Irene Andreua,
Eva Natividad*a,
Costanza Ravaglib,
Miguel Castroa and
Giovanni Baldib
aInstituto de Ciencia de Materiales de Aragón (ICMA), CSIC – Universidad de Zaragoza, Campus Río Ebro, María de Luna, 3, 50018 Zaragoza, Spain. E-mail: evanat@unizar.es
bCentro Ricerche Colorobbia, Sovigliana Vinci, Italy
First published on 19th June 2014
The heating ability of magnetic nanoparticles (MNPs) intended for magnetic hyperthermia is quantified by means of the specific absorption rate (SAR), also referred to as specific loss power. This quantification is mainly performed on ferrofluids, even though the SAR values so obtained are often not representative of the nanoparticle performance inside tissues (solid matrices). In this study, the SAR of cobalt ferrite MNPs with mean crystallite diameters of 5.5 nm and 7.4 nm, functionalized or not, and dispersed in liquid or solid media, was determined in the 180–300 K temperature range. Higher SAR values were systematically obtained for samples in liquid media. On the one hand, heat capacity data together with zero-field-cooled and field-cooled magnetization curves allowed correlation of these results with ferrofluid dynamics originating from viscosity changes in samples dispersed in diethylene glycol. On the other hand, the higher degree of agglomeration attained by the functionalized MNPs after immobilization in paraffin wax seemed responsible for the decrease in SAR values and the shift in blocking temperature. In sum, MNP's spatial arrangements acquired after ferrofluid injection in magnetic hyperthermia should be taken into account to predict SAR values during therapies.
In the last years, great efforts have been made to synthesize nanomaterials with improved heating ability to be used in MFH. This ability is currently evaluated by several methods (magnetic, calorimetric),11 over samples with different MNP concentration, distribution and dispersive medium,12,13 and in different thermal (adiabatic, non-adiabatic, in vitro, in vivo) and magnetic (H0, f) conditions. This spectrum of variables often makes difficult comparing the results. In particular, with respect to concentration, distribution and dispersive medium, most authors measure the heating efficiency in liquid,12–19 where distances between particles are expected to be large and homogeneous throughout the sample. However, MNPs suspended in a liquid carrier are also free to move, form structures and orientate with the alternating magnetic field.20–22 The results obtained in these conditions can strictly only be compared with samples of the same concentration and similar solvent, since inter-particle distances affect interaction effects, and the properties of the dispersive medium also play a role in heat dissipation mechanisms.19,23 Furthermore, given that in real applications the MNP suspension is previously injected into the tissue, results obtained in liquid cannot be directly extrapolated to in vivo experiments. After injection into a solid matrix, MNP distribution varies, interparticle distances turn non-constant, interparticle interactions change, and even some heating mechanisms are no longer active (orientation in liquid, Brown relaxation).24,25 In addition, results of in vivo heating experiments are not usually accompanied by a description of the MNP distribution and concentration acquired upon injection of MNPs. Studies considering all these variables would certainly help to compare heating efficiencies between samples synthesized by different authors.
In this work, we evaluate the heating ability of cobalt ferrite MNPs dispersed in liquid, considering the influence of size, functionalization, dispersive media and particle distribution. We also study and justify the change in their performance upon immobilization in solid matrices on the basis of ferrofluid dynamics and interparticle interactions. Eventually, the importance of measuring the dissipated heating power as a function of temperature to infer the effect of these variables is highlighted.
Batch | Size (nm) | Sample | f.a | Disp. medium | cb (gMNP/g) |
---|---|---|---|---|---|
a f. is functionalization with cetyl phosphate.b c is the sample concentration in grams of nanoparticles per grams of sample. | |||||
CF1 | 5.2 (TEM), 5.5 (XRD) | CF1L | No | DEG | 0.0256 |
CF1S | No | PEG | 0.0089 | ||
CF1fL | Yes | n-dodecane | 0.0596 | ||
CF1fS | Yes | Paraffin wax | 0.0236 | ||
CF2 | 7.4 (XRD) | CF2L | No | DEG | 0.0262 |
CF2S | No | PEG | 0.0142 |
X-ray diffraction (XRD) experiments were performed on a Phillips X'per Pro diffractometer (Cu Kα radiation). Diffraction patterns were recorded in the 2θ range 10–70° with a scan step of 0.05° (2θ) for 5 s, from which the crystalline structure of the samples was identified. Also, the nanoparticle crystallite size was estimated from the peak broadening using the Scherrer method.
Particle size and arrangement were studied by Transmission Electron Microscopy (TEM) with a JEOL 2000 FXII instrument working at an acceleration voltage of 200 kV. Drops of ferrofluids were put onto copper grids with a carbon membrane film. The solid samples were prepared by ultramicrotomy, using epoxy resin (Epofix) as embedding medium. Blocks were sectioned with a diamond knife and slices were also laid on similar copper grids.
![]() | (1) |
Liquid samples were measured inside quartz containers sealed with adhesive to prevent liquid leakage in the vacuum environment required for adiabatic measuring conditions. Solid samples were prepared either in pellets (CF1fS) or embedded in epoxy resin (CF1S, CF2S). Eventually, measurements were performed with ac-field frequency and amplitude of 111 kHz and 3 kA m−1, respectively, in order not to overtake the biological range of ac-field application and study the low-field heating of the nanoparticles.
![]() | ||
Fig. 1 Above: XRD patterns of CoFe2O4 database reference and CF1 (dashed line) and CF2 (continuous line) powder. Below: TEM image and diameter histogram for batch CF1. |
Fig. 1 also shows a TEM image of batch CF1. According to the fit of the histogram to a normal distribution, these nanoparticles have an average diameter of 5.2 ± 1.0 nm. This result indicates that the size of the MNPs synthesized in batch CF1 has a standard deviation less than 20%. Given the small size of these MNPs, we can consider that they present a narrow size distribution, even if they do not fulfill strictly the standard criterion for monodispersity (10%).
Sample CF1L (Fig. 2) presents quasi-2D arrangements onto the copper grid indicative of a slight fractal aggregation. This arrangement is expectable in absence of functionalization since MNPs, due to their magnetic nature, tend to interact and form some structures in liquid.21 The image of sample CF1fL shows a 2D distribution of MNP islands formed upon solvent evaporation. This indicates that the synthesis of batch CF1 has produced well-dispersed MNPs, and that functionalization has enabled maintaining the absence of aggregation. Eventually, sample CF2L displays a raspberry-like arrangement, with visible edge particles similar in size and shape to those of sample CF1L. The successive growth steps of the synthetic method together with an eventual agglomeration of the seeds have induced formation of raspberry-like nanoparticles with an overall mean diameter of 40 nm. Even though these particles show considerable polydispersity, they have proven useful to explain the behavior of samples coming from batch CF1, as it is described below. According to the previous results, crystallite size as obtained by XRD agrees well with TEM particle size for batch CF1, while the results for batch CF2 would indicate that raspberry nanoparticles consist in several nanocrystallites with a certain size distribution, fact that should be reflected in the physical properties.
Fig. 2 also illustrates the MNP distribution within solid samples. The images of samples CF1S and CF1fS show regions of very dense nanoparticle packaging, while the raspberry-like structure of sample CF2S is preserved upon solid matrix transfer.
The concentration of the other samples (Table 1) was calculated with the help of M(H) curves, assuming that MS values are similar for samples of the same synthetic batch (CF1 or CF2). The resulting specific magnetization data are displayed in Fig. 3.
The specific heat of samples CF1fS and CF2S grows monotonically from 150 K to around 300 K, where the melting peak of the solid matrices (paraffin wax and PEG, not shown) starts to contribute. Sample CF1S follows the same trend but, in addition, presents a glass transition at around 172 K. This transition is assigned to the existence of some DEG trapped with the NPs during solidification of PEG. The quantification of this step (heat capacity of 0.27 J g−1 K−1) has allowed estimating the DEG content in the sample between 25–30% of the total weight.
From previous data, the heat capacity of the specimens used in SAR measurements was determined as C = ΣcP,imi, where cP,i and mi, are, respectively, the specific heat and mass of each contribution to C. Table 2 collects the mass of these contributions in each case.
Sample | msample (g) | Container | mcontainer (g) |
---|---|---|---|
CF1L | 0.1146 | Quartz | 0.3799 |
CF1S | 0.1204 | Epoxy | 0.1331 |
CF1fL | 0.0826 | Quartz | 0.3998 |
CF1fS | 0.1545 | — | — |
CF2L | 0.1032 | Quartz | 0.3539 |
CF2S | 0.1233 | Epoxy | 0.1222 |
ZFC-FC magnetization vs. temperature measurements at low field are a useful tool for determining FM/SPM transitions. Three characteristic temperatures can be distinguished within ZFC-FC curves as temperature increases:34 Tsat, temperature above which the FC magnetization starts to decrease, Tmax, temperature corresponding to the maximum of the ZFC curve, and Tbra, temperature at which the ZFC and FC curves join together. Tbra is representative of the temperature at which the majority of the MNPs is unblocked and behaves as a superparamagnet. In this situation, the magnetization of each MNP flips direction due to thermal energy, with a characteristic time for magnetization reversal higher than the measuring time. Tmax determines the onset of the superparamagnetic behavior assuming non-interacting and monodisperse MNPs, and its value is related to Eb. Accordingly, these curves must also reflect differences related to the particle size.
Fig. 5 shows the ZFC-FC M(T) curves of all the samples. From this data, values of Tmax and Tbra were obtained and collected in Table 3. Tmax and Tbra data of samples coming from batch CF1 are lower than those of samples coming from CF2. This result is indicative of a larger average particle volume of the later, as previously deduced from XRD and M(H) measurements. In effect, in our MNPs, main contributions to K are magnetocrystalline, KM, and surface, KS, ones, so K can be expressed as K ≅ KM + KS = KM + λ(S/V), where λ is a positive constant,31 V is the magnetic volume of the MNP and S, its surface. KM is the same for batches CF1 and CF2, but Ks increases for smaller particles due to their higher S/V ratio. This implies that, if K (CF1) > K (CF2) and Tmax, Tbra (CF1) < Tmax, Tbra (CF2), then V (CF1) < V (CF2).
Apart from this difference, ZFC-FC curves reflect other remarkable features. Let us pay attention to the differences between the ZFC-FC curves of samples CF2L and CF2S. While the latter shows a typical ZFC-FC trend for FM/SPM transitions, sample CF2L present several untypical features, related to the temperature dependence of the DEG viscosity. In the ZFC curve, taken upon heating, once DEG undergoes the glass transition, M(T) grows faster than it does in PEG, and at 194 K a great step is observed. This sharp magnetization rise is a consequence of the decrease of the DEG viscosity to a value low enough to allow physical orientation and/or reorganization of the MNPs by the action of the static magnetic field, thus increasing χ0. Such behavior has been also observed in cobalt MNPs in several organic solvents.35,36 Afterwards, at 243 K, the magnetization reaches a local minimum that coincides with the recrystallization phenomenon reflected in the cP of DEG, and afterwards recovers the decreasing trend upon re-melting of the formed crystallites. In FC curves, magnetization of sample CF2L is considerably higher than that of sample CF2S, given that, in this case, the sample is solidified under the static magnetic field that generates the ordered high-χ0 state.
Similar effects are also present in sample CF1L. At 194 K, when the DEG viscosity is low enough to allow physical orientation and/or reorganization of the MNPs, magnetization starts to increase, but the magnetization rise is counteracted by the transition of the sample to the superparamagnetic state, and only a slight bump is observed. In order to highlight this rise, we performed new ZFC measurements over samples CF1L and CF1S (with 25–30% of DEG) stopping at 200 K to record the temporal evolution of M. These results are shown in the inset of Fig. 5. In both cases the magnetization increases when the system is allowed to orientate within the applied magnetic field, although it is sample CF1L which undergoes the higher rise.
Finally, sample CF1fL does not present an increase in χ0 upon melting of n-dodecane. This is probably related to the fact that this sample is already in the superparamagnetic state when the phase transition of the dispersive media takes place.
![]() | (2) |
According to the Linear Response Theory (LRT), valid for low H0 values, SAR depends on the out-of-phase component of the ac magnetic susceptibility as:
![]() | (3) |
Samples CF2L and CF2S show the highest SAR values at room temperature, 1.2 W g−1 and 0.6 W g−1, respectively, measured under 111 kHz frequency and 3 kA m−1 amplitude, while samples obtained from batch CF1 display lower SAR values at room temperature. This fact makes batch CF2 more interesting for MFH applications. However, the SAR(T) trend of these samples does not allow full understanding of the effects studied in the present work, given that their FM/SPM transition is beyond the temperature measuring limits. For this purpose, samples obtained from batch CF1 are of greater interest.
Certainly, the SAR values mentioned above may seem dramatically low compared to other values reported in the literature like, for example, that of bacterial magnetosomes, which show a breathtaking SAR of 960 W g−1 at 410 kHz and 10 kA m−1,37 or the unprecedented SAR of 415 W g−1 displayed by synthetic iron/iron carbide core–shell MNPs38 at 96 kHz and 20 mT (16 kA m−1). However, these values are not directly comparable with our data, as SAR depends strongly on the ac-field parameters, and much greater SAR is obtained for the same MNPs when using higher f or, especially, H0 values, often beyond the physiological limits of the therapy. Comparison then implies the use of extrapolations.
A linear dependence of SAR on f is usually considered, although recent studies contradict this consideration, especially in presence of inter-particle magnetic interactions39 or when high frequencies are used.40 The dependence of SAR with H0 usually shows a H0n relationship, with n ranging between 0 and 3, depending on the characteristics of the MNP assembly and on the value of H0. In the case of magnetosomes an average dependence of H2.5 was found37 using static hysteresis cycles with H0 ranging from 0.5 to 10 kA m−1. Using this fit and the typical lineal relation with f, the SAR of magnetosomes at 111 kHz and 3 kA m−1 is estimated to be a still high but more modest 12.8 W g−1. In the case of synthetic iron/iron carbide core–shell MNPs, measurements at 3.8 mT (≅ 3 kA m−1) derive negligible SAR values.
Due to the variable and often unknown dependency of SAR on H0 and f, fair comparisons between MNP assemblies can be made when SAR values are obtained with more similar parameters of the ac-field. For example, the highest SAR of assemblies of CoFe2O4 powder with different sizes was found to be 0.07 W g−1 at 500 kHz and 5.4 kA m−1 for samples with a mean diameter of 8.4 nm,41 close to the size of the crystallites of samples CF2. Also, highly concentrated Y3Fe5O12 powder in Epoxy resin showed a SAR of 0.2 W g−1 at 200 kHz and 1.77 kA m−1.42 The SAR of assemblies of different ferrite MNPs, measured at 110 kHz and 4.8 kA m−1, was found to be highest for MgFe2O4 particles, 33 W g−1, and more modest for CoFe2O4 MNPs, 1 W g−1.43 Extrapolating these values to 111 kHz and 3 kA m−1, using the Hn dependence obtained in their work (n = 2.7 for MgFe2O4 and n = 2.8 for CoFe2O4), SAR turns into 9.4 W g−1 and 0.3 W g−1 for MgFe2O4 and CoFe2O4, respectively. Eventually, Fe3O4 assemblies with different coatings showed SAR values of 6.7 W g−1 under 750 kHz and 2 kA m−1.44 The extrapolation of these values to 111 kHz and 3 kA m−1, with n = 1.9, gives a SAR of 2.1 W g−1. In sum, samples from batch CF2 show quite high SAR values, compared to samples of akin composition and measured with similar ac-field parameters.
Furthermore, although not among the highest ever found, SAR values of samples CF2 are high enough to achieve therapeutic temperatures in MFH treatments for certain tumor volumes45 and MNP delivery methods.46 For example, therapeutic temperatures were recently achieved in animal bladders by using injected magnetite MNPs,47 whose SAR, measured at 40 kHz and in liquid, was about 0.8 and 5.6 W g−1 at 2 and 5.5 kA m−1, respectively.
Going back to Fig. 6, and just considering SAR values at 300 K, one could conclude that samples obtained from batch CF1, functionalized or not, have similar heating behavior. However, the study of the SAR(T) trends over the whole temperature range reveal interesting differences among the samples. Note that data acquired during the melting of n-dodecane (sample CF1fL) were not considered, since the first-order character of the transition prevents a correct evaluation of the heat capacity at this temperature range.
Tb data obtained from SAR(T) trends are collected in Table 3, for comparison with the characteristic temperatures obtained from ZFC-FC M(T) curves. Given that SAR was measured at high frequency (111 kHz), Tb values are systematically higher than Tmax and Tbra data. However, Tb and Tmax are linearly proportional, confirming the correlation between both magnitudes.
During measurements, taking place on heating, samples CF1L and CF2L, dispersed in DEG, go gradually from a high-viscosity to a liquid state, and the SAR(T) trends of both samples are continuous. The Tb of sample CF1L is 244 K, while sample CF2L is in the FM state within the whole temperature range. Sample CF1fL is in the solid state up to 263 K. In this state, the trend shows a maximum at Tb = 238 K that corresponds to the unblocking of the Néel relaxation mechanism, since the physical movement of the MNPs within the dispersive medium is inhibited. Above 263 K, in the liquid state, the trend is similar to that of sample CF1L. Measurements of samples CF1S, CF2S and CF1fS take place entirely in the solid state. The most outstanding difference between liquid and solid samples is the decrease of SAR values in all cases. Samples CF1S and CF2S present qualitatively the same trend than samples CF1L and CF2L, respectively, but lower values. However, sample CF1fS shows an increase of 11 K in Tb and the strongest reduction in SAR compared to CF1fL. As both CF1fL and CF1fS were solid at Tb, the observed shift cannot be assigned to the suppression of Brown relaxation in sample CF1fS. For the same reason, the SAR reduction undergone by all solid samples cannot be justified on the basis of different relaxation mechanisms.
As explained in the previous section, a great step is observed in the ZFC M(T) curve of the ferrofluid with uncoated raspberry-like MNPs (sample CF2L) once DEG undergoes the glass transition, and its viscosity becomes low enough to allow physical orientation and/or reorganization of the MNPs by the action of the magnetic field. As a consequence, magnetization of sample CF2L gets considerably higher than that of sample CF2S, in which field-induced reorganizations are inhibited due to the solid matrix. This greater static magnetic susceptibility of sample CF2L at 3 kA m−1 with respect to that of sample CF2S would explain the SAR differences between them, although the increase in χ0 may not be fully reflected in ac-measurements due to frequency effects.
The ferrofluid with uncoated and slightly-agglomerated MNPs (sample CF1L) undergoes the same effect, although the M(T) rise is much smaller, since such ferrofluid dynamics are readily inhibited by the transition of the sample to the superparamagnetic state. This effect should be reflected in SAR(T) measurements. Given that, during SAR measurements, the sample is subjected to an alternating magnetic field, and that Tb, unlike Tmax, shifts to higher values as the ac-field frequency increases, the counteraction of the FM/SPM transition in SAR(T) measurements occurs at higher temperature. This fact should allow, according to the results observed in sample CF2L, a more pronounced increase in χ0 for sample CF1L compared to sample CF1S than that observed in static measurements. This could be the reason why the SAR of sample CF1L appears considerably higher than that of sample CF1S, although χ0(T) is similar in both samples in ZFC-FC curves.
The ferrofluid with well-dispersed functionalized MNPs (sample CF1fL) does not show an increase in χ0 upon melting of the solvent, due to the higher melting temperature of n-dodecane, above the blocking temperature. Accordingly, the SAR differences between samples CF1fL, which displays the highest SAR values among the samples of the same particle size, and CF1fS, which is the worst performing sample, cannot be assigned to ferrofluid dynamics. However, the increase of 11 K in Tb of sample CF1fS with respect to that of sample CF1fL points an appreciable modification of Eb, presumably due to the effect of interparticle interactions. Indeed, the most outstanding change in particle distribution upon embedding in a solid matrix was clearly undergone by functionalized samples of batch CF1 (Fig. 2). While sample CF1fL presents well dispersed MNPs in n-dodecane thanks to functionalization, sample CF1fS shows very dense nanoparticle packing in paraffin wax.
In order to justify this change in Tb, the Berkov and Gorn (BG) model33 was considered. This is a numerical simulation of χ′′(T) for various excitation frequencies, and for samples with different homogeneous particle concentration, damping parameter (λ) and magnetic anisotropy. It predicts variations both in the position and in the value of the maximum of χ′′(T). Depending on the value of the adimensional parameter β = 20K(J m−3)/MS(kA m−1)2 two different regimes can be distinguished: high- and moderate-anisotropy regime (β > 1) and low-anisotropy case (β < 1), for which the behavior of χ′′(T) is quite different.
To roughly estimate the regime in which sample CF1fL lies, β can be calculated using the MS value obtained from magnetic measurements (330 kA m−1), and the K value derived from the Néel relaxation time, τN = τ0 exp(Eb/kBT), assuming Eb = KV, that is, overlooking the effect of magnetic interactions. At Tb, τ0 exp(Eb/kBTb) = 1/2πf, and, considering τ0 = 10−9 s, the value of K for sample CF1fL at 238 K is 274.025 kJ m−3 and then β = 50 ≫ 1 at this temperature. Even if the calculation is just an estimate, we can conclude that sample CF1fL lies in the high- and moderate anisotropy regime. For this case, the BG model predicts that inter-particle interactions do not fully govern the energy barriers, but such barriers are already created by the relatively high single-particle anisotropy. Interactions then decrease Eb, shift the χ′′(T) peak to lower temperatures and decrease its value for concentrations cv > 0.08 (volume concentration). This is not consistent with the evolution of the functionalized samples coming from batch CF1f upon embedding in paraffin wax, since the SAR(T) trend of sample CF1fS is lower than that of sample CF1fL, but shifted to higher temperatures.
The BG model attributes this disagreement to the presence of MNP agglomeration within the samples. Such an agglomeration would lead to correlations between the anisotropy axis directions of the neighboring particles, increasing the energy barriers in the system, thus leading to the shift of the χ′′(T) peak towards higher temperatures with increasing particle concentration. In effect, the MNPs in sample CF1fS can be considered aggregated rather than concentrated, so a shift of the χ′′(T) peak towards higher temperatures together with a decrease in its value is predictable.
When comparing samples CF1L and CF1S, no appreciable shift in the position of the maximum of SAR(T) is observed. One possible reason is that, given that sample CF1L presents already some aggregation due to the lack of coating, the embedding in PEG has not substantially changed its agglomeration. This is also the case of samples CF2L and CF2S, for which TEM images reveal similar MNPs arrangement, although the high position of their SAR(T) maximum, out of the measuring range, prevents further discussion.
This journal is © The Royal Society of Chemistry 2014 |