Identification of the rock types is critical for the knowledge of origins and history in geological studies^[1]. Usually, a lithological classificationis performed by experienced researchers based on the colors and morphologies, which is time-consuming with the accuracy not guaranteed^[2]. This fact motivates the development of widely used, efficient methodologies for the lithological description. Elastic properties and lithology can be predicted from seismic data bydata inversion, which is an inverse problem and for a given seismic data set many solutions to exist^[3]. Other techniques have also been used for element identification in the laboratory, including natural gamma radiation, X-ray fluorescence spectral scanning, laser-induced breakdown spectroscopy, Raman spectroscopy, etc^{[4, 5, 6]}.

As a general method of passing information from materials, optics can be served as the appropriate means to promote the development of geological engineering^[7]. Terahertz (THz) radiation, which is located between far-IR (infra-red) and millimeter-wave bands of the spectrum, is willing to be a “ silver bullet” in various aspects of geological and geophysics studies, including mineral characterization, organic evaluation, anisotropy determination, etc^{[8, 9, 10]}. According to our previous study, drill cuttings at several depths from an oil well were studied using THz spectroscopy. The cuttings were clearly classified based on Cluster analysis and principal component analysis methods^[11]. Composition and granularity are the main factors affecting the spectral response for powder samples as drill cuttings. However, for block samples, both the structure and the component are complicated, and the model and mechanism of the THz spectrum collected in one region can be hardly applied to other ones^[12]. Here in this study, THz spectroscopy was employed to study the block samples with variant lithologies from different regions. The key factors thatdetermine the THz spectral response characteristics of rocks with different classifications were stepwise studied and discussed. The samples were classified based on the relations between effective refractive index (n) and attenuation coefficient (a), in which the results were in agreement with the lithology. In addition, relations between the properties of the rocks and their THz parameters were built to explain the mechanism of rock classification. The variation tendency of n and a were plotted with the elemental content, porosity and the oil yield for granite, limestone, sandstone, and oil shale, respectively, indicating that the variant THz responses were resulted by varying compositions and structures for different classifications of rocks. We aim to employ the THz technique as a promising means for lithology determination, which will be a significant supplementary in geology research.

Twelve samples were divided into four groups according to their lithology (three samples per group), including deoiled sandstone, oil shale, granite and limestone. Specimen blocks and cores were cut into slices with the thickness around 2 mm. All investigated samples were summarized and numbered in Table 1. In order to confirm the basic information of the samples, XRD measurements were employed to detect the types of mineral components in the samples. As the most common mineral on earth, quartz was found in all types of rocks, which was the most abundant mineral in the sandstone and the granite. Carbonate minerals were the main components for the limestone and the oil shale, including calcite and dolomite.

Table 1

Table 1

Table 1 General information for all the samples Identifier Specific Name Lithology Sampling Position Other Properties G1 Biotite Granite Granite North Dabieshan Mountain - G2 Yuexi Granite Granite Yuexi County, Dabieshan Mountain - G3 Baimajian Granite Granite Baimajian Peak of Dabieshan Mountain - L1 Middle Cambrian Limestone Limestone Northwest of Hubei Province - L2 Chuanshan Limestone Limestone Chuanshan, Western Hubei Province - L3 Nanling Limestone Limestone Nanling Country, Anhui Province - S1 Low Permeability Sandstone (Well No. WX20) Sandstone Daqing North, Heilongjiang Province Porosity: 16.64% S2 Low Permeability Sandstone (Well No. WX11) Sandstone Daqing North, Heilongjiang Province Porosity: 14.83% S3 Low Permeability Sandstone (Well No. WX14) Sandstone Daqing North, Heilongjiang Province Porosity: 11.69% O1 Longkou Oil Shale Oil Shale Longkou, Shandong Province Oil Yield: 14.16 O2 Huadian Oil Shale Oil Shale Huadian, Jilin Province Oil Yield: 9.96 O3 Barkol Oil Shale Oil Shale Barkol Country, Xinjiang Province Oil Yield: 5.66

Table 1 General information for all the samples

A transmission THz-TDS system was employed for the THz measurements under atmosphere. The slices were placed in the test position with the THz wave focused on the oil shale surface and measured one by one. Structural and morphology features of the samples were examined with a scanning electron microscopy (SEM), and the information on components of the samples were determined by the energy dispersive X-ray spectroscopy (EDS).

Typical THz time-domain spectra acquired by scanning the sliced samples with the THz system were reported in Figs.2(a)— (d). Rocks with various components and structures complexly modulated the incident THz signal, resulting in changes of the amplitude and pulse width of the transmitted signals. The waveforms of the samples were variant from each other at the peak amplitude and the delay time, indicating that THz waves could discriminate between the physical properties of the samples. For further investigation on whether the rocks with different lithologies can be distinguished by their spectroscopic THz properties, the maximum amplitude peaks (A_p) and the corresponding time delay (τ ) of the THz signals were extracted, and here we determined and calculated the attenuation coefficient (a) as well as the effective refractive index (n) calculated by

a=ArefdAsam(1)

n=c(τsam-τref)d+1(2)

in which A_ref and A_sam represent the amplitude peaks, as well as τ _ref and τ _sam represent the time delay of the reference and the samples, respectively. d is the sample thickness, while c is the velocity of light. Then, as shown in Fig.2(e), n and a were plotted together to establish the predictive models among the parameters obtained in this study, in which the scatters with different colors represent the classification by lithology of the samples. As shown in Fig.2(e), a of the oil shale and the granite were larger than that of the limestone and the sandstone, whilen distributions of the samples with variant lithologies have little difference. The sandstones and limestones were located close to each other in the coordinate system, but it is possible to differentiate between them. It is clear from the figure that n correlates linearly in respect to a, with the slope and intercept varied between different classifications. Positive correlations were built between a and n for sandstone, limestone and granite, with the fitting slopes equaled 0.77, 0.23, and 0.27, respectively. Variant from the above three kinds of rocks, n of the oil shale is negatively correlated with a, and the slope was -0.47. In general, based on the values of a, n, as well as the linear relations between them, the scatter mainly divided into four gatherings, with the cluster results in agreement with the lithology of the samples.

The discrepancy in components and structures resulted in the variant THz responses. Visualizing and quantifying the microstructures and elements by SEM and EDS permit an accurate means to determine the petrophysical properties. As a kind of coarse-grained igneous rock, granite is usually tight and hard, which is suitable for building materials. XRD results in Fig.1 has revealed the composition as quartz, mica, zeolum, and fayalite, while the crystalline texture of quartz and orthoclase was clearly shown in Fig.3(a). Limestone composed mainly of calcium carbonate and considerable amounts of magnesium carbonate (dolomite). Those carbonates were distributed like hydrangeas inside the limestone, as shown in Fig.3(b). As a common sedimentary rock consisting of cemented quartz, the composition of sandstone is generally uniform, with varying porosity and permeability under different sedimentary conditions and environments. The micro pores and cracks of the sandstone can be clearly observed in Fig.3(c). Micrograph characterization of the oil shale is depicted in Fig.3(d), in which the organics known as kerogen are clearly observed as the dark area surrounded by clay and carbonate. EDS analysis offers the quantitate results of elements, in which the contents of Fe and Mg are respectively different among the granites and the limestones, corresponding to the Iron oxides and dolomite minerals, as shown in Fig.3(e).

Fig.1

	Figure Option ViewDownloadNew Window
	Fig.1 XRD patterns analysis about the mineral type of the samples with different lithologies

Fig.2

	Figure Option ViewDownloadNew Window
	Fig.2 Time dependent THz waveforms of the rocks (a): Granite; (b): Limestone; (c): Sandstone; (d): oil shale, respectively; (e): Relationship between a and n of rocks corresponding to their lithologies

Fig.3

	Figure Option ViewDownloadNew Window
	Fig.3 Surface morphology of (a) granite, (b) limestone, (c) sandstone, and (d) oil shale characterized by SEM; (e) Elemental analysis of the granite and limestone samples by EDS

Further study of the THz response to the influence factor is essential for the mechanism analysis. As shown in Fig.4(a) and (b), n and a were respectively plotted in the double-y axis with the content of Fe and Mg for granite and limestone. The variation tendency of n and a roughly conformed to that of the Fe and Mg content, indicating that the variant THz responses were resulted by different compositions among those samples. Actually, Iron oxides and compounds were proved to be the key component for the absorbing effect of THz wave. Meanwhile, dolomite also has a stronger absorption of the THz pulse than carbonate and quartz^[15]. Both of those minerals have higher molecular weights and densities, resulting in the positively correlated relations between n and the element content. The porosities of the deoiled sandstones used in this experiment were measured by mercury injection method. The sandstone can be regarded as the combination of pores and rock matrix. Therefore, the porosity of sandstone is inversely proportional to the refractive index and absorption. Pores in oil shalesare always occupied by insoluble organic matters known as kerogen. Owing to the vibration modes caused by intermolecular and intramolecular interactions, photons with the energies in the vicinity of energy levels corresponding to the vibration or rotation mode are more likely to be absorbed by organics. Therefore, strongly depends on the oil yields of oil shale. The dielectric constant of kerogen is much lower than that of the mineral matrix, thus higher oil content is responsible for lower n.

Fig.4

	Figure Option ViewDownloadNew Window
	Fig.4 THz-n and -a versus (a) Fe content for granite, (b) Mg content for limestone, (c) porosity for sandstone, and (d) oil yield for oil shale

It is clear fromthe figures that the elements, components and structural properties correlate with the THz parameters, with variant response modes for rocks with different lithologies. Therefore, the correlations between n and a are different for variant classifications. The mineral composition is the main factor affecting the absorption and refraction of the THz wave. For sandstone with stable elements, both the absorption and refraction of THz wave are negatively correlated with the porosity. Organic content in oil shale has a reverse effect on THz absorption and refraction, in which the higher oil yield results in a stronger absorption and lower refraction. Herein, the refractive index is positively correlated with the absorption for granite, limestone and sandstone, while the relation between them is negatively correlated for oil shale.

In summary, spectral characteristics of rocks with different lithologies were studied by THz-TDS. Calculated THz parameters, including n and a, were employed to classify and analyze the rocks. The samples were clearly classified based on the relations between n and a, in which the results were in agreement with the lithology. Besides, the variation tendency of THz parameters was plotted with the componential and structural properties, indicating that the variant THz responses were resulted by varying compositions and structures for different classifications of rocks. Our results prove that the THz technique isa promising means for determining the lithology as well as the petrophysics properties, which is willing to be a significant qualitative and quantitative means in geology research.