The objective of this thesis was to improve a paraffin actuated micropump design, to be able to pump against extreme pressures (above 100 bar). This was accomplished by initially studying the membrane activation, using video capturing. The micropump has been improved to withstand pressures high enough, to enable use in an high-performance liquid chromatography (HPLC) system. The micropump has been shown to pump against back pressures up to 150 bar, with a positive net-flow. This should be compared with the previously recorded maximum back pressure of 50 bar. The pumping against high back pressures was possible due to an increased understanding of the sealing of the membranes. This resulted in a new design that was manufactured and characterised. Without clamping the pump was measured to manage back pressures of 10 bar, and then starting to leak in a bond at the flow channel. With supporting clamping, the managed back pressures increased ten folded.
When measured on the different valves, pressure above 200 bar has been possible to withhold. Although the valves were below their maximum limit, the pressure was not possible to be further increased due to a limitation in the equipment, i.e. risk of damaging the connections. When examined after pressurised at extreme pressures (above 100 bar) several times, no signs of fatigue or damage of the membrane was seen.
A new behaviour of the valves was discovered. Above certain pressures some designs self sealed, i.e. withholding the pressure after the voltage was turned off. For these valves the pressure had to be released by some other means.
resulterade i god t¨atning, men en op˚alitligt slagvolym vid aktivering. F¨or att komma bort fr˚an plasticeringen byttes membranet ut mot ett st˚almembran.
St˚almembranet visade sig vara mer begr¨ansat i sina r¨orelser och st¨orsta delen av utb¨ojningen skedde i mitten av membranet. Tryckm¨atningar visade p˚a att ett inlopps/utlopps h˚al placerat mot membranets mitt klarade tryck sv˚ara att uppn˚a med experimentutrust-ningen (> 200 bar), medan ett h˚al placerat mot membranets rand medf¨orde l¨ackage redan strax ¨over 20 bar. Den h¨ar uppt¨ackten var orsaken till de nya designerna d¨ar inlopps- och utloppsh˚alen placerades dels p˚a lika avst˚and fr˚an membranets mitt, men ¨aven d¨ar h˚alet riktat mot pumpens utlopp placerades centralt. Tanken bakom den senare designen var att det p˚alagda mottrycket ligger p˚a utloppet, medan inloppet kommer att befinna sig vid atmosf¨arstryck.
N˚agot ¨overraskande uppenbarades ett nytt fenomen d˚a de nya designerna utv¨arder-ades. Tidigare har halvm˚aneformade h˚al anv¨ants vid membranets rand f¨or att leda v¨atskan vidare fr˚an en pumpkammare till n¨asta. I de nya designerna anv¨andes utes-lutande cirkul¨ara h˚al vilka visade sig ha en tendens att l˚asa sig n¨ar de placeras l˚angt ut mot kanten p˚a membranet. Resultatet av detta var att ventilerna ¨over ett visst tryck sl¨ot t¨att och h¨oll t¨att ¨aven n¨ar sp¨anningen st¨angdes av, tills trycket l¨attades p˚a annat h˚all.
De tryckm¨atningar som avbrutits vid 200 bar har alla genomf¨orts utan detta st¨od. N¨ar mikropumpen l¨ats pumpa mot ett mottryck klarade den dock inte tryck ¨over 10 bar.
Med ˚atkl¨amning betedde sig mikropumpen b¨attre och har visat sig kunna pumpa mot tryck ¨over 100 bar. Ett positivt fl¨ode har registrerats med ett mottryck p˚a 150 bar. Enligt vad vi k¨anner till ¨ar detta det h¨ogsta tryck en membrandriven mikropump n˚ansin klarat av. Faktum ¨ar att det endast existerar en typ av mikropumpar, med elektroosmos som drivningsmekanism, som har rapporterats klarat h¨ogre mottryck. Pumpar som anv¨ander elektroosmos begr¨ansas dock av kravet p˚a att v¨atskan som pumpas m˚aste inneh˚alla joner. Mikropumpen i detta arbete har inte n˚agon s˚adan begr¨ansning, ¨aven om endast vatten ¨annu har utv¨arderats f¨or pumpen.
1.2 HPLC Systems . . . 4
1.3 Previous work . . . 5
1.4 Main objectives . . . 5
2 Theory 6 2.1 Actuation using paraffin . . . 6
2.2 The pump . . . 6
2.3 The membrane . . . 7
2.3.1 Initial membrane deflection . . . 7
2.3.2 Plasticising of the membrane . . . 8
2.4 Manufacturing . . . 8
2.5 Pump design . . . 9
2.6 Activation cycle . . . 9
2.7 Supply of energy through resistive heating . . . 11
2.8 Time constant for membrane activation . . . 12
2.9 Material selection of the membrane . . . 12
2.10 Theory of deflecting membrane . . . 14
3 Design 17 3.1 Design O (original design) . . . 17
3.2 Design I . . . 18
3.3 Design II . . . 18
3.4 Design III (IV & V) . . . 18
4 Experimental 20 4.1 Video capture of membrane activation . . . 20
4.3 Pumping against back-pressure . . . 23
4.4 Valve measurements . . . 23
4.5 Optical surface profilometry . . . 24
5 Results 25 5.1 Directional dependency of steel membranes . . . 25
5.2 Membrane activation - Images and video . . . 25
5.3 Valves . . . 25
5.3.1 Voltage dependency . . . 31
5.4 Pumping against back-pressure . . . 31
5.4.1 Unclamped micropump . . . 31
5.4.2 Clamped pumping . . . 32
5.5 Flow and activation of a pump-cycle . . . 35
5.5.1 Activation time of a steel valve . . . 35
5.5.2 Deflection of outlet membrane . . . 35
6 Discussion 39 6.1 A small pump with great potential . . . 39
6.2 Placement of channels towards the membrane . . . 39
6.3 Material selection for the membrane . . . 40
6.4 Steel compared to polyimide as a membrane . . . 40
6.5 Motivation of the new designs . . . 40
6.6 Trapped air . . . 41
6.7 Bonding between channel- and inlet-stencils . . . 42
6.8 Limitations in the equipment . . . 42
6.9 Expansion within the system . . . 43
6.10 Self-sealing valves . . . 43
6.11 Blocked channels . . . 44
6.12 Plateau in pressure measurements . . . 44
6.13 The activation time . . . 44
6.14 Frequency dependency . . . 45
6.15 Future improvements . . . 45
7 Conclusions 46
Acknowledgements 47
to achieve high flow-rates.
1.1
In a world of micropumps
The creativity of engineering micropumps (active parts of sub-cm size) knows no bound-aries. The utilised mechanisms for micropumps ranges from mechanical pumps to pumps using ionic transportation as an actuator mechanism, although very few pumps are of actual use in the world outside the laboratory [2]. Since there is a wide variety of mi-cropump designs, they need to be categorised. The most common way would be to differ between the non-mechanical and the mechanical pumps [3].
The non-mechanical pumps directly provides the driving energy to the working medium using a physical phenomenon (e.g. magnetism, gravity, Coulomb forces). The mechanical pumps on the other hand usually adds the energy by applying pressure on the medium. The function of an actuator in mechanical pumps is to apply a force on a membrane that will do the actual work on the medium. The mechanical pumps often include valves which will have the advantage/disadvantage that the flow will periodically stop. The majority of micropumps today are still mechanical due to the robust designs of these structures, even though the best flow-rates at highest back-pressures have been achieved using valve-less pumps. A comparison by Bod´en, figure 1.1, gives a good overview of the different pumps based on their actuation mechanism.
Looking at the valve-less designs, one of the advantages of these micropumps is that most of the designs will result in a continuos flow. There are also no (or at least a minimal amount of) moving parts, which reduces the risk of fatigue failures and leakages. Among
Figure 1.1: A comparison of the different micropumps based on the actuation mechanism
by Bod´en. The star marks the performance of the first design of the steel-micropump [1].
these pumps, at least three are worth mentioning. The first one is a mechanical valve-less micropump designed by Olsson et al. [4]. This pump is milled out of two 0.5 mm thick blocks of brass. The pump contains two parallel connected pumping chambers and is activated using four piezoelectric elements. This is a famous design because the concept entered new lands by removing the valves and using shaped inlet and outlet channels (denoted nozzle and diffuser) to direct the flow, figure 1.2(a). This pump achieved flow rates of 16 ml min−1 against a back-pressure of 0.17 bar. The downside with this pump
(also being one of its main features) is the lack of valves towards the ”outside”. This makes the pump useless as a dispenser and it is not possible to pump against high pressures because there is nothing to prevent the back-flow.
The second valve-less pump worth mentioning is one using electroosmotic flow as the driving mechanism, figure 1.2(b). The electroosmotic design is based on a spontaneous electric double layer, created at the interface between a surface and a liquid. At the surface of a glass capillary, spontaneously created positive charges will adhere to the anions in the liquid. As a result the cations will be distributed more towards the middle of the capillary. By applying a voltage parallel to the walls, the cations will move towards the negative charge dragging the neutral particles in the liquid with them [5]. This pumping mechanism relies on the pumped liquid being an ionic solution. Electroosmotic pumps are reported to achieve flow-rates of several µl min−1 at back-pressures reaching above
200 bar [6]. There has even been reports of pressures around 500 bar using a more complex variant of the electroosmotic principle denoted packed-bed electroosmotic pump [7].
(a) (b)
(c)
Figure 1.2: (a) The principle of the valve-less micropump designed by Olsson et al. [4].
The shapes of the inlet and outlet directs the flow generated by the movement of a mem-brane activated by a piezoelectric element.
(b) Electroosmotic principle. The negatively charged ions will be attracted towards the walls of the glas capillary leaving the positively charged ions in the middle. By applying a voltage perpendicular to this gradient, the positively charged ions will be moved dragging along the molecules in the fluid.
energy [8], figure 1.2(c). This pump, as well as many other valve-less pumps, have the disadvantage that they can only pump against low back-pressures.
Because the main objective for many micropumps is in biomedical applications (e.g. drug delivery), a well defined pumped volume is of great importance. In those applications, the pressure is not in a region where it would be the main design consideration. For a micropump to be used as an implantable device, the highest pressure to overcome in vivo is 1 bar [2]. However, there are applications where the ability to manage a high pressure is the primary interest.
1.2
HPLC Systems
High-performance liquid chromatography, or high pressure liquid chromatography (HPLC) is a technique to separate different components in a fluid, figure 1.3. A mobile phase is flowed through a stationary phase, often consisting of well defined particles [9] contained in a stainless steel tube. Apart from amplifiers, computer, ports etc. the system mainly consists of three parts: a pump, a separation column and a detector. The pressure in the system is primarily generated in the column (considered by many to be the heart of the system), that consists of some type of porous structure whose task is to make it hard for molecules to pass [10]. This column is usually operated at high pressures (well above 100 bar) which makes it sensitive to uneven flow rates, due to the risk of damaging the column [11]. The column will separate the molecules depending on for example their size, and will result in different times for different molecules to pass through the system to the detector. The result is a chromatogram with the intensity as a function of time. The time for a molecule to pass through the system is specific for that molecule, making separation possible. It is thereby of interest to have as constant flow rate as possible. The function of the pump in an HPLC-system is to provide enough pressure and to accomplish a constant flow through the column. Using displacement pumps, the constant flow rate becomes an obstacle that has to be managed. This can be accomplished by using several pumps connected in parallel, and thereby even out the flow.
Figure 1.3: A commercial HPLC-system [12].
1.3
Previous work
The first micropump, designed by Bod´en et al. [14], was mainly manufactured in poly-meric materials, with the exception of a glass lid and copper coated polyimide heaters. This pump was able to build up a pressure of 2 bar without clamping (a support from a fixture). When clamped it was able to pump up a pressure of 9.2 bar. Although this was considered a high pressure at the time, the pump was not long term stable and the evolution of the pump went towards a steel micropump. The first steel design used the same kind of polyimide membrane as was being used in the polymeric pump. This version of the metal pump could build up a pressure of 50 bar before a leakage was observed in the joints between the top stencil and the flow channel [1].
The polyimide membranes used were not able to cope the high pressures, but plas-ticised into the channels. This gave the pump an unpredictable flow-rate which is very undesirable for most applications. Afterwards a number of different membrane materials have been tested, although not fully evaluated. Among them there are two different steels, showing the most promising results.
1.4
Main objectives
In order to achieve pumping pressures above 100 bar, the main objectives that are to be accomplished:
• Understand the sealing of the polyimide and steel valves.
• To be able to pump against high pressures, the valves have to be designed to with-stand and seal at these pressures.
Theory
2.1
Actuation using paraffin
The common candle wax (paraffin) has some amazing properties. When heated up from room temperature it will undergo a phase transition around 45 ◦C (possible to varying
between -100◦C and 100 ◦C) dependent on the molecular length of the polymer chains,
resulting in a volume expansion in the region of 10 - 20 % of the original volume [15]. Paraffin has a high bulk modulus (i.e. the volume change for a certain applied pressure) about 1.3 GPa [16]. This can be compared with the bulk modulus of air in the region of 0.1 MPa [17]. The high bulk modulus of melted paraffin implies a low compressibility, hence the material can withstand high pressures without significant volume change.
The high bulk modulus and large volume expansion of paraffin, together with the low working temperature gives paraffin the potential of being a powerful actuator. Paraffin has a high heat capacity and low thermal conductivity. This implies that paraffin is most useful as an actuator material in miniaturised systems.
2.2
The pump
The micropump in this thesis is a peristaltic pump, and consists of three cavities with a diameter of 2 mm. The cavities are enclosed by a backplate and a membrane, connected in series, figure 2.1. The cavities are filled with paraffin and inside each cavity a heater is located. When the paraffin is melted it expands, exerting a pressure against the sur-roundings. If the steel-walls and backplate are considered stiff, the volume expansion will deflect the membrane. If enough paraffin is melted, the deflection of the membrane will be large enough to seal the channel and using this principle for all three membranes a pumping cycle can be achieved. The amount of paraffin that needs to be melted to seal the channel will set the lower limit of the power needed for the pump.
The three cavities will each have a specific role in the pumping cycle. The two outer membranes will work as valves, while the membrane in the middle will dispense the liquid. The membranes are connected by channels above the cavities.
(b)
Figure 2.1: (a) A cross-section of the old pump design. Top down: A top-stencil and
a channel stencil in stainless steel, a 50 µm thick PI membrane (yellow) with an initial deflection of 40 µm, a cavity-stencil, a heater (red), a cavity-stencil and at the bottom a back-stencil. (b) Left to right: Back stencil, cavity stencil, heater stencil, cavity stencil, membrane, channel stencil, top stencil.
2.3
The membrane
2.3.1
Initial membrane deflection
2.3.2
Plasticising of the membrane
The previously evaluated polymers showed a prone behaviour of plasticising by the pres-sure during the expansion of paraffin. The channel stencil will act as a stop for the membrane during the expansion, but this support does not exist in the holes for the inlet and outlet. A too ductile membrane will as a result be pressed into the channel and be plasticised. This behaviour has been observed for the polymeric materials but also for steel grade 304. The deformation will in the case of steel result in a changed deflection by a few µm, while the polymeric materials will have a changed deflection of more than 50 µm.
2.4
Manufacturing
Utilising the CAD (Computer Aided Design) technique the blueprints for the different stencils are sent to an external company, HP Etch, Sweden, to be etched out of either 100 or 200 µm thick steel sheets. The etching technique used is an industrial process called photochemical machining (PCM). The use of this technique allows large scale batch production of the steel stencils. Except for the heaters (etched out of copper coated polyimide sheets) and the membrane, this is the material used in the pump. To bond the steel sheets together a 7 µm parylene coating is used. The coating is performed by an external company PlasmaParylene, Germany. The bonding is done in two steps, the first at 200 ◦C is done in a vacuum oven for 30 minutes. The second heat treatment is
done at 240 ◦C for 30 minutes. The heaters used in the pump is a structure etched out
of a copper coated polyimide sheet, and the heaters are also coated with parylene for the bonding process. The heaters are manufactured at the ˚Angstr¨om Laboratory using standard lithography, wet and dry etching [1].
To bond the separate stencils together, a fixture is used where the stencils are stacked in order. The whole pump is bonded in one step excluding the backplate, which is later glued.
One of the previous problems with the pump has been paraffin leakage. Due to the isolation gap at the copper heaters, a channel exists for the paraffin to ”escape”. This has to be sealed and this is done after the pump is bonded together using a low viscosity epoxy glue (EPO-Tek 301-2). The paraffin is introduced to the cavities from the bottom by filling the cavities with melted paraffin. Afterwards, the excess paraffin is scraped off and a back-stencil is glued onto the pump using Loctite 407. The last step is to solder the wires onto the copper pads.
Figure 2.2: The micropump beside a swedish 1 SEK-coin.
2.5
Pump design
The micropump is 15 x 30 mm and 1 mm thick, figure 2.2. Compared to the previous design [1], the amount of paraffin has decreased to 1/3 of the original amount due to the previous amount of paraffin being over-dimensioned. The first metal pump used three 0.2 mm thick cavity-stencils. Today only two 0.1 mm thick cavity-stencils are used. The channel and top-stencil are more rigid than the cavity-stencil being 0.2 mm thick. The standard diameter of the holes for the channels used is 0.3 mm while the paraffin cavities have a diameter of 2 mm.
2.6
Activation cycle
to move in the opposite direction, giving a momentary negative flow.
Ideally the opening and closing of the membranes would be instantaneous, but the heating and cooling of the paraffin are time consuming processes, being the bottle necks in the driving cycle. The cycle time used in previous pump experiments has usually been in the region of 5 s, but this is far from an optimised speed.
The membrane activation is illustrated in figure 2.3(c).
!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~ 0 1 Pumpcycle Period Act iva te d Inlet Outlet Chamber T1 T2 T3 T4 T5 T6 (a) !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~ Flow Time T1 T2 T3-T5 T6 0 (b) (c)
Figure 2.3: (a) A schematic of the pumpcycle. (b) The flow of a whole cycle. (c) A
schematic over the membrane activation by expansion of paraffin for the different periods of the pump cycle.
The frequency is defined as [19]:
f = 1
T (2.1)
where T is the period of a whole pump cycle. A more detailed definition would be to sum the period time of all the cycle-steps and denote the individual parts with an index i. This yields the following equation:
f =X 1 Ti
!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~
R δ
a
Figure 2.4: A spherical cap.
If the volume V will be dispensed every cycle the flow rate q will be described by:
q= f V (2.4)
For the pump there is an optimal frequency that will give the highest flow rate, figure 2.5. At higher frequencies the time is not enough to heat the needed amount of paraffin or cool the paraffin to open the membranes, decreasing the performance of the pump. At lower frequencies the potential of the paraffin cavities is not fully used, and the paraffin is heated too much, resulting in a waste of energy and in worst case damage to the pump.
2.7
Supply of energy through resistive heating
Figure 2.5: A principal graph of the typical frequency dependency of the peristaltic
mi-cropump for atmospheric pressure.
which in this thesis is used denoting pressure): P = UI = U
2
R (2.5)
2.8
Time constant for membrane activation
During activation and deactivation there is a system specific delay before the response actually goes to the desired output. An example is the combustion engine in your car (the rpm does not react instantaneous with the throttle). The declination of the response during deactivation is often possible to fit to an exponential expression:
x(t) = x(0)e−t/τ (2.6)
x(t) is the momentarily response, x(0) the initial response, t is the time and τ is a time constant. For the case when t = τ , the exponential becomes e−1 and x(t)/x(0) is
thereby 0,367. This gives that the time it takes for the signal to become 37% of its initial value will be equal to the time constant τ , and this value can be used to compare different systems.
2.9
Material selection of the membrane
general properties of this material includes a yield strength of 61 MPa and a Young´s modulus of 2.5 GPa which is maintained up to 230 ◦C. Translating these numbers into
words, PI can be described as a rather flexible material that can withstand high loads considering that it is a polymer. Both 50 µm and 25 µm thick membranes have previously been used in the paraffin pump, but is now replaced since they plasticise. Compared to commonly used construction materials like stainless steel, PI has a low yield strength and will thereby plasticise at relatively low stresses. The consequence will be that PI has an unreliable stroke. To get around the problem of deformation of polymeric materials, elastomeres (e.g. PDMS) could be used. However, when pushed against the opposite surface containing the channels, the elastomere would be pushed into the channels by the melted paraffin, due to the inability of the material to support applied loads. Although the material would not permanently deform, it would eventually break due to the elongation. Steel is a very versatile material. It has both a high ductility and a high yield strength. This makes steel a good choice in many applications. The downside of this versatility is that there is almost always another material more suited for a specific task. The knowl-edge of steel is very broad and the properties can be varied using different hardening mechanisms (Hall-Petch strengthening, solid solution strengthening, precipitation hard-ening and cold working). Many engineers find comfort in using steel as a construction material due to the insensitivity to local composition variations due to the ductile nature of the material. Due to its versatility, there are many different categories of steel. The most common is the regular stainless steel usually denoted 18-8 or 304. It is an alloy of 18 % chromium and 8 % nickel. When in contact with an oxygen rich atmosphere, the chromium will form a homogeneous layer of chromium oxide at the surface, preventing the oxygen to come in contact with the rest of the material, and thereby protecting the bulk from corrosion [21].
the membranes used today is parylene coated prior to the bonding process. Disregarding the heaters, 18-8 is the construction material in the micropump. Looking at figure 2.7 in the theory-chapter, showing the different stresses of a simple 10 µm thick membrane, it can be seen that for a deflection of 12 µm the stresses at the edge of the membrane is dangerously close to the yield strength of the material with the present design. This deflection is usually the initial deflection of an unactuated 304-membrane.
One of the biggest problems with the metal membranes may be considered to be fatigue. After long runs cracks appear at the membrane even though the stresses is below the yield strength. It is estimated that 90 percent of all failures of metallic applications is the result of a fatigue failure [23]. The stress at failure in a steel material is a function of the number of cycles. This is not a linear relation and the curve will even out at a specific value of the applied stress. The value is usually in the region above 35 - 60 % of the yield strength for steel. As a construction parameter this value is often a better parameter than the yield strength of a metallic material. 301, another variation within the stainless steel family, has similar properties as 304, but is modified to have a higher fatigue strength.
Titanium is used in applications where high strength is desired in combination with a low weight. Although pure titanium is a relatively soft material, it is possible to reach tensile strengths up to 800 MPa with titanium alloys (Grade 5, Ti6Al4V). The limit where the fatigue strength goes to infinity is about 160 MPa, which is in the region of the tensile strength of steel. A favourable property with titanium as a membrane material would be its lower Youngs modulus (E), which is in the region of 110 GPa for almost all alloys. The consequence of the lower Youngs modulus and high fracture strength is a material that is more durable than steel, but also more prone to deflection, i.e. a membrane that will give a larger stroke and that can withstand higher stresses. However, titanium is an expensive material with a limited number of suppliers, especially for thin alloy foils. Because of this, the membrane material used in this thesis was stainless steel grade 301.
2.10
Theory of deflecting membrane
By simplifying the membrane to a circular disc (see figure 2.6) the membrane stress is described by the following three equations at some key points [24].
a
δ
h
p
where a and h is the radius and thickness of the membrane respectively, µ is the Poissons ratio for the material used and p is the pressure evenly distributed along the membrane.
The characteristic equation for a flat membrane for small displacements (the deflection being less than a third of the membrane thickness) is given by
p= 16Eh
3
δ 3(1 − µ2
)a4 (2.10)
δ is the variable denoting the deflection of the membrane and E is the Youngs modulus of the membrane material. This equation only takes the bending stresses into account. The characteristic equation for the tensile stresses of a deflecting membrane is described by p= 7 − µ 3(1 − µ) Eh4 a4 δ3 h3 (2.11)
Worth noticing is that the tensile stresses increases with the cubic power of the deflection compared to the linear relation of the bending stresses towards the deflection. By the method of super-positioning, equation 2.10 and 2.11 can be combined to an equation which takes into account both bending stresses and the tensile stresses:
p= Eh 4 a4 ( 16δ 3(1 − µ2 )h + (7 − µ)δ3 3(1 − µ)h3) (2.12)
The material constants related to this equations is mainly the Youngs modulus (E) and Poissons ratio µ. These two values will represent the stiffness of the material and the ratio of which a material will expand in one direction when compressed in another direction. One critical variable not mentioned in these equations is the tensile strength of the material σy. This is the maximum stress that a specific material can withstand before
it starts to plasticise.
2 4 6 8 10 12 14 50 100 150 200 250
Stress (Deflection) for SS 304
Deflection [µm]
Stress [MPa]
Stress at center Radial stress at edge Tangential stress at edge
Figure 2.7: A graph of equation 2.7 - 2.9, calculated for different values of the deflection
Inlet Outlet
(a) (b)
(c) (d)
Figure 3.1: The placement of the holes in relation to the paraffin cavities for different
designs of the channel stencils. The dashed circles are the cavities and represent, from left to right, the inlet, chamber and outlet. (a) The original design with crescent shaped holes towards the channels. (b) Design I (c) Design II (d) Design III, IV and V
3.1
Design O (original design)
This is the original design of the channel stencil that was used with the polyimide mem-brane. The inlet and outlet are circular holes and they are located directly above the centre of the membranes. The holes towards the channels has the shape of a crescent, and is located near the rim of the membrane, figure 3.1(a). The pump design is symmetrical, i.e. the pump can be used to pump in both directions.
Table 3.1: The used designs of the different micropumps occurring in this thesis. Micropump Design L56 O L80 III L81 I L82 II L83 V L84 IV L87 II
3.2
Design I
The crescent shaped holes are replaced by circular holes with a diameter of 0.3 mm, and the holes at the inlet and outlet are distributed at an equal distance of 0.3 mm from the centre of the cavities, figure 3.1(b). Symmetrical.
3.3
Design II
The holes at the inlet and outlet are at equal locations as in design I with the diameter of 0.3 mm. The holes in the middle chamber are placed at a distance of 0.3 mm from the centre, figure 3.1(c). Symmetrical.
3.4
Design III (IV & V)
The idea behind design III is based on the results (5.1) indicating that a hole at the rim will manage a lower pressure than a hole in the middle, figure 3.1(d). The outlet hole is placed at the membrane centre, whereas the inlet hole is placed close to the rim, at a distance of 0.65 mm from the membrane centre, figure 3.1(d). This removes the symmetry of the design, and the pump is thereby limited to pump in one direction.
Design IV uses the same placement of the holes as design III, with the only difference that a diameter of 0.5 mm is used instead of 0.3 mm as in design III.
!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~
Top stencil
Channel stencil
Experimental
The capabilities of the micropump was evaluated by letting the micropump work against an applied pressure built up by an external pump. The same set-up was used to charac-terise the individual valves with the difference that the flow sensor was removed in the valve measurements.
4.1
Video capture of membrane activation
A separate steel fixture was fabricated with a milled opening in the middle to allow light passing through. One of the long sides was chamfered at an approximate angle of 45◦to
al-low illumination from the side. The fixture was placed underneath a microscope equipped with a video-recorder connected to a computer. Using the built-in Windows Movie Maker at the maximum possible resolution (720x540) the live footages were obtained. The top-stencil and channel-top-stencil of the micropump was replaced by a channel-top-stencil milled out in polymethylmethacrylate (PMMA), which is optically transparent and makes it possi-ble to watch the membrane during activation. Due to the optical properties of steel, a transparent polyimide membrane was used at first to evaluate the melting behaviour of paraffin. Later a stainless steel membrane was used to evaluate the behaviour of the steel membrane compared to the polyimide membrane.
4.2
Setup
Standard HPLC connectors from Upchurch Scientific was used and assembled according to figure 4.1(a). The pump was mounted inside a fixture to get support by an external clamping over the channel structure. The fixture also had holes to fit connectors towards the tubing. The applied pressure was monitored by connecting the inlet to a pressure sensor. For the flow measurements a flow sensor was connected between the pressure-sensor and micropump to monitor the strokes and thereby also the flow. The signals generated from the pressure sensor and flow sensor were monitored using the Labview 8.5 software.
ference as much as possible, heat-conducting paste is applied in the interface between the fixture and the plate.
4.2.2
Pressure sensor
The pressure sensor used is a Keller PA-11 mounted onto a 3-way junction. It uses a constant current of 1 mA and the voltage varies linear with applied pressure (at a rate of 0.039 V/bar). The limit of the sensor is 400 bar and the output signal is monitored. The constant current was acquired from a TTi QL355TP power supply.
4.2.3
Flow sensor
The flow sensor used is a Sensiron SLG1430. The sensor was connected in the fluidic system using capillaries with a diameter of 75 µm. The range of the sensor is ±40 µl min−1
with a resolution of ±7 nl min−1, and the sampling rate used was 6.25 Hz.
4.2.4
External pump
When measuring which pressures the membrane can withstand, an external pump was used for pressure build-up. In the first experiments a syringe pump (Harvard Instruments PHD 2000) was used. However, since the pressure limit was about 65 bar it was later replaced by a Pharmacia P-3500 pump. Even though the micropump can build up pres-sures of its own, the pumped amount is so small that it would take several hours to reach a pressure of a few bar due to the elasticity of the fluidic system.
4.2.5
Labview
(a) 1. External Pump 2. Pressure Sensor 3. Flow Sensor 4. Micropump mounted inside fixture Computer with Labview 8.5 7. NI SCB-68 8. NI TBX-68 Amplifier Input signals Output signals Inlet Outlet Fluidic Tubing Input signals Output signals (b)
Figure 4.1: (a) Photo of the set-up. (b) Principal diagram. The set-up: 1. HPLC pump
Figure 4.2: The fixture in which the pump is mounted for experiments.
of 50 mV was present between the set value in Labview and the heater. The driving frequency and voltage can be set separately for the different heaters.
In the latest version of the Labview-program, the flow, applied voltages over the heaters and pressure could be registered simultaneous making it possible to determine the response to a membrane activation at a certain pressure.
4.3
Pumping against back-pressure
By letting the P-3500 pump build up a pressure against the pump outlet, the ability of the micropump to pump against a back-pressure was evaluated. This eliminated the need of the micropump to build up the pressure by itself (which is a very time consuming process). The P-3500 pump can be stopped without losing the built up pressure (although a small pressure drop is noticed at higher pressures). When the desired pressure was reached, the P-3500 pump was paused and the pressure and flow logged to a text-file. The flow was measured in two ways: using a flow sensor, and by measuring the time to pump a specified amount through a capillary at the inlet.
4.4
Valve measurements
Figure 4.3: The user interface of the program used for control and monitor of the pump.
The pressure and flow rate was monitored in graphs as well as the applied voltages.
4.5
Optical surface profilometry
To determine the membrane deflection, a white-light interferometer (WYKO NT1100) was used. The objective used was at 10x magnification.
Outlet
Pressurisation from inside Pressurisation
from outside
pressurised from above, it was possible to achieve pressures over 100 bar while pressures from inside had a maximum around 25 bar. After each experiment the valve was checked so that it sealed by pressurising it with a syringe. The resistance of the activated heater was measured to be 1.8 Ω and the applied voltage was in the two first measurements 0.9 V but was afterward increased to 1.0 V.
5.2
Membrane activation - Images and video
The design of the channel-stencil shown is a remake of the crescent shaped holes. The lid was milled out of a PMMA (plexiglass) sheet. This implied that larger holes had to be used due to the limiting mechanical properties of PMMA.
During deactivation it can be seen that the membrane is going down instantaneous, long before the major part of the paraffin is solid, figure 5.2. Paraffin is melting from inside to rim.
The behaviour of polyimide membranes and steel membranes have a major difference. The polyimide membranes are flexible and thereby manage to seal along all of the cavity, whereas the sealing capability of the steel membranes are limited to the largest movement at the centre of the cavity.
5.3
Valves
The outlet membrane was pressurised for the five new designs by building up a pressure until either the membrane or the equipment failed. The pressure was in this case applied
200 250 300 350 400 450 500 550 600 650 700 0
50 100 150
L56 − Pressurized from outside
Time [s] Pressure [bar] Measurement 1, 50 µl min−1 Measurement 2, 50 µl min−1 Measurement 3, 20 µl min−1 500 600 700 800 900 1000 1100 0 5 10 15 20 25
L56 − Pressurized from inside
Time [s]
Pressure [bar]
Measurement 4, 20 µl min−1 Measurement 5, 20 µl min−1
Figure 5.1: Pressure measurement of the pump L56 (Old design). The membrane was
pressurised from the two channel openings. The disturbances in the graphs were later found out to be due to the electrical field emitted from a nearby placed lamp.
from the outside, figure 4.4, and hence the joints of the pumps were loaded in a minimal way. During all of the valve measurements, the pressure was built up with the Pharmacia P-3500 pump at the rate of 10 µl min−1. All the valve measurements have been conducted
with the pump unclamped.
The outlet membrane of design I (L81) sealed at 0.75 V with a heater resistance of 2.5 Ω, figure 5.3. No leakage was observed through the whole experiment, although a small pressure drop was observed at 180 bar. At 200 bar the experiment was aborted due to the risk of breaking the ferrules on the tubings. When the voltage was turned off the membrane was still sealing, preventing the pressure from going down. This self-sealing behaviour was later pinpointed to occur between 35-45 bar. To release the pressure the flow sensor had to be opened.
Design II (L82) had a resistance of 2.2 Ω and sealed at 0.95 V. This valve only managed to withstand a pressure of 120 bar after the voltage was increased to 1.05 V. This was due to a leakage between the ferrules and the pump. The ferrules were tightened, figure 5.4 but the leakage in the ferrules was still present and no further investigations were made to try to increase the pressure.
(a) (b)
(c)
Figure 5.2: Solidification of paraffin. (a) Activated heater. (b) The paraffin solidified at
0 5 10 15 20 0 20 40 60 80 100 120 140 160 180 200 L81 Time [min] Pressure [bar] Flowrate 10 µl min−1 at 0.75 V
Figure 5.3: Valve measurement of the pump design I (L81). At 200 bar the measurement
was stopped due to the risk of damaging the ferrules at higher pressures. No leakages were noticed although a strange bump occured in the measurement around 180 bar. When the valve was deactivated the pressure remained constant.
0 2 4 6 8 10 12 14 16 0 20 40 60 80 100 120 L82 Time [min] Pressure [bar] Flowrate 10 µl min−1 at 1.05 V
Figure 5.4: Valve measurement of design II (L82). The membrane started leaking just
0 2 4 6 8 10 12 14 16 18 20 0
Time [min]
Figure 5.5: Valve measurement of design III (L80). The pressure was increased until the
ferrules burst at 230 bar. No leakage was observed during the build up.
Design IV (L84) sealed at 1.55 V with the resistance of 2.3 Ω and reached 200 bar without visible leakage, figure 5.6. When the membrane was deactivated the pressure dropped instantaneous.
Design V (L83) sealed at 0.96 V and reached the limit of 200 bar, where the experiment was aborted to save the ferrules, figure 5.7. The resistance was measured to be 2.1 Ω. When the valve was deactivated it was still sealing keeping the pressure constant and had to be lowered to 15 bar to open the valve. The pressure was released by opening the pressure sensor.
The results of the pressure measurements on the valves are summarised in table 5.1.
Table 5.1: Summary of the different designs. *No data due to failures in the ferrules. Design Maximum pressure [bar] Self-seal [bar]
I >200 35-45
II 120 (inside 103)
-III >200 *
IV >200
0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 140 160 180 200 L84 Time [min] Pressure [bar] Flowrate 10 µl min−1 at 1.55 V
Figure 5.6: Valve measurement of the pump design IV (L84). The membrane withstood
200 bar and the experiment was aborted. When the valve was deactivated the pressure dropped instantaneously. 0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 140 160 180 200 L83 Time [min] Pressure [bar] Flowrate 10 µl min−1 at 0.96 V
Figure 5.7: Valve measurement of design V (L83). The membrane reached 200 bar and the
500 550 600 650 700 750 800 850 0 20 40 60 80 100 Time [s] Pressure [bar] 1.2 V
Figure 5.8: Pressure measurements for different voltages.
5.4
Pumping against back-pressure
The pumping experiments against an applied back pressure were conducted according to the method described in section 4.3. Both unclamped and clamped pumping experi-ments were conducted and evaluated. Although clamped pumping is preferred, problems emerged when the clamping pressure was applied at the active parts of the pump. This resulted in that the channel stencil was pressed against the membrane, and thus blocking the flow. To avoid this, the membrane was shaped using a stamp with circular bumps, pressed against the membrane foil before bonding.
5.4.1
Unclamped micropump
This was solved by flushing ethanol through the pump.
As in the pressure measurements the Pharmacia P-3500 was used to build up pressure in the oppsite direction in relation to the pumping direction of the micropump. Two experiments were conducted using a frequency of 0.2 Hz.
At first the L82 was used driven with the voltages 1.3 V, 1.1 V and 1.1 V at the inlet, outlet and pump chamber respectively. The pump had a positive flow up to a back-pressure of 6 bar before the joints between the top-stencil and the channel-stencil broke, figure 5.9. The resistance of the heaters were measured to be 1.8 Ω, 2.2 Ω and 2.0 Ω for the inlet, outlet and chamber respectively.
The second experiment used L83 and managed a back-pressure of 10 bar before the pump started to leak between the top-stencil and the channel-stencil, figure 5.10. The applied voltages was 1.1 V, 1.1 V and 1.3 V at the inlet, outlet and chamber respectively. The resistances were measured to be 1.9 Ω, 2.0 Ω and 2.0 Ω at the inlet, outlet and chamber respectively.
5.4.2
Clamped pumping
To examine the performance of the micropump two variables were evaluated, the applied voltage and the driving frequency.
Varying Voltage
Using the pump of design II (L87), with a shaped membrane to allow clamping without blocking the channels, a positive flow rate was observed for pressures up to 140 bar, figure 5.11. Three measurement series were conducted with different voltage on the heaters, table 5.2. The pump operated at a frequency of 0.4 Hz at 19.5◦C. The resistance of the
heaters was for the outlet, inlet and chamber: 2.4 Ω, 2.2 Ω and 2.1 Ω.
The voltage on the heaters were initially set to 0.95 V and achieved flow rates up to 0.31 µl min−1. The flow rate decreased linearly with the applied pressure and above
100 bar the measured flow rate was unreliable. Looking at the inlet tube it was confirmed that the pump did not move any liquid in the positive direction above 110 bar.
When the voltage was increased to 1.05 V a maximum flow rate of 0.33 µl min−1 was
recorded. The micropump managed a back-pressure up to 130 bar.
At a voltage of 1.15 V the maximum flow rate was further increased to 0.35 µl min−1.
The pump managed a maximum back-pressure of 140 bar. When looking at the inlet tube at 150 bar, the pump was seen having a high positive flow rate although the flow sensor registered a high negative flow rate, indicating leakage in the connections between the pump and the flow sensor.
Varying frequency
0 100 200 300 400 0 0.5 Time [s] Volume [µl] (a) 0 100 200 300 400 0 2 Time [s] Pressure [bar] (b) 0 50 100 150 200 250 300 350 400 −5 −4 −3 −2 −1 0 1 2 3 4 L82 − Flow Measurement Time [s] Flow [µl min−1] (c)
0 100 200 300 400 500 0 0.5 1 1.5 L83 − Pumped volume Time [s] Volume [µl] (a) 0 100 200 300 400 500 0 5 10 15 L83 − Pressure Time [s] Pressure [bar] (b) 0 50 100 150 200 250 300 350 400 450 500 −10 −8 −6 −4 −2 0 2 4 6 8 10 L83 − Flow Measurement Time [s] Flow [µl min−1] (c)
0.65 µl min−1 at 0.6 Hz, measured at the inlet capillary.
Driving the pump at 0.4 Hz the flow rate was with a 95% confidence interval estimated to be 0.38±0.03 µl min−1 within 0-130 bar. At 0.5 Hz the flow rate was estimated to
be 0.49±0.03 µl min−1 within 0-80 bar. Between 0-100 bar for 0.6 Hz the flow rate
was estimated to be 0.63±0.05 µl min−1. At 0.7 Hz the pump managed to pump at
atmospheric pressure at the rate of 0.28 µl min−1, deviating from the linear behaviour
and was back-flowing already at 10 bar.
5.5
Flow and activation of a pump-cycle
The voltage applied on the different membranes were logged in a pumping experiment, using L83, and plotted in a diagram against the flow measured at the same moment, figure 5.13. The voltages and flow were normalised in relation to their maximum values of the measurement. It was seen that the activation of the membrane was not instantaneous, but delayed due to the required heating of the encapsulated paraffin.
5.5.1
Activation time of a steel valve
By applying a constant flow through the pump, the response of the flow was measured when a membrane was activated, figure 5.14. Using equation 2.6, the time-constant τ (response-time in a rough approximation) was calculated to be 1.1 s.
5.5.2
Deflection of outlet membrane
0 50 100 150 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4
Pumping against applied pressure of L87
Pressure [bar]
Flow [µl min−1]
0.95 V 1.05 V 1.15 V
Figure 5.11: The flow measurements for L87 at different applied back-pressures for three
different voltages at the heaters at 0.4 Hz.
0 20 40 60 80 100 120 140 160 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Pressure [bar] Flow [µl min−1] Varying frequency − L87 0.4 Hz 0.4 Hz − Flow sensor 0.4 Hz − Measurement 2 0.5 Hz 0.5 Hz − Flow sensor 0.6 Hz 0.6 Hz − Flow sensor
Figure 5.12: Flow measurements of L87 when varying the frequency at constant voltage
258 259 260 261 262 263 264 265 266 267 268 −1 −0.8 −0.6 Time [s] Normalised voltage/flow Flow Inlet Outlet Chamber
Figure 5.13: The normalised flow in relation to the activation of the three membranes
from a measurement of the pump L83.
415 420 425 430 435 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Time [s] Normalised voltage/flow Activation time Flow Voltage τ
However, there is one type of pump, the electroosmotic, that is capable of pumping against back pressures higher than the micropump in this thesis. Electroosmotic pumps require ionic working fluids in order to function. The pump developed in this thesis is to our knowledge the only micropump that can pump over 100 bar for potentially any solution, which opens up the possibility to use the micropump in HPLC systems.
6.2
Placement of channels towards the membrane
The pumping experiments that reached back pressures above 100 bar were conducted using design II. This design has both its holes at an equal distance from the centre of the membrane. This pump still has the symmetric features of the old design, and is able to pump in any direction. However, if bidirectionality can be omitted, this is not the optimal hole placement considering high pressures. The best sealing is achieved at the centre of the membrane, and by locating the hole facing the outlet in the middle, the ability to hold against high pressure will increase even further. Design III, IV and V are all based on this idea, but have not been fully evaluated due to the self sealing behaviour of the valves.
Valve measurements have also indicated that a higher voltage has to be used when pressurising the membrane from inside. Instead of compressive stresses on the bonds, the bonds will be pulled apart requiring the membrane to have a larger deflection to seal. When the membrane is pressurised from above, an additional support will be achieved by the applied pressure, holding the structure together. The upper limit of the valves was not possible to test due to the limitation of the equipment. However, if clamped properly the inlet valve and outlet valve should both manage similar pressures, well above 200 bar.
6.3
Material selection for the membrane
As a membrane, the ideal material would be a stiff material with a high yield strength. In general, polymers do not have this property, thereby lacking the ability to hold up the strong force exerted by paraffin during its expansion when melted.
Two main candidates were considered: steel, and an alloy of titanium, grade 5 (Ti6Al4V). Grade 5 is a high strength material with half of the Youngs modulus compared to steel. This is a favourable combination, and attempts has been made to acquire this material in a foil as thick as the steel foil used. However, one has to consider the price and avail-ability of the materials. It was soon obvious that grade 5 was too expensive to acquire in a suitable thickness, leaving steel as the most suitable high strength material at the moment. It should be added that after the membrane steel was changed from grade 304 to grade 301, the membrane has never failed even once due to fatigue or tensile stresses. The earlier problems with plastic deformation has become negligible, due to the usage of stainless steel (grade 301) membranes.
6.4
Steel compared to polyimide as a membrane
One of the old problems with the pump before the steel membrane was used was perma-nent deformation of the polyimide membrane. As previously mentioned, this resulted in an irregular stroke, but polyimide had one main advantage. Being flexible compared to steel, the polyimide membrane allows the expanding paraffin to seal along the whole mem-brane making the placement of the holes towards the memmem-brane less important. When using the same hole placement with steel membranes, the membrane appeared to have problems of sealing towards the rim of the membrane, having most of its displacement in the middle.
6.5
Motivation of the new designs
Liquids, including water are in general considered to be incompressible. This implies that if a force is applied in one point of a container filled with water, all other points will feel the same increase of force (Pascal’s law). Using this reasoning for the pump, one aspect of the initially bad performance in pumping against a back-pressure may be explained.
If the pump is considered to be a closed compartment with three outlets, each closed by a freely moving piston, the force applied by the back pressure will be transmitted to the inlet chamber when the outlet membrane and channel membrane is not activated. To be able to pump the inlet membrane must withstand a higher pressure than the back pressure, otherwise the membrane will be pushed open, allowing the fluid to flow through the micropump in the negative direction.
The working principle of the pump is based on the assumption of a non-compressible structure, using a non-compressible working fluid. During normal pumping, using design II and IV, some troublesome discoveries were made. Although the valves of the inlet and outlet could be pressurised well above 100 bar, the pump initially did not manage to pump against more than a back pressure of 3 bar at best. The negative flow rate was pin-pointed to the position where the inlet valve opens after all three membranes are activated. Just as the inlet valve opens, a water column shoots back. The water column volume seemed to be twice as large as the largest stroke measured. Consequently, the pump would, even under optimal conditions, only manage a forward stroke amounting to half the back-shooting volume, resulting in a negative flow rate. This was somewhat surprising, as the valves could manage impressive pressures. Based on the fact that all three membranes are activated, it seems unlikely that the water will originate from outside the outlet. Thus, the water must come from the interior of the pump.
Trapped air within the pump could explain this behaviour. Without pin-pointing the exact location of this air, it seems possible that the smaller space between the channels, at the membrane and the channel stencil (due to the low initial deflection of the membrane) will have the highest flow resistance. The change from crescent-shaped holes to circular holes could also be one reason. However, this small gap will reduce the flow through the channel in the channel stencil, making it very hard to flush out the trapped air. When exposed to the external pressure outside the outlet, the air would be compressed by the pressure. After the outlet valve seal, the air will expand when the inlet opens, and the pressure inside of the pump drops to atmospheric pressure. The compression of the trapped air is proportional to the pressure, which would give a pressure-level where the expanded air would push out liquid in the same rate as the stroke of a membrane, and thereby give the pump no pumping capabilities, or even negative flows.
Different approaches to solve this problem has been attempted (filling of water in vacuum, flushing with the high viscous liquid PluronicTM
6.7
Bonding between channel- and inlet-stencils
The major problem at the moment, seems to be insufficient bonding between the top stencil and the channel stencil. When examining L82 after the pump was disassembled, the parylene layer did not adhere to either of the steel surfaces, and was possible to remove in one unharmed piece, figure 6.1. The reason for this behaviour is at the moment unknown. The parylene-coating has previously proven to give a good bonding between steel surfaces, although some problems have occasionally appeared. For the most recent problems there are some possible explanations: either a residual contamination from the etching of the stencils during manufacturing, or a process mishap during the coating of the parylene layer. There were some problem with the coating process, and the stencils used in this project had been recoated to get the right parylene thickness.
If the problems in creating the parylene coating is neglected for a moment, the bond between the top stencil and the channel stencil is the most harshly treated joint in all of the structure. The fact that this bond always is the first to break (even though the whole pump is bonded using parylene) points toward the relatively low stresses acting on the other joints. If pumping against extreme pressures without clamping will be possible, this joint has to be strengthened. The pumping results of L87 (5.4.2) strengthens this statement. L87 uses the same design and stencils from the same batch as L82 with the exception of having a shaped membrane providing the possibility of using clamping which relieves the stresses at the joints. This improvement along with the removal of the trapped air made it possible to pump against back-pressures up to 150 bar.
Figure 6.1: When the L80 micropump was disassembled, it was found that the parylene
layer had a bad adhesion to the stainless steel. In a good bonding, the parylene is expected to stick to both surfaces and be ripped apart when the stencils are separated.
6.8
Limitations in the equipment
the tube were pushed off the PEEK tubes by the high pressure. This limited the pressure measurements to 200 bar to prevent damaging the fittings.
6.9
Expansion within the system
Looking at the pumping experiment of L83 (section 5.10), the stroke is seen to increase with the increased pressure. This has usually been an expected behaviour with an air bubble within the channels. Before this experiment, the air within the micropump was removed, which was seen to be a success due to all air bubbles pouring out of the outlet. When taking the constant flow rate into account, problems with air seems unlikely. Usu-ally, an air bubble will cause the flow rate to decrease with increased pressure, which is not the case in the experiment. Instead the increased strokes could be explained by a more responsive system at higher pressures. At lower pressures expansion within tubes and air may even out some of the flow reaching the sensor. However, this explanation cohere badly with the observations. When the pressure-limit was reached the pump started to leak in the bond between the channel stencil and the top stencil (see section 6.7). This implies that the increased stroke may be caused by an expansion in the parylene bond, yielding a larger stroke due to the larger cavity above the membrane and possibility of the membrane to deflect over its relaxed position. Evidently the stresses in the bond became to large resulting in a failure just below 10 bar.
6.10
Self-sealing valves
During the measurements of the new designs, a new and unexpected behaviour was in-troduced at fairly low pressures. Some of the valves became self-sealed, and managed to withhold the pressure when deactivated. Only a small declination of the pressure was noticed, probably due to a small back-flow in the external pump (P-3500) used to build up the pressure, or a small leakage in some of the connections. There are currently two separate ideas explaining this behaviour:
which will be pushed out towards the edges sealing the adjacent opening. When the heaters is turned off the paraffin closest to the rim and furthest away from the heaters will solidify first, and the valve may withhold the shape even when the membrane is deactivated.
The second explanation is that the high pressure pushes the channel and top stencil against the membrane, which will block the flow. This seems to be a less probable explanation, because of the thick stencils used in comparison to the membrane thickness, and size of the features. There has been no detailed investigation of this problem, although it is a nice feature of the valves that they can be used as both active valves, but also having the possibility to seal when turned off. When used in this way, the usually highly energy consuming valves may be an energy efficient solution.
6.11
Blocked channels
One major obstacle in achieving the high pressure pumping has been (beside bad bonding) the problems in using clamping. The clamping holds the pump together when pressurised, and prevents the micropump from bending with applied pressures. Looking at the cross section of the pump (figure 2.1), the gap at the channel towards the rim of the membrane is small compared to the maximum deflection in the middle. One should consider that the structure shown in the picture is a polyimide membrane, which has a deflection five times larger than the deflection of a steel membrane (typically the steel membrane is deflected approximately 10 µm in the middle). At the rim, the deflection is much smaller (approximately just one or two µm at best), and it does not require much to seal this gap. The problem with blocked channels was solved by shaping the membranes, allowing clamping to be used.
6.12
Plateau in pressure measurements
During the pressurisation of valves a strange drop of 1-5 bar appears in some of the measurements. No leakages was noticed during that moment and afterwards the pressure build-up continues with the same inclination as before. What this movement is caused by is unclear although the best explanation is that the PEEK tube is moved in the connection.
6.13
The activation time
measurement was conducted at that frequency. The flow rate was increased compared to the first measurement. Probably due to harsh treatment of the pump, being pressurised and forced to pump at pressures up to 150 bar several times. This could have slightly deformed the membranes, which would have increased the pumped volume.
6.15
Future improvements
Most of the pumps have been pressurised several times at high pressures and survived unharmed, sealing at the same voltage as before. There are occasions where the microp-ump have been running for days, and when examined afterwards showed to be relatively unaffected. This indicates that except for some minor flaws, the micropump has a durable design, and when a micropump has failed, it has either been the bad bonding at the top stencil or a paraffin leakage at the heaters. The potential of the pump is great, and with some adjustments the performance will be increased even more.
To even out the flow of the pump, several pumps could be connected in parallel. The micropump has a well defined stroke, and would allow for a precise control of the dispensed amount of liquid with an even flow.
The driving sequence of the pump is far from optimised. Due to the many changes of design, and the requirement for a characterisation of each new pump, this would be pointless before the final pump-design is set. Due to this, the flow rates and back-pressures possible to manage today is below the actual potential of the micropump.
To address the problem of being unable to use clamping, the location inlet holes can be changed at the top stencil. By moving the holes from above the cavities, towards the edges of the top stencil, hopefully the channels will be prevented from getting blocked when the ferrules are tightened and the clamping is applied. Another interesting possibility will be to glue fittings directly onto the top stencil, making it possible to evaluate the pump without the usage of a fixture.
Conclusions
The micropump was redesigned and pumping against back-pressures above 100 bar was performed, which was the aim of this thesis. Pressure above 100 bar is a prerequisite for pumps that are going to be used in HPLC, and the potential of reaching even higher back-pressures has been showed during valve measurements.
The previous problem with plasticity of the membranes has been reduced. Even though pressurised above 100 bar several times, no severe damage of the membrane was noticed. Five different valve-designs of the micropump have been measured to withstand pres-sures above 200 bar. Some of the new valve designs were shown to be self-sealing above a certain pressure when pressurised unclamped, managing to withhold the applied pressure when the voltage was turned off.
A new design of the micropump, with circular holes equally distributed from the membrane centres, was able to pump against applied back-pressures up to 150 bar with applied clamping (applied support). Unclamped the pump was recorded to pump against a pressure up to 10 bar.
The weakest link up to date in the micropump has been pin-pointed to be the bond between the channel stencil and the top stencil.
A major problem in the processing technique became apparent. Residual air-pockets after manufacturing initially limited the pumping capabilities to just a few bar. This was solved adding an extra process step: priming the channel with ethanol.
To prevent leakages in the connection to the micropump when reaching high pressures, the PEEK fittings towards the pump had to be replaced by metal fittings. This enabled pressures up to 200 bar without leakages.
Finally I would like to thank the Division of Material Science, The Department of
Engineering Sciences at the ˚Angstr¨om Lab for the welcoming and friendly environment,
being a good foundation of creativity.
[1] Roger Boden, Klas Hjort, Jan-Ake Schweitz, and Urban Simu. A metallic micropump for high-pressure microfluidics. Journal of Micromechanics and Microengineering, 18(11), 2008.
[2] D J Laser and J G Santiago. A review of micropumps. Journal of Micromechanics
and Microengineering, 2004.
[3] A. Nisar, Nitin Afzulpurkar, Banchong Mahaisavariya, and Adisorn Tuantranont. Mems-based micropumps in drug delivery and biomedical applications. Sensors and
Actuators B, 2007.
[4] Anders Olsson, G¨oran Stemme, and Erik Stemme. A valve-less planar fluid pump with two pump chambers. Sensors and Actuators, 1995.
[5] Carl H Hamann, Andrew Hamnett, and Wolf Vielstich. Electrochemistry, pages 128– 130. Wiley-VCH, 2nd edition, 2007.
[6] Lingxin Chen, Yafeng Guan, Jiping Ma, Guoan Luo, and Kehui Liu. Application of a high-pressure electro-osmotic pump using nanometer silica in capillary liquid chromatography. Journal of Chromatography, pages 19–24, 2005.
[7] Lingxin Chen, Yafeng Guan, Jiping Ma, Xin Shu, and Feng Tan. Theory, controls parameter and application of the packed-bed eletroosmotic pump. Chinese Science
Bulletin, 48:2572–77, 2003.
[8] Xiaoyue Zhu, Leonard Yi Chu, Bor han Chueh, Mingwu Shen, Bhaskar Hazarika, Nandita Phadke, and Shuichi Takayama. Arrays of horizontally-oriented mini-reservoirs generate steady microfluidic flows for continuous perfusion cell culture and gradient generation. Analyst, 129:1026–31, 2004.
[9] Daniel C. Harris. Quantitative Chemical Analysis, pages 713 – 718. W. H. Freeman and Company, New York, fifth edition, 2000.
[10] Marvin C. McMaster. HPLC A practical user’s guide, pages 3–9. John Wiley and Sons, Inc., 2nd edition, 2007.
[11] Marek Noga, Filip Sucharski, Piotr Suder, and Jerzy Silberring. A practical guide to nano-lc troubleshooting. Journal of Separation Science, 30(14):2179–2189, 2007.
[16] Stefan Johansson. Actuator materials and microactuation. Department of Engineer-ing Sciences, Uppsala.
[17] Carl Nordling and Jonny ¨Osterman. Physics Handbook for Science and Engineering, page 183. Studentlitteratur, Lund, seventh edition, 2004.
[18] F Goldschmidtb¨oing, A Doll, M Heinrichs, P Woias, H-J Schrag, and U T Hopt. A generic analytical model for microdiaphragm pumps with active valves. Journal of
Micromechanics and Microengineering, pages 673–683, 2005.
[19] Carl Nordling and Jonny ¨Osterman. Physics Handbook for Science and Engineering, page 224. Studentlitteratur, Lund, seventh edition, 2004.
[20] Lennart R˚ade and Bertil Westergren. Mathematics Handbook for Science and
Engi-neering, page 74. Studentlitteratur, Lund, fifth edition, 2004.
[21] Carl H Hamann, Andrew Hamnett, and Wolf Vielstich. Electrochemistry, pages 227– 230. Wiley-VCH, 2nd edition, 2007.
[22] Clemens van Blitterswijk. Tissue Engineering, page 256. Academic Press, 2008. [23] Jr William D. Callister. Materials Science and Engineering An Introduction, pages
211–214. John Wiley and Sons, Inc., 2003.
[24] Mario Di Giovanni. Flat and corrugated diaphragm design handbook. Marcel Dekker, inc, 1982.
