- Email: [email protected]
200 keV cold field emission source using carbon cone nanotip: Application to scanning transmission electron microscopy
Ultramicroscopy 182 (2017) 303–307
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
200 keV cold field emission source using carbon cone nanotip: Application to scanning transmission electron microscopy Shuichi Mamishin a, Yudai Kubo a, Robin Cours b, Marc Monthioux b, Florent Houdellier b,∗ a b
Hitachi High-Technologies Corporation, 882, Ichige, Hitachinaka, Ibaraki 312-8504, Japan CEMES-CNRS, 29 Rue Jeanne Marvig, 31055 Toulouse France
a r t i c l e
i n f o
Article history: Received 19 April 2017 Revised 3 July 2017 Accepted 28 July 2017 Available online 9 August 2017
a b s t r a c t We report the use of a pyrolytic carbon cone nanotip as field emission cathode inside a modern 200 kV dedicated scanning transmission electron microscope. We show an unprecedented improvement in the probe current stability while maintaining all the fundamental properties of a cold field emission source such as a small angular current density together with a high brightness. We have also studied the influence of the low extraction voltage, as enabled by the nanosized apex of the cones, on the electron optics properties of the source that prevent the formation of a virtual beam cross-over of the gun. We have addressed this resolution-limiting issue by coming up with a new electron optical source design. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Since the pioneering work of Crewe [1] in 1968, cold field emission (CFE) sources have opened up the possibilities offered by transmission electron microscope (TEM) and scanning TEM (STEM). CFE sources have the unique ability to combine ultimate probe sizes, high brightness (in the range of 104 A.cm−2 .Sr−1 .V−1 ) and low energy spread electron beam (in the range of 0.3 eV) [2]. This unique combination is sought by specific techniques such as electron energy loss spectroscopy (EELS) combined with STEM [3] but also those that requires high spatial coherence such as electron interferometry [4]. Such high quality electron beams can be generated by CFE sources thanks to their small virtual source size (in the range of 5 nm) and their low angular current density (in the range of 10 μA.Sr−1 ) [2]. But on the other hand, CFE performances are highly sensitive to the quality of the probe-forming optical system, especially through the spherical aberration, which can easily kill their brightness. This weakness is well-known since the work by Crewe, and that is why the source anode shapes are carefully chosen following the well-known Butler solutions [5,6], but this problem needs to be carefully addressed also for the objective lens of the microscope. Recently there has been a resurgence of interest for CFE sources in the S/TEM market thanks to the advent of spherical aberration corrector which allowed easily generating subAngström electron probes and thus making an effective use of the excellent properties provided by CFE sources [7–10].
∗
Corresponding author. E-mail address: fl[email protected] (F. Houdellier).
http://dx.doi.org/10.1016/j.ultramic.2017.07.018 0304-3991/© 2017 Elsevier B.V. All rights reserved.
However another well-known weakness stems from the long and short-term stability of the emitted current [1,2]. These properties are strongly related to the vacuum level of the source [11], the surface quality of the extraction anode [12] and the cathode material [13]. For the latter, major commercial instruments use a monocrystalline tungsten (W) tip oriented either in [310] or [111] zone axis [14]. By using such a cathode along with state of the art ultra high vacuum (UHV) technology, the life cycle of the emission and probe currents can be typically divided into three distinct regions [15]. After a thermal cleaning (usually called flash cleaning) the total and probe currents decrease rapidly by one to two orders of magnitude, which defines the first region; then the rapid downward current drift stabilizes for a period of time depending on the vacuum level, which makes the second region; and then the third region is defined by an increase in both the amplitude of current fluctuations and the total and probe currents. Usually, the tip needs to be flash-cleaned before entering in the third region to avoid any catastrophic arc destruction of the cathode. The figure of merit of a CFE source regarding stability can then be easily determined by the time range associated to region 1 and 2, the amount of current lost in region 1, or more generally by the time necessary between two flash-cleaning. Several manufacturers offer W cathode CFE sources each with their specific vacuum technology, but all the technologies exhibit similar current behaviours [16]. Many studies have been performed trying to change the cathode material to improve the stability property of CFE sources [17– 19], but none of these attempts have been transferred to commercial applications yet. Each of these solutions has their own drawbacks that prevent their full integration in a workable technology. Among them, the carbon nanotube (CNT) cathode studied by
304
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
de Jonge et al. [20] on the one hand exhibits very good stability properties of the emitted current but on the other hand remains strongly limited by the nanotube strength and vibrations. In 2008, Houdellier et al. [21] reported the use of a pyrolytic carbon cone nanotip (CcnT) as a cathode in a high voltage FE-TEM source. Since the apex size of a CcnT is that of a multiwall CNT diameter, the same benefits as for de Jonge’s cathode were observed regarding the emission current stability without the drawbacks inherent to the mere use of a plain CNT, thanks to the cone structure. The excellent properties described in [21] obtained on a 200 kV Hitachi HF-20 0 0 have motivated a collaboration between CEMES-CNRS and the Naka factory of Hitachi High-Technologies (HHT). The purpose of this collaboration reported in this paper was to fully determine CcnT properties used as FE cathode and to try meeting quality criteria to anticipate a future application in STEM. Quality criteria for these CFE sources, in addition to usual parameters such as reduced brightness, energy spread and exit work function already provided in [21–23], must include angular current density, virtual source size, emission and probe current stability over 9 h, emission and probe current noise, a reproducible and technologically accessible preparation method, and finally the range of demagnification values of the gun cross-over necessary to perform high resolution STEM. 2. Cathode preparation The original preparation method, based on a focused ion beam (FIB) procedure, used a micro-tweezer device to grab the cone and approach it to a tungsten tip base suitably prepared. An important and positive side-effect of the tweezers was to protect the CcnT apex against ion irradiation during the ion welding of the cone [21]. Although this method was efficient, it remained difficult to handle and almost impossible to automate. We have developed a new mounting procedure, again based on a FIB, where an omniprobe tip is used to perform the lift off, which is more common on standard FIB machine, and easier to automate for an industrial perspective. The method is described in Fig. 1, where the full sequence is reported. After choosing carefully the suitable CcnT (Fig. 1A), the omniprobe is brought to contact with the rough part surface of the CcnT and then is welded to it thanks to an ion beam-induced deposition of platinum (Pt) (Fig. 1B). The W tip base is brought in the field of the SEM image (Fig. 1C) and positioned to protect the carbon cone against ion irradiation in the ion imaging mode (Fig. 1D). Using the Ga beam the CcnT base is cut while the apex remains protected, and then is brought in contact with and welded to the tip base using the ion beam-induced deposition of Pt before the omniprobe is finally removed (Fig. 1E and F). A part of the omniprobe remains welded on the edge of the rough surface of the CcnT base (see Fig. 1F) but does not induce any disturbance to the emission. Furthermore, on Fig. 1C and D we can notice the presence of a used CcnT formerly mounted on the W tip base coming from a previous emission experiment. Prior to the welding of the fresh CcnT, the used one had been removed by cutting it using the Ga beam. It is worth noting that this is an additional positive point of the CcnT cathodes as the same W tip base can be reused many times. 3. Electron emission properties To determine the current angular density, the CcnT tip was installed inside a dedicated ultra high vacuum (UHV) bench (in the low 10−9 Pa range) equipped with an extracting anode, a fluorescent screen below the anode and a faraday cup located in the centre of the screen. A steady value of 13 μA.Sr−1 was measured which is in the range of common values encountered for W cathode CFE source. This small value makes it possible to conserve a good beam
Fig. 1. New FIB-based preparation method using omniprobe tip to grab the carbon cone and approach it to the W tip base. The final welding step is the same as the one previously reported in [21].
energy spread measured in the range of 0.3 eV [21] minimising the Boersch effect in the apex vicinity. To fully characterize all the FE properties, the CcnT tip was fitted inside a dedicated STEM Hitachi HD-2700 equipped with a faraday cup in the sample position to measure the probe current, and with Bright Field (BF), Dark Field (DF) and High Angle Annular DF (HAADF) detectors to perform STEM imaging (see Fig. 3A). The high voltage configuration of the HD-2700 source, described in Fig. 3B, is comparable to the HF-20 0 0 source configuration used in the previous studies and already reported in [21]. The tip is set at V0 = −200 kV and face the extracting anode set at a voltage V1 relatively to the tip. The distance between the tip and the anode is adjusted in a range of about 4–7 mm. The first anode of the accelerating tube, called focusing anode or gun lens, is set at a potential V2 relatively to the tip. The ratio R = V2 / V1 between the focusing anode and the extracting anode can be adjusted between 2 and 13. The ratio is used to set the position of the full gun cross-over which may be either real (which is the condition selected for TEM applications to maximise the beam intensity) or virtual (in order to maximise the first illumination lens demagnification for STEM applications [24]). Fig. 2 shows the stability data of a 2 nA probe at 200 kV measured on the sample stage over 9 h without any flash cleaning of the tip during the measurement. Fig. 2A corresponds to the CcnT based CFE source with a total emission current set at 5.7 μA, an extraction voltage V1 = 1.3 kV and a ratio of 13, while Fig. 2B refers to the standard W [310] based CFE source with an emission current of 10 μA for an extraction voltage of 2.9 kV and a ratio of 4.5. The vacuum system of the source was the same for the two experiments, fitted with two ion pumps and one non-evaporable getter (NEG) pump reaching a vacuum level below 2.10−9 Pa. As Fig. 2 clearly demonstrates, the long-term stability is strongly improved by using our CcnT as FE cathode. Indeed, using
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
305
Fig. 2. Nine hours stability measurements, performed on a Hitachi HD-2700 dedicated STEM, of a 2 nA probe at 200 kV. Comparison between CcnT CFE source (A) and W [310] CFE source (B). Both measurements were started after a flash-cleaning of the tips.
state of the art W [310] CFE source, the probe current decreases by a factor of 50% after 4 h and by almost 90% after 8 h while in the same emission conditions the CcnT CFE source shows no decrease of the probe current over the full measurement time range i.e. 9 h. We can notice the presence of small peaks, which have no consequence on STEM or TEM imaging. As already reported in [12,21,22], the root mean square value of the probe current noise is around 1%, which is in the order of magnitude of standard CFEG technology. 4. Optical properties As mentioned previously, the key feature of the HD-2700 high voltage gun, described in Fig. 3B, is the possibility to adjust the full gun cross-over position thanks to the ratio value. Compare to the usual W [310] tips the CcnT apex is smaller [21]. As a consequence the extraction voltage required for a specific probe current value will be smaller. This is the case in Fig. 2 where 2 nA are obtained for 1.3 kV of extraction voltage using the CcnT cathode and for 2.9 kV using the W [310] cathode. This difference has an important consequence in the electron optical configuration because it changes the strength of the electrostatic gun lens composed by the tip, the extraction anode and the focusing anode. This effect was simulated and reported in Fig. 4 showing the electrons trajectories calculated for 5 ratio values using the same emission conditions as those described in Fig. 2. We can see that, whatever the ratio value, the 200 kV gun cross-over remains real when the CcnT is used as a cathode. The same calculations for a W [310] cathode show that the gun crossover position switches from a virtual position to a real one for a ratio in between 6 and 8. This effect is clearly demonstrates for the ratio 2, which corresponds to the condition with the minimum strength of the electrostatic gun lens. Even for this condition, the cross-over is real and located below the accelerator with the CcnT cathode while it is virtual for the standard W [310] situation.
This unwanted effect, related to the small apex size of the CcnT, has a direct consequence on the STEM properties. Indeed, for STEM imaging, the virtual position is the most desirable in order to maximize the demagnification power of the first illumination lens which is merely related to the distance between the gun cross-over and this lens. Furthermore the aberration coefficients of the electrostatic gun lens are different in real cross-over and virtual cross-over conditions. In real cross-over condition the spherical and chromatic aberration coefficients of this electrostatic lens are respectively Cs ≈ 60 mm and Cc ≈ 12 mm while in virtual crossover condition they are respectively Cs ≈ 35 mm and Cc ≈ 10 mm. As a consequence, the first spot size formed by the gun lens, starting from an usual virtual source size of CFE tip/extraction anode source assembly in the range of 2–5 nm (see Fig. 5B), will be in the range of 500 nm in real cross-over condition and about half of it in the virtual cross-over condition. That’s why the first illumination lenses, situated after the gun, must demagnify strongly this first spot size to achieve a sub-atomic electron probe on the sample plane. To maximise this demagnification, the distance between the spot size formed by the gun and the first illumination lens must be maximised. To increase this distance using a CcnT source, we could use a ratio 13 condition. It is important to stress that this condition is not optimum for an EELS application point of view as the cross-over is located in a zone where the electrons have a low energy, which is favourable to an energy spreading due to the Boersch effect. Furthermore, as shown in Fig. 3D, even under this ratio condition the demagnification of the gun cross-over by the first illumination lens (called C1) never reaches values otherwise usually obtained by a W [310] cathode for all C1 lens current values. For a maximum C1 lens current of 3 A, the demagnification is 400 with a CcnT source of ratio 13 and almost 700 for a W [310] source of ratio 4.5. The BF-STEM image of gold particles performed with the CcnT CFE source, set to a ratio 13, is reported in Fig. 3C. Although the
306
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
Fig. 4. Electron trajectories inside the HD-2700 source calculated with Simion 8.1 [25] software using the same conditions (meshing, boundary conditions, etc.) as those reported in [26]. The left trajectories correspond to one off-axis electron extract from the CcnT CFE source set to 5.7 μA of emission current at 1.3 kV and 5 ratio values. The initial conditions of the electron remain constants for each ratio. The right trajectories come from one off-axis electron extract from a W [310] cathode set to 10 μA of emission current at 2.9 kV and 5 ratio values. To improve visibility the trajectories are enlarged in the x direction by a factor 500 and superimposed to a sketch of the source.
Fig. 3. A: Configuration of the dedicated STEM Hitachi HD-2700. B: Electron optical configuration of the 200 kV cold field emission source showing the cathode tip, the extraction anode, the focusing anode, the accelerating tube, the condensor aperture and the first illumination lens [24]. C: BF-STEM image of gold nanoparticles obtained with the HD-2700 fitted with CcnT CFE source set to a ratio 13. Emission conditions were the same as the one reported in Fig. 2. D: Comparison of condensor 1 lens demagnification power of the gun cross-over obtained with a CcnT and a W [310] CFE source.
overall STEM image has a good quality, it was impossible to reach the desired atomic resolution due to the lack of demagnification of the gun spot size obtained in real cross-over condition. To address this limiting issue, we propose a new gun configuration based on a magnetic gun lens equivalent to the one already proposed for the design of a 1 MeV CFE source [27]. In Fig. 5A we show that, when the tip is precisely placed in the centre of the magnetic field, virtual cross-over conditions of the gun could be found even for a small extraction voltage. Using such an innovative design, we retrieve the optical flexibility necessary to make an effective use of a CcnT-based CFE source allowing both virtual and real cross-over configurations by adjusting the ratio and the magnetic gun lens strength. Furthermore, a positive side effect of such a magnetic gun lens configuration will be to considerably reduce the spherical aberration contribution of the gun to the electron probe, going from Cs ≈35 mm in the original electrostatic design set in virtual cross-over condition to around Cs ≈ 6 mm with such a magnetic gun lens [27,29]. Finally, all the measured major properties of a CcnT CFE source are summarized in the table reported in Fig. 5B and compared to standard W [310] CFE and W/ZrO [100] Schottky FE sources. The reported reduced brightness is taken from [21] and the exit work function from [23]. The tip/extraction anode assembly virtual source size was estimated combining the brightness and the angular current density measured in our experiments. A source radius of 1.6 nm has been extracted which is in the order of magnitude of standard CFE source size values.
Fig. 5. A. One off-axis electron trajectory calculated with Simion 8.1 [25] software inside the HD-2700 with a magnetic gun lens superimposed to the tip region [27]. The source is set at a ratio of 2 and an extraction voltage of 1.3 kV. The virtual cross-over condition can be obtained when the tip is precisely inserted in the centre of the axial magnetic flux density. The magnetic flux density in Simion has been treated as described in [26] for an excitation of 7.9 × 10−2 T obtained with a 1700turn coil, a current of 0.3 A and a 6 mm gap as proposed in [29]. Due to Larmor rotation, the (x,z) and (y,z) projections of the 3D trajectory are reported in red and green respectively. B: Comparison table of the main source properties. All the numbers reported for standard W [310] CFE source and ZrOW [100] SFE source have been gathered from [2] and [28]. To improve visibility the trajectories are enlarged in the x/y direction by a factor 10 0 0 and superimposed to a sketch of the source.
5. Conclusion Thanks to the use of a carbon cone nanotip as a field emission cathode inside a modern 200 kV dedicated FE STEM, we have considerably improved the probe current stability of a CFE source making it a serious competitor to Schottky field emission source in
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
the TEM, STEM and even SEM market. Besides its very good stability, the CcnT CFE source preserves the excellent properties of a standard cold field emission source i.e. high brightness together with low current angular density, small virtual source size and low energy spread. These excellent properties are however strongly limited by the electron optical configuration of the original CFE source, which is responsible for a lack of demagnification power of the illumination system and a stronger contribution of the gun lens spherical aberration. To take full advantage of the properties provided by the CcnT cathode, a dedicated electron optical configuration should be used. We have shown that the use of a magnetic gun lens could be a valuable solution, which is practically feasible in combination with a 200 kV accelerator. Our future work will move to this direction. Acknowledgment These activities are supported by the French National Research Agency under the ‘Investissement d Avenir’ program reference no. ANR-10-EQPX-38- 01. The authors acknowledge the ‘Conseil Regional Midi-Pyrénées’ and the European FEDER for funding within the CPER program. The authors acknowledge the CNRS and HHT for supporting the CNRS-HHT partnership in which this work was conducted. The authors acknowledge the European Union under the Seventh Framework Programme under a contract for an Integrated Infrastructure Initiative Reference 312483-ESTEEM2. The authors acknowledge Dr. Watanabe of HHT for fruitful discussion. References [1] A.V. Crewe, et al., Rev. Sci. Instrum. 39 (1968) 576. http://dx.doi.org/10.1063/1. 1683435. [2] L.W. Swanson and G. A Schwind, Adv. Imaging Electr. Phys., Vol. 159. [3] O.L. Krivanek, et al., Nature 464 (2010) 571–574. http://dx.doi.org/10.1038/ nature08879. [4] A. Tonomura, T. Matsuda, J. Endo, H. Todokoro, T. Komoda, J. Electron Microsc. 28 (1) (1979) 1–11. http://dx.doi.org/10.1093/oxfordjournals.jmicro.a050142. [5] J.W. Butler, 6 Intern. Cong. Electron Microscopy, 191 (Kyoto, 1966). [6] A. Tonomura, Jpn. J. Appl. Phys. 12 (1973) 1065–1069. http://dx.doi.org/10.1143/ JJAP.12.1065.
307
[7] H. Rose, Optik 85 (1990) 19–24. [8] N. Dellby, N.J. Bacon, P. Hrncirik, M.F. Murfitt, G.S. Skone, Z.S. Szilagyi, O.L. Krivanek, Eur. Phys. J. Phys. 54 (2011) 33505. http://dx.doi.org/10.1051/epjap/ 2011100429. [9] Y. Kohno, E. Okunishi, T. Tomita, I. Ishikawa, T. Kaneyama, Y. Ohkura, Y. Kondo, T. Isabell, Microscopy Anal. Nanotechnol. 24 (2010) S9–S13. [10] H. Müller, I. Maßmann, S. Uholemann, P. Hartel, J. Zach, M. Haider, Nucl. Instr. Meth. A645 (2011) 20–27. http://dx.doi.org/10.1016/j.nima.2010.12.091. [11] E. Martin, J.K. Trolan, W.P. Dyke, J. Appl. Phys. 31 (1960) 782–789. http://dx. doi.org/10.1063/1.1735699. [12] L. Knoop, C. Gatel, F. Houdellier, M. Monthioux, A. Masseboeuf, E. Snoeck, M.J. Hÿtch, Appl. Phys. Lett. 106 (2015) 263101. http://dx.doi.org/10.1063/1. 4923245. [13] H. Zhang, J. Tang, J Yuan, Y. Yamauchi, T.T. Suzuki, N. Shinya, K. Nakajima, L. Quin, Nat. Nanotechnol. (2015) 1–8. http://DOI:10.1038/NNANO.2015.276. [14] L.W. Swanson, Surf. Sci. 70 (1978) 165–180. http://dx.doi.org/10.1016/ 0039- 6028(78)90406- 5. [15] K.S. Yeong, J.T.L. Thong, J. Appl. Phys. 99 (2006) 104903. http://dx.doi.org/10. 1063/1.2197267. [16] K. Kasuya, M. Kimura, S. Katagiri, T. Ohshima, Microsc. Microanal. 16 (Suppl 2) (2010). http://dx.doi.org/10.1017/S1431927610056096. [17] V.I. Kleshch, S.T. Purcell, A.N. Obraztsov, Scient. Rep. 6 (2016) 35260. http://dx. doi.org/10.1038/srep35260. [18] A.K. Singh, J. Kumar, J. Appl. Phys. 113 (2013) 053303. http://dx.doi.org/10. 1063/1.4790323. [19] W.K. Lo, G. Parthasarathy, C.W. Lo, D.M. Tanenbaum, H.G. Craighead, M.S. lsaacson, J. Vac. Sci. Technol. B 14 (6) (1996) 3787. http://dx.doi.org/10.1116/1. 588668. [20] N. Jonge, Y. Lamy, K. Schoots, T.H. Oosterkamp, Nature 420 (2002) 393–395. http://dx.doi.org/10.1038/nature01233. [21] F. Houdellier, A. Masseboeuf, M. Monthioux, M.J. Hÿtch, Carbon 50 (2012) 2037–2044. http://dx.doi.org/10.1016/j.carbon.2012.01.023. [22] F Houdellier, L. de Knoop, C. Gatel, A. Masseboeuf, S. Mamishin, Y. Taniguchi, M. Delmas, M. Monthioux, M.J. Hÿtch, E. Snoeck, Ultramicroscopy 151 (2014) 107–115. http://dx.doi.org/10.1016/j.ultramic.2014.11.021. [23] L. Knoop, F. Houdellier, C. Gatel, A. Masseboeuf, M. Monthioux, M. Hÿtch, Micron 63 (2014) 2–8. http://dx.doi.org/10.1016/j.micron.2014.03.005. [24] H. Murakoshi, M. Ichihashi, S. Isakozawa and Y. Sato, USPatent 5134289A [25] D.A. Dahl, Int. J. Mass Spectrom. 200 (2000) 3–25. http://dx.doi.org/10.1016/ S1387-3806(0 0)0 0305-5. [26] Y. Kubo, C. Gatel, E. Snoeck, F. Houdellier, Ultramicroscopy 175 (2017) 67–80. http://dx.doi.org/10.1016/j.ultramic.2017.01.007. [27] T. Kawasaki, J. Endo, S. Isakozawa, M. Tomita and A. Tonomura, USPatent 4945247A [28] L.W. Swanson, G.A. Schwind, Review Of Zr/O Schottky Cathode, in: John Orloff (Ed.), Handbook of Charged Particle Optics, 2nd ed., CRC Press LLC, New York, 2009, p. 665. ISBN 978-1-4200-45543. [29] K. Kasuya, T. Kawasaki, N. Moriya, M. Arai, T. Furutsu, J. Vac. Sci. Technol. B32 (3) (2014) 0318802.
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
200 keV cold field emission source using carbon cone nanotip: Application to scanning transmission electron microscopy Shuichi Mamishin a, Yudai Kubo a, Robin Cours b, Marc Monthioux b, Florent Houdellier b,∗ a b
Hitachi High-Technologies Corporation, 882, Ichige, Hitachinaka, Ibaraki 312-8504, Japan CEMES-CNRS, 29 Rue Jeanne Marvig, 31055 Toulouse France
a r t i c l e
i n f o
Article history: Received 19 April 2017 Revised 3 July 2017 Accepted 28 July 2017 Available online 9 August 2017
a b s t r a c t We report the use of a pyrolytic carbon cone nanotip as field emission cathode inside a modern 200 kV dedicated scanning transmission electron microscope. We show an unprecedented improvement in the probe current stability while maintaining all the fundamental properties of a cold field emission source such as a small angular current density together with a high brightness. We have also studied the influence of the low extraction voltage, as enabled by the nanosized apex of the cones, on the electron optics properties of the source that prevent the formation of a virtual beam cross-over of the gun. We have addressed this resolution-limiting issue by coming up with a new electron optical source design. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Since the pioneering work of Crewe [1] in 1968, cold field emission (CFE) sources have opened up the possibilities offered by transmission electron microscope (TEM) and scanning TEM (STEM). CFE sources have the unique ability to combine ultimate probe sizes, high brightness (in the range of 104 A.cm−2 .Sr−1 .V−1 ) and low energy spread electron beam (in the range of 0.3 eV) [2]. This unique combination is sought by specific techniques such as electron energy loss spectroscopy (EELS) combined with STEM [3] but also those that requires high spatial coherence such as electron interferometry [4]. Such high quality electron beams can be generated by CFE sources thanks to their small virtual source size (in the range of 5 nm) and their low angular current density (in the range of 10 μA.Sr−1 ) [2]. But on the other hand, CFE performances are highly sensitive to the quality of the probe-forming optical system, especially through the spherical aberration, which can easily kill their brightness. This weakness is well-known since the work by Crewe, and that is why the source anode shapes are carefully chosen following the well-known Butler solutions [5,6], but this problem needs to be carefully addressed also for the objective lens of the microscope. Recently there has been a resurgence of interest for CFE sources in the S/TEM market thanks to the advent of spherical aberration corrector which allowed easily generating subAngström electron probes and thus making an effective use of the excellent properties provided by CFE sources [7–10].
∗
Corresponding author. E-mail address: fl[email protected] (F. Houdellier).
http://dx.doi.org/10.1016/j.ultramic.2017.07.018 0304-3991/© 2017 Elsevier B.V. All rights reserved.
However another well-known weakness stems from the long and short-term stability of the emitted current [1,2]. These properties are strongly related to the vacuum level of the source [11], the surface quality of the extraction anode [12] and the cathode material [13]. For the latter, major commercial instruments use a monocrystalline tungsten (W) tip oriented either in [310] or [111] zone axis [14]. By using such a cathode along with state of the art ultra high vacuum (UHV) technology, the life cycle of the emission and probe currents can be typically divided into three distinct regions [15]. After a thermal cleaning (usually called flash cleaning) the total and probe currents decrease rapidly by one to two orders of magnitude, which defines the first region; then the rapid downward current drift stabilizes for a period of time depending on the vacuum level, which makes the second region; and then the third region is defined by an increase in both the amplitude of current fluctuations and the total and probe currents. Usually, the tip needs to be flash-cleaned before entering in the third region to avoid any catastrophic arc destruction of the cathode. The figure of merit of a CFE source regarding stability can then be easily determined by the time range associated to region 1 and 2, the amount of current lost in region 1, or more generally by the time necessary between two flash-cleaning. Several manufacturers offer W cathode CFE sources each with their specific vacuum technology, but all the technologies exhibit similar current behaviours [16]. Many studies have been performed trying to change the cathode material to improve the stability property of CFE sources [17– 19], but none of these attempts have been transferred to commercial applications yet. Each of these solutions has their own drawbacks that prevent their full integration in a workable technology. Among them, the carbon nanotube (CNT) cathode studied by
304
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
de Jonge et al. [20] on the one hand exhibits very good stability properties of the emitted current but on the other hand remains strongly limited by the nanotube strength and vibrations. In 2008, Houdellier et al. [21] reported the use of a pyrolytic carbon cone nanotip (CcnT) as a cathode in a high voltage FE-TEM source. Since the apex size of a CcnT is that of a multiwall CNT diameter, the same benefits as for de Jonge’s cathode were observed regarding the emission current stability without the drawbacks inherent to the mere use of a plain CNT, thanks to the cone structure. The excellent properties described in [21] obtained on a 200 kV Hitachi HF-20 0 0 have motivated a collaboration between CEMES-CNRS and the Naka factory of Hitachi High-Technologies (HHT). The purpose of this collaboration reported in this paper was to fully determine CcnT properties used as FE cathode and to try meeting quality criteria to anticipate a future application in STEM. Quality criteria for these CFE sources, in addition to usual parameters such as reduced brightness, energy spread and exit work function already provided in [21–23], must include angular current density, virtual source size, emission and probe current stability over 9 h, emission and probe current noise, a reproducible and technologically accessible preparation method, and finally the range of demagnification values of the gun cross-over necessary to perform high resolution STEM. 2. Cathode preparation The original preparation method, based on a focused ion beam (FIB) procedure, used a micro-tweezer device to grab the cone and approach it to a tungsten tip base suitably prepared. An important and positive side-effect of the tweezers was to protect the CcnT apex against ion irradiation during the ion welding of the cone [21]. Although this method was efficient, it remained difficult to handle and almost impossible to automate. We have developed a new mounting procedure, again based on a FIB, where an omniprobe tip is used to perform the lift off, which is more common on standard FIB machine, and easier to automate for an industrial perspective. The method is described in Fig. 1, where the full sequence is reported. After choosing carefully the suitable CcnT (Fig. 1A), the omniprobe is brought to contact with the rough part surface of the CcnT and then is welded to it thanks to an ion beam-induced deposition of platinum (Pt) (Fig. 1B). The W tip base is brought in the field of the SEM image (Fig. 1C) and positioned to protect the carbon cone against ion irradiation in the ion imaging mode (Fig. 1D). Using the Ga beam the CcnT base is cut while the apex remains protected, and then is brought in contact with and welded to the tip base using the ion beam-induced deposition of Pt before the omniprobe is finally removed (Fig. 1E and F). A part of the omniprobe remains welded on the edge of the rough surface of the CcnT base (see Fig. 1F) but does not induce any disturbance to the emission. Furthermore, on Fig. 1C and D we can notice the presence of a used CcnT formerly mounted on the W tip base coming from a previous emission experiment. Prior to the welding of the fresh CcnT, the used one had been removed by cutting it using the Ga beam. It is worth noting that this is an additional positive point of the CcnT cathodes as the same W tip base can be reused many times. 3. Electron emission properties To determine the current angular density, the CcnT tip was installed inside a dedicated ultra high vacuum (UHV) bench (in the low 10−9 Pa range) equipped with an extracting anode, a fluorescent screen below the anode and a faraday cup located in the centre of the screen. A steady value of 13 μA.Sr−1 was measured which is in the range of common values encountered for W cathode CFE source. This small value makes it possible to conserve a good beam
Fig. 1. New FIB-based preparation method using omniprobe tip to grab the carbon cone and approach it to the W tip base. The final welding step is the same as the one previously reported in [21].
energy spread measured in the range of 0.3 eV [21] minimising the Boersch effect in the apex vicinity. To fully characterize all the FE properties, the CcnT tip was fitted inside a dedicated STEM Hitachi HD-2700 equipped with a faraday cup in the sample position to measure the probe current, and with Bright Field (BF), Dark Field (DF) and High Angle Annular DF (HAADF) detectors to perform STEM imaging (see Fig. 3A). The high voltage configuration of the HD-2700 source, described in Fig. 3B, is comparable to the HF-20 0 0 source configuration used in the previous studies and already reported in [21]. The tip is set at V0 = −200 kV and face the extracting anode set at a voltage V1 relatively to the tip. The distance between the tip and the anode is adjusted in a range of about 4–7 mm. The first anode of the accelerating tube, called focusing anode or gun lens, is set at a potential V2 relatively to the tip. The ratio R = V2 / V1 between the focusing anode and the extracting anode can be adjusted between 2 and 13. The ratio is used to set the position of the full gun cross-over which may be either real (which is the condition selected for TEM applications to maximise the beam intensity) or virtual (in order to maximise the first illumination lens demagnification for STEM applications [24]). Fig. 2 shows the stability data of a 2 nA probe at 200 kV measured on the sample stage over 9 h without any flash cleaning of the tip during the measurement. Fig. 2A corresponds to the CcnT based CFE source with a total emission current set at 5.7 μA, an extraction voltage V1 = 1.3 kV and a ratio of 13, while Fig. 2B refers to the standard W [310] based CFE source with an emission current of 10 μA for an extraction voltage of 2.9 kV and a ratio of 4.5. The vacuum system of the source was the same for the two experiments, fitted with two ion pumps and one non-evaporable getter (NEG) pump reaching a vacuum level below 2.10−9 Pa. As Fig. 2 clearly demonstrates, the long-term stability is strongly improved by using our CcnT as FE cathode. Indeed, using
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
305
Fig. 2. Nine hours stability measurements, performed on a Hitachi HD-2700 dedicated STEM, of a 2 nA probe at 200 kV. Comparison between CcnT CFE source (A) and W [310] CFE source (B). Both measurements were started after a flash-cleaning of the tips.
state of the art W [310] CFE source, the probe current decreases by a factor of 50% after 4 h and by almost 90% after 8 h while in the same emission conditions the CcnT CFE source shows no decrease of the probe current over the full measurement time range i.e. 9 h. We can notice the presence of small peaks, which have no consequence on STEM or TEM imaging. As already reported in [12,21,22], the root mean square value of the probe current noise is around 1%, which is in the order of magnitude of standard CFEG technology. 4. Optical properties As mentioned previously, the key feature of the HD-2700 high voltage gun, described in Fig. 3B, is the possibility to adjust the full gun cross-over position thanks to the ratio value. Compare to the usual W [310] tips the CcnT apex is smaller [21]. As a consequence the extraction voltage required for a specific probe current value will be smaller. This is the case in Fig. 2 where 2 nA are obtained for 1.3 kV of extraction voltage using the CcnT cathode and for 2.9 kV using the W [310] cathode. This difference has an important consequence in the electron optical configuration because it changes the strength of the electrostatic gun lens composed by the tip, the extraction anode and the focusing anode. This effect was simulated and reported in Fig. 4 showing the electrons trajectories calculated for 5 ratio values using the same emission conditions as those described in Fig. 2. We can see that, whatever the ratio value, the 200 kV gun cross-over remains real when the CcnT is used as a cathode. The same calculations for a W [310] cathode show that the gun crossover position switches from a virtual position to a real one for a ratio in between 6 and 8. This effect is clearly demonstrates for the ratio 2, which corresponds to the condition with the minimum strength of the electrostatic gun lens. Even for this condition, the cross-over is real and located below the accelerator with the CcnT cathode while it is virtual for the standard W [310] situation.
This unwanted effect, related to the small apex size of the CcnT, has a direct consequence on the STEM properties. Indeed, for STEM imaging, the virtual position is the most desirable in order to maximize the demagnification power of the first illumination lens which is merely related to the distance between the gun cross-over and this lens. Furthermore the aberration coefficients of the electrostatic gun lens are different in real cross-over and virtual cross-over conditions. In real cross-over condition the spherical and chromatic aberration coefficients of this electrostatic lens are respectively Cs ≈ 60 mm and Cc ≈ 12 mm while in virtual crossover condition they are respectively Cs ≈ 35 mm and Cc ≈ 10 mm. As a consequence, the first spot size formed by the gun lens, starting from an usual virtual source size of CFE tip/extraction anode source assembly in the range of 2–5 nm (see Fig. 5B), will be in the range of 500 nm in real cross-over condition and about half of it in the virtual cross-over condition. That’s why the first illumination lenses, situated after the gun, must demagnify strongly this first spot size to achieve a sub-atomic electron probe on the sample plane. To maximise this demagnification, the distance between the spot size formed by the gun and the first illumination lens must be maximised. To increase this distance using a CcnT source, we could use a ratio 13 condition. It is important to stress that this condition is not optimum for an EELS application point of view as the cross-over is located in a zone where the electrons have a low energy, which is favourable to an energy spreading due to the Boersch effect. Furthermore, as shown in Fig. 3D, even under this ratio condition the demagnification of the gun cross-over by the first illumination lens (called C1) never reaches values otherwise usually obtained by a W [310] cathode for all C1 lens current values. For a maximum C1 lens current of 3 A, the demagnification is 400 with a CcnT source of ratio 13 and almost 700 for a W [310] source of ratio 4.5. The BF-STEM image of gold particles performed with the CcnT CFE source, set to a ratio 13, is reported in Fig. 3C. Although the
306
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
Fig. 4. Electron trajectories inside the HD-2700 source calculated with Simion 8.1 [25] software using the same conditions (meshing, boundary conditions, etc.) as those reported in [26]. The left trajectories correspond to one off-axis electron extract from the CcnT CFE source set to 5.7 μA of emission current at 1.3 kV and 5 ratio values. The initial conditions of the electron remain constants for each ratio. The right trajectories come from one off-axis electron extract from a W [310] cathode set to 10 μA of emission current at 2.9 kV and 5 ratio values. To improve visibility the trajectories are enlarged in the x direction by a factor 500 and superimposed to a sketch of the source.
Fig. 3. A: Configuration of the dedicated STEM Hitachi HD-2700. B: Electron optical configuration of the 200 kV cold field emission source showing the cathode tip, the extraction anode, the focusing anode, the accelerating tube, the condensor aperture and the first illumination lens [24]. C: BF-STEM image of gold nanoparticles obtained with the HD-2700 fitted with CcnT CFE source set to a ratio 13. Emission conditions were the same as the one reported in Fig. 2. D: Comparison of condensor 1 lens demagnification power of the gun cross-over obtained with a CcnT and a W [310] CFE source.
overall STEM image has a good quality, it was impossible to reach the desired atomic resolution due to the lack of demagnification of the gun spot size obtained in real cross-over condition. To address this limiting issue, we propose a new gun configuration based on a magnetic gun lens equivalent to the one already proposed for the design of a 1 MeV CFE source [27]. In Fig. 5A we show that, when the tip is precisely placed in the centre of the magnetic field, virtual cross-over conditions of the gun could be found even for a small extraction voltage. Using such an innovative design, we retrieve the optical flexibility necessary to make an effective use of a CcnT-based CFE source allowing both virtual and real cross-over configurations by adjusting the ratio and the magnetic gun lens strength. Furthermore, a positive side effect of such a magnetic gun lens configuration will be to considerably reduce the spherical aberration contribution of the gun to the electron probe, going from Cs ≈35 mm in the original electrostatic design set in virtual cross-over condition to around Cs ≈ 6 mm with such a magnetic gun lens [27,29]. Finally, all the measured major properties of a CcnT CFE source are summarized in the table reported in Fig. 5B and compared to standard W [310] CFE and W/ZrO [100] Schottky FE sources. The reported reduced brightness is taken from [21] and the exit work function from [23]. The tip/extraction anode assembly virtual source size was estimated combining the brightness and the angular current density measured in our experiments. A source radius of 1.6 nm has been extracted which is in the order of magnitude of standard CFE source size values.
Fig. 5. A. One off-axis electron trajectory calculated with Simion 8.1 [25] software inside the HD-2700 with a magnetic gun lens superimposed to the tip region [27]. The source is set at a ratio of 2 and an extraction voltage of 1.3 kV. The virtual cross-over condition can be obtained when the tip is precisely inserted in the centre of the axial magnetic flux density. The magnetic flux density in Simion has been treated as described in [26] for an excitation of 7.9 × 10−2 T obtained with a 1700turn coil, a current of 0.3 A and a 6 mm gap as proposed in [29]. Due to Larmor rotation, the (x,z) and (y,z) projections of the 3D trajectory are reported in red and green respectively. B: Comparison table of the main source properties. All the numbers reported for standard W [310] CFE source and ZrOW [100] SFE source have been gathered from [2] and [28]. To improve visibility the trajectories are enlarged in the x/y direction by a factor 10 0 0 and superimposed to a sketch of the source.
5. Conclusion Thanks to the use of a carbon cone nanotip as a field emission cathode inside a modern 200 kV dedicated FE STEM, we have considerably improved the probe current stability of a CFE source making it a serious competitor to Schottky field emission source in
S. Mamishin et al. / Ultramicroscopy 182 (2017) 303–307
the TEM, STEM and even SEM market. Besides its very good stability, the CcnT CFE source preserves the excellent properties of a standard cold field emission source i.e. high brightness together with low current angular density, small virtual source size and low energy spread. These excellent properties are however strongly limited by the electron optical configuration of the original CFE source, which is responsible for a lack of demagnification power of the illumination system and a stronger contribution of the gun lens spherical aberration. To take full advantage of the properties provided by the CcnT cathode, a dedicated electron optical configuration should be used. We have shown that the use of a magnetic gun lens could be a valuable solution, which is practically feasible in combination with a 200 kV accelerator. Our future work will move to this direction. Acknowledgment These activities are supported by the French National Research Agency under the ‘Investissement d Avenir’ program reference no. ANR-10-EQPX-38- 01. The authors acknowledge the ‘Conseil Regional Midi-Pyrénées’ and the European FEDER for funding within the CPER program. The authors acknowledge the CNRS and HHT for supporting the CNRS-HHT partnership in which this work was conducted. The authors acknowledge the European Union under the Seventh Framework Programme under a contract for an Integrated Infrastructure Initiative Reference 312483-ESTEEM2. The authors acknowledge Dr. Watanabe of HHT for fruitful discussion. References [1] A.V. Crewe, et al., Rev. Sci. Instrum. 39 (1968) 576. http://dx.doi.org/10.1063/1. 1683435. [2] L.W. Swanson and G. A Schwind, Adv. Imaging Electr. Phys., Vol. 159. [3] O.L. Krivanek, et al., Nature 464 (2010) 571–574. http://dx.doi.org/10.1038/ nature08879. [4] A. Tonomura, T. Matsuda, J. Endo, H. Todokoro, T. Komoda, J. Electron Microsc. 28 (1) (1979) 1–11. http://dx.doi.org/10.1093/oxfordjournals.jmicro.a050142. [5] J.W. Butler, 6 Intern. Cong. Electron Microscopy, 191 (Kyoto, 1966). [6] A. Tonomura, Jpn. J. Appl. Phys. 12 (1973) 1065–1069. http://dx.doi.org/10.1143/ JJAP.12.1065.
307
[7] H. Rose, Optik 85 (1990) 19–24. [8] N. Dellby, N.J. Bacon, P. Hrncirik, M.F. Murfitt, G.S. Skone, Z.S. Szilagyi, O.L. Krivanek, Eur. Phys. J. Phys. 54 (2011) 33505. http://dx.doi.org/10.1051/epjap/ 2011100429. [9] Y. Kohno, E. Okunishi, T. Tomita, I. Ishikawa, T. Kaneyama, Y. Ohkura, Y. Kondo, T. Isabell, Microscopy Anal. Nanotechnol. 24 (2010) S9–S13. [10] H. Müller, I. Maßmann, S. Uholemann, P. Hartel, J. Zach, M. Haider, Nucl. Instr. Meth. A645 (2011) 20–27. http://dx.doi.org/10.1016/j.nima.2010.12.091. [11] E. Martin, J.K. Trolan, W.P. Dyke, J. Appl. Phys. 31 (1960) 782–789. http://dx. doi.org/10.1063/1.1735699. [12] L. Knoop, C. Gatel, F. Houdellier, M. Monthioux, A. Masseboeuf, E. Snoeck, M.J. Hÿtch, Appl. Phys. Lett. 106 (2015) 263101. http://dx.doi.org/10.1063/1. 4923245. [13] H. Zhang, J. Tang, J Yuan, Y. Yamauchi, T.T. Suzuki, N. Shinya, K. Nakajima, L. Quin, Nat. Nanotechnol. (2015) 1–8. http://DOI:10.1038/NNANO.2015.276. [14] L.W. Swanson, Surf. Sci. 70 (1978) 165–180. http://dx.doi.org/10.1016/ 0039- 6028(78)90406- 5. [15] K.S. Yeong, J.T.L. Thong, J. Appl. Phys. 99 (2006) 104903. http://dx.doi.org/10. 1063/1.2197267. [16] K. Kasuya, M. Kimura, S. Katagiri, T. Ohshima, Microsc. Microanal. 16 (Suppl 2) (2010). http://dx.doi.org/10.1017/S1431927610056096. [17] V.I. Kleshch, S.T. Purcell, A.N. Obraztsov, Scient. Rep. 6 (2016) 35260. http://dx. doi.org/10.1038/srep35260. [18] A.K. Singh, J. Kumar, J. Appl. Phys. 113 (2013) 053303. http://dx.doi.org/10. 1063/1.4790323. [19] W.K. Lo, G. Parthasarathy, C.W. Lo, D.M. Tanenbaum, H.G. Craighead, M.S. lsaacson, J. Vac. Sci. Technol. B 14 (6) (1996) 3787. http://dx.doi.org/10.1116/1. 588668. [20] N. Jonge, Y. Lamy, K. Schoots, T.H. Oosterkamp, Nature 420 (2002) 393–395. http://dx.doi.org/10.1038/nature01233. [21] F. Houdellier, A. Masseboeuf, M. Monthioux, M.J. Hÿtch, Carbon 50 (2012) 2037–2044. http://dx.doi.org/10.1016/j.carbon.2012.01.023. [22] F Houdellier, L. de Knoop, C. Gatel, A. Masseboeuf, S. Mamishin, Y. Taniguchi, M. Delmas, M. Monthioux, M.J. Hÿtch, E. Snoeck, Ultramicroscopy 151 (2014) 107–115. http://dx.doi.org/10.1016/j.ultramic.2014.11.021. [23] L. Knoop, F. Houdellier, C. Gatel, A. Masseboeuf, M. Monthioux, M. Hÿtch, Micron 63 (2014) 2–8. http://dx.doi.org/10.1016/j.micron.2014.03.005. [24] H. Murakoshi, M. Ichihashi, S. Isakozawa and Y. Sato, USPatent 5134289A [25] D.A. Dahl, Int. J. Mass Spectrom. 200 (2000) 3–25. http://dx.doi.org/10.1016/ S1387-3806(0 0)0 0305-5. [26] Y. Kubo, C. Gatel, E. Snoeck, F. Houdellier, Ultramicroscopy 175 (2017) 67–80. http://dx.doi.org/10.1016/j.ultramic.2017.01.007. [27] T. Kawasaki, J. Endo, S. Isakozawa, M. Tomita and A. Tonomura, USPatent 4945247A [28] L.W. Swanson, G.A. Schwind, Review Of Zr/O Schottky Cathode, in: John Orloff (Ed.), Handbook of Charged Particle Optics, 2nd ed., CRC Press LLC, New York, 2009, p. 665. ISBN 978-1-4200-45543. [29] K. Kasuya, T. Kawasaki, N. Moriya, M. Arai, T. Furutsu, J. Vac. Sci. Technol. B32 (3) (2014) 0318802.