The influence of crystallised Fe3O4 on the magnetic properties of coprecipitation-derived ferrimagnetic glass–ceramics

Acta BIOMATERIALIA Acta Biomaterialia 1 (2005) 421–429 www.actamat-journals.com

The influence of crystallised Fe3O4 on the magnetic properties of coprecipitation-derived ferrimagnetic glass–ceramics O. Bretcanu a

a,*

, S. Spriano a, E. Verne´ a, M. Co¨isson b, P. Tiberto b, P. Allia

c

Materials Science and Chemical Engineering Department, Politecnico di Torino, c-so Duca degli Abruzzi 24, 10129 Torino, Italy b Materials Department, National Electrotechnic Institute Galileo Ferraris, Strada delle Cacce, 91, 10135 Torino, Italy c Physics Department, Politecnico di Torino, c-so Duca degli Abruzzi 24, 10129 Torino, Italy Received 19 November 2004; received in revised form 2 February 2005; accepted 19 April 2005

Abstract Ferrimagnetic glass–ceramics are potential candidates for magnetic induction hyperthermia, which is one form of inducing deepregional hyperthermia, by using a magnetic field. The aim of this work was to analyse the influence of the amount of crystallised magnetite on the magnetic properties of glass–ceramic samples. Thus, two different ferrimagnetic glass–ceramics with the composition of the system Na2O–CaO–SiO2–P2O5–FeO–Fe2O3 were prepared by melting at 1500
C for 30 min of the coprecipitationderived starting products. The X-ray diffraction patterns show the presence of nanometric magnetite crystals in a glassy matrix after cooling from melting temperature. The estimated amount of crystallised magnetite varies between 20 and 45 wt.%, as a function of the chemical composition. The morphology of the crystals was studied by scanning electron micrography and transmission electron micrography. Glass transition temperature and thermal stability were investigated by differential thermal analysis. Magnetic hysteresis cycles were analysed using a vibrating sample magnetometer with a maximum applied field of 17 kOe, at room temperature, in quasi-static conditions. Calorimetric measurements were carried out using a magnetic induction furnace. The power losses estimated from calorimetric measurements under a magnetic field of 40 kA/m and 440 kHz are 65 W/g for the glass–ceramic with lower iron oxides content and 25 W/g for the glass–ceramic with higher iron oxide content.  2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Ferrimagnetic; Glass–ceramic; Coprecipitation; Hyperthermia; Cancer

1. Introduction Hyperthermia is a promising approach in cancer therapy. At present, hyperthermia treatments utilizing regional perfusion with heated blood, implantation of a heat source and electro-coagulation, ultrasound, microwave or another electromagnetic energy source are used clinically [1]. However, the technical problem with these methods is the difficulty of local uniform heating of the tumour only, without damaging normal tissue.

*

Corresponding author. Tel.: +39 011 5644708; fax: +39 011 5644699. E-mail address: [email protected] (O. Bretcanu).

Moreover, most of them involve invasive heat application. It has been reported that ferro or ferrimagnetic particles can heat the tumour locally without damaging normal tissue. These magnetic particles are easily incorporated into a tumour and generate heat under an alternating magnetic field mostly by hysteresis loss [2–7]. It is known that the heat generation depends mainly on the magnetic properties of the implant, the magnetic field parameters and the characteristics of the tissue. One of the most important requirements in the field of materials for magnetic hyperthermia is to tailor the desired magnetic properties of the implant. The mechanism of heat production depends mainly on the microstructure of the implanted magnetic material (crystals structure, magnetic domain structure, magnetic anisotropy,

1742-7061/$ - see front matter  2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actbio.2005.04.007

422

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

residual stress, particles size, grain shape, crystals imperfections such as impurities and dislocations). Therefore, by controlling the magnetic properties of implanted materials we can adjust the heat generation under an oscillating magnetic field. Ferrimagnetic glass–ceramics are potential candidates for magnetic induction hyperthermia [6,8–12]. In our previous study [13] we showed that a ferrimagnetic glass–ceramic containing 45 wt.% of magnetite has a saturation magnetisation of 34 emu/g and a coercive force of 85 Oe. The estimated heat generation of this glass–ceramic using a magnetic field of 40 kA/m and a frequency of 440 kHz was 25 W/g. This material also showed a bioactive behaviour, as after 2 weeks of soaking in a simulated body fluid the formation of a hydroxylapatite layer on their surface was observed [13]. The aim of this work was to study the influence of chemical composition, in particular the concentration of iron ions, on magnetic properties of ferrimagnetic glass–ceramics in the system Na2O–CaO–SiO2–P2O5– FeO–Fe2O3. The knowledge of the material properties plays a key role in the understanding of the temperature distribution within the tissue by applying magnetic induction hyperthermia. Modifying the chemical composition, we are able in principle to adjust the heat generation of these ferrimagnetic glass–ceramics in an alternating magnetic field. Consequently, the first step in controlling the heat generation is to study the effect of the amount of crystallised magnetite on the microstructure of ferrimagnetic glass–ceramics, which has a crucial effect on magnetic properties. In this paper two different compositions in the system Na2O–CaO–SiO2–P2O5–FeO–Fe2O3 were analysed in order to quantify the influence of crystallographic structure and microstructure on magnetic properties. These glass–ceramics, containing different amounts of magnetite, were synthesised by melting the coprecipitationderived starting products at 1500
C for 30 min. Thus, the raw materials were the powders obtained by coprecipitation of soluble salt solutions in distilled water. The coprecipitation method is commonly used to prepare fine-dispersed ceramic powders. Using this procedure, the reagents are mixed at colloidal or molecular scale, and the homogeneity will be improved [14,15]. Our target was to obtain a glass–ceramic, not a ceramic material, so a melting step was necessary to achieve the nucleation of magnetite in a liquid-derived amorphous phase. The coprecipitation step was introduced in order to melt precursors with a high degree of dispersion, homogeneity and purity. X-ray diffraction (XRD), scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), transmission electron microscopy (TEM) and differential thermal analysis (DTA) techniques were used to characterise the ferrimagnetic glass–ceramics. The magnetic properties were evaluated by using a vibrating

sample magnetometer (VSM), with a maximum applied field of 17 kOe, at room temperature. Calorimetric measurements were carried out using a magnetic induction furnace.

2. Materials and methods The compositions of the two ferrimagnetic glass– ceramics are reported in Table 1. The quantities of FeO and Fe2O3 were chosen in order to obtain a theoretical amount of magnetite of 35 and 45 wt.% in the two samples. These theoretical percentage values provide the names for the samples (S35 and S45). The gravimetric ratio between the remaining oxides (Na2O, CaO, SiO2 and P2O5) corresponds to Bioglass (45S5) and was chosen in order to confer to the glass–ceramics the highest bioactivity index. The physical-chemical and magnetic properties of sample S45 were analysed earlier [13]. The raw materials for the preparation of these glass–ceramics were obtained by coprecipitation from aqueous soluble salts solutions. This method is described elsewhere [13,16,17]. The starting reagents were water soluble salts: NaNO3, Ca(NO3)2 Æ 4H2O, (NH4)2HPO4, FeSO4 Æ 7H2O, Fe(NH4)(SO4)2 Æ 12H2O and colloidal SiO2 solution, which coprecipitate in a basic medium of NH4OH. The two resulting precipitates were washed with distillate water and filtered. Then the obtained powders were dried at 150
C in oven. The dried powders were milled in a ball mill for 30 min and then were thermally treated for decomposition in a furnace, in air, at 900
C for 3 h, using a rate of 10
C/min. After heating, the resulting powders were crushed in a mortar to deagglomerate the particles and then melted in a platinum crucible at 1500
C for 30 min in air atmosphere. The melts were poured onto a copper plate at room temperature, obtaining two dark glass–ceramics, with a glassy aspect. The thin, dark red oxidised layer obtained at the surface was eliminated by polishing the samples with SiC paper. Then the samples were washed in distillate water in an ultrasonic cleaner. One part of the glass–ceramic pieces was ground and sieved. The fraction smaller than 45 lm was used for further analysis. In order to obtain good reproducibility, the raw materials, molar ratio between the ions, addition order,

Table 1 Composition of glass–ceramic samples No.

1 2

Sample

S35 S45

Composition (wt.%) Na2O

CaO

SiO2

P2O5

FeO

Fe2O3

15.9 13.5

15.9 13.5

29.3 24.7

3.9 3.3

10.9 14

24.1 31

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

precipitation pH, number of precipitate washing and heat treatment temperatures, must be respected. The crystalline phases were analysed by Philips XPert diffractometer with CuKa radiation, using a step of 0.02
(2h) and a time per step of 15 s. The diffraction lines were identified using the ‘‘XPert HighScore’’ program, with the PCPDFWIN database (2002 JCPDSInternational Centre for Diffraction Data). The profile fitting of the diffraction pattern was performed by using the same software, in order to determine the magnetite crystallographic parameters (crystallite size, lattice strain and lattice constant). The lattice constant was estimated by linear extrapolation to F(2h) = 0, using the Taylor and Sinclair method [18]. Quantitative analyses of glass–ceramic samples were carried out using the MAUD program (Material Analysis Using Diffraction). This software is based on the Rietveld method [19] and approximates a silica glass structure with a nanocrystalline model where the longrange order is lost and a crystallite size is about one cell [20–22]. Pure magnetite powder (Aldrich, 98% purity, grain size <5 lm) and completely amorphous phase were used as references in order to determine the scale factor. The completely amorphous phase was obtained by melting the S35 crushed powders, obtained after the heat treatment at 900
C, at 1600
C and rapid quenching in water. The nanocrystalline model of silica glass phase reported in this software was refined in order to obtain a good fitting for our structure. The microstructure of the samples was examined using a Philips 515 SEM coupled with EDS EDAX PV 9900 and a Philips CM12/STEM TEM PW 6030 working at 120 kV. For SEM analysis, small polished glass–ceramic pieces (5 · 5 · 2 mm3) were chemically treated with a solution 5% volume of HF:HNO3 in a molar ratio 1:1 for 1 min, washed with distilled water and then dried at room temperature. For TEM analysis, powder samples were suspended in isopropanol using an ultrasonic cleaner. A few drops of the suspension were placed on microgrids (copper grids coated with thin carbon film, 200 mesh) and then left to dry at room temperature. Glass transition, endothermic and exothermic process temperatures of the two samples were determined by a Perkin Elmer DTA 7 TAC7/DX in argon atmosphere, using a heating rate of 20
C/min. Pure alumina powder was used as reference material and for base line determination. The magnetic hysteresis cycles for the powder samples were determined by using a VSM at room temperature, in quasi-static conditions. The cycles were measured by applying two different external magnetic fields of 10 kOe and 500 Oe. Magnetite was used as a reference material. The particles of magnetite (Aldrich) had a mean diameter of 5 lm. The calibration curve

423

was made using a nickel sphere as a standard reference material, positioned in the centre of the pick-up coils. Calorimetric measurements were performed using a magnetic induction furnace, with a magnetic field intensity of 40 kA/m and 440 kHz frequency. Small pieces of glass–ceramic samples were immersed in 20 ml water in a plastic container and the initial water temperature Ti was measured with a thermocouple. After then, the assembly was placed in the centre of the coil and the alternating magnetic field was applied for 2 min. Right after the power was switched off, the container was shaken to equilibrate the water temperature and the final water temperature Tf was measured. The specific power loss was calculated using the following equations: Q tm Q ¼ mw  cw  DT

ð2Þ

DT ¼ T f  T i

ð3Þ

P¼

ð1Þ

where P Q t m mw cw DT Ti Tf

specific power loss (W/g) heat absorbed by water (J) time (s) mass of sample (g) mass of water (g) waters specific heat (J/g
C) temperature variation (
C) initial temperature of water (
C) final temperature of water (
C)

For Eq. (2), we consider the waters density to be 1.00 g/ml and its heat capacity 4.2 J/g
C. The samples weight (0.05–0.1 g) was much smaller than the waters mass (20 g) and the heat absorbed by the sample was neglected. 3. Results and discussion The X-ray diffraction patterns of the glass–ceramics after melting are shown in Fig. 1. The picture presents patterns corresponding to the common structure of magnetite (Fe3O4). The diffraction lines of crystallised magnetite are a little shifted, as compared with reference data, indicating a slight variation of the lattice constant of magnetite. Magnetite is the only crystalline phase identified in the two samples. The characteristic halo of the amorphous phase can be observed in the range of 25–35
(2h). S35 sample contains a smaller quantity of iron oxides and hence a higher amount of amorphous residual phase, which is more visible on the XRD pattern. It can be clearly seen that both glass–ceramic samples have a high degree of crystallinity, revealed by sharp peaks. Hence, S35 sample exhibits more broadened peaks than S45, and

424

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429 8.380

M

8.375

Y = 8.3796 - 0.00784*X

8.365

M

M

M

M

M

(b)

M

M

a0 (Å)

Intensity (a.u.)

8.370

8.360 8.355 8.350

M

8.345 M M

M M

M

M

R = - 0.99662 SD = 0.00118

8.340

M

(a)

8.335 8.330

10

20

30

40

50

60

70

0

1

2

3

4

5

6

F (2θ)

2θ

Fig. 1. XRD patterns of glass–ceramics: (a) S35 and (b) S45 (M = magnetite).

Table 2 Magnetite crystallographic parameters estimated from XRD pattern Magnetite parameters

S35

S45

Mean crystals size, d (nm) Lattice strain, e (%) ˚) Lattice constant, a0 (A Theoretical quantity of magnetite (wt.%) Real quantity of magnetite (wt.%)

34 ± 5 0.35 ± 0.01 8.38 35 20

54 ± 9 0.21 ± 0.06 8.40 45 45

consequently, smaller crystal size. The relative intensities of magnetite diffraction lines with respect to the amorphous halo are higher in the S45 sample. Therefore, the quantity of magnetite crystallised in the case of S45 is higher. Magnetite probably crystallised during cooling from melting temperature to room temperature and a successive heat treatment for crystallisation was not necessary. Magnetite crystallographic parameters estimated after profile fitting of XRD patterns for both glass–ceramic samples are presented comparatively in Table 2. The crystal size of magnetite increases from 34 nm in S35 sample to 54 nm in S45, so, the crystallite size grows with the amount of iron oxides. The lattice strain of magnetite decreases from 0.35% in S35 sample to 0.21% in S45, so as the crystal dimensions increase, the lattice distortion decreases. The lattice constant of magnetite increases ˚ in S35 sample to 8.40 A ˚ in S45. These values from 8.38 A are obtained by linear extrapolation to F(2h) = 0 [18]. The extrapolation curve for S35 sample is presented in Fig. 2. The statistical parameters (correlation coefficient R = 0.99662 and standard deviation SD = 0.00118) indicate a good fitting. The lattice constants for the two samples are slightly different from those reported ˚ ). These differences in the literature (a0 = 8.393–8.399 A are related to the small shift of the diffraction patterns of glass–ceramics. The quantitative analysis was carried out by using the Rietveld method. The obtained values are presented on

Fig. 2. Extrapolation curve of lattice parameter a0 of magnetite crystals for S35 sample.

the last row of Table 2. The quantity of magnetite crystallised is 20 wt.% for the S35 sample and 45 wt.% for the S45 sample. As stated earlier, the theoretical stoichiometric quantities of magnetite in the two samples are 35 and 45 wt.%, respectively. Consequently, we can assume that for the S35 sample, 20 wt.% of iron ions crystallises as magnetite and the remaining 15 wt.% are in the glassy matrix. On the other hand, for the S45 sample, all the iron ions crystallised as magnetite. The reason that magnetite is completely crystallised for the S45 sample and partially crystallised for the S35 sample, even though the experimental synthesis conditions were similar, can be explained in terms of chemical composition and glass–ceramic structure. The higher content of the calcium, sodium, silicon and phosphorus oxides presented in the glass–ceramic can prevent the complete crystallisation of magnetite, by forming solid solutions with iron oxides [12,23]. Thus, a part of the iron oxides remain entrapped in the matrix. Also, the lattice strain for the S35 sample is higher than for S45 and the resulting internal forces/stresses can oppose the nucleation and crystal growth of magnetite. Consequently, the quantity of magnetite crystallised in the case of the S35 sample will be lower than the theoretical amount, as was confirmed by experimental data. SEM micrograph of glass–ceramic samples S35 and S45 (after etching) are shown in Fig. 3. The two samples (a and b) show similar characteristics. Small columns with octahedral crystals, uniformly distributed in the matrix, can be observed. This octahedral morphology is typical of magnetite (spinel structure). The crystal size of S35 is smaller than 0.2 lm. S45 has a larger size distribution of crystals, with an average dimension smaller than 0.5 lm. It can be observed that the columns of S45 sample are larger, containing well developed magnetite crystals, while S35 sample contains smaller columns, formed by smaller crystals.

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

425

Fig. 3. SEM micrograph of glass–ceramics: (a) S35 and (b) S45 after etching.

Heat Flow Endo Down (a.u.)

P1

(a) P2 Tg1

P3 P5 P4

Tg2

(b)

P6 P7

200

400

600

800

1000

P8

1200

Temperature (°C) Fig. 4. TEM micrograph of glass–ceramics S45 (magnification 430 · 103).

A characteristic TEM micrograph of S45 glass–ceramic sample is shown in Fig. 4. Although the glass–ceramic particles tend to agglomerate, small dark round particles of magnetite can be distinguished. The black round particles indicated by the arrows were identified as magnetite by electron diffraction. The particle sizes are relatively uniform, with dimensions varying in the range 5–10 nm. DTA curves of S35 and S45 powder samples are presented in Fig. 5. Both glass–ceramic samples exhibit similar behaviour. It should be noted that in the case of the S35 sample, all the exothermic and endothermic peaks are clearly shifted to lower temperatures. This fact indicates that with decreasing the magnetite phase in the glass, the thermal transformation processes occur at lower temperatures. Furthermore, all the peaks of the S35 sample have a larger area with respect to the ones of the S45 sample. Both curves show a glass transition temperature (Tg) typical of an amorphous phase. The

Fig. 5. DTA curves of glass–ceramics: (a) S35 and (b) S45.

presence of the glass transition temperature confirms the presence of a reasonable amount of residual amorphous phase in the glass–ceramic samples. The glass transition temperatures of the glass–ceramic S35 (T g1 ) and S45 (T g2 ) are very similar and occur at 640
C and 638
C, respectively. These temperature values are typical for glasses containing iron ions [24]. In the case of sample S35, one exothermic peak P1 occurring at 870
C and two endothermic peaks, P2 at 1047
C and P3 at 1200
C, are observed. The thermal analysis of S45 shows two exothermic peaks, P4 and P5 at 830
C and 890
C, respectively, as well as three small endothermic peaks P6 at 1190
C, P7 at 1225
C and P8 at 1275
C. Both samples appear to be completely melted after the DTA analysis. This fact does not allow for the thermal transformation processes to be identified. In order to assign the exothermic peaks to the crystallisation of glass–ceramics, sample S45 was heat treated first at 700
C, before crystallisation, and than at

426

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

900
C, after crystallisation (see P4 and P5 exothermic peaks in the DTA curve in Fig. 5), for 6 h and then subjected to XRD analysis. After heat treatment at 700
C the only crystallised phase is magnetite, while the identified crystalline phases at 900
C are magnetite, hematite and calcium sodium iron silicate. Thus, the two exothermic peaks P4 and P5 may be attributed to the crystallisation of hematite and the mixed calcium sodium iron silicate. Conversely, the endothermic peaks may be attributed to the melting of these crystalline phases. In the case of the S35 sample, the XRD analysis of glass–ceramic sample heat treated at 900
C (after crystallisation) reveals the presence of magnetite and calcium sodium iron silicate, while at 750
C (before crystallisation) magnetite is the unique crystalline phase. Therefore, the exothermic peak P1 can be attributed to the crystallisation of calcium sodium iron silicate while the other two endothermic peaks may be attributed to the melting of these crystalline phases. The hysteresis curves for powder glass–ceramic samples and magnetite under a magnetic field of 10 kOe are shown in Fig. 6. It can be observed that the two samples exhibit similar magnetic behaviour, characteristic for soft magnetic materials, with a thin hysteresis cycle and low coercive field (ffi100 Oe). The saturation magnetisation varies from 15 emu/g for S35 to 33 emu/g for S45 sample, while the coercive field changed from 144 Oe for S35 to 82 Oe for S45. The S35 glass–ceramic saturates at a maximum field of 2 kOe, whereas S45 particles are saturated at a value of the field two times higher (4 kOe). In contrast, the remanence magnetisation is about two times higher for S35 (4.45 Oe) than for S45 (2.45 Oe). The reported results are summarised in Table 3. The saturation magnetisation of magnetite particles is around 74 emu/g, and the coercive force is 150 Oe. The coercive fields of these materials are emphasised on the right bottom part of Fig. 6. The saturation magnetisa-

80 60

20 0 4

-20

Ms (emu/g)

Ms (emu/g)

40

Magnetite S45 S35

-40 -60

2 0 -2 -4 -200

-80 -10000

-5000

0

-100

0 100 H (Oe)

5000

200

10000

H (Oe) Fig. 6. Room temperature hysteresis cycle up to 10 kOe for glass– ceramics.

Table 3 Magnetic parameters estimated from hysteresis cycle Magnetic parameters

S35

S45

Saturation magnetisation, Ms (emu/g) Coercive force, Hc (Oe) Remanence magnetisation, Mr (emu/g) Maximum field saturation (kOe) Magnetite quantity (wt.%) Hysteresis area ±10 kOe (erg/g) Hysteresis area ±500 Oe (erg/g)

15 144 4.45 2 20 6150 3850

33 82 2.45 4 45 14,400 1750

tion of magnetite is lower than the reported data due to the powder shape of the sample. As a consequence they have a much higher surface/volume ratio than bulk materials whose saturation magnetisation is reported in the literature. Surface effects can modify the saturation magnetisation of a magnetic material, usually lowering the magnetisation. It is well known that the saturation magnetisation depends almost linearly on the content of the magnetite. Therefore, the quantity of magnetite crystallised in the glass–ceramics can be calculated from the ratio of saturation magnetisation between the samples and magnetite. The estimated amount of magnetite crystallised in the sample S35 is about 20 wt.%, and 45% in S45. These values are in perfectly agreement with the ones estimated by XRD data. Magnetisation increases with the amount of magnetite crystallised in the samples. Glass–ceramic S45 contains a higher quantity of magnetite and, therefore, has a higher value of saturation magnetisation. Consequently, increasing the content of iron oxides, the saturation magnetisation increases. The remanence is the natural quantity expressing the fact that a ferro or ferrimagnetic material can be spontaneously magnetised, even in the absence of external magnetic field. The remanence magnetisation values are much lower than saturation magnetisation values due to structural features of glass–ceramic. Remanence, as well as coercivity, depends strongly on microstructure. The composition has a great effect on the coercive force of glass–ceramics. The coercive force of S35 samples is almost two times higher than S45. The coercive force is influenced in a significant way by the crystal dimensions. As shown in Fig. 7, the coercive field abruptly decreases with an increase in the particle size. Moreover, the internal stress may affect considerably the coercive force. The variation of the coercive field in function of the lattice strain is plotted in Fig. 8. The coercive field is seen to rapidly increase with lattice distortions, as a result of the effect of internal stresses. When glass–ceramics are rapidly cooled to room temperature after melting, internal stresses appear due to the difference of the thermal expansion coefficients between the crystalline phases and the matrix. Such internal stress can inhibit the rotation of the magnetic

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

140 130

Hc (Oe)

120 110 100 90 80 70 34

53

Magnetite crystals size (nm) Fig. 7. Variation of coercive field as a function of size of magnetite crystals.

150 140

Hc (Oe)

130 120 110 100 90 80 70 0.21

0.35

Magnetite lattice strain (%) Fig. 8. Variation of coercive field as a function of lattice strain.

moment, increasing the coercive force [25]. As a consequence, sample S35, having a higher value of lattice strain is expected to have a higher coercive force, as confirmed by experimental data. As is known, in a system with a dominant shape anisotropy, the coercive field decreases as the volume fraction of magnetic particles in the assembly (packing fraction) increases [25]. This statement can be explained considering the magnetic interaction between iron atoms. If the packing fraction increases, the distance between particles decreases, the interactions become stronger and the coercivity decreases [25]. SEM images reported in Fig. 3 show that both glass–ceramic samples contain small columns with octahedral magnetite crystals. Supposing that every column acts as a single crystal with elongated shape, the shape anisotropy will prevail. It can be observed that S45 columns are evidently longer than S35 columns and, therefore, they contain a higher quantity of magnetite. As a result, the packing fraction in S45 is higher and the coercivity will be lower. This explanation is confirmed by considering the influence of the packing fraction on the coercive field.

An additional parameter of basic importance for the magnetic characterisation of these materials is the magnetic loss/cycle or the area of the hysteresis loop. In order to evaluate the magnetic loss/cycle of the glass–ceramic samples, we integrated the loop area of measured hysteresis loops, extrapolating between 10 kOe and +10 kOe (Table 3). Both the saturation magnetisation Ms and applied magnetic field H play an important role on the hysteresis area. S45 is characterised by a magnetic hysteresis area (14400 erg/g) more than twice as high as that for S35 (6150 erg/g). Increasing the content of iron oxides, the saturation magnetisation increases, while coercive force decreases. These opposite effects are reciprocally compensated in terms of magnetic losses. As a result, magnetic loss per cycle, for a maximum applied field of 10 kOe, is higher for S45 glass–ceramic, which contains a higher quantity of magnetite. Assuming that such a high magnetic field is difficult to realise in a clinical laboratory, hysteresis curves were measured using a magnetic field 20 times smaller (500 Oe) (Fig. 9). It can be immediately noticed the influence of magnetic field H on hysteresis area. Whereas for a magnetic field of 10 kOe, the area of S45 sample is higher, for a maximum applied field of 500 Oe, the area of sample S35 is higher. The calculated values of the hysteresis loops area for an applied field of 10 kOe and respective 500 Oe, (reported in Table 3), are plotted in Fig. 10. In this case, when the saturation is not reached, the coercive field has a predominant effect and therefore, the magnetic loss/cycle for S35 sample is higher. These results are confirmed by calorimetric data. The power losses estimated from calorimetric measurements (Eq. (1)) under a magnetic field of 40 kA/m (ffi500 Oe) and 440 kHz are 65 W/g for S35 and 25 W/g for S45 (Fig. 11). The obtained values are an average of five determinations. These values are in agreement with

20

S35 S45

10

M (emu/g)

150

427

0

-10

-20 -600

-400

-200

0

200

400

600

H (Oe) Fig. 9. Room temperature hysteresis cycle at 500 Oe for glass–ceramic samples.

428

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429 18

Hysteresis area (kerg/g)

16

10 kOe

14 12 10 8

10 kOe

6 500 Oe 4 500 Oe

2 0 S35

S45

Composition Fig. 10. Variation of hysteresis area as a function of composition, for applied fields of 10 kOe and 500 Oe.

Specific power loss (W/g)

70 60 50 40 30 20 10 0

S35

S45

Composition Fig. 11. Variation of the specific power loss of glass–ceramic samples as a function of the composition, for an applied field of 40 kA/m and 440 kHz.

magnetic losses measured at 500 Oe where the maximum magnetic loss/cycle is obtained for sample S35. In dynamic conditions, the power losses are determined not only by the classical magnetic losses, but also by eddy current losses, generated by the Joule heating effect. At frequencies higher than zero, the specific energy dissipation will be higher than the sum of magnetic loss and eddy current loss. The term in excess, named the anomalous loss, is a result of non-homogeneous magnetic domain structure and irregular movement of the domain walls [26–29]. Considering these experimental values, we can estimate that, in these conditions, immersing 1 g of S45 sample in 20 ml of distillate water, after 2 min of the application of the magnetic field, the temperature will locally increase by 36
C, whereas 1 g of S35 sample will

produce a temperature increase of 93
C. Obviously, the temperature variation (DT) can be tailored since it is proportional to the samples mass and to the magnetic field characteristics (intensity, frequency). Therefore, using a magnetic field compatible with clinical applications, the maximum heat dissipation will be obtained for sample S35 which has smaller size crystals and a higher coercive force. If a high intensity magnetic field is available, the maximum heat generation will be obtained for sample S45, which presents a higher content of iron oxides. Therefore, there is a strong influence of the amount of the crystallised magnetite on the crystallographic structure, microstructure and magnetic properties of glass–ceramics. This aspect is essential for hyperthermia therapy.

4. Conclusions Ferrimagnetic glass–ceramics containing magnetite as the only crystalline phase are obtained by coprecipitation of the reagent solutions, and melting the thermally treated powders at 1500
C. These glass–ceramics contain nanometric magnetite crystals, which are probably produced during fast cooling from the melting temperature. Magnetite is uniformly distributed in the amorphous matrix. In the case of S35 glass–ceramic, 20 wt.% of iron ions crystallises as magnetite and the rest remains in the glassy matrix. On the contrary, for S45 sample, all the iron ions crystallised as magnetite. The crystallite size of magnetite increases with iron oxides concentration, from 34 nm in S35 to 54 nm in S45. DTA measurements revealed well-defined glass transition temperatures, which confirm the presence of a relevant content of amorphous residual phase in the glass–ceramic samples. The power losses estimated from calorimetric measurements under a magnetic field of 500 Oe and 440 kHz are 65 W/g for S35 and 25 W/g for S45. Therefore, in the case of implanting into a tumour 1 g of S45 sample, a maximum increase in temperature of 36
C will be produced (under a magnetic field of 500 Oe and 440 kHz), while 1 g of S35 sample will locally raise the temperature by a maximum 93
C, depending on the tissue characteristics (blood flow, tissue density, type of tumoural cells, etc.). Using intensive magnetic fields, the maximum heat dissipation is obtained for glass–ceramic samples with a higher content of iron oxides and larger crystal size, for which the contribution of saturation magnetisation on magnetic loss prevails. For lower magnetic fields, when the saturation is not reached, the maximum heat generation is obtained for glass–ceramic samples with small crystal size. In this case, the effect of coercive force is predominant.

O. Bretcanu et al. / Acta Biomaterialia 1 (2005) 421–429

In conclusion, the chemical composition, mainly the iron ions content, affects in a significant way the magnetic properties. The iron ions content has a great influence on glass–ceramic structure and, consequently, on the characteristics of hysteresis cycle. Increasing the content of iron oxides in the materials, the saturation magnetisation increases, while coercive force decreases. Also, raising the amount of iron oxides in the glass– ceramic, increases the crystals size. Modifying the chemical composition, in particular the ratio between the iron oxides, we are able to control the heat generation of glass–ceramics, depending on the magnetic field characteristics and on the nature of neoplastic tissue. In this way, the quality of hyperthermic therapy can be improved.

[10]

[11]

[12]

[13]

[14]

[15]

Acknowledgments The magnetic measurements were carried out at the ‘‘National Electrotechnic Institute Galileo Ferraris’’, Torino, Italy. The calorimetric measurements were performed at Manfredi S.p.A, Pinerolo, Italy.

[16]

[17] [18]

References [1] Moroz P, Jones SK, Gray BN. Status of hyperthermia in the treatment of advanced liver cancer. J Surg Oncol 2001;77:259–69. [2] Gray et al. Targeted hysteresis hyperthermia as a method for treating diseased tissue. United State Patent 6167313, 2000. [3] Matsuki H, Yanada T. Temperature-sensitive amorphous magnetic flakes for intratissue hyperthermia. Mater Sci Eng 1994;A181/A182:1366–8. [4] Lee YK, Kim DH, Lee YJ, Kim KN, Shim IB. Synthesis of Sr and Cr doped barium ferrite microspheres for hyperthermic cancer treatment by sol–gel process. Ceramics, Cells and Tissues, 8th Annual Seminar and Meeting, Faenza, 2003. [5] Matsuki H, Satoh T, Murakami K, Hoshino T, Yanada T, Kikuchi S. Local hyperthermia based on soft heating method utilising temperature-sensitive ferrite rod. IEEE Trans Magn 1990;26:1551–3. [6] Ebisawa Y. Preparation of bioactive and ferrimagnetic glass– ceramics for hyperthermic treatment of cancer. PhD thesis, 2000. [7] Tucker RD. Use of interstitial temperature self-regulating thermal rods in the treatment of prostate cancer. J Endourol 2003;17: 601–7. [8] Borrelli et al. Radio frequency induced hyperthermia for tumor therapy. US Patent 4323056, 1982. [9] Ohura K, Nakamura T, Yamamuro T, Ebisawa Y, Kokubo T, Kotoura Y, et al. A bioactive ferromagnetic glass–ceramic for

[19] [20] [21]

[22] [23] [24] [25] [26] [27]

[28]

[29]

429

hyperthermia. Ceramics in substitutive and reconstructive surgery, Vincenzini, 1991. p. 131–6. Oh SH, Choi SY, Lee YK, Kim KN. Research on annihilation of cancer cells by glass–ceramics for cancer treatment with external magnetic field. I. Preparation and citotoxicity. J Biomed Mater Res 2001;54:360–5. Konaka H, Miyaji F, Kokubo T. Preparation and magnetic properties of glass–ceramics containing a-Fe for hyperthermia. J Ceram Soc Jpn 1997;105:833–6. Arcos D, del Real RP, Vallet-Regi M. Biphasic materials for bone grafting and hyperthermia treatment of cancer. J Biomed Mater Res 2003;65A:71–8. Bretcanu O, Spriano S, Vitale Brovarone C, Verne´ E. Synthesis and characterization of coprecipitation derived ferrimagnetic glass–ceramic. J Mater Sci, accepted for publication. Janasi SR, Emura M, Landgraf FJG, Rodrigues D. The effects of synthesis variables on the magnetic properties of coprecipitated barium ferrite powders. J Magn Magn Mater 2002;238: 168–72. Bretcanu O, Verne´ E, Borello L, Boccaccini AR. Bioactivity of degradable polymer sutures coated with bioactive glass. J Mater Sci: Mater Med 2004;15:893–9. Ferraris M, Bretcanu O, Verne´ E. Biocompatible magnetic materials for hyperthermia cancer treatment and their synthesis processing. Italian Patent IT2002TO00994, 2002. Bretcanu O. Magnetic materials for hyperthermic treatment of cancer. PhD thesis, 2004. Klug H, Alexander L. X-ray diffraction procedures. For polycrystalline and amorphous materials. London: Wiley; 1962. Young RA. The Rietveld method, IUCr. Oxford: Oxford University Press; 1993. Le Bail A. Modelling the silica glass structure by the Rietveld method. J Non-Cryst Solids 1995;183:39–42. Lutterotti L, Scardi P. Simultaneous structure and size-strain refinement by the Rietveld method. J Appl Crystallogr 1990;23:246–52. Available from: http://www.ing.unitn.it/~luttero/Publications/ EPDIC_V/silicate_glass.html. Arcos D, del Real RP, Vallet-Regi M. A novel bioactive and magnetic biphasic material. Biomaterials 2002;23:2151–8. Karamanov A, Pelino M. Crystallization phenomena in iron-rich glasses. J Non-Cryst Solids 2001;281:139–51. Cullity RD. Introduction to magnetic materials. Addison-Wesley; 1972. Fiorillo F. Introduzione ai materiali magnetici. Istituto Elettrotecnico Nazionale Galileo Ferraris, Torino, 2002. Genova E, Gobetto A. Appunti di ferromagnetismo: perdite nei materiali ferromagnetici. Istituto Elettrotecnico Nazionale Galileo Ferraris, Torino, 1984. Bertotti G. Physical interpretation of eddy current losses in ferromagnetic materials. I. Theoretical considerations. J Appl Phys 1985;57:2110–7. Bertotti G. Physical interpretation of eddy current losses in ferromagnetic materials. II. Analysis of experimental results. J Appl Phys 1985;57:2118–26.