pith. sign in

arxiv: 2604.15131 · v1 · submitted 2026-04-16 · ❄️ cond-mat.mtrl-sci

Fully Atomic-Layer-Deposited Vertical Complementary FeRAM with Ultra-High 2Pr > 100 uC/cm2 and High Endurance > 1E10 cycles

Renhao Xue , Ruizhan Yan , Mansun Chan , Xiwen Liu This is my paper

Pith reviewed 2026-05-10 10:33 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords vertical FeRAMcomplementary polarizationatomic layer depositionhafnium oxide ferroelectricdifferential readouthigh endurancecrosspoint array
0
0 comments X

The pith

A vertical stack of two FeRAM cells with complementary polarization sums their responses to deliver an effective window above 100 microcoulombs per square centimeter.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that a fully atomic-layer-deposited vertical complementary FeRAM overcomes the polarization ceiling of thin hafnium-oxide layers by placing two ferroelectric cells on top of each other. Each cell is written so its polarization points opposite to the other, turning a normal single-layer readout into a differential summation of the two polarization charges. This produces a larger effective charge signal without raising the voltage needed to switch either layer or increasing the cell area. The devices reach above 100 microcoulombs per square centimeter differential polarization, keep more than 90 microcoulombs per square centimeter after ten billion cycles, and show stable retention and disturb immunity in small selector-free arrays.

Core claim

The central claim is that an all-ALD-grown vertical stack of top and bottom FeRAM cells, operated with a complementary polarization write-read scheme, converts individual layer responses into a summed differential polarization. Viewed from top to bottom, an up-down pair stores logic 1 and a down-up pair stores logic 0, yielding effective 2Pr above 100 uC/cm2 that remains above 90 uC/cm2 after 1E10 cycles without breakdown, together with retention longer than 1E4 s at 85 C and effective polarization above 80 uC/cm2 after 1E6 disturb pulses under a V/3 scheme.

What carries the argument

The vertical complementary FeRAM architecture in which top and bottom stacked ferroelectric cells hold opposite polarization states so their charges add during differential readout.

If this is right

  • The differential summation increases the memory window per unit area without raising switching voltage or cell footprint.
  • Selector-free crosspoint arrays become feasible, as shown by successful 5 by 5 array operation.
  • High endurance above 1E10 cycles and retention longer than 1E4 s at 85 C become available for embedded non-volatile memory.
  • The all-ALD process allows the stacked structure to be grown uniformly, supporting future scaling.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If interface defects between the two cells can be further reduced, the same complementary scheme might be extended to three or more stacked layers.
  • The approach could be tested with other ferroelectric materials that also face thickness-limited polarization.
  • Larger array sizes would need to confirm that process uniformity preserves the complementary states across many cells.

Load-bearing premise

The top and bottom cells can be set and kept in truly complementary polarization states without interference, depolarization, or process defects that shrink the differential signal.

What would settle it

Measurement showing that the observed differential polarization stays below the arithmetic sum of the two individual cells' polarizations, or that endurance falls short of 1E10 cycles because of interface degradation between the stacked layers.

Figures

Figures reproduced from arXiv: 2604.15131 by Mansun Chan, Renhao Xue, Ruizhan Yan, Xiwen Liu.

Figure 1
Figure 1. Figure 1: (a) Schematic of an all-ALD-grown vertical complementary FeRAM crosspoint array. Two ferroelectric layers are programmed into complementary polarization. (b) Cross-section along AA’ highlighting the alternating MFMFM stacks (M: metal, F: ferroelectric) constituting the array. (c) Cross-sectional TEM image and (d) EDS mapping [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
read the original abstract

A limited remanent polarization (Pr) in HfO2-based FeRAM remains a key obstacle to density scaling and reliability, while material and process optimizations offer only incremental improvements. This limitation fundamentally originates from the thickness-constrained switchable polarization and the intrinsic polarization ceiling of HfO2-based ferroelectrics. Here, we propose an all-ALD-grown vertical complementary FeRAM (VCF) architecture, in which the top and bottom stacked FeRAM cells maintain complementary polarization. This complementary dipole configuration converts the readout from a single-layer polarization response into a differential polarization summation, thereby amplifying the effective charge window without increasing the switching field of each individual layer or incurring area overhead. Viewed from top to bottom, an "up-down" polarization pair stores logic '1', whereas a "down-up" pair stores logic '0'. Using a complementary polarization write-read scheme, the VCF achieves an effective differential polarization above 100 uC/cm^2 and retains above 90 uC/cm^2 after 1e10 switching cycles without electrical breakdown. Robust retention (longer than 1e4 s at 85 degC) and strong disturb immunity are demonstrated, with an effective differential polarization above 80 uC/cm^2 under a V/3 scheme after 1e6 disturb pulses. Array-level operation is validated in a 5 x 5 selector-free crosspoint array. The performance enhancement of the VCF arises from the co-optimization of the all-ALD-grown process, device architecture, and operation scheme, enabling high density, a wide memory window, and strong reliability for scalable FeRAM integration.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript reports an all-ALD vertical complementary FeRAM (VCF) in which two HfO2-based ferroelectric cells are stacked and programmed into opposite polarization states (up-down for logic 1, down-up for logic 0). A complementary write-read scheme is claimed to produce an effective differential remanent polarization >100 μC/cm² that remains >90 μC/cm² after 10^10 cycles, together with retention >10^4 s at 85 °C, V/3 disturb immunity, and functional operation in a 5×5 selector-free crosspoint array.

Significance. If the complementary-polarization mechanism operates without substantial interlayer coupling or depolarization, the architecture would provide a device-level route to enlarge the memory window of HfO2 FeRAM without raising the switching field or cell area. The all-ALD process flow and selector-free array demonstration constitute concrete strengths that could aid integration if the performance metrics are robustly supported.

major comments (2)
  1. [Device characterization and VCF performance results] The central claim of differential polarization >100 μC/cm² (abstract and results) rests on the top and bottom cells maintaining stable, opposite polarization states whose dipoles add constructively. No separate P-E loops for the individual layers after complementary programming, nor direct measurements of interlayer coupling or depolarization fields across the shared electrode, are provided; without these data the additive summation cannot be distinguished from single-layer performance.
  2. [Reliability and array-level measurements] Endurance data (>10^10 cycles with retained >90 μC/cm²) and disturb-immunity results under the V/3 scheme are presented without raw switching-current traces, device-to-device statistics, or explicit confirmation that wake-up/fatigue behavior is symmetric between the two layers. These omissions leave open the possibility that process-induced asymmetry between top and bottom cells limits the differential signal.
minor comments (2)
  1. [Abstract and disturb-immunity subsection] The V/3 disturb scheme is referenced in the abstract and results but is not defined or illustrated in the main text; a short description or reference to the voltage waveform would improve readability.
  2. [Device architecture figures] Figure captions and schematics of the vertical stack would benefit from explicit labels indicating the polarization directions for the two logic states.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify the presentation of our results. We address each major comment point by point below and indicate where revisions will be made to the manuscript.

read point-by-point responses

  1. Referee: [Device characterization and VCF performance results] The central claim of differential polarization >100 μC/cm² (abstract and results) rests on the top and bottom cells maintaining stable, opposite polarization states whose dipoles add constructively. No separate P-E loops for the individual layers after complementary programming, nor direct measurements of interlayer coupling or depolarization fields across the shared electrode, are provided; without these data the additive summation cannot be distinguished from single-layer performance.

    Authors: We acknowledge the value of separate post-programming P-E loops on individual layers to directly confirm opposite states and rule out coupling effects. The effective differential polarization >100 μC/cm² is obtained from direct measurement on the stacked VCF under the complementary write-read scheme, consistently yielding approximately twice the remanent polarization of our single-layer reference devices fabricated with the same ALD process. The retention (>10^4 s at 85 °C) and endurance (>10^10 cycles) data further indicate that depolarization fields across the shared electrode remain small enough not to degrade the differential signal. In the revised manuscript we will add (i) a schematic of the complementary programming/readout sequence, (ii) single-layer P-E loops for comparison, and (iii) a short discussion of possible interlayer coupling based on the observed stability. We cannot, however, retroactively acquire new separate P-E loops on the exact complementary-programmed devices from the original wafer set. revision: partial

  2. Referee: [Reliability and array-level measurements] Endurance data (>10^10 cycles with retained >90 μC/cm²) and disturb-immunity results under the V/3 scheme are presented without raw switching-current traces, device-to-device statistics, or explicit confirmation that wake-up/fatigue behavior is symmetric between the two layers. These omissions leave open the possibility that process-induced asymmetry between top and bottom cells limits the differential signal.

    Authors: We agree that raw current traces and device statistics strengthen the reliability claims. The reported endurance and V/3 disturb results are derived from the differential current of the stacked cells measured across multiple devices in the 5×5 array. In the revised manuscript we will include representative raw switching-current waveforms for the endurance test, add device-to-device variation statistics (at least 20 devices), and explicitly discuss the symmetry of wake-up/fatigue behavior, noting that the identical ALD process for top and bottom layers and the consistent differential signal over 10^10 cycles argue against significant asymmetry. If space permits, we will also add a supplementary figure showing individual-layer current contributions extracted from the differential measurement. revision: yes

Circularity Check

0 steps flagged

No circularity: experimental device demonstration without derivation chain

full rationale

The paper reports fabrication and electrical characterization of an all-ALD vertical complementary FeRAM stack. Claims rest on measured 2Pr values (>100 uC/cm² differential), endurance (>1E10 cycles), retention, and disturb immunity obtained directly from fabricated devices and array tests. No equations, first-principles derivations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text; the architecture is proposed then validated experimentally rather than derived tautologically from its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on experimental fabrication and electrical testing of the VCF structure rather than on theoretical derivation; no free parameters are fitted to produce the reported numbers, and no new physical entities are postulated.

axioms (1)
  • domain assumption HfO2-based thin films can be made ferroelectric by appropriate doping and annealing
    Implicit background assumption required for any HfO2 FeRAM work.

pith-pipeline@v0.9.0 · 5626 in / 1285 out tokens · 30285 ms · 2026-05-10T10:33:36.659522+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    Chen, in 2019 IEEE International Memory Workshop (IEEE, 2019)

    W. Chen, in 2019 IEEE International Memory Workshop (IEEE, 2019)

    work page 2019

  2. [2]

    Haratipour, S-C

    N. Haratipour, S-C. Chang, S. Shivaraman, C. Neumann, Y-C, Liao, B. Alpizar, I-C. Tung, H. Li, V. Kumar, B. Doyle, S. Atanasov, J. Peck, N. Kabir, G. Allen, T. Hoff, A. Oni, S. Dutta, T. Tronic, A. Roy, F. Hamzaoglu, R. Bristol, M. Metz, I. Young, J. Kavalieros, and U. Avci, in 2022 IEEE IEDM Tech. Dig. (IEEE, 2022)

    work page 2022

  3. [3]

    Z. Zhou, L. Li, Y. Feng, X. Wang, Z. Zheng, L. Jiao, H. Zheng, Y. Shi, K-W. Ang, and X. Gong, Adv. Mater. e09525 (2025)

    work page 2025

  4. [4]

    Z. Fu, S. Cao, H. Zheng, J. Luo, Q. Huang and R. Huang, in 2023 IEEE IEDM Tech. Dig. (IEEE, 2023)

    work page 2023

  5. [5]

    Martin, C

    S. Martin, C. Jahan, L. Hosier, F. Grimaud, M. Louro, J. Laguerre, J. Coignus, W. Vandendaele, J. Borrel, N. Castellani, V. Meli, J. Rottner, C. Boixaderas, T. Magis, A. Roman, M. Bedjaoui, J. Mercier, S. Minoret, C. Euvrard, J. Sturm, S. Kerdilès, K. J ullian, J.-P. Mendivil, T. Giammaria, L. Couture, O. Glorieux, K. Azizi -Mourier, T. Billon, S. Dumont,...

    work page 2024

  6. [6]

    Okuno, T

    J. Okuno, T. Kunihiro, T. Yonai, R. Ono, Y. Shuto, and R. Alcala, M. Lederer, K. Seidel, T. Mikolajick, U. Schroeder, M. Tsukamoto, and T. Umebayashi, in 2023 IEEE IEDM Tech. Dig. (IEEE, 2023)

    work page 2023

  7. [7]

    R. Cao, B. Song, D. Shang, Y. Yang, Q. Luo, S. Wu, Y. Li, Y. Wang, H. Lv, Q. Liu, and M. Liu, IEEE Electron Devices Lett. 40, 1744 (2019)

    work page 2019

  8. [8]

    Wu, C -H

    C-H. Wu, C -H. Chang, Y -M. Tseng, T -Y. Lin, K -H. Kao, C -J. Su, and V.P.H. Hu, in 2025 IEEE Symposium on VLSI Technology (IEEE, 2025)

    work page 2025

  9. [9]

    T. Song, H. Tan, R. Bachelet, G. Saint-Girons, I. Fina, and F. Sánchez, Appl. Electron. Mater. 11, 4809 (2021)

    work page 2021

  10. [10]

    Neuber, M

    M. Neuber, M. Lederer, K. Mertens, T. Kämpfe, M. Czernohorsky, and K. Seidel, Crystals 8, 1115 (2022)

    work page 2022

  11. [11]

    Song, M-J

    J. Song, M-J. Oh, and C-B. Yoon, Materials 5, 1959 (2023)

    work page 1959

  12. [12]

    Y. Qi, X. Xu, I. Krylov, and M. Eizenberg, Appl. Phys. Lett. 118(3), 032906 (2021)

    work page 2021

  13. [13]

    Rowtu, P

    S. Rowtu, P. Meihar, A. Pandey, M. H. Ali, S. Lashkare and U. Ganguly, IEEE Trans. on Electron Devices 7, 3536 (2023)

    work page 2023

  14. [14]

    I. Jeon, H. Lim, S. Kim, J. Sur, J. Kim, Y. Lee, J. Choi, Y. Hong, S. Ahn, C. Jung, E. Chung, Y. Goh, K. Lee, W. Kim, H. Kwon, S. Park, D. Kwon, J. Jeon, S. Lee, D. Ha, S. Ahn, S. Hyun, and J. Song, in 2025 IEEE Symposium on VLSI Technology (IEEE, 2025)

    work page 2025

  15. [15]

    K. Kim, S. Oh, M. Fiagbenu, J. Zheng, P. Musavigharavi, P. Kumar, N. Trainor, A. Aljarb, Y. Wan, H. Kim, K. Katti, S. Song, G. Kim, Z. Tang, J. Fu, M. Hakami, V. Tung, J. Redwing, E. Stach, R. H. Olsson III, and D. Jariwala, Nat. Nanotechnol. 18, 1044 (2023)

    work page 2023

  16. [16]

    Ramaswamy, A

    N. Ramaswamy, A. Calderoni, J. Zahurak, G. Servalli, A. Chavan, S. Chhajed, M. Balakrishnan, M. Fischer, M. Hollander, D. P. Ettisserry, A. Liao, K. Karda, M. Jerry, M. Mariani, A. Visconti, B. R. Cook, B. D. Cook, D. Mills, A. Torsi, C. Mouli, E. Byers, M. Helm, S. Pawlowski, S. Shiratake, and N. Chandrasekaran, in 2023 IEEE IEDM Tech. Dig. (IEEE, 2023)

    work page 2023

  17. [17]

    A. M. Walke, M. I. Popovici, S. H. Sharifi, E. C. Demir, H. Puliyalil, J. Bizindavyi, F. Yasin, S. Clima, A. Fantini, A. Belmonte, G. S. Kar, and J. V. Houdt, IEEE Electron Device Lett., 4, 578 (2024)

    work page 2024

  18. [18]

    Chang, N

    S.-C. Chang, N. Haratipour, S. Shivaramani, C. Neumann, S. Atanasov, J. Peck, N. Kabir, I. -C. Tung, H. Liu, B. Krist, A. Oni, S. Sung, B. Doyle, G. Allen, C. Engel, A. Roy, T. Hoff, H. Li, F. Hamzaoglu, R. Bristol, M. Radosavljevic, B. Turkot, M. Metz, I. Young, J. Kavalieros, and U. Avcil, in 2021 IEEE IEDM Tech. Dig. (IEEE, 2021)

    work page 2021

  19. [19]

    J. Yu, S. Guo, J. Zhang, X. Jin, C. Wu, M. Zhao, H. Li, C. Guo, K. Xu, Y. Tian, D. Tian, Z. Li, T. Wang, H. Zhu, Q. Sun, Y. Xie, H. Wang, D. W. Zhang, and L. Chen, in 2024 IEEE IEDM Tech. Dig. (IEEE, 2024)

    work page 2024

  20. [20]

    S Zhang, R. Han, X. Lin, X. Wu, and M. Chan, IEEE Electron Device Lett., 9, 661 (2004)

    work page 2004

  21. [21]

    W. Wei, W. Zhang, L. Tai, G. Zhao, P. Sang, Q. Wang, F. Chen, M. Tang, Y. Feng, X. Zhan, Q. Luo, Y. Li, and J. Chen, in 2021 IEEE IEDM Tech. Dig. (IEEE, 2021)

    work page 2021

  22. [22]

    M. I. Popovici, J. Bizindavyi, P. Favia, S. Clima, M. N. R. Alam, R. K. Ramachandran, A. M. Walke, U. Celano, A. Leonhardt, S. Mukherjee, O. Richard, A. Illiberi, M. Givens, R. Delhoulgne, J. V. Houdt, and G. S. Kar, in 2022 IEEE IEDM Tech. Dig. (IEEE, 2022)

    work page 2022

  23. [23]

    T. Fu, M. Zeng, S. Liu, H. Liu, R. Huang, and Y. Wu, in 2022 IEEE IEDM Tech. Dig. (IEEE, 2022)

    work page 2022

  24. [24]

    Zhao, Y.-R

    Z. Zhao, Y.-R. Chen, Y.-W. Chen, W.-H. Hsieh, J.-F. Wang, J.-Y. Lee, Y. Xing, G.-H. Chen, and C. W. Liu, in 2023 IEEE Symposium on VLSI Technology (IEEE, 2023)

    work page 2023

  25. [25]

    J. Hur, P. Wang, Z. Wang, G. Choe, N. Tasneem, A. I. Khan, and S. Yu, in 2020 IEEE IEDM Tech. Dig. (IEEE, 2020)

    work page 2020

  26. [26]

    W. Shi, D. Zhang, Z. Zheng, Z. Zhou, K. Han, C. Sun, Q. Kong, Y. Feng, G. Liang, K. Ni, J. Chen, and X. Gong, in 2024 IEEE IEDM Tech. Dig. (IEEE, 2024)

    work page 2024

  27. [27]

    Tahara, K

    K. Tahara, K. Toprasertpong, Y. Hikosaka, K. Nakamura, H. Saito, M. Takenaka, and S. Takagi, in 2021 IEEE Symposium on VLSI Technology (IEEE, 2021)

    work page 2021

  28. [28]

    D. Chen, S. Zhong, Y. Dong, T. Cui, J. Liu, M. Si, and X. Li, IEEE Electron Device Lett., 12, 2065 (2022)

    work page 2065