You are on page 1of 21

Journal of the Geological Society, London, Vol. 144, 1987, pp. 327-347, 21 figs, 8 tables.

Printed in Northern Ireland

The movement and entrapment of petroleum fluids in the subsurface


W . A .E N G L A N D ,A . S . MACKENZIE',D.M.MANN & T. M . QUIGLEY Geochemistry Branch, BP Research Centre, Sunbury+n-Thames, Middlesex TW16 7LN, UK
Abstract: This paper discusses the migration of petroleum from its formation in a source rock to its subsequent possible entrapment in a reservoir. The chemical and physical properties of petroleum gases and liquids are stressed, particularly their phase behaviour undersubsurface conditions which is shown to be a very important factor in determining migration behaviour. Engineering correlations are presented for estimating the propertiesof petroleum fluids under geologically realistic conditions. The are deduced from the directions and magnitudes of the forces acting on migrating petroleum combined effects of buoyancy and water flow in compacting sediments. These forces are combined, using a fluid potential description. This procedureallows the direction of migration to be defined. The rate of migration is thenestimatedfromtheproperties of thesedimentsinvolved, allowing a distinction to be made between 'lateral' and'vertical' carrier beds. This simplified approach is suitable for rapid predictive calculationsin petroleum exploration. It is compared with the more complex 3-D computer modelling approaches which are currently becoming available. Migration losses are related to the cumulative pore volume employed by the petroleum in establishing a migration pathway. The petroleum migration mechanism is shown to be predominantly by bulk flow, with a small diffusive contribution for light hydrocarbons over distances less than c. 100m. The loss factors involved in secondary migration are estimated from field evidence. Themechanism of reservoir filling is presented as a logical extensiontothose described formigration.This,together with the inefficiency of in-reservoir mixing by diffusion or convection, is shown to tend to cause significant lateral compositional gradients in reservoirs over and above the gravitationally induced vertical gradients described by other workers.

Symbols and units used


A
BG

Scaling constant in k, = A @ * ,m' formation Gas volume a factor, for phase gas reservoir: volume of gas

single-

+ dissolved condensate

under subsurface conditions volume of gas measured at STP

B0

Oil formation volume factor: volume of oil dissolved gas in subsurface volume of oil at STP

c,
cc

C" CGR d

D
F

F K
F P
Fw

Number of components number, Capillary pq/y Compaction coefficient Coefficient of consolidation Condensate:gas petroa relating ratio, to leum gas, kg kg-' m diameter, Grain Diffusion coefficient, m' s-l Number of degrees of freedom; or force on unit volume of fluid, N m-3 or Pa m-l Rate of enthalpy change caused by unit volume of kerogen breakdown, J m-3 sK1 Force acting on volume unit of petroleum fluid, Nm-3 or Pa m-' Force acting on unit volume of water, N m-3 or Pa m-'

Vertical force acting on unit volume of petroleum, N m-3 or Pa m-' S ' Acceleration due to gravity, 9.81 m Surface gas:oil ratio of petroleum fluids expelled from a source rock, kg kg-' Surface gas: oil ratio of a subsurface petroleum liquid, kg kg-' height of sedimentary column, m Carrier bed thickness, m Maximum height of petroleum column that a seal can support, m Characteristic segregation length for a given compound, m Intrinsic permeability, m* Boltzmann constant, 1.38X 10-23JK-' Molecular weight of ith component, kg (kg mol) Subsurface mass of petroleum liquid or gas, kg Surface mass of petroleum liquid (oil or condensate), kg Surface mass of petroleum gas, kg number of moles Reynolds number, pqtlp Pressure, Pa; or number of phases (for use with phase rule) Darcy flux perunit cross-sectional area of rock, m3 m-' s-l Vertical component of Darcy flux, m3m-'
S-1

'Presentaddress BP PetroleumDevelopment(Norway)plc., Forusbeen 35, P.O. Box 197, 4033 Forus, Norway.

Lateral Darcy flux, m3 s-l Vertical Darcy flux, m3S '

327

328

W. A.ENGLAND
Pore throat radius, mean pore throat radius, m Hydrodynamic radius of the ith component, m d JK-' (kg mo1)-' Gas constant, 8.314 X I Rayleigh number Petroleum saturation of pore space, m3 m-3 Time, S Temperature, K Volume, m3 m3 (kg Molar volume of ith component, moI)-' Length scale for estimating Peclet number Horizontal length of camer bed, m Horizontal dimension, m Mole fraction of ith component Composition Horizontal width of carrier bed, m Vertical dimension, m Compressibility factor to correct ideal gas law for non-ideal behaviour: PV = ZnRT Dip of beds, degrees or radians Coefficient of expansion for water, K-' Contact angle of petroleum-water interface on pore wall, measured through petroleum, degrees, or radians Interfacial tension, or energy, Nm-' or Jm-* Proportions of subsurfacepetroleum fluids, liquids orgases that are gas at the surface, kg kg-' Tortuosity factor, m m-' Dynamic viscosity, Pas or kg m-' sC1 Density, kg m-3 Subsurface petroleum gas density, kg m-3 Surface petroleum gas density, kg m-3 Subsurface petroleum liquid density, kg m-3 Surface petroleum liquid (oil or condensate) density, kg m-3 Subsurface petroleum fluid density, kg m-3 Subsurface rock density, kg m-3 Subsurface water density, kg m-3 Total stress, effective stress, Pa Porosity, average porosity, surface porosity, m3 m-3 Fluid potential, Pa or J m-3 Electrostatic potential, volts, mechanical potential, J or Petroleumpotential,waterpotential,Pa J m-3 Vector differential of vertical and lateral water potential, N m-3 or Pa m-'

ET AL.

r, f

Table 1 . Compositions of subsurface petroleum liquid from the Bruce field, UK continental shev, reservoired at 40 MPa and 105 "C; and surface gases and oils producedat STP (COR = 0.3 kg kg-')
Subsurface petroleum liquid Surface gas Surface Components
M(%)

R R,
S t

oil

mol(%) M(%) mol(%) M(%) mol(%)

T
V

vi
W

W
X Xi

X Y
z

Z
(Y

(YW

2.6 4.3 1.1 1.9 0.4 0.3 0.1 49.9 68.311.8 10.9 0.1 0.02 8.1 11.0 12.6 3.3 0.1 5.8 7.8 13.1 3.4 1.4 0.4 3.8 4.7 10.4 3.0 2.6 0.9 2.5 2.5 6.9 2.5 1.8 2.3 4.3 5.8 2.8 10.8 5.3 3.5 0.8 3.4 4.8 14.9 8.2 4.0 0.2 1.0 6.3 2.90.08 0.4 5.1 10.5 6.7 0.1 3.7 1.9 1.4 3.0 1.2 2.9 1.1 2.9 1.1 3.0 0.9 2.7 0.8 2.6 0.7 2.3 0.5 1.9 0.4 1.4 0.2 0.9 0.2 0.9 0.2 0.7 0.09 0.4 0.3 0.06 0.2 0.04 0.1 0.02 0.1 0.02 0.01 0.07 0.007 0.04 0.003 0.02 26.57 (MW. 450) 4.2 100 100 100 100
C, refers to a molecule with n carbon atoms.

0.4
0.5

1.6

0.02 7.0

4.9 4.1 3.9 3.9 4.0 3.6 3.5 3.1 2.5 1.9 1.2 1.2 0.9 0.5 0.4 0.3 0.1 0.1 0.1 0.05 0.03 16.036.2
l o o

5.3 4.5 4.3 4.1 3.4 3.1 2.5 1.4 0.9 0.9 0.6 0.3 0.3 0.1
0.06 0.06 0.06 0.06

0.03 0.01 100

Understanding the movement of petroleum fluids through the pores of sedimentary rocks is of enormous commercial importance. Much has been written onthe extraction of petroleum fluids from the pores of underground reservoirs (e.g. Dake 1978); but the understanding of how these fluids moved towards and accumulated in the reservoirs is somewhat superficial. An improved appreciation of this process will help to plan extraction programmes, and increase the precision of petroleum exploration. Petroleum is defined as both crude oil and natural gas; 'oil' and 'gas' are descriptions applied to petroleum fluids under surface conditions of pressure and temperature. A typical petroleum composition isshown in Table 1. The thousands of naturally occurring compounds have been

grouped by carbon number. Thus, the C6 fraction includes normal and branched hexanes (C,H,,) as well as the unsaturated compound benzene (C6&). The compounds with carbon numbers of five or less are mainly found in the gaseous phase under surface conditions, while the heavier hydrocarbons are mostly found in the liquid state. Petroleum is formed in the subsurface in finegrained source rocks and its generation is well understood (e.g. Tissot & Welte 1984). Some fraction of the organic remains of dead organisms deposited with the rocks may be preserved to form a solid, insoluble constituent known as kerogen (e.g. Durand 1980). Kerogen ischemically stable until c. lOO"C, at which point some of the bonds within kerogen are broken and mobile petroleum fluids are produced; if their volume within the pores is adequate to form an inter-connected phase, expulsion may occur(Cooles et al. 1985). To create accumulations from which petroleum may be extracted economically, the petroleum must migrate into the pores of coarser, more permeable 'reservoir' rocks. It is not uncommon for petroleum to migrate more than 2 km vertically and 1 0 0kmlaterally from its origin to a reservoir. There are several procedures for quantifying the

MOVEMENT OF PETROLEUM FLUIDS

329

generation andexpulsion of petroleum fromsource rocks. A mass balance calculated from the decreasing kerogen concentration and corresponding increasing petroleum fluid concentration has been described by Cooles et al. (1985). In addition, there are kinetic schemes which relate the conversion of kerogen petroleum to fluids and the temperature history of the source rock (Tissot & Espitalie 1975; A. S. Mackenzie & T. M. Quigley inpreparation). Hence the masses of petroleum fluids produced in a given source rock may be estimated, if its volumeand initial kerogen concentration are known. A petroleum expulsion efficiency must also be estimated to compute the amount of petroleum expelled from the source rock. An overall migration efficiency must be assigned to calculate the amount of petroleum whichfinally reaches a trap from a given source rock. This paper aims to explain the fate of petroleum fluids after expulsion, as they migrate towards and into reservoir rocks. To achieve this we need to: (a) develop a realistic physical model of migration, and (b) make accurate estimates of the properties of the subsurface fluids and the rocks which containthem.Case histories will be used to examine the migration of petroleum and thelosses involved. This will be related to predictions to produce a physical modeldeterminingthedirectionandrange of petroleum migration. The entrapment of commercial quantities of petroleum, and the displacement of water from thereservoir rock pores will be described.Therole of diffusive mixing within a petroleumaccumulation,andtheproduction of compositional gradients by theEarth'sgravitational field will be analysed, together with the effect of lateral pressure gradients created by formation water flow on the equilibrium configuration. The scope for calibrating rigorously the theoretical models of the flow and interaction of petroleum-water systems through porous media on a geological timescale is great, given the largedatabase collected by drilling for petroleum.Theincreasedunderstandingthat is emerging will undoubtedly throwlight on a myriad of other geological processes where multiphase flow in porous media occurs.

(water-saturated)petroleum gas. The gaseous phase will usually contain predominately CH4 with other light hydrocarbons, NZ,CO, and small quantities of H,S and H,O. The terms 'gas' and 'oil' will be reserved for describing petroleum phases under surface conditions, following the somewhat confusing practice of the petroleum industry. The next sections will discuss the generalized properties of petroleum phases and water, and how they are affected by pressure temperatureand composition (P, T and X). Conclusions will be deduced from engineering correlations, rather than thermodynamic calculations. The correlations are based on laboratoryexperiments(e.g. Standing 1952; G l a s ~1980) and are accurate enough for the purposes of this overview, and have been validated by process and plant engineers.

Composition
The relationships between P, T and X for subsurface petroleum gases and liquids are first examinedandthen used as a basis for defining subsurface fluid densities. The greatest compositional influence on subsurface liquid density is the quantity of lighter hydrocarbons (generally those which are gases at STP) dissolved in the liquid phase. This is quoted as the gas :oil ratio (GOR) expressed in kg kg-', as measured at the surface. Figure 1 shows the processes involvedindefining the GOR. First,aquantity of subsurface petroleum liquid of mass Ml is takentothe surface and P-T arereduced, usually resulting intheseparation of a gaseous phase of mass M3 and a liquid phase of mass M,.The GOR is defined as the ratio of the surface masses of gas and oil:

GOR = M J M ,
The corresponding volumes are defined inFig. l as V,, V, and V,. Using the extensive oil industry database of fluid properties, predictive correlations of parameters such as GOR, subsurface density etc. can be made. The correlations use typical surface densities of oil and gas at standard temperatureandpressure(STP), p:' = 800 kg r n p 3 and = 0.8 kg m-3. Predictions from these correlations are showninFig. 2a; GOR increases sharply with increasing

p E z

Properties of petroleum fluids


The phase behaviour of subsurface fluids is one of the main determinants of their properties. The phase rule states: P=C-F+2 where P is thenumber of phases, C is thenumber of components and F is the number of degrees of freedom, usually ~ 3 Due . to the very large number of components present, see Table 1, the phase rule does not impose any severeconstraint on the number of phases whichmayin principle coexist. Under usual subsurface conditions, up to three phases may in fact be in equilibrium: (1) A gaseous phase, referred to as 'petroleum gas'. (2) A petroleum-poor liquid phase rich in water, referred to as 'water'. (3) A petroleum-rich liquid phase, referred to as 'petroleum liquid'. Because the mutual solubility of most petroleum species and water is very low, the petroleum phases will be saturated with respect to water. One need only therefore consider two phases: (water-saturated) petroleumliquid and

Stock Tank

Stock Tank

Surface Subsurface

Surface Subsurface

Liquld

(Liquid-Containing Reservoir)

(Gas-Contaming R e w v o i r )

Definition of terms to relate subsurface petroleum liquids and gases to their surface properties. The reductionP in - T , as oil petroleum is bought to the surface, causes separation into crude and gas whose relative masses and volumes are important parameters. NB: in practice, several separators are used by engineers.

Fig. 1 .

330

W. A . E N G L A N D E T

AL.

U)

A,
200C

50 Pressure (MPa)

100

100 Temperature I ' C J

200

Fig, 2. (a) Variation in the saturation gas:oil ratio of petroleum liquid as a functionof pressure, at 50 and 200 "C, calculated using Glasa's (1980) correlations, p z y = 800 kg m-3 and p:? = 0.8 kg The trends above 30 MPa are extrapolated. (b) Curves of constant gas:oil ratio (GOR) of petroleum liquid in P-Tspace, calculated as described for Fig. 2a. The stippled is chemically unstable above 160 "C on a geological region delimits theP-T range encountered in the subsurface. (Note: oil time scale.)

CGR measurements by Price et al. (1983) for methane pressure but decreases slightly with increasing temperature. gas are shown in Fig. 3a, and demonstrate the rapid increase The relationship between the variables P , T and GOR can in CGR with temperature and pressure. Using a phase be treated as a conventional phase diagram, in which GOR diagram representation (Fig. 3b) it can beshownhow a is contoured in P-T space (Fig. 2b). For a given GOR it condensate-rich petroleum gas at depth (high P - T ) loses its shows the equilibrium pressures and temperatureswhich are heavier components with upwards migration. possible assuming both liquid and gaseous petroleum phases The relative compositions of subsurface gases and liquids are present. For liquid petroleum migrating along a typical are thus liable to continuous adjustment as movement P-T line, defined by the geothennal and pressure gradients, occurs along geological gradients of P - T . In particular this the change in GOR can be estimated. Note that the explains the field observation thatthe richest condensate calculations to give Fig. 2 are for an arbitrary stock tank oil accumulations are generally found at the greatest depths. density of 800 kg m-3 and must be recomputed in order to describe particular field situations. Subsurface petroleum gas will also have physical properties which are strongly influenced by composition, Density - since it always has a finite amount of heavier hydrocarbons The of a subsurface petroleum fluid in equilibrium (i.e. C, and above) in solution. At STP the amount is with gas or liquid will influence greatly its direction of but at pressures and temperatures migration. The subsurface density of gas-saturated oil may the amount may be large* When subsurface gas is brought to be calculated fromthe GOR and oil formation STP a liquid phase may separate and is known by petroleum factor B0 from the mass balance Ml = M2+ M3. This gives: engineersasa gas condensate.Thecondensate: gas ratio (CGR) is defined in Fig. 1:

density

CGR = MJM3

2.0

I
50 Pressure (MPa)

...
100
100

200

Temperature ("C J

Fig. 3.

(a) Variation in the saturation condensate: gas ratio (CGR) of petroleum gas at 50, 150 and 200C with pressure, as reported by Price et al. (1983) for mixturesof methane = 0.68 kg m-') and an oilof surface density 806 kg m-3. (b) Location of curves of constant condensate: gas ratio (CGR)in P-T space, as described for Fig. 3a. The stippledregion delimits the P-Trange encountered in the subsurface. Because the condensate components are thermally unstable at temperatures 6 0"C on a geological timescale, itis unlikely that the high CGRs predicted at these temperatures are ever greater than 1 achieved.

(PE:

FLUIDS PETROLEUM OF MOVEMENT

331

The 'oil formation volume factor' B. represents the reduction in volume observed when a volume V, of subsurface liquid is broughtto surface conditions via a gaslliquid separator-see Fig. 1. Thus, B. = VJV, and typical values of B. are in the range 1.1-2. A similar formula is derived for subsurface gaseous density: ( l + CGR) STP Pgas = PS= B G The 'gas formationvolumefactor', B G may be calculated from the modified ideal gas law, PIVl = ZlnRTl, where 2 , is the empirical compressibility factor. Values for Z, may be obtained from correlation tables (Standing 1952) if p i z , and P-T are known. From Fig. 1:

Figures 5a & b show contours of pgas and poi,on a P-T plot; a typical geological, P-T gradient is shown. Clearly as petroleum migrates upwards, the density of the liquid phase increases, whereas that of the gaseous phase decreases.

Water density
The subsurface density of water isprincipallyaffected by P-T and salinity. Correlations have been given by Schowalter (1979) but the influence of dissolved gases has not been reported.

Interfacial tensions and viscosities


An appreciation of how these parameters varywith P, T and X is essential when considering the mechanismand rates of petroleum migration. Berg (1975) has shown that the oil-water interfacial tension, y , remains reasonably constant with increasing P-T (within the limits experienced in sedimentary basins) at 20-40 X 10-3 Nm-l.Atdepths >2 kmgashas a similar density to that of oil, and the gas-water interfacial tension is similar to the oil-water interfacial tension; at depths <2 km the rapidly decreasing density and water solubility of gas causes y to rise to 70 X lOP3 Nm-' at the surface (Berg 1975). The interfacial tension between gas-saturated petroleum liquid and condensate-saturated petroleum gas decreases from 3 Nm-' at STP to 0.02 Nm-' at 40 MPa and 120C, the two phases approach one another in their properties. The low interfacial tension at high P-T implies that if two phases separate from a single petroleum-rich phase at high P-T the physical unmixing of the two phases may be slow: thesegregated petroleum phases will continueto migrate together, possibly as a foam (Katz et al. 1943). The viscosities of the three phases have been measured in the laboratory for different compositions and ranges are given in Table 2. Viscosity increases in the order petroleum gas << water = petroleum' liquid. Hence viscous fingering of petroleum gas into water, and water into liquid petroleum may occur. This in part explains why the stability of an

thus

B G = 3352, - (assuming Z,= 1, P3 = 0.1 MPa '1 and = 298 K)

(4)

poil and psaswith temperature and The variation in pressure is shown in Figs 4a and 4b. As before p:fp and p E are set to 800 kgm-3 and 0.8 kg m-3 respectively. The most strikingfeature of the subsurface density of petroleum liquid is its decrease with increasing pressure (Fig. 4a). The dissolution of additional low molecular weight components (expressed as an increasing G O R ) lowers the average molecular weight of the petroleum liquid, and thus also lowers its density. Values of poil may be asmuchas 30% lower than those found at STP. The subsurface gas density behaves in a completely different way-see Fig. 4b. Density increases with increasing pressure due to the normal PVT behaviour of gases, and to the increasein CGR with pressure. Values of pgas may reach about half that of liquid petroleum at pressures of c. 40 MPa.Thus in thesubsurface,petroleum gasesmay have properties approachingthose of petroleum liquids.

700 r

50C

B
600

f
500
'

500

2
a
0
(D r

400'

'

'

'

50 Pressure (MPa)

100

9
"I

:/G

50

100

Pressure (MPa)

Fig. 4. (a) Subsurface densityof petroleum liquid, when saturated with petroleum gas, as a function of pressure at 50 and 200 "C, calculated using equation (1). COR was estimated from Fig. 2a; the oil formation volume factors were calculated using Glase's (1980) correlations.Surface densities were as for Fig. 2a. The trends above 30 MPa are extrapolated. (b) Variation in the subsurface density of petroleum gas, when saturated with condensate, as a function of pressure50 atand 200 "C,calculated using equation (2). CGR was estimated from Fig. 3a; the gas formation volume factors were calculated usingGlas~i's (1980) correlations. Surface fluid densities were as for Fig.2a. The trends above 30 MPa are extrapolated.

332

W. A. ENGLAND

ET A L .

interface between two phases is related to the ratio of their viscosities. A large difference in viscositysuch as between gas and water can cause fingers of the less dense phase to move into moredense phase if the interface is moving (Saffmann & Taylor 1958).

Effects on composition and phase behaviour during movement


Typical P-T ranges for sedimentary basins are shownin Figs 2-5: they illustrate thegeneraltrends in behaviour expected as gas or liquid-saturated petroleum fluidsmove upwards. These trends are summarized as follows: ( 1 ) Petroleum liquids lose low molecular weight material to a gaseous phase. ( 2 ) Petroleum gases lose high molecular weight material to a liquid phase. (3) Petroleum liquids increase their density. (4) Petroleum gases reduce their density. (5) Petroleum gas and liquid properties become less similar. It is currently not possible to model the more complex three-phase (petroleum liquid-petroleum gas-water) behaviour of real dynamic systems, that result from continuous variations in P, T and composition. A further complication is the character of the surrounding matrix of porous rock as this may cause changes in behaviour. The discussion above has only concerned systems in which gas and liquid phases are always in equilibrium. Situations are possible in which petroleum gases and liquids become physically separated, which is the geological equivalent of a one-stage gaslliquid separator.
Table 2. Approximate subsurface vbcosities offluids (taken from Frick & Taylor 1962)
Viscosity (Pa Gas
S)

However there is much that may be learnt from an examination of equilibrium phase behaviour. The composition, X , has only been described in terms of three parameters, the surface density ofoil or condensate, the surface density of gas, the GOR or CGR. In reality as shown in Table 1 petroleum has a very large number of significant components, their concentrations can have a great effect on the phase behaviour of particular petroleum systems, which may deviate widely from the typical correlations used above. Wehave discussed elsewhere (A. S . Mackenzie & T. M. Quigley in preparation), the techniques for estimating the masses of petroleum expelled from a given volume of source rock ( M E ) and the ratio of surface gas and oil of the expelled petroleum CF.Defining M3 and M2 by analogy with Fig. 1 , GF= M3/M,. Surface mass balance requires that ME = Mz + M3. For a given P-T and GF it is possible to compute the quantities andnature of the phases present from ME and the subsurface GOR and CGR. Three possible conditions exist, depending onthe relative magnitudes of GF, GOR and IICGR. I f GF is less than the subsurface saturation GOR, calculated from the correlations discussed above, then the petroleum liquid is undersaturated with respect to gas, and no separate petroleum gas phase will be present. I f GFis greater than 1ICGR (i.e. the gas:oil ratio for the petroleum gas phase), then the petroleum gas phase will be undersaturated with respect to oil, and no separate petroleum liquid phase will be present. If, however, GF is greater than the COR and less than lICGR, two phases will be present in equilibrium. They will both be fully saturated with respect to the other phase. The relative masses of the petroleum liquids and gasmaybe computed from mass balance considerations.

Forces controlling petroleum movement


Migrationis the process bywhich petroleum fluidsmove from the low porosity, fine-grained source rocks, where they aregenerated,to higher porosity reservoir rocks, where they may (under suitable circumstances) form a highly concentrated hydrocarbon accumulation. Primary migrationisdefinedas the movement of the newly generated petroleum from the low permeability source rock to its first encounter with higher permeability

Oil
Water

10-~-10-~ 5 X 10-4-5 X 10- 10-~-10-~

Gas viscosity increases with increasing depth; Oil viscosity decreases with increasing depth; Water viscosity increases with increasing depth.

FLUIDS PETROLEUM OF MOVEMENT

333

beds-usually a sandstone or fracturedlimestone body. The distance involved is typically in the range up to 1km. Secondary migration is the subsequent transfer of petroleum through higher permeability strata known as carrier beds. If a suitable reservoir structure is encountered within the range of secondary migration, a petroleum accumulation may be formed. The distance involved in secondary migration is usually up to 100 km, but depends on the volumes and types of petroleumand rocks involved; these factors will be discussed below. Figure 6 illustrates the definition of primary and secondary migration. Early attempts to explain the mechanism of migration were based on the dissolution of petroleum in pore water andlor on diffusion through water-wet rocks. However, attempts to quantify these mechanisms have shown that the solubilities and diffusion constants are far too low to account for the masses of petroleum transported, or the timescales available (Jones 1980; Leythaeuser et al. 1982). This paper demonstrates that models based on the bulk flow of petroleum can quantitatively account for the migration distances and timescales observed in nature (e.g. Durand 1981). The discussion of migration will be split into twomain sections. First the origin and magnitudes of the driving forces which controlpetroleum migration will be discussed. Then these will be used, in conjunction with rock and fluid properties to estimate the rates at which primary and secondary petroleum migration occur, and the distances covered.

The force vector F points along the steepest gradient of the contours of @-the equipotential contours. In electrostatics the potential, QE, is defined as the work done on transferring unit charge from a datum point (usually infinity) to the coordinates of interest.The force vector on unit charge is then given by F = -VQE The negative sign shows that the forces point from areas of high potential to those of low potential. Positions of stable equilibrium in any potential field are defined by: and

VQ=O V2Q 3 0

(6)

Fluid potential: the driving force


In many complicated physical situations a number of forces compete to control a natural or artificial system. Instead of resolving the force vectors at every point, it is often simpler to work with a scalar potential, Q, calculated for each point. This generalapproachtopetroleummovement hasbeen described in more detail by Hubbert (1953); a summary of his work is given here and some alternative techniques and interpretations are offered. In conventional mechanics, the motion of a mass resting on a frictionless surface may be examined mosteasily by associating a mechanical potential QM at each point. QM is defined as the work against gravity neededto bring unit mass from some datum level to the point of interest. The force experiencedby unit mass is then given by:

i.e. zero force andpositive curvature in Q. We will now define Q in a form applicable to petroleum migration, known as the petroleum potential QP. A similar potential, Qw, maybe defined to describe the subsurface movement of water. Note, however, that a certain condition must be satisfied before any force field can be represented as a gradient of asuitablepotential, namely the curl of the force must vanish. This is only strictly true for the fluid force field when the fluid density does not varywith horizontal position. This is generally not the case during long distance petroleum migration where oil and gas densities mayvary considerably. For more local considerations, however, where the petroleum densities can be regarded as approximately constantit is convenient to use the fluid potential concept. The following sections will show methods for calculating the various contributionsto QP. Sinceknowledge of Qpp leads to an understanding of the driving forces acting on petroleum, it leads to a description of the factors controlling petroleum migration. Table 3 compares the mechanical, electrostatic and fluid potentials.

Definition of water and petroleum fluid potentials


The fluid potential is defined as the work necessary to transfer unit volume of fluid from reference conditions to the relevant (subsurface) conditions of interest. The reference condition is taken as bulk fluid at a depthof zo = 0 and a gauge pressure of P O = 0 (NB, z increases, as depth increases). The conditions atadepth z and pressure P include any capillary pressure caused by petro1eum:water interfacial tension y in pores of radius r . The work done (assuming incompressible fluids) is therefore:

[P- PO]V - mg[z - 201 + 2y -

K1v
in scient@c

The first two terms are the work done against pressure and gravity respectively; the final term is the work done against capillary forces in transferring bulk fluid into a porous rock
Table 3. Examples of the use of potential fields
applications

Force,
Primary Migration Secondary Migration

UnitsPotential Application Mechanics Joules energy Potential Electrostatics Potential Hydrology Water potential
d%/&

F ,
Volts Pascals (Nm-*)
q d@,/&

Fig. 6. Definitions of primary and secondary migration after Tissot & Welte (1984).

dQW/&

334

W . A . ENGLAND E T A L .
HYDROSTATIC ENVIRONMENT Petroleum Petroleum Water liquid gas

medium. (Strictly speaking the capillary pressure termshould be multiplied by cos /3 where /3 is the angle the petroleum-water meniscus makes with the pore wall, measuredthroughthepetroleum. In practice, p = 0-30: cosB will tendto unity and may be ignored.) Since the potentials will be defined aswork per unit volume, and noting that pp = m / V , P* = 0 and zo = 0:

@ p = P - p p g z + 2Y r
by analogy, the water potential is defined by:
@W

HYDRODYNAMIC ENVIRONMENT

= P - pwgz

(8)

Petroleum Water liquid gas

Note the definitions introduced by Hubbert (1953) are for unit mass of petroleum or water, and therefore differ by a factor of p from our definitions. The use of unit volume as reference a results in potentials being measured in convenient units of pressure. Our water potentialis identical to the overpressure, used by soil scientists, or the hydrologists piezometric pressure. Since @ , , the water potential, is widely applied in hydrodynamics, is often relatedto by substitutingequation (7) intoequation (8):
@P

Fig. 7. The forces acting on unit volumes of water, petroleum liquid


and petroleum gas may be obtained from the vector equations (11) and (12). The resultant force F Por F, acting on petroleum or water is that given by vector summationof the pressure gradient, V P , and the appropriate density-relatedterm ppg or pwg. representation. Differentiating equations (7) and (8) gives:

= @W

+ (Pw - P P k Z + 2 y / r (9)

It only is in small-pored rocks-notably source rocks-that the 2 y l r term in equation (7) or (8) will be significant. For example in a clay with 60 nm pores it may reach a value of 1MPa ( = l 0 atm), assuming an interfacial tension of y = 0.03 N m-. However, in rocks with larger pores, such as sandstones, the capillary contributions to @p are insignificant, being <0.01MPa (-0.1 atm). This behaviour of the capillary contributionto fluid potential means that capillary effects can be safely ignored in rocks with large pores such as sandstones, butwill be significant in clays. Since many source rocks are organic-rich clays or silts, primary migration will be influenced by capillary effects. In contrast, capillary effects will be negligible for secondary migration because of the largepores found in the camer beds. The seals, which prevent petroleum leaking out of reservoirs are often fine-grained rocks which generate sufficient capillary pressures to present a barrierin (Pp which prevents further movementof petroleum. In ordertointerpretpetroleumand water potentials, consider the water potential and its behaviour in hydrostatic and hydrodynamic systems. In a normally pressured (hydrostatic) environment, the pressure is given by the equivalent columnof water; thus P = pwgz and equation (8) shows that = 0 at any position.There will be no driving forceto cause movement of water in any direction since V a W is everywhere zero. However, petroleum oil orgas, will experienceastrongupwardforceona unit volume given by (neglecting capillary effects):

Fw = - V P

+ pwg

(12)

This is known as the buoyant force, and is larger for petroleum gases than petroleum liquids since pgas < pail. In a hydrodynamic regime, W W # 0 at all points, and the forces experienced by petroleum will no longer only be vertical. One way of visualizing the interplay of forces experienced by migrating petroleum, is by using a vector

Figure 7 shows these force components and the resultant FP and Fw for hydrostatic andhydrodynamic environments, ignoring the capillary term 2 y V ( l / r ) which is small except at lithological boundaries involving fine-grained rocks. In a hydrostatic environment there is no flowof water and Fw = 0. Since V P = pwg, theresultant force l$ = (pw - ppg) and is known as the buoyant force. The resultant forces on migrating petroleumarethus always vertical: petroleum gases and liquids will migrate in the same direction. Because of their lower density, petroleum gases experience larger values of FP. Thus once a free gas phase separates it will migrate considerably faster than its associated liquid phase, particularly since the relative permeability of petroleum gases through water-wet rocks is usually greater than that for petroleum liquids. In a hydrodynamic environment, water will experience an additional force G . This force may be in any direction, depending onthenature of the flow involved. Figure 7 illustrates that the resultant forces on petroleum fluids and water will be in different directions. Thus petroleum liquids and gases may migrate in different directions. This has important consequences for petroleum exploration, where it is vital to assess the migration pathways of petroleum in order to compute drainage areas and potential reserves. If contours orthogonal toFPare drawn, migration occurs only at right angles to these equipotential contours, along lines of steepest descent. If a suitable rock structure, with a seal or cap rock is present, migrating petroleum fluids will accumulate in such a way that the tilt of their oil-water or gas-water contacts are along an oil or gas equipotential surface. Figure 8a shows that in a hydrostatic environment,

FL PU EIT D R SO L E U M O FM O V E M E N T

335

Gas-waterequipotentialcontour

----=

Gasliquid equipotential contour Impermeable rock

important: consider an initially water-saturated rock, through which petroleum may flow. The rock will contain pores, whose radii vary greatly in size. Thus as the petroleum potential gradient, VaP, is increased across the rock, petroleum will startto flow into the pore space. However, flow will occur only along a given tortuous pathway through the rock if the applied pressure is greater than the opposing capillary pressure 2 y l r . (i.e. flow will occur only if VcPp for a given pathway is always negative.) For relatively low pressures, no continuous pathway through the rock is possible, and no flow will occur. As the pressure difference across the rock increases, a critical value will be reached at which continuous pathways through the rock occur. This is the minimum pressure at which bulk flow can start. This process is illustrated in Fig. 9. This type of non-linear behaviour is obviously inconsistent with Darcys law, which only allows for linear flux-V@ relationships. However, in the absence of a better theory, modified versions of Darcys law will be used in the remainder of this paper. For practical purposes, the intrinsic permeability of a pore network filled with moving petroleum may be estimated from its mean pore radius f (Amyx et al. 1960) according to Pouisseilles law:

Fig. 8. Equipotential contours for petroleum liquid and petroleum gas under hydrostatic (a) and hydrodynamic (b) conditions. Equipotential contours are perpendicular to the forces acting on the fluids (see Fig. 7). Thus, to be at rest, a petroleum-water boundary (or contact) must be parallel to an equipotential surface.

where 8 is the tortuosity of the network, defined as the averaged ratio of the path-lengths travelled by petroleum fluid to the geometrical length of the region of rock considered. Substitution into equation (13) yields:

all petroleum-water contacts are horizontal. However, in a hydrodynamic environment, Fig. 8b, tilted contacts occur. The recognition of tilted contacts hasbeen predicted and observed for many years (Hubbert 1953).

Direction, rates and ranges of petroleum movement


Having discussed the forces acting on petroleum, wewill show how its rateof movement may be calculated from rock and petroleum properties.

where q now refers to the flux per m of the petroleum-filled network. Thus the petroleum flux per square metre of rock may berelatedtoequation (15) by including factors of porosity, G, and the fraction of the porosity that is

Flow through porous media


For single-phase flow through a porousrock, Darcyslaw has been found to be an accurate description: -k, q =-V@
P

q isknown as the superficial Darcy velocity,since it has

units of ms-, but is better thought of as the volume flux of fluid passing across unit area of the rock (m m- S - ) . V@ is the fluid potential gradient discussed previously, and k,lp defines the constant of proportionality between q and V@. k, is the intrinsic permeability of the rock measured in m (1 Darcy = 9.689 X lO-I3 m). p is the dynamic viscosity of the fluid measured in Pas (1 centipoise = 10-3 Pas). Darcys lawissuccessful for single phase flow; unfortunately itis inadequate for describing multiphase flow. In single phase flow V@ only contains the fluid pressure gradient term of equation 7: i.e. the capillary term may be ignored since no water-petroleum interfaces are present. In two-phase systems, the capillary term will become

Fig. 9. Sketch toshow the nature of petroleum flow through a core of porous water-filled rock.

336

W. A. ENGLAND E T A L .

petroleum-filled, known as the petroleum saturation S:

q=

r2
802p
(16)

Inordertoestimatethe flux of petroleum flowing (i.e. migrating persquare metre of rock) one must have good estimates of the quantities on the right hand side of equation (16). C # J and p are measured by standard techniques (see Amyx et al. 1960). The tortuosity factor, 6,for most rocks is taken to be about d 3 , the theoretical prediction for a loose random pore structure (Li & Gregory 1974), and the dynamic viscosities of petroleum under subsurface conditions have been discussed above. S, the fractional oil saturationat which the sample of rock first contains interconnecting pathways, may be estimated using percolation theory (Stauffer 1979). Percolation theory attempts to describe complex interconnected networks, such as the pore space of a rock. Calculations for idealized networks indicate that 20-30% of a rocks pores (on a number basis, as opposed to a volume basis) must become petroleum-filled before petroleum can flow through the rock. Because of the effect of capillary pressure, when conduction begins the petroleum will be contained in the largest pores-i.e. only those with an entry radius above some value. Thus the critical volume saturation will exceed 20-30%. Once interconnection is complete,breakthrough occurs, and the saturation of the network will remain constant. We have carried out experiments to investigate this effect, by studying oil breakthrough in laboratory cores of typical reservoir rocks. Liquid petroleum at a few atmospheres pressure was supplied toone end of a core sample and when breakthrough occurred, the petroleum saturation was calculated from knowledge of the porosity and volume of petroleum supplied. The results are presented in Table 4: they confirm that S is always greater than 20%. The maximum values found reached 90%. The mean radius, F, of the petroleum-filled network may be estimated by mercury porosimetry (e.g. Ritter & Drake 1945). By injecting pressurized mercury into dried core plugs, the relationship between pressure and mercury saturation may be obtained. Since the non-wetting mercury will enter the largest porethroatsfirst,the effective pore radius atthat pressure can be ascertained from P = 2 y / rthroat,the capillary term of equation (7). A distribution function, S ( r ) , whichisdefined as the cumulative volume fraction of rock porosity occupied by pores whose radius is
Table 4. Pore space saturations ( S ) required for oil flow through rocks

greater than r , may be deduced from the mercury porosimetry data. Figure 10 shows two typicalcumulative pore volume, versus poreentry radius curves. The average porethroat radius for interconnected flow across a rock may be estimated from the critical saturation, S . Since our experimental results (Table 4) suggest S = 50% on a volume basis, the pores generally involved in bulk flow are represented by the shaded areas. From these measurements the average pore throat radius F may be deduced after some mathematical computation. Equation (16) may then be used to estimate the petroleum fluxinprimary and secondary migration.

Variation in water fluid potential caused by sediment compaction and petroleum generation
In this section it is shown how the water potential may be calculated in a sediment which is undergoing both compaction and petroleum generation. The principle of the method, Terzaghi (1948), is to consider a volume element of variable size, which always contains the same mass of rock. Thus, as compaction occurs, the volume elements reduce their size appropriately. Figure 11 defines the volumes involved: in this one-dimensional model, only vertical water flux, qz (normalized to unit area of rock), need to be considered with any internal volume generating processes. The rate of change in volume may be written:

I d --(AV

v dt

d - V )=(qJ dz

V is the resulting volume change in an element, and AV represents the volume created by petroleum generation and thermal expansion. (In practice the effects of thermal expansion are generally small and were ignored.) The flows, qL, were calculated using Darcys Law, equation (13).

a -

Porosity

Sample 9.2 6.5 3.9 6.7 15.1 29.0* 20.0 24.5 18.0

(%)

. ... . ... .... ... . . .. . ....... ..\.

. . . . . . .:::X* .. ......

.
5

Yorkshire Deltaic 59.6 sandstone series 91.0 Grit Millstone 65.3 Cotswold Oolites Sandstone Berea (US) St Bees Sandstone

10 Pore throat radius ( r ) - Llm

15

56.0 47.8

*Average value of three results: 17.6%, 31.0%, 38.4%.

Fig. 10. Cumulative plots of pore volume filled relativeto pore throat radius for two sandstones. Sandstone A hasa porosity of 19.7% and a permeability of 5.1 X 1O-l m; for sandstone B the values are 20.3% and4.01 X l0-l m respectively. Since our experimental breakthrough results suggest that S > 0.5 for breakthrough of petroleumto occur, pore throats between 0.5 and 3 pm in sandstone A, and between 8 and 13 pm in sandstone B, must be filledfor flow of petroleum to occur.

M O V E M E N T O F LP U EIT DR SO L E U M
42+62

337

Petroleum

Equation (21) is written in terms of ad/&, i.e. the rate of increase in effective stress with time. Since this is principally caused by the extra loading of freshly deposited sediment, itis more convenient to workwith (dhldt) the rate of increase in sediment thickness at the surface, i.e. the burial rate. U is the pressure due to the combined weight of water and rock.

Expansion

where pR is the rock mineral grain density. Substituting U into equation (21), and differentiating with respect to time gives:

42

Fig. 11. Definition of terms used in the calculation of water

potential in compacting sediments. q. is the vertical pore water f l u into a volume elementof variable size;qz+sris the vertical flux out of the volume element; and AV is the extra volume created by the generation of petroleum and the thermal expansion of water in the volume element. Of course, in a compacting sediment, the intrinsic permeability, k,, is itself a function of depth and time. This is modelled by taking the localvalues of porosity, G, as estimated for a given element, and using a porositypermeability correlation for shales and mudstones:
k, = A @

which will be used latertorepresent do'/& in terms of known geological quantities. Experimentally it has been found that the equilibrium porosity can be calculated from the effective stress by the equation:

1 -G G

1-40 Go - - c c l o g l o ( ~ )

where Go and U; are, respectively, the values of porosity and effective stress at some reference level. For convenience thislevelmay te taken as a depth of 10 m with the overlying sediment being normally pressured and having an average porosity of 0.5. Thus 5 6 is given by: The best fit to experimental data is given when Go = 0.55 for shales and 0.49 for sandstones. It is now possible to re-express equation (17) using equation (18); the volume generated by kerogen breakdown FK is calculated as indicated above:

A is a lithology-dependent constant, for which 4 X 10-15m2 for shales and 4 X 10-'' m' for siltstones were, taken, unless otherwise stated. This correlation (Smith 1971) gives a reasonable behaviour for k, as a function of Cp. The rather low value of A = 4 X 1O-l' m' was found to be necessary in order to give agreementbetween calculated overpressures and field measurements. For sandstones, the relationship reported by Berg (1975) was used: k, = 0.084d2G5.' (20)

01

where d is the average grain diameter in metres. The rate of generation of petroleum from kerogen was calculated using an Arrhenius law formalism developed for oil exploration (A. S. Mackenzie & T. M. Quigley in preparation). l/V(aV/at), the rate of change of sediment volume with respect to time is, by definition, equivalent to the rate of change of linear strain in the z direction i.e.

where e is the base of thenatural logarithm and C, is a constant known in soil mechanics as the compaction coefficient. Typical values of C, are 0.42 for shale, 0.88 for chalk and 0.25 for sandstone. The effective vertical stress U' representsthe pressure due to the overlying rock supported by the sediment. The total pressure due to rock and water is defined by U. Thus the following definition of U' holds:

u'=u-@w-pwgz

(22)

In other words the rock supports the total weight of the overburden less the weight supported by the pore fluids.

since u'/(Cc(l - G)logloe) is nearly constant. C,. is a variable in time and z . It is known in soil mechanicsas the coefficient of consolidation. Our studies have shown that FK issmall compared to 3/az{G(a@,,/az)} and may in most cases be ignored. In order to solve equation (28) it is necessary to define the initial conditions at the surface, in the rock column and at the bottom of the column, where the sediments make contact with the basement. These assumptions and constraints are known as boundary conditions, and play an important role in determining the outcome of a porepressure calculation. The following boundary conditions have been assumed: (a)an initially hydrostatic environment (aw = 0 at all points), (b) a zero-valued water potential at the surface at all times, and (c) either a constant basement overpressure, or an impermeable basement with no flow (i.e. V @ , . , = 0 at the bottom of the sediment column). Equation (28) was solved using finite difference numerical methods; U' and 9 are calculated from the

338

W . A . ENGLAND E T A L .
Burial rate (m Ma-'J

200 r

Depth (km)

4
Depth(km)

Fig. 12. Predicted present day variation in water potential (aw) as


a function of depth, for a shaley sequence deposited onto an impermeable basement, currently buried to depths >l0 km, at burial ratesof 10,35, 100,250 and loo0 m Ma-'. The trends were predicted using equation (28). equations throughout a compacting column of rock for small increments in sediment loading. The output consists of aW, 4 and U as a function of both depth and time. Figure 12 shows the results of calculations based on this method for water potentials aW, which result when shale is deposited at between 10 and 1000 mper million years onto a deep impermeable basement (at a depth greater than 10 km). It is clear that at depths greater than c. 3 km the slope d@,/az is essentially independent of deposition rate. This can be explained in terms of equations (22) and (25). At great depths, permeability and porosity have been reduced by compaction to the point where water flow rates are negligibly small compared to the deposition time scale. All the additional stress due to sediment deposition is then borne by the pore water as a', the effective stress on the rock matrix has become effectively constant. Rearranging equation (22) and making use of equation (23)gives the slope d@,/dz:

Fig. 14. Comparison of predicted variation in pore pressure as a function of depth, as for Fig.13, except the pressure data shown are for the Gulf of Mexico Coast (taken from Dickenson 1953) and the predicted trend is fora deposition rateof 150 m Ma-'.

In Figs 13 and 14 measured pore water pressures (i.e. + pwgz) obtainedfrompetroleum exploration in the NorthSea and the Gulf of Mexico are shown.Assuming deposition rates of 35 and 150 m per million years
cPw

respectively in the North Sea and the Gulf of Mexico, we used our model to predict the pressures expected as a function of depth at the present-day. It can be seen that although the model appears to fit the trend there are many significant deviations from it. This 'scatter'is attributable to the presence of dipping high permeability features, such as sand lenses not considered in ourone dimensional model. These porous bodies will have a very small gradient in across them, which will disturb local water flows as sketched in Fig. 15. This may increase or decrease at a given depth from the trend if shale alone were present. Since pressures in the field may be measured accurately onlyin permeable rockssuch as sandstones, drilling into points A or C will give, respectively, over-estimates or under-estimates of the undisturbed sediment pressure in the absence of the sandstone lens. Only point B shows the undisturbed pressure of the low permeability sediment. These considerations can explain the wide ranges of pressures measured in the North Sea of Gulf of Mexico and shown in Figs 13 and 14. Figure 16 shows the results of our calculation procedure for the more complicated case of a petroleum source rock sandwiched between equal thicknesses of shales and with a sandstone overburden. The geological column is assumed to have a normally-pressured basement.

lI2
150

3 m E a

? !

100

4
Depth(km1

Fig.U. Comparison of predicted variation in pore pressure as a


function of depth with observed pressures for the North Sea, using equation (B), and a burial rateof 35 m Ma-'.

Fig. 15. Perturbation to water potentialsand water flows caused by a dipping sandstone lens within a shaley sequence. This causes QW at C to be lower than predictedby our model, which assumes only vertical flow. Conversely aW at A will be higher than that

predicted.

MOVEMENT OF FLUIDS PETROLEUM

339

0.1 km and k,, = 10-" m2 (i.e. typical value for a shale) equation (32) becomes:

--

QLAT

QVERT

- 10-15kLAT

(33)

v)

Therefore, if kLATis greater than 10-15m* ( = l mD), then lateral flows will be moreimportant than vertical where laterally extensive carrier beds are present. Although permeabilities measured on sandstone cores may substantially exceed l O - I 5 m', the presence of faults and the tortuosity of high permeability streaks within extensive sandstone formations may decrease their overall effective permeability.

rvJ
v)

m c

a -

Migration directions
Having discussed the calculation of and hence the flow of water in compacting sediments; we will now describe how the direction of water movement may be related to that of petroleum. First V @ . , must be related to VmP (which determines the forces acting on petroleum) and (via permeability) to migration rates. Equations (11) and (12) define the forces acting on petroleum and water; by substituting one into the other FP can be related to Fw: since: FP=&+(Pp-Pw)g Fw= -V@., and F P = -V@, (34) (35)

(a)

(b)

(C)

F % . 16. Predicted variation in QW through threeshaley sequences that contain a source rock. The water potential was calculated by assuming the sequences were buriedto 3.5 km at 100 m Ma-' by sandstone with pore pressure closeto hydrostatic; the sequences are underlain by similar sandstones. See text for full explanation.
Figure 16a shows how the water potential attains a maximum value near the middle of the source rock portion, which is caused by the extra volume created by petroleum generation. Thus water will flow both upwards and downwards away from the centre of the source rock unit. Figure 16b was calculated for similar conditions as Fig. 16a, except thatthe overlying shale istwice as thick:all water flow through the source rockisnow downwards. In Fig.16c the overlying shale in Fig.16a has been replaced with a more permeable siltstone: more water can now escape upwards, and most of the water flowing through the source rock will flow upwards. We have discussed modelling of the flow of water vertically; however our previous discussion of the North Sea and Gulf of Mexico results suggested there is often a significant lateral component, at least in sandstone. We must therefore examine situations where it may be assumed that water flow is mainly vertical, and where it is mainly lateral. The competition between vertical flow and lateral flow may be quantified using Darcy's law, equation (13), and Fig. 17. Thelateraland vertical flows, QLAT and Q m R T (in m3S-') are estimated by multiplying the Darcy fluxes q L A T and qWRTby the corresponding cross-sectional areas H Y and Wlcos a Y .
QLAT

+ ( P P - Pw)g

For vertical flows in rocks less permeable than 10-15m' (this includes primary migration) petroleum will move in the same direction as water provided that the buoyancy term (pp- pw)g is less than VQW (or in the vertical case V@,""). Taking pw = loo0 kg m-3, pp= 600 kg m-3 for petroleum liquid and pp= 200 kg m-3 for petroleum gas, the buoyancy term has value of c. 4000 Pa m-' and 8000 Pa m-l for liquid and gaseous petroleum respectively.Figures 12, 13, 14 and 16, suggest that in over-pressured sequences at depths greater than2-3 km d@.,/dr exceeds 10,000 Pa m-', reaching 15,000Pa m-'. Most petroleum is generated and expelled from source rocksbelow 2 km; therefore during verticalmigration the buoyancy term is smaller than the V@., term, and as seen from Fig. 7 petroleum willflowin thesame direction as water. Extrapolating from Figs 15 and 16 implies that petroleum will migrate out of the source rock, up or down, towards the nearest continuous horizon of permeability greaterthan10-'5mZ; henceforth called a lateral carrier. However, if the overlying strata are relatively impermeable

= -.

~ L A T

H . Y * V@bAT

(30)

V@,",^' isgivenby(@2-@l)cos a / W while V@FRT is equal to (@z-@l)/zor (@+D,)/W tan a: Thus the ratio of lateral and vertical fluxes is estimated to be:
kLAT QLAT

--

QVERT

H --. tan aW VERT

(32)
=

Assuming realistic values of a = 2", W = 300 km, H

Fig. 17. Competition between vertical and lateral flows,QWRTand QLAT, through a sandstone embedded in shale.

340

W . A . ENGLAND ET A L .

compared to the underlying strata, a large gradient in will produce a powerful force causing downward migration of petroleum into more permeable lateral carrier beds. The buoyancy force acting updip on petroleum migrating in lateral carriers is given by multiplying the vertical buoyancy force by the sine of the dip of the beds. Taking a typical dip of 2O, and the above densities this yields 140 Pa m-' for liquid and 280 Pa m-' for gas. Case history in the North Sea suggests values of 500-1000 Pa m-' for V@&AT are common at depths >3 km. Boththe water potential gradient and the force due to buoyancy push petroleum (and water) in the same direction-updip. At depths >3 km the major force driving petroleum is water potential gradients; buoyancy takes over at shallower depths. Until this point, capillary effects have been ignored-the 2 y / r term in equation (7). This term will generate significant forces only when there is a lithology change which causes a gradient in r . This important effect will occur at the boundary between fine and coarse-grained rocks. In a simplified picture of capillary forces, there will be a large capillary pressure difference which drives petroleum out of source rocks into carrier beds, e.g. Hubbert (1953). By makinggeochemical measurements on actively generating source rocks we have found (Mackenzie et al. 1986) that source rocks are more efficiently depleted of petroleum within 5 m of their margins which are in contact with sandstone strata. We attributethis to capillary effects.

Table 6. Petroleum liquid mean subsurface superjicial velocities and dimensionless numbers

m-'s-')(m3 Vertical migration in shales Lateral migration in sandstones

c,

NRe

4 X 10-l~
8 X 10-'O 10-'0

7 X 10-l~. 5 X 10-l~

10-'O

9 = 0.2 (sandstones), 0.1 (shales); (sandstones),10-8 m (shales); y = 0.03 Nm-': . , .= 650 kg m-3; p = 5 X 10-3Pa S-'; V @ , = 103Pam-l (LAT), 104 Pam-' (VERT).
Typical' values: S = 0.5;
f = 10-6 m

Nature of petroleum flow


We havesuggested that petroleum generated in a source rock will migrate vertically up or down towards a neighbouring horizon of high lateral permeability, i.e. permeability is morethan 10-15 m* (or 1 millidarcy). This horizon is called a lateral carrier. Petroleum will remain in the lateral carrier unless it overcomes the excess capillary pressure that opposes its entry into the smaller pores of the overlyingseals. If this happens, the petroleum will then movevertically into and through the overlying rock until another lateral carrier/seal systemis encountered, inwhich case it will again migrate laterally updip. We have shown that the typical magnitude of V@, for vertical migration is about 104Pam-'; the magnitude of V@, for lateral migration along carrier beds is about 103Pa m-'. The variables required to solve equation (16) for q , the flux of petroleum per m' of rock, are now defined. Taking the mean values of p@, S and f from Tables 2 and 5, and the typical values of 103Pa m-' and 104Pam-' determined above for V@kAT and V@FRT respectively, the average values for q may be calculated for lateral migration of petroleum liquids in sandstones and verticalmigrationin shales. The values are given in Table 6. A measure of the significance of these flow rates is given
Table 5. Typicalproperties of rocks important for migration (approximate burial depth of 3 k m )

Sandstones

Shales

9 (porosity)
(mean radius of petroleumpores filled in 10-6 metres) S (saturation)
f

0.2 0.5

0.1
10-8

0.5

by the capillary number, C, = p q / y , which represents the ratio of viscous to capillary (or surface tension) forces in establishing pathways through pore networks. Experiments with oil-water mixtures in rocks have shown that at capillary numbers greater than 10-4, viscous forces become important. Hinch (1985) has related the small value of this number to thepore geometry: values in this range are expected if the pores have throatsa tenth of the size of their lengths. However, Table 6 shows that at geological flow rates the capillary number is never greater than 1 0 l ' : capillary forces therefore dominate at all times. The capillary numbers for our experiments, designed to calculate the poresaturation required for petroleum flow through porous media, are about 10? capillary forces still dominate and the measured values for saturation at breakthrough will thus apply under geological timescales. The Reynolds number, NRe = p q P / p , measures the ratio of inertial to viscous forces. Taking an arbitrary density for subsurface petroleum liquid of 650 kg m-3, the values calculated range from 10-l' to 5 X 10-l6 (Table 6). These may be regarded as small: they show that during petroleum migration, allflows occur in the 'laminar' (non-turbulent) regime. A third parameter of interest is the Peclet number qw/D, Lerman (1979), which measures the relative importance of bulk and diffusive mass transport over a particular length scale defined by W. D is the effective diffusion coefficient for a given component in the water-filled rock: it is distinct from the molecular diffusion constant measured in a single phase. If the Peclet number is c. 0.5 diffusive mass transport will be more significant than bulk flow (Mackenzie et al. 1986). Estimates of the Peclet number, using Leythaeuser et al's. (1982) field estimates of D in shales for Cl-C, hydrocarbons are shown in Table 7, whichalsoshows the Peclet number for length scales of W = 10 m, 100 m and 1000m, using a value of q appropriate for liquidphase vertical migration in shales (Table 6). Table 7 suggests that, with the exception of methane and ethane, diffusion is insignificant during vertical migration in shales over distances greater than 10-100m for all the components of petroleum. The Peclet numbers (and hence the diffusive length scale) for lateral migration in sandstones will be the same order of magnitude as for vertical migrationinshales: although q increases by about three orders of magnitude for sandstones (Table 6), laboratory experiments (Krooss 1985)suggest a similarincreasein D

MOVEMENT OF FLUIDS PETROLEUM

341

Table 7. Calculation of Pecletnumbers for verticalmigration petroleum liquids

of

Peclet numbers (qw/D) for different length scales (W) (m2s-')D* Methane Ethane Propane Isobutane n-butane n-pentane n-hexane n-heptane n-decane n-tncosane
= -0.283C,,m-2
S

w=10 wm =100m

w=lOOOm

2 X 10-10 1 X 10-10 6 X 10-" 4 X 10-11 3 X 10-" 2 X 10-" 8 X 10-'' 4 X 10-'' 6 X 10-'3t 1 X 10-I3t

0.02 0.04 0.07 0.10 0.13 0.20 0.50 1.0 6.7


40

0.20 0 . 4 0 0.70 1.0 1.3 2.0 5.0

1 0
67

400

2.0 4.0 7.0 10 13 20 50 100 670 4Ooo


q =4 X

* From Leythaeuser et al. (1982); t Extrapolated using log,, D


10-13 m3
-1

10.39 where is n carbon number. (Table 6).

fromshales tosandstones.Hence because petroleumcan migrate more than lOOkm from its source rock (e.g. Tissot & Welte 1984), these results support our earlier assumption that diffusive transport is only importantfor very short distance migration of light n-alkanesduring expulsion of petroleum and vertical migration. Otherwise bulk transport is the dominant process of migration.

history from regions whose petroleum geology is well understoodandwhere most accumulations of petroleum have been discovered. By subtracting the volumes of discovered petroleum from theexpelled petroleum volumes, calculated using methods reported elsewhere (Cooles et al. 1985; A. S . Mackenzie & T. M. Quigley in preparation), the lost petroleum volumes can be estimated. These volumes can be ratioed to the pore volume available for migration (the total volume of rock along the migration pathway multiplied by the average porosity). Detailed results will be reported elsewhere (MacGregor & Mackenzie1986; A. S . Mackenzie & T. M. Quigley in preparation);these will demonstrate that theaverage ratio of lost petroleum to pore volume is of the order of a few percent. Our experiments on cores reported in (Table 4) suggest that flowing petroleum onaverage exploits about 50% of the available porosity. In order to reduce the effective overall saturation to a few percent, this implies that the flowing petroleum uses less than 10% of the rock area. This observation is in agreement with our analysis of the relative migration fluxes, whereit was concluded that migrating petroleum exploits at least between l and 10% of the rock area available for flow. The comparison between theory and observation assumes thatpetroleum migrates by forging an interconnected path of petroleum-filled porosity leading away from mature source rock: petroleum will only migrate as far (and as fast) as the volume of petroleum expelled from the source rock can spread out, whilst remaining fully interconnected.

Losses during migration


The migration rates reported in Table6 must be affected by the rate at which petroleum may be supplied by a source rock. Typically source rocks for oil are 100 m thick and have potential yields of petroleum fluid relative to rock weight of 0.02 kg kg-'. Most of thispotential is realized between 120-150C (Cooles er al. 1985). Geological heatingrates range mostly between 1 and 10 "CMa-l. Assuming a rock density of 2400 kgm-' and a petroleum density of 650 kg m-3, then the flux across the boundary of the source rock is about 8 X 10-" to 8 X lO-I4 m3 m-' S-'. This flux is compared with the values for q inTable 6. If vertical migration throughshales,relative to rock area, occurs at 4 X 10-13m3m-'s-l (Table 6) then the cross-sectional area of rock filled with flowingpetroleum must be greater than or equal to between two hundredths and two tenths the areaof thesource rock interface.A similar argumentforlateral migration through sandstones at8 X 1O-I' m3 m-' S-' (Table 6) suggests that the cross sectional area of rock filled with flowing petroleum must be greater than or equal to 10-' to 10-4 the area of the source rock interface. The area of rock available for vertical migration will be similar to the area of the source rock interface. The above analysis suggests therefore that the petroleum must exploit at least two hundredths to two tenths of this area. In our experience, the ratio of the area of mature source rock to the cross sectional area of a lateral carrier (perpendicular to the direction of flow) is about 1000: 1. Therefore,the minimum area that must be petroleum-filled during lateral migration will correspond to one hundredth to one tenth of the area available for petroleum flow. Perhaps a better approach to estimate the area of rock exploited by flowing petroleumistomakeuse of case

F i g .18. An example of an eroded migration pathway. Oil stains the coarser partsof a turbidite sequence belonging to the Socorro Formation of Eocene Age.The picture taken by Martin Heffernan near the Ancon oilfieldson the Santa Elena Peninsula, Ecuador; the hammer handle measures0.25 m in length.

342

W . A . ENGLAND E T A L .

It also assumes, after migration, a volume of petroleum will remain in the sequence. This residual volume will be similar to that necessary to connect both sides of the sequence with flowing petroleum. There are two supporting arguments for these assumptions: the dominance of capillary forces implied by low capillary numbers (see above) and the field evidence of petroleum-stained rock. Because capillary forces dominate over viscous forces, water is unlikely to displace petroleum after the supply of petroleum from the source rock has dried up. The amount of petroleum left behind in a bed after migration will be approximately equal to the amount necessary to form an interconnected pathway initially. Migration pathways may be examined either by drilling, or by analysis of uplifted anderoded pathways. Inboth cases there is strong circumstantial evidence that migration involves considerable losses. The frequency with which stained rock is encountered suggests that petroleum occupies between 1 and 10% of the rock along the migration pathways. Figure 18 showssuch an observation; migrating petroleum seeksoutthe larger poresand coarser-grained regions of the rock.

Entrapment, filling and mixing of petroleum fluids


The final stages of migration, in which an initially water-wet reservoir rock is transformed into a petroleum accumulation now is considered. The initial trapping and filling mechanisms, and the mechanical and chemical adjustment of petroleum to its new reservoir environment is described. The effects of convective and diffusive mass transport as well as gravitational segregation are considered.

Entrapment and filling


There are no characteristic physical properties to distinguish petroleum reservoir rocks fromcarrier beds or migration pathways. It isonly the behaviour of the forces acting on petroleum which defines the location of possible a petroleum reservoir or trap. A trap may be loosely thought of as a dead-end or cul-de-sac; it is a regionwhere the forces controlling petroleum migration converge. A trap is always associated with fine-grained a caprock, which prevents vertical movement by capillary forces. The capacity of caprock will be discussed below. Figure 19 shows the directions of the forces in the

Fig. 19. The direction of forces acting on petroleum fluids in the f a trap (after Dahlberg 1982). The combination of water vicinity o potential, capillary pressure andbuoyancy forces petroleum towards the reservoir in the crest of the structure.

vicinity of a hypothetical trap, together with the associated contours of @p. The much higher capillary forces acting on petroleum in the fine-grained caprock causes a very sharp gradient to occur in QPpat the reservoir rock-caprock interface. The relevant contribution to is governed by 2ylr as shown in equation (7). The caprock provides a potential energy barrier to petroleum leakage, the capacity of this barrier may be estimated from rock properties as shown later. Equation (11) also shows how the forces acting on petroleum (h) depend on the hydrological environment and the density of the liquid or gaseous petroleum phases present. The forces experienced by petroleum in hydrodynamic environments may cause verysignificantshiftsin the positions of petroleum liquid and gas accumulations compared with predictions made assuming hydrostatic conditions (see Fig. 8). Fuller details are given by Dahlberg (1982), who considers in more detail its importance to petroleum exploration. We have previously shown that secondary migration may be thought of as involving two levels of focusing into (a) the beds with the largest average pore entry radii and (b) the mostaccessible pores of the beds described in (a). Figure 20a illustrates the situation envisaged by thismodelwhen migrating petroleum first encounters aregionwhichmay subsequently become a petroleum reservoir. The migrating petroleum will movealonga dendritic network from the source rock which demarcates the coarser-grained parts of the carrier bed. The threedimensional structure of the migration pathway will obviously dependonthe depositional environment of the rocks involved. From our experiments reported above, the petroleum saturation of the dendritic network itself will be c. 50% whichis thesaturation level at which petroleum entersthetrapforthe first time. Within thetrap itself, petroleum will not behave as a continuous fluid. One therefore expects thatthe petroleum will initially fill the reservoir at the crest of a trap as an advancing front-in rather the same way as a chromatographic front advances along a chromatographic column. This first stage of reservoir filling is shown in Fig. 20b. Because the pore sizes of the various layers of a reservoir rock are not uniform, migrating petroleum will enter first the layers with the largest pores. This situation is shown in Fig. 20b, in which several layers with the highest pore sizes have filled with petroleum. The oil saturation within a layer is 4 0 % but the overall saturation on a reservoir-wide basis is about, say, 2%. Since exit from the trap is prevented by an overlying seal, fresh petroleum arriving from the source will be forced into successively smaller and smaller pores. Thus new bands of reservoir rock will become petroleumfilled (Fig. 20c). As more and more petroleum fills the trap, the buoyant pressure exerted by the petroleum increases with the growing height of the inter-connected petroleum stringers. This increasing pressure will be able to overcome the larger capillary pressure of the smaller water-filled pores, and petroleum will displace water from these pores. This can continue until the height of interconnected petroleum exerts sufficient buoyant pressure at its uppermost point to overcome the capillary pressure acting between the reservoir rock and the overlying watersaturated seal. The seal must, of course, have a significantly smaller pore size than the reservoir rock so that 2y/r is much greater for the seal than the caprock. At this point,

MOVEMENT O F PETROLEUM FLUIDS

343

1 covered Area

by

fh>-frll

Water Bearing Sand Petroleum Bearing Sand

0 0

Migrating Petroleum Shale

F i g .2 0 . Proposed mechanism of trap-filling.(a) Petroleum


advances into the reservoir from the source rock;a dendritic pathway of petroleum-filled pores connects thetrap with the source rock. (b) Petroleum advances into the trap via a series of fronts, that reflect the changing composition of petroleum leaving the source rock. (c) & (d) The fraction of petroleum filled pores increases by petroleum displacing water downwards, until only small amountsof low pore-size rock remain unfilled. (assuming that the reservoir is not already filled to spill) petroleum will flow into the seal and no further increase in petroleum saturation within the trap is possible. As shown by Schowalter (1979), the maximum height of petroleum which may be supported by a caprock or seal is given by:

If lateral updip transmission of pressure occurs through the reservoir, a positive pressure or potential difference may exist between seal and reservoir; hence a term equivalent to VQw/r(p, - pp)g must be subtracted from the right hand side of equation (36). Lateral transmission of pressure in the reservoir rock will cause the seal to fail under some circumstances; this is predicted when the calculated h,, is negative.

Figure 20d suggests the emergence of a transition zone at the bottom of the petroleum accumulation. This is caused by the reduced buoyant pressure of petroleum near the bottom being increasingly unable to overcome the capillary pressure term (2ylr) of the petroleum potential. Thus only the very largest pores are oil-filled near the base of the trap. As petroleum continues to migrate into the reservoir, a point will be reached when the petroleum saturation is sufficiently great forthe petroleum to behave in a more liquid-like manner. From previous arguments this point is taken arbitrarily as corresponding to an overall saturation of -50%. This will be achieved when the petroleum column has sufficient height so that the force due to buoyancy at the top of the column exceeds the capillary pressure of the smallest pores that must become petroleum-filled to achieve an overall saturation 4 0 % . Equation (36) can be used to calculate the height where r is the critical pore radius: the height required increases with increasing petroleum density and decreasing critical pore radius (broadly equivalent to reservoir quality). Most petroleums require heights <l0 m, but heavy petroleum liquid in mediocre reservoirs may require 100 m. When -50% saturation is achieved throughout the reservoir, the petroleum contents of the reservoir will behave like a body of free fluid; the petroleum now becomes relatively mobile, and a condition for mechanicalstability will apply. This requires that the density of the petroleum increases with depth within the reservoir and withinany horizontally communicating parts of it. This may be achieved by a gradual overturning of the layers in the reservoir-bearing in mind that in mostcases the densest petroleum will be that generated first, and also reservoired first. These stages in the development of a reservoir are illustrated in Fig.20.Assuming that the reservoir isfilled from only one side, as is often the case, this model predicts thatthere will be a general trend in maturity across a reservoir, with most recently generated petroleum being found nearest its source rock. Once the reservoir is filled to thedepth necessary foran overall saturation of =50%, lateral maturity gradients may still persist if mixing is poor. Petroleum only arrives at one side of the growing accumulation, butit canstillachievemechanicalstability without chemical equilibrium: for example one petroleum density can be achieved by wide a range of natural petroleum compositions. In reality, however, the movement of fluids through porous media inevitably involves some mixing along and at right angles to the flow direction. This is essentially due to the random choice of pathways through the rock matrix by different parts of the moving fluid. The effect has been much studied by chemical engineers dealing with packed catalyst beds, (Denbigh 1965), as well as by geochemists, (Lerman 1979). It will cause some smearing out of the initial simple picture of a clear front of advancing petroleum moving through a reservoir structure. In conclusion, the manner in which a petroleum phase achieves sufficient saturation to behave as an interconnected fluidmayreadily be understood in terms of capillary and buoyant forces. However, the mechanism by which the system readjusts towards mechanicalequilibrium oncethe contents behave in a fluid-like fashion is not clear. In particular, the exact consequences for reservoirfilling of successive shots of increasingly mature and possibly less dense petroleum are hard to predict.

344

ENGLAND W. A.

ET AL.

One does, however, expect reservoir petroleum to preserve some lateral maturity differences, representing the history of the filling process, although mixing processes and the possible overturning of density gradients will also cause some degree of smearing of the initial simpler picture. In our experience these maturity differences can be detected by geochemical measurements made on petroleumsamples.

Gravitational and thermal segregation


In petroleum reservoirs, because of the often considerable height of petroleum columns, the effect of the Earths gravitational field must be taken into account. This significantly alters the equilibrium concentrations of the various hydrocarbons so that they vary systematically with depth (assuming sufficient time for the petroleum reservoir to come to thermodynamic equilibrium). This results in the denser (high molecular weight) components tending to be moreconcentrated toward thebottom of thepetroleum column. In the presence of a geothermal gradient (typically 30 Ckm-), thermally induced concentration gradientsmay become established: this is known as the Soret effect. This effect has been estimated to be of a similar magnitude to gravitational segregation (Holt et al. 1983), but has not been widely studied. The magnitude of the gravitationally induced concentration gradient may be estimated from chemical thermodynamics. If one assumes ideal mixingbehaviour-i.e.no interactions between the molecules, an exponential concentration gradient is predicted, Hirschberg (1984).

Diffusion. Molecular diffusionis a processwhich tends to reduce and eventually eliminate chemical potential gradients by the random motions of molecular species. Consider the case of a freshly filled reservoir with an initially non-uniform distribution of chemical components. Diffusion will cause a redistribution of matter so that horizontal concentration gradients will be eliminated and vertical, gravitational or thermally induced gradients become established. In order to estimate the rates and timescales of diffusion, it is essential to have accurate values for the diffusion constants of hydrocarbons in rocks. Unfortunately these are not available tobetterthan order-of-magnitude estimates, so our analysis is based on more accurate laboratory measurements in pure liquids. A reasonable first-order correlation of diffusion constants with viscosity, temperature and molecular dimension is given by the Stokes equation, Ghai et al. (1973):

xi(hl) = xi(h2)e(kl-k2)hg
where RT
(~resV,

(37)

This implies that Di is proportional to absolute temperature, but inversely proportionaltodynamic viscosity-ri is the hydrodynamic radius of the ith component. Since the viscosity of oil in an oil reservoir is approximately equal to that of pure water at 20 C, one can make use of results fromexperimentscamedout in water at 20 C without correction, as only order-of-magnitude estimates of diffusion rates are required. Table 8 shows the values of Di estimated for various hydrocarbons in pure liquids (i.e. in the absence of a rock matrix). The time necessary to achieve equilibration by diffusion, teq, may be estimated from the relation, e.g.Shulte (1980): (39) l is the length scale over which diffusion is considered to takeplace. In an individual petroleum column, l will be taken as 1 0 0m for the purpose of estimating zeqfor vertical diffusion: l will be taken as 2000mfor estimating tcq for lateral diffusion on a reservoir-wide basis. A factor of O2 is included to account for the increased length through which matter must diffuse in a tortuous porous medium, compared
Table 8. Diffusion coefficientsfor petroleum components in rock-free liquids and equilibration times (res) for selected components and length scales (1)

h, =

- M&

The mole fractions of the ith component at heights hl and h2 above a reference height are related to the reservoir fluids average density prcs andthe molecular weight, Mi, and partial molar volume, V,.Thus concentrationdifferences will be greatestfor high molecular weight species in low density reservoirs. In practice, however, real petroleum reservoirs often exhibit concentration gradients up tofive times greater than predicted by equation (37). This is a reflection of the non-ideal behaviour of multi-component mixtures. This type of system is best dealt with by using an equation of state approach in which an empirical mathematical model is used to describe the non-ideal behaviour of the mixtures. Schulte (1980) for example found that improved (though not complete) agreement with field resultpwas obtained by assuming non-ideal mixing. For example in the Brent field (UK NorthSea)themethane mole fraction changes by 5 mole % over 150 m, and in the Statfjord reservoir gravitationally induced concentration changes may have been sufficient to eliminate a sharp petroleum gas-petroleum liquid interface in the reservoir.

res (Ma) from equation (39)


Component* CH4
Di liquid) (pure
1=100m

1=2000m 42 84 422

c 1 2

tGm

In-reservoir mass transport processes


There are two possible mechanisms which could cause movement of material within a petroleum reservoir; these are molecular diffusion and thermal convection.

1 X 1 0 - ~(m2S-) Sahores & Witherspoon (1970) 0.5 X 10-9 Interpolated 1 X 10-10 Balthus & Anderson (1983)

0.1 0.2 1.0

* C, refers to a molecule with n carbon atoms; t Typical molecular weight for high molecular weight natural petroleum constituents known as asphatenes by the petroleum industry.

MOVEMENT OF FLUIDS PETROLEUM

345

with within a pure liquid. A value of 8 = v 3 is assumed for rock matrices which is similar to values found experimentally by Li & Gregory (1974), and is based on that predicted theoretically for a loose random pore structure. However, it is possible that when considering 8 over lengths measuredin metres or kilometres, additional larger-scale tortuosities operate.It should also be rememberedthat hydrocarbon molecules will only be diffusing through that partof the total pore network whichis petroleum-saturated, this will also affect e. The values for Diin pure liquids were calculated by making viscosity correlations, if needed, to data from the literature. The results are shown in Table 8 which shows the times necessary forvariouscomponentsto diffuse across 1 0 0m or 2000 m. These distances are representative of the vertical and lateral extents of a typical petroleum reservoir. It is clear that molecules will, in general, have sufficient time to equilibrate by diffusion to set up gravitationally or thermally induced concentration gradientswithin a reservoir in any given vertical petroleum column, provided no disturbance to the reservoir occurs during =l Ma. The case is different, however, on the larger length scales associated with horizontal diffusion over 2000 m. It is very unlikely that a reservoir will exist without disturbance for sufficient time to allow the hydrocarbon components to equilibrate horizontally, although this is often assumed to be the case. Convection. Thermal convection may occur when a body of fluid experiences a heat flow across it. The Rayleigh number, R , determines the onset of thermal convection: if R , > 40 for a body of fluid in a horizontal rock formation, convection currents will be established. It is important to note, that in contrast to diffusion, convection will completely homogenize a reservoir, removing any thermally or gravitationally induced concentration gradients.Figure 21 illustrates schematically the different effects of reservoir
(b

mixing by convection or diffusion on lateral and horizontal concentration gradients. Figure 21ashows the variation in composition between wells, and within an individual column of oil 'inherited' from the filling process shown in Fig. 20. Diffusion will rapidly cause thermodynamic equilibration within individual wells on a geological timescale, but not an a well-to-well basis. Thus Fig. 21b shows that wells 1, 2 and 3 have different average compositions. Given sufficient time for diffusion to occur, the compositions of the different wells would eventually approach each other, but the gravitationally induced compositional gradient would persist. This is shown in Fig. 21c. If, however, thermal convection occurs, the entire reservoir will become homogenized, Fig. 21d. Field evidence suggests thatpetroleum accumulations do not convect and that most accumulations reach the state exemplified by (b). However, in the absence of careful well testing and analysis, (b) may be hardto distinguish from (c). In practice, the Rayleigh number of a liquid petroleum reservoir is c. 0.1, so convection is not expected to occur except under exceptional conditions. This is in fact borne out by field observations, which show that significant lateral and vertical concentration gradientscan occur in reservoirs. Field evidence. Some of our studies of petroleum reservoir compositions have confirmed that they are indeed not well-mixed. This appears to be a new theoretical and observational conclusion as far as the oil industry is concerned. In previous studies of petroleum reservoirs apparent inconsistencies in analyses between wellsin the same reservoir, have often been ascribed to samplingand analytical errors. As far as vertical gravitationally induced concentration gradients are concerned,we have observed similar values to those of Shulte (1980), Hirschberg (1984), Creek &

Vertical Diffusion

ea
0 c a

WllS

1,2&3

Wells
1 2 3

Inherited composition variations

Composition

I Composition

(d)

Wells 1,263

Convection

Composition

21. The effects of various mass transfer processes on three dimensional reservoir composition for a hypothetical petroleum accumulation.

F i g .

346

W . A . ENGLAND E T A L .

Schrader (1985),Monte1 & Gouel (1985), and Riemens & de Jong (1985). We have also observed significant lateral differencesin the composition of samples from different parts of a reservoir. The petroleum which is in regions of the reservoir closest to actively generating source rocksis found to be more mature in its chemical composition than the petroleum in more distant parts. This is powerful evidence that petroleum reservoirs arenot well-mixed or homogenized laterally. This non-uniformity clearly has important repercussions in deciding on the best methods to extract the petroleum from the reservoir, as well as on reserve estimates. Because of the greater lateral extentof reservoirs compared with theirheight, reservoirs will usually be in vertical (gravitational and thermal) equilibrium, but horizontal dis-equilibrium.

Discussion and conclusions


We have demonstrated that the migration of petroleum fluids in compacting sediments can be described adequately by the physics and chemistry of multiphase fluid flow. The separation of this movement by previous authors into different processes, called primary migration, secondary migration and accumulation, artificial. is Nearly all petroleum flow in the subsurface occurs by the same mechanism-bulk transport driven by gradients in petroleum fluid potential. Capillary forces dominate over viscous forces and the flow is not turbulent. Diffusion is only significant in the immediate vicinity of the source rock, and within a petroleum accumulation where it is important for mixing petroleums of different composition on a c. 100 m length scale. Thenature of the flowitselfis non-turbulent,and is dominated by capillary forces in the rock pore network. Thus the migrating petroleum exploits the larger rock pores in preference the to smaller pores, which remain water-filled. The petroleum must fill c. 50% of the available rock pore volume in order to create an interconnected pathway, allowing movement to occur. Because of natural variations in theaverage pore sizes of the rocks involved, on a 10 m length scale petroleum movement is concentrated, or focused, intobetween 1 and 10% of the cross-sectional area of the rocks through which if passes: i.e. in the coarser, larger-pored beds. The direction in which petroleum moves may be conveniently described in terms of fluid potential. It is defined in such a way that petroleum fluids move from high to low fluid potential. Fluid potential gradients are caused by: variations in excess water pressures, the natural buoyancy of the less dense petroleum in a medium otherwise filled with denserwater, and capillary pressure differences. Insedimentary basins, excess water pressures (pressures greater than those needed to support the overlying water column) are induced by the finite rates at which sediments dewater in response to compaction forces. This is the dominant mechanism controlling fluid potentials at depths greater than 3 km. In contrast, buoyancyeffects predominate at shallower depths. Capillary pressure differences encourage the petroleum fluids to move from small pores to large pores, at lithological boundaries. We have chosen to investigate petroleum movement usingsimple models. Two classes of rocks have emerged.

Allfluids in rockswitheffective permeabilities >10-5m2 will move chiefly laterally; uncemented fractured limestones, and partially cemented sandstones fall in this category. Fluids in rocks with effective permeabilities <lO-mZ will move chiefly vertically; only when these fluids reach laterally continuous beds of permeability >lO- m* are they diverted laterally. In theory these models could be extended to three dimensions and coupled to models of the thermal history of a compacting sequence (e.g. McKenzie1981) and of the kinetics of kerogen to petroleum conversion (e.g. Tissot & Espitalie 1975), so as to predict the location of petroleum accumulations, their volumes and composition (e.g. Chenet et al. 1984). We feel that at our present stage in understanding thiswould not be sensible. We believe our simplified treatment is more useful (A. S. Mackenzie & T. M. Quigley in preparation). Increased knowledge of the distributions of pore sizes and permeability in sedimentary sequences is required before more complexmodelswould be justified. In particular, we have deliberately not considered therole of faults and fractures. Most of the joints and fractures that form during the burial of siliciclastic sediments are compressional and do not help flow (Price & Cosgrove in press): the situation may be different for carbonates. Although there is strong evidence that the faults have lowintrinsic permeability parallel to the fault plane, the permeability and capillary pressures in direction a sub-perpendicular to the plane are in our experience, highly variable.Solving the controls on thisvariabilityiscrucial before we can attempt to model flow across faults, and hence in three-dimensions within a sedimentary basin. As petroleum fluids migrate, vertical movement will generally lead to changes in pressure and temperature: this will cause changes in the relative volumes of petroleum liquid and gas. Our physical understanding is not at present sufficient to describe accurately the complex three-phase flows involving petroleum gases and liquids through water-saturated rocks. However, muchcan be learned by constructing two-phase models involving water and one other petroleum gas or liquid phase. The final stage of petroleum movement is its accumulation in a reservoir or trap, overlain by a finer-grainedlithologyknown as a seal, which holds the petroleum within the reservoir by virtue of the capillary pressure difference across the lithological boundary. Unless sufficient petroleum fills the reservoir to produce a significant column height, its buoyant pressure will not overcome the capillary pressure of the majority of the pores in the reservoir, and a poorly-filled accumulation will result. Once a petroleum accumulation has achieved mechanical stabilization, it can only mix by diffusion or thermal convection. However, in liquid petroleum accumulations, convection is not significant a process. Thus the only available mixing mechanism diffusion. is Except for distancesless than c. loom, diffusionis too slow to cause compositional mixing in geologically accepted time-scales. Thus in most commercial petroleum reservoirs lateral compositional variations acquired during the filling process will be retained. However, since the column heights of typical reservoirs are measured in tens to hundreds of metres, the vertical compositions may reach thermodynamic equilibrium in a reasonably short time. Thus petroleum reservoirs are ina state of lateral disequilibrium, but are vertically equilibrated.

MOVEMENT OFL P U EIT D R SO L E U M We are grateful to Stephen W. Richardson and D a n P. McKenzie for helpful discussions. We thank Gary P. Cooles for donation of the results of his oil saturation experiments (Table 4) and British Petroleum p.1.c. for permission to publish.

341

crude oils containing dissolved gases. American Institute of Mining and Metallurgical Engineers, Technical Publication No. 1624. Kmoss, B. 1985. Experimentelle Untersuchung der D i m i o n Niedrigmolekulerer Kohlenwasserstoffe in Wassergesattigken Sedimentgestein. PhD thesis, University of Aachen. LERMAN, A. 1979. Geochemical Processes, Water and Sediment Environments. Wiley, New York. 64-5. LEYTHAEUSER, D., SCHAEFER, R. G . & YUKLER, M. A. 1982. Role of diffusion in primary migration of hydrocarbons. American Association of Petroleum Geologists Bulletin, 66, 408-29. Y.-H. LI & GREGORY, S . 1974. Diffusionof ions in sea water and in deep-sea sediments. Geochimica et Cosmochimica Acta, 38,703-14. MACGREGOR, D. S. & MACKENZIE, A. S. 1986. Quantification of Oil Generation and Migration in the Malacca Strait Region. Proceedings of the15th Annual Convention of theIndonesianPetroleum Association, 7-9 October 1986, Jakata. MCKENZIE, D. 1981. The variation of temperature with time and hydrocarbon maturation in sedimentary basins formed by extension. Earth and Planetary Science Letters, 55, 87-98. MACKENZIE, A. S . , LEYTHAEUSER, D., MULLER, P., RADKE, M. & SCHAEFER, R. G.1986.GenerationandmigrationofpetroleumintheBraearea, CentralNorthSea. Proceedings of the 3rd Conference on Petroleum Geology of NW Europe, London, 26-29 October 1986. MONTEL, F. & GOLJEL, P. L. 1985. Prediction of compositional grading in a reservoir fluid column. Society of Petroleum Engineers of American Institute of Mining and Metallurigical Engineers, Paper No. 14410. PRICE, L. C, WEGNER, L.M,, GING, T.& BLOW,C.W.1983.Solubility of crude oil in methane as a function of pressure and temperature. Organic Geochemistry, 4, 201-21. PRICE, N. J. & COSGROVE, J. inpress. Analysis ofGeologicalStructures, Cambridge University Press. RIEMENS, W. G. & DE JONG, L. N. J. 1985. Birba field pVT variations along Sociery of the hydrocarbon column and confirmatory field tests. Petroleum Engineers of American Institute of Mining and Metallurgical Engineers Paper No. 13719. -R, L. C. & DRAKE, R. L. 1945. Pore-size distribution in porous materials. Industrial & EngineeringChemistry (Analytica[Edition), 17, 782-6. SAFFMAN,P. G. & TAYLOR, G. I. 1958. The penetration of afluidintoa porous medium or Hele-Shawcellcontainingamoreviscousliquid. Proceedings of the Royal Society of London, Series A. 245, 312-29. SAHORES, J. J. & WITHERSPOON, P. A. 1970. Diffusion of light paraffin G. D. & SPEERS, hydrocarbons in water from 2C to 80C. In: HOBSON, G. C. (eds) Advances in Organic Geochemistry 1966 Pergamon Oxford. 219-30. SCHOWALTER, T. T. 1979. Mechanics of secondary hydrocarbon migration and entrapment. American Association of Petroleum Geologists Bulletin, 63, 723-60. SCHULTE, A. M. 1980. Compositional variations within a hydrocarbon column due to gravity. Society of Petroleum Engineers of American Institute of Mining and Metallurgical Engineers Paper No. 9235. SMITH, J. E. 1971. The dynamics of shale compaction and evaluation of pore fluid pressures. Journal of Mathematical Geology, 3, 239-63. Volumetric and Phase Behaviour of Oilfield STANDING, M. B. 1952. Hydrocarbon System. Reinhold, New York. STAUFFER, D. 1979.Scalingtheoryofpercolationclusters. PhysicsReports (Review Section of Physics Letters), 54, 1-74. TERZAGHI, K. 1948. Theoretical Soil Mechanics. Chapman & Hall, London. TISSOT,B. P. & ESPITALIE, J. 1975. LCvolutionthermiquedelamatiere organiquedessediments:applicationsdunesimulationmathkmatique. Revue de llnstitut Francais du Petrole, 30, 743-77. - & WELTE, D. H. 1984. Petroleum Formation and Occurrence. Springer-Verlag, Berlin.

References
AMYX, J. W.,BASS,D.M.
& WHITTNG, R. L.1960. PetroleumReservoir Engineering, Physical Properties. McGraw-Hill, New York. BALTHUS,R. E. & ANDERSON, J. L. 1983.Comparisonofgelpermeation

chromatography elution characteristics and diffusion coefficients of asphatenes. Fuel, 63, 530-5. BERG, R. R. 1975. Capillary pressures in stratigraphic traps. American Association of Petroleum Geologists Bulletin, 59, 939-56. F., UNGERER, P,, NOGARET, E. & PERRIN, J. F. 1984. CHENET,P.Y., BESSIS, Commentlesmodelesmathematiques en geologiepeuventreduirele risque en exploration petroliere. Eleventh World Petroleum Congress, 2, 385-404. Wiley, Chichester. COOLES, G. P., MACKENZIE, A. S. & QUIGLEY, T. M.1985.Calculation of masses of petroleumgeneratedandexpelledfromsourcerocks. In: LEYTHAEUSER, D. & RULLKOTER,J. (eds) Advances in Organic Geochemistry 1985. Pergamon, Oxford (in press). CREEK, J . L. & SCHRADER, M. L. 1985. East Painter Reservoir: an example of a composition gradient from a gravitational field.Society of Petroleum Engineers of American Institute of Mining and Metallurgical Engineers Paper No. 14411. DAHLBERG, E. C. 1982. Applied HydrodynamicsinPetroleumExploration. Springer-Verlag, New York. DAKE, L. P. 1978. Fundamentals of Reservoir Engineering. Elsevier, Amsterdam. DENBIGH, K. 1965. ChemicalReactor Theory: A n Introduction. Cambridge University Press, Cambridge. DICKENSON, G. 1953.Geologicalaspectsofabnormalreservoirpressures. American Association of Petroleum Geologists Bulletin, 37, 41C-32. DURAND, B. (ed.) 1980. Kerogen. Editions Technip, Paris. -(1981). Advances in Organic Geochemistry, 17, Wiley, Chichester. FRICK,T.C. & TAYLOR,R. W. 1962. Petroleum Production Handbook. McGraw-Hill, New York. GHAI,R. K., ERTL,H. & DULLIEN, F. A. L. 1973. Liquid diffusion of non-electrolytes. American Institute of Chemical Engineers Journal, 1 9 , 881. GLAs~, 0.1980. Generalized pressure-volume temperature correlations. Journal of Petroleum Technology, 32, 785-95. HINCH, E. J. 1985. The recovery of oil from underground reservoirs. Journal of Physico-chemical Hydrodynamics, 6, 601-22. HIRSCHBERG, A. 1984. The role of asphaltenes in compositional grading of a of American reservoirsfluidcolumn. SocietyofPetroleumEngineers Institute of Mining and Metallurgical Engineers PaperNo. 13171. HOLT, T., LINDEBERG, E. & RATKJE, S. K. 1983. The effect of gravityand temperature gradients on methane distribution in oil reservoir. Society of PetroleumEngineersof American Institute of MetallurgicalEngineers, Unsolicited paper 11761. Entrapment of petroleum under hydrodynamic HUBBERT, M. K. 1953. conditions American Association ofPetroleumGeologistsBulletin, 37, 1954-2026. JONES,R. W. 1980. Some mass balance and geological constraints on migrationmechanisms. In: ROBERTS, W. H. & CORDEU,R. J. (eds) Problem of PetroleumMigration, American Association of Petroleum Geologists Studies in Geology, 10, 47-68. K A n , D.L., MONROE,R.R. & TRAINER,R.P. 1943.Surfacetension of

Received 30 April 1986; revised typescript accepted 12 September 1986.

You might also like