Three-dimensional reconstruction of surface nanoarchitecture from two-dimensional datasets
© Boshkovikj et al.; licensee Springer. 2014
Received: 8 October 2013
Accepted: 15 December 2013
Published: 10 January 2014
The design of biomaterial surfaces relies heavily on the ability to accurately measure and visualize the three-dimensional surface nanoarchitecture of substrata. Here, we present a technique for producing three-dimensional surface models using displacement maps that are based on the data obtained from two-dimensional analyses. This technique is particularly useful when applied to scanning electron micrographs that have been calibrated using atomic force microscopy (AFM) roughness data. The evaluation of four different surface types, including thin titanium films, silicon wafers, polystyrene cell culture dishes and dragonfly wings confirmed that this technique is particularly effective for the visualization of conductive surfaces such as metallic titanium. The technique is particularly useful for visualizing surfaces that cannot be easily analyzed using AFM. The speed and ease with which electron micrographs can be recorded, combined with a relatively simple process for generating displacement maps, make this technique useful for the assessment of the surface topography of biomaterials.
KeywordsThree-dimensional visualization Scanning electron microscopy Atomic force microscopy Surface topographical analysis
Advances in microscopic technology have revolutionized the way objects can be perceived on the nanoscale. Powerful instruments such as the scanning electron microscope (SEM) provide the ability to rapidly characterize samples using high magnification, producing two-dimensional (2D) image representation of the samples (Cizmar et al., 2008; Schatten, 2011), making it highly a suitable analytical technique in both biological and materials sciences (Coelho et al., 2009; Kang et al., 2009). The main drawback of the technique however, is that the data obtained is limited to two dimensions (i.e. height and width, but not depth) which can often limit the extent of understanding that can be obtained regarding the morphology and topographical features of many samples.
In order to perform an in-depth characterization of material surface structures and architecture, a number of different methods have emerged for the purposes of 3D surface visualization. Atomic force microscopy (AFM) is perhaps the most useful of these techniques, as it can be employed to generate topographical maps of a surface with sensitive and accurate measurement in the depth dimension (Binnig et al., 1986). This data can be employed to reconstruct three-dimensional surface models (Schift et al., 2009). AFM surface scans are physical in nature, in that the AFM tip physically interacts with a sample surface, producing 3D coordinates. Some drawbacks of the technique are that in cases where the surface topography is extremely complex or heterogenous in its mechanical properties, the resulting data may not be accurate. In addition, AFM scans are typically only able to sample a small section of the surface, and take a relatively long time to obtain a representation of a surface. In contrast, SEM can be used to sample much larger fields of view with high resolution over a shorter period of time. The main drawback of SEM is that unlike AFM, the height values are not measured in the depth dimension, and so it cannot be readily utilized for the 3D visualization of samples. To overcome this limitation, stereo scanning electron microscopes have been developed (Marschall et al., 2000). In this technique, two electron beams on two different angles are focused at the same point on a sample and the three-dimensional coordinates are measured, based on the data obtained from the two different perspectives (Ostadi et al., 2011). The ability to accurately compute the 3D points strongly depends on the ability to accurately match the two images (Samak et al., 2007).
The aim of this study was to develop and evaluate a technique for the effective three-dimensional reconstruction of the nanoarchitecture of a surface, based on two-dimensional electron micrographs. This technique combines the rapid data collection capability of SEM with the accurate three-dimensional measurement ability of AFM. As a result, large areas of material surfaces can be rapidly analyzed and subsequently presented as high-resolution 3D models.
Materials and methods
Four substrata surfaces were analyzed using SEM to demonstrate the versatility of the technique: 150 nm-thick thin titanium films (prepared as described previously (Wang et al., 2008), dragonfly (Hemianax papuensis) wings, standard polystyrene petri plates (Cellstar, Greiner Bio-One) and silicon wafers. Prior to analysis, all samples were rinsed with 70% ethanol and then MilliQ H2O, except in the case of the dragonfly wings, which were rinsed with water, as ethanol would damage the waxy surface structures of the wing. The dragonfly wings and small excised pieces of polystyrene were also coated with ~10 nm of gold using a Dynavac CS300 prior to electron microscopy in order to make the surfaces conductive (Mitik-Dineva et al., 2009; Truong et al., 2009).
Scanning electron microscopy (SEM)
The electron micrographs of all four sample types were recorded using a FESEM (ZEISS SUPRA 40VP) at 3 kV at 70000× magnification. In addition, titanium films were analysed at 30000×, 90000× and 150000× magnifications. All samples were initially viewed at lower magnification in order to identify suitable regions of the surface for analysis, prior to analysis at higher magnification.
AFM surface calibration
All AFM scans were conducted using an Innova scanning probe microscope (Veeco, U.S.A.). Scans were performed in tapping mode at ambient temperature and pressure, using silicon cantilevers (MPP-31120-10, Veeco, U.S.A.) with a spring constant of 0.9 N m-1 and a resonance frequency of approximately 20 kHz. Scanning was performed perpendicular to the axis of the cantilever at a scan speed of 1 Hz. Initially, 10 μm × 10 μm fields of view were scanned in order to identify suitable analysis regions of the surface, prior to analysis at higher resolution. Scan areas were selected to closely match with the resolution of electron micrographs, as the distance between individual sampling points can affect the calculated roughness parameters (Brune et al., 1997; Crawford et al., 2012).
Three-dimensional visualization of SEM images
The 3D visualization of SEM images was realized in the 3D animation software package Autodesk Maya® (http://usa.autodesk.com/maya/). Maya’s 3D modeling and shading/texturing capabilities were used for construction of displacement map from the SEM images and conversion into 3D polygonal geometry.
Construction of displacement maps
Displacement maps were constructed according to the following procedure: A polygonal plane was created as a reference point for the creation of 3D images from the electron micrographs. By default, the plane was a two-dimensional object. The resolution and dimension attributes of each plane were set according to the corresponding SEM image being used for modeling. The resolution of each plane was adjusted to match that of the SEM image, and the x- and y-dimensions (i.e. height and width) were set to correspond with those of the actual areas being analyzed. Data representing the depth dimension was extracted from the alpha values of each pixel in the SEM images. A simple script was developed using Python programming language (Python Software Foundation) in order to obtain the height values at each point of the surface topography. The script recorded the alpha values of each pixel as a relative translation attribute value for each vertex. The data was then stored in an empty comma-separated values file (.csv).
Roughness data calculated from AFM scans were used to calibrate the depth dimension scale of the displacement maps. The average value of the .csv files were set to match the average roughness (Ra) determined for each of the corresponding sample surfaces, and all other values were scaled proportionate to this value.
Conversion of displacement maps into 3D polygonal geometry
A Lambert shading material was assigned to the polygonal plane. The type of material selected does not affect the displacement mapping. A new 2D file texture node was created and connected to the Lambert Material as a displacement map attribute. The file node allows the importation of an image into Maya. A filter type can be chosen for the image, which will affect the quality of the image. However, it will also have an effect on the 3D geometry shape when the image is converted into a 3D object. For the SEM image the Gaussian filter was selected. The filter will give the best image quality, resulting in a smooth 3D model. It did not affect the main features of the object.
There are 2 options in Maya for conversion of an image into a polygonal geometry. One is ‘convert - displacement to polygons’ and the other is ‘convert- displacement to polygons with history.’ The first option converts the image, however, it adds additional subdivisions to the surface and the resolution will be excessively high to process. This option is sufficient to convert polygonal objects with low starting subdivision value. The resulting 3D object was also very messy when the unnecessary resolution was added to it. The second option performs the same task, but only deforms the already defined subdivisions of the plane, without adding additional geometry. The conversion process also triangulates the subdivisions.
Additional file 1: Three-dimensional surface models of titanium thin films, dragonfly wings and silicon wafer based on two-dimensional data. Comparative visualization of material surfaces of 150 nm-thick titanium thin films, dragonfly (Hemianax papuensis) wings and silicon wafers derived from atomic force microscopy (AFM) roughness data and scanning electron microscopy (SEM) imaging. (MP4 15 MB)
Visualization of titanium surfaces
In the figures presented here, the height scale of the displacement maps was calibrated according to the AFM data. AFM analyses were performed on each of the samples visualized, and the average roughness (Ra) was calculated. Average roughness is defined as the average deviation of the height values from the mean height (Stout et al., 1993; Brune et al., 1997; Webb et al., 2012). Each pixel within the displacement maps was assigned a proportional value between 0 and 1, therefore the average deviation from the mean of the pixel values was calculated, and the height values for each pixel were scaled so that the average deviation matched the Ra of the sample. It should be noted that the calculated Ra of a surface is dependent on the sampling interval, i.e. the resolution of the AFM scan, therefore scan areas were chosen in order to match the resolution of the AFM scans to that of the electron micrographs. Calibration of the height scales in this manner requires only a few time-consuming AFM scans, and the application of average roughness data can be applied to many SEM displacement maps. AFM is by no means the only technique available that can be used to calibrate the height scales; stylus and optical profilometers, for example, can also be used to record topographical data.
Other materials and sample types
In addition to the titanium surfaces, three different substrata surfaces were used to produce displacement maps in order to assess the versatility of the technique. The three surfaces chosen were specifically selected for their highly diverse chemical compositions and surface topographies. The surfaces included dragonfly wings, unmodified silicon wafers, and polystyrene (PS) plastic excised from standard cell culture Petri plates (Figure 2). Dragonfly wings are known to have relatively large surface features (Ivanova et al., 2013; Nguyen et al., 2013), while polystyrene and silicon wafers are known to be quite smooth (Decuzzi and Ferrari, 2010; Gentile et al., 2010; Zeiger et al., 2013), and all three samples are less conductive than the previously utilized metallic titanium. In the case of dragonfly wings, the SEM displacement maps provide a clear advantage over AFM scans. The large feature size, and the inherent ‘stickiness’ of the epicuticular lipids that are present on the surface of the wing make it quite difficult to produce accurate AFM scans that are free from artifacts. The SEM displacement map presents a clearly defined structure. The SEM displacement map and AFM scan of the silicon surface are also relatively comparable; both techniques produce a smooth, feature-less surface.
The advantage of this displacement map technique is in its ability to quickly generate 3D model representations of surfaces that can be more easily analyzed. In contrast to previous studies involving stereoimaging (Cuijpers et al., 2011), only a single SEM micrograph is required for the reconstruction of the 3D model. An additional advantage of using the displacement map technique is that complex mathematical computations are not required for the visualization process (Marschall et al., 2000; Samak et al., 2007). As a result, topographical surveys can be performed quickly, bearing in mind that the depth dimension of the data is an approximation.
To illustrate the advantages and limitation of the displacement map technique for generating 3D maps, consider the following example. An aerial-view photograph is recorded of a room full of people with a distribution of heights, colored in grey-scale according to the heights of each person. It is not a simple task to draw conclusions on the relative heights of different people, especially without reference to the color scale, however, it is of course much easier to obtain knowledge on the relative heights of the people when viewing from the side. The extra dimension allows the potential for insight into the height distribution to increase. The absolute heights of each individual could also be obtained by directly measuring them, however if this were time consuming the heights of a subsection of the people in the room could be measured, and provided that the distribution of heights throughout the population were homogeneous and random, data from that subsection could be applied to approximate the heights of the rest of the group. Thus, the relative heights of the people are accurately represented, and their absolute heights are approximated in a short time. Provided one keeps in mind the limitations of the technique, this can be a highly effective tool in analyzing large populations of data.
Finally, the results presented here demonstrate that displacement maps cannot always be applied to the surfaces of insulating materials. Samples that are weakly or moderately insulating can be coated with a thin layer of gold or carbon, which is common practice in electron microscopy (Wang et al., 2008). This process improves the conductivity of the surface and enables micrographs of adequate quality to be obtained. The thickness of the conductive coating must be minimized as much as practically possible in order to avoid losing surface details (Nowell and Pawley, 1980). For strongly insulating materials however, electron micrographs do not contain sufficient topographical detail to produce useful 3D displacement maps.
The ability to produce three-dimensional displacement maps based on two-dimensional data widens the range of analytical techniques that can be applied for the visualization and assessment of surface topography. This technique is particularly applicable to scanning electron micrographs, when calibrated appropriately using data collected from AFM scans. Displacement maps are particularly effective for visualizing conductive surfaces such as titanium, or for viewing surfaces that cannot be easily examined by other topography analysis tools, such as AFM.
This study was supported in part by Australian Research Council (ARC). Autodesk, Maya are registered trademarks or trademarks of Autodesk, Inc., and/or its subsidiaries and/or affiliates in the USA and/or other countries.
- Binnig G, Quate CF, Gerber C: Atomic force microscope. Phys Rev Lett 1986, 56(9):930–933. doi: 10.1103/PhysRevLett.56.930 10.1103/PhysRevLett.56.930View ArticlePubMedGoogle Scholar
- Brune D, Hellborg R, Whitlow HJ, Hunderi O: Surface Characterization. Weinheim: Wiley-VCH; 1997.View ArticleGoogle Scholar
- Cizmar P, Vladár AE, Ming B, Postek MT: Simulated SEM images for resolution measurement. Scanning 2008, 30(5):381–391. 10.1002/sca.20120View ArticlePubMedGoogle Scholar
- Coelho PG, Granjeiro JM, Romanos GE, Suzuki M, Silva NRF, Cardaropoli G, Van Thompson P, Lemons JE: Basic research methods and current trends of dental implant surfaces. J Biomed Mater Res B Appl Biomater 2009, 88(2):579–596.View ArticlePubMedGoogle Scholar
- Crawford RJ, Webb HK, Truong VK, Hasan J, Ivanova EP: Surface topographical factors influencing bacterial attachment. Adv Colloid Interface Sci 2012, 179–182: 142–149.View ArticlePubMedGoogle Scholar
- Cuijpers VMJI, Walboomers XF, Jansen JA: Scanning electron microscopy stereoimaging for three-dimensional visualization and analysis of cells in tissue-engineered constructs: technical note. Tissue Eng Part C Methods 2011, 17(6):663–668. 10.1089/ten.tec.2010.0455View ArticlePubMedGoogle Scholar
- Decuzzi P, Ferrari M: Modulating cellular adhesion through nanotopography. Biomaterials 2010, 31: 173–179. 10.1016/j.biomaterials.2009.09.018View ArticlePubMedGoogle Scholar
- Dmitriev K, Makarov E: Generating displacement from normal map for use in 3D games. ACM Special Interest Group on Computer Graphics and Interactive Techniques Conference, SIGGRAPH’11, Vancouver, BC, 2011 2011.Google Scholar
- Gentile F, Tirinato L, Battista E, Causa F, Liberale C, di Fabrizio EM, Decuzzi P: Cells preferentially grow on rough substrates. Biomaterials 2010, 31: 7205–7212. 10.1016/j.biomaterials.2010.06.016View ArticlePubMedGoogle Scholar
- Hiraiwa A, Nishida A: Noise filtering of scanning-electron-microscope images for accurate analysis of line-edge and line-width roughness. J Micro Nanolithogr MEMS MOEMS 2012a., 11(4): doi: 10.1117/1.jmm.11.4.043010Google Scholar
- Hiraiwa A, Nishida A: Scanning-electron-microscope image processing for accurate analysis of line-edge and line-width roughness. San Jose, CA: Metrology, Inspection, and Process Control for Microlithography XXVI; 2012b.View ArticleGoogle Scholar
- Ivanova EP, Nguyen SH, Webb HK, Hasan J, Truong VK, Lamb RN, Duan X, Tobin MJ, Mahon PJ, Crawford RJ: Molecular organization of the nanoscale surface structures of the dragonfly Hemianax papuensis wing epicuticle. PLoS One 2013, 8: e67893. 10.1371/journal.pone.0067893PubMed CentralView ArticlePubMedGoogle Scholar
- Jang H, Han J: Feature-preserving displacement mapping with graphics processing unit (GPU) tessellation. Comput Graph Forum 2012, 31(6):1880–1894. 10.1111/j.1467-8659.2012.03068.xView ArticleGoogle Scholar
- Joy DC, Joy CS: Low voltage scanning electron microscopy. Micron 1996, 27: 247–263. 10.1016/0968-4328(96)00023-6View ArticleGoogle Scholar
- Kang B-S, Sul Y-T, Oh S-J, Lee H-J, Albrektsson T: XPS, AES and SEM analysis of recent dental implants. Acta Biomater 2009, 5(6):2222–2229. doi:10.1016/j.actbio.2009.01.049 10.1016/j.actbio.2009.01.049View ArticlePubMedGoogle Scholar
- Lu TK, Low KL, Zheng J: Fast visualization of complex 3D models using displacement mapping. 35th Graphics Interface Conference, GI 2009, Kelowna, BC, 2009 2009, 7–14.Google Scholar
- Marschall HU, Danzer R, Pippan R: Three-dimensional analysis of decorated ceramic fracture surfaces by automatic stereophotogrammetry. J Am Ceram Soc 2000, 83(1):223–225. 10.1111/j.1151-2916.2000.tb01176.xView ArticleGoogle Scholar
- Mitik-Dineva N, Wang J, Truong VK, Stoddart P, Malherbe F, Crawford RJ, Ivanova EP: Escherichia coli , Pseudomonas aeruginosa , and Staphylococcus aureus attachment patterns on glass surfaces with nanoscale roughness. Curr Microbiol 2009, 58: 268–273. 10.1007/s00284-008-9320-8View ArticlePubMedGoogle Scholar
- Nguyen SHT, Webb HK, Hasan J, Tobin MJ, Crawford RJ, Ivanova EP: Dual role of outer epicuticular lipids in determining the wettability of dragonfly wings. Colloids Surf B: Biointerfaces 2013, 106: 126–134.View ArticlePubMedGoogle Scholar
- Nowell JA, Pawley JB: Preparation of experimental animal tissue for SEM. Scanning Electron Microsc 1980, Pt 2: 1–19.Google Scholar
- Ostadi H, Rama P, Rui C, Xiaoxian Z, Yuan G, Jiang K: 3D visualization and characterization of nano structured materials. Nanotechnology (IEEE-NANO), 2011 11th IEEE Conference on, 15–18 Aug. 2011 2011, 770–775. doi:10.1109/nano.2011.6144402Google Scholar
- Samak D, Fischer A, Rittel D: 3D reconstruction and visualization of microstructure surfaces from 2D images. CIRP Ann Manuf Technol 2007, 56(1):149–152. doi:10.1016/j.cirp.2007.05.036 10.1016/j.cirp.2007.05.036View ArticleGoogle Scholar
- Schatten H: Low voltage high-resolution SEM (LVHRSEM) for biological structural and molecular analysis. Micron 2011, 42(2):175–185. 10.1016/j.micron.2010.08.008View ArticlePubMedGoogle Scholar
- Schift H, Kim G, Lee J, Gobrecht J: 3D visualization of mold filling stages in thermal nanoimprint by white light interferometry and atomic force microscopy. Nanotechnology 2009., 20(35): doi:10.1088/0957–4484/20/35/355301Google Scholar
- Shiramine K-i, Muto S, Shibayama T, Sakaguchi N, Ichinose H, Kozaki T, Sato S, Nakata Y, Yokoyama N, Taniwaki M: Tip artifact in atomic force microscopy observations of InAs quantum dots grown in Stranski-Krastanow mode. J Appl Phys 2007, 101: Art. no. 033527.View ArticleGoogle Scholar
- Stout KJ, Sullivan PJ, Dong WP, Mainsah E, Luo N, Mathia T, Zahouani H: Development of methods for the characterisation of roughness in three dimensions. Publication no. EUR 15178. Luxembourg: EN of the Commission of the European Communities Dissemination of Scientific and Technical Knowledge Unit; 1993.Google Scholar
- Szirmay-Kalos L, Umenhoffer T: Displacement mapping on the GPU state of the art. Comput Graph Forum 2008, 27(6):1567–1592. 10.1111/j.1467-8659.2007.01108.xView ArticleGoogle Scholar
- Tabet MF, Urban FK III: Comparison of atomic force microscope and Rutherford backscattering spectrometry data of nanometre size zinc islands. Thin Solid Films 1996, 290–291: 312–316.View ArticleGoogle Scholar
- Truong VK, Rundell S, Lapovok R, Estrin Y, Wang JY, Berndt CC, Barnes DG, Fluke CJ, Crawford RJ, Ivanova EP: Effect of ultrafine-grained titanium surfaces on adhesion of bacteria. Appl Microbiol Biotechnol 2009, 83(5):925–937. 10.1007/s00253-009-1944-5View ArticlePubMedGoogle Scholar
- Wang JY, Ghantasala MK, McLean RJ: Bias sputtering effect on ultra-thin SmCo 5 films exhibiting large perpendicular coercivity. Thin Solid Films 2008, 517: 656–660. 10.1016/j.tsf.2008.07.033View ArticleGoogle Scholar
- Webb HK, Truong VK, Hasan J, Fluke C, Crawford RJ, Ivanova EP: Roughness paramet zers for standard description of surface nanoarchitecture. 2012. Scanning: [Early online edition]Google Scholar
- Zeiger AS, Hinton B, Van Vliet KJ: Why the dish makes a difference: quantitative comparison of polystyrene culture surfaces. Acta Biomater 2013, 9: 7354–7361. 10.1016/j.actbio.2013.02.035View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.