Fabrication of integrated micromachined polymer magnet

Olarn Rojanapornpun and Chee Y. Kwok

School of Electrical Engineering and TelecommunicationsUniversity of New South Wales, Sydney, Australia

ABSTRACT

In this paper, the technique for fabricating integrated polymer magnet suitable for microelectromechanical sys-tem (MEMS) is discussed. Spin coating and screen printing techniques have been reported by other researchers,however, obtaining small feature sizes and good alignment accuracy are difficult. Furthermore, from our exper-iments, spin coating of composite material results in unacceptable surface roughness and it is difficult to obtainhigh powder loading. The paper discusses the technique of implementing ‘template printing’ where the sacri-ficial layer is used to form the template for polymer magnet structure. Instead of dispersing magnetic powderinto the matrix material, magnetic powder is filled into template follow by embedding of matrix material. Thistechnique has the advantages of improved alignment accuracy, minimum structure size achievable, and highmagnetic powder loading. Strontium ferrite/polyimide composite structure has been fabricated. Integratedpolymer magnet as small as 100 µm in diameter has been achieved. Alignment is only limited by alignment ofthe template using mask aligner used by typical photolithography step. Volume loading of 65%, Residual mag-netic flux density Br of 0.21 T, Intrinsic coercivity Hci of 325 kA/m, and maximum energy product (B·H)max

of 8145 T A/m were achieved.

Keywords: Composites, polymer magnetic, polyimide, ferrite, micromachining

1. INTRODUCTION

In the early days of MEMS devices, external magnets had been used in many magnetic devices due to thesimplicity in transition from macro world to micro world. These external magnets were attached to the de-vices by mean of adhesive. To further miniaturise the device, external magnet has to be scaled down whichincreases difficulties in handling and manufacturing. Accuracy in positioning the external magnet also becomesmore important. Therefore, fabrication technique capable of fabricating fully integrated magnetic structure inmicrometer scale is required to realise truly miniaturised devices.

Electroplating1–4 and polymer magnet1, 5–7 fabrication had been used to fabricate integrated thick mag-netic structures suitable for microsensors and microactuators. These two techniques complement each otherin term of magnetic properties. Polymer magnet fabrication has the potential of producing high materialresistivity (suitable for high frequency devices) and high coercivity (in case of permanent magnet).

Spin coating and screen printing of uncured composites (consist of magnetic particles and polymer matrices)had been reported1, 5–7 to be compatible with MEMS processing. In this work, both processes will be evaluatedand the new technique called ‘template printing’ capable of producing fine structures with high alignmentaccuracy will be proposed.

2. SPIN COATING OF MAGNETIC COMPOSITES

2.1. Powder loading concentration

Filler loading concentration was defined in term of percent by weight of the total composition of compositematerial.8 This number is used by the plastic industry to calculate material cost. However, in this work,we emphasise on the magnetic properties of polymer magnet which should be related to the volume packingof magnetic particles and voids between particles. Therefore, particle loading by volume will be used as thereference property.

Further author information: E-mail: olarnr@ee.unsw.edu.au, Telephone: 61 2 9385 5395

Table 1: Properties of magnetic powders.

HM170 72500

specify gravity (g/cc) 5.07 5.20

average fisher particle size (µm) 1.39 0.98

average coulter particle size (µm) NA 3.56

coulter particles sizes range (µm) NA 0.4–7.0

Table 2. Properties of polyimide.

PI-2555

solids content (%) 18.8

viscosity (Poise) 13.4

density (g/cc) 1.072

cured density (g/cc) 1.42

Table 3: Properties of titanate coupling agents.

KR-TTS LICA-44

Chemical Description (%) isopropyl triisostearoyltitanate

neopentyl(diallyl)oxy,tri(N-ethylenediamino)ethyl titanate

viscosity (Poise) 1.25 100

density (g/cc) 0.95 1.17

In order to obtain high magnetic properties, high volume loading is required. The maximum achievablevolume loading of polymer magnet depends on many factors such as, shape, size, and size variation of theparticles,8 interaction between particles, etc. As with particles generated by ball milling, there will be variationin particle sizes. Hence, the calculation of maximum volume loading is complicate and will not be studied inthis work.

The matrix material used in this work is PI-2555 polyimide from HD MicroSystems. Supplied in liquidform, it consists of 18.8% solid content and N-Methyl-2-Pyrrolidone/Aromatic Hydrocarbon solvent. To formrigid, intractable polymer, PI-2555 is thermally cured at 350 ◦C to completely drive off solvents and completeimidization of polyimide. Since the solid content of PI-2555 is very small, the volume of cured PI-2555 is severaltimes smaller than the original liquid form volume.

At high volume loading, as the particles are tightly packed, voids between particles cannot be shrunk asmuch as the shrinkage of PI-2555 during thermal curing. Therefore, it’s likely that there will be air voidsinside the fully cured composite. Since air and polyimide have the same magnetic permeability, the inclusionof air voids has no effect on magnetic properties of polymer magnet. However, volume of air voids should beconsidered in the calculation of powder volume loading of polymer magnet.

2.2. Composite preparation

Material used in this work comprise of Strontium ferrite (SrFe12O19) powder HM170 from Hoosier magneticsgroup, Nickel Zinc ferrite 72500 from Steward company, polyimide PI-2555 from HD Microsystems, and couplingagent (KR-TTS and LICA-44) from Kenrich Petrochemicals, inc. KR-TTS is the standard coupling agent forferrite and LICA-44 has amino chemistry as does polyimide. Table 1–3 show properties of HM170/72500,PI-2555, and KR-TTS/LICA-44 respectively.

Sample A composed of 10.5% volume loading of 1% LICA-44 treated 72500 powder in uncured PI-2555.If there is no air void created in sample A after curing, volume loading should be 45%. 1% (by weight of thepreparing 72500 powder) LICA-44 was diluted in Isopropanol Alcohol (IPA). 72500 powder was treated bydispersing into diluted LICA-44 using SPEX8000 paint mixer (provides vigorous shake in three dimensions)for 10 min. 3 12 mm ceramic balls (Coorstek AD-85) were added into sample A container before mixing, toaid the dispersion. IPA was allowed to evaporate leaving a dry 72500 powder. Ball Milling was then used toslowly disperse treated 72500 powder into PI-2555 for 50 h at around 40 rpm. 72500/PI-2555 composite wasspin coated onto silicon wafer at 3000 rpm for 40 s. The composite film was then harden by baking at 120 ◦C.

Figure 1. SEM micrograph of 45% volume loading72500/PI2555 composite film fabricated by spin coating.

Figure 2. SEM micrograph of 54% volume loadingHM170/PI2555 composite film fabricated by spin coating.

Sample B composed of 14% volume loading of 1% KR-TTS treated HM170 powder in uncured PI-2555.After full curing of polyimide (assuming that there is no air void created in sample B), volume loading shouldbe 54%. Xylene was used as the solvent for KR-TTS. Sample B composite is more viscous than sample A andball milling did not seem to disperse HM170 properly. Dispersion was done by SPEX8000 for 75 minute instead.Sample B composite was then spin coated on silicon water at 3000 rpm for 60 s and baked at 120 ◦C.

2.3. Results and discussion

Composite film from sample A is reasonably uniform in thickness. Upon close inspection of the film surface,there were many agglomerates (Figure 1) of various sizes. Cross section views of the agglomerates show thatthere are more coarse and medium sizes particles at the base than the peak. When sample A was spin coated atlower speed (2000 rpm) for a shorter period (30 s), there was less number of large and medium sizes agglomerates.However, there are still small size agglomerates. It seems that agglomerates cannot be eliminated completely.There is one possible explanation, as the 72500 particle can be as large as 7 µm, during spin coating, manyparticles that had been moved by outward force may be blocked and agglomerate on large particles. Withprolong spinning, further agglomeration occurs.

Since the agglomeration may caused by the variation of particle sizes, a more uniform size powder is prefer-able. It has been suggested9 that the reduction in variation cannot be accomplished by sieve technique due tovery small sizes of particles. The only technique that may work is air classification.

Composite film from sample B is not smooth (Figure 2), which could be due to the high viscosity of thecomposite. There are some areas where there is no powder, which should be the results of bubbles generatedby SPEX8000 paint mixer. Note that PI-2555 is very viscous, which is why volume loading of more than 50%is not suitable for spin coating. It is possible to reduce the viscosity of PI-2555 polyimide by thinning withHD Microsystems T-9039 thinner (composes of N-Methyl-2-Pyrrolidone and Propylene Glycol Methyl Ether).However, sedimentation of powder occurs with thinned PI-2555 composite. Spin coating has to be done rightafter dispersion.

3. TEMPLATE PRINTING

Screen printing is based on paint printing in the macro world. Therefore, alignment of the screen to thedevices may not be very accurate. In certain applications, alignment of polymer magnet structures may be veryimportant. As the minimum feature size for screen printing technique depends on patterning of screen in use,obtaining very small magnets can be very difficult to achieve.

In order to improve alignment accuracy and minimum feature size capability, fabrication techniques thathave been used in MEMS field are adopted. Template is used to replace the screen. It can be fabricated by

photolithography technique to provide good feature size and alignment accuracy. Thick polymer magnets canbe achieved by using thick template, fabricated by processing techniques, such as, LIGA10 process or thickphotoresist process (Microlithography Chemical Corporation SU-8,11, 12 Clariant AZ456211, 13 and AZ9260,14

Micro Resist Technology ma-P 10011). With these techniques, the thickness of template is easy to control andthe thickness of more than 500 µm can be achieved. Depending on the requirement of the application, templatecan be removed after polymer magnet is formed, by wet etching or plasma etching.

Since homogenised dispersion of magnetic powder for high volume loading composite is difficult to achieve,an alternative technique will be used to fabricate polymer magnets. Instead of mixing magnetic powder withmatrix polymer and then fill the composite into the template, magnetic powder is filled into template andfollowed by filling of binding polymer. Since the powder does not have to be dispersed into matrix materialmechanically, coupling agent treatment can be omitted.

3.1. Experiment

Many polymers with low viscosity can be used as binding material. Since PI-2555 and Photosensitive polyimidePI-2732 from HD Microsystems are readily available in our laboratory, they will be used in this work. PI-2732was spin coated onto silicon wafer at 2000 rpm for 30 s and soft-cured at 110 ◦C for 10 min. The second layer ofPI-2732 was then spin coated with the same parameters and soft-cured at 110 ◦C for 20 min. Since PI-2732 isnegative acting, it was exposed to UV light without mask for 30 s. PI-2732 film had been cured using hot platein air up to 200 ◦C. 6000 A thick titanium was then evaporated onto PI-2732 film. Ti film was patterned usingstandard photolithographic steps to form the mask for plasma etching of PI-2732 film. Plasma etching wasdone under 60 W of RF power with gas mixture of 3.5 Pa O2 and 0.35 Pa CF4 until the underlying Si surfaceis exposed. This PI-2732 template consists of various circular patterns from 100 µm to 1 mm in diameter. Thethickness is around 35 µm.

In order to improve adhesion between PI-2555 and magnetic particles, HM170 was treated in 0.25% HDMicrosystems VM651 (α-amino propyltriethoxysilane) in De-ionised (DI) water. HM170 was stirred continuouslyfor 2 min. Excess VM651 solution was then removed and HM170 slurry was dried in oven at 90 ◦C for 30 min.

PI-2732 template wafer was treated with 0.5% VM651 water solution prior to filling by the VM651 treatedHM170, so that the fabricated HM170/PI-2555 polymer magnet will adhere to the silicon. Plastic spatula wasused to pack the treated HM170 powder into the template. Excess powder outside the template was removedby shaving the surface of template wafer with flat edge of the spatula. The result is a clean wafer surface,relatively flat surface of powder filled area, and completely filled template.

As the viscosity of PI-2555 is relatively high, PI-2555 was thinned by T-9039 (2:1 PI-2555:T-9039 by weight),so that it can flow easily to replace air voids between particles. The wafer was slanted slightly to assist theflow of thinned PI-2555 into HM170 filled template. This will reduces the amount of polyimide required andassists in gently removing of air voids in HM170 filled template. If polyimide is filled in too quickly, air voidsmay be trapped. During curing of polyimide, this trapped air may expand and leave the template, resulting insurface defects on magnetic structure surface. After the filling of thinned PI-2555 is completed, the polyimidewas cured in vented oven at up to 200 ◦C in air.

The excess PI-2555 layer outside template was removed by plasma etching (60 W RF power, 3.5/0.35 PaO2/CF4 gas mixture). Since O2/CF4 gas mixture does not attack magnetic ferrite significantly and plasmaetching conditions used exhibits anisotropic etching, the PI-2555 matrix inside the template is protected by thelayer of top particles. Titanium etchant (with HF as main component) was used to remove titanium layer. Toinspect the side wall of the fabricated HM170/PI-2555 composite, PI-2732 template was removed by plasmaetching.

3.2. Results and discussion

Figure 3 and 4 shows cylindrical HM170/PI-2555 polymer magnet with 100 and 850 µm in diameter and 35 µmin thickness fabricated by template printing. As can be seen, the structures are completely filled with HM170and followed the shape of templates. Top surfaces of these magnets are reasonably good considering that they

Figure 3. SEM micrograph of 100 µm × 35 µmHM170/PI2555 composite structure fabricated by tem-plate printing.

Figure 4. SEM micrograph of 850 µm × 35 µmHM170/PI2555 composite structure fabricated by tem-plate printing.

are composed of 1–3 µm particles. No agglomerate is visible. Structures with feature sizes ranging from 20 µmto 1 mm had been successfully fabricated.

The processing required can be simplify significantly by the use of high-aspect ratio thick photosensitivematerial as template and low curing temperature polymer matrix. Note that template material has to bechemically resistance to solvent (if any) in polymer matrix and relatively tough physically as ferrite powders areabrasive. If the selected polymer matrix contains solvent that attacks photosensitive material or requires highcuring temperature, electroplated metal template can be used. Metal wet etchant that does not attack polymermatrix chosen can be used to remove the metal template.

It is obvious that, using this technique, only high powder loading magnet can be fabricated. In most cases,good magnetic properties are desirable, and therefore it should be applicable in most applications. Power loadingcannot be controlled because it depends on many factors including pressing pressure and particles orientationduring filling. Therefore there will be variation in magnetic properties of fabricated magnets even though theyhave been fabricated with the same conditions. The variation has not yet been studied as the magnets fabricatedare very small and cannot be measured easily.

3.3. Magnetic properties

In measuring the magnetic properties of polymer magnets, the Vibrating Sample Magnetometer (VSM) systemrequires sample sizes much larger than those fabricated in the previous section. A special template was con-structed from two 2 mm thick aluminium sheets. The top sheet was milled to create a 2×10 mm slot. Thetwo sheets are attached together by bolts and nuts. HM170 was treated with VM651 the same way as normaltemplate process. However, no VM651 was applied onto the Al template as the magnets have to be removedfor measurement. 10 mm wide, 2 mm thick Al sheet was used to compact the treated HM170 in the slot. Thematrix used is 2.5:1 by weight of PI-2555/T9039 mixture. Since the sample magnets will be 2 mm thick, thefilling of powder/packing/filling of polyimide were done in multiple small steps. Sample C was prepared withoutexternal magnetic field, whilst the easy axes of HM170 particles in sample D were magnetically aligned along10 mm length by an external field of 0.15 T before packing and filling of polyimide. The filling of HM170 wasdone outside magnetic field because otherwise free HM170 particles will all be attracted by external magnet.

Sample C and D were removed from templates after soft-cured at 120 ◦C. Both samples were then curedslowly up to 200 ◦C in air and 350 ◦C in nitrogen atmosphere. These samples are physically stronger than fullycured KR-TTS treated HM170/PI-2555 composite.

Table 4. Magnetic properties and volume loading of strontium ferrite polymer magnetsfabricated by template printing and screen printing.

sample C sample D 65%6 80%6 bulk ferrite6

powder volume loading 68% 65% 65% 80% NA

particle alignment no yes unknown unknown NA

Br (T) 0.150 0.213 0.2 0.28 0.36

Hci (kA/m) 314 326 ∼ 320 ∼ 320 > 320

(B·H)max (T A/m) 3 870 8 150 6 370 11 900 25 000

The powder volume loading can be calculated from mass and volume of fabricated magnet. Since the sizes ofthese samples are relatively small, vernier and microscope were used to measure the dimensions. As the magnetscannot be made perfectly flat and rectangular block shape, there will be errors in dimensions measurement.The error of just 100 µm on each dimensions can results in 18% error in total volume. However, exact volumemeasurement can be very difficult. To calculate the volume loading, first assumes that air voids between particleswere replaced by liquid matrix material completely and that liquid matrix material only exists in those voids.The effective density of cured matrix material can be calculated from

ρcured eff =mcured

Vuncured=

solid content×muncured

Vuncured

= solid content× ρuncured

where muncured and mcured are masses of matrix material before and after cure. Vuncured is volume of matrixmaterial before cure and assumes that Vuncured = Vcomposite −Vpowder. ρuncured is the density of matrix materialused. Solid content is the solid content (by weight) of matrix material. Powder volume loading can be calculatedfrom

ρcomposite = VLOAD× ρpowder + (1−VLOAD)× ρcured eff

therefore VLOAD =(

ρcomposite − ρcured eff

ρpowder − ρcured eff

)

where ρcomposite is the density of cured composite. Note that the actual density of cured polymer matrix hasno effect on the volume loading after cure.

VSM measurements of Sample C and D were conducted at CSIRO Telecommunications & Industrial Physics,Linfield. The magnetic properties were measured along the length of the samples. The measured data are mag-netic moments m (emu) of test samples induced at various applied magnetic field. Magnetisation 4πM(T) can becalculated from 4πm/104V where V is the volume of test samples. From the magnetisation curves (Figure 5), theintrinsic coercivities Hci (A/m) and residual flux densities Br (T) can be obtained from the negative X-axis andpositive Y-axis intersections, respectively. Energy product B·H(T ·A/m) can be calculated from the productof flux density B = 4πM(T) + µ0H(A/m) and the applied field H (A/m). Maximum energy product (B·H)max

can be determined from negative peak of energy product curve (Figure 6) derived from second quadrant of B-Hcurve (demagnetisation curve).

Figure 5 and 6 show that alignment of magnetic particles to required magnetisation direction is very impor-tant to achieve high residual flux density and energy product. Table 4 shows summarised properties of sampleC and D, compared to bulk ferrite and composites reported by other work. Note that the values given forsample C and D have not been compensated for demagnetisation factors of the magnets due to difficulties incalculations. As can be seen from table 4, magnetic properties of sample D compared quite well with otherwork.

Figure 5. Magnetisation 4πM versus applied magneticfield H of samples fabricated by template printing.

Figure 6. Energy product B·H versus magnetic field den-sity B of samples fabricated by template printing.

4. CONCLUSION

It has been shown that spin coating is not suitable for fabrication of high volume loading integrated polymermagnet due to it’s problem with composite viscosity, difficulty and time consuming of composite dispersion,agglomeration, and patterning. Template printing had been developed to provide superior structure definingand aligning than screen printing. Magnetic features as small as 20 µm had been fabricated successfully. Themagnetic properties achieved are comparable to other techniques. Other template and matrix materials will bestudied to simplify process steps required.

ACKNOWLEDGMENTS

The authors acknowledge the material samples provided by Hoosier magnetics group, Steward company andKenrich Petrochemicals, inc. The authors would like to thank Dr. Eric Gauja, Dr. Tom Puzzer, Prof. ChrisSorrell, and John Sharp for their assistance. The authors would also like to thank Dr. Stephen Collocott,CSIRO, Linfield, for conducting the VSM measurements.

REFERENCES1. J. Y. Park and M. G. Allen, “Development of magnetic materials and processing techniques applicable to

integrated micromagnetic devices,” Journal of Micromechanics & Microengineering 8(4), pp. 307–16, 1998.Publisher: IOP Publishing, UK.

2. W. P. Taylor, M. Schneider, H. Baltes, and M. G. Allen, “Electroplated soft magnetic materials for mi-crosensors and microactuators,” Tranducers 97. 1997 International Conference on Solid-State Sensors andActuators. Digest of Technical Papers (Cat. No.97TH8267). IEEE. Part vol.2, 1997 , pp. 1445–8 vol. 2. NewYork, NY, USA. Proceedings of International Solid State Sensors and Actuators Conference (Transducers’97). Chicago, IL, USA. IEEE Electron Devices Soc. 16-19 June 1997.

3. H. J. Cho and C. H. Ahn, “A novel bi-directional magnetic microactuator using electroplated permanentmagnet arrays with vertical anisotropy,” Proceedings IEEE Thirteenth Annual International Conferenceon Micro Electro Mechanical Systems (Cat. No.00CH36308). IEEE. 2000 , pp. 686–91. Piscataway, NJ,USA. Proceedings IEEE Thirteenth Annual International Conference on Micro Electro Mechanical Systems.Miyazaki, Japan. IEEE Robotics & Autom. Soc. Micromachine Center. 23-27 Jan. 2000.

4. E. Fullin, J. Gobet, H. A. C. Tilmans, and J. Bergqvist, “A new basic technology for magnetic micro-actuators,” Proceedings MEMS 98. IEEE. Eleventh Annual International Workshop on Micro Electro Me-chanical Systems. An Investigation of Micro Structures, Sensors, Actuators, Machines and Systems (Cat.No.98CH36176). IEEE. 1998 , pp. 143–7. New York, NY, USA. Proceedings IEEE Eleventh Annual Inter-national Workshop on Micro Electro Mechanical Systems An Investigation of Micro Structures, Sensors,Actuators, Machines and Systems. Heidelberg, Germany. IEEE Robotics & Autom. Soc. ASME DynamicSyst. & Control Div. Ministr. Econ. Affairs of Baden-Wuttemberg. 25-29 Jan. 1998.

5. L. K. Lagorce, O. Brand, and M. G. Allen, “Magnetic microactuators based on polymer magnets,” Journalof Microelectromechanical Systems 8(1), pp. 2–9, 1999. Publisher: IEEE, USA.

6. L. K. Lagorce and M. G. Allen, “Magnetic and mechanical properties of micromachined strontium fer-rite/polyimide composites,” Journal of Microelectromechanical Systems 6(4), pp. 307–12, 1997. Publisher:IEEE, USA.

7. B. M. Dutoit, P. A. Besse, H. Blanchard, L. Guerin, and R. S. Popovic, “High performance micromachinedsm/sub 2/co/sub 17/ polymer bonded magnets,” Sensors & Actuators A-Physical A77(3), pp. 178–82,1999. Publisher: Elsevier, Switzerland.

8. T. H. Haberberger, “Magnetic fillers,” in Handbook of fillers for plastics, H. S. Katz and J. V. Milewski,eds., Van Nostrand Reinhold Company Inc., New York, 1987.

9. Scott Smith (Steward Co.) : private communication.10. E. W. Becker, W. Ehrfeld, P. Hagmann, A. Maner, and D. Munchmeyer, “Fabrication of microstructures

with high aspect ratios and great structural heights by synchrotron radiation lithography, galvanoforming,and plastic moulding (liga process),” Microelectronic Engineering 4(1), pp. 35–56, 1986. Netherlands.

11. B. Loechel, “Thick-layer resists for surface micromachining,” IOP Publishing. Journal of Micromechanics& Microengineering 10(2), pp. 108–15, 2000. UK. 10th Micromechanics Europe Workshop (MME’99). Gifsur Yvette, France. 27-28 Sept. 1999.

12. M. Despont, H. Lorenz, N. Fahrni, J. Brugger, P. Renaud, and P. Vettiger, “High-aspect-ratio, ultrathick,negative-tone near-uv photoresist for mems applications,” Proceedings IEEE. The Tenth Annual Interna-tional Workshop on Micro Electro Mechanical Systems. An Investigation of Micro Structures, Sensors,Actuators, Machines and Robots (Cat. No.97CH36021). IEEE. 1997 , pp. 518–22. New York, NY, USA.Proceedings IEEE The Tenth Annual International Workshop on Micro Electro Mechanical Systems. An In-vestigation of Micro Structures, Sensors, Actuators, Machines and Robots. Nagoya, Japan. IEEE Robotics& Autom. Soc. ASME Dynamic Syst. & Control Div. Micromachine Center. 26-30 Jan. 1997.

13. S. Roth, L. Dellmann, G. A. Racine, and N. F. de Rooij, “High aspect ratio uv photolithography forelectroplated structures,” Journal of Micromechanics & Microengineering 9(2), pp. 105–8, 1999. Publisher:IOP Publishing, UK.

14. V. Conedera, B. Le Goff, and N. Fabre, “Potentialities of a new positive photoresist for the realizationof thick moulds,” Journal of Micromechanics & Microengineering 9(2), pp. 173–5, 1999. Publisher: IOPPublishing, UK.