5744
J. Phys. Chem. B 1998, 102, 5744-5753
Modified Sol-Gel Based Approaches for Synthesizing Borophosphosilicate Glasses and Glass-Ceramics Jin Yong Kim and Prashant N. Kumta* Department of Materials Science and Engineering, Carnegie Mellon UniVersity, Pittsburgh, PennsylVania 15213 ReceiVed: September 17, 1997; In Final Form: April 20, 1998
Modified sol-gel based processes were used to synthesize borophosphosilicate glasses and glass-ceramics, utilizing boron oxide and phosphorus pentoxide as starting precursors. In these processes, the alkoxides of boron and phosphorus were synthesized in situ using oxide precursors, which were then subjected to controlled or rapid hydrolysis and condensation reactions to form either gels (modified oxide sol-gel process: MOSG) or precipitate powders (modified oxide sol-precipitation process: MOSP). The dried gels and precipitates obtained in both processes were heat-treated to 800 °C to crystallize the BPO4 phase. The crystallinity of the as-prepared and heat-treated powders was analyzed using X-ray diffraction, while Fourier transform infrared spectroscopy (FTIR) was employed to identify the molecular linkages of the various components present. Results of these studies show that B-O-B species form during heat treatment of the MOSG-derived gels because of the presence of unreacted B-OH groups, leading to microsegregation of these units in the gels. However, these units are not detected in the powders obtained using the MOSP process, thereby suggesting differences in the gel structures. These structural differences lead to different pathways for the formation of BPO4 in the two processes. Nevertheless, independent of the process, both samples, although structurally different, exhibit almost the same extent of crystallization of BPO4 after heat treatment at 800 °C.
Introduction Glasses and glass-ceramics have received considerable attention in the field of microelectronic packaging for high-speed devices because of their low dielectric constant and good thermal expansion match to silicon.1 Over the years, several oxidebased glasses and glass-ceramics have been identified and studied as substrate materials for electronic packaging.2-6 Among these amorphous oxide materials, borosilicate glasses have received the most attention.7-12 Borosilicate glasses exhibit optimum dielectric constant, but they also possess low strengths. This drawback has been overcome to a large extent by the addition of phosphorus pentoxide, leading to the formation of the crystalline borophosphate (BPO4) phase in the borosilicate glass matrix. Glass-ceramics in the borophosphosilicate system were initially identified by MacDowell and Beall.13,14 They synthesized glasses in this system using conventional glass-melting techniques and obtained the glassceramics by post heat-treatment of the glasses. Conventional glass-melting techniques, involving melting of individual elements or compounds, are most commonly used and represent an economical method for processing large batches of glasses. However, these methods suffer from the problems of volatilization and phase separation, generally leading to the formation of inhomogeneous glasses. These problems tend to be more severe in the borophosphosilicate system because of the volatility of B2O3 and P2O5. Many of these limitations, commonly observed in conventional glass-processing methods, can be eliminated by using low-temperature solution-chemistrybased processing techniques such as the well-known sol-gel process. The traditional solution sol-gel process consists of using metal alkoxides as precursors in alcoholic solutions, which undergo hydrolysis and condensation reactions to form poly-
meric species.15 These species on heat treatment eventually condense to form the core oxide network structural units, resulting in the formation of an amorphous glass or the desired crystalline ceramic. The technique works very well for binary systems. However, there are problems, when synthesizing materials containing several components, due to differences in the rates of hydrolysis exhibited by alkoxides of various metals. The problem becomes particularly noteworthy in the case of reactive metals with a large affinity for oxygen such as those from the transition series, the group III elements (e.g., B, Al) and the rare earths. Once in solution, they tend to undergo preferential hydrolysis in the presence of relatively stable alkoxides such as silicon. This leads to molecular inhomogeneity in the solid precursor, promoting the formation of undesired phases. Prolonged heat treatments at high temperatures are therefore required to promote adequate diffusion of the components in the solid state to form the desired singlephase material. Despite these treatments, the presence of trace amounts of secondary phases can never be discounted, as we have seen in our prior work on sol-gel synthesis of cordierite.16 One way of counteracting the problem is to complex the alkoxide and thus reduce the reactivity of the alkoxides. This control of reactivity is mainly brought about by a reduction in the partial positive charge on the metal species, thereby rendering it more stable toward hydrolysis. The most common complexation or chelating agents studied in the literature has been acetylacetonate (acacH).17 There have also been other approaches utilizing carboxylic acids such as acetic acid.18,19 An alternative approach would be to synthesize the alkoxide in situ using stable oxide precursors. In earlier papers, we have shown the use of metal oxides as starting precursors to form metal alkoxides in situ.20-22 The approach demonstrated a novel and economical methodology to form homogeneous gels in the
S1089-5647(97)03055-1 CCC: $15.00 © 1998 American Chemical Society Published on Web 07/07/1998
Sol-Gel Based Synthesis
J. Phys. Chem. B, Vol. 102, No. 30, 1998 5745
case of reactive group III and V alkoxides. Specifically, the technique consists of reacting oxides of boron and phosphorus with alcohol to form partially hydrolyzed alkoxides, as shown by the well-understood reactions below:23
B2O3 + 6ROH T 2B(OR)3 + 3H2O B(OR)3 + xH2O T B(OR)3-x (OH)x + xROH P2O5 + 6ROH T 2[OP(OR)3] + 3H2O OP(OR)3 + xH2O T OP(OR)3-x(OH)x + xROH where 0 < x < 3 and R is an alkyl group. The kinetics of the reactions can be either controlled to result in the formation of amorphous gels via the solution sol-gel process or enhanced to form gel powders directly via the solprecipitation process. The two processes are therefore aptly called modified oxide sol-gel (MOSG) and the modified oxide sol-precipitation (MOSP), respectively. The partially hydrolyzed nature of the in situ synthesized alkoxides, however, largely affects the structure of the MOSG and MOSP-derived gel powders. The xerogels produced by the oxide sol-gel approach can then be heat-treated and sintered to arrive at the desired glass-ceramic. We have already reported on the sintering characteristics and the dielectric properties of the bulk glass-ceramics derived using the two processes.20,21 In this paper, the salient structural differences between the resultant borophosphosilicate glasses derived using the MOSG and the MOSP processes have been analyzed. Both the asprepared and heat-treated powders have therefore been examined for their crystallinity and molecular structure using X-ray diffraction and Fourier transform infrared spectroscopy (FTIR). The as-prepared powders have also been chemically analyzed for the presence of residual carbon and individual metal contents. The apparent differences in the structures of the two gels have been analyzed, while highlighting the impact of these differences on the crystallization of BPO4 and the formation of the glassceramic. Experimental Section Glass Synthesis. Borophosphosilicate glasses were synthesized using both modified oxide sol-gel (MOSG) and modified oxide sol-precipitation (MOSP) processes employing commercially obtained boron oxide (99.98%, Johnson Matthey) and phosphorus pentoxide (99.998%, Aldrich), while using tetraethyl orthosilicate (TEOS; 99.999%, Aldrich) for silicon. Figure 1 shows a schematic of the procedures followed to synthesize borophosphosilicate powders using both MOSG and MOSP processes. Stoichiometric quantities of the oxides and TEOS were reacted in a molar proportion of B2O3/P2O5/SiO2 ) 1:1:4 and 1:1:8, using both approaches. Details of both the processes are given below. (i) MOSG Process. The MOSG process consists of mixing equimolar amounts of B2O3 and P2O5 in a Nalgene beaker. All the necessary materials were then placed in a glovebag filled with nitrogen in order to create an inert atmosphere. This was necessary to prevent the volatilization of the in situ synthesized boron alkoxide. The glovebag was sealed, and the remainder of the experiment was conducted inside the bag itself. The oxide powders (B2O3 and P2O5) were then dissolved in 100 mL of ethyl alcohol. The beaker was covered with Parafilm, and the contents were stirred for 10 min in order to dissolve the powders completely. At this stage, TEOS was added to the beaker
Figure 1. Flow sheet showing the procedure followed in the MOSG and MOSP processes for synthesizing borophosphosilicate gels.
containing B2O3 and P2O5, and the solution was mixed for 15 min. Water was then added to the resulting solution in two stages corresponding to an “r” ratio of 20. The first stage consisted of adding 16 mL and stirring the solution for 45 min at 50 °C. The second stage consisted of adding a basic solution of water (16 mL) containing 10 drops of 14.8 N ammonium hydroxide. The entire solution was then mixed for 2 h at 50 °C. Gelation of the solution occurred after a period of 6 days, and the gels were then dried in air at 150 °C for 2 h to obtain the xerogel. (ii) MOSP Process. The procedure followed in the MOSP approach is very similar to the MOSG process except that the kinetics of the sol-gel reaction have been accelerated to induce precipitation. This was done by adding excess ammonium hydroxide to the water that is introduced in the second stage (25 mL of 14.8 N ammonium hydroxide and 16 mL of water). After the precipitation reaction, the precipitate was collected using a centrifugal separator and dried in air at 150 °C for 2 h. To compare the nature of the BPO4 phase crystallized from the synthesized borophosphosilicate glasses, borophosphate gels were also prepared using a procedure similar to that described in our earlier publication.22 Stoichiometric amounts of B2O3 and P2O5 in a ratio of 1:1 were accordingly dissolved in a solution of ethyl alcohol acidified with HCl, and the mixture was stirred at room temperature for 5-6 h in a beaker provided with a Parafilm seal. Water was then added corresponding to an “r” ratio of 20, and the mixture was stirred for approximately 8 h until the solution became increasingly viscous. The viscous solution was subsequently poured into a Petri dish and heated to approximately 70 °C until it transformed into a dry xerogel. Finally, the xerogel was heat-treated in air at 800 °C for 24 h to form the crystalline BPO4 phase. Heat-Treatment. Heat-treatment was performed on the basis of DTA results that have been reported by us earlier.20 Heat treatments were correspondingly conducted in air at 450, 500, and 800 °C for both the MOSG- and MOSP-derived samples to investigate the following three aspects: first, to follow the structural changes occurring in the glasses, second, to identify the crystallization of BPO4, and third, to analyze the removal
5746 J. Phys. Chem. B, Vol. 102, No. 30, 1998
Kim and Kumta
TABLE 1: Heat-Treatment Schedule Followed for Processing Borophosphosilicate Glass Ceramics duration (h) temp (°C)
114SG-450 and 114SP-450a,b
114SG-500 and 114SP-500a,b
5 5 10 10 24
5 5 10 10
150 200 300 400 450 500 800
114SG-800 and 118SG-800a,b
114SP-800 and 118SP-800a,b
5 5 10 10
5 5 10 10
96 (regrinding twice)
24
24
a 450, 500 and 800 indicate the final temperature of heat treatments in °C. Heating rate was set at 1 °C/min for all heat treatments and all samples were furnace-cooled after the desired heat-treatment period. b 114 and 118 indicate the initial stoichiometric ratio of B2O3, P2O5, and SiO2. “SG” and “SP” refer to sol-gel samples and sol-precipitated samples, respectively.
of residual carbon. According to our previously published thermal analysis results, the glass transition temperature (Tg) of borophosphosilicate gels with a composition of B2O3/P2O5/ SiO2 ) 1:1:4 and the temperature of crystallization (Tc) of BPO4 are 430 and 490 °C, respectively. Thus, both the gel and precipitated powders were subjected to identical heat treatments in air up to 400 °C as shown in Table 1 before being heattreated at 450 and 500 °C for 24 h, respectively. The heattreatment procedures at 800 °C were, however, different for the samples derived using the two processes. In the case of the MOSG process, the gel powders were heat-treated at 800 °C and reground twice in order to remove the carbon entirely and to induce crystallization of the borophosphate phase. Hence, the total time for heat treatment at 800 °C was 96 h. On the other hand, the precipitates from the MOSP process were crushed with a mortar and pestle and heat-treated up to 800 °C for 24 h. The complete heat-treatment profile and nomenclature for each heat-treated sample are shown in Table 1. The modified oxide sol-gel and sol-precipitation derived samples are accordingly denoted as “SG” and “SP”, respectively. The numbers to the left and right of these letters correspond to the stoichiometry of the samples and the heat-treatment condition employed, respectively. The letters “AP” represent the asprepared state of the powders obtained via both processes. Materials Characterization. X-ray diffraction using a Rigaku θ/θ diffractometer was performed on the as-prepared and heat-treated gels to identify the presence of any crystalline phases. The molecular linkages of the various components in the as-prepared and heat-treated powders were investigated using a Fourier transform infrared spectrometer (Galaxy series FTIR 5030, ATI Mattson). Infrared absorption spectra were collected on the powders using the KBr pressed-pellet method in the spectral range 400-4000 cm-1. The KBr pellets were prepared by utilizing a constant sample weight (2 wt % of the powder in 800 mg of the KBr pellet) and applying a constant pressure (30 MPa) for all the samples analyzed. This was done to minimize any inconsistencies and to facilitate good comparison of the spectra between different samples. The as-prepared powders prepared using both processes were chemically analyzed for carbon, hydrogen, nitrogen, and metal content (Galbraith laboratories, Knoxville, TN). Results The as-prepared and heat-treated powders, derived using both the modified oxide sol-gel (MOSG) and modified oxide sol-
Figure 2. X-ray diffraction trace of borophosphate gels heat-treated at 800 °C for 24 h, indicating the formation of the crystalline BPO4 phase (all peaks correspond to the BPO4 phase).
precipitation (MOSP) processes, were characterized using XRD, FTIR, and chemical analyses. The results of the studies conducted on the two processes will be presented and discussed in two sections. (1) MOSG Process. X-ray Analysis. The resultant asprepared and heat-treated gel powders derived using the MOSG process were first characterized for the presence of a crystalline BPO4 phase using X-ray diffraction. A borophosphate gel sample was also synthesized, mainly as a reference to compare the nature of the crystallized BPO4 phase present in the synthesized borophosphosilicate glasses. The crystallinity of this gel was also examined using X-ray diffraction. As shown in Figure 2, the borophosphate gel powder shows the presence of single-phase crystalline BPO4 after heat treatment at 800 °C for 24 h in air. Figure 3 shows the XRD traces obtained on the gel samples derived using the MOSG process. XRD profiles obtained on the as-prepared powders with B2O3/P2O5/SiO2 ) 1:1:4 composition and B2O3/P2O5/SiO2 ) 1:1:8 stoichiometry indicate the typical amorphous nature characteristic of the gel powders (see parts a and e of Figure 3). The gel powders heattreated at 450 °C, below the crystallization temperature of 490 °C, still exhibit the amorphous nature (Figure 3(b)), while the sample heat-treated at 500 °C, which is just above the crystallization temperature, reveals the formation of BPO4 (Figure 3c). When the gel powders were heat-treated to a higher temperature of 800 °C (Figure 3d), the peak intensities corresponding to crystalline BPO4 are observed to increase in the samples corresponding to the stoichiometry of B2O3/P2O5/SiO2 ) 1:1:4 (114SG-800). The gel samples containing B2O3/P2O5/SiO2 ) 1:1:8 (118SG-800), on the other hand, exhibit decreasing intensities of the BPO4 phase in comparison to the “114SG800” sample due to the presence of a relatively smaller molar fraction of B2O3 and P2O5 available for the formation of BPO4 (refer to parts d and f of Figure 3). Analysis of IR Spectra. To investigate the structure of the glasses synthesized using the MOSG process, the gel powders were characterized for their molecular linkages using FTIR. The IR absorption spectra collected on the as-prepared and heattreated gel powders, derived using the MOSG process, are presented in Figure 4. The figure also shows, for comparison, the IR spectra collected on the borophosphate gel powders heattreated at 800 °C for 24 h in air. The borophosphate gel powder, which exhibits strong XRD peaks characteristic of single-phase crystalline BPO4 (Figure 2), shows a weak shoulder at 490 cm-1 related to BO4 group24 while exhibiting intense peaks corresponding to B-O-P (624 cm-1)25 and P-O-P (952 and 1149 cm-1)24,25 (see Figure 4a). The as-prepared powder, derived
Sol-Gel Based Synthesis
J. Phys. Chem. B, Vol. 102, No. 30, 1998 5747
Figure 3. X-ray diffraction traces of as-prepared and heat-treated borophosphosilicate gel powders derived using the MOSG process: (a) 114SGAP; (b) 114SG-450; (c) 114SG-500; (d) 114SG-800; (e) 118SG-AP; (f) 118SG-800.
using the MOSG process corresponding to the B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry, exhibits various absorption peaks as shown in Figure 4b. The assignment for these peaks and the corresponding references are listed in Table 2. As indicated in the table, the as-prepared powders derived from the MOSG process show intense peaks corresponding to Si-O-Si (466 cm-1), O-Si-O (800 cm-1), Si-O (1100 cm-1), B-O (1400 cm-1), and NH2 or OH (3210 cm-1) linkages and weak peaks corresponding to B-OH (560 cm-1), B-O-Si (670 cm-1), CH3 (1193, 1450, and 2960 cm-1), C-O (1630 cm-1), and C-H (2920 cm-1) linkages. In addition, there is a wide band spanning from 3300 cm-1 to 3700 cm-1 that could be assigned to the unreacted Si-OH and P-OH groups.31 All the heat-treated gel powders, derived using the MOSG process corresponding to B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry, show Si-O-Si (466 cm-1), B-O-P (624 cm-1), O-Si-O
(800 cm-1), B-O-Si (930 cm-1), Si-O (1100 cm-1), and B-O-B (1350 cm-1) linkages and a shoulder at 1400 cm-1 corresponding to B-O bonds (see curves c-e of Figure 4 and the expanded view shown in the inset). In addition, there is a wide band spanning from 3300 to 3600 cm-1 for gel samples heat-treated to 500 °C, which is related to the presence of residual Si-OH and P-OH groups. This band disappears in the sample that was heat-treated at 800 °C (see Figure 4e). The peak corresponding to the B-O-P (624 cm-1) linkage is also seen to appear at temperatures of 450 °C (below the crystallization temperature of BPO4, Tc ) 490 °C) and above, and the intensity of the peak continues to increase with the corresponding increase in the heat-treatment temperature. On the other hand, absorptions related to the CH3 groups (1193, 1450, and 2960 cm-1) almost disappear and the presence of only a small peak corresponding to C-O (1630 cm-1) vibration is seen after
5748 J. Phys. Chem. B, Vol. 102, No. 30, 1998
Kim and Kumta
Figure 4. Infrared absorption spectra collected on the as-prepared and heat-treated borophosphosilicate gel powders synthesized using the MOSG process as well as the heat-treated borophosphate gels: (a) BPO4; (b) 114SG-AP; (c) 114SG-450; (d) 114SG-500; (e) 114SG-800; (f) 118SG-800.
TABLE 2: Assignment of Peaks Observed in the IR Spectra Collected on the As-Prepared Powders Derived Using the MOSG and MOSP Processes
a
wavenumber (cm-1)
assignment
114SG-APa
114SP-APa
ref
466 560 670 800 1100 1193 1400 1450 1630 2920 2960 3210 3300-3700 (wide band)
Si-O-Si (bending) B-OH B-O-Si (symmetric stretching) O-Si-O (bending) Si-O (stretching) CH3 (rocking) B-O (stretching) CH3 (asymmetric bending) C-O (stretching) C-H (stretching) CH3 and CH2 NH2 (or possibly due to the residual OH) (possibly due to residual) P-OH (3300), Si-OH (3500, 3680)
s w w s s w s s w w w s vw (shoulder)
s vw (shoulder) w s s vw (shoulder) s vw (shoulder) w N vw (shoulder) s vw (shoulder)
15, 26 8, 27 8, 9, 15, 27 15, 27 9, 15, 27 27 9, 27 28 29 28 28 30, 31 31
s: strong intensity. w: weak intensity. vw: very weak intensity. N: no intensity.
heat treatment in air at 450 °C. This absorption peak decreases with a gradual increase in the heat-treatment temperature, as one would expect for gel samples heat-treated in air, and is completely eliminated after heat treatment at 800 °C. All the heat-treated samples also show the absorption centered at 545 cm-1. Figure 4f shows the IR spectra obtained from the MOSGderived samples corresponding to the B2O3/P2O5/SiO2 ) 1:1:8 composition, heat-treated at 800 °C. In comparison with the sample of B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry (see 114SG800 in Figure 4e), the relative intensities of absorptions related to B and P (B-O-P at 624 cm-1) are reduced owing to a reduction in the amounts of B2O3 and P2O5, while the relative intensities of the B-O (1400 cm-1) and B-O-B (1350 cm-1) linkages appear to remain almost unchanged. Thus, similar to the samples of B2O3/P2O5/SiO2 ) 1:1:4 composition, the MOSG-derived sample corresponding to B2O3/P2O5/SiO2 ) 1:1:8 stoichiometry also shows the presence of the B-O-B linkage at 1350 cm-1. At the same time, the absorption centered at 545 cm-1 is also lower in intensity in comparison to the B2O3/ P2O5/SiO2 ) 1:1:4 composition (114SG-800).
(2) MOSP Process. X-ray Analysis. The XRD traces obtained on the MOSP-derived samples are presented in Figure 5. The MOSP-derived samples also exhibit the same trend with respect to the formation of the BPO4 phase as the MOSGderived samples. However, there is a distinct difference in the peak intensities of the BPO4 phase observed in the MOSGderived and MOSP-derived samples (see 114SG-500 in Figure 3c and 114SP-500 in Figure 5c, respectively), which are heattreated at 500 °C. The MOSP-derived sample shows less intense peaks corresponding to the BPO4 phase than the MOSG-derived sample when the as-prepared powders are heat-treated at 500 °C for 24 h. However, the resultant peak intensities of the BPO4 phase observed in both the MOSG-derived (114SG-800, Figure 3d) and MOSP-derived (114SP-800, Figure 5d) samples after heat treatment at 800 °C are almost identical. Another important difference between the MOSG- and MOSP-derived samples corresponding to the B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry is the appearance of Si3(PO4)4. This phase is seen to appear only in the MOSP-derived sample heat-treated at 500 °C (see 114SP500 in Figure 5c), while it is not detected in the MOSG-derived
Sol-Gel Based Synthesis
J. Phys. Chem. B, Vol. 102, No. 30, 1998 5749
Figure 5. X-ray diffraction traces of as-prepared and heat-treated borophosphosilicate precipitated gel powders derived using the MOSP process: (a) 114SP-AP; (b) 114SP-450; (c) 114SP-500; (d) 114SP-800; (e) 118SP-AP; (f) 118SP-800.
TABLE 3: Chemical Analysis Conducted on the As-Prepared Powders Synthesized Using Both MOSG and MOSP Processes Corresponding to the Stoichiometry of B2O3/P2O5/SiO2 ) 1:1:4
a
sample
borona
phosphorusa
silicona
carbona
hydrogena
nitrogena
B/Pb
114SG-AP 114SP-AP
2.84 2.28
8.87 10.53
15.77 17.61
7.75 0.65
3.27 2.26
0.57 4.35
0.92 0.62
All analyses are in wt %. b “B/P” represents the molar ratio of boron to phosphorus. The initial molar ratio B/P is 1:1 for both samples.
sample heat-treated at the same temperature (see 114SG-500 in Figure 3c). We had previously reported that the formation of silicon phosphate could occur because of the loss of boron in the borophosphosilicate glass gels, which were initially synthesized using an identical molar ratio of boron and phosphorus.22 To confirm this hypothesis, chemical analysis was conducted on the as-prepared powders corresponding to an initial composition
of B2O3/P2O5/SiO2 ) 1:1:4, derived using the MOSP process. Results of the chemical analyses confirm the deficiency of boron compared with that of phosphorus in the MOSP-derived asprepared powder (refer to Table 3). The results of the chemical analysis shown in Table 3 reflect a molar ratio of B2O3/P2O5 ) 0.62:1 in the case of the MOSP-derived sample, although the starting nominal composition of B2O3/P2O5 was selected corresponding to a 1:1 molar ratio. Thus, there is a 38% loss of
5750 J. Phys. Chem. B, Vol. 102, No. 30, 1998
Kim and Kumta
Figure 6. Infrared absorption spectra collected on the as-prepared and heat-treated borophosphosilicate gel powders synthesized using the MOSP process as well as the heat-treated borophosphate gels: (a) BPO4; (b) 114SP-AP; (c) 114SP-450; (d) 114SP-500; (e) 114SP-800; (f) 118SP-800.
boron in the MOSP-derived as-prepared powder. On the other hand, the MOSG-derived as-prepared gel powder corresponding to the same initial composition shows a molar ratio of B2O3/ P2O5 ) 0.92:1, which reflects a smaller loss of boron (only 8%). Nevertheless, the formation of silicon phosphate is not seen to be detrimental, since it appears to be consumed into the glass phase during further heat treatment of the gel powders to 800 °C, leading to the formation of only BPO4 as the crystalline phase (see Figure 5d). The 118SP-800 sample exhibits lower intensities of the BPO4 phase than the 114SP800, consistent with the 118SG-800 sample (see Figures 5f and 3f). Analysis of IR Spectra. Powders synthesized using the MOSP process were also characterized for their molecular linkages using FTIR. Figure 6 shows the IR absorption spectra collected on the as-prepared and heat-treated powders, derived using the MOSP process. The IR spectrum, collected on the as-prepared powder corresponding to the B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry, is presented in Figure 6b. The assignment for these peaks and the corresponding references are also listed in Table 2. Similar to the MOSG-derived sample, the as-prepared powder obtained using the MOSP process shows intense peaks corresponding to Si-O-Si (466 cm-1), O-Si-O (800 cm-1), Si-O (1100 cm-1), B-O (1400 cm-1), and NH2 or OH (3210 cm-1) linkages and weak peaks corresponding to B-O-Si (670 cm-1) and C-O (1630 cm-1) linkages. In addition, there is also a wide band spanning the 3300-3700 cm-1 wavenumber range, which is related to the unreacted Si-OH and P-OH groups.31 However, the MOSP-derived as-prepared powder shows much less (or no) intensity of peaks corresponding to B-OH (560 cm-1), CH3 (1193, 1450, and 2960 cm-1), and C-H (2920 cm-1) linkages. Reduced peak intensities, corresponding to CH3 and C-H vibrations in the MOSP-derived asprepared powder compared with those of the MOSG-derived sample, imply less carbon in the MOSP-derived sample. Results of chemical analysis conducted on the as-prepared powders do confirm the above result (refer to Table 3). Table 3 clearly shows that the MOSP-derived sample contains significantly reduced amounts of carbon and hydrogen compared to the MOSG-derived sample. However, the nitrogen content of the
MOSP-derived as-prepared powder is much higher owing to ammonium hydroxide, which was added in excess to induce precipitation of the gel powders. All the heat-treated powders, derived using the MOSP process corresponding to the B2O3/P2O5/SiO2 ) 1:1:4 stoichiometry, show Si-O-Si (466 cm-1), O-Si-O (800 cm-1), B-O-Si (930 cm-1), Si-O (1100 cm-1), and B-O (1400 cm-1) linkages similar to the MOSG-derived gel powders (see curves c-e of Figure 6). The expanded view shown in the inset, however, reflects the strong predominance of B-O linkages in the heattreated MOSP-derived powders. Moreover, three distinct differences are detected in the IR spectra collected on the heattreated samples derived using the MOSG and MOSP processes. Contrary to the heat-treated powders obtained from the MOSG process, the MOSP-derived heat-treated samples do not exhibit the peak corresponding to B-O-B (1350 cm-1) linkages. Furthermore, the B-O-P (624 cm-1) linkage is absent in the MOSP-derived sample heat-treated at 450 °C (see Figure 6c) and can only be seen in the samples heat-treated above the crystallization temperature (Tc ) 490 °C) of BPO4 (refer to curves d and e of Figure 6). In addition, it can be seen that upon heat treatment to 450 °C, the pre-existing B-O-Si (670 cm-1) linkages become more clear and continue to remain at 500 °C as well. Further heat treatment to 800 °C causes a reduction in the intensity of this B-O-Si linkage, which is now only seen as a shoulder. Similar to the MOSG-derived gels, the heat-treated powders synthesized using the MOSP process also show the peak centered at 545 cm-1. However, similar to the B-O-P linkage, this peak is also seen to emerge only in samples heat-treated beyond 450 °C. The IR spectrum obtained from the MOSP-derived samples, corresponding to the stoichiometry of B2O3/P2O5/SiO2 ) 1:1:8 heat-treated at 800 °C, is presented in Figure 6f. In comparison to the samples of B2O3/P2O5/SiO2 ) 1:1:4 molar ratio (114SP800), the relative intensities of absorptions related to B and P (B-O-P at 624 cm-1) are decreased owing to the reduced initial amounts of B2O3 and P2O5, similar to the MOSG process. At the same time, the absorption centered at 545 cm-1 is also lower in intensity in comparison to the “114SP-800” sample. However, similar to the heat-treated MOSP-derived samples
Sol-Gel Based Synthesis
J. Phys. Chem. B, Vol. 102, No. 30, 1998 5751
corresponding to B2O3/P2O5/SiO2 ) 1:1:4 composition (114SP800), the MOSP-derived sample corresponding to the B2O3/ P2O5/SiO2 ) 1:1:8 stoichiometry (118SP-800) shows the strong B-O absorption at 1400 cm-1 while not exhibiting the presence of the B-O-B linkage at 1350 cm-1. Discussion The XRD and FTIR results indicate several distinct differences between the MOSG and MOSP processes which are related to the amorphous structure, and the corresponding differences in the evolution of crystalline BPO4. In relation to the crystallization of BPO4, the MOSP-derived sample heattreated at 500 °C shows two distinct differences in comparison with the MOSG-derived sample heat-treated at the same temperature. The first difference observed in the X-ray diffraction analysis is the presence of crystalline Si3(PO4)4 in the “114SP-500” sample. As mentioned in the previous section, the appearance of crystalline Si3(PO4)4 only in the MOSPderived sample could be a reflection of the significant loss of boron during the synthesis of the as-prepared powder using the MOSP process. Results of chemical analysis shown in Table 3 indicate a 38% loss of boron in the MOSP-derived as-prepared powders in comparison to the initial nominal starting composition. This loss of boron can be expected to occur during the centrifugal separation of the precipitate from the solution that contains the unbridged boron groups. On the other hand, there is very little loss of boron in the MOSG-derived gels, since this process was performed in a well-sealed glovebag. More importantly, most of the boron remains confined within the gel, since the procedure involves no extraction of solvent. Hence, the unbridged boron groups, if any, remain to undergo further reaction within the framework of the gel structure. The significant loss of boron in the MOSP-derived asprepared powder promotes the formation of Si3(PO4)4 at 500 °C, possibly via the reaction of uncondensed phosphorus [OP(OR)3-x(OH)x, where R ) C2H5] and silicon [Si(OR)4-x(OH)x, where R ) C2H5] groups. The presence of silicon phosphate, however, does not adversely affect the generation of BPO4 in the glass matrix, since it appears to be consumed into the glass phase during further heat treatment of the gel powders to 800 °C, resulting in the formation of only BPO4 as the crystalline phase (see Figure 5d). It can be therefore construed that the excess P remains bonded to silicon in the amorphous phase. On the other hand, the MOSG samples derived using the same initial composition do not show the crystallization of Si3(PO4)4, since the unreacted boron groups still exist within the gel and probably prevent its crystallization. The other difference in the X-ray diffraction data between the MOSG- and MOSP-derived samples heat-treated at 500 °C is the variation in the amount of the crystallized BPO4 phase. Although the ultimate amounts of crystallized BPO4 observed in both the MOSG- (see 114SG800 in Figure 3d) and MOSP-derived (see 114SP-800 in Figure 5d) samples are almost identical after heat treatment at 800 °C, the difference in the extent of crystallization of BPO4 at 500 °C suggests a possible variation in the initial glass structure of the as-prepared powders and the consequent steps involved in its crystallization. These differences in molecular linkages and structure can be better explained by analyzing the IR spectra, which provide information on the structure of the as-prepared and heat-treated powders. The most important difference in the IR spectra collected on the as-prepared powders derived using both MOSG and MOSP processes is the intensity of the absorption peak centered at 560 cm-1. This absorption peak is due to the presence of unreacted
Figure 7. Infrared absorption spectra collected on the as-prepared borophosphosilicate powders showing the expanded view of the molecular linkages in the 400-600 cm-1 wavenumber range: (a)114SG-AP; (b) 114SP-AP.
B-OH groups that do not undergo condensation. Figure 7 shows an expanded view of the IR spectra spanning across the 560 cm-1 region collected from the as-prepared powders derived using both MOSG and MOSP processes. Only the as-prepared powder obtained from the MOSG process exhibits an intense peak at this wavenumber. This is an indication that unreacted B-OH groups remain in the gel samples derived using the MOSG process, while these unreacted B-OH groups are largely removed in the MOSP process during centrifugation. This distinction is quite important, since the retention and presence of these unreacted B-OH groups in the MOSG-derived as-prepared powder could induce significant changes in the chemical homogeneity and gel structure after heat treatment, affecting the crystallization of BPO4. The presence of these unreacted B-OH groups observed only in the as-prepared powders derived from the MOSG process causes two main differences in the structure of the gels after heat treatment. One distinct difference between the heat-treated powders synthesized using the MOSG and MOSP processes is that the B-O-B linkages (1350 cm-1) appear only in the IR spectra of the samples obtained using the MOSG process. In addition, the B-O bond is seen as a shoulder at 1400 cm-1 in the MOSG-derived samples, while the MOSP-derived samples exhibit only the B-O linkage at 1400 cm-1 without showing the presence of any B-O-B linkage at 1350 cm-1 (see insets in Figures 4 and 6). The peak corresponding to the B-O-B linkage at 1350 cm-1 appears in all the heat-treated samples synthesized using the MOSG process and does not seem to change significantly with heat-treatment temperature. Moreover, it appears to be independent of the composition of the gel. This can be inferred with conviction, since the heat-treated gel sample corresponding to the initial starting compositions of B2O3/P2O5/ SiO2 ) 1:1:8, i.e., “118SG-800” sample, also exhibits B-O-B linkages at 1350 cm-1 as shown in Figure 4f and in the expanded view displayed in the inset. These results therefore relate to the fact that only the MOSG-derived as-prepared gel powder contains unreacted B-OH groups as can be seen in Figure 7. These unreacted B-OH groups in the as-prepared powders derived using the MOSG process can be more favored to undergo the expected condensation reaction during subsequent heat treatments, thus forming the B-O-B linkages in the glasses, which could eventually lead to segregation of crystalline B2O3 as reported by Nogami and Moriya9 and also as reported by us in an earlier publication.32 On the other hand, the IR
5752 J. Phys. Chem. B, Vol. 102, No. 30, 1998 spectra collected on the as-prepared powder derived using the MOSP process do not indicate the presence of B-OH groups, owing to the removal of these unreacted B-OH groups during extraction of the liquid after sol-precipitation. Therefore, the heat-treated samples in this case do not exhibit B-O-B linkages in the IR spectra. The other distinct difference between the heat-treated samples derived using the MOSG and MOSP processes is that the MOSG-derived sample heat-treated at 450 °C (below the crystallization temperature, Tc ) 490 °C) shows the presence of B-O-P (624 cm-1) linkages, a characteristic of the crystalline phase of BPO4 (see Figure 4c), while the MOSPderived sample heat-treated at the same temperature exhibits only weak absorption corresponding to B-O-Si (670 cm-1) linkages without showing the presence of B-O-P units. It should be remembered that at this temperature, both the MOSGand MOSP-derived samples are X-ray amorphous (Figures 3b and 5b). Furthermore, the MOSG-derived sample heat-treated at 500 °C (just above the crystallization temperature) shows a more intense peak corresponding to the B-O-P linkages in comparison with MOSP-derived sample heat-treated at 500 °C (refer to Figures 4d and 6d). This result is also consistent with the X-ray data, which show that the MOSG-derived sample reveals more intense peaks corresponding to crystalline BPO4 than the MOSP-derived sample heat-treated at 500 °C (refer to Figures 3c and 5c). Consequently, it can be concluded that the unreacted B-OH groups present in the MOSG-derived gels not only result in the formation of B-O-B units but also react with the unreacted P-OH groups to form B-O-P linkages during heat treatments even below the crystallization temperature (Tc ) 490 °C) of the BPO4 phase. The formation of these B-O-P linkages facilitates the nucleation and crystallization of BPO4 just above its crystallization temperature. Thus, the differences observed in the XRD peak intensities in the samples heat-treated at 500 °C can also be related to the formation of these B-O-P linkages (see Figures 3c and 5c). The XRD and IR results of the MOSP-derived samples suggest a different mechanism for the formation of BPO4. The precipitation reaction leads to the separation of the condensed alkoxide units, leaving most of the unreacted groups in solution. Hence, the as-prepared powder contains boron largely bonded to silicon (B-O-Si linkages) initially in the glassy borophosphosilicate phase. Heat treatment of the gel powders subsequently leads to an increase in the peak intensity of B-O-P linkages (624 cm-1) along with the decrease in the intensity of B-O-Si linkages (670 cm-1) as shown in the IR spectra (Figure 6). Consequently, the formation of B-O-P linkages and crystallization of BPO4 are largely affected by the extraction of boron from the B-O-Si groups in the glass matrix and its simultaneous reaction with Si3(PO4)4. The generation of BPO4 via this mechanism is relatively slow at 500 °C, since the formation of B-O-P linkages via separation of boron from B-O-Si groups is more difficult than that via the reaction of unreacted B-OH and P-OH groups. However, heat treatment at a high temperature of 800 °C helps to accelerate the kinetics of this reaction, thereby rendering the extent of crystallization of BPO4 to be much larger compared to that at 500 °C. Therefore, the XRD data for both the MOSGand MOSP-derived samples heat-treated at 800 °C show almost identical peak intensities of BPO4, although the MOSG-derived samples contain a larger content of boron and reveal much more intense peaks corresponding to BPO4 in comparison with the MOSP-derived powders when heat-treated at 500 °C. The removal of unreacted B-OH groups during centrifugal separa-
Kim and Kumta tion in the MOSP process is beneficial, since there is no B-O-B segregation seen in the MOSP-derived powders, unlike the case of the heat-treated gel powders derived using the MOSG process (compare the IR spectra in Figures 4 and 6). At the same time, almost identical amounts of BPO4 appear to be crystallized in the two processes, despite the loss in boron causing a change in the composition of the as-prepared powders. The other interesting aspect is the presence of the unknown absorption centered at 545 cm-1, which is seen in the heattreated samples derived using both MOSG and MOSP processes containing B-O-P linkages (624 cm-1), as well as in the borophosphate gel powders heat-treated at 800 °C. This absorption is also seen to increase in both the MOSG- and MOSP-derived samples with a corresponding increase in the heat-treatment temperature (along with an increase in crystallinity of the BPO4 phase and intensity of B-O-P linkages). Therefore, this unknown absorption can be assigned to be characteristic of the borophosphate phase. Finally, aside from the differences in the mechanism of formation of BPO4 in the two processes, it should be also noted that the as-prepared powder synthesized using the MOSP process exhibits much reduced peak intensities corresponding to CH3 vibrations (1193, 1450, and 2960 cm-1; refer to Figures 4b and 6b) in comparison with the as-prepared powder derived using the MOSG process. This result as well as the chemical analysis data in Table 3 implies that the as-prepared powder derived using the MOSP process contains less carbon than the MOSG-derived gel. This is because the precipitation reaction and the centrifugation step in the MOSP process separate the residual alcohol and the soluble unreacted alkoxide units from the condensed solid precipitate. A lower carbon content indicates that the MOSP-derived powders should be more amenable for sintering to form glass ceramics with minimum or significantly reduced porosity in comparison to the MOSGderived samples. Porosity in sintered gel-derived glass ceramics is known to occur mainly owing to loss of carbon through the gel during sintering. A lower carbon content would, therefore, lead to sintered compacts containing largely reduced porosity. This result has been observed and also reported by us in earlier publications.20,21 The reduced carbon content makes the MOSP process very viable to process dense low-dielectric-constant glass-ceramics for microelectronic application as we have shown in our earlier publications.20,21 Conclusions Borophosphosilicate glasses and glass-ceramics were synthesized using both modified oxide sol-gel (MOSG) and modified oxide sol-precipitation (MOSP) processes. The two approaches lead to gels and precipitated powders possessing different molecular structures. As a result, they display different crystallization behaviors. The as-prepared powders, corresponding to the B2O3/P2O5/SiO2 ) 1:1:4 and 1:1:8 stoichiometry derived using the MOSG process, contain more unreacted B-OH groups compared to those derived using the MOSP process. These unreacted B-OH groups in the as-prepared powder undergo condensation reaction during heat treatments and cause microsegregation of B-O-B units that could eventually lead to formation of crystalline B2O3. At the same time, the B-OH and P-OH groups react to form B-O-P linkages below the crystallization temperature (Tc) of BPO4. The formation of these B-O-P linkages in the MOSG-derived samples facilitates the nucleation and crystallization of BPO4 in comparison to the MOSP-derived samples. As a result, gels derived using the MOSG process exhibit a larger amount of
Sol-Gel Based Synthesis the crystallized BPO4 phase in comparison to the MOSP-derived samples, when heat-treated just above the crystallization temperature. The MOSP process, on the other hand, leads to gel powders with significant loss of boron. This is because the precipitation reaction forces the separation of the condensed alkoxide units from alcohol and water. As a result, there is considerably less amount of solvent and residual alkoxy groups retained in the as-prepared powders. This reduction in the amount of boron due to the removal of unreacted B-OH species alters the mechanism of formation of BPO4. The formation of BPO4 is now mainly driven by extraction of boron from the borosilicate glass and its reaction with Si3(PO4)4. However, the extent of crystallization of BPO4 at 800 °C is similar to that of the MOSGderived gel powders. Thus, the MOSG- and MOSP-derived gel powders at 800 °C reveal almost the same extent of crystallized BPO4 except that the MOSG-derived powders contain segregated B-O-B units due to the presence of excess unreacted B-OH groups. Moreover, the precipitation process (MOSP) results in powders with much less carbon content, thereby rendering the powders more amenable for forming dense glass ceramics in comparison to the MOSG process. Consequently, the MOSP process offers an easy, promising, and viable solution for synthesizing homogeneous glasses and glass ceramics in the borophosphosilicate system. Acknowledgment. The authors acknowledge the donors of the National Science Foundation (Grant DMR-9301014), Research Initiation Award (RIA) from the Engineering Division of NSF (Grant CTS-9309073) and NSF (Grant CTS-9700343), and the Faculty Development Fund of Carnegie Mellon University. The authors also thank Dr. M. A. Sriram and Mr. R. Hsu whose initial experiments formed the basis of the present work. References and Notes (1) Tummala, R. R., Rymaszewski, E. J., Eds. Microelectronic Packaging Handbook; Van Nostrand Reinhold: New York, 1989. (2) Geiger, G. Bull. Am. Ceram. Soc. 1990, 69, 1131. (3) Rabinovich, E. M. Trans. ASME: J. Electron. Packag. 1989, 111, 183. (4) Tummala, R. R.; Shaw, R. B. High Technology Ceramics; Vincenzini, P., Ed.; Elsevier Science: Amsterdam, 1987; p 75. (5) Sprague, J. L. IEEE Trans. Compon., Hybrids, Manuf. Technol. 1990, 13, 390.
J. Phys. Chem. B, Vol. 102, No. 30, 1998 5753 (6) Kinsman, K. R. J. Met. 1988, 6, 7. (7) Jabra, R.; Phalippou, J.; Zarzycki, J. J. Non-Cryst. Solids 1980, 42, 489. (8) Irwin, A. D.; Holmgren, J. S.; Zerda, T. W.; Jones, J. J. Non-Cryst. Solids 1987, 89, 191. (9) Nogami, M.; Moriya, Y. J. Non-Cryst. Solids 1982, 48, 359. (10) Villegas, M. A.; Capel, F.; Fermandez Navarro, J. M. J. Mater. Sci. Lett. 1988, 7, 791. (11) Matsuyama, I.; Susa, K.; Suganuma, T.; Katouyama, T.; Obayashi, H. U.K. Patent Application GB 2075003, 1981. (12) Yoldas, B. E. J. Mater. Sci. 1979, 14, 1843. (13) MacDowell, J. F.; Beall, G. H. Materials Research Society Symposium Proceedings, AdVanced Electronic Packaging Materials; Barfknecht, A. T., Partridge, J. P., Chen, C. J., Li, C.-Y., Eds.; Materials Research Society: Pittsburgh, 1989; Vol. 167, p 43. (14) MacDowell, J. F.; Beall, G. H. Proceedings of the 1st International Science and Technology Conference, Materials and Processes for Microelectronic Systems, Ceramic Transactions; Nair, K. M., Pohanka, R., Buchanan, R. C., Eds.; American Ceramic Society: Columbus, OH, 1989; Vol. 15, p 259. (15) Brinker, J. C.; Scherer, G. W. Sol-Gel Science: The Physics and Chemistry of Sol-Gel Processing; Academic Press: Boston, 1990. (16) Kumta, P. N.; Hackenberg, R. E.; McMichael, P. H.; Johnson, W. C. Mater. Lett. 1994, 20, 355. (17) Guglielmi, M.; Carturan, G. J. Non-Cryst. Solids 1988, 100, 16. (18) Phule, P. P.; Deis, T. A.; Dindiger, D. G. J. Mater. Res. 1991, 6, 1567. (19) Livage, J.; Sanchez, C.; Henry, H.; Doeuff, S. Solid State Ionics 1989, 32/33, 633. (20) Hsu, R.; Kumta, P. N.; Feist, T. P. J. Mater. Sci. 1995, A30, 3123. (21) Hsu, R.; Kim, J. Y.; Kumta, P. N.; Feist, T. P. Chem. Mater. 1996, 8, 107. (22) Kumta, P. N.; Sriram, M. A. J. Mater. Sci. 1993, A28, 1097. (23) Cotton, F. A.; Wilkinson, G. AdVanced Inorganic Chemistry, 5th ed.; Wiley: New York, 1988; pp 168, 400. (24) Ray, N. H. Phys. Chem. Glasses 1975, 16, 75. (25) Izawa, T.; Shibata, N. Appl. Phys. Lett. 1977, 31, 33. (26) Brinker, C. J.; Haaland, D. M. J. Am. Ceram. Soc. 1983, 66, 758. (27) Prassas, M.; Hench, L. Ultrastructure Porcessing of Ceramics, Glasses and Composites; Hench, L., Ulrich, D. R., Eds.; Wiley: New York; 1984; p 100. (28) McKenzie, D. R.; McPhedran, R. C.; Savvides, N.; Cockayne, D. J. H. Thin Solid Films 1983, 108, 247. (29) Babonneau, F.; Coury, L.; Livage, J. J. Non-Cryst. Solids 1990, 121, 153. (30) Han, H.-X.; Feldman, B. J. Solid State Commun. 1988, 65, 921. (31) Woignier, T.; Phalippou, J.; Zarzycki, J. J. Non-Cryst. Solids. 1984, 63, 117. (32) Sriram, M. A.; Kumta, P. N. Materials Research Society Symposium Proceedings, Synthesis and Processing of Ceramics: Scientific Issues; Rhine, W. E., Shaw, T. M., Gottschall, R. J., Chen, Y., Eds.; Materials Research Society: Pittsburgh, 1992; Vol. 249, p 59.