Seems rather couter-intuitive but:
Original article here
http://iopscience.iop.org/0022-3727/43/35/354001
The full text appears to be freely available.
From the abstract:
"Abstract
Artificial light has long been a significant factor contributing to the quality and productivity of human life. As a consequence, we are willing to use huge amounts of energy to produce it. Solid-state lighting (SSL) is an emerging technology that promises performance features and efficiencies well beyond those of traditional artificial lighting, accompanied by potentially massive shifts in (a) the consumption of light, (b) the human productivity and energy use associated with that consumption and (c) the semiconductor chip area inventory and turnover required to support that consumption. In this paper, we provide estimates of the baseline magnitudes of these shifts using simple extrapolations of past behaviour into the future. For past behaviour, we use recent studies of historical and contemporary consumption patterns analysed within a simple energy-economics framework (a Cobb–Douglas production function and profit maximization). For extrapolations into the future, we use recent reviews of believed-achievable long-term performance targets for SSL. We also discuss ways in which the actual magnitudes could differ from the baseline magnitudes of these shifts. These include: changes in human societal demand for light; possible demand for features beyond lumens; and guidelines and regulations aimed at economizing on consumption of light and associated energy."