Achieving robust ferroelectric polarization in uncapped Hf_0.5Zr_0.5O₂ thin films by incorporating Al dopant Open Access

As the demand for miniaturization, faster operation speed, and cost efficiency continues to drive advances in integrated circuits (ICs), modern complementary-metal–oxide–semiconductor (CMOS) technology has achieved a significant increase in the transistor density. This advance promotes the development of nanoelectronics with more efficient gate modulation and lower power consumption. In addressing these challenges, two-dimensional (2D) materials have emerged as promising candidates for next-generation nanoelectronics due to their atomic-level thickness, tunable bandgap, and high carrier mobility.^1,2 These unique properties can be further optimized by integrating a ferroelectric (FE) gate, which enhances the electrical, optical, and mechanical properties of 2D materials.³ Among various ferroelectric materials incorporated into 2D devices, such as poly(vinylidene fluoride-trifluoroethylene) [P(VDF-TrFE)]⁴ and PbZr_xTi_1−xO₃ (PZT),⁵ HfO₂-based ferroelectric films have recently gained attention as a superior option. The excellent compatibility with CMOS technology and the scalability down to sub-10 nm thickness of HfO₂-based FE films make them particularly advantageous for 2D device development.^6,7 The ability of HfO₂-based ferroelectrics to maintain strong ferroelectric properties at reduced dimensions benefits large-scale production and facilitates the integration of ferroelectric-enhanced 2D devices, such as phototransistors, memories, and neuromorphic devices, into the IC industry.^8,9 This compatibility and scalability promise HfO₂-based FE films as key materials in advancing the performance and manufacturability of 2D-based nanoelectronics.

The ferroelectric field effect transistor (FeFET) structure is widely considered a superior choice for ferroelectric-enhanced 2D devices due to its nondestructive readout capability and smaller overall footprint. As a result, the 2D/ferroelectric (FE) heterojunction structure is commonly employed in such devices. However, for HfO₂-based FE films, a top electrode (TE) is typically required to achieve strong ferroelectricity. The ferroelectric behavior of HfO₂ films has been attributed to the formation of the metastable orthorhombic Pca2₁ phase (o-phase), which is stabilized only under specific conditions.^10,11 Several factors, including doping,¹² annealing,¹³ and the presence of TE,¹⁴ need to be optimized to induce ferroelectricity. Among these, TE plays a critical role by applying mechanical stress that promotes the formation of the desired ferroelectric o-phase during annealing. However, the mechanical stress induced by the TE depends on the material of the TE, limiting the choice of suitable TEs for HfO₂-based FE systems.¹⁵ After the post-metallization annealing (PMA) process, a complex interface with abundant oxygen vacancies may be created between FE and electrode, when the conventional electrodes, such as TiN, are used. This is supposed to be a major contributing factor for the wake-up effect of the HfO₂ FE films.¹⁶ Furthermore, integrating an electrode between 2D materials and HfO₂ ferroelectric films introduces additional interfaces and complicates the fabrication of ferroelectric-enhanced 2D devices, further increasing the complexity of their integration into advanced CMOS manufacturing. In contrast, a post-deposition annealing (PDA) process, allowing the ferroelectricity in HfO₂ films without a TE, has shown great potential to reduce the wake-up effect and simplifies the fabrication process by removing the requirement for a TE. This approach not only enhances the development of ferroelectric-enhanced 2D devices but also facilitates their seamless integration into existing IC processes.

Despite the advantages of uncapped HfO₂ FE films, their poor ferroelectric performance hinders broader applications. Notably, a robust remanent polarization (2P_r) of approximately 50.4 μC/cm² has been achieved in Al-doped HfO₂ films by optimizing the distribution of Al dopants.¹⁷ The incorporation of suitable Al dopants is believed to introduce microscopic stress within the films, promoting the formation of the FE o-phase, akin to the effect of a TE.¹⁸ However, this process requires a high thermal budget of 800 °C. On the other hand, Hf_0.5Zr_0.5O₂ (HZO) FE films can be processed at lower annealing temperatures (400–600 °C) to exhibit strong ferroelectricity.¹⁹ However, uncapped HZO films are supposed to exhibit poor ferroelectricity. Recent advancements have demonstrated significant improvements in the 2P_r of the uncapped HZO films through techniques such as atomic layer annealing and introducing dielectric layers.^20,21 Despite these enhancements, the endurance of these films remains a critical area of concern. For instance, uncapped Al:HfO₂ films have been observed to suffer from severe fatigue, with their 2P_r significantly decreasing after 10⁶ cycles.²² Recent research demonstrated that incorporating dopants into HZO films, such as La, Al, and Y, enhanced the ferroelectric performances and endurance of the HZO film.^23–25 However, research studies remain to focus on the capped films, and research focused on optimizing HfO₂ films in a top electrode-free configuration remains limited. Accordingly, introducing dopants into HZO films may be a promising strategy for developing HfO₂-based FE films in a top electrode-free configuration. This approach not only enhances the ferroelectric properties but also facilitates the integration of these materials into advanced IC processes without the complications associated with a top electrode.

In this work, we present a straightforward method for developing uncapped HfO₂-based FE films by incorporating Al dopants into HZO films. We examined the effects of doping concentration and annealing temperature on the ferroelectric performances of the uncapped samples. The incorporation of Al dopants was confirmed by x-ray photoelectron spectroscopy (XPS). The uncapped Al:HZO films with a doping ratio of 1:24 demonstrated a high 2P_r of ∼39.5 μC/cm² at a low thermal budget of 500 °C. Additionally, the endurance performance of the Al films was significantly improved, showing a nearly wake-up-free behavior and robust polarization stability up to 10⁸ cycles.

The key process flow for metal–ferroelectric–silicon (MFS) capacitors is illustrated in Fig. 1(a). Al:HZO films were deposited on highly doped p-type (100) silicon wafers at 280 °C using plasma-enhanced atomic layer deposition (PEALD). Prior to deposition, the native SiO₂ layer on the Si wafers was removed with a hydrofluoric acid (HF) solution. The precursors used for Hf, Zr, and Al were [(CH₃)₂N]₄Hf (TDMAH), [(CH₃)₂N]₄Zr (TDMAZ), and (CH₃)₃Al (TMA), respectively, while oxygen plasma served as the oxygen source. HfO₂ and ZrO₂ layers were alternately deposited with a total of 96 ALD cycles for all samples. The Al cycles were set as 2, 4, and 6, achieving different Al:(Hf + Zr) cycle ratios of 1:48, 1:24, and 1:16, respectively. The final thickness of all samples was ∼13.5 nm, as measured by an inline spectral ellipsometer. Subsequent to deposition, rapid thermal annealing was conducted in an oxygen environment at temperatures ranging from 400 to 700 °C for 30 s. A 50 nm platinum layer was then deposited using ion sputtering with shadow mask patterning techniques (Φ400 μm), resulting in the formation of a Pt/Al:HZO/Si MFS capacitor, referred to as uncapped samples. The chemical compositions of the films were analyzed using x-ray photoelectron spectroscopy (XPS, PHI-5400, PE). The ferroelectric properties were assessed at room temperature by employing a ferroelectric tester (TF Analyzer 2000, Aixacct Systems) with triangular pulses at a frequency of 1 kHz. Grazing incidence x-ray diffraction (GIXRD) was performed using an x-ray diffractometer (Xpert Pro, Panalytical) with an incidence angle of 0.5° to analyze the crystalline structures of the Al:HZO films. Additionally, piezoresponse force microscopy (PFM) imaging was conducted using a scanning probe microscope (Dimension Icon, Bruker).

Figures 1(b) and 1(c) present the x-ray photoelectron spectroscopy (XPS) spectra of Al 2p and O 1s, respectively. The peak positions are calibrated by referencing the C 1s peak of adventitious carbon at 284.8 eV.²⁶ The Al 2p peaks, observed around 73.90 eV in Fig. 1(b) for all doped samples, confirm the successful incorporation of Al dopants into the HZO films. The intensity of the Al 2p peaks increases with the cycle ratio, indicating a corresponding rise in dopant concentration. However, due to the nonuniform distribution of Al dopants within the 13.5 nm thick films, XPS cannot detect all Al dopants accurately. The typical probing depth of XPS is approximately 6–9 nm,²⁷ which limits precise quantification of the doping concentration. Previous reports suggest that the doping concentration is proportional to the doping cycle ratio.²⁸ Thus, we use the cycle ratio as a proxy for doping concentration in this study for convenience. Notably, the binding energy (BE) of Al exhibits a slight shift to higher values with an increasing cycle ratio. This shift is less than 0.1 eV and may result from experimental uncertainty. Similar negligible shifts are observed in the O 1s spectra (<0.1 eV). Consequently, these minor variations, which may be attributable to measurement errors, should not be considered significant in the discussion. The O 1s spectra are consistent across all samples, revealing two distinct O shoulders upon deconvolution. The dominant shoulder is located around 529.98 eV, corresponding to the typical binding energy of lattice oxygen atoms in Hf–O bonds.²⁹ A minor shoulder, accounting for approximately 5% in all samples, appears around 531.8 eV, potentially indicating the presence of oxygen vacancies. However, the Hf 4f core-level patterns presented in Fig. S1 in the supplementary material are only deconvoluted into a spin–orbit doublet, suggesting that the films are well oxidized. Therefore, the O 1s peak at 531.8 eV is likely attributed to C–O or O–C = O components from adventitious carbon coming from exposure to air.²⁷ The use of oxygen-rich annealing may facilitate the recovery of oxygen vacancies resulting from the Al dopant incorporation. Consequently, less oxygen vacancies should be formed in all samples after annealing.

Figure 2 presents the polarization–voltage (P–V) loops for the uncapped samples with different doping concentrations. Prior to measurement, all samples underwent a wake-up process. Due to the use of the MFS structure in this study, higher applied voltages are required to switch the ferroelectric domains compared to that in the metal–ferroelectric–metal (MFM) configuration. Significant differences in ferroelectric performance are observed between the samples before and after the incorporation of Al dopants. For the undoped samples, hysteresis loops were observed with a maximum 2P_r of ∼9.5 μC/cm². However, no clear switching current peak was observed in the current–voltage (I–V) curves shown in Fig. S2(a) in the supplementary material. The observed hysteresis loops in the undoped samples can be attributed to significant leakage currents at elevated applied voltages, as evidenced by the rise in current in the I–V curves. Consequently, the ferroelectric characteristics were negligible in the uncapped HZO samples, consistent with results in previous works. In contrast, pronounced hysteresis loops were found in the doped samples, as illustrated in Figs. 2(b)–2(d). Corresponding switching peaks were also observed in the I–V curves presented in Figs. S2(b)–S2(d) in the supplementary material. These results verify the presence of ferroelectricity in the uncapped samples incorporated with Al dopants.

Figure 2(e) summarizes the 2P_r of the uncapped samples subjected to different annealing temperatures. Despite the fact that the doping concentrations were different, the 2P_r exhibited a consistent trend that 2Pr initially increased first and then decreased with increasing doping concentration. The largest 2P_r was obtained at an annealing temperature of 600 °C. However, the 2P_r of samples annealed under 500 °C was found to be comparable to that annealed at 600 °C. This indicates that uncapped Al:HZO films can achieve significant ferroelectricity even at low temperatures. Regarding the influence of doping concentration, the 2P_r was also found to increase first and then decrease with the doping concentration. The sample with a doping ratio of 1:24 exhibited the largest 2P_r of 39.5 μC/cm² when annealed at 600 °C, which is comparable to the ferroelectric performance of HZO films annealed with a TE (30–50 μC/cm²). The robust ferroelectricity in the 1:24 is confirmed by the PUND measurement as shown in Fig. S3 in the supplementary material. Besides, the 2P_r decreased a little after 10³ s at 75 °C as shown in Fig. S4 in the supplementary material, demonstrating the good retention performance of the 1:24 samples. These results demonstrate that Al dopants can be used to achieve good ferroelectricity in uncapped HZO films, even under a low thermal budget.

To investigate the effects of doping concentration on the crystal structure, the GIXRD patterns in the 2θ range of 20°–40° for the uncapped samples with different cycle ratios are presented in Fig. 3(a). The annealing temperature was set at 500 °C for all samples. The undoped samples displayed two prominent peaks at 28.8° and 31.4°, typical peaks of the monoclinic (m) phase, indicating that the m-phase was dominant in the undoped samples. Due to the absence of TEs, the ferroelectric o-phase is not easily generated in the undoped HZO samples, which is consistent with their paraelectric performances. In the case of the 1:48 samples, the peak at 28.8° was still identified, implying that the m-phase is still formed. However, the peak at 31° became dominant with larger intensity, demonstrating the generation of o-phase or t-phase. Due to the clear ferroelectric properties, the presence of the o-phase is expected in the 1:48 samples. Consequently, the peak near 31° should be assigned to the o(111) reflection. The emergence of the o-phase resulted in a reduction in the content of the m-phase in the 1:48 samples. Moreover, the characteristic peak at 28.8° was completely absent when the cycle ratio further increased to 1:24 and 1:16, which implies the suppression of the m-phase by Al dopant. These results favor the fact that the formation of m-phase was suppressed by high concentrations of Al dopants.

For both the 1:24 and 1:16 samples, a single prominent diffraction peak near 31° is observed, which is attributable to the overlapping o(111) and t(101) reflections due to their structural similarities.¹³ Due to the comparatively poor ferroelectric properties of the 1:16 films, a higher tetragonal (t) phase content is inferred in the 1:16 samples relative to the 1:24 films. The slight positive shift (0.2°) of this diffraction peak further suggests an increased proportion of the t-phase in the 1:16 samples. Additionally, the t-phase appears to exhibit greater stability within the 1:16 films, as depicted in Fig. 3(b). Notably absent in the 1:16 films, the peak at 28.8° emerged in the 1:24 samples upon annealing at 700 °C, indicating the formation of the m-phase in the 1:24 films. This emergence of the m-phase corresponds with a reduction in the o-phase, in accordance with the observed decrease in 2P_r for the 1:24 samples. Furthermore, the m-phase was only generated in the 1:16 films at an elevated annealing temperature of 800 °C, underscoring the greater stability of the t-phase in heavily doped films. This finding suggests that Al dopants may suppress the transition from the t-phase to the m-phase.

To further confirm the ferroelectricity in the uncapped samples, PFM was carried on the 1:24 HZO films annealed at 500 °C as shown in Fig. 4. To perform the PFM measurements, a bias voltage of −10 V was first applied to a 6 × 6 μm² square, then a bias voltage of +10 V was applied to a 4 × 4 μm² inside concentric square, and, finally, a 8 × 8 μm² concentric square was swept to obtain the local phase and amplitude images. The out-of-plane PFM phase image in Fig. 4(a) displayed high dark and bright contrasts in the scanned squares, which suggests the presence of opposite polarization directions in the dark and bright regions. Clear phase boundaries between dark and bright regions can be observed in Fig. S5 in the supplementary material. The out-of-plane PFM amplitude image in Fig. 4(b) exhibited clear boundaries between opposite bias regions, indicating the existence of ferroelectric domains with opposite polarization directions in 1:24 Al:HZO films. These results demonstrate good ferroelectricity in the 1:24 Al:HZO films.

In addition to ferroelectric performance, good endurance characteristics are essential for HfO₂-based FE films. The endurance behavior of the 1:24 film annealed at 500 °C was evaluated by applying a cyclic electric field, as depicted in Fig. 5(a). The cyclic electric field was maintained at an amplitude of 6 V, with a frequency of 1 kHz for the wake-up process, and 100 kHz for the fatigue process, as illustrated in Fig. 5(a). The read pulse of 8 V was applied during both processes. For the wake-up process, no significant increase in 2P_r was observed in the 1:24 uncapped sample after 10⁵ cycles. A potential explanation may lie in the enhanced stability of the t-phase, as suggested by Fig. 3. The depinning of domains and the phase transition from the m/t-phase to the o-phase during cycling can lead to an increase in 2P_r, which is often cited as the origin of the wake-up effect in HfO₂-based FE films.^30,31 The stabilization of the t-phase could suppress this phase transition, reducing the typical wake-up response. Instead, the 2P_r was observed to be slightly reduced from 42.1 to 39.5 μC/cm², as presented in Fig. 5(b). This reduction in 2P_r is likely to result from a decrease in leakage current, as indicated in the inset of Fig. 5(b). The cyclic electric field appears to promote the diffusion and redistribution of oxygen vacancies into the bulk of the film, which may contribute to the reduced leakage current after cycling.^31,32 Additionally, the shifting of switching current peaks toward the lower field region, which may also result from the oxygen vacancy migration and redistribution, further supports this hypothesis. Thus, the initial switching behavior of the fresh sample may account for the observed changes in 2P_r.

For the fatigue process, the 2P_r decreased from 39.5 to 34.1 μC/cm², reflecting a reduction of approximately 14% after 10⁸ cycles, as shown in Fig. 5(c). Unlike the slit-up of switching current peaks observed in HZO films after fatigue,^31,33 the switching current peaks in the 1:24 sample shifted slightly toward the negative region, resulting in only minor distortions in the P–V loop. This negative shift in switching current peaks may be due to the generation of defects, particularly oxygen vacancies, during the long field cycling. These newly generated oxygen vacancies likely migrate toward the top/bottom electrode interface under the influence of the cyclic field, resulting in a localized bias field that could partially shift the switching peaks.^33,34 The nearly wake-up-free behavior and the robust 2P_r retention demonstrate the good endurance performance of the uncapped films.

In this study, we demonstrated the effective development of uncapped HfO₂-based FE films by incorporating Al dopants into Hf_0.5Zr_0.5O₂ films. Through systematic optimization of doping concentration and annealing conditions, a remanent polarization of approximately 39.5 μC/cm² was achieved in the 1:24 Al:HZO films at a low thermal budget of 500 °C. The ferroelectricity in the 1:24 uncapped films was confirmed by the PFM. The XRD results indicated that the Al doping suppresses the formation of the monoclinic phase and enhances the stability of the tetragonal phase. Furthermore, the 1:24 Al:HZO film exhibits minimal wake-up effects and substantial retention of polarization up to 10⁸ cycles, underscoring the good endurance performances of the uncapped films by incorporating the Al dopants. Our findings facilitate the development of HfO₂-based FE films without the need for top electrodes, which benefits their applications in advanced nanoelectronic devices and integrated circuits, and promotes their compatibility with existing CMOS processes.

See the supplementary material for the XPS of Hf, I–V loops measured at 8 V after wake-up, the PUND results, the retention results measured at different temperatures, and the phase regions of the out-of-plane PFM phase image.

This work was financially supported by Natural Science Basic Research Program of Shaanxi Province (No. 2024JC-YBQN-0447) and the Fundamental Research Funds for the Central Universities (No. XRT062023001).

The authors have no conflicts to disclose.

Xin Liu: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Weidong Zhao: Data curation (equal); Investigation (equal). Jiawei Wang: Data curation (equal); Investigation (equal). Lulu Yao: Data curation (equal); Investigation (equal). Man Ding: Methodology (equal); Project administration (equal). Yonghong Cheng: Supervision (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.