基于机器学习的锆石成因分类研究

锆石是在自然界中多种温压条件下能够稳定保存,并记录原岩年龄信息的副矿物。锆石微量元素能完整记录地质演化过程信息。通过微量元素分析锆石成因的研究已久,通常利用Th-U图解和La_N-(Sm/La)_N图解等二元图解对锆石进行分类研究。然而,随着锆石研究的深入,以及二元图解无法呈现数据高维度信息的局限性,传统图解已经不能满足对锆石类型进行准确判别,且对已知类型的锆石出现判定偏差。因此,本文将地质大数据与机器学习相结合,训练出高维度锆石成因分类器。文中收集了3 498条不同成因类型的锆石微量元素数据,并通过测试和运用随机森林、支持向量机、人工神经网络和k近邻等4种机器学习算法,最终得出准确率为86.8%的线性支持向量机锆石成因分类器,用于锆石类型的判定与预测。这项工作为锆石分类研究提供了更高维度的判别手段,极大提高了微量元素分析成因结果的精度。将锆石微量元素数据与机器学习方法相结合,是大数据分析与机器学习技术在地球化学研究中的积极探索。     

磷灰石Eu/Y-Ce:基于大数据的源区类型判别新图解SCIEI北大核心CSCD

磷灰石广泛分布于火成岩、沉积岩和变质岩中,是一种常见的、包含丰富微量元素的副矿物。磷灰石晶格可容纳丰富的微量元素,且因其形成的物理化学条件不同会表现出差异明显的微量元素特征。利用磷灰石微量元素特征可以追踪物质来源和演化。现在常用的方法是利用磷灰石的微量元素绘制二元判别图解,经典判别图解包括Sr-Y、Sr-Mn、Y-(Eu/Eu^(*))和(Ce/Yb)_(N)-REE图解。随着微区测试技术发展,磷灰石微量元素数据日渐丰富,同时由于磷灰石化学成分的复杂性,传统图解已逐渐无法有效利用这些数据所携带的信息,进而无法准确判别其生成环境。建立能准确判别磷灰石物源的新型判别图解故而迫切。近年来,磷灰石微量元素数据的大量积累,为运用以大数据为依托,准确判别磷灰石物源奠定了数据基础。本研究将大数据技术与地球化学数据相结合,共收集整理了1925个代表性磷灰石测试点的微量元素数据,对富碱性火成岩、超镁铁质岩石、镁铁质火成岩、长英质花岗岩、中-低级变质岩、高级变质岩六种类型中磷灰石微量元素数据进行穷举端元处理,共获得7140个磷灰石物源判别图解端元组合,在轮廓系数限定下,进一步有效筛选并提取出能判别磷灰石物源类型的最优图解端元。本文构建了Eu/Y-Ce磷灰石判别新图解,相较于之前的磷灰石判别图解,其涵盖了更全面的物源类型,可以更准确地判别源区类型。     

正、副角闪岩判别标志的新认识--赣中前寒武纪正、副角闪岩的矿物-岩石地球化学特征对比研究

通过对赣中相山地区前寒武纪正、副角闪岩矿物学、岩石-地球化学特征系统对比研究,得出角闪岩成因判别标志的几点新认识:①提出了识别正、副角闪岩的角闪石矿物学标志;②阐明了区分正、副角闪岩的稀土元素地球化学特征;③对角闪岩微量元素地球化学特征提出采用模糊聚类分析与蛛网图相结合进行成因判别的有效方法;④提出综合应用15种元素对的比值(Ti/V,Ce/Yb,La/Yb,Th/Yb,La/Sm,La/Nb,Ce/Y,Rb/Sr,Nb/Y,Sm/Nd,Sr/Y,Sr/Ba,Zr/Y,Ti/Zr,Ti/Y)作为判别角闪岩成因的标志.     

锆石微量元素对成矿岩体的判别——来自大数据思维的应用

华南钦杭结合带燕山期岩浆活动异常活跃,且具有较明显的成矿专属性。近年来微区测试技术日益成熟,积累了大量锆石微量数据。通过全体数据挖掘的思维方法,对前人发表的数据进行了进一步数据挖掘,利用锆石稀土元素对岩体成矿潜力进行判别,探讨有效的找矿地球化学标志。利用Python语言编程,对采用的13种稀土元素及元素比值进行穷举式组合,获得了4095个二元图解及121485个三元图解,并设计筛选算法,自动筛选出能有效区分锆石母岩成矿类型的图解。结果表明,锆石稀土元素含量及比值图解对不同成矿类型岩体的区分程度各异:与Ce、Eu有关的地球化学指标可以较清晰地对斑岩铜矿和钨锡(锡)矿床进行判别,这可能与岩体的氧逸度和含水量有关。此外,还挖掘出一些新的元素组合图解,如Dy/Lu-Er/Lu、Gd/Dy-Er/Yb等,可以有效区分岩体成矿类型,其隐含的地球化学机制尚待进一步解释。地球化学数据挖掘结果可以作为找矿标志使用,为华南燕山期岩浆-热液矿床研究及找矿勘查提供了科学依据,也是大数据技术在矿床学方面应用研究的积极探索。     

运用机器学习和锆石微量元素构建花岗岩成矿潜力判别图解：以东昆仑祁漫塔格为例

由于锆石在中酸性岩中广泛存在且成分稳定、不易受到后期热液活动的扰动,因此锆石成分可以有效记录成矿岩浆信息。其中,锆石的Ce4+/Ce3+、Ce/Ce*、Eu/Eu*和Ce/Nd值可以反映岩浆氧逸度和含水量等成矿信息,已被广泛用于花岗岩类成矿潜力评价。然而,随着研究的深入发现,这些地球化学指标并不完全具有普适性。此外,以往研究均是根据对成矿岩体的“已知认识”提出成矿潜力判别方法,但考虑到成矿过程的复杂性,许多反映岩浆成矿能力的地球化学信息可能均尚未被揭露。为此,笔者以东昆仑祁漫塔格成矿带为例,借助当前广泛应用的机器学习算法之一——支持向量机,对来自该成矿带斑岩−矽卡岩Cu−Fe−Pb−Zn多金属矿床成矿岩体和全球非成矿岩体的锆石数据开展机器学习训练,目的在于挖掘能够反映岩浆成矿能力的锆石微量元素特征,从而构建花岗岩成矿潜力判别图解。模型训练结果显示,在21个常见的锆石微量元素特征中,Gd、Dy、Yb、Y、Tm等5种元素特征对识别岩浆成矿能力最为重要。在此基础上,笔者新建立了10个二元判别图解,它们在识别成矿岩体和非成矿岩体时的准确率均接近1。研究表明,利用机器学习方法和地质大数据,可以挖掘传统研究方法难以发现的新的地球化学指标和图解,这对深入认识矿床成因、指导找矿勘查具有重要意义。     

镁铁质-超镁铁质岩浆岩中单斜辉石的智能分析研究

区分不同构造环境中岩浆作用是认识地幔中岩浆形成过程的基本手段。目前较为成熟的是利用全岩去区分判别,而利用造岩矿物去判别构造环境、演绎岩浆演化的研究还不够深入。本文运用机器学习的方法,以全球新生代大洋中脊、洋岛以及岛弧构造背景中的镁铁质-超镁铁质岩浆岩中单斜辉石的地球化学数据为研究对象,试图区分这3 种不同构造环境的单斜辉石。通过机器学习方法中K-邻近（K-Nearest Neighbor,KNN）和随机森林（Random Forest,RF）的计算和比较,认为RF 是一种有效的地球化学区分方法,它的结果不仅可用来判别构造环境,同时还能够提取特征元素。同时我们发现,在镁铁质-超镁铁质岩浆岩单斜辉石构造环境判别图解中,Rb、La、Ba、Cr、Sr、Yb、V、Ti、Nd、Eu、Gd等微量元素具有较高的贡献率,而主量元素贡献率较低。在此基础上,我们结合前人的对单斜辉石的构造环境判别图的研究成果,提出几个判别效果较好的判别图解。但是整个研究由于缺少进一步可视化的成果,限制了机器学习方法的推广,这也是今后需要进一步研究的课题。     

内蒙古浩布高铅锌矿床闪锌矿微量元素特征及其地质意义

内蒙古浩布高铅锌矿床位于大兴安岭南段黄岗-甘珠尔庙成矿带, 该带是我国北方重要的铅锌多金属成矿带之一。目前, 浩布高矿床闪锌矿微量元素赋存机制尚不清晰, 矿床成因类型亦存在一定的争议。本研究采用激光剥蚀电感耦合等离子体质谱仪(LA-ICP-MS)原位微量元素分析手段, 结合机器学习方法, 对浩布高闪锌矿微量元素组成特征、赋存机制以及矿床成因类型进行了探讨。LA-ICP-MS分析结果表明, 浩布高矿床中闪锌矿以富集Fe、Mn、Co、Cu、Se、Ag、Cd、In、Sn等元素, 贫Ni、Ga、Ge、As、Mo、Sb、Au、Tl、Pb、Bi等元素为特征。其中, Fe、Mn、In等元素主要以类质同象的方式替代Zn赋存于闪锌矿中, Cu、Ag、Sn等元素含量变化范围较大, 可能部分以微粒包裹体的形式存在。In与Cu含量具有一定的正相关, 推测In在闪锌矿中富集机理可能为Cu⁺+In³⁺↔2Zn²⁺。Cd与Fe含量呈一定的正相关性, 而与Zn含量呈负相关性, 这可能暗示在闪锌矿中Cd主要以类质同象的方式置换Zn元素而非Fe元素。通过穷举闪锌矿微量元素图解发现, 即使得分最高的Co/Ag-Mn图解依然存在较大范围的重叠区域, 因此不能简单地用闪锌矿微量元素二元图解来判别矿床类型。通过测试四种机器学习中经典的不同核函数的支持向量机算法, 最终得出准确率为91.5%的高斯内核支持向量机闪锌矿成矿类型分类器, 可以用于多种矿床类型闪锌矿的研究。浩布高矿床中闪锌矿微量元素组成明显不同于密西西比河谷型(MVT)、沉积喷流型(SEDEX)、火山块状硫化物型(VMS)和浅成低温热液型铅锌矿床, 而与矽卡岩型接近。综合已有矿物学、年代学以及本研究获得的闪锌矿地球化学证据表明, 浩布高应是一个典型的矽卡岩型铅锌矿床。

     

四川底苏沉积改造层控型铅锌矿床地球化学初步研究

本文对底苏铅锌矿床微量元素地球化学、稀土元素地球化学、成矿热液地球化学性质及稳定同位素地球化学等进行了研究，并提供了首批研究数据，同时首次对该类型矿床的成因提出了一定认识。     

四川底苏沉积改造层控型铅锌矿床地球化学初步研究

本文对底苏铅锌矿床微量元素地球化学、稀土元素地球化学、成矿热液地球化学性质及稳定同位素地球化学等进行了研究，并提供了首批研究数据，同时首次对该类型矿床的成因提出了定认识。     

磁铁矿元素地球化学大数据构建及其在矿床成因分类中的应用

磁铁矿广泛分布在岩浆、热液及沉积等各类矿床中,其地球化学元素组成往往受温度、氧逸度等物理化学条件的影响,能反映矿床形成环境并指示矿床成因类型,是一种重要的勘查指示矿物.自20世纪60年代以来,磁铁矿的主微量元素数据被用来构建不同的判别图,试图来区分矿床的成因类型.然而,由于矿床成因类型的多样性以及同一类型矿床的磁铁矿的...     

Discrimination of Pb-Zn deposit types using sphalerite geochemistry: New insights from machine learning algorithm

Due to the combined influences such as ore-forming temperature, fluid and metal sources, sphalerite tends to incorporate diverse contents of trace elements during the formation of different types of Lead-zinc (Pb-Zn) deposits. Therefore, trace elements in sphalerite have long been utilized to distinguish Pb-Zn deposit types. However, previous discriminant diagrams usually contain two or three dimensions, which are limited to revealing the complicated interrelations between trace elements of sphalerite and the types of Pb-Zn deposits. In this study, we aim to prove that the sphalerite trace elements can be used to classify the Pb-Zn deposit types and extract key factors from sphalerite trace elements that can discriminate Pb-Zn deposit types using machine learning algorithms. A dataset of nearly 3600 sphalerite spot analyses from 95 Pb-Zn deposits worldwide determined by LA-ICP-MS was compiled from peer-reviewed publications, containing 12 elements (Mn, Fe, Co, Cu, Ga, Ge, Ag, Cd, In, Sn, Sb, and Pb) from 5 types, including Sedimentary Exhalative (SEDEX), Mississippi Valley Type (MVT), Volcanic Massive Sulfide (VMS), skarn, and epithermal deposits. Random Forests (RF) is applied to the data processing and the results show that trace elements of sphalerite can successfully discriminate different types of Pb-Zn deposits except for VMS deposits, most of which are falsely distinguished as skarn and epithermal types. To further discriminate VMS deposits, future studies could focus on enlarging the capacity of VMS deposits in datasets and applying other geological factors along with sphalerite trace elements when constructing the classification model. RF’s feature importance and permutation feature importance were adopted to evaluate the element significance for classification. Besides, a visualized tool, t-distributed stochastic neighbor embedding (t-SNE), was used to verify the results of both classification and evaluation. The results presented here show that Mn, Co, and Ge display significant impacts on classification of Pb-Zn deposits and In, Ga, Sn, Cd, and Fe also have relatively important effects compared to the rest elements, confirming that Pb-Zn deposits discrimination is mainly controlled by multi-elements in sphalerite. Our study hence shows that machine learning algorithm can provide new insights into conventional geochemical analyses, inspiring future research on constructing classification models of mineral deposits using mineral geochemistry data.     

Distribution and petrogenetic behaviour of trace elements in granitic pegmatite quartz from South Norway

The present study documents that the trace-element distribution in granitic quartz is highly sensitive to CAFC processes in granitic melts. Igneous quartz efficiently records both the origin and the evolution of the granitic pegmatites. Aluminium, P, Li, Ti, Ge and Na in that order of abundance, comprises >95% of the trace elements. Most samples feature >1 ppm of any of these elements. The remnant 5% includes K, Fe, Be, B, Ba and Sr whereas the other elements are present at concentrations lower than the detection limit. Potassium, Fe, Be and Ti are relatively compatible hence obtain the highest concentrations in early formed quartz. Phosphorous, Ge, Li and Al are relatively incompatible and generally obtain the highest concentrations in quartz that formed at lower temperatures from more evolved granitic melts. The Ge/Ti, the Ge/Be, the P/Ge and the P/Be ratios of quartz are strongly sensitive to the origin and evolution of the granitic melts and similarly the Rb/Sr and the Rb/K ratios of K-feldspars may be utilised in petrogenetic interpretations. However, the quartz trace element ratios are better at distinguishing similarities and differences in the origin and evolution of granitic melts. After evaluating the different trace element ratios, the Ge/Ti ratio appears to be most robust during subsolidus processes in the igneous systems, hence probably should be the preferred ratio for analysing and understanding petrogenetic processes in granitic igneous rocks.Editorial responsibility: J. Hoefs     

基于PCA-SVM算法对稀土元素与稀土判别指标耦合数据集的铀矿床分类

不同类型铀矿床的沥青铀矿/晶质铀矿具有不同的稀土元素组成,其组成可作为判别铀矿床类型的重要指标。采用基于Python语言的主成分分析（principal component analysis,PCA）与支持向量机（support vector machines,SVM）结合的分类模型,对收集到的全球已知6种类型铀矿床的216组沥青铀矿/晶质铀矿稀土元素数据进行研究。以216组数据为训练集,通过数据清洗、特征缩放、PCA特征提取、网格搜索和交叉验证参数寻优构建SVM分类模型,对24组同变质型胡家峪晶质铀矿进行智能识别。研究结果显示：仅使用稀土元素的14维训练集最优模型判定胡家峪晶质铀矿类型的测试准确率为0.4%;由稀土元素、稀土总量、轻重稀土比、铕异常组成的17维训练集最优模型的测试准确率为75.0%,较14维训练集提高74.6%,模型泛化能力强;而通过传统稀土元素配分曲线、w（ΣREE）-（LREE/HREE）_N图解不能有效判定胡家峪晶质铀矿类型。本次研究表明,PCA-SVM算法对增有传统稀土判别指标数据集进行挖掘可有效厘定铀氧化物成因类型,效果明显优于单纯的稀土元素数据集以及传统的稀土配分曲线、w（ΣREE）-（LREE/HREE）_N图解。     

三类构造背景辉长岩单斜辉石主量元素和微量元素的数据分析研究

判别岩浆岩产出的构造环境已经成为岩石学、地球化学及其地球动力学研究的重要内容。作为岩浆岩中的一种喷出岩,玄武岩被视为判别构造环境的最佳成员。对其中单斜辉石的研究,由于其数据本身的利用程度有限而效果欠佳。理论上,不同构造环境的辉长岩也会存在一定差异。为此,利用机器学习算法研究全球新生代辉长岩的单斜辉石势在必行。本文主要针对岛弧(IAB)、洋岛(OIB)及大洋中脊(MORB)3种构造背景辉长岩的单斜辉石进行特征筛选和数据分类。从GEOROC数据库中,经数据收集与清洗,我们分别获得岛弧辉长岩单斜辉石数据385条,洋岛辉长岩单斜辉石数据756条,大洋中脊辉长岩单斜辉石数据5 500条。其中绝大部分为主量元素数据,其余为微量元素数据。在特征提取部分,我们选用卡方检验判断特征独立性,F检验估计两个随机变量之间的线性依赖程度,互信息法捕获其他种类的统计相关性。3种检验方法互相印证,得出了统计学可靠的重要分类特征。在数据分类过程中,本文对比了K-近邻、决策树和支持向量机3种主流机器学习分类算法在辉长岩数据上的表现。研究表明,对于上述3种构造背景,Al2O3、TiO2为最有区分度的辉长岩单斜辉石主量元素成分,Sr为最有区分度的微量元素成分。另外,对于3种构造背景的辉长岩单斜辉石主量元素和微量元素数据,机器学习模型分类准确率均达94%。     

鄂东南付家山脉石英矿杂质特征及其用作高纯石英原料的潜力

作为战略性矿产资源之一,高纯石英已广泛应用于集成电路、半导体芯片、太阳能等高新技术产业中,但是能够生产高纯石英的原料矿床极为稀缺,我国尤为紧缺高纯石英原料矿。鄂东南地区是湖北省脉石英矿床的主要分布区。本文针对鄂东南付家山脉石英矿床,通过光学显微镜、扫描电子显微镜观察了脉石英的脉石矿物类型和包裹体特征,采用电感耦合等离子发射光谱法(ICP-OES)对原矿进行了微量元素分析,旨在获得付家山脉石英矿床的杂质元素特征,进而评价矿床用作高纯石英原料的潜力。结果表明,付家山脉石英矿石SiO₂含量大于99.95%,杂质元素主要为Al、K、Fe、Ti、Ca等,脉石矿物主要有白云母、钾长石、铁氧化物等,流体包裹体较为发育。杂质元素分析结果表明,付家山脉石英原矿质量达到低端高纯石英标准,经传统工艺提纯后,可能具有生产中高端高纯石英的潜力。     

Trace element chemistry of pyrite: A useful guide to the occurrence of sulfide base metal mineralization

Trace element contents for pyrite from a range of sulfide mineral occurrences in the Kangiara region, eastern Australia, illustrate two main groups of pyrite. The first group, with higher Ag, Cu, Pb and Mo contents, corresponds to samples from sulfide base metal deposits and the second group, with higher Mn, Ti and Ni contents, contains samples from skarn mineralization, volcanic rocks and quartz veins. The model proposed for the development of pyrite in the Kangiara region is that the first group was formed from base metal-bearing solutions, while the second group reflects diagenetic pyrite and metamorphic pyrite. Thus, the pyrite trace element chemistry may provide a means of distinguishing types of mineral occurrences, in particular, those containing significant base metal mineralization.     

In-situ LA-ICP-MS trace elemental analyses of magnetite: Fe–Ti–(V) oxide-bearing mafic–ultramafic layered intrusions of the Emeishan Large Igneous Province,SW China

The Taihe, Baima, Hongge, Panzhihua and Anyi intrusions of the Emeishan Large Igneous Province (ELIP), SW China, contain large magmatic Fe–Ti–(V) oxide ore deposits. Magnetites from these intrusions have extensive trellis or sandwich exsolution lamellae of ilmenite and spinel. Regular electron microprobe analyses are insufficient to obtain the primary compositions of such magnetites. Instead, laser ablation ICP-MS uses large spot sizes (~ 40 μm) and can produce reliable data for magnetites with exsolution lamellae. Although magnetites from these deposits have variable trace element contents, they have similar multi-element variation patterns. Primary controls of trace element variations of magnetite in these deposits include crystallography in terms of the affinity of the ionic radius and the overall charge balance, oxygen fugacity, magma composition and coexisting minerals. Early deposition of chromite or Cr-magnetite can greatly deplete magmas in Cr and thus Cr-poor magnetite crystallized from such magmas. Co-crystallizing minerals, olivine, pyroxenes, plagioclase and apatite, have little influence on trace element contents of magnetite because elements compatible in magnetite are incompatible in these silicate and phosphate minerals. Low contents and bi-modal distribution of the highly compatible trace elements such as V and Cr in magnetite from Fe–Ti oxide ores of the ELIP suggest that magnetite may not form from fractional crystallization, but from relatively homogeneous Fe-rich melts. QUILF equilibrium modeling further indicates that the parental magmas of the Panzhihua and Baima intrusions had high oxygen fugacities and thus crystallized massive and/or net-textured Fe–Ti oxide ores at the bottom of the intrusive bodies. Magnetite of the Taihe, Hongge and Anyi intrusions, on the other hand, crystallized under relatively low oxygen fugacities and, therefore, formed net-textured and/or disseminated Fe–Ti oxides after a lengthy period of silicate fractionation. Plots of Ge vs. Ga + Co can be used as a discrimination diagram to differentiate magnetite of Fe–Ti–(V) oxide-bearing layered intrusions in the ELIP from that of massif anorthosites and magmatic Cu–Ni sulfide deposits. Variable amounts of trace elements of magmatic magnetites from Fe–Ti–(P) oxide ores of the Damiao anorthosite massif (North China) and from Cu–Ni sulfide deposits of Sudbury (Canada) and Huangshandong (northwest China) demonstrate the primary control of magma compositions on major and trace element contents of magnetite.