Non-contact electrical conduction measurement for insulating films

[0002] This application is a divisional application of U.S. patent application Ser. No. 08/841,501, now U.S. Pat. No. 6,097,196 filed Apr. 23, 1997.

[0003] The present invention relates to the field of semiconductor wafer testing and, more particularly, to the measurement of the electronic conduction behavior of various insulating layers.

[0004] The production of insulating layers, particularly, thin oxide layers, is basic to the fabrication of integrated circuit devices on semiconductor wafers. A variety of insulating dielectric layers are used for a wide range of applications. These insulating layers can be used, for example, to separate gate layers from underlying silicon gate regions, as storage capacitors in DRAM circuits, for electrical device isolation and to electrically isolate multilayer metal layers. The electrical insulating properties of these layers is of great interest. Some of the measures of the electrical insulating quality of these layers are (1) the conduction current at a given applied voltage or applied electric field strength, (2) the voltage or electric field strength corresponding to a given applied conduction current and (3) the terminal value of a saturating increase in voltage or electric field strength (tunneling field) corresponding to a regime of rapidly increasing conduction with increasing voltage or field.

[0005] Another property of interest, related to the insulating properties of an insulating layer is the electrical breakdown voltage of the dielectric layer. Voltage can be increased across an insulating layer until a sudden increase in conduction current is observed. If this current is due to a localized fault or a physical pinhole in the dielectric, then it is often referred to as a defect breakdown. If the current is due to a field induced impact ionization mechanism that takes place rather uniformly over the entire cross section of the dielectric, then the field produced by the applied voltage (i.e., voltage over thickness of the dielectric) is considered to be the intrinsic breakdown field of the material.

[0006] It is noted that for thermal oxides (e.g., silicon dioxide), for example, the tunneling currents can be so high as to tend to obscure the intrinsic breakdown characteristic of the material.

[0007] Various methods involving the use of contacts at the oxide surface have been used for measurement of current-voltage behavior. These methods suffer from sensitivity to pinholes in the oxide and from localized breakdown effects associated with the contacting electrodes.

[0008] A typical approach has been to deposit a metal layer on top of the oxide and to then apply a test voltage to the oxide. This not only affects the measurement, but also will typically spoil the test wafer.

[0009] The current-voltage behavior of an insulator can be influenced by the composition of the layer as well as the thermodynamic growth conditions of the layer. Other perturbing influences can be electron trapping states in the layer, polarity dependent carrier injection from the silicon and the particular type of deposited electrode that is used to perform the measurements. Quite often, particularly for thin oxides, the ability to make measurements is thwarted by pinholes in the layer that become localized highly conductive paths after a measurement electrode is deposited. The present invention is advantageously insensitive to localized faults and pinholes.

[0010] U.S. Pat. No. 5,498,974 is incorporated herein in its entirety by reference. The patent discloses an apparatus for depositing corona charge on an insulating layer and for measuring the voltage on the surface of the layer.

[0011] A method for measuring a current-voltage characteristic for an insulating layer on a substrate includes depositing increments of corona charge on the layer, measuring the derivative of a voltage resulting from a reduction of the charge with respect to time, and determining from the voltage and voltage derivative the current-voltage characteristic.

[0012] A method for measuring tunneling field for an oxide layer on a semiconductor wafer including (a) depositing an increment of corona charge on the layer, (b) measuring a voltage across the layer, (c) pausing an increment of time, (d) repeating steps b and c until the voltage saturates, and (e) using the saturation voltage to determine the tunneling field.

[0013] A method for measuring tunneling field for an oxide layer on a semiconductor wafer including depositing a predetermined value of excess charge on the layer, measuring a voltage across the layer, and determining tunneling field in accordance with the voltage.

[0014] FIG. 1 is a schematic diagram of an apparatus for practicing the invention.

[0015] FIG. 2 is an exemplary graph of current versus voltage for a thin oxide under test.

[0016] FIG. 3 is an exemplary graph of current versus voltage for a thick oxide under test.

[0017] FIG. 4 is an exemplary graph of electric field versus corona charge for a tunneling field test on an oxide.

[0018] Referring to FIG. 1, an apparatus 10 for measuring the current-voltage characteristic for an oxide 12 on a substrate 14 of a semiconductor wafer 15 may include a wafer chuck 16, corona gun 18, a Kelvin probe 20, a Kelvin control 22, a current to voltage converter 24, a current integrator 26, a position actuator 28, a high voltage supply 30, and a controller 32.

[0019] The wafer chuck 16 holds the wafer 15 during the measurement process and provides a grounding contact to the substrate 14 of the wafer 14. The grounding contact can be obtained, for example, from a high pressure contact using a sharp tungsten carbide needle.

[0020] The high voltage supply 30 supplies high voltage (e.g., 6-12 Kv to the corona gun 18 to produce positive or negative corona charges depending on the polarity of the supply 30. In the preferred embodiment, the gun 18 includes one or more needles 19 connected to the high voltage supply 30.

[0021] The Kelvin probe 20 is an electrode attached to a vibrator 21. Movement of the probe 20 above a charged surface results in an AC voltage representative of the potential of the charged surface. The Kelvin control 22 converts the AC voltage to a signal corresponding to the voltage of the surface.

[0022] Current flowing through the wafer 15 from the corona gun 18 is converted to a voltage by the current to voltage converter 24. This voltage (current) is integrated by the charge integrator 26 to provide a measure of the charge deposited by the corona gun 18 on the oxide 12. The circuits thus provide a coulombmeter.

[0023] A position actuator 28 may be provided to move the corona gun 18 and the Kelvin probe 20 over the wafer 15.

[0024] The controller 32 controls the operation of the apparatus 10. The controller 32 controls the position actuator 28 and the high voltage supply 30 in response to the Kelvin controller 22, the current to voltage converter 24 and the current integrator 26. Based on the method set forth below, the controller 32 can provide a measurement 34 of the I-V behavior. The controller 32 may be, for example, a dedicated microprocessor-based controller or a general purpose computer.

[0025] For ease of understanding the description below we will assume the measurement of the I-V characteristic of a thermal oxide on a P minus silicon substrate with a negative polarity corona. However, it should be understood that the method of the invention is applicable to a variety of insulating or dielectric layers grown and/or deposited on substrates of semiconductor materials or metals. The charge used can be positive or negative, as appropriate.

[0026] The oxide current I_OXmay be expressed as the product of the oxide capacitance per unit area C_OXand the derivative with respect to time of the voltage across the oxide (dV_OX/dt).

[0027] C_OXcan be calculated from E_O•E_OX/T_OX. Where E_Ois the permitivity of free space, 8.86E-14 farads/cm, E_OXis the relative dielectric constant of the oxide (3.9 for thermal oxide), T_OXis the thickness of the oxide in centimeters.

[0028] The derivative of V_OXcan be approximated by the change in V_OX, ΔV_OXduring a time Δt in seconds.

[0029] An increment of charge as determined by the current integrator 26 is deposited on the oxide surface by the corona gun 18 and the voltage V_OXmeasured by the Kelvin probe 20. After a delay, Δt, V_OXis measured again. This then gives a value for dV_OX/dt which is used to determine I_OX. This change in voltage results from the reduction of charge over the interval Δt. The time Δt used varies with the oxide thickness and is picked to provide the desired sensitivity for measuring the oxide current. The increments of charge can also be summed to provide the cumulative deposited charge Q_OX.

[0030] In general, it may also be necessary to remove any undesired charge from the surface of the dielectric layer before starting the measurements. This may be accomplished by measuring the voltage with the Kelvin probe 20 and applying charge with the corona gun 18 until the oxide voltage is at a typically low value (e.g., −2 volts for a 1,000 Å oxide) corresponding to a low field and a slight degree of silicon accumulation. As a first approximation, the voltage reading of the Kelvin probe, V_KPwill be equal to V_OX. This approximation will hold very well for thick oxides (e.g., greater than 2,000 Å), where the work function difference, V_WF, between the Kelvin probe and the silicon bulk can sometimes be ignored and where the silicon surface potential, V_Si, (which is in series with V_KP) can also sometimes be ignored.

[0031] In order to estimate the actual value of the oxide voltage, V_OX, V_KPmust be corrected for the fact that V_KP=V_OX+V_WF+V_Si. Therefore, V_WFand V_Siare first estimated and subtracted from V_KP. V_WFcan be estimated by first substituting a material, with a predetermined, known work function, such as gold or graphite, in place of the wafer and then measuring V_KP.

[0032] V_Sican be estimated from a surface photovoltage, measurement, SPV, using a very high light intensity source such as xenon flash. Devices for making SPV measurements are well-known in the art. For a given value of SPV and a reasonably estimated value of excess optically induced carrier generation, delta n, there will be an approximate corresponding value of V_Sithat can be calculated from a theoretical model, such as that of E. O. Johnson, Phys. Rev., Vol. 111, No. 1. The first order effect in the Johnson model is that the magnitude of SPV tends to approach V_Siand delta n becomes comparable to and larger than the silicon doping concentration. It is noted that delta n can also be estimated from Johnson, based on a SPV measurement in strong accumulation and in. strong inversion.

[0033] For more accuracy and/or thinner oxides, the above corrections can be employed.

[0034] In the preferred embodiment, the oxide layer 12 starts at zero volts, either inherently or by the application of the appropriate polarity and quantity of corona charge by the corona gun 18.

[0035] Then a negative increment of corona charge ΔQ_cis deposited onto the oxide surface by the corona gun 18. This results in a first oxide voltage V_OX1being measured by the Kelvin probe 20.

[0036] After a pause, Δt, the Kelvin probe 20 measures a second oxide voltage V_OX2. The difference between the voltages is used with Δt to determine I_OX.

[0037] The incrementing of the corona charge ΔQ_cand the calculation of the resulting I_OXcontinues until an I-V characteristic of interest has been generated.

[0038] Referring to FIG. 2, a graph of I-V results from the above method for a thermal oxide layer with a thickness of 143 ∈ is shown. The Δt used was 15 seconds. The charge increment was about 3.3E-7 coulombs/cm².

[0039] Referring to FIG. 3, a graph of I-V results from the above method for a thermal oxide layer with a thickness of 1,000 Å is shown. The Δt used was 60 seconds. The charge increment was about 3.3E-7 coulombs/cm².

[0040] The V-I graphs in FIGS. 2 and 3 can be used to identify the oxide voltage or electric field at a given conduction current. Conversely, FIGS. 2 and 3 can also be used to identify the conduction current corresponding to a given value of oxide voltage or field strength. In addition, the likely existence of a particular conduction mechanism of interest, can be postulated by curve-fitting various oxide conduction models to the I-V behavior in FIGS. 2 and 3. For example, conduction due to tunneling. behavior would tend to exhibit an I-V characteristic I=V_OX²* exp(−b/V_OX), where b is a constant. This tunneling behavior occurs, for example, in thermal oxides. Conduction due to field enhanced thermal excitation of carriers from bulk oxide traps (known as Frenkel-Poole Emission) would tend to exhibit and I-V characteristic I=V*exp(f(T,V)), where f(T,V) is a function of temperature and the voltage across the dielectric, V. Frenkel-Poole Emission is observed, for example, for silicon nitride layers.

[0041] The accuracy of the I-V characteristic can be further improved by correcting for V_WFand V_Si(in case of substrates other than silicon, other surface potentials can be evaluated).

[0042] The work function of a material is defined as the energy required to remove an electron from the Fermi level, physically extract it from the material and then move it an infinite distance away from the material. V_WFcan-be expressed as the work function difference between the work function of the Kelvin probe electrode, W_KP, and the work function of the silicon bulk, W_Si, of the wafer under test. While W_Siis known to be about 4.8 eV, W_KPis usually unknown and may drift due to dipole effects from adsorbed air molecules.

[0043] W_KPcan be determined by making a calibrating Kelvin probe measurement, V_KP2of a material having a predetermined effective work function, W_REF, in place of the wafer 12. For example, highly oriented pyrolytic graphite may be used as a reference material. This graphite has the advantage that a freshly cleaved surface can be obtained by applying and removing a piece of adhesive tape from the surface. This fresh surface allows for very repeatable measurements of W_KP, where W_KP=VKP2−W_REF. It follows then that V_WF=V_KP2+W_REF−W_Si.

[0044] The resulting value for V_WFis then subtracted from V_KPby the controller 32 to provide a corrected value for V_OX.

[0045] The SPV tool 34 is used to make an SPV measurement using a very high intensity light source such as a xenon flash tube. The resulting value of SPV is then used to estimate V_Siwhich is then subtracted from V_KPby the controller 32 to provide a corrected value for V_OX.

[0046] The light intensity must be sufficient for creating a concentration of excess light induced carriers that is comparable to or greater than the doping concentration of the wafer (e.g., 1E15 carriers/cm³). The excess carriers (electrons and holes) will separate in the silicon surface field, due to V_Si, and then set up an opposing field that will tend to reduce V_Sitoward zero. Therefore, the magnitude of the SPV (actually, the change in V_Si) will tend to be a significant fraction of V_Si. For silicon, in the depletion regime, the SPV can be as much as 80% of V_Si. For the accumulation regime, the SPV will tend to be about 30% of V_Si.

[0047] The need to correct for V_WFand V_Sibecomes greater for thinner oxides. An uncorrected error in V_WFcould be as high as one volt. For a 2,000 Å oxide this could correspond to a tunneling field error of 0.05 Mv/cm which would be a 0.6% error for a nominal tunneling field of about 8 Mv/cm. In contrast, for a 50 Å oxide, the error would go up to 25%. By correcting for V_WF, this latter error can be reduced to about 5%.

[0048] An uncorrected error in V_Sican also be significant. Assuming a true oxide conduction current of about 46E-9 amps/cm², for a 2,000 Å thick oxide, the theoretical drop in oxide voltage per second would be about 265 mv/sec. Without using V_WFor V_Sicorrection results, for example, in a measured drop in oxide surface voltage of 283 mv/sec or a 7% error.

[0049] In the case of a 50 Å thick oxide, the theoretical drop in oxide voltage per second would be about 7 mv/sec. Without using V_WFor V_Sicorrection results, for example, in a measured drop in oxide surface voltage of 30 mv/sec, or 400% error. Correcting the measurement for V_Siand V_WFrealistically reduces the error by a factor of ten.

[0050] The theoretical values are based on the Johnson Model, assuming that the excess light induced carrier concentration was a reasonable value equal to ten times the doping concentration of the wafer, which was assumed to be 1E15 toms/cm³.

[0051] The measured data can also be advantageously used to determine the tunneling field for an oxide. Referring to FIG. 4, electrical field strength (V_OX/T_OX), as a function of deposited corona charge, Q_OXis shown. This is a convenient way of finding the saturating, terminal value of the oxide field corresponding to increasing deposited corona charge density. In FIG. 4, the electric field strength in a 28.8 Å thermal oxide is tending to reach a terminal value as the oxide conduction current (coulombs/sec-cm²), due to tunneling, approaches the rate of corona deposition (coulombs/sec-cm²). This terminal value (here, about 8 Mv/cm) is referred to as the tunneling field.

[0052] Tunneling field may also be measured more directly with the apparatus 10. The corona gun 18 is used to deposit a charge on the oxide 12 that would be sufficient to establish a field strength greater than the expected tunneling field (e.g., 7-10 Mv/cm). This predetermined value of excess charge only creates a field corresponding to the tunneling field due to the resulting tunneling current. The resulting V_OXis measured with the Kelvin probe 20 and the tunneling field determined from V_OX/T_OX. The repeatability and accuracy of the measurement can be improved by controlling the corona deposition rate, total corona charge and elapsed time before making the V_OXmeasurement.

[0053] The method of the invention allows measurement of I-V characteristics and tunneling field without spurious results due to localized defects such as pinholes. Because no conductor is applied to the surface of the dielectric, the localized defects stay localized as only the local corona charge is available to pass through the defect. It is also noted, that for tunneling current measurements, this invention offers the added advantage of not having otherwise, undesirable, enhanced tunneling around the abrupt edge of a MOS electrode. This edge effect problem was discussed in T. B. Hook and T. P. Ma, J. Applied Physics. 59 (11), Jun. 1, 1986. For the Corona-Oxide-Semiconductor electrode used in this invention, the charge density around the effective edge of the electrode will tend to be tapered, as opposed to a MOS electrode.

[0054] It should be evident that this disclosure is by way of example and that various changes may be made by adding, modifying or eliminating details without departing from the fair scope of the teaching contained in this disclosure. The invention is therefore not limited to particular details of this disclosure except to the extent that the following claims are necessarily so limited.