Fundamental Manufacturing Process Innovation Changes the World
We describe fundamental manufacturing process innovation (FMPI) and show that it is central to technological and societal progress. We identify the mechanisms by which FMPI occurs and also explain why it is frequently overlooked and/or misunderstood. By analyzing eight such innovations that shaped the modern world, we highlight the scope of FMPI and reveal its common characteristics. We aim to stimulate a conversation about FMPI, assess its frequency and range of impact, and consider methods for incentivizing it in the future.
Introducing FMPI
Technological innovation has dramatically expanded the availability and lowered the cost of energy, materials, food, healthcare, and information. Life expectancy is now two times greater than in 1900 [1], the percentage of people living below the poverty line has plummeted from 90% in 1820 to less than 10% in 2015 [2], and global conflict is at or near an all-time low [3]. At the same time, there is increasing concern that societal progress may be in jeopardy due to a slowdown in the rate of technological innovation [4–6]. Thus, debate has renewed about innovation’s underlying mechanisms and how to incentivize them. Here, we argue that recognizing and focusing on fundamental manufacturing process innovation (FMPI) — rethinking the basic strategy for organizing and executing a series of manufacturing steps — represents a significant opportunity to enhance our future.
Let us start with an everyday example of fundamental process innovation to emphasize its essential nature. Shopping is a strikingly pervasive yet almost entirely hidden process. At the turn of the 20th century, a single shopkeeper fetched multiple items from shelves behind a counter for each customer. Shopping transitioned to multiple customers traversing rows of aisles, each with a cart into which they place items. This process innovation refactored the tasks of the shopping process, with the customer taking on tasks previously carried out by the shopkeeper. The innovation created a significant cost savings for the shopkeeper that was passed on the customer. This change also impacted the supply chain and customer behavior [7]. FMPIs are changes in the way we generate new configurations of matter with a similar character and magnitude of impact.
FMPI fuels technological and societal transformation by opening a path to scalability. Scalable manufacturing processes can decrease costs and increase production rates by orders of magnitude relative to the prior state-of-the-art. Like many innovations, scalability enables quantitative change — more products at lower cost. More significantly, scalability can propel qualitative change because the magnitude of the decrease in cost or increase in deployment can be so large — effectively infinite compared to that previously possible. Qualitative change opens the door to entirely new capabilities and use cases.
Despite its potential for tremendous impact, FMPI is frequently overlooked and/or misunderstood. Why? There are analogous reorganizations of process steps in almost all areas of the economy: finance [8], transportation [9], and information technology [10]. While innovations in these domains may be referred to as models, strategies, or algorithms, the underlying mechanisms are similar. One major reason stems from the type of “product” in each case. Process innovations in the service sector are inherently more visible because the process — exchanging money, moving goods, or processing data — is the product. In manufacturing, however, FMPIs are obscured because the end goal is a physical product. Thus, the focus naturally falls on the product rather than the process that created it.
In the present work, we define FMPI and propose an initial framework to understand how these innovations occur. We examine eight manufacturing innovations that shaped the modern world and show how they fit within the FMPI framework. These examples are not usually examined in the context of process innovation; however, when viewed in that light, they reveal a set of characteristics common to FMPI. To be clear, not all world changing innovations are FMPIs. However, many important ones are far more about “process” than is usually recognized and increasing our focus on these mechanisms promises to accelerate technological and societal progress. To this end, we seek to build a community interested in this topic that will allow us to 1) identify FMPI across a range of science and engineering disciplines, 2) further understand the underlying mechanisms and advance a set of core principles, 3) determine whether FMPI is uncommon or simply hidden, thus enabling a more accurate assessment of its overall importance to technological progress, and, assuming it is important, 4) develop methods for incentivizing it in the future.
The iceberg of FMPI
Like an iceberg, the results of FMPI are in plain sight but the innovation itself is hidden below the surface. To reveal the iceberg’s depth, we identify and discuss the hurdles of 1) transiency, 2) specialization, and 3) trivialization.
Transiency. FMPI rarely leaves an enduring mark in history. Materials and products, on the other hand, can remain visible long after their production. Periods of human civilization were named for the prominent materials of the day (e.g., stone, bronze). However, as presented in Figure 1, the processes enabling the efficient manufacture of those materials (e.g., cutting, smelting) are hidden. Early manmade tools (e.g., arrowheads, knives) are still found, yet the process innovations that enabled them are invisible. The fleeting nature of process is further exemplified by the Romans’ advent and use of concrete [11]. Because its production process was eventually forgotten, many stood in front of the remarkable Roman structures with no ability to replicate them. The embedding of processes in products resigns the process to obscurity or speculation.
Specialization. FMPI operates in every science or engineering discipline and technical domain where matter is manipulated or transformed. The breadth of its application makes it easy for specialized practitioners to miss it. Today, when FMPI is considered, it usually occurs within a specific discipline or technical domain. For example, process engineering is frequently discussed and used in chemical, industrial, and systems engineering. Similarly, the steam engine is usually included in discussions of energy and the integrated circuit is considered within electronics. The stumbling block, we argue, is that FMPI does not belong to any specific field of endeavor but is an enabler of them all. Mathematics is a good analogy. It is both a separate discipline and an essential component of many others.
Trivialization. FMPI is thought to be ‘just’ a rearrangement of steps, too simple to be responsible for major technological change. This misperception is not surprising. Many tales are of heroic experiments, legions of robots, or enormous computational power. Thus, it is easy to dismiss FMPI as irrelevant, uninteresting, or unworthy of attention. While FMPIs may be straightforward in terms of their overall mechanism, their identification and implementation require science and engineering to converge in new and unanticipated ways. When viewed from this perspective, FMPI offers a complexity and richness that rivals any other class of innovation.
The debate is not immune to iceberg problems. It is a struggle to see FMPI; it is fleeting and we all wear the blinders of our own training. We believe that only through a multidisciplinary community can the study of FMPI rise to the surface.
A framework to understand FMPI
Understanding how FMPI occurs is the first step toward recognizing and nurturing it. To this end, we propose an initial set of schemas to classify FMPI. First, we define what we mean by “process,” which, in its most general sense, is an interconnected series of steps that transforms an initial state to a final state. In manufacturing, a process converts inputs (a collection of raw materials or precursors) to outputs (products). The first process innovation was a recognition of the process itself, separate from the artisan who created it. Subsequently, the modularization and standardization of the parts and the process steps themselves opened the door to mass-production. Rather than relying on a small number of craftspeople, each skilled in many aspects of production, tremendous efficiencies resulted from the work of highly-specialized artisans, each of whom focused on the production of specific, standard parts. Evidence from the Terracotta Army unearthed in Shaanxi Province, China suggests that these ideas emerged as early as the 2nd century BCE [12]. The Venetians applied modular and standard parts to revolutionize ship building during the early 16th century [13], supporting their naval dominance in the Mediterranean for two centuries. The same concepts were deployed for artillery by Jean-Baptiste Vaquette de Gribeauval in France [14] and for rifles by Simeon North and John Hall in the United States [15] in the 18th and 19th centuries. Henry Ford later popularized the use of modular parts in his assembly line for automobile manufacturing [16].
FMPI changes the manner of defining, connecting, and arranging basic process steps in time and space. FMPI usually occurs through one or more of the schemas, or mechanisms, illustrated in Figure 2. Parallelization (schema I) allows multiple instances of the same step to be simultaneously carried out. To enable parallelization, flows of material to and from process steps may be split (schema II) or merged (schema III). A separation (schema IV) occurs when flows are selectively partitioned to send entities with different properties to different downstream process steps. Multiple process steps may be combined (schema V) or individual process steps may be factored (schema VI). Note that the simultaneous use of schema V and VI results in refactoring, where process steps are first factored and then combined in new ways. Process steps may themselves change from being subtractive (schema VII) to additive (schema VIII) or vice versa. Subtractive processes selectively remove (e.g., using a mill or lathe) unwanted material from a larger starting volume whereas additive processes combine smaller volumes of material to construct a larger part (e.g., attaching one sub-component to another). Different schemas become more or less preferred over time, as the available technology evolves or external incentives change. Moreover, the advent of new process steps often goes hand in hand with FMPI. A new process step can clarify a schema’s feasibility. Inversely, the recognition of a particular schema can drive the development of a new process step.
Printing processes, which have benefitted from three FMPIs over the last millennium, followed several of the identified schemas. Prior to the 11th century, documents and books were copied by hand in a serial fashion. Movable type was invented in 11th century China [17–18]. It was introduced to Europe by Gutenberg three centuries later [19]. By simultaneously arranging, then inking, and finally pressing the characters or letters required for an entire printed page, movable type ushered in the era of low-cost and high-throughput printing. Moveable type, in its initial, fully manual embodiment, exemplifies parallelization (schema I). Gutenberg’s printing process, whereby the inking and pressing are completed in a single process step, is an example of combination (schema V). In another example of combination (schema V), xerography further accelerated the reproduction of documents by eliminating the need to manually arrange characters or letters. The automation of document feeding further parallelized (schema I) printing by allowing multiple pages to be loaded at the same time. As is common with FMPI, Gutenberg required and made key engineering advances in the materials and processes for fabricating the letters used in the printing process. Similarly, xerography is enabled by new process steps — the charging of a photoconductive film, electrostatic attachment of toner to charged regions, and transfer of toner to a sheet of paper — that became possible due to scientific and engineering advances. Finally, note the modern focus on the printing press or xerox machine rather than the underlying processes that enabled their efficiencies.
Eight FMPIs that changed the world
We now highlight eight FMPIs from the last 250 years that have shaped or are shaping the modern world. These examples establish the extent of FMPI and its common characteristics. In many cases, these innovations are generally not considered process innovations; the original FMPI has been hidden, replaced by the story of the product or the machine that enabled it. These innovations are also usually considered disparately, within the specialized discipline or technical domain from which they emerged. However, we show that they fit into a single framework. We are, by necessity, intentionally revisionist and do not attempt to understand or express the motivations of the original innovators.
1) The Watt process for power generation. James Watt’s process for extracting work from coal by manipulating steam was a core enabler of the industrial revolution [20]. Watt’s innovation, conceived in 1765, reduced the coal used for pumping water from mines by approximately three quarters compared to Thomas Newcomen’s process. The greater efficiency meant that coal could be moved further from the mine head, leading to new forms of industrial mechanization and locomotion. In the Newcomen process, as illustrated in Figure 3, the work is generated by creating a vacuum in a single vessel by condensing its steam. However, this process required the heating and cooling of the singular vessel each cycle [21]. In contrast, the Watt process factors (schema VI) vacuum pulling from steam condensation, with each step being carried out in a separate vessel [20].
2) The Fourdrinier process for continuous papermaking. Paper was originally made through a slow, singular sheet process in which each sheet was pressed in a mold and dried. The continuous papermaking process dramatically lowered the cost and increased the availability of paper and hence mass information. The first patent of the concept was issued to Louis-Nicholas Robert in France in 1799 [22], but the process was first implemented in England by the Fourdrinier brothers a few year later [23]. The sheet forming, pressing, and drying processes were factored (schema VI) into distinct steps along the direction of sheet movement. In the first step, a thin slurry of pulp was poured onto a moving fabric or web. The fabric both supported the slurry during its forward motion and allowed water to drain, consolidating the pulp and forming an initial sheet. The final sheet emerged from a combination of rollers that further dried the paper and controlled its thickness. Papermaking was one of the first continuous manufacturing processes and it later inspired processes for iron and steel sheet-making.
3) The Bessemer process for steel manufacturing. Henry Bessemer’s process for steel manufacturing allowed for the production of high-quality, low-cost steel at scale [24]. First announced in 1856, the Bessemer process allowed steel to be used in applications such as railway tracks, expanding their reach by lowering the cost per mile of track and allowing higher railcar loads.
In the first FMPI, as illustrated in Figure 4, Bessemer combined (schema V) the heating of the raw “pig” iron with the removal of impurity atoms (schema IV), such as carbon, phosphorus, and silicon. Although counterintuitive at the time, blowing air through the system caused impurity oxidation that increased the melt temperature. Impurities, such as carbon, formed gaseous oxides and were separated with the air, combining separation and heating. Less volatile oxides and other impurities were removed in a low-density slag phase that floated on top of the steel.
In a second FMPI, Bessemer completely separated all impurities from the melt (schema IV) and then re-added impurities in specific amounts (schema III) as necessary to achieve the desired properties [25]. In a departure from the traditional processes that attempted to remove impurities in excess of those desired, Bessemer’s process significantly improved the reproducibility of steel manufacture. Notably, the idea of complete separation followed by controlled remixing or blending is widely used in manufacturing processes today from Si wafer production to gasoline formulation.
4) The Houdry process for the catalytic cracking of crude oil. Catalytic cracking of crude oil was first demonstrated by Almer McAfee in 1915, but Eugene Houdry laid the foundation for its commercial success in the 1920s and 1930s. This process enabled high octane gasolines and aviation fuels to be produced from the heavy components of crude, enhancing both the yield and quality of the resulting fractions. During the first two years of World War II, 90% of catalytically cracked gasoline was produced in Houdry plants [26]. The performance of Allied aircraft was substantially enhanced by these fuels. Payloads, for example, could be up to 25% larger per aircraft.
In catalytic cracking, feed oil is vaporized and then cracked over a static bed of acid clay particle catalyst. Deactivation of the catalyst by coke required a periodic regeneration step, where high temperature oxidation removed the coke and reactivated the catalyst. Houdry realized the benefits of factoring (schema VI) the cracking and catalyst regeneration steps into two separate vessels. A bucket elevator system mechanically removed deactivated catalyst, transferred it to the regenerator, and returned reactivated catalyst to the reactor. Fluidized catalytic cracking (FCC) was developed with a similarly factorized (schema VI) catalysis and regeneration system, but where the catalyst particles were fluidized rather than being mechanically transported. FCC allowed for heat integration — recovering the heat of the coke combustion and using it to vaporize the incoming oil. FCC eventually displaced the Houdry process due to the greater efficiencies of mass and heat transport and now produces more than half the gasoline fractions in modern refineries.
5) The planar process for fabricating integrated circuitry. Global information technology spending reached $4.8 trillion in 2018 [27]. Integrated circuits, the processing machines that underpin our information technology infrastructure, are approaching global annual shipments of 1 trillion units and are found in technologies as diverse as data centers, mobile phones, vehicle safety systems, and children’s toys. Jean Hoerni, working at Fairchild Semiconductor in the late 1950s, realized that transistors fabricated entirely inside a Si wafer dramatically improved electronic performance [28]. Robert Noyce then recognized that devices embedded in the wafer left the top surface planar (hence the “planar” process), greatly simplifying their interconnection [28]. As opposed to the fabrication and packaging of discrete transistors, the monolithic integration of transistors (and other devices), which now number in the billions, on a single substrate is an example of parallelization (schema I). A number of new process steps were needed, including photolithography. Parts of the planar process are subtractive (schema VII) while others are additive (schema VIII). Construction of each circuit layer relies on the removal of regions of an initially continuous film to create key circuit features (e.g., transistors and interconnects). Multiple layers of circuitry are sequentially added to create a complete chip. The planar process remains at the heart of integrated circuit manufacturing, today providing remarkable performance per unit cost, functional diversity (e.g., microprocessors, memory, sensors, etc.), and physical compactness.
The next two FMPIs, while fundamentally distinct, are collectively providing unprecedented insights into biology by helping to exonerate the innocent and convict the guilty, and enabling entirely new routes to diagnose and treat disease.
6) The shotgun process for sequencing of DNA. The rapid, low cost sequencing of DNA is often described in the context of sequencing machines and supercomputers; the approach itself is enabled by a FMPI. In the late 1970’s, Joachim Messing and Peter Seeburg recognized that the enormous length of DNA in full genomes would prevent its sequencing with the slow, labor-intensive methods of the time. Messing and Seeburg showed that DNA could be cut into large numbers of fragments and simultaneously sequenced [29]. The full DNA sequence is reconstructed with high fidelity by identifying overlapping sequences from different fragments [30]. Shotgun sequencing combined known methods for cutting and sequencing DNA with a new process step that reconstituted the overall sequence. Such a massively parallelized (schema I) process provides efficiency gains that far outweighs the cost of sequencing base pairs in the overlap regions multiple times.
7) The polymerase chain reaction for DNA amplification. The amplification of DNA via the polymerase chain reaction (PCR) was invented and refined by Kary Mullis, for which he shared the 1993 Nobel Prize in Chemistry [31], and co-workers at the Cetus Corporation in the mid-1980’s. PCR produces double-stranded DNA molecules in vitro with an enzyme that attaches nucleotides to a small starting piece of single-stranded DNA (the “primer”). The resulting sequence is defined by another “template” strand of DNA. In an example of parallelization (schema I), heating separates each as-produced, double-stranded DNA molecule into two single strands that then become templates for the next round of amplification. PCR offered a number of advantages relative to traditional approaches to DNA cloning, the predecessor technology for DNA production. Cloning uses the natural machinery in a (bacterial) cell to replicate, and thus amplify, DNA by culturing the bacteria. The per strand production rate of PCR is orders of magnitude faster than DNA cloning. A single DNA strand can be amplified ~9 orders of magnitude within tens of minutes. DNA amplification using cells only produces additional strands when the entire organism replicates itself, which takes tens of minutes for a single cell division. PCR is also simpler to implement and automate, which makes it useful in many more research and development settings.
8) 3-D printing. 3-D printing is the broad class of processes where parts or entire products are fabricated by sequentially adding material (schema VIII). While additive manufacturing techniques have long been taken for granted at the atomic (e.g., molecular synthesis) and macroscopic (e.g., building construction) scales, they have expanded to the mesoscale (e.g., nm to cm) in recent decades. Hideo Kodama and Alan Herbert explored the use of ultraviolet light to locally cure photosensitive polymers in the early 1980’s. Charles Hull received the first patent for a complete stereolithography process in 1984 [32]. Sales of 3-D printers, the products that embed additive manufacturing processes, doubled between 2005 and 2011 and are projected to top $10 billion by 2021 [33]. Conventional manufacturing techniques, including machining or casting/molding, fall within the subtractive schema. Such techniques often result in material waste and impose constraints on the structure and complexity of parts. Changing from a subtractive to additive schema allows for structural complexity never before possible or achievable in many fewer steps. The simultaneous digitization of design and manufacturing enables highly distributed part production, mass customization, rapid prototyping, smaller production runs, and reduced inventories. The new structures and material combinations possible with 3-D printing are also driving work to fabricate, understand, and leverage materials with never-before-possible properties.
The characteristics of FMPI
FMPIs can set in motion deep technological, scientific, economic, and ultimately societal changes. Using our eight process innovations as a guide, we now distill and discuss these characteristics.
Enabling a pathway to scalability. Process innovations provide the path to scale-up, resulting in widespread adoption. Technological adoption may mean commercialization in some cases or mainstreaming of a new R&D tool in others.
- In the case of continuous papermaking, increasing the speed of the web and making the machine wider supported further scale-up. Modern papermaking lines routinely operate at linear speeds in excess of 1000 meters per minute and approach 10 meters in width.
- The Bessemer process cut the cost of steel by a factor of five in about five years [34]. As displayed in Figure 5(a), it allowed the production volume of steel to increase by three orders of magnitude between 1870 and 1900 [35].
- The initial implementation of the Houdry process permitted a production rate greater than 3,000 barrels/day in 1941; however, rapid improvements in the speed of catalyst transfer increased capacity 20-fold in two years. The first FCC unit was deployed in 1942 and “by the end of World War II, 34 FCC units were operating, with capacity exceeding 500,000 barrels per day in the United States for 20 different companies” [36]. Today, 5 million barrels per day are processed in the United States alone.
- The planar process provided a route to integrate ever-increasing numbers of transistors as well as reduce their size. More than 1 billion transistors are commonplace in modern integrated circuits. As shown in Figure 5(b), this extreme degree of integration has enabled a ~8 orders of magnitude reduction in the cost per transistor since the 1960s [37–38].
- The cost of DNA sequencing has dropped, as seen in Figure 5(c), more than five orders of magnitude, from about $5,000 per megabase pair in 2001 to less than $0.02 today [39]. The cost inflection between 2007 and 2008 results from a transition (in the data) from first to second generation sequencing machines; however, the basic shotgun-type approach remains the same.
- After the initial demonstration of PCR, improvements to the rate of thermocycling and use of the thermostable Taq enzyme [40] reduced process times from hours to minutes.
The process innovation, because it provides a path to scale-up, also marshals resources along a well-defined trajectory. By traversing this path, the research and development community can leverage each other’s efforts to accelerate progress. Key examples of this are the former International Technology Roadmap for Semiconductors (ITRS) [41] and current International Technology Roadmap for Devices and Systems (IRDS) [42], which describe important directions and associated technical barriers in integrated circuit manufacturing via the planar process.
S-curve behavior. FMPIs, like any disruptive technological innovation, exhibit the so-called “S-curve” behavior displayed in Figure 6 [43]. The uptake of the innovation is initially slow; however, after early issues are worked out and its value is more widely recognized and rapidly adopted, it ultimately saturates the market. The causes of the slow uptake period are numerous. Technological innovations cannot initially compete on the basis of conventional metrics [44]. For instance, catalytic cracking initially suffered from catalyst deactivation and limited process up-time. Subtle process modifications or product architectural changes are often needed for the initial process innovation to achieve its full potential but can further slow initial uptake or make early implementations more expensive. Bessemer’s original steel manufacturing process design did not entirely remove impurities, which led patent licensees to struggle with reproducibility. Only after implementing complete impurity separation and controlled reincorporation did Bessemer’s process take off [25]. During the innovation uptake phase, new bottlenecks are often recognized. For example, the advent of continuous paper manufacturing exposed the limitations of rags (i.e., cloth) as the feedstock for paper fibers and, over time, drove the development of wood pulp as a more scalable alternative [22]. 3-D printing technologies are currently near the first inflection point in Figure 6 and will undoubtedly face similar challenges as commercial deployment continues.
All successful technologies eventually reach the top of the S-curve, where the benefits of the innovation have been wrung out. Incumbent technologies, because the cost of further innovation steadily increases during this stage, are particularly at risk for replacement or obsolescence by a process innovation. Consider integrated circuit manufacturing. Today, planar processing is almost universally assumed. However, chip speeds plateaued nearly a decade ago and, more recently, the cost per transistor has struggled to continue its downward march. State-of-the-art fabs now exceed $10B to build, and the development costs for successive chip generations are increasing at a rate of ~50% [45]. Such a price tag is off-limits to all but the largest companies, a situation that creates barriers for new entrants, stifles innovation, increases financial risk, and ultimately drives interest in replacement processes.
Compounding benefits. FMPIs lead to subsequent innovations that compound the benefits of the original. To return to the Bessemer process, the complete separation of impurities from the iron melt allowed for a much wider range of steel compositions and thus properties. Later on, the basic oxygen process, a sustaining innovation of the Bessemer process, emerged. The replacement of air by oxygen further decreased smelting time and drove down capital costs, but the separation of impurities via oxidation remained unchanged. Similarly, early integrated circuits contained a single layer of transistors interconnected with a single layer of metal wiring. However, the addition of more layers was a straightforward extension of the planar process. Today’s integrated circuits contain as many as 10 layers of metal wires. Moreover, the planar process is now used to fabricate products other than integrated circuits, including microelectromechanical machines (MEMs) and, most recently, integrated photonics. In another example, the original shotgun DNA sequencing methods have been largely supplanted by more advanced sequencing techniques. For instance, hierarchical shotgun sequencing reduced the computational intensity and improved the reliability of sequencing by first creating a low-resolution genome map that reduced the number of fragments to be sequenced.
Enabling emergent phenomena. FMPIs can reveal previously unrecognized phenomena because they permit a scale of behavior never before possible. We are comfortable with this “more is different” [46] mentality in scientific disciplines. For example, biology is distinct from chemistry, although it emerges from combinations of, and interactions between, molecules. However, we often struggle to translate that appreciation to technology. For example, the proliferation of personal computers and telecommunications made possible by the planar process has revolutionized interpersonal communications and led to new human behaviors and group dynamics. The mass production of steel via the Bessemer process was central to the spread of the railroad, which was a major driver of urbanization, just as the mass production of gasoline fuels via FCC enabled suburbanization.
Introducing different biases. A FMPI often establishes a new set of biases within or across technologies. Biases are advantageous in certain situations but can be detrimental in others. For example, the high cost of single-crystal, nearly impurity-free Si wafers at the heart of planar processing biases integrated circuits toward the smallest possible devices at the highest possible density. This bias is synergistic with, and has enabled today’s remarkable high-speed computer systems. At the same time, the per-step perfection demanded of planar processing inadvertently biases us against many alternatives. For instance, it has created a ‘lock-in’ effect for Si. Competing materials, which may offer distinct functionality or better performance, are not afforded the time (i.e., decades) needed to perfect their processing.
Materials discovery is another area where the biases introduced by a process innovation can leave their mark. The search for materials with improved performance often occurs in conjunction with an existing process paradigm or a limited number of possible alternatives. Process innovations, because they can change the constraints of the problem, frequently push the search toward new sets of materials. For instance, 3-D printing of polymers with hierarchical porosity is creating new opportunities for catalytic or adsorbent materials that would otherwise be off-limits due to heat management challenges [47].
Catalyzing new science and engineering. Perhaps the most underappreciated impact of process innovation is that it motivates scientific and engineering pursuits that were not obvious prior to the innovation. This “to the lab” behavior, which has been discussed in other contexts [48–50], is quite distinct from the common academic view that technological innovation begins with scientific discovery and continues with a one-way translation “from the lab.” We identify three categories of new science and engineering: a) integrative, b) advancing, and c) emergent:
- Integrative science and engineering are required to combine existing disciplines to initially demonstrate the process innovation. For example, the advent of FCC forced particle fluidization (which was not previously thought possible) to be understood and advanced. Demonstration of the planar process necessitated surface passivation, junction diffusion, and photolithography to construct the embedded transistors and electrically isolate neighbors.
- Advancing science and engineering translates an initial process innovation to maturity. For instance, the need to improve the efficiency of steam power generation drove the development of thermodynamics. The planar process made clear the need to investigate, for example, light-matter interactions (for photolithography), atomic diffusion (for doping), plasmas (for reactive ion etching), and surface reactivity (for chemical vapor deposition). 3-D printing is motivating the study of new structures and materials combinations, new polymer chemistry and physics, and the rheology of complex inks flowing in confined spaces.
- Emergent science and engineering can take place once a process innovation reaches maturity. The rapid, low cost sequencing of DNA enables identification of different cancer cell genotypes which has transformed cancer science. The same capability is also propelling the exploration of DNA as a data storage material [51]. Directed evolution is another example. This method, for which Frances Arnold shared the 2018 Nobel Prize in Chemistry [52], leverages errors during PCR to produce proteins with performance superior to, or fundamentally distinct from, their naturally occurring counterparts [53]. Like the technological benefits that compound following a process innovation, science is impacted well beyond the initial innovation.
As can be seen from the above examples, much of the science required is basic in nature. Moreover, since FMPI changes society, it has profound impacts on the social sciences as well.
Open questions
The present discussion is not definitive and should be viewed as the starting point for further elaboration. In this spirit, we present several questions that naturally emerge from our analysis and which we seek to understand going forward.
How frequent are FMPIs and what is their range of impact? If FMPI is common, can we develop methods to more easily recognize it? If FMPI is uncommon, does that stem from the fact that it is inherently difficult or that we simply do not focus on it? Even if FMPI turns out to be rare, its potential for far reaching impact means it is still worthy of attention. To this end, we also seek to assess the range of possible impacts, from the inconsequential to those leading to seismic shifts in technology and society, and understand the factors driving either end of this spectrum. Here, to make our point, we intentionally catalog process innovations with significant impact. However, one must be careful not to succumb to confirmation or survivorship bias. We do not know the number of unsuccessful process innovations that faded away and were forgotten.
What is the breadth of disciplines and technical domains in which FMPI operates and how is it distributed among them? We have already cataloged manufacturing innovations that span a variety of disciplines and technical domains, and one does not have to look far to find others. Consider George Eastman, who is famous for democratizing photography through his invention of the Kodak camera [54]. Prior to Eastman, photography was highly specialized, requiring expertise in optics and composition but also in chemically developing the photographic plates on which images were acquired. This challenge limited the number of photographers. Eastman factored plate production, image acquisition, and photograph development. It is worth noting that we tend to focus on the Kodak camera, rather than the photograph manufacturing process it enabled. We have also mentioned several examples of process innovation outside of manufacturing — in finance, transportation, and information technology. We suspect that even in service-type domains, where the process is more out-front, there is still room to improve our understanding of the mechanisms and, in doing so, accelerate these innovations.
Are there other schemas by which FMPI occurs? Do impactful FMPIs emerge more frequently from certain schema? In all cases, how useful are the schemas for a priori driving FMPI? We have outlined an initial set of schemas but expect that others exist and still others may emerge. Moreover, our definition of FMPI and the initial set of schemas focus on the way process steps are arranged and interconnected. However, there may be broader notions of process innovation that apply at a hierarchy of spatial and temporal scales. For instance, what factors influence where and how fast an FMPI propagates from its initial realization to widespread practice? In terms of the prevalence of certain schema, four of our eight manufacturing innovations utilize factoring. Is this a general trend or simply a result of our limited sample size? In addition, the schemas we have defined are general, and do not reveal how or what is required to enable them in a given situation. Are the schema simply a good way to understand process innovation after the fact or can they be a framework for driving process innovation? If the latter, what additional information is needed?
How does FMPI interact with the science and engineering that enable it? There is often an intimate connection between process innovation and one or more enabling scientific discoveries or inventions: cracking catalysts in FCC; dopant diffusion and photolithography in the planar process; polymerases in PCR; photocurable polymers in 3-D printing. Our contention is that the process innovation is as central to technological advances as the discovery or invention. However, the elevation of process innovation creates a “chicken or egg” problem. Does process innovation or science and engineering occur first? Can they occur simultaneously? Is the order case specific? Does the order of innovation matter? We fully appreciate, on the one hand, that many times discovery or invention comes first, and the process innovation only becomes clear afterward. A focus on possible process innovations, on the other hand, will allow us to efficiently channel our resources by clarifying the need for particular enabling capabilities, and the associated science and engineering.
Can we predict FMPIs, from where they will emerge, or their ultimate importance? We have been careful to look at the mechanisms by which process innovation has occurred, which are more fixed and visible than associated economic, social, political, or environmental forces. The COVID-19 pandemic highlights the challenge of predicting when and how process innovations will emerge. Brick and mortar retailers switched from a process where a small number of employees collected items for each customer to one where a large number of customers fetched items themselves. The fear of COVID-19 has accelerated the rise of companies such as Instacart, who offer a service qualitatively similar to the former “shopkeeper” process. While predicting future process innovations with any specificity is unlikely, we are interested in establishing both cultural and physical environments to increase the probability of their occurrence and potential for impact.
What are the next steps? Answering these complex and multifaceted questions will require contributions from a diversity of practitioners. This reflects the state of process innovation today, a diffuse body of knowledge that is distributed across a range of disciplines and technical domains. Given the breadth of specializations in which process innovation operates coupled with its potential for transformational change, we raise the possibility that “Process Studies” should be established as its own discipline. An independent discipline would go well beyond the process engineering that exists today, embracing a diversity of domains, including business, software, service, and manufacturing, to define a cohesive cannon that can be leveraged to systematically shape future process systems. We look forward to engaging on these subjects with innovators across industry, academia, and government.
What will be the next big FMPI?
What will be the big FMPIs of the next 250 years? The cost per transistor has decreased precipitously since the advent of the integrated circuit. However, the rate of decrease has slowed, and by some measures even ceased, in recent years. Will an alternative process for integrated circuit manufacturing replace the planar process? If so, what new use-cases for electronic systems might it enable? The need to mitigate global climate change is driving demand for renewable energy generation and storage technologies. Wind and solar energy, and their combination with Li ion batteries, are now making zero-carbon all-electric vehicles a reality. Will FCC be made obsolete by these technologies? To mitigate COVID-19 and future pandemics, how can process innovation be used to accelerate the development and manufacturing of diagnostic tests and vaccines by orders of magnitude? What industries will be created by future FMPIs and how will they drive technological and societal change? What new science and engineering will be stimulated? How can we accelerate FMPI, and will we invest in the mechanisms required to do so? And, in the products of the future, will people recognize the process innovations that took place to get there?
Acknowledgements
We are grateful to the myriad of colleagues, friends, and family whose insightful feedback strengthened our ideas and improved this work: David McDowell, David Sholl, Christopher Jones, Ryan Lively, Seth Marder, Baratunde Cola, Jeff Siirola, David Rampulla, Mark Styczynski, Naresh Thadhani, Charles Bennett, Jacqueline Mohalley-Snedeker, Phillip Szuromi, Jake Yeston, Amar Mohabir, Dave Anderson, Sarah MacLeod, and Marshall Filler.
About the authors
Michael A. Filler is an Associate Professor and the Traylor Faculty Fellow in the School of Chemical & Biomolecular Engineering at the Georgia Institute of Technology. His research lies at the intersection of chemical engineering and materials science, focusing on large-scale nanomanufacturing processes for applications in electronics, photonics, and energy conversion.
Email: michael.filler@chbe.gatech.edu
Matthew J. Realff is a Professor and the David Wang Senior Faculty Fellow in the School of Chemical & Biomolecular Engineering at the Georgia Institute of Technology. He is currently the editor-in-chief of the Journal of Advanced Manufacturing and Processing. From 2005–2007, Dr. Realff was a program officer at the National Science Foundation, where he helped develop programs in resilient and sustainable infrastructure and environmentally benign processing.
Email: matthew.realff@chbe.gatech.edu
References
- OECD, Life Expectancy at Birth (Indicator). (accessed July 28 2020).
2. Ravallion, M., The Economics of Poverty: History, Measurement, and Policy. Oxford University Press: 2016.
3. Roser, M. War and Peace. https://ourworldindata.org/war-and-peace.
4. Cowen, T., The Great Stagnation: How America Ate All the Low-hanging Fruit of Modern History, Got Sick, and Will (Eventually) Feel Better. Dutton: 2011.
5. Gordon, R. J., The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War. Princeton University Press: 2016.
6. Bhattacharya, J.; Packalen, M., Stagnation and Scientific Incentives. National Bureau of Economic Research Working Paper Series 2020, № 26752.
7. The Internet is now driving another process innovation in shopping. Web browsers are replacing the physical traverse of a shopping cart in aisles of prearranged goods. This change radically alters the logistics of product storage and delivery and enables dynamic placement of items expected to be desirable ‘near’ the consumer.
8. Bessen, J. Toil and Technology Finance & Development [Online], 2015. https://www.imf.org/external/pubs/ft/fandd/2015/03/bessen.htm (accessed October 10, 2019).
9. Levinson, M., The Box: How the Shipping Container Made the World Smaller and the World Economy Bigger Princeton University Press: 2008.
10. Domingos, P., The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World. Basic Books: 2015.
11. Jackson, M. D.; Mulcahy, S. R.; Chen, H.; Li, Y.; Li, Q.; Cappelletti, P.; Wenk, H.-R., Phillipsite and Al-tobermorite Mineral Cements Produced through Low-temperature Water-rock Reactions in Roman Marine Concrete. American Mineralogist 2017, 102 (7), 1435–1450.
12. Williams, D., Mass-Produced Pre-Han Chinese Bronze Crossbow Triggers: Unparalleled Manufacturing Technology in the Ancient World. Arms & Armour 2008, 5 (2), 142–153.
13. Lane, F. C., Venetian Ships and Shipbuilders of the Renaissance. Johns Hopkins University Press: 1992.
14. Chartrand, R.; Hutchins, R., Napoleon’s Guns 1792–1815 (1): Field Artillery Osprey Publishing: 2003.
15. Gordon, R. B., Simeon North, John Hall, and Mechanized Manufacturing. Technology and Culture 1989,30 (1), 179.
16. Hounshell, D., From the American System to Mass Production, 1800–1932: The Development of Manufacturing Technology in the United States. Johns Hopkins University Press: 1985.
17. Ronan, C. A., The Shorter Science and Civilisation in China. Cambridge University Press: 1980; Vol. 1.
18. Ronan, C. A., The Shorter Science and Civilisation in China. Cambridge University Press: 1985; Vol. 2.
19. Martin, D., Johannes Gutenberg: the Man and His Invention. Routledge: 1996.
20. Dickinson, H. W., James Watt: Craftsman and Engineer. Cambridge University Press: 1935.
21. Morris, C. R., The Dawn of Innovation: The First American Industrial Revolution. PublicAffairs: 2012.
22. Hunter, D., Papermaking: The History and Technique of an Ancient Craft. Dover: 1978.
23. Hills, R. L., Papermaking in Britain 1488–1988: A Short History. Bloomsbury Publishing: 2015.
24. Schauwinhold, D.; Toncourt, M., Steel, 1. Introduction and History. In Ullmann’s Encyclopedia of Industrial Chemistry, 2011.
25. Jeans, W. T., The Creators of the Age of Steel. Charles Scribner’s Sons: 1884.
26. Sun Company, The Houdry Process for Catalytic Conversion of Crude Petroleum to High-Octane Gasoline. American Chemical Society: 1996.
27. CompTIA Properties. IT Industry Outlook; 2018.
28. Lécuyer, C.; Brock, D. C., Makers of the Microchip: A Documentary History of Fairchild Semiconductor. MIT Press: 2010.
29. Messing, J.; Crea, R.; Seeburg, P. H., A System for Shotgun DNA Sequencing. Nucleic Acids Research 1981, 9 (2), 309–321.
30. Dunham, I., Genome Sequencing. In eLS, 2005.
31. The Nobel Prize: Kary B. Mullis Facts. Available at: https://www.nobelprize.org/prizes/chemistry/1993/mullis/facts/ (accessed June 5, 2019).
32. Bártolo, P. J.; Gibson, I., Stereolithography: Materials, Processes and Applications. Springer Science & Business Media: 2011; pp 37–55.
33. OECD, The Next Production Revolution: Implications for Governments and Business; Paris, 2017.
34. Heilbroner, R.; Singer, A., The Economic Transformation of America: 1600 to the Present. 4th ed.; Cengage Learning: 1998.
35. United States Bureau of the Census, 12th Census of the United States, Taken in the Year 1900, Manufactures. United States: 1900.
36. Fletcher, R. P., The History of Fluidized Catalytic Cracking: A History of Innovation: 1942–2008. In Innovations in Industrial and Engineering Chemistry, American Chemical Society: 2008; Vol. 1000, pp 189–249.
37. Kurzweil, R., The Singularity Is Near: When Humans Transcend Biology. Penguin: 2005.
38. Boahen, K., A Neuromorph’s Prospectus. Computing in Science & Engineering 2017, 19 (2), 14–28.
39. Wetterstrand, K. A. DNA Sequencing Costs: Data from the NHGRI Genome Sequencing Program (GSP). Available at: www.genome.gov/sequencingcostsdata (accessed August 21, 2019).
40. Saiki, R. K.; Gelfand, D. H.; Stoffel, S.; Scharf, S. J.; Higuchi, R.; Horn, G. T.; Mullis, K. B.; Erlich, H. A., Primer-directed Enzymatic Amplification of DNA with a Thermostable DNA Polymerase. Science 1988, 239 (4839), 487.
41. International Technology Roadmap for Semiconductors; 2015.
42. International Technology Roadmap for Devices and Systems; IEEE: 2017.
43. Rogers, E. M., Diffusion of Innovations. 5th ed.; Free Press: 2013.
44. Christensen, C. M., The Innovator’s Dilemma. Harvard Business Review Press: 1997.
45. After Moore’s Law. The Economist 2016.
46. Anderson, P. W., More Is Different. Science 1972, 177 (4047), 393.
47. Zhang, F.; Ma, Y.; Liao, J.; Breedveld, V.; Lively, R. P., Solution-Based 3D Printing of Polymers of Intrinsic Microporosity. Macromolecular Rapid Communications 2018, 39 (13), 1800274.
48. Kealey, T., The Economic Laws of Scientific Research. Macmillan: 1996.
49. Taleb, N. N., Antifragile: Things That Gain from Disorder. Random House: 2012.
50. Narayanamurti, V.; Odumosu, T., Cycles of Invention and Discovery: Rethinking the Endless Frontier. Harvard University Press: 2016.
51. Organick, L.; Ang, S. D.; Chen, Y.-J.; Lopez, R.; Yekhanin, S.; Makarychev, K.; Racz, M. Z.; Kamath, G.; Gopalan, P.; Nguyen, B.; Takahashi, C. N.; Newman, S.; Parker, H.-Y.; Rashtchian, C.; Stewart, K.; Gupta, G.; Carlson, R.; Mulligan, J.; Carmean, D.; Seelig, G.; Ceze, L.; Strauss, K., Random Access in Large-scale DNA Data Storage. Nature Biotechnology 2018, 36, 242.
52. The Nobel Prize: Frances H. Arnold Facts. Available at: https://www.nobelprize.org/prizes/chemistry/2018/arnold/facts/ (accessed June 5, 2019).
53. Chen, K.; Arnold, F. H., Tuning the Activity of an Enzyme for Unusual Environments: Sequential Random Mutagenesis of Subtilisin E for Catalysis in Dimethylformamide. Proceedings of the National Academy of Sciences 1993, 90 (12), 5618.
54. Jenkins, R. V., Technology and the Market: George Eastman and the Origins of Mass Amateur Photography. Technology and Culture 1975, 16 (1), 1–19.
