BACKGROUND OF THE INVENTION The present invention relates to fluid handling devices. More particularly, it relates to reliable, accurate, fluid flow regulators which are capable of handling fluid flow rates which are so low that they may be measured in hundredths of a cubic centimeter per day; which have a zero electrical energy consumption; and which may be economically mass produced by using micromachining processes.

SUMMARY OF THE INVENTION In many medical situations, it is desirable to continually administer fluid medication to a patient over an extended period of time at a relatively low flow rate. Examples of such cases are the use of morphine for the treatment of malignant or non-malignant pain; the use of FUDR for cancer chemotherapy; and the use of baclofen for the treatment of intractable spasticity. This manner of administering the medication is desirable because the level of the fluid medication in the patient's blood remains at a relatively constant, medically effective level. By way of contrast, if the medication was administered periodically in larger doses, such as in tablet form by mouth, the level of the medication in the patient's blood may tend to fluctuate markedly over time, from too little to too much, rather than staying at the desired medically effective level.

Accordingly, one general aspect of the present invention may be to provide fluid flow regulators which are capable of continually handling fluids over an extended period of time at relatively low flow rates, which may be as low as about 0.01 cc/day.

In many medical situations it is desirable to have fluid flow regulators which are extremely small, so that they may be implanted within a patient's body. Accordingly, another general aspect of the present invention may be to provide fluid flow regulators which are

so small that they may be easily implanted within a patien 's body.

However, when making a fluid flow regulator which is so small, it becomes relatively easy to inadvertently clog any openings in the fluid flow regulator (such as its ports, channels, cavities, or gaps) if a bonding material is used to bond the various parts of the fluid flow regulator together. Accordingly, another general aspect of the present invention is to provide fluid flow regulators in which at least some of their parts are anodically bonded together, thereby eliminating the need to use a separate bonding material to bond those parts together.

If the fluid flow regulators are intended to be implanted in a patient's body, it is preferable that they be passive devices, which do not consume any electrical energy at all.

In medical situations, the reliability of the fluid flow regulators which handle the fluid medication must be very high. In general, reliability may be enhanced by simplifying the fluid flow regulators to have as few total parts as possible; to have as few moving parts as possible; and to have nominal operating pressures which are as low as possible. Accordingly, one general aspect of the present invention may be to provide fluid flow regulators which are inherently highly reliable because they may have a total of as few as two parts, as few as one of which may be a moving part. Another general aspect of the present invention may be to provide fluid flow regulators which may operate at pressures which are relatively low. In medical situations, it is desirable to have fluid flow regulators which have a fail-safe design so that if they are subjected to overpressures in excess of their nominal design limits, they are very resistant to failure, so that they do not deliver excessive amounts of the medication to the patient. Accordingly, another general aspect of the present invention may be to provide fluid flow regulators which, when subjected to an overpressure, may be resistant to being ruptured because their flexures or membranes may be at least partially supported by at

least one other element of the fluid flow regulators; and which may reduce the flow of the medication, or shut it off altogether.

In view of the generally high cost of medical care, it is desirable to provide high quality, accurate, reliable fluid flow regulators at a price which is so economical that the fluid flow regulators may be considered to be disposable. Accordingly, another general aspect of the present invention may be to achieve these goals by using micromachining processes to mass produce the fluid flow regulators, or parts thereof.

Since many medications and body fluids are corrosive, particularly where the fluid flow regulator is used for an extended period of time, or is implanted in a human or an animal, it is important that the fluid flow regulator be corrosion-resistant. Accordingly, three further general aspects of the present invention may be to provide a fluid flow regulator having a layer of one or more corrosion- resistant substances; to bond such a corrosion-resistant layer to the fluid flow regulator by using anodic bonding; and to automatically perform such anodic bonding of the corrosion-resistant layer at the time certain other parts of the fluid flow regulator are being anodically bonded together. In many medical situations, it may be desirable to maintain the flow rate of the liquid medication to a patient at a predetermined rate, despite any fluctuations (either increases or decreases) in the pressure of the supply of the medication. For example, if the supply of the medication comprises a reservoir in which the medication is pressurized with a gas, as the reservoir is emptied the gas expands, thereby reducing the pressure on the decreasing amount of medication remaining in the reservoir. Accordingly, in addition to one or more of the above general aspects of the present invention, a specific aspect of the present invention may be to provide a fluid flow regulator which will maintain the flow of the medication within predetermined parameters, despite fluctuations in the pressure of the medication which is

received by the fluid flow regulator, or fluctuations in the pressure at the medication outlet port.

Such a fluid flow regulator may comprise a substrate having fluid inlet means, a regulator seat, and fluid outlet means. The fluid flow regulator may further comprise a flexure bonded to the substrate, and a regulator gap which is located between the flexure and the regulator seat.

During use, both the fluid inlet means and the outer surface of the flexure may be exposed to a source of liquid medication under pressure. The medication will flow sequentially through the fluid inlet means, the regulator gap and the fluid outlet means. As the medication flows through the regulator gap, the height of the regulator gap will tend to decrease as the driving pressure difference across the regulator increases, and will tend to increase as the driving pressure difference across the regulator decreases. As a result, the fluid flow regulator will tend to hold the flow rate of the medication constant, despite any fluctuations in the driving pressure difference of the medication across the fluid flow regulator.

Another specific aspect of the fluid flow regulator of the present invention may be that its flow rate may be selectively increased or decreased by selectively increasing or decreasing the number, size, shape and length of its fluid inlet means.

A further specific aspect of the fluid flow regulator of the present invention may be that its characteristic flow rate verses its applied driving pressure difference response may be chosen by selectively adjusting the fluid flow resistances of the fluid inlet means and the regulator gap with respect to each other.

Other specific aspects of the fluid flow regulator of the present invention may be that it may be a radial flow regulator, in which at least a portion of its fluid inlet means extend at least substantially around its regulator seat's periphery; in which its flexure overlies its regulator seat and extends outwardly past its regulator seat's periphery; in which at least part of its fluid

outlet means are located within its regulator seat; and in which the medication flows from its fluid inlet means radially inwardly across its regulator seat's top surface, from its regulator's periphery.to its fluid outlet means. Further specific aspects of the flow regulator of the present invention may be that it may be a linear flow regulator which may have an elongated regulator seat and an elongated flexure which extend between the inlet means and the outlet means. Other specific aspects of the linear flow regulator of the present invention may be that it may have a length to width ratio (L/ ) in the range of from about 5:1 to about 1000:1, and preferably about 20:1; that its flexure be unrestrained at its inlet means; and that its regulator seat and flexure may follow a straight course, a non- straight course (such as circular, spiral, or sinuous) , or a combination thereof.

Two further specific aspects of the linear flow regulator of the present invention may be that its regulator seat may have a contoured shape, such as the shape its flexure would assume if its flexure was unsupported by its substrate, and was subjected to a certain driving pressure difference across the regulator; and that such a contoured shape may be imparted to its regulator seat by pressure deflecting the flexure down into the substrate while the substrate is in a softened state, maintaining such pressure while the substrate is hardened, and then releasing such pressure and permitting the flexure to return to its original, undeflected configuration. Other aspects of the linear flow regulator of the present invention may be that its regulator seat may comprise a channel in its substrate; that its channel may not have a contoured shape; and that its channel may be micromachined into its substrate by being etched into its substrate.

Although all of the forgoing comments regarding the fluid flow regulators of the present invention have been with reference to handling medicinal fluids in a medical context, it is understood that the fluid flow regulators of

the present invention may also be used to handle any type of non-medicinal fluid, in both medical and non-medical contexts. In addition, although the fluid flow regulators which were mentioned above may be very small and may handle fluids at very low flow rates and at relatively low pressures, it is understood that, by applying scaling laws, the fluid flow regulators of the present invention may, in general, be scaled up to any desired size; to handle any desired fluid flow rate and pressure. Further, the term "fluid" is used in its broad sense, and includes both liquids and gasses.

It should be understood that the foregoing summary of the present invention does not set forth all of its features, advantages, characteristics, structures, methods and/or processes; since these and further features, advantages, characteristics, structures, methods and/or processes of the present invention will be directly or inherently disclosed to those skilled in the art to which it pertains by the following, more detailed description of the present invention.

BRIEF DESCRIPTION OF THE FIGURES Fig. 1 is a top elevational view of a micromachined radial flow regulator of the present invention;

Fig. 2 is a cross-sectional view thereof, taken substantially along line 2 - 2 of Fig. 1;

Fig. 3 is a graph depicting certain of the fluid flow characteristics thereof;

Fig. 4 is a schematic diagram of some of the fluid characteristics thereof; Figs. 5 and 6 are graphs depicting certain further fluid flow characteristics thereof;

Fig. 7 is a perspective view, partly in a cross- section taken substantially along line 7 - 7 of Fig. 8, of a micromachined linear flow regulator of the present invention having a contoured regulator seat;

Fig. 8 is a top elevational view thereof; Figs. 9 - 11 are top elevational views of three additional embodiments thereof;

Fig. 12 is a graph depicting certain fluid flow characteristics thereof;

Fig. 13 is a perspective view, partly in a cross- section taken substantially along line 13 - 13 of Fig. 14, of a micromachined linear flow regulator of the present invention having a non-contoured regulator seat;

Fig. 14 is a top elevational view thereof; and Fig. 15 is a is a graph depicting certain fluid flow characteristics thereof. DESCRIPTION OF THE PREFERRED EMBODIMENTS

MICROMACHINED RADIAL FLOW REGULATOR 32 (FIGS. 1 - 6):

STRUCTURE Turning now to Figs. 1 - 2, the micromachined radial flow regulator 32 of the present invention may be used to control the flow rate of a fluid medication 12 passing through it. The term "medication" is used in its broad sense throughout this document, and may be any fluid, whether or not the fluid is medicinal in nature; unless the context should indicate otherwise. Similarly, the term "fluid" is also used in its broad sense throughout this document, and may include both liquids and gasses; unless the context should indicate otherwise.

The radial flow regulator 32 may comprise a substrate 34 and a membrane 36. The substrate 34 may have four radially oriented inlet channels 38, a ring-shaped inlet cavity 40, a ring-shaped regulator seat 42, a cylindrical outlet cavity 52, and a venturi-shaped outlet port 54. In Fig. 1, the membrane 36 is depicted as being transparent, for clarity, so that the substrate 34's various features may be seen more easily.

The membrane 36 may have four mounting portions 26, which are mounted to respective portions of the substrate 34*s top surface 46; a circular, flexible flexure 28, which lies over the inlet cavity 40 and the regulator seat 42; and four inlet channel cover portions 30, each of which lie over a respective inlet channel 38. Although the membrane 36 is illustrated as being of uniform thickness, and as having flat bottom and top surfaces 50, 62, the membrane 36 may not be of uniform thickness, and may have bottom and

top surfaces 50, 62 which are not flat. Although the membrane 36 is illustrated as having four mounting portions 26, it may have fewer or more mounting portions 26.

A ring-shaped regulator gap 48 is provided between the regulator seat 42 and the flexure 28.

Although four, straight, radially oriented inlet channels 38 are illustrated, each having a rectangular cross-sectional configuration and a respective cover portion 30, there may be fewer or more inlet channels 38, each having a respective cover portion 30; any particular inlet channel 38 may have any other suitable size and cross-sectional configuration, such as square or rounded; the length of any particular inlet channel 38 may be varied; and any particular radial inlet channel 38 need not follow a straight, radially oriented course, but may follow a circular, spiral, serpentine, or other non-straight, non- radially oriented course, such as do the channels 86 of the regulators 80 of Figs. 9 - 11. The use of one or more inlet channels 38 following a circular, spiral, serpentine, or other non-straight, non-radially oriented course may be desirable since it may permit the manufacture of a radial flow regulator 32 which is more compact, as compared to a radial flow regulator 32 having straight, radially oriented inlet channels 38. Although the ring-shaped inlet cavity 40 is illustrated as having a circular or cylindrical configuration, and a uniform depth, it may have any other suitable size and configuration, and a non-uniform depth. In addition, although the inlet channels 38 and the inlet cavity 40 are illustrated as being separate elements, it is understood that these elements may be merged into each other so that they are no longer distinct elements. This may be done in any suitable way, such as by enlarging the inlet channels 38 until they perform most, if not all, of the functions of the inlet cavity 40; by enlarging the inlet cavity 40 until it performs most, if not all, of the functions of the inlet channels 38; or by any combination of the forgoing two ways.

Although the ring-shaped regulator seat 42 is illustrated as having a circular or cylindrical configuration, a flat top surface 44, and a uniform thickness, it may have any other suitable size and configuration, a top surface 44 which is not flat, and a non-uniform thickness.

Although the ring-shaped regulator gap 48 is illustrated as having a circular or cylindrical configuration, and a uniform height, it may have any other suitable size and configuration, and a non-uniform height. Although the regulator gap 48 is illustrated as being formed by selecting the regulator seat 42 to have a thickness such that its top surface 44 is lower than the portions of the substrate 34's top surface 46 to which the membrane 36's mounting portions 26 are secured, the regulator gap 48 may be formed in any other suitable way. For example, the regulator seat 42's top surface 44 and the substrate 34's top surface 46 may be selected to be co- planar, and the regulator gap 48 may be formed by reducing the thickness of the portion of the flexure 28 which overlies the regulator 42 by an amount equal to the desired height of the regulator gap 48. Alternatively, the regulator gap 48 may be formed by a combination of the two forgoing ways. Although a single outlet cavity 52, and a single outlet port 54 are illustrated, there may be more than one of each of these elements.

Although an outlet cavity 52 having a circular or cylindrical configuration and a uniform depth is illustrated, it may have any other suitable size and configuration, and a non-uniform depth. The outlet cavity 52 may be used to define a clean outer perimeter for the outlet port 54, particularly if the outlet port 54 is drilled with a laser. However, the outlet cavity 52 may be eliminated, and the outlet port 54 may be extended upwardly so that it communicates directly with the regulator gap 48. Alternatively, the outlet port 54 may be eliminated, and the outlet cavity 52 may be extended downwardly so that it

communicates directly with the regulator 32's bottom surface 56.

Although an outlet port 54 having a venturi-shaped configuration is illustrated, it may have any other suitable configuration, such as round or cylindrical.

Although the inlet cavity 40, the regulator seat 42, the regulator gap 48, the outlet cavity 52 and the outlet port 54 are illustrated as being uniformly arranged with respect to each other around a common center, they may be arranged with respect to each other in any other suitable way, and may not have a common center.

The regulator 32 may have a flow rate for medical applications in the range of from about 0.01 cc/day to about 20 cc/day; and preferably in the range of from about 0.1 cc/day to about 2.0 cc/day. However, it is understood that, in view of all of the disclosures contained in this document, the radial flow regulator 32 may be scaled up or down in size to regulate higher or lower flow rates of the medication 12. By way of example, the chip bearing the radial flow regulator 32 may be a square having sides about 4.83 mm long. Its membrane 36 may be manufactured from silicon; and may have a thickness of about 25 microns. Its substrate 34 may have a thickness of about 0.5 mm, and may be manufactured from 7740 Pyrex glass, manufactured by the Corning Company located in Corning New York. The inlet channels 38 may have a length of about 2.54 mm, a width of about 107 microns, and a depth of about 5.65 microns. The ring-shaped inlet cavity 40 may have a depth of about 5.65 microns, an O.D. (outer diameter) of about 2.29 mm, and an I.D. (inner diameter) of about 1.52 mm, (i.e., the ring- shaped cavity 40 may have a radial width of about 0.77 mm). The ring-shaped regulator seat 42 may have an O.D. of about 1.52 mm and an I.D. of about 0.5 mm, (i.e. the ring-shaped regulator seat 42 may have a radial width of about 1.02 mm) . The ring-shaped regulator gap 48 may have a height of about 2.5 microns, when the driving pressure difference (P) across the flexure 28 is zero; and may have a radial width of about 1.02 mm. The outlet cavity 52 have a width of 0.5

mm, and a depth of about 3.15 microns. The outlet port 54 may have z minimum diameter of about 100 microns, and a depth of about 494 microns. The flow characteristics of this example radial flow regulator 32 are illustrated in Figs. 5 and 6.

As will be appreciated from all of the disclosures in this document, the fact that the regulator 32 may, as in the example set forth above, have an extremely small size, be extremely light weight, have only two parts, and have a zero electrical energy consumption, offer numerous advantages over a regulator 32 which was physically much larger, much heavier, more complex or consumed electrical energy. For example, the regulator 32 may be ideal for use as part of a miniaturized medication delivery device which is to be implanted in a human or animal for delivery of constant flows of the medication 12 at flow rates as low as about 0.01 cc/day — flow rates which are so low that they may be impossible for a physically larger flow regulator of a different design to reliably and accurately deliver. MICROMACHINED RADIAL FLOW REGULATOR 32 (FIGS. 1 - 6):

OPERATION AND DESIGN The radial flow regulator 32 may be installed in its intended location of use in any suitable way. Any suitable medication supply means may be used to connect the radial flow regulator 32's inlet channels 38 to a source of the medication 12; and any suitable medication delivery means may be used to connect the radial flow regulator 32's outlet port 54 to whatever person, animal or thing is to receive the medication 12 from the outlet port 54. In some cases, the medication supply means may also be used to supply the medication 12 to the flexure 28's top surface 62, at a pressure which may or may not be the same as the pressure at which the medication 12 is supplied to the inlet channels 38. For example, the radial flow regulator 32 may be installed within any type of reservoir means for the medication 12 by any suitable means, such as by locating the radial flow regulator 32's outlet port 54 over the reservoir means's outlet, and by using an adhesive face

seal between the radial flow regulator 32's bottom surface 56 and the inside of the reservoir means to hold the radial flow regulator 32 in place. As a result, when the reservoir means is filled with the medication 12, the radial flow regulator 32 will be immersed in the medication 12, with its inlet channels 38 and its flexure 28's top surface 62 in fluid communication with the medication 12 within the reservoir means, and with its outlet port 54 in fluid communication with the reservoir means' outlet. Such an installation for the radial flow regulator 32 has numerous advantages.

For example, it is quick, easy, reliable and inexpensive, because no additional medication supply means (such as supply conduits) are needed to supply the medication 12 to the radial flow regulator 32's inlet channels 38 and to the flexure 28's top surface 62 (since they are already immersed in the medication 12) ; and because no additional medication delivery means (such as delivery conduits) are needed to convey the medication 12 away from radial flow regulator 32's outlet port 54 (since the reservoir means' outlet is used for this purpose) . Such additional inlet and outlet conduits may be undesirable since it may be relatively time consuming, difficult and expensive to align and connect them to radial flow regulator 32, due to the extremely small size of its inlet channels 38, flexure 28, and outlet port 54. Such additional inlet conduits may also be undesirable because they may tend to trap a bubble when being filled with a liquid medication 12, which bubble might then be carried into the radial flow regulator 32 and cause it to malfunction.

In the discussion which follows it will be assumed, for clarity and simplicity, that during operation of the radial flow regulator 32, the flexure 28's top surface 62 and the entrances of the inlet channels 38 are all exposed to a pressurized source of the medication 12 from the medication supply means. It will also be assumed, for clarity and simplicity, that the driving pressure difference (P) of the medication 12 across the radial flow

regulator 32 is the pressure difference between the medication 12 at the membrane 36's top surface 62, and the medication 12 at the outlet port 54; which is the same as the pressure difference between the medication 12 at the entrances of the inlet channels 38 and the outlet port 54. However, it is understood that during operation of the radial flow regulator 23, these pressure differences need not be equal, and the flexure 28's top surface 62 does not necessarily have to be exposed to the pressurized source of the medication 12 from the medication supply means.

When the flexure 28's top surface 62 is exposed to the medication 12, the driving pressure difference (P) across the flexure 28 may be the dominant factor in determining the amount of deflection of the flexure 28, and thus, the size of the regulator gap 48. On the other hand, if the flexure 28's top surface 62 is not exposed to the medication 12, then the velocity of the medication 12 through the regulator gap 48 may be the dominant factor in determining the amount of the deflection of the flexure 28, and thus, the size of the regulator gap 48.

During operation, as a driving pressure difference (P) is applied across the radial flow regulator 32, such as by pressurizing the source of the medication 12 with respect to the radial flow regulator 32's outlet port 54 by any suitable means, the medication 12 will pass sequentially through the radial flow regulator 32's inlet channels 38, inlet cavity 40, regulator gap 48, outlet cavity 52, and outlet port 54. The inlet cavity 40 may serve to more or less equally distribute the flow of the medication 12 from the inlet channels 38 to the entire circumference of the regulator gap 48, for more predictable operation of the radial flow regulator 32.

Referring now to Fig. 3, the regulator curves 64, 66, 68 and 70 are illustrated for a radial flow regulator 32 having one, two, three and four radial inlet channels 38, respectively. The triangular, diamond, circular and square data points on the regulator curves 64, 66, 68 and 70 are for the measured flow rate (Q) of an actual radial flow

regulator 32 having the physical parameters of the example radial flow regulator 32 which was set forth above.

The regulator curves 64, 66, 68 and 70, as well as all of the data points in Fig. 3, are plots of the flow rate (Q) of the medication 12 through the radial flow regulator 32 in microliters per day (μL/day) , as a function of the driving pressure difference (P) across the radial flow regulator 32 in millimeters of mercury (mm Hg) .

As seen in Fig. 3, at a zero driving pressure difference (P) there is no flow of the medication 12 through the radial flow regulator 32, regardless of how many inlet channels 38 there may be. Then, as the driving pressure difference (P) is increased from zero, the radial flow regulator 32 exhibits four flow regimes, again regardless of the number of inlet channels 38 which it may have.

That is, as the driving pressure difference (P) is increased from zero, there is a corresponding increase of the flow rate (Q) ; but there is also a gradual lessening of the flow rate's (Q's) sensitivity to the driving pressure difference (P) . For example, this is seen on the curve 66 from about a 0 mm Hg to about a 200 mm Hg driving pressure difference (P) .

At intermediate driving pressure differences (P) there is a "control zone" wherein the flow rate (Q) is relatively insensitive changes in the driving pressure difference (P) . For example, this is seen on the curve 66 from mm Hg to about 300 mm Hg.

Then, at driving pressure differences (P) higher than the "control zone", the flow rate (Q) actually decreases as the driving pressure difference (P) increases. For example, this is seen on the curve 66 from about 300 mm Hg to about 450 mm Hg.

Finally, at very high driving pressure differences (P) , not illustrated in Fig. 3, the flow rate (Q) may gradually decrease to near zero as the driving pressure difference (P) of the medication 12 acting on the flexure 28's top surface 46 drives the flexure 28 down against the regulator seat 42.

It has been discovered that the radial flow regulator 32 has a built-in, fail-safe characteristic, due to its structure, that may provide the user with exceptional protection against catastrophic failure of the flexure 28, when the flexure 28 is subjected to driving pressure differences (P) that are far in excess of the regulator 32's designed driving pressure difference (P) range.

This fail-safe characteristic exists because, as has been mentioned, at very high driving pressure differences (P) the medication 12 acting on the flexure 28's top surface 46 may drive the flexure 28 down against the regulator seat 42's top surface 44. When this happens, the regulator seat 42 then acts as a support for the flexure 28 and prevents its further downward deflection; which further deflection might otherwise cause the flexure 28 to crack or rupture. As a result, a much higher driving pressure difference (P) is required to rupture the flexure 28 than would otherwise be the case, since the largest unsupported span of the flexure 28 is reduced in size from the maximum overall diameter of the inlet cavity 40, to the much smaller radial width of the ring-shaped inlet cavity 40. For example, for a flexure 28 which was a membrane of silicon about 25 microns thick, a driving pressure difference of at least about 100 psi (5,171 mm Hg) would be required to crack or rupture the flexure 28. By way of comparison, as seen in Fig. 3, the regulator 32's typical operating driving pressure difference (P) may be only about 300 mm Hg. Thus, in this instance, the regulator 32 would have about a 17 times overpressure safety factor. The type of response curves 64, 66, 68, 70 shown in

Fig. 3 is highly desirable for many applications. This is because the radial flow regulator 32 will deliver a relatively constant flow rate (Q) of the medication 12 in its nominal "control zone", despite a substantial range of variations in the driving pressure difference (P) . In addition, if the nominal "control zone" driving pressure difference (P) is exceeded, then the flow rate (Q) of the medication 12 will not increase, but will actually decrease; thereby avoiding the possibility of damage which

might otherwise be caused if the flow regulator 32 permitted more than the desired amount of the medication 12 to flow.

For example, let us assume that a medication delivery device, having a source of medication 12 under pressure, was equipped with a radial flow regulator 32 in order to control the flow rate (Q) of the medication 12 from the medication delivery device. As a result, such a medication delivery device may be designed for operation in the radial flow regulator 32's above nominal "control zone" where the flow rate (Q) is relatively insensitive to changes in the driving pressure difference (P) . This may be highly desirable, since the patient will receive the medication 12 at the needed flow rate (Q) ; despite any variations in the driving pressure difference (P) , such as may be caused by the gradual emptying of the medication delivery device. In addition, if the nominal "control zone" driving pressure difference (P) were to be substantially exceeded, such as if a medical person accidentally overfilled the medication delivery device, then the medication flow rate (Q) will actually fall, thereby significantly reducing the possibility of injury or death to the patient, due to an overdose of medication 12, which might otherwise occur. The radial flow regulator 32 tends to maintain the flow rate (Q) of the medication 12 at a relatively constant value in its "control zone", despite changes in the driving pressure difference (P) , in the following way. If the driving pressure difference (P) increases, then the flexure 28 will tend to be deflected downwardly towards the regulator seat 42 an increased amount, thereby reducing the height of the regulator gap 48. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the increased driving pressure difference (P) . This is because the reduced height of the regulator gap 48 will tend to compensate for the increased driving pressure difference (P) by reducing the flow rate (Q) which would otherwise occur at that increased driving pressure difference (P) .

On the other hand, if the driving pressure difference (P) decreases, then the flexure 28 will tend to be deflected downwardly towards the regulator seat 42 a decreased amount, thereby increasing the height of the regulator gap 48. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the reduced driving pressure difference (P) . This is because the increased height of the regulator gap 48 will tend to compensate for the reduced driving pressure difference (P) by increasing the flow rate (Q) which would otherwise occur at that reduced driving pressure difference (P) .

Lastly, at driving pressure differences (P) above the regulator 32's "control zone", the flow rate (Q) is gradually reduced to near zero, as the flexure 28 is driven down closer and closer to the regulator seat 42.

Although the radial flow regulator 32 is deceptively simple in appearance, it has been discovered that it is not possible to develop simple, generally applicable design rules for its construction. This is in large part attributable to the close and nonlinear compensatory coupling between the flow rate (Q) through the radial flow regulator 32's regulator gap 48 and the flow resistance of the regulator gap 48 (R _B ) ; which, in turn, makes it difficult to separate cause and effect in a mathematical sense. In addition, certain other problems arise in developing simple, generally applicable design rules for the radial flow regulator 32 because the flow rates (Q) of the medication 12 through the regulator 32 may be so low, (as low as about 0.01 cc per day), and because the dimensions of the inlet channels 38, the inlet cavity 40, the regulator gap 48, the outlet cavity 52 and the outlet port 54 may be so small, (as small as about 0.1 microns).

As a result of such low flow rates (Q) and such small dimensions, certain fluid flow effects (such as the viscous shear forces of the medication 12 acting to deform various parts of the radial flow regulator 36) , which are normally negligible in predicting the performance of physically larger flow regulators, which handle flow rates of over about 0.1 cc per minute, for example, may become very

important. In addition, certain other fluid flow effects (such as the Equation of Continuity and Bernoulli's Equation) , which are normally important for physically larger flow regulators, handling such higher flow rates (Q) , may become negligible for a regulator 32 having such low flow rates (Q) and such small dimensions. And, at intermediate flow rates (Q) and dimensions, a combination of the pertinent small scale and large scale fluid flow effects may have to be taken into account. Because of all of the above problems, it has been discovered that two quite different strategies may be used to assist in designing a radial flow regulator 32 which has any particular desired flow regulation characteristics. The first strategy is one which is empirical in nature. That is, a series of flow regulators 32 may be built, and one feature at a time may be varied, so that the effects of changing that particular feature may be determined.

For example, the series flow resistance (R _B ) of the regulator gap 48 may be independently varied by holding constant the regulator gap 48's initial height (when the driving pressure difference (P) is equal to zero) ; while varying the width of the ring-shaped regulator seat 42, such as by varying the I.D. and O.D. of the ring-shaped regulator seat 42. Similarly, the regulator gap 48's initial height (when the driving pressure difference (P) is equal to zero) may be varied; while holding constant the width of the ring shaped regulator seat 42, such as by holding constant the I.D. and O.D. of the ring-shaped regulator seat 42.

By building and testing a large number of radial flow regulators 32; by then plotting data points for each of them for their various flow rates (Q) versus their driving pressure differences (P) ; and by then curve-fitting the plotted data points, it has been discovered that it is possible to generate an empirical model for the performance of the radial flow regulator 32 which shows the relationships between key features of the radial flow regulator 32 and the operating behavior of the radial flow

regulator 32. These empirical relationships may then be used to interpolate or extrapolate from known design cases to predict the behavior of a new radial flow regulator 32. For example, it has been discovered that the radial flow regulator 32's flow rate set point (Q _set ) (the average flow rate (Q) of the radial flow regulator 32 over its "control zone") obeys power-law relationships with respect to many of the radial flow regulator 32's design features. That is, it has been discovered that if the series flow resistance (R ₈ ) of the regulator gap 48, and the regulator gap 48's initial height (when the driving pressure difference (P) is equal to zero) , are independently varied, then the radial flow regulator 32's flow rate set point (Q _set ) may be described over a considerable range of values by an equation of the form: aG m

"" f

where (Q _set ) and (R _s ) are as have been defined above; where (G) is the regulator gap 48's initial height (when the driving pressure difference (P) is equal to zero) ; and where (a) is a constant.

For example, for a radial flow regulator 32 having a silicon flexure 28 with a thickness of about 25 microns; having channels 38 with a width of about 107 microns, with a depth of about 5.65 microns, and a length of about 2.54 mm; having a ring-shaped inlet cavity 40 with an O.D. of about 2,300 microns, and a depth of about 5.65 microns; having a ring-shaped regulator seat 42 with a width (as measured between its I.D. and O.D.) of about 750 - 2,000 microns; having an initial regulator gap 48 height (when the driving pressure difference (P) is equal to zero) in the range of about 2 - 3 microns; having an outlet cavity 52 about 5.65 microns deep; and having an outlet port 54 with a minimum diameter of about 100 microns and a length of about 494 microns, it has been

discovered that ( ) is on the order of about 2.4 and (n) is on the order of about 2/3.

The second strategy which may be used to assist in designing a radial flow regulator 32 which has any particular desired flow regulation characteristics is to develop a sophisticated physical model using numerical methods.

The starting point for formulating the model may be that, for any particular radial flow regulator 32, the flow of the medication 12 through it may be generally governed by the following equation: r>

(R _ch +R _S (Q) )

where (Q) , (P) , and (R _B ) are as has been defined above; where ( _Ch ) is the combined flow resistance across the radial inlet channels 38; where (R _ch ) is a direct function of the length (L) and the wetted perimeter (C) of each of the radial inlet channels 38; where (R _ch ) is an inverse function of the cross-sectional area (A) of each of the radial inlet channels 38; and where (R _s ) is a nonlinear function of the flow rate (Q) .

That is, the flow rate (Q) is proportional to the driving pressure difference (P) , and is inversely proportional to the sum of the two flow resistances (R _Ch ) and (R _s ) . A circuit diagram illustrating this behavior is shown in Fig. 4.

It has been found that accurate prediction of the nonlinear flow resistance (R _β ) of the regulator gap 48 may require the consideration of at least the following four factors. First, the nonlinear flow resistance (R _s ) of the regulator gap 48 may be a function of the pressure drop across the inlet channels 38, due to their flow resistance (R _ch ) . This is because the greater the pressure drop across the inlet channels 38, the greater the driving pressure difference (P) across the flexure 28, and the greater the amount of the deflection of the flexure 28 (and

vice versa) . This, in turn, generally decreases the height of the regulator gap 48; thereby generally increasing the flow resistance (R _s ) of the regulator gap 48 (and vice versa) . However, such deflection of the flexure 28 is not uniform, since the deflected flexure 28 is not flat, but instead assumes a convex, or bowed shape. This results in the nonlinear flow resistance (R _s ) of the regulator gap 48 being a relatively complex function of the flow resistance (R _ch ) of the inlet channels 38. Second, the nonlinear flow resistance (R _s ) of the regulator gap 48 may be a function of the viscous shear forces of the medication 12 acting on the flexure 28*s bottom surface 50 as the medication 12 flows radially inwardly through the regulator gap 40, from the inlet cavity 40 to the outlet cavity 52. Such viscous shear forces are, in turn, a function of such things as the viscosity and velocity of the medication 12 in the regulator gap 48. Such viscous shear forces are directed radially inwardly on the flexure 28's bottom surface 50, and tend to twist or distort the flexure 28's bottom surface 50 with respect to the flexure 28's top surface 62. Such twisting or distorting of the flexure 28 tends to vary the size and shape of the regulator gap 48 which, in turn, varies the flow resistance (R _B ) of the regulator gap 48. This results in the nonlinear flow resistance (R _s ) of the regulator gap 48 being a relatively complex function of the viscous shear forces of the medication 12 acting on the flexure 28.

Third, the nonlinear flow resistance (R _B ) may be a function of the velocity of the medication 12 passing through the regulator gap 48. Such velocity is, in turn, a function of such factors as the flow resistance (R _Ch ) of the inlet channels 38; the driving pressure difference (P) across the regulator 32; the height, size and shape of the regulator gap 48; and the flow resistance (R _s ) of the regulator gap 48.

Because of the Equation of Continuity and Bernoulli's Equation, as the velocity of the medication 12 through the regulator gap 48 increases, the pressure of the medication

12 within the regulator gap 48 tends to decrease (and vice versa) . This is because the Equation of Continuity requires that the velocity of the medication 12 must increase at a restriction. Thus, since the regulator gap 48 is a restriction (as compared to the inlet cavity 40) , the velocity of the medication 12 must increase as it flows through the regulator gap 48. Bernoulli's equation then requires that the pressure of the medication 12 in the regulator gap 48 must fall, due to its increased velocity as it flows through the regulator gap 48.

That is, as the velocity of the medication 12 in the regulator gap 48 increases, the pressure of the medication 12 in the regulator gap 48 decreases. This increases the amount of the deflection of the flexure 28; which, in turn, generally decreases the height of the regulator gap 48 and increases the flow resistance (R _B ) (and vice versa) .

Fourth, the nonlinear flow resistance (R _B ) may be a function of the flexure 28's thickness, resiliency, elasticity and stiffness. This is because for any given forces acting on the flexure 28, the amount of the deflection of the flexure 28, and the shape (or radial profile) of the deflected flexure 28, may be a function of the flexure 28*s thickness, resiliency, elasticity and stiffness. In the forgoing discussion, it was assumed that the membrane 36's inlet channel cover portions 30 were selected such that they would not be deflected a substantial amount downwardly into the inlet channels 38 by the medication 12 during the operation of the radial flow regulator 32. Thus, the foregoing discussion assumed that substantially none of the regulation of the flow of the medication 12 by the flow regulator 32 was done by any such deflection of the 36's inlet channel cover portions 30.

However, this need not be the case since, as will be made apparent by all of the disclosures in this document, regulation of the flow rate (Q) of the medication 12 through the radial flow regulator 32 may be at least partially done by such deflection of the membrane 36's inlet channel cover portions 30.

From the foregoing, it is seen that the primary difficulty in developing a sophisticated physical model for the radial flow regulator 32 which uses numerical methods is that the amount of the deflection or bowing of the flexure 28, and the shape of the flexure 28 as it deflects or bows, are closely coupled and nonlinear in interaction. As a result, iterative techniques may be used to generate a solution for the flexure 28's deflected or bowed shape (a) that is correct at every point of contact between the medication 12 and the flexure 28 in terms of the forces applied to the flexure 28; and (b) that simultaneously provides a consistent radial gradient of the driving pressure difference (P) of the medication 12 across the flexure 28 which obeys the laws of fluid flow. To do this, a free body diagram may first be developed for an arbitrary ring-shaped segment of the flexure 28. This model creates the following governing fourth-order differential equation set that identifies the relationship between the flexure 28's local curvature and its thickness; Young's modulus and Poisson's ratio; and exterior forcing functions that include the shear between the medication 12 and the flexure 28, any driving pressure difference (P) across the flexure 28, and the flexure 28's radial tension. The governing fourth-order differential equation set may be determined in the following way.

The first derivative of the deflection (θ ) of the flexure 28 towards the regulator seat 42 is given by the following second order differential equation:

Equation 1 where r is radial position with respect to the center of the flexure 28; ΔP(r) is the driving pressure difference across the flexure 28 at the radial position r; and D _t is given by:

D _t = Et-

12 ( 1- ² )

where E is Young's Modulus; t is the thickness of the flexure 28; and υ is Poisson's ratio.

The actual deflection Y(r) of the flexure 28 is obtained by integrating above Equation 1: r

Equation 2 where Y ₀ is the centerline deflection of the flexure 28. Hence the governing differential equation set is fourth order.

Next, a first-order differential equation may be developed for the flexure 28's driving pressure difference (P) pressure drop across the annular ring-shaped portion of the regulator gap 48 which is directly under the above mentioned arbitrary ring-shaped segment of the flexure 28. This differential equation must take into account not only the change in the regulator gap 48 caused by the amount of the deflection or bowing of the flexure 28, but also any change in the regulator gap 48 caused by the substrate 34's inlet channels 38 and inlet cavity 40. The first-order differential equation for the driving pressure difference ΔP(r) is:

Equation 3 where ΔP _α is the driving pressure difference (P) at the outer rim of the inlet cavity 40, referenced to the constant pressure of the medication 12 external to the flexure 28; Q is the volumetric flow rate of the medication 12; μ is the viscosity of the medication 12; and H(r) is the height of the regulator gap 48 when the driving pressure difference (P) is equal to zero.

In short, the above Equation 2 describes the amount and shape of the deflection or bowing of the flexure 28; while the above Equation 3 provides a means to calculate the driving pressure difference (P) of the medication 12 across the flexure 28 at any radial position with respect to the flexure 28. To be a physically correct representation of the interaction between the driving pressure difference (P) across the flexure 28, and the amount and shape of the deflection or bowing of the flexure 28, the solutions of the above Equations 2 and 3 must be consistent with each other in a point-to-point sense across the entire radius of the flexure 28.

To achieve this goal, the above Equation 2 may be converted to a finite difference form, and a guess may be made as to the radial profile of the driving pressure difference (P) of the medication 12 across the flexure 28. This is necessary since a solution of the bowing or deflected flexure 28's above Equation 2 requires knowledge of the driving pressure difference (P) of the medication 12 across the flexure 28 at each radial position.

This results in a so-called tridiagonal array of coupled equations that may be solved recursively for the flexure 28's radial slope at each point. Then this may be integrated once to yield the flexure 28's deflection at each radial position. Once this is known _/ the radial profile of the driving pressure difference (P) across the flexure 28 may be recalculated using these new deflection values for the flexure 28 by integration of the above Equation 3. The above iterative process may then be continued, as necessary, until the calculated position of the flexure 28 does not change by some arbitrarily set small amount per iteration, thereby signalling a consistent solution set.

Any number of different cylindrically-symmetric fluid flow devices may be modeled with the above equation set.

By changing the sign of Q, both inward and outward flow of the medication 12 through the devices may be modeled. By adjusting the function H(r) to reflect the height of the gap between the particular fluid flow device's flexure and

the corresponding part of its substrate, the above equation set may be equally useful for modeling other flow regulators (such as the flow regulators 80, 110) .

Fig. 5 shows the above mathematical model being used to predict the response of a typical design for the flow regulator 32. The solid curve 72 in Fig. 5 shows a plot of the above mathematical model for the radial flow regulator 32, in terms of flow of the medication 12 through the regulator 32 in μL/day as a function of the driving pressure difference (P) across the flexure 28.

The diamond-shaped data points 74 which are plotted in Fig. 5 are for a typical radial flow regulator 32 having a radial array of four rectangular inlet channels 38, each having a width of about 70 microns, a length of about 1270 microns, and a depth of about 5.7 microns; an inlet cavity having a maximum diameter of about 2290 microns, and a depth of about 5.7 microns; a regulator seat 42 having an I.D. of about 508 microns, and an O.D. of about 1780 microns; a flexure manufactured from a membrane of silicon having a thickness of about 25 microns; a regulator gap 48 having a height of about 2.5 microns (when the driving pressure difference (P) across the flexure 28 is zero) ; an outlet cavity having a diameter of about 508 microns, and a depth of about 5.7 microns; and an outlet port having a minimum diameter of about 100 microns, and a length of about 494 microns.

The dashed curve 76 is an empirical curve which is derived by applying curve-fitting techniques to the plotted data points 74. As seen in Fig. 5, agreement between the theoretical curve 72 and the plotted data points 74 is very good.

The curve 78 in Fig. 6 shows the above mathematical model being used to predict the reduction in the medication 12's flow rate setpoint (Q _set ) caused by sealing off the four inlet channels 38 one-by-one. The diamond-shaped data points 74 which are plotted in Fig. 6 are for a typical radial flow regulator 32 having the physical parameters which were set forth above.

As seen in Fig. 6, when one inlet channel 38 is sealed off, the effective combined flow resistance (R _ch ) of the remaining three channels 38 increases by about 33%, as compared to the combined flow resistance (R _ch ) of the original array of four inlet channels 38. Similarly, plugging two and three of the inlet channels 38 will increase the effective combined flow resistance (R _ch ) of the remaining inlet channel(s) 38 by about 100% and about 400%, respectively. As seen in Fig. 6, the mathematical model curve 78 predicts this behavior of the radial flow regulator 32 very well.

The qualitative effect of closing off one or more of the inlet channels 38 is to increase the driving pressure difference (P) across the flexure 28 which is needed for any given flow rate (Q) of the medication 12 through the radial flow regulator 32. This causes the flexure 28 to come into closer proximity to the regulator seat 42 at a lower flow rate (Q) , and hence biases the regulator 32's flow rate setpoint (Q _set ) to a value for the flow rate (Q) which is lower than would otherwise be the case.

It has been discovered that, as seen in Fig. 3, as the number of the inlet channels 38 is decreased, the "control zone" of the flow rate (Q) occurs at lower and lower flow rates (Q) for any given driving pressure difference (P) across the radial flow regulator 32. It has also been discovered that, as is also seen in Fig. 3, as the number of the inlet channels 38 is decreased, there may be a reduced sensitivity in the rate of change of the flow rate (Q) for any given change in the driving pressure difference (P) in the "control zone" of the flow rate (Q) . However, the theoretical grounds for this behavior of the regulator 32 are not clear, and it is possible that this behavior may be due to an artifact in a regulator 32 which is imperfect. All of the forgoing is very important, since it has been discovered that a single radial flow regulator 32 may actually have the properties of four different regulators 32, depending on whether none, one, two or three of its inlet channels 38 are sealed. That is, as seen in Fig. 3, such regulators 32 have quite different regulation curves

64, 66, 68, 70; have quite different "control zones" and flow rate set points (Q _set ) ; have quite different flow rates (Q) for any given driving pressure difference (P) ; and have quite difference sensitivities to changes in their flow rates (Q) for any given change in the driving pressure difference (P) .

This makes the present invention much more versatile, since a single radial flow regulator 32 may be easily modified to do the work of four single-function flow regulators. Of course, as was mentioned above, there may be fewer, or more, than four inlet channels 38; so one radial flow regulator 32 may be easily modified to do the work of fewer or more single-function flow regulators 32. From the disclosures in this document, it is possible to selectively design a radial flow regulator 32 for any particular desired flow regulation characteristics or driving pressure difference (P) . This may be done by selectively adjusting one or more of the pertinent parameters, such as: (a) the number, length, size, and cross sectional configuration of the radial inlet channels 38; (b) the number, size, cross-sectional configuration and location of the cavities 40, 52 and the outlet port 54; (c) the number, size, cross-sectional configuration and height of the regulator seat 42; (d) the number, size, cross- sectional configuration and height of the regulator gap 48; and (e) the thickness, resiliency, elasticity and stiffness of the membrane 36. For example, it has been discovered that by adjusting the fraction of the driving pressure difference (P) that is dropped across the radial inlet channels 38 in relation to the fraction of the driving pressure difference (P) which is dropped across the regulator gap 48 (by adjusting the flow resistance (R _ch ) of the inlet channels 38 and the flow resistance (R _B ) of the regulator gap 48 with respect to each other) , two things may be selectively modified. First, the degree of control of the regulator flow (Q) versus the driving pressure difference (P) may be selectively varied; and second, the amount of regulator flow (Q) versus the driving pressure difference (P) may also be selectively varied.

^*"• Q

MICROMACHINED RADIAL FLOW REGULATOR 32 (FIGS. 1 - 6):

MANUFACTURE The substrate 34 may be manufactured from any suitable strong, durable material which is compatible with the medication 12, and in which the inlet channels 38, the inlet cavity 40, the regulator seat 42, the outlet cavity 52, and the outlet port 54 may be manufactured in any suitable way, such as by using any suitable etching, molding, stamping and machining process. Such a machining process may include the use of physical tools, such as a drill; the use of electromagnetic energy, such as a laser; and the use of a water jet.

The membrane 36 may be manufactured from any suitable strong, durable, flexible, material which is compatible with the medication 12.

If the radial flow regulator 32 is intended to regulate a medication 12 which is to be supplied to a human or an animal, then any part of the regulator 32 which is exposed to the medication 12 should be manufactured from, and assembled or bonded with, non-toxic materials. Alternatively, any toxic material which is used to manufacture the regulator 32, and which is exposed to the medication 12 during use of the regulator 32, may be provided with any suitable non-toxic coating which is compatible with the medication 12.

Suitable materials for the substrate 34 and the membrane 36 may be metals (such as titanium) , glasses, ceramics, plastics, polymers (such as polyimides) , elements (such as silicon) , various chemical compounds (such as sapphire, and mica) , and various composite materials.

The substrate 34 and the membrane 36 may be assembled together in any suitable leak-proof way. Alternatively, the substrate 34 and the membrane 36 may be bonded together in any suitable leak-proof way, such as by anodically bonding them together; such as by fusing them together (as by the use of heat or ultrasonic welding) ; and such as by using any suitable bonding materials, such as adhesive, glue, epoxy, solvents, glass solder, and metal solder.

Anodically bonding the substrate 34 and the membrane 36 together may be preferable for at least four reasons. First, anodic bonding is relatively, quick, easy and inexpensive. Second, an anodic bond provides a stable leak-proof bond.

Third, since an anodic bond is an interfacial effect, there is no build-up of material at the bond; and the bond has essentially a zero thickness, which desirably creates no essentially no spacing between the substrate 34 and the membrane 36. As a result, an anodic bond does not interfere with the desired height of the regulator gap 32.

Fourth, an anodic bond may be preferable since it eliminates the need for any separate bonding materials, which might otherwise clog or reduce the size of the inlet channels 38, the inlet cavity 40, the regulator seat 42, the regulator gap 48, the outlet cavity 52, and the outlet port 54; or which might lead to corrosion of the joint between the regulator 32's substrate 34 and membrane 36.

One example of how the radial flow regulator 32 may be manufactured will now be given. The starting point may be a 76.2 mm diameter wafer of Corning 7740 Pyrex glass, which will form the regulator 32's substrate 34.

The glass wafer may be cleaned in any suitable way, such as by immersing it in a buffered hydrofluoric acid (BHF) etchant for two minutes, rinsing it with distilled water, and drying it.

A thin chrome metallization layer may then applied to the top surface of the glass wafer by any suitable means, such as with an electron beam evaporator. The chrome layer may provide a good adhesion surface for the subsequent application of photosensitive resist (photoresist) to the glass wafer's top surface.

Following this, a thin layer of any suitable photoresist may be applied on top of the chrome layer, such as Microposit 1650 photoresist made by the Shipley Company, located in Newton, Massachusetts. The layer of photoresist may be dried in any suitable way, such as by baking it at about 90°C for about 25 minutes.

An image of the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52 may then be exposed onto the photoresist in any suitable manner, such as by using a first mask and a mask aligner. This image may be developed (that is, the exposed photoresist may be removed) , by using any suitable photoresist developer, such as 351 developer, made by the above Shipley Company. The glass wafer may then be then rinsed in distilled water and dried. As a result of the forgoing procedure, the chrome layer will now bear an image, unprotected by the photoresist, of the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52. The unprotected portions of the chrome layer may then be removed by using any suitable chrome etch solution, such as Cyantek CR-7, made by the Cyantek Company, located in Fremont, California.

The forgoing procedure will result in an image of the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52 having been formed on the top surface of the glass wafer, which image is unprotected by the layers of photoresist and chrome which cover the rest of the glass wafer's top surface. The image may then be etched into the glass wafer's top surface to any desired depth by any suitable means, such as by immersing the glass wafer's top surface in BHF etchant; rinsing the glass wafer in distilled water, and drying it. A suitable depth may be about 6.0 microns.

Next, an image of the regulator seat 42 may then be exposed onto the photoresist on the glass wafer's top surface using a second mask. The newly exposed photoresist may then be developed; and the newly exposed portions of the chrome removed. Then the image of the regulator seat may be etched into the top surface of the glass wafer to any desired depth in any suitable manner, in order to define an elevation difference between the regulator seat 42's top surface 44 and the top surface of the glass wafer. A suitable elevation difference may be about 2.5 microns.

At this point, it may be noted that the depths of the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52 will also have been automatically increased by about 2.5 microns, since they are also unprotected by the layers of chrome and photoresist. In other words, the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52 may intentionally be initially etched to a depth less than their desired final depth, in order to permit them to automatically and simultaneously reach their desired final depth while the regulator seat 42 was being etched.

Note that the above procedure is unusually economical and quick, since if the four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52 were originally etched to their desired final depth, then the additional steps of re-coating the entire glass wafer with photoresist (in order to protect the etched four radial inlet channels 38, the inlet cavity 40, and the outlet cavity 52), and then baking the photoresist, would have to be done prior to the exposing, developing and etching of the regulator seat 42.

After the regulator seat 42 has been etched, the regulator 32's outlet port 54 may be formed by any suitable means, such as by drilling it with a focused beam from a 25 C0 ₂ laser, with a physical drill, or with a water jet drill. It has been discovered that when using a laser to form the outlet port 54, heating the glass wafer to near its anneal point improves the quality of the outlet port 54, and also reduces undesirable cracking of the glass wafer adjacent to the outlet port 54.

Preferably, as seen in Fig. 2, the outlet port 54 may have a venturi-like shape, rather than being cylindrical in shape, for better fluid flow through it. It has also discovered that the outlet port 54 may be given its preferred venturi-like shape by drilling the outlet port 54 with a laser in the manner discussed above. The desired venturi-like shape may be automatically formed during the laser drilling process, and apparently results from the thermal effects of the laser beam interacting with the

87 , -9

glass wafer as the outlet port 54 is being drilled through it with the laser beam. After the outlet port 54 has been drilled, the glass wafer may be lightly etched with BHF etchant, in order to remove any volatilized glass which ma have condensed on the glass wafer adjacent to its outlet port 54.

After the outlet port 54 has been formed, a nominal layer of one or more corrosion-resistant substances may be deposited onto the top surface of the glass wafer by any suitable means, such as by sputtering using an e-beam evaporator. As a result, the four radial inlet channels 38, the inlet cavity 40, the regulator seat 42, the outlet cavity 52, and the outlet port 54 will have been coated with a layer of the corrosion-resistant substance(s) . Suitable corrosion-resistant substances may be silicon, or may be metals, such as gold, platinum, chrome, titanium and zirconium, or may be the oxides of silicon or such metals. Such oxides may be formed by thermally oxidizing the corrosion-resistant substance(s) in air afte it has been applied to the substrate 34. However, other suitable corrosion-resistant substances may be used, depending on the particular medication 12 with which the radial flow regulator 32 is designed to be used. The oxides of metals such as titanium and zirconium are well- known to be stable against water solutions over a wide pH range. The thickness of the layer of the corrosion- resistant substance(s) may be from 200 A - 1000 A; but the thickness may depend on the particular corrosion-resistant substance(s) being used, and on the particular medication 12 with which the radial flow regulator 32 is designed to be used.

Alternatively, the layer of corrosion-resistant substance(S) may comprise a donut-shaped disk of such corrosion-resistant substance(s) , such as silicon, which may be bonded to the regulator seat 42's top surface 44 by any suitable means, such as by using any of the means whic have been mentioned for bonding the substrate 34 and the membrane 36 together.

Such a corrosion-resistant donut-shaped disk may be formed in any suitable way, such as by using a masking and etching process which is similar to that described above regarding the substrate 34. The starting point may be a clean epitaxial-coated silicon wafer, to which is applied a thin chrome metallization layer and a layer of photoresist. After the photoresist is dry, an image of the donut-shaped disk may then be exposed onto the photoresist. The exposed photoresist, and the underlying portions of the chrome layer, may then be removed, resulting in an image of the donut-shaped disk on the surface of the silicon wafer, which image is not protected by the photoresist or by the chrome layer. The exposed portions of the silicon wafer may then etched to a depth in excess of the desired thickness of the desired donut-shaped disk in any suitable way, such as by the use of an isotropic silicon etchant. For example, if a donut-shaped disk having a thickness of about 1 micron was desired, then the exposed portions of the silicon wafer may be etched to a depth of about 5 microns.

The silicon wafer may then be cleaned; the etched faces of the silicon and glass wafers may then be aligned with each other, so that the donut-shaped disk on the silicon wafer is aligned with the regulator seat 42 on the glass wafer; and the silicon and glass wafers may then be bonded together in any suitable way, such as by using an anodic bonding process like that which will be described below regarding anodically bonding together the silicon and glass wafers that will form the regulator 32's substrate 34 and membrane 36. Next, the silicon wafer with the donut- shaped disk may then etched again in any suitable way, such as by the use of an anisotropic ethylene diamine etchant, until the desired ultimate thickness of the donut-shaped disk of silicon is obtained. The manufacture of only one substrate 34 and only one donut-shaped disk of corrosion-resistant material for the substrate 34's regulator seat 42 was described above. However, it will be appreciated that on any pair of glass and silicon wafers respective arrays of substrates 34 and

corresponding donut-shaped disks of corrosion-resistant material may be manufactured simultaneously in a manner which is similar to that described above. If such is the case, the array of substrates 34 on the glass wafer may be aligned with, and then bonded to, the corresponding array of donut-shaped disks of corrosion-resistant material on the silicon wafer. After the final etching of the silicon wafer, the manufacture of each regulator 32 may then be completed in the manner which is set forth below. If a donut-shaped disk or layer of corrosion-resistant substance(s) is bonded or applied to the regulator seat 42, then the regulator seat 42 may have to be etched an additional amount during its above etching step prior to applying the donut-shaped disk or layer of corrosion- resistant substance(s) to the regulator seat 42's top surface 44. The additional amount of etching may be equal to the thickness of the donut-shaped disk or layer, in order to end up with the desired elevation difference between the regulator seat 42's top surface 44 and the top surface of the glass wafer (which will form the substrate 34) .

It has been discovered that the donut-shaped disk or layer of corrosion-resistant substance(s) on the regulator seat 42's top surface 44 may serve an unexpected further function in addition to its corrosion-resistant function.

That is, it may also prevent the regulator 32*s membrane 36 from being inadvertently bonded to the regulator seat 42's top surface 44 when the membrane 36 is being bonded to the glass wafer, such as when the membrane 36 is being anodically bonded in the manner which will be described below.

After the regulator seat 42 has been etched, and after any desired layer or disk of corrosion-resistant substance(s) has been applied to etched portions of the glass wafer, the photoresist and chrome which remain on the unetched portions of the glass wafer may be removed by any suitable means, such as by using standard lift-off techniques.

Fabricating the membrane 36 and mounting it to the glass wafer (which is the substrate 34) , may be done in any suitable way. One suitable way is to start with a prime silicon wafer having a boron-doped epitaxial silicon layer which has been deposited onto its top surface. Since the boron doped epitaxial silicon layer will ultimately form the regulator 32's membrane 36, the layer's thickness will depend on the desired thickness of the membrane 36. The boron-doped epitaxial silicon layer, and thus the membrane 36, may be from 1 - 50 microns thick, for example. The boron doping may be in excess of 3 X IO ¹⁹ atoms of boron per cubic centimeter, which conveys a dramatic etch- resistance to the epitaxial silicon layer in silicon etchants based on ethylene diamine. The glass and silicon wafers may then be cleaned; dried; and anodically bonded together. The anodic bonding may be performed in any suitable way, such as by placing the respective top surfaces of the glass and silicon wafers in contact with each other in a vacuum chamber in an oven maintained at a temperature of about 500°C. A DC voltage of approximately 1000 volts may then be applied to the two wafers for a period of about 15 minutes, with the silicon wafer at a positive potential relative to the bottom surface of the glass wafer. An exponentially decaying current will flow through the wafers over this time period, at the end of which the two wafers will have been anodically bonded together, i.e., they will have been hermetically bonded to each other to form a silicon/glass sandwich. It has also been discovered that the anodic bonding process may also ensure that any corrosion- resistant substance(s) which were applied to the etched surfaces of the glass wafer are firmly attached to it.

Upon cool-down, after the anodic bonding process is complete, the silicon wafer portion of the anodically bonded silicon/glass sandwich may be ground down; but preferably, the silicon wafer is not ground down so much that any of its boron doped epitaxial silicon layer is removed. For example, if the boron doped epitaxial silicon layer is from about 1 - 50 microns in thickness, then the

silicon wafer portion of the bonded silicon/glass wafer sandwich may be ground down to about 125 microns in thickness. The remaining non-doped silicon in the silicon wafer may then be removed in any suitable way, such as by placing the ground down silicon/glass sandwich in an ethylene diamine etchant maintained at 112°C for 3.5 to 4.0 hours. A suitable ethylene diamine etchant may be PSE-300, manufactured by the Transene Corp. located in Rowley, Massachusetts. At the end of this time, the boron doped epitaxial silicon layer will be exposed in the form of a continuous, flat membrane 36 which is anodically bonded to the glass wafer (the substrate 34) , thereby forming the completed radial flow regulator 32, which is then cleaned and dried. The purpose of the above grinding step is simply to reduce the amount of silicon which needs to be etched away. Accordingly, as alternatives, the grinding step may be eliminated, with all of the undesired silicon being etched away; or a thinner silicon wafer may be used, so that there is less undesired silicon to begin with.

The manufacture of only one radial flow regulator 32 was described above. However, it will be appreciated that on any pair of glass and silicon wafers numerous regulators 32 could be manufactured simultaneously in a manner similar to that described above. If such is the case, an array of substrates 34 may be simultaneously etched in the glass wafer before the silicon and glass wafers are aligned and bonded together. Then, all of the membranes 36 may be formed simultaneously by grinding and etching the silicon wafer to its desired final thickness. The silicon/glass sandwich may then be divided by any suitable means (such as dicing) into individual chips, each chip bearing at least one radial flow regulator 32.

One of the advantages of using the etching and anodic bonding process which was described in detail above is that such a process enables high quality, very reliable radial flow regulators 32 to be mass produced in great numbers at a cost so low that the regulator 32 may be considered to be disposable. In addition, it should also be noted that the

regulator 32 is stunning in its simplicity since it may have as few as only two parts (the substrate 34 and the membrane 36) ; and since it may have only one moving part (the membrane 36's flexure 28). Further, because the raw materials from which the regulator 32 may be manufactured may be very inexpensive, such as glass and silicon, the cost of the regulator 32 may held to a very low level. MICROMACHINED LINEAR FLOW REGULATOR 80 HAVING A CONTOURED REGULATOR SEAT 90 (FIGS. 7 - 12): STRUCTURE

Turning now to Figs. 7 - 8, the micromachined linear flow regulator 80 of the present invention is illustrated. The linear flow regulator 80 may be used to control the flow rate of a fluid medication 12 passing through it, and may comprise a substrate 82, and a membrane 84. The substrate 82 may include a straight, elongated channel 86 having a contoured regulator seat 90 and an outlet port 92. The membrane 84 may have a mounting portion 97, which is mounted to a corresponding portion of the substrate 82's top surface 96; an inlet port 94; and a flexure 98 which overlies the channel 86. A regulator gap 99 lies between the flexure 98 and the regulator seat 90.

Although only one outlet port 92 in the substrate 82 is illustrated, there may be more outlet ports 92. Although, as seen, the outlet port 92 is preferably located near one end of the channel 86, so that during operation of the regulator 80 the downwardly deflecting flexure 98 will not seal off the outlet port 92, each outlet port 92 may be positioned in any other suitable location in the channel 86. Although preferably the outlet port 92 may have a venturi-like shape, for better flow of the medication 12 through it, the outlet port 92 may have any other suitable shape. Although the outlet port 92 is illustrated as being located in the substrate 82, the outlet port 92 may be wholly or partially located in the membrane 84.

Although the channel 86 and the regulator seat 90 which are illustrated in Figs. 7 - 8 follow a straight course, the channel 86 and the regulator seat 90 may follow a circular (Fig. 9) , spiral (Fig. 10) , serpentine (Fig.

11) , or other non-straight course. The use of a regulator 80 having a channel 86 and a regulator seat 90 which follow a circular, spiral, serpentine, or other non-straight course may be desirable. This is because, for any given length of channel 86 and regulator seat 90, such courses may permit the manufacture of a linear regulator 80 which is more compact, as compared to a linear regulator 80 having a straight channel 86 and regulator seat 90.

Preferably, the contour of the regulator seat 90 may approximate, or duplicate, the contoured shape that the flexure 98 would assume if the flexure 98 were entirely unrestrained by any part of the substrate 82 when the linear flow regulator 80 is subjected to the regulator 80's maximum designed driving pressure difference (P) of the medication 12 between the regulator 80's inlet port 88 and outlet port 92.

Although the membrane 84 is illustrated as being of uniform thickness, and as having flat top and bottom surfaces 100, 101, the membrane 84 may not be of uniform thickness, and may have top and bottom surfaces 100, 101 which are not flat.

Although the membrane 84 shown has one rectangular inlet port 88, there may be more than one inlet port 88, and each inlet port 88 may have any other suitable shape. Although the inlet port 88 is illustrated as being in the membrane 84, the inlet port 88 may be wholly or partially located in the substrate 82.

By way of example, the linear flow regulator 80's substrate 82 may be manufactured from 7740 Pyrex glass, and may have a thickness of about 0.5 mm. The channel 86 and the regulator seat 90 may each have a length of about 10 mm, and a maximum width of about 480 microns. The regulator gap 99 may have a maximum height of about 6.65 microns (when the driving pressure difference (P) is equal to zero) . The outlet port 92 may have a minimum diameter of about 100 microns, and may have a length of about 496 microns. The membrane 84 may be manufactured from silicon, and may have a thickness of about 4.0 microns. The inlet port 88 may have a width of about 480 microns, and a length

of about 500 microns. The flow characteristics of this example linear flow regulator 80 are illustrated in Fig. 12, which will be discussed below.

Turning now to Figs. 9 - 11, the linear flow regulators 80 illustrated therein are the same as, or at similar to, the linear flow regulator 80 of Figs. 7 - 8, in their structure, operation, theory, and manufacture, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document. Accordingly, the respective parts of the linear flow regulators 80 of Figs. 9 - 11 have been given the same reference numerals as the corresponding parts of the linear flow regulator 80 of Figs. 7 - 8, for clarity and simplicity. One of such differences is that the linear flow regulators 80 illustrated in Figs. 9 and 11 have a pair of outlet ports 92, instead of the single outlet port 92 of the regulator 80 of Figs. 7 - 8. Other of such differences are that the channel 86 and the regulator seat 90 of the regulator 80 illustrated in Figs. 7 - 8 follow a straight course, while the channels 86 and the regulator seats 90 of the regulators 80 of Figs. 9 - 11 follow a circular, spiral, and serpentine course, respectively.

As will be appreciated from all of the disclosures in this document, the fact that the linear flow regulators 80 may, as in the example set forth above, have an extremely small size, be extremely light weight, have only two parts, and have a zero electrical energy consumption, offer numerous advantages over a regulator 80 which was physically much larger, much heavier, more complex, or which consumed electrical energy. For example, the regulators 80 may be ideal for use as part of a miniaturized medication delivery device which is to be implanted in a human or animal for delivery of constant flows of the medication 12 at flow rates as low as about

0.01 cc/day — flow rates which are so low that they may be impossible for a physically larger flow regulator of a different design to reliably and accurately deliver.

MICROMACHINED LINEAR FLOW REGULATOR 80 HAVING A CONTOURED REGULATOR SEAT 90 (FIGS. 7 - 12):

OPERATION AND THEORY The linear flow regulator 80 may be installed in its intended location of use in any suitable way. Any suitable medication supply means may be used to connect the regulator 80's inlet port 88 to a source of the medication 12; and any suitable medication delivery means may be used to connect the regulator 80's outlet port 92 to whatever person, animal or thing is to receive the medication 12 from the outlet port 92. In some cases, the medication supply means may also be used to supply the medication 12 to the flexure 98' top surface 100, at a pressure which may or may not be the same as the pressure at which the medication 12 is supplied to the inlet port 88.

For example, the regulator 80 may be installed within any type of reservoir means for the medication 12 by any suitable means, such as by locating the regulator 80's outlet port 92 over the reservoir means's outlet, and by using an adhesive face seal between the regulator 80's bottom surface 104 and the inside of the reservoir means to hold the regulator 80 in place. As a result, when the reservoir means is filled with the medication 12, the regulator 80 will be immersed in the medication 12, with its inlet port 88 and its flexure 98's top surface 100 in fluid communication with the medication 12 within the reservoir means, and with its outlet port 92 in fluid communication with the reservoir means' outlet. Such an installation for the regulator 80 has numerous advantages.

For example, it is quick, easy, reliable and inexpensive, because no additional medication supply means (such as supply conduits) are needed to supply the medication 12 to the regulator 80's inlet port 88 and to the flexure 98's top surface 100 (since they are already immersed in the medication 12) ; and because no additional medication delivery means (such as delivery conduits) are needed to convey the medication 12 away from regulator 80's outlet port 92 (since the reservoir means' outlet is used

for this purpose) . Such additional inlet and outlet conduits may be undesirable since it may be relatively time consuming, difficult and expensive to align and connect them to regulator 80, due to the extremely small size of its inlet port 88, flexure 98, and outlet port 92. Such additional inlet conduits may also be undesirable because they may tend to trap a bubble when being filled with a liquid medication 12, which bubble might then be carried into the regulator 80 and cause it to malfunction. In the discussion which follows it will be assumed, for clarity and simplicity, that during operation of the regulator 80, the flexure 98's top surface 100 and the inlet port 88 are both exposed to a pressurized source of the medication 12 from the medication supply means. It will also be assumed, for clarity and simplicity, that the driving pressure difference (P) of the medication 12 across the regulator 80 is the pressure difference between the medication 12 at the flexure 98's top surface 100, and the medication 12 at the outlet port 92; which is the same as the pressure difference between the medication 12 at the entrance of the inlet port 88 and the outlet port 92. However, it is understood that during operation of the regulator 80, these pressure differences need not be equal, and the flexure 98*s top surface 100 does not necessarily have to be exposed to the pressurized source of the medication 12 from the medication supply means.

During operation, as a driving pressure difference (P) is applied across the regulator 80, such as by pressurizing the source of the medication 12 with respect to the regulator 80's outlet port 92 by any suitable means, the medication 12 will pass sequentially through the regulator 80's inlet port 88, regulator gap 99, and outlet port 92. Referring now to Fig. 12, the regulator curve 106 is illustrated for the regulator 80. The regulator curve 106 is a plot of the flow rate (Q) of the medication 12 through a regulator 80 having the physical parameters of the example regulator 80 which was set forth above. In Fig. 12, the flow rate (Q) is plotted in terms of microliters per day (μL/day) , as a function of the driving pressure

difference (P) across the regulator 80 in mmHg (millimeters of mercury) .

As seen in Fig. 12, at a zero driving pressure difference (P) , there is no flow of the medication 12 through the regulator 80. Then, as the driving pressure difference (P) is increased from zero, the regulator 80 exhibits four flow regimes.

That is, as the driving pressure difference (P) is increased from zero, there is a corresponding increase of the flow rate (Q) ; but there is also a gradual lessening of the flow rate's (Q's) sensitivity to the driving pressure difference (P) . For example, this is seen on the curve 106 at driving pressure differences (P) from about 0.0 mmHg to about 160 mmHg. At intermediate driving pressure differences (P) there is a "control zone" wherein the flow rate (Q) is relatively insensitive changes in the driving pressure difference (P) . For example, this is seen on the curve 106 at driving pressure differences (P) from about 160 mmHg to about 250 mmHg.

Then, although not illustrated in Fig. 12, at driving pressure differences (P) higher than the "control zone", the flow rate (Q) may actually decrease as the driving pressure difference (P) increases. Finally, at very high driving pressure differences (P) , the flow rate (Q) may gradually decrease to near zero as the driving pressure difference (P) of the medication 12 acting on the flexure 98's top surface 100 drives the flexure 98 down against the regulator seat 90. It has been discovered that the regulator 80 has a built-in, fail-safe characteristic, due to its structure, that may provide the user with exceptional protection against catastrophic failure of the flexure 98, when the flexure 98 is subjected to driving pressure differences (P) that are far in excess of the regulator 80's designed driving pressure difference (P) range.

This fail-safe characteristic exists because, as has been mentioned, at very high driving pressure differences (P) the medication 12 acting on the flexure 98's top

surface 100 may drive the flexure 98 down against the regulator seat 90. When this happens, the regulator seat 90 then acts as a support for the flexure 98 and prevents its further downward deflection; which further deflection might otherwise cause the flexure 98 to crack or rupture. As a result, a much higher driving pressure difference (P) is required to rupture the flexure 98 than would otherwise be the case.

The type of response curve 106 shown in Fig. 12 is highly desirable for many applications. This is because the regulator 80 will delivery a relatively constant flow rate (Q) of the medication 12 in its nominal "control zone", despite a substantial range of variations in the driving pressure difference (P) . In addition, if the nominal "control zone" driving pressure difference (P) is exceeded, then the flow rate (Q) of the medication 12 will not increase, but may actually decrease; thereby avoiding the possibility of damage which might otherwise be caused if the flow regulator 32 permitted more than the desired amount of the medication 12 to flow.

For example, let us assume that a medication delivery device, having a source of medication 12 under pressure, was equipped with a linear flow regulator 80 in order to control the flow rate (Q) of the medication 12 from the medication delivery device. As a result, such a medication delivery device may be designed for operation in the regulator 80's above nominal "control zone" where the flow rate (Q) is relatively insensitive to changes in the driving pressure difference (P) . This may be highly desirable, since the patient will receive the medication 12 at the needed flow rate (Q) ; despite any variations in the driving pressure difference (P) , such as may be caused by the gradual emptying of the medication delivery device. In addition, if the nominal "control zone" driving pressure difference (P) were to be substantially exceeded, such as if a medical person accidentally overfilled the medication delivery device, then the medication flow rate (Q) will actually fall, thereby significantly reducing the

possibility of injury or death to the patient, due to an overdose of medication 12, which might otherwise occur.

The regulator 80 tends to maintain the flow rate (Q) of the medication 12 at a relatively constant value in its above "control zone", despite changes in the driving pressure difference (P) , in the following way. If the driving pressure difference (P) increases, then the flexure 98 will tend to be deflected towards the regulator seat 90 an increased amount, thereby reducing the height of the regulator gap 99. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the increased driving pressure difference (P) . This is because the reduced height of the regulator gap 99 will tend to compensate for the increased driving pressure difference (P) by reducing the flow rate (Q) which would otherwise occur at that increased driving pressure difference (P) .

On the other hand, if the driving pressure difference (P) decreases, then the flexure 98 will tend to be deflected towards the regulator seat 90 a decreased amount, thereby increasing the height of the regulator gap 99. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the reduced driving pressure difference (P) . This is because the increased height of the regulator gap 99 will tend to compensate for the reduced driving pressure difference (P) by increasing the flow rate (Q) which would otherwise occur at that reduced driving pressure difference (P) .

Lastly, at driving pressure differences (P) above the regulator 80's "control zone", the flow rate (Q) is gradually reduced to zero, as the flexure 98 is driven down closer and closer to the regulator seat 90.

It has been discovered that two quite different strategies may be used to assist in designing a regulator 80 which has any particular desired flow regulation characteristics.

The first strategy is one which is empirical in nature. That is, a series of regulators 80 may be built, and one feature at a time may be varied, so that the

effects of changing that particular feature may be determined.

For example, the series flow resistance (R _B ) of the channel 86 and the regulator gap 99 may be independently varied by holding constant the regulator gap 99's initial height (when the driving pressure difference (P) is equal to zero) ; while varying the length of the channel 86 and the regulator gap 99. Similarly, the regulator gap 99's initial height (when the driving pressure difference (P) is equal to zero) may be varied by varying the depth and shape of the channel 86; while holding constant the length of the channel 86 and the regulator gap 99.

By building and testing a large number of regulators 80; by then plotting data points for each of them for their various flow rates (Q) versus their driving pressure differences (P) ; and by then curve-fitting the plotted data points, it may be possible to generate an empirical model for the performance of the regulator 80 which shows the relationships between key features of the regulator 80 and the operating behavior of the regulator 80. These empirical relationships may then be used to interpolate or extrapolate from known design cases to predict the behavior of a new regulator 80.

The second strategy which may be used to assist in designing a regulator 80 which has any particular desired flow regulation characteristics is to develop a mathematical model using numerical methods.

The starting point for formulating the model is that, for any particular regulator 80, the flow of the medication 12 through it may be generally governed by the following equation:

Q=

R _S (L,A, C, Q)

where (Q) , (P) , and (R _B ) are as has been defined above; where (R _B ) is a direct function of the length (L) and the wetted perimeter (C) of the channel 86 and the regulator gap 99; where (R _s ) is an

inverse function of the cross-sectional area (A) of the channel 86 and the regulator gap 99; and where (R _B ) is a nonlinear function of the flow rate (Q) .

That is, the flow rate (Q) is proportional to the driving pressure difference (P) , and is inversely proportional to the flow resistance (R _s ) of the channel 86 and the regulator gap 99.

It has been found that accurate prediction of the nonlinear flow resistance (R _B ) of the channel 86 and the regulator gap 99 may require the consideration of at least the following four factors. First, the nonlinear flow resistance (R _B ) of the channel 86 and the regulator gap 99 may be a function of the pressure drop across the length of the channel 86 and the regulator gap 99, due to their flow resistance (R _B ) . This is because the greater the pressure drop across the length of the channel 86 and the regulator gap 99, the greater the driving pressure difference (P) across the flexure 98, and the greater the amount of the deflection of the flexure 98 towards the regulator seat 90 (and vice versa) . This, in turn, generally decreases the height of the regulator gap 98; thereby generally increasing the flow resistance (R _B ) of the channel 86 and the regulator gap 99 (and vice versa) . However, such deflection of the flexure 98 is not uniform, since the deflected flexure 98 is not flat, but instead assumes a convex, or bowed shape. This results in the nonlinear flow resistance (R _B ) of the channel 86 and the regulator gap 48 being a relatively complex function of the pressure drop across the length of the channel 86 and the regulator gap 99.

Second, the nonlinear flow resistance (R _B ) of the regulator gap 99 may be a function of the viscous shear forces of the medication 12 acting on the flexure 98's bottom surface 101 as the medication 12 flows through the length of the regulator gap 99, from the inlet port 88 to the outlet port 92. Such viscous shear forces are, in turn, a function of such things as the viscosity and velocity of the medication 12 in the regulator gap 99. Such viscous shear forces are directed along the length of

the flexure 98's bottom surface 101 and tend to twist or distort the flexure 98's bottom surface 101 with respect to the flexure 98's top surface 100. Such twisting or distorting of the flexure 98 tends to vary the size and shape of the regulator gap 99 which, in turn, varies the flow resistance (R _B ) of the regulator gap 99. This results in the nonlinear flow resistance (R _B ) of the regulator gap 99 being a relatively complex function of the viscous shear forces of the medication 12 acting on the flexure 98. Third, the nonlinear flow resistance (R _B ) may be a function of the velocity of the medication 12 passing through the regulator gap 99. Such velocity is, in turn, a function of such factors as the driving pressure difference (P) across the regulator 80; the height, size, shape and length of the regulator gap 99; the flow resistance (R _β ) of the channel 86 and the regulator gap 99; and the size, shape, length and location of the inlet and outlet ports 88, 92.

Because of the Equation of Continuity and Bernoulli's Equation, as the velocity of the medication 12 through the regulator gap 99 increases, the pressure of the medication 12 within the regulator gap 99 tends to decrease (and vice versa) . This is because the Equation of Continuity requires that the velocity of the medication 12 must increase at a restriction. Thus, since the regulator gap

99 is a restriction (as compared to the inlet port 94) , the velocity of the medication 12 must increase as it flows through the regulator gap 99. Bernoulli's equation then requires that the pressure of the medication 12 in the regulator gap 99 must fall, due to its increased velocity as it flows through the regulator gap 99.

That is, as the velocity of the medication 12 in the regulator gap 99 increases, the pressure of the medication 12 in the regulator gap 99 decreases. This increases the amount of the deflection of the flexure 98; which, in turn, generally decreases the height of the regulator gap 98 and increases the flow resistance (R _B ) of the channel 86 and the regulator gap 99 (and vice versa) .

Fourth, the nonlinear flow resistance (R _B ) may be a function of the flexure 98's thickness, resiliency, elasticity and stiffness. This is because for any given forces acting on the flexure 98, the amount of the deflection of the flexure 98, and the shape (or radial profile) of the deflected flexure 98, may be a function of the flexure 98's thickness, resiliency, elasticity and stiffness.

From the forgoing, it is seen that the primary difficulty in developing a mathematical model for the regulator 80 which uses numerical methods is that the amount of the deflection or bowing of the flexure 98, and the shape of the flexure 98 as it deflects or bows, are closely coupled and potentially nonlinear in interaction. However, in developing the mathematical model for the regulator 80, let it now be assumed that the regulator seat 90 has an x,y,z coordinate system in which the z axis lies along the longitudinal centerline of the regulator seat 90; in which the x axis is transverse to the z axis, extends left and right from the z axis, and equals zero at the z axis; and in which the y axis is transverse to the x and z axes, measures the height above the regulator seat 90, and equals zero at the z axis.

In such a coordinate system, the most fundamental flow element consists of a local fluid slice (dx) wide which spans the gap from the regulator seat 90 to the flexure 98. If the bowing or deflection of the flexure 98 is very slight, then it may be assumed that this local fluid slice has negligible shear along its sides and is dominated by viscous drag at its top and bottom. This is a local fluid slice flow approximation. With this assumption, it may be f ound that the local fluid slice contribution to flow (dQ) is given by:

Equation 4

where μ is the viscosity of the medication 12; and Y _m (x) is the height of the local fluid slice.

The bowing or deflection of the flexure 98 towards the regulator seat 90 may be dictated by the flexure 98's stiffness. For a pressure difference (P _B - p(z)) across the flexure 98, the amount (Y) of the bowing or deflection of the flexure 98 is given by:

Y=Y ₀ (1+X ⁴ -2X ² )

Equation 5 where

_v _ (p _s -p) ^w *

2Et ³

Equation 5A and where (E) is Young's Modulus; (t) is the thickness of the flexure 98; (w) is the half-width of the regulator seat 90; and (X) = x/w. This assumes that the regulator seat 90 has fixed edges, and that there is a guided condition at the regulator seat 90's center, or at the first point of contact of the flexure 98 with the regulator seat 90. A guided condition means that dY/dx = 0 at that point.

If we now solve for the total flow across the local fluid slice cross-section by integrating the above Equation 4, and by incorporating the above Equation 5 to provide the height of the local fluid slice, this is found to be:

_Q= Et 3 h4 ^ dP x P _C1 _ _P . _(i+ χ4_ _2χ 2 _j -x 3 _dχ μw ³

Equation 6

and where (h) is equal to the height of the regulator gap 99. The dimensionless pressure (P) is a variable that is zero when there is no differential pressure, and is 1.0

when the bowing or deflection of the flexure 98 is equal to the height of the regulator gap 99 when there is a zero driving pressure difference (P) across the flexure 98. The integral's value is 0.2781 if P = l. Otherwise, a simple polynomial in P is obtained. It is clear that the total flow (Q) of the medication 12 is a constant throughout the length of the regulator gap 99. This means that for P ■< 1, variables can be separated in the above Equation 6 so that the left side is (z) , (Q • dz) , and the right side is a polynomial function of P. This can then be integrated to yield a relationship between the flow rate-channel length product, Q • Z ₀ , and pressure at a given axial position:

Equation 7 where (Z ₀ ) is the length of the regulator seat 90. This yields a relation between pressure and total flow up to the point where P = 1. However at this point, since the regulator seat 90 is assumed to be a perfect replica of the bowed or deflected flexure 98, at a pressure that causes the flexure 98 to touch the center of the regulator seat 90, the flexure 98 is also in contact with the entire regulator seat 90, from side to side. Accordingly, the flow rate-channel length product Q • Z ₀ is:

The above model may now be used to predict the performance of the linear flow regulator 80.

Turning again to Figs, 7 and 8, the flexure 98's inlet port edge 94 may overlie the channel 86, as seen therein, and thus is not affixed to or restrained by any portion of the substrate 82. As a result, all of the flexure 98 which is located adjacent to the inlet port 88 (including the inlet port edge 94) , is free to fully flex in response to changes in the flow rate (Q) of the medication 12 through

the channel 86; as compared to if the flexure 98's inlet port edge 94 were affixed to or restrained by any portion of the substrate 82. This may be desirable for at least two reasons. First, it may result in smoother, more predictable, regulation of the flow rate (Q) of the medication 12 by the flexure 98 over the regulator 80's designed flow and regulation parameters.

Second, it may also make possible a more compact regulator 80, since the part of the flexure 98 which is located adjacent to the inlet port 88 is not rendered wholly or partially inoperative by any portion of the substrate 82.

Alternatively, the substrate 82 may be manufactured so that part or all of the flexure 98's inlet port edge 94 may be affixed to or restrained by some part of the substrate 82. In order to compensate for such a structure, the channel 86 and the flexure 98 may be lengthened so as to provide the desired overall length of the flexure 98 which is effectively unrestrained by any portion of the substrate 82.

The channel 86 and the flexure 98 may both have a large length to width ratio (L/ ) , with "large" being defined in this context to be a L/W in the range of about 5:1 to 1000:1. Preferably, the L/W may be about 20:1. A large L/W may be desirable because it may allow the linear flow regulator 80 to have a more robust flexure 98, since the membrane 84's regulator function is distributed over a longer length, as compared to a regulator 80 with a flow channel 86 and a flexure 98 having a small L/W ratio, such as 1:1.

A more robust flexure 98 may be desirable since it may be more durable, it may be less likely to rupture due to undesired operating overpressures, it may be easier to manufacture, and it may be easier to handle during the assembly of the flow regulator 80, as compared to a less robust flexure 84.

A channel 86 and a flexure 98 having a large L/W ratio also allows the use of a channel 86 and a regulator gap 99 having a larger cross-sectional area, for any particular

designed flow and regulation parameters, as compared to a channel 86 and a flexure 98 having a small L/W ratio. This is because as the channel 86 becomes longer, its fluid resistance (R _B ) becomes greater. Thus, for any particular desired operating pressure, in order to obtain a particular desired flow rate (Q) , the cross sectional area of the channel 86 and the regulator gap 99 will have to be made larger, as the channel 86 and the flexure 98 become longer, in order to compensate for the otherwise increased fluid resistance (R _s ) of the elongated channel 86 and the regulator gap 99.

Such an elongated channel 86 and regulator gap 99, having a larger cross-sectional area for a particular desired operating pressure and flow rate, may be desirable because their larger cross-sectional area may be less likely to foul due to contaminants in the medication 12, due to corrosion of the channel 86, the regulator seat 90, and the flexure 98.

From the disclosures in this document, it is possible to selectively design a linear flow regulator 80 for any particular desired flow regulation characteristics or driving pressure difference (P) . This may be done by selectively adjusting one or more of the pertinent parameters, such as: (a) the number, size, shape, length and location of the inlet port 88 and the outlet port 92; (b) the number, size, cross-sectional configuration, and length of the channel 86 and the regulator gap 99; and (c) the length, thickness, resiliency, elasticity and stiffness of the flexure 98. MICROMACHINED LINEAR FLOW REGULATOR 80

HAVING A CONTOURED REGULATOR SEAT 90 (FIGS. 7 - 12):

MANUFACTURE The substrate 82 may be made from any suitable strong, durable material which is compatible with the medication 12, and in which the channel 86 and the outlet port 54 may be manufactured in any suitable way, such as by using any suitable etching, molding, stamping and machining process. Such a machining process may include the use of physical

tools, such as a drill; the use of electromagnetic energy, such as a laser; and the use of a water jet.

The membrane 84 may be made from any suitable strong, durable, flexible, material which is compatible with the medication 12. The membrane 84 may also be elastic. If the regulator 80 is intended to regulate a medication 12 which is to be supplied to a human or an animal, then any part of the regulator 80 which is exposed to the medication 12 should be made from, and assembled or bonded with, non-toxic materials. Alternatively, any toxic material which is used to manufacture the regulator 80, and which is exposed to the medication 12 during use of the regulator 80, may be provided with any suitable non-toxic coating which is compatible with the medication 12. Suitable materials for the substrate 82 and the membrane 84 may be metals (such as titanium) , glasses, ceramics, plastics, polymers (such as polyimides) , elements (such as silicon) , various chemical compounds (such as sapphire, and mica) , and various composite materials. The substrate 82 and the membrane 84 may be assembled together in any suitable leak-proof way. Alternatively, the substrate 82 and the membrane 84 may be bonded together in any suitable leak-proof way, such as by anodically bonding them together; such as by fusing them together (as by the use of heat or ultrasonic welding) ; and such as by using any suitable bonding materials, such as adhesive, glue, epoxy, glass solder, and metal solder.

Anodically bonding the substrate 82 to the membrane 84 may be preferable for at least four reasons. First, anodic bonding is relatively, quick, easy and inexpensive.

Second, an anodic bond provides a stable, leak-proof bond. Third, since an anodic bond is an interfacial effect, there is no build-up of material at the bond; and the bond has essentially a zero thickness, which desirably creates no essentially no spacing between the substrate 82 and the membrane 84. As a result, an anodic bond does not interfere with the desired height or shape of the regulator gap 99.

Fourth, an anodic bond may be preferable since it eliminates the need for any separate bonding materials, which might otherwise clog or reduce the size of the inlet port 88, the channel 86, the regulator gap 99, and the outlet port 92; or which might lead to corrosion of the joint between the substrate 82 and the membrane 84.

One example of how the linear flow regulator 80 may be manufactured will now be given.

The starting point may be a 76.2 mm diameter wafer of Corning 7740 Pyrex glass, which will form the regulator

80's substrate 82. The channel 86 and its regulator seat 90 may then be formed in the substrate 82 in any suitable way. One suitable way may be to first etch a generally rectangular channel into the substrate 82, wherein the rectangular channel has a length and a depth about equal to the length and the depth of the desired channel 86 and regulator seat 90. The rectangular channel may be etched into the substrate 82 in any suitable way, such as by using a process which is the same as, or at least similar to, that used to etch the radial flow regulator 32's inlet channels 38 into its substrate 34. After the rectangular channel has been etched, the regulator 80's outlet port 92 may be formed. The structure, operation, theory and manufacture of the linear flow regulator 80's outlet port 92 is the same as, or at least similar to, that of the radial flow regulator 32's outlet port 54, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document. After the outlet port 92 has been formed, a nominal layer of one or more corrosion-resistant substances may be deposited onto the top surface of the glass wafer in any suitable way. As a result, the channel 86, the regulator seat 90, and the outlet port 92 will have been coated with a layer of the corrosion-resistant substance(s) . The structure, operation, theory and manufacture of such a layer of one or more corrosion-resistant substances for the linear flow regulator 80 is the same as, or at least similar to, that of the radial flow regulator 32's layer of

one or more corrosion-resistant substances, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document. After the channel 86 and the regulator seat 90 have been etched, and after any desired layer of corrosion- resistant substance(s) has been applied to etched portions of the glass wafer, the photoresist and chrome which remain on the unetched portions of the glass wafer may be removed by any suitable means, such as by using standard lift-off techniques.

At this point, certain work on the manufacture of the membrane 84 may be done before the desired channel 86 and regulator seat 90 may be completed. Manufacturing the membrane 84 and bonding it to the glass wafer (which is the substrate 82) may be done in any suitable way. The structure, operation, theory and manufacture of the linear flow regulator 80's membrane 84 and the bonding of the membrane 84 to its substrate 82 to form a silicon/glass sandwich is the same as, or at least similar to, the manufacture of the radial flow regulator 32's membrane 36, and the bonding of the membrane 36 to its substrate 34 to form a silicon/glass sandwich, except for those differences, which will be made apparent by an examination of all of the Figures and disclosures in this document.

Some of those differences will now be addressed. It may be recalled that the starting point for the regulator 80's membrane 84 may be a prime silicon wafer having a boron-doped epitaxial silicon layer which has been deposited onto one of its surfaces. Since the boron doped layer will ultimately form the regulator 80's membrane 84, the boron-doped layer's thickness will depend on the desired thickness of the membrane 84. The boron-doped layer has a dramatic etch-resistance to silicon etchants based on ethylene diamine; but is easily etched by isotropic etchants. The isotropic etchants will also easily etch the non-doped layer of the silicon wafer. Accordingly, the first step in manufacturing the regulator 80's membrane 84 is to first clean the silicon

wafer; apply a thin chrome metallization layer to both of the wafer's surfaces; and then apply and dry a thin layer of any suitable photoresist on top of the chrome layer.

An image of the inlet port 88 may then be exposed onto the photoresist on the silicon wafer's boron-doped layer, and then developed; after which the silicon wafer may be cleaned and dried.

As a result of the forgoing procedure, the chrome layer over the boron-doped layer will now bear an image, unprotected by the photoresist, of the inlet port 88. The unprotected portion of the chrome layer may then be etched away; resulting in an image of the inlet port 88 having been formed on the boron-doped layer, which image is unprotected by the layers of photoresist and chrome which cover the rest of the silicon wafer. The image of the inlet port 88 may then be etched into the boron-doped layer to a depth which is at least slightly greater than the thickness of boron-doped layer in any suitable way, such as by using an isotropic etchant. The photoresist and the chrome layers may then be removed from the silicon wafer, which may then be cleaned and dried.

The silicon and glass wafers may then be aligned, so that the etched image of the inlet port 88 in the boron- doped layer is in proper registry with the etched image of the channel 86 in the glass wafer. The etched surfaces of the silicon and glass wafers may then be anodically bonded together; after which the non-doped layer of the silicon wafer may then be etched and ground to produce the desired membrane 84, with the desired inlet port 88. The desired inlet port 88 is automatically formed during the etching and grinding of the non-doped layer of the silicon wafer because the isotropic etchant had previously completely etched away the image of the inlet port 88 in the boron- doped layer of the silicon wafer; so that when the non- doped layer of the silicon wafer is etched and ground away, the inlet port 88 is automatically formed.

The desired contour may be imparted to the channel 86's regulator seat 90 in any suitable way. One suitable way may be to use a pressure forming method, in which the

O

substrate 82 may first be placed into a forming device which restrains the substrate 82's bottom surface 104 and its lateral edges. A flat stencil having a cutout whose length and width corresponds to that of the desired channel 86 may be provided; the membrane 84 may be tightly sandwiched between the stencil and the substrate 82; and at least the substrate 82 may be heated to an elevated process temperature at which the substrate 82 is softened. For example, the elevated process temperature for a Pyrex glass substrate may be about 600°C.

The portion of the membrane 84 which is exposed through the stencil (i.e., the flexure 98), may then exposed to a pressure about equal to the regulator 80's maximum designed driving pressure difference (P) , thereby deflecting the flexure 98 down into the softened substrate 82 and forming the channel 86's contoured regulator seat 90. For example, for a Pyrex glass substrate which is heated to about 600°C, the pressure may be about 100 pounds per square inch (psi) . The rectangular channel mentioned above, which was etched into the substrate 82, may aid in the formation of the desired contoured regulator seat 90 since all the deflected flexure 98 needs to do is to mold the rectangular channel's sides and bottom into the desired contour. The forming device for the substrate 82 may have a relief hole which permits the exit of any excess substrate 82 material which is displaced by the deflected membrane 84. Alternatively, the rectangular channel mentioned above may be dispensed with, and the flexure 98 may be deflected with pressure down into the softened substrate 82 to form the contoured regulator seat 90.

The desired pressure may then be maintained on the flexure 98 while the substrate 82 is cooled and hardened, thereby forming the substrate 82's channel 86 with the desired contour in its regulator seat 90. The pressure may then be released, allowing the elastic flexure 98 to return to its undeflected configuration. The stencil, the forming device and any undesired displaced substrate 82 material may then be removed.

Although the use of a heat softened substrate 82 was described above, the contoured regulator seat 90 may be pressure formed into the substrate 82 in any other suitable way. For example, the substrate 82 may be selected to be made from a material, such as an epoxy, a ceramic material or a solvent-softened material, which is soft at room temperature, and which is then hardened after the flexure 98 is deflected down into it by a chemical reaction, by the application of heat, or by the evaporation of the solvent, respectively.

Another way to form the contoured regulator seat 90 may be to micromachine the desired contour into the substrate 82 by use of a laser beam. Such a laser beam may be regulated, by any suitable means, to have an intensity gradient similar to that of the desired contour of the channel 86's regulator seat 90, such as by projecting the laser beam through one or more suitable lenses and/or gradient filters.

Another way of using a laser beam to micromachine the desired contour into the substrate 82 may be to project the laser beam through a mask to give at least a portion of the laser beam a cross-sectional configuration which is similar to that the desired contour of the regulator seat 90.

The manufacture of only one linear flow regulator 80 was described above. However, it will be appreciated that on any pair of glass and silicon wafers numerous regulators 80 could be manufactured simultaneously in a manner similar to that described above. If such is the case, an array of substrates 82 may be simultaneously etched in the glass wafer before the silicon and glass wafers are aligned and bonded together. Then, all of the membranes 36, and their inlet ports 88 may be formed simultaneously. The silicon/glass sandwich may then be divided by any suitable means (such as dicing) into individual chips, each chip bearing at least one linear flow regulator 80.

One of the advantages of using the etching, anodic bonding, and pressure forming process which was described in detail above is that such a process enables high quality, very reliable linear flow regulators 80 to be mass

produced in great numbers at a cost so low that the regulator 80 may be considered to be disposable. In addition, it should also be noted that the regulator 80 is stunning in its simplicity since it may have as few as only two parts (the substrate 82 and the membrane 84) ; and since it may have only one moving part (the membrane 84's flexure 98) . Further, the cost of the regulator 80 may held to a very low level because the raw materials from which the regulator 80 may be made may be very inexpensive, such as glass and silicon.

MICROMACHINED LINEAR FLOW REGULATOR 110 HAVING A NON-CONTOURED REGULATOR SEAT 90 (FIGS. 13 - 15):

STRUCTURE The linear flow regulator 110 which is illustrated in Figs. 13 - 14 is the same as, or at least similar to, the linear flow regulators 80 of Figs. 7 - 11 in its structure, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document. Accordingly, the respective parts of the linear flow regulator 110 of Figs. 13 - 14 has been given the same reference numerals as the corresponding parts of the linear flow regulators 80 of Figs. 7 - 11, for clarity and simplicity.

As seen in Figs. 13 - 14, the linear flow regulator 110 has a pair of outlet ports 92; and a channel 86 having a pair of sides 112 which are at right angles to the non- contoured regulator seat 90.

As used herein, the term "non-contoured" means that the regulator seat 90 does not approximate, or duplicate, the contoured shape that the flexure 98 would assume if the flexure 98 were entirely unrestrained by any part of the substrate 82 when the linear flow regulator 110 is subjected to the regulator 110's maximum designed driving pressure difference (P) of the medication 12 between the regulator 110's inlet port 88 and outlet port 92.

It should be understood that Figs. 13 - 14 illustrate only one example of a non-contoured regulator seat 90, i.e., a flat non-contoured regulator seat 90. Naturally,

the non-contoured regulator seat 90 could have any of a variety of other configurations, shapes, or forms.

By way of example, the linear flow regulator 110's substrate 82 may be manufactured from 7740 Pyrex glass, and may have a thickness of about 0.5 mm. The channel 86 and the regulator seat 90 may have a length of about 1.0 cm, and may have a maximum width of about 508 microns. The regulator gap 99 may have a maximum height of about 4.2 microns (when the driving pressure difference (P) is equal to zero) . Each outlet port 92 may have a minimum diameter of about 100 microns, and may have a length of about 496 microns. The membrane 84 may be manufactured from silicon, ar.d may have a thickness of about 9.0 microns. The inlet port 88 may have a width of about 508 microns, and a length of about 500 microns. The flow characteristics of this example linear flow regulator 110 are illustrated in Fig. 15, which will be discussed below.

MICROMACHINED LINEAR FLOW REGULATOR 110 HAVING A NON-CONTOURED REGULATOR SEAT 90 (FIGS. 13 -15): OPERATION AND THEORY

The linear flow regulator 110 which is illustrated in Figs. 13 - 14 is the same as, or at least similar to, the linear flow regulators 80 of Figs. 7 - 11 in its operation and theory, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document. During operation, as a driving pressure difference (P) is applied across the regulator 110, such as by pressurizing the source of the medication 12 with respect to the regulator 110's outlet port 92 by any suitable means, the medication 12 will pass sequentially through the regulator 110's inlet port 88, regulator gap 99, and outlet port 92.

Referring now to Fig. 15, the flow rate (Q) of the medication 12 through the regulator 110 is plotted in terms of microliters per day (μL/day) , as a function of the driving pressure difference (P) across the regulator 110 in mm Hg. The regulator curve 114 is a plot of a theoretical mathematical model of the flow rate (Q) of the medication 12 for a regulator 110 having the physical parameters of

the example regulator 110 which was set forth above. The theoretical mathematical model will be discussed below. The seven square data points 116 seen in Fig. 15 are for the measured flow rates (Q) of an actual flow regulator 110 having the physical parameters of the example regulator 110 which was set forth above. As seen, the theoretical model does quite well in predicting the performance of the flow regulator 110.

Fig. 15 reveals that, at a zero driving pressure difference (P) , there is no flow of the medication 12 through the regulator 110. Then, as the driving pressure difference (P) is increased from zero, the regulator 110 exhibits two main flow regimes.

That is, as the driving pressure difference (P) is increased from zero, there is a corresponding increase of the flow rate (Q) ; but there is also a gradual lessening of the flow rate's (Q's) sensitivity to the driving pressure difference (P) . For example, this is seen on the curve 114 at driving pressure differences (P) from about 0.0 mm Hg to about 120 mm Hg.

Then, at higher pressure differences (P) , there is a "control zone" wherein the flow rate (Q) is relatively linear function of changes in the driving pressure difference (P) , and the sensitivity of the regulator 110 to changes in the driving pressure difference (P) is reduced by about 45%, as compared to a device having no flow regulation properties at all.

This behavior of the flow regulator 110 during operation is due to the fact that, as the driving pressure difference (P) across the regulator 110 is increased, the flexure 98 tends to be deflected towards the regulator seat 90 an increased amount, thereby reducing the height of the regulator gap 99. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the increased driving pressure difference (P) . This is because the reduced height of the regulator gap 99 will tend to compensate for the increased driving pressure difference (P) by reducing the flow rate (Q) which would otherwise occur at that increased driving pressure difference (P) .

On the other hand, if the driving pressure difference (P) decreases, then the flexure 98 will tend to be deflected towards the regulator seat 90 a decreased amount, thereby increasing the height of the regulator gap 99. This, in turn, tends to maintain the flow rate (Q) at a relatively constant value, despite the reduced driving pressure difference (P) . This is because the increased height of the regulator gap 99 will tend to compensate for the reduced driving pressure difference (P) by increasing the flow rate (Q) which would otherwise occur at that reduced driving pressure difference (P) .

However, as seen in Fig. 13, even when the driving pressure difference (P) has been increased to the point where the flexure 98 is deflected so much that it starts to contact the regulator seat 90, the medication 12 will still be permitted to flow through the two side channels 118 which are formed between the deflected flexure 98, the regulator seat 90's side portions 120, and the channel 98' side walls 112. Lastly, as the driving pressure difference (P) is increased still further, the flexure 98 will be flattened against the regulator seat 90 an increased amount, thereby gradually decreasing the size of the side channels 118 (and vice versa) . In order to assist in designing the regulator 110 to have any particular desired regulator curve 114, either an empirical strategy or a mathematical model may have to be employed.

The starting point for the mathematical model is the mathematical model which was set forth above regarding the regulators 80. It will be recalled that the above equation 7 yielded a relation between the driving pressure difference (P) and the total flow rate (Q) of the medication 12 up to the point P = 1, the dimensionless pressure at which the flexure 98 first contacts the regulator seat 90.

However, beyond that pressure the flexure 98 will spread laterally across the regulator seat 90; and the first point of contact between the flexure 98 and the

regulator seat 90 will expand and move along the z axis towards the inlet port 88.

In this mode of operation, the dimensionless pressure (P) still cannot exceed 1.0 since the depth of the channel 86 is fixed. Instead, as has been mentioned, two parallel side channels 118 are formed, each having a size which is a function of the driving pressure difference (P) . That is, as the driving pressure difference (P) increases, the size of the side channels 118 decreases, (and vice versa) . For a channel 86 having a length Z ₀ , the point of first contact between the flexure 98 and the regulator seat 90 is some fraction of that length, i.e., n • Z ₀ . Since P = 1 at each local fluid slice in the two side channels 118, the effective width of each of the side channels 118 is:

^e ^725

where P 1.

From the above Equation 6, for the boundary condition P = 1, the following differential equation for pressure drop can be obtained:

^=21.57. (^?£._ _) ^d z Et ³ h ⁴ nP ' ²⁵

where (Q ₀ ) is the flow when the contact between the flexure 98 and the regulator seat 90 is at z = Z ₀ . That is, when P = 1 is the above Equation 7. Upon separation of variables, and some algebraic manipulation, a surprisingly simple relationship for flow versus pressure is found in the mode where there is contact between the flexure 98 and the regulator seat 90:

Q=Q ₀ [l+.427i(p ¹ - ²⁵ -i)]

where P > 1.

The above model may now be used to predict the performance of the linear flow regulator 110.

From the disclosures in this document, it is possible to selectively design a linear flow regulator 110 for any particular desired flow regulation characteristics or driving pressure difference (P) . This may be done by selectively adjusting one or more of the pertinent parameters, such as: (a) the number, size, shape, length and location of the inlet port 88 and the outlet port 92; (b) the number, size, cross-sectional configuration, and length of the channel 86 and the regulator gap 99; (c) the length and shape of the regulator seat 90; (d) the length and shape of the side channels 118; and (e) the length, thickness, resiliency, elasticity and stiffness of the flexure 98.

MICROMACHINED LINEAR FLOW REGULATOR 110 HAVING A NON-CONTOURED REGULATOR SEAT 90 (FIGS. 13 - 15):

MANUFACTURE The manufacture of the linear flow regulator 110 of Figs. 13 -14 may be the same as, or at least similar to, the linear flow regulators 80 of Figs. 7 - 11, except for those differences, if any, which will be made apparent by an examination of all of the Figures and disclosures in this document.

Since the regulator 110 has a non-contoured regulator seat 90, the steps relating to pressure forming the regulator 80's contoured regulator seat 90 may be eliminated. Instead, the rectangular channel is etched into the substrate 82 so that the rectangular channel has a length, width and depth which are equal to that of the desired channel 86. In other words, the etched rectangular channel becomes the desired channel 86, and the etched rectangular channel's bottom forms the non-contoured regulator seat 90; with no further work being needed in order to form the desired channel 86 and the non-contoured regulator seat 90. Another way to form the channel 86 and the non- contoured regulator seat 90 may be to use a laser beam or a water jet.

It is understood that the foregoing forms of the invention were described and/or illustrated strictly by way of non-limiting example.

In view of all of the disclosures herein, these and further modifications, adaptations and variations of the present invention will now be apparent to those skilled in the art to which it pertains, within the scope of the following claims.