Circuit for Detecting Both Charge Gain and Charge Loss Properties in a Non-Volatile Memory Array

The invention relates generally to integrated circuit development and manufacturing tools for facilitating the design and testing of selected semiconductor structures. The present invention has specific applicability to test structures that are capable of detecting both charge gain and charge loss properties in an array of Flash EPROM memory cells or other programmable logic cells.

In the design of semiconductor devices, it is often advantageous (if not necessary) to monitor the behavior of some physical property of such device to characterize and thus determine whether a particular physical implementation is superior to another implementation. A substantial amount of research and effort is invested and undertaken by design and process engineers simply to create better, simpler, and more accurate test vehicles for observing the characteristics of a proposed design. The monitoring of device behavior and characteristics is done in a variety of ways, and one common method practiced today includes the use of test structures which are used either (or sometimes both) in the design and manufacturing of the devices in question. These test structures are usually implemented directly on the same substrate as the device under investigation, and are configured to be more easily accessible for measuring the property of interest.

An example of this is the threshold voltage of a Flash memory cell. This voltage is widely considered as the most critical parameter in determining the memory state that is being stored by the cell. The threshold voltage is controlled by the amount of charge that is placed on the floating gate of the Flash memory cell. In other words, this voltage is a direct function of the amount of charge stored within a particular cell. During the operation of the memory cells in an array, the floating gate for a given cell can inadvertently gain or lose charge. This unintentional charge gain or charge loss can change the memory state of the cell and become a cause of reliability failure in the Flash memory array. It is crucial to have a good monitor of this behavior for a large array of memory cells in a given Flash memory technology. As is also apparent, as the number of memory cells increases for a particular device, the expected variations between poor charge storing devices (leaky bits) and poor charge discharging devices (stubborn bits) are going to increase. For this reason, a testing mechanism which leads to more tightly controlled charge distribution populations is highly useful.

The most straightforward method to evaluate the charge-gain/charge-loss behavior of a device is to monitor the threshold voltage of each cell in the large memory array. This is often not convenient since it involves the use of a fully functional circuit onboard the device with all the proper decoding to address and access each individual cell in the memory array. However, this solution is not extremely practical from a functional or manufacturing perspective. It is substantially more useful for the population characteristics to be observed prior to the release of such device for use, and some manufacturing test procedures are known in the art for monitoring the devices as they are manufactured in a fabrication facility. Even this approach, however, is non-optimal, and for that reason, some degree of device behavior is both simulated and tested during the design phase of a particular flash cell, so that the charge population distribution of a particular physical cell implementation can be measured and controlled before full scale manufacturing is undertaken using such implementation.

Accordingly, it has become common in the industry to try and monitor the charge-gain/charge-loss behavior to evaluate the feasibility and reliability of an unproven Flash technology, prior to committing the design of a full functional circuit. In previous test structures of this type, an array of Flash EPROM memory cells are all connected together to provide a simple way of evaluating the behavior of the memory cell population. One drawback of this approach is that it is often capable of only evaluating the charge loss behavior of the memory array. The monitor of the charge gain behavior still has to resort to the use of a full functional memory array circuit. In such contexts, while it has been hitherto impossible, it would be very desirable to have a simple test structure capable of monitoring both the charge gain and charge loss behavior, without the use of full functional circuit.

FIG. 1 illustrates a test structure in the prior art that is capable of only monitoring the charge loss behavior. This structure typically consists of many cells (100, 101, 102, etc.) with a shared gate 100a, shared drain 100b, and shared source 100c. If all the cells are fixed at a same threshold voltage, the sub-threshold characteristics of the array will be equivalent to the sub-threshold characteristics of a single cell having a size equal to the total aggregate width of all the cells in the array. FIG. 2A illustrates an approximate graphical depiction of the sub-threshold behavior of an array with all the cells fixed at the same threshold voltage. (It should be noted that the Y-axis on the curve is in logarithmic scale.)

As is to be expected as a consequence of normal manufacturing variations, however, in the population of cells there will be cells having slightly different floating gate charge retention characteristics. This is inevitable given typical semiconductor processes which result in slightly different floating gate sizes, thicknesses, coupling to source/drains, etc. Thus, it is expected that there will be some random distribution (rather than a fixed or completely uniform distribution) of cell charge characteristics, with some being below a target charge retention characteristic, and some being above such target.

Accordingly, if a single cell in the test array experiences significant charge loss and results in a lower threshold voltage, the sub-threshold characteristics of the array will exhibit a different behavior than that expected for a completely uniform distribution. FIG. 2B illustrates a situation where a small number of cells are leaky (i.e., they have lost some charge) which results in a lower threshold voltage for such cells. These cells conduct current at a lower gate voltage V.sub.L before the main array begins to conduct current at a target gate voltage V.sub.target. The resulting sub-threshold characteristics will show a kink in the overall I-V curve, as illustrated in FIG. 2B. An observation of the heavy solid line of FIG. 2B, therefore, permits designers and manufacturers of flash cells to see and directly observe the size of this population, and study their characteristics, so that improvements (i.e., adaptations and refinements) can be made to the cell structure, or to the cell manufacturing process.

Through such improvements, of course, the cell populations should become more uniform and thus result in improved performance of devices utilizing such structures and processes. This is because the more uniform the distribution of cell charge populations (i.e., the more each cell is made to be identical to every other cell in the array) the more likely it is that various electrical operations occurring during the normal use of such cells in actual devices (such as erase and program) are going to be implemented correctly and successfully for each cell in the array.

In examining the solid line portion of the graph in FIG. 2B, however, it is apparent that users of such test structures are only able to glean information from one segment of the defective cell population. If, for example, some of the cells in this test structure array start to gain charge on their respective floating gate (as would be expected again as a result of manufacturing variations or as a result of defects and damage caused by operational stress), it will result in these cells having higher threshold voltages. These cells will start conduction at a higher gate voltage V.sub.H than the rest of the population. As shown in FIG. 2B, however, their contribution to the characteristics will not be noticed. This is because by the time these cells start to conduct a very small amount of subthreshold current, the rest of the population together, having a lower threshold voltage, will be conducting a current that is at least a few orders of magnitude higher than these cells. Thus, the charge gain behavior of cell populations in a Flash memory array cannot be observed using this kind of traditional test structure. This is true even though it is conceivable that the cell charge retention characteristics of a particular structure and process implementation may tend to result in a larger relative population of charge gainers versus charge losers.

Accordingly, a design that appears otherwise acceptable based on observations of prior art test structures may in fact have substantial charge gain defects that would go undetected. What is needed is an extension of the test structure to include the capability of evaluating the charge gain behavior, so that the true properties of various cell designs and processing techniques can be more accurately determined.

An object of the present invention, therefore, is to provide a test structure that is capable of evaluating and monitoring the charge-gain and charge-loss behavior of a flash memory array.

Another object of the present invention is to provide a test structure that permits designers and manufacturers of flash memory cells to better optimize cell designs and cell manufacturing techniques so that improved cell charge retention characteristics are achieved resulting in correspondingly improved cell voltage threshold populations.

A further object of the present invention is to provide a test structure that eliminates the need for a full function circuit to access each individual cell later in an actual device.

Yet another object of the present invention is to provide flash cell architecture designers, flash cell process designers and flash cell manufacturers with a method for evaluating and controlling the distribution of cell charge characteristics during design, test and manufacturing of such cells.

Another object of the present invention is to provide a method for manufacturing flash cells which results in more uniform voltage threshold characteristics so that operation of such devices is improved.

These and other objects are effectuated by the present invention, which includes a circuit for observing charge retention characteristics of cells in a memory array. A first portion of the circuit is used for monitoring charge loss of any of the cells while a second portion monitors charge gain of any of the cells. In this way, cells having conduction thresholds below a desired target threshold and cells having conduction thresholds above a desired target threshold can both be observed. The first portion of the circuit can be activated during a first period for observing cells having conduction thresholds below the target threshold, and the second portion can be activated during a second period for observing cells having conduction thresholds above the target threshold. In a preferred embodiment, the first and second portions are isolated from each other to ensure more accurate modeling.

Structurally, the first portion of the circuit is comprised of a number of flash cells each having a floating gate and a first type conductivity channel, and the second portion is comprised of a number of flash cells each having a floating gate and a second type conductivity channel opposite that of the first conductivity channel. The cells in both arrays have their own common sources, common drains, and common gates.

Corresponding ones of the floating gates of the first and second portions are connected to each other, which permits the evaluation of the charge-gain behavior. The first portion is a regular n-channel array, and the second portion is a mirror p-channel array which is inactivated except during testing of the n-channel array. In this manner, cells having conduction thresholds below the target threshold are observed during a first period in which only the n-channel devices are activated, and cells having conduction thresholds above the target threshold are observed during a second period in which only the p-channel devices are activated.

It can be seen that the present invention permits a more accurate evaluation of cell population voltage threshold distributions. The distribution can be studied to determine both positive and negative deviations between the actual cell voltage thresholds and target thresholds so that adjustments can be made during a design phase to a cell structure or cell process to minimize such deviations and therefore effectuate a more uniform distribution of voltage thresholds for said cells.

The invention is also useful during manufacture of a memory array using non-volatile memory cells, because cell population voltage threshold distributions can be improved by identifying and discarding defective cells. The regular array and a mirror array can be formed in the same substrate. In a later test stage of manufacturing, the voltage thresholds of the cells (i.e., whether they are above or below the target threshold) can be measured and evaluated to identify such defective cells. Integrated circuits made using such manufacturing processes will exhibit minimized threshold deviations and better performance and reliability since they will effectuate a more uniform distribution of voltage thresholds for the cells in the array.

FIG. 1 is a depiction of a prior art test structure for monitoring charge loss behavior of an array of flash memory cells; and

FIG. 2A is a graphical illustration of an I-V curve for an ideal array of memory cells which have a single, uniform target voltage threshold; and

FIG. 2B is a graphical illustration of an I-V curve for an actual array of memory cells in which a noticeable portion of such cells have either a sub-target threshold voltage or an over-target threshold voltage; and

FIG. 3 is an electrical diagram of an embodiment of the present invention; and

FIG. 4 is an I-V curve graphically illustrating how the present invention monitors charge gain behavior of flash memory cells.

To evaluate the charge gain behavior of cells in a Flash memory array, the applicants have devised a mechanism so that conduction for such cells initiates only at a higher threshold voltage. In another words, a physical phenomenon is selected wherein the device conduction is stronger when the floating gate potential is more negative, rather than more positive. A P-channel transistor exactly fits this type of conduction mechanism, in that a P-channel transistor conducts more current when the gate voltage is more negative with respect to the channel potential. For various reasons, however, a pure P-channel Flash cell would not be suitable to evaluate the charge-gain/charge-loss property, since a P-channel Flash cell by itself would most likely require a different set of program/erase mechanisms, and for that reason is likely to be incompletely correlated in behavior during normal operation with typical N-channel devices. Due to the different mechanisms used, the result might not be easily translated and be applicable to actual program/erase conditions an actual product array will experience, thus rendering the test results meaningless.

To overcome these roadblocks, the present invention uses the circuit shown in FIG. 3. As shown, there are two separate arrays: a regular, typical flash array inside the P-substrate (or P-well), and a mirror array inside an N-Well. These two arrays generally correspond to a first portion (N-channel) which, when activated, monitors charge loss of any of the cells, and a second portion (P-channel) for monitoring charge gain of any of said cells. Using these two test structures, the behavior of the entire array can be observed and more accurately characterized, because cells having conduction thresholds below a desired target threshold as well as cells having conduction thresholds above a desired target threshold can be identified.

From a structural perspective the present circuit shares some similarities with the prior art. The flash cells of the N-channel array have a common gate 100a, common source 100b, and a common drain 100c. Similarly, the flash cells of the mirror P-channel array have a common gate 300a, a common source 300b and a common drain 300c. These features make it possible to activate, cycle and stress all of the respective cells in each of these arrays simultaneously.

A key aspect of the present invention lies in the fact that floating gates of each Flash memory cell in the regular array (100, 101, etc.) are connected to corresponding floating gates of a corresponding cell (300, 301, etc.) in the mirror array. For ease of manufacturing, the source/drain implant of the mirror cell array inside the N-Well is implemented typically with a P+ implant. Thus the mirror cell array inside the N-Well is an array of P-channel Flash memory cells, but since they share their floating gates with the regular N-channel devices in the regular array, their electrical behavior is more closely reflective of the behavior of the N-channel devices.

To monitor both the charge loss and charge gain of the regular array, the following steps are taken. First, the N-channel array is cycled and stressed during a first period of time, so that a certain degree of charge is built up on the floating gates. During this first period, the P-channel array is left unconnected, and thus is not cycled or stressed. Following this step, the existence and behavior of cells in the regular N channel array having sub-threshold voltage characteristics can be determined in the same manner as previously performed in the prior art systems.

As explained above, during cycling and stressing of the regular array in the P-substrate, the mirror array inside the N-Well is not connected and has no impact on the regular array. During a third period, however, the P-channel devices are activated so that the identity and behavior of cells in the N-channel array which have higher threshold voltages can be monitored and measured. This part of the circuit performs this previously unavailable monitoring operation because of the following principle: if any cell inside the regular N channel array has a higher threshold voltage, then the potential on the floating gate for that cell will be more negative. The corresponding P-channel cell in the mirror array, which is connected through the common floating gate line, will be more conductive since its floating gate voltage is more negative with respect to the N-Well. Thus once the stress is done on the regular cell array, the mirror array is connected for charge gain evaluation while the regular array is left unconnected. The cell in the mirror array corresponding to the cell in the regular array with the most unintentional charge gain on its floating gate will conduct the most current.

FIG. 4 illustrates the characteristics that can be monitored and observed from the mirror array where a small number of cells in the regular array have gained some charge during the cycling or stressing conditions. These cells in the regular array have a higher threshold voltage than the rest of the population. Due to their higher threshold voltage, which is equivalent to a more negative floating gate potential, these cells in the mirror array will conduct first as the common control gate on the mirror array is ramped down from high positive voltage to negative voltage. In this fashion, the charge gain behavior of the regular array as a function of typical cycling conditions can be monitored and studied. It should be clear to those skilled in the art that with the mirror array unconnected, the test structure is just the same as in the one in the prior art. Thus the charge gain/loss behavior can be monitored by studying the actual regular array, and not a p-channel substitute that is likely to give inaccurate information. Thus, the circuit of the present invention has the unique capability to monitor both the charge gain and the charge loss behavior using a simple array test structure.

It can be further seen that the above circuit provides a more accurate evaluation of cell population voltage threshold distributions, heretofore unavailable since only one aspect of cell charge characteristics was easily determinable. By obtaining these distributions, both positive and negative deviations between cell voltage thresholds and target threshold can be measured. In this manner, cell structure designs, and cell manufacturing processes can be more accurately implemented because a more accurate test is available for measuring the true performance of such designs and processes.

This unique feature, therefore, permits the present invention to be used both at the development phase for designing the basic structures and processes for flash cells, as well as in the manufacturing phase as well. Since more accurate modeling is obtained for the behavior of the cells, it is easier to construct populations with tighter, more uniform voltage threshold distributions. The behavior of various combinations of cell structures and cell processes can be better characterized so that typical split lot testing will yield faster and more accurate information for proposed designs.

Similarly, in the manufacturing context, defective cells of both types (both leaky cells and stubborn cells) can be more easily identified. The regular array can be formed in a typical substrate, along with the second array of mirror cells. During a testing stage, cells having voltage thresholds either above or below a target threshold can be identified. Such cells can be disabled using conventional techniques, and replaced with repair cells so that the overall behavior of the array is improved. In some cases it may even be acceptable to include such testing and replacement capability onboard a fully functioning device in the field (after manufacture) as is known in the art. In this way, the cell voltage threshold optimization is performed dynamically, taking into consideration the fact that different cells may "age" differently and need replacement at different times.

In any event, finished integrated circuit articles embodying the present invention will exhibit superior performance since better, more uniform voltage threshold populations will be implemented during the manufacturing process.

The above test structure is provided merely by way of example, and is not intended to be limiting of the present invention in any respect. Other variations of the test structure for effectuating the same purpose will become apparent to those skilled in the art based on the discussion above. Furthermore, the sequence of steps for monitoring the charge-loss and charge gain is not critical, and can be performed in any order most suitable for the particular application.

It will also be appreciated by those skilled in the art that the above discussion can be applied to a number of different semiconductor manufacturing operations where considerations of charge-gain and charge loss are applicable, such as in EPROMs and the like. Also, it is apparent that the principles of the present invention could be used in any environment which would benefit from better control of voltage threshold characteristics of FETs. Accordingly, it is intended that the all such alterations and modifications be included within the scope and spirit of the invention as defined by the following claims.