Fabrication of a Microchannel Plate From a Perforated Silicon

The invention pertains to a method for fabricating a microchannel plate (MCP) from a perforated silicon workpiece produced by photoelectrochemical etching (PECE).

There are various known methods for manufacturing microchannel plates (MCPs). It is instructive to begin by examining the relative capabilities of such known methods. The traditional method for manufacturing an MCP substrate uses a multi-fiber draw (MFD) technique employing reduced-lead silicate glass (RLSG). Another method employs reactive particle etching (RPE) and thin-film activation. The table below summarizes the relative capabilities, and cost of these methods.

It is desirable to improve and expand each of the categories to thereby improve the resulting products. It is also desirable that newly devised process steps may be employed to improve the manufacturing process. Such process steps may result in a new process category or may be combined with satisfactory elements of known processes. In either case, improvements in quality and productivity may be achieved.

The present invention is based upon the discovery that the manufacture of a microchannel plate may be improved using photoelectrochemical etching and thin film activation.

In an exemplary embodiment, the invention comprises the steps of defining a pore pattern on an etchable wafer; photoelectrochemical etching (PECE) the wafer to produce an array of anisotropic pores in the wafer corresponding to the defined pattern; and activating the pores. The activation step includes forming an electron emissive surface on the walls of the pores. In the case where the wafer is a conductive material, such as silicon (Si), an isolation layer is formed on the walls of the pores to separate the wafer from the emissive layer. Alternatively, the conductive wafer may be transformed to a less conductive material prior to activation.

FIG. 1 is a schematic illustration of a method employing photoelectrochemical etching (PECE) for manufacturing an MCP;

FIG. 2A is a cross-section of a Si substrate containing pores produced by PECE;

FIG. 2B is a cross-section of the Si substrate illustrated in FIG. 2 which has been partially oxidized to produce a thermally grown SiO.sub.2 layer;

FIG. 2C is a cross-section of the substrate illustrated in FIG. 2B which has been etched in an HF solution to remove the SiO.sub.2 layer;

FIG. 3 is a schematic diagram of a fixture employed to prevent warping of the Si substrate during oxidation;

FIG. 4A is a cross-section of the substrate illustrated in FIG. 2C which has been oxidized to produce an electrical isolation layer;

FIG. 4B is a cross-section of the substrate illustrated in FIG. 4A which has been oxidized to fully consume the Si;

FIG. 4C is a top plan view of an oxidized Si substrate with a rectangular array of pores with unoxidized Si along diagonals between the pores;

FIG. 4D is a top plan view of an oxidized Si substrate with a hexagonal array of pores where the space between pores is more uniform and hence fully oxidized;

FIG. 5A is a cross-section of a substrate in which a dielectric film has been formed on the walls of the pores;

FIG. 5B is a cross-section of the substrate illustrated in FIG. 5A in which an insulating oxidized layer and an overlying semiconductive film have been formed as a composite isolation layer;

FIG. 5C is a cross-section of the substrate illustrated in FIG. 5B in which a semiconductive film has been formed on the isolation layer;

FIG. 5D is a cross-section of the substrate illustrated in FIG. 5C in which an emissive layer has been formed on the semiconductive layer;

FIGS. 6A and 6B are schematic illustrations of a fusion bonding technique herein described;

FIGS. 7A and 7B are respective exploded and assembled side sectional views of an integrated device manufactured in accordance with the teachings of the invention; and

FIG. 8 is a cross-section of a stack of MCPs arranged in a multistage device.

The process for fabricating a microchannel plate is shown schematically in FIG. 1. According to the invention, an array of pores is produced by photoelectrochemical etching (PECE) in an etchable substrate or wafer to form a perforated workpiece. The pores are then activated by thin film processing techniques.

Photoelectrochemical etching (PECE) is a processing method illustrated schematically in FIG. 1 for etching a silicon (Si) substrate or wafer 10 with deep, highly anisotropic pores 12. In the process, an array of etch pits 14 is defined on one side of a n-type silicon wafer 10. The wafer 10 is then connected as an anode 16 in a electro-chemical cell and the patterned surface is exposed to a hydrofluoric-acid 18 containing electrolyte. The wafer 10 is simultaneously illuminated from the backside, i.e., the side without the etch pits by a source 20. The illumination produces a high concentration of minority charge carriers 22 (holes) at the base 24 of the etch pit 14 due to a focusing effect of the electric field lines in the space-charge region, whereby concave surface regions are more efficient in collecting holes than convex ones. The plane separating the space-charge region from the bulk lies below the tips 24 of the pores 12 advancing into the bulk. Accordingly, the region 26 between the pores 12 is depleted and all minority carriers 22 (holes) which are effectively collected at the tips or base 24 of the pores 12 with none left to promote etching of the sidewalls 28 of the pores. The base 24 of the pits 14 thus etch rapidly while the sidewalls 28 of the pits 14 do not etch, allowing for the formation of highly anisotropic pores 12. A typical etch rate for such a process is 0.5 .mu.m/min.

PECE is ideally suited for producing arrays of pores in silicon with diameters (D) ranging from 1 .mu.m to 10 .mu.m and lengths (L) ranging from a few microns to hundreds of microns. In addition, channels may be etched at a bias angle ranging from 0.degree.-10.degree. with respect to the substrate normal. Pores may be etched entirely through a wafer 10, thereby producing a perforated silicon workpiece. Alternatively, the pores may be etched partly through the wafer and the workpiece released. PECE can thus be used to produce a regular array of deep, highly anisotropic channels in a silicon substrate. According to the invention, the etched substrate is activated by thin film processing techniques, hereinafter described, to produce a micromachined electron multiplier.

The fabrication sequence described herein includes a sequence of thin film processing steps designed to produce layers of materials with the electrical and physical properties necessary to support electron multiplication. These layers include current-carrying semiconductive films with specific sheet resistances and films that have suitable secondary electron emissive characteristics. In addition, thin-film methods are described to modify the dimensions of the pores in the workpiece after the PECE step. Finally, techniques are described to overcome the mechanical limitations of the lamina during processing.

According to one aspect of the invention, the pore size and placement are characterized by the dimensions defined in FIG. 2A. Each pore 12 is characterized by the width (D) and the length (L). The spacing between the pores is characterized by pitch (P). (In a circular pore, D is the diameter.) The channel wall thickness (W) is equal to the pitch minus the pore diameter (P-D). The aspect ratio of the channel is defined as .alpha.=L/D. In one embodiment of the invention, the pores 12 are arranged in a rectangular array with the pore width (D) equal to the wall thickness (W), as shown. Thus, P=2D. It should be apparent to those skilled in the art that other spatial arrangements of pores 12, pore geometries, and combinations of pore width/wall thickness are possible. For example, a hexagonal arrangement may be produced. The pore width may be larger than the wall thickness or smaller than the wall thickness. Also, the pore arrangement may be aperiodic.

As shown in FIG. 2B, after the perforated silicon work-piece 10 is produced, the wall surface 28 may be converted to silicon dioxide (SiO.sub.2) 30 by thermal oxidation. This is done either to electrically isolate the silicon, or to modify the array dimensions. Generally, oxidation may be conducted via pyrogenic steam in the temperature range 1000.degree.-1200.degree. C. Typically, a H.sub.2 /O.sub.2 ratio of about .about.2:1 may be used, and a chlorine (TCA, HCl) source may also be included.

Silicon dioxide is less dense than silicon. Thus, there is a volume expansion associated with the oxidation of the silicon wall 28. The change in volume when the surface 28 of a material is oxidized is known as the Pilling-Bedworth ratio, which equals 2.2 for a flat silicon surface. For concave silicon surfaces such as the interior wall of pores 12, this ratio is higher than that for a flat surface. Thus, for every micron of silicon consumed, 2.2 microns or more of oxide are grown; and the diameter of the pore decreases during the oxidation, as shown in FIG. 2B.

Volume expansion during oxidation produces compressive stress in the sample surface, which can cause the workpiece to bow. In the case of workpieces having 2.5 .mu.m pores on 4 .mu.m centers with an .alpha. of .about.30, bowing takes place as the thickness of the oxide exceeds .about.0.5 .mu.m. The bowing can be so severe so as to cause the piece to roll-up upon itself if the oxidation is allowed to continue. If the wall thickness is substantially greater, or the overall length of the pores is higher, it is likely that substantially more oxide could be grown without bowing of the workpiece. Also, as discussed hereinafter, the workpiece may be completely converted to oxide, thus removing the stress caused by materials differences and allowing the workpiece to remain flat.

One method successfully employed to prevent bowing is to physically constrain the sample during oxidation with a weighted fixture 40, as shown in FIG. 3. In an exemplary embodiment, the pressure exerted on the workpiece by the fixture presently used is .about.100 g/cm.sup.2. The fixture 40 may be in the form of a vessel 42 having an opening 44 for receiving a weight 46. The workpiece or wafer 10 is constrained between respective, confronting surface portions 48 and 50 of the vessel 42 and the weight 46. The fixture 40 should be of a material that is compatible with the thermal oxidation process, such as fused silica. The surface portions 48 and 50 of the fixture 40 that contact the workpiece 10 must be flat to properly constrain the workpiece. Unless precautionary steps are taken, it is possible that the workpiece 10 may bond to the surface portions 48 and 50 of the fixture during oxidation. This can be eliminated by choosing a material and surface texture that prevents this bonding. In an exemplary embodiment of the invention, the surface portions 48 and 50 each comprise a silicon wafer coated with silicon nitride. The surface of each wafer 48 and 50 that contacts the workpiece 10 is not polished. This configuration successfully constrains the sample without allowing the workpiece to bond to the fixture.

In one embodiment of the invention, it may be desirable to thin the silicon wall 28 of the workpiece 10. This may be done by consuming the silicon by thermal oxidation, as illustrated in FIG. 2B, and then, removing the oxide 30 by etching in hydrofluoric acid (HF), which etches SiO.sub.2 without etching Si. This decreases the thickness W of wall 28 and increases width D of the silicon pores 12 without changing the pitch P of the array, as shown in FIG. 2C.

The steps described thus far produce an array of pores with the desired dimensions of D, P, and W. The following describes the steps for activating the channels to form continuous dynode electron multipliers.

As shown in FIG. 4A, when a conductive substrate 50 such as Si is employed, it first becomes necessary to electrically isolate the substrate. One method for isolating the substrate 50 is by direct thermal oxidation of the silicon, as described previously. The pore diameter is decreased as the silicon is converted to oxide and .alpha. increases, as shown in FIG. 4A. For the embodiment of a rectangular array of pores 52, it has been found that such a workpiece may be oxidized until the silicon 51 between adjacent pores 52 is consumed, as shown in FIG. 4B. However, unconsumed silicon 54 may be present along the diagonal between pores 52, as shown in FIG. 4C. The stress thus created by oxidation of this residual Si may cause the piece to warp and it may even crack. It is possible to eliminate this effect by changing the placement of the pores in the array. For example, the pores 52' may be arranged in a hexagonal arrangement, shown in FIG. 4D. With this configuration, the distance between nearest-neighbors is more uniform and may allow oxidation to continue until the entire silicon structure is consumed and converted to silicon dioxide. Other methods of isolating the Si substrate are described in U.S. Pat. No. 5,378,960, herein incorporated by reference.

Another method for electrically isolating the substrate shown in FIG. 5A, is to deposit a dielectric film 60 directly on the channel walls 62 by chemical vapor deposition (CVD). Material systems which are suitable for this application include silicon oxide (Si.sub.x O.sub.y), silicon nitride (Si.sub.x N.sub.y), and silicon oxynitride (Si.sub.x O.sub.y N.sub.z). These materials may be deposited from the precursor system of SiH.sub.2 Cl.sub.2 /N.sub.2 O/NH.sub.3, where the fraction of ammonia and nitrous oxide precursors are varied to produce films with the desired fractions of oxygen and nitrogen respectively. This precursor system deposits conformally in high-aspect-ratio channels, thus making it a viable system for isolating channels by CVD. In general, CVD processing may be carried out at reduced pressures (generally between 0.1 torr-3 torr) in a temperature range between about 800.degree. and about 900.degree. C. Isolating the substrate by CVD deposits material directly on the channel wall 62 and thus reduces the width (D) of the pore without consuming the silicon wall. Thus, .alpha. increases as does the overall wall thickness.

The two methods described herein for isolating the substrate may be combined to form a composite isolation layer, as shown in FIG. 5B. The silicon substrate 50 is first thermally oxidized, to produce oxide layer 64; and then, additional insulation may be provided by the layer of film 60 deposited on the thermal oxide layer 64 by CVD. The deposition of material directly on the wall may be used to increase the aspect ratio to a value necessary to support electron multiplication.

After electrical isolation, a semiconductive layer 68 may be deposited on the insulated substrate by CVD, as shown in FIG. 5C. Nitrogen-doped silicon is one material that can be used. The film may be formed from a SiH.sub.2 Cl.sub.2 /NH.sub.3 precursor system, to produce a highly conformal film in the high-aspect-ratio channels. This process generally may be carried out at reduced pressures (30 to 300 mtorr) and in a temperature range between about 750.degree. and about 850.degree. C. The electrical resistivity of these films may be controlled by varying the ratio of the precursors during deposition, generally between .about.10:1 to .about.4:1 SiH.sub.2 Cl.sub.2 /NH.sub.3. Sheet resistances (R.sub.s =10.sup.11 -10.sup.14 .OMEGA./Sq.) of this semiconductor layer are suitable for MCPs. The addition of this material to the channel walls increases the aspect ratio of the channel. See also, U.S. Pat. No. 5,378,960 which describes other methods of producing thin-film dynodes.

The final step in activating the channels is to form a secondary-electron emissive layer 70 on the conductive layer 68, as shown in FIG. 5D. This may be done by chemical vapor deposition of a dielectric, including silicon oxide, silicon nitride, or silicon oxynitride from the SiH.sub.2 Cl.sub.2 /N.sub.2 O/NH.sub.3 precursor system. The CVD may generally be conducted at a relatively low temperature (.about.700.degree. C.) to reduce the deposition rate and to produce maximum conformality in the channel. Other methods of producing thin-film dynodes are incorporated herein by reference to U.S. Pat. No. 5,378,960.

Another method of forming an emissive layer 70 is by direct thermal conversion of the surface 72 of the silicon-containing conductive film 68 by oxidation or nitridation. A silicon oxide emissive layer 70 may thus be formed by oxidizing the same conductive layer in oxygen or steam. A silicon nitride emissive layer 68 may be formed by heating the semiconductive layer in ammonia. Silicon oxynitride may be formed by heating the semiconductive layer in nitrous oxide. An oxynitride film may also be formed by heating an existing silicon dioxide layer in ammonia to incorporate nitrogen in the film.

Listed below are a series of exemplary process steps in the various embodiments described above:

A distinct advantage to silicon based MCPs is the ability to apply other bulk and surface micromachining techniques. This has application in two specific areas both of which take advantage of direct fusion bonding and field-assisted thermal or anodic bonding techniques illustrated in FIGS. 6A and 6B. Direct fusion bonding is a thermally driven process by which two clean, flat surfaces are fused. By heating the two surfaces 80 and 82 in contact to high temperature (800.degree. to 1200.degree. C.), generally for several hours, a permanent bond is form between the two pieces. Anodic bonding is an electrochemical process for low temperature (300.degree. to 600.degree. C.) fusion or sealing of alkali containing glass 82 to metal or semiconductors 80, 80A and 80B. It is accomplished by heating the materials in contact and then applying a positive bias 86 to the metal or semiconductor 80A, relative to the previously bonded glass semiconductive pair 82-80B, of the order of 100 to 1000 volts. Bonding can then take as little as several minutes depending primarily on the voltage, temperature, and component stack.

An application of bonding techniques available with a Si based micromachined MCP manufactured according to the invention relates to assembly of a low-cost, ultra-compact image or photomultiplier tubes using anodic bonding. It should be understood that an image tube and a photomultiplier tube differ mainly in application and in photocathode material. For example, the semiconductor substrate in FIG. 6A may be an MCP bonded or sandwiched to a glass window by means of anodic bonding. Additionally, substrates can be bonded to the fused stack of substrates and windows, one at a time by alternating glass and semiconductor substrates, as shown in FIG. 6B. Important considerations in employing such a bonding method are the relative thermal expansion of the materials used and the cleanliness and flatness of the surface to bonded.

The components layers for an image tube or photomultiplier 90 tube would include an MCP as herein described or as described in U.S. Pat. No. 5,086,248, the teachings of which are incorporated herein by reference. The arrangement, illustrated in FIGS. 7A and 7B in respective exploded and assembled forms, includes a glass window 92, with photocathode 94, micromachined MCP 96, a glass spacer 98, a silicon spacer 100, and a glass window 102 having a phosphor screen 104. Once the elements are bonded in vacuum, the image tube having a integrated structure results, as shown in FIG. 7B. The bonding procedure can be practiced to produce micromachined image tubes or photomultiplier tubes with electronic readouts (e.g., CCDs and CIDs). The image and photomultiplier tubes thus produced are compact and modular in design. Exemplary dimensions of the various layers are set forth below.

Another application of bonding techniques, available with Si based micromachined MCPs, is for multiply stacked, high gain (10.sup.6 -10.sup.8), high resolution (.gtoreq.50 lp/mm), detectors. Traditionally, MCPs are stacked together to form high gain detector free from ion feedback. This is accomplished by physically holding two or more MCPs together with opposing bias angles. The lack of long range order in the arrangement of channels in RLSG MCPs militates against one-to-one channel registration between the two or more stacked MCPs, resulting in a reduction in spatial resolution. For example, a single RLSG MCP with D=10 .mu.m and P=12 .mu.m would have a resolution of about 25 to 40 lp/mm depending on the assembly and operation of the detector. Two such MCPs, arranged in confronting relation with reverse bias angles, would have a spatial resolution of about 16 to 25 lp/mm. An arrangement of three such MCPs, necessary for very high gains (10.sup.8), results in a resolution of about 8 to 12 lp/mm.

By relying on the long range order of channels defined by lithographic techniques available with micromachined MCPs, a stack 110 of MCPs 112A-112C each having the same bias angle B may be bonded, as shown in FIG. 8, which maintain one-to-one channel registration of the respective channels 114A-114C through the stack. This results in no loss of spatial resolution. Thus, it is possible to have an micromachined MCP based detector with 10.sup.8 gain and a spatial resolution >100 lp/mm.

While there have been described what are at present considered to be the preferred embodiments of the present invention, it will be apparent to those skilled in the art that various changes and modifications may be made therein without departing from the invention, and it is intended in the appended claims to cover such changes and modifications as fall within the spirit and scope of the invention.