Manuscript Received

The Detection, Removal and Effect on Damage Thresholds
of Cerium Impurities on Fused Silica

1-17-89

R. C. Estler, N. S. Nogar, and R. A. Schmell
Chemical and Laser Sciences Division
Los Alamos National Laboratory
Los Alamos, New Mexico 87545

We describe the detection, by resonance ionization mass spectrometry (RIMS), of cerium impurities on fused silica optical substrates. The impurity is desorbed into the gas phase by an infrared laser (1.06 μm), and detected, as both free cerium atoms and cerium oxide, by RIMS. In addition, we describe an acid wash procedure that both improves the damage threshold, and reduces significantly the measured levels of near-surface cerium.

Key words: cerium; cleaning; damage; fused silica; impurities; lasers; mass spectrometry.

Introduction

Low-optical-absorption fused silica is a commonly used substrate material for ultraviolet optics. It has good optical properties, and is resistant to both chemical and physical abuse. The measured damage thresholds for this material are widely scattered, and usually lie significantly below the intrinsic damage threshold. This, and other evidence, implies that the dominant damage mechanism in fused silica is dielectric breakdown initiated by absorption at impurity sites [1-2]. Further, recent evidence [1] indicates that absorption in fused silica is not due to bulk impurities such as Al or OH. This suggests that damage may be initiated by the presence of inhomogeneously distributed absorbing impurities or inclusions. Small levels of chemical impurities including residual polishing compound and absorbing inclusions, or physical imperfections, such as grain boundaries or misoriented microcrystals, are among the candidates for initiation sites. These impurities can be detected with a high degree of sensitivity using resonance ionization mass spectrometry (RIMS) [3].

Fused silica is routinely polished with cerium oxide [4], a compound known to absorb at ultraviolet wavelengths. It has been speculated that reduced optical damage thresholds may result from cerium contamination left as polishing artifacts in surface dislocations. We have recently examined the effect of various surface treatment processes on both damage thresholds and the concentration of detectable cerium on the optic surface.

Experimental

Two-inch-diameter by 3/8-inch-thick Corning 7940 fused silica windows from a commercial vendor were prepared using standard optical fabrication techniques.

Cerium was detected as both free Ce atoms and CeO by resonant, or quasi-resonant ionization processes. The apparatus has been described previously [5]. Briefly, desorption was initiated by 10-nsec pulses from a Q-switched Nd-YAG laser (Quanta Ray DCR-2). Fundamental or frequency multiplied pulses from this laser were focused to a fluence of 5-20 J/cm2 at the surface of the optical substrate. At a variable time delay, typically 10-15 μsec, an excimer-pumped dye laser was triggered. Pulses from this laser (15 nsec, 1-3 mJ, 2-435 nm, Av=0.3 cm11) passed through the ionization region of the time-of-flight mass spectrometer parallel to, and 1 cm from, the substrate surface. Photoions were repelled down the flight tube at right angles to both incident lasers. lons were identified by both optical and mass spectra.

*Permanent address: Department of Chemistry, Fort Lewis College, Durango, Colorado 81301.

The mass spectrometer has a 0.4-m flight tube, and was maintained at a vacuum better than 107by a liquid-nitrogen trapped oil diffusion pump. lons were detected by a channel electron multiplier, and the signals processed by a series of amplifiers and either a transient recorder or a boxcar averager.

Surface roughness was measured using a heterodyne surface profilometer. The average surface roughness before cleaning was 3.7 Å RMS, with a peak-to-valley maximum of 21 Á. Photos of the surface were taken before and after cleaning treatments, and before and after laser irradiation, with a Nomarski microscope. The parts were checked before and after cleaning for cerium by RIMS, and also with the KrF optical damage test stand [6].

The cleaning procedure used consisted of an acid wash in 1:2 mixture (volume) of aqueous HNO, in H2SO at 90°C for 12 hours. The samples were then rinsed with distilled water and recleaned with acetone. The samples were then rechecked on the profilometer. The average surface roughness increased 1 Å rms.

Results and Discussion

Figure 1 shows a schematic energy level diagram for Ce and CeO in the energy (wavelength) region of interest, while figure 2 displays the experimental spectra in that region for masses 140 and 156. It should be noted that Ce has ≥12 states lying below 3500 cm, all of which may have substantial population following laser desorption [8]. In addition, the density of even levels near the one-photon resonance is 0.1/cm1. The net effect is the very congested, nondescript spectrum observed in figure 2a, exacerbated by power broadening at the intensities (10 MW/cm2) required for efficient ionization and detection. In addition, some of the signal detected at mass 140 may be due to simultaneous dissociation/ionization [8] of a CeO, precursor. Similarly, the CeO ground state is split into at least four levels [9,10] (probably 32, 33, 3, and ',), of generally unknown splitting, the chief exception being a well-defined 2060 cm1 splitting for 30,-1,. The resonance state is an ill-defined level of B symmetry. In all cases, the vibrational constant is ~600-800 cm3, and the rotational constant is 0.35 cm. The result is another very congested spectrum, as shown in figure 2b. Although the optical spectra cannot be used by itself to "fingerprint" the species, the combination of optical and mass spectra can be used to positively identify the desorbates, as shown in figure 3.

Figure1. Schematic energy level diagrams for Ce and CeO, showing potential resonant levels. The cross-hatched area indicates the ionization continuum.

Figure 2. Ionization spectra, obtained at intensity of 10 MW/cm2, for laser-desorbed neutrals from a ceria slurry: a) mass 156 (CeO*) and b) for mass 140 (Ce*).

Figure 3. The top trace shows the single-shot time-of-flight RIMS spectrum obtained from material laser desorbed from a fused silica optic, while the lower trace shows a calibration mass spectrum from a ceria slurry.

In figure 3a, obtained by laser desorption/RIMS from a prepared ceria slurry deposisted on glass, both mass 140 (Ce) and 156 (CEO) are easily identified in the time-of-flight spectrum. In figure 3b, obtained from the ceria-polished fused silica substrate, only CeO is observed.

For materials desorbed from the optical surface, ion signals were typically processed in a pulse-counting mode, sent to either a transient recorder, as in figure 3, or to a gated integrater set to accept signals from CeO, and displayed on an X-Y recorder. Single ion signals were easily observed above the a.c. noise level, as seen in figure 4. The relative concentration of cerium impurity was determined from the number of single ion spikes observed as the sample was rotated to sweep the laser across a predetermined arc of the optic.

Figure 4. Laser-desorbed CeO signal a) for untreated sample and b) for acid washed samples. Fluence levels were 1) 6 J/cm2, 2) 12 J/cm2, and 3) 19 J/cm2.

Figure 4a shows the cerium/cerium oxide signal obtained by sweeping the laser beam across≈180° (~0.5-in. radius) on an untreated SiO, optic, for several incident 1.06-μm laser intensities. Figure 4b shows a similar trace for an optic subjected to the acid wash procedure described previously. For the unbtreated optic, 11 CeO ions were detected (the third, fourth, and sixth peaks correspond to two ions), while for the acid-washed samples, no ions were detected at 6 J/cm2. At 12 J/cm2, two ions were detected in each case, so that the overall ratio of detected cerium is 12:1 for these two samples. These results are typical of those obtained for six pairs (treated and untreated) of samples, where the average ratio was 11±1.2:1 (O). It should also be noted that the number of CeO ions observed increased with the maximum fluence to which the sample was exposed. This result is consistent with models [11,12] of damage based on a distributed, or nondegenerate defect ensemble. Low fluence irradiation removes surface or nearsurface contaminants with minimal damage [3], while higher-fluence irradiation may remove more deeply embedded, or lower susceptibility, contaminants, with concommitant removal of surrounding coating material.

Based on previous results with this apparatus [5,13] and the magnitude of the observed ceria signals, the number of cerium oxide fragments removed from the sample lies in the range 107-10 for each detected CeO ion. This estimate assumes that the distribution is spatially isotropic and kinetically thermal (2000 K), and that the fraction of molecular quantum states addressed by

the laser of 103(partition function). It should be noted that this does not place an upper limit on the total amount of material removed in the desorption process. Cerium clusters, oxides, and other impurities may also be desorbed.

The results of damage threshold measurements are displayed in figure 5 for both untreated and acid washed samples. The difference in measured damage threshold indicated in this figure is representative of the difference in four pairs of samples: 7.57 ±1.77 J/cm2 (,) for untreated samples, and 9.72 ± 1.09 J/cm2 for the acid-washed samples.

1

Figure 5. Threshold damage measurements for unwashed (upper) and acid washed (lower) fused silica samples. Solid circles indicate macroscopic damage, while the open circles indicate microscopic damage.

Conclusions

We have demonstrated a correlation between an acid wash procedure, and the reduction of measured cerium impurities on polished fused silica optical flats. At the same time, the acid wash procedure affects a substantional, -30%, improvement in the measured damage threshold at 248 nm. It is tempting to speculate that damage was initiated in unwashed samples by absorption at inhomogeneously distributed cerium inclusions introduced by the polishing procedure.

A choice must then be made concerning the general utilization of ceria as a polishing substance. A substantial body of evidence now suggests that ceria inclusions generally tend to render optics more susceptible to damage, particularly at UV wavelengths. Should one then give up the convenience, speed and finish avaiable from ceria polishing in return for increased damage thresholds? Should this, or some other cleaning procedure [14] be used routinely with ceria-polished optical components? These choices must be made by the individual user or user-group, in recognition of the particular application intended, and lifetime needed.