Electrical breakdown and ESD phenomena for devices with nanometer-to-micron gaps

Transcription

1 Electrical breakdown and ESD phenomena for devices with nanometer-to-micron gaps Al Wallash *a and Larry Levit b a Maxtor Corporation, 500 McCarthy Blvd, Milpitas, CA, USA b Ion Systems, 1005 Parker Street, Berkeley, CA, USA ABSTRACT The current vs. voltage and electrical breakdown behavior for devices with micron and sub-micron gaps between conductors is studied. The limitations of the well-known but often-misinterpreted Paschen curve are discussed. The little-known modified Paschen curve, that includes field emission effects so important in understanding breakdown behavior for devices with sub-micron gaps, is described. Current vs. voltage measurements across metal-air-metal, metal-insulator-metal and metal-insulator-air-insulator-metal gaps with gaps ranging from 4 nm to 4 µm are reported. The breakdown voltage for an air gap of 0.9 µm was found to be 150 V, far below the Paschen curve minimum breakdown limit, and field emission behavior was confirmed via the Fowler-Nordheim plot. Metal-insulator-metal gaps with a diamond-like carbon thin-film with a thickness of 4 nm had a breakdown voltage of only 1V. SEM and AFM analysis show that the breakdown damage is crater-like and through the carbon layer. Other characterization of the damage caused by breakdown is presented. Tribocharging, electrostatic induction, and other ESD-related phenomena, are discussed for several devices with sub-micron gaps. It is concluded that devices with sub-micron gaps can face a serious challenge due to electrical breakdown during manufacturing, handling and operation. These devices include photolithographic reticles, magnetic recording heads, MEMS and field emission displays. Keywords: Electrical breakdown, Paschen curve, spark gap, ESD, electrostatic discharge, GMR, MEMS, reticle INTRODUCTION The understanding and control of electrical breakdown across the gap between two conductors is very interesting and critically important for the microelectronics industry. With each new generation of device, the gap spacing between conductors is often reduced by scaling down of device feature sizes. A prime example of this is the hard disk drive magnetic recording industry, which today routinely utilizes devices with air gaps as small as 10 nm and insulators as thin as 6 Angstroms! The integrated circuit, MEMS and flat panel display industries, to name a few, also have devices with micron and sub-micron size gaps between conductors. Downscaling of devices can result in a reduced electrical breakdown voltage, and, if ignored, this lower breakdown voltage can cause yield issues or even problems during device operation. The well-known Paschen curve describes the gaseous breakdown voltage as a function of the reduced variable of the pressure-gap spacing product (ρ=p*d). [1] Breakdown occurs when gas ions accelerate across the gap resulting in a Townsend avalanche. [2] Figure 1 shows the Paschen curve for the most relevant case of air at a pressure of 1 atmosphere and a geometry of two 30 cm diameter brass spheres. [3] For gaps greater than about 10 µm, breakdown has been well studied and occurs when the electric field exceeds about 3 V/µm (30 kv/cm). Note the minimum breakdown voltage of about 360 V at a gap spacing of 5 µm, and an increase in the breakdown voltage for even smaller gaps. * al_wallash@maxtor.com; phone ; fax

2 After seeing the Paschen curve, one may be left with the strong impression that there is an absolute minimum breakdown voltage of about 360V and that breakdown across small (< 5 µm) air gaps at atmospheric pressure is difficult and would only occur at a very high voltage. While the left, rising part of the Paschen curve can be observed at low pressure and large gap spacing, it is not observed for air gaps at atmospheric pressure. [3] One proof of this is the 80V breakdown voltage reported for a 0.12 µm air gap in ref. 4. Nevertheless, there is a widespread misconception that there exists a minimum breakdown voltage for air gaps and that if the spacing of a device is scaled-down below ~ 5 µm then breakdown will not occur. It is important to understand that electrical current can result from mechanisms other than the Townsend avalanche within an ionized gas, and that these other mechanisms are not considered in the Breakdown Voltage (V) Breakdown Voltage (V) Air at 1 atmosphere ~ 360 V minimum ~ 3 V/µm Gap spacing (µm) Fig. 1. Paschen curve: Gaseous breakdown voltage vs. gap spacing between two conductors. Gas is air at a pressure of 1 atmosphere. Note that this curve can be extremely misleading when predicating breakdown at gaps less than 5 µm. theory behind the Paschen curve. Thus, while it is true to conclude from the Paschen curve that gaseous breakdown at 1 atmosphere pressure does not occur below ~ 360V, it is false to conclude that breakdown of any kind does not occur below 360V. The other mechanisms for prebreakdown current production at a small gap spacing are the closely related phenomena of electron field emission and tunneling. For electric fields across gaps less than 5 µm, the electric field becomes quite large and electrons can tunnel through the deformed surface potential barrier. This is called field emission. [5] Electrons from field emission are one reason why breakdown and sparks occur in a vacuum, which of course is not possible if one only considers the Townsend Avalanche mechanisms for gas ionization used to generate the Paschen curve. In addition, the Paschen curve was developed for the geometry of two large spheres and is effectively parallel plate geometry for a gap spacing of several microns. For modern microelectronic devices, such a geometry does not apply and the inhomogeniety of fields from tiny devices results in areas of higher field-line concentration, particularly at sharp edges. In addition, the work of Paschen involved the reduced variable of the pressure-gap size product. For very low gaps it was easier to make measurements at lower pressure rather than the very small gap sizes relevant to today s nanodevices. Figure 2 shows the little known modified Paschen curve for air at 1 atmosphere. [6] The modified Paschen curve retains the right part of the Paschen curve that is still valid at larger gap spacings, but discards the minimum and upturn in the pure Paschen curve at small gap spacings and replaces these with a plateau and steep decline to zero. Moving from about 10 µm smaller gaps, the modified Paschen curve shows a plateau where the pure Paschen curve would have a minimum, at a gap of around 5 µm. This plateau is interpreted as the transition region between the gaseous, Townsend Avalanche and Knee Air at 1 atmosphere Plateau ~ 75 V/µm Gap spacing (µm) ~ 3 V/µm Fig. 2. Modified Paschen curve: breakdown voltage vs. gap spacing. Note plateau, knee and steep decline in breakdown voltage for a gap spacing less than 5 µm.

3 field emission induced breakdown. Below approximately 2.5 µm, there is a knee in the curve where and field emission takes over completely and the breakdown voltage drops precipitously to zero. The exact location of the transition to a steep drop will be related to the details of the geometry and metal electrode properties. The voltage for Townsend avalanche from sharp objects is expected to be raised to higher levels, whereas the voltage for field emission would be lowered. Both of these effects are related to inhomogeniety of fields from sharp objects as compared to the fields between two large (30 cm diameter) spheres. In addition, electrode material and surface roughness will also affect the breakdown voltage for both field emission. The field emission current, I, is described by the Fowler Nordheim equation I = ae 2 e b Φ E 3/ 2, (1) where E is the electric field, Φ is the work function, and a and b` are constants. [7] If current is due to field emission, a plot of 1/E vs. ln(i/e 2 ) (or 1/V vs. ln(i/v 2 ) should yield a straight line with a negative slope that is proportional to the work function of metal. This plot is known as the Fowler-Nordheim (F-N) plot and can be used to determine if the current flow is due to field emitted electrons. To complete the picture, for ultra-thin gaps smaller than ~2 nm, there is a finite probability that electrons can tunnel across an insulating barrier and result in a significant tunneling current. To summarize, there are three distinct mechanisms that produce current flow due to an applied electric field prior to breakdown. They are: 1. Townsend Avalanche of gas ions for gaps > 5 mm, 2. Field emission of electrons from metals for gaps larger than 2 nm but less than 5000 nm, and 3. Tunneling across a barrier thinner than about 2 nm Table 1 summarizes the current flow mechanisms and applicable gap length. It is clear that one needs to tailor the measurement technique and data analysis to the appropriate current generating mechanism. The goal of this work is to show that electrical breakdown across micron and sub-micron gaps does occur at voltages far below the pure Paschen curve minimum and that the modified Paschen curve should be used instead for submicron gaps. In this work, emphasis is placed on field emission effects for gap spacing ranging from 4 nm to 4 µm. The current vs. voltage (I-V) curve, breakdown voltage and physical damage are measured. I-V curve data is analyzed in terms of the Fowler-Nordheim equation to confirm field emission. In this paper we will also discuss how electrostatics and ESD can cause electrical breakdown across the sub-micron gaps in reticles, MEMS, magnetic recording and flat panel devices. Current Mechanism Townsend Avalanche of gaseous ions Field Emission of electrons Tunneling of electrons Gap Distance > 5 µm 5 nm to 5 µm < 2 nm Table 1. List of three different mechanisms and appropriate length scale that generate prebreakdown current across the gap between two conductors.

4 EXPERIMENTAL The setup for measuring the current vs. voltage (I-V) curve is shown in Fig. 3. A Keithley 2400 Source- Measure unit was used to source voltage and measure current of the device with a gap. This enabled measurement of the I-V characteristics across the gap with a sensitivity of 30 pa. This sensitivity is required to measure the expected low currents (~ na) associated with field emission. The compliance current was set at 100 ma. A TEKTRONIX CT-6 current probe and LeCroy 9362 digital oscilloscope was also used to measure the current transient at the point of breakdown. The single-shot system bandwidth for measuring the breakdown transient was 750 MHz. A probe station was used with micropositioners and probes to make contact with the device under test. - + Keithley 2400 Source/Measure GPIB PC / LabView Probes Device with gap U TEK CT-6 current probe to LeCroy 9362 Fig. 3. Schematic representation of the experimental setup to measure the current-voltage (I-V) behavior of a device with a gap. The test procedure involved making electrical contact to the terminals at each side of the gap and then ramping the voltage from zero volts in small steps while measuring the current. A custom LabView program was used to automate data acquisition. If breakdown was not observed upon reaching the maximum voltage output of the Keithley 2400 of 210V, testing was continued from 210V to 1000V using the DC voltage output from a Keithley 6517A electrometer. Three types of devices were selected to represent a wide range of gap spacing and gap material. Characterizing them by the type of material in the gap between the two metal surfaces, the three types of devices were: 4 µm gap 500X 250 µm 1. Metal-Air-Metal 2. Metal-Insulator-Metal 3. Metal-Insulator-Air-Insulator-Metal Figure 4 shows two photos of the Metal-Air-Metal device. This device consisted of chrome deposited on an insulating glass substrate commonly used for reticles. [8] These devices had gaps in the chrome with 0.9, 1.0, 1.5, 2.0, 3.0 and 4.0 µm spacing. All chrome was 100 nm thick. Two devices at each gap distance were tested. 4 µm gap 2 µm Glass Chrome Figure 5 shows a schematic representation of the Metal- Insulator-Metal case. The insulating layer was a 4 nm thick diamond-like carbon (DLC) film on a metallic substrate that formed one of the metal interfaces. A probe tip with a radius of 0.35 µm formed the other metal interface. The DLC film and substrate comprised a thin-film disk commonly used in today s hard disk drives. 1000X Fig. 4. Optical photographs of metal-air-metal device on reticle with 4 µm gap at two magnifications: 500X (top), 1000X (bottom).

5 + + Tungsten probe (0.35 µm tip radius) 4 nm thick carbon Air gap GMR sensor Conductive slider 4 nm thick carbon Metal substrate (CoPtCr/Aluminum) - Fig. 5. Schematic diagram of the probe on carbon film that forms a metal-insulator-metal interface with a 4 nm gap. Metal substrate (CoPtCr/Aluminum) - Fig. 6. Schematic diagram of the GMR sensor resting on the thin film disk. This arrangement formed a metal-insulator-airinsulator-metal interface. Gap consisted of 8 nm of DLC insulator with variable air gap between carbon insulators. The Metal-Insulator-Air-Insulator-Metal system is shown in Fig. 6 and consisted of a magnetic recording slider resting on the thin-film magnetic recording disk. The slider is again coated with a 4 nm thick DLC film. Even though the surfaces of the slider and disk are lapped smooth, when the slider was placed on the disk it is expected that some small air gap, on the order of 1 to 10 nm, was present. 1. I-V Curve Analysis Metal-Air-Metal Device RESULTS Figure 7 shows the I-V curve from 0 to 200V across a device with an air gap of 0.9 µm. The current spike at 151V coincided with an abrupt current transient and melting damage to the device. The other 0.9 µm device broke down at 135 V and the average breakdown voltage for the 0.9 µm device was taken to be 141 V. The breakdown voltage vs. gap spacing for all of the metalinsulator-metal devices is shown in Fig. 8. It is clear that the devices with a 0.9 and 1.0 µm gap experienced breakdown at 141 and 150 V, far below the pure Paschen curve limit of 360V. There is a clear knee in the breakdown curve at 2.0 µm gap spacing followed by a plateau from spacings ranging from 2 to 4 µm. The knee and plateau are consistent with the modified Paschen curve shown in Fig. 2. Current (na) Gap spacing = 0.9 µm Voltage (V) Fig. 7. Current vs. voltage (I-V) curve for metal-air-metal-gap device on glass with 0.9 µm gap. Current spike at 151 V coincided with breakdown current transient and melting of device.

6 Breakdown Voltage (V) Gap spacing (µm) Fig. 8. Breakdown voltage vs. gap spacing for metal-air-metal devices. A conservative estimate of the electric field in the 0.9 µm gap at breakdown is 156 V/ µm (E=V/d). This is conservative since it neglects enhancement of the E-field due to the radius of curvature in the gap and the 200 nm thickness of the film. Nonetheless, this is still much higher than the 3 V/ µm slope in the pure Paschen curve shown in Fig. 1 at larger gap spacings. Since most of the current below 100V is leakage in the instrument or through the glass substrate, the I-V curve data was fit to a polynomial and subtracted from the total current for voltages greater than 100V. Figure 9(a) shows the remainder current after subtraction. For voltages greater than 120V, there is a clear increase in current that grows larger as breakdown is approached.. Using a value for the electric field, E estimated by using E=V/d, where V is the voltage and d is the gap spacing, the Fowler-Nordheim (F-N) plot from 120 to 150V was generated and is shown in Fig. 9(b). Note the linear relationship between 1/E and ln(i/e 2 ) and the negative slope. This linear behavior and negative slope indicate that field emission is the main cause of the current close to breakdown µm gap µm gap Current (na) ln(i/e^2) y = x R 2 = Voltage (V) /E (µm/v x 10-3)) Fig. 9. (a) Left: Corrected current vs. voltage (I-V) curve for 0.9 µm gap device near breakdown. Background current was subtracted out to reveal current close to breakdown. (b) Right: Fowler-Nordheim plot for 0.9 µm gap device near breakdown.

7 Current (A) 1.E+00 1.E-02 1.E-04 Metal-Insulator-Metal 4 nm insulator thickness 1.E Voltage (V) Fig. 10. Current vs. voltage curve for metal probe on diamondlike carbon (DLC) insulator. Breakdown across the 4 nm thick DLC layer occurred at 0.91 V. Metal-Insulator-Metal Interface A typical I-V curve for the case of metal-insulator-metal gap is shown in Fig. 10. The current starts at ~ 1 µa, increases to about 1 ma at 0.95V and then goes to the instrument compliance Current (ma) V breakdown = 1.1 V Time (µs) Fig.11. Current vs. time measured at breakdown at 1.1V. Measured using the CT-6 current probe in series with the K2400 Source/Measure unit. current of 100 ma indicating that a conductive short had developed through the 4 nm insulating DLC layer. Coincident with the abrupt increase in current, the current transient shown in Fig. 11 was measured. Therefore, the point where the current increased to compliance current was taken to indicate breakdown of the DLC layer. The average breakdown voltage across the 4 nm DLC carbon film was 1.1 V. Figure 12 shows the Fowler-Nordheim plot for the metal-insulator-metal device data shown in Fig. 10. Breakdown occurred at 0.91V, or 1/V = 1.1. Note the change from a positive to a negative slope, indicating the development of significant field emission induced prebreakdown current as the voltage was increased from 0.3V to breakdown Breakdown ln(i/v 2 ) /V (1/Volt) Fig. 12. Fowler-Nordheim plot for metal probe on 4 nm-thick diamond-like carbon (DLC) insulator. Breakdown occurred at 0.91V, or 1/V = 1.1. Note change in the slope of curve, indicating the development of significant field emission induced current flow as voltage was increased from 0.3V to breakdown.

8 Metal-Insulator-Air-Insulator Interface A typical I-V curve for the metal-insulator-air-insulator-metal gap case is shown in Fig. 13. The current first starts at ~ 0.1 na, increases to about 0.1 ma at the gap breakdown at 3.8V. The F-N plot for this device is shown in Fig. 14, were we again see the appearance of field emission current (negative slope) for voltages greater than 0.5V. For all three types of devices, field emission current was observed at voltages close to and below the breakdown voltage. Current (A) 1.E-02 1.E-04 1.E-06 1.E-08 Breakdown ln(i/v 2 ) Breakdown 1.E Voltage (V) /V (1/Volt) Fig. 13. I-V curve for the metal-insulator-air-insulator-metal gap case. Breakdown across the ~ 8 nm thick DLC thin-films occurred at 3.8V when current when to compliance. Fig. 14. Fowler-Nordheim plot for metal-insulator-air-insulatormetal situation. The change of slope to a negative value signifies the onset of field emission prebreakdown current. 2. Damage and Failure Analysis after Breakdown It is interesting and important to study and understand the damage caused by breakdown in these devices. Each device is unique and can result in different types of damage. Figure 15 shows a photo at 500X of the metal-air-metal device with a 1.5 µm gap after breakdown. Comparing this device after breakdown with a similar device before breakdown shown in Fig. 4 (top), we can clearly see new fractal-shaped damage in the chrome layer at the lower left that is reminiscent of lightning bolt behavior in air. The chrome finger is also completely melted or perhaps even vaporized. Two examples of the breakdown damage to the DLC carbon film for the metal-insulator-metal case are shown in the atomic force microscope (AFM) scans shown in Fig. 16. The damage is crater-like a diameter about 1 µm and has sharp peaks and valleys. AFM profile scans across these craters show the peaks and valleys can be +/- 25 nm, which far exceeds the film thickness of 4 nm. Fig. 15. Optical photo at 500X magnification of the metal-air-metal device with 1.5 µm gap after breakdown. Compare with Fig. 4 (top) for a comparison to a similar undamaged device before breakdown.

9 Fig. 16. Atomic force microscope (AFM) scans of two different craters caused by breakdown through the DLC carbon. ESD IMPLICATIONS FOR DEVICES WITH SUB-MICRON GAPS Once the breakdown voltage is known for a device with a small gap, it would seem a simple exercise to limit the voltage across the gap during operation to a lower value and to be able to forget about device breakdown. Unfortunately, this would completely ignore electrostatics and ESD during handling, assembly and operation of a device. Once it is understood that electrostatic phenomena can easily induce hundreds and even thousands of volts on a floating metal object, [9] it should be apparent that electrostatic considerations should not be ignored with a device that has a low breakdown voltage. The most fundamental and surprisingly powerful electrostatic phenomenon is tribocharging, which is the transfer of electrons from object to another upon contact. This net charge on each object will produce an electric field that will induce a voltage on other nearby ungrounded metal objects. If the difference in the induced voltage on two conductors separated by a sub-micron gap exceeds the breakdown voltage, breakdown and its associated damage will occur. This field-induced breakdown has already been demonstrated to occur for a magnetic recording head with a 200 nm gap. [10] A related scenario where field-induced breakdown is possible is when the device with a sub-micron gap is situated on an insulating substrate. Two such cases are reticles used in photolithography [8] and flat panel displays. In these devices, any contact (i.e. rubbing) of the insulating substrate can result in a highly charged substrate. This charge will induce a different voltage on all of the metal structures and devices that can easily result a voltage across the gap that exceeds electrical breakdown. MEMS devices face their own unique and only recently recognized ESD-related challenges and it has already been reported that they are very susceptible to ESD events, with a failure voltage ~ 150 V. [11]

10 SUMMARY AND CONCLUSIONS A review of the literature and experimental data both show that electrical breakdown across sub-micron air gaps at atmospheric pressure does not follow Paschen s law. Instead, there is a plateau in the breakdown voltage for 2 to 3 µm gaps, followed by a precipitous drop as the gap spacing is reduced even further. The current generating mechanism before breakdown is field emission, as verified by a linear Fowler-Nordheim plot with a negative slope. The field strength at breakdown in the sub-micron regime is on the order of 150 V/µm. Breakdown damage in these devices ranged from outright melting of the device to micron-sized craters in the 4 nm thick carbon films. It is concluded that breakdown in air at atmospheric pressure can occur well below the Paschen curve minimum of 360V and should be considered in the processing, handling and operation of devices with micron and sub-micron gaps. REFERENCES 1. F. Paschen, 1889 Wied. Ann 37, pp J.S. Townsend, J. Franklin Inst., 200, 563, ELECTROPHOTOGRAPHY by R.M. Schaffert, John Wiley and Sons, 1975, p.514, or the 1965, Focal Press Limited version, pp Capacitive Coupling Effects in Spark Gap Devices, A. Wallash and T. Hughbanks, Proc EOS/ESD Symposium, EOS-16, pp , FIELD EMISSION AND FIELD IONIZATION, F. Gomer, Harvard University Press, 1961, p. 6. ELECTROPHOTOGRAPHY by R.M. Schaffert, John Wiley and Sons, 1975, p Fowler and L. Nordheim, Proc. Roy. Soc. (London), A119, 173, 1928; A124, 699, Englisch, A., van Hesselt, K., Tissier, M., Wang, K.C., "CANARY: a high-sensitive ESD test reticle design to evaluate potential risks in wafer fabs", Proceedings of SPIE, 19th Annual Symposium on Photomask Technology, BACUS, Volume II, p ELECTROSTATIC DISCHARGE CONTROL, Owen. J. McAteer, 1990, Chapter A. Wallash and M. Honda, Field induced breakdown ESD Damage of Magnetoresistive Recording Heads EOS/ESD Symp. Proc. EOS-19, pp , J.A. Walraven, J.M. Soden, D. M. Tanner, P. Tangyunyong, E. I. Cole, Jr., R.R. Anderson and L.W. Irwin, Electrostatic Discharge/Electrical Overstress Susceptibility in MEMS: A New Failure Mode, Proc. SPIE 2000, V. 4180, pp