《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_Properties of Laminated Silicon Nitride

J.Am. Ceram.Soe,88p02713-2721(2005 DOI:10.1.151-29162005.00512.x ournal c 2005 The American Ceramic Society Processing, Microstructure, and Properties of Laminated Silicon Nitride Stacke Bernd Bitterlich and Jurgen G. Heinrich* Department for Engineering Ceramics, Institute for Nonmetallic Materials, Clausthal University of Technology Clausthal-Zellerfeld Germany By lamination of silicon nitride tapes, components with demonstrated mainly with binder syste at are dissolved in geometries can be produced. Unstructured tapes can 6 Becau the complete mass loss of the nated by common thermal compressi tructured t anic binders during binder burnout, porosity between the tap d by pressureless develops, which can still be present after sintering, thus influ- e.g. pastes as lamination aids because deformation of the struc- encing the pi Some authors report that the properties of tures would occur. These pastes usually contain a a laminated stack depend on the direction of measuring, i.e maintaining the mechanical contact between the t arallel or perpendicular to the layers. Therefore, a ceramic event the loss of binders during burnout, pre-ceramic polymers were use ork. These ceramic precursors convert partly into an con nitride tapes. These precursors have already been used material during heat treatment with a significant reduced mass successfully as binders for processes like injection molding loss compared with common organic binders. Thus, the porosity In this work, aqueous binder systems have been used to in the interlayer of a laminated stack is strongly decreased manufacture silicon nitride tapes. These tapes have been lami which should be favorable for the mechanical and thermal prop- nated by using pastes on the basis of a ceramic precursor and erties. This work discusses the resulting microstructure, Iso by the compression method for comparison. The rheolog strength, and thermal diffusivity data of stacks laminated with ical characteristics of the pastes, their behavior during screen pastes containing various precursor contents. These results are printing, the thermal treatment, the resulting microstructure, the compared with those obtained by samples prepared by compres- trength, and the thermal conductivity of sintered stacks parallel sion of green tapes. It is found that except for some large pores, and perpendicular to the laminating direction will be discussed the microstructure of the precursor-derived interlayers is qual- itatively the same as in the tape material For stacks made by both lamination methods, strength measurements reveal that the Il. Experimental Procedure properties parallel and perpendicular to the layers are different It is shown that the same strength level can be obtained both by (1) Materials using the pressureless route and by the compression metho a direct nitrided silicon nitride powder (Silzot HQ, SKw Unlike the strength, the thermal conductivity does not change Trostberg Metallchemie, Trostberg, Germany) was homoge- with the direction of measuremen nized in water with 6 mass% yttria(Meldform metals, Royston UK) and 4 mass% alumina (CT3000 SG, Alcoa Chemie, hafen, Germany) using an attrition mill (Lme 1. Net L. Introducti zsch, Selb, Germany). The data of the as-delivered and milled silicon nitride powders are summarized in Table I. For tape S ICON nitride is a candidate material for engine applications casting, the powders were dispersed in water with a combination cause of the interesting combination of thermal and me of an amino alcohol and a synthetic polyelectrolyte(dolapix chanical properties that it possesses. The starting powders, sin- A88 and PC 33. Zschimmer Schwarz Chemische Fabriken tering, and machining are very cost intensive and prevent serial application. For prototyping of complex-shaped components raspum con, Zschimmer& Schwarz Chemische Fabriken)and the effort of machining can be reduced by tape casting and lam- as a binder, a commercial polymer emulsion(Mowilith DM 765. inating. Tape casting is a widespread and cost-effective method Clariant, Frankfurt, Germany)were used. Because of the dom- to produce fat, large area components with high quality. B ating influence of the binder emulsion, the slurry exhibited a structuring and laminating tapes, complex-shaped components stinctive shear-thinning behavior. Negligible thixotropy was can be manufactured. Lamination can be carried out by thermal observed. Tapes were manufactured by using a continuous mpression or by using laminating pastes. Lamination by ther- working tape-casting machine with stationary blades(Hutter, mal compression works well when the tapes have a simple ge- Lohmer, Germany ). a pol ometry and when they have a certain range of properties such as Hoechst Trespaphan, Neunkirchen, Germany) with a thickness porosity and binder content, respectively. .But the applied of 40 um was used as a carrier. Drying was performed under air pressure prevents the buildup of more complex-shaped geome- convection at room temperature. a detailed characterization of tries because of deformation of the structures during lamination the electrokinetic and rheological behavior of the dispersed Pressureless lamination using pastes to join the tapes allows powders and the description of the tape-casting procedure can the design of very complex-shaped geometries, which has been be found elsewhere. For parts with simple geometries, these tapes can be laminat- K.J. Bowman-contributing editor ed by compression at room temperature without any additional materials(see Section II(2). More complex geometries would deform during compression, and therefore have to be built up by pressureless joining of tapes using laminating pastes. The pastes used for lamination of the green tapes in this work consist of several components: a pre-ceramic polymer (liquid polysilazane, Ceraset SN, AlliedSignal Composites Inc, Morristown, NJ

Processing, Microstructure, and Properties of Laminated Silicon Nitride Stacks Bernd Bitterlichw and Ju¨rgen G. Heinrich* Department for Engineering Ceramics, Institute for Nonmetallic Materials, Clausthal University of Technology, Clausthal-Zellerfeld, Germany By lamination of silicon nitride tapes, components with complex geometries can be produced. Unstructured tapes can be lami￾nated by common thermal compression. Structured tapes, how￾ever, have to be joined by pressureless processes using e.g. pastes as lamination aids because deformation of the struc￾tures would occur. These pastes usually contain a binder for maintaining the mechanical contact between the tapes during processing. To prevent the high mass loss of typical organic binders during burnout, pre-ceramic polymers were used in this work. These ceramic precursors convert partly into an inorganic material during heat treatment with a significant reduced mass loss compared with common organic binders. Thus, the porosity in the interlayer of a laminated stack is strongly decreased, which should be favorable for the mechanical and thermal prop￾erties. This work discusses the resulting microstructure, strength, and thermal diffusivity data of stacks laminated with pastes containing various precursor contents. These results are compared with those obtained by samples prepared by compres￾sion of green tapes. It is found that except for some large pores, the microstructure of the precursor-derived interlayers is qual￾itatively the same as in the tape material. For stacks made by both lamination methods, strength measurements reveal that the properties parallel and perpendicular to the layers are different. It is shown that the same strength level can be obtained both by using the pressureless route and by the compression method. Unlike the strength, the thermal conductivity does not change with the direction of measurement. I. Introduction SILICON nitride is a candidate material for engine applications because of the interesting combination of thermal and me￾chanical properties that it possesses. The starting powders, sin￾tering, and machining are very cost intensive and prevent serial application.1 For prototyping of complex-shaped components, the effort of machining can be reduced by tape casting and lam￾inating.2 Tape casting is a widespread and cost-effective method to produce flat, large area components with high quality.3 By structuring and laminating tapes, complex-shaped components can be manufactured. Lamination can be carried out by thermal compression or by using laminating pastes. Lamination by ther￾mal compression works well when the tapes have a simple ge￾ometry and when they have a certain range of properties such as porosity and binder content, respectively.4,5 But the applied pressure prevents the buildup of more complex-shaped geome￾tries because of deformation of the structures during lamination. Pressureless lamination using pastes to join the tapes allows the design of very complex-shaped geometries, which has been demonstrated mainly with binder systems that are dissolved in organic solvents.6 Because of the complete mass loss of the or￾ganic binders during binder burnout, porosity between the tapes develops, which can still be present after sintering, thus influ￾encing the properties. Some authors report that the properties of a laminated stack depend on the direction of measuring, i.e. parallel or perpendicular to the layers.7,8 Therefore, a ceramic precursor with a yield of about 70–80 mass% during binder re￾moval seems to be an appropriate material for joining these sil￾icon nitride tapes. These precursors have already been used successfully as binders for processes like injection molding.9 In this work, aqueous binder systems have been used to manufacture silicon nitride tapes. These tapes have been lami￾nated by using pastes on the basis of a ceramic precursor and also by the compression method for comparison. The rheolog￾ical characteristics of the pastes, their behavior during screen printing, the thermal treatment, the resulting microstructure, the strength, and the thermal conductivity of sintered stacks parallel and perpendicular to the laminating direction will be discussed. II. Experimental Procedure (1) Materials A direct nitrided silicon nitride powder (Silzots HQ, SKW Trostberg Metallchemie, Trostberg, Germany) was homoge￾nized in water with 6 mass% yttria (Meldform metals, Royston, UK) and 4 mass% alumina (CT3000 SG, Alcoa Chemie, Ludwigshafen, Germany) using an attrition mill (LME 1, Net￾zsch, Selb, Germany). The data of the as-delivered and milled silicon nitride powders are summarized in Table I. For tape casting, the powders were dispersed in water with a combination of an amino alcohol and a synthetic polyelectrolyte (Dolapix A88 and PC 33, Zschimmer & Schwarz Chemische Fabriken, Lahnstein, Germany). As a defoamer, a fatty alcohol (Cont￾raspum con., Zschimmer & Schwarz Chemische Fabriken) and, as a binder, a commercial polymer emulsion (Mowilith DM 765, Clariant, Frankfurt, Germany) were used. Because of the dom￾inating influence of the binder emulsion, the slurry exhibited a distinctive shear-thinning behavior. Negligible thixotropy was observed. Tapes were manufactured by using a continuous working tape-casting machine with stationary blades (Hutter, Lohmer, Germany). A polypropylene polymer film (NNA 40, Hoechst Trespaphan, Neunkirchen, Germany) with a thickness of 40 mm was used as a carrier. Drying was performed under air convection at room temperature. A detailed characterization of the electrokinetic and rheological behavior of the dispersed powders and the description of the tape-casting procedure can be found elsewhere.10 For parts with simple geometries, these tapes can be laminat￾ed by compression at room temperature without any additional materials (see Section II(2)). More complex geometries would deform during compression, and therefore have to be built up by pressureless joining of tapes using laminating pastes. The pastes used for lamination of the green tapes in this work consist of several components: a pre-ceramic polymer (liquid polysilazane, Cerasett SN, AlliedSignal Composites Inc., Morristown, NJ, 2713 Journal J. Am. Ceram. Soc., 88 [10] 2713–2721 (2005) DOI: 10.1111/j.1551-2916.2005.00512.x r 2005 The American Ceramic Society K. J. Bowman—contributing editor *Member, American Ceramic Society. w Author to whom correspondence should be addressed. e-mail: b.bitterlich@ ceramtec.de Manuscript No. 20494. Received May 13, 2004; approved May 2, 2005

Journal of the American Ceramic Society--Bitterlich and Heinrich 88,No.10 Table l. Si3N4 Powder Data(Silzot HQ, SKw, Germany) (um) As delivered (suppliers data) After milling Specific surface area(m/g) 4.5 9.2 阝 content(%) 12 Amount of nitrogen N(%) 38.6(40.2) O(%) 0.50(084) C(ppm) 0.085 0.119 Lasergranulometer (Typ LS 230, Coulter. Krefeld, Germany). Leco, TC-436. St Josephs, MI.Leco C400, St Josephs. MI with I mass% catalyst, Dicumyl peroxide( Aldrich Chemicals St. Louis, MO), a solvent (Terpineol, CioHiso, purum. Fluka Fig. 1. Surface of green tapes(left: top surface; right: bottom sur Chemie ag, Buchs, Switzerland), the powder mixture described face)(scanning electron microscopy, SEM). above, and I mass% of dispersant based on the powder content of polycarbonate(Trusan 450, Trukem Gmbl All processing steps and measurements were carried out in a Worms, Germany) in various ratios(see Table II glove box with controlled low oxygen and low water concen The pre-ceramic polymer was used as a binder for the powder tration to prevent any reactions of the pre-ceramic polymer. mixture. The solvent had to be added to optimize the viscosity of the compositions. In this work, terpineol was used because of its low evaporation rate and its chemical compatibility with the (2) Lamination and Thermal Treatment precursor. The dispersant was chosen after several screening Lamination was performed both by compression of green tapes tests to improve the stability of the pastes. These compor t 14 MPa for 20 s at room temperature and by using the pastes were homogenized using a triple-roller mill(Exakt 80S, Exakt consisting of the pre-ceramic polymer. The pastes were applied Apparatebau, Norderstedt, Germany), which is frequently used to the tapes by screen printing(mesh diameter=90 um, wire for high-viscosity mixtures thickness= 40 um). The laminated stack was heat treated in the During the screen-printing process, the shear rates in the paste change very fast. To analyze the behavior of the laminat Because of the observed shrinkage anisotropy of single tapes ing paste, this process was simulated in the viscosimeter(VT550, during sintering, 0 the alternating tapes were turned around by Haake, Karlsruhe, Germany) at 25"C using the following pro- 90 to their casting direction. Also, care was taken to ensure that cedure: after pre-shearing at 50 s for 30 s, a holding time of the bottom side of the green tapes always met the top side be 120 s followed. This pre-treatment is necessary to obtain repro- cause of their different surface properties( Fig. 1) Pyrolysis and binder burnout, respectively, were carried out during increasing shear rate in a logarithmic scale up to a max simultaneously under a flowing nitrogen atmosphere at a heat imum of 200s-l in 320s. Then the shear rate was decreased with ing rate of 5K/min and a soaking time of I h at a maximum the same ramp to observe any time-dependent behavior - Di- mperature of 750C. The pyrolyzed tapes were sintered rectly after this program, the shear rate was set to 30 s as fast in a gas pressure furnace(FPW, FCT, Rodental, Germany) as possible. After 30 s, the shear rate was decreased rapidly t 1800C for 60 min under a nitrogen pressure of 5 MPa. to only I s. The response of the viscosity was measured with The heating rate was 20 K/min up to 1100 C and 10 K/min up to 1800.C. The cooling rate was set to 10 K/min Composition of Pastes us heological examination an Amounts(vol%) Fraction of precursor to precursor plus powder Used for lamination In mass% Cer2041.8 14 11.0 45.8 Cer2041.0 Cer2040.6 Cer30424 424 37.2 Cer30419 419 14 38.0 Cer3041.3 41.3 1.3 38.8 Cer3040.2 1.3 40 Cer4041.9 41.9 1.3 16.8 Cer4040.8 40 1.3 28.4 9.4 Cer40-39.6 .3 274 31.7 Cer5042.1 42.1 14 43.8 12.7 23.3 Cer50-39.5 39.5 1.3 394 18.0 Cer50-394 394 1.3 39.5 18.2 38.5 1.3 60.2 31.3 0.3 3.2

with 1 mass% catalyst,11 Dicumyl peroxide (Aldrich Chemicals, St. Louis, MO), a solvent (Terpineol, C10H18O, purum, Fluka Chemie AG, Buchs, Switzerland), the powder mixture described above, and 1 mass% of dispersant based on the powder content on the basis of polycarbonate (Trusan 450, Trukem GmbH, Worms, Germany) in various ratios (see Table II). The pre-ceramic polymer was used as a binder for the powder mixture. The solvent had to be added to optimize the viscosity of the compositions. In this work, terpineol was used because of its low evaporation rate and its chemical compatibility with the precursor. The dispersant was chosen after several screening tests to improve the stability of the pastes. These components were homogenized using a triple-roller mill (Exakt 80S, Exakt Apparatebau, Norderstedt, Germany), which is frequently used for high-viscosity mixtures. During the screen-printing process, the shear rates in the paste change very fast. To analyze the behavior of the laminat￾ing paste, this process was simulated in the viscosimeter (VT550, Haake, Karlsruhe, Germany) at 251C using the following pro￾cedure: after pre-shearing at 50 s
1 for 30 s, a holding time of 120 s followed. This pre-treatment is necessary to obtain repro￾ducible results. After this, the viscosity values were recorded during increasing shear rate in a logarithmic scale up to a max￾imum of 200 s
1 in 320 s. Then the shear rate was decreased with the same ramp to observe any time-dependent behavior.12 Di￾rectly after this program, the shear rate was set to 30 s
1 as fast as possible. After 30 s, the shear rate was decreased rapidly to only 1 s
1 . The response of the viscosity was measured with time. All processing steps and measurements were carried out in a glove box with controlled low oxygen and low water concen￾tration to prevent any reactions of the pre-ceramic polymer. (2) Lamination and Thermal Treatment Lamination was performed both by compression of green tapes at 14 MPa for 20 s at room temperature and by using the pastes consisting of the pre-ceramic polymer. The pastes were applied to the tapes by screen printing (mesh diameter 5 90 mm, wire thickness 5 40 mm). The laminated stack was heat treated in the glove box at 1501C for 1 h. Because of the observed shrinkage anisotropy of single tapes during sintering,10 the alternating tapes were turned around by 901 to their casting direction. Also, care was taken to ensure that the bottom side of the green tapes always met the top side be￾cause of their different surface properties (Fig. 1). Pyrolysis and binder burnout, respectively, were carried out simultaneously under a flowing nitrogen atmosphere at a heat￾ing rate of 5 K/min and a soaking time of 1 h at a maximum temperature of 7501C. The pyrolyzed tapes were sintered in a gas pressure furnace (FPW, FCT, Ro¨dental, Germany) at 18001C for 60 min under a nitrogen pressure of 5 MPa. The heating rate was 20 K/min up to 11001C and 10 K/min up to 18001C. The cooling rate was set to 10 K/min Table I. Si3N4 Powder Data (Silzots HQ, SKW, Germany) Grain size (mm) As delivered (suppliers data) After milling D10 0.7 0.1w D50 1.6 0.4w D90 4.1 1.9w Specific surface area (m2 /g) 4.5 9.2 b-content (%) 12 Amount of nitrogen N (%) 38.6 (40.2z ) 38.2z Amount of contaminations O (%) 0.50 (0.84z ) 1.50z C (ppm) 0.085y 0.11y w Lasergranulometer (Typ LS 230, Coulter, Krefeld, Germany). z Leco, TC-436, St. Josephs, MI. y Leco, C-400, St. Josephs, MI. Table II. Composition of Pastes used for Rheological Examination and for Lamination Descriptionw Used for lamination Amounts (vol%) Fraction of precursor to precursor plus powder Powder Dispersant Precursor Solvent In vol% In mass% Cer 0 32.5 1.1 — 66.5 — — Cer 20–41.8 41.8 1.4 11.0 45.8 20 7.0 Cer 20–41.0 O 41.0 1.4 10.8 46.8 Cer 20–40.6 40.6 1.3 10.7 47.3 Cer 30–42.4 42.4 1.4 19.0 37.2 30 11.5 Cer 30–41.9 O 41.9 1.4 18.7 38.0 Cer 30–41.3 41.3 1.3 18.6 38.8 Cer 30–40.2 40.2 1.3 18.1 40.3 Cer 40–41.9 41.9 1.3 29.2 27.6 40 16.8 Cer 40–40.8 O 40.8 1.3 28.4 29.4 Cer 40–39.6 39.6 1.3 27.4 31.7 Cer 50–42.1 42.1 1.4 43.8 12.7 50 23.3 Cer 50–39.5 O 39.5 1.3 39.4 18.0 Cer 50–39.4 39.4 1.3 39.5 18.2 Cer 60 O 38.5 1.3 60.2 — 60 31.3 Cer 90 9.6 0.3 90.1 — 90 73.2 Cer100 — — 100 — 100 100 w The description contains (a) the precursor content in vol% based on precursor and powder and (b) the powder amount to distinguish pastes with a different solvent content. Fig. 1. Surface of green tapes (left: top surface; right: bottom sur￾face) (scanning electron microscopy, SEM).10 2714 Journal of the American Ceramic Society—Bitterlich and Heinrich Vol. 88, No. 10

October 2005 laminated silicon nitride stacks down to 1500.C. At this temperature, the furnace was switched pressureless method using lamination aids. Use of a slurry, how- ever, with the same components as that of the original tape casting slurry, did not show success d lamination of (3) Characterization of the Multi-Layer Stacks the silicon nitride green tapes. After applying the slurry onto the The density of the sintered stacks was measured by the Ar- green tape, the layer dries rapidly because the porous green tape paks up the water immediately. The dry binder emulsion does chimedes principle in water with a wetting agent to avoid air not have the ability to join a second tape anymore. Thus, scanning electron microscopy(sEM, CamScan, Cambridge lamination with this system is not possible. This problem can UK). The 4-point bending strength was measured according be overcome by using terpineol as a viscous organic solvent to DIN/EN 843, part I with an sample geometry of 2.5 mm x 2 or the powder suspension. The processing is further improved mm×30mm. These les were prepared parallel and pe when pastes containing the liquid precursor are used because dicular to the direction of the laminated layers. In the fo the precursor increases the adhesion between the laminated owing text, parallel" and"perpendicular" are used when the resulting tensile stress is parallel and perpendicular, respective An efficient way to apply a thin layer of a slurry onto a ubstrate is screen printing. For this process, the pastes must to the stacking direction during strength measurement(compare exhibit a certain rheological behavior: the viscosity should be within certain limits and the pastes should be shear thinning Thermal diffusivity was measured in the Netzsch application To optimize the rheological behavior, different pastes ranging lab using the laser-Flash method (LFA 427, Netzsch). Disks from precursor contents of 0% to 100% with varying solvent having a diameter of 12 mm and a thickness of approximately 2 contents have been investigated (Fig 3). Because the pastes con- mm were used. These samples were cut out of sintered stacks arallel and perpendicular to the laminated layers. The specific sist of several components, a complex rheological behavior can heat was evaluated by differential sc be expected. The composition of Cer100(i. e. pure liquid pre- 404. Netzsch). The thermal conductivity i was calculated from cursor) was analyzed to understand the effects of the single the diffusivity a, density p, and specific heat cp using the fol components. The liquid polymer shows Newtonian behavior lowing equation: i=apcp. X-ray diffraction patterns of the viscosity does not change with the applied shear rate(at low planes parallel to and perpendicular to the casting plane shear rates, the limit of the equipment was reached, resulting in determined by a Philips diffractometer(PW 1050, Almelo, great scattering of the measured values). The viscosity is only 0. 15 Pa s at 25C and therefore the pure polymer cannot be used for screen printing. On adding powder, the viscosity in- (Cer90, i.e. 10 vol% er). The composition with 60 II. Results and discussion vol% of precursor( Cer60)still does not contain an additional solvent. At this higher solid content, the rheological behavior () Lamination and Thermal Treatment changes to slightly shear thinning and additional shear thicken- Unstructured green tapes can be joined at room temperature at ing at higher shear rates. Both effects become stronger with in- a pressure of 14 MPa without any additional aid. By the applied creased solids content( Cer40 and Cer 20). If more solvent pressure, the tapes are pressed into each other and a homoge- added, it only acts as a diluent and reduces the effects. It can also neous material is obtained(Fig. 2). Because of its low glass be observed that in the shear-thickening range, a hysteresis de- transformation temperature of-6C, the binder in the tapes can velops: the measured viscosities during increasing shear rates are be easily deformed plastically. Because of the tape-casting higher than that of the ramp downward. The slurry without the process, the "bottom" surface, which had contact with the pre-ceramic polymer(CerO), i.e. only the powder dispersed in the solvent, shows strong shear thinning and no further shear ). Lamination is not possible when two rough top surfaces thickening, but a slight hysteresis. This thixotropic behavior meet during joining. Because of the compression during lami- typical for powder dispersions nation. a shrinkage in the thickness direction of about 8% takes Shear thinning can be explained by breaking up an internal lace,which slightly increases the green density of the whole structure that builds up when the suspension is in rest. The stack. The applied pressure also influences the shrinkage during strength of the internal structure is dependent on the attraction tering in the x-y and the z direction. and repulsion forces between the particles. The more this struc- Complex-shaped geometries, where no pressure can be ap- ture is destroyed, i.e. the higher the applied shear rate, the lower plied because of expected deformation, can be produced by a he viscosity. When the shear rate decreases again, the structure begins to build up again. If this buildup need time, a hysteresis will be observed. This is the case for th der suspension without a precursor(Fig 3). Slurries containing the pre-ceramic polymer also show shear thinning at low shear rates. But because of the long molecules, hich are present between the powder particles, the strength of he internal structure is lower, leading to a lower viscosity in the condition of rest and to a less steep decrease in viscosity wit increasing shear rate. At slightly higher shear rates. however, the viscosity increases again. Such shear thickening can be explained by a buildup of a new internal structure under the influence of e applied shearing. The resulting viscosity is determined by the competing processes of structure destruction and buildup The viscosity measurements comprise a second part to ob- serve the response of viscosity to a simulated screen-printing process. During screen printing, the paste is placed onto the screen. Then, the paste is abruptly sheared by the blade. The neasurements show that the pastes in this work respond with some delay to a sudden change in shear rate(Fig 4). The sus- pension without a precursor, which exhibits a strong shear-thin- Fig. 2. Fracture surface of green tapes, laminated by compression at ning behavior (Fig. 3), needs almost 10 s to gain a constan room temperature(13 layers)(scanning electron microscopy, SEM); bar viscosity. This confirms the time-dependent behavior( thixotro- on the right top corner represents the thickness of a single layer. py) of this slurry. Because of the strong shear thinning, the vis-

down to 15001C. At this temperature, the furnace was switched off. (3) Characterization of the Multi-Layer Stacks The density of the sintered stacks was measured by the Ar￾chimedes principle in water with a wetting agent to avoid air bubbles in the samples. The microstructure was investigated by scanning electron microscopy (SEM, CamScan, Cambridge, UK). The 4-point bending strength was measured according to DIN/EN 843, part 1 with an sample geometry of 2.5 mm 2 mm 30 mm. These samples were prepared parallel and per￾pendicular to the direction of the laminated layers. In the fol￾lowing text, ‘‘parallel’’ and ‘‘perpendicular’’ are used when the resulting tensile stress is parallel and perpendicular, respectively, to the stacking direction during strength measurement (compare Fig. 11). Thermal diffusivity was measured in the Netzsch application lab using the Laser–Flash method (LFA 427, Netzsch). Disks having a diameter of 12 mm and a thickness of approximately 2 mm were used. These samples were cut out of sintered stacks parallel and perpendicular to the laminated layers. The specific heat was evaluated by differential scanning calorimetry (DSC 404, Netzsch). The thermal conductivity l was calculated from the diffusivity a, density r, and specific heat cp using the fol￾lowing equation: l 5 arcp. X-ray diffraction patterns of the planes parallel to and perpendicular to the casting plane were determined by a Philips diffractometer (PW 1050, Almelo, The Netherlands). III. Results and Discussion (1) Lamination and Thermal Treatment Unstructured green tapes can be joined at room temperature at a pressure of 14 MPa without any additional aid. By the applied pressure, the tapes are pressed into each other and a homoge￾neous material is obtained (Fig. 2). Because of its low glass transformation temperature of
61C, the binder in the tapes can be easily deformed plastically.10 Because of the tape-casting process, the ‘‘bottom’’ surface, which had contact with the tape carrier, is much smoother than the ‘‘top’’ surface (see Sec￾tion II). Lamination is not possible when two rough top surfaces meet during joining. Because of the compression during lami￾nation, a shrinkage in the thickness direction of about 8% takes place, which slightly increases the green density of the whole stack. The applied pressure also influences the shrinkage during sintering in the X–Y and the Z direction.13 Complex-shaped geometries, where no pressure can be ap￾plied because of expected deformation, can be produced by a pressureless method using lamination aids. Use of a slurry, how￾ever, with the same components as that of the original tape￾casting slurry, did not show any success during lamination of the silicon nitride green tapes. After applying the slurry onto the green tape, the layer dries rapidly because the porous green tape soaks up the water immediately. The dry binder emulsion does not have the ability to join a second tape anymore. Thus, lamination with this system is not possible. This problem can be overcome by using terpineol as a viscous organic solvent for the powder suspension. The processing is further improved when pastes containing the liquid precursor are used because the precursor increases the adhesion between the laminated tapes. An efficient way to apply a thin layer of a slurry onto a substrate is screen printing. For this process, the pastes must exhibit a certain rheological behavior: the viscosity should be within certain limits and the pastes should be shear thinning.14 To optimize the rheological behavior, different pastes ranging from precursor contents of 0% to 100% with varying solvent contents have been investigated (Fig. 3). Because the pastes con￾sist of several components, a complex rheological behavior can be expected. The composition of Cer100 (i.e. pure liquid pre￾cursor) was analyzed to understand the effects of the single components. The liquid polymer shows Newtonian behavior: the viscosity does not change with the applied shear rate (at low shear rates, the limit of the equipment was reached, resulting in great scattering of the measured values). The viscosity is only 0.15 Pa s at 251C15 and therefore the pure polymer cannot be used for screen printing. On adding powder, the viscosity in￾creases (Cer90, i.e. 10 vol% powder). The composition with 60 vol% of precursor (Cer60) still does not contain an additional solvent. At this higher solid content, the rheological behavior changes to slightly shear thinning and additional shear thicken￾ing at higher shear rates. Both effects become stronger with in￾creased solids content (Cer40 and Cer20). If more solvent is added, it only acts as a diluent and reduces the effects. It can also be observed that in the shear-thickening range, a hysteresis de￾velops: the measured viscosities during increasing shear rates are higher than that of the ramp downward. The slurry without the pre-ceramic polymer (Cer0), i.e. only the powder dispersed in the solvent, shows strong shear thinning and no further shear thickening, but a slight hysteresis. This thixotropic behavior is typical for powder dispersions.14 Shear thinning can be explained by breaking up an internal structure that builds up when the suspension is in rest. The strength of the internal structure is dependent on the attraction and repulsion forces between the particles. The more this struc￾ture is destroyed, i.e. the higher the applied shear rate, the lower the viscosity. When the shear rate decreases again, the internal structure begins to build up again. If this buildup needs some time, a hysteresis will be observed. This is the case for the pow￾der suspension without a precursor (Fig. 3). Slurries containing the pre-ceramic polymer also show shear thinning at low shear rates. But because of the long molecules, which are present between the powder particles, the strength of the internal structure is lower, leading to a lower viscosity in the condition of rest and to a less steep decrease in viscosity with increasing shear rate. At slightly higher shear rates, however, the viscosity increases again. Such shear thickening can be explained by a buildup of a new internal structure under the influence of the applied shearing.14 The resulting viscosity is determined by the competing processes of structure destruction and buildup. The viscosity measurements comprise a second part to ob￾serve the response of viscosity to a simulated screen-printing process. During screen printing, the paste is placed onto the screen. Then, the paste is abruptly sheared by the blade. The measurements show that the pastes in this work respond with some delay to a sudden change in shear rate (Fig. 4). The sus￾pension without a precursor, which exhibits a strong shear-thin￾ning behavior (Fig. 3), needs almost 10 s to gain a constant viscosity. This confirms the time-dependent behavior (thixotro￾py) of this slurry. Because of the strong shear thinning, the vis￾Fig. 2. Fracture surface of green tapes, laminated by compression at room temperature (13 layers) (scanning electron microscopy, SEM); bar on the right top corner represents the thickness of a single layer. October 2005 Laminated Silicon Nitride Stacks 2715

2716 Journal of the American Ceramic Society-Bitterlich and Heinrich Vol. 88. No. 10 ◇cer20-418 cer0-32.5 △cer20-40.6 10 ◇cer40-419 100●Cer40-40.8 100 △cer40-396 cer60-38.5 U 10 Cer 1 (c) shear rate/s-1 shear rate /s Fig 3. Viscosity measurements of laminating pastes with varying contents of pre-ceramic polymer and solven osity increases by a factor of 10 after jumping back to a low When the lamination paste, consisting of the ceramic powder, shear rate. The compositions containing the pre-ceramic poly- the pre-ceramic polymer, and the solvent, is applied onto the mer behave differently: at the jump to a higher shear rate, the green tapes, the liquid components spread into the porosity of viscosity overshoots before it becomes constant. The subsequent the tapes, which is about 30 vol%. In this stage, the tapes jump back to the low shear rate leads to a small decrease held together only by adhesion On heating the laminated tapes scosity because of the shear-thickening behavior(see Fig. 4) to 150.C, the solvent evaporates and cross-linking of the pre- The time necessary to reach a constant value is much shorter ceramic polymer begins. This cross-linked polymer is rigid and than that for the suspension without a polymer, which is because eads to a good green strength of the laminated stack. Further of the weaker inter-particle forces combined with the smaller more, it reduces the chemical sensitivity of the precursor to the change in viscosity. All slurries with a precursor show the same atmosphere (especially water), which is important because the characteristics amples had to be moved from the glove box to the pyre All graphs show a peak during the jump to the low shear rate furnace This effect occurs because the adjustment of the shear rate in the During heating of the laminated stack, the binder burns out viscous paste cannot follow the set value immediately. n the temperature range from roughly 300 to 450C with a of precursor po maximum rate at 393.C(Fig. 5(a)). The pyrolysis of the pre- was varied and the quality of the printed layers was assessed ceramic polymer takes place over a broader temperature range qualIt itatively. A viscosity of 6-10 Pa s at a shear rate of 50 s n two steps(Fig. 5(b)). It begins at approximately 350C and at shown to be best suited. The thickness of the dried screen round 800C, the mass loss is almost completed. during printed layers was around 30 um pyrolysis, the pre-ceramic polymer is converted into an inor

cosity increases by a factor of 10 after jumping back to a low shear rate. The compositions containing the pre-ceramic poly￾mer behave differently: at the jump to a higher shear rate, the viscosity overshoots before it becomes constant. The subsequent jump back to the low shear rate leads to a small decrease in viscosity because of the shear-thickening behavior (see Fig. 4). The time necessary to reach a constant value is much shorter than that for the suspension without a polymer, which is because of the weaker inter-particle forces combined with the smaller change in viscosity. All slurries with a precursor show the same characteristics. All graphs show a peak during the jump to the low shear rate. This effect occurs because the adjustment of the shear rate in the viscous paste cannot follow the set value immediately. For different ratios of precursor:powder, the solvent content was varied and the quality of the printed layers was assessed qualitatively. A viscosity of 6–10 Pa s at a shear rate of 50 s
1 was shown to be best suited. The thickness of the dried screen￾printed layers was around 30 mm. When the lamination paste, consisting of the ceramic powder, the pre-ceramic polymer, and the solvent, is applied onto the green tapes, the liquid components spread into the porosity of the tapes, which is about 30 vol%.10 In this stage, the tapes are held together only by adhesion. On heating the laminated tapes to 1501C, the solvent evaporates and cross-linking of the pre￾ceramic polymer begins.15 This cross-linked polymer is rigid and leads to a good green strength of the laminated stack. Further￾more, it reduces the chemical sensitivity of the precursor to the atmosphere (especially water), which is important because the samples had to be moved from the glove box to the pyrolysis furnace. During heating of the laminated stack, the binder burns out in the temperature range from roughly 3001 to 4501C with a maximum rate at 3931C (Fig. 5(a)). The pyrolysis of the pre￾ceramic polymer takes place over a broader temperature range in two steps (Fig. 5(b)). It begins at approximately 3501C and at around 8001C, the mass loss is almost completed. During pyrolysis, the pre-ceramic polymer is converted into an inor￾Fig. 3. Viscosity measurements of laminating pastes with varying contents of pre-ceramic polymer and solvent. 2716 Journal of the American Ceramic Society—Bitterlich and Heinrich Vol. 88, No. 10

October 2005 laminated silicon nitride stacks s-I 1000 Cer 60 Cer 20.4 0.1 Fig, 4 of the screen-printing process and measured viscosi- Fig. 6. Microstructure of sintered tapes laminated by compression ties as a to the set change of shear rates for laminating pastes coarse B-Si3N4 grains surrounded by fine grains(light gray) and ith different contents of the pre-ceramic polymer condary phase(white)(scanning electron microscopy, SEM, plasma ganic material by separation of organic molecules of low mass The pyrolysis of a laminated stack may lead to stresses and such as hydrocarbons and ammonia. This is accompanied by cracks in the interlayer because of the shrinkage of the highly mass loss of about 22 mass% and a linear shrinkage of 25% precursor-containing interlayer and(almost)no shrinkage of the A normal organic binder, in comparison, burns out completely, green tapes. Such cracks were indeed observed for thick inter- leaving behind porosity. Thus, by using the ceramic precursor, a layers (interlayer thickness a tape thickness)and high amour dense interlayer can be expected. The product after pyrolysis is an amorphous SiC,N, material. During a further heat Thus, cracks during pyrolysis can be avoided by thin interlayers treatment-in this work, the subsequent sintering step--it crys- and medium precursor contents allizes to Si3N4, SiC, and C. These phases will undergo com- In conclusion, the pastes containing 30-50 vol% of the ce- plex reactions with the atmosphere and/or the glassy phase ramic precursor(Cer 30-50) turned out to be best suited for during sintering. lamination of the green tapes With lower amounts of precursor, the effect of the pre-ceramic polymer as a binder and therefore the green strength become too low. For higher fractions of pre- cursors(>50 vol%), problems arise because the sintering be- havior differs very much from that of the tape material a15 (2) Microstructural Development After lamination by compression, the single layers cannot be distinguished in the fracture surface of green parts(Fig. 2). The analysis of polished samples by SEM at high magnification re- veals a homogeneous material as well( Fig. 6) In terlayers of a paste-laminated stack 005 be easily detected after pyrolysis because their density is higher 14% than that of the pyrolyzed green tapes(Fig. 7). The green tapes contain about 25 vol% binder and 30 vol% porosity, -or Temperature/C about 55 vol% porosity, respectively, after binder burno (assuming no shrinkage). In comparison, the measured tot 015 porosity of a sample Cer60(prepared from a mixture of 60 vol% precursor and 40 vol% silicon nitride)is only 33 vol%because of the high ceramic yield of the ceramic precursor. The different densities after pyrolysis affect the sintering be -6vol% precur vior. Generally, the lower the starting density, the higher the 005o shrinkage to achieve full density. Thus, tape material with lower density needs a higher shrinkage for complete densi- fication than the interlayers. During sintering of a stack, the 扌2 difference in shrinkage can lead to internal stresses, incomplet densification of the tape material, or even cracks in the tapes perpendicular to the layers, which in fact had been observed under non-optimized processing conditions(Fig. 8). The pre- cautions mentioned above(thin interlayers, medium precursor contents) helped to prevent cracks. 0.10 The overall densities of sintered stacks decrease slightly with ncreasing precursor content of the laminating pastes: stack T/C Fig. 5. Mass loss of (a)the green tape and(b) the cured ceramic pre- hereas the densities of stacks made with the pastes decrease to cursor in mixtures from 100 vol%(Cer100) to 50 vol% precursor and 3.247(Cer30)and 3. 241( Cer50)g/cm. This proves that the powder(Cer50)(flowing nitrogen atmosphere, 5 K/min heating rate). ering behavior is negatively influenced by the presence of the

ganic material by separation of organic molecules of low mass such as hydrocarbons and ammonia.16 This is accompanied by a mass loss of about 22 mass% and a linear shrinkage of 25%.17 A normal organic binder, in comparison, burns out completely, leaving behind porosity. Thus, by using the ceramic precursor, a dense interlayer can be expected. The product after pyrolysis is an amorphous SiCxNy material. During a further heat treatment—in this work, the subsequent sintering step—it crys￾tallizes to Si3N4, SiC, and C.16 These phases will undergo com￾plex reactions with the atmosphere and/or the glassy phase during sintering. The pyrolysis of a laminated stack may lead to stresses and cracks in the interlayer because of the shrinkage of the highly precursor-containing interlayer and (almost) no shrinkage of the green tapes. Such cracks were indeed observed for thick inter￾layers (interlayer thickness  tape thickness) and high amounts of pre-ceramic precursors in the laminating paste (  70%). Thus, cracks during pyrolysis can be avoided by thin interlayers and medium precursor contents. In conclusion, the pastes containing 30–50 vol% of the ce￾ramic precursor (Cer30–50) turned out to be best suited for lamination of the green tapes. With lower amounts of precursor, the effect of the pre-ceramic polymer as a binder and therefore the green strength become too low. For higher fractions of pre￾cursors (450 vol%), problems arise because the sintering be￾havior differs very much from that of the tape material. (2) Microstructural Development After lamination by compression, the single layers cannot be distinguished in the fracture surface of green parts (Fig. 2). The analysis of polished samples by SEM at high magnification re￾veals a homogeneous material as well (Fig. 6). In comparison, the interlayers of a paste-laminated stack can be easily detected after pyrolysis because their density is higher than that of the pyrolyzed green tapes (Fig. 7). The green tapes contain about 25 vol% binder and 30 vol% porosity,10—or about 55 vol% porosity, respectively, after binder burnout (assuming no shrinkage). In comparison, the measured total porosity of a sample Cer60 (prepared from a mixture of 60 vol% precursor and 40 vol% silicon nitride) is only 33 vol%17 because of the high ceramic yield of the ceramic precursor. The different densities after pyrolysis affect the sintering be￾havior. Generally, the lower the starting density, the higher the shrinkage to achieve full density. Thus, tape material with a lower density needs a higher shrinkage for complete densi- fication than the interlayers. During sintering of a stack, the difference in shrinkage can lead to internal stresses, incomplete densification of the tape material, or even cracks in the tapes perpendicular to the layers,18 which in fact had been observed under non-optimized processing conditions (Fig. 8). The pre￾cautions mentioned above (thin interlayers, medium precursor contents) helped to prevent cracks. The overall densities of sintered stacks decrease slightly with increasing precursor content of the laminating pastes: stacks made by compression have a sintering density of 3.252 g/cm3 , whereas the densities of stacks made with the pastes decrease to 3.247 (Cer30) and 3.241 (Cer50) g/cm3 . This proves that the sintering behavior is negatively influenced by the presence of the Fig. 4. Simulation of the screen-printing process and measured viscosi￾ties as a response to the set change of shear rates for laminating pastes with different contents of the pre-ceramic polymer. Fig. 5. Mass loss of (a) the green tape and (b) the cured ceramic pre￾cursor in mixtures from 100 vol% (Cer100) to 50 vol% precursor and powder (Cer50) (flowing nitrogen atmosphere, 5 K/min heating rate). Fig. 6. Microstructure of sintered tapes laminated by compression: coarse b-Si3N4 grains surrounded by fine grains (light gray) and secondary phase (white) (scanning electron microscopy, SEM, plasma￾etched). October 2005 Laminated Silicon Nitride Stacks 2717