Fractures and naturally fractured reservoirs

Almost every core and every outcrop is cut by fractures, and almost every layer has them. The question is not whether there are fractures. It is whether they affect the flow of fluids. To answer it, we first have to understand what kind of break we have. In fracture mechanics, fractures are divided by how their walls moved. Opening is mode I. Sliding along the surface is mode II (across the edge of the fracture) or mode III (along the edge). For a reservoir it also matters what fills the break, and what has happened to the grains at its edges.

What we can read on the walls of a fracture in core:

• Plumose structure (a feather-like pattern): fan-shaped grooves on a joint. They spread from the place where the fracture started, in the direction in which it grew, and arc-shaped lines mark stops of its front.
• Slickensides and slickenlines (polished surfaces and the grooves on them): a sign of a shear fracture. They show the direction of the last slip (parts of a fault).
• Druses: crystals that grew freely on the wall. Their growth needs an open cavity filled with mineral-rich water. So the fracture was open at least at some time.
• Fibrous fill across a vein: the fracture opened a little at a time, cracking and healing again and again (crack-seal).
• Mineral bridges between the walls of a partly healed fracture. They work as natural proppant (grains that hold a fracture open) and keep the fracture open as the load grows (see the section on fracture closure).

Intact rock breaks when the Mohr circle of effective stresses touches the failure envelope (Mohr circle). Effective stresses are the stresses carried by the rock itself, after the pore pressure is subtracted. The failure envelope is the limit line beyond which the rock breaks. Under compressive normal stress the envelope is close to the Coulomb straight line (Mohr–Coulomb criterion). In the tensile region it bends towards the point $\sigma_n' = -T$, where $T$ is the tensile strength of the rock, usually a few megapascals. A handy model of this part of the envelope is the Griffith parabola.

From this comes the main rule of orientation. A joint opens in the direction of least compression, so it is at right angles to $\sigma_3$ and lies in the plane of $\sigma_1$ and $\sigma_2$. In an extension setting, where $\sigma_1$ is vertical, joints are vertical and run along the greatest horizontal stress $\sigma_H$. In compression, where $\sigma_3$ is vertical, they are horizontal. Shear fractures form in pairs, as conjugate sets, symmetric about $\sigma_1$. The acute angle between them, about 60°, is cut in half by the direction of $\sigma_1$ (Anderson angles).

At depth all the principal stresses are compressive, so a joint needs a “helper”: pore pressure. It moves the Mohr circle to the left without changing its size. When $p$ exceeds $\sigma_3$ by $T$, the rock breaks in the same way as in hydraulic fracturing in a well (fracture pressure). Such natural hydraulic fracturing happens where the pore pressure is abnormally high (Abnormal formation pressures: origin, prediction and recognition): in quickly sinking shales, in source rocks while they are making oil and gas, and under reliable seals (rocks that hold fluids in). The second way is to reduce the stresses. When rock is uplifted and eroded, it cools and shrinks, and the removal of load from above also lowers the horizontal stresses. The smallest of them can become tensile, and unloading joints open in the rock.

The mechanics is one, but the settings in which rock reaches failure are different, and each leaves its own pattern of fractures. Once we know the pattern, we can predict fractures between wells.

The intensity of fractures depends on the rock too. The stronger and stiffer it is, and the fewer pores and clay it has, the more fractures it gets for the same deformation. Dolomites, silicified limestones, cherts and cemented quartz sandstones crack more densely than porous limestones, weak sandstones and clays. In clays fractures either do not form or close up. So in a layered sequence the fractures are often confined to strong layers and stop at their boundaries with clays. Each such layer works as a separate mechanical layer. The alternation of layers by their mechanical properties is called mechanical stratigraphy.

In folds we predict fracturing from curvature. The stretching in the outer arc of a layer is proportional to its thickness and curvature (curvature and stretching), so there are more fractures in crests and where a fold bends sharply. Maps of the curvature of horizons (the mapped tops of layers) from seismic data are a common way to look for such zones. Near faults, the fracture density is greatest in the damage zone (the cracked rock around a fault) and falls quickly with distance. At the tips of faults and in relay ramps, where the slip passes from one segment to the next, fracturing is especially strong (relay ramps).

Fractures of one orientation and origin form a set. Systematic fractures are long, planar and almost parallel. Non-systematic fractures are short and curved and end against the systematic ones. We show the orientation of sets by poles on a stereographic projection: each set is a cluster of points (stereographic projection). For vertical sets a rose diagram of strikes (a circular histogram of directions) is handy too. The relative age of sets is found from their junctions: a younger fracture ends against an older one, making a T-shaped junction, or offsets it.

How many fractures there are in the rock is described by the fracture intensity. The easiest to measure is the linear intensity $P_{10}$ (the density): the number of fractures per metre of a line of observation, such as a well, a core, or a line on an outcrop. One divided by it is the average spacing $s$ of the set, if the line is at right angles to the fractures. Some measures do not depend on the direction of the line: the total length of fracture traces per unit area ($P_{21}$, m/m²) and the total area of fractures in a unit volume ($P_{32}$, m²/m³). It is $P_{32}$ that is usually set in fracture models.

In layered sequences, the joints of one set are most often limited by a mechanical layer, and their spacing grows with its thickness: in thin layers the fractures are dense, and in thick layers they are sparse. The explanation is mechanical. The stretching passes into the layer through its boundaries, and next to each fracture the stress in the layer is relieved over a distance of about the thickness of the layer. A new fracture between two neighbours can appear only if they are farther apart than about this thickness. As the strain (the amount of stretching) grows, the spacing falls to a limit, the state of saturation, in which the ratio of spacing to layer thickness is close to one (0.8–1.2 in observations). After that, the stretching goes into opening the existing fractures. The rule is not universal. In thick layers (thicker than one and a half to two metres), the spacing, according to a number of observations, hardly depends on thickness, and fractures of natural hydraulic fracturing do not follow it.

Fluid flows through a fracture between two walls. If the walls are smooth and parallel, an exact solution describes the flow. The flow rate through the fracture is then proportional to the cube of the aperture $w$ (the opening; in Russian-language literature the term “openness of a fracture” is more common). This is the cubic law. The permeability of the fracture itself is $w^2/12$ (permeability of a fracture). For a rock cut by a set, we multiply it by the share of volume taken by fractures, and get the sum of the cubes of the apertures.

The aperture that we see on core or estimate from an image is the mechanical one: the average distance between the walls. Real walls are rough and touch each other at separate points, and the fluid goes round the contacts through winding channels. So the hydraulic aperture, the one that gives the measured flow rate when it is put into the cubic law, is smaller than the mechanical one. The rougher the walls and the larger the area of contacts, the smaller it is. The aperture on core is also distorted: when the core is brought up, the stresses are removed, and the fractures open wider than in the reservoir.

A fracture stays open because its uneven walls rest on each other at their bumps, and sometimes on minerals that have grown on the walls. As the effective normal stress grows, the bumps are crushed and the fracture closes. First this happens fast, then more and more slowly, because the contact area grows and the fracture becomes stiffer. This relation is often described by a hyperbolic law.

During development reservoir pressure falls, and the effective normal stress on the fractures grows. But not by the whole drop in pressure: the horizontal stress also falls with depletion (the fall of pressure as fluids are produced), by a share $A$ of the pressure drop (poroelastic coefficient). For vertical fractures at right angles to $\sigma_h$, the effective normal stress grows by $(1 - A)\,\Delta p$.

Orientation decides which fractures are open. Fractures at right angles to the least stress $\sigma_h$ are squeezed less than the others and are more open. So the permeability of a fractured layer is usually greatest along $\sigma_H$ (fractures and the direction of stresses). Observations in deep wells suggest a second rule: the conductive fractures are often the critically stressed fractures. These are fractures on which the ratio of shear stress to effective normal stress is close to the coefficient of friction (how close a fracture is to slip). Small slips along them do not let the uneven walls close. The rule is not universal. In some rock masses, fractures inherited from earlier stages conduct, whatever the present stresses.

Almost every layer has fractures, but a reservoir is called fractured when open natural fractures clearly change its behaviour. A widely used classification splits such reservoirs into four types, by the share of storage and permeability that the fractures give.

• Anisotropy. Permeability along the fractures can be tens and hundreds of times higher than across them. Wells interact more strongly along the fractures, the drainage area is elongated, and horizontal wells are better drilled across the fractures (direction of the well).
• Early breakthrough of water and gas. Injected water, and water from the water-bearing zone, go along the fractures and leave the oil in the matrix. The water cut (the share of water in the produced fluid) rises early and fast. We place injection wells in rows along the strike of the main fracture set, so that water moves across the fractures. We do not put producers on the continuation of the fractures from the injectors (flood patterns).
• Exchange between matrix and fractures. Oil leaves the matrix blocks for the fractures by capillary imbibition in water-wet rock (the rock soaks up water, like a sponge, and pushes the oil out), by the difference in densities, and by elastic expansion. The rate and completeness of recovery depend on the size of the blocks and on wettability. In models, such layers are described by two systems of pores, the matrix and the fractures (dual porosity).
• Fall of permeability when pressure falls (the section above). So for fractured layers it is especially important when and how to maintain pressure.
• Drilling. Open fractures take drilling mud, up to a complete loss of circulation, and when the pressure in the hole is lower than reservoir pressure, they give inflows. Mud losses are one of the most reliable signs of conducting fractures.
• Well tests. Permeability from a test is often many times higher than from core. On the diagnostic plot (the plot used to recognise flow regimes), the pressure derivative of a layer with dual porosity makes a dip between two stretches of radial flow (pressure derivative).

In Kazakhstan, fractures are especially important in the subsalt carbonates of the Pre-Caspian Basin (subsalt tier). At Tengiz, most of the porosity is vugs, the primary porosity does not exceed 2–3%, and from the available data the permeability of the carbonates is controlled mainly by fractures. The fractures there are of several generations, from early ones that formed before the rocks were buried deep to late tectonic ones. Researchers rate differently which of them matter more for the present inflow. The general picture is this. The early fractures are steep and run parallel or perpendicular to the edge of the platform. There are most fractures in the strong framework rocks, the boundstones, of the outer platform and the upper slope, and they can be followed into the platform up to 2 km from the edge. In the inner part of the platform there are few fractures, and the inflow is set by the properties of the matrix. Some of the fractures have been widened by dissolution up to vugs. Such zones give both strong mud losses in drilling and highly productive wells.

At Kashagan, three generations of fractures have been identified from core. The first are early ones, linked to the instability of the edge of the buildup, and more frequent along it. The second are burial fractures, linked to stylolites. The third are thin fractures that formed while the accumulation was filling, when the fluid pressure had risen enough to break the seal. The large-scale permeability of the edge of the buildup is linked with the early fractures. Mud losses are used as an indirect measure of the conductivity of fractures. On an image, a high or low resistivity of a fracture does not by itself say whether it conducts.

Another example is the Triassic deposits of South Mangyshlak. The accumulations in the Middle Triassic carbonates are linked to fractured dolomites with vuggy porosity. The matrix of the dolomites is almost impermeable, with porosity of not more than 3–4%, and the voids were dissolved out (leached) along fracture sets that formed after the Early Miocene. At the Oymasha field the oil accumulation lies in fractured and weathered granites of the basement (the old crystalline rock beneath the sedimentary layers), productive down to 300 m below the top of the basement. The voids in such a reservoir were made mainly by fractures and weathering. In neighbouring areas the accumulations are controlled less by uplifts than by zones of fracturing and leaching that strike north-east and north-west. The north-east zones are linked with faults that have a considerable strike-slip component. So we can expect such accumulations outside uplifts too (strike-slip zones).

Core is the only source where we can touch a fracture: we see whether it is open or healed, what fills it, and whether its walls have slickenlines or a plumose pattern. But core is small, and it is not always oriented. We restore its orientation from the image of the same interval, or we take oriented core (coring). It is important to tell natural fractures from induced ones. Drilling and bringing up the core create their own. These are discs into which the core splits when it is drilled under high stress (core discing), fractures along the core axis, and breaks made when the core is pulled out and laid in boxes. A natural fracture is given away by mineral fill, traces of slip, altered walls, and an orientation that agrees with other fractures of the set.

Borehole images (borehole image log) give the orientation of every fracture from its sinusoid (attitude from an image) and a continuous picture along the whole well. On electrical images in water-based mud, an open fracture filled with mud filtrate (the liquid from the mud that has entered the rock) conducts current and looks like a dark sinusoid. A fracture healed with calcite, quartz or anhydrite looks like a light, high-resistivity one. Fractures with clay or pyrite fill, and thin shale beds, can be dark too. So we tell a fracture from a bed by its angle and direction of dip relative to the bedding, and by whether the sinusoid breaks off. In oil-based mud, open and healed fractures are equally high in resistivity on ordinary electrical images, and they are harder to tell apart. On acoustic images, open fractures show up by a weaker reflected signal.

Drilling itself creates fractures and breakouts on the wall of the hole, and this is valuable information about stresses (direction of stresses). In a vertical well, thin tensile fractures induced by drilling run along the hole in pairs on opposite sides, at the azimuth of $\sigma_H$. Breakouts (pieces of the wall that have broken off, seen as wide smeared bands) lie 90° from them, at the azimuth of $\sigma_h$. Natural fractures parallel to $\sigma_H$ are more open than the rest.

Other observations reveal fractures too. One sign is mud losses and drops of the tool. Another sign is a flow profile from production logging with a spinner flowmeter (a tool that measures the flow in the well), when the inflow comes from separate narrow intervals. A third is permeability from a well test that is higher than the core permeability. A fourth is shear-wave anisotropy in dipole sonic logging (Sonic log). Between wells, seismic signs help: curvature of horizons, breaks in the continuity of reflections, and azimuthal anisotropy. No single sign proves that fractures conduct. Mud losses and inflow are checked, and images are calibrated against core and tests.