index.html

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=Edge" />
<meta property="og:type" content="website" />
<meta property="og:title" content="Materials Platform for Data Science" />
<meta property="og:site_name" content="MPDS.IO"/>
<meta property="og:description" content="Online PAULING FILE materials database with nearly 2 million data entries (physical properties, crystal structures, phase diagrams) available via API, ready for modern data-intensive applications." />
<meta property="og:image" content="https://mpds.io/wm_logo_opt.png" />
<meta property="og:image:type" content="image/png" />
<meta property="og:image:width" content="225" />
<meta property="og:image:height" content="224" />
<meta property="og:url" content="https://mpds.io">
<meta name="twitter:card" content="summary" />
<meta name="twitter:site" content="@mpdsio" />
<meta name="twitter:title" content="Materials Platform for Data Science" />
<meta name="twitter:creator" content="@tildepro" />
<meta name="twitter:image" content="https://mpds.io/wm_logo_opt.png" />
<meta name="twitter:image:alt" content="Online materials data-mining platform" />
<meta name="twitter:description" content="Online PAULING FILE materials database with nearly 2 million data entries (physical properties, crystal structures, phase diagrams) available via API, ready for modern data-intensive applications." />
<meta name="description" content="Online PAULING FILE materials database with nearly 2 million data entries (physical properties, crystal structures, phase diagrams) available via API, ready for modern data-intensive applications." />
<meta name="keywords" content="materials platform, physical properties, phase diagrams, crystalline structures, materials database, PAULING FILE" />
<meta name="subject" content="materials informatics" />
<meta name="url" content="https://mpds.io" />
<meta name="rating" content="General" />
<meta name="author" content="feedback@tilde.pro" />
<meta name="robots" content="notranslate" />
<title>MPDS API docs</title>
<style>
/* COMMON */

@font-face {
font-family:Exo2;
src: url('data:application/x-font-woff;base64,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') format('woff');
}

*{font-family:Exo2,Georgia;margin:0;padding:0;}
html, body{height:100%;-webkit-text-size-adjust:100%;text-size-adjust:100%;}
sup, sub{font-size:12px;}
body.sky{background:repeating-linear-gradient(60deg, #fff, #fff 44%, rgba(255,255,255,0) 44%, rgba(255,255,255,0) 78%), linear-gradient(to bottom, #DBF2FF 0%, #fff 6%);background-attachment:fixed;}
#overlay{display:none;position:fixed;left:0;top:0;width:100%;height:100%;text-align:center;z-index:100;background:#fff;opacity:0.75;overflow:hidden;}

/* COMMON NAVBAR START */

#nav{position:fixed;top:0;left:0;height:34px;z-index:101;background:#fff;font-size:0.9em;padding:0 3px;border-bottom-right-radius:4px;line-height:33px;border-bottom:1px solid #888;border-right:1px solid #888;cursor:pointer;box-shadow:6px 6px 4px -3px rgba(0,0,0,0.4);}
#nav li{list-style:none;float:left;font-weight:bold;margin:0 7px;}
#nav a{text-decoration:none;color:#555;}
#nav a:hover{border-bottom:1px solid #000;}
li.nav_fluid{display:none;}

/* COMMON NAVBAR END */

#menu{display:none;position:fixed;left:0;top:170px;width:12.5%;height:auto;z-index:100;}
#menu li{text-align:left;list-style:none;letter-spacing:-0.5px;}
#menu a{display:block;}
#menu > ul > li > a{text-decoration:none;color:#900;border-bottom:1px solid #900;padding:4px 0;font-size:2vmin;padding-left:5px;line-height:2.5vmin;}
#menu > ul > li > a:hover, #menu > ul > li > a.active {color:#fff;background:#900;}
#menu > ul > li > ul > li > a{text-decoration:none;color:#222;border-bottom:1px solid #222;padding:4px 0;font-size:1.8vmin;padding-left:18px;line-height:2.3vmin;}
#menu > ul > li > ul > li > a:hover, #menu > ul > li > ul > li > a.active {color:#fff;background:#222;}

#bookmarks{display:none;position:fixed;right:0;top:171px;width:12.5%;height:auto;z-index:100;}
#bookmarks li{text-align:left;list-style:none;letter-spacing:-0.5px;}
#bookmarks a{display:block;}
#bookmarks > ul > li:first-child {border-top:1px solid #222;}
#bookmarks > ul > li > a{text-decoration:none;color:#222;border-bottom:1px solid #222;font-size:0.9em;padding:3px 1px;line-height:2.3vmin;}
#bookmarks > ul > li > a:hover{color:#fff;background:#222;}

div.modal_dialogue{position:fixed;top:10%;left:50%;border-radius:4px;border:1px solid #999;box-shadow:15px 15px 10px -5px rgba(0,0,0,0.4);}
div.cross{float:right;width:16px;height:16px;font-size:10px;text-align:center;cursor:pointer;background:url(data:image/gif;base64,R0lGODlhEAAQAJEAAAAAAP///1VVVf///yH5BAEAAAMALAAAAAAQABAAAAIjnI+py6L+BDRy0gqxHVqvLjGcJ3Yl+IEhQpLX+m7wMYv2PRQAOw==) no-repeat;}
span.href{cursor:pointer;text-decoration:none;color:#3e3f95;border-bottom:1px solid #3e3f95;}

#response{white-space:pre;}
#logo_bg{display:none;width:250px;height:250px;position:fixed;z-index:-1;right:5px;bottom:20px;background:url(https://tutorial.mpds.io/mpds.png) left top no-repeat;background-size:100%;opacity:0.1;}
#notifybox{z-index:201;position:fixed;top:12px;right:4px;min-width:150px;background:#fc6;border-radius:5px;padding:20px 25px;font-size:1.2em;line-height:1.45em;}
#main{width:100%;min-height:100%;height:auto !important;height:100%;margin-bottom:-33px;}
h1{font-size:100%;font-weight:normal;display:inline;}
h2, h3{line-height:2.2em;font-size:1.6em;font-weight:bold;color:#900;}
h2:first-letter , h3:first-letter {color:#222;}
h3{line-height:2.2em;font-size:1.2em;}

div.spacer{width:790px;margin:0 auto;height:1px;border-top:1px solid #ccc;}

#content{width:790px;margin:0 auto;background:#fff;padding:0 15px 40px;border-radius:15px;}
#content > p{font-size:1.05em;line-height:1.7em;color:#666;margin:18px 0;padding:20px 6px;border-radius:4px;background:#f9f9f9;}
#content > p:first-letter {font-size:1.6em;font-weight:bold;}
#content a{color:#222;text-decoration:none;border-bottom:1px solid #222;}
#content > table{width:auto;margin:20px 0;font-size:0.9em;text-align:left;color:#666;border-collapse:collapse;border-spacing:0;}
#content > table td{line-height:2em;border:1px solid #ddd;padding:4px 2px;}
#content > table th{line-height:2.2em;border:1px solid #ddd;background:#f6f6f6;padding:4px 2px;font-size:1.1em;}
span.t{font-family:'Courier New',Courier;background:#aaa;color:#fff;padding:1px 2px;}

div.blackbg{background:#2b2b2b;font-size:0.9em;color:#fff;font-family:'Courier New',Courier;line-height:1.6em;margin:20px 0;padding:10px 5px;border-radius:4px;white-space:pre;}
div.caption{text-align:center;font-size:0.9em;line-height:2em;color:#666;margin:20px 0 30px;}
div.caption b{font-style:normal;}

/* LOGOS */

#landing_box{width:590px;margin:0 auto;height:170px;padding:20px 0;}
#logo_l{float:left;width:230px;margin-left:15px;font-size:3em;text-align:right;padding-top:38px;letter-spacing:24px;color:#555;font-weight:normal;}
#logo{float:left;width:140px;height:130px;background:url(https://tutorial.mpds.io/mpds.png) center center no-repeat;background-size:85%;}
body.darkmode--activated #logo{filter:invert(100%);}
#logo a{display:block;width:100%;height:100%;border:0;}
#logo_r{float:left;width:175px;margin-left:15px;font-size:1.2em;line-height:1.3em;text-align:left;font-weight:normal;padding-top:42px;color:#555;}
#motto{clear:both;width:100%;height:33px;text-align:left;padding-top:20px;padding-left:80px;letter-spacing:0.5px;font-size:0.95em;color:#333;}
#motto a{text-decoration:none;color:#333;border-bottom:1px solid #333;}

div.darkmode-layer{z-index:9999;}

/* FOOTER */
/* HEIGHT 32px */

#footer{width:100%;height:24px;padding-top:8px;text-align:center;font-size:0.85em;color:#333;background:#f6f6f6;border-top:1px solid #333;}
#footer a, #footer span{text-decoration:none;color:#333;border-bottom:1px solid #333;cursor:pointer;}

/* MEDIA-QUERIES */

@media all and (min-width:1100px){
#menu, #bookmarks{display:block;}
li.nav_fluid {display:inline-block;}
}

/* DOCSON */

.docson {
font-family: 'Droid Sans',verdana,helvetica;font-size: 13px;padding: 6px 6px 6px 6px;
}
.docson .box {
position: relative;float: left;background-color:rgba(255, 255, 255, 0.2);border: 1px solid lightgrey;border-radius: 4px;-moz-border-radius: 4px;-webkit-border-radius: 4px;padding-bottom: 10px;-moz-box-shadow:    1px 1px 1px darkgray;-webkit-box-shadow: 1px 1px 1px darkgray;box-shadow:         1px 1px 1px darkgray;
}
.docson .expand-button {
position: absolute;margin-top: 15px;padding: 2px;right: 2px;float: right;width: 18px;text-align: center;cursor: pointer;font-weight: bold;font-family: "Lucida Console", Monaco, monospace;font-size: 15px;color: darkgray;display: none;
}
.docson .source-button {
position: absolute;padding: 2px;margin-top: 3px;right: 2px;width: 18px;text-align: center;cursor: pointer;font-weight: bold;font-family: "Lucida Console", Monaco, monospace;font-size: 10px;color: darkgray;display: none;
}
.docson .box-header {
padding: 6px;border-top-left-radius: 4px;border-top-right-radius: 4px;border-bottom: 1px solid lightgrey;background: whitesmoke;
}
.docson .box-name {
color: darkblue;padding: 5px 40px 2px 3px;float: left;
}
.docson .title {
margin-top: -3px;font-size: 120%;
}
.docson .link {
text-decoration: underline;
}
.docson .box-description {
color: dimgray;float: left;max-width: 600px;
}
.docson .end {
clear: both;
}
.docson .error {
color: red;
}
.docson .box-body {
padding-left: 10px;padding-right: 28px;clear: both;
}
.docson .box-1 {
background: whitesmoke;
}
.docson .box-2 {
background: beige;background: -webkit-linear-gradient(white, beige, beige, gainsboro);background: -moz-linear-gradient(white, beige, beige, gainsboro);
}
.docson .box-3 {
background: azure;background: -webkit-linear-gradient(white, azure, azure, gainsboro);background: -moz-linear-gradient(white, azure, azure, gainsboro);
}
.docson .box-4 {
background: mistyrose;background: -webkit-linear-gradient(white, mistyrose, mistyrose, gainsboro);background: -moz-linear-gradient(white, mistyrose, mistyrose, gainsboro);
}
.docson .box-5 {
background: #ddffee;background: -webkit-linear-gradient(white, #ddffee, #ddffee, gainsboro);background: -moz-linear-gradient(white, #ddffee, #ddffee, gainsboro);
}
.docson .box-6 {
background: ghostwhite;background: -webkit-linear-gradient(white, ghostwhite, ghostwhite, gainsboro);background: -moz-linear-gradient(white, ghostwhite, ghostwhite, gainsboro);
}
.docson .box-7 {
background: lavenderblush;background: -webkit-linear-gradient(white, lavenderblush, lavenderblush, gainsboro);background: -moz-linear-gradient(white, lavenderblush, lavenderblush, gainsboro);
}
.docson .box-8 {
background: aliceblue;background: -webkit-linear-gradient(white, aliceblue, aliceblue, gainsboro);background: -moz-linear-gradient(white, aliceblue, aliceblue, gainsboro);
}
.docson .signature {
padding-top: 5px;clear: left;
}
.docson .signature-header {
padding-top: 3px;height: 20px;line-height: 24px;
}
.docson .property-name {
float: left;font-family: "Lucida Console", Monaco, monospace;min-width: 130px;
}
.docson .required {
font-weight: bold;
}
.docson .type-keyword {
font-size: 80%;color: dimgray;
}
.docson .type-pattern {
font-size: 80%;color: darkolivegreen;
}
.docson .type-enum {
color: purple;font-family: "Lucida Console", Monaco, monospace;font-size: 90%;padding-left: 2px;
}
.docson .type-default {
color: darkslategrey;font-family: "Lucida Console", Monaco, monospace;
}
.docson .signature-type {
padding-left: 6px;float: left;min-width: 160px;max-width: 320px;
}
.docson .signature-type-any {
font-style: italic;
}
.docson .signature-type-string {
color: green;
}
.docson .signature-type-number {
color: blue;
}
.docson .signature-type-integer {
color: blueviolet;
}
.docson .signature-type-boolean {
color: orangered;
}
.docson .signature-type-null {
font-family: "Lucida Console", Monaco, monospace;color: red;
}
.docson .signature-type-date-time {
color: #224466;
}
.docson .signature-type-email {
color: cornflowerblue;
}
.docson .signature-type-hostname {
color: darkolivegreen;
}
.docson .signature-type-ipv6 {
color: darkslateblue;
}
.docson .signature-type-ipv4 {
color: indigo;
}
.docson .signature-type-uri {
color: saddlebrown;
}
.docson .signature-button {
cursor: pointer;padding: 4px;border-radius: 4px;border: 1px solid lightgrey;-moz-border-radius: 4px;
}
.docson .signature-type-ref {
color: darkblue;
}
.docson .signature-type-expanded {
padding: 5px 3px 3px 5px;vertical-align: -1px;opacity: 0.6;-moz-box-shadow:    inset 1px 1px 1px darkgray;-webkit-box-shadow: inset 1px 1px 1px darkgray;box-shadow:         inset 1px 1px 1px darkgray;
}
.docson .signature-description {
padding-left: 6px;padding-bottom: 3px;color: dimgray;float: left;max-width: 600px;
}
.docson .signature-box-container {
clear: both;
}
.docson .box-container {
float: left;padding-top: 4px;padding-left: 8px;display: none;
}
.docson .button {
-webkit-touch-callout: none;-webkit-user-select: none;-khtml-user-select: none;-moz-user-select: none;-ms-user-select: none;user-select: none;
}
.docson .button:hover {
color: darkred;
}
@media all and (min-width: 20px) and (max-width: 500px) {
.docson .property-name {min-width: 0px;}
.docson .signature-type {min-width: 0px;}
.docson .source {width: 100%;}
.docson .box-container, .box {width: 100%;}
}
.docson .desc {
padding-left: 6px;line-height: 17px;
}
.docson .desc p {
margin: 0;margin-top: 3px;
}
.docson .desc code {
color: purple;font-size: 90%;
}

.docson .desc table {
margin-top: 4px;margin-bottom: 4px;border-collapse:collapse;
}
.docson .desc th {
padding: 0px;font-size: 80%;font-weight: normal;background: rgba(231, 231, 231, 0.5);
}
.docson .desc table,
.docson .desc th,
.docson .desc td {
border: 1px solid lightgrey;
}
.docson .desc tr > td {
padding: 2px 6px 2px 6px;
}
.docson .desc h1 {
font-size: 120%;margin-top: 0px;margin-bottom: 3px;
}
.docson .desc h2 {
font-size: 115%;margin-top: 0px;margin-bottom: 3px;
}
.docson .desc h3 {
font-size: 110%;margin-top: 0px;margin-bottom: 3px;
}
.docson .desc h4 {
font-size: 100%;margin-top: 0px;margin-bottom: 3px;
}
.docson .desc pre {
border: 1px solid lightgray;border-radius: 4px;-webkit-border-radius: 4px;-moz-border-radius: 4px;padding: 4px;margin: 3px 0px;
}
.docson .desc ul {
padding-left: 16px;margin: 3px 0px;
}
.docson .desc ol {
padding-left: 16px;margin: 3px 0px;
}
.docson .desc blockquote {
margin: 3px 0px 3px 10px;padding-left: 6px;border-left: 1px solid lightgray;
}
.docson .desc img {
vertical-align: middle;
}
.docson .desc hr {
border: 0;height: 1px;background: lightgrey;color: lightgrey;width: 50%;
}
.docson .source {
display: none;padding-left: 16px;padding-right: 30px;
}
.docson .json, .json-schema {
font-family: "Lucida Console", Monaco, monospace;
}
.docson .json-punctuation {
font-weight: bold;
}
.docson .json-null, .json-true, .json-false {
font-weight: bold;
}
.docson .json-true {
color: #080;
}
.docson .json-false {
color: #800;
}
.docson .json-object-key {
color: #246;
}
.docson .json-keyword {
font-weight: bold;color: #036;
}
.docson .json-number {
font-weight: bold;color: #660;
}
.docson .json-string {
color: #800;
}
.docson .json-schema-map > .json-object-key {
color: #080;font-style: italic;
}

/* Code highlighting */

.hljs {
display: block;
overflow-x: auto;
padding: 0.5em;
color: #fff;
background:#2b2b2b;font-size:0.9em;font-family:'Courier New';line-height:1.6em;border-radius:4px;
}

.hljs-strong,
.hljs-emphasis {
color: #a8a8a2;
}

.hljs-bullet,
.hljs-quote,
.hljs-link,
.hljs-number,
.hljs-regexp,
.hljs-literal {
color: #6896ba;
}

.hljs-code,
.hljs-selector-class {
color: #a6e22e;
}

.hljs-emphasis {
font-weight: bold;
}

.hljs-keyword,
.hljs-selector-tag,
.hljs-section,
.hljs-attribute,
.hljs-name,
.hljs-variable {
color: #cb7832;
}

.hljs-params {
color: #b9b9b9;
}

.hljs-string {
color: #6a8759;
}

.hljs-subst,
.hljs-type,
.hljs-built_in,
.hljs-builtin-name,
.hljs-symbol,
.hljs-selector-id,
.hljs-selector-attr,
.hljs-selector-pseudo,
.hljs-template-tag,
.hljs-template-variable,
.hljs-addition {
color: #e0c46c;
}

.hljs-comment,
.hljs-deletion,
.hljs-meta {
color: #7f7f7f;
}

/* "Fork me on GitHub" CSS ribbon v0.1.1 */

.github-fork-ribbon {
position: fixed;
padding: 2px 0;
background-color: #900;
background-image: -webkit-gradient(linear, left top, left bottom, from(rgba(0, 0, 0, 0)), to(rgba(0, 0, 0, 0.15)));
background-image: -webkit-linear-gradient(top, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.15));
background-image: -moz-linear-gradient(top, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.15));
background-image: -ms-linear-gradient(top, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.15));
background-image: -o-linear-gradient(top, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.15));
background-image: linear-gradient(to bottom, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.15));
-webkit-box-shadow: 0 2px 3px 0 rgba(0, 0, 0, 0.5);
-moz-box-shadow: 0 2px 3px 0 rgba(0, 0, 0, 0.5);
box-shadow: 0 2px 3px 0 rgba(0, 0, 0, 0.5);
font: 500 9px "Helvetica Neue", Helvetica, Arial, sans-serif;
z-index: 9999;
pointer-events: auto;
}
.github-fork-ribbon a,
.github-fork-ribbon a:hover {
color: #fff;
text-decoration: none;
text-shadow: 0 -1px rgba(0, 0, 0, 0.5);
text-align: center;
width: 100px;
line-height: 12px;
display: inline-block;
padding: 2px 0;
border-width: 1px 0;
border-style: dotted;
border-color: #fff;
border-color: rgba(255, 255, 255, 1);
}
.github-fork-ribbon-wrapper {
width: 100px;
height: 100px;
position: fixed;
overflow: hidden;
top: 0;
z-index: 9999;
pointer-events: none;
}
.github-fork-ribbon-wrapper.right {
right: 0;
}
.github-fork-ribbon-wrapper.right .github-fork-ribbon {
top: 16px;
right: -22px;
-webkit-transform: rotate(45deg);
-moz-transform: rotate(45deg);
-ms-transform: rotate(45deg);
-o-transform: rotate(45deg);
transform: rotate(45deg);
}
</style>
<script type="text/javascript" src="https://developer.mpds.io/webassets/js/jquery.min.js"></script>
<script type="text/javascript" src="https://developer.mpds.io/webassets/js/highlight.min.js"></script>
<script type="text/javascript" src="https://developer.mpds.io/webassets/js/jquery.nearest.min.js"></script>
<script type="text/javascript" src="https://developer.mpds.io/webassets/js/darkmode.js"></script>
<script type="text/javascript" src="https://developer.mpds.io/webassets/js/require.min.js"></script>
<script type="text/javascript">
"use strict";

function load_custom_schema(addr, el){
    $.get(addr).done(function(schema){
        for (var type in schema.definitions){
            if (type == "input_query" || type == "chemical_elements")
                continue;
            try {
                docson.doc("schema_" + type, schema.definitions[type]);
            } catch (e){
                console.error("Could not parse JSON schema!");
            }
        }
    }).fail(function(xhr, status, err){
        console.error("Could not load JSON schema");
    });
}

function debounce(func, wait, immediate){
    var timeout;
    return function(){
        var context = this,
            args = arguments;
        var later = function(){
            timeout = null;
            if (!immediate) func.apply(context, args);
        }
        var callNow = immediate && !timeout;
        clearTimeout(timeout);
        timeout = setTimeout(later, wait);
        if (callNow) func.apply(context, args);
    }
}

function is_inview(element){
    var pageTop = $(window).scrollTop(),
        pageBottom = pageTop + $(window).height(),
        elementTop = $(element).offset().top,
        elementBottom = elementTop + $(element).height();
    return ((elementTop <= pageBottom) && (elementBottom >= pageTop));
}

function update_menu(){
    $('#menu a').removeClass('active');
    var tol = 3, cnt = 0;
    $('#content .view').each(function(){
        if (is_inview(this)){
            var h3 = $(this).nearest('h3', {onlyY: true, directionConstraints: ['top']}).attr('id');
            var h2 = $(this).nearest('h2', {onlyY: true, directionConstraints: ['top']}).attr('id');
            $('#menu a[href="#' + h3 + '"]').addClass('active');
            $('#menu a[href="#' + h2 + '"]').addClass('active');
            cnt++;
            if (cnt == tol) return false;
        }
    });
}

$(document).ready(function(){
    if ($('#menu').is(':visible')){
        var menu = '<ul>';
        $('h2, h3').each(function(){
            var that = $(this),
                active_cls = '',
                hlevel = that.prop('tagName'),
                value = that.text(),
                href = that.attr('id');
            if (is_inview(that)) active_cls = ' class="active"';

            if (hlevel == 'H2') menu += '<li><a' + active_cls + ' href="#' + href + '">' + value + '</a></li>';
            else if (hlevel == 'H3') menu += '<li><ul><li><a' + active_cls + ' href="#' + href + '">' + value + '</a></li></ul></li>';
        });
        menu += '</ul>';
        $('#menu').append(menu);

        $(window).scroll(debounce(update_menu, 500));
    }

    if ($('#bookmarks').is(':visible')){
        var menu = '<ul>';
        $('div.caption').each(function(){
            var that = $(this),
                value = that.html(),
                href = that.attr('data-id');
            if (value.substr(-1, 1) == '.') value = value.substr(0, value.length - 1);
            menu += '<li><a href="#' + href + '">' + value + '</a></li>';
        });
        menu += '</ul>';
        $('#bookmarks').append(menu);
    }

    require.config({
        baseUrl: 'https://developer.mpds.io'
    });
    require(['webassets/js/docson'], function(docson){
        load_custom_schema('https://developer.mpds.io/mpds.schema.json', 'mpds_json_schema');
    });

    /*if (window.location.host.indexOf('developer') !== -1){ // FIX LINKS IN GH
        $("#nav > ul > li > a").each(function(item){
            $(this).attr('href', 'https://mpds.io' + $(this).attr('href'));
        });
    }*/
    window.addEventListener('hashchange', function(evt){
        var url = window.location.hash.substr(1);
        if (url == 'start' || url == 'products'){
            window.location.href = 'https://mpds.io/#' + url;
        }
    });

    window.darkmode = new Darkmode({saveInCookies: false});
    document.getElementById('darkmode_trigger').innerHTML = window.darkmode.isActivated() ? 'enabled' : 'disabled';

    // theme switcher
    document.getElementById('darkmode_trigger').onclick = function(){
        window.darkmode.toggle();
        document.getElementById('darkmode_trigger').innerHTML = window.darkmode.isActivated() ? 'enabled' : 'disabled';
        document.getElementById('comms').contentWindow.postMessage(JSON.stringify({darkmode: window.darkmode.isActivated() ? 1 : 2}), '*');
    }

    // cross-site comms
    window.addEventListener('message', function(message){
        //if (message.origin === 'https://trusted.com') {
        var received;
        try {
            received = JSON.parse(message.data);
        } catch (ignore){}
        var darkmode = received.darkmode;
        if (darkmode !== undefined){
            darkmode = (darkmode === 1);
            if (darkmode !== window.darkmode.isActivated()) window.darkmode.toggle();
            document.getElementById('darkmode_trigger').innerHTML = window.darkmode.isActivated() ? 'enabled' : 'disabled';
        }
        //}
    });

    hljs.initHighlightingOnLoad();
});
</script>
</head>
<body class="sky">

    <div id="nav"><ul>
        <li><a href="/#start">Start</a></li>
        <li class="nav_fluid"><a href="/tutorial">Tutorial</a></li>
        <li class="nav_fluid"><a href="/developer">API</a></li>
        <li><a href="/#products">Products</a></li>
    </ul></div>

    <div id="menu"></div>

    <div id="bookmarks"></div>

    <div id="main">
        <div id="landing_box">
            <div id="logo_l">MPDS</div><div id="logo"><a href="/"></a></div><div id="logo_r"><h1>Materials Platform<br />for Data Science</h1></div>
            <div id="motto">based on the <a href="http://paulingfile.com" target="_blank" rel="noopener noreferrer">PAULING FILE</a> experimental inorganic database</div>
        </div>

        <div class="spacer"></div>

        <div id="content">
            <p style="background:#f4fbff;">The MPDS application programming interface (API) presents the materials data of the <a href="http://paulingfile.com" target="_blank">PAULING FILE</a> database <i>online</i> in the machine-readable formats, suitable for automated processing. The intended audience is software engineers and data scientists. The API is available by a subscription (SLA). To start using the API the reader needs a valid API key from the <a href="/#modal/menu">MPDS account</a>. We encourage to <a href="/#products">contact us</a> for opening such the account.</p>

            <p style="background:#fdffe8;">Some parts of the data are opened and freely available. In particular, these are: (<i>a</i>) <strong>cell parameters - temperature diagrams</strong> and <strong>cell parameters - pressure diagrams</strong>, (<i>b</i>) all data for compounds containing <strong>both Ag and K</strong>, (<i>c</i>) all data for <strong>binary</strong> compounds of <strong>oxygen</strong>, (<i>d</i>) all data generated by <strong>machine learning</strong>, and (<i>e</i>) all data generated by <strong>first-principles calculations</strong>. <!-- Please, login via <a target="_blank" href="/github_oauth.html" rel="nofollow noopener noreferrer">GitHub</a> or <a target="_blank" href="/orcid_oauth.html" rel="nofollow noopener noreferrer">ORCID</a> or <a target="_blank" href="/linkedin_oauth.html" rel="nofollow noopener noreferrer">LinkedIn</a> to get the API access to these data. --> </p>

            <p style="background:#f4fbff;">A tech-savvy reader may also explore <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/notebooks">Jupyter notebooks</a> (also in <a target="_blank" href="https://mybinder.org/v2/gh/mpds-io/mpds-api/gh-pages?filepath=notebooks">Binder</a>) and <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/kickoff">kickoff Python demos</a>. <strong>Important</strong>: using the API keys within the third-party environment like Binder is potentially insecure and should be done with the greatest care. Should the reader be not familiar with the materials science, just wishing to crunch the data, this <a href="https://tutorial.tilde.pro">quick tutorial</a> can be recommended.</p>

            <h2 id="Downloading-data">&sect;1. Downloading data</h2>

            <h3 id="Overview-1">&sect;1.1. MPDS data model</h3>

            <p class="view">The standard unit of the MPDS data is an <strong>entry</strong>. All the MPDS entries are subdivided into three kinds: <strong>crystalline structures</strong>, <strong>physical properties</strong>, and <strong>phase diagrams</strong>. They are called S-, P- or C-entries, correspondingly. Entries have persistent identifiers (similar to DOIs), <i>e.g.</i> <span class="t">S377634</span>, <span class="t">P600028</span>, <span class="t">C100026</span>. These identifiers are used in the stable URLs, <i>e.g.</i> <span class="t">https://mpds.io/entry/C100026</span> and therefore citable, their persistence is guaranteed.</p>

            <p class="view">Another dimension of the MPDS data is the <strong>distinct phases</strong>. The three kinds of entries are interlinked via the distinct materials phases they belong. A tremendous work was done by the PAULING FILE team in the past 20 years to manually distinguish about 200&nbsp;000 inorganic materials phases, appearing in the literature. Each phase has a unique combination of (<i>a</i>) chemical formula, (<i>b</i>) space group, (<i>c</i>) Pearson symbol. Each phase has the permanent integer identifier called <span class="t">phase_id</span>, also used in the stable citable URLs, such as <span class="t">https://mpds.io/phase_id/7268</span>. An analogous concept is the Materials Project IDs, such as <span class="t">mp-2657</span>, matching the above mentioned MPDS <span class="t">phase_id</span> <span class="t">7268</span>.</p>

            <p class="view">Consider the following example of the <strong>entries</strong> and <strong>distinct phases</strong>. There can be the following distinct phases for the titanium dioxide: rutile with the space group <i>136</i> (let us say, <span class="t">phase_id&nbsp;1</span>), anatase with the space group <i>141</i> (<span class="t">phase_id&nbsp;2</span>), and brookite with the space group <i>61</i> (<span class="t">phase_id&nbsp;3</span>). Then the S- and P-entries for the titanium dioxide must refer to either <span class="t">1</span>, or <span class="t">2</span>, or <span class="t">3</span>, and the C-entries must refer to <span class="t">1</span>, <span class="t">2</span>, and <span class="t">3</span> simultaneously.</p>

            <p class="view">Our relatively new development is the <strong>machine-learning</strong> data generated from the original <strong>peer-reviewed</strong> data for the less known phases. Such data are always clearly attributed and not supplied by default. So far only the P-entries (10 physical properties) were predicted and thus can be of the machine-learning data type. For some of these properties we also provide our in-house <strong><i>ab&nbsp;initio</i> calculations</strong> data, based on the experimental (carefully relaxed) crystalline structures. See below how to enable all these new data types in our API outputs.</p>

            <p>Additionally we support the Optimade unified access interface. The Optimade stands for the Open Databases Integration for Materials Design and defines a standardized way for a programmatic access to any online materials database. Please, refer to the <a href="https://www.optimade.org">Optimade website</a> for the details.</p>

            <h3 id="Categories">&sect;1.2. Categories of data</h3>

            <p class="view">The most common task is to get the MPDS data according to some search criteria. We support 14 basic criteria for searching and downloading the entries:</p>

            <table class="view" id="T1">
                <thead>
                    <tr>
                    <th>Facet</th>
                    <th>Machine-readable name</th>
                    <th>Examples</th>
                    <th>Case sensitivity</th>
                    </tr>
                </thead>
                <tbody>
                    <tr>
                    <td>Physical properties (see the <a href="/#hierarchy">MPDS physical properties hierarchy</a>,<br />numerical properties specifically)</td>
                    <td><span class="t">props</span></td>
                    <td><span class="t">conductivity</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Chemical elements</td>
                    <td><span class="t">elements</span></td>
                    <td><span class="t">Cs</span>, <span class="t">Fr-O</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Materials classes (various groups of terms: <i>i.e.</i> mineral names, periodic groups, physical classes, element count keywords <i>etc.</i>, see <a href="https://tutorial.mpds.io/#Classes">examples</a>)</td>
                    <td><span class="t">classes</span></td>
                    <td><span class="t">perovskite</span>, <span class="t">quaternary</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Crystal system</td>
                    <td><span class="t">lattices</span></td>
                    <td><span class="t">orthorhombic</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Chemical formula</td>
                    <td><span class="t">formulae</span></td>
                    <td><span class="t">SrTiO3</span>, <span class="t">O3Al2</span>, <span class="t">A2B</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Space group (number or international short symbol)</td>
                    <td><span class="t">sgs</span></td>
                    <td><span class="t">221</span>, <span class="t">Pm-3m</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Prototype system (<a href="#L3">"Strukturbericht"</a> or formula, Pearson symbol, space group number)</td>
                    <td><span class="t">protos</span></td>
                    <td><span class="t">D51</span>,<br /><span class="t">SiS2 oI12 72</span></td>
                    <td align="center">yes</td>
                    </tr>
                    <tr>
                    <td>Polyhedron atoms (center, ligands)</td>
                    <td><span class="t">aeatoms</span></td>
                    <td><span class="t">U</span>, <span class="t">X-Se</span>, <span class="t">UO6</span>, <span class="t">UX7</span><br />(<span class="t">X</span> = any atom)</td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Polyhedral type (see <a href="#L4">all supported types</a>)</td>
                    <td><span class="t">aetypes</span></td>
                    <td><span class="t">icosahedron 12-vertex</span></td>
                    <td align="center">yes</td>
                    </tr>
                    <tr>
                    <td>Publication author</td>
                    <td><span class="t">authors</span></td>
                    <td><span class="t">Evarestov</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>CODEN (code of publication journal, see <a href="http://cassi.cas.org">CODEN index</a>)</td>
                    <td><span class="t">codens</span></td>
                    <td><span class="t">PPCPFQ</span></td>
                    <td align="center">no</td>
                    </tr>
                    <tr>
                    <td>Publication years</td>
                    <td><span class="t">years</span></td>
                    <td><span class="t">1960-1990</span></td>
                    <td align="center">&mdash;</td>
                    </tr>
                    <tr>
                    <td>Author location</td>
                    <td><span class="t">geos</span></td>
                    <td><span class="t">United Arab Emirates</span></td>
                    <td align="center">yes</td>
                    </tr>
                    <tr>
                    <td>Author organization</td>
                    <td><span class="t">orgs</span></td>
                    <td><span class="t">CSIRO</span></td>
                    <td align="center">yes</td>
                    </tr>
                </tbody>
            </table>

            <div class="caption" data-id="T1"><b>Table 1.</b> Categories of MPDS data.</div>

            <p class="view">Thus, in an API request it is possible to combine all these criteria in a reasonable manner. Combination is always done implying conjunctive <span class="t">AND</span> operator. The reasonable manner means one can combine <i>e.g.</i> <strong>publication authors</strong> with <strong>publication years</strong>, but cannot combine <strong>chemical formulae</strong> with <strong>chemical elements</strong> or <strong>space group</strong> with <strong>crystal system</strong>, <i>i.e.</i> the common sense rules in chemistry and physics are implied.</p>

            <p>We distinguish the single-term categories, such as the physical properties and crystal systems, and many-term categories, such as the materials classes and publication authors. A combination of terms from the single-term category is <i>not</i> supported. That is, one <i>cannot</i> search for both <span class="t">band gap</span> and <span class="t">conductivity</span>. A combination of terms from the many-term category is easily possible. That is, one <i>can</i> search for both <span class="t">binaries</span> and <span class="t">perovskites</span>.</p>

            <p>The single-term categories are <span class="t">props</span>, <span class="t">lattices</span>, <span class="t">formulae</span>, <span class="t">sgs</span>, <span class="t">protos</span>, <span class="t">aeatoms</span>, <span class="t">codens</span>, <span class="t">geos</span>, and <span class="t">orgs</span>. The many-term categories are <span class="t">elements</span>, <span class="t">classes</span>, <span class="t">aetypes</span>, <span class="t">authors</span>, and <span class="t">years</span>.</p>

            <h3 id="REST-endpoint">&sect;1.3. REST-alike download endpoint</h3>

            <p class="view">To serve data we employ the single <a href="https://en.wikipedia.org/wiki/Representational_state_transfer">REST</a>-alike endpoint, following the <a href="https://en.wikipedia.org/wiki/Hypertext_Transfer_Protocol">HTTP</a> protocol basics. We support input parameters in the query string and request header. We provide the <a href="https://en.wikipedia.org/wiki/List_of_HTTP_status_codes">standard HTTP status codes</a> in response. The address of the data download endpoint is the following:</p>

            <div class="blackbg">https://api.mpds.io/v0/download/facet</div>

            <p class="view">The HTTP verb to be used is always <span class="t">GET</span>. The header must always contain the <span class="t">Key</span> string with the valid MPDS API key, obtained from the corresponding <a href="/#modal/menu">MPDS account</a>. A minimal <span class="t">curl</span> example reads:</p>

            <div class="blackbg">curl -H Key:a_long_secret_string https://api.mpds.io/v0/download/facet</div>

            <p class="view">The following query string parameters are supported:</p>

            <table class="view" id="T2">
                <thead>
                    <tr>
                    <th>Parameter</th>
                    <th>Type</th>
                    <th>Description</th>
                    </tr>
                </thead>
                <tbody>
                    <tr>
                    <td><span class="t">q</span></td>
                    <td><span class="t">JSON-serialized object</span></td>
                    <td>The object <span class="t">q</span> may contain any reasonable combination of criteria, given in the <a href="/#T1">Table 1</a>, column "Machine-readable name". Always required.</td>
                    </tr>
                    <tr>
                    <td><span class="t">phases</span></td>
                    <td><span class="t">string or integer</span></td>
                    <td>Distinct phase identifiers (called <span class="t">phase_id</span>'s) separated by commas can be optionally supplied to limit the search. This is convenient for programmatic chained searches one after another. A relevant <span class="t">q</span> parameter must be always given. The count of <span class="t">phase_id</span>'s per a single request cannot exceed <span class="t">1000</span>.</td>
                    </tr>
                    <tr>
                    <td><span class="t">dtype</span></td>
                    <td><span class="t">integer</span></td>
                    <td>Data type. Possible values:<br />
                        <span class="t">1</span> or <span class="t">MPDSDataTypes.PEER_REVIEWED</span> (original data),<br />
                        <span class="t">2</span> or <span class="t">MPDSDataTypes.MACHINE_LEARNING</span> (data generated by machine learning),<br />
                        <span class="t">4</span> or <span class="t">MPDSDataTypes.AB_INITIO</span> (in-house simulations data), and<br />
                        <span class="t">7</span> or <span class="t">MPDSDataTypes.ALL</span> (all mentioned types).<br />
                        Default value: <span class="t">1</span> (original data).</td>
                    </tr>
                    <tr>
                    <td><span class="t">pagesize</span></td>
                    <td><span class="t">integer</span></td>
                    <td>The maximum number of hits per a single response (pagination). Default value: <span class="t">10</span>. Allowed values: <span class="t">10</span>, <span class="t">100</span>, <span class="t">500</span>, and <span class="t">1000</span>.</td>
                    </tr>
                    <tr>
                    <td><span class="t">page</span></td>
                    <td><span class="t">integer</span></td>
                    <td>The page number, either in the selected or in the default pagination. Counts from <span class="t">0</span>. Defaults to the first page, which has number <span class="t">0</span>, <i>i.e.</i> the default value is <span class="t">0</span>.</td>
                    </tr>
                    <tr>
                    <td><span class="t">fmt</span></td>
                    <td><span class="t">integer</span></td>
                    <td>Output data format. Can be <span class="t">json</span> or <span class="t">cif</span>. The latter <span class="t">cif</span> only makes sense for the <strong>crystalline structures</strong>. Default value: <span class="t">json</span>.</td>
                    </tr>
                </tbody>
            </table>

            <div class="caption" data-id="T2"><b>Table 2.</b> Query string parameters for download endpoint.</div>

            <p class="view">An example of the <span class="t">q</span> parameter to request the <strong>crystalline structures</strong> (with the spaces and line breaks added for readability):</p>

            <div class="blackbg view">{
    "elements": "Sr-Ti-O",
    "classes": "perovskite, conductor",
    "sgs": "Pm-3m",
    "props": "structural properties"
}</div>

            <p class="view">Another example of the <span class="t">q</span> parameter to request the <strong>physical properties</strong>:</p>

            <div class="blackbg view">{
    "elements": "Zr",
    "classes": "oxide",
    "props": "thermal expansion"
}</div>

            <p class="view">Please, note: we do not support querying by entry IDs in the API intentionally. Although the entry IDs are stable, the entry contents might be updated, changed, or deactivated with time, therefore we advice not to rely on their IDs in the API data mining in the mass manner. Use <a href="https://mpds.io/entry/S377634">GUI</a> to get the particular entries by their IDs.</p>

            <p class="view">By default the response has <span class="t">json</span> format, containing the following properties: <span class="t">out</span> (list <i>i.e.</i> array of the MPDS entries), <span class="t">npages</span> (number of pages in the selected or default pagination), <span class="t">page</span> (current page number), <span class="t">count</span> (total number of hits), and <span class="t">error</span> (should be <span class="t">null</span>). Some other properties might be added in future.</p>

            <p class="view">Each kind of entries (<i>i.e.</i> S- or P-entries) has its own specific properties in <strong>JSON</strong> (see below).</p>

            <div class="blackbg view">{
    "error": null,
    "out": [
        {
            "entry_one_properties": "entry_one_values"
        },
        {
            "entry_two_properties": "entry_two_values"
        }
    ],
    "npages": 1,
    "page": 0,
    "count": 2
}</div>

            <p class="view">And if no results are found:</p>

            <div class="blackbg view">{
    "error": null,
    "out": [],
    "npages": 0,
    "count": 0,
    "page": 0
}</div>

            <p class="view">The output in <span class="t">cif</span> format presents nothing more than concatenated <strong>CIF</strong>s in a plain text.</p>

            <p class="view">The response for a correctly understood request will always have the status code <span class="t">200</span>, whereas the status codes <span class="t">400</span> ("Wrong parameters") or <span class="t">403</span> ("Forbidden") may signal about an error in the request. Also, <span class="t">429</span> ("Too many requests") tells that the request rate should be decreased. Generally, checking the response status code should be always done first.</p>

            <h3 id="JSON-schemata">&sect;1.4. JSON schemata of the MPDS entries</h3>

            <p class="view" id="L1">JSON is the first-class-passenger format in the MPDS API. We provide <a href="http://developer.mpds.io/mpds.schema.json">JSON schemata</a> for all kinds of the MPDS entries in the API output. This is visualized below:</p>

            <div id="schema_crystal_structure" class="view"></div>

            <div class="caption" data-id="L1" id="L2"><b>Listing 1.</b> Interactive JSON schema for crystalline structure or S-entry (<span class="t">object</span>s are clickable).</div>

            <div id="schema_physical_property" class="view"></div>

            <div class="caption" data-id="L2" id="L3"><b>Listing 2.</b> Interactive JSON schema for physical property or P-entry (<span class="t">object</span>s are clickable).</div>

            <div id="schema_phase_diagram" class="view"></div>

            <div class="caption" data-id="L3"><b>Listing 3.</b> Interactive JSON schema for phase diagram or C-entry (<span class="t">object</span>s are clickable).</div>

            <p class="view">Any MPDS API JSON output can be validated against the above schemata <i>e.g.</i> using a Python script below. In the following example the external <span class="t">httplib2</span> and <span class="t">jsonschema</span> Python packages are used:</p>

            <div id="json_validate" class="view">
            <pre id="L4"><code>#!/usr/bin/env python

import sys
import json
import httplib2
from jsonschema import validate, Draft4Validator
from jsonschema.exceptions import ValidationError

try: input_json = sys.argv[1]
except IndexError: sys.exit("Usage: %s input_json" % sys.argv[0])

network = httplib2.Http()
response, content = network.request("http://developer.mpds.io/mpds.schema.json")
assert response.status == 200

schema = json.loads(content)
Draft4Validator.check_schema(schema)

target = json.loads(open(input_json).read())

if not target.get("npages") or not target.get("out") or target.get("error"):
    sys.exit("Unexpected API response")

try:
    validate(target["out"], schema)
except ValidationError, e:
    raise RuntimeError(
        "The item: \r\n\r\n %s \r\n\r\n has an issue: \r\n\r\n %s" % (
            e.instance, e.context
        )
    )

            </code></pre>
            </div>

            <div class="caption view" data-id="L4"><b>Listing 4.</b> Python console script for JSON validation.</div>

            <p class="view">All the MPDS data without exceptions are expected to be valid against the supplied schemata.</p>

            <h3 id="Example-scripts">&sect;1.5. Example scripts</h3>

            <p class="view">Below are simple examples for the MPDS data retrieval in two programming languages: <strong>Python</strong> and <strong>JavaScript</strong>. More languages can be added by request. Note, that the Python example requires an external package <span class="t">httplib2</span>. We provide more convenient <a href="https://github.com/mpds-io/mpds-client">Python client library</a>, so these examples serve demonstration purposes only.</p>

            <div id="example_python" class="view">
            <pre id="L5"><code>#!/usr/bin/env python

from urllib.parse import urlencode
import httplib2
import json


api_key = "" # your key
endpoint = "https://api.mpds.io/v0/download/facet"
search = {
    "elements": "Mn",
    "classes": "binary, oxide",
    "props": "isothermal bulk modulus",
    "lattices": "cubic"
}

req = httplib2.Http()
response, content = req.request(
    uri=endpoint + '?' + urlencode({
        'q': json.dumps(search),
        'pagesize': 10,
        'dtype': 2 # see parameters documentation above
    }),
    method='GET',
    headers={'Key': api_key}
)

if response.status != 200:
    # NB 400 means wrong input, 403 means authorization issue etc.
    # see https://en.wikipedia.org/wiki/List_of_HTTP_status_codes
    raise RuntimeError("Error code %s" % response.status)

content = json.loads(content)

if content.get('error'): raise RuntimeError(content['error'])

print("OK, got %s hits" % len(content['out']))
            </code></pre>
            </div>

            <div class="caption view" data-id="L5"><b>Listing 5.</b> Sample Python script for working with the MPDS API.</div>

            <div id="example_js" class="view">
            <pre id="L6"><code>#!/usr/bin/env node

var https = require('https');
var qs = require('querystring');

var api_key = ""; // your key
var host = "api.mpds.io", port = 443, path = "/v0/download/facet";
var search = {
    "elements": "Mn",
    "classes": "binary, oxide",
    "props": "isothermal bulk modulus",
    "lattices": "cubic"
};

https.request({
    host: host,
    port: port,
    path: path + '?' + qs.stringify({q: JSON.stringify(search), pagesize: 10, dtype: 2}),
    method: 'GET',
    headers: {'Key': api_key}
}, function(response){
    var result = '';
    if (response.statusCode != 200){
        // NB 400 means wrong input, 403 means authorization issue etc.
        // see https://en.wikipedia.org/wiki/List_of_HTTP_status_codes
        return console.error('Error code: ' + response.statusCode);
    }
    response.on('data', function(chunk){
        result += chunk;
    });
    response.on('end', function(){
        result = JSON.parse(result);
        if (result.error) console.error(result.error);
        else {
            console.log('OK, got ' + result.out.length + ' hits');
        }
    });
}).on('error', function(err){
    console.error('Network error: ' + err);
}).end();
            </code></pre>
            </div>

            <div class="caption view" data-id="L6"><b>Listing 6.</b> Sample JavaScript code for working with the MPDS API.</div>

            <h3 id="API-GUI">&sect;1.6. API and GUI</h3>

            <p class="view">The MPDS data provided in API and <a href="https://mpds.io">GUI</a> (graphical user interface) are generally the same. There are however two important points to consider.</p>

            <p class="view">First, the GUI is intended for the human researchers, not for the automated processing. Therefore the data in GUI are less rigorously formatted, being thus closer to the original publications (which, of course, impedes machine analysis).</p>

            <p class="view">Second, some portion of data (~15%) present in the GUI is absent in the API. This is because currently a number of MPDS entries has only textual content or even no content at all (in process of preparation), <i>i.e.</i> present little value for the data mining. However, we make our best to expose the maximum MPDS data in the API, and this is the work in progress.</p>

            <p class="view">To summarize, the API differs from GUI in the following: (<i>a</i>) machine-readable (and to a certain extent <i>machine-understandable</i>), (<i>b</i>) programmer-friendly, pluggable and integration-ready, (<i>c</i>) massively exposing the data, <i>i.e.</i> all the MPDS content can be analyzed in minutes. Thus, thanks to the API, the MPDS platform can be used in variety of the other scientific ways, which we had never designed. So the researchers get unprecedented flexibility and power, which is unthinkable within the GUI.</p>

            <h2 id="Processing-data">&sect;2. Processing data</h2>

            <h3 id="Overview-2">&sect;2.1. Overview</h3>

            <p class="view">To facilitate starting with the MPDS API we provide several exercises (or miniature case studies): (<i>a</i>)&nbsp;distribution density analysis, (<i>b</i>)&nbsp;clustering, and (<i>c</i>)&nbsp;correlation search. To a certain extent they could be considered as recipes, however aiming to illustrate the informatics aspect, not the materials science aspect. All these exercises are described below and can be found in the <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/kickoff">Github repository</a>.</p>

            <p class="view">There is also a number of <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/notebooks">Jupyter notebooks</a>: (<i>a</i>)&nbsp;basic plotting using the periodic table, (<i>b</i>)&nbsp;machine-learning and peer-reviewed data, (<i>c</i>)&nbsp;unusual materials phases from the machine learning, (<i>d</i>)&nbsp;pVT-data and EoS fitting. They can be launched online from GitHub using <i>e.g.</i> <a href="https://mybinder.org/v2/gh/mpds-io/mpds-api/gh-pages?filepath=notebooks">Binder environment</a>.</p>

            <p class="view" style="background:#fdffe8;">The reader now may decide, whether to follow a text with the code listings below ("classical way"), or to download and run the <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/kickoff">Python scripts</a> ("Python way"), or play with the Jupyter notebooks <a target="_blank" href="https://mybinder.org/v2/gh/mpds-io/mpds-api/gh-pages?filepath=notebooks">interactively online</a> ("ultra-modern way"). <strong>Important</strong>: using the API keys within the third-party Jupyter environment is potentially insecure and should be done with the great care (see <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/notebooks">notebooks</a>).</p>

            <h3 id="Client-library">&sect;2.2. Client library</h3>

            <p class="view">As mentioned, any programming language, able to execute HTTP requests and handle the JSON output, can be employed. However, one of the most frequently used languages in data processing is Python. Therefore we provide a <a href="https://github.com/mpds-io/mpds-client">client library</a> for Python. This library takes care of many aspects of the MPDS API, such as pagination, error handling, validation, proper data extraction and more. We encourage our users to adopt this library for their needs. It is installed as any other Python library:</p>

            <div class="blackbg">pip install mpds_client</div>

            <p class="view">The exercises below use this library. Some of them might also employ <span class="t">numpy</span>, <span class="t">scipy</span>, and <span class="t">pandas</span> libraries (a popular data science toolbox). As a necessary preparation, we save the MPDS API key in an environment of our terminal session. Later the Python API client reads the key from there, if not explicitly provided. For Bash:</p>

            <div class="blackbg">export MPDS_KEY=...</div>

            <p class="view">Note, that there are three fundamental data types at the MPDS concerning the data origin: <strong>peer-reviewed</strong>, <strong>in-house machine-learning</strong>, and <strong>in-house first-principles simulations</strong>. By default only the peer-reviewed data are always supplied in the API. To change this using the Python library, one has to setup the <span class="t">dtype</span> attribute in advance. As seen below, at first all the data types are enabled, then only the machine-learning type is enabled:</p>

            <div class="blackbg">from mpds_client import MPDSDataRetrieval, MPDSDataTypes
client = MPDSDataRetrieval()
client.dtype = MPDSDataTypes.ALL
...
client.dtype = MPDSDataTypes.MACHINE_LEARNING
...</div>

            <p class="view">The data type is set up by a client with the <span class="t">dtype</span> query string parameter.</p>

            <h3 id="Probability-density">&sect;2.3. Bond length analysis</h3>

            <p class="view">Let us say we want to study the chemical bond of A and B elements, <i>e.g.</i> uranium and oxygen. What is the length or lengths of this bond the most frequently reported in the world scientific literature? MPDS contains more than 2500 crystalline structures with uranium and oxygen, so why should not we perform a quick data investigation to answer this question. Starting with the necessary imports in Python:</p>

            <div id="exercise_a1" class="view">
            <pre><code>#!/usr/bin/env python
import pandas as pd
from mpds_client import MPDSDataRetrieval</code></pre>
            </div>

            <p class="view">We use a naive brute-force approach to calculate bond lengths between particular atom types in a crystalline environment. Obviously, we are interested in neighboring atoms only, so we do not consider interatomic distances more than, let us say, 4 &#8491;. We then represent a crystalline structure with the <span class="t">ase</span>'s <span class="t">Atoms</span> class and calculate distances using its <span class="t">get_distance</span> method. Note rounding of distances marked by comment <i>NB</i> in the code:</p>

            <div id="exercise_a2" class="view">
            <pre><code>def calculate_lengths(ase_obj, elA, elB, limit=4):
    assert elA != elB
    lengths = []
    for n, atom in enumerate(ase_obj):
        if atom.symbol == elA:
            for m, neighbor in enumerate(ase_obj):
                if neighbor.symbol == elB:
                    dist = round(ase_obj.get_distance(n, m), 2) # NB occurrence <-> rounding
                    if dist < limit:
                        lengths.append(dist)
    return lengths</code></pre>
            </div>

            <p class="view">
            Note that the crystalline structures are not retrieved from MPDS by default, so we need to specify additional four fields: <span class="t">cell_abc</span>, <span class="t">sg_n</span>, <span class="t">basis_noneq</span>, <span class="t">els_noneq</span>.

            On top of that, we also ask for the <span class="t">phase_id</span>'s, MPDS entry numbers, and chemical formulae. Note that <span class="t">get_data</span> API client method returns a usual Python list, whereas <span class="t">get_dataframe</span> API client method returns a Pandas dataframe. We use the former below:
            </p>

            <div id="exercise_a3" class="view">
            <pre><code>client = MPDSDataRetrieval()

answer = client.get_data(
    {"elements": "U-O", "props": "atomic structure"},
    fields={'S':[
            'phase_id',
            'entry',
            'chemical_formula',
            'cell_abc',
            'sg_n',
            'basis_noneq',
            'els_noneq'
            ]}
)</code></pre>
            </div>

            <p class="view">
            The <span class="t">MPDSDataRetrieval.compile_crystal</span> API client method helps us to handle the crystalline structure in the <span class="t">ase</span>'s <span class="t">Atoms</span> flavor. A popular <a href="https://github.com/materialsproject/pymatgen">Pymatgen</a> library can be also used instead. Mind however that these two libraries are generally incompatible, so never mix crystalline structures in <span class="t">ase</span> and <span class="t">Pymatgen</span> flavors.</p>

            <p class="view">
            We then call the <span class="t">calculate_lengths</span> function defined earlier:
            </p>

            <div id="exercise_a4" class="view">
            <pre><code>lengths = []

for item in answer:
    crystal = MPDSDataRetrieval.compile_crystal(item, 'ase')
    if not crystal: continue
    lengths.extend( calculate_lengths(crystal, 'U', 'O') )</code></pre>
            </div>

            <p class="view">
            That runs a little bit slow (about five minutes), since <span class="t">ase</span>'s <span class="t">Atoms</span> are expectedly not performing very well on dozens of thousands of bond length calculations. We may want to either use the <span class="t">ase</span>'s <span class="t">neighbor_list</span> method, or employ a C-extension, but this is outside the scope of this exercise. Anyway now we have a flat list <span class="t">lengths</span>. Let's convert it into a Pandas Dataframe and find which <strong>U-O</strong> distances occur more often than the others.
            </p>

            <div id="exercise_a5" class="view">
            <pre><code>dfrm = pd.DataFrame(sorted(lengths), columns=['length'])
dfrm['occurrence'] = dfrm.groupby('length')['length'].transform('count')
dfrm.drop_duplicates('length', inplace=True)</code></pre>
            </div>

            <p class="view">
            What did we do here? We calculated the numbers of occurrences (counts) of each particular <strong>U-O</strong> length and then augmented our dataframe with this new information, creating a new column <span class="t">occurrence</span>. (The reader remembers that we have rounded the interatomic distances calculated.)
            </p>

            <p class="view">
            Now it's up to the reader to visualize the result (this is absolutely a matter of taste). Below is a bar chart rendered using <a href="https://plot.ly">Plotly</a> and <a href="https://d3js.org">D3</a> JavaScript libraries. We provide <a href="#Visualizations">visualization exporting toolbox</a> in Python, allowing the export of the results into <strong>CSV</strong> and <strong>JSON</strong> formats and reproducing the figures, although its usage is optional.
            </p>

            <img loading="lazy" decoding="async" id="F1" src="https://developer.mpds.io/figures/1.png" alt="Probability density distribution of U-O bond length across the MPDS database" width="800" />

            <div class="caption" data-id="F1"><b>Fig. 1.</b> Distribution of U-O bond lengths across MPDS database.</div>

            <p class="view">
            We can see now that the most frequent bond lengths between the neighboring uranium and oxygen atoms are 1.78 and 2.35 &#8491;. This excellently agrees with the well-known <a href="http://canmin.geoscienceworld.org/content/35/6/1551">study of Burns <i>et al.</i></a>, done in 1997. However, Burns considered only 105 structures, and we did more than 2500, confirming very well although his findings on the <a href="https://en.wikipedia.org/wiki/Uranyl">uranyl ion</a> geometry.
            </p>

            <h3 id="Clustering">&sect;2.4. Clustering the band gaps</h3>

            <p class="view">
            Unsupervised grouping the values into the clusters in a many-dimensional space is a classical mathematical problem. In an applied data science it is successfully managed using <i>e.g.</i> such algorithms as <a href="https://en.wikipedia.org/wiki/K-means_clustering"><i>k</i>-means</a> and <a href="http://scikit-learn.org/stable/modules/mixture.html">Gaussian mixtures</a>.
            </p>

            <p class="view">
            Taking the non-conducting binary compounds, let us find the clusters in their band gaps and the periodic groups of their elements. That is, we obtain the band gaps and corresponding pairs of the chemical elements, forming a binary compound, and then, for the series of values <strong>group of the first element</strong> &mdash; <strong>group of the second element</strong> &mdash; <strong>band gap</strong> we apply <i>k</i>-means. Given a set of points (in our case, in a three-dimensional space), <i>k</i>-means aims to partition the points into a number of sets, in order to minimize the within-cluster sum of distance functions of each point to the cluster center.
            </p>

            <p class="view">Here we explicitly specify some fields of P-entries we are interested in. As said, all the fields are given in the <a href="#JSON-schemata">JSON schema</a>. Filtering by the units (eV), we get rid of some band gap derivatives, which are also present in the output. Then we remove a couple of unphysical values (negative or more than 20 eV) that are probably wrongly reported:</p>

            <div id="exercise_b1" class="view">
            <pre><code>#!/usr/bin/env python
from ase.data import chemical_symbols
from retrieve_MPDS import MPDSDataRetrieval
from kmeans import Point, kmeans, k_from_n # standalone K-means implementation

client = MPDSDataRetrieval()

dfrm = client.get_dataframe(
    {"classes": "binary", "props": "band gap"},
    fields={'P': [
        'sample.material.chemical_formula',
        'sample.material.chemical_elements',
        'sample.material.condition[0].scalar[0].value',
        'sample.measurement[0].property.units',
        'sample.measurement[0].property.scalar'
    ]},
    columns=['Formula', 'Elements', 'SG', 'Units', 'Bandgap']
)
dfrm = dfrm[dfrm['Units'] == 'eV']
dfrm = dfrm[(dfrm['Bandgap'] > 0) & (dfrm['Bandgap'] < 20)]</code></pre></div>

            <p class="view">
            Note, that there might be different values of the band gaps reported in the literature for each particular compound. Clearly, this is because we have not specified the character of the desired values (direct or indirect band gap), but even apart of that variations are expected, <i>e.g.</i> due to slightly different experimental conditions. Here we simply take the average of all the values per a compound, although there are of course better ways to handle that:
            </p>

            <div id="exercise_b2" class="view">
            <pre><code>avgbgfrm = dfrm.groupby('Formula')['Bandgap'].mean().to_frame().reset_index() \
    .rename(columns={'Bandgap': 'AvgBandgap'})
dfrm = dfrm.merge(avgbgfrm, how='outer', on='Formula')
dfrm.drop_duplicates('Formula', inplace=True)</code></pre></div>

            <p class="view">We now have a lot of different binary compounds with the averaged values of band gap. So our data are ready for clustering. This is how we get the periodic element groups:</p>

            <div id="exercise_b3" class="view"><pre><code>def get_element_group(el_num):
    if el_num == 1: return 1
    elif el_num == 2: return 18

    elif 2 < el_num < 19:
        if (el_num - 2) % 8 == 0: return 18
        elif (el_num - 2) % 8 <= 2: return (el_num - 2) % 8
        else: return 10 + (el_num - 2) % 8

    elif 18 < el_num < 55:
        if (el_num - 18) % 18 == 0: return 18
        else: return (el_num - 18) % 18

    if 56 < el_num < 72: return 3 # custom group for lanthanoids
    elif 88 < el_num < 104: return 3 # custom group for actinoids
    elif (el_num - 54) % 32 == 0: return 18
    elif (el_num - 54) % 32 >= 17: return (el_num - 54) % 32 - 14
    else: return (el_num - 54) % 32</code></pre></div>

            <p class="view">
            And this is how we define the points for clustering based on our data:
            </p>

            <div id="exercise_b4" class="view"><pre><code>fitdata = []

for n, row in dfrm.iterrows():
    groupA, groupB = \
        get_element_group(chemical_symbols.index(row['Elements'][0])), \
        get_element_group(chemical_symbols.index(row['Elements'][1]))
    fitdata.append(
        Point(
            sorted([groupA, groupB]) + [round(row['AvgBandgap'], 2)],
            reference=row['Formula']
        )
    )
clusters = kmeans(fitdata, k_from_n(len(fitdata)))</code></pre></div>

            <p class="view">We use a <a href="https://github.com/mpds-io/mpds-api/tree/gh-pages/kickoff/kmeans.py">standalone</a> <i>k</i>-means Python implementation <span class="t">kmeans</span>, although the reader might want to employ <span class="t">sklearn</span> or <span class="t">scipy</span>. Note, that the <i>k</i>-means algorithm does not provide the number of clusters to divide our data. We try to guess this number from the size of our data naively using <span class="t">k_from_n</span> function. For homogeneous data this guess is likely to be wrong, but for heterogeneous data it should make sense.</p>

            <p class="view">We then export the data in a flat Python list for plotting as follows. The plot below is prepared using the <a href="#Visualizations">visualization exporting toolbox</a>.</p>

            <div id="exercise_b5" class="view"><pre><code>export_data = []

for cluster_n, cluster in enumerate(clusters, start=1):
    for pnt in cluster.points:
        export_data.append(pnt.coords + [pnt.reference] + [cluster_n])</code></pre></div>

            <br />

            <img loading="lazy" decoding="async" id="F2" src="https://developer.mpds.io/figures/2.png" alt="Binary compounds clustering by their band gaps and constituent element groups" width="600" />

            <div class="caption" data-id="F2"><b>Fig. 2.</b> Binary compounds clustering by their band gaps and constituent element groups.</div>

            <br />

            <h3 id="QSAR-QSPR">&sect;2.5. Structure-property correlations</h3>

            <p class="view">In this excercise we quantitatively estimate the relationship between a physical property of interest and the atomic structure. This is the very basic task in chemoinformatics, called QSAR/QSPR study, which stands for the quantitative structure-activity or structure-property relationship modeling. Technically, we request some physical property from MPDS and try to figure out, whether it depends on the crystalline structure, and if yes, how much.</p>

            <p class="view">Here the question "how much" is answered by a correlation coefficient, which can be defined either as a measure of how well two datasets fit on a straight line (<a href="https://en.wikipedia.org/wiki/Pearson_correlation_coefficient">Pearson coefficient</a>), or difference between concordant and discordant pairs among all the possible pairs in two datasets (<a href="https://en.wikipedia.org/wiki/Kendall_rank_correlation_coefficient">Kendall's tau coefficient</a>). Both Pearson and Kendall's tau correlation coefficients are defined between <span class="t">-1</span> and <span class="t">+1</span>, with <span class="t">0</span> implying no correlation. For the Pearson coefficient, values of <span class="t">-1</span> or <span class="t">+1</span> imply an exact linear relationship. Positive correlations imply that as <i>x</i> increases, so does <i>y</i>. Negative correlations imply that as <i>x</i> increases, <i>y</i> decreases. For the Kendall's tau coefficient, values close to <span class="t">+1</span> indicate strong agreement, values close to <span class="t">-1</span> indicate strong disagreement. Calculation of both these correlation coefficients is provided by <span class="t">pandas</span> Python library.</p>

            <p class="view">As usual, we start with the Python imports:</p>

            <div id="exercise_c1" class="view"><pre><code>#!/usr/bin/env python
from __future__ import division

import numpy as np
import pandas as pd
from ase.data import chemical_symbols, covalent_radii

from retrieve_MPDS import MPDSDataRetrieval</code></pre></div>

            <p class="view">We then present two <i>descriptors</i> for the crystalline structures. The term <i>descriptor</i> stands for the compact information-rich number, allowing the convenient mathematical treatment of the encoded complex data (crystalline structure, in our case). We employ an <a href="https://en.wikipedia.org/wiki/Atomic_packing_factor">APF</a> descriptor, as well as topological <a href="https://en.wikipedia.org/wiki/Wiener_index">Wiener index</a>, defined per a unit cell. It is up to the reader to check the <a href="https://doi.org/10.1103/PhysRevB.89.205118">other descriptors</a> for the crystalline structures.</p>

            <div id="exercise_c2" class="view"><pre><code>def get_APF(ase_obj):
    volume = 0.0
    for atom in ase_obj:
        volume += 4/3 * np.pi * covalent_radii[chemical_symbols.index(atom.symbol)]**3
    return volume/abs(np.linalg.det(ase_obj.cell))

def get_Wiener(ase_obj):
    return np.sum(ase_obj.get_all_distances()) * 0.5</code></pre></div>

            <p class="view">Note that our descriptors generally depend on whether we use <span class="t">compile_crystal(item, 'ase')</span> or <span class="t">compile_crystal(item, 'pmg')</span>. This is because by default <span class="t">ase</span> and <span class="t">Pymatgen</span> libraries handle crystalline structures in a slightly different manner. The reader is welcomed to edit the code above for the <span class="t">Pymatgen</span> library.</p>

            <p class="view">Concerning the property being investigated, we restrict ourselves with some materials classes, in order to reduce the time needed for descriptor calculations:</p>

            <div id="exercise_c3" class="view"><pre><code>client = MPDSDataRetrieval()

dfrm = client.get_dataframe({
    "classes": "transitional, oxide",
    "props": "isothermal bulk modulus"
})
dfrm = dfrm[np.isfinite(dfrm['Phase'])]
dfrm = dfrm[dfrm['Units'] == 'GPa'] # cleaning
dfrm = dfrm[dfrm['Value'] > 0] # cleaning

phases = set(dfrm['Phase'].tolist())
answer = client.get_data(
    {"props": "atomic structure"},
    phases=phases,
    fields={'S':[
            'phase_id',
            'entry',
            'chemical_formula',
            'cell_abc',
            'sg_n',
            'basis_noneq',
            'els_noneq'
            ]}
)
descriptors = []

for item in answer:
    crystal = MPDSDataRetrieval.compile_crystal(item, 'ase')
    if not crystal: continue
    descriptors.append(( item[0], get_APF(crystal), get_Wiener(crystal) ))

descriptors = pd.DataFrame(descriptors, columns=['Phase', 'APF', 'Wiener'])</code></pre></div>

            <p class="view">Now we have got the physical property values in the <span class="t">dfrm</span> dataframe and the structure descriptors in the <span class="t">descriptors</span> dataframe. The reader remembers that in MPDS the physical properties and crystalline structures are related via the distinct phase concept, <i>i.e.</i> the same phase identifiers, which we have saved in advance in the <span class="t">Phase</span> columns of <span class="t">dfrm</span> and <span class="t">descriptors</span> dataframes. As the MPDS might have several crystalline structures per a physical property value (or vice versa), we take an average (mean) of all the respective quantities within a distinct phase. After that we join our dataframes using the <span class="t">Phase</span> column:</p>

            <div id="exercise_c4" class="view"><pre><code>d1 = descriptors.groupby('Phase')['APF'].mean().to_frame().reset_index()
d2 = descriptors.groupby('Phase')['Wiener'].mean().to_frame().reset_index()
dfrm = dfrm.groupby('Phase')['Value'].mean().to_frame().reset_index()

dfrm = dfrm.merge(d1, how='outer', on='Phase')
dfrm = dfrm.merge(d2, how='outer', on='Phase')

dfrm.drop('Phase', axis=1, inplace=True)
dfrm.rename(columns={'Value': 'Prop'}, inplace=True)</code></pre></div>

            <p class="view">And this is how the both correlation coefficients are calculated:</p>

            <div id="exercise_c5" class="view"><pre><code>corr_pearson = dfrm.corr(method='pearson')
corr_kendall = dfrm.corr(method='kendall')

print("Pearson. Prop vs. APF = \t%s" % corr_pearson.loc['Prop']['APF'])
print("Pearson. Prop vs. Wiener = \t%s" % corr_pearson.loc['Prop']['Wiener'])
print("Kendall Tau. Prop vs. APF = \t%s" % corr_kendall.loc['Prop']['APF'])
print("Kendall Tau. Prop vs. Wiener = \t%s" % corr_kendall.loc['Prop']['Wiener'])</code></pre></div>

            <p class="view">Expectedly, we can see some correlation between the selected property (namely, <i>isothermal bulk modulus</i> for the <i>oxides of transitional elements</i>) and two structural descriptors. However it varies considerably with the different materials classes, since the crystalline structure also changes considerably. Using the intuition, may the reader suggest the <a href="/#hierarchy">physical property</a> showing higher correlation with either of two structural descriptors within a certain materials class &mdash; or no correlation at all?</p>

            <br />

            <h3 id="Visualizations">&sect;2.6. Visualizations</h3>

            <p class="view">Although the reader is encouraged to visualize the data using his habitual tools, we provide a set of helper utilities. Using a simple exporting toolbox in our Python client library each of the exercises considered above may output two files for the further plotting: <strong>CSV</strong> and <strong>JSON</strong>.</p>

            <p class="view">By default these two files are written in a system-wide temporary directory <span class="t">/tmp</span> (subdirectory <span class="t">_MPDS</span>). <strong>CSV</strong> is commonly used in the electronic sheets (such as OpenOffice Calc), and <strong>JSON</strong> has a custom self-explanatory layout suitable for the <a href="https://visavis.mpds.io">Vis-&agrave;-vis web-viewer</a>. This is a browser-based JavaScript application, heavily used inside the MPDS GUI. It employs <a href="https://plot.ly">Plotly</a> and <a href="https://d3js.org">D3</a> visualization libraries. All <strong>JSON</strong> produced with the exporting toolbox may be simply drag-n-dropped in the browser window with the loaded Vis-&agrave;-vis.</p>

            <p style="background:#f4fbff;">We thank the reader for the time and interest! Any questions or feedback is <a href="mailto:feedback@tilde.pro">very welcomed and greatly appreciated</a>.</p>
        </div>
    </div>

    <div id="logo_bg"></div>

    <div id="footer">Created by <a href="https://mpds.io" target="_blank">Materials Platform for Data Science</a> and <a href="https://tilde.pro" target="_blank">Tilde MI</a>. Dark mode is <span id="darkmode_trigger"></span>.</div>

    <div class="github-fork-ribbon-wrapper right">
        <div class="github-fork-ribbon">
            <a href="https://github.com/mpds-io/mpds-api" target="_top">See on GitHub</a>
        </div>
    </div>

    <!-- CROSS-SITE COMMS -->
    <iframe id="comms" src="https://absolidix.com/comms.html" style="width:0;height:0;border:none;position:absolute;"></iframe>
</body>
</html>