Domain investigation by the conventional Bitter pattern technique with digital image processing ∗ ) W. Szmaja, J. Balcerski Department of Solid State Physics, University of L ´ od´z Pomorska 149/153, PL 90–236 L ´ od´z, Poland Received 8 October 2001 In the cases of soft magnetic materials or complex domain configurations, investigations of the domain structure by the conventional Bitter pattern technique are generally difficult. It is demonstrated in this paper, referring to the examples of domain images of thin permalloy films and the basal surface of bulk cobalt single crystals, that this problem can be overcome by the application of digital image processing (DIP) system. In particular, the visibility limit in domain observation was expanded by an order of magnitude and high quality domain images could be obtained. Improvements over earlier results were achieved. PACS : 75.60.–d Key words: Bitter pattern method, magnetic domain structure, digital image processing

1

Introduction

Among various techniques for magnetic domain imaging, the conventional (or classical) Bitter pattern technique, in which the wet colloid and the conventional optical microscope are used, is the oldest one [1] and of great historical significance. But even today it remains an important and often used method owing to its low cost, ease of application and good surface sensitivity [2, 3]. Nevertheless, it is known that observations with the conventional Bitter pattern method are in general difficult for soft magnetic materials or complicated domain structures [2, 3]. In the present paper we demonstrate that this problem can be overcome thanks to the introduction of DIP system. As a consequence, it is possible to obtain high quality domain images and to analyze them in more detail. The domain structures in thin nickel-iron films and on the basal surface of bulk cobalt single crystals are used as examples. 2

Experimental

The domain images shown in this paper were recorded with the conventional Bitter pattern method supported by a DIP system. We used water-based colloidal suspension of magnetite (Fe3 O4 ) particles. Experimental conditions, concerning the optical microscope, the specimens and the colloid, were optimized to improve the Bitter pattern contrast. The optimum concentration of the colloid was determined experimentally. In the case of cobalt single crystals, the observed surface ∗)

Presented at 11th Czech and Slovak Conference on Magnetism, Koˇsice, 20–23 August 2001

Czechoslovak Journal of Physics, Vol. 52 (2002), No. 2

223

W. Szmaja and J. Balcerski

was smoothed, to reduce the topographic contrast, by mechanical polishing using standard metallurgical procedure. The DIP system consists of a monochrome CCD camera of high sensitivity (10 mlux), an IBM PC computer equipped with a frame grabber, and software. 3

Results and discussion

Domain studies in soft magnetic materials by the conventional Bitter pattern method are difficult due to the fact that these materials produce only weak stray fields. In practice, low contrast domain images are obtained. An example is shown in Fig. 1a which presents the original image (i.e. obtained directly from the optical microscope) of the domain structure in a thin permalloy film. But with the use of DIP system, we could easily produce high quality images. Application of a simple digital procedure for contrast improvement [4] to the original image from Fig. 1a resulted in an image with considerably enhanced contrast (not shown). However, the latter image was still somewhat unsharp and could be improved by further processing. After application to this image a procedure for sharpness improvement [4], the image of high quality, shown in Fig. 1b, was obtained, where the cross-tie walls, displayed as black lines, are clearly seen.

Fig. 1. (a) Original image of the domain structure in a thin Ni81 Fe19 film recorded by the conventional Bitter pattern method; (b) image obtained after applying digital procedures for contrast and sharpness enhancement to the original image.

With the use of the conventional Bitter pattern technique it is also difficult to investigate bulk cobalt single crystals of basal orientation, which exhibit strong stray fields above the basal surface. This problem was reported in a number of papers in the past (e.g. Ref. [5]), and is due to the complex character of the surface domain structure, with the magnetization direction changing continuously between adjacent domain regions [6, 7]. As a consequence, the local maxima of the stray field near the crystal surface are not clearly marked and the original images are of poor quality. However, as previously for permalloy, the difficulty can be overcome with the help of the DIP system. An example is shown in Fig. 2. The application of a simple procedure for contrast enhancement [4] to the original image of Fig. 2a 224

Czech. J. Phys. 52 (2002)

Domain investigation by Bitter pattern technique . . .

produced the image presented in Fig. 2b. It is now clearly visible that the latter image suffers from the lack of sharpness and therefore needs further processing. After applying a procedure for sharpness improvement [4] to the image from Fig. 2b, the image shown in Fig. 2c was obtained. In this image, the complicated character of the surface domain structure is visible with high contrast and clarity; the surface domains are displayed approximately as circles, stars or flowers.

Fig. 2. (a) Original image of the domain structure at the basal surface of a bulk cobalt single crystal taken by the conventional Bitter pattern method; (b) image obtained by applying a digital procedure for contrast enhancement to the original image; (c) image obtained by applying a digital procedure for sharpness improvement to the image (b); (d) image obtained by digital processing of the image (c).

In the context of complicated domain configurations, as for example that shown in Fig. 2, it is practically impossible to determine the domain width in a reliable way by visual measurements. In such a case employing DIP techniques is certainly much more appropriate. Here we present an example of determining the domain width by digital means using the stereologic method of Bodenberger and Hubert [8] (see also Ref. [2]), which appears to be the most universal and commonly used approach. The image considered is that shown in Fig. 2c. What we need is to determine the positions of domain walls. For this purpose, a simple thresholding operation on the level of the average image intensity and then a 3 × 3 pixel median filter were applied, resulting in the black-and-white image. From the latter image, the corresponding image shown in Fig. 2d was derived in a straightforward way, in which the domain boundaries are represented by black lines. To the image from Fig. 2d, the stereologic method with 1000 test lines was applied (in this image, Czech. J. Phys. 52 (2002)

225

W. Szmaja and J. Balcerski: Domain investigation . . .

4 test lines are additionally superimposed); the domain width D was determined

from the formula D = (2/π) ( i li ) / ( i ni ), where li is the length of the ith test line and ni is the number of intersections of the ith test line with domain walls. To evaluate the accuracy of determining the domain width, the procedure was repeated 20 times. The obtained average surface domain width is DS = (2.72 ± 0.03) µm. 4

Conclusions

We have shown in the present paper that the usefulness of the conventional Bitter pattern method can be extended by the introduction of DIP system. The domain structures in thin permalloy films and on the basal surface of bulk cobalt single crystals were used as examples. The visibility limit in domain observation was expanded by an order of magnitude. Poor quality original images of the Bitter patterns were enhanced through software processing to such a degree that the domain structure (in zero or near zero external magnetic fields) could be observed with high contrast and good sharpness. In the case of complex domain structures, the domain width can be determined by DIP methods more precisely and objectively than by visual measurements. Improvements over earlier results were achieved. Only thanks to the application of the DIP system, detailed domain studies, which were not possible before, could be performed [7, 9]. Despite the great advantages of using DIP system, we have to be aware that if the feature of interest is not present in the original image, then no amount of digital processing will be able to detect it. Therefore we must not forget about suitable specimen preparation and optimum experimental operating conditions. References [1] F. Bitter: Phys. Rev. 38 (1931) 1903. [2] A. Hubert and R. Sch¨ afer: Magnetic Domains: The Analysis of Magnetic Microstructures, Springer, Berlin, 1998, p. 12. [3] R.J. Celotta, J. Unguris, M.H. Kelley, and D.T. Pierce: in Methods in Materials Research: A Current Protocols Publication, Wiley, New York, 2000, Unit 6b.3. [4] T. Pavlidis: Algorithms for Graphics and Image Processing, Wydawnictwa NaukowoTechniczne, Warsaw, 1987, pp. 56, 72 (in Polish). [5] R. Gemperle, A. Gemperle, and I. Bursuc: Phys. Status Solidi 3 (1963) 2101. [6] J. Unguris, M.R. Scheinfein, R.J. Celotta, and D.T. Pierce: Appl. Phys. Lett. 55 (1989) 2553. [7] W. Szmaja: J. Magn. Magn. Mater. 219 (2000) 281. [8] R. Bodenberger and A. Hubert: Phys. Status Solidi A 44 (1977) K7 (in German). [9] W. Szmaja: J. Magn. Magn. Mater. 234 (2001) 13.

226

Czech. J. Phys. 52 (2002)