*3.1. XPS Analyses*

To evaluate the chemical bonding states of various stacked gate dielectrics, XPS measurements were carried out. Figure 2 displays the In 3d, P 2p, and O 1s XPS spectra of three samples with various stacked gate dielectrics. It can be noted that In 3d spectra can be deconvoluted into four components that represent the InP, InPO4, In(PO3)3, and In2O3, respectively. The relative intensity values of the different components have been extracted and are shown in Figure 3a. For S2 and S3, the contents of In(PO3)3 and In2O3 shows a decreasing trend, indicating that the ALD-derived Al2O3 passivation layers prior to Sm2O3 deposition can significantly prohibit the formation of In and P suboxides, which can be attributed to the interface cleaning function of plasma O2 [21]. Compared to S2, the peak areas of InPO4 corresponding to S1 and S3 remain approximate at about 7.89% and 5.20%, which is much lower than that of S2 (19.41%), indicating that double deposition of ALD-derived Al2O3 may accelerate the diffusion of oxygen in the substrate and the formation of indium phosphate. During the secondary deposition of Al2O3, more oxygen vacancies may generate, which can be ascribed to plasma O2 acting as an oxygen source, and promote the oxygen interdiffusion between Al2O3 passivation layers and the InP substrate. In(PO3)3 can react with In to produce InP and InPO4 using the following reaction Equation [22].

$$\text{3In} + 4\text{In} (\text{PO}\_3)\_3 \to \text{3InP} + \text{9InPO}\_4 \tag{1}$$

**Figure 2.** (**a**) In 3d, (**b**) P 2p, and (**c**) O 1s XPS spectra for S1, S2, and S3 sample.

**Figure 3.** Peak area ratio histograms of (**a**) In 3d, (**b**) P 2p, and (**c**) O 1s spectra.

More importantly, sample S3 shows a tendency to decrease the content of In(PO3)3 and AlPO4 compared to sample S2, which can give a detailed illustration from the phenomenon of P 2p spectral changes. As shown in Figure 2b, it can be noted that P 2p spectra can be deconvoluted into four components, which represent InP, InPO4, In(PO3)3, and AlPO4, respectively. No P2O5 was detected in all samples, which can be attributed to the fact that gaseous P2O5 generated during the deposition can easily diffuse through the defects in the gate dielectric [22]. The peak area ratio of In(PO3)3 for S2 and S3 showed a significant decreasing trend compared to S1, indicating that Al2O3 prior to the deposition of Sm2O3 gate dielectric can inhibit the formation of P-O bound states and improve the interfacial quality. The detection of AlPO4 in P 2p spectra can be attributed to the reaction equation described below [23].

$$4\text{Al} + 7\text{O}\_2 + 2\text{In}(\text{PO}\_3)\_3 + 2\text{InP} \to 4\text{AlPO}\_4 + 4\text{InPO}\_4\tag{2}$$

Based on the mentioned reaction above, it can be inferred that two depositions of Al2O3 passivation layers increase the formation of AlPO4, which is confirmed by the change in peak area ratio shown in Figure 3c. In order to systematically explore the interfacial chemistry of various stacked gate dielectrics, O 1s spectra were investigated and are shown in Figure 2c. O 1s spectra can be deconvoluted into Sm2O3, Al2O3, InPO4, In(PO3)3, and AlPO4. According to the reaction Equation (2), in the plasma O2 atmosphere, In(PO3)3 can react with O2 to produce AlPO4 and InPO4 and leads to the disappearance of In(PO3)3. In agreement with the previous In 3d and P 2p spectra, S1 has the largest In(PO3)3 content, leading to a decrease in interfacial quality and deterioration of electrical properties. Meanwhile, AlPO4 of S2 is the highest, originating from the second deposition of Al2O3. For S3 sample, the contents of InPO4, AlPO4, and In2O3 were significantly controlled, indicating that the addition of an ALD-derived Al2O3 layer prior to the deposition of Sm2O3 gate dielectric could reduce the generation of suboxides and improve the interfacial quality.
