1994 atomic structure of longitudinal sections of a pitch based carbon fiber studied by stm

applied surface science

ELSEVIER Applied Surface Science 74 (1994) 73-80

Atomic structure of longitudinal sections of a pitch-based carbon fiber studied by STM

P.W. de Bont a, P.M.L.O. Scholte a,*, M.H.J. Hottenhuis b, G.M.P. van Kempen a, J.W. Kerssemakers a, F. Tuinstra a

n Solid State Group, Department of Applied Physics, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands b Akzo Research Laboratories Arnhem, Corporate Research, P.O. Box 9300, 6800 SB Arnhem, The Netherlands

(Received 6 May 1993; accepted for publication 14 September 1993)

Abstract

Longitudinal sections of pitch-based carbon fibers have been studied with scanning tunneling microscopy. A hexagonal superstructure due to a rotation of 9.5” of the top graphitic plane with respect to the underlying bulk was observed. Remarkably this superstructure was modulated near defects by a (6 x fi)R30” modulation. The same modulation was found on the images with atomic resolution. It was concluded that the atomic structure of the fiber resembles the hexagonal structure of graphite. But locally this structure is disturbed. From the modulation of the superstructure it is deduced that this disturbance extends at least two layers into the bulk.

1. Introduction

Carbon fibers form a class of carbon modifica- tions, with remarkable mechanical properties that make them attractive for applications in compos- ite materials. The structure and morphology of carbon fibers have been studied extensively [1,21. It has been shown conclusively that they consist of graphitic layers that are preferentially oriented parallel to the fiber axis; however, the mechanical properties of the fibers do not resemble those of graphite as can be seen from the observed high Young’s moduli up to 800 GPa. This difference in mechanical properties can be ascribed to the

* Corresponding author.

orientation distribution of the graphitic layers in the fiber. In a frequently used model the fiber is thought to consist of a disordered core that is surrounded by an ordered mantle. Both the core and the mantle consist of graphitic layers that are preferentially oriented parallel to the fiber axis. At the atomic level these graphitic layers are thought to be connected by interlinking, i.e. merging of different layers 111, or by covalent cross-linking [2]. In the latter case the layers are connected by sp3 bonds between some of the C atoms in adjacent layers. Through this connection the weak van der Waals interaction, which is present in ordinary graphite crystallites, is re- placed by strong chemical bonds. This immobi- lizes the layers with respect to each other and consequently increases the shear modulus be- tween the graphitic layers.

0169-4332/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved SSDI 0169-4332(93IE0221-7

74 P. W. de Bent et al. /Applied Surface Science 74 (1994) 73-80

The atomic structure of carbon fibers has been investigated before with scanning tunneling mi- croscopy @TM) [3,4]. In these studies either the outer surface or sections perpendicular to the fiber axis were analyzed. However, longitudinal cuts should be studied in order to understand the relation between the high moduli and the atomic structure of the graphitic layers. Because the graphitic layers are oriented parallel to the fiber axis, only a longitudinal section allows the atomic structure of the layers to be imaged, while per- pendicular cuts do not. In this paper we present the results of a STM study of longitudinal sec- tions of a pitch-based carbon fiber.

The paper is organized as follows. After a short introduction into the experimental details, first a superstructure that has been observed on highly oriented pyrolytic graphite (HOPG) will be discussed. In the subsequent sections the results on the longitudinal cuts are presented. In the last section the superstructure on HOPG is used to draw conclusions about the atomic structure of the fibers.

2. Experimental details

The samples used in this study were pitch- based carbon fibers of the Carbonic HM70 type produced by Kashima oil. This fiber has a Young’s modulus of 716 GN m-’ [2]. The structural pa- rameters of this fiber were determined by Northolt et al. [2] with X-ray diffraction. It was concluded that the fiber consisted of small graphite crystallites with lattice parameters de- pendent on the crystallite size, but very close to the values of the lattice parameters of the hexag- onal graphite lattice. The size of the crystallites in the HM70 fiber parallel and perpendicular to the c-axis of the graphite lattice was found to be 17.0 and 59.2 nm, respectively. The structural p%rame- ters were determined as d(10) = 2.131 A and d(002) = 3.411 A [21.

For the STM experiments a bundle of fila- ments was embedded in a resin, each of the filaments with a diameter of approximately 10 pm. Subsequently a longitudinal section was made through the resin and fibers with an ultramicro-

tome (LKB, 2128) using a diamond knife. Electri- cal contact between the filaments and the STM sample holder was made at one end of the fiber bundle, with a small drop of Eccobond 66C.

The scanning tunneling microscope used in this study was of the Beetle type [5]. Constant- current images were taken in air using a Pt/Ir tip. The tunnel current was set between 1 and 10 nA and the bias voltage of the tip was in between -0.33 and 0.33 V. No difference was observed between empty state and filled state images. Each image consisted of 512 x 512 pixels. Scans were made over areas from 40 x 40 A2 up to 4900 x

4900 AZ. The surfaces of the longitudinal sections ap-

peared to be very rough. In order to be able to obtain lateral atomic resolution, it was necessary to apply a hardware high-pass filter. The filter enhances features in the image with high fre- quencies, such as step edges, and removes low- frequency features, such as a tilt. The effect is similar to the effect of a derivative filter. As a result the images show the corrugation of the gradient of the height, rather than the corruga- tion of the height itself. Apart from this hardware filtering all images in this work represent raw data.

3. Results

3.1. Superstructures on HOPG

In Fig. la a superpericdic structure is shown with periodic&y 40 f 3 A, which has been ob- served near a defect on HOPG. The resolution of the image is sufficient to observe the atomic periodicity of the graphite net, in addition to the superstructure. Similar superstructures on graph- ite (HOPG) have been observed earlier by Oden et al. [6] and Kuwabara et al. [7]. These images can be understood as an atomic moire pattern due to a rotation of the top graphite plane with respect to the underlying bulk [71.

This can be seen most easily in the reciprocal space. The two-dimensional power spectrum of the hexagonal graphite net contains six wavevec- tors. These wavevectors are related by symmetry.

P. W. de Bent et al. /Applied Surface Science 74 (1994) 73-80 75

Fig. 1. (a) 153 X 153 A* constant-current image of HOPG (I,,, = 2.0 nA, V,, = 0.50 V) near a defect on the surface. A modulation with a periodicity of 40 f 3 w is observed of the atomic hexagonal graphite net. (b) Power spectrum calculated from (a). Two groups of wavevectors are observed. The six peaks at small wavevector values are due to the superperiodic structure. The six broad peaks at large wavevector values are due to the periodicity of the graphite net. Note that the relative contrast of the latter set of peaks has been increased to make them visible.

In Fig. 2 three wavevectors of a graphite net are shown in the reciprocal space, together with the three wavevectors of the same graphite net that has been rotated over a small angle 6. The result-

Fig. 2. Generation of a moire pattern in reciprocal space. Two hexagonal nets, each represented by three respectively dashed and dashed-dotted wavevectors, are rotated with respect to each other over an angle 0. The resulting moire pattern is generated by the small solid difference wavevectors.

ing moire pattern is represented by the small solid difference vectors in Fig. 2. From this figure it is immediately obvious that the resulting super- structure has hexagonal symmetry and from sim- ple goniometry it follows that the periodicity P in real space of the superstructure can be expressed as:

P = +p/sin( +8),

where p is the periodicity of the graphite net as observed with an STM (p = 2.46 A) and 0 is the rotation angle of the uppermost graphite layer. For symmetry reasons this formula is correct only if -6O”IeI60”.

In Fig. lb the power spectrum is shown of the moire pattern in Fig. la. Six peaks are observed at small wavevector values from the superperiod- icity and six very broad peaks from the periodicity of the graphite net. The broadening in the verti- cal direction of the latter peaks is due to the limited correlation between subsequent scan lines. From the ratio between the superperiodicity and the periodicity of the graphite net a rotation angle of 0 = 3.5” k 0.3” can be calculated.

76 P. U? de Bont et al. /Applied Surface Science 74 (1994) 73-80

Because of the atomic resolution that has been achieved in Fig. la, the rotation angle may be determined also in a different, independent way. From Fig. 2 it can be deduced that the angle between the orientations of the wavevectors of the super-periodic structure and those of the atomic structure should be (90 - $9). From the maxima in the power spectrum shown in Fig. lb this angle was found to be 88.5”. From this value a rotation angle 0 = 3” is calculated. This value is in good agreement with the value 8 = 3.5” f. 0.3” deduced from the relative length of the wavevec- tors of the graphite lattice and the superstruc- ture. Therefore, we conclude that the interpreta- tion of the superstructure in terms of an atomic moire pattern is justified.

Fig. 3. 610X610 A* constant-current image (ZrUn = 9.1 nA, V,, = - 0.31 V) of a Carbonic HM70 carbon fiber withOsuper- structure. The period of the superstructure is 14.9 A. The square regions have been used to calculate the power spectra from.

3.2. Superstructures in fibers

In Fig. 3 a 610 x 610 A2 scan of Carbonic HM70 carbon fiber is shown. The STM image

Fig. 4. (a) Power spectrum calculated from the region at the lower region in the lower right corner of Fig. 3. The inset identifies the mo lines are just to guide the eye.

left corner of Fig. 3. (b) Power spectrum calculated from the st significant peaks. The symbols are explained in the text. The

P. W. de Bont et al. /Applied Surface Science 74 (1994) 73-80 17

shows a hexaEona1 superstructure with a periodic- ity of 14.9 A. This periodicity is too large to represent the translation symmetry of the graphite net. It is a superstructure similar to the atomic moire pattern observed in HOPG. This is corrob- orated by the large defect area that is visible in the upper right corner. Usually on HOPG the moire superstructure is observed near defects such as steps or grain boundaries [6,7].

Immediately below the defect area a (6 X &)R30” modulation of the intensities of the superstructure is visible in Fig. 3. The amplitude of this modulation decreases from the defect area.

From the superperiodicity of 14.9 A it can be estimated that the top layer in this longitudinal section of the fiber is rotated over 9.5” with respect to the underlying bulk.

To analyze the STM images in more detail, regions of interest were defined of 128 x 128 pixels in each image. The power spectra of these regions were calculated.

in Fig. 4a the power spectrum of the region in the lower left corner of Fig. 3 is shown. Only the six peaks of the hexagonal net of the superperiod- icity are visible, and two peaks in the center that are artefacts of the FFT routine used to calculate the power spectrum. The resolution of Fig. 3 is not sufficient to resolve the underlying periodicity of the graphite net, therefore the wavevectors of the atomic graphite net are also missing in the power spectrum.

Fig. 4b shows the power spectrum of the re- gion just below the large defect area in the upper right corner of Fig. 3. This spectrum displays many more peaks. The most significant peaks are identified in the inset of Fig. 4b. The peaks near the center of Fig. 4b (open circles in the inset) are artefacts due to the FFT transformation. Six peaks are at the same positions as in Fig. 4a and are due to the hexagonal superperiodic net (indi- cated by solid squares). Six peaks at small wavevector values (marked by solid circles) origi- nate from the (6 X fi)R30” modulation of the corrugation in this part of Fig. 3. In addition six higher-order peaks (open squares) are visible that can be attributed as the sum of a wavevector of the hexagonal superperiodic net and a wavevec- tor of the (fi X J?;)R30” modulation.

3.3. Atomic structure of fibers

Atomic resolution could be obtained on a small fraction of the exposed longitudinal sections only. This is attributed at least partly to the roughness of the samples and to the contamination of the surface during the cutting process. The region over which atomic resolution could be obtained, is limited also by the random orientation of the graphitic layers. Although the graphitic layers are aligned parallel to the fiber axis, they need not to be parallel to the stanning ptane of the STM tip.

In Fig. 5 a 76 A by 76 A area is shown on which atomic resolution was achieved. The upper left area again shows a defect area. On other parts of the image clearly the hexagonal pattern of the graphite net can be observed. The corruga- tion and the apparent periodicity change over the displayed area. At some parts in this figure the atomic structure is blurred due to contamination. On close inspection, however, it can be seen that the atomic structure continues in registry with the parts with full atomic resolution.

In Fig. 6 the power spectrum of the indicated area in the upper right corner of Fig. 5 is shown.

Fig. 5. 16.3 X 76.3 A2 constant-current image (It,, = 2.0 nA, VtiP = -0.33 V) of a Carbonic HM70 carbon fiber, showing the atomic structure of the graphitic layers.

78 P. W de Bont et al. /Applied Surface Science 74 (1994) 73-80

Six peaks from the hexagonal graphitic net are observed. But additionally four extra peaks emerge at smaller wavevector values. These peaks are at the positions of a (6 X &)R30” modula- tion, although for the fully symmetric modulation six (fi x fi)R30” peaks should have been ob- served. Also a number of peaks is visible that are understood to be linear combinations of wavevec- tors from the hexagonal graphitic net and the (6 x &)R30” wavevectors.

In different regions in Fig. 5 additional peaks were always found at the same positions in recip- rocal space. Only the relative intensities of the peaks changed from one region to the other. In Fig. 7 the positions of the peaks corresponding to the five regions indicated in Fig. 5 are given. The solid circle represents the length of the wavevec- tor of the ideal graphite net. The peaks related to the (fi x fi)R30” modulation on ideal graphite should all lie on the dashed circle. From this figure we conclude that the atomic structure shown in Fig. 5 is compatible with the translation symmetry of the graphite net. This is in accor- dance with the X-ray observations by Northolt et al. [2]. The altered appearance of the atomic structure in different regions of Fig. 5 is mainly

Fig. 6. Power spectrum calculated from the region in the

upper right corner of Fig. 5.

Fig. 7. Superimposed power spectra calculated from the five

regions indicated in Fig. 5. The light squares represent the

wavevectors from the hexagonal graphitic net, the dark squares

are due to the (fixfi)R30” modulation. The circles are

explained in the text.

due to the changes in relative intensities of the fourier components [8,9]. Its is not due to changes in the atomic structure.

The (fi X 6)R30” modulation turned out to be a general feature. A defect area was visible on all the images with atomic resolution. In all those image peaks (fi X &)R30” modulation could be identified in the power spectrum. So we conclude that the (fi X fi)R30” modulation is at least partially present at the atomic level.

4. Discussion and conclusion

On HOPG the (6 x &)R30” reconstruction is often observed close to defects or adatoms 181. It is not a reconstruction in the sense that atoms are displaced over or removed from the surface. But rather the electronic charge density is modu- lated due to the presence of an impurity or a defect. This is similar to the Friedel oscillations in a charge density around an impurity. The charge density tries to screen the defect or adatom. So away from the impurity the amplitude of the density modulation will decrease, as can be

P. W de Bont et al. /Applied Surface Science 74 (1994) 73-80 79

observed in Fig. 3. The (fi x fi)R30” wavevec- tors represent the first-order components of such a modulation in reciprocal space. Higher compo- nents do not appear in the power spectra shown in Figs. 4 and 6, because of the limited resolution of the STM.

In the previous section we pointed out that the (6 x fi)R30” modulation is present in the fibers, although the intensities of the Fourier components in the power spectrum change be- tween different parts of the surface. An STM image is a convolution of the point-spread func- tion of the tip and the (electronic) structure of the surface. Therefore, at first sight it is not clear whether the changes in intensity are due to the surface structure or to random changes in the tip state. But the intensities of the (fi X fi)R30” Fourier components do not change randomly over different parts of the image. If one considers individual scan lines, the Fourier intensities change at the same point for a large number of consecutive scan lines. Therefore, we are confi- dent that the variations in intensity are due to a surface effect. Xhie showed that combinations of (6 x fi)R30” Fourier components with differ- ent intensities, give rise to totally different real- space images [8]. This explains the different ap- pearance of the atomic structure in the different regions in Fig. 5. The intensity changes of the Fourier components can be understood to arise from the differences between the defects that cause the (a X fi)R30” modulation in the re- spective regions. As in the case of the electro- static screening of an impurity, the spatial distri- bution of the screening charge density will adapt itself to the symmetry of the defect. The Fourier components of the (6 x &)R30” modulation will be affected especially, since they arise di- rectly from the presence of the defect. This ex- plains the presence of only four (6 x fi)R30 wavevectors in Figs. 6 and 7, while all six wavevectors from the graphitic net are present.

the defect must extend at least two layers deep. An example of such a defect are cross-links be- tween neighboring graphitic layers. Northolt et al. concluded from their X-ray experiments that the graphitic layers were cross-linked by covalent sp3 bonds, near the edge of the layers [2]. Such a bond will deform the graphite net locally and may give rise to the (fi x fi)R30” modulation over the atomic structure. Also the second graphite plane that is connected to the top layer will contain a similar defect. If only one bond is present the two involved graphitic layers are free to rotate around the cross-link with respect to each other. A similar construction as in Fig. 2 shows that the superposition of two (6 x &)R30”-modulated hexagonal nets results in a superstructure with a similar modulation. Such a superstructure has been observed and is shown in Fig. 3. A superposition of two (6 x fi)R30” modulated graphitic nets is the simplest model to explain the (6 X &)R30” modulation of the moire pattern. It can be seen in Fig. 3 that the (6 x &)R30” modulation is constrained around the large defect. This suggests that the cross-links are located in the neighborhood of the large defect.

We conclude that in the pitch-based carbon fibers Carbonic HM70 the atomic structure is similar to the atomic structure in graphite. How- ever, the graphitic nets contain a large number of defects as can be deduced from the presence of the (6 x fi)R30” modulation of the STM im- age on the atomic scale. But these defects are not limited to the top layer. From the presence of the modulation on the superstructure it can be con- cluded that at least also the second layer contains sufficient defects to modulate the electronic charge density.

Acknowledgments

It should be noted that we cannot decide con- Mr. A.J. van de Berg is acknowledged for clusively the nature of the defect that causes the (6 X &)R30” modulation. But since the (fi

enlightening discussions and assistance with the

X filR30” modulation has been observed on the analysis of the images. One of us (P.d.B.1 grate- fully acknowledges the financial support of Akzo

superperiodic structure also, we conclude that Research Laboratories.

80

References

P. W. de Bont et al. /Applied Surface Science 74 (1994) 73-80

[5] K. Besocke, Surf. Sci. 181 (1987) 145. [6] P.I. Oden, T. Thundat, L.A. Nagahara, S.M. Lindsay, G.B.

ill

PI

[31

[41

J.-B. Donnet and R.C. Bansal, Carbon Fibers (Dekker, New York, 1984) chs. 1 and 2. M.G. Northolt, L.H. Veldhuizen and H. Jansen, Carbon 29 (1991) 1267. W.P. Hoffman, W.C. Hurley, T.W. Owens and H.T. Phan, J. Mater. Res. 26 (1991) 4545. S.N. Magonov, H.-J. Contow and J.-B. Donnet, Polym. Bull. 23 (1990) 555.

Adams and O.F. Sankey, Surf. sci. L&t. 254 (1991) L454. [7] M. Kuwabara, D.R. Clarke and D.A. Smith, Appl. Phys.

Lett. 56 (1990) 2396. [S] J. Xhie, K. Sattler, U. Mueller, N. Venkateswaran and G.

Raina, Phys. Rev. B 43 (1991) 8917. [9] G.M. Shedd and P.E. Russell, Surf. Sci. 266 (1992) 259.