See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/364613964

The Taxonomy of Mineral Occurrence Rarity and Endemicity

Article  in  The Canadian Mineralogist · October 2022

DOI: 10.3749/canmin.2200010

CITATIONS READS

0 104

4 authors:

Liubomyr Gavryliv Vitalii Ponomar

Comenius University Bratislava University of Oulu
23 PUBLICATIONS   14 CITATIONS    30 PUBLICATIONS   210 CITATIONS   

SEE PROFILE SEE PROFILE

Marko Bermanec Marián Putiš

Universität Bern Comenius University Bratislava
13 PUBLICATIONS   1 CITATION    180 PUBLICATIONS   2,314 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Data analysis in Mineralogy View project

VEGA 1/0257/13: Accessory minerals under rock-fluid interaction conditions in magmatic and metamorphic systems View project

All content following this page was uploaded by Liubomyr Gavryliv on 26 November 2022.

The user has requested enhancement of the downloaded file.

 THIS IS AN AUTHOR'
S VERSION OF THE PAPER
The final accepted paper is here
1
https://pubs.geoscienceworld.org/canmin/article-abstract/60/5/731/618246/The-Taxonomy-of-Mineral-Oc
currence-Rarity-and?redirectedFrom=fulltext

THE TAXONOMY OF MINERAL OCCURRENCE RARITY AND ENDEMICITY

LIUBOMYR GAVRYLIV§
Comenius University, Faculty of Natural Sciences, Department of Mineralogy, Petrology and Economic Geology, Ilkovičova 6,
841 04, Bratislava, Slovakia

VITALII PONOMAR
University of Oulu, Faculty of Technology, Fiber and Particle Engineering Research Unit, Pentti Kaiteran katu 1, 90 014 Oulu,
Finland

MARKO BERMANEC
University of Bern, Institute of Geological Sciences, Baltzerstrasse 1þ3, CH-3012 Bern, Switzerland

MARIÁN PUTIŠ
Comenius University, Faculty of Natural Sciences, Department of Mineralogy, Petrology and Economic Geology, Ilkovičova 6,
841 04, Bratislava, Slovakia

ABSTRACT

Nearly a half of known IMA-approved minerals (as of November 2021) are reported from four localities or fewer and so
may be considered rare mineral species. These minerals form a continuum with more common species (e.g., rock-forming
minerals), all of which constitute important constituents of Earth and contributors to its dynamics. To better understand the
taxonomy of mineral rarity, evaluations have been made on the basis of k-means clustering and kernel density estimation of
one-dimensional data on mineral occurrence metrics. Results from second- and third-degree polynomial regression analyses
indicate the presence of a divergence between the observed number of endemic minerals discovered since 2000 and those that
are likely to represent ‘‘true’’ endemic species. The symmetry index, calculated using the approach of Urusov for each rarity
cluster, reveals a gradual decrease from ubiquitous to endemic from 0.64 to 0.47. A network analysis of element co-occurrences
within each rarity cluster suggests the existence of at least three different communities having similar geochemical affinities;
the latter may reflect the relative abundance of minerals their elements tend to form. The analysis of element co-occurrence
matrices within each group indicates that crustal abundance is not the only factor controlling the total number of minerals each
element tends to form. Other significant factors include: (1) the geochemical affinity to the principal element in the group (i.e.,
sulfur for chalcophile and oxygen for lithophile elements) and (2) dispersion of the principal element through geochemical
processes. There is a positive correlation between the lithophile element group’s abundance in the Earth’s crust and the number
of common minerals they tend to form, but a negative correlation with the number of rare species.

Keywords: rarity, endemic minerals, data analysis, symmetry index.

INTRODUCTION characteristic associated with a high risk of extinction,

From the biological point of view, rarity may be
considered evidence that a species has characteristics
that differ from a more common taxon (Kunin &
Gaston 1993). In bioecology, rarity is a population
implying that once a biological species becomes
extinct, they will not re-emerge (Vermeij &
Grosberg 2018). In mineral ecology, most of what
are considered endemic minerals are not likely to
disappear, except those falling into category 3 of

§
Corresponding author e-mail address: liubomyr.gavryliv@uniba.sk

Page 1
2
Hazen & Ausubel (2016). Further, metamict minerals
and those formed in the mantle may be considered
exceptions. However, the question of our current
ability to recognize ‘‘mineral fossils’’ should be raised
apart from these minerals. Possibly, the general rarity
of the bulk of endemic minerals will be diminished
through the application of new analytical methods to
identify such minerals and the study of hard-to-reach
locations, including that which have an extraterrestrial
origin. For instance, ferroskutterudite, (Fe,Co)As, was
first reported from dolomite-calcite veins of the
Noril’sk ore field (Russian Federation; Spiridonov et
al. 2007) and was considered endemic until it was
described from a second locality, the Puy-les-Vignes
breccia pipe, Massif Central, France (Staude et al.
2012). It has been subsequently discovered at
Wittichen, Schwarzwald, SW Germany (Harlaux et
al. 2015), and from Pira Inferida Yard, Fenugu Sibiri
Mine (Sardinia, Italy). By contrast, harstigite, Ca6Mn
Be4(SiO4 )2(Si2O7)2(OH)2 , known only from its type
locality at Harstigen Mine, Persberg ore district
(Sweden), has been an endemic mineral for 136 years
(Flink 1886).
Nearly 21% (1232) of all IMA-approved minerals
(5762 as of November 23, 2021) are known from a
single locality, thus making them endemic mineral
species. Another 28% (1616) occur at between 2 and 5
localities, meaning that almost half of all known
minerals are rare according to the systematics of
Hazen & Ausubel (2016), the other half being
considered common or widespread. Those that are
considered as being widespread are typically rock-
forming minerals, which play critical roles in reflect-
ing the Earth’s composition and dynamics. The crust
comprises almost 2.5% of the Earth’s volume, with
fewer than 100 minerals making up 99% of it (Hazen
& Ausubel 2016). Of these, 90% are silicates and 75%
are tectosilicates (Deer et al. 2004). Despite being
widespread in the crust, 22% of the known tectosili-
cates (i.e., 40 species) are recorded at only one locality
on the Earth’s surface, indicating the importance of
discriminating between the rarity taxonomy of those
minerals directly accessible to us and those that are
not.
Another example is bridgmanite, which has been
estimated to constitute up to 93% of the lower mantle
(Tschauner et al. 2014). However, it has been recorded
at only one locality at the surface of the Earth,
probably owing to its instability under ambient
conditions (category 2 according to Hazen &
Ausubel 2016). As such, bridgmanite may be consid-
ered an endemic mineral sensu stricto, but a
widespread mineral sensu lato. Here, mineral rarity
is defined based on the distribution of a mineral within
the available geospatial data, the latter representing a
metric based on the number of localities where
minerals have been recorded close to, or at, the
surface of the Earth.
Hazen & Ausubel (2016) suggested a prototype of
a taxonomy of mineralogical rarity based primarily on
the taxonomy of biological rarity. According to this, a
mineral may be considered rare if it is known from five
or fewer localities. Furthermore, it was proposed that
rare species generally conform to one of four distinct
categories (with slight overlaps) based on the reason
for their rarity:: (1) a restricted P-T-X range, (2) rare
combinations of essential elements, (3) instability
under prolonged exposure to ambient conditions, and
(4) difficulties in the recognition of the mineral or
challenges associated with its discovery.
As large-number-of-rare-event (LNRE) models
depend primarily on the distribution of rarer species
(Hazen et al. 2016, 2015a, Hystad et al. 2019b, Grew
et al. 2016, 2017, Liu et al. 2018, Hazen et al. 2017,
Morrison et al. 2017), it is noted that rare minerals are
vital to developing a more complete understanding of
overall mineral diversity and in predicting the total
number of mineral species yet-to-be discovered. Based
on predictive population models, the total number of
‘missing’ minerals in several different subsets has
been evaluated for a range of elements, including: C
(Morrison et al. 2020, Hazen et al. 2016), Co (Hazen
et al. 2017), Cr (Liu et al. 2017), B (Grew et al. 2016),
Li (Grew et al. 2019), and V (Liu et al. 2018) minerals.
Since the predicted number of undiscovered minerals
is believed to have been underestimated, owing to the
acceleration in the rate of the mineral discovery over
the last decade (Hazen et al. 2015b), Hystad et al.
(2019a) employed Bayesian methods in an effort to
estimate the total number of mineral species that may
exist within the Earth’s crust, obtaining a number of
~3700 mineral species which have not yet been
discovered. Hazen & Ausubel (2016) emphasized that
the precision of mineral ecology studies, which
employ statistical methods, is primarily dominated
by those mineral species found at one or two localities.
Similar to biological ecosystems, where only a few
species are widespread, but most species are rare, it
was posited that the Earth’s minerals might also follow
a similar distribution model.
Furthermore, Hazen et al. (2015a) note that up to
15,000 mineral species could plausibly be discovered
on Earth-like planetary objects. Of course, this
depends on the possibility that these bodies possess
environments with P-T-X conditions that are either not
present, or remain to be discovered on Earth. However,
even if these bodies began as did Earth, it might be
expected that there would be a difference in the
number of rare minerals between those found on them
and those found on Earth. The latter implies that
Page 2

mineral rarity and abundance cannot yet be explained
by a completely deterministic model, at least not
without the introduction of additional parameters
which are not yet known or sufficiently understood.
The issue of mineral rarity was also highlighted
from the crystal symmetry point of view in the works
of Urusov (2002, 2007), where the concept of
‘‘evolutionary desymmetrization of mineral matter’’
was proposed. This underpins the gradual increase in
the percentage of low-symmetry (particularly triclinic)
mineral species as compared to those of higher-
symmetry classes. In this context, Urusov (2007)
mentions several mechanisms by which secondary,
lower-symmetry minerals form at the expense of
higher-symmetry species. The general process here is
one that follows a passive mineral and symmetry
evolutionary model, similar to that as proposed by
Krivovichev et al. (2018): it implies that while new
mineral species of lower symmetry are formed, the
precursor, older mineral species continue to be form
and be produced, i.e., they do not disappear entirely.
This phenomenon has led to a relative increase in the
percentage of rare minerals (mostly lower symmetry)
discovered (often within a single deposit and in very
low abundance) over the last five decades primarily
through the application of high-resolution analytical
techniques. This suggests that a hypothesis of a
‘‘natural selection’’ algorithm may be applied to the
mineral kingdom, under which every mineral tends to
minimize desymmetrization to ensure stability. The
latter is the principle of minimum desymmetrization as
proposed by Urusov (2002). Accordingly large-scale
global processes in the Earth’s crust favor desymmet-
rization in the minerals that develop. However, this
tendency could be reduced in specific environments
with extreme geochemical conditions, precisely those
where most rare species have formed.
Rarity, in the context of the discussion presented, is
considered more as a relative concept than an absolute
one.. Mineral species are considered relatively rare
compared with the abundance of other species, or
based on their sharing unique crystal-chemical prop-
erties and restricted P-T-X stability ranges. Therefore,
the demarcation point wherein a mineral species is
defined as being rare becomes subjective. The fact that
a mineral species has been discovered at many
geographic localities cannot be considered a valid
indicator as to whether a mineral may be considered as
endemic or not and thus is insufficient to completely
address the taxonomy issue. In this context, a new,
alternative point of view on the issue is provided,
including the presentation of a taxonomy proposal
based on the study of mineral rarity as an absolute
measure and deciphering the similarities within the
rarity classes using machine-learning (ML) techniques.
3
The present paper: (1) provides a thorough and
reliable taxonomy of mineral rarity and classification
of IMA-approved minerals into rarity groups based on
data analysis; (2) compares these categories in the light
of their discovery history and composition in order to
reduce the dimensionality, increase interpretability of
the taxonomy, and minimize the information loss; and
(3) assesses the evolution of mineral rarity.
MATERIALS AND METHODS
The basic data used in this study comes from the
list of IMA-approved mineral species, including their
mineral formulae, crystal systems, locality counts,
discovery years, and classification within the Nickel-
Strunz classification scheme. These were appended to
through subsequent data searches and extraction from
other resources. Some of this data is provided by web
resources and some is accessible through hard-copy
publications. The bulk of information on the crystal
chemistry of minerals, their occurrence, associated
minerals, and discovery localities was accessed
through digital versions of The American Mineralo-
gist, The Canadian Mineralogist, and Mineralogical
Magazine journals and reviewed afterward. The IMA
list served as a primary skeleton of information and
was supplemented with data coming from the original
literature.
Materials
The IMA List of Mineral Species. The official list of
IMA-approved mineral species (http://cnmnc.main.jp/
imalist.htm) is also accessible via the RRUFF Project
(https://RRUFF.info/IMA) (Lafuente et al. 2016). The
RRUFF Project IMA list allows users to search over
5740 (as of November 12, 2021) species by mineral
name, composition, crystal system, space group, point
group, unit-cell parameters, origin, paragenetic mode,
IMA status, and even water solubility. Additionally,
the RRUFF resource provides data on crystal structure
groups, oldest known age, Raman spectra, and other
physical properties.
The Mineral Evolution Database (MED). The
Mineral Evolution Database (Golden et al. 2016),
designed primarily for supporting mineral evolution
and ecology studies, is another resource freely
accessible through RRUFF Project (https://RRUFF.
info/Evolution). The MED contains mineral locality
data extracted from mindat.org. As of November 14,
2021, the MED data on 810,907 mineral-locality pairs
is available. Additionally, the database allows retrieval
of subsets of mineral-locality pairs through years
between 2016 and 2020. The latter is extremely
Page 3
4
helpful in assessing the evolution of rarity and
endemicity with time.
Athena. The Athena mineral database was the first
complete mineral database on the web, created during
1986–1987 at the Department of Mineralogy at the
Geneva Natural History Museum. In 1994, the
database was published online; since then, it has
become accessible to the geoscientific community.
The Nickel-Strunz classification used in the current
research was kindly provided by Pierre Perroud, the
founder of Athena (https://athena.unige.ch/athena/),
with his written permission.
Handbook of Mineralogy. Handbook of Mineralogy
(http://www.handbookofmineralogy.org) is another
web resource for accessing data on 4988 IMA-
approved minerals (as of December 2, 2021), main-
tained by the Mineralogical Society of America
(MSA) since 2001. A variety of data is stored under
the website: crystal data, physical properties, optical
properties, unit-cell data, occurrence, association,
distribution, name origins, and references.
Mindat. In rare cases, the number of occurrences
and mineral formulae of some rare minerals was cross-
checked with the online mineral database Mindat
(www.mindat.org) in November 2021. The localities
returned were then screened visually and analyzed
through external resources to decide whether these
localities constitute different geological units or not.
All web resources cited above, other sources used
in research negligibly, and online and hard-copy
publications used to develop the core analytics of the
research, were accessed under ‘‘fair use’’ conditions
governed by The Copyright Act of 1976 — a United
States copyright law (17 USC § 107) and what is
known as ‘‘fair dealing’’ in other countries (Canada,
Australia, UK, EU, and its Member States). All the
data used in this research resides in the public domain
and is freely accessible through web interfaces.
All the code developed during the research is
provided under MIT license in the public GitHub
repository ‘‘mineral-rarity’’, accessible through an
open-source geoscientific computing organization
called ‘‘mineralogy-rocks’’ (https://github.com/
mineralogy-rocks/mineral-rarity).
Methodology
Python 3.10.1 was used along with shared scientific
libraries (NumPy, Pandas, scikit-learn, NetworkX,
matplotlib) to build up the analytics warehouse and
compute the dependencies within the collected data.
Several data analysis algorithms were employed for
parsing text data like mineral formulae, encoding
categorical data into binary (‘‘one-hot encoding’’ or
OHE), and calculating the co-occurrence matrices
within the defined rarity groups. They are: (1) data
cleaning, (2) data transformation, and (3) data parsing
and normalization to at least a second normal form
(2NF) where appropriate. In addition, two unsuper-
vised machine learning (ML) techniques were used for
testing the hypotheses in regard to data distribution:
(1) k-means clustering and kernel density estimation
(KDE) of unlabeled one-dimensional data for data
classification and (2) ordinary least-squares linear
regression models with second- and third- degree
polynomial feature transforms for predicting the
‘‘true’’ rate of mineral discoveries. Furthermore, a
network analysis of element co-occurrences was
employed to encode the interactions between mineral
associations of several rarity groups.
Network community structures were computed and
measured through agglomerative and division com-
munity detection algorithms known as the Girvan-
Newman algorithm and a greedy modularity maximi-
zation based on the Clauset-Newman-Moore algorithm
(Clauset et al. 2004). Greedy modularity maximization
focuses on identifying those communities that produce
the greatest increase in modularity, beginning with
each node in its community. On the other hand, the
Girvan-Newman algorithm detects communities by
steadily removing edges from the original graph. At
each step, the algorithm removes the ‘‘most valuable’’
edge of the network, traditionally this being the edge
with the highest betweenness centrality. The number
of elements co-occurrence was used as the attribute to
measure the weight of each edge.
The work follows PEP-8 rules of Rossum et al.
(2001) as much as possible, and the authors recom-
mend that other developers in geoscientific areas
follow generally accepted code quality guides. A
standardized way of building the data warehouses and
analytics pipelines is not only a suggestion, but rather
a demand dictated by programming ethics. Moreover,
storing the code and history of commits in version-
control systems (e.g., GitHub, GitLab, or Bitbucket)
under open-access conditions to track and to manage
changes to code developed within the scope of
research, is strongly encouraged. The code repository
provided here could be treated as a guide to setting up
the local development environment, analyzing miner-
alogical data using modern tools, and writing descrip-
tive and meaningful commits history.
Key definitions
A few key definitions are provided in this section to
emphasize the importance of understanding the
terminology used in this paper.
Rarity. While rarity is considered as an absolute,
not relative, metric, an idea is put forward that once a
Page 4
 THE TA
species’ rarity status is assigned, it can shift toward a
more common category (case 1) and, in very rare
cases, toward a rarer one (case 2) with time. For case
1, the rarity of a mineral follows a ‘‘normal rarity
flow’’, which is typical behavior of the majority of
species. When a mineral is discovered at n localities, it
may subsequently be discovered at other localities
(nþ1), thus shifting its rarity index toward that of a
more common mineral. However, in exceptional cases,
a mineral recorded at n localities may have been
erroneously reported from one locality, meaning that
the absolute number of localities where the mineral is
recorded from is n–1. For instance, in 2011, jacutin-
gaite was erroneously reported from Libštát copper
mines (Košťálov, Semily District, Liberec Region,
Czech Republic), due to a mistake developed during
thin-section polishing (Malec et al. 2012).
Another example is samarskite-(Yb), an endemic
mineral that was erroneously reported as occurring at a
secondary locality in the Pryor Mountains (Big Horn
Co., Montana, USA). Upon detailed examination, it
was demonstrated that the chemical composition did
not in fact match that of samarskite-(Yb). As such, the
mineral rarity index would necessarily shift toward
that of a rarer one, i.e., following an ‘‘inverse rarity
flow’’.
Locality. Particular attention should be drawn to
what defines a mineral locality, especially when one
has not yet been standardized. In some instances, this
term can have a minimal meaning, such as a locality of
a mineral found in a thin section from a sample
collected at a set of exact coordinates. Other times, this
term has a much broader and less defined meaning, for
example, an outcrop, a mine, or a geologic formation.
The latter introduces bias in the data set, which prefers
micro localities, these often corresponding to different
outcrops of the same rock formation or to a different
level within an underground mine.
The data modeling presented in this paper is
dependent on the correct number of the localities for
a given mineral species and other data used in
conjunction with the locality counts. While the
localities used in this research constitute a data point
(geographical location of mineral occurrence), it also
introduces an immeasurable bias into the results for the
reasons mentioned above. Additional caveats include:
(1) the number of occurrences does not necessarily
reflect the abundance of a given mineral in the Earth’s
crust; (2) the abundance of a single phase at the
locality is not considered, and (3) frequently, a few
localities represent different portions of a given
quarry, open pit, or mine, all related to the same ore
body, geological formation, or stratigraphic unit. In the
first case, when considering two different rare species
recorded at fewer than five localities, they are very
c.
5
similar in terms of rarity. For instance, charoite is
recorded from only three localities, and abramovite
occurs only at Kudriavy volcano, Kuril Islands,
Russian Federation. Both are considered rare species
and are therefore similar from a taxonomical point of
view.
However, charoite’s abundance at the recorded
localities is so high that the name ‘‘charoitite’’ has been
applied to those rocks consisting of more than 50% of
charoite, along with variable proportions of quartz,
aegirine, and K-feldspar. In contrast, abramovite
occurs as tiny elongated lamellar-shaped crystals, up
to 1 3 0.2 mm in size, that are barely visible to the
naked eye. Therefore, while charoite and abramovite
are similar in terms of the locality-counts metric, they
are different in terms of their relative abundances in
the upper crust. A rarity taxonomy must consider the
overall abundance of a mineral in addition to its spatial
distribution to be as robust and complete as is possible.
However, an abundance parameter could not be
introduced into the research due to a lack of data
relating to the relative size or morphology of all the
minerals considered.. Of course, these discrepancies
impact more greatly on relatively rare mineral species
than those that are widespread.
Discovery year. In order to assess the rate at which
minerals have been discovered, which can then be
employed to develop regression models, the year in
which a mineral was first discovered was considered as
a parameter. Importantly, the discovery year is defined
as the year of the first publication or report about the
mineral, not the IMA approval or redefinition year.
Mineral abundance. When comparing the abun-
dance of elements in the Earth’s crust to the number of
minerals where a given element is present in essential
quantities, a simple rule was followed: if a chemical
element is present in a stoichiometric formula, it is
considered an essential constituent. This metric is
called a ‘‘mineral clarke’’ (Krivovichev et al. 2018), a
proportion of minerals containing a specific element
relative to all those minerals present in the data set.
The Krivovichev et al. (2018) approach was followed,
and this parameter was calculated as a proportion of
minerals containing the element but only within the
members of a specific rarity group, called ‘‘mineral
abundance’’ for clarity (i.e., a mineral abundance of Si
means a proportion of minerals where Si is present in
all minerals within the rarity group under consider-
ation). Technically, the mineral abundance of each
element was calculated using a regular expression
extraction of unique elements from every IMA-
approved mineral formula. This data, grouped by
mineral species, was used to calculate co-occurrence
matrices of elements within each rarity group and
Page 5
6
FIG. 1. Mineral-locality pairs using: (a) raw data and (b) log-transform (base 10).
compare their crustal abundance to the observed
distribution within minerals.
Data clustering and classification
The raw localities data is hard to visualize and
interpret because of its complexity and a vast spread of
data (Fig. 1)—some minerals are recorded at only one
locality (gaotaiite) while others are recorded at
.60,000 localities (quartz). A common technique to
delineate raw one-dimensional data into categories is
to calculate the frequency distribution over the 25%,
50%, and 75% quantiles. In the case of mineral-
locality pairs, 25% corresponds to one locality, 50% to
4, and 75% to 18 localities. The latter appears
reasonable enough and suggests a preliminary skeleton
for the taxonomy as follows:
(1) minerals occurring at 1–4 localities make up 46%
of the data volume. All of these species are rare,
while those found at exactly one locality are
endemic;
(2) minerals found at 5–18 localities constitute nearly
26% of data volume and could be regarded as
transitional between rare and common minerals;
(3) minerals recorded at .18 localities make up 28%
of data volume and are relatively common.
Based on the preliminary examination of the
quartiles, a set of clustering techniques of unlabeled
data was employed to separate samples into n groups
of equal variance, to identify similar rarity groups
. Page 6
within this type of data, and to further interpret the
results in light of the classification obtained.
The mineral versus locality counts pairs were used
as entry data for the analysis, since the goal during this
step was to find the similarities within localities data
rather than test hypotheses. Data preparation involved
data cleaning and transformation, including feature
scaling—a common requirement for many machine
learning estimators, so that individual features look
like standard normally distributed data (e.g., Gaussian
with zero mean and unit variance).
Two ways to place the data attributes on the same
scale were considered: so-called min-max scaling
(normalization) and standardization. Considering that
mineral-locality count pairs are left-skewed (owing to
a large group of minerals occurring at only a few
localities), log-transformation (base 10), furthered by
standardization, was employed to achieve the unit
variance and to reduce the impact of outliers on the
clustering. The standardization component included
removing the mean and scaling to unit-variance.
K-means is an unsupervised machine learning
technique commonly used to organize a sizable
multivariate dataset into classes and recognizes typical
characteristics. For the k-means algorithm, often
referred to as Lloyd’s algorithm (Lloyd 1982), there
are three principal hyperparameters to set up to define
the model’s configuration: (1) initial values of clusters,
(2) distance measures, and (3) the number of clusters
(k value). The k-meansþþ initialization was used to
select initial cluster centers for clustering, with 10 as

FIG. 2. Visual representation of the Elbow method: within-
cluster sum of squares (WCSS) and the number of
clusters.
the number of times the k-means algorithm will be run
with different centroid seeds, and 300 as the maximum
number of iterations of the k-means algorithm for a
single run. In addition, the k-meansþþ allows the
initialization of the centroids distant from each other.
The number of clusters (k) is the most critical
hyperparameter in k-means clustering. In order to
calculate the optimal k, both elbow and silhouette
(Rousseeuw 1987) methods were used.
Several k-means were run, incremented k with each
iteration, and recorded the sum of the squared
distances (within-cluster sum of the square) with each
run to perform the empirical elbow method (Fig. 2).
The ‘‘elbow point’’ is where the within-cluster sum of
square (WCSS) curve starts to bend. The x-value of
this point for the used dataset is between k ¼ 4 and k ¼
7, meaning that the optimal number of clusters would
be within this range.
The silhouette coefficient is another method to find
the optimal number of clusters and measures cluster
cohesion and separation. The technique allows the
identification of how well each object has been
classified based on two factors: (1) how similar the
data object is to other objects in its cluster (cohesion)
and (2) how different the data object is from objects in
other nearest cluster (separation).
Despite all of the clusters being above the average
silhouette scores, the silhouette analysis shows that the
cluster values of 2–4 are a poor choice for the given
data, due to wide fluctuations in the size of the clusters
that result. Silhouette analysis is more ambivalent in
deciding between 5 and 7. When the cluster number is
equal to 5, all the clusters are more or less of similar
thickness. For this study, and to better differentiate
minerals occurring at precisely one locality, a k value
¼ 5 was selected for the initial number of clusters used
7
FIG. 3. KDE plots calculated using the Epanechnikov (1969)
method on one-dimensional standardized and scaled
locality counts with various bandwidths.
in the analysis. Results coming from this choice
suggest the existence of the following clusters within
the data:
(1) highly dense cluster with minerals occurring at 1
locality;
(2) 2–4 localities;
(3) 5–16 localities;
(4) 17–70 localities;
(5) minerals occurring at .70 localities.
Another powerful method to identify intervals
within a one-dimensional array of data is KDE, often
referred to as the Parzen-Rosenblatt window method
(Rosenblatt 1956, Parzen 1962), a non-parametric way
to estimate the probability density function of a
random variable. The KDE method is also regarded
as an unsupervised clustering approach and is
instrumental in estimating the class-conditional mar-
ginal densities of data. During the study, Gaussian,
Epanechnikov, and tophat (rectangular) methods were
used to calculate KDE over various bandwidth
parameters from 0.1 to 0.9 at 0.05 step. The
Epanechnikov (1969) method revealed the highest
performance and data fit (Fig. 3). The bandwidth
affects how ‘‘smooth’’ the resulting KDE curve is,
controlling the tradeoff between bias and variance in
the result. This approach was taken to determine the
optimal number of clusters by first finding dense
regions in data and focusing on features described by
local maxima (modes) and local minima (anti-modes)
(Bugrien et al. 2014). Accordingly, KDE using the
Epanechnikov method implies the existence of at least
6 margins within the data: 1, 2, 4, 5, and between 12
and 20 localities, depending on the bandwidth
parameter.
Page 7
8
TABLE 1. THE TAXONOMY OF RARITY GROUPS
Rarity Rarity Number of Number of
group subgroup Index localities minerals
Rare Endemic RE 1 1232
Rare RR 2–4 1473
Transitional Rare TR 5–16 1313
Ubiquitous TU 17–70 840
Ubiquitous U .70 701
THE TAXONOMY OF MINERAL RARITY
It must be noted that the categorization based on
the k-means clustering results, the KDE implications,
and quantile assessment broadly intersects, especially
in the range of rarer species, since more than a half of
the data volume is concentrated within the range of 1–
4 localities. Therefore, all the results were taken into
account, and the locality counts data were classified
into the following groups (Table 1), each of which
represents the rarity taxonomy of a mineral species:
1. R: rare group – minerals occurring at ,5 localities
(2705 minerals)
This group is well-described by Hazen & Ausubel
(2016), who defined ‘‘rare’’ minerals as those recorded
from five or fewer localities. While they were the first
to make an in-depth analysis of mineral rarity and to
pose the question in regard to taxonomy, this paper
recommends limiting this range to 1–4, since all
algorithms, especially the KDE, detect an extremum
between 4 and 5 localities. Additionally, the rare group
is further split into two distinct and dense subcatego-
ries:
a) E: endemic subgroup
Once a new mineral is discovered, it is often
reported as endemic unintentionally. For instance, of
65 minerals approved by IMA in 2021, 57 were
reported from only one locality (as of November 11,
2021) and thus constitute endemic species. However,
nitscheite, a new endemic mineral with a uranyl-
sulfate sheet, was first reported from the Green Lizard
Mine (Utah, USA) and subsequently discovered at the
Krunkelbach Valley Uranium deposit (Freiburg Re-
gion, Germany), shortly after the discovery of the
mineral at the type locality. The latter indicates the
importance of finding the criteria to discriminate
between newly discovered species that are ‘‘endemic’’
conventionally (RE) and ‘‘true’’ endemic minerals
(tRE). Accordingly, two more factors should be
considered when assessing ‘‘true endemicity’’: the rate
of mineral discovery and the discovery year. The latter
is covered in the following subsection of the paper.
b) R: rare subgroup
These minerals are not endemic anymore, but are
still very rare, taxonomically.
2. T: transitional group – minerals occurring at 5–70
localities (2153 minerals):
The term ‘‘transitional species’’ designates those
species that are no longer rare but are not yet common
enough at the same time, i.e., they are transitional
between these two categories. Generally, these species
can relatively quickly transform through a ‘‘normal
rarity flow’’, due to their discovery at new locations or
the intensive application of new analytical methods.
For instance, more than 4500 IMA-approved species
were discovered during the period 1950–2021. Once
discovered, almost every mineral was regarded as
endemic or rare. However, some of them were later
discovered at enough additional localities to shift the
rarity status of these species toward a more common
state, into a ‘‘transitional state’’, in this case. The
transitional species are divided into two subcategories,
delineating the type of the transition:
a) R: rare subgroup
The minerals of this subgroup should be considered
rare within a broader extent. However, they are highly
likely to become common and transfer into the next
stage with the continuation of mineral exploration.
b) U: ubiquitous subgroup
In contrast, the minerals of this subgroup are no
longer considered rare. They are not likely to become
rare by ‘‘inverse rarity flow’’, and they tend to become
ubiquitous.
3. U: ubiquitous group – minerals occurring at .70
localities (701 in total).
These may be considered as rock-forming minerals
that are relatively widespread in distribution.
‘‘True’’ endemicity
It is noted that the proportion of observed endemic
minerals has increased steadily since 1800 (Fig. 4),
increasing from 12.5% in 1839 to 87.6% in 2021, with
only slight deviations from the general trend prior to
2011. A steep increase is observed starting from 2012.
The latter could be attributed to a general climb in rate
of mineral discovery during this period. However,
upon closer examination, it can be seen that despite a
similar rate of mineral discovery over the period 2009–
Page 8

FIG. 4. Discovery rate of endemic minerals.
FIG. 5. Polynomial regression model predictions with 95% confidence intervals of endemic mineral discovery rates using 2nd
and 3rd degree polynomial transform.
9
2011, the proportion of endemic minerals was
relatively lower compared to that over the period
2012–2021. Considering the principle of ‘‘normal
rarity flow’’, a hypothesis can be made that a portion
of the observed RE minerals discovered during the last
decade will become RR over the next few years. To
decipher the discovery rate of ‘‘true’’ endemicity and
determine the optimal features required for training,
the ML model was trained, and several linear
regression models were tested, using four different
spans of years: (1) 1900–1990, (2) 1900–2000, (3)
1900–2011, and (4) 1900–2021. The presumption is
that the discovery year and the number of minerals
discovered during a specific year are independent
features during this analysis, while the number of
endemic minerals discovered depends on them. Before
training the models, the nonlinear feature of the input
data was considered, and a second-degree polynomial
feature transformation was applied. After the models
were trained, the results were compared with those
trained with third-degree polynomial transformed
features (Fig. 5).
Page 9
10
TABLE 2. EFFICIENCY OF POLYNOMIAL REGRESSION MODELS WITH DIFFERENT
TRAINING SETS
Model estimates
nd
2 degree polynomial transform
Explained
Training set R2 RMSE variance
1900–1990 0.6444 1.1685 0.6444
1900–2000 0.9432 0.7781 0.9432
1900–2011 0.6889 2.7114 0.6900
1900–2021 0.7527 12.4175 0.7556
In some cases, models trained on third-degree
polynomial transformed features lead to ‘‘overfit-
ting’’, specifically when training is performed using
data over the range of 1900–2021. Additionally,
modeling is somewhat dependent on the set of initial
features that were used. Therefore, the initial data for
each run was split into a training set (90% of data)
and a testing set (10%), which was followed by
features shuffling, model training, running the
predictions, and finally, evaluating the efficiency of
the model (Table 2).
Taking into account the results of model fits and a
general presumption that the number of endemic
minerals observed during the last decade has been
overestimated, the features set for the period 1900–
2000 transformed to a second-degree polynomial was
chosen for estimating the preliminary gap between the
values of ‘‘true’’ endemicity and those observed. The
predicted target values for the period of 2000–2021 are
17–33 for ‘‘true’’ endemic minerals, with ~550 total
endemic species. This number is only approximate, but
the approach and the results obtained generally imply
that: (1) the ‘‘true’’ endemicity rate follows the same
general trend as rate of discovery, but with a tendency
to increase at a rate lower in magnitude, and (2) the
~380 minerals discovered after 2000 are more likely
to be classified as RR rather than RE, taxonomically.
The latter indicates that the total number of tRE
minerals might be lower than that of RE (~850 versus
1232), while the number of RR could be higher
(~1750 versus 1473). However, these observations do
not influence the proposed classifications, since any
deviations only relate to rare group species, specifi-
cally RE and tRE. Taking into account that the ‘‘true’’
discovery rate of endemic mineral species clearly
deviates from the observed rate of discovery, separate
data analytics are provided for all endemic minerals
(RE), along with data for those discovered before 2000
(tRE), where appropriate.
3rd degree polynomial transform
Explained
R2 RMSE variance
0.7519 0.9759 0.7659
0.9360 0.8259 0.9360
0.6805 2.7477 0.6824
0.7940 11.3352 0.7977
COMPOSITION OF RARITY GROUPS
Rare group R(RE,RR,TR)
Over the period 1800–1950, 323 rare (20 RE, 75
RR, and 228 TR) minerals were discovered, and
another 3639 (1165 RE including 276 tRE, 1306 RR,
and 1168 TR) were discovered after 1950. The
majority of rare group minerals are silicates (RE:
317 including 78 tRE, RR: 379, TR: 332), along with
phosphates, arsenates, and vanadates (RE: 286 includ-
ing 72 tRE, RR: 264, TR: 212). The most uncommon
are so-called organic minerals (RE: 24 including 12
tRE; RR: 16, TR: 12) and those containing native
elements (RE: 29 including 5 tRE, RR: 37, TR: 34)
(Fig. 6). Each group of RE, RR, and TR exhibits similar
distributions of those minerals representing different
crystal-chemical groups, thus suggesting a general
homogeneity within values for (RE,RR,TR).
Nearly 70% of RE, RR, and TR mineral species
exhibit low symmetry, i.e., belonging to the triclinic,
monoclinic, and orthorhombic crystal systems, with
only 7.3% of those minerals in the RE and RR groups
being cubic, rising slightly to 8.9% in the TR group
(Fig. 7). As in the case of chemical groups, the
distribution patterns in the RE, RR, and TR groups
follow similar patterns when the proportions of those
minerals with different crystal systems are compared.
Following the approach used by Urusov (2002), the
symmetry index is defined as (HþM)/L, where H is the
percentage of species with the highest symmetry
category and M and L are the percentages of the
intermediate and lowest categories. Applying this, the
symmetry indices vary between 0.47 and 0.48 for RE
and RR groups, respectively, while TR exhibits a
slightly higher value of 0.54. The latter is still below
the index of 0.55 calculated by Urusov (2007) for 3314
rare species.
Taking into account the chemical similarities
among values for R(RE,RR,TR) and the difference in
symmetry indices between R(RE,RR) and TR, further
Page 10

FIG. 6. The share of (a) RE, (b) RR, and (c) TR minerals between Nickel-Strunz classes.
analysis is separately provided for R(RE,RR) and
compared with R(RE,RR,TR), where appropriate.
The rarest of the rarest. The class of the rarest
species (tRE) were further analyzed to determine
which species are even rarer within their respective
taxonomy group when their chemistry is considered
(Table 3). Here a metric is formulated that designates
how often a particular element occurs within the
mineral formula of a given tRE mineral species (e.g.,
mineral abundance), and the data set is filtered by only
those that occur in ,2 minerals. For instance, only one
tRE mineral with Au in its stoichiometric mineral
formula is known to date: penzhinite, (Ag,Cu)4
Au(S,Se)4, which is only known from a single
occurrence (the Sergeevskoye Au-Ag occurrence,
FIG. 7. The share of (a) RE, (b) RR, and (c) TR minerals between crystal systems.
11
Penzhina River, Penzhina Bay, Kamchatka Krai,
Russia). Lepersonnite-(Gd) and mineevite-(Y) are the
only ‘‘true’’ endemic minerals that contain Dy and Gd
in their respective formulae. These minerals were
discovered after 1950, occurring at precisely one type
locality. Both söhngeite and gallobeudantite are
reported from the same type locality—the Tsumeb
mine (Oshikoto Region, Namibia). The relative
abundances of all these species at each of their
localities is extremely low, since, in most cases, they
were identified on only one sample or in one alteration
zone, combined with the fact that their relative sizes
are generally too fine to allow them to be recognized
macroscopically.
Page 11
 TABLE 3. ESSENTIAL DATA ON THE RAREST SPECIES 12

Abundance in
Critical Earth’s crust in ppm Mineral Discovery
Element (Haynes 2016) Name Formula year Relative size Reference
Rh 0.001 rhodplumsite Rh3Pb2S2 1983 as grains, to 40 lm, in a platinum Genkin et al. (1983)
nugget
Ir 0.001 gaotaiite Ir3Te8 1995 as veinlets, 0.05–0. 02 mm wide and Zuxiang (1995)
0.3–1.0 mm long and as equant
hypautomorphic granular aggregates
with individual grains 0.05–0.2 mm in
diameter
Au 0.004 penzhinite (Ag,Cu)4Au(S,Se)4 1984 intergrowths of elongate or platy grains, Bochek et al. (1984)
up to 7 lm
Pd 0.015 borishanskiite Pd1þx(As,Pb)2 x ¼ 0–0.2 1975 grains of 20–30 lm in size Razin et al. (1975)
Br 2.4 kelyanite [Hg2]2þ6Sb3þBrCl2O6 1983 irregular grains up to 1–2 mm in size Vasil’ev et al. (1983)
Cs 3 cesplumtantite (Cs,Na)2(Pb,Sb3þ)3Ta8O24 1986 veinlets and elongated concretions Voloshin et al. (1986)
measuring up to 0.3 mm
mccrillisite NaCs(Be,Li)Zr2(PO4)41–2H2O 1994 crystals occur in mm- to cm-sized Foord et al. (1994)
cavities; individuals are up to 1.2 mm
in maximum dimension
Dy-Gd 5.2–6.2 lepersonnite-(Gd) Ca(Gd,Dy)2(UO2)24(SiO4)4 1982 as mammillary crusts or spherules, up Delines & Piret (1982)
(CO3)8(OH)2448H2O to 5 mm in diameter
mineevite-(Y) Na25Ba(Y,Gd,Dy)2(CO3)11 1992 identified as two inclusions of irregular Khomjakov et al. (1992),
(HCO3)4(SO4)2F2Cl shape, 0.5 and 1 cm in size in K- Yamonva et al. (1992)
feldspar
Sm 7.05 shabaite-(Nd) Ca(Nd,Sm,Y)2(UO2)(CO3)4 1989 pale yellow micaceous flakes and as Deliens & Piret (1989)
(OH)26H2O rosettes up to 5 mm in diameter
Ga 19 gallobeudantite PbGa3(AsO4)(SO4)(OH)6 1996 identified in a single, blackish specimen Jambor et al. (1996)
of massive copper ore, approximately
3 3 5 3 1.5 cm
söhngeite Ga(OH)3 1965 crystalline, twinned groups and Strunz (1965)
aggregates, up to 1 cm; single
crystals are very rare

* Each critical element is present only in mineral formulas of the corresponding tRE minerals.

Page 12

FIG. 8. The share of (a) TU and (b) U minerals between Nickel-Strunz classes.
Ubiquitous group R(TU,U)
There is a substantially lower number of generally
common minerals (1726 in total, with 919 TU and 807
U species). In contrast to the rare group, most of these
minerals (1080 species) were discovered before 1950,
and the other 571 shortly after 1950. The bulk of these
are silicates (231 TU, 215 U), along with lesser
numbers of phosphates, arsenates, and vanadates (186
TU, 113 U). The most uncommon mineral species are
those considered organics (4 TU, 1 U), along with
those containing native elements (34 TU, 26 U) (Fig.
8). Both TU and U groups exhibit similar proportions
of minerals over the various chemical classes except
for borates and phosphates, which are primarily
classed within the TU group.
Nearly 60% of those minerals that class within the
TU and U groups are of low symmetry (i.e., belong to
the triclinic, monoclinic, and orthorhombic crystal
systems). In comparison, the proportion of cubic
minerals increases from 8.7% in TU to 13.2% in the
U group (Fig. 9). The symmetry index for TU species is
0.55; for U it is even higher at 0.64—this is the highest
13
value among all the rarity groups. Since TU and U
groups exhibit a similar distribution over the various
chemical classes and have elevated symmetry indices,
these groups were further analyzed (see below).
All rock-forming minerals, as well as the majority of
accessory phases, belong to the R(TU,U) rarity group:
e.g., pyroxene group (aegirine, augite, diopside, pi-
geonite, enstatite, etc.), olivine group (fayalite, forster-
ite, calcio-olivine, liebenbergite, etc.), amphibole
supergroup (actinolite, hastingsite, tremolite, etc.).
Therefore, this particular rarity group includes minerals
which can be key in the classification of certain rock
types (e.g., Le Bas & Streckeisen 1991). This
observation reflects that the fact that the division of
minerals based on their occurrence on the Earth’s
surface generally corresponds to their relative rarity in
the Earth’s crust.
NETWORK ANALYSIS OF RARITY GROUPS
Rare group R(RE,RR,TR)
According to results from examination of the co-
occurrence matrix of elements and network analysis,
Page 13
14
FIG. 9. The share of (a) TU and (b) U minerals between crystal systems.
the top 10 elements that co-occur with the greatest
number of other elements (see Supplementary mate-
rials 3 and 41) are O and H (both co-occur with 63
elements or 91% of the network), followed by S (60 or
87%), Fe (59 or 86%), Cu (58 or 84%), As and Pb
(both co-occur with 57 elements or 82%), Al (54 or
78%), Ca (53 or 77%), and Cl (52 or 75%). These
elements have the highest degree centrality (impor-
tance score) and form the most extensive chemical
diversity within R(RE,RR) species
The top 10 elements that exhibit the biggest
betweenness centrality (i.e., the proportion of shortest
paths passing through each vertex) among R(RE,RR)
species are H and O (both 4%), followed by S (3.8%),
Cu (4%), Se (3.5%), Fe (3.1%), Pb (3%), As (2.3%),
Al (1.9%), and Si (1.5%). EvaluationPof the same
parameter for these elements in the (RE,RR,TR)
group is somewhat increased for O (6%), Fe (5%), H
(4%), and As (4%), but decreased for S (3.5%). In
graph and network theory, the latter parameter
indicates the degree to which nodes (elements) stand
between each other. Therefore, a node with higher
betweenness centrality would have more control over
the network because more information will pass
through that node. Additionally, the closeness central-
ity and eigenvector centrality were also analyzed and
evaluated. In both cases, the results obtained resemble
those for the betweenness centrality. The eigenvector
1
Supplementary Data are available from the Depository of
Unpublished Data on the MAC website (http://
mineralogicalassociation.ca/), document ‘‘Taxonomy of
Mineral Occurrence Rarity and Endemicity, CM60, 22-
00010’’.
Inc.
centrality is a measure of the influence of a node in a
network. A high eigenvector value means a node is
connected to many other nodes with high scores. It is
noted that O, H, S, Fe, and Cu exhibit the highest
degree centrality and betweenness centrality, suggest-
ing that these elements are the most common among
the entire R(RE,RR) and R(RE,RR,TR) networks.
Therefore, the co-occurrence of these elements with
the matrix elements of each other provides evidence
for the development of patterns among the chemical
entities within the network.
The elements with the lowest-degree centralities
are the rarest and most uncommon for the entire group.
On the other hand, rare elements such as the REE and
PGEs, are always located at the network’s periphery.
Taking this into account and from the viewpoint of the
unique combinations of elements, the rarest and most
unique mineral species among the entire R(RE,RR)
and R(RE,RR,TR) groups is hafnon, HfSiO4—the only
Hf-dominant mineral known to date (Miyawaki et al.
2020).
Three-element communities are detected within the
R(RE,RR) and R(RE,RR,TR) networks using the
Clauset-Newman-Moore greedy modularity maximi-
zation algorithm (Fig. 10). The lanthanides, Sm, Gd,
Dy, and Pr represent the first community, with very
low, but similar, crustal abundances. These elements
always occur together, forming only ten minerals,
primarily carbonates and phosphates. The second
community is composed primarily of siderophile and
chalcophile elements, these frequently forming as
sulfides, sulfosalts, oxides, phosphates, arsenates,
vanadates, and sulfates (blue nodes in Fig. 10a and
green nodes in Fig. 10b). Most of the elements of this
community have a decreased affinity for O, but
Page 14

FIG. 10. A spring network diagram of elements co-occurrences within (a) R(RE,RR) and (b) R(RE,RR,TR) group (size – the
degree centrality, color – the community).
increased affinity for S and Se, so often develop as
sulfides and sulfosalts, halides, and PGM. The third
community consists primarily of lithophile elements;
these exhibit elevated crustal abundances and a
tendency to form as silicates, phosphates, and oxides
(green nodes in Fig. 10a and blue nodes in Fig. 10b).
This group consists of elements with an increased
affinity to O. In addition, OH– and F– tend to occur on
the same sites in the crystal structure as O, their
substitutions leading to development of complex solid
solutions in silicates and oxides. For example, that
involving berborite, Be2(BO3)(OH,F)H2O; bideauxite,
Pb2AgCl3(F,OH)2; panasqueiraite, CaMg(PO4)(OH,F);
and reinhardbraunsite, Ca5(SiO4)2(OH,F)2. Despite
being classified as siderophile elements, Mn and Fe
are commonly associated with O.
tRE group. The analysis of mineral co-occurrences
within this rarity group (see Supplementary material
5) suggests that the top 10 elements responsible for
the chemical diversity within the tRE group are O
(co-occurs with 56 elements or 90% of the network)
followed by H (53 or 85%), S (48 or 77%), Ca (46 or
74%), Na (45 or 72%), Si (42 or 68%), Fe (40 or
64%), Cu (39 or 62%), and Cl and Al (both 38 or
61%). In this scope, the tRE group differs from the
R(RE,RR) group by the co-occurrence of Na and Si at
the center of the network and as the absence of Pb and
As. However, the importance of O, H, and S is
equally the same. The top 10 elements are very
common within the group and therefore appear at the
center of the network (Fig. 11). Additionally, they are
often found to occur with one another. For instance:
15
O occurs with H 201 times, the maximum for the
entire matrix (see Supplementary material 5). Almost
every element is more frequently found with O and H
than with other elements. However, this pattern is not
followed by Au-Ag and PGEs (which tend to form as
sulfides, sulfarsenides, and alloys), Cd, Co, Nd, Se
and Tl.
The top 10 elements with the biggest betweenness
centrality are: O (10%), S (8%), H (6%), Pb (5%), Cu
(4%), As (4%), Te (3%), Ca (3%), Cl (3%), and Na
(3%). Accordingly, the most common elements for the
entire tRE network are O, H, S, Ca, and Na. The
network is generally somewhat decentralized, indicat-
ing that elements within this group are not uniformly
interconnected. Considering the network analysis of
mineral formulae and the presence of only one ‘‘end-
bridge’’ element for the entire tRE group (Te),
gaotaiite is considered the rarest and most unique
mineral from this point of view.
Based on results from the community detection
analyses, at least three different groups of elements are
present in this rarity group. The first one consists of
lanthanides (La, Ce, Nd, Sm, Dy, Gd), Y, Nb, N, C,
Ta, U, and Th, being primarily represented by uranyl
phosphates and sulfates, REE carbonates, REE oxa-
lates, niobates, and tantalates. Carbon occurs in
carbonates with REE, along with carbides and
extremely rare oxalates and cyanates, while N occurs
mainly in uranyl phosphates and organic minerals
containing NH4þ. As was seen in the generally rare
group, the second community is represented by
elements with generally lower oxygen affinities (blue
Page 15
16
FIG. 11. A spring network diagram of elements co-occurrences within tRE group (size – the degree centrality, color – the
community).
nodes in Fig. 11). The minerals classifying in this
group mainly include arsenides, antimonides, bismuth-
ides, sulfites, selenides, tellurides, and sulfides. The
last community consists of minerals containing
elements that are relatively abundant elements in the
Earth’s crust (green nodes in Fig. 11), these also
having high affinities for O.
Ubiquitous group R(TU,U)
The top 10 elements that exhibit the highest degree
centrality are the most common and co-occur with the
highest number of other elements (see Supplementary
material 8). These include O (co-occurs with 54
elements or 85% of the network), Fe (53 or 84%), H
(52 or 83%), S (49 or 78%), Cu (46 or 73%), As (44 or
70%), Ca and Si (both 43 or 68%), Mn (42 or 67%),
and Al (40 or 63%). The elements that show elevated
betweenness centrality are the same, except for
ordering and replacement of Al by Pb: S (9%), Fe
(7%), O (7%), As (6%), H (5%), Cu (5%), Pb (3%), Si
(2%), Mn (,2%), and Ca (,2%). Accordingly, the
most common elements for the entire network are O,
S, H, Fe, and Cu.
Generally, the network is decentralized and three
different communities of nodes are identifiable (Fig.
12). As in other rarity groups, the first community
consists of relatively rare elements: the lanthanides
(La, Ce, Nd), U, Th, Ta, Y, Sr, Nb, Cr, Ti, and V. The
second community consists of elements with increased
affinity to sulfur or selenium (blue nodes in Fig. 12).
Finally, the third community consists primarily of
Page 16

FIG. 12. A spring network diagram of elements co-occurrences within the R(TU,U) group (size – the degree centrality, color –
the community).
lithophile elements (green nodes in Fig. 12). This
community generally represents the most abundant
elements in the crust, essential elements of most rock-
forming silicates, phosphates, oxides, sulfates, carbon-
ates, and nitrates.
The most distinctive elements for the entire
network are the REE in the first community, and the
PGE, and halogens in the second community. These
elements are rare and occur in only a few minerals. For
instance, the PGE form 63 widespread minerals, while
Br and I are present only in five minerals: bromargyr-
ite, AgBr; iodargyrite, AgþI–; murdochite, Cu2þ12Pb4þ2
O15 Cl2 ; perroudite, Ag1þ4 Hg2þ5 S2–5 (I,Br)2 Cl2 ; and
schwartzembergite, Pb2þ5H2I3þO6Cl3. The presence of
some dispersed elements (Cs, Sc, Zr, Be) in the third
17
community, along with abundant elements, is ques-
tionable. However, upon closer examination, it was
noted that: (1) Cs is represented by only two
minerals—pollucite, (Cs,Na)2 Al2 Si4 O12 2H2 O, , and
rhodizite, (K,Cs)Al4Be4(B,Be)12O28; (2) Sc is repre-
sented by three minerals occurring in Sc-bearing
granitic pegmatites—bazzite, Be3 (Sc,Fe3þ,Mg)2 Si6
O18 Na0.32 nH2 O; kolbeckite, Sc(PO4 )2H2 O; and
thortveitite, Sc2Si2O7; (3) minerals containing essential
Zr (e.g., baddeleyite, zircon, zirconolite) are common
accessory minerals in many magmatic rocks; and (4)
even though Be is not uncommon in silicate minerals,
it is not taken to represent either the beryl group or the
beryllonite group, as members of these groups are
quite rare.
Page 17
18
FIG. 13. The correlation between the logarithm (base 10) of crustal abundance of elements and the logarithm (base 10) of the
number of minerals they form.
GEOCHEMISTRY OF RARITY GROUPS
The correlation between the crustal abundance of
elements and the number of minerals these elements
form was probably first observed by Povarennykh
(1966), and several attempts have been made to
explain this correlation (Yaroshevsky & Bulakh
1994, Wenk & Bulakh 2004, Higgins & Smith
2010). The most thorough analysis was probably
provided by Christy (2015).
A log-log plot of the most abundant elements and
the minerals they form (Fig. 13) highlights a linear
correlation between the two. Christy (2015) attributes
this scatter to: (1) the substitution of one element for
another in a solid-solution series, (2) the geochemical
behavior and intermediate electronegativity, and (3)
the ability of an element to exist in a variety of
oxidation states. Data-specific reasons can be suggest-
ed for this scattering and these might contribute even
more bias than those related to crystal-chemistry or
geochemistry. Additionally, while all previous studies
have calculated the number of minerals where a
specific element is recorded, the abundance of
essential and non-essential (impurities) elements
within a specific mineral has not been full considered.
These factors would shift the points in X space only,
but as the number of rare minerals discovered
increases substantially annually, the proportions of
mineral species found to contain rarer elements will
also change. Therefore, the lower part of the plot,
where scatter is present, is expected to change
drastically. The plot of crustal abundance and mineral
abundance is provided here to ensure the used dataset
is diverse and, in general, reflects the relations
highlighted by Christy (2015), so can be used in
further analyses.
A total number of minerals
Figure 14 shows the percentage of minerals
containing each chemical element and the distribution
among rarity groups, sorted by the abundance of the
elements in the Earth’s crust in descending order and
divided over appropriate Goldschmidt groups (i.e.,
lithophile, siderophile, chalcophile, and atmophile).
The first column of this plot depicts each element’s
total number of minerals. There is a general tendency
for a decrease in the number of minerals within each
geochemical group that relates to decreasing the
crustal abundance of an element. For example, for
the siderophiles such as Fe, Mn, Ni and Co, the
number of mineral species drops as the crustal
abundance decreases in the sequence. However, the
percentage of minerals containing Fe and Mn is higher
than that those containing other siderophile elements,
an explanation for which is linked to the mixed
lithophile–siderophile nature of these elements. The
latter implies a high affinity for O, which is the
dominant element in the Earth’s crust (Wedepohl
1995). Additionally, Fe and Mn co-occur with O 1148
and 660 times, respectively, which represents the local
Page 18
 19

FIG. 14. Proportion of IMA-approved minerals present in different rarity groups: (a) all IMA-approved minerals, (b) tRE/all, (c)
R(RE,RR)/all, (d) R(RE,RR,TR)/all, (e) T/all, (f) U/all, (g) R(TU,U)/all. Based on the elements present in their stoichiometric
formula, abundance of these elements in Earth’s Upper Crust (Haynes 2016), Goldschmidt classification, and ion radius,
sorted in descending order of elements abundance in Crust within the same geochemical affinity groups.

maxima for all siderophile elements (see Supplemen-
tary material 1).
The same tendency is observed for lithophile
elements, but this is periodically interrupted, produc-
ing a step-like pattern (Fig. 14a). The latter is mainly
controlled by the co-occurrence of these elements with
the most abundant and principal element of the
group—O (Table 4). Thus, elements like Ca, P, Cl,
B, and U tend to produce positive anomalies, likely
reflecting the greater affinity of these elements for O or
O-based anions. This includes phosphates, primarily in
the form of PO43– and HPO42–; oxyhalides, hydro-
halides, and related double halides for Cl; borates and
borosilicates for B mainly in the form of BO33–; uranyl
phosphates and arsenates, sulfates, and hydroxides for
U mainly in the form of UO22þ. These positive

TABLE 4. CO-OCCURRENCE OF LITHOPHILE ELEMENTS WITH OXYGEN, SORTED

BY THEIR ABUNDANCE IN THE EARTH’S CRUST IN DESCENDING ORDER

Number of Number of Number of

co-occurrences co-occurrences co-occurrences
Element with oxygen Element with oxygen Element with oxygen
Si 1587 V 264 Sm 7
Al 1141 Cr 90 Gd 4
Ca 1440 Rb 6 Dy 3
Na 1092 Ce 181 Er 1
Mg 815 Nd 50 Yb 6
K 556 La 87 Hf 1
Ti 397 Y 140 Cs 30
P 663 Sc 21 Be 127
F 410 Li 121 U 298
Ba 261 Nb 187 Br 11
Sr 155 B 295 Ta 70
Zr 137 Th 41 W 43
Cl 369 Pr 1 I 16

Page 19
20

TABLE 5. CO-OCCURRENCE OF CHALCOPHILE Common minerals

ELEMENTS WITH SULFUR, SORTED BY THEIR
ABUNDANCE IN THE EARTH’S CRUST IN THE The subplot (Fig. 14g) represents the percentage of
DESCENDING ORDER minerals occurring at more than 16 localities, based on
the total number of accepted mineral species. For most
Number of Number of siderophile elements, the percentage of common
co-occurrences co-occurrences minerals is relatively close, with a median value of
Element with sulfur Element with sulfur 36%. However, the percentage of Mn and Mo mineral
Zn 87 In 10 species exhibit negative anomalies, which must in part
Cu 322 Sb 184 be related to the increase in rare O-bearing minerals.
Ga 5 Cd 18 Beyond the widespread occurrence of molybdenite,
Pb 259 Hg 50 Mo can develop as molybdates and oxides within a
Sn 42 Ag 155 rarer group of minerals, these typically being associ-
As 189 Se 22 ated with O2– and OH– anions. On the other hand, Mn
Ge 22 Bi 125 is more often present in silicates and phosphates of
Tl 68 Te 39 generally rare species R(RE,RR,TR) as Mn2þ, com-
monly co-occurring with Ca2þ, Naþ, Mg2þ, and Fe3þ,
along with OH–, O2–, and PO43–. Rhenium is not
anomalies in part reflect the ability of the elements to present in the common group of minerals, and it forms
form strong bonds and complex units with O (Liebau only two minerals: rheniite, ReS2, and tarkianite,
1985). (Cu,Fe)(Re,Mo)4S8. According to the proposed classi-
For chalcophile elements, the relationship is fication, both are metal sulfides and belong to the TR
different, showing a greater number of mineral species group.
containing S, Cu, Pb, As, and Sb. This may be For lithophile minerals, the percentage of common
explained by the affinity of these elements for S and, minerals decreases from O (31%) to B (18%), with
consequently, their co-occurrence with the principal minor deviations attributed to the absence of Rb
element in the group (Table 5). Accordingly, the minerals and positively located Th. Despite the high
increase in the abundances of mineral species content in the Earth’s crust, Rb almost does not form
containing Cu, Pb, As, and Sb can be explained by its minerals, instead most commonly substituting for K
the tendency of these cations to bond with S. Thus, in minerals (owing to it having nearly the same ionic
nearly half of all Cu-bearing minerals are bond to S, radius as K and because the latter is more than 2000
either as S2– and SO42–. As well, Pb, Sb, and Ag tend times more abundant than Rb; Butterman & Reese Jr
to form sulfosalts based on the PbS and SnS 2003). Only six valid IMA minerals contain essential
archetypes, and more rarely as phosphates and oxy- Rb, none of which belong to the common rarity class
halides. Nevertheless, the most common anion for all of minerals. However, of these, two belong to common
Pb, Sb, and Ag minerals is S2–. Also, almost half of the rock-forming groups of minerals, e.g., rubicline is
As minerals are sulfosalts based on the SnS archetype, feldspar, and voloshinite is a mica. The group of
where the most common anion is again S2–. Despite elements from Pr to Hf does not form common
their relatively high abundance in the crust, some other minerals, as they are rather dispersed in the crust.
elements, such as Zn, Ga, Sn, and Ge, develop in fewer However, Cs, as an alkaline metal, does not belong to
distinct mineral species. For Ga and Ge, this is due to this group and does form 2 common minerals (6%):
their geochemical behavior (substituting for Al and Si, pollucite, (Cs,Na)2(AlSi2O6)22H2O, and rhodizite,
respectively), so neither exhibits a strong enrichment (K,Cs)Al4Be4(B,Be)12O28. In comparison, the majority
resulting from processes relevant to mineral formation of Cs minerals are rare (94%), these including minerals
(Burton et al. 1959, Frenzel 2016). that are phyllosilicates, tectosilicates, and phosphates.
As all atmophile elements occur either in the Among mineral containing chalcophile elements,
gaseous state or form volatile hydrides, they are most (10–40%) are assigned to the common-mineral
strongly depleted in the Earth’s crust relative to their group, with a median of 24%, and 60–90% of the rare-
solar abundances, owing to losses from the atmosphere mineral group, with a median of 74%. Minerals
during the formation of Earth. However, H, C, and N containing elements from S to Ge generally are
are widely represented in minerals, forming hydrox- classed into the common-mineral group, but with
ides, carbonates, and nitrides, respectively. The variable, individual distributions.. The elements Ga,
number of minerals decreases in the following Ge, and Tl almost exclusively develop in mineral
sequence: H – C – N, which correlates with the species that class in the rare-mineral group. The
abundances of these elements in Earth’s crust. percentage of mineral species containing these ele-

Page 20

ments differs from those of chalcophile elements in
general, a fact evident when one considers the
presence of negative anomalies in light of the general
trends. As Ga and Ge easily substitute for Al and Si,
respectively, they do not tend to form as distinct
mineral species (Frenzel 2016). The greatest abun-
dance of minerals wherein essential quantities of Ga
and Ge occur, occurs in minerals formed at the
Tsumeb mine, Namibia. Typically, Ge and Ga co-
occur with Cu2þ, Fe3þ, Pb2þ, and Zn2þ as well as S2–,
OH–, and SO42– anions. Accordingly, these elements
commonly form as sulfides and sulfates. It is noted that
sulfide deposits developed from solutions with higher
S activity facilitated the formation of Ge-dominant
sulfides, particularly minerals with Cu2þ [e.g., barquil-
lite, Cuþ2CdGeS4; briartite, Cuþ2(Fe2þ,Zn)GeS4; zin-
cobriartite, (Cuþ,Cu2þ)2(Zn,Fe2þ)(Ge,Ge)S4] or to
substitute for As or Sn in sulfosalts (Bernstein 1985).
Thallium is another rare chalcophile element normally
found as a trace constituent in a variety of rock-
forming minerals, typically being incorporated through
substitutions involving alkali elements like Rbþ and
Kþ in silicates and sulfates as well as for Agþ and Pb2þ
in sulfides and sulfosalts (Rader et al. 2018). It is an
essential element in only three minerals that occur at
more than 16 localities: cuprostibite, Cu2(Sb,Tl);
lorándite, TlþAs3þS2–2; and thalcusite, (Cu,Fe)4Tl2S4.
Most other Tl minerals are classified as rare sulfosalts
of the SnS archetype: sulfarsenides, sulfantimonides,
and sulfbismuthides.
The number of common minerals number contain-
ing atmophile elements correlates with the crustal
abundance of these elements, decreasing from 28% for
H to 15% for N with a median of 25%.
Endemic minerals
The number of true rare endemic minerals (tRE;
Fig. 14b) is relatively close among all the geochemical
groups, with an average value of 7.8% and a median of
6%. Generally, the slightly higher content of true
endemic minerals can be attributed to more wide-
spread distribution of the chemical elements that are
required for their formation. The latter is illustrated
when such minerals developed from siderophile
elements are considered: the greatest number of tRE
are those containing elements that range from Fe to
Mo, these corresponding to the most widespread
elements in this group. Most of these minerals are
silicates, phosphates, oxides, and sulfides. For those
elements that are represented by a low number of
minerals, it is possible that no tRE may be found.
These could include Pt, Os, Ru, Re for siderophile
elements; Rb and most the REE elements for the
lithophile group; and In for chalcophile group.
21
On the other hand, some REE reveal that there is a
higher percentage of some tRE containing REE,
although this is likely due to the low total number of
minerals that exist (see subsection ‘‘Composition of
rarity groups. The rarest of the rarest.’’). Also, the
higher percentage of tRE is strongly influenced by
specific chalcophile elements, namely Ge, Ga, and Cd.
Due to their geochemical behavior, these elements are
rarely concentrated in their minerals, instead they are
more likely to be dispersed in Zn-bearing minerals.
Rare minerals
A considerably higher number of minerals can be
attributed to the R(RE,RR) group that comprises the
minerals that have been found at four localities or
fewer (i.e., rare minerals). In general, a decrease of
element abundance in the Earth’s crust corresponds to
an increase in the percentage of rare minerals that form
from these elements. This pattern is apparent when
data for lithophile elements is compared, as in Figures
14c and Figure 14g. For chalcophile elements, the
percentage of minerals that class into the in the rare-
mineral group ranges from 40 to 50%, many of these
being minerals that contain Ga, Ge, Tl, In, Cd, Hg, and
Se. As mentioned previously, this distribution is
heavily influenced by the geochemical behavior of
the elements involved (notably, dispersed elements),
along with their relatively low abundances in the
Earth’s crust. For minerals containing Se, this may be
connected with the geochemical similarity between Se
and S and the prominently higher concentration of the
latter in the crust; thus, relatively few geochemical
processes are sufficiently effective at fractionating
between the two, a key step required for subsequent
formation of discrete selenides and selenates. Indeed,
the combination of Se with the PGE (especially Pd and
Pt), As, Sb, Sn, Cd, Te, Hg, Cl, U, and Zn leads to the
formation of almost exclusively rare minerals.
The number of lithophile elements in the rare-
mineral group presents a complex behavior. By
comparison with the total number of minerals, a
generally inverse relationship between the percentage
of minerals and their rarity is observed. For example,
the number of rare minerals increases from 47% for O
to 60% for Cr, while the percentage of R(TU,U)
minerals decreases in this range from 28% to 19%
respectively. Additionally, subgroups within the
broader groups of elements are also evident, these
being connected with the co-occurrences of the
elements. Typically, it is predicated that the potential
number of minerals that could develop containing a
given element would correlate with the abundance of
the element in the Earth’s crust. This can be seen in
Figure 14g, where the number of minerals gradually
Page 21
22 TH
decreases from O to I with some deviations from the
trend. However, since R(RE,RR)/Rall minerals ¼ (Rall
minerals – R(T,U))/Rall minerals, an opposite behav-
ior for rare species is observed. In effect, the ability of
an element to develop in a high number of individual
mineral species is tied to its abundance in the Earth’s
crust: the rarer the element is, the greater is it chance
to develop as a rare mineral species rather than a
common one. As an example, I and Br form rare
minerals that belong to the lithophile group of
elements, thus highlighting the dispersion effect. This
is noted in the fact that 35% of rare Br-I minerals
contain Cl– and .75% of their halides contain Hg
cations Hgþ or Hg2þ. The most prominent rare halides
of Hg are the members of the calomel group:
kuzminite, Hg2þ2(Br,Cl)2, and moschelite, Hg2þI–2.
Other naturally occurring Hg iodates include coccinite,
Hg2þI–2, and tedhadleyite, Hg2þHgþ10O4I–2(Cl,Br)2.
However, the majority of Hg halides are exclusively
rare, occurring in the low temperature, oxidized
alteration zones of Hg ore deposits, where the richest
Hg ores can contain up to 2.5% Hg. As a rule, these
minerals often co-occur at the same localities, i.e.,
aurivilliusite, gaildunningite, mosesite, tedhadleyite,
and vasilyevite which all are occur at the Clear Creek
Hg mine (Table 6), presumably developing in response
to changes in fluid composition (Cooper et al. 2019).
Both Cs and Rb form more minerals in the rare-
mineral group than do those elements that neighbor
them on the periodic table. There are only six Rb
minerals—three of these are RR, and the other three
are TR species, so Rb minerals must be considered
extremely rare. The same applies to Cs minerals—75%
(or 25 minerals) of these minerals are R(RE,RR), and
an additional 18% (or six minerals) are TR the sum of
which (93%) classifies them as being generally rare.
Both are considered as large lithophile elements and
their large ionic radii result in them behaving as
incompatible elements. The latter results in them being
concentrated in the melt portions developed during
late-stages of crystal fractionation. Accordingly, five
of six Rb minerals—londonite, (Cs,K,Rb)Al4 Be4 (B,
Be)12 O28 ; masutomilite, (K,Rb)(Li,Mn3þ,Al)3 (AlSi3
O10 )(F,OH)2 ; ramanite-(Rb), Rb[B5 O6 (OH)4 ]2H2 O;
rubicline, Rb(AlSi3 O8 ); and voloshinite, Rb(LiAl1.5
A1.5 )(Al0.5 Si3.5 )O10 F2 are found solely in the inner
zones of highly evolved rare-element, Li- and B-rich
granitic pegmatites (Nagashima 1975, Simmons et al.
2001, Teertstra et al. 1998, Thomas et al. 2008, Pekov
et al. 2010). The sixth Rb mineral, aleutite, [Cu5
O2](AsO4)(VO4)(Cu,K,Pb,Rb,Cs)Cl, was identified as
a product of fumarolic activity (Siidra et al. 2019). The
bulk of Cs minerals was also identified in rare-element
granitic pegmatites (Agakhanov et al. 2021, Foord et
al. 1994, Matzko et al. 1989) and fumarole sublimates
(Pekov et al. 2018).
For siderophile elements, the number of R(RE,RR)
species decreases as does their respective concentra-
tion in the Earth’s crust. This phenomenon is opposite
in behavior when compared to lithophile and atmo-
phile elements. This can be attributed to an increased
influence of O on the ability of siderophile elements to
form rare minerals, which is reflected in their co-
occurrence matrices (see Supplementary material 3
and 4). The percentage of rare minerals is observed to
decrease in the following sequence: Mo  Mn  Ni
 Fe  Co. Both Mo and Co have similar crustal
abundances, but the number of rare minerals drops
drastically from 70% for Mo to 34% for Co, which
may be connected to differences in the geochemical
behavior of these elements in forming minerals. For
example, the rare minerals of Mo incorporate MoO4
and MoO6 structural units more often than widespread
minerals. Conversely, Co does not form ions with O
and primarily combines with S and As to form
sulfarsenides and arsenides such as cobaltite, CoAsS;
safflorite, CoAs2; glaucodot, (Co,Fe)AsS; and skutter-
udite, CoAs3 (Holleman 2019). Another possible
explanation for the increased number of rare minerals
of Co is the dispersion of Co in Ni-bearing minerals
due to similarities in their geochemical behaviors
(Feng et al. 2016).
An average of 51% of minerals containing
atmophile elements were classified into the rare group.
There is a tendency for H to occur in rarer minerals
along with a significantly greater number of co-
occurring elements (see Supplementary materials 4
and 8). This leads to the formation of a large and
diverse group of minerals containing complex anions,
e.g., borates (mono-, di-, tri-, tetra-, and hexaborates),
which are present primarily in the rare group of
minerals. Carbon also forms a diverse range of rare
minerals—carbonates and organic compounds, sili-
cates, and minor sulfates—although the bulk of C-
bearing common minerals is represented by carbon-
ates. Nitrogen follows a similar pattern, although it
typically forms as rare minerals rather than common
ones. This is attributed to the fact that N occurs not
only in nitrates and nitrides that are both rare and
common, but also in organic compounds that are
exclusively rare; therefore, that influences the more
extensive distribution of rare species.
CONCLUSIONS
This study presents a statistical classification of the
rarity of minerals and a meaningful definition of their
true endemicity. In addition, minerals’ chemical and
structural features in the rarity groups were analyzed in
Page 22
 TABLE 6. ESSENTIAL DATA ON THE RARE Hg HALIDES

Mineral name Formula Locality Associated minerals at locality Reference

aurivilliusite Hg1þHg2þ(I,Br,Cl)2O2 Clear Creek mercury mine, New Idria calcite, globules of native mercury, Roberts et al. (2004)
Mining District, California, USA cinnabar, and edgarbaileyite
comancheite Hg2þ55N3–24(NH2,OH)4(Cl,Br)34 Mariposa mercury mine, Terlingua calcite, quartz, cinnabar, calomel, and Roberts et al. (1981)
district, Texas native mercury
gaildunningite Hg2þ3[NHg2þ2]18(Cl,I)24 Clear Creek mercury mine, New Idria quartz, CCUK-18 Cooper et al. (2019)
Mining District, California, USA
kadyrelite [Hg2]2þ3OBr3(OH) Kadyrel mercury occurrence, Uyuk eglestonite, cinnabar, calomel, Vasil’ev (1987)
Range, Tuva, Russian Federation kuzminite
kelyanite [Hg2]2þ6Sb3þBrCl2O6 Kelyan antimony-mercury deposit, calomel, eglestonite, Native mercury, Vasil’ev et al. (1983)
Baunt District, Republic of Buryatia, shakhovite, cinnabar, quartz
Russian Federation
kuzminite [Hg2]2þ(Br,Cl)2 Kadyrel mercury occurrence, Uyuk eglestonite, lavrentievite, cinnabar, Vasil’ev et al. (1986)
Range, Tuva, Russian Federation calomel
moschelite [Hg2]2þI2 Backofen mercury mine, Landsberg, eglestonite, cinnabar, native mercury Krupp et al. (1989)
Rhineland-Palatinate, Germany
mosesite (Hg2N)(Cl,SO4,MoO4)H2O Clear Creek mercury mine, New Idria eglestonite, calomel, native mercury Canfield et al. (1910)
Mining District, California, USA
tedhadleyite Hg2þHg1þ10O4I2(Cl,Br)2 Clear Creek mercury mine, New Idria globules of native mercury, calomel, Roberts et al. (2002)
Mining District, California, USA traces of cinnabar, eglestonite, and
montroydite
tocornalite (Ag,Hg)I Kintore open cut, Broken Hill, Australia Birch (1997)
vasilyevite [Hg2]2þ10I3Br2Cl[CO3]O6 Clear Creek mercury mine, New Idria globules of native mercury, eglestonite, Roberts et al. (2003)
Mining District, California, USA montroydite, cinnabar, and an
undefined Hg oxyhalide
23

Page 23
24
connection with the geochemistry and co-occurrence
of these elements.
It is suggested that ‘‘true’’ endemicity be defined
based on the number of localities and the year of
discovery. While RE minerals discovered in the early
20th century and before are likely to stay endemic, RE
minerals discovered in recent years are often subse-
quently discovered in other localities, resulting in them
being classed as RR instead of RE, taxonomically.
Therefore, the relatively large increase in the rate at
which new RE minerals have been discovered in the
last 20 years does not directly correspond to the
increase in the rate of tRE minerals discovered.
Instead, the majority of current RE minerals are more
likely to be better assigned to the RR subgroup.
The preliminary gap between ‘‘true’’ endemicity
and that observed suggests that the number of tRE
mineral species is overestimated (around 850 in
contrast to 1232), while the number of RR is likely
underestimated (nearly 1750 in contrast to 1473).
While the fluctuation in the number of species that are
generally rare is probably overestimated for each rarity
group and subgroup, this variation is the highest in the
rare group. This can be explained by the fact that it is
easier for mineral species to transition from RE to RR
subgroup than it is for rare or transitional mineral to
transfer to a lower rarity group. Only one additional
locality discovery is needed for the mineral to
transition from RE to RR, while for a mineral to
transition from TR to TU, it needs to be discovered in
up to 12 additional localities, which is challenging. For
this to occur, considerably more time and investigation
of more mineral localities is required, which in turn
makes the number of minerals constituting lower rarity
groups relatively constant. Additionally, a much
higher stability rate is observed within rarity groups
than between them. Therefore, while the rarity index
of a given mineral might change, it is unlikely that the
taxonomy of the rarity groups will change substantial-
ly.
Compositionally, the common minerals (TU and U
rarity groups) differ based on the lower number of
halides, organic compounds, and borates, but also the
increased number of carbonates, native elements,
sulfides, and sulfosalts, as well as sulfates. The number
of organic compounds is the highest for the RE
group—the rarest one. Most of these minerals are
oxalates and hydrocarbon-bearing minerals. The
number of halides is the highest for the RR group.
The highest number of cubic minerals are present in
the U group (13.2%), while in the RE group this
number is almost half (7.3%). The number of
monoclinic and triclinic minerals gradually decreases
from the RE to U group from 48 to 38%. The
symmetry index, calculated for each rarity group,
gradually decreases from U to RE group: U – 0.64, TU
– 0.55, TR – 0.54, RR – 0.48, RE – 0.47. Changes in
the symmetry indices among the groups provides
support for the proposal of Urusov (2007) that rare
minerals generally are of lower symmetry than more
abundant ones, with the most stable minerals being
characterized by higher symmetry. Abnormally high
values of the symmetry index correlate with the time
period in which the greatest number of widely
distributed minerals was discovered. Accordingly, it
indicates that the average symmetry index of a rare-
mineral species is significantly lower than that of
common minerals. Statistically, this reflects the fact
that the mean symmetry index gradually decreases
with the number of new minerals that are discovered,
most of these being rare or endemic.
The network analysis, furthered by the application
of community detection algorithms, suggests the
existence of at least three different communities
within each rarity group: (1) the first and the smallest
involving the rarest elements (i.e., lowest abundance in
the Earth’s crust). This group is dominated by
lanthanide-bearing minerals and generally corresponds
to those minerals that are uncommon in all networks;
(2) the second community is characterized by an
increased affinity for S (including PGEs, halogens,
Co–Ni–Cr–V); and (3) the third community, that
includes minerals dominated by lithophile elements,
these having a high affinity for O (typically, these
include elements essential for the formation of rock-
forming minerals). All of these communities are
represented by elements having similar geochemical
affinities and crustal abundances; thus, they likely
reflect the relative abundance of minerals their
elements tend to form. The latter effect is the most
prominent for the U group, where the third community
is composed of elements that occur in elevated crustal
abundances; therefore, this community represents
major rock-forming silicates, phosphates, oxides,
sulfates, carbonates, and nitrates. Accordingly, the
co-occurrence matrices of essential elements used in
the community analyses reflects the abundance of
these elements in the crust and the relative abundance
of the minerals they form. The latter provides for
opportunities for using methods like these in deci-
phering common mineral-forming processes and
environments by combining essential elements data
with ions data, which is currently a work in progress.
It must be noted that it is not just the crustal
abundances of elements that are critical to determining
the abundance of a given mineral species as a function
of the elements present, but there are also other factors,
including the co-occurrence of elements in minerals,
their geochemical affinity to the principal element in
the group (i.e., S for chalcophile and O for lithophile
Page 24

elements), and processes related to geochemical
dispersion or concentration. The distribution of
minerals related to each chemical element among the
various rarity groups shows a strong influence of all
these factors. Generally, the abundance of common
minerals increases as does the abundance of the
element in the Earth’s crust. At the same time, an
opposite behavior is seen in rare minerals. However,
these geochemical factors have little effect on true rare
endemic minerals since the percentage of true rare
endemic minerals is quite close among all geochem-
ical groups, with an average value of 7.8%.
ACKNOWLEDGMENTS
The project received funding from the European
Union’s Horizon 2020 research and innovation
Programme based on a grant agreement under the
Marie Skłodowska-Curie scheme No. 945478 and was
supported by the Slovak Research and Development
Agency (contract APVV-19-0065).
REFERENCES
AGAKHANOV, A.A., KASATKIN, A.V., BRITVIN, S.N., SIIDRA,
O.I., PAUTOV, L.A., PEKOV, I.V., & KARPENKO, V.Y.U.
(2021) Cesiokenopyrochlore, the first natural niobate with
an inverse pyrochlore structure. Canadian Mineralogist
59, 149–157.
BERNSTEIN, L.R. (1985) Germanium geochemistry and
mineralogy. Geochimica et Cosmochimica Acta 49,
2409–2422.
BIRCH, W.D. (1997) Minerals of the Kintore and Block 14
open cuts at Broken Hill, New South Wales. Australian
Journal of Mineralogy 3, 23–71.
BOCHEK, L.I., SANDOMIRSKAYA, S.M., CHUVIKINA, N.G., &
KHVOROSTOV, V.P. (1984) A new selenium-containing
sulfide of silver, gold, and copper — penzhinite
(Ag,Cu)4Au(S,Se)4. Zapiski Vsesoyuznogo Mineralogi-
cheskogo Obshchestva 113, 356–360.
BUGRIEN, J., MWITONDI, K., & SHUWEIHDI, F. (2014) A kernel
density smoothing method for determining an optimal
number of clusters in continuous data. Risk Analysis 9,
165.
BURTON, J.D., CULKIN, F., & RILEY, J.P. (1959) The
abundances of gallium and germanium in terrestrial
materials. Geochimica et Cosmochimica Acta 16, 151–
180.
BUTTERMAN, W.C. & REESE, JR, R.G. (2003) Mineral
Commodity Profiles — Rubidium. USGS Open-File
Report.
CANFIELD, F.A., HILLEBRAND, W.F., & SCHALLER, W.T. (1910)
Mosesite, a new mercury mineral from Terlingua, Texas.
American Journal of Science 30(177), 202–208.
25
CHRISTY, A.G. (2015) Causes of anomalous mineralogical
diversity in the Periodic Table. Mineralogical Magazine
79, 33–49.
CLAUSET, A., NEWMAN, M.E.J., & MOORE, C. (2004) Finding
community structure in very large networks. Physical
Review E - Statistical Physics, Plasmas, Fluids, and
Related Interdisciplinary Topics 70, 6.
COOPER, M.A., HAWTHORNE, F.C., ROBERTS, A.C., STANLEY,
C.J., SPRATT, J., & CHRISTY, A.G. (2019) Gaildunningite,
ideally Hg2þ3[NHg2þ2]18(Cl,I)24, a new mineral from the
Clear Creek mine, San Benito County, California, USA:
Description and crystal structure. Canadian Mineralogist
57, 295–310.
DEER, W.A., HOWIE, R.A., WISE, W.S., & ZUSSMAN, J. (2004)
The rock-forming minerals: Silica minerals. Volume 4B.
In Framework Silicates: Silica Minerals, Feldspathoids
and the Zeolites (R.A. Howie, J. Zussman, & W.S. Wise,
eds.). Geological Society of London, London, UK (1–
982).
DEER, W.A., HOWIE, R.A., & ZUSSMAN, J. (1992) An
Introduction to the Rock-Forming Minerals–Chain Sili-
cates. Longman, London, UK.
DELIENS, M. & PIRET, P. (1989) La shabaı̈te-(Nd), Ca(TR)2
(UO2 )(CO3 )4 (OH)2 A6H2 O, nouvelle espèce minérale de
Kamoto, Shaba, Zaı̈re. European Journal of Mineralogy
1, 85–88.
DELINES, M. & PIRET, P. (1982) Bijvoetite and lepersonnite,
hydrated uranyl and rare earth carbonates from Shinko-
lobwe, Zaire. Canadian Mineralogist 20, 231–238.
EPANECHNIKOV, V.A. (1969) Non-parametric estimation of a
multivariate probability density. Theory of Probability &
Its Applications 14, 153–158.
FENG, C.Y., ZHAO, Y.M., LI, D.X., LIU, J.N., & LIU, C.Z.
(2016) Mineralogical characteristics of the Xiarihamu
nickel deposit in the Qiman Tagh mountain, East Kunlun,
China. Geological Review 62, 215–228.
FLINK, G. (1886) Mineralogiska notiser 1. Harstigit från
Pajsberg. Bihang till Kongl. Vetenskaps-Akademiens
Förhandlingar 12, 59–63.
FOORD, E.E., BROWNFIELD, M.E., LICHTE, F.E., DAVIS, A.M., &
SUTLEY, S.J. (1994) McCrillisite, NaCs(Be,Li)Zr2 (PO4 )4
A1-2H2O, a new mineral species from Mount Mica,
Oxford County, Maine, and new data for gainesite.
Canadian Mineralogist 32, 839–842.
FRENZEL, M. (2016) The Distribution of Ga, Ge and In in
Conventional and Non-Conventional Resources - Impli-
cations for Global Availability. PhD Thesis, Faculty of
Geosciences, Geoengineering and Mining at the Techni-
sche Universität Bergakademie Freiberg, Germany.
GENKIN, A.D., VYAL’SOV, L.N., EVSTIGNEEVA, T.L., LAPUTINA,
I.P., & BASOVA, G.V. (1983) Rhodplumsite Rh3Pb2S2 —
A new sulfide of rhodium and lead. Mineralogicheskii
Zhurnal, Kiev 5, 87–91.
Page 25
26
GOLDEN, J.J., PIRES, A.J., HAZEN, R.M., DOWNS, R.T., RALPH,
J., & MEYER, M. (2016) Building the Mineral Evolution
Database: Implications for Future Big Data Analysis. In
Mineralogical Evidence for the Co-Evolution of the
Geosphere and Biosphere: In Honor of Robert M. Hazen.
Proceedings from the GSA Annual Meeting in Denver,
Colorado, USA – 2016.
GREW, E.S., HYSTAD, G., HAZEN, R.M., KRIVOVICHEV, S.V., &
GORELOVA, L.A. (2016) Counting Boron minerals in
Earth’s crust: Can contradictory evidence be reconciled?
In Mineralogical Evidence for the Co-Evolution of the
Geosphere and Biosphere: In Honor of Robert M. Hazen.
Proceedings from the GSA Annual Meeting in Denver,
Colorado, USA – 2016.
GREW, E.S., HYSTAD, G., HAZEN, R.M., KRIVOVICHEV, S.V., &
GORELOVA, L.A. (2017) How many boron minerals occur
in Earth’s upper crust? American Mineralogist 102, 1573–
1587.
GREW, E.S., HYSTAD, G., TOAPANTA, M.P.C., ELEISH, A.,
OSTROVERKHOVA, A., GOLDEN, J., & HAZEN, R.M. (2019)
Lithium mineral evolution and ecology: Comparison with
boron and beryllium. European Journal of Mineralogy 31,
755–774.
HARLAUX, M., MARIGNAC, C., CUNEY, M., & MERCADIER, J.
(2015) The Puy-les-Vignes breccia pipe (Massif Central,
France): A unique occurrence of polymetallic W-Nb6Ta-
HREE-Bi-Cu-As6Au-Ag mineralization in the Variscan
belt. In Sciences of the Universe. Proceedings from the
13th Biennial SGA Meeting, Aug 2015, Nancy, France
(749–752).
HAYNES, W.M. (2016) Abundance of Elements in the Earth’s
Crust and in the Sea. CRC Handbook of Chemistry and
Physics, 95th edition, Internet Version.
HAZEN, R.M. & AUSUBEL, J.H. (2016) On the nature and
significance of rarity in mineralogy. American Mineral-
ogist 101, 1245–1251.
HAZEN, R.M., GREW, E.S., DOWNS, R.T., GOLDEN, J., &
HYSTAD, G. (2015a) Mineral ecology: Chance and
necessity in the mineral diversity of terrestrial planets.
Canadian Mineralogist 53, 295–324.
HAZEN, R.M, HYSTAD, G., DOWNS, R.T., GOLDEN, J.J., PIRES,
A.J., & GREW, E.S. (2015b) Earth’s ‘‘missing’’ minerals.
American Mineralogist 100, 2344–2347.
HAZEN, R.M., HUMMER, D.R., HYSTAD, G., DOWNS, R.T., &
GOLDEN, J.J. (2016) Carbon mineral ecology: Predicting
the undiscovered minerals of carbon. American Mineral-
ogist 101 889–906.
HAZEN, R.M., HYSTAD, G., GOLDEN, J.J., HUMMER, D.R., LIU,
C., DOWNS, R.T., MORRISON, S.M., RALPH, J., & GREW, E.S.
(2017) Cobalt mineral ecology. American Mineralogist
102, 108–116.
HIGGINS, M.D. & SMITH, D.G.W. (2010) A census of mineral
species in 2010. Elements 6, 346.
HOLLEMAN, A.F. (2019) Lehrbuch der anorganischen Chemie.
GmbH & Co KG, Berlin, Germany, 1454 pp.
HYSTAD, G., ELEISH, A., HAZEN, R.M., MORRISON, S.M., &
DOWNS, R.T. (2019a) Bayesian estimation of Earth’s
undiscovered mineralogical diversity using noninforma-
tive priors. Mathematical Geosciences 51, 401–417.
HYSTAD, G., MORRISON, S.M., & HAZEN, R.M. (2019b)
Statistical analysis of mineral evolution and mineral
ecology: The current state and a vision for the future.
Applied Computing and Geosciences 1, 100005.
JAMBOR, J.L., OWENS, D.R., GRICE, J.D., & FEINGLOS, M.N.
(1996) Gallobeudantite, PbGa3[(AsO4),(SO4)]2(OH)6, a
new mineral species from Tsumeb, Namibia, and
associated new gallium analogues of the alunite-jarosite
family. Canadian Mineralogist 34, 1305–1315.
KHOMJAKOV, A.P., POLEZHAEVA, L.I., YAMNOVA, N.A., &
PUSHCHAROVSKY, D.Y. (1992) Mineevite-(Y) — Na25
Ba(Y,Gd,Dy)2 (CO3 )11 (HCO3 )4 (SO4 )2 F2 — A new
mineral. Zapiski Vserossijskogo Mineralogicheskogo Ob-
shchestva 121, 138–142 (in Russian).
KRIVOVICHEV, V.G., CHARYKOVA, M.V., & KRIVOVICHEV, S.V.
(2018) The concept of mineral systems and its application
to the study of mineral diversity and evolution. European
Journal of Mineralogy 30, 219–230.
KRUPP, E., NOTTES, G., & HEIDKE, U. (1989) Moschelite
(Hg2I2): A new mercury mineral from Landsberg-
Obermoschel. Neues Jahrbuch für Mineralogie Monat-
shefte, 524–526.
KUNIN, W.E. & GASTON, K.J. (1993) The biology of rarity:
patterns, causes and consequences. Trends in Ecology &
Evolution 8, 298–301.
LAFUENTE, B., DOWNS, R.T., YANG, H., & STONE, N. (2016)
The power of databases: The RRUFF project. In
Highlights in Mineralogical Crystallography (L. Bindi,
S.V. Churakov, R. Downs, F. Gfeller, S.V. Krivovichev,
B. Lafuente, & H. Yang, eds). Walter de Gruyter GmbH,
Bern, Switzerland (1–29).
LE BAS, M.J. & STRECKEISEN, A.L. (1991) The IUGS
systematics of igneous rocks. Journal of the Geological
Society 148, 825–833.
LIEBAU, F. (1985) Chemical bonds in silicates. In Structural
Chemistry of Silicates (F. Liebau, ed.). Springer, Berlin,
Germany (14–51).
LIU, C., ELEISH, A., HYSTAD, G., GOLDEN, J.J., DOWNS, R.T.,
MORRISON, S.M., HUMMER, D.R., RALPH, J.P., FOX, P., &
HAZEN, R.M. (2018) Analysis and visualization of
vanadium mineral diversity and distribution. American
Mineralogist 103, 1080–1086.
LIU, C., HYSTAD, G., GOLDEN, J.J., HUMMER, D.R., DOWNS,
R.T., MORRISON, S.M., RALPH, J.P., & HAZEN, R.M. (2017)
Chromium mineral ecology. American Mineralogist 102,
612–619.
Page 26
 X 27
LLOYD, S. (1982) Least squares quantization in PCM. IEEE
Transactions on Information Theory 28, 129–137.
MALEC, J., VESELOVSKY, F., BOHMOVA, V., & PROUZA, V.
(2012) Jacutingaite, palladium gold and Pd-selenides in
Cu-ore from carbonaceous sediments from Ko álova u
Semil (Podkrkonoska basin). Zprávy o geologických
výzkumech, 189–193 (in Czech).
MATZKO, J.T., EVANS, H.T., MROSE, M.E., ARUSCAVAGE, P.,
CHESNOKOV, B.V., & POLYAKOV, V.O. (1989) New mineral
names. American Mineralogist 74, 500–505.
MIYAWAKI, R., HATERT, F., PASERO, M., & MILLS, S.J. (2020)
IMA Commission on New Minerals, Nomenclature and
Classification (CNMNC) — Newsletter 56. Mineralogical
Magazine 84, 623–627.
MORRISON, S.M., BUONGIORNO, J., DOWNS, R.T., ELEISH, A.,
FOX, P., GIOVANNELLI, D., GOLDEN, J.J., HUMMER, D.R.,
HYSTAD, G., KELLOGG, L.H., KREYLOS, O., KRIVOVICHEV,
S.V., LIU, C., MERDITH, A., PRABHU, A., RALPH, J., RUNYON,
S.E., ZAHIROVIC, S., & HAZEN, R.M. (2020) Exploring
carbon mineral systems: Recent advances in C mineral
evolution, mineral ecology, and network analysis. Fron-
tiers in Earth Science 8, 208.
MORRISON, S.M., LIU, C., ELEISH, A., PRABHU, A., LI, C.,
RALPH, J., GOLDEN, J.J., FOX, P., HUMMER, D.R., MEYER,
M.B., & HAZEN, R.M. (2017) Network analysis of
mineralogical systems. American Mineralogist 102
1588–1596.
NAGASHIMA, K. (1975) A new mineral masutomilite from Mt
Tanakami, Otsu City, Shiga Prefecture. Chigaku Kenkyu
26, 319–324.
PARZEN, E. (1962) On estimation of a probability density
function and mode. The Annals of Mathematical Statistics
33 1065–1076.
P EKOV , I.V., K ONONKOVA , N.N., A GAKHANOV , A.A.,
BELAKOVSKY, D.I., KAZANTSEV, S.S., & ZUBKOVA, N.V.
(2010) Voloshinite, a new rubidium mica from granitic
pegmatite of Voron’i Tundras, Kola Peninsula, Russia.
Geology of Ore Deposits 52, 591–598.
P E K O V , I.V., Z U B K O V A , N.V., A G A K H A N O V , A.A.,
PUSHCHAROVSKY, D.Y., YAPASKURT, V.O., BELAKOVSKIY,
D.I., VIGASINA, M.F., SIDOROV, E.G., & BRITVIN, S.N.
(2018) Cryptochalcite, K2Cu5O(SO4)5, and cesiodymite,
CsKCu5O(SO4)5, two new isotypic minerals and the K–Cs
isomorphism in this solid-solution series. European
Journal of Mineralogy 30, 593–607.
POVARENNYKH, A.S. (1966) On the abundance of the chemical
elements in the Earth’s crust and on the quantity of
mineral species. Mineralogicheskiy Sbornik 20(2), 178–
185.
RADER, S.T., MAZDAB, F.K., & BARTON, M.D. (2018)
Mineralogical thallium geochemistry and isotope varia-
tions from igneous, metamorphic, and metasomatic
systems. Geochimica et Cosmochimica Acta 243, 42–65.
RAZIN, L.V., DUBAKINA, L.S., MESHCHANKINA, V.I., & BEGIZOV,
V.D. (1975) Borishanskiite — A new plumboarsenide of
palladium for the copper-nickel sulfides ores of the Talnakh
differentiated intrusive. Zapiski Vsesoyuznogo Mineralogi-
cheskogo Obshchestva 104, 57–61 (in Russian).
ROBERTS, A.C., ANSELL, H.G., & DUNN, P.J. (1981) Coman-
cheite, a new mercury oxychloride-bromide from Terlin-
gua, Texas. Canadian Mineralogist 19, 393–396.
ROBERTS, A.C., COOPER, M.A., HAWTHORNE, F.C., CRIDDLE,
A.J., STIRLING, J.A.R., & DUNNING, G.E. (2002) Tedhad-
leyite, Hg2þHg1þ10O4I2(Cl,Br)2, a new mineral species
from the Clear Creek Claim, San Benito County,
California. Canadian Mineralogist 40, 909–914.
ROBERTS, A.C., COOPER, M.A., HAWTHORNE, F.C., STIRLING,
J.A.R., PAAR, W.H., STANLEY, C.J., DUNNING, G.E., &
BURNS, P.C. (2003) Vasilyevite, (Hg2)2þ10O6I3Br2Cl(CO3),
a new mineral species from the Clear Creek Claim, San
Benito County, California. Canadian Mineralogist 41,
1167–1172.
ROBERTS, A.C., STIRLING, J.A.R., CRIDDLE, A.J., DUNNING,
G.E., & SPRATT, J. (2004) Aurivilliusite, Hg2þHgIþOI, a
new mineral species from the Clear Creek claim, San
Benito County, California, USA . Mineralogical Maga-
zine 68, 241–245.
ROSENBLATT, M. (1956) Remarks on Some Nonparametric
Estimates of a Density Function. The Annals of Mathe-
matical Statistics 27, 832–837.
ROSSUM, G. VAN, WARSAW, B., & COGHLAN, N. (2001) PEP
0008: Style guide for python code. Python.org 1565, 1–25.
ROUSSEEUW, P.J. (1987) Silhouettes: A graphical aid to the
interpretation and validation of cluster analysis. Journal
of Computational and Applied Mathematics 20, 53–65.
S IIDRA , O.I., NAZARCHUK , E.V., AGAKHANOV , A.A., &
POLEKHOVSKY, Y.S. (2019) Aleutite [Cu5 O2 (AsO4 )
(VO4)(Cu0.5A0.5) Cl, a new complex salt-inclusion
mineral with Cu2þ substructure derived from a Kagome-
net. Mineralogical Magazine 83, 847–853.
SIMMONS, W., PEZZOTTA, F., FALSTER, A.U., & WEBBER, K.L.
(2001) Londonite, a new mineral species: The Cs-
dominant analogue of rhodizite from the antandrokomby
granitic pegmatite, Madagascar. Canadian Mineralogist
39, 747–755.
SPIRIDONOV, E.M., GRITSENKO, Y.D., & KULIKOVA, I.M. (2007)
Ferroskutterudite (Fe,Co)As3: A new mineral species
from the dolomite-calcite veins of the Noril’sk ore field.
Doklady Earth Sciences 417, 1278–1280.
STAUDE, S., WERNER, W., MORDHORST, T., WEMMER, K., JACOB,
D.E., & MARKL, G. (2012) Multi-stage Ag-Bi-Co-Ni-U
and Cu-Bi vein mineralization at Wittichen, Schwarz-
wald, SW Germany: Geological setting, ore mineralogy,
and fluid evolution. Mineralium Deposita 47, 251–276.
STRUNZ, H. (1965) Söhngeit, Ga(OH)3, ein neues Mineral.
Naturwissenschaften 52, 493.
Page 27
 28

TEERTSTRA, D.K., ČERNÝ, P., HAWTHORNE, F.C., PIER, J., WANG,
L.M., & EWING, R.C. (1998) Rubicline, a new feldspar
from San Piero in Campo, Elba, Italy. American
Mineralogist 83, 1335–1339.
THOMAS, R., DAVIDSON, P., & HAHN, A. (2008) Ramanite-(Cs)
and ramanite-(Rb): New cesium and rubidium pentaborate
tetrahydrate minerals identified with Raman spectroscopy.
American Mineralogist 93, 1034–1042.
TSCHAUNER, O., MA, C., BECKETT, J.R., PRESCHER, C.,
PRAKAPENKA, V.B., & ROSSMAN, G.R. (2014) Discovery
of bridgmanite, the most abundant mineral in Earth, in a
shocked meteorite. Science 346, 1100–1102.
URUSOV, V.S. (2002) Principle of minimal structural dis-
symmetrization and the influence of rare minerals.
Proceedings of the Russian Academy of Sciences 386,
379–383 (in Russian).
URUSOV, V.S. (2007) Symmetry statistics of mineral species
and the evolutionary dissymmetrization of mineral matter.
Geology of Ore Deposits 49, 497–504.
VASIL’EV, V.I. (1987) Kadyrelite Hg4(Br,Cl)2O — A new
oxyhalide of mercury from the Kadyrelsky ore occur-
rence. Zapiski Vsesoyuznogo Mineralogicheskogo Ob-
shchestva 116, 733–737 (in Russian).
VASIL’EV, V.I., LAVRENT’YEV, Y.G., & PAL’CHIK, N.A. (1983)
Kelyanite, Hg36Sb3(Cl,Br)9O28 — A new mineral.
International Geology Review 25, 864–868 (in Russian).
Zapiski Vsesoyuznogo Mineralogicheskogo Obshchestva
5, 595–598 (in Russian).
VERMEIJ, G.J. & GROSBERG, R.K. (2018) Rarity and persis-
tence. Ecology Letters 21, 3–8.
VOLOSHIN, A.V, PAKHOMOVSKII, Y., BAKHCHISARAITSEV, AY., &
DEVINIA, I.I. (1986) Cesplumtantite — A new cesium-
antimony tantalate from granitic pegmatites. Mineralog-
ical Journal 8(5), 92–98 (in Russian).
WEDEPOHL, K.H. (1995) The composition of the continental
crust. Geochimica et Cosmochimica Acta 59, 1217–1232.
WENK, H.-R. & BULAKH, A. (2004) Minerals. Cambridge
University Press, Cambridge, UK, 668 pp.
YAMONVA, N.A., PUSHCHAROVSKII, D.Y., VYATKIN, S.V., &
KHOMYAKOV, A.P. (1992) Crystal structure of the new
native sulfate-carbonate Na25BaTR2(CO3)11(HCO3)4(SO4)2
F2Cl. Soviet physics. Crystallography 37, 753–756 (in
Russian).
YAROSHEVSKY, A.A. & BULAKH, A.G. (1994) The mineral
composition of the Earth’s crust, mantle, meteorites,
moon, and planets. In Advanced Mineralogy (A.A.
Yaroshevsky & A.G. Bulakh, eds.). Springer, Berlin,
Germany (27–36).
ZUXIANG, Y. (1995) Gaotaiite — A new iridium telluride. Acta
Mineral Sinica 15, 1–4.

VASIL’EV, V.I., LAVRENTEV, Y.G., & PALCHIK, N.A. (1986)

Kuzminite Hg2(Br,Cl)2 — A new natural mercury halide.

View publication stats Page 28