1 Introduction

Mineral chemistry data of flood basalt provinces often play an important role as significant petrogenetic assessor (Marzoli et al. 1999; Dey et al. 2021). Although for flood basalt provinces, high precision major, trace and isotopic data are regarded as sensitive petrogenetic indicators, mineral chemistry data serves as significant tool (in association with whole-rock geochemical data) for finger printing petrogenetic processes (Moraes et al. 2018; Macedo Filho et al. 2019). In some cases, the mineral chemistry data alone have been rewardingly used to unravel certain primary magmatic features like appearance of liquidus temperatures of different minerals, oxygen fugacity and magma density (Ganguly et al. 2012; Rao et al. 2012). In the case of Deccan flood basalt province, importance of using such mineral chemical data has recently been highlighted by Dey et al. (2021). The mineral chemistry data interpretation can be extended to almost all flood basalt provinces. For recognition of magma chamber processes, the importance of mineral chemistry data has been dealt with in detail for Deccan Volcanic Province (Melluso and Sethna 2011; Ganguly et al. 2012; Rao et al. 2012). Considering the significance of mineral chemistry data (for understanding the petrogenetic processes of flood basalt provinces) in the backdrop, the approach of cluster analysis (involving mineralo-chemical data) for several volcanic rocks was taken up by Tibaldi (1995), Mazzarini (2004), Mazzarini et al. (2010). In fact, cluster analysis refers to a simple process of grouping a set of objects in such a way so that the objects in the same group known as cluster, are more similar in some sense to each other than those in other groups (clusters) (Le Maitre 1982). A good resultant of cluster analysis will give rise to a number of clusters where within each such cluster some observations remain as similar as possible. Cluster analysis does not differentiate between dependent and independent variables and after cluster analysis, the number of observations or cases becomes consolidated (or amalgamated) into smaller set of clusters. The basic flow of cluster analysis involves decisive algorithm where all the data are initially considered as a single group, and subsequently, it is subdivided according to some specified parameter (or criterion) until each group contain only one object. This method of clustering involves construction of several dendrograms and may be successfully employed to understand magma source (Valentine and Perry 2007; Germa et al. 2013; Tadini et al. 2014; Cortes et al. 2015). In a multiphased lava system and associated dykes, cluster analysis may reveal useful information on feeder dyke orientation (Tibaldi 1995; Korme et al. 1997; Corazzato and Tibaldi 2006; Paulsen and Wilson 2010; Bonali et al. 2011). In Deccan basalts, use of statistical analyses has been gainfully used to build up a cogent petrogenetic model in recent years by Shrivastava et al. (2014).

In the present contribution, an attempt has been made to employ cluster analysis of a huge number of available mineral chemistry data (Dey et al. 2021) of parts of Eastern Deccan Volcanic Province (EDVP) near Khandwa (21°49′N, 76°21′E), Madhya Pradesh, India. The outcome of such cluster analysis will be interesting in the sense that it will help to understand the inherent magmatic processes of specific parts of Deccan Volcanic Provinces (DVP) (having silica saturated/oversaturated characters) based on multivariate analysis or ‘grouping’ method. Moreover, the outcome of cluster analysis for basaltic rocks of Khandwa will address almost all of its petrogenetic tenets except the remaining portion that involves elucidation of whole rock geochemical model.

2 Regional geology

The Deccan Volcanic Province (DVP) has profound importance in the global context because of its large spatial distribution, huge volume and specific eruption-duration, which broadly corresponds to K–T boundary (Duncan and Pyle 1988; Baksi 1994; Hull et al. 2020). Mantle melting processes, development of lava flows and plumbing systems of dyke-sills offer great scope for understanding petrogenetic processes in continental flood basalt provinces (Coffin and Eldholm 1994; Self et al. 1997; Eldholm and Coffin 2000; Ernst and Buchan 2001; Rajan et al. 2005; Sheth et al. 2009).

Within the vast expanse of DVP, the quantum of studies vary from one sector to another. For example, the southwestern and western parts are geologically well studied (Mahoney et al. 1982; Lightfoot and Hawkesworth 1988; Peng et al. 1994; Bondre et al. 2006) and the development of lavas in the western and southwestern sector has been relegated to Satpura Narmada Tapi rift system as suitable channel ways (Crookshank 1936; West 1958; Baksi 1994; Sen and Cohen 1994; Bhattacharjee et al. 1996; Kumar and Shrivastava 2009; Kashyap et al. 2010; Shrivastava et al. 2014, 2017; Pathak et al. 2017). The Eastern Deccan Volcanic Province (where the present study area belongs to) is a relatively lesser attended section where geological studies have been carried out in sporadic sectors at different point of time (Crookshank 1936; Alexander and Paul 1977; Deshmukh et al. 1996; De 1996; Nair et al. 1996; Yedekar et al. 1996; Peng et al. 1998; Pattanayak and Shrivastava 1999; Ahamad and Kumar 2008). However, during the last 20 years, for EDVP, a plethora of information involving petrogenetic aspects, mineral chemistry, existence of liquid immiscibility and lava morphology has been generated (Mahoney et al. 2000; Sengupta and Ray 2007; Duraiswami et al. 2008; Kashyap et al. 2010; Sengupta and Ray 2011a, b; Ganguly et al. 2014; Srinivas et al. 2019; Dey et al. 2020, 2021; Kale et al. 2020). However, a well-demarcated ‘chemostratigraphy’ (as existing in Western Deccan Volcanic Province; Cox and Hawkesworth 1985; Beane et al. 1986) in EVDP is yet lacking. Shrivastava et al. (2015) presented 40Ar–39Ar age data from basaltic lavas of the Mandla lobe (64.21 ± 0.33 Ma) occurring at the eastern margin of Deccan Volcanic Province. According to Shrivastava et al. (2015), the lavas of Mandla lobe appears to be younger than the main Deccan volcanic activity of the Western Ghats (~67–65 Ma).

3 Present study

The present investigation was carried out around Khandwa (21°49′N, 76°21′E), which forms a part of Eastern Deccan Volcanic Province, lying at the Nimur district, Madhya Pradesh, India (figure 1a). For the preparation of geological map of the area (figure 1b), three-tier classification of De (1974) was used. The physical volcanological characters of this study area were earlier given by Dey et al. (2020, 2021).

Figure 1
figure 1

(a) Overall distribution of Deccan Volcanic Province (stippled) in India. Green areas mark the extent of Eastern Deccan Volcanic Province (EVDP) after Kashyap et al. (2010). Solid circle indicates present study area. (b) Geological map of the present study area (prepared by present authors). Locations of study samples (e.g., PD 48, PD 41, etc.) have been shown in this map. Star in the inset map denotes the study area.

Full size image

3.1 Mode of occurrence

The study area reveals characteristic development of three flows (flow I, flow II and flow III) with different degrees of prominence and development of chilled facies for flow I and flow II. Coarse-grained gabbroic cumulate rock (apparent basement to the lava flows) has also been noted. Characteristic presence of feeder dyke (FD) and chilled dyke (CD) was also recorded in this area. The feeder dyke is found to be associated with flow II and maintains a strike direction of 75°–78° being located at 327 m height. The chilled dyke, on the other hand, is also associated with flow II and runs with almost east-west strike at an elevation of 309 m. In the field, the chilled dyke gives a cross-cutting relation with flow II. The sample locations (corresponding to several structural zones belonging to different flows) have been shown in figure 1(b) for easy understanding for the readers. Moreover, presence of two small but prominent marker horizons is found. These are intertrappean beds between flow I and flow II and between flow II and flow III (figure 1b). A geological map has been prepared based on the identification of several characteristic zones within each lava flow, which are Lower Vesicular Zone (LVZ) grading upward to Lower Colonnade Zone (LCZ), the Entablature Zone (EZ) and the Upper Colonnade Zone (UCZ) grading upward to Upper Vesicular Zone (UVZ) (De 1974) (figure 1b). In majority of the places, different zones within the lava flows show good exposure, but in some cases, their presence can be studied in subcrop as well. The lava flows are by and large fresh and unaltered but in some cases, they show development of spheroidal weathering and highly altered vesicular zone. Representative photographs highlighting prominent field features have been given in figure 2(a–f).

Figure 2
figure 2

Field photographs showing (a) Development of UCZ and EZ in flow III near Nagchun. (b) Prominent EZ in flow III near Nagchun Village. (c) Upper vesicular zone filled in with abundantly developed zeolite amygdules in flow II. (d) Prominent Feeder dyke associated with flow II. (e) Laterally extensive chilled dyke found to be closely associated with flow II near Panjra area. (f) Coarse-grained gabbroic apparent basement in Chamathi village.

Full size image

3.2 Petrography

For all three lava flows, clinopyroxene, plagioclase (and rarely olivine) are the main phenocrystal phases. The groundmasses are represented by smaller-sized clinopyroxene, tiny plagioclase, opaque mineral, microlites and glass (often devitrified). The proportions of individual minerals are not constant; on the contrary, their presence is zone-specific (for example, glass is more dominant in vesicular zones). Ophitic, subophitic, porphyritic, glomeroporphyritic, intersertal and vitrophyric textures are commonly displayed in these lava flows. For both feeder and chilled dykes, plagioclase and clinopyroxene occur both as phenocryst and groundmass. Textures for dyke rocks are similar to that of the lava flows. The apparent basement rock (coarse-grained gabbro) consists of relatively coarse grains of subhedral plagioclase, prismatic clinopyroxene and some opaque minerals, which correspond to cumulus minerals. The intercumulus (relatively fine-grained) minerals are represented by smaller plagioclase and clinopyroxene. Overall this gabbroic basement represents cumulus and intercumulus textural patterns. Representative photomicrographs of different units in the study area have been shown in figure 3(a–f).

Figure 3
figure 3

Photomicrographs showing (a) Large zoned phenocryst of plagioclase (set in a finer groundmass) in feeder dyke. Relatively fine-grained plagioclase grains are also noticed in the groundmass. (b) Dispersed elongated plagioclase phenocrysts in chilled groundmass material (flow I chilled zone). Few clinopyroxene (micro) phenocrysts are noticed. (c) Phenocrysts of relatively coarse clinopyroxene and plagioclase showing subophitic relation. These phenocrysts are set in chilled groundmass (flow I chilled zone). (d) Large strongly zoned plagioclase phenocrysts embedded in cryptocrystalline and glassy groundmass (flow I chilled zone). (e) Number of plagioclase phenocrysts are clustered together (and set in a finer groundmass) to give rise to glomeroporphyritic texture (flow II chilled zone). (f) Cumulus and intercumulus textural pattern developed in coarse-grained gabbroic rock (apparent basement). The cumulus (C) portion is mostly marked by coarse plagioclase, while intercumulus (IC) portion is represented by smaller clinopyroxene and plagioclase.

Full size image

3.3 Mineral chemistry

Electron microprobe analysis (EPMA) of constituent mineral phases of the three lava flows, dyke rocks and apparent cumulate gabbro basement were performed at the Department of Earth Sciences, IIT Bombay, India, using CAMECA SX-Five Electron Probe MicroAnalyser (EPMA). The analytical conditions for the instrument were: acceleration voltage of 15 kV, beam current of 20 nA and beam diameter of 1 μm. Both natural and synthetic standards were used for calibration of the elements and data correction was done by X-PHI method.

It needs to be mentioned that different constituent minerals, namely, olivine, clinopyroxene, and plagioclase are not developed in different flows (or zones) in equal prominence. The flow/zone/dyke/or cumulate rock-wise distributions of different minerals of the investigated area are given in table 1. EPMA of olivine, clinopyroxene and plagioclase (phenocrystal data) has been furnished in Supplementary tables S1–S3, respectively.

Table 1 Flow/zone-wise occurrences of different constituent phenocrystal minerals in the study area.
Full size table

4 Selection and significance of suitable major elements

Correlation coefficients of several major oxides are often determined to understand their nature of significance to elucidate petrogenetic history of a given unit of rock (Saha et al. 1974; Hazra et al. 2010). The routine task is to find out the significance levels of obtained correlation coefficient values in respect of several major oxides bivariate diagrams for mineral phases. ‘Highly significant’ or ‘significant’ major oxides are next shortlisted (and used) in the present study because such highly significant/significant correlation values are considered to be helpful to elucidate petrogenetic history (Saha et al. 1974). Supplementary table S4 gives such shortlisted ‘highly significant’ (or ‘significant’/occasionally non-significant) correlation coefficient for different minerals. Based on these significance levels of ‘r’ (Snedecor and Cochran 1967), for olivine, two important oxides MgO and FeO have been selected. For plagioclase, four variables, namely, SiO2, Al2O3, Na2O and CaO were chosen. Following six variables namely, SiO2, TiO2, Al2O3, FeO, MgO and CaO for clinopyroxene were found to be important.

4.1 Multivariate statistical analysis and dendrogram construction

In our present study, major element mineral chemical data were subjected to multivariate statistical analysis, which involve hierarchical division, cluster analysis and construction of dendrogram using PAST software. Details of PAST software used in the present study is given in Supplementary materials.

5 Dendrograms and mineral-wise pattern identification

In the present study, hierarchal clustering method with agglomerative algorithm was employed for major oxides of clinopyroxene, plagioclase and olivine. Standardized Euclidean distances between samples were computed following Le Matrie (1982). In all cases, the horizontal (base level) axis denotes the point numbers of analyses of different relevant minerals (see Supplementary tables S1–S3), while the vertical axis refers to Euclidian distance. This distance, in fact, represents progressive magmatic differentiation and consequent fall of symmetry patterns with respect to the initial ones.

5.1 Olivine

Olivine phenocryst is recorded only from flow II and that too in small amount. The dendrogram pattern of olivine shows 10 initial data points with gradual amalgamation giving rise to smaller number of clusters. At a moved distance of 3 (out of 9), the initial 10 clusters reduced down to two, and this corresponds to about 33% fractional crystallization of parent magma. In fact, interrelation between magmatic differentiation and clusters (distance) was depicted earlier by Owen (1989). At a moved distance of 8.5 (out of 9), the number of clusters reduces to two and this corresponds to 94% fractional crystallization. The extreme right-hand side (RHS) dendrogram pattern (figure 4) shows no change in the clustering pattern from the beginning to the end of crystallization. This undisturbed pattern (22/1) possibly represents xenocryst of highly differentiated olivine grain inherited from any precursor rock.

Figure 4
figure 4

Dendrogram pattern of olivine found in flow II. The bulk level of crystallization is ~33%. The vertical axis represents Euclidian distance (~ degree of differentiation). Numerals in the horizontal axis indicate data points of olivine analyses (Supplementary table S1).

Full size image

5.2 Clinopyroxene

Clinopyroxene phenocrystal data are available for flow I, flow II, flow III, feeder dyke, chilled dyke and the (apparent) gabbroic basement rock. For the clinopyroxene of flow I, in general, there are numerous clustering (figure 5a) at the initiation of magmatic crystallization. From the huge number of initial clusters, the pattern reduces drastically at a moved distance of 3.5 (out of 12) which corresponds to 29% fractional crystallization. The differentiation was totally completed at ~92% (at a moved distance of 11 out of 12). The two clusters lying at the extreme left have high FeO content and for the major part of the fractional crystallization, they remained undisturbed. This corresponds to the presence of local ferrohypersthene crystal.

Figure 5
figure 5

(a–f) Dendrogram patterns for clinopyroxene corresponding to different flows/feeder dyke/chilled dyke and gabbroic apparent basement. In each case, vertical axis refers to Euclidian distance (~ degree of differentiation). Numerals in the horizontal axis indicate data points of clinopyroxene analyses (Supplementary table S2).

Full size image

Clusters of clinopyroxene for flow II (figure 5b) on the other hand, show an almost equally evolved pattern and it appears that the major portion of the crystallization was completed at ~50% (10 moved distance out of 20). The differentiation was finalized at almost ~95% (19/20×100).

Patterns of the clinopyroxene from flow III show numerous initial clusters (figure 5c). But depending upon their later amalgamation, the cluster pattern can be subdivided into right-hand side (RHS) cluster and left-hand side (LHS) cluster. For the RHS cluster, the bulk of differentiation was complete at a moved distance of 2 (out of 10.5), which corresponds to nearly 20% of fractional crystallization. For the LHS clusters, the ultimate evolution continued up to a more advanced stage of differentiation at a moved distance of 7.3 (out of 10.5) which suggests about 70% fractional crystallization. At the end, the differentiation reached almost 95% (9.5 moved distance out of 10.5), which represents the (extreme) final stage of crystallization.

The cluster pattern for feeder dyke (figure 5d) starts with similar geometry; however, most of the differentiation was completed at ~45% (moved distance 9 out of 20). The differentiation progressed and finally ceased at ~89% (moved distance 17.8 within 20).

For the cluster pattern of chilled dyke (figure 5e), only one data point from the beginning was highly differentiated (extremely RHS data point), whereas the other data points showed equal level of heterogeneity (LHS cluster pattern). For this data point (LHS), differentiation in bulk level was almost complete at 25% (2.5 moved distance out of 10). The end part of differentiation is marked at 92% (9.2 moved distance out of 10).

Clinopyroxene from the apparent basement characteristically shows three cluster patterns (figure 5f). The middle dataset suggests cluster pattern that corresponds to effective cumulate portion having relatively higher MgO. The LHS clusters show an apparent grouping having relatively higher FeO, which has been caused due to differentiation of the parent magma from the beginning. It corresponds to crystallization from the left-over (residual) evolved intercumulus liquid. The extreme RHS data points suggest its most primitive nature having the highest MgO and relatively higher FeO, possibly signifying olivine relict. This might have reacted later with the ambient liquid and finally disappeared from the magmatic scenario.

5.3 Plagioclase

Plagioclase of flow I (Chilled Zone) shows two broad clusters; which may be divided into extreme RHS cluster and middle cluster (figure 6a). For these two clusters, plagioclase composition becomes broadly identical at about 20% fractional crystallization (6 moved distance out of 30). Finally, plagioclase data points corresponding to these two clusters converge where crystallization ended at almost 80% fractional crystallization (moved distance 24.5 out of 30). One data point in the plagioclase dataset of flow I shows no evidence of clustering and remains undisturbed from the beginning. This possibly suggests an accidental potash feldspar inclusion within plagioclase where the inclusion is marked by lower CaO (wt.%), lower Na2O (wt.%) and relatively higher K2O (wt.%).

Figure 6
figure 6

(a–f) Dendrogram patterns for plagioclase corresponding to different flows/feeder dyke/chilled dyke and gabbroic apparent basement. In each case, vertical axis refers to Euclidian distance (~ degree of differentiation). Numerals in the horizontal axis indicate data points of plagioclase analyses (Supplementary table S3).

Full size image

In the case of flow III (figure 6b), two data points form a single cluster (extreme LHS) having very low CaO and low Na2O contents (CaO: 8.2 wt.%, Na2O: 6.9 wt.% and CaO: 6.8 wt.%, Na2O: 6.1 wt.%, respectively). This suggests presence of local andesine fragments within highly calcic plagioclase. Besides this, there are two clusters present in this diagram (figure 6a) which may be termed as the middle cluster and the RHS cluster. For both of these clusters, plagioclase composition attained unique (identical) chemistry. The bulk of the differentiation is complete when ~38% crystallization took place (4 moved distance out of 10.5). The differentiation finally ended at 97% crystallization (10.2 moved distance out of 10.5).

The dataset corresponding to flow II (figure 6c) shows strong initial heterogeneity giving rise to numerous initial clusters. However, despite having initial heterogeneity, the differentiation was completed at ~25% (10 moved distance out of 40). Final (extreme) point of differentiation was achieved at 90% fractional crystallization (36 moved distance out of 40). It is to be noted that one data point (46/1.77) remains undisturbed from the beginning to the end of crystallization. Highest K2O of the data point possibly represents a potash feldspar inclusion.

The plagioclase dataset corresponding to feeder dyke is shown in figure 6(d). One data point shows alien character (extreme RHS) and it does not show any trend of differentiation. In all probability, this data point represents xenocrystic plagioclase inherited from a more primitive source. The remaining two clusters (figure 6d), namely the middle cluster and the LHS cluster, show effect of plagioclase differentiation. The differentiation has been significantly advanced at ~44% (1.75 moved distance out of 4). Finally, the plagioclase crystallization reached its acme at 80% (3.2 moved distance out of 4). Another (almost undifferentiated) cluster (close to data point 3/1) corresponds to extremely differentiated plagioclase having the highest Na2O and lowest CaO contents.

Plagioclase datasets from chilled dyke form two clusters (LHS cluster and RHS cluster) (figure 6e). The plagioclase composition attained its maximum differentiation at ~87%, corresponding to both the clusters (3.5 moved distance out of 4). The bulk level of crystallization was, however, noted to be 38% (1.52 moved distance out of 4).

Plagioclase datasets for the apparent gabbroic basement (figure 6f) have two parts: the RHS cluster represents intercumulus portion and denotes relatively restricted range of composition. On the other hand, the LHS cluster shows a variable composition. For this, the most evolved plagioclase composition attained ~60% fractional crystallization of the parent magma (3.70 moved distance out of 6).

6 Discussion

It is apparent that multivariate statistical analysis (and construction of dendrograms and mineral-wise pattern identification) involving mineral chemical data for the present study indicates mineral-specific cluster patterns. The primitive (early formed) olivine is quite restricted in occurrence and has been detected only in flow II with a less modal proportion. It may happen that olivine-impoverishment may be related to subsequent destruction or decadence of early formed olivine by reaction with the ambient magma (Jennings et al. 2019). For all the three phenocrystal phases (namely, olivine, clinopyroxene and plagioclase), in some of the lava units or dyke system, there are some alien data points in the dendrogram patterns. These alien data points do not show any further segmentation during the crystallization process and can be interpreted as representing some xenocrystic mineral composition or some stray inclusion. Apart from this alienation, the other majority of the datasets show progressive amalgamation of dendrogram clusters and decrease in symmetry patterns. For the convenience of understanding, the degree of bulk crystallization pattern as well as terminal stage of crystallization for olivine, clinopyroxene and plagioclase have been given in table 2 and figure 7(a–c). Corresponding thermometric values have also been shown in figure 7(a–c) to decipher the interrelation between degree of crystallization and thermal ambience. Figure 7(a) shows that the modally deficient olivine could complete its bulk crystallization at almost ~33% fractional crystallization; although there is an indication from dendrogram pattern (figure 4) that the olivine crystallization might have continued up to 90% differentiation. This may be discounted considering modal decadence of olivine in the present case. Figure 7(b) shows that for the bulk of the crystallization pattern (for clinopyroxene), there is a positive correlation between the relative degrees of crystallization and deduced thermometric values. For terminal crystallization behaviour of clinopyroxene, however, there is not much change and it is varying from ~90 to ~95%. The patterns of clinopyroxene crystallization for the feeder dyke (FD) and chilled dyke (CD) are distinctly different. For the FD, the crystallization continued up to ~50%, whereas for the CD, the crystallization pattern could proceed up to ~30%, where the chance of advanced crystallization was too low for quick chilling.

Table 2 Degree of bulk level and ultimate crystallization (in %) as revealed from dendrogram analyses.
Full size table
Figure 7
figure 7

Plot of bulk level of crystallization and ultimate crystallization for (a) olivine (Ol), (b) clinopyroxene (Cpx) and (c) plagioclase (Pl). For both figure 7(b and c), percentage of bulk level of crystallization increases from flow I to III. In case of flow I, flow II and flow III, the deduced thermometry (Putirka 2008) has been denoted. Pattern of crystallization for feeder dyke and chilled dyke has been marked with a different field. Again for figure 7(b and c), it may be noted that fields of maximum extent of crystallization (ultimate crystallization) for lava flows have been shown by separate fields. The ultimate extent of crystallization represents theoretical possibility up to which crystallization may continue. In reality, in all probability, crystallization reached its acme only at its bulk level of crystallization. For both feeder dyke and chilled dyke, only bulk levels of crystallization were the appropriate cases and shown accordingly. For figure 7(b and c), the pink (almost horizontal) elongated field represents ultimate crystallization for lavas, whereas straight line indicates their bulk level of crystallization; the green lensoidal field denotes bulk crystallization of dyke (feeder dyke and chilled dyke). The deduced thermometric values have been incorporated from Dey et al. (2021). For figure 7(a–c), FI, FII and FIII represents lava flow I, lava flow II and lava flow III, respectively; FD: feeder dyke, CD: chilled dyke (for details see text).

Full size image

Plagioclase crystallization (table 2 and figure 7c) also shows a similar pattern as that of clinopyroxene. In figure 7(c), the terminal stage of crystallization for plagioclase shows a spectrum of ~80 to ~97% crystallization. The pattern for plagioclase crystallization in feeder dyke and chilled dyke is drastically different from that of the lava flows. Both the dykes show moderate to low degree of crystallization (~35 to ~40%) because of their early chilling behaviour. However, for the feeder dyke, it appears to be active and capable of supplying magma even at a later period of magmatic history.

Apparent basement in the present study is represented by coarse gabbroic rocks. Textural studies further show that these rocks have two distinct parts: (i) the cumulate part corresponding to crystal settling and (ii) intercumulus part that corresponds to the derived or residual liquid. For clinopyroxene datasets, the intercumulus dendrogram pattern is relatively homogeneous, which suggests more restricted composition in the derived liquid. The clinopyroxene cumulus dataset show a relatively greater inhomogeneity which indicates a slight diversity among the settled cumulus clinopyroxene compositions. From the plagioclase datasets, similar restricted nature of intercumulus crystals has been suggested. Plagioclase cumulus crystals also show relative compositional diversity similar to that noted from clinopyroxene dendrogram pattern. The compositional difference between the intercumulus liquid and cumulus crystals can be ascribed to rising magma and sudden transfer below solidus.

This study based on multivariate analysis, therefore effectively helps to identify the cluster patterns for different rock-forming minerals within a portion of Eastern Deccan Volcanic Province. The distinct nature of hierarchical cluster patterns involving different minerals, for several units of the study area can be well discriminated in terms of their bulk and terminal crystallization history.

7 Conclusion

Based on the above study, the following conclusions can be drawn.

  • For available mineral chemistry data, some highly significant major oxide variations could be categorically identified.

  • Dendrogram pattern analyses indicate that at the outset of magmatic crystallization, the mineralogical parameters display quite heterogeneous (greater symmetry) clusters (though different minerals have their discrete liquidus temperatures).

  • For the lava flows, both clinopyroxene and plagioclase suggest bulk level of crystallization and terminal crystallization. The bulk level of crystallization varies from ~30 to ~60%, whereas termination of crystallization for the lava flows appears to be quite high (~80 to ~97%). The bulk level of crystallization shows a broad control of ambient temperature.

  • For the dyke system [feeder dyke (FD) and chilled dyke (CD)], the bulk level of crystallization is comparable to that of lavas. However, the chilled dyke (CD) suggests early quenching, whereas the crystallization of feeder dyke (FD) was to some extent prolonged possibly because of feeding the lava flows.

  • For the apparent basement rock, the cluster patterns for both clinopyroxene and plagioclase strongly suggest a relative compositional spectrum for the cumulus settled crystals, whereas the intercumulus portion corresponds to a restricted compositional range. The compositional difference between the intercumulus liquid and cumulus crystals can be ascribed to rising magma and sudden transfer below solidus.