Thermally induced permittivity enhancement in La-doped ZrO2 grown by atomic layer deposition on Ge(100)

La-doped ZrO 2 thin ﬁlms grown by O 3 -based atomic layer deposition directly on Ge (cid:1) 100 (cid:2) exhibit a dielectric constant of 29. Upon annealing in N 2 at 400 °C, a high (cid:1) value (cid:2) 40 is extracted for ﬁlm thickness below 15 nm. Compositional depth proﬁling allows to correlate this observation with a remarkable Ge interdiffusion from the substrate which is consistent with the stabilization of the tetragonal ZrO 2 phase. Ge interaction with the oxide stack and the formation of a germanate-like interfacial region, which acts as an electrical passivation for the Ge surface, are also investigated.

La-doped ZrO 2 thin films grown by O 3 -based atomic layer deposition directly on Ge͑100͒ exhibit a dielectric constant of 29. Upon annealing in N 2 at 400°C, a high value Ͼ40 is extracted for film thickness below 15 nm. Compositional depth profiling allows to correlate this observation with a remarkable Ge interdiffusion from the substrate which is consistent with the stabilization of the tetragonal ZrO 2 phase. Ge interaction with the oxide stack and the formation of a germanate-like interfacial region, which acts as an electrical passivation for the Ge surface, are also investigated. © 2009 American Institute of Physics. ͓doi:10.1063/1.3227669͔ As alternative to Si in high-speed logic devices, Ge is widely considered due to its higher carrier mobilities. 1,2 Coupling Ge channel with high dielectric constant ͑͒ material is a promising strategy for ultrascaled logic devices. Among highoxides, ZrO 2 has been proved to be a promising insulator. Numerous results were presented for ZrO 2 / Ge stacks grown by sputtering, 3,4 atomic layer deposition ͑ALD͒ 5-7 and molecular beam deposition ͑MBD͒. 8 Interface engineering still remains an open issue as the Ge/oxide interface exhibits an unacceptably large density of interface traps ͑D it ͒, which can be related to Ge dangling bond interface defects. 9,10 A broad range of solutions was considered, e.g. Ge controlled oxidation 6,11 and deposition of passivating interfacial layer ͑IL͒. 12 Recently, a study on ZrO 2 / GeO 2 stacks prepared by atomic oxygen-assisted MBD ͑Ref. 8͒ correlates a very high value ͑ϳ44͒ with the promotion of tetragonal ZrO 2 . Such a large is there associated with Ge incorporation into the ZrO 2 due to thermal decomposition of the GeO 2 IL. In our work, we focus on O 3 -based ALD of La-doped ZrO 2 ͑La-ZrO 2 ͒ on Ge͑100͒ therein targeting the growth of a high-material through a direct deposition approach without IL. As-grown and annealed La-ZrO 2 films were characterized by means of ͑i͒ high resolution transmission electron microscopy ͑HRTEM͒ coupled to electron energy-loss spectroscopy ͑EELS͒, and selected area electron diffraction ͑SAED͒, ͑ii͒ x-ray photoemission spectroscopy ͑XPS͒, ͑iii͒ time of flight secondary ion mass spectroscopy ͑ToF SIMS͒, and ͑iv͒ capacitance-voltage ͑C-V͒ measurements. La-ZrO 2 films were grown at 300°C in a Savannah 200 ͑Cambridge Nanotech Inc.͒ ALD reactor on n-type Ge͑100͒. Native GeO 2 was removed in a diluted HF solution ͑1:25͒. La incorporation in ZrO 2 ͑ϳ5% mean value͒ should enhance the value through doping induced effect. 13,14 La can also favor the interfacial La-O-Ge bonds, thus reducing Ge suboxides and electronic traps. 15,16 In this respect, O 3 oxidation was demonstrated to opportunely passivate Ge interface. 17,18 This motivates our choice to start the ALD with a La 2 O 3 step and to choose O 3 ͑ϳ200 g / Nm 3 ͒ as oxidizer. Precursors were ͑ i PrCp͒ 3 La and ͑MeCp͒ 2 ZrMe͑OMe͒ ͑SAFC Hitech͒. The ALD cycle number varied from 8 to 80, stack thickness was estimated using spectroscopic ellipsometry and x-ray reflectivity ͑XRR͒ in the 3-30 nm range. La doping was tuned by the ALD La/Zr ratio. Postgrowth rapid thermal annealing ͑RTA͒ was performed for 60 s in N 2 at 400°C.
Room temperature C-V measurements were performed on metal insulator semiconductor capacitors; capacitance equivalent oxide thickness ͑CET͒ was determined from the accumulation capacitance ͑C acc ͒ at 100 kHz. Figure 1͑a͒ illustrates multifrequency C-V curves obtained for an asgrown La-ZrO 2 film. Only weak frequency dispersion is observed both in accumulation and depletion, indicating a relatively low D it ͑ϳ8 Ϯ 1 ϫ 10 11 eV −1 cm −2 evaluated with the Hill-Coleman method 19 at 500 kHz͒. For the as-grown stack a =18Ϯ 1 can be calculated from the C acc by applying a single-layer capacitor model. Figure 1͑b͒ displays multifrequency C-V curves obtained after annealing. A clockwise hysteresis of 550 ͑as-grown͒ and 850 ͑annealed͒ mV is measured. C-V curves appear more stretched and only a moderate increase of the D it up to ϳ2 Ϯ 1 ϫ 10 12 eV −1 cm −2 is observed. After annealing, the C acc is noticeably higher than the one related to the as-grown state thus suggesting a significant impact of the RTA on the dielectric properties. Applying a capacitor model a =21Ϯ 1 for the annealed stack is calculated.
A structural and chemical investigation of the Ge͑100͒/La-ZrO 2 interface and within the film was per-a͒ Author to whom correspondence should be addressed. Electronic mail: luca.lamagna@mdm.infm.it. formed with HRTEM and EELS on the 14 nm thick sample before and after RTA ͑for details on TEM see Ref.
20͒. As-grown films are crystalline for all the analyzed thicknesses, as also proved by x-ray diffraction ͑not shown͒, with a cubic ͑c-͒ and/or tetragonal ͑t-͒ phase of ZrO 2 which cannot be univocally assigned ͓Fig. 3͑c͔͒. Differently from La 0.25 Zr 0.75 O 2−␦ / Si ͑Ref. 21͒ the very low La content does not promote the amorphization of the ZrO 2 matrix. A direct contact between the La-ZrO 2 crystals and Ge͑100͒ is observed also after RTA ͓Fig. 2͑a͔͒. Consequently, there is no IL from the structural viewpoint. A chemical interfacial region is revealed as evidenced on the La, Zr, O, and Ge elemental STEM-EELS profiles as a function of the distance from the substrate by scanning a 1 nm probe diameter up to the film surface ͓Fig. 2͑b͔͒. The Zr and O profiles increase just after the substrate and beyond few nm become constant with a Zr/O ratio ϳ0.5. The La profile has a peak just above the interface, the intensity at the maximum being La/ Zr ϳ 0.1 and then continuously decreases. The La, Zr, and O profiles are very similar in the as-grown and annealed state. First, Ge is present above the interface with a non-negligible content in the region of the La peak. After RTA, Ge profile has a similar shape but spreads over a larger distance from the substrate. STEM-EELS profiles reveal a La and Ge rich region near the interface different from the following region of nearly constant composition. Hence, the as-grown and annealed films can be described by a two layers structure from the chemical viewpoint. From the Ge profiles, we can infer that a 2 nm thick IL ͑as-grown͒ and a nearly 2.5 nm thick IL ͑after RTA͒ is present ͓dots and lines respectively, Fig. 2͑b͔͒. The chemical bonding at the interface were probed by XPS analyses of the Ge 2p line performed on 3 nm thick as-grown and annealed samples ͓Fig. 2͑c͔͒ in normal photoemission geometry through a standard Mg K␣ ͑1253.6 eV͒ radiation source. Both the Ge 2p lines can be deconvoluted in two different contributions from the elemental Ge bonding of the substrate ͑Ge 0 ͒ and Ge-O like bonding ͑Ge ‫ء‬ ͒. The mutual shift between the two components is 2.3 eV, which deviates from the expected shift due to GeO 2 and suggests the formation of a germanate-like region before and after RTA. 22 The increased intensity of the Ge 0 contribution after annealing reflects a thermally induced reduction of the stack thickness. XRR analysis ͑same samples, not shown͒ also revealed a reduction of total thickness and a corresponding 10% electronic density increase upon RTA related to a densification of the whole layer. The formation of a germanate IL, associated to the use of O 3 , might help to keep the D it value relatively low.
Electrical results are discussed based on the previous analysis. Figure 3͑a͒ shows a linear dependence of the CET versus La-ZrO 2 physical thickness plot, leading to =29Ϯ 1 for the as-grown films. This value resembles the one expected for c-and is higher than the one of monoclinic ͑m-͒ZrO 2 ͑ϳ20͒, thermodynamically stable on Si͑100͒. 23,24 An active role in determining the of as-grown films might be played by the large ionic radius ͑i rad ϳ 1.23 Å͒ of La atoms, inserted into the ZrO 2 lattice, softening the ZrO 2 IRactive modes. 21,25 The intercept of the linear fit to the CET data confirms the presence of an IL between La-ZrO 2 and Ge͑100͒ and clarifies the higher values of La-ZrO 2 compared to the one of the stack. Given a 2 nm thick IL, as revealed by STEM-EELS, a low IL ͑ IL =4͒ is calculated. The use of this IL in a two capacitors model ͓inset of Fig.  3͑a͔͒ results in a =28Ϯ 1 for the as-grown La-ZrO 2 . After RTA, CET versus La-ZrO 2 thickness plot ͓Fig. 3͑b͔͒ cannot be interpolated by a linear fit. This finding can be clarified by structural and chemical modifications within the films depending on the thickness as illustrated in Figs. 3͑c͒ and 4. Figure 3͑c͒ shows the SAED analyses performed on the thickest ͑29 nm͒ and the thinnest ͑14 nm͒ stacks before and after RTA. The m-ZrO 2 phase is revealed in the thick ͑19 and 29 nm͒ as-grown samples and such m-fraction increases upon annealing. Differently, 14 nm thick film is free of m-ZrO 2 . A linear estimation of the , for the two thickest annealed stacks results in a =34Ϯ 1 ͓Fig. 3͑b͔͒; given a 2.5 nm thick IL ͑from STEM-EELS͒ a IL = 6.2 is calculated from the fit intercept. This value is in agreement with the presence of a comparatively denser germanate compound at the interface or with values proposed for germanates. 22,26 The use of these IL parameters in a two capacitors model provides a =32Ϯ 1 for the thick annealed films ͑19 and 29 nm͒ whereas for the thinner annealed La-ZrO 2 the is remarkably increased up to ϳ42Ϯ 1. Such result confirms that the 14 nm thick film is structurally different. To assess the elemental composition after RTA, ToF SIMS depth profiles were acquired in negative polarity on an ION-TOF IV instrument using Cs + ions at 1 keV for sputtering and Ga + ions at 25 keV for analysis. Upon RTA, LaO and ZrO 2 signals are not modified ͓Fig. 4͑a͔͒ whereas Ge diffusion throughout the oxide is evidenced by the 70 Ge signal. The level of Ge should be low because it was not observed by STEM-EELS with a sensitivity limit for Ge around 3-5%. Ge can be inferred to diffuse up to the top of the La-ZrO 2 layer directly from the substrate; its diffusion strongly depends on film thickness as shown by the different 70 Ge profiles ͓Figs. 4͑a͒-4͑c͔͒. Actually, the presence of a Ge-poor region is more pronounced in thicker films and therefore these dissimilar chemical profiles might be associated with variable crystalline mixtures in the films. Hence, Ge diffusion is expected to cause the absence of m-ZrO 2 phase in the 14 nm sample and thus the increase of La-ZrO 2 after RTA. Interdiffused Ge might play a role in stabilizing the t-ZrO 2 phase, as proposed in Refs. 8 and 13, since the of the thicker samples also slightly increases after RTA despite the development of the m-phase. Indeed, pronounced electrical differences are accompanied in the thinnest annealed sample by a remarkable Ge diffusion. It should be noted that, when ZrO 2 was deposited on a La 2 O 3 IL, only a ϳ 32 for ZrO 2 was measured. 22 Differently from Refs. 8, 22, and 27, La-ZrO 2 was directly deposited on Ge͑100͒ without IL. The observed enhancement is consistent with the prediction by Fischer et al. 13 that, upon introduction of Ge substitutional atoms, t-ZrO 2 can become energetically more favorable than m-or c-phases. Ge atoms, with a small i rad ͑ϳ0.53 Å͒, are readily incorporated in the thin La-ZrO 2 layer therein promoting the stabilization of t-ZrO 2 . Although the revealed Ge concentration in La-ZrO 2 is lower than the one theoretically predicted, this amount is enough to stabilize t-ZrO 2 . 28 In summary, we showed that the combination of a direct O 3 -based ALD of La-ZrO 2 and RTA can lead to a Ͼ 40. It is proved that Ge atoms are supplied by the substrate and penetrate into the oxide upon annealing. The observed large value after RTA is consistent with Ge-induced stabilization of the t-ZrO 2 phase in thin film. Ge diffusion occurs without affecting the interfacial details qualified by a germanate-like region and an acceptably low D it value.