1. Managing Pre-Technological Knowledge: A Multi-Dimensional Approach Charles Weber INFORMS Meeting, Pittsburgh, PA November 6, 2006

2. Bohn’s Eight Stages of Knowledge Quality of knowledge improves as process matures. From R. Bohn, "Measuring and Managing Technological Knowledge,“ Sloan Management Review, Fall 1994, p. 63.

3. Learning in High Tech Manufacturing (Bohn, 1994, p. 64) “High tech manufacturing requires rapid learning about multiple variables in new products and processes. High tech processes are those in which many of the important variables are at stage 4 or below. –This makes the process difficult to control and to work with, –so a lot of effort goes into raising the knowledge level as quickly as possible. Because of customer and competitive pressures, –no sooner is knowledge raised for one product than higher performance products are demanded, –which brings in new low stage variables. Thus managing in high tech industries requires both –rapid learning and –the ability to manufacture with ‘immature’ (low stage of knowledge) processes.”

4. Research Question: How does one manage low-stage (“pre-technological”) knowledge?

5. VLSI Semiconductor Manufacturing Bohn (1994) specifically cites VLSI semiconductor design and fabrication. –Hundreds of non-linear, potentially interdependent variables –More variables are added as new products and processes are introduced. –New variables start at low stages of knowledge. –This requires many changes in product and process design. –Existing variables “regress” by a knowledge stage or two.

6. Integrated (3-D) Framework for Concurrent Process Development (Weber, Moslehi, Dutta,1995) Discrete learning sectors {Q,PI,S} Learning experience in each sector is unique. Equipment Pass Station Pass Work Cell Pass Process Layer Process Module VLSI Fab Cycle Packaged Part 1 Month 1 Week 1 Day 1 Shift 1 Hour Process Integration Axis Intrinsic Data Cycle Time Scaling Axis 0.7 um 1.0 um 0.5 um 0.35 um 0.25 um 0.175 um G5 G4 G3 G2 G1 SystematicMixedRandom Process Research Pilot Development Commercial Startup Volume Production 1.5σ3.0σ4.5σ 101.00.1 Control FD (cm -2 ) Dominant Defect Type Development Phase (Pisano, 1994) Quality Axis Quality Milestones VPRe CuSa ESa WSa PrRe

7. Concurrent IC Process Development (Weber, Moslehi, Dutta, 1995; Weber and Utterback, 1996) Learning by conquering volume The output of each sector feeds into adjacent sectors. Development Phase Quality Axis PRPDCSVP Scaling Axis G1 G2 G3 G4 G5 Process Generation Process Integration Axis Equipment Pass Station Pass Work Cell Pass Process Layer Unit Process VLSI Fab Packaged Part Level of Integration Isochrones

8. Specific Research Questions How mature is knowledge in each learning sector? How does the quality of knowledge evolve as a function of process maturity?

9. Preliminary Empirical Investigation Data come from (my dissertation) –69 cases of learning and problem solving –in semiconductor manufacturing and process development. –that transpired at 35 semiconductor facilities in Asia, Europe and North America. –Each case is clearly associated with a particular learning sector. Formal analysis has not yet taken place. This is a work in progress.

10. Stages of Process Development (after Pisano, 1994, pp. 90-91) Process Research (PR) – “involves defining the basic structure of the process. … – The goal of process research is to define the basic process architecture rather than the details.” Pilot Development (PD) –Scale up the process to some intermediate scale –Select reaction parameters (e.g. timing, temperature, pressure), –which optimize the efficiency of the process –Much more empirical in nature than process research –Relies on the analysis of the output of pilot production runs, –which are subjected to conditions that reflect actual production environment more accurately. Commercial startup (CS) involves ramping up the VLSI circuit manufacturing process to commercial scale. Volume Production (VP) at commercial scale

11. Levels of Integration (Weber, 1996) Full VLSI process (FP) yields VLSI circuits. Unit process (UP) –Consists of multiple process steps –yields electrically testable structures –3 to 10 unit processes in a full process. Single process steps (PS) –50 to 500 process steps in a full process

12. ‘Baselining’ (Weber, 1996, 2003) Run current generation in same manufacturing line as previous generation Current generation shares ~75% of all process steps with previous generation. Advantages –Most key variables have high level of knowledge early on in process development. –Problem solving: solution space shrinks If new VLSI process has low yield, but the old one has high yield, Then the problem is associated with new process steps (~25%). Share whole unit processes with previous generation, if you can.

13. Knowledge Sectors and Stages of Knowledge Current Generation FP1-33-54-64-6 UP2-54-64-74-7 PS4-74-75-75-7 PhasePRPDCSVP Previous Generation FP1-33-54-64-6 UP2-54-64-74-7 PS4-74-75-75-7 PhasePRPDCSVP time Scaling Quality Integration Observations Level of knowledge decreases with level of integration. Level of knowledge increases with process quality more rapidly at high levels of integration than at low levels. Baselining: taking advantage of higher levels of knowledge of previous generation. All more so in PR and PD than in CS and VP

14. Propositions ‘Knowledge bricks’ are intact. Knowledge architecture is not. How can we show this?

15. Organizational Differentiation and Integration (Lawrence & Lorsch, 1967-1969) Differentiation – is defined as “the state of segmentation of the organizational system into subsystems (e.g. sales, research and production), –each of which tends to develop particular attributes in relation to the requirements posed by its relevant external environment.” Integration is defined as “the process of achieving unity of effort among the various subsystems in the accomplishment of the organization’s task.” Completing a task requires a significant amount of knowledge that is differentiated with respect to relevant external environment of the various subsystems, and the differentiated knowledge of the various subsystems must be integrated to achieve unity of effort for the organization.

16. Organizational Differentiation and Integration (Lawrence & Lorsch, 1967-1969) High performing organizations are required to be both highly differentiated and well integrated. These two goals are inherently at cross- purposes, unless individuals, teams or departments act as intermediaries – or integrating devices – between the various subsystems. The presence of integrating devices would thus suggest low levels of integration knowledge. Their absence would suggest –high levels of integration knowledge or –that little integration knowledge is needed.

17. Data from my Dissertation (Weber, 2003) Availability of Integrating Devices in Process Engineering Subsystems Number of Work Environ- mentsSituation at Work Environment 29Wafer factory runs full, state-of-the-art process. Integrating device is available. 3Wafer factory does not run full process. No integrating device. 2Wafer factory does not run state-of- the-art process. No integrating device.

18. Implications Running full, state-of-the-art process requires integration knowledge, which tends to be at pre-technological levels of maturity (Stage 2 – 3) Is this inherently so? Subject warrants further investigation. I’ll give you an update in two years.