Polycrystallization effects on the nanoscale electrical properties of high-k dielectrics
- 3.8k Downloads
In this study, atomic force microscopy-related techniques have been used to investigate, at the nanoscale, how the polycrystallization of an Al2O3-based gate stack, after a thermal annealing process, affects the variability of its electrical properties. The impact of an electrical stress on the electrical conduction and the charge trapping of amorphous and polycrystalline Al2O3 layers have been also analyzed.
atomic force microscopy
conductive atomic force microscope
contact potential difference
Kelvin probe force microscope
root mean square
rapid thermal process
ramped voltage stresses
transmission electron microscopy
ultra high vacuum.
To reduce the excess of gate leakage currents in metal-oxide-semiconductor (MOS) devices, the ultra thin SiO2 gate oxide is replaced by other high-k dielectric materials . However, high-k-based devices still show some drawbacks, and therefore to have a better knowledge of their properties and to improve their performance, a detailed electrical characterization is required. Many researches have been devoted to the study of the electrical characteristics of high-k gate dielectrics, mainly using standard wafer level characterization techniques on fully processed MOS capacitors or transistors [1, 2, 3, 4]. However, since the lateral dimensions of complementary MOS devices are shrinking to a few tens of nanometers or below, for a detailed and profound characterization, advanced methods with a large lateral resolution are required. In this direction, conductive atomic force microscope (CAFM), as demonstrated for SiO2 and other insulators [5, 6, 7, 8, 9, 10, 11, 12, 13, 14], is a very promising tool which allows for a nanometer-resolved characterization of the electrical and topographical properties of the gate oxide. Characterization at the nanoscale allows us to study which factors determine the electrical properties of the dielectric stack, and details on how they affect them. For example, some manufacturing processes (such as high-temperature annealing) can alter their electrical properties because of the polycrystallization of the high-k dielectric, which can affect its electrical homogeneity . Recently, the CAFM has been started to be used to evaluate the electrical conduction of polycrystalline high-k dielectrics. While in some polycrystalline materials the electrical conduction seems to be mainly related to the bulk of grains , in others, current can flow preferentially through grain boundaries (GBs) [17, 18, 19, 20]. Since this topic can be crucial for the successful inclusion of high-k dielectrics in electron devices, in this study, AFM-related techniques have been used to investigate, at the nanoscale, the effect of the high-k material polycrystallization (derived from an annealing process) on the conductivity and charge trapping of Al2O3-based stacks for Flash memories.
Gate stacks, which consist of a nominal 10-nm-thick Al2O3 layer and a 1-nm-thick SiO2 interface layer on top of a p-type Silicon substrate, have been analyzed. After the Al2O3 deposition, some of the samples were annealed by rapid thermal process (RTP) in nitrogen at 750 or at 950°C. The electrical properties of the stack were measured using a Dimension 3100 AFM provided with CAFM and Kelvin probe force microscope (KPFM) modules. The CAFM allows us to obtain, simultaneously to the topography, current images of the structures, by means of applying a constant voltage between the tip and the sample during a surface scan, and I-V characteristics on fixed locations, by means of applying ramped voltage tests. The KPFM allows us to obtain, simultaneously to the topography, images of the contact potential difference (CPD) between the tip and the substrate. For all the current and CPD measurements, Si tips with a Pt-Ir or diamond coating were used. Topographic images have been obtained in tapping mode using Silicon ultra sharp tips without coating, which offer a better spatial resolution. Other techniques such as transmission electron microscopy (TEM) and X-ray reflectometry have been also used to perform a physical analysis of the structures.
Thicknesses of the Al2O3 and SiO2 layers obtained from X-ray reflectrometry on the amorphous and polycrystalline samples
The impact of the polycrystallization of the Al2O3 layer on the electrical conduction of the gate stack has been analyzed at the nanoscale from current and CPD images obtained on fresh structures (before an electrical stress). Figure 1e,f shows two current images obtained on the amorphous and polycrystalline sample, respectively, at 10.25 V (their average and rms values are included in the figure). Note that smaller currents (in average) are measured in the polycrystalline stack. However, since the EOT of the polycrystalline sample is smaller (see Table 1), its lower conductivity can only be attributed to the crystallinity of the stack and not to the reduction of the oxide thickness. Figure 1f also shows that the rms value of the current and, therefore, the electrical inhomogeneity of the polycrystalline stack is larger. Both samples have also been analyzed with KPFM , which can provide information about the presence of charge and trapping centers in the stack. Figure 1 shows two CPD images obtained on the amorphous (g) and polycrystalline sample (h). Their rms value is also included. Again, after crystallization, the deviation increases, suggesting larger inhomogeneities in its trapping properties.
It is important to emphasize that the correlation of the leaky positions with the grain boundaries is a qualitative result, since the resolution in these experiments is not high enough to resolve grain boundaries. This is because the CAFM measurements presented in this section have been performed with Si tips coated with a metallic layer in ambient conditions, drastically reducing its lateral resolution to approximately 20 nm . Note, however, that other experiments, with sufficient resolution, have shown the relation between leaky sites and grain boundaries [27, 28]. The section "Influence of the environment on the resolution of grain boundaries" will be devoted to investigate how the CAFM resolution can be improved.
Comparing Figure 3a,d, which corresponds to the first image (fresh area) measured on amorphous and polycrystalline structures, respectively, results similar to those shown in the previous section are obtained. On polycrystalline samples, background conduction is smaller (table of Figure 3, first scan). However, the leaky sites of polycrystalline structures (spot S3, S4, and S5) have a larger conductivity compared to those of amorphous samples (S1 and S2). As discussed in the previous section, the larger current differences in the polycrystalline structure could be attributed to the differences in the conductivities between the crystals (background) and grain boundaries (leaky sites).
The effect of the stress has been analyzed from the images measured during the zoom-outs. On the amorphous sample (Figure 3b), the central area (which was previously pre-stressed, Figure 3a) shows smaller currents than the rest of the scanned region. A similar behavior can be observed in Figure 3c, where three concentric areas can be distinguished: a first central area with the smallest current value (three scans), another second area with a larger current (two scans), and the peripheral and the most conductive area (only one scan, that is, a fresh area). The quantitative values of the background current on the three concentric areas are shown in the table of Figure 3. In the amorphous sample, the background current decreases significantly as the stress proceeds, making the structure less conductive. This behavior, as already pointed out for SiO2 layers  or high-k dielectrics, can be related to negative charge trapping in the native defects or in traps generated during the stress. In the case of polycrystalline structures (Figure 3d,e,f and table), the decrease in the background conductivity is less important when compared to amorphous samples. This result suggests a smaller impact of the stress at the positions where crystals are present.
Although, in polycrystalline samples, regions with background currents (which can be probably related to positions with a crystal under the CAFM tip) seem to be more resistive and robust to an electrical stress than amorphous oxides, this behavior cannot be extended to the weak spots (leaky sites). As an example, the table in Figure 3 shows the maximum current in different spots and its evolution with the stress on amorphous (spots S1 and S2) and polycrystalline samples (S3, S4, and S5). Note that the weak spots in the polycrystalline structure show, before the stress ("first scan"), larger leakage currents compared with the amorphous gate oxide. However, after the stress ("second scan" and "third scan"), the reduction of current through these spots is larger than those in amorphous structures. Therefore, initially, the leaky sites of polycrystalline samples are, from an electrical point of view, weaker (their conductivity is higher) than those in amorphous oxides (because the dielectric is thinner or because of the presence of defects that enhance tunneling). However, as the stress proceeds, a larger amount of charge is trapped in the as-grown or generated defects on the leaky positions, leading to a higher reduction of the current compared to amorphous oxides. Since these leaky regions could be related to the grain boundaries between nanocrystals, charge trapping (in as-grown or generated defects) mainly occurs at those locations, leading to a higher reduction of the conductivity compared to the background areas. In amorphous samples, no distinction between crystals and grain boundaries can be observed, and so trapping is randomly distributed in the gate area.
Influence of the environment on the resolution of grain boundaries
Some authors have suggested that, when working with a CAFM in air, the tip-sample contact area increases, probably due to the presence of a water layer on the sample (and, therefore, a water meniscus between the tip and the surface), which can reduce the lateral resolution of the measurements [19, 30, 31]. Since the grain boundaries width is in the range of few nanometers [32, 33] and the CAFM lateral resolution in air when using metal coated tips is about 10 nm, grain boundaries could not always be resolved. This would explain why in the sections "As-grown dielectrics" and "Stressed dielectrics" when working with a CAFM in ambient conditions, a point-to-point correlation between the topographical and electrical properties (in particular, between the leaky sites and the grain boundaries position) was not possible. For this reason, when a higher resolution is needed, CAFM in vacuum or UHV has been used [12, 24, 34, 35]. In this section, the impact of environmental conditions on the CAFM electrical resolution for the study of polycrystalline structures will be analyzed.
The conductivity and charge trapping of amorphous and polycrystalline Al2O3 layers stacks for memory applications have been studied before and after an electrical stress at nanometer scale using AFM-related techniques in ambient conditions. The current measurements obtained with CAFM before an electrical stress show that the polycrystallization of the Al2O3 leads to a smaller average and a larger spatial inhomogeneity of the sample conductivity. A statistical analysis of the current images registered on polycrystalline samples has been compared to the measurements obtained with TEM, showing that the mean size of the less conductive areas is similar to the dimensions of the crystals. Therefore, the regions with a smaller conductivity could be related to the grains of the polycrystalline structure: the polycrystals are more insulating whereas the grain boundaries show a larger conductivity. The charge-trapping properties of amorphous and polycrystalline samples were also investigated after an electrical stress. The results suggest that, although the crystals are more resistive and robust (from an electrical point of view) than the amorphous oxide, the grain boundaries of the polycrystalline samples seem to be more sensitive to an electrical stress than those of the non-crystallized structures: grain boundaries would initially act as conductive paths, but would favor a faster charge trapping. Therefore, polycrystallization strongly contributes to the inhomogeneity increase of the conduction and trapping properties of the stacks, which could reduce the reliability of the MOS devices due to the weaker dielectric strength of the grain boundaries. Finally, the influence of the environment conditions on the study of polycrystalline high-k dielectrics was also analyzed. The results demonstrate that the reduction of the water meniscus can be a determinant factor for a precise study in detail on the electrical properties of the grain boundaries.
This study has been partially supported by the Spanish MICINN (TEC2007-61294/MIC research project and HA2007-0029 Integrated Action), and the "La Caixa" and Deutscher Akademischer Austausch Dienst (DAAD) pre-doctoral fellowships program. The authors are also grateful to P. Michalowski and L. Wilde from Fraunhofer Centre Nanoelectronic Technology (Dresden) for help in carrying out TEM experiments, to G. Benstetter and D.P. Yu from Hochschule Deggendorf and Peking University, respectively, for hosting M. Lanza to carry out some of the presented AFM experiments and to G. Jaschke and S. Teichert (Qimonda, Germany) and G. Bersuker (Sematech, USA) for sample provision. The authors also want to acknowledge T. Schroeder (IHP) for hosting V. Iglesias in their facilities in Frankfurt Oder (Germany). The authors are also indebted to them for valuable discussions.
- 2.Degraeve R, Aoulaiche M, Kaczer B, Roussel P, Kauerauf T, Sahhaf S, Groeseneken G: Review of reliability issues in high-k/metal gate stacks. 15th Physical and Failure Analysis of Integrated Circuits (IPFA) 2008, 1–6.Google Scholar
- 3.Kittl JA, Opsomer K, Popovici M, Menou N, Kaczer B, Wang XP, Adelmann C, Pawlak MA, Tomida K, Rothschild A, Govoreanu B, Degraeve R, Schaekers M, Zahid M, Delabie A, Meersschaut J, Polspoel W, Clima S, Pourtois G, Knaepen W, Detavernier C, Afanas'ev VV, Blomberg T, Pierreux D, Swerts J, Fischer P, Maes JW, Manger D, Vandervorst W, Conard T, Franquet A, Favia P, Bender H, Brijs B, Van Elshocht S, Jurczak M, Van Houdt J, Wouters DJ: High-k dielectrics for future generation memory devices (Invited Paper). Microelectron Eng 2009, 86: 1789–1795.CrossRefGoogle Scholar
- 10.Yanev V, Rommel M, Lemberger M, Petersen S, Amon B, Erlbacher T, Bauer AJ, Ryssel H, Paskaleva A, Weinreich W, Fachmann C, Heitmann J, Schroeder U: Tunneling atomic-force microscopy as a highly sensitive mapping tool for the characterization of film morphology in thin high-k dielectrics. Appl Phys Lett 2008, 92: 252910.CrossRefGoogle Scholar
- 11.Weinreich W, Wilde L, Kucher P, Lemberger M, Yanev V, Rommel M, Bauer AJ, Erben E, Heitmann J, Schroder U, Oberbeck L: Correlation of microscopic and macroscopic electrical characteristics of high-k ZrSi x O 2-x thin films using tunneling atomic force microscopy. J Vac Sci Technol B 2009, 27: 364–368.CrossRefGoogle Scholar
- 16.Menou N, Wang XP, Kaczer B, Polspoel W, Popovici M, Opsomer K, Pawlak MA, Knaepen W, Detavernier C, Blomberg T, Pierreux D, Swerts J, Maes JW, Favia P, Bender H, Brijs B, Vandervorst W, Van Elshocht S, Wouters DJ, Biesemans S, Kittl JA: 0.5 nm EOT low leakage ALD SrTiO 3 on TiN MIM capacitors for DRAM applications. Int Electron Devices Meeting Tech Dig 2008, 929.Google Scholar
- 17.Lanza M, Porti M, Nafria M, Aymerich X, Benstetter G, Lodermeier E, Ranzinger H, Jaschke G, Teichert S, Wilde L, Michalowski PP: Conductivity and charge trapping in amorphous and polycrystalline Al 2 O 3 based devices studied with AFM related techniques. IEEE Trans Nanotechnol 2010, 99: 1–9.Google Scholar
- 18.Yanev V, Rommel M, Lemberger M, Petersen S, Amon B, Erlbacher T, Bauer AJ, Ryssel H, Paskaleva A, Weinreich W, Fachmann C, Heitmann J, Schroeder U: Tunneling atomic-force microscopy as a highly sensitive mapping tool for the characterization of film morphology in thin high-k dielectrics. Appl Phys Lett 2008, 92: 252910.CrossRefGoogle Scholar
- 21.Lanza M, Porti M, Nafria M, Aymerich X, Benstetter G, Lodermeier E, Ranzinger H, Jaschke G, Teichert S, Wilde L, Michalowski P: Crystallization and silicon diffusion nanoscale effects on the electrical properties of Al 2 O 3 based devices. Microelectron Eng 2009, 86: 1921–1924.CrossRefGoogle Scholar
- 23.Kim Y, Gebara G, Freiler M, Barnett J, Riley D, Chen J, Torres K, JaeEun L, Foran B, Shaapur F, Agarwal A, Lysaght P, Brown GA, Young C, Borthakur S, Hong-Jyh L, Nguyen B, Zeitzoff P, Bersuker G, Derro D, Bergmann R, Murto RW, Hou A, Huff HR, Shero E, Pomarede C, Givens M, Mazanez M, Werkhoven C: Conventional n-channel MOSFET devices using single layer HfO 2 and ZrO 2 as high-k gate dielectrics with polysilicon gate electrode. Proceedings of the IEEE International Electron Devices Meeting 20.2.1 2001, 455.Google Scholar
- 30.Kremmer S, Peissl S, Teichert C, Kuchar F: Conducting atomic-force microscopy investigations on thin silicon gate oxides: influence of tip shape and humidity. Proceedins of the 28th International Symposium of Testing and Failure Analysis, EDFAS 2002, 473–482.Google Scholar
- 31.Israelachvili JN: Intermolecular and Surface Forces. London: Academic Press; 1992.Google Scholar
- 32.Bersuker G, Yum J, Iglesias V, Porti M, Nafría M, McKenna K, Shluger A, Kirsch P, Jammy R: Grain boundary-driven leakage path formation in HfO2 dielectrics. Proceedings of the 41th Solid-State Device Research Conference ESSDERC 2010, 333–336. ISSN 1930–8876 ISSN 1930-8876Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.