadre of ODC centers are shown to be suppressed. By examining the difference spectra, we have been able to identify four defect centers that are related to excess oxygen defects at 4.76 eV, 5.42 eV, 5.75 eV and 6.25 eV. Using previously reported optical absorption energy

47

and energy of formation calculations, we conclude that the band at 4.76 eV is due to the ozone interstitial, the 5.42 eV band to POLs and the 5.75 eV band to the OS center, a charged POL structure. While we cannot specify the structure of the band at 6.25 eV, we suggest it is diamagnetic and related to an oxygen excess structure.

References [1] K. Kajihara, M. Hirano, L. Skuja, H. Hosono, Intrinsic defect formation in amorphous SiO2 by electronic excitation: bond dissociation versus Frenkel mechanism, Phys. Rev. B 78 (2008). 09421-1. [2] K. Kajihara, T. Miura, H. Kamioka, A. Aiba, M. Uramoo, Y. Morimoto, M. Hirano, L. Skuja, H. Hosono, Diffusion and reactions of interstitial oxygen species in amorphous SiO2: a review, J. Non-Cryst. Solids 354 (2008) 224. [3] L. Skuja, K. Kajihara, M. Hirano, H. Hosono, Visible to vacuum-UV range optical absorption of oxygen dangling bonds in amorphous SiO2, Phys. Rev. B 84 (2011). 205206-1. [4] L. Skuja, Optically active oxygen deficiency related centers in amorphous silica, J. Non-Cryst. Solids 239 (1998) 16. and references there in. [5] L. Skuja, K. Kajihara, M. Hirano, H. Hosono, Oxygen excess related point defects in glassy/amorphous SiO2 and related materials, Nucl. Instrum. Methods Phys. Res. Sect. B 286 (2012) 159. [6] R.H. Magruder III, R.A. Weeks, R.A. Weller, New intrinsic oxygen related defect bands in oxygen implanted silica, J. Non-Cryst. Solids 357 (2011) 1615. [7] R.H. Magruder III, A. Stesmans, K. Clémer, R.A. Weeks, R.A. Weller, Origins of optical absorption between 4.8 and 4.9 eV in silica implanted with Si and with O ions, J. Non-Cryst. Solids 352 (2006) 3027. [8] R.H. Magruder III, A. Stesmans, K. Clémer, R.A. Weeks, R.A. Weller, Sources of optical absorption between 5.7 and 5.9 eV in silica implanted with Si or O, J. Appl. Phys. 100 (2006) 033517. [9] E.J. Friebele, D. Griscom, Treatise on Materials Science and Technology, in: M. Tomozawa, R.H. Doremus (Eds.), vol. 17, Academic Press, New York, 1979, p. 257. [10] R.A. Weeks, Optical and magnetic properties of ion implanted glasses, in: J. Zarzychi (Ed.), Materials Science and Technology, vol. 9, VCH, Weinheim, 1991, p. 331. and references therein. [11] P.D. Townsend, P.J. Chandler, L. Zhang, Optical Effects of Ion Implantation, Cambridge University Press, Cambridge, 1994. and references therein. [12] R.H. Magruder III, R.A. Weeks, Y. Morimoto, Si related defects in the VUV in silicon multi-energy implanted type III silica, Nucl. Instrum. Methods Phys. Res. Sect. B 269 (2011) 2532. [13] L. Skuja, B. Guttler, D. Shiel, A.R. Silin, Infrared photoluminescence of preexisting or irradiation-induced interstitial oxygen in glassy SiO2 and a quartz, Phys. Rev B 21 (1998) 14296. [14] L. Skuja, B. Guttler, D. Shiel, A.R. Silin, Quantitative analysis of the concentration of interstitial molecules of O2 in SiO2 glass using luminescence and Raman spectrometry, J. App. Phys. 83 (1998) 6106. [15] H. Hosono, N. Matsunami, Structural defects and chemical interactions of implanted ions with substrate structure in amorphous SiO2, Phys. Rev B 48 (1993) 13469. [16] R.H. Magruder III, R.A. Weeks, R.A. Weller, Luminescence and absorption in type III silica implanted with multi-energy Si, O and Ar ions, J. Non-Cryst. Solids 322 (2003) 58. [17] Y. Morimoto, R.A. Weeks, R.H. Magruder III, R.A. Zuhr, Optical and electron paramagnetic resonance spectra of silica glass implanted with oxygen Ions, J. Non Cryst. Solids 203 (1996) 55. [18] M. Szymanski, A.L. Shluger, M. Stoneman, Role of disorder in incorporation energies of oxygen atoms in amorphous silica, Phys. Rev. B 63 (2001). 224207-1. [19] F. Galeener, in: R.A.B. Devine (Ed.), The Physics and Technology of Amorphous SiO2, Plenum Press, New York, 1988. page 1. [20] J. Simon, in: J.D. MacKenzie (Ed.), Modern Aspects of the Vitreous State, vol. I, Butterworth and Co., London, 1960. [21] G.W. Arnold, P.M. Mazzldi, in: P.M. Mazzoldi, G.W. Arnold (Eds.), Ion Beam Modification of Insulators, Elsevier, Amsterdam, 1987. [22] H. Hosono, Fourier transform infrared attenuated total reflection spectra of ion implanted silica glasses, J. Appl. Phys. 69 (1991) 8079. [23] Profile Code copyrighted by Implant Sciences Corp., 107 Audubon Road, Wakefield, MA. [24] J.F. Ziegler, M.D. Ziegler, J.P. Biersack, SRIM – The stopping and range of ions in matter, Nucl. Instrum. Methods Phys. Res. Sect. B 268 (2010) 1818. [25] A. Savitzky, M.J.E. Golay, Smoothing and differentiation of data by simplified least squares procedures, Anal. Chem. 36 (1964) 1627. [26] IGOR Pro 6.0, Wavemetrics, Inc., Lake Oswego, OR, 2007. [27] R.H. Magruder III, A. Stesmans, K. Clémer, R.A. Weeks, R.A. Weller, The effect of implanting nitrogen on the optical absorption and electron paramagnetic resonance spectra of silica, J. Non-Cryst. Solids 354 (2008) 3580. [28] R.H. Magruder III, A. Stesmans, R.A. Weeks, R.A. Weller, The effect of implanting boron on the optical absorption and electron paramagnetic resonance spectra of silica, J. Appl. Phys. 104 (2008) 054110–54111. [29] PeakFit 4.12 SeaSolve Software, Inc., Framington MA, 2003. [30] R.H. Magruder III, S.H. Morgan, R.A. Weeks, R. Zuhr, Effect of ion implantation range order: IR spectrum of silica, J. Non-Cryst. Solids 120 (1990) 241.

48

R.H. Magruder III, S.J. Robinson / Nuclear Instruments and Methods in Physics Research B 375 (2016) 40–48

[31] Q. Williams, R. Jeanloz, Spectroscopic evidence for pressure-induced coordination changes in silicate glasses and melts, Science 239 (1988) 902. [32] A. Halabica, J.C. Idrobo, S.T. Pantelides, R.F. Haglund, R.H. Magruder, S.J. Pennycook, Pulsed infrared laser annealing of gold nanoparticles embedded in a silica matrix, J. Appl. Phys. 103 (2008). 083545-1. [33] K. Hubner, L. Schumann, A. Lehmann, H.H. Vajen, G. Zuther, Detection of LO and TO phonons in amorphous SiO2-films by oblique-incidence of IR light, Physi. Status Solidi B-Basic Res. 104 (1981) K1. [34] R.H. Magruder III, Y. Morimoto, unpublished data. [35] C. Sousa, C. de Graaf, G. Pacchioni, Optical absorption of peroxy radicals in silica, multi-configurational perturbation theory calculations, J. Chem. Phys. 114 (2001) 6259. [36] B.B. Stefanov, K. Raghavachan, Photoabsorption of the peroxide linkage defect in silicate glasses, J. Chem. Phys. 111 (1999) 8039.

[37] N. Richard, L. Martin-Samos, G. Roma, Y. Limoge, J. Crocombette, J, Non-Cryst, Solids 351 (2005) 1825. [38] L. Skuja, M. Hirano, H. Hosono, Oxygen related intrinsic defects in glassy SiO2: interstitial ozone molecules”, Phys. Rev. Lett. 84 (2000) 302. [39] D.R. Hamann, Diffusion of atomic oxygen in SiO2, Phys. Rev. Lett. 81 (1998) 3447. [40] R.A. Weeks, R.H. Magruder III, P.W. Wang, Some effects of 5 eV photons on defects in SiO2, in: J. Non-Cryst. Solids 149 (1992) 122. [41] M. Guzzi, M. Martini, A. Paleari, F. Pio, C.B. Azzoni, Neutron irradiation effects in amorphous SiO2: optical absorption and electron paramagnetic resonance, J. Phys. Condens. Matter 5 (1993) 8105.