1. Field of the Invention
The present invention is directed to micro-electromechanical systems. More particularly, the present invention relates to a method of depositing a low charging dielectric for capacitive micro-electromechanical systems.
2. Description of the Related Art
Dielectrics may be used in many different applications—as an insulator or barrier layer in semiconductor devices, as an active element of a micro-electromechanical systems (MEMS) device, etc. When a dielectric traps a charge therein, it can diminish the dielectrics desired functionality.
MEMS have been developed for use in a number of electronic devices and components, such as phase shifters, tunable filters, and resonators. MEMS switches operate through the electrostatic actuation of a beam to achieve physical contact with an electrode. An exemplary capacitive MEMS switch is shown in
Loss of bandwidth of switch 100 is defined by the RF coupling through the dielectric layer 4. The down capacitance of the switch 100 is determined based on the thickness and dielectric constant of the dielectric layer 4. The choice of dielectric, however, is constrained by many of the switch properties such as the actuation voltage, as this sets the field across the dielectric. The field strength must remain below the breakdown voltage of the dielectric. Silicon nitride is a compound that has been found to have a relatively high dielectric constant (κ˜7) and a relatively high dielectric strength (˜6000 kV/cm). Based on its combination of properties, silicon nitride is used extensively in MEMS devices.
It is known to use plasma enhanced chemical vapor deposition (PECVD) to deposit these nitrides as the dielectric layer in MEMS devices because of the ability to deposit these films at relatively low temperatures (˜200-300° C.), which is compatible with the materials and substrates used to fabricate MEMS devices.
Capacitive MEMS switches utilizing silicon nitride as a dielectric layer have shown failure mechanisms associated with the charging of the dielectric. This failure mechanism manifests itself as an increased “open state” capacitance resulting from the accumulation of charge trapped within the dielectric film. This trapped charge can exert enough force on the beam to decrease the air-gap between the beam and the dielectric or to keep the keep the beam in contact with the dielectric in the “open” state (i.e., after the charge has been removed from electrode 3).
Silicon nitride films are capable of storing charge for extended periods of time. Charge can be trapped in both shallow surface states and deep bulk traps. The density of surface states can be impacted by material properties, deposition conditions, and environmental conditions such as subsequent processing steps, humidity, oxidation, and surface contamination. The bulk traps can be impacted by both the deposition conditions and the material properties of the dielectric. As shown in
The behavior of the dielectric in the MEMS switch is difficult to determine because mechanical, electrical, material, and environmental complications can be associated with testing a complete MEMS switch.
Thus, there is a continued need for new and improved dielectric layers for use in MEMS devices and methods for testing and fabricating the same.
According to a preferred embodiment of the present invention, a dielectric film is provided for use in MEMS devices. The dielectric is compatible with MEMS fabrication techniques, decreases the rate of charge accumulation in the bulk dielectric by greater than 95% and increases switch lifetime reliability by 40 times relative to standard silicon nitride films.
According to another embodiment of the present invention, a test structure is provided for monitoring the impact of the MEMS switch fabrication process on charge accumulation in the nitride films. The test structure includes an M-I-S (Metal-Insulator-Semiconductor) structure.
According to another embodiment of the present invention, a method of fabricating a MEMS device is provided. The method includes a step of depositing a dielectric film on an electrode of the MEMS device. An amount of trapped charge within the dielectric film is determined during the depositing step. At least one process parameter is adjusted in the deposition step in order to reduce the amount of trapped charge within the dielectric film.
According to another embodiment of the present invention, a method is provided for determining an amount of trapped charge in a dielectric film of a MEMS device. The method includes a step of depositing a dielectric film on a silicon wafer. The dielectric film is deposited under the same conditions as the dielectric film of the MEMS device. A metal layer is deposited on top of the dielectric film. The resulting M-I-S structure is biased with a bias voltage. The flatband voltage and capacitance of the M-I-S structure is measured. The amount of trapped charge is calculated based on the flatband voltage and capacitance measured.
According to another embodiment of the present invention, a method is provided for fabricating a capacitive MEMS switch having a dielectric film. The method includes a step for fabricating a M-I-S structure on a dielectric film; a step for determining an amount of trapped charge in the dielectric film; a step for determining optimum process parameters associated with depositing the dielectric film to minimize the amount of trapped charge in the dielectric film; and a step for fabricating the MEMS switch utilizing the optimum process parameters to deposit the dielectric film.
Two strong correlations were discovered relative to the observed molecular bonding and the charging behavior of nitrides. The desirable low charging behavior of a nitride is believed to correspond to a high number of Si:Si bonding, as well as a large ratio of Si:H bonds compared to N:H bonds. Improved nitrides have Si:H/N:H ratios preferably greater than 1, and more preferably greater than 3; and extinction coefficients (at 248 nm) preferably greater than 0.06, and more preferably greater than 0.1.
Further applications and advantages of various embodiments of the invention are discussed below with reference to the drawing figures.
Trapped charges within a dielectric can be caused by many factors. Detailed capacitance-voltage (C-V) measurements may be conducted on metal-dielectric-semiconductor structures as a means of quantifying the amount of charge trapped in a dielectric film. C-V methods for characterizing dielectrics used in MEMS capacitive switches allow the behavior of the dielectric to be separated from the mechanical, electrical, material, and environmental complications associated with testing a complete MEMS switch. C-V methods allow the bulk and interfacial sheet charging of the dielectric to be separated and establish a rapid and inexpensive technique for surveying dielectrics for various applications.
Dielectric films (i.e., nitrides) currently used in the fabrication of MEMS devices are often deposited by plasma enhanced chemical vapor deposition (PECVD). An exemplary PECVD device that can be used with the present invention is the PlasmaTherm® Model 730 PECVD device, which is manufactured and marketed by PlasmaTherm®. Of course, one skilled in the art will readily understand that different PECVD devices can be used and how to adjust the parameters from device to device to achieve the present invention.
The properties of a deposited film, such as silicon nitride, are controlled through several processing parameters and can be dependent upon the device used to deposit the film. For example, with the Model 730 PECVD device, the following seven process parameters control the deposition of a silicon nitride film: silane (SiH4) flow rate, ammonia (NH3) flow, nitrogen (N2) flow, helium (He) flow, RF power, and deposition chamber pressure and temperature. A method for determining the optimum values for the process parameters associated with depositing a film is described below.
As will be shown in further detail below, it was discovered that the charge trapping characteristics of a nitride can be controlled by controlling the amount of Si:H, Si:Si, and N:H bonds in a nitride. A number of experiments were performed to show the correlation between the Si:H, Si:Si, and N:H bonds in a nitride and charge trapping. Below, the experiments are described and experimental data is reported. One having ordinary skill in the art will readily understand how to achieve the present invention after reviewing the present disclosure.
A metal-insulator-semiconductor (M-I-S) structure including the dielectric was constructed on a silicon wafer with the PECVD device. An exemplary capacitor is shown in
A bias voltage was applied to the capacitor, and flatband measurements can be taken to quantify the amount of charge trapped in the dielectric. The total charge trapped in the dielectric 4 can be extracted from the flatband voltage (VFB) measurements taken from the C-V curves of the M-I-S structure by the formula:
VFB=ØMS−(QF/CFB).
ØMS is the work function of the metal-semiconductor system, VFB and CFB are the measured flatband voltage and the flatband capacitance of the MIS structure, respectively, used to simulate the electric environment of a nitride dielectric incorporated into a MEMS device. Preferably, the time dependence of the total charge trapped in the dielectric films is measured under a −50 Volt bias.
Using an iterative scientific method, the process parameters used for depositing the dielectric film were incrementally varied. A number of M-I-S structures were created, each with a dielectric layer having different properties. Flatband measurements are taken for each M-I-S structure. The impact of each process parameter on the behavior of the dielectric can be determined based on the measurements, and the process parameters was incrementally varied for each iteration until the amount of trapped charge measured in the dielectric film of the M-I-S structure is minimized.
For example, a thin film measurement system was used to determine the following quantities for the nitride films: thickness (d), refractive index (n), extinction coefficient (k), and energy band gap (Eg). An n&k Analyzer was used to obtain these quantities and measures the reflectance spectrum of the nitride and the substrate over optical wavelengths from 190 to 1000 nm. The reflectance spectrum represents the interaction of the light with both the nitride film and the silicon substrate and is dependent, in this case, on the nitride film thickness, refractive index spectra, and the extinction coefficient spectra, as well as the energy band gap of the nitride material. The physical properties of the nitride are extracted from the measured reflectance data using the Forouhi-Bloomer formulation for the dispersion relation of n(λ) and k(λ) and the Fresnel equation to describe the reflectivity of the thin film system. This measurement technique has been used to characterize amorphous and polycrystalline silicon films, carbon overcoats, and Cr—SiOx and SiC thin film resistors.
The reflectance spectrum has been used to determine the thickness of silicon nitride films deposited on silicon substrates.
The n&k Analyzer was used to extract the refractive index and extinction coefficient dispersion curves for each nitride film. The refractive index, n(λ), at 633 nm has traditionally been used as a means of process monitoring to ensure constant composition from lot to lot. The refractive index, however, of PECVD silicon nitrides has been shown to only roughly correlate with the film composition and gives limited information concerning the relative abundance of Si—N, Si—H, N—H, and Si—Si bonding in the film.
The refractive index dispersion curve for each nitride film outlined in
The behavior of the extinction coefficient shown in
The silicon content of the various nitrides correlates very well with the measured band-gap and conductivity of these films. The band gap, Eg, measured by the n&k Analyzer represents the minimum photon energy required to induce a direct electronic transition from the valence to the conduction band. Note that E=hc/λ, and for the case of E<Eg, absorption of light due to direct electronic transitions does not occur.
However, the silicon substrates used for this measurement are not ideal for silicon nitride spectra because of the over lap of bonds (Si—Si, Si—H, and Si—O) existing in the silicon background and bonds existing in the nitride sample. This is particularly evident in measurements on the Si—Si peak at 450 cm−1, which were inconclusive. The Si—N peak proved difficult to measure to get meaningful bond densities because of the large number of bonds that absorb near 850 cm1, including the N—H stretch, Si—O, and a Si—H mode (not shown).
Hysteresis curves were measured between −100 V and +100 V for each nitride film in this study. The structures used for this measurement were metal/insulator/metal fabricated on GaAs wafers with a passivating nitride used to insulate the bottom metal of the MIM from the semiconducting substrate. The nitride film thickness were approximately 2.5 kA for this measurement.
The relative conductivities of each film correlate reasonably well with the band gap energy and UV extinction coefficient measurements described earlier. The films having a higher extinction coefficient (see
The flatband voltage of each dielectric has also been measured under varying bias conditions. In this set of measurements, the C-V curves were traced from 0 Volts to Vmax, where Vmax was incremented from 5 V to 100 V in 5 volt steps.
Detailed capacitance-voltage measurements have been conducted on metal-dielectric-semiconductor (M-I-S) structures as a means of quantifying the charge trapped in stressed nitride films. The impetus behind developing a C-V method of characterizing dielectrics used in MEMS capacitive switches is to separate the behavior of the nitride from the mechanical, electrical, material, and environmental complications associated with testing a complete switch. The C-V measurements offers the opportunity to separate the bulk and interfacial sheet charging of the dielectric as well as establishing a rapid and inexpensive technique for surveying dielectrics for various applications.
In order to quantitatively compare the total trapped charge among the various nitrides, the flatband results must be normalized for constant field strengths across the dielectric. This normalization is accomplished by scaling the bias voltage by the inverse of the nitride thickness: E−V/d.
Once the amount of trapped charge in the dielectric is minimized, the process parameters are considered to be optimum. MEMS devices can be fabricated using these optimum process parameters for depositing the dielectric layer 4.
The charge accumulation data extracted from flatband measurements on M-I-S structures show a significant decrease in charging of the improved dielectric relative to the standard silicon nitride deposition process. The slope of the accumulation curve decreased from ˜20 to ˜0.5. The slope of these curves is proportional to the concentration of trap states in the dielectric based on a model by Buchanan et al. (Solid State Electronics, Vol. 30, No. 12, pp. 1295-1301, 1987, the entire contents of which are incorporated herein by reference).
Buchanan et al. describes a simplified model for depicting charge transfer in and out of traps in the silicon nitride through a tunneling process, which shows that the charge (Q) trapped in the dielectric increases logarithmically with time (t):
Based on this model, the relative concentration of trap states (N) in a given dielectric can be obtained from the slope of a plot of Q(t) against 1 n(t/to), where to is a constant to make the quantity (t/to) dimensionless. The slopes of the curves shown in
MEMS capacitive switches have been fabricated to demonstrate the improvement in switch performance and reliability based on the use of the improved dielectric.
The flatband data shown in
Thus, new and improved dielectric suitable for use in electronic and MEMS devices have been provided herein. Two strong correlations were discovered relative to the observed molecular bonding and the charging behavior of the resulting nitrides. Desirable low charging behavior of a nitride corresponds with a high number of Si:Si bonding (demonstrated by the extinction coefficient measurements taken using 248 nm light), as well as a large ratio of Si:H bonds compared to N:H bonds. However, there is more than one possible explanation for why an increase in Si:Si bonds and a large Si:H/N:H ratio would cause the improved (i.e., decreased) charging behavior in the nitrides. Improved nitrides had Si:H/N:H ratios greater than 1, and preferably greater than 3 (e.g., three test nitrides that had good performance had ratios of 3.67, 9.75, 5.32 respectively). The extinction coefficients for improved nitrides measured at 248 nm (which correlates to Si:Si bonding) were greater than 0.06 and preferably greater than 0.1 (e.g., three test nitrides that had good performance had extinction coefficients of 0.363, 0.116 and 0.572 respectively). No poor performing nitrides were determined to have an Si:H/N:H ratio's greater than 1, or extinction coefficients at 248 nm greater than 0.06.
Thus, a number of preferred embodiments have been fully described above with reference to the drawing figures. Although the invention has been described based upon these preferred embodiments, it would be apparent to those of skilled in the art that certain modifications, variations, and alternative constructions would be apparent, while remaining within the spirit and scope of the invention.