The morphology of dust grains, including individual particle size and shape as well as agglomeration state, has a great influence on optical properties. Therefore, it is essential to have a well-defined experimental data and theoretical models in order to identify the observed infrared (IR) emission spectra as accurate as possible since observed spectra provide information on not only the chemical composition and proportion of elements in chemical compounds, but also the physical properties of dust grains (e.g., size). However, it is not easy to incorporate all physical conditions (e.g., individual grain shape, size heterogeneity, agglomerate structure) into light scattering simulations. Thanks to advanced technology, we are able to perform the spectroscopic measurements of nano- and micron-sized particles using a synchrotron beam. Small-sized dust grains (less than 2 micron) are ideal targets for a highly collimated IR beam from a synchrotron which can focus a sample spot close to the diffraction limit, achieving high spatial resolution. Hence, we have carried out the spectroscopic extinction measurements of systematically arranged agglomerates using an IR microscope at the Super Photon ring-8 GeV (SPring-8), the Japanese synchrotron facility located in Hyogo. Our target samples are silicon compounds including SiO<font size="1">2</font>, Mg<font size="1">2</font>SiO<font size="1">4</font>, and MgSiO<font size="1">3</font>. We focused on the 8 to 12 micrometers wavelength range where the strongest stretching vibration bands appear. Here, we clearly demonstrate (1) to what extent the agglomeration and orientation affects the stretching vibration band profile, and (2) the deviation between the experimental and theoretical outcomes from the light scattering theories (Mie theory, the discrete dipole approximation (DDA), the transition matrix (T-matrix)) via the comparative study. We conceive that this will provide clues for clarifying the interpretation of the physical conditions of dust grains, for example in protoplanetary disks.