METHOD FOR MANUFACTURING AN OXRAM TYPE RESISTIVE MEMORY CELL

Abstract

A method for manufacturing an OxRAM type resistive memory cell including a silicon oxide layer, the method including determining manufacturing parameter values enabling the resistive memory cell to have an initial resistance between 10.sup.7Ω and 3.Math.10.sup.9Ω; and forming on a substrate a stack successively including a first electrode, the silicon oxide layer and a second electrode, by applying the manufacturing parameter values.

Claims

1. A method for manufacturing an OxRAM type resistive memory cell comprising a silicon oxide layer, said method comprising: determining values of manufacturing parameters enabling the resistive memory cell to have an initial resistance comprised between 10.sup.7Ω and 3.Math.10.sup.9Ω; forming on a substrate a stack successively comprising a first electrode, the silicon oxide layer and a second electrode, by applying said manufacturing parameter values.

2. The method according to claim 1, wherein the manufacturing parameters are a thickness of the second electrode, a thickness of the oxide layer and a proportion of oxygen in the oxide layer.

3. The method according to claim 2, wherein the silicon oxide is porous and wherein the proportion of oxygen in the silicon oxide layer is comprised between 1.6 and 2.

4. The method according to claim 2, wherein the silicon oxide is porous and wherein the thickness of the silicon oxide layer is comprised between 4 nm and 7 nm.

5. The method according to claim 2, wherein the silicon oxide is porous and wherein the thickness of the second electrode is comprised between 3 nm and 7 nm.

6. The method according to claim 2, wherein the silicon oxide is non-porous and wherein the proportion of oxygen in the silicon oxide layer is comprised between 1 and 1.6.

7. The method according to claim 2, wherein the silicon oxide is non-porous and wherein the thickness of the silicon oxide layer is comprised between 3 nm and 4 nm.

8. The method according to claim 2, wherein the silicon oxide is non-porous and wherein the thickness of the second electrode is comprised between 4 nm and 6 nm.

9. The method according to claim 1, wherein the silicon oxide layer is formed by cathodic sputtering.

10. The method according to claim 1, wherein the first and second electrodes are formed by cathodic sputtering.

11. The method according to claim 1, wherein the first electrode is made of titanium nitride and the second electrode is made of titanium.

12. The method according to claim 1, wherein the resistive memory cell has an initial resistance comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω.

13. The method according to claim 3, wherein the proportion of oxygen in the silicon oxide layer is comprised between 1.8 and 1.9.

14. The method according to claim 6, wherein the silicon oxide is non-porous and wherein the proportion of oxygen in the silicon oxide layer is comprised between 1.2 and 1.4.

Description

BRIEF DESCRIPTION OF THE FIGURES

[0056] Other characteristics and advantages of the invention will become clear from the description that is given thereof below, for indicative purposes and in no way limiting, with reference to the appended figures, among which:

[0057] FIG. 1 schematically represents a first embodiment of a method for determining a manufacturing parameter value of a resistive memory cell;

[0058] FIG. 2 illustrates different deposition regimes during the sputtering of a silicon source in the presence of oxygen;

[0059] FIG. 3 represents, fora plurality of reference TiN/SiO.sub.x/Ti memory cells, the resistance in the high resistance state as a function of the initial resistance;

[0060] FIG. 4 represents, for the same reference memory cells, the resistance in the high resistance state as a function of the forming voltage.

[0061] For greater clarity, identical or similar elements are marked by identical reference signs in all of the figures.

DETAILED DESCRIPTION

[0062] [FIG. 1]: FIG. 1 represents steps S1 to S7 of a method for determining a value of at least one manufacturing parameter of a resistive memory cell, according to a first embodiment of the invention. When the resistive memory cell is manufactured following this parameter value, the programming window of the memory cell reaches a maximum value or a value close to the maximum value. The programming window PW (also called “memory window”) of a resistive memory cell is equal to the resistance of the cell in the high resistance state (HRS), hereafter noted R.sub.HRS, divided by the resistance of the same cell in the low resistance state (LRS), hereafter noted R.sub.LRS (PW=R.sub.HRS/R.sub.LRS).

[0063] The resistive memory cell of which it is sought to improve the programming window comprises a stack of thin layers (each of <100 nm thickness). Conventionally, this stack is formed on a substrate, for example made of silicon, and comprises: [0064] a first electrode 11 arranged on the substrate and hereafter called “lower electrode”; [0065] a layer of material with variable electrical resistance 12, also called “resistive material”, arranged on the first electrode 11; and [0066] a second electrode 13 arranged on the layer of resistive material and hereafter called “upper electrode”.

[0067] The resistive memory cell is preferably an oxide-based random access memory cell, commonly called “OxRAM”. The resistive material is then an oxide, for example a transition metal oxide (e.g. HfO.sub.2, Ta.sub.2O.sub.5, TiO.sub.2 . . . ) or a silicon oxide. The electrodes may be formed of doped silicon, a silicide, a metal (e.g. titanium, tantalum, tungsten . . . ) or a material with metallic character (also called metallic nature), such as titanium nitride (TiN) or tantalum nitride (TaN).

[0068] The first step S1 of the method consists in providing a number n of reference memory cells 10, where n is a natural integer greater than or equal to 2, preferably greater than or equal to 20. The greater the number n of reference memory cells 10, the more precise will be the determination method. To avoid needlessly complicating FIG. 1, only three reference memory cells 10 have been represented. The reference memory cells 10 and the resistive memory cell to manufacture comprise a same type of stack of thin layers. The stacks are designated of same type when the number of active layers is identical and the materials used are of same nature. For example, the stack of reference memory cells 10 (and the resistive memory cell to manufacture) comprises a lower electrode 11 made of titanium nitride, a layer of resistive material 12 made of silicon oxide (SiO.sub.x) and an upper electrode 13 made of titanium (TiN/SiO.sub.x/Ti type stack).

[0069] The reference memory cells 10 differ in the values of their manufacturing parameters. Among these manufacturing parameters may be cited as examples the thickness t.sub.OX of the oxide layer 12, the thickness t.sub.TE of the upper electrode 13 and the stoichiometric coefficient x of the oxide layer 12 (corresponding to a proportion of oxygen compared to the other elements forming the resistive material). The reference memory cells 10 may have different values of a same manufacturing parameter (for example the thickness t.sub.TE of the upper electrode) or different values of several manufacturing parameters. Each reference memory cell 10 is manufactured following a set of manufacturing parameters and at least one manufacturing parameter of each set differs from the other sets of parameters. In this sense, the reference memory cells 10 represent technological alternatives of a same stack of thin layers.

[0070] In the case of a TiN/SiO.sub.x/Ti type stack, the manufacturing parameters that vary between the n reference memory cells 10 are preferably the thickness t.sub.OX of the SiO.sub.x layer 12, the thickness t.sub.TE of the upper titanium electrode 13 and the proportion of oxygen x of the SiO.sub.x (the thickness of the lower electrode has no influence on the initial resistance, its thickness is for example of the order of 40 nm). For example, the thickness t.sub.OX of the SiO.sub.x layer 12 varies between 1 nm and 20 nm, the thickness t.sub.TE of the upper titanium electrode 13 varies between 1 nm and 20 nm and the proportion of oxygen x of the SiO.sub.x varies between 1 and 2. The silicon oxide may thus be sub-stoichiometric (x<2) or be silicon dioxide (x=2). The silicon dioxide is preferably porous, whereas the sub-stoichiometric silicon oxide may be porous or non-porous (i.e. exempt of pores).

[0071] In the case of a metal “high-k” dielectric material/metal stack, where the metal of the upper electrode (e.g. Hf, Ti, Ta . . . ) plays the role of oxygen scavenger, also called oxygen getter, for the “high-k” dielectric material (i.e. with high dielectric permittivity, e.g. HfO.sub.2, TiO.sub.2, Ta.sub.2O.sub.5 . . . ), the thickness of the layer of “high-k” dielectric material can vary between 1 nm and 20 nm and the thickness of the upper electrode (oxygen scavenger layer) can vary between 1 nm and 20 nm. In the case of a “high-k” dielectric material of metal oxide “MO.sub.x” type, where M is a transition metal (e.g. Hf, Ti, Ta . . . ), the proportion of oxygen x may further vary between 1 and a value corresponding to the stoichiometric oxide (x=2 for HfO.sub.2 or TiO.sub.2, x=2.5 for Ta.sub.2O.sub.5 . . . ).

[0072] The initial resistance R.sub.i of each reference memory cell 10 is next measured during a step S2. The initial resistance is the electrical resistance obtained at the end of the manufacture of the memory cell, before the conductive filament is formed for the first time (in other words, before the “forming” step). The initial resistance R.sub.i may be measured by applying a measuring voltage U.sub.1 (for example 100 mV) between the electrodes 11 and 13 of the memory cell 10, by measuring the current I.sub.1 of the cell (through the oxide layer 12) subjected to this voltage U.sub.1 then by calculating the ratio of the measuring voltage U.sub.1 over the measured current I.sub.1 (R.sub.i=U.sub.1/I.sub.1).

[0073] The reference memory cells 10 are next programmed in the high resistance state (“HRS”) during a step S3. A first so-called “forming” voltage (for example of the order of 3 V) is applied between the electrodes of the memory cells 10 to activate the resistive material and to place the memory cells 10 in the low resistance state (“LRS”), then a second so-called erasing voltage, of lower absolute value than the first voltage is applied to switch the reference memory cells 10 from the low resistance state to the high resistance state (the erasing voltage is generally negative, for example comprised between −1 V and −2 V).

[0074] Then, the resistance in the high resistance state R.sub.HRS is measured for each reference memory cell 10 during a step S4. In an analogous manner to the initial resistance R.sub.i, the resistance R.sub.HRS may be measured by applying a measuring voltage U.sub.2 (for example 100 mV) between the electrodes of the memory cell 10 (in the high resistance state), by measuring the current I.sub.2 of the cell subjected to this voltage U.sub.2 then by calculating the ratio of the measuring voltage U.sub.2 over the measured current I.sub.2 (R.sub.HRS=U.sub.2/I.sub.2).

[0075] At step S5, a relationship R.sub.HRS(R.sub.i) between the resistance in the high resistance state R.sub.HRS and the initial resistance R.sub.i is established from the resistance values R.sub.i and resistance values R.sub.HRS measured respectively during steps S2 and S4. For example, the resistance values R.sub.HRS and R.sub.i of the reference memory cells 10 may be plotted on a graph. Each point of the graph corresponds to a reference memory cell 10 and thus to a technological alternative of the stack (i.e. a combination of technological parameters). The points of the graph are next described, during a so-called fitting operation, by a curve or an equation of the type R.sub.HRS=f(R.sub.i). The relationship R.sub.HRS(R.sub.i) may thus take the form of a curve or an equation. The relationship between the resistance in the high resistance state R.sub.HRS and the initial resistance R.sub.i is preferably written in the form of a second degree polynomial, with as variable the logarithm of the initial resistance R.sub.i.

[0076] Step S6 consists in determining, using the relationship R.sub.HRS(R.sub.i), at least one value R.sub.i_opt of the initial resistance R.sub.i for which the resistance in the high resistance state R.sub.HRS is greater than or equal to a predetermined target value R.sub.HRS_tg. This target value R.sub.HRS_tg may be defined as a function of a target value of the programming window (preferably the maximum) or may be equal to a percentage of the maximum of the resistance in the high resistance state R.sub.HRS (for example 90% of the maximum of the resistance R.sub.HRS). The maximum of the resistance R.sub.HRS may be deduced from the relationship R.sub.HRS(R.sub.i) established at step S5.

[0077] In this first embodiment of the method, the resistance in the low resistance state R.sub.LRS of the reference memory cells 10 is assumed constant (and thus independent of the technological parameters). Indeed, the resistance R.sub.LRS of OXRAM cells programmed in the low resistance state is controlled by the programming current in the low resistance state. For example, for a TiN/SiO.sub.x/Ti type stack, the resistance R.sub.LRS is equal to around 10.sup.4Ω when the programming current is equal to around 100 μA. A maximum of the resistance in the high resistance state R.sub.HRS then corresponds to a maximum of the programming window.

[0078] One value, several distinct values or a range (continuous) of values of the initial resistance R.sub.i may thus be obtained at the end of step S6, depending on the target value selected or the resistance values R.sub.HRS taken into consideration (greater than the target value R.sub.HRS_tg and/or equal to the target value R.sub.HRS_tg). All these values may be qualified as “optimal” or “optimised” in so far as they make it possible to approach or even reach a maximum of the programming window.

[0079] Finally, at step S7, at least one optimal value t.sub.TE_opt/t.sub.OX_opt/x.sub.opt of one or more manufacturing parameters is determined from the optimal initial resistance value R.sub.i_opt (or optimal values). These manufacturing parameters are not necessarily the same as those that differentiate the reference memory cells 10. They are preferably selected from among the thickness t.sub.OX of the oxide layer 12, the thickness t.sub.TE of the upper electrode 13 and the proportion of oxygen x in the oxide layer 12.

[0080] In a preferential embodiment of step ST the values of all the manufacturing parameters having an influence on the initial resistance R.sub.i are determined from the optimal initial resistance value R.sub.i_opt. In an alternative embodiment, the values of a part only of these manufacturing parameters are determined from the optimal initial resistance value R.sub.i_opt. The values of the other manufacturing parameters (including those not having any influence on the initial resistance, for example the thickness t.sub.BE of the lower electrode 11, the role of which is to ensure good electrical contact) may be determined in another manner. They may notably be imposed by integration constraints.

[0081] The optimal value of a manufacturing parameter may be determined from an optimal initial resistance value R.sub.i_opt knowing the dependency of this parameter on the initial resistance R.sub.i. For example, the initial resistance R.sub.i of a resistive memory cell increases with the thickness t.sub.OX of the oxide layer 12 and with the proportion of oxygen x. Conversely, it decreases when the thickness t.sub.TE of the upper electrode 13 increases (up to a certain threshold).

[0082] An experimental design may be implemented in order to establish dependency relationships between the initial resistance R.sub.i and the different manufacturing parameters. This experimental design may notably consist in varying the aforementioned three manufacturing parameters (thickness t.sub.OX of the oxide layer 12, thickness t.sub.TE of the upper electrode 13 and proportion of oxygen x in the oxide layer 12), preferably by combining all the parameter values, and by measuring the initial resistance corresponding to each set of values.

[0083] In the case of the TiN/SiO.sub.x/Ti stack, the following relationships have been obtained by setting two parameters then by varying the final parameter (with R.sub.i in Ω, x without units, t.sub.OX and t.sub.TE in nm):

log(R.sub.i)=12.6.Math.x−16.6  [Math 1]

[0084] with t.sub.OX=t.sub.TE=5 nm.

[0085] Thus, equation Math 1 above expresses the variation in the initial resistance R.sub.i as a function of the stoichiometric coefficient x of oxygen and where the thicknesses t.sub.OX of the oxide layer 12 and t.sub.TE of the upper electrode 13 have been set at 5 nm.

R.sub.i=4.Math.10.sup.−6×exp(5.4099−t.sub.OX)  [Math 2]

[0086] with t.sub.TE=5 nm and x=1.8.

[0087] Thus, equation Math 2 above expresses the variation in the initial resistance R.sub.i as a function of the thickness t.sub.OX of the oxide layer 12 and where the stoichiometric coefficient x of oxygen has been set at 1.8 and the thickness t.sub.TE of the upper electrode 13 has been set at 5 nm.

R.sub.i=9.Math.10.sup.9×exp(−0.97.Math.t.sub.TE)  [Math 3]

[0088] with t.sub.OX=5 nm and x=1.9.

[0089] Thus, equation Math 3 above expresses the variation in the initial resistance R.sub.i as a function of the thickness t.sub.TE of the upper electrode 13 and where the thickness t.sub.OX of the oxide layer 12 has been set at 5 nm and where the stoichiometric coefficient x of oxygen has been set at 1.9.

[0090] The above equations have been obtained from experimental values and are dependent on the deposition equipment used.

[0091] When several optimal initial resistance values R.sub.i_opt are available, several optimal values of the manufacturing parameter(s) may be obtained.

[0092] In a second embodiment of the method, not represented by the figures, it is assumed that the resistance in the low resistance state R.sub.LRS of the reference memory cells 10 varies. The method then comprises, in addition to steps S1-S4 described previously, a step of programming the reference memory cells 10 in the low resistance state, a step of measuring the resistance R.sub.LRS of the reference memory cells 10 in the low resistance state and a step of calculating the programming windows of the reference memory cells 10 from the measured resistances values R.sub.LRS and R.sub.HRS. The resistance R.sub.LRS of the reference memory cells 10 in the low resistance state is advantageously measured before step S3 of programming the reference memory cells 10 in the high resistance state, after the forming step (which thus constitutes the step of programming the reference memory cells 10 in the low resistance state).

[0093] Instead of determining a relationship R.sub.HRS(R.sub.i) between the resistance in the high resistance state R.sub.HRS and the initial resistance R.sub.i, at step S5 a relationship is determined between the programming window and the initial resistance. A target value of the programming window is then considered during step S6 (instead of a target value of the resistance in the high resistance state R.sub.HRS).

[0094] An exemplary embodiment of the determination method according to the invention will now be described.

[0095] The resistive memory cell of which it is sought to optimise the programming window as well as the reference memory cells 10 provided for this purpose comprise the stack of TiN/SiO.sub.x/Ti thin layers described previously.

[0096] The silicon oxide is in this example porous and has been obtained by reactive cathodic sputtering in a vacuum deposition chamber. The deposition chamber is equipped with a silicon target and comprises two gas inlets, one for oxygen (O.sub.2), the other for a neutral gas such as argon. The sputtering reactor comprises a direct current (DC) voltage generator and a magnetron. The bias of the source supplied by the DC generator is advantageously pulsed. The parameters having an influence on the proportion of oxygen x of SiO.sub.x are the power applied by the DC generator, the working pressure, the flow rates of the neutral gas and oxygen, the frequency, the ratio T.sub.ON/T.sub.REV of the duration of the deposition phases (“ON” state of the generator) over the duration of the electrostatic discharge phases (“OFF” state of the generator) and the duty cycle of the pulses of the DC generator (equal to T.sub.ON/(T.sub.REV+T.sub.ON)).

[0097] [FIG. 2]: FIG. 2 represents the effect of the bias voltage applied to a silicon target (by the DC generator) as a function of the flow rate of oxygen entering into the deposition chamber (expressed in sccm, the abbreviation for “Standard Cubic Centimetres per Minute”, i.e. the number of cm.sup.3 of gas flowing per minute under standard pressure and temperature conditions, i.e. at a temperature of 0° C. and a pressure of 1013.25 hPa) on the state of the silicon target. The relationship between the bias voltage of the target and the flow rate of oxygen forms a hysteresis that sets the state of the silicon target: amorphous silicon (a-Si) for low flow rates of oxygen (<7 sccm), sub-stoichiometric silicon oxide (SiO.sub.x, with x comprised between 1 and 2 excluded) for intermediate flow rates of oxygen (7-18 sccm) and silicon dioxide (SiO.sub.2) for high flow rates of oxygen (>18 sccm). The stoichiometry of the silicon oxide deposited may thus be controlled thanks to the flow rate of oxygen entering into the deposition chamber.

[0098] Eight reference memory cells have been manufactured following different manufacturing parameter values listed in table 1 below. The stoichiometry x of the SiO.sub.x is controlled via the flow rate of oxygen injected into the chamber. The other deposition parameters are identical between the 8 reference memory cells (temperature in the chamber: 25° C.; power of the DC generator 1 MW, main flow rate of argon: 50 sccm; flow rate of argon on the rear face of the substrate: 15 sccm; pressure in the chamber: 1 to 3 mTorr depending on the flow rate of oxygen; valve of the cryogenic pump in intermediate position).

TABLE-US-00001 TABLE 1 Flow rate of t.sub.OX oxygen Proportion of t.sub.TE R.sub.i R.sub.HRS cell no (nm) D.sub.O2 (sccm) oxygen x (nm) (Ω) (Ω) 1 5 6 1.9 5 5E6 5E4 2 7 6 1.9 5  4E10 1E5 3 5 6 1.9 7 9E7 8E5 4 7 6 1.9 7  3E10 9E4 5 5 7 2 7 2E8 1E6 6 7 5 1.8 5  4E10 7E4 7 5 7 2 5 1E9 1E6 8 5 5 1.8 5 3E6 1E5

[0099] Table 1 also gives for these 8 reference memory cells the measured values of the initial resistance R.sub.i and the resistance in the high resistance state R.sub.HRS. The resistance in the low resistance state R.sub.LRS is assumed constant and equal to 10.sup.4Ω. The relationship that links the values of flow rate of oxygen D.sub.O2 (between 4 sccm and 7 sccm) and the values of the proportion of oxygen x is the following:

D.sub.O2=10.Math.x−13  [Math 4]

[0100] For a flow rate of oxygen D.sub.O2 greater than or equal to 7 sccm, the proportion of oxygen x is equal to 2.

[0101] [FIG. 3]: FIG. 3 is a graph on which have been plotted the measured resistance values R.sub.i and R.sub.HRS of the 8 reference memory cells. These points have next been extrapolated by means of a curve C. The equation of curve C (obtained experimentally) is the following:

R.sub.HRS=−2.2.Math.10.sup.5.Math.(log(R.sub.i)).sup.2+4.Math.10.sup.6.Math.log(R.sub.i)−2.Math.10.sup.7  [Math 5]

[0102] with R.sub.i and R.sub.HRS in Ω.

[0103] Curve C, bell or parabolic shaped, shows that there exists a maximum of resistance in the high resistance state R.sub.HRS—and thus a maximum of the programming window—for an initial resistance R.sub.i of around 10.sup.8Ω. An explanation for this bell shaped dependency could be the following: at low initial resistance R.sub.i, it is not possible to reach a high resistance value R.sub.HRS due to an intrinsic limitation of the resistance of the memory cell. At high initial resistance R.sub.i, a high forming voltage is necessary to be able to use the memory cell and this high voltage generates an important quantity of defects in the SiO.sub.x layer. Since the defects are still present during the erasing of the memory cell (return to the high resistance state caused by a dissolution of the conductive filament), the resistance of the high resistance state is reduced.

[0104] According to curve C of FIG. 3, an initial resistance R.sub.i comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω corresponds to a resistance in the high resistance state RRs greater than or equal to R.sub.HRS_tg1=10.sup.6Ω. To obtain a programming window greater than or equal to 100, the resistive memory cell (TiN/SiO.sub.x/Ti) will thus be manufactured such that its initial resistance R.sub.i is comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω.

[0105] Multiple combinations of parameter values exist to obtain an initial resistance R.sub.i comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω. As indicated previously, one of these combinations may be obtained by setting a parameter (for integration reasons for example), then by varying the two other parameters. To facilitate the search for an initial resistance R.sub.i comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω, and given the targeted application, the thickness t.sub.OX_opt of the oxide layer 12 may be set at a value comprised between 4 nm and 7 nm. In an alternative or additional manner, the thickness t.sub.TE_opt of the upper electrode 13 may be set at a value comprised between 3 nm and 7 nm. In an alternative or additional manner, the concentration of oxygen x.sub.opt may be set at a value comprised between 1.6 and 2 (i.e. a flow rate of oxygen comprised between 5 sccm and 8 sccm), preferably between 1.8 and 1.9.

[0106] [FIG. 4]: FIG. 4 represents the resistance values in the high resistance state R.sub.HRS of the 8 preceding reference memory cells, associated with the forming voltage values V.sub.f that have been applied to these cells. This figure shows as a comparison that, when the manufacturing parameters are adjusted with the aim of reaching a forming voltage less than or equal to 2 V (typical value to be compatible with the power supply circuit of the memory), a programming window is obtained around ten times less than the maximum programming window (reached fora forming voltage of around 3 V). The determination method according to the invention thus enables a significant improvement in the programming window of resistive memory cells compared to current practice.

[0107] Still according to curve C of FIG. 3, an initial resistance R.sub.i comprised between 10.sup.7Ω and 3.Math.10.sup.9Ω corresponds to a resistance in the high resistance state R.sub.HRS greater than or equal to R.sub.HRS_tg2=5.Math.10.sup.5Ω. To obtain a programming window greater than or equal to 50, the resistive memory cell (TiN/SiO.sub.x/Ti) will thus be manufactured such that its initial resistance R.sub.i is comprised between 10.sup.7Ω and 3.Math.10.sup.9Ω.

[0108] The silicon oxide SiO.sub.x may also be non-porous and sub-stoichiometric (x<2). To facilitate the search for an initial resistance R.sub.i comprised between 3.Math.10.sup.7Ω and 10.sup.9Ω, and given the targeted application, the thickness t.sub.OX_opt of the oxide layer 12 may be set at a value comprised between 3 nm and 4 nm. In an alternative or additional manner, the thickness t.sub.TE_opt of the upper electrode 13 may be set at a value comprised between 4 nm and 6 nm. In an alternative or additional manner, the concentration of oxygen x.sub.opt may be set at a value comprised between 1 and 1.6, preferably between 1.2 and 1.4.

[0109] More generally, an extrapolation of curve C makes it possible to determine the thickness of the oxide layer t.sub.OX (non-porous) fora desired value of R in the following manner: log(R.sub.i)=1.1.Math.t.sub.OX+0.7 where the thickness of the upper electrode t.sub.TE has been set at 10 nm and where the stoichiometric coefficient x has been set at 1.2.

[0110] In the same way, an extrapolation of curve C makes it possible to determine the stoichiometric coefficient x for a desired value of R.sub.i in the following manner: log(R.sub.i)=19.Math.x+16 where the thickness of the oxide layer t.sub.OX (non-porous) has been set at 3 nm and where the thickness of the upper electrode t.sub.TE has been set at 5 nm.

[0111] Another aspect of the invention relates to a method for manufacturing a resistive memory cell, and more particularly of an OxRAM memory cell comprising a TiN/SiO.sub.x/Ti type stack.

[0112] The manufacture of the resistive memory cell successively comprises a step of deposition of the lower electrode 11 on a substrate (for example made of silicon), a step of deposition of the oxide layer 12 on the lower electrode 11 and a step of deposition of the upper electrode 13 on the oxide layer 12. By following at each step the parameter value(s) obtained using the determination method according to the invention, the resistive memory cell will have a large programming window.

[0113] The silicon oxide (whether it is porous or non-porous) of the TiN/SiO.sub.x/Ti stack may be obtained by cathodic sputtering of a silicon source in the presence of oxygen. The lower titanium nitride electrode and the upper titanium electrode may be formed by cathodic sputtering (reactive in the case of TiN).

[0114] It will be noted that the invention is not limited to the embodiments described with reference to the figures and alternatives could be envisaged without going beyond the scope of the invention.