Extreme UV secondary electron yield measurements of Ru, Sn, and Hf oxide thin films

Abstract. Background: The secondary electron yield (SEY) of materials is important for topics as nanoparticle photoresists and extreme ultraviolet (EUV) optics contamination. Aim: Experimentally measure SEY and secondary electron energy distributions for Ru, Sn, and Hf oxide. Approach: The SEY and energy distribution resulting from 65 to 112 eV EUV radiation are measured for thin-film oxides or films with native oxide. Results: The total SEY can be explained by EUV absorption in the topmost nanometer of (native) oxide of the investigated materials. Conclusions: Although the relative SEY of Ru and Sn is well-explained by the difference in EUV absorption properties, the SEY of HfO2 is almost a factor 2 higher than expected. Based on the energy distribution of secondary electrons, this may be related to a lower barrier for secondary electron emission.

Extreme UV secondary electron yield measurements of Ru, Sn, and Hf oxide thin films 1 Introduction Irradiation of materials with energetic photons, ions, or electrons results in generation of secondary electron emission from the surface. Secondary electron emission is important for devices such as photon counters, plasma display panels, and electron multipliers, 1,2 whereas for plasma devices, secondary electrons can have adverse effects on stable operation. 3,4 Knowledge about the secondary electron emission of materials is, thus, highly relevant for various device applications. In addition, secondary electrons also play an important role in chemical reactions at surfaces. In the x-ray and extreme ultraviolet (EUV) wavelength range (around 13.5 nm), the typical attenuation length of photons in matter is much longer than that of the generated photoelectrons and secondary electrons. Therefore, it is often assumed that secondary electrons play a much more important role in driving chemical reactions as compared to direct photoabsorption processes. 5 Quantitative measurements of secondary electron yields (SEYs) induced by EUV radiation were performed on EUV mirrors and thin films of protective coating materials, most notably Ru, in order to understand the role of secondary electrons in EUV-induced contamination processes. 6 Although EUV-induced reactions of contaminants on mirrors are mostly undesirable, the process of exposure of photoresist-coated wafers for lithographic patterning requires efficient use of the available EUV power to ensure high throughput in the chip manufacturing process. Since EUV enables printing of features down to 7 nm lateral size, photoresist layers should be relatively thin (on the order of 40 nm) to prevent collapse defects, but should still absorb a sufficiently high fraction of the available light in order to ensure high production throughput. [7][8][9][10] In recent years, researchers have, therefore, investigated resists with metal oxide nanoparticles composed of ZrO 2 , HfO 2 , or SnO 2 , which have higher EUV absorption coefficients compared to organic resists. In photoresists of this type, the relatively high EUV absorption in the metal oxides efficiently generates photoelectrons and secondary electrons, which then induce chemical reactions in organic ligands on the particle, thereby inducing the desired chemical modification upon exposure. 9 Recently, Fallica et al. 8 did a detailed study of the EUV absorption of different photoresists containing metal oxide particles or clusters, using spin-coated resist films on EUV transmissive membranes. In view of the before-mentioned importance of secondary electrons for inducing chemical reactions, the SEY of metal oxides used in resists may also play a role in their sensitivity. In this paper, we present a model study on the SEY and energy distribution of (native) oxide films of Hf and Sn, compared to a Ru thin film as reference material. 6 Measurements of the total SEY versus photon energy show that the SEY of a Ru thin film can be explained by absorption of EUV radiation in the native oxide present on the film. The relative SEY of native SnO 2 , as compared to native RuO 2 , is well explained by the difference in EUV absorption of both materials. The measured SEY of HfO 2 is, however, almost a factor 2 higher than expected based on the EUV absorption. This study shows that the SEY of materials may be, along with EUV absorption, an important factor for the sensitivity of EUV photoresists.

Sample Preparation and Precharacterization
Ru thin films were used as reference for the SEY measurements, since the EUV optical properties of Ru are well known and the SEY has been studied previously. 6 20-nm Ru films were deposited by direct current magnetron sputtering on a Si wafer substrate. A 20-nm HfO 2 film was deposited by atomic layer deposition (ALD) in a Picosun R200 ALD reactor, using TEMAHf (tetrakis(ethylmethylamido)hafnium(IV)) and ozone as precursors at a growth temperature of 250°C, which gave a growth per cycle of 0.086 nm. As substrate, a Si wafer precoated with 20 nm Mo (deposited by direct current magnetron sputtering) was used. The Mo film has no particular function in the current experiment, but was deposited for other planned experiments on samples from the same batch, for which a good conductive layer below the oxide was required. A native SnO 2 film on Sn was obtained by depositing 20 nm of Sn metal onto a Si wafer piece precoated with 20 nm Mo (from the same batch as used for the HfO 2 sample) with a thermal evaporator. Details on Sn deposition can be found in Ref. 11. The chemical composition of all samples was precharacterized using a Thermo Scientific ThetaProbe x-ray photoelectron spectroscopy (XPS) instrument, employing monochromatic AlK α radiation (photon energy hν ¼ 1486.7 eV). All samples could be measured without using a flood gun for charge compensation, indicating that the oxide films were thin enough to prevent significant sample charging. The Hf 4f spectral region of the HfO 2 sample could be fitted with a single doublet with peak position 18.4 eV for the Hf 4f 7∕2 peak, which indicates that stoichiometric HfO 2 was produced. Angle-resolved XPS calculations according to the method and attenuation lengths described by Cumpson and Seah 12 indicated that the Ru film had a native RuO 2 thickness of 0.8 nm, whereas the Sn film was covered with about 2 nm of native SnO 2 .

Secondary Electron Measurements
EUV-induced SEY was measured at the ID beamline 13 at the metrology light source (MLS) of the Physikalisch-Technische Bundesanstalt, Berlin, Germany. A Scienta R4000 hemispherical electron analyzer, operated at a pass energy of 20 eV, was used for measurements of the kinetic energy distribution of emitted photoelectrons and secondary electrons at three different wavelengths λ of synchrotron light: λ ¼ 18.0 nm (hν ¼ 68.88 eV), λ ¼ 13.5 nm (hν ¼ 91.84 eV), and λ ¼ 12.0 nm (hν ¼ 103.33 eV). For all results shown in this paper, the synchrotron light was incident at an angle of 45 deg relative to the sample surface normal, whereas photoelectrons emitted along the sample surface normal were detected. The typical photon flux applied for these measurements was 1.4 × 10 10 s −1 , corresponding to a radiation power of 0.21 μW, focused on a spot on the sample of ∼1 mm × 0.015 mm. The instantaneous peak irradiance from a single-bunch pulse from the synchrotron is a factor 1.25 higher than the specified time-averaged radiation power. Typical measurement times of electron spectra were 2 min. In order to check the effect of limited sensitivity and/or analyzer transmission for low-energy secondary electrons, all secondary electron spectra were measured both with the sample grounded and with a negative sample bias between 8.52 and 8.64 V, in order to accelerate electrons toward the detector. The sample bias was checked with a digital volt meter for each measurement. The pressure in the spectrometer system during operation ranged from 5 × 10 −10 to 1 × 10 −9 mbar. Most dominant residual gas components were water and hydrogen (partial pressures between 1.5 × 10 −10 to 3 × 10 −10 mbar), whereas C and CO showed an order of magnitude lower partial pressure. SEY spectra measured on the Ru reference sample were repeated to check for possible influence of photoinduced contamination, but no change in SEY caused by previous measurements on the same spot was found.
Quantitative total yield SEY measurements were performed by measuring the sample current (which was on the order of 1 nA) as a function of wavelength of the incident light, which was scanned from 11 to 19 nm (photon energy 111.9 to 65.4 eV), while simultaneously measuring the ring current of the synchrotron. In order to calibrate the ring current against the photon flux incident on the sample, a separate wavelength sweep was performed where the signal of a calibrated EUV photodiode (type SXUV) positioned at the sample position was measured as function of wavelength and ring current. From these characteristics, the SEY was calculated in units of electrons per incident photon. All samples were measured as introduced, no surface cleaning was applied.   function of the analyzer. The RuO x sample shows metallic characteristics: the Fermi level (or high kinetic energy on-set of the spectrum) is at 0 eV binding energy within the precision of the energy calibration of the analyzer. The metallic characteristics can be assigned to the very small thickness of the native oxide and/or the conductive nature of RuO 2 . 14, 15 The HfO 2 and SnO x samples show a band gap: the highenergy on-set of the photoelectron spectrum is significantly below 0 eV binding energy. The measured position of the Hf4f 7∕2 peak in the case of the HfO 2 sample was 18.4 eV, identical to the precharacterization with the laboratory XPS. This shows that possible charging is similar (or absent) as in the reference XPS experiment. Also the Sn 4d level observed for the SnO x sample was in line with the XPS precharacterization.

Secondary Electron Energy Distribution
Low-energy secondary electron spectra for the three samples, acquired with 91.84 eV photons, are plotted in Fig. 2.
Where applicable, the kinetic energy has been corrected for an applied sample bias (used to accelerate electrons toward the analyzer) and a normalized intensity scale is used in order to facilitate comparison of the spectra. When comparing the low energy part of spectra acquired without (a) and with (b) sample bias, the most obvious difference is that measurements with sample bias show a much higher signal of electrons up to about 2.5 eV kinetic energy. This is attributed to limited transmission of the energy analyzer for such lowenergy electrons and/or deflection of low-energy electrons by stray magnetic and electric fields in case no accelerating bias is used. In addition, it cannot be excluded that, in the absence of sample bias, the emitted secondary electron flux reduces secondary electron emission due to electrostatic repulsion. This may be an additional reason for a reduced signal of low-energy electrons if no sample bias is used.
In comparison with the Ru film, which only has a thin native oxide, the SnO x and HfO 2 sample showed a higher total SEY (as shown in Sec. 3.3) and a narrower electron energy distribution with a peak at lower energy. This is in line with previous reports for samples with (native) oxide films and may be attributed to two effects. First, it has been proposed that electron emission causes a positive surface charge. This charge may attract electrons from below the surface and thereby enhance the emission of secondary electrons. 15,16 Alternatively, it has been demonstrated that under conditions for photoelectron emission, downward band bending can occur in (high bandgap) semiconductors, as schematically indicated in Fig. 3(a). 17 This will also lower the barrier for secondary electron emission and make the energy distribution more narrow with a peak at lower kinetic energy. 15 For both the unbiased and biased cases, the kinetic energy of the peak maximum and width of the secondary electron distribution are highest for the RuO x . However, the kinetic energy peak maximum and width for SnO x is greater than that for HfO 2 only in the unbiased case. The larger relative shift of the SnO x peak kinetic energy (and decrease in width) upon application of sample bias, indicates that the barrier for electron emission is more significantly influenced by the application of sample bias, compared to HfO 2 . This change of peak maximum and width was consistent for all three photon energies studied. The origin of this effect may be related to the before-mentioned influence of band bending on the barrier for emission of secondary electrons. A negative sample bias applied to the substrate holder will enhance the band bending over the ∼2 nm insulating native SnO x film [see Fig. 3(b)], thereby enhancing the transport of electrons toward the surface and lowering the barrier for emission of secondary electrons. On the contrary, the same substrate bias generates an order of magnitude lower electric field over the HfO 2 layer, which was a 20-nm-thick oxide film. In that case, a negative sample bias thus has a smaller effect on lowering the barrier for secondary electron emission. Figure 4 shows secondary electron energy distributions for the HfO 2 sample acquired with three different photon energies, both with and without sample bias. The shape of the secondary electron distributions is virtually independent of the photon energy used for excitation, which can be explained by the fact that low-energy secondary electrons are generated by multiple excitations where primary photoelectrons generate secondary electrons that may themselves generate lower energy secondaries. Similarly, no significant influence of the photon energy was found for the secondary electron spectra of RuO x and SnO x samples.

Quantitative Total Secondary Electron Yields
Quantitative measurements of the SEY were performed by measuring sample current, without application of sample bias. Figure 5 shows the obtained SEY in electron per photon as a function of photon energy (solid lines). The measured SEY for the RuO x sample (black line) is in good agreement with previous measurements of Ru with native oxide by Yakshinskiy et al. 6 Within the investigated photon energy range, the SEY drops with increasing photon energy. Since generation of photoelectrons and secondary electrons is the main process by which EUV is absorbed in matter, 5 we investigated whether the energy dependence of EUV absorption in the sample can explain the measured SEY curve. The attenuation length of the employed synchrotron light, defined as the thickness of material needed to attenuate normally incident light to 1∕e intensity, ranges from roughly 15 to 64 nm for the materials studied in this work, as outlined in Table 1. 18,19 This means that all investigated thin films absorb only part of the incoming synchrotron light. On the contrary, the attenuation length for low-energy electrons is much shorter. Taking a kinetic energy of 50 eV as reference value, the attenuation length of electrons in RuO 2 , HfO 2 , and SnO 2 is around 0.25 nm as calculated according to the method of Cumpson and Seah. 12,20 Consequently, it is expected that most of the photoelectrons and secondary electrons emitted from the sample surface will originate from the topmost nm of the sample. We, therefore, compare the measured SEY with the calculated absorbed photon power density in 1 nm of RuO 2 , HfO 2 , and SnO 2 . To this end, the transmission T of 1 nm of these materials is calculated as function of photon energy according to the x-ray absorption data from Henke et al. 18,19 The absorbed photon energy density is then obtained from hνð1 − TÞ, where hν is the photon energy. The absorbed photon energy density for 1-nm RuO 2 is plotted with the dashed black line in Fig. 5, on the right vertical axis. The scale of this axis is adjusted such that an optimal correspondence of the absorbed photon energy with the measured SEY curve is obtained. As can be observed from this figure, the dependence of the SEY on photon energy can be nearly perfectly described by photon absorption in the topmost ∼1 nm of RuO 2 on the sample. On the contrary, when the SEY is compared to the absorbed photon energy in 1 nm of Ru metal (black dotted line), no good agreement of the dependence of the SEY on photon energy can be achieved. It is thus unlikely that photon absorption in the Ru metal below the ∼0.8 nm native oxide film has a significant influence on the emitted secondary electrons, since in that case one would expect more similarity of the SEY curve to the energy dependence of photon absorption in Ru metal. From this, we conclude that photon absorption in the topmost nm of material is decisive for the emitted secondary electrons, in line with expectations based on the attenuation length of low-energy electrons generated by EUV light. Subsequently, we compare the SEY of the SnO x and HfO 2 samples versus photon energy with absorbed photon energy in 1 nm films of SnO 2 and HfO 2 . It should be noted that all dashed curves representing absorbed photon energy density are scaled with the same factor as the absorbed photon    Fig. 4 Energy distribution of secondary electrons from the HfO 2 thin-film sample for different photon energies. Solid lines indicate measurements without sample bias, and dashed lines indicate measurements with bias. Similarly, also for the RuO x and SnO x samples, no significant influence of the photon energy on the shape of the energy distribution was found. energy curve for RuO 2 . Figure 5 thus provides a direct comparison of the expected SEY based on photon absorption and the actual measured total SEY, using the RuO x sample as a reference. For the Sn film with native oxide, the photon energy dependence of the SEY roughly follows the expected dependence from the absorbed photon energy, only the SEY is about 10% lower than expected. For the HfO 2 film, the measured SEY is almost a factor 2 higher than expected based on the absorbed photon energy. This high SEY is probably related to a low energy barrier for secondary electron emission from HfO 2 , which is consistent with the observed low peak energy and small peak width of the SEY curve without bias in Fig. 2(a).

Summary and Conclusions
In this paper, we presented measurements of the SEY and secondary electron energy distribution of thin films of Ru and Sn with native oxide, as well as a stoichiometric HfO 2 film. The shape of the low-energy secondary electron energy distribution up to 20 eV kinetic energy is virtually independent of exact photon energy within the probed range from 69 to 103 eV. This is expected, since these low-energy secondary electrons result from multiple excitations induced by higher energy photoelectrons. The shape and peak position of the energy distribution is material dependent, which is related to differences in the energy barrier for secondary electron emission. [15][16][17] Photon energy-dependent quantitative measurements of the total SEY show that the SEY of Ru can be explained by photons absorbed in the topmost nanometer of native oxide on top of the thin film. This is a consequence of the short attenuation length of about 0.25 nm for photo-and secondary electrons in the studied energy range, which is roughly an order of magnitude less than the attenuation length of EUV photons. The SEY of Sn with native oxide is well-explained by the higher EUV absorption compared to RuO 2 , whereas the total yield of HfO 2 is a factor 2 higher than would be expected based on EUV absorption. These measurements show that the energy barrier for secondary electron emission is an important factor to consider for processes where secondary electrons can play a decisive role, such as contamination of EUV optical elements and inorganic nanoparticle photoresists. Additionally, the observation that the outermost (few) nm of material appears decisive for the SEY indicates that photoresists with metal oxide nanoparticles, in which secondary electrons generated in the metal oxide induce reactions in organic ligands, will make most efficient use of the available EUV photons when the size of the metal oxide particles is on the order of a few nm.