- Email: [email protected]
Evaporation mechanisms of MgO in laser assisted atom probe tomography
Ultramicroscopy 111 (2011) 571–575
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
Evaporation mechanisms of MgO in laser assisted atom probe tomography B. Mazumder a,n, A. Vella a, B. Deconihout a, Tala’at Al-Kassab b,c a b c
Groupe de Physique des Mate´riaux, University of Rouen, France Institut f¨ ur Materialphysik, G¨ ottingen, Germany King Abdullah University of Science and Technology (KAUST), Thuwal, KSA
a r t i c l e in f o
abstract
Available online 19 November 2010
In this paper the field evaporation properties of bulk MgO and sandwiched MgO layers in Fe are compared using laser assisted Atom Probe Tomography. The comparison of flight time spectra gives an estimate of the evaporation times as a function of the wavelength and the laser energy. It is shown that the evaporation takes place in two steps on two different time scales in MgO. It is also shown that as long as the MgO layer is buried in Fe, the evaporation is dominated by the photon absorption in Fe layer at the tip apex. Eventually the evaporation process of MgO is discussed based on the difference between the bulk materials and the multilayer samples. & 2010 Elsevier B.V. All rights reserved.
Keywords: MgO Field evaporation Atom probe tomography
1. Introduction Among the relevant new techniques available in nanoscience, atom probe tomography (APT) provides key information due to its uniqueness in three dimensional atomic scale imaging of materials. Long limited to metals or highly doped semi-conductors, the implementation of ultra fast laser assisted APT has considerably widened the field of application of the technique, making it applicable to semi-conductors, oxides and ceramic materials, which are key materials for micro electronics [1–4]. It is now possible to analyse high band gap materials like Al2O3, HfO2 and MgO [2,5,6]. The study of metal–oxide interfaces is important because of the inter-reaction and inter-diffusion phenomena occurring along the interface between thin layers. Although APT has provided a large number of results on a wide panel of materials, interpretation of 3D images are sometimes difficult due to a lack in the understanding of the different evaporation processes involved in semi-conductors (SC) and oxides. For metals, the roles of the laser wavelength, energy and polarization are quite well understood and can be interpreted by ¨ the ultra fast heating and cooling mechanisms on the Muller emitter after the laser excitation. Typical evaporation times can be predicted and optimised as a function of analysis parameters by studying the confinement of the absorption at the tip apex [7,8]. For SC and insulators, the field evaporation process is still unclear. Several behaviours have been evidenced such as surface photo resonant ionisation at low laser intensities or two time-scale evaporation for higher energies and for certain wavelengths [9,10]. For instance, by changing the laser energy silicon shows
n
Corresponding author. E-mail address: [email protected] (B. Mazumder).
0304-3991/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2010.11.017
different regimes of field evaporation [10,11]. At high laser intensities and photon energy near the band gap energy, a very long evaporation time was reported [11]. This work is focussed on the field evaporation of oxides through the investigation of MgO. The choice of MgO is motivated by ever increasing interest in this oxide due to their promising applications in magneto electronic devices [12–14] By analysing bulk MgO, it is shown that the evaporation is similar to that of SC taking place on two different time scales. The relative importance of the two processes is found to depend on the laser parameters (wavelength and energy). For a photon energy near the band gap of MgO, the delayed evaporation disappears resulting in the disappearance of the wide hump after the main peaks and consequently to a good mass resolution. 32 and 4 nm MgO multilayers are also investigated in detail for different laser parameters. The differences in the evaporation process between the bulk and the multilayer structure is then discussed and interpreted.
2. Material and methods The multi layered structures of Fe/MgO/Fe are prepared by means of argon plasma sputter deposition on a pre-patterned substrate consisting of an assembly of flat-topped Si (1 0 0) pillars (5 5 100) mm3. The schematic and instrumental details of the sputter deposition are described elsewhere [5]. It is very important to clean the substrate by ion beam prior to deposition to achieve good mechanical stability of the specimen [15]. The base pressure of the deposition chamber before deposition is 10 5 Pa. Ar pressure during the sputtering is 2 10 2 Pa. The MgO layer is deposited using an MgO target with no further additional oxidation and
572
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
without any oxygen floating during deposition. The sputtering rate for Fe and MgO are 2.26 and 0.3 nm/min, respectively. The first and third layers sputtered are Iron (Fe), and in between, the insulating barrier of MgO is deposited. Each layer is 32 nm thick. Silicon pillars are then removed from the wafer and mounted on a fine tungsten needle with conductive commercial adhesive under the optical microscope. They are then shaped by Focussed Ion Beam (FIB) milling to obtain tips suitable for APT analysis. The bulk samples are obtained from a MgO substrate (from the company CAMECA) by lift out and prepared in the same way. The end radius of the tips measured in SEM before the analysis is about 40 75 nm with a cone angle below 101. For the experiments, all the tips were prepared in the same way guarantying an end radius almost constant from one tip to another. Analyses were performed at 80 K in an ultrahigh vacuum chamber at a pressure of 10 9–10 10 Pa. The instrument is a Laser Assisted Wide Angle Tomography Atom Probe (LAWATAP). Details of the instrument are given in literature [16].The femtosecond laser pulse system used is an amplified ytterbium-doped laser (AMPLITUDE SYSTEM s-pulse) delivering 350 fs pulses at 100 KHz. The laser wavelength of 1030 nm can be changed to green (515 nm) and UV (343 nm) by Second Harmonic Generation and sum generation, respectively. For our work 515 and 343 nm are used.
3. Results and discussions 3.1. Field evaporation of bulk MgO We can expect that due to the high resistivity of MgO tip, the surface field is lower than the filed measured for metallic tip [17]. However the specimen is about few mm long and the applied voltage difference (VDC) is about 10 kV, which corresponds to an electric field inside the oxide of a few V/nm. This field is higher than the break down field of MgO [18]; hence, the oxide tip becomes conducting and can be analysed in APT by monitoring the standing voltage. Using green light (photon energy 2.45 eV), bulk MgO sample analyses are performed at a constant detection rate (0.005– 0.01 atoms/pulse). The polarization was set axial and the laser fluence was tuned from 6 to 51 J/m2. Fig. 1a shows that, for low laser fluence, both the Mg2 + main peak and its 2 isotope peaks are clearly resolved in the mass spectrum. The mass resolution is good (m/Dm¼210) enough to allow the chemical identification of atoms.
As the fluence increases, it becomes more and more difficult to distinguish the isotopes because of the appearance of a wide peak just spreading on several amu. For the highest laser fluence, this hump almost hides the isotopes. The same experiments were carried out using UV light. As it can be seen from Fig. 1b, for photon energy of 3.5 eV, all the peaks are clearly resolved without any trace of hump even for the highest laser fluence used (50 J/m2). This behaviour was already observed by authors in silicon [11]. When a Si tip (band gap energy: 1.12 eV) is illuminated by IR laser with photon energy of 1.2 eV, a similar wide peak appears at high laser fluences. In addition, for green light (photon energy above the gap) this hump does not exist. It is clear that the presence of two peaks (one narrow and one wide) at high fluences is the proof of two different heating processes. The first one is rapid and can be attributed to electron–electron thermalization in the conduction band. The second one is very long and due to inter-band electron–hole recombination. Since the recombination of charges increases the lattice temperature, ions are evaporated by a two steps thermal process. More details of our model can be found in Ref. [11] To apply this model to the present work, we have to consider the band gap of bulk MgO. MgO is an insulator having a band gap of 7.8 eV [19]. However, like other SC or insulators, the presence of lattice defects in MgO has been shown to give rise to a variety of optical and electrical conductivity phenomena, which are not found in perfect crystalline materials [20–22]. The presence of surface oxygen vacancies affects the fundamental band gap. When an oxygen atom is removed, a doubly occupied electronic state appears in the fundamental gap, which is called F center. F2 state center exists when two oxygen vacancies are located either on neighbouring lattice oxygen sites or on second nearest neighbour sites of the atom lattice [23] as is shown in Fig. 2a. In Fig. 2b. the band structure of MgO with F2 center defects is presented. A 1% oxygen deficiency in the MgO layer correspond to a 5 1020 cm 3 oxygen vacancies, i.e. a concentration sufficient to induce defect in the band structure resulting in new states close to the bottom of the conduction band. As a consequence, the band gap of bulk MgO with F2 defects is close to 2.5 eV as reported in [24]. This is consistent with our explanation of a resonant absorption of green light, i.e. 2.5 eV photon energy whereas, using UV light, the resonant excitation does not occur, since the photon energy is much larger than the gap.
3.2. Field evaporation of sandwiched MgO layer
Fig. 1. Mass spectrum showing Mg2 + as obtained from bulk MgO analyses: (a) With green light (l ¼515 nm) at different laser fluencies and (b) with UV (l ¼ 343 nm), with high laser fluence.
Fig. 3 shows 3D images from the analysis of the 32 nm(Fe)/ 32 nm (MgO)/32 nm (Fe) multilayer specimen using green light and UV light. The selected volume shown in this image is 8 8 50 nm3. It is worth noting that, due to the difference in the evaporation fields of the layers, the DC field and the laser fluence has to be changed during the analysis when the interfaces are crossed. Using green light, the laser fluence is increased close to the second interface to avoid a too large increase of the voltage, which might cause the tip rupture. The analysis is performed with a fixed flux (0.005–0.01 atoms/pulse). The laser fluence of 18 J/m2 used for the first Fe layer was increased up to 42 J/m2 in the MgO layer. Hence, at the second (MgO/Fe) interface an increase of laser fluence about 40% was applied. Fig. 4 shows that when the laser fluence increases, the Mg2 + main peak for the 32 nm thick sample shows the same behaviour as for bulk MgO analyses. It is followed by a wide peak spreading over isotopes at high fluences. However, flight time spectra of the Fe2 + peak for the first and second layer do not have the same shape, as shown in Fig. 5a. Only a wide peak is observable, with a very long rise time of 7.5 ns, corresponding to a long heating process.
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
573
Fig. 2. (a) Lattice vacancies in MgO [23] (b) Band structure of MgO with F2 centers. New states appear below the conduction band making resonant absorption of green photons possible.
Fig. 3. 3D distribution of Mg (magenta), O (sky blue) and Fe (blue) atoms resulting from the analysis of a 4 nm MgO barrier obtained by LAWATAP analysis with (a) UV light (l ¼ 343 nm) and (b) green light (l ¼515 nm), volume : 8 8 50 nm3. (b) 3D distribution of Mg, O and Fe atoms in the 32 nm Fe/MgO/Fe multilayer with (c) UV light and (d) Green light, volume : 8 8 50 nm3. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
The laser spot size on the tip is very large (80 mm in diameter); hence, the whole multilayer structure is uniformly illuminated. The density of electrons in the Fe layers is two orders of magnitude higher than the MgO defect density so that the absorption is higher in Fe. In addition, the light diffraction at the apex generates an absorbing zone at a distance of about d¼ l/3 nm from the apex, as shown by authors for several metals in Ref. [25]. The long rise time of Fe peak (7.5 ns) is due to the time necessary for the heat to reach the tip apex given by t ¼d2/D¼8 ns with D, the thermal diffusivity of Fe (5 10 6 m2/s) and d ¼187 nm. Furthermore, the rise time is the same for the first and the second layer of Fe because the thermal diffusivity of MgO is almost equal to that of Fe, as reported in the literature (D of MgO 5 10 6 m2/s a 200 K). The heating of the specimen, due to the recombination of charges generated by absorption in the MgO layer is very low, and the resulting evaporation rate is completely flattened by the exponential behaviour of the evaporation rate. However, when the MgO is at the tip apex, its absorption increases due to the presence of a strong electric field; hence, a strong band bending, which changes locally the materials properties to that of a semi-metals
[17]. Hence, the apex of the tip becomes very absorbing, and the evaporation of the MgO layer follows the same behaviour of the MgO bulk, discussed in the previous section. An interesting aspect is the change in the background noise level in spectra related to the investigation of the first and second Fe layers. The background noise is lower for the second Fe layer because the laser energy used was higher as compared to the energy used to investigate the first Fe layer. Indeed, the DC voltage (adjusted to have a constant flux) was 20% lower at the second interface increasing the ionic barrier (Qn). This increase of the ionic barrier is consistent with the observed lower background noise. The Fe peak before the MgO layer is wider than the Fe peak after. The width of the peak is related to the cooling time (tcooling), the temperature rise (Trise) and the ionic barrier (Qn) height following the equation below
fpexp
Qn KB TðtÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi with TðtÞ ¼ T0 þ Trise = 1 þ 2t=tcooling
ð1Þ
574
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
Fig. 4. Flight time peak of Mg2 + in Fe/MgO/Fe for different laser fluences with green light (l ¼515 nm).
Fig. 6. Mg2 + flight time peak (red) and Fe peaks for the first layer (black) and second one (blue) (a) for green light (l ¼ 515 nm) and (b) UV light (l ¼343 nm) obtained from the investigation of a 4 nm thick MgO tunnel barrier in iron. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 5. Mg2 + flight time peak (magenta) and Fe peaks for the first layer (black) and second one (blue) (a) for green light (l ¼ 515 nm) and (b) UV light (l ¼343 nm) on a 32 nm thick MgO layer in iron. Red and green lines are theoretical fits using Eq. (1) with Qn ¼0.3 and 0.4 eV, Trise ¼200 and 150 K, respectively and ttcooling ¼50 ns. Red, cyan and blue lines are theoretical fits using Eq. (1) with Qn ¼ 0.15, 0.11 and 0.09 eV, Trise ¼ 80, 200 and 200 K and ttcooling ¼ 0.8, 1.5 and 0.8 ns, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
From these equations, the decay of the temperature and thus, that of the evaporation flux can be calculated and compared to experimental data (smooth solid lines in Fig. 5 correspond to the fit using Eq. (1)). In both cases, the cooling time is the same; only the Trise and Qn are changed. The knowledge of the laser fluence used for each layer allows fitting the experimental data (Trise 10 and 180 K and Qn of 0.3 and 0.4 eV for the first and the second layer, respectively). It can be seen that the model and the experiments are in good agreement. Using UV light, the laser fluence is always increased close to the second interface to limit the increase of the voltage which can cause the tip rupture. However, in contrast to the green light case, only an increase of less than 30% is possible due to the optical setup limitation. The analysis was performed at a fixed flux (0.005–0.01 atom/pulse)
and the laser fluence was fixed at 9 J/m2 for the first Fe layer and at 13 J/m2 for the second one. The TOF spectra of the three Mg2 + peaks and Fe for the first and second layer are shown in Fig. 5b. The background noise is higher for the Fe peak corresponding to the second layer. Because the laser fluence underwent only a small increase at the second interface, the voltage changed a lot, to keep the flux constant ( 450% of change). Hence, the ionic barrier decreased by about 50% increasing the background noise. Moreover this noise is higher for Mg peak due to its lower evaporation field as compared to Fe. Consequently, the value of the ionic barrier in Eq. 1 was fixed to 0.11 eV for Mg, 0.155 eV for the first Fe layer, and 0.09 eV for the second one. The fits of flight time spectra (Fig. 5) as obtained by Eq (1), were obtained for a Trise ¼80 K in the first layer, consistent with the low laser fluence used for the evaporation. The temperature rise is larger (200 K) in the second Fe layer and in the Mg layer owing to the increase in laser fluence. The cooling time is very short (0.8 ns), corresponding to a heated zone (w) of almost 70 nm following the equation:tcooling ¼ w2 =D. In other words, using UV light, the heating zone due to light diffraction is closer to the tip apex; hence, very fast heating and cooling processes are observed. Moreover, in the MgO layer, the UV photons have an energy higher than the gap so that the evaporation due to the long recombination process cannot be evidenced. Fig. 3(a) and (b) shows images of a 4 nm MgO tunnel barrier obtained in green and UV. We found that the behaviour of the evaporation of Fe and MgO layers are exactly the same for that of the 32 nm. In particular, Fe peaks (Fig. 6) related to each Fe layer have a rise time of 7.5 ns. In addition the Mg2 + peak is followed by a hump disappearing in UV.
4. Conclusion By changing the laser fluence and wavelength different evaporation behaviours of the MgO have been evidenced. For MgO bulk sample, for the case of green light the evaporation mechanisms and the presence of a long evaporation process are explained taking into account the existence of inter gap energy levels due to oxygen vacancies ( 41%). Hence, a resonant excitation at l ¼ 515 nm can
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
give rise to two charges relaxation processes, as previously reported for silicon and SiC. For MgO layer sandwiched between two Fe layers the evaporation mechanism of the metal is dominant for the tip owing to the higher absorption coefficient of Fe as compared to MgO both in the green and UV; reflecting an evaporation behaviour of the tip as if there was no MgO layer involved in the process. Moreover, the change in the thickness of the MgO layer (from 4 to 32 nm) does not change the evaporation behaviour. Only when the MgO layer is at the surface, its absorption increases due to the band bending; hence, the evaporation mechanism becomes similar that of the bulk MgO. More investigations, changing the nature of the metal layers, are planed to support our study. In the case of UV analyses, the narrow mass peaks suggest a very small heating zone and, consequently, a short cooling time.
Acknowledgments We acknowledge the ESP Carnot, the TAPAS, the ANR and Cameca France for supporting our work. We thank L. Renaud for provided bulk MgO samples. We would like to thank Ryota Gemma for helping in MgO sample preparation.
575
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]
B. Gault, et al., Appl. Phys. Lett. 88 (2006) 114101. T.F. Kelly, M.K. Miller, Rev. Sci. Inst. 78 (2007) 031101. E. Cadel, et al., J. Appl. Phys. 106 (2009) 044908. Y.M. Chen, et al., Scripta Mater. 61 (2009) 693. G. Mario Kuduz, Schmitz, R. Kirchheim, Ultramicroscopy 101 (2004) 197–205. D. Blavette, J. Int., Mater. Res. 99 (2008) 454. J. Houard, et al., APL 94 (2009) 121905. F. Vurpillot, et al., J. Phys. D: Appl. Phys. 42 (2009) 125502. T.T. Tsong, et al., J. Chem. Phys. 65 (1976) 2469. B. Mazumder, et al., New J. Phys. 12 (2010) 113029. B. Mazumder, et al., Appl. Phys. Lett. 97 (7) (2010) 073104. W.J. Gallagher, J. Appl. Phys. 81 (1997) 3741. J.M. Daughton, J. Appl. Phys. 81 (1997) 3758. M. Peter., Levy, A. Fert, Phys. Rev. B 74 (2006) 224446. J. Schleiwies, G. Schmitz, Mater. Sci. Eng. A327 (2002) 94. B. Gault, et al., Rev. Sci. Instrum. 77 (2006) 043705. M.K. Miller, et al., Atom Probe Field Ion Microscopy, Oxford University Press, Oxford, 1996. Y. Irokawa, et al., Appl. Phys. Lett. 84 (15) (2004) 2919–2921. D.M. Roessler, W.C. Walker, Phys. Rev. 159 (1967) 733. G.H. Rosenblatt, et al., Phy. Rev. B 39 (1989) 10309. M..Freund, M.F. Freund, F. Batllo, PRL 63: 2096. J. Tombrello, et al., Phy. Rev. B 30 (1984) 7374. A. Gibson, R. Haydock, J.P. LaFemina, Phy. Rev. B 50 (1994) 2582. S. Gue´zo, et al., APL 93 (2008) 172116. J. Houard, et al., Phy. Rev. B 81 (2010) 125411.
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
Evaporation mechanisms of MgO in laser assisted atom probe tomography B. Mazumder a,n, A. Vella a, B. Deconihout a, Tala’at Al-Kassab b,c a b c
Groupe de Physique des Mate´riaux, University of Rouen, France Institut f¨ ur Materialphysik, G¨ ottingen, Germany King Abdullah University of Science and Technology (KAUST), Thuwal, KSA
a r t i c l e in f o
abstract
Available online 19 November 2010
In this paper the field evaporation properties of bulk MgO and sandwiched MgO layers in Fe are compared using laser assisted Atom Probe Tomography. The comparison of flight time spectra gives an estimate of the evaporation times as a function of the wavelength and the laser energy. It is shown that the evaporation takes place in two steps on two different time scales in MgO. It is also shown that as long as the MgO layer is buried in Fe, the evaporation is dominated by the photon absorption in Fe layer at the tip apex. Eventually the evaporation process of MgO is discussed based on the difference between the bulk materials and the multilayer samples. & 2010 Elsevier B.V. All rights reserved.
Keywords: MgO Field evaporation Atom probe tomography
1. Introduction Among the relevant new techniques available in nanoscience, atom probe tomography (APT) provides key information due to its uniqueness in three dimensional atomic scale imaging of materials. Long limited to metals or highly doped semi-conductors, the implementation of ultra fast laser assisted APT has considerably widened the field of application of the technique, making it applicable to semi-conductors, oxides and ceramic materials, which are key materials for micro electronics [1–4]. It is now possible to analyse high band gap materials like Al2O3, HfO2 and MgO [2,5,6]. The study of metal–oxide interfaces is important because of the inter-reaction and inter-diffusion phenomena occurring along the interface between thin layers. Although APT has provided a large number of results on a wide panel of materials, interpretation of 3D images are sometimes difficult due to a lack in the understanding of the different evaporation processes involved in semi-conductors (SC) and oxides. For metals, the roles of the laser wavelength, energy and polarization are quite well understood and can be interpreted by ¨ the ultra fast heating and cooling mechanisms on the Muller emitter after the laser excitation. Typical evaporation times can be predicted and optimised as a function of analysis parameters by studying the confinement of the absorption at the tip apex [7,8]. For SC and insulators, the field evaporation process is still unclear. Several behaviours have been evidenced such as surface photo resonant ionisation at low laser intensities or two time-scale evaporation for higher energies and for certain wavelengths [9,10]. For instance, by changing the laser energy silicon shows
n
Corresponding author. E-mail address: [email protected] (B. Mazumder).
0304-3991/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2010.11.017
different regimes of field evaporation [10,11]. At high laser intensities and photon energy near the band gap energy, a very long evaporation time was reported [11]. This work is focussed on the field evaporation of oxides through the investigation of MgO. The choice of MgO is motivated by ever increasing interest in this oxide due to their promising applications in magneto electronic devices [12–14] By analysing bulk MgO, it is shown that the evaporation is similar to that of SC taking place on two different time scales. The relative importance of the two processes is found to depend on the laser parameters (wavelength and energy). For a photon energy near the band gap of MgO, the delayed evaporation disappears resulting in the disappearance of the wide hump after the main peaks and consequently to a good mass resolution. 32 and 4 nm MgO multilayers are also investigated in detail for different laser parameters. The differences in the evaporation process between the bulk and the multilayer structure is then discussed and interpreted.
2. Material and methods The multi layered structures of Fe/MgO/Fe are prepared by means of argon plasma sputter deposition on a pre-patterned substrate consisting of an assembly of flat-topped Si (1 0 0) pillars (5 5 100) mm3. The schematic and instrumental details of the sputter deposition are described elsewhere [5]. It is very important to clean the substrate by ion beam prior to deposition to achieve good mechanical stability of the specimen [15]. The base pressure of the deposition chamber before deposition is 10 5 Pa. Ar pressure during the sputtering is 2 10 2 Pa. The MgO layer is deposited using an MgO target with no further additional oxidation and
572
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
without any oxygen floating during deposition. The sputtering rate for Fe and MgO are 2.26 and 0.3 nm/min, respectively. The first and third layers sputtered are Iron (Fe), and in between, the insulating barrier of MgO is deposited. Each layer is 32 nm thick. Silicon pillars are then removed from the wafer and mounted on a fine tungsten needle with conductive commercial adhesive under the optical microscope. They are then shaped by Focussed Ion Beam (FIB) milling to obtain tips suitable for APT analysis. The bulk samples are obtained from a MgO substrate (from the company CAMECA) by lift out and prepared in the same way. The end radius of the tips measured in SEM before the analysis is about 40 75 nm with a cone angle below 101. For the experiments, all the tips were prepared in the same way guarantying an end radius almost constant from one tip to another. Analyses were performed at 80 K in an ultrahigh vacuum chamber at a pressure of 10 9–10 10 Pa. The instrument is a Laser Assisted Wide Angle Tomography Atom Probe (LAWATAP). Details of the instrument are given in literature [16].The femtosecond laser pulse system used is an amplified ytterbium-doped laser (AMPLITUDE SYSTEM s-pulse) delivering 350 fs pulses at 100 KHz. The laser wavelength of 1030 nm can be changed to green (515 nm) and UV (343 nm) by Second Harmonic Generation and sum generation, respectively. For our work 515 and 343 nm are used.
3. Results and discussions 3.1. Field evaporation of bulk MgO We can expect that due to the high resistivity of MgO tip, the surface field is lower than the filed measured for metallic tip [17]. However the specimen is about few mm long and the applied voltage difference (VDC) is about 10 kV, which corresponds to an electric field inside the oxide of a few V/nm. This field is higher than the break down field of MgO [18]; hence, the oxide tip becomes conducting and can be analysed in APT by monitoring the standing voltage. Using green light (photon energy 2.45 eV), bulk MgO sample analyses are performed at a constant detection rate (0.005– 0.01 atoms/pulse). The polarization was set axial and the laser fluence was tuned from 6 to 51 J/m2. Fig. 1a shows that, for low laser fluence, both the Mg2 + main peak and its 2 isotope peaks are clearly resolved in the mass spectrum. The mass resolution is good (m/Dm¼210) enough to allow the chemical identification of atoms.
As the fluence increases, it becomes more and more difficult to distinguish the isotopes because of the appearance of a wide peak just spreading on several amu. For the highest laser fluence, this hump almost hides the isotopes. The same experiments were carried out using UV light. As it can be seen from Fig. 1b, for photon energy of 3.5 eV, all the peaks are clearly resolved without any trace of hump even for the highest laser fluence used (50 J/m2). This behaviour was already observed by authors in silicon [11]. When a Si tip (band gap energy: 1.12 eV) is illuminated by IR laser with photon energy of 1.2 eV, a similar wide peak appears at high laser fluences. In addition, for green light (photon energy above the gap) this hump does not exist. It is clear that the presence of two peaks (one narrow and one wide) at high fluences is the proof of two different heating processes. The first one is rapid and can be attributed to electron–electron thermalization in the conduction band. The second one is very long and due to inter-band electron–hole recombination. Since the recombination of charges increases the lattice temperature, ions are evaporated by a two steps thermal process. More details of our model can be found in Ref. [11] To apply this model to the present work, we have to consider the band gap of bulk MgO. MgO is an insulator having a band gap of 7.8 eV [19]. However, like other SC or insulators, the presence of lattice defects in MgO has been shown to give rise to a variety of optical and electrical conductivity phenomena, which are not found in perfect crystalline materials [20–22]. The presence of surface oxygen vacancies affects the fundamental band gap. When an oxygen atom is removed, a doubly occupied electronic state appears in the fundamental gap, which is called F center. F2 state center exists when two oxygen vacancies are located either on neighbouring lattice oxygen sites or on second nearest neighbour sites of the atom lattice [23] as is shown in Fig. 2a. In Fig. 2b. the band structure of MgO with F2 center defects is presented. A 1% oxygen deficiency in the MgO layer correspond to a 5 1020 cm 3 oxygen vacancies, i.e. a concentration sufficient to induce defect in the band structure resulting in new states close to the bottom of the conduction band. As a consequence, the band gap of bulk MgO with F2 defects is close to 2.5 eV as reported in [24]. This is consistent with our explanation of a resonant absorption of green light, i.e. 2.5 eV photon energy whereas, using UV light, the resonant excitation does not occur, since the photon energy is much larger than the gap.
3.2. Field evaporation of sandwiched MgO layer
Fig. 1. Mass spectrum showing Mg2 + as obtained from bulk MgO analyses: (a) With green light (l ¼515 nm) at different laser fluencies and (b) with UV (l ¼ 343 nm), with high laser fluence.
Fig. 3 shows 3D images from the analysis of the 32 nm(Fe)/ 32 nm (MgO)/32 nm (Fe) multilayer specimen using green light and UV light. The selected volume shown in this image is 8 8 50 nm3. It is worth noting that, due to the difference in the evaporation fields of the layers, the DC field and the laser fluence has to be changed during the analysis when the interfaces are crossed. Using green light, the laser fluence is increased close to the second interface to avoid a too large increase of the voltage, which might cause the tip rupture. The analysis is performed with a fixed flux (0.005–0.01 atoms/pulse). The laser fluence of 18 J/m2 used for the first Fe layer was increased up to 42 J/m2 in the MgO layer. Hence, at the second (MgO/Fe) interface an increase of laser fluence about 40% was applied. Fig. 4 shows that when the laser fluence increases, the Mg2 + main peak for the 32 nm thick sample shows the same behaviour as for bulk MgO analyses. It is followed by a wide peak spreading over isotopes at high fluences. However, flight time spectra of the Fe2 + peak for the first and second layer do not have the same shape, as shown in Fig. 5a. Only a wide peak is observable, with a very long rise time of 7.5 ns, corresponding to a long heating process.
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
573
Fig. 2. (a) Lattice vacancies in MgO [23] (b) Band structure of MgO with F2 centers. New states appear below the conduction band making resonant absorption of green photons possible.
Fig. 3. 3D distribution of Mg (magenta), O (sky blue) and Fe (blue) atoms resulting from the analysis of a 4 nm MgO barrier obtained by LAWATAP analysis with (a) UV light (l ¼ 343 nm) and (b) green light (l ¼515 nm), volume : 8 8 50 nm3. (b) 3D distribution of Mg, O and Fe atoms in the 32 nm Fe/MgO/Fe multilayer with (c) UV light and (d) Green light, volume : 8 8 50 nm3. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
The laser spot size on the tip is very large (80 mm in diameter); hence, the whole multilayer structure is uniformly illuminated. The density of electrons in the Fe layers is two orders of magnitude higher than the MgO defect density so that the absorption is higher in Fe. In addition, the light diffraction at the apex generates an absorbing zone at a distance of about d¼ l/3 nm from the apex, as shown by authors for several metals in Ref. [25]. The long rise time of Fe peak (7.5 ns) is due to the time necessary for the heat to reach the tip apex given by t ¼d2/D¼8 ns with D, the thermal diffusivity of Fe (5 10 6 m2/s) and d ¼187 nm. Furthermore, the rise time is the same for the first and the second layer of Fe because the thermal diffusivity of MgO is almost equal to that of Fe, as reported in the literature (D of MgO 5 10 6 m2/s a 200 K). The heating of the specimen, due to the recombination of charges generated by absorption in the MgO layer is very low, and the resulting evaporation rate is completely flattened by the exponential behaviour of the evaporation rate. However, when the MgO is at the tip apex, its absorption increases due to the presence of a strong electric field; hence, a strong band bending, which changes locally the materials properties to that of a semi-metals
[17]. Hence, the apex of the tip becomes very absorbing, and the evaporation of the MgO layer follows the same behaviour of the MgO bulk, discussed in the previous section. An interesting aspect is the change in the background noise level in spectra related to the investigation of the first and second Fe layers. The background noise is lower for the second Fe layer because the laser energy used was higher as compared to the energy used to investigate the first Fe layer. Indeed, the DC voltage (adjusted to have a constant flux) was 20% lower at the second interface increasing the ionic barrier (Qn). This increase of the ionic barrier is consistent with the observed lower background noise. The Fe peak before the MgO layer is wider than the Fe peak after. The width of the peak is related to the cooling time (tcooling), the temperature rise (Trise) and the ionic barrier (Qn) height following the equation below
fpexp
Qn KB TðtÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi with TðtÞ ¼ T0 þ Trise = 1 þ 2t=tcooling
ð1Þ
574
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
Fig. 4. Flight time peak of Mg2 + in Fe/MgO/Fe for different laser fluences with green light (l ¼515 nm).
Fig. 6. Mg2 + flight time peak (red) and Fe peaks for the first layer (black) and second one (blue) (a) for green light (l ¼ 515 nm) and (b) UV light (l ¼343 nm) obtained from the investigation of a 4 nm thick MgO tunnel barrier in iron. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 5. Mg2 + flight time peak (magenta) and Fe peaks for the first layer (black) and second one (blue) (a) for green light (l ¼ 515 nm) and (b) UV light (l ¼343 nm) on a 32 nm thick MgO layer in iron. Red and green lines are theoretical fits using Eq. (1) with Qn ¼0.3 and 0.4 eV, Trise ¼200 and 150 K, respectively and ttcooling ¼50 ns. Red, cyan and blue lines are theoretical fits using Eq. (1) with Qn ¼ 0.15, 0.11 and 0.09 eV, Trise ¼ 80, 200 and 200 K and ttcooling ¼ 0.8, 1.5 and 0.8 ns, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
From these equations, the decay of the temperature and thus, that of the evaporation flux can be calculated and compared to experimental data (smooth solid lines in Fig. 5 correspond to the fit using Eq. (1)). In both cases, the cooling time is the same; only the Trise and Qn are changed. The knowledge of the laser fluence used for each layer allows fitting the experimental data (Trise 10 and 180 K and Qn of 0.3 and 0.4 eV for the first and the second layer, respectively). It can be seen that the model and the experiments are in good agreement. Using UV light, the laser fluence is always increased close to the second interface to limit the increase of the voltage which can cause the tip rupture. However, in contrast to the green light case, only an increase of less than 30% is possible due to the optical setup limitation. The analysis was performed at a fixed flux (0.005–0.01 atom/pulse)
and the laser fluence was fixed at 9 J/m2 for the first Fe layer and at 13 J/m2 for the second one. The TOF spectra of the three Mg2 + peaks and Fe for the first and second layer are shown in Fig. 5b. The background noise is higher for the Fe peak corresponding to the second layer. Because the laser fluence underwent only a small increase at the second interface, the voltage changed a lot, to keep the flux constant ( 450% of change). Hence, the ionic barrier decreased by about 50% increasing the background noise. Moreover this noise is higher for Mg peak due to its lower evaporation field as compared to Fe. Consequently, the value of the ionic barrier in Eq. 1 was fixed to 0.11 eV for Mg, 0.155 eV for the first Fe layer, and 0.09 eV for the second one. The fits of flight time spectra (Fig. 5) as obtained by Eq (1), were obtained for a Trise ¼80 K in the first layer, consistent with the low laser fluence used for the evaporation. The temperature rise is larger (200 K) in the second Fe layer and in the Mg layer owing to the increase in laser fluence. The cooling time is very short (0.8 ns), corresponding to a heated zone (w) of almost 70 nm following the equation:tcooling ¼ w2 =D. In other words, using UV light, the heating zone due to light diffraction is closer to the tip apex; hence, very fast heating and cooling processes are observed. Moreover, in the MgO layer, the UV photons have an energy higher than the gap so that the evaporation due to the long recombination process cannot be evidenced. Fig. 3(a) and (b) shows images of a 4 nm MgO tunnel barrier obtained in green and UV. We found that the behaviour of the evaporation of Fe and MgO layers are exactly the same for that of the 32 nm. In particular, Fe peaks (Fig. 6) related to each Fe layer have a rise time of 7.5 ns. In addition the Mg2 + peak is followed by a hump disappearing in UV.
4. Conclusion By changing the laser fluence and wavelength different evaporation behaviours of the MgO have been evidenced. For MgO bulk sample, for the case of green light the evaporation mechanisms and the presence of a long evaporation process are explained taking into account the existence of inter gap energy levels due to oxygen vacancies ( 41%). Hence, a resonant excitation at l ¼ 515 nm can
B. Mazumder et al. / Ultramicroscopy 111 (2011) 571–575
give rise to two charges relaxation processes, as previously reported for silicon and SiC. For MgO layer sandwiched between two Fe layers the evaporation mechanism of the metal is dominant for the tip owing to the higher absorption coefficient of Fe as compared to MgO both in the green and UV; reflecting an evaporation behaviour of the tip as if there was no MgO layer involved in the process. Moreover, the change in the thickness of the MgO layer (from 4 to 32 nm) does not change the evaporation behaviour. Only when the MgO layer is at the surface, its absorption increases due to the band bending; hence, the evaporation mechanism becomes similar that of the bulk MgO. More investigations, changing the nature of the metal layers, are planed to support our study. In the case of UV analyses, the narrow mass peaks suggest a very small heating zone and, consequently, a short cooling time.
Acknowledgments We acknowledge the ESP Carnot, the TAPAS, the ANR and Cameca France for supporting our work. We thank L. Renaud for provided bulk MgO samples. We would like to thank Ryota Gemma for helping in MgO sample preparation.
575
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]
B. Gault, et al., Appl. Phys. Lett. 88 (2006) 114101. T.F. Kelly, M.K. Miller, Rev. Sci. Inst. 78 (2007) 031101. E. Cadel, et al., J. Appl. Phys. 106 (2009) 044908. Y.M. Chen, et al., Scripta Mater. 61 (2009) 693. G. Mario Kuduz, Schmitz, R. Kirchheim, Ultramicroscopy 101 (2004) 197–205. D. Blavette, J. Int., Mater. Res. 99 (2008) 454. J. Houard, et al., APL 94 (2009) 121905. F. Vurpillot, et al., J. Phys. D: Appl. Phys. 42 (2009) 125502. T.T. Tsong, et al., J. Chem. Phys. 65 (1976) 2469. B. Mazumder, et al., New J. Phys. 12 (2010) 113029. B. Mazumder, et al., Appl. Phys. Lett. 97 (7) (2010) 073104. W.J. Gallagher, J. Appl. Phys. 81 (1997) 3741. J.M. Daughton, J. Appl. Phys. 81 (1997) 3758. M. Peter., Levy, A. Fert, Phys. Rev. B 74 (2006) 224446. J. Schleiwies, G. Schmitz, Mater. Sci. Eng. A327 (2002) 94. B. Gault, et al., Rev. Sci. Instrum. 77 (2006) 043705. M.K. Miller, et al., Atom Probe Field Ion Microscopy, Oxford University Press, Oxford, 1996. Y. Irokawa, et al., Appl. Phys. Lett. 84 (15) (2004) 2919–2921. D.M. Roessler, W.C. Walker, Phys. Rev. 159 (1967) 733. G.H. Rosenblatt, et al., Phy. Rev. B 39 (1989) 10309. M..Freund, M.F. Freund, F. Batllo, PRL 63: 2096. J. Tombrello, et al., Phy. Rev. B 30 (1984) 7374. A. Gibson, R. Haydock, J.P. LaFemina, Phy. Rev. B 50 (1994) 2582. S. Gue´zo, et al., APL 93 (2008) 172116. J. Houard, et al., Phy. Rev. B 81 (2010) 125411.