Method for producing and testing a corrosion-resistant channel in a silicon device

TECHNICAL FIELD

0001This invention relates generally to silicon devices (including MEMS devices) and more specifically to a method for producing and testing a corrosion-resistant channel in a silicon device.

BACKGROUND ART

0002A relatively recent development in the semiconductor industry is to use microelectromechanical systems (MEMS) in semiconductor and pharmaceutical manufacturing processes. MEMS devices are typically silicon chips that include miniaturized mechanical components, such as actuators, mirrors, levers, diaphragms, or sensors. MEMS devices may also include electronic circuitry.

0003When MEMS devices are employed in semiconductor and pharmaceutical manufacturing processes, they are exposed to the chemical and biochemical substances used in such processes. The part of the MEMS device exposed to fluids <i>(i.e.,</i> gases or liquids) during operation is commonly referred to as the "wetted path." The wetted path may be different from the primary flow path <i>(i.e.,</i> the path along which the fluid is intended to travel) because fluids sometimes can enter into open spaces other than the primary flow path, referred to as the "dead volume."

0004The materials of the MEMS device that form the wetted path must be able to withstand corrosion or attack from fluids flowing through the device. In applications where corrosive fluids are present, the materials in the wetted path are critical, and compatibility of all the materials present is essential. In products requiring high purity, such as those used in the semiconductor or pharmaceutical industries, even a small amount of corrosion is unacceptable.

0005In many MEMS devices the wetted path is formed from a silicon channel, as MEMS devices are usually comprised at least in part by silicon wafers. The microvalve illustrated in Figures <b>1(a)</b> and <b>(b)</b>, in "off" and "on" states, respectively, is an example of a MEMS device with silicon in the wetted path. The valve is used to finely control the flow of fluids. The microvalve includes a heater plate <b>22</b>, a diaphragm plate <b>28</b>, and a channel plate <b>30</b>.

0006The channel plate <b>30,</b> which is formed from a silicon wafer, includes an input port <b>32</b> and an output port <b>34.</b> The illustrated valve is a normally-open valve in that fluid entering input port <b>32</b> normally is able to travel freely through the valve 100 and out via output port <b>34,</b> as depicted in Figure <b>1(a).</b> An example of a normally-closed valve is described in <patcit id="pcit0001" dnum="US6149123A"><text>U.S. Patent No. 6,149,123</text></patcit> (the "'123 patent"), the contents of which are incorporated by reference as if fully disclosed herein.

0007The diaphragm plate <b>28</b> includes a cavity <b>41</b> which holds a thermopneumatic liquid. The thermopneumatic liquid also extends up through channels <b>56</b> in the heater plate <b>22.</b> When control circuitry (not shown) associated with the valve indicate the valve should close, the heater plate <b>22</b> warms the thermopneumatic liquid. The diaphragm plate <b>28,</b> which is formed from a silicon wafer, includes a flexible diaphragm <b>44.</b> When the thermopneumatic liquid is heated, it expands, causing the diaphragm <b>44</b> to bend and block input port <b>32.</b> As illustrated in Figure <b>1(b),</b> when the input port <b>32</b> is blocked, the valve is closed and any fluid flow is severely restricted (e.g., less than 1 sccm).

0008The wetted path of valve <b>100</b> is cavity <b>43,</b> the input and output ports <b>32, 34,</b> and any exposed surfaces around the foregoing, all of which are formed from channel plate <b>30</b> and diaphragm plate <b>28.</b> As these two plates <b>28, 30</b> are made of silicon wafers, the wetted path is a silicon channel.

0009A valve similar in operation to valve <b>100</b> is described in <patcit id="pcit0002" dnum="US4996646A"><text>U.S. Patent No. 4,996,646</text></patcit> (the "'646 patent"). Another example of a normally-open valve is described in <patcit id="pcit0003" dnum="US6129331A"><text>U.S. Patent No.6,129,331</text></patcit> (the "'331 patent").

0010As stated above, the fluids flowing through MEMS devices, such as the valve illustrated in Figures <b>1(a)</b> and <b>(b),</b> must not corrode the device. For instance, if fluids were to sufficiently corrode the valve of Figures <b>1(a)</b> and <b>(b),</b> the diaphragm <b>44,</b> which is made up of a thin layer of silicon, would eventually break under operation. In addition, the cleanliness of the semiconductor or pharmaceutical process may be compromised by the products of the reaction of such fluids with the silicon. While silicon is non-reactive with most process gases and single constituent acids, it reacts with atomic fluorine, F, and other compounds which can spontaneously dissociate to atomic fluorine. A silicon atom, Si, will react with fluorine atoms to form SiF₄, a volatile component which vaporizes off the surface, thereby corroding the silicon. Consequently, there is a need to protect the wetted path from fluorine.

0011Also, some liquid bases (e.g., pH >8) or mixed acids will corrode silicon, and, therefore, there is also a need to protect the wetted path from such fluids.

0012In semiconductor manufacturing processes that etch silicon with fluorine, a mask is often used to cover those portions of the wafer where etching is not desired. Such masks are made of materials which are unreactive or react very slowly with fluorine. Examples of such materials are SiO₂, Si₃N₄, photoresist, or metal films of aluminum or nickel. However, these masks and the corresponding processes are used to selectively etch silicon and have not been employed to provide long-term protection of the wetted path of a MEMS device from corrosion by fluorine or other elements. In addition, such methods do not provide a means for identifying devices with inadequate coverage of the protective material.

0013Furthermore, such methods typically entail creating a protective metal film of aluminum or nickel by exposing aluminum or nickel layers to ClF₃ gas or F₂ gas, where the fluorine in these gases reacts with the metal to create a film, consisting of a non-volatile fluorine compound, over the metal. The creation of the film provides a "passivating layer" on the aluminum or nickel. Materials, like aluminum and nickel, with which fluorine reacts to create a nonvolatile compound, are known to form these passivating layers. The problem with using ClF₃ or F₂ is that such gases are corrosive and highly toxic, rendering the passivation process dangerous, difficult, and expensive. For instance, exposure of silicon to ClF₃ can produce extreme heat and may result in catastrophic failure of the MEMS device and associated equipment.

0014Applying materials, such as aluminum, nickel, or other protective layers, to the wetted path of a multilayer silicon MEMS devices presents an additional challenge. Some MEMS devices, such as valve 100, are comprised of two or more silicon wafers fusion bonded together. The fusion bonding creates hidden flow passages which are difficult to access using conventional deposition or electroplating techniques, and, thus such techniques are not suitable for multi-layer MEMS devices. Atomic layer deposition ("ALD") processes can more easily reach such hidden passages, but known, true ALD techniques do not enable materials like aluminum to be deposited in layers thick enough to adequately protect the silicon.

0015The hidden passages in a MEMS device also present a challenge in ensuring complete protection of the wetted path. It is very important that potential defects in the protective film be screened out prior to use in a hostile environment.

0016Therefore, there is a need for a process for depositing, passivating, and testing a fluorine-resistant (and/or base or mixed acid resistant) material in the wetted path of a single or multilayer MEMS device that is reliable and complete and preferably employs less toxic and corrosive compounds than ClF₃ or F₂ to achieve the passivating layer.

0017<patcit id="pcit0004" dnum="US4325984A"><text>US 4,325,984</text></patcit> discloses a method for preventing post-etch corrosion of an aluminum or an aluminum-alloy film which has been etched utilizing chlorinated plasma. In said method, post-etch corrosion is prevented by a step of exposing the etch film to sulfur hexafluoride plasma.

0018<patcit id="pcit0005" dnum="JP03136240B"><text>JP 03136240</text></patcit> comprises a method in which an aluminum film is applied to a semiconductor substrate and selectively removed by etching. The surface of the aluminum film is fluorinated according to a desired etching pattern and a dry etching is performed on the film using the fluorinated aluminum film as a mask.

DISCLOSURE OF INVENTION

0019According to one embodiment, the wetted path of a MEMS device is coated with a material capable of being passivated by fluorine. The device is then exposed to a gas that decomposes into active fluorine constituents either spontaneously or, preferably, when activated by a plasma or other energy source. One example of such a gas is CF#, an unreactive gas which is easier and safer to work with than reactive gases like ClF₃. The gas will passivate the material and corrode any exposed silicon. The device is tested in a manner in which any unacceptable corrosion of the wetted path will cause the device to fail. If the device operates properly, the wetted path is deemed resistant to corrosion by fluorine:

0020As discussed above, many MEMS devices are comprised of two wafers bonded together. In one embodiment, each of the wafers, prior to bonding, is coated at least in part with a material capable of both being passivated by fluorine and forming a eutectic bond with silicon. The wafers are then attached by a eutectic bond between the material and the silicon before being exposed to CF₄ or other similar gas.

0021By applying the selected material prior to bonding, it is more likely that all parts corresponding to the wetted path, including the hidden channels formed after bonding, will be adequately coated with the material. Furthermore, choosing a material that can form a eutectic bond with silicon means the bond will be achieved at a lower temperature than the melting point of the material, thereby ensuring the wafers can be attached without destroying the material. Aluminum and nickel are examples of the materials which can be applied to passivate fluorine and form a eutectic bond with silicon.

0022In an alternate embodiment, an organic bonding compound, instead of a metal, is deposited in those areas of the wafer(s) corresponding to the wetted path. The organic compound creates a polymer film over the silicon in the wetted path (which acts as a barrier against fluorine or other compounds) and, in multilayer applications, it can be used to bond the wafers together. An example of such an organic bonding compound is benzocyclobutene (BCB). The steps after application of the organic bonding compound (e.g., exposure to CF₄ or another gas with fluorine) are the same as described above.

0023An optional step that can be added to both of the above-described methods is to place the processed and bonded wafer structure in a plasma activated C₄F₈ gas or similar compound. The step, which preferably occurs after the exposure to a fluorine-based gas, provides a protective, Teflon-like film which acts as a further barrier to attack of the silicon by fluorine compounds.

BRIEF DESCRIPTION OF THE DRAWINGS

0024<ul id="ul0001" list-style="none" compact="compact"><li>Figures 1(a) and (b) illustrate cross-sectional views of a known MEMS microvalve in the "off" and "on" states, respectively.</li><li>Figure 2 illustrates a method for producing and testing a corrosion-resistant silicon device in accordance with one embodiment of the present invention.</li><li>Figure 3 illustrates a method for producing and testing a corrosion-resistant silicon device using a protective metal layer in accordance with one embodiment of the present invention.</li><li>Figure 4 illustrates a method for producing and testing a corrosion-resistant microvalve in accordance with one embodiment of the present invention.</li><li>Figure 5 illustrates the portions of an upper and lower wafer that correspond to a microvalve.</li><li>Figure 6 illustrates a method for producing and testing a corrosion-resistant silicon device using an organic bonding compound in accordance with one embodiment of the present invention.</li><li>Figures 7(a) and 7(b) illustrate an example process flow for producing a corrosion-resistant microvalve in accordance with the method illustrated in Figure 4.</li></ul>

MODES FOR CARRYING OUT THE INVENTION

0025Figure 2 illustrates a method for producing and testing a corrosion-resistant wetted path in a silicon device according to one embodiment of the present invention. Starting with a silicon wafer(s) in which the wetted path has been formed, the protective material is applied 210 to at least those portions of the silicon wafer(s) corresponding to the wetted path. As described in further detail below, examples of such protective material include (1) a metal, such as aluminum or nickel, that can be passivated by fluorine compounds or (2) an organic compound, such as BCB, that is either resistant to fluorine or can be passivated by fluorine.

0026The wafer (or multi-layer wafer structure if applicable) is then exposed 220 to a gas that decomposes into active fluorine compounds, either spontaneously or when activated by a plasma or other energy source. An advantage of using an unreactive gas that requires an energy source to decompose into fluorine compounds, such as CF₄, is that these gases usually are safer and easier to work with than those that decompose spontaneously. The purpose of the gas exposure is to (1) passivate the deposited material (if applicable, as some organic compounds may be non-reactive to fluorine without passivation) and (2) cause failure in any silicon region in the wetted path not protected by the deposited material. The conditions of the gas exposure are preferably optimized for etching silicon, thereby rendering it probable that any exposed silicon will be attacked and can be identified through inspection or testing of the completed wafer. The exposure to the gas should be long enough to cause failure of a silicon region unprotected by the passivating material during subsequent inspection or testing.

0027Examples of gases that may be used are CF₄, SF₆, NF₃, and ClF₃, but those skilled in the art will appreciate that some of the other fluorine-based gases, such as certain fluorinated hydrocarbons (especially those with an effective fluorine-to-carbon ratio > 2), also will decompose into active fluorine compounds that will passivate the deposited metal and etch any exposed silicon. For reference, the Semiconductor Equipment and Materials International (SEMI) draft document 3520 titled "Guidelines for Gas Compatibility with Silicon MEMS Devices" dated September 24, 2002 lists many known fluorinated hydrocarbons, and those skilled in the art will appreciate that some of these will etch silicon and passivate metals like aluminum and nickel. The <nplcit id="ncit0001" npl-type="s"><text>D.H. Flamm, et al. article "The Design of Plasma Etchants," Plasma Chemistry and Plasma Processing, Vol. 1(4), 1981; p. 317</text></nplcit>, also includes the etching properties of select fluorinated hydrocarbons.

0028After the wafer is exposed to the gas, it is subsequently tested 230 in such a way that any unacceptable corrosion of the wetted path would likely cause failure of the device. If the wetted path is not adequately protected, the fluorine compounds will corrode it during the gas exposure, and the device will not operate properly during the test. If the device operates properly, the wetted path is deemed resistant to fluorine (and possibly other fluids, depending on the type of protective material deposited).

0029As stated above, examples of the deposited protective material include (1) a metal that can be passivated by fluorine (e.g., aluminum, nickel) and (2) an organic compound like BCB. The above-described method is set forth in greater detail below with respect to using a metal or an organic compound as the protective material.

1. Metal Film

0030Figure <b>3</b> illustrates a method, according to one embodiment of the present invention, for protecting a MEMS device (or other silicon device) from corrosion by fluorine compounds by applying and passivating a metal layer in the wetted path. This method is described with respect to a single wafer, but the method is applicable to either single-layer or multilayer-wafer silicon devices. The method is as follows: <ol id="ol0001"><li>a. Starting with a completely processed wafer <i>(i.e.,</i> complete except for protection of the wetted path), the wafer is stripped 305 of any oxide or nitride and cleaned using conventional techniques. A layer of SiO₂ at least 5 nanometers ("nm") thick then is grown 310 on both sides the wafer. The maximum thickness of the SiO₂ layer is not necessarily critical and will depend on whether there are any thickness restrictions for the wafer in order for the device being manufactured to operate properly. However, the oxide layer should not be so thick that removing exposed regions of it would damage the metal film created as set forth below.</li><li>b. At least a 50nm layer of a metal is deposited 320 on the wafers using conventional physical vapor deposition ("PVD") or other well known techniques. As stated above, such metal should be capable of being passivated by fluorine compounds. The metal is deposited on at least those surfaces of the wafer that need to be protected from corrosion by fluorine atoms. The maximum thickness of the deposited metal layer on each wafer will depend on any thickness restrictions for the wafers due to device operation requirements. Examples of metals which may be used are aluminum or nickel. Aluminum will react with fluorine molecules to create AlF₃ (Al + 3/2 F₂ → AlF₃). AlF₃ is a non-volatile compound which forms a film over the aluminum, thereby protecting the aluminum (and, therefore, also the silicon) from reacting completely with fluorine compounds. Nickel also reacts with fluorine molecules to form a protective film, but aluminum is preferred to nickel because the nickel film does not adhere to silicon as well as the aluminum film.</li><li>c. Any additional wafer-level processing required is completed <b>330.</b></li><li>d. At any time deemed appropriate after depositing the metal layer (including prior to step <b>330</b>), the wafer structure is exposed <b>340</b> to a gas that decomposes to active fluorine compounds. As stated above, one example is CF₄ gas in a plasma discharge, but there are other fluorine-based gases which can passivate the deposited metal and etch exposed silicon. During the exposure, the active fluorine compounds react with the deposited material in the wetted path to create a non-volatile compound, which forms a protective film over the deposited material. The active fluorine compounds also reacts with any exposed silicon not covered by the protective film. Thus, the select gas, and the conditions of the gas exposure, are such that the gas passivates the deposited metal and etches any exposed silicon.</li><li>e. The wafer structure is tested <b>350</b> in such a way that any unacceptable corrosion of the wetted path likely would cause failure of the device. If the wetted path is not adequately protected, the fluorine will corrode it, and the device will not operate properly during the test. If the device operates properly, the wetted path is deemed to be resistant to corrosion by fluorine. The wafer structure also may be visibly inspected for any gaps in the coverage of the deposited material.</li></ol>

0031In an alternate embodiment, preferably after step (d) and prior to step (e), the wafer is placed in a plasma activated gas of C₄F₈. This process creates a polymer, Teflon-like film over the wafer and in the hidden channels, which provides an additional barrier to attack by fluorine compounds. During this step, the wafer temperature is maintained below 50°C. Other fluorinated hydrocarbons with an effective fluorine-to-carbon ration ≤ 2 generally may be used as an alternative to C₄F₈ (for instance, a CHF₃ and Argon mix, which has an effective fluorine-to-carbon ratio of 2 because the hydrogen atom is treated like a carbon for determining the ratio). For reference, the article dated <nplcit id="ncit0002" npl-type="s"><text>16 October 2000 and titled "Hydrophobic valves of plasma deposited octafluorocyclobutane in DRIE channels" by Helene Andersson, Wouter van der Wijngaart, Patrick Griss, Frank Niklaus, and Goran Stemme</text></nplcit> (the "Andersson <i>et al.</i> article").

0032Figure <b>4</b> illustrates an application of the method of Figure <b>3</b> to the manufacturing of a multilayer-wafer microvalve, such as the valve illustrated in Figures <b>1(a)</b> and <b>1(b).</b> The microvalve application is merely an example, and the method described with respect to Figure <b>4</b> can be applied to other multilayer devices having two or more wafers. For reference, Figure <b>5</b> illustrates the portions of the silicon membrane and channel plate wafers that correspond to the microvalve <b>100</b> illustrated in Figures <b>1(a)</b> and <b>(b)</b>. The method is as follows: <ol id="ol0002"><li>a. Prior to bonding, completed membrane and lower wafers <b>528</b>, 530 are stripped <b>400</b> of any oxide or nitride and cleaned <b>400</b> using conventional techniques. A 5-50 nm layer of SiO₂ is then grown <b>405</b> on both sides of the membrane wafer <b>528</b>. The SiO₂ is preferably no more than 50 nm thick on the membrane wafer <b>528</b> so as not to compromise the flexibility of the diaphragm <b>544</b>. On both sides of the lower wafer <b>530</b> (the channel plate), a 5 nm or thicker layer of SiO₂ is grown <b>410</b>.</li><li>b. Anywhere from 50-300 nm of a suitable metal (<i>i.e.,</i> one that can be passivated by fluorine, preferably aluminum) is deposited <b>415</b> on all parts of the flow side of membrane wafer <b>528</b> except for bonding regions, which are protected preferably by a shadow mask. Depositing more than 300 nm may cause the passivated metal film to peel as the diaphragm <b>544</b> moves or may impair movement of the diaphragm <b>544</b> itself. On the lower wafer <b>530</b>, 50 nm or more of the metal is deposited <b>420</b> uniformly on both sides of the wafer. Care is taken to cover completely all of the vertical channels. The metal may be applied to the wafers using conventional PVD or other well known techniques. By applying the metal prior to bonding the silicon wafers together, all parts corresponding to the wetted path are more likely to be adequately coated with the material. Since the metal is applied prior to bonding, it is advantageous to select a metal, such as aluminum or nickel, that can form a eutectic bond with silicon. Such a metal allows the wafers to be bonded without fusion bonding. Fusion bonding typically requires temperatures greater than 900°C, and the preferred metals for passivating against fluorine will melt at such temperatures. Eutectic bonding occurs at a temperature lower than the melting point of the metal and the silicon. For instance, aluminum and silicon form a eutectic bond at ~577°C, which is lower than the melting point of aluminum (~660°C) and the melting point of silicon (1430°C), enabling the wafers to be bonded without harm to either the aluminum or the silicon. In steps <b>415</b> and <b>420,</b> the purpose of applying metal to the bonding region of the lower wafer, but not the membrane wafer, is to enable the two wafers to be joined by a metal-silicon eutectic bond. In this microvalve example, the wafers are eutectically bonded together, but, in general, the wafers may be bonded in other ways, such as by an organic bonding compound, provided that the bonding process does not destroy the deposited metal in the wetted path. If the wafers are joined by a means other than through a eutectic bond, then the metal is not deposited in the bonding region of either wafer unless required by the selected bonding process.</li><li>c. The exposed SiO₂ in the bonding region of the membrane wafer <b>528</b> is then removed <b>425</b> by conventional wet or dry techniques just prior to the eutectic bonding step. Care is exercised that none or very little of the aluminum deposited on the membrane wafer <b>528</b> is removed. Subsequently, if required, the wafers <b>528</b>, <b>530</b> are cleaned <b>430</b> using conventional techniques.</li><li>d. The two wafers are aligned and mated <b>435</b> through a eutectic bonding process. The bonding is done in a non-oxidizing and, preferably, reducing atmosphere (which is typically achieved by adding hydrogen to a gas stream to prevent oxidation) in a furnace. Placing the wafers on a flat surface in the furnace with additional weight, such as a one pound quartz disc, on top of them can facilitate the bonding step. For aluminum, temperatures above 577°C, but below 660°C, are preferred and may be necessary depending on the flatness of the wafers and the weight applied to them. Some example processing steps for mating the membrane wafer and the lower wafer include aligning the wafers; placing them in contact; inserting the wafer pair in the furnace on a flat surface with weight on top of them; purging the furnace with an inert gas; purging the furnace with a reducing (without being explosive) gas, such as a forming gas (10% hydrogen in nitrogen); heating the furnace to between 577°C and 650°C for one minute or longer, and then removing the bonded wafers.</li><li>e. A Pyrex layer (e.g., heater plate <b>22</b>), which includes the heating unit for the microvalve, is anodically bonded <b>440</b> to the bonded wafer structure using conventional techniques. Also, any additional wafer-level required processing is completed.</li><li>f. Same steps as those described above with respect to step 340 in Figure <b>3</b>. In one embodiment, the wafer structure is loaded into a plasma reactor and exposed <b>445</b> to CF₄ at two torr for one hour at 350 watts.</li><li>g. The microvalves in the wafer structure are tested <b>450</b> to ensure proper operation. In one embodiment, the testing occurs after the wafer has been diced into individual valves. If the wetted path is not protected properly and completely by a protective film, then the fluorine compounds will have corroded the silicon in the wetted path during the fluorine-based gas exposure. As the diaphragm <b>544</b> is thin, it will break during operation if it is significantly corroded. In one embodiment, a valve die is tested by extending the membrane pneumatically to 200 psig or more. Any microvalve operating properly during testing is assumed to have resisted corrosion by fluorine components. In one embodiment, the valves are also visibly inspected for any gaps in the coverage of the protective material.</li></ol>

0033For reference, Figures <b>7(a)</b> and <b>(b)</b> illustrate a more detailed example of a process flow, at the wafer and die level, for creating a microvalve in accordance with the embodiment illustrated in Figure <b>4</b>. The steps from the section titled "Eutectic Bond Process Flow" and below relate to protection of the wetted path. The steps in the sections "Upper Silicon Process" and "Lower Silicon Process" relate to etching of the wafers and not to protection of the wetted path.

2. Organic Bonding Compound

0034In an alternate embodiment, the silicon-formed wetted path is protected from a corrosive fluid by applying an organic bonding compound to the wafers prior to bonding. The compound creates a polymeric film over the silicon in the wetted path and, in multilayer applications, bonds the wafers together. An example of such an organic bonding compound is benzocyclobutene (BCB).

0035An organic bonding compound can be used to protect against fluids other than fluorine. Liquid environments of pH > 8 and mixed acids will corrode silicon, and, unlike the metal films, compounds like BCB will protect the wetted path against such acids and bases, although such organic bonding compounds often are not as resistant to attack by fluorine as the metal films.

0036As illustrated in Figure <b>6</b>, an organic bonding compound, such as BCB, is applied <b>610</b> to at least the areas of the silicon wafer corresponding to the wetted path. In multilayer applications, the organic bonding compound may be applied to all the wafers, including the bonding regions. For example, in the microvalve application, the organic bonding compound is deposited evenly on the flow side of the membrane wafer and both sides of the lower wafer, with care taken to ensure that the vertical channels in the lower wafer are completely covered. The organic bonding compound is applied as thin as can easily be done without compromising coverage or the ability to form a hermetic seal between the wafers. In one embodiment, the thickness of the organic compound is about 2 microns.

0037Methods for applying an organic bonding compound to a silicon wafer are known to those skilled in the art. For reference, one such method is described in the article "<nplcit id="ncit0003" npl-type="s"><text>Void-Free Full Wafer Adhesive Bonding," presented at the 13th IEEE Conference on Microelectromechanical Systems in Miyazaci, Japan, January 23-27, 2000, pp. 247-252, by Frank Niklaus, Peter Enoksson, Edvard Kalveston, and Goran Stemme</text></nplcit>. The Dow Chemical Company also has made publicly available methods for applying its BCB compound "CYCLOTENE," which is one type of BCB that may be used.

0038In the multilayer applications, the wafers may be bonded 620 together with the organic bonding compound. When additional hermeticity is required, the organic bonding material is circumscribed by a hermetic epoxy compound in accordance with the teaching of <patcit id="pcit0006" dnum="US6325886B1"><text>U.S. Patent 6,325,886 B1</text></patcit>.

0039The wafer structure is subsequently exposed <b>630</b> to a gas that decomposes to active fluorine compounds either spontaneously or when activated by a plasma or other energy source. As discussed in more detail above, one example of such a gas is CF₄ (in a plasma discharge), but there are other fluorine-based gases which can be used. The conditions of the gas exposure are preferably optimized for etching silicon, thereby rendering it probable that any exposed silicon will be corroded. In one embodiment, the wafer structure is loaded into a plasma reactor and exposed to CF₄ at two torr for one hour at 350 watts.

0040If the organic bonding compound has been properly and completely applied to the wafers, it will act as a barrier between the fluorine compounds and the silicon in the wetted path. Specifically, the organic bonding compound will either be passivated by the fluorine, non-reactive with the fluorine, or slowly reactive with the fluorine (slow enough such that it will withstand attack from fluorine during the gas exposure in step <b>630</b>), depending on the particular organic compound used. To ensure complete protection, the wafer structure is tested <b>640</b> in such a way that significant corrosion of the wetted path would cause failure of the device. If the device operates properly, the organic bonding compound is presumed to completely cover the wetted path. In one embodiment of the microvalve application, the testing occurs after the wafer has been diced into valves, which are each tested by extending the membrane pneumatically to some appropriate minimum pressure (for example, 200 psig). The wafer structure may also be visibly inspected for any gaps in the coverage of the organic bonding compound.

0041If a particular organic compound is merely slowly reactive to fluorine compounds (as opposed to being essentially non-reactive), then such organic bonding compound will likely be used to protect the wetted path against fluids other than fluorine. Nevertheless, the aforementioned process of exposing the wafer(s) to a fluorine-based gas is still applicable (provided the reaction of the organic compound with the fluorine compounds is slow enough to withstand attack during such exposure) as a way to determine whether there are any gaps in the coverage of the organic compound. Step <b>630,</b> however, is not necessarily limited to fluorine-based gases in such cases, as the step also can be accomplished by exposing the wafer to another type of fluid that is corrosive to silicon but not to the organic bonding compound.

0042In an alternate embodiment, preferably after step <b>630</b> and prior to the testing step, the wafer structure is placed in a plasma activated gas of C₄F₈. This process creates a polymer, Teflon-like film over the wafer, which provides an additional barrier to attack by fluorine compounds. During this step, the wafer temperature is maintained below 50°C. As stated above, other fluorinated hydrocarbons with an actual or equivalent fluorine-to-carbon ratio ≤2 generally may be used as an alternative to C₄F₈ (for instance, a CHF₃ and Argon mix).