PaperlayerView exact revision on arXiv

Source-linked AI summary

Review on Physically Flexible Nonvolatile Memory for Internet of Everything Electronics

Mohamed T. Ghoneim, Muhammad M. Hussain

arXiv:1606.08404v1cs.ET

TL;DR

Physically flexible electronics for IoT and healthcare need dense, low-power memory, yet future systems depend on integrating transistors with nonvolatile storage. This review surveys flexible NVM fabrication approaches and benchmarks their performance and reliability; it concludes that commercialization challenges remain, with flexible FeRAM and flash having advantages tied to their rigid-state maturity.

  • Problem

    Physically flexible IoT devices require reliable, ultra-low-power, high-density NVM alongside processing and wireless communication.

    Source: §1 · p. 4

  • Method

    The review surveys flexible NVM architectures and fabrication approaches, benchmarks devices across all-organic, hybrid, and inorganic routes, and discusses reliability.

    Source: §1 · p. 2, §6 · p. 36

  • Results

    Flexible FeRAM has an edge because its rigid form is used as embedded NVM in microprocessors, while flash has high potential because of its rigid-state maturity.

    Source: §6 · p. 36

  • Takeaways & Limitations

    Flexible ReRAM, FeRAM, PCRAM, MRAM, and flash still face challenges before commercialization; transfer and progress in flexible flash may speed its introduction.

    Source: §6 · p. 36

    Flexible flash remains far from state-of-the-art bulk properties, and demonstrated flexible flash can suffer from low yield.

    Source: §5.4 · pp. 35–36

Abstract

from arXiv · show

Solid-state memory is an essential component of the digital age. With advancements in healthcare technology and the Internet of Things (IoT), the demand for ultra-dense, ultra-low-power memory is increasing. In this review, we present a comprehensive perspective on the most notable approaches to the fabrication of physically flexible memory devices. With the future goal of replacing traditional mechanical hard disks with solid-state storage devices, a fully flexible electronic system will need two basic devices: transistors and nonvolatile memory. Transistors are used for logic operations and gating memory arrays, while nonvolatile memory (NVM) devices are required for storing information in the main memory and cache storage. Since the highest density of transistors and storage structures is manifested in memories, the focus of this review is flexible NVM. Flexible NVM components are discussed in terms of their functionality, performance metrics, and reliability aspects, all of which are critical components for NVM technology to be part of mainstream consumer electronics, IoT, and advanced healthcare devices. Finally, flexible NVMs are benchmarked and future prospects are provided.

1. Introduction

IoT and wearable applications call for physically flexible systems that combine reliable, low-power, high-density memory with processing and communication. This review focuses on flexible NVM, surveying fabrication approaches and technologies alongside performance, reliability, and integration considerations.

1. Introduction

  • Flexible-electronics approaches include all-organic systems, inorganic devices transferred onto organic substrates, and transfer-free thinning of inorganic substrates.The review also describes organic materials and nanoscale materials used in reported memory devices.Source: §1 · pp. 2, 4
  • Future electronic systems may rely on transistors and NVM, as emerging NVM can serve both long-term storage and quick-access memory roles.Memory arrays contain the system’s largest concentration of transistors because each cell uses an access device.Source: §1 · p. 2
  • The review surveys flexible resistive, ferroelectric, phase-change, and charge-trapping NVM, including device reliability and commercialization prospects.It emphasizes recent techniques for producing flexible NVM and benchmarks performance across approaches.Source: §1 · pp. 2–3
  • The review focuses on demonstrated flexible rewriteable NVM from January 2010 to May 2015 and combines flexibility, performance, reliability, and integration perspectives.Earlier reviews often focused on particular NVM types or omitted flexibility, reliability, access transistors, or monolithic integration.Source: §1 · pp. 3–4
  • IoT devices are expected to combine physical flexibility with real-time processing, reliability, ultra-low power, dense NVM, and wireless communication.Potential applications include healthcare, infrastructure, vehicles, and wearable devices.Source: §1 · pp. 4–5

2. Approaches for Making Flexible Devices

The review contrasts all-organic, hybrid, spalling, and transfer-free silicon routes to flexible devices, weighing fabrication approaches against performance, reliability, and mechanical constraints. The transfer-free approach seeks to retain silicon electronics’ high performance and integration advantages by reducing substrate thickness.

2.1. The All-Organic Approach

  • All-organic devices combine polymeric channel and ferroelectric materials with flexible insulating substrates, but face lower channel mobility, reliability, and thermal stability than inorganic silicon electronics.Their flexibility and conformal abilities are benefits, while low dielectric constants make it challenging to achieve the capacitance needed for high-performance devices.Source: §2.1 · p. 5

2.2. The Hybrid Systems Approach

  • Hybrid systems combine organic and inorganic materials, including by transferring high-performance inorganic devices to plastic, but transfer adds steps and can reduce yield.Direct deposition on plastic avoids transfer but restricts fabrication temperatures, challenging high-quality films, while solvents constrain plastic choices.Source: §2.2 · p. 6

2.3. Spalling Technology

  • Spalling uses a stressor layer to initiate cracks and peel thin device-bearing semiconductor layers onto flexible substrates, but its tuning, etching, and dexterous peeling challenge mass production.Reported demonstrations include room-temperature controlled spalling of full SOI wafer circuitry onto plastic.Source: §2.3 · p. 9

2.4. The Complementary Transfer-free Inorganic Approach

  • Transfer-free silicon flexibility comes from thinning the host substrate, preserving silicon’s performance, reliability, ULSI density, low cost, and monolithic integration.Soft back etching avoids the chemical-mechanical polishing, residual stress, and substrate damage associated with back grinding; substrate thickness must also suit the bending radius to remain below fracture stress. Common challenges include high speed, low power, and reliability under flexing beyond studied planar stresses; organic electronics additionally need temperature stability and satisfactory integration density.Source: §2.4 · pp. 10, 12

3. NVM Operational Principles and Architectures

The review distinguishes memories by retention and operating principle, then describes common cell architectures. Nonvolatile memory retains information without power, with an industry-standard retention duration of about 10 years.

3.1. NVM Operational Principles

  • Nonvolatile memory (NVM) retains information without power for long periods (~10 years is the industry standard), unlike volatile memory, which loses information when power is removed.The review lists ReRAM, FeRAM, MRAM, PCRAM, and flash as major NVM classes.Source: §3.1 · p. 13
  • PCRAM stores data by switching a localized material between crystalline, low-resistance and amorphous, high-resistance phases using current or laser pulses.Short pulses above the melting temperature produce the crystalline-to-amorphous change.Source: §3.1 · p. 15
  • Floating-gate flash uses a three-layer gate dielectric; programming tunnels carriers from the channel to the floating gate, shifting the transistor state, while reverse bias erases.The floating-gate layer may use doped polysilicon, quantum dots, or metallic nanoparticles.Source: §3.1 · p. 15

3.2. NVM Architectures

  • Memory cells use 1T, 1T1C/1T1R, or 2T2C architectures, with different interconnect arrangements including NOR, NAND, and crossbars.In 1T1C/1T1R cells, an access transistor accompanies a storage capacitor or resistor; 2T2C uses two transistors and two capacitors per bit.Source: §3.2 · p. 16, Figure 7 · p. 18

4. Flexible Field-Effect Transistors (FETs) for Logic and NVM Arrays

The section distinguishes switching transistors for logic and memory access from transistors that store charge, then surveys flexible transistor technologies. All-organic devices offer strong bending flexibility but generally have low mobility and other performance trade-offs.

4. Flexible Field-Effect Transistors (FETs) for Logic and NVM Arrays

  • Logic and access FETs switch without storing charge, unlike 1T memory transistors, which are assessed by retention, endurance, and memory window.Access and logic devices are evaluated by switching measures including subthreshold slope, drain-induced barrier lowering, and threshold-voltage stability.Source: §4 · p. 19

4.1. All-Organic Transistors

  • All-organic flexible transistors generally have mobility below 10 cm2/V.s, except graphene TFTs at 340 cm2/V.s, while other metrics trade off against one another.Increasing BST gate-dielectric thickness can reduce threshold voltage and increase effective mobility, but may reduce ION/IOFF and increase subthreshold slope.Source: §4.1 · p. 19, Table 2 · p. 20
  • All-organic devices demonstrated bending radii down to 100 μm and persistence through 10,000 bending cycles, but often had large features and operating voltages.The reported maximum yield was 66%, for ink-jet-printed devices.Source: §4.1 · p. 19, Table 2 · p. 20

min bending radius

Table 3 summarizes demonstrated hybrid flexible devices.

min bending radius

  • Table 3 covers demonstrated hybrid flexible devices.The caption supplies the table's scope but no bending-radius values.Source: Table 3 · p. 22

Structure

Hybrid transistors show improved electrical performance over all-organic devices but generally lower flexibility. Flexible silicon transistors offer strong electrical properties and integration potential, while mechanical and electrical stress can degrade reliability.

4.2. Hybrid Transistors

  • More than 80% of demonstrated hybrid devices had minimum bending radii of about 2 mm or larger, despite mobility often exceeding 10 cm2/V.s.One device reached 141 cm2/V.s, while another demonstrated up to one million bending cycles at a 10-mm radius.Source: §4.2 · p. 24
  • Hybrid devices also demonstrated features of a few microns, an order-of-magnitude improvement over typical all-organic features, and reported yield reached 90%.Some hybrid devices reached submicron gate lengths.Source: §4.2 · p. 24

4.3. Inorganic Transistors on Flexible Silicon

  • Flexed inorganic silicon gate stacks showed lifetime degradation from electrical stress, attributed to increased interfacial charges and a 20% decrease in safe operating voltage.The reported voltage reduction still met the projected ten-year industry lifetime standard.Source: §4.3 · p. 24, Table 4 · p. 25
  • Dynamic mechanical stress caused the most severe reported degradation in flexible silicon transistors, while planar MOSFET functionality persisted through 200 bending cycles.The review identifies CMOS-based flexible transistors as promising for IoT because of monolithic integration and bulk-like electronic properties, while noting that degradation should inform system design.Source: §4.3 · p. 24

5. Flexible NVM Technologies

The section assesses flexible NVM through device performance metrics and reviews flexible ReRAM and FeRAM approaches. It relates memory-cell characteristics to array efficiency, power, and reliability.

5. Flexible NVM Technologies

  • Memory assessment uses form factor, density, cost per bit, endurance, retention, operation voltage, speed, and memory window.Form factor is cell area divided by the square of the technology node; memory window reflects the distinguishability of memory states.Source: §5 · pp. 25–26
  • Lower form factor and higher density can improve array fit and reduce bit cost, while endurance and retention characterize repeated use and data preservation.Ten years is identified as the industry retention benchmark for NVMs.Source: §5 · p. 26

5.1. Flexible ReRAM

  • Flexible ReRAM approaches include organic devices, transferred silicon devices, low-temperature inorganic deposition on plastic, and inorganic devices on flexed silicon.Cellulose-nanofiber ReRAM on aluminum foil was reported with a 0.35 mm bending radius and ±0.5 V operation.Source: §5.1 · pp. 26–27

5.2. Flexible FeRAM

  • FeRAM uses bistable ferroelectric polarization for data storage, and its endurance and variability are strengths relative to redox memristive memories.PZT-based flexible FeRAM is limited on polymer platforms by its crystallization temperature above 600 °C; hybrid devices transferred to polymer have reported 20,000 bending cycles.Source: §5.2 · p. 27

Flexible Final

The supplied table fragments list flexible ReRAM device materials, approaches, and selected performance entries, but do not preserve enough context to associate individual values with specific devices.

Flexible Final

  • The fragments include structures such as PES/Al/graphene-O/Al and graphene/PMMA/Al, alongside organic and hybrid approaches.The rows do not preserve enough context to match each structure to a specific approach.Source: Flexible Final · p. 29
  • Listed operating-voltage entries are −4, −4 to +10, −5, −2 to +3, −0.5 to 2.8, and −1 to +1 V.The fragment does not identify which device corresponds to each entry.Source: Flexible Final · p. 29
  • Form-factor, memory-window, speed, and yield entries are present, but most are marked with dashes in the supplied fragment.The fragment reports yield entries of 80% and 60% without preserving their device associations.Source: Flexible Final · p. 29

Flexible Final

The supplied table fragments identify flexible-memory structures and approaches, but provide little recoverable performance information for comparing the listed devices.

Flexible Final

  • The listed structures include PES/Cu/TiO2/Cu and Al foil/Ag-CNP with cellulose.The fragment does not preserve device-specific associations for all entries.Source: Flexible Final · p. 30
  • The approaches named in the fragments include low-temperature hybrid deposition, inorganic devices, and all-organic devices.The entries do not provide enough context to match each approach to a particular device.Source: Flexible Final · p. 30
  • Form factor, memory window, speed, and yield appear as table metrics, but the supplied entries are all dashes.Source: Flexible Final · p. 30

Flexible Final

The section compares flexible PCRAM and flash with reported performance and reliability challenges. It concludes that flexible NVM still faces difficulties matching bulk devices, while flash remains far from bulk properties in part because research has focused on new materials and processes.

Flexible Final

  • The supplied flexible-memory table fragment lists operating-voltage, form-factor, memory-window, retention, bending, and yield fields, but the entries lack device labels.The fragment reports 90% yield without preserving its device association.Source: Flexible Final · p. 32

5.3. Flexible PCRAM

  • 176 Gbit/square inch was the highest reported flexible PCRAM density, while the highest reported bending endurance and yield were 1,000 cycles and 66%.The 176 Gbit/square inch device was reported by Yoon et al., and Mun et al. reported the bending endurance and yield.Source: §5.3 · p. 33

5.4. Flexible Flash

  • Flexible flash reports operation voltages from ±5 to ±90 V and minimum channel lengths of 2 μm, alongside bendability up to 5 mm for 2,000 cycles, endurance of 100,000 cycles, and retention of 10^6 s.The review says flexible flash remains far from state-of-the-art bulk flash properties, partly because research focused on new materials and processes.Source: §5.4 · pp. 35–36

Flexible Final

  • Flexible NVMs face challenges matching bulk counterparts, and flexible FeRAM's projected 10-year retention and 130 nm CMOS integration have not been demonstrated.The review also notes a CMOS-integration challenge for cellulose-nanofiber ReRAM on aluminum foil.Source: §5.4 · p. 35

6. Conclusions and Future Prospects

The review benchmarks flexible NVM approaches and highlights their prospects for commercialization. It also places transfer-free inorganic silicon-based flexible electronics as a newer approach than flexible organic electronics.

6. Conclusions and Future Prospects

  • Transfer-free inorganic silicon-based flexible electronics is relatively new compared with flexible organic electronics, despite high expectations for silicon-based performance and reliability.Source: §6 · p. 36
  • The review benchmarks transistors and storage devices across all-organic, hybrid, and inorganic approaches, with Table 9 covering flexible Flash work from the past five years.Source: §6 · p. 36, Table 9 · p. 37
  • Flash has the highest commercialization potential because of rigid-state maturity, while flexible FeRAM has an edge from its use as embedded NVM in microprocessors.ReRAM, PCRAM, and MRAM remain emerging even in bulk rigid form, though their scaling, speed, and low-power potential may help them catch up in flexible electronics.Source: §6 · p. 36

Flexible Final

Flexible NVMs could support low-power IoT storage, including data retained in both powered and unpowered states, but current flexible devices remain behind bulk devices. Closing that gap requires either new technologies that rival CMOS integration and processes at reasonable cost, or flexible devices that adopt CMOS technology without compromising critical performance metrics.

Flexible Final

  • The comparison fragment lists deposition and transfer approaches, operating-voltage and speed entries, and unreported yield entries.It includes a hypothesized read time of ~14 ns, but the supplied fragments do not establish which approach or device it refers to.Source: Flexible Final · p. 38
  • Flexible NVMs could store instructions and data in powered and unpowered states, while high-density storage could defer data transmission until external charging.The review also describes storing information for later analysis rather than continuously transmitting it by antenna.Source: Flexible Final · p. 39
  • Current flexible NVMs lag bulk devices, and the review identifies CMOS-compatible flexibility or rival technologies as routes to address the gap.The proposed challenge is to rival CMOS integration ability and standard processes at reasonable cost, or obtain flexibility without affecting critical device performance metrics.Source: Flexible Final · p. 39

Conflicts of Interest

  • The authors report no competing financial interest and state that copyrights for previously published materials were secured.Source: Conflicts of Interest · p. 39
Loading 1606.08404v1…