Apparatus and Method for Improving the Endurance of Floating Gate Devices

Floating gate MOS transistors making use of Fowler-Nordheim tunneling of electrons are useful in a plurality of applications, particularly programmable memory devices such as EPROMs and EPALs. In typical applications, a large number of floating gate MOS transistors are formed onto a substrate in an integrated circuit thereby forming a high-density memory array. The reliability of such an array of programmable memory devices is generally dependent upon the failure of single, isolated bits in the array.

FIG. 1a shows a typical floating gate transistor 10 having source 12 and drain 14, generally formed of N+ doped regions; floating gate 15, generally formed of a group I polysilicon; control gate 16, generally formed of a group II polysilicon; and dielectric silicon oxide (SiO.sub.2) layer 18.

To be useful, the floating gate transistor 10 is charged and discharged in accordance with its use in the particular application.

In order to charge or program a particular floating gate transistor, a programming voltage V.sub.pp, on the order of 5-20 volts, is applied to the control gate electrode 16, while source 12 and drain 14 are held at ground or allowed to float. Upon applying V.sub.pp to the control gate 16, capacitive coupling between the control gate 16 and the floating gate 15 causes the floating gate 15 to acquire electrons from the drain region 14 through the process of Fowler-Nordheim tunneling. When V.sub.pp is removed, the floating gate 15 assumes a negative voltage due to the presence of the trapped electrons on the floating gate 15.

Discharge of electrons from the floating gate is generally accomplished by grounding the control gate 16, and applying the programming voltage V.sub.pp to the drain 14 while allowing the source 12 to float.

In a typical high-density memory array, a certain percentage of floating gate transistors used in the memory cells will exhibit atypical charge/discharge characteristics. These atypical cells comprise "weak" bits in the memory array. An atypical cell is defined as a cell which, after a number of charge/discharge cycles, fails to charge/discharge properly because of a breakdown of the tunnel dielectric caused by excessive electron tunneling. A typical bit performs charge/discharge functions for the period of time expected by the hardware designer, taking into account the particular materials used and the parameters of the floating gate transistor, such as oxide thickness, length, width, and implantation concentration. In general, the endurance of a typical bit in the array is generally orders of magnitude better than those of the weak bit.

The endurance of an oxide layer depends on the quality of the SiO.sub.2 tunneling dielectric, which is a function of the amount of charge which can be made to pass through the dielectric before breakdown of the floating gate occurs. This is normally referred to as "fluence" or Q.sub.BD. When the term "endurance" is used with respect to a memory array, the term refers to the length of time the particular array will function before a failure in a "weak" bit occurs.

A failure of any particular weak bit in a memory array thus reduces the endurance of the array. It is then desirable to reduce the failure rate of weak bits within an array to improve the overall endurance of the memory chip.

It is thus an object of the invention to improve the reliability of a floating gate transistor.

It is a further object of the invention to improve the endurance of a memory array using floating gate transistors.

It is another object of the present invention to provide the above objects in an electrically erasable programmable memory cell.

It is a further object of the invention to improve the endurance of a floating gate transistor by controlling the tunneling current through the dielectric layer of the transistor during discharge of the floating gate.

It is a further object of the invention to provide the above objects in a floating gate transistor configuration which optimizes the electric field potential at the tunnel region of the dielectric insulating layer during discharge of the floating gate.

The invention comprises an apparatus and method for improving the endurance and reliability of a floating gate transistor often used in memory applications. Specifically, a method and means are provided for controlling the electric field induced across the tunnel oxide region of a floating gate memory transistor when discharging electrons from the floating gate. By controlling the electric filed across the tunnel region of a particular floating gate apparatus during the discharge of electrons from the floating gate, the tunneling current through the oxide region is optimized for the particular gate oxide thickness of individual floating gate transistors. In large memory arrays wherein a multitude of floating gate devices are formed on a particular substrate, those devices with thinner-than-expected oxide regions in localized areas resulting in high voltage potentials during discharge, and thus earlier floating gate breakdown under previous conditions, have been better optimized for higher localized electric fields during discharge.

The apparatus of the invention can utilize a variable resistance coupled to the drain electrode of the floating gate device or a variable voltage source coupled to the source electrode. In the preferred embodiment of the invention wherein the invention is utilized in an EEPROM memory cell, both a variable resistance means and variable voltage source are utilized simultaneously. In such a preferred embodiment, the resistance means is coupled to the drain electrode of the select transistor of the EEPROM, while the variable voltage source is coupled to the source electrode of the floating gate transistor.

The invention will be described with respect to the specific embodiments thereof. Other objects, embodiments, and features of the invention will be apparent from a reading of the detailed description in conjunction with the drawings.

FIG. 1a is a schematic diagram of a configuration of a floating gate transistor;

FIG. 1b is a schematic diagram showing a floating gate transistor in an electrically erasable, programmable, read only memory (E.sup.2 PROM) demonstrating a prior-art method for discharging a floating gate transistor in an E.sup.2 PROM;

FIG. 2a is a schematic diagram showing a first embodiment of the present invention in an E.sup.2 PROM cell;

FIG. 2b is a graph representing the relationship between the cell threshold voltage (V.sub.TD) on discharge and the voltage potential applied to the source of the floating gate (V.sub.s).

FIG. 3a shows a second embodiment of the present invention in an E.sup.2 PROM;

FIG. 3b is a graph representing the relationship between the cell threshold voltage (V.sub.TD) on discharge and a resistance (R.sub.D) applied at the drain of a select transistor in an E.sup.2 PROM; and

FIG. 4 shows the preferred embodiment of the present invention in an E.sup.2 PROM.

The subject matter of the invention will be described herein with reference to use of the floating gate transistor in an electrically erasable programmable read-only memory (E.sup.2 PROM) cell. However, upon reading of the detailed description of the invention, it will be apparent to a person of ordinary skill that the invention is applicable in numerous applications wherein a floating gate transistor using Fowler-Nordheim tunneling is used, for example EPALs, EPLDs and Flash-EEPROMs.

It has been determined that Q.sub.BD is typically much lower during the discharge of electrons from the floating gate into the doped drain region than it is during the charging process. Thus, the amount of charge which can be made to pass through the dielectric before breakdown of the tunneling dielectric occurs is lower on discharge than on charge. Further, the charge on the floating gate 15 is related to the external cell threshold voltage, V.sub.T.

The probable cause of single floating gate transistor failures in atypical cells in a typical memory array arise from:

(1) a cell with a higher-than-average ratio of control gate--floating gate capacitance over the total capacitance or (C.sub.g /C.sub.t); or

(2) a cell with a lower-than-average ratio of drain--floating gate capacitance over the total capacitance of the cell (C.sub.d /C.sub.t): or

(3) a cell with a thinner-than-typical tunnel dielectric thickness (caused by local variations in cell oxide thickness during formation of the chip). These characteristics give rise to a higher-than-typical electric field across the tunneling dielectric during charge/discharge cycling.

With reference to FIG. 1b, a floating gate transistor 30 is shown in a typical application--an E.sup.2 PROM cell. The E.sup.2 PROM cell generally comprises the floating gate transistor 30 with source 32, drain 34, and control gate 36, and a MOS select transistor 20 having source 22, drain 24, and gate 26.

For purposes of this discussion and the examples shown herein, the floating gate transistor 30 is assumed to have an intrinsic threshold voltage of V.sub.TO .perspectiveto.0 . The charged threshold value V.sub.TC is, therefore, greater than zero and the discharged threshold value V.sub.TD is less than zero, due to the presence and absence, respectively, of charge on the floating gate. In practice, an intrinsic V.sub.TO .perspectiveto. 0 is not essential; the intrinsic V.sub.T of the floating gate can be either positive or negative.

FIG. 1b demonstrates the prior-art method for discharging the electrons stored on the floating gate 35 of a floating gate transistor 30 when used in an E.sup.2 PROM cell. A program voltage V.sub.pp on the order of 15-20 volts is applied to the drain electrode of select transistor 20. Typically, gate 26 of the select transistor 20 is generally coupled to a word line (not shown) of an array of E.sup.2 PROM cells. A voltage V.sub.wl, greater than or equal to the programming voltage V.sub.pp, is applied to the gate 26 of the select transistor via the word line rendering select transistor 20 conductive. The gate 36 of floating gate transistor 30 is held at ground while the source 32 of floating gate transistor 30 is allowed to float or also held at ground.

Upon application of V.sub.pp and V.sub.WL, the electrons will be discharged from the floating gate transistor 30. In the case of an atypical cell, discharge of the floating gate occurs more quickly than in typical cells. Thus, while there is no need for the continued application of V.sub.pp to discharge the floating gate 35, because the duration of application of this voltage is standard for both typical and atypical cells, the application of V.sub.pp will continue.

If V.sub.s floats, then atypical cells will, after the discharge pulse V.sub.pp has been applied for a set time, exhibit a more negative V.sub.TD (or greater .vertline.V.sub.TD .vertline.). In such situations, Fowler-Nordheim tunneling across the tunnel dielectric will continue until such time as either (1) the discharge pulse is terminated, or (2) the floating gate voltage is sufficiently positive that tunneling across the tunnel dielectric either ceases or becomes negligibly small. This continued application of V.sub.pp results in degradation of the floating gate eventually leading to a failure of the "weak bit".

FIG. 2a shows a first embodiment of the present invention wherein a floating gate transistor 30 is used in an E.sup.2 PROM cell. The E.sup.2 PROM cell is essentially the same as the E.sup.2 PROM cell shown in FIG. 1b, however, an adjustable voltage source V.sub.s is coupled to the source of the floating gate transistor 30.

The V.sub.T of each cell is related to the amount of charge on the floating gate as discussed above. Coupling voltage source V.sub.s to source 32 of the floating gate transistor 30 allows one to lower the electric field across the tunneling region, and lower the .vertline.V.sub.TD .vertline. for atypical cells.

In the region where potential applied by voltage source V.sub.s is in the range of 0.ltoreq.V.sub.s <5 volts, the source potential controls the linear turn-on point of the E.sup.2 PROM cell which in turn controls the voltage on the drain tunnel region. (FIG. 2b.)

The effect of controlling voltage source V.sub.s during discharge is shown in FIG. 2b. FIG. 2b is a graph representing the relationship of V.sub.TD versus V.sub.s. By controlling the linear turn-on point of the E.sup.2 PROM for atypical cells, .vertline.V.sub.TD .vertline., is reduced because less tunneling occurs due to the lower electric field caused by using V.sub.s control.

In effect, the source voltage V.sub.s is capacitatively coupled onto the floating gate 35 of the floating gate transistor 30. Both floating gate transistor 30 and select transistor 20 have an effective resistance associated with them. As V.sub.s at source 32 is increased, the resistance of the total E.sup.2 PROM cell increases and thus the voltage appearing on the drain 34 of floating gate transistor 30 also increases. An increase in V.sub.s leads to a higher tunneling field and more negative discharge threshold voltages (V.sub.TD). For V.sub.s> 5 volts, capacitive coupling of the source voltage becomes significant and the discharge threshold voltage (V.sub.TD) of the cell begins to increase.

FIG. 3a shows a second embodiment of the present invention for controlling tunneling current through the oxide layer of a floating gate transistor during discharge. FIG. 3 a shows a variable resistance R.sub.D coupled in series to the drain 24 of the select transistor 20 in an E.sup.2 PROM cell. Applying a resistance in series with the floating gate transistor drain allows for control of current at drain 34 and through floating gate transistor 30. In this manner, control of the tunneling current through the oxide layer in the tunneling region 37 of the floating gate transistor 30 is achieved. In the embodiment shown in FIG. 3a, the variable resistance R.sub.D can comprise, for example, an external resistance or may be fabricated on the same substrate as the E.sup.2 PROM cell. Variable resistance R.sub.D could also, for example, comprise the select transistor 20. It should also be noted that, while FIG. 3a shows the source 32 of floating gate transistor 30 coupled directly to ground, in practice the source 32 may be at virtual ground, and in fact coupled to high capacitative loads which act as a virtual ground.

FIG. 3b graphically shows the relationship between the variable resistance R.sub.D and the cell threshold voltage on discharge V.sub.td obtained for the following values: R.sub.D =10K.OMEGA. to 1M.OMEGA.; V.sub.pp =14.5 volts applied for 1200 .mu.s; V.sub.wl =20 volts; and V.sub.s at ground.

FIG. 4 shows the preferred embodiment for implementing the present invention. Both a variable resistance R.sub.D coupled to the drain 24 of the select transistor 20 and variable voltage source V.sub.s are coupled to the source 32 of the floating gate transistor 30. By controlling both source voltage V.sub.s and drain current R.sub.D, the tunneling current at the tunnel region of the dielectric layer of a floating gate transistor 30 can be optimized. Experimental results using a varying resistance R.sub.D and varying V.sub.s between 0 and 5 volts on a population of E.sup.2 PROM cells which had seen about 200 charge/discharge cycles yielded optimal values for R.sub.D =100 Kohms and for V.sub.s =5 volts to obtain the least standard deviation amongst .vertline.V.sub.TD .vertline. for the E.sup.2 PROM cells. The triple standard deviation of V.sub.TD for all cells in the array using the aforesaid optimal values was 0.75 volts.

Using the source bias and series resistance coupling method of the proposed invention, .vertline.V.sub.TD .vertline. for atypical cells can be reduced, thus causing the floating device to conduct earlier and the field across the tunneling dielectric to collapse.

Thus, under the proposed method, an atypical cell (or "weak" bit) will have a cell threshold voltage, on discharge (V.sub.TD) which is much closer to that of typical cells because the maximum field across the tunneling oxide regions of the atypical cells is reduced. This leads to a higher endurance for each atypical floating gate transistor with thinner-than-average Si02 layers and such atypical transistors disposed within a typical memory array will not fail as quickly as with prior-art discharge methods.

Furthermore, once the floating gate becomes discharged, the transistor turns "on" thereby collapsing the electric field across the gate "automatically". Effectively, the total tunneling current through the oxide layer is "self limited" by the floating gate transistor's lower .vertline.V.sub.TD .vertline..

The invention has been described with respect to particular embodiments thereof. Numerous variations are possible as will be apparent to a person of ordinary skill in the art after reading the present specification. For example, though the invention has been particularly described with reference to an E.sup.2 PROM cell, those skilled in the art will recognize that the invention is applicable to all particular applications using floating gate transistors operating with Fowler-Nordheim tunneling through an oxide region. This variation and others are intended to be within the scope of the present invention as defined by the claims.