Enhanced Performance of FeFET Gate Stack via Heterogeneously co-doped Ferroelectric HfO$_2$ Films

This study demonstrates that spatially controlled heterogeneous co-doping of Zr and Al in HfO$_2$ films via atomic layer deposition effectively tunes phase evolution and domain nucleation, thereby significantly enhancing the remanent polarization and endurance of FeFET gate stacks.

The Gist

The Big Picture: Building a Better "Memory Switch"

Imagine you are trying to build a super-fast, super-small computer that never forgets anything, even when the power is turned off. To do this, scientists use a special type of switch called a FeFET (Ferroelectric Field-Effect Transistor). Think of this switch as a tiny light switch that can stay "on" or "off" without needing electricity to hold the position.

The heart of this switch is a thin layer of a material called Hafnium Oxide (HfO₂). However, pure Hafnium Oxide isn't naturally a good switch. To make it work, scientists add "impurities" (called dopants) to it, like adding spices to a soup to change its flavor.

For a long time, they used just one type of spice: either Zirconium (Zr) or Aluminum (Al).

• Zirconium makes the switch very sensitive (it turns on easily), but it wears out quickly, like a cheap door hinge that breaks after a few thousand uses.
• Aluminum makes the switch very durable (it lasts a long time), but it's a bit stiff and hard to turn on.

The Innovation: The "Layer Cake" Strategy

The researchers in this paper asked a clever question: "What if we don't just mix the spices randomly, but arrange them in specific layers?"

Instead of mixing Zirconium and Aluminum together like a smoothie, they baked them into a layer cake. They used a precise machine (Atomic Layer Deposition) to stack thin layers of Zirconium-rich material and Aluminum-rich material on top of each other in different orders.

They tested three main "cake recipes" for a 10-nanometer-thick film:

1. Recipe A: Zirconium at the bottom, Aluminum in the middle and top.
2. Recipe B: Zirconium in the center, Aluminum on top and bottom.
3. Recipe C: Zirconium at the top, Aluminum at the bottom and middle.

What They Discovered: The "Goldilocks" Zone

The results were fascinating. The order of the layers changed the internal structure of the material, which is like changing the crystal structure of ice.

1. The "Center" is King for Durability
When they put the Zirconium layer right in the middle of the film (Recipe B), something magical happened. The material became incredibly tough. It could be switched on and off millions of times without breaking.

• The Analogy: Think of the Zirconium in the middle as a "shock absorber" in a car. It absorbs the stress of the switching, protecting the rest of the structure. Even though it created a slightly less "sensitive" crystal structure (more of the "monoclinic" phase), the trade-off was worth it because the device lasted so much longer.

2. The "Bottom" is a Disaster Zone
When they put the Zirconium layer at the very bottom (touching the silicon base), the device failed quickly.

• The Analogy: Imagine building a house on a foundation that is slowly dissolving. The Zirconium atoms were "leaking" into the silicon base (the foundation), weakening the whole structure. This caused electricity to leak out (like a leaky roof), and the switch stopped working after just a few thousand tries.

3. The "Top" is Okay, but Not Great
Putting Zirconium at the top was better than the bottom, but it didn't offer the same durability boost as putting it in the center.

The Secret Sauce: Why It Works

The researchers found that by controlling where the Zirconium sits, they could control how the material "freezes" (crystallizes) when heated.

• If Zirconium is in the middle, it crystallizes freely, creating a tough, stable structure.
• If Zirconium is at the bottom, it messes up the interface with the silicon, causing leaks.

The Bottom Line

This paper proves that how you arrange the ingredients matters just as much as what ingredients you use.

By creating a "heterogeneous" (mixed but layered) stack, the scientists managed to get the best of both worlds:

• High Performance: The switch still turns on easily (good memory).
• Long Life: The switch lasts for millions of cycles (high endurance).

The Takeaway for the Future:
To build the next generation of computers that can remember everything instantly and last forever, we shouldn't just mix our materials randomly. We need to be like master chefs, carefully layering our ingredients to create the perfect structure. Specifically, never put the Zirconium layer right against the silicon base, or the whole device will leak and fail.

This simple change in design could help us build computers that are faster, use less energy, and never lose your data.

Technical Summary

1. Problem Statement

Ferroelectric Field-Effect Transistors (FeFETs) based on doped Hafnium Oxide (HfO$_2$) are promising candidates for non-volatile memory and neuromorphic computing due to their scalability and CMOS compatibility. However, current HfO$_2$-based FeFETs face significant reliability challenges, including:

• Small Memory Windows: Limited remanent polarization ($2P_r$).
• Poor Endurance: Rapid fatigue (degradation of polarization) during switching cycles.
• Leakage Current: High leakage leading to early breakdown.
• Phase Instability: The ferroelectric orthorhombic (o-) phase is metastable and competes with the non-ferroelectric monoclinic (m-) phase.

Existing strategies often rely on uniform doping (e.g., Zr or Al throughout the film), which fails to simultaneously optimize phase stabilization and interface quality. The authors aim to address these limitations by exploring spatially controlled heterogeneous co-doping in Metal-Ferroelectric-Insulator-Semiconductor (MFIS) gate stacks.

2. Methodology

The study utilized a systematic approach combining material synthesis, structural characterization, and electrical testing:

• Sample Fabrication:

  • Substrate: Highly p-doped Si wafers with a chemically grown SiO$_2$ and a thermally grown SiO$_2$/SiON interfacial layer (IL).
  • Deposition: 10 nm HfO$_2$ films were deposited via Atomic Layer Deposition (ALD) at 250°C.
  • Doping Strategy: The films were engineered with distinct vertical arrangements of Zirconium (Zr) and Aluminum (Al) dopants.
    • Control Groups: Uniformly mono-doped Hf$_x$Zr$_{1-x}$O$_2$ (HZO) and Hf$_x$Al$_{1-x}$O$_y$ (HAO).
    • Experimental Groups: Heterogeneously co-doped stacks with Zr/Al layers placed at the Bottom (C), Center (B), or Top (A) of the HfO$_2$ film (e.g., IL/HZO/HAO/HAO, IL/HAO/HZO/HAO, etc.).
  • Post-Processing: A 10 nm TiN capping layer was deposited, followed by Rapid Thermal Processing (RTP) at 800°C for 20 seconds to induce crystallization.

• Characterization Techniques:

  • Structural: Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) to verify dopant depth profiles; Grazing Incidence X-ray Diffraction (GIXRD) to analyze phase composition (o-phase vs. m-phase) and crystallographic texture.
  • Electrical:
    • Fatigue Measurements (FM): Cycling at $\pm$5 V, 10 kHz to assess endurance.
    • Positive-Up Negative-Down (PUND) Tests: To extract true remanent polarization ($P_r$) and leakage current ($J$) while distinguishing switching current from leakage.

3. Key Contributions

• Novel Doping Architecture: The paper introduces a "heterogeneous co-doping" strategy where the vertical sequence of Zr and Al dopants is precisely controlled via ALD cycle sequencing, rather than using uniform mixtures.
• Mechanism Elucidation: It establishes a direct correlation between the spatial location of dopants and the resulting crystallization dynamics, specifically how interface constraints and local crystallization temperatures influence phase competition.
• Interface Engineering: The study identifies the critical role of the interface between the interfacial layer (SiON) and the ferroelectric layer, revealing that the specific dopant at this interface dictates leakage and reliability.

4. Key Results

Structural Findings (GIXRD & ToF-SIMS)

• Phase Control: The vertical arrangement of dopants significantly alters the ratio of the ferroelectric orthorhombic (o-) phase to the monoclinic (m-) phase.
  • Placing HZO (Zr-doped) in the center of the film promotes the formation of the m-phase (monoclinic) due to reduced interfacial confinement and lower nucleation temperatures (~300°C for HZO vs. >650°C for HAO).
  • Placing HZO at the top or bottom interfaces enhances the stabilization of the o-phase due to stronger capping effects from adjacent layers.
• Texture: Samples with HZO at the top or HAO mono-doped films showed increased intensity in the o{200} peak, suggesting improved crystallographic texture or semi-epitaxial growth (templating effect).

Electrical Findings

• Endurance vs. Polarization Trade-off:
  • Mono-doped HZO: High initial $2P_r$ but very poor endurance (rapid fatigue).
  • Mono-doped HAO: Lower $2P_r$ but excellent endurance.
  • Co-doped Stacks: Achieved a balance, but performance was highly dependent on dopant position.
• The "Bottom Interface" Criticality:
  • Failure Mode: Any stack with HZO directly contacting the SiON interfacial layer (e.g., IL/HZO/...) exhibited rapid degradation, high leakage, and early breakdown.
  • Cause: ToF-SIMS confirmed Zr$^{4+}$ diffusion from the HZO layer into the SiON layer, degrading the insulator quality and increasing leakage current.
• Optimal Configuration:
  • The IL/HAO/HAO/HZO stack (where HZO is at the top and HAO is at the bottom interface) demonstrated the best overall performance.
  • Results: It combined high remanent polarization ($P_r$) with no degradation up to $10^4$ cycles.
  • Switching Behavior: This configuration showed the most concentrated switching current and the lowest coercive voltage, attributed to higher o{200} phase fractions and reduced imprint (likely due to Al-induced oxygen vacancies counteracting built-in fields).

5. Significance

This work demonstrates that spatial engineering of dopants is a superior strategy to simple compositional tuning for FeFET gate stacks.

• Reliability Breakthrough: By avoiding the direct SiON/HZO interface, the study solves a major leakage and endurance bottleneck, enabling FeFETs to meet the rigorous requirements for in-memory and neuromorphic computing.
• Design Rule: The findings provide a clear design rule for future MFIS devices: Place Al-doped layers at the bottom interface to stabilize the interface and prevent diffusion, while placing Zr-doped layers at the top or center to optimize ferroelectric phase stabilization and polarization magnitude.
• Scalability: The approach leverages standard ALD processes, making it highly compatible with existing CMOS manufacturing lines for next-generation non-volatile memory.