Correlation analysis on partition of rare earth in ion-exchangeable
phase from weathered crust ores
CHI Ru-an(池汝安), DAI Zu-xu(戴祖旭), XU Zhi-gao(徐志高),
WU Yuan-xin(吴元欣), WANG Cun-wen(王存文)
Key Laboratory of Hubei Province for Novel Chemical Reactor and Green Chemical Technology,
Wuhan Institute of Technology, Wuhan 430074, China
Received 2 March 2006; accepted 17 July 2006
Abstract: The rare earth(RE) in weathered crust ores mainly exists as ion-exchangeable phase, approximately 80%. The correlation analysis on partition of 376 samples in ion-exchangeable phase from weathered crust ores was conducted. The results show that partition both among heavy RE elements and light RE elements with high partition appears positive correlation, but partition sums between the heavy RE elements and the light RE elements appear close negative correlation obviously. Clear negative correlations exist between the light RE elements (except Ce) and yttrium(Y). Matrix of correlation analysis on this partition can be divided into three zones. The correlated coefficient variation from negative to positive in zones B and C occurs at Gd, so does that in zones B and A (except Ce, Eu, and Sm), suggesting that RE elements can be divided into two groups with Gd as border. This phenomenon is called Gadolinium-broken effect.
Key words: weathered crust; ore; rare earth partition; correlation analysis
1 Introduction
Weathered crust ores which are widely deposited in Jiangxi, Fujian, Guangdong, Hunan, Yunnan and Guangxi provinces in the south of China[1-3] are the main resources of mid-heavy RE. Their development and utilization have solved the shortage problem of mid-heavy RE for a long time[4]. NESBIT et al[5], PRICE et al[6] and BRAUN et al[7] have researched systematically the theory of RE elements in weathered crust ores since 1960, concluding that RE adsorbed on clay minerals mobilized, fractionated and enriched during adsorption and desorption process. After RE minerals formed by mobilization and fractionation process of clay minerals were found in 1970s, an extraction technique was then explored to recover mid-heavy RE. It is well known that weathered crust rare earth ores, existing as RE aqueous or RE hydroxyl aqueous ions, come from the weathered rock such as bastnasite and ytterbite. The clay minerals are formed by the action of physics, chemistry and biology from lava and granite under warm and humid climate[8-14]. Therefore, exchangeable RE ions adsorbed on clay minerals can be dissolved by cation exchange with ammonium salt, such as ammonium sulfate, and then RE is recovered with oxalic acid[15]. In order to further investigate the rule on partition of exchangeable RE ions adsorbed on clay minerals, the mathematic regression method is applied to describe the partition of RE products extracted from weathered RE ores. In this work, the regular pattern of mobility and fractionation of RE in weathered crust ores is explained and the theory of inorganic chemistry and geochemistry on RE elements is focused.
2 Existing state of RE in weathered ores
RE elements in the weathered crust ores mainly exist as the following four states: water soluble phase, ion-exchangeable phase, colloidal sediment and mineral phase[16-19]. The content of RE in four phases is in the order of ion-exchangeable phase (aqueous or hydroxyl aqueous)＞mineral phase＞colloid sediment?water soluble phase, in which RE mainly exists as ion-exchangeable phase in the ores up to 80%.
Since kaolinite and halloysite are the main clay minerals in the weathered crust rare earth ores (more than 80%), it can be conceived that the rare earth primarily exists as RE aqueous or RE hydroxyl ions adsorbed in clay minerals.
The chemical adsorption model can be expressed as follows[18-21]:
Type of halloysite:
[RE(mH2O, nOH???)](3-n)+[Al4Si4O10(OH)(5-n)]·4H2O
Type of kaolinite:
[RE(mH2O, nOH???)](3-n)+[Al4Si4O10(OH)(5-n)]
For heavy rare earth:
m+n= 6, 7; n=0, 1, 2
For light rare earth:
m+n=6, 7, 8; n=0,1,2
As can be seen from the model, hydroxyl or hydroxyl aqueous rare earth ions can take part in the ion-exchange reaction. This is the reason for recovering RE adsorbed on clay minerals with ammonium salt in industry, which demonstrates that RE product from weathered crust ores is ion-exchangeable RE.
3 Mathematics model on partition of RE
3.1 Correlation analysis of partition among 15 RE elements
Linear correlation analysis can be used to find out the dependency relationship between two random variables. Correlation coefficient is an important index that explains both intensity and direction of the dependency relationship[21,22].
Let x and y are random variables which are assumed values in {
} and {
}, respectively. The Pearson correlation coefficient of x and y is defined as follows[22]:
=
(1)
where 
=
，
=
Generally, the closer
is to 1, the stronger linear correlation between x and y is.
The t-examination is usually used to test significance of the correlation coefficient, and the formula is
=
(2)
where n, n-2 denote specimen capacity and degree of freedom, respectively. When the confidence level α is given, the critical value t(n-2,α) can be obtained from t-table according to degree of freedom (n-2) and α at the same time. If t≥t(n-2,α), rxy is significant, and the linear correlation is thus closed between x and y.
3.2 Regression equation of RE partition
Once linear dependency relationship of random variables is known, a quantitative relationship of variables will be obtained. Regression analysis is a helpful tool for us to establish an equation of variables. The model is[22]
(3)
where 
is the estimated value of y determined by x1, ???, xn, b0 is a constant, and b1,…,bn are partial regression coefficients corresponding to x1, ???, xn. Partial regression coefficient is the ratio of change of dependent variable due to the change of one independent variable with other variables fixed.
R, R2 and 
are the three primary statistic indexes of multiple linear regression, multiple correlation denoting coefficient of determination, and adjusted coefficient of determination, respectively. The definitions are given as follows[22]:
=
(4)
=
(5)
=
(6)
where m and n are specimen capacity and the number of variables, respectively.
The closer R2 is to 1, the more capable the variable interpreting y is. Usually, the percentage of change of y can be illustrated with R2 multiplied by 100%. With the increase of the number of independent variables, the fitting accuracy of regression equation is always overestimated by R2, although the linear correlation between some variables and dependent variable is insignificant.
4 Results and discussion
4.1 Correlation coefficients of RE partition
The Pearson correlation analysis on partitions of 15 RE elements from 376 weathered crust RE samples was conducted by adopting soft ware “SPSS”(statistical package for social science) (version11.0). The results are listed in Table 1.
As can be seen from Table 1, the matrix of correlation analysis on partition of RE in ion-exchangeable phase is divided into three zones. The correlated coefficient variation from negative to positive in zones B and C occurs at Gd, so does that in zones B and A (except Ce, Eu, and Sm). Therefore, RE elements can be divided into two groups by Gd as border. The elements before Gd are called light RE elements and the elements from Gd to Lu, including Y, are called heavy RE elements. This phenomenon is called gadolinium- broken effect, which results from the electron structure of RE ion.
RE elements in the nature exist as third valence in ion compounds (Ln3+) and aqueous or hydroxyl ions adsorbed on clay minerals (except Ce). Consequently, the RE elements mainly manifest the properties of the ions of third valence, which are shown in Table 2.
As shown in Table 2, from La to Gd, the number of the 4f orbit electron is from 1 to 7; while from Gd to Lu, it is from 7 to 14. The non-paired electrons, spectral basic term and colors of RE ions are symmetrical with Gd as border. It also can be seen from Table 1 that the correlation variation between negative and positive appears at Gd. Based on the electron structure of RE, the correlated relationship of RE elements takes symmetrization bounding on Gd as well. That is to say, RE elements can be divided into two groups by Gd, which is called gadolinium-broken effect.
Y without 4f electron is a special element which does not belong to lanthanide series. However, its ion radium, 0.88×10-10m, is interposed between erbium (0.881×10-10m) and thulium (0.869×10-10m), and it usually coexists with lanthanide series in the nature. Therefore, Y, together with scandium, is also called rare earth element. In general, the partition of yttrium is above 10%, even up to 50% in some weathered crust ores such as Longnan ore. Y is regarded a heavy rare earth element because its properties and radium are similar to those of heavy ones, which is also supported by correlated relationship of rare earth partition.
Table 1 Pearson correlated coefficient matrix among 15 kinds of RE partition
Table 2 Properties of RE ions
4.2 Correlation analysis of light and heavy RE groups
Separated by Gd, the elements from La to Eu belong to light RE group, and those from Gd to Y belong to heavy RE group. The statistic vectors of light RE group and heavy RE group are obtained by summing the partition of every sample in the same group, and the scatter graph (Fig.1) and regression equation are as follows.
Fig.1 Scatter graph of light RE and heavy RE
The regression equation is
ΣCeO2=-0.996ΣY2O3+99.954 (7)
The R, R2 and values of the equation are 0.993, 0.986 and 0.986, respectively, which implies the close linear correlation between light RE elements and heavy ones.
4.3 Regression equations of RE element partition
To further clarify quantitative correlation among RE partitions, the regression analysis was done based on the mathematic model in 3.2. Given the partitions of 15 RE elements as dependant variables and the others as independent variables, a multiple linear regression equation can be established using stepwise regression in platform of SPSS (version 11.0), which are as follows:
Y2O3=63.063-0.701Pr6O11-0.597La2O3-0.840Nd2O3 (8)
R=0.953, R2=0.909,=0.908
La2O3=83.339-0.873Y2O3-1.596Sm2O3-
0.900Nd2O3-2.288Dy2O3 (9)
R=0.954, R2=0.911,=0.910
Pr6O11=10.184-9.902×10-2Y2O3-0.415Dy2O3 (10)
R=0.912, R2=0.832,=0.831
Nd2O3=59.732-0.614Y2O3-0.563La2O3-
1.454Dy2O3-1.018Er2O3 (11)
R=0.954, R2=0.910,=0.909
Er2O3=6.904+0.472Yb2O3+0.158Dy2O3-
7.556×10-2La2O3-0.213Gd2O3-8.640× (12)
10-2Nd2O3-5.604×10-2Y2O3
R=0.897, R2=0.805,= 0.802
Dy2O3=1.775-0.262Pr6O11+1.477Tb4O7+0.340Yb2O3+
0.372Gd2O3+0.610Ho2O3 (13)
R=0.936, R2=0.875,= 0.874
Yb2O3=1.083+0.499Er2O3+0.144Dy2O3+1.226Tm2O3-
0.257Eu2O3-2.090×10-2La2O3-0.137Gd2O3 (14)
R=0.886, R2=0.784,= 0.780
Analyzing the above correlations, the relationship of RE partition is obtained. The positive correlation emerges among the heavy RE elements and so does the light ones with high partition.
The RE elements containing high oxide, that is to say, with high partition, appear marked correlation among them, and the RE elements with low partition appear relatively bad correlation, as well as the correlation between heavy RE elements and other RE elements. Especially, the negative correlation is distinguishable between Y and light RE elements except Ce.
5 Conclusions
1) The correlation analysis on partition of RE in weathered crust ores indicates that the partition of heavy RE elements and light ones with high partition appears positive correlation. On the contrary, the apparent negative correlation can be found between the total partition value of light RE elements and heavy ones.
2) The matrix of correlation analysis about partition of RE in ion-exchangeable phase is divided into three zones. The correlation variation from negative to positive in both zone B and C occurs at Gd, so does that in zone B and A (except Ce, Eu, and Sm). Therefore, the RE elements are divided into two groups with Gd as border. The elements before Gd are called light RE elements and the elements from Gd to Lu including Y are called heavy RE elements. This phenomenon is called gadolinium- broken effect.
3) The positive correlation between Y and the heavy RE elements of lanthanides is very close, so is the negative correlation between Y and the light RE elements, which further confirms that Y should be classified into the heavy RE elements group.
References
[1] WANG Z G, YU X Y, ZHAO Z H. Geochemistry of Rare Earth Elements [M]. Beijing: Science Press, 1990. 1-495. (in Chinese)
[2] WU C Y, HUANG D H, GUO Z X. Geochemistry of REE from granite weathering crusts in Longnan of Jiangxi Province [J]. Geology Acta, 1989, 63(4): 349-360. (in Chinese)
[3] LIN C X, ZHENG Z P. Experimental studies on metalogenetic mechanism of weathering crust leaching REE deposit in Southern China [J]. Geochimica, 1994, 23(2): 189-198. (in Chinese)
[4] CHI R A, WANG D Z. Beneficiation of Rare Earth Ore and Extraction Technology [M]. Beijing: Science Press, 1996. 27-241. (in Chinese)
[5] NESBIT H W. Mobility and fractionation of rare earth elements during weathering of a granodiorite [J]. Nature, 1979, 279: 206-210.
[6] PRICE R C, GRAY C M, WILSON R E. The effects of weathering on rare earth element, Y and Ba abundances in Tertiary basalts from southeastern Australia [J]. Chemical Geology, 1991, 93: 245-265.
[7] BRAUN I J, PAGEL M, HERBILLA A. Mobilization and redistribution of REEs and thorium in a sysnitic latertic profile: A mass balance study [J]. Geochim Cosmochim Acta, 1993, 57: 4419-4434.
[8] CHI R A, HE P J, ZHU Y J. REE mobility and enrichment in clay minerals [J]. Chinese Rare Earths, 1992, 13(5): 67-71. (in Chinese)
[9] YANG Z M. Clay mineral in the REE-weathering crust, Longnan, Jiangxi [J]. Science Geologica Sinica, 1978, 22(1): 70-74. (in Chinese)
[10] YANG X M, YANG X Y, ZHANG P S, YANG X M, YANG X Y. Geochemistry of REE from granite weathering crusts in Dajishan of Jiangxi Province [J]. Chinese Rare Earths, 1999, 20(1): 1-5. (in Chinese)
[11] CHI R A, XU J M, HE P J, ZHU Y J. REE geochemistry of granitoid weathering crust and properties of ores in Southern China [J]. Geochimica, 1995, 24(3): 261-269. (in Chinese)
[12] DUDDY I R. Redistribution and fractionation of rare earth and other elements in a weathering profile [J]. Chem Geol, 1980, 30: 363-381.
[13] PRUDENCIO M I, BRAGA M A S, GOUVEIA M A. REE mobilization, fractionation and precipitation during weathering of basalts [J]. Chemical Geology, 1993, 107: 251-254.
[14] CHI R A, WANG D Z. Study of adsorption properties and enriching RE on clay minerals by quantum chemical calculation [J]. Journal of Rare Earth, 1993, 11(2): 142-147.
[15] CHI R A. Extraction of rare earths from a low-grade, kaolinitic ore by percolation leaching [A]. BAUTISTA R G., WONG M M. Rare Earths Extraction Preparation and Application [C]. Reno, Nevads, TMS. 1989, 227-236.
[16] CHI R A, ZHU G C. Rare earth partitioning of granitoid weathering crust in southern China [J]. Trans Nonferrous Met Soc China, 1998, 8(4): 693-699.
[17] CHI R A, ZHU G C, ZHANG P, XU Y H. Chemical behavior of cerium element in rock weathering system [J]. Trans Nonferrous Met Soc China, 1999, 9(1): 158-164.
[18] CHI R A, LI Z J, PENG C, ZHU G C, XU S M. Partitioning properties of rare earth ores in China [J]. Rare Metals, 2005, 24(3): 205-209.
[19] SUN C X, WANG S J, JI H B. Formation mechanism of the superhigh concentration of REE and the strong negative Ce anomalies in the carbonate rock weathering profiles in Guizhou Province, China [J]. Geochimica, 2002, 31(2): 119-128. (in Chinese)
[20] CHI R A, TIAN J, LI Z J, PENG C, WU Y X, LI S R, WANG C W, ZHOU Z A. Existing state and partitioning of rare earth on weathered ores [J]. Chinese Journal of Rare Earth, 2005, 23(6): 756-759.
[21] GUO Z G. Statistical Package for Social Science—Usage of SPSS Soft [M]. Beijing: Renmin University of China Press, 1999. 5. (in Chinese)
[22] LU W D. SPSS for Windows Statistical Analysis (Second edition) [M]. Beijing: Electron Industry Press, 2002. 1. (in Chinese)
(Edited by YANG Bing)
Foundation item: Project(59725408) supported by the National Outsanding Younger Scholar Foundation of China; Projects(59674021, 50474022) supported by the National Natural Science Foundation of China
Corresponding author: CHI Ru-an; Tel/Fax: +86-27-87194500; E-mail: rac@mail.wit.edu.cn
