The advent of non-volatile flash memory — semiconductor memory that does not lose its data when the power is turned off — revolutionized consumer electronics. It is now used to store the numbers in mobile phones, the pictures taken with digital cameras, and the music tracks in MP3 players. Similar types of memory based on organic semiconductors, rather than traditional silicon-based semiconductors, would make possible entirely new concepts: intelligent food packaging, for instance, that could be used by retailers to control their inventories and to alert consumers when the food is getting close to its 'use-by' date. Writing in Advanced Materials, Baeg et al.1 describe an organic thin-film memory transistor that brings such intriguing possibilities a little closer.

Two particular advantages of organic thin-film transistors over solid-state transistors are that they are simpler to make, and can be fabricated on thin, flexible plastic substrates. They could thus form the backbone of low-cost microelectronics ranging from radio-frequency identification tags to flexible, large-area active-matrix displays2. Many of these applications require non-volatile (stable) data storage, preferably with memory elements that can be programmed, erased and read electrically. Baeg and colleagues' transistor1 not only fulfils these requirements, but can also, owing to its similar architecture, be simply integrated into existing technology based on organic transistors.

In order to function as a memory, a device must be observed in two (or more) different states. In flash memories, this is achieved by introducing a second 'floating gate' to the silicon transistor between the normal control gate, which regulates the flow of current through the transistor, and the semiconducting substrate (Fig. 1a). This floating gate is insulated all around by an oxide layer. A high-voltage pulse applied to the control gate places charge on the floating gate, where it becomes trapped. This partially cancels out the electric field coming from the control gate, and so modifies the threshold voltage of the transistor — that is, the voltage required before it lets current flow. Thus, when the transistor is 'read' by placing a specific voltage on the control gate, electrical current will either flow or not flow, depending on the threshold voltage, and so the number of electrons on the floating gate. The resulting current level flowing through the transistor can be used to define boolean '0' and '1' memory states. The charge trapped on the floating gate persists long after all voltages are removed: the information is thus retained after power shutdown, and the transistor functions as a non-volatile memory device.

Figure 1: Methods against memory loss.
figure 1

The basic transistor is a device in which a small voltage applied at the control gate (G) modulates a much larger current flow from source (S) to drain (D) through a semiconductor substrate. a, In flash memories, an amount of charge is trapped on a floating gate (FG) that modifies the control voltage required for current to flow from S to D. Whether current flows or not defines a boolean '1' or '0'. The memory of this state persists as long as the charge remains trapped on the floating gate. b, In Baeg and colleagues' organic device1, the same principle is used, but the charge is trapped locally on a thin 'electret' of chargeable polymer, rather than on an isolated floating gate.

The non-volatile memory technology developed by Baeg et al.1 takes a slightly different approach (Fig. 1b). Instead of charge being trapped on a floating gate, it is trapped locally on a thin layer of a chargeable polymer. This 'electret' is inserted between the insulating material (silicon dioxide) that makes up the gate and the organic semiconductor (which is pentacene, C22H14). When a high-voltage pulse is applied to this electret, it charges up. Applying a reverse voltage pulse either decharges it, restoring the initial state, or charges it to the opposite polarity. The trapped charge imposes an added voltage on the threshold gate voltage, as a floating gate does in a flash memory. This too can be sensed and translated into '1's and '0's, reproducing the stored data, by measuring the current flowing through the transistor.

Memory based on such organic field-effect transistors (OFETs) has been investigated before3,4, but Baeg et al.1 are the first to report programming speeds of around 1 microsecond — a million times faster than the previous best time of around a second. That marks a decisive step towards making organic memory technology fit for technological purposes.

The new speed record is the result of fast and efficient charge transfer from the organic semiconductor into the polymer electret by means of the electric gate field. The gate field lowers the energy barrier at the interface of the semiconductor and the electret, and so facilitates charge transfer. Once transferred, most of the charge is trapped deeply in the electret. This model explains Baeg and colleagues' most important observations: a critical gate field, caused by the energy barrier, for transfer and trapping of the charges; the lowering of this critical field when the device is illuminated with visible light; a long retention time of the order of hours in the dark; and a decrease of the retention time upon illumination.

These observations also indicate moot points. Can these devices reach data retention times necessary for practical applications — typically years for non-volatile memory? Furthermore, can the device be scaled down to more practical voltages — to 10 V from the 100 V used in the present work — without sacrificing device speed and stability? Such questions remain unanswered, but these encouraging results will without doubt spur intensive investigations into this approach.

The relevance of these results could also go beyond the scope of just memory. Baeg and colleagues1 aim to produce non-volatile memories that exploit charge trapping and storage, but others are concerned with the converse problem: eliminating charge trapping as much as possible where it gives rise to undesired shifts in threshold voltages that can suppress 'n-channel' (electron) mobility5 in OFETs, and limit the operational lifetime of 'p-type' (electron-hole-based) logic circuitry6. Whatever the intent, all will benefit from a thorough understanding of charge-trapping effects at the crucial interface between a semiconductor and its gate.