1. Why USB Drives and SD Cards Suddenly Stop Working: The Tiny Rooms Inside Flash Memory
Inside SD cards, USB drives, and SSDs, there are countless tiny "rooms" that can hold electrons. This collection of rooms is called NAND flash memory [1]. The smaller these rooms are, the more of them can fit on a card of the same size. The reason a card the size of a fingernail tip can hold tens of thousands of family photos is that the technology to shrink and stack these rooms has advanced.
However, these rooms can break. A card might stop working when plugged in, or the computer might display a message asking you to "format" the drive. In this notebook, we will first look at the technology used to shrink these rooms, and then examine how they break and where the data goes.
2. How Each Room in Flash Memory Traps Electrons to Store Data
Each room in flash memory is surrounded by a wall made of an insulator. By putting very small particles called "electrons" into or out of the room enclosed by this wall, the room represents either a "0" or a "1" [1]. Even when the power is turned off, the electrons remain in the room, so the data does not disappear.
The person who conceived this system was Fujio Masuoka, an engineer at Toshiba. In 1980, he invented a memory that could be rewritten using electricity. The NOR type, which simplified the readout circuit, was announced in 1984, and the NAND type, designed to be cheaper and mass-produced, was presented at the International Electron Devices Meeting (IEDM) in 1987 [1].
Later, the amount of information each room could record increased. Initially, it was just 1 bit (either "0" or "1"). In 1998, NEC demonstrated a 2-bit chip, and in 2009, Toshiba and SanDisk announced a 4-bit chip. In 2010, Samsung began mass-producing a 3-bit method [1]. However, as the number of bits packed into a room increases, the number of electrical strength levels that must be distinguished grows. This makes misreading more likely due to slight shifts, and the number of times rewriting can be withstood decreases [1].
3. Why Shrinking the Rooms Makes Their Walls Thinner and Causes Problems
Making the rooms smaller allows more of them to fit in the same space. But if they get too small, problems arise. The walls between neighboring rooms become thin, and the electrons meant to be trapped can leak into the adjacent room [2].
Kioxia (the company that was formerly Toshiba's flash memory division) explains that shrinking rooms on a flat plane became difficult when a single room reached about 16 nanometers (about 1/60,000th of a millimeter) [3][4]. The walls became too thin to hold electrons stably.
The walls themselves also suffer. Every time electrons are moved in or out, a strong voltage is applied to the wall. This gradually breaks the bonds inside the wall and weakens its insulating power [1]. As engineers try to shrink rooms and control electron movement more precisely, the impact of this burden becomes more significant.
4. How Stacking Rooms in 3D Solved the Problem of Shrinking Flat Rooms
Since it became difficult to shrink rooms while keeping them flat, engineers devised another method. Instead of arranging them side by side, they stack them in layers vertically. This is called "3D NAND." Toshiba announced this stacking concept in 2007 [3].
Samsung was the first to commercialize it, mass-producing 24-layer V-NAND in 2013 [3]. Kioxia then increased the number of layers: 48 layers in 2015, 96 layers in 2018, 112 layers in 2020, and 162 layers in 2022. By March 2023, they achieved stacking over 200 layers [4]. The manufacturing method involves stacking plate-like electrodes and insulators, then drilling vertical holes through them to pass charge-storing films and electrodes, creating cells in all layers at once [4].
This is similar to building a taller apartment building to house more people. However, the higher you stack, the harder it becomes to drill straight, uniform vertical holes. Even by changing the strategy from "shrinking" to "stacking," engineers face new difficulties.
5. Why Flash Memory Wears Out with Use and Fades Over Time
Each time electrons are moved in or out, the walls of the rooms undergo stress. Just as a paperclip bent repeatedly eventually breaks, rooms used many times become unable to write new data [5].
For example, the Samsung 1TB SSD "970 EVO," released in 2018, lists a lifespan indicator of "600TBW" (capable of withstanding a total of 600 terabytes of writing). Since the capacity is 1TB, dividing the total write limit by the capacity means the entire drive can be rewritten approximately 600 times [6].
Even if the write limit is not reached, data naturally fades over time because trapped electrons slowly leak from the walls. Industry standards (JEDEC) state that if an SSD near the end of its life is placed in a high-temperature location without power, data loss begins in about 3 months for enterprise drives and about 1 year for consumer drives [6]. It may seem surprising that enterprise drives fail faster, but it is said that designs optimized for heavy writing sacrifice the ability to retain data for long periods. Note that this specific reason is not stated in the sources used here. Researchers in digital preservation also state that flash memory can begin to lose data about 1 year after last use, depending on temperature [7]. If there is plenty of life left and it is stored at room temperature, it usually lasts much longer than this period, but it is not a basis for claiming "flash memory is safe for decades."
6. Why the Controller Chip Can Fail Even When the Rooms Are Fine
SD cards and USB drives contain not only the rooms that trap electrons but also a small semiconductor called a "controller." This guide decides which room to write to and avoids damaged rooms. Many failures where a computer suddenly stops recognizing the device are said to be caused by defects in this controller or manufacturing variations, rather than the rooms themselves [5].
The USB drive's connector itself also has a lifespan. Standards assume that contact may worsen after about 1,500 insertions and removals [5]. Even if the data inside is safe, if the port breaks, it cannot be read.
Therefore, when a storage medium "breaks," at least three types of failure are mixed: (1) the rooms themselves reach their lifespan and cannot be written to, (2) the controller fails and cannot direct reading and writing, and (3) physical parts like connectors break. Even if the symptom is "it cannot be read," what can be done next depends on the cause.
7. Why Deleted Files May Remain and Saved Files May Disappear
When you "delete" a file on a hard disk, the data itself often remains on the disk. The computer only deletes the "pointer" (a marker indicating the file's location). Until new data overwrites it, specialized software can sometimes read it [8]. This point was examined in detail in another notebook, "Where Do Deleted Photos Go?". Here, as a continuation, we look at what happens when the storage medium itself breaks.
However, SSDs and USB drives are different. If a mechanism called "TRIM" is active, the computer tells the drive, "This area is no longer in use," and the drive erases that area in the background to prepare for new writes. This erasure can happen within minutes. If the data is checked immediately after deletion, it might still be readable, but after a short time, software may no longer be able to restore it [6]. A mechanism called wear leveling also makes it difficult to know where rewritten data remains [9]. Things that remain on hard disks disappear quickly on SSDs.
On the other hand, if the controller fails and the data cannot be read, the story is different. The flash chip itself may be intact. Specialists sometimes remove just the chip from the board (a method called "chip-off") and connect it directly to a specialized reader to extract data [10]. However, this method carries risks, such as the "popcorn effect," where moisture inside the chip expands rapidly due to heat, breaking components. Without experienced professional work, this can destroy the chip and permanently lose the data [10].
Furthermore, many recent SSDs automatically encrypt data before writing it (self-encrypting drives). In these cases, the key to decrypt the data is often stored only within the controller. If the controller breaks and the key is lost, even if the chip is read via "chip-off," the output may be nothing but meaningless symbols [11].
8. What to Keep in Mind If You Want to Investigate Flash Memory Further
As seen in this notebook, flash memory is a tool where things that "should not disappear" can disappear depending on conditions. Among those involved in digital preservation, the concept of "migration" is used: rather than trusting one medium forever, records are regularly copied to new media or file formats before they degrade [7].
It might be interesting to ask your family if they have ever had trouble with a broken USB drive or SD card. Discussing whether important photos and videos are copied to other places (like another card or an internet storage service) helps you see where the "conditions for disappearing" apply to your own life.
The record for the number of layers in flash memory continues to be updated. If you check the technical information pages of companies like Kioxia [4] in a few years, you will see how outdated the numbers in this notebook have become.