Combination of a getter pump with turbomolecular pumps in UHV applications

Vacuum 55 (1999) 27}37 Combination of a getter pump with turbomolecular pumps in UHV applications Roberto Giannantonio*, Magda Bovisio, Andrea Conte SAES Getters S.p.A., Viale Italia, 77-20020-Lainate (MI), Italy Received for publication 13 February 1999 Abstract This work reports on RGA studies relative to pump-down and ultimate pressure experiments made on a 40 l UHV chamber with a specially designed non-evaporable getter (NEG) pump operated in series with the main pumping group, comprising a 600 l/s hybrid turbo pump backed by a 12 m /h rotary pump. Experiments were also made with a 50 l/s turbomolecular drag pump inserted between the exhaust of the hybrid turbo pump and the intake pipe of the rotary pump. The NEG pump, located on top of the main turbo pump, was realized using "ve NEG elements arranged around the walls of a cylindrical cartridge inserted inside a stainless-steel nipple having an inner diameter of 150 mm. A single NEG element, consisting in a E 5 mm, 720 mm long rod, shaped as a sinusoid, was prepared by sintering ca. 20 g of pure titanium on a E 0.5 mm nichrome wire. The NEG pump allowed for a tenfold reduction of H O partial pressure and a reduction of the ultimate pressure from 1 10 to 1 10 mbar. In particular, because of the sorption characteristics for hydrogen of NEG pumps, the H partial pressure could be signi"cantly reduced by e!ectively dealing with the hydrogen backstreaming, typical of turbo pumps. 1999 Elsevier Science Ltd. All rights reserved. Keywords: Non-evaporable getter pump; Turbomolecular drag pump; Rotary pump; UHV chamber 1. Introduction Turbomolecular pumps (TMPs) are increasingly being used as ultrahigh vacuum (UHV) pumping systems mainly because of their high reliability, low system and operating cost, ease of operation, almost oil-free operating conditions, constant pumping speed, etc [1]. One of the major limitations of TMPs is due to their rather low compression ratio for light gases, in particular for H [2}5]. Hydrogen being the main constituent of the residual atmosphere in stainless-steel vacuum chambers operating under UHV conditions, the lowest ultimate pressure here attainable is mainly determined by the hydrogen partial pressure. Generally speaking, the partial pressures of the residual gases are determined not only by surface outgassing from the walls of the vacuum chamber but also by the backstreaming from the pump itself (and, for trapped systems, also from the trap) [6]. Back#ow of H from turbo pumps is due both to a rela- *Corresponding author. E-mail address: roberto}giannantonio@saes-group.com (R. Giannantonio) tively high exhaust-to-inlet transmission probability [2] and to surface desorption from the inner surfaces of the pump, the latter being di$cult to reduce as the baking temperature of a TMP never exceeds ca. 1003C [7]. The reduction of the ultimate partial pressure of H can be achieved by means of: (i) improvement of the compression ratio of the TMP for H [5}7], (ii) reduction of the partial pressure of H in the fore vacuum line (e.g. using baking pumps with higher pumping speed [8}9], through ballasting [10] or using forepump oils having low H solubility [5]) or (iii) combination of TMPs with other UHV pumping systems (namely, non-evaporable getter (NEG) pumps or titanium sublimation pumps [10]). Aiming at evaluating the combination of a TMP with a NEG pump both for industrial and research applications, an in-line NEG pump was specially designed to operate in series with the turbomolecular pump. In fact, assembling the NEG pump between the TMP and the vacuum chamber, i.e. as a trap, allows to increase both the net pumping speed in the chamber and to cope with hydrogen backstreaming from the TMP, as suggested by a previous work [11]. 0042-207X/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2-2 0 7 X ( 9 9 ) 0 0 1 2 0-7

28 R. Giannantonio et al. / Vacuum 55 (1999) 27}37 2. Experimental setup and methods To evaluate the performances of the TMP/NEG pumping system, the test apparatus was assembled schematically as shown in Fig. 1. It consisted of a stainlesssteel cylindrical chamber having a volume of ca. 40 l and an inner surface area of ca. 6000 cm. A blank #ange was connected at one side of the cylinder, where a ionization gauge (Leybold IE-414), an extractor gauge (Leybold IE-514) and a quadrupole mass spectrometer equipped with 903 SEM (Balzers QMA-125/QMG-421C) were mounted. The other side of the chamber was connected through an 8-in Con#at #ange to the NEG pump, isolated from the main TMP (TMP M) by means of the gate valve 9. The main turbomolecular pump was a leybold turbovac 600 C with grease-lubricated ceramic ball bearings. Nominal pumping speed for N and H were 620 and 590 l/s, respectively. Compression ratios for N and H were about 10 and 1.1 10, respectively. The exhaust of the main pump could be connected either to a 12m /h rotary vane pump (Edwards RV12) or to a 50 l/s Pfei!er TMU-064 turbomolecular drag pump (TMP B), featuring pumping speeds for N and H of 53 and 31 l/s, respectively and compression ratios for the same gases of about 10 and 4 10, respectively. Back-di!usion of oils from the rotary pump was prevented using the catalytic trap 14 (Pfei!er URB-040). All of the cited instrumentation was interfaced to a computer to achieve complete automation (i.e. good repeatability) of measurement cycles. RGA, pump-down times and ultimate pressure meassurements were performed with: (i) TMP M directly baked by the rotary pump, (ii) TMP M baked by TMP B, (iii) TMP M in combination with the NEG pump, and (iv) TMP M baked by TMP B in combination with the NEG pump. The following experimental procedure was used: (i) venting of the vacuum chamber in ambient air for 15 min, (ii) pumpdown of the system for 15 min (iii) bake-out at 2003C (mass spectrometer at 1203C, gate valve at 1003C) for 20 h; if present, the getter pump was activated (heating from room temperature to 5503C in 1 h and heating at 5503C for 1 h) after 18 h of baking, (iv) cooling of the system to 503C (the chamber was cooled to 503C instead of room temperature to enhance the hydrogen desorption rate thus highlighting the role of the NEG pump), (v) pumpdown of the system for 20 h and measurement of the pumpdown time to base pressure, (vi) ultimate pressure and RGA measurements, and (vii) isolation of TMP M from the system (valve 9 was closed) and rate-of-rise test. All of the experiments were executed twice to check for repeatability. 3. In-line NEG pump Fig. 1. Layout of the experimental apparatus. 1"Bayard}Alpert gauge; 2"extractor gauge; 3"quadrupole mass spectrometer; 4"vacuum chamber; 5"In-line NEG pump; 6, 10}12, 15, 17}19, 21, 22"valves; 7"micrometric leak valve; 8, 20"Pirani gauges; 9"gate valve; 13"TMP M; 14"catalytic trap; 16"TMP B; 23"rotary vane pump. A new NEG pump was suitably developed to work in combination with a turbomolecular pump. The NEG pump structure here (and elsewhere ) described was optimized in order to operate the pump both as a stand-alone pump, exploiting its pumping action on the vacuum chamber, and as a trap for the H back#owing from turbomolecular pumps. In particular, being the Ho coef- "cient [12] of a TMP strongly dependent on the pump inlet conductance, the conductance drop due to the NEG pump, located on top of the TMP inlet #ange, was properly minimized. An assembly of the pump is shown in Fig. 2. A single NEG element was prepared by sintering ca. 20 g of pure titanium on a E 0.5 mm nichrome wire. Each NEG element consists in a E 5 mm, 720 mm long rod, shaped as a sinusoid. A total of "ve NEG elements, connected in series, were arranged around the walls of a cylindrical cartridge, having an outer diameter of 143 mm. The cartridge can easily be inserted inside an AISI 304L stainless steel nipple having an inner diameter of 150 mm and Filed Italian patent application MI97A 001420.

R. Giannantonio et al. / Vacuum 55 (1999) 27}37 29 Fig. 2. Assemblies of the in-line NEG pump. A, detail of the NEG pump cartridge; B, detail of the pump housing; C, detail of the electrical connections; D, assembled pump. mounting at each open side a 150 mm Con#at #ange. Two ends of the NEG elements chain are "xed to a molybdenum plug located on the cartridge base structure. While inserting the cartridge inside the nipple, the plug "ts inside a socket placed at the bottom of the nipple, thus establishing electrical contact between the NEG elements and the external power feedthroughs. The activation of the NEG pump was usually carried out by means of a DC power supply. An activation temperature of ca. 5503C on the NEG rods could easily be attained by operating the power supply at 40V/11 A. In Fig. 3, the sorption characteristics of the pump for H and CO, measured according to the standard conductance method, are shown. 4. Experimental results and discussion In Fig. 4 typical total pressure pro"les, measured with the extractor gauge during pumpdown of the system, are ASTM procedures F798-82.

30 R. Giannantonio et al. / Vacuum 55 (1999) 27}37 Fig. 3. H and CO gettering rates vs. sorbed quantities at 25 and 2003C of the tested in-line NEG pump. Activation conditions: 5503C/2 h. Sorption pressure: 4.0 10 mbar. reported. The steep rise of pressure after 18 h of bake-out is due to the gas, which is almost H, released during the NEG pump activation step. The sudden change of slope of the three pressure vs. time curves corresponds to the initial cooling of the chamber, i.e. to a reduced outgassing throughout. Fig. 4, clearly shows that with the NEG pump operating in combination with TMP M, a lower ultimate pressure (5.3 10 mbar) could be attained compared with the pressure reached with TMP M (1.9 10 mbar) and TMP M#TMP B (1.5 10 mbar). Pumpdown times to a reference pressure of 1.9 10 mbar could also be greatly reduced when using the NEG pump (23 h) with respect to those related to the TMP M (40 h) and TMP M#TMP B (26 h) tests. The slight di!erence between the ultimate pressure attained with TMP M and the one obtained by backing TMP M with TMP B is due to a higher compression ratio for H of the tandem TMP con"guration. In Fig. 5 the evolution of the mass spectrometer ion current i due to water vapour is shown. As expected, the ultimate partial pressure of water in the chamber pumped by TMP M baked by TMP B is essentially identical to the pressure reached when the exhaust of TMP M is directly connected to the rotary vane pump, due to a rather high maximum compression ratio of TMP M for H O [8, 9]. The use of the NEG pump in combination with TMP M results in a net reduction of the partial pressure of water by almost an order of magnitude. Similar results were observed for N /CO. Ion current i pro"les, due to hydrogen, are represented in Fig. 6. As discussed below, the insertion of TMP B in series with TMP M reduces the #ux of hydrogen backstreaming from the exhaust of TMP M to the vacuum chamber, thus halving the "nal hydrogen partial pressure. A further reduction of the hydrogen partial pressure could be obtained by using the NEG pump. After 43.5 h from the start of the experiments, valve 9 was suddenly shut allowing for pressure rate-of-rise measurements. Some of the results drawn from these tests are shown in Fig. 7, where the variation of the mass spectrometer ion current i due to hydrogen is plotted. A "rst observation will drive the following discussion. Looking at curve C in Fig. 7, one can clearly see that, after the sudden shutting of valve 9, i drops from a steady-state value of ca. 5 10 A down to ca. 2 10 A. In steady-state conditions, where the usual relation P "F /S holds (F is the net #ow of hydrogen into the chamber and S is the total pumping speed in the chamber), this abrupt variation of the hydrogen partial pressure can only be due to a sudden reduction of the hydrogen #ow. As valve 9 separates the line comprising the chamber and the NEG pump from the turbomolecular pump(s), this extra #ow of hydrogen can only come from the turbomolecular pump(s) itself. A part of the total amount of hydrogen in the chamber is therefore due to the backstreaming of H from the turbomolecular pump(s). In these conditions, the turbomolecu- lar pump acts as a source of hydrogen instead of acting as

R. Giannantonio et al. / Vacuum 55 (1999) 27}37 31 Fig. 4. Total pressure vs. time pro"les for different system con"gurations. a sink. To quantify the in#uence of H backstreaming on the H ultimate partial pressure attainable in the vacuum chamber, material balances for hydrogen can be made around di!erent regions of the experimental apparatus, indicated by dashed closed lines as R1}R4 in Fig. 8 (system with the NEG pump fully passivated, thus represented without the NEG pump) and Fig. 9 (system with the NEG pump fully activated, thus represented with the NEG pump). The equations originating from these mass balances are collected in Table 1. Pressures, #uxes and pumping speeds appearing in Table 1 are referred to hydrogen. All of the pumping speeds S appearing in Table 1 are e+ective pumping speeds in the vacuum chamber, a!ected by the "nite conductance of the line separating TMP M from the vacuum chamber itself, this line comprising also the NEG pump. Hydrogen partial pressures are expressed as "ki, where k is a constant containing both the calibration factor for the extractor gauge and the calibration factor for the mass spectrometer. Aiming at determining only relative estimates for pumping speeds and #uxes, the explicit value for k will not be reported here. It is however worth noticing that

32 R. Giannantonio et al. / Vacuum 55 (1999) 27}37 Fig. 5. Mass spectrometer ion current i (H O) vs. time pro"les for di!erent system con"gurations. the conversion factor k should be considered as a mean value, averaged over the whole set of experiments, its dispersion being $10%. F "F #F is the (steadystate) throughput due to outgassing (F ) and to a leak found to be present in the vacuum line (F ). S and S* are the apparent pumping speeds of TMP M, when the pump is baked by the rotary pump and by TMP B (with or without the NEG pump), respectively. S is the real pumping speed of TMP M. The relationships de"ning S, S and S* (together with the corresponding backstreaming #uxes F and F* ) are represented by Eqs. (10) and (11). Eqs. (10) and (11) are related to the hydrogen mass balance in region R4, when TMP M is directly baked by the rotary pump and when TMP M is baked by TMP B, respectively (these equations apply also when the NEG pump is under operation i.e. also for system con"gurations C and D). Eqs. (10) and (11) express the well-known relation used by Kruger and Shapiro to model their experimental single-rotor turbomachine [2]. The mass-balance equation used by Kruger and Shapiro

R. Giannantonio et al. / Vacuum 55 (1999) 27}37 33 Fig. 6. Mass spectrometer ion current i (H ) vs. time pro"les for di!erent system con"gurations. (herein after called KS equation) reads, using the same symbols adopted by the two researchers, as: = "!, where is the pressure at the highvacuum side of the compressor, is the pressure at the exhaust, and are transmission probabilities and = is the Ho coe$cient of the pump. As ="S, where S is the apparent pumping speed of the pump and S is the pumping speed of an aperture having the same size of the free entrance of the pump (i.e. S is the maximum theoretical pumping speed of the pump, acting as a pure conductance), the KS equation can be rewritten as S "S!S. Comparing this equation with Eq. (10), we can see that S &S, S &S and F &S (i.e. S* &S and F* &S, if Eq. (11) is considered), thus highlighting the dependence of F (and F* )on. Expressions similar to the KS equation and to Eqs. (10, 11) can also be found elsewhere in the literature [13]. Under system con"guration C (and D), the NEG pump reduces the hydrogen partial pressure in the

34 R. Giannantonio et al. / Vacuum 55 (1999) 27}37 Fig. 7. Mass spectrometer ion current i (H ) vs. time pro"les during rate-of-rise tests for di!erent system con"gurations. chamber, with respect to the value reached under system con"guration A (and B), due to the added pumping speed S. As the compression ratio K (or K* )" / and the pumping speed S (or S* ) of TMP M are strictly correlated, the H partial pressure "K at the exhaust of TMP M for system con"guration C (and D) should be lower than the pressure found for system con"guration A (and B). Nevertheless, the partial pressure of hydrogen in equilibrium with hydrogen dissolved inside the rotary pump oil is almost constant during the whole duration of the experiments, the oil being saturated with hydrogen (the rotary pump oil can thus be regarded as a reservoir of hydrogen). This means that the partial pressure of hydrogen at the exhaust of TMP M (and at the exhaust of TMP B) is almost constant throughout the experiments. Therefore, a #ux F (and a #ux F* ), due to outgassing of the rotary pump oil, logically represented as a #ux entering into the system, is introduced in the equations so that the hydrogen partial pressure at the exhaust of TMP M is the same both in system

R. Giannantonio et al. / Vacuum 55 (1999) 27}37 35 Fig. 8. Layout of the system con"gurations without the NEG pump (NEG pump fully passivated): TMP M (A) and TMP M#TMP B (B). Mass balances around system regions R1, R2 and R4, surrounded by dashed closed lines, give rise to the equations in Table 1. Fig. 9. Layout of the system con"gurations with the NEG pump (NEG pump fully activated): TMP M#NEG (A) and TMP M#TMP B #NEG (B). Mass balances around system regions R1}R4, surrounded by dashed closed lines, give rise to the equations in Table 1.

36 R. Giannantonio et al. / Vacuum 55 (1999) 27}37 Table 1 Equations derived from hydrogen mass-balances around regions R1}R4 de"ned in Figs. 8 and 9. Eqs. (12)}(19) are the solution of the set of Eqs. (1)}(9). F "F #F (F "H #ow due to outgassing of vacuum chamber, manifolds, etc.; F "H #ow due to leaks); F, F* "backstreaming H #ows, recycled by TMP M; F, F* "backstreaming H #ows, due to outgassing of the rotary pump oil (non recycled by TMP M); S, S* "apparent, e!ective pumping speeds of TMP M; S "real, e!ective pumping speed of TMP M; S "e!ective pumping speed of the NEG pump. Symbols containing the asterisks are related to the series combination TMP M#TMP B. Symbols without the asterisks are related to TMP M directly baked by the rotary pump System con"guration Mass-balance Mass-balance equations Main results region Valve 9 open (A) TMP M (R2) P #F )/S k2 10 (1) S 10 (12) (R1) P "F k2 10 (2) S /S* 4.5 (13) (R4) P S "P S #F (10) S /S 4.5 (14) (B) TMP M#TMP B (R2) P #F* )/S k9 10 (3) S* 2.2 (15) (R1) P "F /S* k9 10 (4) F 0.7 (16) (R4) P S "P S* #F* (11) F* 0.2 (17) (C) TMP M#NEG (R2) P #F #F )/(S #S ) k5 10 (5) F 1.2 (18) (R1) P #F )/(S #S ) k5 10 (6) F* 0 (19) (D) TMP M#TMP B#NEG (R2) P #F* #F* )/(S #S ) k2 10 (7) (R1) P #F* )/(S* #S ) k2 10 (8) Valve 9 closed (C) TMP M#NEG and (D) TMP M#TMP B#NEG (R3) P "F k2 10 (9) con"gurations A and C (and the same both in system con"gurations B and D). Of course, both hydrogen streams F (or F* ) and F (or F* ) pass through the exhaust of TMP M towards the vacuum chamber but, in the present model, the former should be considered as being e!ectively recycled by the turbopump while the latter should be regarded as just crossing TMP M to be trapped by the NEG pump. The solution of the set of Eqs. (1)}(9) gives the results (12)}(19), also summarized in Table 1. Even if the hydrogen partial pressure is greatly reduced by chaining TMP M with TMP B, the apparent pumping speed of TMP M being almost equal to its real pumping speed, a net reduction in hydrogen concentration can be attained only using the NEG pump, exploiting a pumping speed S 4.5S. However, despite its impact on the hydrogen ultimate pressure, the NEG pump does not act as a perfect trap. If so, the ultimate pressure attained under system con"guration C should be very close to the one reached under system con"guration D. On the contrary, a rather high fraction of the total hydrogen back#ow, represented by both ratios F and F, pass through the NEG pump, thus contributing to the observed value of the hydrogen ultimate pressure. The NEG pump structure described in the present work was chosen in order to maximize both the net pumping speed of the NEG pump itself and the conductance of the line separating TMP M from the vacuum chamber (i.e. the e!ective pumping speed of TMP M). Due to the cylindrical symmetry of the NEG pump, obtained by arranging the NEG elements around the inner walls of the pump cartridge, the performances of TMP M were maintained as high as possible but the trapping e.ciency of the NEG pump could not be optimized. It is worth remarking that an independent estimate for the ratio S* can be obtained by means of the KS equation. In fact, for any given #ux F of hydrogen passing through TMP M and its baking pump, we have, in steady-state conditions, that F"S P "S P, where S and S are the (e+ective and apparent) pumping speeds for H of TMP M and the baking pump, respectively. Therefore, the compression ratio of TMP M, de"ned as K " / can also be written as K "S. Generally speaking, S depends on K in accordance with the KS equation, written as = /=3 "S /S3 "(K3!K )/ (K3!1)"(K3!S )/(K3!1) where =, =3, S3 and K3 are the Ho coe$cient, the maximum Ho coe$cient, the maximum pumping speed and the maximum compression ratio of TMP M, respectively [7]. As in the present case K3 <1, we have that 1 1/S3 #1/(K3 S ), so that the ratio S* can be calculated by simply setting in the above expression the nominal pumping speeds of the rotary pump (for S ) and of TMP B (for S* ). A ratio S* 2.2, in accordance with the value (15), is obtained, provided that a value S 0.5 l/s is used for the rotary pump. The oil of the rotary pump being saturated with hydrogen, this "gure,

R. Giannantonio et al. / Vacuum 55 (1999) 27}37 37 compared with a nominal pumping speed of 3.3 l/s, seems to be reliable. elements arranged horizontally, thus intercepting backstreaming hydrogen, are now running. 5. Conclusions The combination of a specially designed in-line NEG pump with a TMP was evaluated by means of residual gas analysis, ultimate pressure and pump-down experiments. The use of the NEG pump allows for a reduction of both H partial pressure and total ultimate pressure greater than that obtainable by the insertion of a second turbopump on the exhaust line of the primary turbo pump. The use of a NEG pump, in combination with a turbomolecular pump, seems to be a suitable solution to cope with H back-streaming, typical of all TMPbased pumping systems (in particular, when TMPs not equipped with molecular-drag stages are used). When great amounts of hydrogen are present in the vacuum chamber, NEG pump structures having higher trapping e$ciencies than that observed for the pump model discussed in this work, should probably be preferred. Further experimental work on a NEG pump having NEG References [1] La!erty JM, editor. Foundations of vacuum science and technology. New York: Wiley, 1998. Chap 9. [2] Kruger CH, Shapiro AH. Trans Natl Vac Symp 1960;7:6. [3] Bernhardt KH, J Vac Sci Technol A 1983;1:136. [4] Becker W. Vacuum 1966; 16:625. [5] Henning J. Vacuum 1971; 21:523. [6] Santeler DJ. J Vac Sci Technol 1971;8:299. [7] O'Hanlon JF. A user's guide to vacuum technology. New York: Wiley, 1980. Chap 7 and 11. [8] Ishimaru H. J Vac Sci Technol A 1989;7:2439. [9] Cho B, Lee S, Chung S. J Vac Sci Technol A 1995;13:2228. [10] Bernardini M, Bradaschia C, Pan HB, Pasqualetti A, Torelli G, Zhang Z. Problem of H back-stream from turbo pumping systems. Paper Presented at the XIV3 National Congress on Vacuum Science and Technology, Vicenza (Italy), 5}8 May, 1998. [11] Pozzo A, Bo$to C, Mazza F, Vacuum 1996;47:783. [12] Ho TL. Physics 1932;2:386. [13] Roth A. Vacuum technology. 2nd ed. Amsterdam: North-Holland, 1982. Chap 7.