Composition study of CoPt bimetallic nanocrystals of 2 nm

Abstract

The synthesis of bimetallic alloy nanocrystals with a well-controlled relative composition is a real challenge and requires chemical analysis techniques with high accuracy. A new chemical route has been used to synthesize cobalt–platinum nanocrystals of 2-nm diameter in a wide range of relative stoichiometry. A study of the elemental composition of the nanoalloy was carried out by X-ray fluorescence (XRF) spectroscopy and energy-dispersive X-ray analysis. We have developed a set-up for XRF analysis using a silicon wafer as a support to determine the elemental composition with only a small amount of sample. The calibration step and the measurement capabilities are described. In a composition range of 25–75% cobalt, the results of both analytical methods are discussed and compared in detail.

Introduction

Bimetallic alloy nanocrystals represent a particularly interesting class of materials owing to their potential use in ultra-high-density magnetic recording [1], in magnetic biosensing [2] and in catalysis [35]. The specific properties imparted upon alloying owing to the large variety of compositions, structures and properties of metallic alloys have led to a wide variety of applications [6]. In nanoalloys, the composition and the segregation properties are important determinants of the chemical reactivity and the crystalline phase transformation. For instance, CoPt alloy has an ordered crystalline phase (L10) around the equiatomic composition, which creates interesting magnetic behaviours owing to a large magnetocrystalline anisotropy constant (4.9 × 107 erg/cm3) [7, 8]. CoPt spherical nanocrystals in the L10 phase have the potential to exceed the superparamagnetic limit for an average diameter close to 5 nm [9]. However, to reach the L10 ordered phase, a high level of control of the chemical composition of bimetallic nanocrystals is crucial. In fact, a narrow composition range (approximately 3% cobalt) around the 50:50 value is necessary to achieve this magnetic nanoscale system (L10 phase) [10].

The accurate determination of the relative composition of bimetallic alloy nanocrystals remains complicated [11]. The accuracy depends mainly on the characterization techniques and on the set-up used during the analysis. In the case of nanoparticles with an average diameter of 2 nm, the crystals analysed consist of approximately 300 atoms [12]; lowering the data acquisition time and the required amount of nanocrystal is important. Recently, new techniques have been introduced, including scanning transmission electron microscopy with electron energy loss spectroscopy. Subnanometre resolution is coupled with Z-contrast measurements and provides an insight into the structures and relative elemental speciation within these bimetallic clusters [13]. However, the common techniques for stoichiometric analysis used for nanocrystals are energy-dispersive X-ray (EDX) analysis [14], Rutherford back scattering spectrometry (RBS) [15] and extended X-ray absorption fine structure spectroscopy [16]. These techniques are able to obtain quantitative measurements of low accuracy and have a drawback; they require a large quantity of sample. The analysis of local chemical composition, as in the samples of alloy nanoparticles, is difficult to achieve. Hence, X-ray fluorescence (XRF) spectroscopy appears to be an alternative technique. The XRF measurements can be carried out with small quantities of particles (nanograms) and with good accuracy for the heavy elements [17, 18]. If the measurements are performed under thin-layer conditions, no matrix effect correction is required.

In this paper, we report the development and the application of a protocol adapted to the analysis of the elemental composition of small quantities of nanoparticles by XRF spectroscopy using a silicon wafer as the sample support. The calibration was performed with pure metallic salts of both elements present in the bimetallic nanoparticles (cobalt and platinum). Using the XRF setup, we carried out the stoichiometric analyses of samples of nanocrystals in the composition range 25–75%. Finally, these XRF results were compared with those obtained by EDX measurements.

Experimental

Chemical synthesis of CoPt nanocrystals

In the colloidal approach, several chemical methods are usually used to synthesize nanoalloys, such as inverse micelles [19, 20] and organometallic decomposition [21]. In this study, the liquid–liquid phase transfer method was used to synthesize CoPt nanocrystals of 2-nm diameter with fine control of the composition [22], which was achieved with the process described in [23]. The synthetic route was developed by adjusting the initial platinum salt, PtCl6(TDA)2 (TDA is tetradecylammonium; 6.25 × 10-3 M), and cobalt salt, CoCl4(TDA)2 (6.25 × 10-3 M), ratio in the organic phase using tetradecylammonium bromide (Aldrich, 99%) as a phase transfer agent. Hexadecylamine was used as a coating agent to stabilize the nanocrystals in solution and prevent aggregation. The reduction step was carried out by introducing an aqueous solution of NaBH4 (378 mg, 10 mL of H2O) (Aldrich, 99%) into the organic mixture containing the metal salts [20 mL of toluene, 10 mL of PtCl6(TDA)2, 10 mL of CoCl4(TDA)2 and hexadecylamine (1 mL of C16H33NH2; Fluka, 99%)]. After 16 h of reaction, the organic phase was evaporated, ethanol was added in excess and the particles were separated by centrifugation. The procedure was repeated twice, and finally the nanocrystals were dispersed in 4 mL of toluene.

X-ray fluorescence spectroscopy

XRF measurements were performed with a wavelength-dispersive Philips Pananalytical X-ray PW2404 spectrometer with a primary rhodium X-ray source [18, 24]. The spectrometer has an autosampler, which allows convenient and quick measurements. The analytical conditions are reported in Table 1. A common background correction method was used to study the peaks obtained. For each sample, three replicates were done. Silicon wafers (100) of 2.54-cm diameter were used to support the calibration and nanoparticle solution deposits. A package of ten wafers from the same batch was employed for each measurement. The fluorescence spectra of clean wafers and the measured elements did not interfere with each other.

Table 1 Analytical conditions for X-ray measurements

Cleaning step

To reuse the silicon wafers after each measurement set, an effective washing process was established to maintain a non-contaminated support. The wafers were soaked in 100 mL of toluene under sonication (10 min), and then rinsed with water and dried. Then, the silicon wafers were rinsed with aqua regia (a mixture of 66% HCl and 33% HNO3). Next, we used a diluted RBS cleaning agent (SocoChim) and rinsed the wafers with ethanol. Finally, all the wafers were dried in a nitrogen stream. Analyses of used wafers blanked with this protocol did not exhibit any differences compared with virgin wafers, demonstrating that the washing process is effective.

Calibration

The calibration of the analytical technique is an essential step for the good measuring accuracy of quantification. A dilute (6.25 × 10-3 M) solution of metallic salts of platinum, PtCl6(TDA)2, and cobalt, CoCl4(TDA)2, dissolved in toluene was used for the calibration. Precise volumes of these standard solutions were deposited at the midpoint of a silicon wafer using a micropipette. The concentration of the solution must be low enough to maintain "thin-layer" conditions in the dried drop, avoid matrix effects caused by X-ray absorption and provide a linear calibration. To obtain the calibration lines, volumes of 10, 20, 30 and 40 µL, respectively, were deposited. The point corresponding to 20 µL was triplicated. At the beginning of the experiment, all the clean wafers were analysed and did not exhibit any differences. Then, one of the wafers was kept clean and used as a zero for each batch. Calibration lines were then calculated by linear least-squares regression (Fig. 2). More concentrated solutions or larger volumes exhibit clear absorption effects and are not recommended (see “Results and discussion”).

Energy dispersive X-ray analysis

The EDX analyser of a JEOL 5510LV scanning electron microscope was used to determine the average composition of the CoPt nanocrystals. A thin film made up of nanocrystals was obtained by evaporation of a concentrated nanocrystal solution on silicon wafers at room temperature. The resulting film was thicker than 10 μm. In the case of EDX analysis (scanning electron microscopy), a 1-μm thickness is required to obtain a reliable analysis [25]. Several areas of the film were measured and EDX spectra were treated by the ZAF analysis technique.

Results and discussion

CoPt alloy nanocrystals were synthesized by the liquid–liquid phase transfer method [23]. Figure 1a shows a transmission electron microscope (TEM; JEOL 1011, 100 kV) image of Co50Pt50 nanocrystals with an average diameter of 2 nm. These crystals are quasi-spherical (Fig. 1c) and have a homogeneous size with a polydispersity of 12%, which was determined from measurements of TEM images of 500 nanocrystals [23]. The selected-area electron diffraction measurements exhibited an interlattice distance of 2.2 Å, corresponding to the (111) face-centred-cubic plane of the CoPt nanocrystals in disordered structure A1 (Fig. 1b). In addition, in a previous article [23], we demonstrated that Co x Pt(100-x) nanocrystals obtained by this method are a nanoalloy. The linear variation of the lattice parameter as a function of the relative composition determined by X-ray diffraction and high-resolution TEM measurements confirmed the formation of CoPt nanoalloy and suggests negligible platinum or cobalt surface segregation. Indeed, in the bimetallic nanoparticles, the segregation effects can occur and drastically modify the particle properties [26]. Nanocrystals of different compositions were synthesized by varying the concentration ratio of cobalt and platinum salts in the initial solution in the range 25–75%.

Fig. 1
figure1

a Transmission electron microscope (TEM) image of Co50Pt50 nanocrystals with an average diameter of 2 nm. Inset: the nanocrystal size histogram (polydispersity of 12%). b Selected-area electron diffraction pattern of CoPt nanocrystals. c High-resolution TEM image of CoPt nanocrystals of 2 nm

The calibration lines resulting from cobalt and platinum salt deposition are plotted in Fig. 2. A good linear relationship was obtained. We notice a weak shift of the straight line at the origin for both curves probably due to the linear regression process. The slopes of the lines are 2.36 cps/µL for cobalt (Fig. 2a) and 0.684 cps/µL for platinum (Fig. 2b). After the calibration stage, the quantitative measurements of the CoPt nanoalloy sample were performed. The measured standard deviations of the method for both the linear regression and the triplicate results were 6 and 8% for cobalt and platinum, respectively.

Fig. 2
figure2

Calibration curves of a cobalt [CoCl4(C40H84N)2, M = 1,356.73 g/mol] and b platinum [PtCl6(C40H84N)2, M = 1,563.78 g/mol] from salt solutions: fluorescence intensity versus deposited volume

Considering an analysis time of 90 s (split into 3×30 s shot), the total counts for the 20-µL triplicate are 4,250 and 1,231 for cobalt and platinum, respectively. Here, the background can be neglected for this calculation. The minimum standard deviation of a count is the square root of the count. The relative standard deviations induced by this raw variation were 1.5 and 3%, respectively. The remaining part of the measured deviation may be attributed to the dispersion of the sample composition, and was evaluated to about 5% for both cobalt and platinum. This deviation should be caused by the variation of the deposited volume.

An increase in the counting time or the excitation power will reduce the counting variation and may decrease the measured standard deviation by up to 5%, but improving the precision of the deposited volume may also reduce this deviation. Moreover, for relative measurements, the accuracy of the deposited volume does not matter as much as the deposited quantity staying in the thin-layer range. In this case, only the counting uncertainty remains. For the method explained here, the expected standard deviation for the cobalt-to-platinum ratio can be estimated to be the sum of the counting deviation of platinum and cobalt, which is 4.5%, and can be improved by increasing the counting time or the X-ray power delivered.

We analysed the composition of Co x Pt(100-x) nanocrystals obtained with various cobalt salt to platinum salt ratios in the initial solution. Five nanoparticle samples were analysed: the first four with 25, 33.3, 66.6 and 75% cobalt composition (expected) deposited once each on a silicon wafer and the last one with 50% cobalt composition triplicated on three wafers. The equiatomic composition is the most interesting for the magnetic properties; thus, we focused our analyses on this sample to obtain better precision for its composition and associated uncertainty

Table 2 shows the XRF measurement data of nanocrystals samples. The quantification analysis allows us to determine the composition ratio of the nanocrystals and the corresponding standard deviation. A plot of the cobalt content of the precursor salt in solution versus the cobalt content in CoPt nanocrystals determined by XRF spectroscopy is shown in Fig. 3. A linear variation can be observed with a slope of 1, indicating good control of the nanoparticle composition by the initial salt ratio. For the value with 50% cobalt, we notice a weak variation from 48.3 to 50.5% for the three points measured.

Table 2 Data obtained from X-ray fluorescence measurements on a silicon wafer of nanocrytals in the range from 25 to 75% cobalt composition
Fig. 3
figure3

Experimental composition measured by X-ray fluorescence (XRF) spectroscopy versus the theoretical composition for cobalt in CoPt nanocrystals. A dotted line with slope 1 is also plotted

These measurements remain in the range of the expected values and the standard deviation obtained by salt deposition. Concerning the expected values of 25, 33.3, and 66.6%, the XRF measurements are in good agreement (Table 2). But for the expected composition with 75% cobalt, the analysis of nanoparticles gives a value of 70% which does not include in its confidence interval the expected 75% value. This effect cannot be explained by a simple absorption effect of the matrix because the elements themselves have very low absorption in XRF and thus an increase in the proportion of cobalt induces a decrease in the autoabsorption of the sample. However, this stoichiometry variation could be also due to a limitation of the composition control of the synthesis method for cobalt-rich particles. Furthermore, for this composition of 75% cobalt, structural and segregation effects can appear and induce important modifications of the nanocrystals.

To verify the influence of the absorption effect on these measurements of the composition, calculations of the decrease of the X-ray emission were carried out as a function of the thickness of the measured layer (Table 3) [27, 28]. In the case of the metallic salts (Pt1Cl6C80H168N2 and Co1Cl4C80H168N2) with a deposited surface of 1 cm2, thicknesses of 2.6 µm for platinum and 2.9 µm for the cobalt were found assuming a decrease of signal of 1%. Table 3 shows that the absorption effect becomes significant (more than 1%) up to 30 µL for platinum and 40 µL for cobalt. Thus, the thickness of the layers used for the calibration is just in the limit to avoid important absorption effects. In the case of nanoparticles, the core–shell CoPt–hexadecylamine (about 70 molecules per nanoparticle) was taken into account with a deposited surface of 1 cm2. Supposing a decrease of signal of 1%, we found thicknesses of 0.614 µm (platinum) and 0.395 µm (cobalt) for the Co50Pt50 nanoparticles. As shown in Table 3, the thickness of the measured layer of nanoparticles is very low. Here, in these measurement conditions, the absorption effects could not influence the X-ray analysis. A significant but weak difference is observed for the 75% cobalt sample; this cannot be well explained by a matrix effect of the analytical method but rather by a shift in the chemical composition of the particles.

Table 3 Calculations of the decrease of the X-ray fluorescence emission by the absorption effect according to the thickness of the deposited layer (metallic salts and Co50Pt50 nanoparticles)

The average composition of the CoPt nanocrystals was also analysed by another method, EDX analysis with a scanning electron microscope. A film of nanocrystals with thickness of 5 µm was made on a silicon wafer (Fig. 4b, d) by deposition of colloidal solution. Figure 4a shows a typical EDX analysis obtained for Co50Pt50 nanoparticles with characteristic peaks of Kα and Kß cobalt and Mα and Lα platinum. These analyses were carried out on the previous samples. The measurement data are in good agreement with the expected values, as shown in Fig. 4c.

Fig. 4
figure4

a Energy-dispersive X-ray (EDX) analysis of CoPt nanocrystals on a silicon wafer showing the characteristic peaks of Kα and Kß cobalt and Mα and Lα platinum (accelerating voltage from 1 to 10 kV). b, d Scanning electron microscope images of nanocrystal film (accelerating voltage of 5 kV). c Percentage of cobalt in CoPt nanocrystals cobalt versus percentage of cobalt in the precursor salt

Indeed, the linear variation between the composition of the nanocrystals and the metallic salt solution indicates a precise control of the cobalt-to-platinum ratio. These analyses were achieved in global (5 µm × 5 µm) and local (point analysis) mode with similar results. The software used to analyse the EDX spectra corrects standard artefacts such as the spectrum background and the effect of close peaks. The concentration profile EDX spectra were treated by the ZAF analysis technique [22], which performs computing corrections of spectra by iterations from the first approximate composition. However, in the case of diffusion, there is an artefact whose effects on the concentration profiles are very important and that is generally not treated by software. This artefact could be called pear effect and is caused by the shape of the volume of the X-ray emission, whose dimensions can be of the same order of magnitude as the penetration length of diffusion (a thickness of 1 µm). Thus, matrix effects such as absorption and fluorescence were partially evaluated, and afterwards corrected to obtain a final converged composition. In contrast to the XRF analysis, the EDX analysis does not show variations for samples of nanocrystals with a high content of cobalt. Figure 5 shows the intercalibration graph for cobalt for EDX and XRF data. The EDX microanalysis exhibits a systematically higher cobalt value than XRF analysis. In the EDX analysis, the pear-shaped volume of X-ray emission has a depth of about 1 µm. Thus, a large amount of matter is probed in an important thickness. This induces standard phenomena during the measurement: XRF and a matrix effect. These effects can explain the composition difference between both techniques. In our XRF device, the analyses were carried out using a very small amount (less than 1 µg) of nanoparticles. This parameter of chemical analysis is crucial because a small quantity of matter avoids the matrix effect and allows a more local measurement. Therefore, in this case, the analysis becomes more sensitive to local composition defects. In addition, this suggests that the segregation effects, which can occur in bimetallic nanoparticles such as CoPt nanoalloy, can affect the chemical analysis. The composition ratio of 50:50 is recovered with a difference of 1.5% with XRF spectroscopy but 2.8% with EDX analysis (Fig. 5). Here, the accuracy of nanoalloy composition analysis seems to be better for the XRF technique. Finally, it is worth noting that the accurate determination of nanocrystal composition is increasingly important and that this XRF set-up appears to be an efficient method.

Fig. 5
figure5

Intercalibration graph for XRF and EDX data. A dotted line with slope 1 is also plotted

Conclusion

Thin-layer XRF spectroscopy appears to be an efficient, easy to use and accurate analytical method to determine the chemical composition of CoPt bimetallic nanoparticles. The results obtained by XRF spectroscopy were found to be in accordance with the EDX results more widely used by our community. Both analysis methods indicate a good controlled composition for CoPt nanoparticles synthesized by a biphasic method. The limiting factors for the accuracy of the determination of absolute quantities of cobalt and platinum by the XRF technique are the homogeneity of the deposited amount of the nanoparticle suspension (small drop) and also the calibration step depending on the thickness of the metallic salts. Only a small quantity of sample is necessary for the XRF analysis and a large number of samples can be routinely and automatically processed. This possibility to measure a small amount of nanoparticles avoids the matrix effect and allows a more local measurement. The proposed method can be translated to a small desktop energy-dispersive XRF machine, to make X-ray determination of the composition of nanoparticles by XRF spectroscopy a cheap method. Finally, for the samples of nanoalloys, XRF spectroscopy on a silicon wafer appears as a good complementary technique to standard techniques such as EDX analysis.

References

  1. 1.

    Grundy PJ (1998) J Phys D Appl Phys 31:2975

    Article  CAS  Google Scholar 

  2. 2.

    Sandhu A (2007) Nat Nanotechnol 2:746–748

    Article  CAS  Google Scholar 

  3. 3.

    Snytnikov PV, Yusenko KV, Korenev SV, Shubin YV, Sobyanin VA (2007) Kinet Catal 48(2):276–281

    Article  CAS  Google Scholar 

  4. 4.

    Demortière A, Launois P, Goubet N, Albouy PA, Petit C (2008) J Phys Chem B 112(46):14583–14592

    Article  Google Scholar 

  5. 5.

    Wilcoxon J (2009) J Phys Chem B 113:2647–2656

    Article  CAS  Google Scholar 

  6. 6.

    Weller D, Moser A (1999) IEEE Trans Mag 35(6):4423

    Article  CAS  Google Scholar 

  7. 7.

    Sun X, Jia ZY, Huang YH, Harrel JW, Nikles DE, Sun K, Wang LM (2004) J Appl Phys 95:6747

    Article  CAS  Google Scholar 

  8. 8.

    Tsukamoto R, Muraoka M, Seki M, Tabata H, Yamashita I (2007) Chem Mater 19:2389–2391

    Article  CAS  Google Scholar 

  9. 9.

    Kim JH, Kim J, Oh N, Kim YH, Kim CK, Yoon CS, Jin S (2007) Appl Phys Lett 90:023117

    Article  Google Scholar 

  10. 10.

    Kim J, Rong C, Lee Y, Liu JP, Sun S (2008) Chem Mater 20(23):7242–7245

    Article  CAS  Google Scholar 

  11. 11.

    Valcárcel M, Simonet BM, Cárdenas S (2008) Anal Bioanal Chem 391:1881–1887

    Article  Google Scholar 

  12. 12.

    Ferrando R, Jellinek J, Johnston RL (2008) Chem Rev 108(3):847

    Article  Google Scholar 

  13. 13.

    Sanchez SI, Small MW, Zuo JM, Nuzzo RG (2009) J Am Chem Soc 131(24):8683–8689

    Article  CAS  Google Scholar 

  14. 14.

    Abdelsayed V, Aljarash A, El-Shall MS, Al Othman ZA, Alghamdi AH (2009) Chem Mater 21(13):2825–2834

    Article  CAS  Google Scholar 

  15. 15.

    Ouerghi A, Penuelas J, Andreazza-Vignolle C, Andreazza P, Bouet N, Estrade-Szwarckopf H (2006) J Appl Phys 100:124310

    Article  Google Scholar 

  16. 16.

    Park JI, Kim MG, Jun YW, Lee JS, Lee WR, Cheon J (2004) J Am Chem Soc 126:9072

    Article  CAS  Google Scholar 

  17. 17.

    Losno R, Bergametti G, Mouvier G (1987) Environ Technol Lett 8:77–87

    Article  CAS  Google Scholar 

  18. 18.

    de Chateaubourg P, Quisefit JP, Garivait S, Steiner E, Goyon C (1993) Analusis 21:293–298

    Google Scholar 

  19. 19.

    Petit C, Rusponi S, Brune H (2004) J Appl Phys 95:4251

    Article  CAS  Google Scholar 

  20. 20.

    Lisiecki I, Parker D, Salzemann C, Pileni MP (2007) Chem Mater 19(16):4030–4036

    Article  CAS  Google Scholar 

  21. 21.

    Shevchenko EV, Talapin DV, Schnablegger H, Kornowski A, Festin Ö, Svedlindh P, Haase M, Weller H (2003) J Am Chem Soc 125(30):9090–9101

    Article  CAS  Google Scholar 

  22. 22.

    Brust M, Bethell D, Schiffrin DJ, Kiely CJ (1995) Adv Mater 7:795

    Article  CAS  Google Scholar 

  23. 23.

    Demortière A, Petit C (2007) Langmuir 23:8575–8584

    Article  Google Scholar 

  24. 24.

    Quisefit JP, Garivait S, Losno R, Steiner E (1995) Nucleus 32:135–142

    Google Scholar 

  25. 25.

    Nuspl M, Wegscheider W, Angeli J, Posch W, Mayr M (2004) Anal Bioanal Chem 379:640–645

    Article  CAS  Google Scholar 

  26. 26.

    Cheng G, Guo T (2002) J Phys Chem B 106:5833–5839

    Article  CAS  Google Scholar 

  27. 27.

    Speiser C, Baumann T, Niessner R (2001) J Anal Chem 370:752–759

    Article  CAS  Google Scholar 

  28. 28.

    Quisefit JP, de Chateaubourg P, Garivait S, Steiner E (1994) X Ray Spectrom 23:59–64

    Article  CAS  Google Scholar 

Download references

Acknowledgement

The authors would like to express their gratitude to Dr. Anh-tu Ngo, LM2N (UMR 7070), Paris, France, for his work on the EDX analysis of nanocrystal samples.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Arnaud Demortière.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Demortière, A., Losno, R., Petit, C. et al. Composition study of CoPt bimetallic nanocrystals of 2 nm. Anal Bioanal Chem 397, 1485–1491 (2010). https://doi.org/10.1007/s00216-010-3689-5

Download citation

Keywords

  • CoPt nanoparticles
  • Nanoparticle composition
  • X-ray fluorescence spectroscopy
  • Energy-dispersive
  • X-ray analysis
  • Nanoalloy
  • Bimetallic nanocrystals