Chips are too difficult.

The environmental impact of the digital industry has increasingly drawn attention, and the growth of digital products and services in a globalized society has exacerbated this impact. The materiality of the digital industry is typically reflected in the environmental impact of mining activities, indicating that digitalization does not mean dematerialization. Although the mining industry is crucial, this discussion is usually limited to a few minerals that have become symbols of the extractive industry (e.g., cobalt, lithium).

This article will further explore the materiality of the digital industry based on the diversity of elements in the semiconductor industry and its purity requirements. Semiconductors are responsible for manufacturing the key components of the digital industry, namely microchips. Given the very specific demand for ultra-high-purity materials in the semiconductor industry, we conducted research on some global companies, revealing new key players in complex supply chains. This highlights the strong dependence of the mining industry on other large-scale manufacturing industrial sectors and the need for a more in-depth study of its interaction with the chemical industry (which complements the mining industry).

Introduction

In recent years, there has been growing concern about the environmental impact of the digital domain, including infrastructure (such as data centers, communication networks), products (such as smartphones), and services (such as video-on-demand streaming). Recently, the trend of large-scale integration of artificial intelligence and edge artificial intelligence in many fields has become increasingly evident, further intensifying concerns about environmental issues. Large technology companies have pointed out in their environmental reports that their footprint is expanding.

So far, the materiality of the digital sector (here defined as "all extraction and production chains capable of manufacturing final digital products") has mainly been discussed from the perspective of "mining", indicating that mining activities require some specific raw materials (e.g., lithium/cobalt for batteries) and a large amount of electricity and water for the manufacturing process. However, the latest data from the United Nations shows that the value of key elements used in information and communication (ICT) technologies, such as gallium (Ga), germanium (Ge), indium (In), rare earth elements (REE), selenium (Se), tantalum (Ta), and tellurium (Te), only accounts for 0.77% of all elements mined in 2018 (excluding coal).

Although it is undeniable that ICT products and services would not exist without the production of the mining industry, its mineral demand seems limited at first glance. To gain a more comprehensive understanding of the materiality of the digital industry, it is necessary to further explore the production chain from a complementary perspective. As the core of the digital industry, semiconductor manufacturing has received increasing attention due to its material flow and environmental impact. However, despite growing evidence of environmental impact and pollution, it has not been properly examined. The materiality aspect of this industry has been overlooked, mainly because its supply chain is extremely difficult to identify and analyze. This is mainly due to its complexity and breadth, combined with the opacity of upstream players (miners, refiners, etc.) and technology companies.

However, since all ICT products and services basically rely on these hardware components called microchips, the materiality of the digital industry is deeply rooted in the materiality of the semiconductor industry, as shown in Figure 1.

Figure 1: A simplified view of the semiconductor industry supporting digital goods and services

Microchips can be said to be the most complex products ever manufactured by humans. They not only require the greatest variety of basic materials but also extremely high purity. In fact, technological progress in the semiconductor industry is mainly driven by miniaturization, performance, and cost-effectiveness. These improvements require more diverse materials to improve physical and electrical performance while reducing size.

As technology nodes shrink, not only do transistors (FEOL: front-end process) require new (and old) materials with the highest purity; the interconnect stack (BEOL: back-end process) has also become increasingly special (in terms of material use) and complex. Each generation requires new alloys to achieve interconnects with lower resistivity in smaller technology nodes. In the early 2000s, copper interconnects had only 3 - 4 layers, while now, the metal interconnects in the latest Nvidia or Apple microchips have more than 20 layers, with wiring lengths reaching thousands of kilometers. Figure 2 shows the elemental breakdown of the common BEOL metal stack in modern CMOS technology nodes and demonstrates the complexity of this evolution process. Any information about the most advanced nodes, whether in the production stage (e.g., 2 nanometers) or in the research stage (e.g., 0.8 nanometers), is highly confidential and will definitely add more elements to this figure.

Figure 2: Evolution of BEOL materials by technology node

The common perception that microchips are mainly composed of semiconductors (silicon, germanium, etc.) and a few scattered elements is long outdated. In fact, even industry insiders find it difficult to list all the elements required for modern processors. This list should not only include the elements inside the microchip but also many other elements necessary for manufacturing nanoscale transistors (billions of such transistors are required in a typical high-end processor).

According to our best estimate (Figure 3), the semiconductor industry now requires more than 85% of the non-radioactive elements in the periodic table. In the past 30 years, this shift from a few elements to almost all possible elements has relatively gone unnoticed. Microchip manufacturing not only requires a large amount of materials, but these materials must be extremely pure.

In fact, this means that the concentration of impurities (unwanted elements and defects) must be controlled and reduced to extremely low levels. This is mainly because advanced technologies rely on complex manufacturing processes, and these processes have increasingly high requirements for purity: the smaller the feature size, the relatively larger the size of a single impurity.

For example, the global semiconductor industry association SEMI has developed guidelines for chemical processes and gases from grade A to E. Grade A, B, and C chemicals are suitable for geometric sizes between 1.2 microns and 90 nanometers. For sizes smaller than these, new grades D, E, and above are required, and the impurity detection concentration is required to be lower than one part per trillion (less than 0.1 ppt). These grades could also be called electronic grade (EG), VLSI in the past, and XLSI for sizes below 90 nanometers, defining the chemical purity level required for each size.

The closest equivalent to the purity requirements may come from the pharmaceutical industry, where ppm and ppb purity levels are common, while the requirements in the semiconductor industry are ppt or even more stringent. It is also obvious that producing elements with higher purity requires more processing steps, including the use of other high-purity (usually toxic) gases, as well as energy and water consumption. Given the special and crucial role of purity requirements in microchip manufacturing, ignoring purity considerations will lead to an incomplete understanding of the importance of the digital domain and may underestimate the environmental impact of the semiconductor industry.

Figure 3: Elements used in the semiconductor industry from the 1980s to the 2010s, adapted and updated from O'Connor's previous work

Therefore, this article aims to explore how purity considerations reshape and enhance the current understanding of the materiality of digital products and services. We explore how to examine microchip production from the perspective of purity, revealing its secondary materialization and bottlenecks, which are not only related to environmental impact but also to the industry's supply chain. To this end, we propose a complementary approach that relies on the material characteristics unique to the microchip industry, which has very specific requirements for materials in two aspects: (i) diversity, i.e., the number of elements contained; (ii) (ultra-high) purity requirements. This applies to many materials involved in the semiconductor manufacturing process, including metals and gases, whether they are used in subtractive manufacturing processes or permanently exist on the wafer (and ultimately in the final product). We show how these two characteristics become the key to a more refined approach, enabling the identification and separate analysis of the materiality upstream of the semiconductor industry.

The rest of this article is structured as follows. The "Method" section introduces the research scope, the data collection process, and a case study focusing on four elements. Then, the results of the element purity requirements and the case study analysis are presented respectively. Finally, the "Discussion" section elaborates on the advantages and limitations of the purity-based approach to better understand the importance of the upstream materials in semiconductor manufacturing.

Method

When exploring the importance of the semiconductor industry from the perspective of purity requirements, several questions arise: Which elements are currently used in the manufacturing of digital components? What purity levels do microchip manufacturers need to achieve today? What industrial processes are involved in purification? Where are these processes carried out? To what extent does the high-purity demand change the criticality of elements? How does it differ from other industrial demands? The purity approach proposed in this article involves the estimation of three main characteristics:

• Comparison between the standard-grade purity level and the required semiconductor-grade purity level;
• The industrial manufacturing processes required to achieve high purity levels;
• The material requirements of wafer fab equipment (e.g., the gases used in excimer lasers in DUV lithography machines).

The list of elements in Figure 3 comes from a small amount of information released by TSMC and Intel, the two largest foundries in the world. However, it may be extremely difficult to further break down the specific uses of these elements (e.g., for doping, deposition, patterning, etching, substrates). There may also be significant differences in use cases between different technology nodes and foundries.

It should be noted that although we did not find a similar high-level periodic table overview in the scientific literature, this is still a non-exhaustive overview of the elements used in current semiconductor manufacturing, which can be expanded through further research and supplemented with industrial data. Once purity requirements are overlaid, such an analysis can help understand the material requirements of the semiconductor industry from a completely different perspective.

Since it is difficult to find a unified information source on industrial and semiconductor purity levels, we had to reconstruct a database from heterogeneous sources. We first collected information from the catalogs of major suppliers: gas suppliers include Linde Gas, Sumitomo Seika, and Airgas; suppliers of elements used in sputtering targets include Umicore, Stanford Materials, Applied Materials, Honeywell, and Materion; compound suppliers include Solvay and the American Chemistry Council.

Subsequently, based on the preliminary investigation, we conducted a specialized search for each element in the professional scientific literature and patent databases on purification processes. In addition, we also used government reports or institutional literature from research institutions as much as possible to fill in the gaps.

A: Scope

This study focuses on semiconductor manufacturing because it provides some key hardware components required for digital products and services. Since the semiconductor industry concentrates an important part of the material flow in the digital domain, it becomes an area worthy of attention. More specifically, we focus on microchip production and its related industrial manufacturing processes, because many of these processes are subtractive manufacturing, i.e., only a small part of the materials will ultimately be used in the final product.

B: Data collection on purity requirements

Like many industries, the semiconductor industry also has its trade secrets. Considering that modern microchips may be the most complex mass-produced devices ever, their confidentiality is usually much higher than that of other industries. Since a few major players (TSMC, Samsung, SMIC, Intel, etc.) are capable of producing the most advanced technology nodes, it is not surprising that there is little detail about the actual process steps and the formulations used.

To understand which elements are currently used in semiconductor manufacturing and the purity levels currently required by microchip manufacturers, we referred to the public documents of the major players in the industry (TSMC, STMicroelectronics, ASML, Applied Materials) and the catalogs and data sheets of industrial suppliers (Umicore, Materion, Linde, SAMaterials, Solvay, Air Products). Purity specifications vary by material and are usually provided in industrial catalogs.

The standard grade (also known as industrial grade, corresponding to the purity of the maximum production of elements) has lower purity requirements (95 - 99%), while the highest requirements for the semiconductor grade are much higher (usually higher than 99.999% or 5N). Wafer fabs that manufacture microchips usually source ultra-pure materials from other suppliers upstream in their value chain. Since the purity requirements of other industries are not as high as those of the semiconductor industry, it can be assumed that most of these upstream industries are mainly used for microchip production.

Based on these data, it is possible to estimate which elements are used, the purity levels, and the standard industrial purity levels outside the semiconductor industry. A lot of work is still needed to understand the specific role of key wafer fab equipment in different technology nodes. This involves continuously monitoring the generations of the fab's manufacturing equipment to determine the basic elements that need attention.

Results

This section first presents the results of our analysis of all elements in the periodic table and their purity requirements in semiconductor manufacturing. Then, the case study results provide a more in-depth analysis of four specific elements used in semiconductor manufacturing.

A: Elements and purity requirements in semiconductor manufacturing

Figure 4 summarizes our best speculation on the purity requirements for modern microchip manufacturing. Although it is impossible to retrieve data for all elements in the periodic table, this list covers most non-radioactive elements. Obviously, the purity requirements for semiconductor manufacturing depend on the element itself. For example, the requirement for Si may be higher than 11N (99.999999999%), while the requirement for Au is limited to 4 - 5N. The purity levels reported in Figure 4 consider the commonly used semiconductor-grade purity levels.

However, it should be noted that the same material may have different uses in different wafer fabs and technology nodes. For example, ultra-pure silicon is used as a substrate throughout the semiconductor industry, but it is also used to coat the mirrors in EUV lithography. An important piece of information is that the high-purity requirements of the semiconductor industry are not limited to a few subsets of elements; they cover almost the entire periodic table. It can be safely assumed that no other industry uses such a wide range of elements in its supply chain, and certainly not at this level of purity.

Figure 4: A non-exhaustive overview of the main elements in the semiconductor industry (upstream). The purity requirements of elements are indicated by orange shading

If many other industries also have similar requirements, the demand for such high-purity materials may not necessarily be large. Therefore, we compared the common semiconductor-grade purity requirements with the highest purity levels in non-semiconductor industrial literature. As shown in Figure 5, the comparison results show an additional increase of N in purity. For example, the maximum purity requirement for sulfur in the semiconductor industry is 3N higher than that in other industries. First, the figure shows that the gap between industrial and semiconductor purity requirements is uneven, ranging from a small gap for some elements (e.g., Cu, Ne) to a significant gap for other elements (e.g., Si, Ge, Ga). Although further research on each element is needed to understand the potential manufacturing processes for improving purity, this ranked list provides valuable insights into element-level purity requirements and is a good starting point for prioritizing data collection and model improvement in the context of environmental LCA databases. The purity requirement for silicon is the highest so far, but the purity requirements for many other elements have also increased by 2N or more than 3N.

Finally, Figure 5 also highlights the bottleneck of data scarcity, because for many elements, it is not possible to obtain both industrial-grade and semiconductor-grade purity requirements (with sufficient credibility). Although we do not have reliable data on tracking the purity requirements of individual technology nodes, it can be assumed that smaller nodes require higher-purity materials in each process step, and this trend will continue in the future.