Open Access
Add to BibSonomy
Abstract
Hafnium oxide thin films with varying oxygen content were investigated with the goal of finding the optical signature of oxygen vacancies in the film structure. It was found that a reduction of oxygen content in the film leads to changes in both, structural and optical characteristics. Optical absorption spectroscopy, using nanoKelvin calorimetry, revealed an enhanced absorption in the near-ultraviolet (near-UV) and visible wavelength ranges for films with reduced oxygen content, which was attributed to mid-gap electronic states of oxygen vacancies. Absorption in the near-infrared was found to originate from structural defects other than oxygen vacancy. Luminescence generated by continuous-wave 355-nm laser excitation in e-beam films showed significant changes in the spectral profile with oxygen reduction and new band formation linked to oxygen vacancies. The luminescence from oxygen-vacancy states was found to have microsecond-scale lifetimes when compared with nanosecond-scale lifetimes of luminescence attributed to other structural film defects. Laser-damage testing using ultraviolet nanosecond and infrared femtosecond pulses showed a reduction of the damage threshold with increasing number of oxygen vacancies in hafnium oxide films.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
During the last two decades, hafnium dioxide (HfO2) thin films have been the subject of extensive studies because of their prospective application in the semiconductor industry in metal–oxide–semiconductor transistors. In the same period, owing to its relatively large band gap of ~5.7 eV, HfO2 became the preferable high-refractive-index material used in optical thin-film coatings for high-power, high-energy laser applications with stringent requirements regarding laser-induced damage thresholds [1]. Structural defects (vacancies, interstitials, etc.) in HfO2 material are of special interest because of their impact on electrical properties, such as carrier mobility and threshold voltage shifts in semiconductor devices [2,3]. In the case of high-power pulsed-laser optics applications, structural defects can be a source of enhanced absorption, leading to laser damage [4–6], which limits the overall system energy output. Consequently, finding theoretical and experimental means of characterization of structural defects in HfO2 thin films is of great importance. Significant advances have been made in modeling electronic properties of HfO2 structural defects, particularly oxygen vacancies [7–9]. These calculations revealed the presence of deep oxygen-vacancy defect energy levels inside the band gap and shallow populated and unpopulated states located near the bottom edge of the conduction band, which might play an active role in the trapping of electrons and holes. Modeling [8] also produced data on optical transitions with the largest oscillator strength involving oxygen-vacancy states, most of which are thermodynamically stable, excluding only double-negatively charged vacancy states. On the other hand, experimental data on optical characteristics of the oxygen vacancies, especially in the spectral range with energies <4 eV, are very limited. Takeuchi et al. [10], using spectroscopic ellipsometry for 9.4-nm HfO2 thin films deposited on silicon substrates, found an additional absorption peak in the range of 4.5 eV to 5 eV, below the onset of the fundamental absorption edge around 5.7 eV. Rapid thermal annealing in an oxygen atmosphere at temperatures of 500°C to 700°C led to a drastic reduction in peak intensity, which allowed linking of the peak origin to oxygen vacancies. As was demonstrated in [11], luminescence measurements combined with thermal annealing of HfO2 films can also be used to characterize oxygen-vacancy defects. In particular, e-beam−deposited, 120-nm-thick HfO2 films that were first annealed in an argon atmosphere using temperatures between 200°C and 600°C were found to be amorphous at lower annealing temperatures and crystalline when annealed at temperatures ≥400°C. Broad luminescence (550-nm to 850-nm spectrum) excited in these films by a 514-nm line of an Ar+ laser was strongly dependent on the films’ annealing temperature (degree of crystallinity) being two orders of magnitude more intense for T = 600°C than as deposited. Further annealing of these films in an oxygen atmosphere has led to the full quenching of the luminescence, thereby confirming a link between luminescence and oxygen vacancies. It should be noted that nearly all previous studies of HfO2 thin-film photoluminescence (PL) [12–16] used ultraviolet (UV) (4.5-eV to 7.5-eV) excitation sources, and the resulted PL spectra were mostly represented by a broad band centered at 2.7 eV (~460 nm). This band was also observed using cathodoluminescence [17] but the link of this band to oxygen vacancies remains unclear. For instance, Ito et al. [13] found no correlation of 2.7-eV band intensity with oxygen content in hafnium oxide film. Regarding the role of oxygen vacancies in HfO2 film’s laser-damage performance, the only study conducted (to our knowledge) involved film annealing in air at 473 K as a tool to controlling the oxygen content [18]. This study demonstrated reduced absorption and increased 1064-nm, 12-ns pulse damage thresholds after annealing, which was attributed to a reduced number of oxygen vacancies in the film’s structure.
In the current study, HfOx thin films with an oxygen content varying from stoichiometric [oxygen-to-hafnium (O/Hf) atomic ratio x = 2] to nonstoichiometric with ratio x < 1.7 were prepared to find characteristic optical signatures of oxygen vacancies in the film’s structure in a wide spectral range from near-ultraviolet to near-infrared. Optical absorption spectroscopy performed using nanoKelvin calorimetry [19] revealed an enhanced absorption attributed to oxygen vacancies in the spectral range of 325 nm to 750 nm. Enhanced absorption at 355 nm in samples with a reduced oxygen content was confirmed using photothermal detection instrumentation. Luminescence excited by a 355-nm continuous-wave (cw) laser showed the presence of optical bands linked to oxygen-vacancy states. Finally, pulsed laser–induced damage threshold measurements confirmed the detrimental role of oxygen vacancies to the laser-damage resistance of hafnium oxide thin films.
2. Experimental
Hafnium oxide thin films with 175-nm physical thickness were deposited on polished fused-silica substrates containing an additional 500-nm-thick SiO2 film layer deposited in the same coating run. Conventional electron beam (e-beam) evaporation and dual ion-beam–sputtering (IBS) systems were used for film deposition with the goal of comparing the impact of oxygen content in porous (e-beam) and bulk-like (IBS) films. The e-beam system used Hf metal as a starting material that was oxidized into hafnium oxide because of the presence of oxygen in the vacuum chamber. The IBS deposition used an Ar+ ion beam, a Hf target, and oxygen flow into the chamber to produce an oxide film. In the case of e-beam deposition, the oxygen content in the film was controlled by the deposition rate and oxygen back-fill pressure. An increase in the deposition rate or a reduction in oxygen pressure results in a reduced oxygen presence in the film. For the IBS system, oxygen flow rate was the main parameter controlling oxygen content in the film. The oxygen content in the films was evaluated using x-ray photoelectron spectroscopy (XPS). Before XPS measurements, the film samples were subjected to ion-beam etching inside the XPS vacuum chamber to remove a few nanometers of surface material containing organic contamination manifested by the presence of the carbon peak in the XPS signal. Each XPS data point is a result of averaging over five sample sites. All relevant deposition parameters for both systems and XPS measurement results are summarized in Table 1. As denoted in Table 1 and for future reference, the e-beam film samples will be designated as EB1 (standard), EB2 and EB3 (both with reduced oxygen content), and IBS film samples as IBS1 (standard), IBS2 (with excessive oxygen content), and IBS3 (with reduced oxygen content). For better comparison, O/Hf atomic ratios are also shown normalized to an O/Hf ratio of 2 of the standard samples EB1 and IBS1.
Table 1. HfOx film deposition conditions and x-ray photoelectron spectroscopy (XPS) measurement results. Column xnorm shows XPS data normalized to perfectly stoichiometric O/Hf atomic ratios for standard samples EB1 and IBS1.
The microstructure of the films was characterized by measuring the x-ray diffraction using a Bruker D8 diffractometer at grazing angles varying from 5° to 90°.
Optical absorption measurements in the range of 325 nm to 1400 nm were performed at cryogenic temperatures in vacuum using a custom-built nanoKelvin calorimeter system, schematically shown in Fig. 1. It measures the temperature increase of a sample caused by absorption of light delivered from a light source tunable from about 300 nm to 1700 nm with a bandwidth of ~5 nm and a spectral power of ~2 μW/nm. The source consists of a commercial laser-plasma emitter (Energetiq, Eq. (-99) followed by a monochromator. The light is coupled into the calorimeter using a multimode fiber. The cryostat operates at 4 K and works with a closed-cycle refrigerator (TransMIT, PTD406-RV). Overall sensitivity in terms of measurable dissipated power is about 5 pW, corresponding to a temperature increase of ~10 nK. The sensitivity limit for 10 μW excitation is ~0.1 ppm (parts per million) over the entire spectral range, or 0.1 ppb (parts per billion) for 10 mW of excitation using a laser. The temperature is measured using paramagnetic sensors and SQUID (superconducting quantum interference device) readouts [20]. Calibration is performed using ohmic resistors as a heat source.
Fig. 1 Schematic of the nanoKelvin calorimeter setup. SQUID: superconducting quantum interference device.
In addition, a photothermal heterodyne imaging (PHI) system [21,22] described in detail in [22], was employed for absorption mapping with 0.4-μm spatial resolution. In short, the 355-nm pump- and 633-nm probe-laser beams are collinearly combined on the entrance of the high-numerical-aperture (N.A.) objective and focused onto the sample in a submicrometer spot (see Fig. 2). Absorption of the modulated pump beam produces local modulation of the refractive index, giving rise to modulated scattering of the probe beam detected by the photodiode and lock-in amplifier. Mounting the sample on the nanopositioning stage permits raster-type scanning and the creation of absorption maps with submicrometer resolution. Inherent to the photothermal pump/probe technique, PHI data are very difficult to convert to absolute absorption values; even the comparison between different materials may be complicated because of different thermal and optical responses of the materials. For that reason, we consider data obtained by the PHI technique complementary to nanoKelvin calorimetry and used primarily to qualitatively evaluate absorption changes with oxygen content in the film and mapping the spatial distribution of absorption.
Fig. 2 Schematic of the photothermal heterodyne imaging and luminescence setup. For luminescence measurements, the high-numerical-aperture (N.A.) objective is replaced by a 6-mm fused-silica focus lens.
To perform the photoluminescence studies, a 355-nm pump laser with a maximum power on a sample of ~10 mW was used as the excitation source. To optimize the signal, the laser beam was focused into an ~0.1-mm-diam spot on HfOx film using a fused-silica lens with a 6-mm focal length. The lens was employed instead of a high-N.A. objective, which appeared to produce weak luminescence when 355-nm light was passing through it. Luminescence measurements were also performed with a large, 5-mm-diam laser beam spot on the sample to avoid intensity-dependent annealing effects, which will be discussed later in Sec. 3.2. The luminescence signal was detected using a flip mirror on the PHI setup to redirect light coming from the sample onto the entrance slit of a spectrometer (Cornerstone MS260i, 15-nm/mm resolution) equipped with an avalanche array detector working in the photon-counting mode (see Fig. 2). To evaluate the luminescence lifetime, an acousto-optical modulator produced 500-ns-long square pulses with a frequency of 24 kHz (40-μs period) and a trailing edge shorter than 10 ns. Luminescence decay profiles were recorded by a multichannel scaler (Ortec MCS), whose temporal resolution was limited by a minimum channel width of 100 ns.
To evaluate the role of oxygen vacancies in the laser-damage process, the films were damage tested using 351-nm, 1-ns and 1053-nm, 600-fs laser pulses. The 1-on-1 protocol, when single-pulse irradiation is conducted for each sample site, was adopted. Damage was defined as any sample site modification after irradiation recorded by an imaging system with ~3-μm resolution using an image subtraction technique. In addition, all irradiated sites were analyzed with a Leica microscope at 1500 × magnification in order to differentiate damage initiated in the hafnium oxide film from damage initiated by substrate defects introduced during the glass-finishing process [23] and residing in the subsurface layer adjacent to the coating. This type of damage results in much larger damage craters, compared to craters originating from the hafnium oxide, owing to the additional 500-nm-thick SiO2 layer between HfOx film and the substrate. Consequently, substrate-defect–induced damage can be eliminated from damage statistics by using 1500 × -magnification optical microscopy.
3. Results and Discussion
3.1 Structural characterization
X-ray diffraction measurements (see Fig. 3) showed a high degree of crystallinity (identified as monoclinic form) in the e-beam films, with increased intensity in the diffraction peaks for samples with reduced oxygen content. IBS films remain amorphous for all three samples, and only the angular position and width of the broad peak change with oxygen content, indicative of a change in the average interatomic distances [24]. In particular, as compared to the standard IBS1 film, peak broadening for film IBS2 with excessive oxygen content indicates an increase in the average interatomic distances. In the case of the film IBS3 with reduced oxygen content, the peak transformation flattens and shifts to larger angles. The latter indicates the presence of areas with denser packing of the film material.
Fig. 3 X-Ray diffraction spectra of (a) e-beam films; standard EB1 film and EB2 film with reduced oxygen content show predominantly monoclinic crystalline structure. (b) Ion-beam sputtered films; standard (middle trace), with excessive oxygen (upper trace), and reduced oxygen content (lower trace) showed pure amorphous structure. The shift in the broad peak position reflects a change in the average interatomic distance.
3.2 Absorption measurements
Measurements performed using the nanoKelvin calorimeter (see Fig. 4) demonstrated increased absorption (roughly 2 × ) for the sample with reduced oxygen content (EB2) as compared to the standard sample (EB1) for wavelengths shorter than 750 nm (energy E > 1.65 eV). According to results of ab initio theoretical calculations [8], this enhanced absorption can be attributed to occupied oxygen-vacancy defect states located deep in the band gap and depicted schematically in Fig. 5. Absorption from these states can lead to the excitation of resonance states (predicted in [8]), located near the bottom of the conduction band, and followed by radiative (luminescence) or nonradiative relaxation. The other absorption pathway, at energies >3.2 eV, is optical transition of an electron into the conduction band followed by fast, <1-ns [4] relaxation out of the conduction band through direct recombination with holes in the valence band or to either the unoccupied mid-gap defect states or the shallow trap states located inside the band gap, below the bottom of the conduction band. The latter states may exhibit longer lifetimes on the millisecond scale [4]. These states can play an important role in the pulsed-laser damage process, providing an additional channel for ionization during damage initiation [5]. It should be noted that in the spectral range from 800 nm to 1400 nm, absorption is dominated by the 3-mm-thick fused-silica substrate (see Fig. 4). After subtracting the substrate contribution, no measurable difference between samples EB1 and EB2 can be found in this portion of the spectrum corresponding to transitions from shallow occupied oxygen-vacancy states [8]. It suggests that near-IR absorption is from either intrinsic, polaron-type states [25] or states other than oxygen-vacancy defect states. This conclusion is supported by theoretically predicted [8] low oscillator strengths (~0.005) for transitions from shallow oxygen-vacancy states.
Fig. 4 Absorption spectra of e-beam films, EB1 and EB2, and a substrate recorded in the range 325 nm to 1400 nm by the nanoKelvin calorimeter. A bare substrate was measured as a reference. The absorption spectra of the films were obtained after subtracting the substrate absorption and taking into account film interference effects.
Fig. 5 Energy diagram based on results of theoretical ab initio calculations [8] illustrating possible optical excitation and relaxation (nonradiative and radiative) processes.
The results of 355-nm absorption measurements using the PHI system were obtained from raster scanning the film samples. One of the characteristic features of 355-nm laser interaction with e-beam–deposited HfO2 films is that both absorption (reported earlier in [26]) and luminescence signals showed annealing effects. The annealing effect is dependent on both the intensity in the focal spot and the exposure time (dose of laser energy passing through the film). Figure 6(a) shows a typical example of an absorption map recorded in this work for sample EB2 with reduced oxygen content, where the central part was exposed before raster scanning for 5 min by a laser beam with a few milliwatts of power focused into a 0.5-μm spot. The cross-sectional profile shows an order-of-magnitude reduction in the photothermal signal at the exposed location. Other (peripheral) portions of the map show a signal characteristic of a film that is unaffected as a result of a very small film exposure of ~100 ms during the scanning process, owing to a small beam size of 0.5 μm. In contrast with e-beam films, IBS films showed either no annealing under exposure by the 355-nm laser (standard IBS1 film and IBS3 film with strongly reduced oxygen content) or relatively small, ~25% absorption annealing for sample IBS2 with excessive oxygen content [see Figs. 6(b) and 6(c)]. The difference might be attributed to the bulk-like (void-free) structure of IBS films versus the very porous structure of e-beam–prepared HfOx films [27]. A possible explanation is that laser light’s interaction with surface defects (surface dangling bonds, etc.) of e-beam films might be responsible for the observed annealing process. It should be noted that absorption maps for all samples were obtained using laser power and scan speeds that minimized annealing effects to a negligible level. These maps also showed relatively small ( ± 5% for a 20-μm × 20-μm scan) absorption spatial variation, which suggests an average absorber separation much less than the 0.4-μm system spatial resolution. The results of PHI measurements presented in Table 2 provided relative data on absorption at 355-nm wavelength for e-beam and IBS films. One can see that the e-beam films with reduced oxygen content showed significantly increased absorption as compared to standard (stoichiometric) films. This trend is in good agreement with absolute absorption data obtained using the nanoKelvin calorimeter. Considering IBS films, the IBS3 film with significantly reduced oxygen content showed two-orders-of-magnitude–larger absorption as compared to the standard IBS1 film. The only sample with increased oxygen content, IBS2, also showed increased absorption, presumably caused by the formation of oxygen interstitial defects (oxygen and peroxy bridges). When comparing standard stoichiometric films, the e-beam film showed several-times-lower absorption than the IBS film.
Fig. 6 Photothermal maps and corresponding cross sections for e-beam and ion-beam−sputtered (IBS) films exposed for 5 min at the map center by 355-nm, few-mW laser radiation focused into a 0.5-μm spot. (a) EB2 film showing strong absorption annealing effect, (b) IBS1 film with no absorption annealing, and (c) IBS2 film with ~25% reduction in photothermal signal.
Table 2. Photothermal measurements show enhanced absorption at 355 nm for films with reduced oxygen content. Enhanced absorption for ion-beam–sputtering (IBS) film with extra oxygen is attributed to structural defects other than oxygen vacancy
3.3 Photoluminescence spectroscopy results
Luminescence excited in HfOx films by a 355-nm laser was investigated only for the e-beam films since no measurable luminescence at a maximum laser power of ~10 mW was detected in the IBS films. A similar observation was reported in another luminescence study [14], where 266-nm laser excitation produced intense luminescence in e-beam films but practically no luminescence in IBS films. It was suggested that high compressive strain, typical for IBS films, contributes to the luminescence quenching. Another possibility might be linked to the amorphous structure in IBS films, which provides a broad host matrix phonon continuum that increases the nonradiative energy relaxation rate and effectively quenches the luminescence. This mechanism is known for impurity-doped laser crystals [28] and can be considered for oxygen vacancies that, to a first approximation, can be structurally treated as impurities. The luminescence signal in the e-beam films also showed annealing effects (signal reduction) under continuous 355-nm laser exposure. The kinetics of annealing depicted in Fig. 7 for EB2 film shows a rapid reduction during initial (a few seconds) exposure followed by a slow decline during several tens of minutes. After approximately 20 min, the signal relaxation becomes sufficiently slow, <1%/min, which enables one to record the luminescence spectrum (requiring ~8 min) at conditions close to steady state. Corrected for sensitivity and the substrate contribution (<5%), the luminescence spectra in the range from 400 nm to 780 nm, as shown in Fig. 8, exhibit a complex profile. This indicates the presence of several overlapping bands corresponding to transitions from different defect states. The best fit, which provided reasonable agreement with the following quantitative analysis of the spectra, was achieved using four bands with Gaussian profile designated as L1, L2, L3, and L4. The broad luminescence spectrum observed in this study and spectra reported in earlier work [11–16] indicate strong electron–phonon interaction in HfO2 material, which makes the choice of the Gaussian band profile for the fitting procedure fully justified [29]. By overlapping the spectra on a single plot (see Fig. 9), one can see that a significant change in the contribution by each band takes place with reduction of the oxygen content in the film. In particular, there is a notable reduction in intensity at shorter wavelengths and a significant increase in the wavelength range >500 nm. To analyze the spectral content of different film samples, four wavelength positions—440 nm, 510 nm, 565 nm, and 660 nm—were chosen (see Fig. 8) to quantify the effect of different oxygen contents. It should be noted that the selected wavelengths are not reflecting peak positions of the four Gaussian bands (L1, L2, L3, and L4) but are rather linked to the spectral areas where oxygen reduction caused the strongest spectrum transformation. Each wavelength position has a 15-nm bandwidth defined by the 1-mm opening of the spectrometer slits and a 15-nm/mm instrument resolution. Considering the overlap of these bands, 440 nm from bands L1 and L2; 510 nm mostly reflects the contribution from band L3; 565 nm from a mix of bands L3 and L4; and 660 nm from band L4 only.
Fig. 8 Corrected for sensitivity luminescence spectra of (a) stoichiometric EB1 film and (b) EB3 film with reduced oxygen content. Spectra can be well approximated by convolution of the four Gaussian bands (dashed lines) shown on the graphs as L1, L2, L3, and L4. Dots represent experimental points. The solid line is a sum of Gaussian components.
Fig. 9 Comparison of luminescence spectra for stoichiometric film (EB1) and film with reduced oxygen content (EB3). Oxygen reduction leads to lower luminescence intensities at shorter wavelengths (reduction of oxygen interstitials) and enhanced luminescence at wavelengths >500 nm attributed to oxygen-vacancy formation.
Several types of luminescence measurements were conducted using this choice of wavelengths. First, luminescence annealing kinetics was recorded after opening the laser shutter. For each wavelength, a new sample site was utilized and reproducibility of the kinetics was confirmed by exploring several sites. Typical examples of relaxation kinetics for sample EB2 depicted in Fig. 10 demonstrate faster relaxation and stronger signal reduction for shorter wavelengths as compared to longer wavelengths. Taking into account that the luminescence kinetics from a particular excited state follows the kinetics of the excited state population, this observation provides initial evidence that several transitions from different defect states contribute to the luminescence excited by a 355-nm laser.
Fig. 10 Luminescence annealing kinetics for EB2 film sample at four selected wavelength positions. The difference in kinetics suggests contributions to luminescence from several different defect states.
The presence of luminescence annealing creates a challenge for comparing luminescence properties of different films. To accomplish this task, the peak luminescence signal was measured for each selected wavelength immediately after opening the laser shutter (opening time <0.5 s). The measurements were performed using either a focused laser beam with an ~0.1-mm spot on the sample (high-intensity regime) or a diverging beam with an ~5-mm spot on the sample (low-intensity regime). The latter provided a minimum annealing impact but dramatically reduced the signal, which prohibited one from collecting reliable data for the 660-nm wavelength. The results of luminescence-peak measurements presented in Fig. 11 clearly show a random change of the 440-nm signal with oxygen content. This rules out a link of L1 and L2 bands to oxygen-vacancy states and suggests a connection to other types of electronic defects. Possible candidates are dangling bonds at the surfaces of the porous film structure and oxygen bridges, where the bond connects two oxygen atoms instead of connecting oxygen to a hafnium atom. The latter defect is a good candidate since films with lower oxygen content and, consequently, lower probability of oxygen bridge formation show a much lower signal for 440 nm (L1 and L2 bands) as compared to the standard film (see Fig. 9). In the case of a 510-nm wavelength corresponding to the L3 band, the result is not consistent. There is some correlation of luminescence signal with oxygen reduction in the case of high intensity (0.1-mm beam spot) and an absence of correlation for low intensity (5-mm beam spot). A possible explanation might rest on a larger contribution to the signal at this wavelength from the L2 band than shown in Fig. 8, provided there is some uncertainty in the four-band deconvolution. In this case, under high-intensity exposure, the L2 band can experience a much stronger signal relaxation during the shutter’s opening time (see Fig. 10) than the L3 band, which preserves the dominant contribution of the L3 band and results in a better correlation with oxygen content. On the other hand, under low-intensity exposure and in the absence of relaxation, the L2 band can make a much stronger contribution, leading to “no correlation” with oxygen content.
Fig. 11 Luminescence peak signal for e-beam films EB1, EB2, and EB3 at four selected wavelength positions. [(a)–(d)] Small laser beam spot (high-intensity regime). Notable is the absence of luminescence signal correlation with oxygen content at 440 nm, modest correlation at 510 nm, and strong correlation at longer wavelength, 565 nm and 660 nm. [(e)–(g)] Large laser beam spot (low-intensity regime). No correlation with oxygen content at 440 nm and 510 nm and strong correlation at 565 nm.
Peak luminescence signals at longer wavelengths, 565 nm and 660 nm, show good correlation with a reduction of oxygen in the film in both high- and low-intensity regimes. This makes the link to oxygen-vacancy formation very likely for the L3 band and certain for the L4 band. Especially notable is the change in signal at 660 nm, which is comparable to noise (~250 dark counts) for the standard sample EB1 and increases by an order of magnitude for samples EB2 and EB3 with reduced oxygen content.
Luminescence lifetime measurements provide an additional insight into optical transitions involving oxygen-vacancy states. Table 3 provides a summary on lifetime measurements for the selected set of wavelengths. Each measurement was carried out on a fresh site not affected by previous laser irradiation. It should be noted that detection temporal resolution was limited to 100 ns (minimum channel width in the multichannel analyzer detection system). Consequently, any luminescence signal with decay faster than 100 ns was detected only in one channel and is denoted in Table 3 as <0.1 μs. Several important trends emerge from lifetime data. First, only fast nanosecond-scale lifetimes were measured at all selected wavelengths for standard stoichiometric sample EB1, except for 660 nm for which the luminescence signal was too small to obtain reliable data. Similarly, only nanosecond-scale luminescence decay was observed at 440 nm in all four samples, rendering L1 and L2 bands being associated with states characterized by short lifetimes. On the other hand, luminescence at 660 nm corresponding to the L4 band, previously assigned to oxygen vacancies, showed only microsecond-scale decay (see Fig. 12). Luminescence at intermediate wavelengths of 510 nm and 565 nm, influenced by the contribution of bands L3 and L4, contained both nanosecond- and microsecond-decay components, with a smaller nanonsecond-component contribution at 565 nm, as compared to 510 nm. The energy-level diagram presented in Fig. 13 offers a possible explanation of the origin and properties of L1, L2, L3, and L4 bands. Absorption of 355-nm (3.5-eV) laser light excites electrons from mid-gap defect states (intrinsic and oxygen-vacancy states) into the conduction band. This process is followed by a subpicosecond intraband relaxation R1 of electrons to the bottom of the conduction band resulting from electron–phonon collisions [30]. Further relaxation can go through radiative L1, L2, and L3 and nonradiative R2 and R3 pathways. The R2 corresponds to fast (<100-ps) [4] population of the shallow oxygen-vacancy trap states, characterized by long lifetimes [4] and giving rise to L4 band luminescence. R3 corresponds to nonradiative electron transition into the valence band, possibly through mechanism of electron-hole recombination. Based on observed luminescence annealing effects, this nonradiative relaxation mechanism should be more effective for intrinsic defect states, as compared to oxygen-vacancy states, which is also reflected in shorter radiation lifetimes measured for standard EB1 film. To summarize, luminescence excited by a 355-nm laser in the standard e-beam film is dominated by electronic defects other than oxygen vacancies and is characterized by excited states with nanosecond-scale lifetimes. In the case of films with reduced oxygen content, luminescence in the wavelength region >510 nm is strongly enhanced compared to the standard film, exhibits a microsecond-scale lifetime, and can be assigned to oxygen vacancies in the film structure.
Table 3. Luminescence lifetimes (in μs) measured for e-beam films at four selected wavelengths. Only nanosecond-scale lifetimes were recorded for standard (EB1) film. Films with reduced oxygen showed both nanosecond- and microsecond-scale luminescence. The latter is attributed to oxygen vacancies.
Fig. 12 Example of 660-nm wavelength luminescence kinetics with ~1.1-μs lifetime for film EB3 with reduced oxygen content.
Fig. 13 Schematic presentation of the luminescence band L1, L2, L3, and L4 formation resulting from electron excitation from defect states to the conduction band by a 355-nm photon.
3.4 Pulsed-laser damage thresholds
The results of laser-damage testing by 351-nm, 1-ns and 1053-nm, 600-fs pulses presented in Table 4 allow one to elucidate the role of oxygen vacancies in the laser-damage initiation process. In the earlier discussion of absorption results (see Sec 3.2), it was noted that near-UV absorption from occupied mid-gap oxygen-vacancy states can enhance the overall rate of electron transitions into the conduction band, which, in the case of pulsed-laser excitation, may be followed by electron heating, avalanche formation through collisional ionization, and damage [5]. Moreover, the population of the long-lived shallow trap states after electron relaxation from the conduction band with a time constant <1 ns may provide an additional source of ionization during a nanosecond laser pulse. Note that the absence of a contribution to absorption from shallow oxygen-vacancy states in the steady-state regime (see Sec. 3.2) does not rule out a contribution to ionization from shallow states under high intensity, far from the equilibrium pulse regime. In the case of femtosecond pulses, the presence of the mid-gap states effectively reduces the number of IR photons needed for multiphoton ionization, which seeds free electrons for avalanche formation, leading to damage [5,31]. All of this points to oxygen vacancies as possible initiators of the damage process. One can see in Table 4 that the reduction of oxygen in the film leading to oxygen-vacancy formation has a detrimental impact on both nanosecond- and femtosecond-pulse damage thresholds. The only film with a high oxygen content, IBS2, showed damage thresholds practically equal to thresholds for the standard stoichiometric film IBS1, which suggests that an excess of oxygen during IBS film deposition is not a concern considering film damage performance. A similar result was reported for IBS-deposited Sc2O3 films with various densities of oxygen interstitial point defects [32]. The measured 800-nm, 40-fs single-pulse damage thresholds appeared to be independent of this type of defect number densities for oxygen to scandium atomic ratios in the range of 1.86 to 2.05. Looking at the performance of standard samples, the e-beam film showed a higher nanosecond threshold than the IBS film, and the thresholds were equal in the case of femtosecond pulses. This difference in pulse-length dependence of damage performance may be attributed to a different mechanism of laser interaction with the film material, localized absorption and matrix heating controlled by thermal diffusion for nanosecond pulses, and multiphoton absorption followed by electron heating and collisional ionization in the case of femtosecond pulses.
Table 4. Single-shot nanosecond and femtosecond damage thresholds.
4. Conclusions
The e-beam and IBS-deposited HfOx films with varying oxygen content were characterized with the goal of finding an optical signature for oxygen-vacancy defects. The study showed that reducing the oxygen content in HfO2 films leads to structural changes, an increased crystallinity for e-beam films, and a change in the average interatomic distances in amorphous IBS films. Absorption spectroscopy revealed enhanced near-UV and visible-light absorption in films with reduced oxygen content. This result was linked to the formation of oxygen vacancies with occupied electronic states residing deep inside the band gap. No contribution to absorption from oxygen vacancies was found in the near-infrared portion of the spectrum corresponding to transitions from shallow defect states. Near-UV absorption was two orders of magnitude higher in IBS film with strongly reduced oxygen content, as compared to film with standard stoichiometry.
Luminescence excited by a 355-nm laser in e-beam films with reduced oxygen content revealed that the formation of luminescence bands in the wavelength region >510 nm can be clearly attributed to oxygen vacancies. Luminescence lifetime measurements established that luminescence originating from oxygen-vacancy excited states is characterized by microsecond-scale (~1-μs) lifetimes. In comparison, the luminescence linked to other structural film defects showed only nanosecond-scale (<100-ns) lifetimes. No measurable luminescence was detected for IBS films with the exact luminescence-quenching mechanism yet to be understood.
Finally, our measurements of nanosecond and femtosecond pulse damage thresholds showed that the formation of oxygen vacancies and associated mig-gap states in hafnium oxide thin films provides additional channels for material ionization by pulsed laser radiation, which facilitates plasma formation and damage.
Funding
Department of Energy National Nuclear Security Administration under Award Number DE-NA0001944; University of Rochester; New York State Energy Research and Development Authority.
Acknowledgment
This report was prepared as an account of work sponsored by an agency of the U.S. Government. Neither the U.S. Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the U.S. Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the U.S. Government or any agency thereof.
References and links
1. J. B. Oliver, S. Papernov, A. W. Schmid, and J. C. Lambropoulos, “Optimization of laser-damage resistance of evaporated hafnia films at 351 nm,” Proc. SPIE 7132, 71320J (2008). [CrossRef]
2. G. Ribes, J. Mitard, M. Denais, S. Bruyere, F. Monsieur, C. Parthasarathy, E. Vincent, and G. Ghibaudo, “Review on high-k dielectrics reliability issues,” IEEE Trans. Device Mater. Reliab. 5(1), 5–19 (2005). [CrossRef]
3. E.-A. Choi and K. J. Chang, “Charge-transition levels of oxygen vacancy as the origin of device instability in HfO2 gate stacks through quasiparticle energy calculations,” Appl. Phys. Lett. 94(12), 122901 (2009). [CrossRef]
4. D. N. Nguyen, L. A. Emmert, D. Patel, C. S. Menoni, and W. Rudolph, “Transient phenomena in the dielectric breakdown of HfO2 optical films probed by ultrafast laser pulse pairs,” Appl. Phys. Lett. 97(19), 191909 (2010). [CrossRef]
5. L. A. Emmert, M. Mero, and W. Rudolph, “Modeling the effect of native and laser-induced states on the dielectric breakdown of wide band gap optical materials by multiple subpicosecond laser pulses,” J. Appl. Phys. 108(4), 043523 (2010). [CrossRef]
6. S. Papernov, E. Shin, T. Murray, A. W. Schmid, and J. B. Oliver, “355-nm absorption in HfO2 and SiO2 monolayers with embedded Hf nanoclusters studied using photothermal heterodyne imaging,” Proc. SPIE 8530, 85301H (2012). [CrossRef]
7. A. S. Foster, F. Lopez Gejo, A. L. Shluger, and R. M. Nieminen, “Vacancy and interstitial defects in hafnia,” Phys. Rev. B 65(17), 174117 (2002). [CrossRef]
8. D. Muñoz Ramo, J. L. Gavartin, A. L. Shluger, and G. Bersuker, “Spectroscopic properties of oxygen vacancies in monoclinic HfO2 calculated with periodic and embedded cluster density functional theory,” Phys. Rev. B 75(20), 205336 (2007). [CrossRef]
9. T.-J. Chen and C.-L. Kuo, “First principles study of the structural, electronic, and dielectric properties of amorphous HfO2,” J. Appl. Phys. 110(6), 064105 (2011). [CrossRef]
10. H. Takeuchi, D. Ha, and T.-J. King, “Observation of bulk HfO2 defects by spectroscopic ellipsometry,” J. Vac. Sci. Technol. A 22(4), 1337–1341 (2004). [CrossRef]
11. J. Ni, Q. Zhou, Z. Li, and Z. Zhang, “Oxygen defect induced photoluminescence of HfO2 thin films,” Appl. Phys. Lett. 93(1), 011905 (2008). [CrossRef]
12. M. Kirm, J. Aarik, M. Jürgens, and I. Sildos, “Thin films of HfO2 and ZrO2 as potential scintillators,” Nucl. Instrum. Methods Phys. Res. A 537(1), 251–255 (2005). [CrossRef]
13. T. Ito, M. Maeda, K. Nakamura, H. Kato, and Y. Ohki, “Similarities in photoluminescence in hafnia and zirconia induced by ultraviolet photons,” J. Appl. Phys. 97(5), 054104 (2005). [CrossRef]
14. A. Ciapponi, F. R. Wagner, S. Palmier, J.-Y. Natoli, and L. Gallais, “Study of luminescent defects in hafnia thin films made with different deposition techniques,” J. Lumin. 129(12), 1786–1789 (2009). [CrossRef]
15. V. A. Gritsenko, D. R. Islamov, T. V. Perevalov, V. Sh. Aliev, A. P. Yelisseyev, E. E. Lomonova, V. A. Pustovarov, and A. Chin, “Oxygen vacancy in hafnia as a blue luminescence center and a trap of charge carriers,” J. Phys. Chem. C 120(36), 19,980–19,986 (2016). [CrossRef]
16. A. A. Rastorguev, V. I. Belyi, T. P. Smirnova, L. V. Yakovkina, M. V. Zamoryanskaya, V. A. Gritsenko, and H. Wong, “Luminescence of intrinsic and extrinsic defects in hafnium oxide films,” Phys. Rev. B 76(23), 235315 (2007). [CrossRef]
17. T. V. Perevalov, V. Sh. Aliev, V. A. Gritsenko, A. A. Saraev, V. V. Kaichev, E. V. Ivanova, and M. V. Zamoryanskaya, “The origin of 2.7 eV luminescence and 5.2 eV excitation band in hafnium oxide,” Appl. Phys. Lett. 104(7), 071904 (2014). [CrossRef]
18. C. Xu, P. Yi, H. Fan, J. Qi, Y. Qiang, J. Liu, C. Tao, and D. Li, “Correlations between the oxygen deficiency and the laser damage resistance of different oxide films,” Appl. Surf. Sci. 289(Supplement C), 141–144 (2014). [CrossRef]
19. B. Roshanzadeh, S. T. P. Boyd, and W. Rudolph, University of New Mexico, (private communication, 2017).
20. T. C. P. Chui, D. R. Swanson, M. J. Adriaans, J. A. Nissen, and J. A. Lipa, “Temperature fluctuations in the canonical ensemble,” Phys. Rev. Lett. 69(21), 3005–3008 (1992). [CrossRef] [PubMed]
21. S. Berciaud, D. Lasne, G. A. Blab, L. Cognet, and B. Lounis, “Photothermal heterodyne imaging of individual metallic nanoparticles: Theory versus experiment,” Phys. Rev. B 73(4), 045424 (2006). [CrossRef]
22. S. Papernov, A. Tait, W. Bittle, A. W. Schmid, J. B. Oliver, and P. Kupinski, “Near-ultraviolet absorption and nanosecond-pulse-laser damage in HfO2 monolayers studied by submicrometer-resolution photothermal heterodyne imaging and atomic force microscopy,” J. Appl. Phys. 109(11), 113106 (2011). [CrossRef]
23. Y. Zhou, P. D. Funkenbusch, D. J. Quesnel, D. Golini, and A. Lindquist, “Effect of etching and imaging mode on the measurement of subsurface damage in microground optical glasses,” J. Am. Ceram. Soc. 77(12), 3277–3280 (1994). [CrossRef]
24. A. Guinier, X-Ray Diffraction: In Crystals, Imperfect Crystals, and Amorphous Bodies (W. H. Freeman, San Francisco, 1963).
25. D. Muñoz Ramo, A. L. Shluger, J. L. Gavartin, and G. Bersuker, “Theoretical prediction of intrinsic self-trapping of electrons and holes in monoclinic HfO2.,” Phys. Rev. Lett. 99(15), 155504 (2007). [CrossRef] [PubMed]
26. S. Papernov, A. A. Kozlov, J. B. Oliver, T. J. Kessler, A. Shvydky, and B. Marozas, “Near-ultraviolet absorption annealing in hafnium oxide thin films subjected to continuous-wave laser radiation,” Opt. Eng. 53(12), 122504 (2014). [CrossRef]
27. H. K. Pulker, “Film deposition methods,” in Optical Interference Coatings, N. Kaiser and H. K. Pulker eds., Springer Series in Optical Sciences, A. Adibi, T. W. Hänsch, F. Krausz, B. R. Masters, H. Venghaus, H. Weber, H. Weinfurter, and K. Midorikawa eds. (Springer-Verlag, Berlin, 2003), pp. 131−154.
28. D. M. Calistru, S. G. Demos, and R. R. Alfano, “Dynamics of local modes during nonradiative relaxation,” Phys. Rev. Lett. 78(2), 374–377 (1997). [CrossRef]
29. L. Skuja, “Optical properties of defects in silica,” in Defects in SiO2 and Related Dielectrics: Science and Technology, G. Pacchioni, L. Skuja, and D. L. Griscom eds., Nato Science Series II (Kluwer Academic Publishers, The Netherlands, 2000), pp. 73−116.
30. M. Mero, A. J. Sabbah, J. Zeller, and W. Rudolph, “Femtosecond dynamics of dielectric films in the pre-ablation regime,” Appl. Phys., A Mater. Sci. Process. 81(2), 317–324 (2005). [CrossRef]
31. B. C. Stuart, M. D. Feit, S. Herman, A. M. Rubenchik, B. W. Shore, and M. D. Perry, “Nanosecond-to-femtosecond laser-induced breakdown in dielectrics,” Phys. Rev. B Condens. Matter 53(4), 1749–1761 (1996). [CrossRef] [PubMed]
32. C. S. Menoni, P. F. Langston, E. Krous, D. Patel, L. Emmert, A. Markosyan, B. Reagan, K. Wernsing, Y. Xu, Z. Sun, R. Route, M. M. Fejer, J. J. Rocca, and W. Rudolph, “What role do defects play in the laser damage behavior of metal oxides?” Proc. SPIE 8530, 85300J (2012). [CrossRef]
