Thermochemical Reactions and Composition Analysis

Table 5.4 provides a comprehensive list of the accessible solid state reactions involved in the Ga–As–O phase diagram [33], [34] to aid us in determining the transformation phenomena that occurred during annealing. The critical temperature Tc is defined as the temperature at which the reaction becomes thermodynamically favorable, i.e., where the Gibbs free energy of formation (^△Gf) is less than zero; we obtained the values of Tc from experimental results reported in the literatures [24]-[26], [37]. We also calculated the value of

△Gf for each of these stoichiometric equations [35], [36]. If ^△Gf is greater than zero, the chemical reaction will not proceed; for example, the mechanisms associated with Ga2O formation are inhibited at RT, but an increase in temperature enhances the driving kinetics of

the reaction. Because various kinds of GaAs chemical products desorb and/or transform during high-temperature processing, we performed XPS and SIMS analyses to better characterize the electrical differences.

The As 2p3/2 and Ga 2p3/2 spectra in Fig. 5.11 indicate that the diffusion of GaAs chemical species was enhanced within the Al2O3 gate dielectric after thermal annealing; we extracted the contribution of each chemical component. In the following discussion we denote the N2-annealed capacitor that underwent sulfide treatment as the “SN-sample” and denote the O2-annealed capacitors without and with sulfide treatment as the “O-sample” and

“SO-sample,” respectively. The As 2p3/2 spectra revealed an increased incorporation of arsenic oxides, including As2O5, As2O3, and AsOx, relative to their content in the as-deposited Al2O3

film. We observed similar behavior in the Ga 2p3/2 spectra: both the Ga2O3 and GaOx oxides were slightly enriched after annealing, with the SN-sample having relative higher concentrations. From the atomic quantification, Table 5.5 indicates that the amount of diffused arsenic oxides was larger, by about one order of magnitude, than that of the gallium oxides in the annealed Al2O3; this behavior is similar to that of the oxide layer formed during thermal oxidation of a GaAs surface [39]. In addition, only the O-sample after annealing at 600 °C presented an oxygen-enriched Al2Ox film; we attribute this finding to a portion of the sulfur atoms filling a number of vacancies inside the high-k gate dielectric, thereby avoiding over-oxidization in the sulfidized samples during PDA.

Table 5.4 (part A) indicates that the value of Tc for the desorption of arsenic oxides ranges from 150 to 350 °C; these transformations are thermodynamically favorable at RT. In contrast, the higher thermal stability of the Ga-based oxide compounds inhibits the reduction of nonvolatile Ga2O3 into Ga2O (Table 5.4, part D); consequently, its value of Tc is higher (400–600 °C). Because of the low sublimation point of the oxides of arsenic, volatile As2O3

oxide first reacts with either the GaAs substrate or interstitial Ga atoms to form a Ga2O3 layer and segregated As close to the interface. It is possible that the highly stable Ga2O3 oxide also

reacts with the substrate again to produce non-stoichiometric Ga2O along with the As co-products under high-temperature annealing at 600 °C. Most of these As species—in the form of elemental As and/or gaseous As4—are oxidized during the desorption event and then trapped inside the dielectric film. Note that Ga2O3 formation can arise through several transformation paths; for example: As2O3 + 2GaAs → Ga2O3 + 4As or As4. Because the O2

environment provides additional reaction paths (Table 5.4, part B), the content of chemical products is likely to be somewhat different to that in the N2 annealing case. This situation can be explained in terms of the chemical reaction principle that kinetics naturally drive several oxidation mechanisms—e.g., the oxidation of GaAs, As, and As4—resulting in a deceleration of the rate of transformation of As2Ox compounds as well as the generation of As defects in the O2 ambient [41]. In addition, even if molecular oxygen prefers to react with Ga atoms over As atoms, the As species will still have a higher probability of oxidation because they are more readily desorbed [42]. When the content of arsenic oxides increases in the upper area of the dielectric film they can act as a block, which in turn hinders the diffusion of oxygen into the dielectric–substrate region. These phenomena result in our observation of an enriched amount of As2Ox with a low Ga2Ox concentration for both the SO- and O-samples, a finding that is opposite to the behavior of the SN-sample. Another interesting phenomenon visible in the 2p3/2 spectra is that PDA under N2 resulted in the formation of some metallic As, whereas some elemental Ga appeared after PDA under O2. This finding provides direct evidence of the fast generation of As species in the N2 ambient. On the other hand, as far as the source of these Ga atoms is concerned, we speculate that the oxidation of the Al2O3/GaAs interface and decomposition of the GaAs substrate are responsible for the supply of free Ga atoms [26], [41]. It should be pointed out that as regarding the reactions related to Al and these in-diffused Ga or As oxides, most possible reaction we presume is the reaction of Al2O3 with Ga2O3, due to the higher thermal stability relative to As2O3. However, in our study the highest RTA temperature is 600 °C, probably not enough to make Al react with Ga oxides [43]; these

thermal processes are excluded here and still need further investigation.

Figure 5.12 presents the highly bulk-sensitive As 3d and Ga 3d photoemissions; Table 5.6 provides the chemical bonding ratio of the contributed chemical components for convenient characterization. We observe that annealing at 600 °C further decreased the percentage of interfacial As presented in the overall As 3d spectra, with respect to that in the as-deposited Al2O3/GaAs structures (Fig. 5.5, Table 5.3). Even though the desorption event contributed to the formation of either As or As4, most of the As atoms escaped into the atmosphere or were further oxidized into various oxidation states. A substantial amount of Ga2Ox oxides was generated accordingly through chemical transformation. We also observed that the additional sulfide pretreatment of the SO-sample made it (relative to the O-sample) more highly resistant to the growth of these undesirable components. Fig. 5.13 illustrates the respective SIMS depth profiles; the thickness of the Al2O3 thin film, which was deposited at 300 °C for 50 cycles, was ca. 50 (±3) Å after PDA at 600 °C. Of primary interest is the distribution of the As, Ga, and S species in the overlying Al2O3 high-k film. We found that a higher concentration of As was presented around the bottom of the Al2O3 bulk film, coinciding with the diffusion of As-enriched oxides and the interfacial arsenic layer. Another feature is the observation of a small tails of Ga species at a depth of ca. 10–30 Å for the SO- and O-samples only, probably related to a trace of out-diffused Ga elements, the detailed mechanism of which requires further investigation. On the other hand, the S signal detected in the O-sample was due to mass interference of molecular oxygen in the SIMS analysis of the oxide film. We believe that the S atoms did exist even after high-temperature processing, as determined after subtracting the induced signal of the mass interference used as the baseline background. The corresponding XPS analysis in Table IV indicates that the sulfidized samples were relatively less oxygen-excessive (Al/O ratio = 1.78) with respect to the HCl-last sample (Al/O ratio = 2.2). Meanwhile, the smaller C–V hysteresis width after sulfidization is indicative of less charge trapping in the Al2O3 dielectric (not shown here). Thus, we conclude

that most of the S atoms diffused into the high-k film and probably occupied either the vacancy sites or pre-existing defects; as a result, this process facilitated the improvement in the quality of the Al2O3 gate dielectrics as well as their electrical performance.