Improved pumping speeds of oxygen-free palladium/titanium nonevaporable getter coatings and suppression of outgassing by baking under oxygen Open Access

Nonevaporable getter (NEG) coating is a technique in which the inner walls of a vacuum chamber are coated with a thin film of an NEG material. NEG coatings are widely used in accelerators, because they are oil-free, magnetic-field-free, vibration-free, economical, space saving, and energy efficient.^1–8 We recently developed a new NEG coating consisting of an oxygen-free thin film of Ti covered with a thin film of Pd, which we named the “oxygen-free palladium/titanium (Pd/Ti) coating.”^9–12 The advantages of the oxygen-free Pd/Ti coating are: (1) it can be activated by baking at 133 °C for 12 h;¹⁰ (2) the pumping speed for H₂ and CO does not decrease, even on repeated introduction of air and activation;¹¹ (3) the initial costs and running costs are low;¹⁰ (4) the coating processes do not require skilled technicians;^9–11 (5) the coating can be applied to large chambers with complex structures⁹ or to bellows;¹⁰ (6) the coating can be applied to NEG pumps;¹¹ and (7) oxygen-free Pd/Ti is thermally stable up to 260 °C.¹² 

When a thin film of oxygen-free Pd/Ti is exposed to air outside a clean room, the Pd surface becomes contaminated to some degree with carbon.¹⁰ This contamination is probably responsible for a decrease in the pumping speed of oxygen-free Pd/Ti for H₂ and CO, but this can be ameliorated by baking under O₂ atmosphere.¹¹ Here, we report an effective and simple method for removing carbon contamination from the Pd/Ti surface to improve the pumping speeds of oxygen-free Pd/Ti for H₂ and CO and to suppress the formation of H₂, CO, H₂O, and CH₄ by means of catalytic chemical reactions on Pd.

The apparatus used for the preparation of oxygen-free Pd/Ti samples for x-ray photoelectron spectroscopy (XPS) measurements has been previously described.¹⁰ Thin films of Ti and Pd were sequentially deposited on stainless-steel 304L (SS304L) substrates (8 × 8 × 1 mm) under ultrahigh vacuum (UHV) in the pressure range of 10⁻⁷–10⁻⁸ Pa. The purity of the Ti filament (Canon Anelva, 956-0010) was 98.85% or higher and that of the Pd filament (Nilaco, PD-341465) was 99.95%. The samples were then exposed to the atmosphere and installed in the sample bank of a photoelectron spectroscopy (PES) apparatus at beam line BL-13B of a synchrotron radiation (SR) facility (Photon Factory, KEK).¹³ A schematic structure of the oxygen-free Pd/Ti sample is shown in Fig. 1(a). The samples were treated in situ by one of the following procedures: (1) heating at 150 °C for 3 h under UHV (UHV-heated) or (2) heating at 150 °C for 3 h under an O₂ pressure of 1.3 × 10⁻⁴ Pa (O₂-heated). The samples were heated by electron bombardment in the preparation chamber of the PES apparatus. The treated samples and an unheated sample were then transferred to an analyzer chamber for XPS measurements.

XPS measurements were carried out at room temperature by using p-polarized SR light with a photon energy (hν) of 700 eV.¹³ The samples were irradiated with SR light at an incidence angle of 65° to the surface normal, and photoelectrons emitted in the surface-normal direction were detected. Photoelectron kinetic energies were analyzed by using a hemispherical electron-energy analyzer (Scienta SES200; Gammadata). The instrumental energy resolution with this energy analyzer and the SR light was estimated to be about 130 meV. The base pressure of the analyzer chamber was 2 × 10⁻⁸ Pa.

The surface morphologies of the unheated and UHV-heated oxygen-free Pd/Ti samples were observed by scanning electron microscopy (SEM) at the National Institute of Material Science (Tsukuba). The SEM observations were carried out by using a JSM-7000F instrument (JEOL Ltd., Tokyo) operated at an accelerating voltage of 15 kV.

The inner walls of a chamber made from SS304L stainless-steel were coated with an oxygen-free Pd/Ti thin film as previously described.¹⁰ The inner diameter and depth of the chamber were 147 and 236 mm, respectively, as shown in Fig. 1(b). The volume of the chamber was approximately 4 l, and the area of its inner walls was 1310 cm², assuming that it had ideally smooth surfaces.

Applications of NEG pumps can be roughly classified into two groups. The first involves evacuation in an isolated vacuum system, and the second involves evacuation in combination with a second vacuum pump, such as a turbomolecular pump (TMP). To investigate the vacuum quality in the oxygen-free Pd/Ti-coated chamber in these two cases, we carried out two experiments. In the first, the total and partial pressures in the coated chamber pumped by a TMP before and after closing a UHV valve between the chamber and the TMP were measured. In the second, we measured the total and partial pressures in the coated chamber, without a UHV valve, pumped by a TMP before, during, and after UHV or O₂ baking.

The total and partial pressures of the oxygen-free Pd/Ti-coated chamber before and after closing the valve were investigated by using the apparatus shown in Fig. 1(c). First, the coated chamber was pumped with a TMP (STP301; Edwards) through the pneumatic UHV gate valve. The TMP was pumped with an oil-sealed rotary pump (RP) through the foreline trap. The coated chamber was baked for 12 h at 150 °C and then cooled to room temperature. The gate valve was closed 12 h after the baking off. No gas was introduced during the UHV baking, whereas O₂ gas at a pressure of 1.3 × 10⁻⁴ Pa was introduced during the O₂ baking. The total and partial pressures were measured by using a Bayard–Alpert (B–A) nude ionization gauge (NIG-2F; Canon Anelva) and a quadrupole mass spectrometer (QMS) (Prisma 80; Pfeiffer Vacuum), respectively, before and after valve closure.

Before, during, and after UHV or O₂ baking, the total and partial pressures in the oxygen-free Pd/Ti-coated chamber with the TMP were measured by using the Prisma 80 QMS and the apparatus shown in Fig. 1(d). One hour after the commencement of the pressure measurements, the coated chamber was baked under UHV or O₂ for 3 h at 150 °C. The coated chamber was exposed to the air for seven days before the measurements.

Pumping speeds of the oxygen-free Pd/Ti-coated chamber for H₂ or CO were measured by the orifice method¹⁴ using the apparatus shown in Fig. 1(e). The apparatus consists of a UHV chamber with a B–A gauge and a leak valve for gas inlet (chamber A), a UHV chamber with another B–A gauge (chamber B), and an orifice (φ 10 mm × 0.5 mm) between chambers A and B.^15,16 The oxygen-free Pd/Ti-coated chamber was connected to chamber B as shown in Fig. 1(e). The B–A gauge in chamber A was calibrated by the manufacturer (Canon Anelva), and this calibrated gauge was used to calibrate the B–A gauge in chamber B. Chambers A and B were pumped by a TMP and baked at 150 °C for 24 h, whereas the coated chamber was left unbaked. The coated chamber and chambers A and B were then baked under UHV at 150 °C for 12 h. When the equipment had cooled to room temperature, the pumping speed for H₂ was measured. H₂ was introduced through a variable leak valve into chamber A [see Fig. 1(e)]. The coated chamber and chambers A and B were again baked under UHV at 150 °C for 12 h, and the pumping speed for CO was then measured by the same procedure as H₂. The pumping speed measurements after O₂ baking were carried out by a similar procedure. O₂ was introduced through the port with a 70 mm outer diameter (OD) ConFlat flange of the coated chamber. The pressures in chambers A and B during O₂ baking were about 1.3 × 10⁻⁵ and 1.3 × 10⁻⁴ Pa, respectively.

In the orifice method, the pumping speed (S) of a vacuum pump is given by the following equation:¹⁴ 

where P_A and P_B are the pressures in chambers A and B under the gas inlet, P_A0 and P_B0 are the base pressures in chambers A and B without gas inlet after closure of the gate valve, respectively, and C is the conductance of the orifice. The values of C for H₂ and CO at 26 °C were calculated to be 33.27 and 8.89 l s⁻¹, respectively.^15,16 The pumped quantity [Q(t)] at time t is given by the following equation:¹⁶ 

Figure 2 shows wide-scan XPS spectra of the unheated, UHV-heated, and O₂-heated oxygen-free Pd/Ti samples. The Fermi level was taken as the origin of the binding energy axis. The depth of probing was estimated to be 0.6–1.2 nm. The spectra are normalized by the intensities of the Pd 3d peaks. The C 1s peak is considered to be derived from graphene^17,18 and from other molecules containing carbon on the Pd surface, because the samples were exposed to the air outside a clean room before each measurement. Figure 2 shows that there was considerable carbon on the unheated sample, whereas levels of carbon on the UHV- and O₂-heated samples decreased to about a half and about one-tenth of the previous level, respectively.

Figures 3(a) and 3(b) show the deconvoluted Pd 3d_5/2 and C 1s peaks for the unheated, UHV-heated, and O₂-heated oxygen-free Pd/Ti samples. The Pd 3d_5/2 peaks of each sample were deconvoluted by using Voigt function. The Pd 3d_5/2 peak with a binding energy (BE) of 334.9 eV was assigned to Pd in the bulk,^19,20 whereas the Pd 3d_5/2 peak with a BE of about 335.4 eV probably arises from Pd covered by graphene or Pd with adsorbed carbon-containing molecules (Pd_C-ad).^17,18,21,22 The surface core-level shift of the Pd 3d_5/2 peak of the unheated sample (BE = 334.3 eV) was –0.6 eV and that of the UHV- or O₂-heated samples (BE = 334.5 eV) was –0.4 eV. These results suggest that the number of less-coordinated Pd atoms decreased after heating, because the surface core-level shifts of Pd 3d_5/2 of Pd(110), Pd(100), and Pd(111) surfaces have been reported to be −0.55, −0.47, and −0.22 eV, respectively.^23,24 The C 1s peaks with BE values of 283.8 to 284, 285.0 to 285.5, and 286.7 to 287.3 eV were assigned to graphene, CO-containing molecules on Pd(CO_ad), and CO₂-containing molecules on Pd(O–C = O_ad), respectively.^20,25–27

Figures 4(a) and 4(b) show SEM images of the Pd surface of the unheated and UHV-heated (at 150 °C for 12 h) oxygen-free Pd/Ti samples. The Pd surface of the unheated sample had an uneven structure with irregularities of several tens to several hundreds of nanometers and numerous steps, whereas the surface of the UHV-heated sample had a leaf-like structure with relatively large flat surfaces. These results are consistent with the XPS spectra, which suggest that the number of less-coordinated Pd atoms decreases after heating.

The photoemission intensity of a Pd peak of a clean Pd surface (I_Pd) is given by the following equation:²⁸ 

where N_Pd is the atomic density of Pd, σ_Pd is the photoionization cross section of Pd, λ_Pd is the inelastic mean free path (IMFP) in the Pd layers of a photoelectron emitted from Pd, and φ is the emission angle of a photoelectron in relation to the surface normal. When the total carbon coverage on Pd surface (θ_C) is less than one monolayer, the peak intensities of Pd (⁠$IPdC$⁠) and C (I_C) are given by the following expressions:

where d is the thickness of the carbon monolayer, $λPdC$ is IMFP in the C layer for a photoelectron emitted from Pd, and n_C and σ_C are the atomic densities in a monolayer and in a cross section of C, respectively. One monolayer (1 ML) of carbon coverage is defined as one monolayer of graphene covering the Pd surface. When φ is 0°, the value of θ_C is given by the following equation:

Assuming that the carbon on the Pd surface is in the form of graphene,^17,18,21,22 the value of n_C is 3.82 × 10¹⁵ cm⁻² and that of d is 0.4 nm.^29,30 The value of N_Pd is 7.03 × 10²² cm⁻³. The value of σ_C of C 1s is 0.15 Mb atom⁻¹ and that of σ_Pd of Pd 3d is 1.7 Mb atom⁻¹ when hν is 700 eV.^31,32 The values of λ_Pd and $λPdC$ were calculated to be 0.69 and 1.28 nm, respectively, by using the TPP-2M IMFP predictive equation.³³ The values of $IPdC$ and I_C were obtained from the background subtracted peak areas of the measured Pd 3d and C 1s spectra. According to the Lagrange–Helmholtz relationship, the detection efficiency of photoelectrons in a hemispherical electron-energy analyzer is proportional to E_k^−1/2, where E_k is the kinetic energy of the photoelectron. Because the values of E_k for Pd 3d and C 1s are approximately 360 and 410 eV, respectively, the corrected ratio of the XPS peak intensities of Pd 3d and C 1s [(I_C/$IPdC$⁠)_corr] was obtained by multiplying the measured intensity ratios by a factor of 1.07 [(I_C/$IPdC$⁠)_corr = I_C/$IPdC$ × 1.07]. The values of (I_C/$IPdC$⁠)_corr for the unheated, UHV-heated, and O₂-heated samples were 0.102, 0.039, and 0.009, respectively. By using the above equations, the corresponding values of θ_C were estimated to be 1, 0.5, and 0.1 ML, respectively.

So far, we have discussed the total carbon coverage. Next, we will discuss the graphene coverage, because the carbon on Pd/Ti samples consists of graphene, CO-containing molecules, and CO₂-containing molecules, as shown in Fig. 3(b). Graphene does not diffuse on Pd surfaces at room temperature, whereas carbon-containing molecules, such as CO, do so. Consequently, graphene is expected to have a large effect in reducing the pumping speeds for H₂ and CO. The graphene coverage of the Pd surface (θ_G) can be calculated by using the expression (I_G/$IPdC$⁠)_corr [(I_G/$IPdC$⁠)_corr = I_G/$IPdC$ × 1.07], where I_G is the peak area for C 1s, assigned to graphene. The values of θ_G for the unheated, UHV-heated, and O₂-heated samples were estimated to be 0.9, 0.3, and 0.04 ML, respectively. The heating conditions, the ratios of the XPS peak intensities, and the estimated total carbon and graphene coverages on the Pd surface are listed in Table I.

The total and partial pressures in the oxygen-free Pd/Ti-coated chamber were measured by the following procedure using the apparatus shown in Fig. 1(c). The coated chamber was baked at 150 °C for 12 h under UHV or under an O₂ pressure of 1.3 × 10⁻⁴ Pa and then cooled to room temperature. Twelve hours after the baking off, the pneumatic UHV gate valve between the chamber and the TMP was closed. Figure 5 shows the total and partial pressure curves for the coated chamber before and after closing the gate valve. The time when the valve was closed is taken as the origin of the time axis. The molecule with m/z = 28 was assigned to CO, because the amount of N₂ is usually negligible after baking if the vacuum chamber has negligible leakage. In the case of UHV baking the total, H₂, and CO pressures 5 h after the valve was closed were 4.2 × 10⁻⁶, 1.4 × 10⁻⁷, and 8.9 × 10⁻⁷ Pa, respectively, whereas in the case of O₂ baking, these pressures were 1.0 × 10⁻⁶, 3.6 × 10⁻⁸, and 5.8 × 10⁻⁸ Pa, respectively. The partial pressures of H₂ and CO were therefore improved by a factor of 3.9 for H₂ and of 15.3 for CO. These factors are much larger than those expected from the values of (1 – θ_G) (0.7 ML for UHV-heated and 0.96 ML for O₂-heated samples; see Table I). These results therefore suggest that the O₂ baking not only improves the pumping speeds of oxygen-free Pd/Ti thin films for H₂ and CO, but also suppresses outgassing of H₂ and CO from the coated chamber.^34,35

Carbon-containing contaminants on the Pd surface are considered to have been removed by catalytic chemical reactions such as

during UHV baking, and

during O₂ baking,^34,35 where the subscripts “ad” and “cat” refer to adsorption on the Pd surface and the catalytic chemical reaction, respectively. The CO, H₂, and CO₂ reaction products can be pumped by the TMP.

Figure 6 shows the partial pressures in the oxygen-free Pd/Ti-coated chamber before, during, and after UHV and O₂ baking. The apparatus used is shown in Fig. 1(d). During UHV baking, the partial pressure of CO₂ was 1 × 10⁻⁶–2 × 10⁻⁸ Pa, whereas that during the O₂ baking was 6 × 10⁻⁶–3 × 10⁻⁷ Pa. The partial pressures of CO₂ during the O₂ baking were larger by a factor of 6−15 than those during the UHV baking. This shows that O₂ reacts with the carbon on the Pd surface to form CO₂. During UHV baking, the partial pressure of H₂O was 2 × 10⁻⁵–9 × 10⁻⁷ Pa, whereas that during the O₂ baking was 3 × 10⁻⁵–1 × 10⁻⁶ Pa. The partial pressures of H₂O during the O₂ baking were larger by a factor of 1.1−1.5 than that during the UHV baking. This result suggests that H on Pd surface was removed by catalytic chemical reactions such as

during O₂ baking.³⁶ In the case of UHV baking, the H₂O, CO, CH₄, and H₂ pressures 0.1 h after baking off were 5.1 × 10⁻⁷, 4.8 × 10⁻⁹, 1.2 × 10⁻⁷, and 3.2 × 10⁻⁸ Pa, respectively, as shown in Fig. 6(c). In the case of the O₂ baking, the H₂O and CH₄ pressures were 1.3 × 10⁻⁷ and 4.0 × 10⁻⁸ Pa, respectively, whereas the CO and H₂ pressures were below their respective detection limits [see Fig. 6(c)]. These results show that O₂ baking is effective not only in reducing the H₂ and CO pressures, but also in reducing the H₂O and CH₄ pressures, suggesting that residual H₂O and CH₄ are derived from H and C adsorbed on the coated chamber. Catalytic chemical reactions such as those shown in Eqs. (7)–(9) appear to remove adsorbed carbon and hydrogen, and consequently suppress the formation of H₂, CO, H₂O, and CH₄. The total pressures 6  h after the end of UHV baking and O₂ baking were about 4 × 10⁻⁸ and 2 × 10⁻⁸ Pa, respectively. An improvement in the pumping speeds for H₂ and CO and the suppression of the formation of H₂, CO, H₂O, and CH₄ seem to be responsible for the improvement in the total pressure.

Figure 7 shows the measured pumping speeds of the oxygen-free Pd/Ti-coated chamber for H₂ and CO as a function of the pumped quantity after UHV or O₂ baking. The apparatus used is shown in Fig. 1(e). Pumping speeds for H₂ after UHV or O₂ baking were measured as 510–120 and 990–400 l s⁻¹, respectively, in the pumped-quantity range 0.01–10 Pa l. Those for CO were determined to be 960–40 and 1250–140 l s⁻¹, respectively in the pumped-quantity range 0.005–1 Pa l. The initial pumping speeds were improved by factors of 1.9 for H₂ and 1.3 for CO. These results indicate that O₂ baking improves the pumping speeds for H₂ and CO owing to removal of carbon contaminants. The improvement factor for CO (1.3) had a similar value to that of the ratio of 1 – θ_G for the O₂-heated sample (0.96 ML; see Table I) to that of 1 – θ_G for the UHV-heated sample (0.7 ML, see Table I) $(0.960.7=1.4)$⁠, whereas that for H₂ (1.9) is rather larger than the ratio of (1 – θ_G) (1.4). These results suggest that carbon contamination preferentially blocks the sites responsible for H₂ absorption. The pumping speeds, pumped quantities, P_A0, P_B0, P_A, and P_B for each measurement are listed in Table II.

We have previously described a new NEG pump with a 203 mm OD ConFlat flange using oxygen-free Pd/Ti thin films, for evacuating residual H₂ and CO.¹¹ The best pumping speeds for H₂ and CO of 680 and 900 l s^–1, respectively, were realized after baking at 150 °C for 12 h under an O₂ pressure of 1.3 × 10^–4 Pa (the same conditions as in the present O₂ baking).¹¹ These results are consistent with the present results that O₂ baking improves the pumping speeds of oxygen-free Pd/Ti for H₂ and CO.

We report a method in which UHV or O₂ heating is used to remove carbon contaminants from a Pd/Ti surface to improve pumping speeds of oxygen-free Pd/Ti for H₂ and CO and to suppress the formation of H₂, CO, H₂O, and CH₄. Samples of oxygen-free Pd/Ti that were unheated, heated in UHV at 150 °C for 3 h, or heated at 150 °C for 3 h under an O₂ pressure of 1.3 × 10⁻⁴ Pa were analyzed by XPS. The graphene coverage of the unheated sample was estimated to be 0.9 ML, whereas those of the UHV-heated and O₂-heated samples were estimated to be 0.3 and 0.04 ML, respectively.

The total and partial pressures of the oxygen-free Pd/Ti-coated chamber before and after valve closure were investigated after UHV or O₂ baking at 150 °C for 12 h. In the case of UHV baking, after the valve closure, the total pressure was 4.2 × 10⁻⁶ Pa and the partial pressures of H₂ and CO increased to 1.4 × 10⁻⁷ and 8.9 × 10⁻⁷ Pa, respectively, in 5 h. On the other hand, in the case of the O₂ baking, the corresponding pressures were improved to 1.0 × 10⁻⁶, 3.6 × 10⁻⁸, and 5.8 × 10⁻⁸ Pa in 5 h. These improvements are much larger than would be expected from the values of θ_G (0.3 ML for UHV-heated and 0.04 ML for O₂-heated samples). The results therefore indicate that the removal of carbon contamination not only improves the pumping speeds of oxygen-free Pd/Ti thin films for H₂ or CO, but also suppresses outgassing from the chamber.

Partial pressures in the oxygen-free Pd/Ti-coated chamber with a TMP before, during, and after UHV or O₂ baking were measured. During the UHV baking, the partial pressure of CO₂ was 1 × 10⁻⁶–2 × 10⁻⁸ Pa, whereas that during O₂ baking was 6 × 10⁻⁶–3 × 10⁻⁷ Pa. These results indicate that O₂ reacts with carbon on the Pd surface, producing CO₂. In the case of UHV baking, the total H₂O, CO, CH₄, and H₂ pressures 0.1 h after baking off were 5.1 × 10⁻⁷, 4.8 × 10⁻⁹, 1.2 × 10⁻⁷, and 3.2 × 10⁻⁸ Pa, respectively. On the other hand, in the case of the O₂ baking, the H₂O and CH₄ pressures were 1.3 × 10⁻⁷ and 4.0 × 10⁻⁸ Pa, respectively, whereas the CO and H₂ pressures were negligible. The total pressures 6 h after the end of UHV or O₂ baking were about 4 × 10⁻⁸ and 2 × 10⁻⁸ Pa, respectively. These results suggest that O₂ baking removes adsorbed carbon and hydrogen and consequently suppresses the formation of H₂, CO, H₂O, and CH₄. Catalytic chemical reactions on the Pd surface appear to be responsible for the removal of adsorbed carbon and hydrogen.

The pumping speeds of the oxygen-free Pd/Ti-coated chamber were measured for H₂ or CO. The initial pumping speeds for H₂ and CO were 510 and 960 l s⁻¹ after UHV baking and 990 and 1250 l s⁻¹ after O₂ baking. These results demonstrate that O₂ baking is useful for removing the carbon contaminants from oxygen-free Pd/Ti and for recovering pumping speeds of oxygen-free Pd/Ti for H₂ and CO.

The authors are grateful to Takahiro Koyama, Shigeji Sugimoto, Masashi Iguchi (Osaka Vacuum, Ltd.), Mitsuyoshi Sato (Scienta Omicron), Hiromu Nishiguchi, Eriko Kazama (Baroque International Inc.), Yasunori Tanimoto, Akio Toyoshima, and Hirokazu Tanaka (KEK) for their invaluable advice and support. This work was partly supported by a Grant-in-Aid for scientific research (JSPS KAKENHI under Grant No. JP17K05067) and a TIA-Kakehashi grant. The XPS measurements were performed with the approval of the Photon Factory Program Advisory Committee (PF-PAC Nos. 2016T002 and 2017PF-23). Part of the present research was conducted in collaboration with Osaka Vacuum, Ltd. Part of this work was carried out by using the facility of NIMS TEM station. This work was partially supported by the Global Research Center for Environment and Energy based on Nanomaterials Science.