Extension of Murray's law using a nonNewtonian model of blood flow
 Rémi Revellin†^{1}Email author,
 François Rousset†^{1},
 David Baud^{2} and
 Jocelyn Bonjour^{1}
https://doi.org/10.1186/1742468267
© Revellin et al; licensee BioMed Central Ltd. 2009
Received: 09 April 2009
Accepted: 15 May 2009
Published: 15 May 2009
Abstract
Background
So far, none of the existing methods on Murray's law deal with the nonNewtonian behavior of blood flow although the nonNewtonian approach for blood flow modelling looks more accurate.
Modeling
In the present paper, Murray's law which is applicable to an arterial bifurcation, is generalized to a nonNewtonian blood flow model (powerlaw model). When the vessel size reaches the capillary limitation, blood can be modeled using a nonNewtonian constitutive equation. It is assumed two different constraints in addition to the pumping power: the volume constraint or the surface constraint (related to the internal surface of the vessel). For a seek of generality, the relationships are given for an arbitrary number of daughter vessels. It is shown that for a cost function including the volume constraint, classical Murray's law remains valid (i.e. ΣR^{ c }= cste with c = 3 is verified and is independent of n, the dimensionless index in the viscosity equation; R being the radius of the vessel). On the contrary, for a cost function including the surface constraint, different values of c may be calculated depending on the value of n.
Results
We find that c varies for blood from 2.42 to 3 depending on the constraint and the fluid properties. For the Newtonian model, the surface constraint leads to c = 2.5. The cost function (based on the surface constraint) can be related to entropy generation, by dividing it by the temperature.
Conclusion
It is demonstrated that the entropy generated in all the daughter vessels is greater than the entropy generated in the parent vessel. Furthermore, it is shown that the difference of entropy generation between the parent and daughter vessels is smaller for a nonNewtonian fluid than for a Newtonian fluid.
Introduction
Since several decades, many studies have been carried out on the optimal branching pattern of a vascular system. Based on the simple assumption of a steady Poiseuille blood flow, the well known Murray's law [1] has been established. It links the radius of a parent vessel R_{0} (immediately upstream from a vessel bifurcation) to the radii of the daughter vessels R_{1} and R_{2} (immediately downstream after a vessel bifurcation) as R_{0}/R_{1} = R_{0}/R_{2} = 2^{1/3}. From Murray's analysis, the required condition of minimum power occurs when Q ∝ R^{3} where Q denotes the volumetric flow. This relation, called "cube law", is determined assuming that two energy terms contribute to the cost of maintaining blood flow in any section of any vessel: (i) the pumping power and (ii) the energy metabolically required to maintain the volume of blood which is referred to as "volume constraint". A generalization of this relation can be proposed as Q ∝ R^{ c }where c is determined from the condition of minimum power by assuming other constraints (for instance surface constraint yields Q ∝ R^{2.5} [2]). Under the condition c = 3, the shear stress on the vessel walls is uniform and independent of vessel diameter [3]. Several studies have been carried out to determine the value of c [4–8] which usually ranges between 2 and 3. The influence of the value of c from 2 to 4 has also been investigated [9]. The in vivo wall shear stress in an arterial system has been measured [10]. It was found that mean wall shear stress was far from constant along the arterial tree, which implied that Murray's cube law on flow diameter relations could not be applied to the whole arterial system. According to the authors, c likely varies along the arterial system, probably from 2 in large arteries near the heart to 3 in arterioles. A method allowing for estimation of wall shear rate in arteries using the flow waveforms has been developed [11]. This work allowed to determine the timedependent wall shear rates occurring in fully developed pulsatile flow using Womersley's theory. They found a nonuniform distribution of wall shear rates throughout the arterial system.
Following the cubic law, Murray [12] proposed the optimal branching angle. Optimally, the larger branch makes a smaller branching angle than the smaller branch. This work was extended to nonsymmetrical bifurcations [13]. The arterial bifurcations in the cardiovascular system of a rat have been investigated [14]. The results were found to be consistent with those previously reported in humans and monkeys. Murray's optimization problem has also been reproduced computationally using a three dimensional vessel geometry and a timedependent solution of the NavierStokes equations [15].
From Murray's law, some relationships have been proposed between the vessel radius and the volumetric flow, the average linear velocity flow, the velocity profile, the vesselwall shear stress, the Reynolds number and the pressure gradient [9]. In the same way, based on the Poiseuille assumptions, scaling relationships have been described between vascular length and volume of coronary arterial tree, diameter and length of coronary vessel branches and lumen diameter and blood flow rate in each vessel branch [16, 17].
It is also possible to determine Murray's law using other approaches. A model have been suggested based on a "delivering" artery system of an organ characterized, (i) by the spacefilling fractal embedding into the tissue and (ii) by the uniform distribution of the blood pressure drop over the artery system [18]. The minimalist principles were not used but the result remains the same. Murray's energy cost minimization have been extended to the pulsatile arterial system, by analysing a model of pulsatile flow in an elastic tube [19]. It is found that for medium and small arteries with pulsatile flow, Murray's energy minimization leads to Murray's Law.
Surprisingly, so far, none of the existing methods on Murray's law deal with the nonNewtonian behavior of blood flow although, the nonNewtonian approach for blood flow modeling looks more accurate. Blood is a multicomponent mixture with complex rheological characteristics. Experimental investigations showed that blood exhibits nonNewtonian properties such as shearthinning, viscoelasticity, thixotropy and yield stress [20–22]. Blood rheology has been shown to be related to its microscopic structures (e.g. aggregation, deformation and alignment of red blood cells). The nonNewtonian steady flow in a carotid bifurcation model have been investigated [23, 24]. The authors showed that in that case, viscoelastic properties may be ignored. The fact that blood exhibits a viscosity that decreases with increasing rate of deformation (shearthining or so called pseudoplastic behavior) is thus the predominant nonNewtonian effect. There are several inelastic models in the literature to account for the nonNewtonian behavior of blood [25, 26]. The most popular models are the powerlaw [27, 28], the Casson [29] and the Carreau [30] fluids. The powerlaw model is the most frequently used as it provides analytical results for many flow situations. On the usual loglog coordinates, this model results in a linear relation between the viscosity and the shear rate. Blood viscosity have been measured by using a fallingball viscometer and a coneplate viscometer for shear rate from 0.1 to 400 s^{1} [31]. For both techniques, the authors found that measured values are aligned on a straight line suggesting that the powerlaw model fits experimental data with sufficient precision.
From the literature review, it can be established that none of the existing studies deal with the minimalist principle along with nonNewtonian models. The combination of both aspects will be studied and are presented hereafter. For a seek of generality, the relationships will be given for an arbitrary number of daughter vessels.
NonNewtonian model of blood flow
where we have used standard cylindrical coordinates such that the zaxis is aligned with the pipe centerline. It means that the only nonzero velocity component is the axial component which is a function of the distance to the pipe centerline only.
where is the generalized viscosity and is the effective deformation rate which is given here by .
It features two parameters: dimensionless flow index n and consistency m with units Pa.s^{n}. On loglog coordinates, this model results in a linear relation between viscosity and shear rate. The fluid is shearthinning like blood (i.e. viscosity decreases as shear rate increases) if n<1 and shearthickening (i.e. viscosity increases as shear rate increases) if n > 1. When n = 1 the Newtonian fluid is recovered and in that case parameter m represents the constant viscosity of the fluid. This model is very popular in engineering work because a wide variety of flow problems have been solved analytically for it.
which reduces in the Newtonian case to the classical HagenPoiseuille relation .
Extension of Murray's Law
where B_{ k }' = B_{k·}π·l_{ k }and A_{ k }' = A_{ k }·Ψ_{ k }.
which is found to be always positive because b>0, and so are A_{k}' and B_{k}'. As a result, the extremum is a minimum.
This expression may be useful when one wants to include mass and/or heat transfer through the vessel wall. Actually, when the vessel diameter decreases, blood catch up with a nonNewtonian fluid and the heat and mass transfer through the vessel wall becomes more and more significant.
Equation (18) is the generalization of the cube law. In case of laminar Newtonian flow (a = 1 and b = 4, thus c = 3), the classical cube law is recovered. Note that whatever the value of n in the Poiseuille case, c is equal to 3.

The area ratio (expansion parameter) β which is the ratio of the combined crosssectional area of the daughters over that of the parent vessel. Values of β greater than unity produce expansion in the total crosssectional area available to flow as it progresses from one of the tree to the next. It can be written as:
Until now, all equations are formulated with a parent vessel that divid into a finite number of daughter vessels (1,..., j). However, since a parent vessel divide into two daughter vessels in animals and humans, further equations will be formulated with j = 2.
Meaning of the results
In this section, we will examine the meaning of the results obtained in the general case. Particularly, we will focus on the variation of each parameter with the radius of the vessel.
Volumetric flow
Velocity of flow
Velocity profile
The maximum velocity, denoted by v_{max}, is attained at the center of the vessel. It can thus be obtained by setting r to 0 in Eq. (5).
As a result, the velocity profile is only function of the fluid properties and remains the same whatever the radius.
Vessel wall shear stress
In the particular case of c = 3, the vessel wall shear stress remains unchanged all along the vascular system.
If c<3, when blood flows from the parent to the daughter vessels, the vessel wall shear stress increases because the vessel radii decrease in the flow direction. On the contrary, if c>3, the vessel wall shear stress decreases because the vessel radii increase in the flow direction.
Reynolds number
Since c is often greater than two, the Reynolds number will always increases in the direction of the blood path.
Pressure gradient
Conductance and resistance
Cross sectional area
Entropy generation
For the Newtonian case, c = 2.5 and = 1.52. For the nonNewtonian case, n = 0.74 for instance and = 1.50. Whatever the fluid, as >1, the entropy generated in all the daughter vessels is greater than the entropy generated in the parent vessel. Furthermore, this result means that the difference of entropy generation between the parent and daughter vessels is smaller for a nonNewtonian fluid than for a Newtonian fluid. This behaviour can be related to the velocity profile, which is blunter for a nonNewtonian fluid, as shown by Eq. (5).
Illustrating example
Influence of n on the c parameter for the two different constraints.
Nominal value of n  n  Volume constraint  Surface constraint 

c  C  
1  1  3  2.5 
0.81  0.78  3  2.44 
0.81  3  2.45  
0.84  3  2.46  
0.74  0.72  3  2.42 
0.74  3  2.43  
0.76  3  2.43 
Influence of n on different parameters
Parameters  Newtonian n = 1  NonNewtonian n = 0.81  NonNewtonian n = 0.74  

ΣR^{3}  ΣR^{2.5}  ΣR^{3}  ΣR^{2.45}  ΣR ^{3}  ΣR ^{2.43}  
Volumetric flow  R ^{3}  R ^{2.5}  R ^{3}  R ^{2.45}  R ^{3}  R ^{2.43} 
Velocity of flow  R  R ^{0.5}  R  R ^{0.45}  R  R ^{0.43} 
Vessel wall shear stress  1  R ^{0.5}  1  R ^{0.55}  1  R ^{0.57} 
Reynolds number  R ^{2}  R ^{1.5}  R ^{2}  R ^{1.45}  R ^{2}  R ^{1.43} 
Pressure gradient  R ^{1}  R ^{1.5}  R ^{1}  R ^{1.45}  R ^{1}  R ^{1.42} 
Conductance  R  R ^{1.5}  R ^{0.43}  R ^{0.98}  R ^{0.22}  R ^{0.79} 
Resistance  R ^{1}  R ^{1.5}  R ^{0.43}  R ^{0.98}  R ^{0.22}  R ^{0.79} 
Cross sectional area  R ^{1}  R ^{0.5}  R ^{1}  R ^{0.45}  R ^{1}  R ^{0.43} 
Entropy generation  R 2  R 2  R ^{2}  R  R ^{2}  R 
1.26  1.52  1.26  1.51  1.26  1.50 
Conclusion
Blood is a multicomponent mixture with complex rheological characteristics. Experimental investigations have shown that blood exhibits nonNewtonian properties such as shearthinning, viscoelasticity, thixotropy and yield stress. Blood rheology is shown to be related to its microscopic structures (e.g. aggregation, deformation and alignment of blood cells and plattelets). Shearthinning is the predominant nonNewtonian effect in bifurcations of blood flows.
In this study, we have proposed for the first time an analytical expression of Murray's law using a nonNewtonian blood flow model (power law model), assuming two different constraints in addition to the pumping power: (i) the volume constraint and (ii) the surface constraint. Surface constraint may be useful if one wants to include heat and/or mass transfer in the cost function, specially in capillaries. For a seek of generality, the relationships have been given for an arbitrary number of daughter vessels. Note that there is an alternative formulation of the constrained optimization problem using the Lagrange multipliers, as discussed in [32]. However, using this approach, the results presented in this paper would not have been modified.
It has been showed that for a cost function including the volume constraint, classical Murray's law remains valid (i.e. ΣR^{ c }= cste with c = 3 is verified). In other words, the value of c is independent of the fluid properties. On the contrary, for a cost function including the surface constraint, different values of c may be calculated depending on the fluid properties, i.e. the value of n. The fluid is shearthinning if n<1 and shearthickening if n>1. When n = 1 the Newtonian fluid is recovered. In the present study, we have used two different blood values of n found in the literature, namely n = 0.81 and n = 0.74. In summary, it has been found that c varies from 2.42 to 3 depending on the constraint and the index n. For the particular Newtonian model, the surface constraint leads to c = 2.5.
Entropy generation has several origins: heat transfer, mass transfer, pressure drop, etc. The cost function (based on the surface constraint) can be related to entropy generation by dividing it by the temperature. It has been demonstrated that the entropy generated in all the daughter vessels is greater than the entropy generated in the parent vessel. Furthermore, it is shown that the difference of entropy generation between the parent and daughter vessels is smaller for a nonnewtonian fluid than for a Newtonian fluid. This behaviour can be related to the velocity profile, which is blunter for a nonNewtonian fluid, as shown by Eq. (5).
Based on the literature review and on our work, we propose in the following, further possible investigations:
The effect of singularities on the cost function has hardly ever been investigated. Few works exist on this aspect [33] and Tondeur et al. (Tondeur D, Fan Y, Luo L: Constructal optimization of arborescent structures with flow singularities. Chem. Eng. Sci. 2009, submitted.). However, the effect of the singularities on the entropy generation might not be negligible. This contribution should be added in the cost function in the future.
In reality, heat and mass might be transfered through the vessel wall, leading to resistance that should be included in the cost function. Indeed, gases, nutrients and metabolic waste products are exchanged between blood and the underlying tissue. Substances pass through the vessels by active or passive transfer, i.e. diffusion, filtration or osmosis. Moreover, pathologic states such as edema and inflammation might increase such phenomena.
It is accepted that pulsatile blood flow is more realistic than steadystate flow. The cost function should also include the effect of pulsatile flow in an elastic tube.
Blood is essentially a twophase fluid consisting of formed cellular elements suspended in a liquid medium, the plasma. The corpuscular nature of blood raises the question of whether it can be treated as a continuum, and the peculiar makeup of plasma makes it seem different from more common fluids. In particular, when the vessel radius decreases down to the smallest capillaries, the continuum approach diverges from the reality. Treating blood as a noncontinuum fluid should be a possible next step.
Notes
Declarations
Authors’ Affiliations
References
 Murray CD: The physiological principle of minimum work. I. The vascular system and the cost of blood volume. Proc Natl Acad Sci U S A. 1926, 12 (3): 207214.PubMed CentralView ArticlePubMedGoogle Scholar
 Tondeur D, Luo L: Design and scaling laws of ramified fluid distributors by the constructal approach. Chem Eng Sci. 2004, 59: 17991813.View ArticleGoogle Scholar
 Karau KL, Krenz GS, Dawson CA: Branching exponent heterogeneity and wall shear stress distribution in vascular trees. Am J Physiol Heart Circ Physiol. 2001, 280 (3): H1256H1263.PubMedGoogle Scholar
 Dawson CA, Krenz GS, Karau KL, Haworth ST, Hanger CC, Linehan JH: Structurefunction relationships in the pulmonary arterial tree. J Appl Physiol. 1999, 86: 569583.PubMedGoogle Scholar
 Griffith TM, Edwards DH: Basal EDRF activity helps to keep the geometrical configuration of arterial bifurcations close to the Murray optimum. J Theor Biol. 1990, 146: 545573.View ArticlePubMedGoogle Scholar
 Horsfield K, Woldenberg MJ: Diameters and crosssectional areas of branches in the human pulmonary arterial tree. Anat Rec. 1989, 223: 245251.View ArticlePubMedGoogle Scholar
 Mayrovitz HN, Roy J: Microvascular blood flow: evidence indicating a cubic dependence on arteriolar diameter. Am J Physiol. 1983, 245 (6): H1031H1038.PubMedGoogle Scholar
 Suwa N, Niwa T, Fukasuwa H, Sasaki Y: Estimation of intravascular blood pressure gradient by mathematical analysis of arterial casts. Tohoku J Exp Med. 1963, 79: 168198.View ArticlePubMedGoogle Scholar
 Sherman TF: On connecting large vessels to small: the meaning of Murray's law. J Gen Physiol. 1981, 78: 431453.View ArticlePubMedGoogle Scholar
 Reneman RS, Hoeks APG: Wall shear stress as measured in vivo: consequences for the design of the arterial system. Medical and Biological Engineering and Computing. 2008, 46: 499507.PubMed CentralView ArticlePubMedGoogle Scholar
 Stroev PV, Hoskins PR, Easson WJ: Distribution of wall shear rate throughout the arterial tree: A case study. Atherosclerosis. 2007, 191: 276280.View ArticlePubMedGoogle Scholar
 Murray CD: The physiological principle of minimum work applied to the angle of branching of arteries. J Gen Physiol. 1926, 9: 835841.PubMed CentralView ArticlePubMedGoogle Scholar
 Zamir M: Nonsymmetrical bifurcations in arterial branching. J Gen Physiol. 1978, 72: 837845.View ArticlePubMedGoogle Scholar
 Zamir M, Wrigley SM, Langille BL: Arterial Bifurcations in the Cardiovascular System of a Rat. J Gen Physiol. 1983, 81: 325335.View ArticlePubMedGoogle Scholar
 Marsden AL, Feinstein JA, Taylor CA: A computational framework for derivativefree optimization of cardiovascular geometries. Comput Methods Appl Mech Eng. 2008, 197: 18901905.View ArticleGoogle Scholar
 Kassab GS: Scaling laws of vascular trees: of form and function. Am J Physiol Heart Circ Physiol. 2006, 290 (2): H894H903.View ArticlePubMedGoogle Scholar
 Kassab GS: Design of coronary circulation: A minimum energy hypothesis. Comput Methods Appl Mech Eng. 2007, 196: 30333042.View ArticleGoogle Scholar
 Gafiychuk VV, Lubashevsky IA: On the Principles of the Vascular Network Branching. J Theor Biol. 2001, 212: 19.View ArticlePubMedGoogle Scholar
 Painter PR, Edén P, Bengtsson HU: Pulsatile blood flow, shear force, energy dissipation and Murray's Law. Theor Biol Med Model. 2006, 3: 31PubMed CentralView ArticlePubMedGoogle Scholar
 Chien S, Usami S, Dellenback RJ, Gregersen MI: Sheardependent deformation of erythrocytes in rheology of human blood. Am J Physiol. 1970, 219 (1): 136142.PubMedGoogle Scholar
 Thurston GB: Frequency and shear rate dependence of viscoelasticity of human blood. Biorheology. 1973, 10: 375381.PubMedGoogle Scholar
 Thurston GB: Rheological parameters fort he viscosity, viscoelasticity and thixotropy of blood. Biorheology. 1979, 16 (3): 149162.PubMedGoogle Scholar
 Gijsen FJH, van de Vosse FN, Janssen JD: The influence of the nonNewtonian properties of blood on the flow in large arteries: steady flow in a carotid bifurcation model. J Biomech. 1999, 32 (6): 601608.View ArticlePubMedGoogle Scholar
 Gijsen FJH, Allanic E, van de Vosse FN, Janssen JD: The influence of the nonNewtonian properties of blood on the flow in large arteries: unsteady flow in a 90 degrees curved tube. J Biomech. 1999, 32 (7): 705713.View ArticlePubMedGoogle Scholar
 Yeleswarapu KK: Evaluation of Continuum Models for Characterizing the Constitutive Behavior of Blood. PhD Thesis. 1996, University of Pittsburgh, Pittsburgh, PAGoogle Scholar
 Sequeira A, Janela J: An Overview of Some Mathematical Models of Blood Rheology. A Portrait of StateoftheArt Research at the Technical University of Lisbon. Edited by: Seabra Pereira M. 2007, Springer Netherlands, 6587.View ArticleGoogle Scholar
 Ostwald W: About the rate function of the viscosity of dispersed systems. Kolloid Z. 1925, 36: 99117.View ArticleGoogle Scholar
 De Waele A: Viscometry and plastometry. Oil Color Chem Assoc J. 1923, 6: 3388.Google Scholar
 Casson N: A flow equation for pigmentoil suspensions of the printing ink type. Rheology of Disperse Systems. Edited by: Mill CC. 1959, Pergamon Press, New York, NYGoogle Scholar
 Carreau PJ, De Kee D, Daroux M: An Analysis of the Viscous Behavior of Polymeric Solutions. Can J Chem Eng. 1979, 57: 135141.View ArticleGoogle Scholar
 Egushi Y, Karino T: Measurement of rheologic property of blood by a fallingball blood viscometer. Ann Biomed Eng. 2008, 36 (4): 545553.View ArticleGoogle Scholar
 Luo L, Tondeur D: Optimal distribution of viscous dissipation in a multiscale branched fluid distributor. Int J Thermal Sci. 2005, 44: 13311141.View ArticleGoogle Scholar
 Wechsatol W, Lorente S, Bejan A: Treeshaped flow structures with local junction losses. Int J Heat Mass Transfer. 2006, 49: 29572964.View ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.