Analyzing Si-based structures in 3D with a laser-pulsed local electrode atom probe
06/01/2006
The addition of high-speed laser pulsing and local electrode geometry has transformed the 3D atom probe into a metrology tool that is capable of analyzing Si-based nanostructures on an atom-by-atom basis. In this work, analysis of heavily doped Si using both the voltage-pulsed and laser-pulsed scheme showed that laser-pulsed analysis provided superior mass resolution and sensitivity and was capable of analyzing undoped Si-SiGe structures. Compared to secondary ion mass spectrometry (SIMS), the laser-pulsed local electrode atom probe (LEAP) provided superior interface analysis in terms of spatial and depth resolutions, as well as quantification of Ge and B atoms within multilayer Si-SiGe stacks.
According to the International Technology Roadmap for Semiconductors (ITRS), by 2013 the physical gate lengths of planar CMOS devices will shrink to 18nm, requiring junction depths of ≤9nm [1, 2]. To date, the most common means of maintaining device integrity during scaling has been to maximize the active dopant concentration in the critical device regions and to minimize the redistribution of these dopants during subsequent thermal treatments. However, the ITRS projects dopant levels that are either physically or practically unrealistic given the known limitations to ion implantation and rapid thermal annealing [1-5]. Thus, alternative device strategies are clearly needed.
The selective use of silicon germanium (SiGe) provides much needed relief to critical areas of device design and fabrication. For example, the addition of Ge to Si to form a SiGe alloy allows the amount of strain in the Si to be engineered so that the stress state varies from tensile at high Si levels to compressive at high Ge levels. Creating such a strained state in the channel-namely, by depositing a SiGe layer beneath the channel-significantly increases the hole mobility within the channel. This generates a direct increase in drive current with no reported increase in power dissipation [6].
Additionally, selective replacement of the crystalline-Si in the source/drain regions with doped SiGe [7, 8] reduces the problems associated with insufficient electrical activation of dopants within the source/drain and excessive redistribution of dopant atoms from the source/drain into the channel region. This technique also offers the possibility of markedly reduced thermal budgets, which are likely necessary given the limitations on thermal processing of high-k gate dielectrics. Furthermore, the strategic placement of specific atomic species (i.e., F, C, etc.) at the transition regions between the source/drain and channel has also been correlated (via electrical measurements) to a reduction of short channel effects and leakage current [9]. These advanced device schemes, in conjunction with high-k gate dielectrics, promise to extend the life of planar CMOS technology and further improve the performance of advanced devices still in the research stage.
The implementation of such advanced process schemes (and subsequent maintenance of the manufacturing process) implies the need for radically advanced metrology techniques. Proper nanodevice engineering of the magnitude described herein requires metrology solutions that provide the location and elemental identification of the atoms within a device of interest. As yet, no technique has been reported that satisfies this need on a fully 3D scale. The recently introduced laser local electrode atom probe offers a potential method of meeting these criteria.
LEAP analysis
Atom-probe analysis techniques have existed for several decades [10]. While the use of atom probes by the metallurgical community is growing, this analysis technique has failed to gain a significant foothold within the semiconductor community. The reasons are many: slow collection rates, limited field of view, complex and nonstandard hardware configurations, difficulty in sample preparation, an inability to analyze samples from a silicon wafer, and so forth. While the implementation of LEAP [11, 12] has solved a number of these issues, resulting in widespread adoption of the atom probe throughout the metallurgical community, its use in semiconductor technology is still limited by the inability of a voltage-pulsed atom probe to analyze non-electrically conducting samples.
The atom probe achieves field evaporation through the application of a short-term (1 nsec), high-magnitude voltage pulse (up to ~3000V) to the base of a needle-shaped specimen [10, 13]. Quality data collection requires that this pulse propagate to the tip of the specimen without significant attenuation or temporal broadening. This has limited the analysis class of materials to those that are efficient electrical conductors, i.e., primarily metallic specimens. While Si-based structures of interest (primarily ultra-shallow junctions) composed entirely of heavily doped Si have been successfully analyzed in an atom-probe system that utilizes the local electrode geometry [14, 15], Si samples that are composed of lightly-doped or dielectric regions cannot be analyzed in a voltage-pulsed atom probe. The use of laser pulsing to achieve field evaporation of high-resistivity materials [16] shows that, in principle, there is considerable promise for the atom probe analysis of Si-based semiconductors and dielectrics.
This paper demonstrates, for the first time, the unparalleled capabilities of a system that uses both the local electrode geometry and very high-speed laser pulsing. The result is the efficient analysis of undoped Si-SiGe test structures and Si-SiGe test structures in which a single layer is doped. Correlations to SIMS data are also presented.
Experimental details
Two atom-probe systems were used in the experiments presented in this paper. The first atom probe is a voltage-pulsed system that employs the local electrode geometry. The temperature was controlled by a He cold finger and could be varied from room temperature down to ~20K. The base pressure of the analysis chamber was ≤10-8 Pascal during all analyses. No gases were intentionally introduced into the system during analysis. The voltage was pulsed at a duty cycle of 200kHz, and the pulse duration was ~1 nsec. The maximum DC voltage applied was 15kV with a maximum pulse voltage of 2.2kV. The second atom probe is identical to the first system except the field evaporation pulsing mechanism was achieved via a high-speed laser. The specimen base temperature was set at 40K for all laser-pulsed experiments. All concentrations in this paper are quoted in atomic percent.
Multilayer film stacks of Si-SiGe were deposited at a low enough temperature that minimal diffusion was expected to have occurred during the deposition process. These structures provide an excellent test stack for correlations between various analytical techniques since the layer transition should be nominally abrupt with the exception of the surface roughness of the layers (typically <1nm). As a comparison to local electrode atom probe data, SIMS analysis of the stacks was performed.
Results
Heavily doped Si. Voltage-pulsed atom probe analysis of undoped Si structures at cryogenic temperature generally is not possible. To create a direct comparison of voltage- and laser-pulsed modes, a heavily Sb-doped Si sample was analyzed in each system. The comparative mass spectra are shown in Fig. 1. It is clear that the laser system provides a considerably higher mass resolution and a greater sensitivity to dopants (i.e., a greater signal-to-noise ratio). The sensitivity of the voltage-pulsed and laser-pulsed systems are ~10,000:1 and ~50,000:1, respectively. The mass resolution improvement is a result of the shorter pulse duration for the laser as compared to the voltage-pulsed system and of the elimination of the “energy deficits” typically observed in voltage-pulsed systems [13]. In traditional systems, the pulse duration is the limiting factor in resolution. Because of the short duration of the laser pulse, the burden of mass resolution is shifted from the pulsing mechanism to the timing electronics of the detector.
Undoped Si-SiGe. The first high-resistivity structure analyzed was a series of Si-SiGe layers. None of the layers in the structure were doped. Field evaporation was not achieved in the voltage-pulsed system, regardless of the specimen temperature or pulse magnitude; width and duty cycle were not varied. Based on the results of others, it is unlikely that such a high-resistivity sample would see any improvement in field evaporation even if these parameters were changed [17, 18].
The same high-resistivity structure was analyzed in the laser-pulsed system. An atom map of the test structure is displayed in Fig. 2a, shown in 2D projection for clarity. The red dots represent the positions of Si atoms and the green dots the positions of Ge atoms. A 1D profile of the atomic composition in the analysis direction was extracted from this 3D image for comparison with traditional SIMS analysis. This 1D composition was achieved by creating a 15nm dia. analysis cylinder in the center region of the sample, shown by the dashed region in Fig. 2a. The cylinder is sliced into ~0.3nm sections, and the various atoms in each section are counted. The number of Ge atoms is divided by the total number of atoms in order to extract the %Ge concentration as a function of depth. The results, along with the corresponding SIMS analysis, are displayed in Fig. 2b. The composition of the various layers in the stack is evident: 60nm SiGe at 6% Ge; 35nm Si; and another SiGe layer with 9% Ge content.
The axes of Fig. 2b are registered in terms of both concentration and number of atoms per slice. By quantifying the data in terms of atoms, one may better appreciate the type of analysis capable with the atom probe. In the Si-only regions, there are 10,000±100 Si atoms in each slice of the 1D analysis cylinder. In the first and second SiGe layers, there are 600±25 Ge atoms and 900± 30 Ge atoms per slice, respectively. The statistical variation in Ge atoms is evident in the 1D composition profile and is the result of natural atomic fluctuations in any material. This natural fluctuation is well defined by a binomial distribution function. For dilute species (i.e., <10%), it can be reasonably approximated by a Poisson distribution. This natural variation can be reduced by increasing the number of atoms/slice, that is, by increasing the volume of the analysis slice. This can be achieved either by taking thicker slices, thereby reducing the spatial sensitivity in depth analysis (z direction), or by increasing the diameter of the analysis cylinder, which would reduce the spatial sensitivity in the x-y dimensions.
While the atom probe and SIMS data shown in Fig. 2b correlate well, there are two distinct differences between the analysis techniques. First, the slope “roll-off” of the SIMS analysis is approximately 7nm/decade as compared to the atom probe, which shows a slope roll-off of only 1nm/decade. Second, both analytical techniques detect an accumulation of Ge atoms at the SiGe-Si interface. The SIMS analysis quantifies this build-up at ~9% while the atom-probe analysis quantifies it at ~12.75%. The width of the build-up is 2nm according to the atom probe data. This length scale cannot be accurately quantified by SIMS due to the high slope roll-off at the peak. The magnitude of Ge accumulation can be quantified as a function of local position by creating several analysis cylinders of 5nm radius and examining the 1D composition profiles extracted from each cylinder. The results indicate that the Ge accumulation does vary locally; the peak %Ge concentration for each analysis cylinder is 11.8, 14.0, 14.8, 11.1, and 12.6. The average of the 5nm cylinders, 12.8%, correlates well with the 12.75% reported from the 15nm radius cylinder. Finally, because the location and identity of the atoms are known, the average chemical roughness of the SiGe-Si interface can be quantified by drawing an isochemical surface at this interface. The chemical roughness for this interface is 0.52nm.
Because of the nature of the two analytical techniques, it is likely that the analysis of the Si-SiGe interfaces is more accurate with the atom probe than with SIMS. Besides the additional information provided in the atom-probe data, it is difficult to achieve high-quality depth profiling with SIMS in multielement materials (such as SiGe) and at considerable depths of material (i.e., >50μm). The considerable amount of material inter-mixing that occurs as a function of the energetic ion beam used to sputter remove the material “blurs” the interface between such layers. Also, the SIMS sputter and ionization rates both vary with material composition. As a result, quantitative analysis of buried interfaces with SIMS is fundamentally challenging. Because the atom probe utilizes field evaporation and does not involve an energetic ion beam, parameters such as ionization yield, sputter rate, and ion-mixing do not factor in the analysis. While the theoretical evaporation field may vary between material species (e.g., W at 44V/nm vs. Al at 22V/nm), the field evaporation threshold for Si and Ge are both approximately 33V/nm [13]. Because of this, the resolution of the atom probe in the z direction is considerably more sensitive than that of SIMS, particularly in multicompositional materials.
Si-SiGe-Si:B. The third test structure analyzed was another series of Si-SiGe layers, except this time one of the Si layers was doped with B. Once again, field evaporation was not achieved in the voltage-pulsed system, regardless of the specimen temperature or pulse magnitude.
The laser-pulsed atom probe was again successful in analyzing this test structure. Figure 3a displays the 3D atom map of the multilayer test structure. Again, the Si atoms are displayed as red dots and the Ge atoms as green dots. The B dopant atoms are displayed as blue spheres, sized larger for emphasis. Figure 3b displays a 1D profile of the atomic composition of the test structure in the analysis direction. The first SiGe layer is 30nm thick with a Ge concentration of 28%. The second SiGe layer is 20nm thick and has a Ge concentration of 20%. The B-doped Si region is apparent with a peak concentration of 1021/cm3 (plotted on right axis in number of atoms/slice). There appears to be some diffusion of the dopant atoms from the Si region into the adjacent SiGe regions.
 Figure 4. Laser LEAP mass spectrum showing Si and B atom counts from Figure 3a. |
The mass spectrum from this analysis that corresponds to the B and Si sections is shown in Fig. 4. The Si atoms were detected primarily at a 2+ charge state, and the isotopes at 14, 14.5, and 15 are evident. The B atoms were detected in both the 1+ and 2+ charge states, as shown. The individual B isotopes are evident. It is worth noting that there is no overlap between the isotope peaks, indicating that the mass resolution of this system is more than adequate to differentiate atoms whose isotopes are closely spaced. The noise floor is approximately 10 counts while the primary Si peak is approaching one million counts. This indicates a dopant sensitivity down to ~1018/cm3.
Conclusion
Comparison to SIMS data shows that the laser atom probe provides superior interface analysis in these systems. The laser atom probe had a depth resolution of 1nm/decade vs. 7nm/decade for SIMS. With the addition of the local electrode geometry and the fast-pulsing laser scheme, the 3D atom probe provides a superior technique for the analysis of dopant atoms in Si- and SiGe-based materials. In addition to providing a more accurate quantification of the Ge accumulation at the interface, the atom-probe analysis allows one to gather information regarding the chemical roughness of the interface and discrete quantification of atomic distributions on a very local scale.
The ability to analyze Si and SiGe structures on a local scale, the atomic level, and in 3D offers a number of unique opportunities for characterizing Si-based devices. The most striking is that one can now analyze the 2D and 3D distributions of dopant atoms in actual devices. Monitoring of the lateral diffusion of dopants from the source/drain region into the channel region is one such possibility. One can also now monitor impurities at interfaces, such as those accumulated during the SiGe deposition process, on the scale of features within the device itself.
Focused ion beam lift-out techniques, already popularized for TEM analysis, allow one to form samples from small (<1×1μm) areas of a device wafer. The possibility of benignly extracting a small section of Si from a test structure region of the wafer at different process points would allow for the creation of distinct atomic maps of a specific region as the wafer is processed. The end result would be a process history of a specific device or region of the wafer. These sorts of advantages should lead to more rapid development and deployment of new process technologies as well as a rich source of information for failure analysis.
Acknowledgment
The authors acknowledge the assistance of S. Corcoran, Intel Corp.
References
1. International Technology Roadmap for Semiconductors, published by the Semiconductor Industry Association (http://www.itrs.net).
2. V. Meyssen, P. Stolk, J. van Berkum, W. van de Wijgert, R. Lindsay, C. Dachs, et al., “Si Front-end Processing: Physics and Technology of Dopant Defect Interactions III,” Materials Research Society, Pittsburgh, Vol. 669, p. J.3.5.1, 2001.
3. K. Thompson, J.H. Booske, Y.B. Gianchandani, R.F. Cooper, “Electromagnetic Annealing for the 100nm Technology Node,” Electron Device Letters, 23(3), p. 127, March 2002.
4. K. Thompson, J.H. Booske, R.F. Cooper, Y.B. Gianchandani, “Electromagnetic Fast Firing For Ultrashallow Junction Formation,” Trans. Semiconductor Mfg., 16(3), p. 460, Aug. 2003.
5. K. Thompson, J.H. Booske, R.L. Ives, J. Lohr, Y.A. Gorelov, K. Kajiwara, “Millisecond Microwave Annealing: Driving Microelectronics Nano,” J. Vac. Sci. & Tech. B, Vol. 23(3), pp. 970-978, May 2005.
6. M.L. Lee, E.A. Fitzgerald, M.T. Bulsara, M.T. Currie, A. Lochtefeld, “Strained Si, SiGe, and Ge channels for High-mobility Metal-Oxide-Semiconductor Field-effect Transistors,” J. Appl. Phys., 97 011101-1, 2005.
7. R. Loo, M. Caymax, P. Meunier-Beillard, I. Peytier, F. Holsteyns, S. Kubicek, et al., “A New Technique To Fabricate Ultra-Shallow-Junctions, Combining In Situ Vapour HCl Etching and In Situ Doped Epitaxial Sige Re-growth,” Appl. Surface Sci., 224, p. 63, 2004.
8. T. Yamashiro, T. Kikuchi, M. Ishii, F. Honma, M. Sakuraba, T. Matsuura, et al., “Super Self-aligned Technology of Ultra-shallow Junctions in MOSFETs Using Selective Si1-xGex CVD,” Materials Science and Engineering, B89 120, 2002.
9. P.R. Chidambaram et al., “Advanced Short-Time Thermal Processing for Si-based CMOS Devices,” ed. by F. Roozeboom, E.P. Gusev, L.J. Chen, M.C. Ozturk, D.L. Kwong, P.J. Timans, Electrochemical Society, Pennington, NJ, 14, pp. 93-98, 2003.
10. E.W. Müller, J.A. Panitz, S.B. McLane, “The Atom-probe Field Ion Microscope,” Rev. Sci. Instrum., Vol. 39, p. 83, 1968.
11. T.F. Kelly, D.J. Larson, “Local-electrode Atom Probes,” Mat. Char., Vol. 44/1-2, p. 59, 2000.
12. T.F. Kelly et al., “First Data from a Commercial Local Electrode Atom Probe,” Micro Microanal., 10, pp. 373-383, 2004.
13. M.K. Miller et al., Atom Probe Field Ion Microscopy, Oxford U. Press, 1996.
14. K. Thompson J. Booske, D. Larson, T. Kelly, “Three-dimensional Atom Mapping of Dopants in Si Nanostructures,” Appl. Phys. Lett. 87, p. 1, 2005.
15. K. Thompson, J.H. Bunton, T. Kelly, D. Larson, “Characterization of Ultralow-Energy Implants and Towards the Analysis of Three-Dimensional Dopant Distributions Using Three-Dimensional Atom-Probe Tomography,” J. Vac. Sci. & Tech. B, Vol. 24(1), pp. 421-428, January 2006.
16. G. Kellogg, T. Tsong, “Pulsed-laser Atom-probe Field-ion Microscopy,” J. Appl Phys. 51(2), p. 1184, February 1980.
17. T. Sakata, “Field Evaporation of Si Surfaces,” Surface Science, 130, p. 313-325, 1983.
18. T. Sakata, J.H. Block, “Field Evaporation of Silicon (111) Surfaces in the Presence of Hydrogen,” Surf. Sci. Lett., 116 L183-189, 1982.
K. Thompson received his PhD in electrical engineering from the U. of Wisconsin. He is a staff scientist at Imago Scientific Instruments Corp., 6300 Enterprise Lane, Madison, WI 53719; ph 608/274-6880.
D.J. Larson received his BS in physics and mathematics, and his MS and PhD in materials science from the U. of Wisconsin. He is a staff scientist at Imago Scientific Instruments Corp.
R.M. Ulfig received his BS in nuclear engineering and his MS in materials science and engineering from UW-Madison. He is a senior applications engineer at Imago Scientific Instruments Corp..
J.H. Bunton received his bachelors in materials science and engineering, and his masters in materials science from UW-Madison. He is a senior project engineer, R&D, at Imago Scientific Instruments.
T.F. Kelly received his BS from Northwestern Univesity and his PhD in materials science from MIT. He is the founder and CTO of Imago Scientific Instruments.